Evaluating the Pressor Effects of Drugs & Ambulatory Blood Pressure Monitoring Studies

– Good morning everyone. Everybody hear me, am I on? Okay, great. I’m Mark McClellan, I’m the Director of the Duke-Margolis
Center for Health Policy, and on behalf of the
Duke-Margolis Center and the FDA, I’d like to welcome all of
you to today’s public event on evaluating the pressor effects of drugs and ambulatory blood
pressure monitoring studies, which we at Duke-Margolis are convening under a cooperative
agreement with the FDA. As you all know, the issues surrounding blood pressure monitoring are of substantial importance
across a wide range of clinical studies and have
important downstream impacts on patient care and
potentially health outcomes for populations. In May of 2018, FDA
released a draft guidance for industry on these issues titled “Assessment of Pressor Effects of Drugs”. The guidance advises industry on the pre-marketing assessment of a drug’s effects on blood pressure and addresses the precision
of blood pressure measurement. The guidance specifically
highlighted four areas in which FDA welcomes
feedback from experts. One is the temporal relationship between changes in blood
pressure and changes in risk; second, the precision
study and its relevance to FDA’s recent draft guidance; third, diverse developmental programs and the evaluation of
blood pressure effects among different, diverse, potentially diverse patient populations; and then fourth, these placebo groups and ambulatory blood
pressure monitoring studies. Following the public comment period on the guidance document,
today’s public meeting is an opportunity for stakeholders to provide some further input
on these subjects for FDA in this interactive environment as they continue to refine the guidance. In addition, today,
our colleagues from FDA will present some new internal analyses of ambulatory blood
pressure monitoring studies with the goal of further
discussing their role in clinical development. We’re going to have
time throughout the day for audience question and answers and feedback during each of the sessions. And then we have a final
session at the end of the day that is an open comment period to make sure everything that
you all would like to raise has gotten covered
during our time together. Throughout the day we hope to move towards more consensus where possible and engage in discussions that foster the regulatory
science advances that FDA wants to use to
guide their evaluation of the pressor effects of drugs. So, this is an important
step in that overall process and an important area for
clinical product development and management of potential safety risks. So thank you all for joining us today. Before we get started, just a
couple of housekeeping notes. As I said, this is an interactive meeting. We’re gonna have time in each session for discussion with those
of you who are here today. For those of you who are in attendance, we have a wireless microphone
set up at each table for use during the day. When you have a question in
these parts of the sessions, please wave, stand, do something to indicate, to let us know
that you’d like to be called on and the moderator for the session will make every effort
to get everyone engaged in the discussion. When you use that microphone, there are a couple of
different types here. I see some in the front tables that have kind of a square at the bottom. There’s a button on the
bottom that you press to turn it on. The mics are all live. And some of the mics have a switch. Just be sure to turn it on when
you’re making your comment. And, for those of you who
are joining us by webcast, thank you for joining us as well. And we encourage you to participate in today’s discussion as well. If you have a question for a panel, you can send them to us,
that’s at [email protected], [email protected], that’s the email address for questions or comments,
and we’ll try to incorporate as many of those as possible
into the meeting as well. For those of you who are here, wifi information’s on
the table in the foyer outside the meeting space
if you’re having any trouble getting connected. And again, we really
appreciate all of you, here in the room and online, taking the time to join us today. I wanna remind everyone that
this is a public meeting and the event is being broadcast, so everything you say is
going to be part of the record for the event. For those of you who
are in the room today, feel free to help yourself
to coffee and beverages, located just outside. Lunch will be on your own. We have a break period for that and there are plenty of very good options in the nearby area. There are some restaurant maps available at the registration desk. And finally, I just
wanna remind you all too that although this
meeting is being convened under a cooperative
agreement with the FDA, it is not a federal advisory committee or a full part 15 hearing. There will be no votes
or formal consensus, determination, et cetera. The meeting will be a success if there is a good,
strong exchange of ideas and open discussion. So, that’s the framing for today. We’d like to get going right now with some opening remarks
from Dr. Ellis Unger, who’s the Director of the
Office of Drug Evaluation-I and the Office of New Drugs, and then from Dr. Norman Stockbridge, the Director of the
Division of Cardiovascular and Renal Products, to help set the stage for the day’s discussion. So, Ellis, if you can come on up and I’ll turn this over to
you for your opening comments. Thank you very much. – Thank you very much, Mark. And I’d like to thank Duke-Margolis
Center for Health Policy and I’m thanking all
of you for being here, and also those who are
joining us on the web. I think that everyone in the room understands the importance
of blood pressure as a risk factor for
cardiovascular disease and some of you might wonder why we put out a draft guidance last May about assessing the blood
pressure effects of drugs because this is hardly new. We’ve been assessing blood
pressure effects of drugs probably since we’ve
been assessing new drugs. But we have a huge problem in this country in terms of unrecognized, undertreated, and untreated hypertension and the reality is that, as all of you know, there are millions of people
walking around in this country with blood pressures that are too high. And we’d like to do everything that we can to lessen the problem. So we regulate the drugs that are used to treat hypertension. Norm Stockbridge is gonna come up, he is the division director of the Division of
Cardiovascular and Renal Drugs and they have an assortment of drugs that are highly effective
in treating hypertension. Well, we also regulate
drugs throughout the FDA, throughout Office of New Drugs that actually contribute to the problem, that raise blood pressure. And, given that one of our charges at FDA is to ensure that when we label the drug, we provide adequate instructions for use. It’s critically important that
the blood pressure effects of drugs are well characterized and well described in labeling. The magnitude of the effects should be described in labeling, the attendant risks should
probably be described in labeling and then some management strategies should be provided in labeling for drugs that can cause or exacerbate hypertension. And, we recognize that through the years we haven’t always gotten this right, which I’m sorry to say,
but it happens to be true. What we’ve typically
done through the years is companies do, as all
of you know, do studies. You get cuff blood pressure measurements at multiple visits through the studies and you put them together
and you analyze them and the FDA puts their heads together with the company’s heads, figures out what the blood pressure effects are and makes some attempt to write a label, which may or may not talk about specific blood pressure changes. In preparation for this workshop, I took a look at four
new drug applications that are sitting on our
server waiting to be analyzed and I counted anywhere from 11,000 blood pressure
measurements for one drug to 40, about 45,000 for another drug. And that is very typical. So all these patients
put out their arms and, well, have a machine,
a Dinamap or whatever, measure blood pressure. And then the question is
well what does the FDA do with all this stuff and
what do the companies do with all this stuff. And I think that sometimes
we don’t do as good a job as we could in actually
trying to untangle those and figure out what the blood
pressure effect a drug is. About a year and a half
ago a friend came to me because she, well she has osteoarthritis and she’d been taking
meloxicam for a while for her osteoarthritis and
her arthritis got worse and somebody increased her dose from 7 1/2 to 15 milligrams a day, and she
was subjected to some stress in her life and her blood pressure, which had been pretty
well controlled I guess, got out of control. So all of the sudden her
blood pressures were high and she’s on a higher dose of meloxicam and she asked me well,
could it be the meloxicam. So I took a dive and looked
at the meloxicam label, which I think we approved in 2000. There was nothing in that label about blood pressure
effect to be expected. There is a box warning that says the drug can cause cardiovascular,
I don’t know, demise. It doesn’t say MACE exactly. Serious cardiovascular and
gastrointestinal events. Well great, thank you. Maybe all that is mediated
by the blood pressure effect. We don’t know. We’re gonna talk about that when we’re gonna talk about precision. But, at least tell me what
the blood pressure effect is. You probably got 15,000
blood pressure assessments during the trial. I went to the FDA review in our archives and I found nothing in the review about specific blood pressure changes. So I found this rather disconcerting. We have all these data and we don’t always
take advantage of them. So, my opinion is we need
to try to do this better and this was part of the
reason for the genesis of the guidance and why you’re here today. And we have some specific
questions for you. Dr. McClellan mentioned them. One is how best do we
analyze the cuff data. We’re gonna talk about ambulatory
blood pressure monitoring and how to analyze that. The draft guidance has a
sentence or two in there about how to use the ambulatory
blood pressure monitoring. It talks about a time weighted
average over 24 hours. Okay, and then that presupposes that if you have a 10 millimeter
elevation for 12 hours, it’s the same as having a
five millimeter elevation for 24 hours. Well, it integrates the same. Well, is that true? You guys know more than we do, but, you know, drugs don’t
necessarily, you know, drugs with shorter half-lives
might have shorter effects. So there are all kinds
of things to consider in terms of how to analyze the data. We wanna talk about temporal relationships between changes in blood
pressure and changes in risk. When we drafted the guidance, we started out kind of in the
mold of the thorough QT study. We called the guidance the thorough blood pressure
guidance for a while and we had the idea that there
might be some threshold of, a regulatory threshold of
concern for blood pressure just as there is for QT. And I think we lost our
enthusiasm as time went on and we basically came to the conclusion that we should probably assess the blood pressure effects of a drug and then look at the blood
pressure effects in context and make case by case decisions
in terms of what to do. And you all know and as
articulated in the guidance, one would think about the chronicity, well the blood pressure effect, the chronicity of use, the indication for which the drug is approved, and then the baseline risk
of the patient population. So, we have to think
about all those things. Interestingly enough,
after the guidance went out or around that time, FDA approved testosterone
enanthate preparation for testosterone replacement therapy in adults males for conditions associated with deficiency or absence of endogenous testosterone. And the drug increases blood pressure and the division and office
responsible for that drug approved that drug with a boxed warning, which is probably the first
time we’ve put a boxed warning on a drug for increasing blood pressure. We have a lot of drugs that
increase blood pressure. We have a drug for orthostatic
hypotension, midodrine, which increases blood pressure. We have a warning, although
I don’t think it’s boxed. But, at any rate, there are
a number of drugs out there that are used chronically
that increase blood pressure and we don’t have boxed warnings. So something we wanna think about, labeling is our responsibility but we wanna think about well,
how should drugs be labeled. So these are some of the questions that we wanna try to tackle today and we really look forward
to all of your insights this morning and this afternoon. So I’ll thank all of you in
advance for your thoughts and we’re looking forward
to a very interesting and deep discussion. Thank you. (audience clapping) – All right, thanks very much, Ellis. And so next is Norman Stockbridge, who’s gonna provide
some additional context for today’s meeting, talk
about some of the key points from the guidance document where FDA would like further feedback and share some insights from his division. Norm, thanks for being here. – Good morning. I wanna add my thanks to people who helped put this together, the Duke-Margolis people. I want to call out particularly Naomi Lowe who helped from the FDA side
getting this put together; all of you who traveled from all across the country at least, some from overseas to get
here and participate in this; the weather gods for having
the Polar Vortex last week and not this week. So, great, and I’m
looking forward to this. We’ve had an interest in trying to get some guidance
around the pressor effects of drugs for some time now. Rick Turner here reminds me that the CSRC meeting on this topic, how many of you were here for that? That was 2012 and since 2012, Ellis, Bob Temple, Doug Throckmorton, a bunch of us have been
meeting for three hours a day, weekdays, weekends, and holidays to try and hammer out a
guidance on this topic. We didn’t entirely succeed. We never got to a place where
we had consensus on everything that needed to go into a guidance and the first group of, first group of sessions here this morning are around four areas that we call out in the draft guidance where we didn’t, couldn’t even get
consensus in a small room. We’ll see if we do better today. Otherwise, I’m not sure how
many years down the road a real guidance is going to be. As you’ve heard, there’s a second session on trying to do ABPM,
which is what we think is critical technology
for careful evaluation of pressor effects. That’ll be in the second session. And then at the end of the day, we’ve allocated some time
for you to talk about issues that haven’t come up. The four issues that
come up early in the day are the ones we identified. You may have many more and
we’ll need to hear those. And then finally I
wanted to point out that Ellis pointed out some
analogies with the QT business. We have already gotten clearance through the Office of New Drugs to start treating consults relating to pressor effects of
drugs much the same way we are, in fact, treating the QT business. Many of you know we
have a particular group of reviewers, some of whom
are in the room here today, who work on the QT business, have developed a set of data standards, a backend database, have
published a large number of papers that helped move that field forward and allowed people to do
those studies efficiently. We’re gonna do the same thing with the pressor business. We’re gonna be collecting data, trying to give consistent advice, researching the data that we collect to make the advice we give more efficient. So, I am hopeful that we’ll resolve at least some of the issues that we had outstanding
and that we won’t need another meeting in 2024 to talk about how to get a guidance done. Thank you. (audience clapping) – All right, thanks, Norm, very
much for those remarks too. So, an important meeting today and again, glad that you all are here. We wanna go ahead and
start our first session on understanding the temporal relationship between changes in blood
pressure and changes in risk. This session will include a presentation from Dr. Michael Weber who’s
a professor of medicine at State University of New York, Downstate College of Medicine. He’s gonna provide an overview
of the temporal relationship between changes in blood
pressure and changes in risk, which is, as you just heard, an issue that’s very relevant to
the discussion today. Then we’re gonna have a panel discussion related to this topic. And I’d like to go ahead
and ask the panelists to come on up so we
can minimize disruption as we get through this session and I’ll introduce you
while you’re on your way up. William Cushman’s the Chief of
Preventative Medicine Section at the Memphis Veterans
Affairs Medical Center and Professor of Preventative Medicine, Medicine and Physiology at the University of Tennessee
Health Science Center. Patrick Twomey is Medical Director in the Oncology branch of Licensing and Early Development
Safety Science at Genentech. And Hector Ventura is the section head of advanced heart failure
and heart transplant at the John Ochsner Heart
and Vascular Institute. And so, Dr. Weber is going to kick us off. Thanks, Michael. – Well thank you, Mark,
and a real pleasure to be part of these deliberations today. And as you heard, I’m going
to give a very quick overview of some of the issues that we face in trying to link
changes in blood pressure during treatment of
non-cardiovascular diseases that might have important
cardiovascular consequences. Oh. I’m sure this button works eventually. I don’t know if someone is– – [Mark] Need a little help? – I seem to have gotten… Do we have a computer
where these slides are on? – [Mark] The computer
seems to have gotten off of slideshow mode. – Looks like a–
– Just a minute. – Of course there could be
a new way of presenting, put all your slides up at once and say there it is and sit down. (Mark laughs) Okay, so here we go. We all know that
epidemiology has taught us, very clearly, there’s
a strong relationship between levels of blood pressure and cardiovascular outcomes. And these are data from Lewington, based on a million
people in placebo groups followed for several years. And you can see regardless of age, blood pressure’s a powerful
determinant of outcomes and a 20 millimeter of mercury difference in systolic pressure doubles
the risk of major events. The problem with epidemiology,
fascinating as it is, is that it doesn’t tell you what happens if you do something to
change the blood pressure. And that’s why we need to study each drug and each situation separately. And to make things as simple as possible, I’ve tried to divide what happens when you use drugs that may have off-target blood pressure effects. I’ve divided into three
temporal categories. The first is the obvious connection, drugs that raise the blood
pressure dramatically and where we would expect
major bad things to happen within days or weeks. Then you have the second category where you have modest changes in pressure, three, four millimeters of mercury. Does that matter over a
period of weeks or months? How long does it take
to see bad events occur? And then finally, the most
difficult and frustrating, what do you do with drugs
that only occasionally, in some vulnerable people,
might raise blood pressure or may only raise blood pressure slightly but in a lot of people. So let’s have a look a little more closely at these things. And the first category is the drugs that have big effects and that should lead to early cardiovascular events. And the best example, I’ve just chosen a few
examples along the way, of all of these categories is Avastin, the drug that’s used for
treating solid tumor cancer. Very effective. It’s a VEGF inhibitor. And it raises a complication
right from the beginning because yes, it does raise the
blood pressure dramatically, but it also has other
cardiovascular effects. It destroys vascular endothelium, it causes tissue ischemia, it
has all sorts of toxic effects separate from blood pressure. So what’s hypertension or
changes in blood pressure, what are the factors? No question that it has dramatic results. And here is a meta-analysis showing that it increases by three or fourfold the probability of
patients having strokes. And we all know that,
it’s a well-worn story with the anti-cancer drugs like Avastin, and cerebral hemorrhage, again, dramatic increase in probability. No big surprise. In fact, a very nice
article from Mayo Clinic, just their own in-house data, showed that they were
able to report 10 strokes, all within three weeks
of starting therapy. Seven of the 10 did have
a history of hypertension. So people with hypertension
were particularly vulnerable and nine out of the 10 patients had severe hypertension
soon after starting the drug and most of them died. So, no question, this is an easy problem to at least understand. What to do about it is another story, but at least we don’t
have to spend too long arguing what to do or how to
measure the blood pressure. What about category two? Blood pressure effects sufficient
to be measurable perhaps, particularly in a cohort,
and we can, obviously, note the cardiovascular events if we’re doing large clinical trials. And these are events we expect to see within weeks or months. Perhaps the best known of these studies, because it helped actually
the cardiorenal division approve two drug combinations for initiating the therapy
for hypertension was VALUE, because VALUE was a trial where two drugs were being compared. There was a difference in
the blood pressure reductions between the two drugs, and within weeks there was a difference in
cardiovascular outcomes. Here are the blood pressure data and on the top you can see that Amlodipine in brown
was a little more effective than Valsartan, the angiotensin
receptor blocker in blue, particularly in the first two
or three months of treatments. And in the bottom you see
that yellow line shows indeed there was about a four
millimeter of mercury difference just in the first few weeks and then the difference
got progressively less. And the result is shown
here and you can see that during that early three month period when there was that three to four millimeter of mercury difference there was a significant difference between the drugs in major cardiac events. So it didn’t take long for a four millimeter
of mercury difference, in this large cohort, there
were several thousand patients in each group, to demonstrate
a cardiovascular effect. So this was very crisp and clean and the FDA was impressed by this and agreed that getting
treatment started effectively and quickly was a high priority and for the first time, they
approved two drug combinations as initial therapy for hypertension. A nice piece of history there. ALLHAT, another study well-known to most of my colleagues in this room, comparing three different drugs to see if there was any difference in cardiovascular outcomes. And the most interesting part, in a way, was the stroke outcome, when Lisinopril, an ACE inhibitor was
compared with Chlorthalidone, a powerful diuretic. And you can see, overall a benefit to Chlorthalidone, a lower stroke rate. But when you look at the details and compare black and non-black patients, you can see the non-black patients had identical stroke
effects regardless of drug, but the African American patients in whom a drug like Lisinopril
doesn’t work very well, in fact, it’s in the
label for ACE inhibitors that they do not work as
well in black patients as in others, you can see there was a
40% excess stroke rate based on a four millimeter
of mercury difference between the effects of the drugs. So here again, four
millimeters of mercury, huge impact on a major
event, in this case, stroke. And the last thing I want to show in this category of three or
four millimeters of mercury is the SGLT2 drug Empagliflozin. As you know, new diabetes
drugs now have to go through a cardiovascular safety study to demonstrate they do not cause an excess of cardiovascular events. Well, it turned out and
this is no surprise, Empagliflozin did reduce blood pressure by three to four millimeters
of mercury throughout the trial and that began very early in the trial. How exactly it lowers blood pressure, other than perhaps getting rid of sodium, isn’t clear and it’s a fascinating story. But here we have a three or four millimeter of mercury difference in blood pressure almost immediately and major cardiovascular effects, beneficial to the drug in this case. I just want to point out
that the effects on death, which was probably the most
dramatic result of the trial and on heart failure was seen
within the first few months. So again, three to four
millimeters of mercury quickly gives you a powerful result. I don’t want to dwell on
this but Empagliflozin obviously doesn’t just
change blood pressure, it has effects on lipids,
it has effects obviously on glucose, it has effects on uric acid and other metabolic factors. Is it all blood pressure that explains the cardiovascular outcomes? Not clear. And finally, category
three, the most difficult. And minor changes in blood pressure, obviously we don’t measure
them in clinical practice. Is ABPM the answer? And, we’re going to talk
about NSAIDs a little later, where Jefferey Borer is
going to be talking about it and it’s a fascinating area. So let me talk about
another drug, Mirabegron, which is a drug used
for overactive bladder. It’s a beta-3 agonist,
but it has a little bit of beta-1 agonist effect as well, and in young, healthy volunteers, it raised blood pressure about
three millimeters of mercury. In older people, the people who usually will take a drug like this, it didn’t raise the blood pressure as much because as we get older, we
tend to lose our beta receptors to some extent. So how do we figure out is
there a blood pressure problem and what can we do about it? Well, this is ABPM and this is a study that I did with Billy White with the people who
manufacture Mirabegron. We have a beautiful placebo
group of 80 patients and two doses of Mirabegron. And you can see, no
difference from baseline to I think it was 12 weeks of treatment with between placebo and Mirabegron. A bit of a surprise. I thought there’d be a
little bit of a difference, but you can see what doesn’t help at all is the big, standard deviations, even when you have 80 people in a group. Well, we thought let’s
get a little more specific and look at it hour by hour. Could it be that for the day
as a whole you don’t see much, getting back to what
Ellis was say early on, but maybe four or five hours
after the drug is taken, when you get the maximum
plasma concentration, there you might see
some troubling effects. And these data are all starting from the time of dosing
and you can see placebo, lower dose, and higher dose of Mirabegron, nothing really there. Frustrating. That fall in blood pressure
in placebo follows lunch, in fact, that’s the postprandial
fall in blood pressure. You don’t see it with Mirabegron. That’s the only thing
I could find (laughs) that would discriminate. And, very difficult, isn’t it? So we went to, let’s
go to something cruder. Let’s just go to outliers. How many people using ambulatory
blood pressure monitoring had an increase of at least
15 millimeters of mercury in their systolic pressure? Placebo, 11%, which is amazing. Isn’t it? Because if placebo didn’t
change pressure much, it meant that at least the 11% of people had a big increase in blood pressure, there had to be another 11% who had a big decrease in
blood pressure to offset it. And this is the frustration of ABPM. For a cohort, there’s almost
no change with placebo. But for individual patients
all over the place, Mirabegron absolutely not
different from placebo. So, that was good news for
the manufacturer of Mirabegron and they got approval for their product and for some of their combinations. These are data I have shown before and Ray Townsend was the
first author of this paper and I don’t wanna waste time
telling you the whole origin of this study, but those are individual
patients in a placebo group subjected to ambulatory
blood pressure monitoring. All over the darn place. Up and down. Almost nobody stays exactly the same. The cohort as a whole, no change. It was a wonderful success
from the point of view of a clinical trial because it allowed us to compare it with an intervention and show a significant difference. But for individual patients,
very, very difficult. So, that’s the story and in a sense, the problem we’re facing. So, it’s easy with severe
blood pressure changes like with Avastin, we
know what to do there. Not so easy with the NSAIDs, which we’ll hear about in a moment. Not so easy with other drugs
that perhaps have three to four millimeter of mercury effects, and very difficult for
minor blood pressure effects in large numbers of people, because as we’ve heard before, and Norm Stockbridge and Bob Temple have made this point many times, from a public health point of view, if you’ve got hundreds of thousands or millions of people taking a drug and it only has a half
millimeter of mercury effect, that’s gonna translate into many people having strokes or heart attacks. So how do you find who those people are? And we have to remember that patients do have their own risk factors which make them more vulnerable to a drug and also remember that drugs don’t just change blood pressure, there are other cardiovascular
effects as well. Let me stop there. Thank you, Mark.
– Thank you very much. (audience clapping) Thanks very much, Dr. Weber. So, we’re now gonna turn to the panelists for some initial reactions
and further thoughts about this topic, and we’ll
start with Dr. Cushman. – I think I have some slides that I’d like to go
through pretty quickly. So, Michael has done a very nice job and I’ll go through this quickly. I wanna say that I’m coming at this from the perspective of a
hypertension clinical trialist. I’ve been on leadership of trials like the ACCORD blood
pressure and ACCORD trial and SPRINT trial and others. But I’m not a statistician, and so look at it from that perspective. I do think, I believe that any drug that chronically raises blood pressure or chronically lowers blood pressure, depending on what other off-target effects or other effects it may have,
is likely to change the risk. I do wanna point out that
we’ve seen over and over again in trials that when you,
and this was illustrated in the VALUE trial and the ALLHAT trial is that even if you are
trying to treat everybody to the same goal, you
can’t totally overcome, at least in populations, a
difference in blood pressure. And that’s something to keep in mind. So… Yeah, and I think pretty
much I’ve said that for the second thing, other than to say that we can’t always
explain the differences in blood pressure. It may not be things that we can measure, it may be how long the
hypertension was going on, it may be adherence, and we see that a lot and certainly non-adherers do worse in terms of outcomes as
well as blood pressure. So we have to keep that in mind. In most randomized controlled trials of blood pressure
lowering, event differences often begin to appear
over one to two years. And most of these trials
we’re putting thousands and thousands of patients
into those trials with an expected outcome within
maybe three to five years, except for the most severe
types of situations. And I wanna quickly go
through some Kaplan-Meier or similar kind of curves to
illustrate the time course. This is one of the most extreme examples. This was the VA study. We don’t have a Kaplan-Meier curve for the 115 to 129 group. They were stopped after about 18 months, but there wasn’t a DSMB
and it wasn’t looked at. So, who knows? It might’ve been stopped at six months. But clearly, for anybody with
that high of a blood pressure, with huge differences in blood pressure, and for example, even the mild
to moderate diastolic group, the differences were 31
over 19 at four months. So huge differences leading
to marked differences in outcomes, and again, this
was stopped at 3.3 years and was looked at somewhat earlier. But it’s a pretty small study, so clearly, big differences
when you have big differences in blood pressure, as Mike has alluded to. This HDFP study looking at mortality, and this just starts to
illustrate that it really, and notice that these are,
this is not 100% access. So these are very small percentages even though you start seeing differences after one to two years, but
HDFP was not stopped early. It went to 4 1/2 years. Here’s the SHEP study, after a year is when the
stroke started separating. Their other studies, I’ll
just look at briefly, where it took even a lot longer than that to see differences. Clearly though, it takes a while to see some of these outcome differences. Here’s from the HYVET study and again, if you look at the stroke over there, it’s somewhat after a year where
the curve starts separating and did go on to its four or five year, I’m sorry, this particular study was stopped at about two years average
follow-up, median follow-up, but it took a very long time
to recruit the patients. So, there were a lot of patients that were in for a lot longer, which is why you see
that the events mostly were separating beyond two years. I did wanna point out that
there are some outcomes that there are differences
and how quickly the outcomes seem to show up in hypertension studies. Heart failure, we have several examples where after about six months, and sometimes even within six months, we see differences in
heart failure outcomes. Whereas for stroke, it
may take a lot longer for coronary events, it
may take even longer. So, here’s the SPRINT Primary
Outcome, comparing 120 to 140, about a 14 millimeter
of mercury difference in blood pressure, and the
primary cardiovascular outcome, we started to see separation
after about a year. The O’Brien-Fleming boundary
started being exceeded, suggesting significance
at about two years. And then ultimately when
the study was stopped, the average or median
follow-up was 3.26 years. And then here’s the all-cause mortality. One of the outcomes that not unexpectedly takes a little bit longer
when there is a difference to see a difference and
this also was significant, and the curve started separating
after about two years. Here’s another example. The heart failure, this is in SPRINT, where you see the curves start
separating after six months. Now these are still fairly small. They become highly significant, but notice that they’re
pretty small differences in percentages and so that’s something that does have to be considered. This compares SPRINT to the
standard glycemia arm in ACCORD where the outcomes were virtually similar in terms of a 26% reduction and again, you’re seeing the separation in curves after about a year or so. I did wanna point out, in ACCORD we did see a significant
reduction in stroke in the blood pressure
study but nothing else. Again, fairly low event rates. But it took over three years
to see that separation. We did not see a significant
difference in SPRINT and maybe that’s because
we had to stop that study before that time period. HOPE-3 I wanted to end with as an example to give us some caution. And that is, this is a very large study. Over 12,000, almost 13,000 individuals, given somewhat modest doses
of blood pressure medications, getting about a six over three
millimeter mercury difference and over quite a few years,
did not show a difference. I think most of us would expect that over time, and certainly
the investigators thought there would be a difference, but I think when you get even this small, but this much of a
difference, six over three, if you’re looking at a drug
and expecting a company for example, to do a long-term trial, this is a very large study. Didn’t happen to show a difference, and therefore, you might
miss what might really be an important difference. And I think that either
a higher risk population or a bigger difference in blood pressure or longer follow-up, I think most of us would expect that this
would have eventually been significantly different. But I think it raises a caution about even having a very large
study showing, if you will, non-inferiority to a drug that
does raise blood pressure. I’ll stop there with those comments. – Thank you. (audience clapping) Thanks, Dr. Cushman. Next is Dr. Twomey. – Up or stay here?
– Whichever you’d rather. – I may just stay, do it from here. I don’t have any slides so,
(laughs) it’s much easier. Thank you, Dr. Cushman. I think that’s a great overview of many of the studies
that I’ve been looking at in preparation for this. I think from my background,
more in the oncology setting and in the safety setting, I think more what I will bring up today is just more of another facet of this, more of another question that I think of as a safety science officer, as we approach these
types of side effects, is not just the temporal aspect but is there something of
the mechanism of action for how blood pressure, for how the hypertension comes around that could be influencing this? Is there a difference between a medicine that might cause a temporal
increase in blood pressure because of an interaction
with nitric oxide, receptor blocking, not nitric oxide? Or is there something that’s impacting the renin-angiotensin-aldosterone system that may actually have
much more lasting effects, even if the blood pressure is only raised for a short period of time? So those types of questions are things that we tend to look at
from the safety science way. And how are these, how can
we think of the mechanism for how it might be
increasing blood pressure and what can we do to either mitigate that or to be able to predict
that it might happen in certain patients and
potentially address it there with new drugs? But I think from my standpoint, that’s what I would
offer for the discussion. – Great. Thanks very much, Dr. Twomey, thank you. (audience clapping) – Next is Dr. Ventura. – It’ll be brief for me. The only thing I wanted
to add in regards to this, I think it’s a very
important clinical question, at least like here in the clinic, every time somebody would ask, “If I take this medicine, “what is gonna happen
to my blood pressure?” You know, I was one of them. In the ’80s I didn’t pay attention to it. I said, “Don’t worry
about it, just take it.” Well, might not be right. But then it came out to
the world of cyclosporine and Prograf because of my
work in heart transplants. And then I realized what I was
saying before, it was wrong. You can’t stop cyclosporine,
you can’t stop Prograf. You can’t. Now I didn’t become a nephrologist, George knows that quite
well, but you know, the nephrotoxicity of
the drugs are important and how do you mitigate those problems in patients that they
have to take it forever? Now, I take away cyclosporine at least, I’ll say 100% of people
that take cyclosporine they have hypertension, it
seems that 110%, they had it. Then Prograf will lower it a little bit, but still, the effects are important. You can’t stop it. And think about this. If you look at the transplant world, more likely the people
that die were long-term, heart transplants, at
least that’s what I do. They tend to die of problems that are not related to the heart. Right? They’re related to blood pressure, renal failure, so on and so on. Obviously in the 1980s, we
didn’t pay enough attention because people didn’t live
20 years post-transplant. But now we have to pay
attention more and more. So, I think it’s an important question. One or two millimeters,
and compounded to this, the people that have a transplant and they have cyclosporine,
well not anymore, Prograf at least, they also take NSAIDs because they have pain. So, you know, (laughs) I
thought for a while in the ’80s that was not an important question and then the guidance is great because it’s a very common
question patients ask. And you cannot treat cancer
because of blood pressure, so it’s a very interesting question. And I’m glad now, I have to mention that Michael is knowledgeable
about this, it’s difficult, but in cyclosporine,
Prograf, you know, seriously, you go to the clinic and you take, most of them take a calcium channel block and most of them take an ACE inhibitor, most of them take everything. So, that is a, my contribution, a small contribution to this. I think it’s an important question, it’s here to stay, it’s not gonna go away, I know, I know all of you
know what I just said, and last but not least, I wanted to thank for the invitation to this meeting. – All right, thank you all
for your opening comments. (audience clapping) Let me first ask the panel
if they’ve got any reactions to what they’ve heard this morning so far from the other panelists or otherwise? Any additional points that
you would like to make about this topic of the
change in blood pressure, change in the risk relationship? Seems like the–
– I think. – Go ahead. – George, did you have a? – And then yes, that was the next thing is to open this up to comments from all of you who are here. So just put your hand up or something so I know to call on you
if you’re in the room. If you are not in the room
but have a comment for us, again, you can send any
questions or comments to [email protected], [email protected] And, George, over to you. – [George] So, thank you very
much for a nice overview. I wanna add another perspective to this, not that there’s not enough
perspectives on this. But I think it’s important to understand that one size does not fit all. So if you come up and say, you know, anything greater than two
millimeters, your risk goes up, I think is a mistake because that’s naive. And Bill did a beautiful
example by showing HOPE-3, which is what I’ve made
the point of as well. So I think we need to
look at this in terms of what is your established, absolute risk for cardiovascular events,
day one, when you’re starting? If you’re at very high risk, then maybe two or three
millimeters is what you need. If you’re at very low
risk, clearly it’s not. Or it may take you a lot
longer to get to that point. I think that’s an important
perspective that we’ve missed in a lot of these data analyses. So I would actually go back and reiterate the latest
blood pressure guidelines that say it’s not about the
numbers as much as it is about what your risk is
when you’re starting. And I think that’s really a key point. – I thought the fascinating
thing about HOPE-3, and as Bill pointed out, they
did not find a difference between placebo and active treatment, was they did have a
pre-specified subgroup analysis. And people who were truly hypertensive did get a big benefit, and people who had low blood
pressures to begin with actually were trending the wrong way and were having an adverse
effect from the treatment, which I really don’t understand. I have trouble with that. And in fact, sometimes
these large, simple trials are perhaps a bit too simple. – [Mark] Yes, over here. (Hector laughing) – That’s right. – There is a button somewhere there. – Ask a little question. – No. (laughs)
– No. – [Jeff] Hello? – There we go.
– There ya go. – [Jeff] Yeah, no, hi. Jeff Heilbraun, Bioclinica. I have to agree with what
Dr. Bakris just said, and I think when we’re looking at risk, not only is it not one size fits all, there is an immediate risk concept based upon the therapeutic
area that you’re treating and then there’s the long-term. So as Dr. Twomey talked
about your area’s oncology, if you start to look at the
different patient populations that you’re working
with, whether it’s adult, adolescent, or pediatric,
it’s the long-term risk that can be associated with that increase in blood pressure. Not so much hypertension,
but the actual increase in blood pressure and
looking long-term downfield so you have to look at
both the population, the therapeutic indication
as part of the risk profile when you’re assessing what level of risk and what level of risk
benefit should we accept. And what therapeutically can we do to treat that and we know that
coming from the oncology area that there is a lot of
prophylactic work that can be done in cardiac safety when we’re
dealing with oncology drugs. So, just wanted to add
that to the discussion. – All right, I see a lot of nodding and it seems like a comment
in a similar direction. Further comments? – (stammers) You’re right. A change in blood pressure does not have to make you hypertensive
to be potentially dangerous if there are other risk factors going on. Nevertheless, when you start
at a high blood pressure, even modest changes are gonna
have a relatively large effect because you’re in a region now where absolute risk is high. So, those are patients, again, I think everyone’s agreeing on, people at high risk obviously
are gonna have more events and high risk includes
being hypertensive already, being old is another bad thing to be, people with renal insufficiency. You can make this list,
having diabetes and so forth, and so, I think we can define the people who you want to worry about and I’m sure that comes into the thinking at the agency when they figure out how big a risk is and how they should put out warnings. – [Mark] I see a couple more
comments here and then there. Go ahead. – [William] I think we
are a little too quick to, I think there’d been a failure
to make a distinction here between elevations in blood
pressure caused by drugs, and elevation in blood pressure caused by a variety of factors, exercise, excitement, what have you,
and essential hypertension. And the overwhelming amount
of evidence we’re showing here is if you treat essential hypertension, you get various benefits. But, it’s not the same thing. But drug-induced blood pressure rises are not the same thing. In many ways, hypertension’s a biomarker for things like atherosclerosis. And one of the reasons that
we see in some studies, for example, well, you see the HOPE study where the people who had
high blood pressure benefited but not necessarily the people
with lower blood pressure from lowering blood pressure. You’re just picking out the
people who have this disease of essential hypertension and, you know, there are other studies
where you only see benefits, for example, in stroke, and
maybe just hemorrhagic stroke. And you don’t see benefits
in things as clearly or as much benefit in reduction in terms of myocardial infarction
or ischemic stroke or thromboembolic events
from the carotid arteries. And I think that’s
important to keep in mind and because the numbers were there, and we can speak to numbers,
and we can measure numbers, and we can treat numbers, I think we make the assumption that it’s all about the numbers. – Yes? – Two quick comments. One is to, don’t know
why, but it turned out that in SPRINT it was almost the reverse in terms of the baseline
blood pressure relationship to benefit, and so, and I don’t
know why that’s different. It would be nice if the whole picture was, you could put it together and say gee, if you’re normal tensive
then it doesn’t matter, but I don’t think we can quite say that. Obviously it’s a different
population than SPRINT, but I think some things may
have happened by chance, which happens in various studies that we can’t really explain. The other thing is that, and I’m sure we’ll discuss this further, I mean, on the one hand, to do a study to exonerate and show safety for a drug, then you probably need a
very high risk population because those are people that
probably are gonna be treated long-term with that drug. And to do a feasible trial, to be short enough and adequate size, you need to do it in a
high risk population. I think many of us though believe that if you started making changes of a few millimeters
of mercury in children and they were on some medicine
for decades and decades that were causing that
blood pressure difference, that’s very likely to
have a profound effect on risk long-term. It’s a hard thing to study,
other than observationally. But I do think you have to
keep those two things in mind in terms of who’s gonna get it. And I’m sure you all have
talked extensively about this, who’s gonna get it and how
long is gonna be taken. And certainly drugs that are approved based on a few years or five year trials often are taken for decades. And we don’t always know the risk. – Questions? Yeah, over there. – [Philip] Excellent discussion. Philip Sager, Stanford University. This session dealt with
temporal relationships. I’d like to try to maybe
bring back the panel and get a little more feedback on that. I mean, I took from the
presentations and the data that a drug that increases blood pressure a relatively small amount, except maybe in a very high risk population, taken for maybe a few weeks
probably isn’t gonna have any major effect. I mean, it seems like it
takes really a period of time, even in relatively high risk populations. Obviously something that has profound blood
pressure increases, that’ll be a different story, but I mean, I think this is
kinda important to talk about as we think about where
are we gonna put our energy in terms of evaluating drugs? A five day course of antibiotics that increases blood pressure a little bit probably isn’t gonna be a
major public health issue. So I think it’d be great to just get the panel’s input on this. – Yeah, sure, I agree. Short-term treatment is one issue, and also there were some drugs, some of the anti-depressants,
figure things like ketamine, that raise blood pressure quite sharply but then it comes back down again within 30 to 60 minutes. Does that matter? Or do you say it went up 20, 30 points, divide that by 24, that’s an average of one millimeter of mercury an hour? I mean, and I think Ellis brought that up, and I think that’s. – [Mark] Other comments? – Yeah, I would definitely agree that there are differences there. I think just from the
oncology world as well, there’s, it still is important, I think, to recognize which drugs may
even have those small increases even if they are given,
I mean, with our drugs, I think much like Dr.
Ventura was talking about, with oncology drugs these
patients do have to take them regardless of what the
side effects might be and understanding them helps us mitigate potential downstream effects. But, what’s important to recognize is that the changes in blood pressure,
if they’re occurring, how do I word this? (laughs) If they’re occurring over
a period of time that, I had the thought in my mind (laughs) and I’m trying to think about how to, how to actually word it correctly. But, I guess, the point
really just remains that there are differences between drugs that might just increase blood pressure for a short period of time versus ones that patient will be on chronically, or for an extended period of time, several months that may have
an impact down the road. I would agree. – I wanted to make the comment
about the anti-depressants, which is the most common
medication prescribed in the United States. I don’t know about you,
if you see patients, but they’re taking 15 medicines but one of them is a
statin and the second one is an anti-depressant. And so the anti-depressants had a question that had been asked several times. And what Michael just
said in regards to that, you know, there’s some, a few studies, small studies I guess, the one
I remember was three months and there’s no change in blood pressure with the anti-depressants,
so it might increase it and then they make it the same. I think it was randomized
in a psychiatric population. And the last thing I wanna
say in regard to the risk, for example, if you take people
on Prograf or cyclosporine, and you have the same age,
one they said is non-ischemic versus ischemic, the
progression is the same. Though as far as I know, at least, again, it may be that you have in the people that have vascular disease they’re worse, but you know, it’s very common to see these things and we make assumptions that are not, for example, nobody
knows what blood pressure in a transplant patient should be. 140 over 90, 130 over 80, you know. So there’s a lot of
assumptions that we make and it’s a very important question. The more I hear the conversation, the more I read the whole thing, it’s a very important
question that we will have. Sometimes we don’t have a clue. Regardless of the
trials, I’m talking about from the clinical point of view. – Thank you. Oh, one more comment? – Yeah, just a quick addition. I think we’re gonna get to this later but I did wanna point
out that all those trials that I showed in blood
pressure differences, pretty much most of them, if not all, used very standardized methodology for measuring blood pressure even though they were clinic
blood pressures, not ABPM. Some of them had ABPM sub-studies, but all of the eligibility
and intervention was based on well-done
clinic blood pressures. Whether that’s the ideal or
not, it certainly though, we know that if the blood
pressure’s taken sloppily that it’s really difficult
to know what that means. – And we are gonna get to
that issue to that issue in a later session. So, over here and then, I
see ya, and more to come. Great discussion. Thank you all. Button on the bottom, that one. (laughs) (person speaking of mic) – There ya go.
– I guess that works. Hi, Bob Kleiman, ERT. There’s a temptation to
sort of simplify this and view it as akin to the problem of drug-induced delays in
repolarization and sudden death, what we all measure as QT. And we’ve learned that QT is a biomarker that has some weaknesses and I think we’re not really discussing very much that blood pressure is a biomarker and is a surrogate marker for the things that we do care about,
which are renal failure, an MI, and stroke, and death. But, you know, people have said a little bit about the time course as an hour a day versus a mean of 24 hour blood
pressure increase different. There are also very large differences in the way drugs may have
effects on blood pressure. And I’m wondering if any of you know if there’s any data that would show that of an X millimeter
increase in blood pressure, independent of mechanism, has
the same effect on outcome or is there any evidence of the opposite? That with different mechanisms,
same blood pressure increase has very different results. Is there any data? – Well, some people will
justify the innocence of an increase in probably
by saying golly, you know, every time you go out and
exercise, you go to the gym and workout hard on a
treadmill or an elliptical for 40 or 50 minutes, your pressure, your systolic pressure’s going up 30, 40 millimeters of mercury, that
doesn’t seem to do any harm. But of course, it’s a
totally different mechanism of raising blood pressure. It’s an increase in cardiac output, it’s driven by tachycardia and so forth. It’s basically dependent on epinephrine, whereas I think most of
the drugs we worry about are raising blood pressure
through vasoconstriction, more of the norepinephrine kind of effect. And as far as I know exercise is harmless. I hope so. (panelists laughing) – [Mark] Um, yeah, over here. – I put in my 40 minutes
this morning (laughs) in the hotel. – [Vasilios] Papademetriou
of Washington, DC at the VA in Georgetown. I just wanted to comment a little bit about the ground rules on
how to evaluate patients who are with hypertension and what kind of blood pressure we need
to consider in the long run to determine whether it has
adverse effects on outcomes. We talk about the blood pressure of two millimeter, five
millimeter increase and having detrimental effects and that’s probably true
for large populations and large studies. However, the important thing
is for the individual patient that we need to consider how much change in the blood pressure do we need to see in order to worry that
has an effect on outcomes. We know from many, many studies we’ve done through the years, even the SPRINT study, the
patients have different pressures at different days, even
without changing their regimen. I mean, I looked at several
patients in the SPRINT study, without changing their regimen, in one visit they were
130, the other were 145, and the third they were 128. So which pressure should we consider to determine whether we should worry about that pressure that day? And, the second thing is, which pressure should we consider? The systolic or diastolic? Both? The average? How should we measure the pressure? Should we measure with osculatory methods, with the automated device,
with the ABPM at home? How? Which pressure? I think we need to
consider all those things before we decide what
we consider important and what we don’t. We know from all these
studies that have been shown and we have a lot more to show that changes in population of
five millimeters in systolic or three millimeters in diastolic can translate into
change in 15% of strokes. But in the individual patient, how many pressures do we have
to see elevated to worry? I did a little survey in
our large VA population of 3 1/2 million patients
and we tried to determine the effect of blood pressure
changes on outcomes. Considering the last
pressure in the records, or the last one before they have an event, a stroke or a heart
attack, or before they die, and we found that there
was a big difference in the determination of the outcomes if we consider the last pressure
versus the average pressure over a 10 year period. And the average is a much
more stronger predictor of mortality and cardiovascular events and I think we should
take this into account when we are deciding how to
evaluate the blood pressure of our patients and what
to consider important. – Very nice comment. And we are gonna come back to this issue later on during the day too. Thank you for teeing it up. Over here. Okay, so I have, maybe the remaining hands,
so Bob here and there, and there. So okay, I’ll go this way
and we got about 10 minutes before a break so try to
get in these four comments. This is your first cognitive
test of the morning. (panelists laughing) – Too early in the morning
to make a microphone work. – [Bob] Just following up on what Dr. Weber was talking about. I don’t think these are
mostly individual decisions about whether your blood pressure went up four millimeters of mercury or three, or what it’s like in
the morning and stuff. What I think you’re talking
about is populations. We’re talking about long-term outcomes. This isn’t gonna be acute. And, it’s sort of been shown if a drug like ibuprofen, say, that you take for five years,
raises your blood pressure by four millimeters of mercury, that had a consequence,
at least in the one study where that was done. And two similar drugs with many similar pharmacologic properties that didn’t raise the blood
pressure didn’t do that. We don’t know how they compare to placebo ’cause there was no
placebo or a path study. I think mostly what you’re interested in is looking at the effects of drugs that are going to be used chronically, because anything used acutely isn’t gonna, a modest effect isn’t
gonna change anything, but you really do need to know what to do about drugs like NSAIDs that raise your blood pressure and seems to me the obvious thing is start somebody on a
low dose of a diuretic to control that. The other thing is, the
early curves you showed show that depending on where you start, there’s a huge difference in
how bad the place you get to is but it’s worsened at all levels. And for a drug that’s
gonna be used for 20 years, maybe that matters. I mean, that’s not trivial. It’s true that the effect
in any one year is nothing, but over the course of
time for a chronic drug, these things matter. Especially since it’s not that hard to take care of a small increase. There are benign anti-hypertensives and things like that and so on. It’s not the individual
case so much, I don’t think. It’s the chronic effect
of these long-term drugs. – If I may, isn’t that the question that we’re trying to address though, is when is that cutoff for
an acute treatment period versus a chronic treatment period? Is it, do we all feel comfortable enough to say that if you’re on a
treatment for three to six months and it increases your
blood pressure X amount, that that’s okay, if especially, if we can put you on other
drugs to mitigate that? Versus if you’re going to be on something for years and years? I think we can all agree that yes, for the chronic drugs that
might raise blood pressure, of course, but that’s the question is, at what point does having
your blood pressure raised for a certain amount of time impact mechanisms further downstream? If you’re affecting the
renin-angiotensin-aldosterone system for several months, that
might have an impact even when you stop the drug, drive you into further
blood pressure issues. I think that’s the question
that we’re trying to address. – [Bob] It would have a further effect because of something else it did, not the blood pressure effect. – True. – [Bob] But that’s a
perfectly good question. I mean, we sort of know that raising blood pressure a little bit for a couple of weeks
isn’t gonna do anything, we sort of know that. If you raise it by 20 or 30, maybe. But you’re right. I mean, that is an important question everybody has to talk about,
but I do think we have to worry and I have to say I don’t think we worried enough about the NSAIDs and what they did and how they did it, in retrospect. It’s easy to get smart after somebody does a five year study. But that’s sort of a message. Drugs with modest but persistent
effects on blood pressure need to have attention paid to them. Even people who don’t get up to 150, just going to 140 turns out to be bad. We know that from the curves you showed. In fact, one of my questions is whether those curve suggest
that most people ought to be pushed down a little bit. Maybe 130’s not so great
either, over 20 years. – Yeah, so great points. I wanna try to get in three more comments, then here, there, and over there. We are gonna have some
more time later in the day, well, we’re gonna keep coming back to some of these same issues
that have been brought up here and allot some time later in the day to get in other comments too. Really glad to see the dialog here though. Please go ahead. – [Karen] Good morning. Karen Hicks from the Division of Cardiovascular and Renal Products. Dr. Weber, I’m so glad you brought up the example that you brought up, which is really the fourth scenario, where a drug product
could be given chronically but it might only be administered
once or twice a week, yet there are changes in blood pressure, on average, of over 10
millimeters of mercury that last for a number of hours. And, by the way, also
there’s a lot of outliers with systolic blood
pressures greater than 180. And so, I hope that today we
can hear a little bit more from the panel members about what to do with these individuals as well. – Good point. – Thank you. – Me?
– No. – Not yet. – No you can’t. – [Frederick] Can you hear me? – Yeah.
– Okay. I’m Dr. Frederick Sannajust from Merck Research
Laboratory in West Point. I’m more a practical pharmacologist
but I would like to say we are here to understand that we define a threshold in systolic or
diastolic blood pressure effect of any drugs, but we are looking
at the integrative system and for example, ursine,
pre-clinical, injured pig, dog, or monkey temporary studies, as well as human, we
have the technology now. Why we don’t consider outright viability? Why we don’t look at circadian rhythm because people that are
less deeper at night for me are more at risk. You know, we are testing drug but it is in this drug that are canceling the circadian rhythm of blood pressure, systolic or diastolic, and
that I am very concerned. Secondly, you could look at the bioeffects by measuring Finapres
and doing some tests. That can give you a better predictivity of what could be the impact, whatever the population of patient. You can have young patient, can older one. So if you integrate systolic
measurement reliably, diastolic, heart rate, and you look at heart rate variability, you got software that can do that easily, you can look at bioeffects too, extracted, I think we need to integrate that to have a good index of predictivity. I’m surprised that we are not considering this integrative platform
as a better tool. I don’t know what’s the–
– Integrate. – [Frederick] Thinking
of the panel about that. I know it’s more work, more expenses when you have some 150 phase I study and you have to that, it costs a lot, but I think we should consider
a more integrative approach and a very predictional approach. – Thank you for bringing up that even more integrative approach. I’m gonna try to get in a
couple more comments here. Please, go ahead. – [Peter] Peter Barker
at VA Medical Affairs. One element I think Dr.
Weber showed very nicely was this placebo circadian curve. And I think an impact on, for example, blood pressure drops at night could also have a
significant effect over time if that doesn’t occur, versus just measuring peak levels. So I think the overall
circadian’s component, I think, could be a very important element and also the measurements from patients who are normal tensive or
have no high blood pressure versus those who start a
study with hypertension, I think there also needs
to be a differentiation between those different populations. And I think we need to be more
differentiated on that end. – Thank you. Was there another, yeah,
did you have a comment? – I will say if the
question will be simple, we will be here, because, you know, while we’re sitting here, I
think you mentioned the point about QT and the simplicity. It’s not as simple. If it would be simple,
we wouldn’t be talking here in Washington, DC
today on February 4th. And so, yeah, it’s complicated. Let me say one more thing. A heart transplant for example, you put a 45 year old heart
in a 65 year old vasculature, so it’s essentially called,
that is a, you know, there’s a lot of interactions. Blood pressure might be related to that. Who knows? I mean, so yes, I think I like the concept that it’s not a simple concept. (panelists laughing) – [Vasilios] The comment I will make is on the short increases of blood pressure versus the long-term and
sustained, sustained increases. I think the short-term increases probably are benign and
they do not have any effect on cardiovascular morbidity and mortality and as an example I wanted
to offer the exercise. Patients who exercise or
individuals who exercise, they always have increased blood pressure. In fact, it can got to
170, 180 and if they’ve run for half an hour or 45 minutes, it’s sustained to that level during their run around
Washington, especially. (panelists laughing) And I have to say that
these patients actually become lower cardiovascular risk. They don’t have increase in their risk of cardiovascular outcomes. So, (laughs) I think it’s
the long-term increase of blood pressure, the sustained
increase of blood pressure as much as it may be that increases the cardiovascular risks. – Absolutely.
– Thank you. So, great, great discussions. Wanna see if any of our panelists have any quick final comments. So it seems like a lot of consensus around long-term, sustained,
even small increases and some concerns about
even in lower risk patients. Trying to get evidence on that, but understand the difficulties from the standpoint of trial design. Some questions that you all raised about, well, when does short-term become moderate and long enough to care about? Any other final thoughts
you’d like to raise? – Well, (stammers) one of the things that
Bob Temple mentioned about a mean change or population
change in blood pressure, three or four millimeters,
whatever it might be, of course, what does that mean? It’s made up of many patients, some of whom actually
have falls in pressure, no change in pressure, small increases, medium increases, and large increases. And I think in many ways we
have not helped ourselves by overanalyzing data and providing means and that sort of thing. We should actually present results from our clinical trials
that involve hypertension or blood pressure in very nice increments, whether it’s stratified to get in different categories
of change in blood pressure. I think that would give a
much better understanding and perhaps in the end what we need to do is to find ways to advise
clinicians how to handle this. As long as we warn them that this drug may increase blood pressure, then at least they’ll be on the lookout for changes in blood pressure and be able to do something about it. – [Mark] And final thought? – Yeah, I think this was brought up. I couldn’t, I apologize, I
couldn’t hear all the comments but clearly we know observationally that nighttime blood pressure is as or more important than
the rest of the 24 hours. We’ve never done our
clinical trials though to assess what our therapies did on nighttime blood pressures. Hopefully that will be done at some point. But so, a drug that is taken
once a day in the morning and raises blood pressure for
a few hours during the day may have much less increase in risk than a drug that is taken at
night, or whenever it’s taken, and it raises blood pressure
consistently at night. We don’t know that for sure, but the observational
data might suggest that. And that certainly, I think, a good aspect of looking at 24 ambulatory monitoring when assessing drug safety. – And I do think that will come up in our ambulatory blood
pressure monitoring session, if not before. I wanna thank all of our panelists for an excellent start to the day and all of you for the contributions. Thank you very much. (audience clapping) And we’re gonna take
about a nine minute break, so start around 10:30
or there about. (laughs) – [Man] Thank you, Michael. – Thanks.
– That was good. – Panelists, go ahead
and join me up on stage. Oh, just um, okay. I think we’re waiting for Dr. Borer. Okay, we’re, (laughs) thank you. So good morning everyone. Welcome back from the break. I’m Greg Daniel, I’m
Deputy Center Director in the Duke-Margolis
Center for Health Policy and I want to echo Mark’s
welcome from this morning. During our second session we’ll
be turning to the discussion of the PRECISION-ABMP trial which has already been
referenced pretty extensively, a sub-study of PRECISION to determine the blood pressure effects of selective COX-2 inhibitors
versus non-selective NSAIDs. We’re gonna be discussing that trial and whether its results and interpretation provide useful context for
FDA’s guidance for industry. Our presenter kicking us
off will be Jeffrey Borer, Professor of Medicine,
Cell Biology, Radiology, Surgery, and Public Health at
State University of New York Downstate Medical Center. And then following his presentation we’ll have two panelists,
Robert Blankfield, Clinical Professor of Family Medicine at Case Western Reserve University, and then unable to make it in person but joining us via phone,
I think is on the line, William White, Professor of Medicine and Chief of the Division of Hypertension and Clinical Pharmacology at the University of
Connecticut School of Medicine. So I’ll go ahead and turn
things over to Dr. Borer. – Okay, I think probably everyone here is familiar with PRECISION
and how it came to be. It was mandated by the
FDA after a 2005 meeting about NSAIDs because
there was a controversy, with one school of thought being that NSAIDs were very bad because, that non-selective NSAIDs were maybe bad but selective, COX-2
selectives were very bad because of imbalance between thromboxane and prostacyclin activity
caused by those drugs. That was called the Fitzgerald Hypothesis. But some of us looked at those data and said no, hey, all these drugs make, many of them make blood pressure go up, maybe that’s the problem, not
the Fitzgerald Hypothesis. And so, we put together, after
the FDA mandated PRECISION so that it could know whether
there was a difference between COX-2 selective
and non-selective NSAIDs, we put together a sub-study looking specifically at blood pressures to see whether maybe blood
pressure really was the culprit. And that study was completed, was reported shortly after PRECISION, the sub-study was reported
shortly after PRECISION. I’m gonna talk a little bit about that. Here is the design of the
overall PRECISION study. As you know, there were
three drugs compared. The FDA wanted more than
three drugs compared. They wanted a placebo,
which was not possible. And the size of this study turned out to be 25,000 patients. So to add another drug would have meant the study would still be going on now. The overall design was here,
it was nothing unusual. The minimum follow-up was 18 months. The patients included had either osteoarthritis or rheumatoid arthritis, which was a problem because the NSAIDs were being prescribed at different doses for those two types of arthritis and that created a little
bit of a difficulty in analyzing the data. But the important point
is that all of them had to have established cardiovascular disease or have a clearly increased
cardiovascular risk because of abnormal, because of multiple risk factors. They had to require
non-steroidal anti-inflammatories for at least six months
for symptom relief. So these had to be symptomatic patients who also had cardiovascular disease or at high risk of cardiovascular disease. They were randomized to
one of three regimens. The regimens were chosen because that’s the way
the labels were written. The Celecoxib dose seemed to
be a little low to some people who practice in rheumatology and think you should give
more, but that’s fine. The label said you couldn’t
give more than 200 a day to somebody with osteoarthritis and 400 a day to somebody
with rheumatoid arthritis. So that’s the way it was done. There was optimal preventive care for cardiovascular risk,
according to local standards, including giving aspirin and whatever else you chose
to give, statins, whatever. And there was an option
to increase the dosage for unrelieved symptoms,
to the maximum approved by local regulatory authorities. Everybody was given a
proton pump inhibitor because of the concern about effects on the gastrointestinal tract and some data were collected
regarding that problem, but I won’t talk about those here. The trial was event-driven,
follow-up was 18 months. Now, that was the overall PRECISION trial, which included 25,000 patients. The ABPM sub-study to determine whether a blood pressure increase might have been a reason
for whatever we found from the big trial
included only 444 patients. Everybody in a single
center had to be included for the center to be included so no cherry picking could go on. This study only went on for four months. So this was short-term and what we found would qualify as a
short-term study, I think. And here were the results from the ABPM. There was clearly a difference in the blood pressure effect of Celecoxib versus ibuprofen, and a
little bit of a difference versus naproxen, but the
difference versus ibuprofen was about 4 millimeters
of mercury, not quite, of systolic pressure, which is
what we’ve heard about before as being associated with
some bad things happening to some people. Here are the changes in mean
24-hour arterial pressure from baseline at four months. And again, you can see that ibuprofen, and to a lesser extent naproxen, caused more of an increase in ABP, in arterial blood pressure,
than did Celecoxib. And in fact, that the difference in the least squares mean change
in arterial blood pressure followed the same pattern. Now how about patients who started with normal blood pressure
and who had hypertension at month four. That was significantly greater. People who were normal tensive
and developed hypertension did so much more frequently with ibuprofen and also with naproxen than with Celebrex. Twice as frequently with ibuprofen, a little less than that with naproxen. This, however, speaks to
one of the major issues that was discussed earlier. How long does it take to see the effect? Here we have months since randomization here on the abscissa. I don’t have a, and patients with the event, the percentage of patients
with the event on the ordinate. And you can see that
it took several months for the curves to separate. The outcome event here is the hospitalization with hypertension. It was increased by 69% with ibuprofen compared with Celecoxib, but you really didn’t see that happen until several months
after the trial began. It took at least 10 months to 12 months before the difference
became clearly apparent. During the mean follow-up of 2 1/2 years, there were 22 APTC events, seven with Celecoxib, nine with ibuprofen, and six with naproxen, clearly
not a difference there. Just a tendency towards more ibuprofen. So, with these 444 patients
followed for 2 1/2 years, we weren’t able to see
a difference in events and the overall event
risk was relatively low. From the overall PRECISION
trial, the big trial however, with more numbers, the
tendency that was suggested by the ABPM sub-study
was seen more clearly. Here for example, non-inferiority trial, intention-to-treat for the APTC endpoints, Celecoxib was non-inferior to the others or the others were
non-inferior to Celecoxib, but it looked like something
was separating there and so a superiority analysis here, what we called modified intent-to-treat, there were fewer events on
Celecoxib than on ibuprofen, no real difference compared with naproxen. And, when all-cause mortality
on treatment was assessed using a superiority design, the Celecoxib clearly had
fewer deaths than the others. So, the conclusions that
we drew and would draw is that the nested blood pressure study showed that Celecoxib is
less hypertension-inducing than the two NSAID comparatives. And though cardiac outcomes
were relatively few, there were fewer with Celecoxib than with the other comparatives. The outcomes of the nested
study were thoroughly consistent with the results of the larger study, of which the nested was a part. So for future evaluation,
nesting a blood pressure study within a larger phase III trial for whatever the primary
outcome of that trial might be and pre-specifying blood pressure
in cardiovascular outcomes should help to define
a regulatory position about approval or at least about the need to warn and label about
blood pressure effects. Thank you. – Okay, great. (audience clapping) Thanks, Dr. Borer. We’re gonna go ahead and turn things over to our first panelist, Dr. Blankfield. – I brought some slides. I wanted to thank the
Duke-Margolis Institute for inviting me today. I hope that by the end of my presentation there might be a little agreement as to what thresholds would warrant a requirement
for cardiovascular safety data and I’m gonna make my
case that those thresholds should be two millimeters systolic and one millimeter diastolic. These three drugs have
been taken off the market in recent years because of
cardiovascular safety concerns. Could I see a show of
hands how many people in hindsight wished that the FDA had obtained cardiovascular safety data before approving these drugs? How many people wish that
that data had existed prior to approval? All right. If one is gonna use blood pressure data as a determinant of
whether to obtain data, then the threshold for all these drugs is a systolic of two or more, or a
diastolic of one or more. And any other criteria
would not have generated a requirement for
cardiovascular safety data. So, two and one is a
sensitive screening test for cardiovascular risk, if one
is gonna use blood pressure. The TZDs Actos and Avandia
have generated a lot of concern about cardiovascular risk. I won’t elaborate much
other than to point out that these drugs have a negative or zero effect on blood pressure. A number of diabetic
medications have required, the FDA has required
cardiovascular safety data. Most of these drugs, apart
from low-dose Onglyza, have a negligible or negative
effect on blood pressure. Other diabetic medications, again, negative or negligible effects. Cardiovascular data has been
obtained on all these drugs and they have negative
blood pressure effects. So, this is the point in the presentation, I know you’re eager to begin asking and wondering when is he
gonna get to naproxen. So, this is that portion
of the presentation. This is not a definitive slide. I probably missed some studies, but the important part of this slide is that there’s some data here comparing blood pressure changes
to baseline and to placebo. And, the effect of
naproxen in these studies on systolic and diastolic
blood pressure is variable. Most of them are different lengths. The PRECISION trial is the last slide. PRECISION trial when changed to baseline was 1.9 systolic, and 0.7 diastolic. And so, if one is gonna use
a threshold of two and one, then one could make the case that Celebrex shouldn’t warrant requirement for cardiovascular safety data. I think that’s the wrong
interpretation of the data and the reason it’s the wrong
interpretation of the data is because you can’t use the study drug against baseline as the criteria. If one looks at some other studies, the Schwartz study in 2002, the change in systolic blood
pressure compared to baseline was three and compared to placebo is four. In that same study, the change
in diastolic blood pressure compared to baseline
was a negative number, but compared to placebo,
it actually goes over that one point threshold. The Baerwald study, similar pattern. Compared to baseline,
the changes are minimal, but the changes are higher
compared to placebo. And in those instances, if one were using a two and one threshold, then one would conclude
that using baseline data does not require
cardiovascular safety data, whereas comparing the drug to placebo does require cardiovascular safety data. So, I think it is a flaw
of the PRECISION trial to make the case that
one can compare a drug against its baseline. So, it’s my recommendation that the FDA require cardiovascular
safety data for drugs, at least drugs used for long-term,
that compared to placebo, raise systolic blood
pressure two or more points or diastolic blood
pressure one or more point. Now if I have another couple minutes, there are a number of drugs on the market for which there is no
cardiovascular safety data. A number of the SNRI antidepressants all have considerable
effects on blood pressure. The only one is Bupropion. I remember as a young physician hearing Dr. Thayes talk about Effexor XR, and it was brought on market
and in his presentation he said there’s data that
the 375 milligram dose increases blood pressure, diastolic blood pressure seven points. And I thought holy mackerel,
that sounds like a lot. And I said well, to
myself, I guess that’s okay because the FDA, they must
have safety data on that. Little did I know. A number of the ADD/ADHD medications, and to the FDA’s credit, they
have safety data for children but I assure you that
adults are using these drugs and there’s no safety data with adults. And these drugs have significant
effects on blood pressure. And then, some miscellaneous drugs. Some of the drugs for
narcolepsy, Nuvigil, Modafinil have considerable effects
or noticeable effects on blood pressure and there
is nothing on the warnings of those drugs regarding
cardiovascular safety. Thank you. (audience clapping) – Okay, thank you, Dr. Blankfield. So we’re gonna turn to our
next panelist, William White, who is, I think, joining us on the line. Dr. White, can you hear us? – [William] Yes. – Okay, go ahead.
– I can hear you. Can you hear me? – Yes, we can. – [William] You can hear me? Oh, that’s just wonderful. (laughs) I’m really sorry I can’t
be there in person. I worked on the planning
committee for this meeting for many months with
Duke-Margolis and Norm Stockbridge and his team. I really am sorry to not be there. So, if we could have my first slide. I only have four slides. And I can see them on the Webex if you can bring that first slide up. – Uh, okay, we’re gonna get to those. – [William] So let me
just say parenthetically that Jeff Borer and I worked together on off-target effects of
NSAIDs and COX-2 inhibitors starting about, I don’t
know, 17 or 18 years ago. And we pooled data, way before
PRECISION was even a concept and actually found a signal for ibuprofen back in those days, from a
pooling of about 20,000 people in development programs
for these COX-2 inhibitors. So the findings–
– Dr. White, just to let you know, we’re still trying to pull up the slide. You can go ahead but
just wanna let you know that your slide’s not quite up yet. – [William] Yeah, I realize
the slides aren’t up. I’m making some introductory comment. – [Greg] Okay. – [William] And, we also knew that we saw no real signal with Celecoxib. The problem was is that
the proportion of people who were on placebo in
those development studies was smaller and also shorter term because arthritis patients
can’t take placebo for more than a few weeks or months before they will bail out of the study because they’re in pain. This is a big challenge of doing cardiovascular safety studies for patients who have symptoms. So, do you see my slide now? – [Greg] Not yet. I’m sorry for the glitch. We thought they were in line, but they’re getting loaded
up relatively soon. (laughs) – [William] Well, they’re
on YouTube right now. The slide is on YouTube
but it’s not on the Webex. YouTube won. And it looks like 130 people are watching this program on YouTube. I have no idea how many people are actually signed into the Webex. But, you can’t see them
in the room, I assume? – Right. (laughs)
– Yeah. We’re getting ’em. I’ll let you know as soon as it’s up. – [William] Okay. You know, that’s a great
advertisement for YouTube that you can actually
see the slide on YouTube but you can’t see it in the live room where the meeting is occurring. – [Greg] Okay, we got it up now. – [William] Okay, that’s wonderful. All right.
– Okay, here we go. – [William] So, thank you. So a number of years ago there was a study done
by a fellow from Scotland named Tom MacDonald and there was a drug that was a little bit like Diclofinac but it had more COX-2
selectivity called Lumiracoxib that was being developed for arthritis. Unfortunately, the drug had
significant hepatotoxicity so it never came to light. But this drug was studied
in a pretty sizeable ambulatory blood pressure
monitoring analysis involving several hundred patients and published in the
Journal of Hypertension. And one of the things I wanted to show you was that it does matter who you are when you’re getting these drugs, as far as the excursion
upwards of blood pressure. So, for example, if you’re older versus if you’re younger,
your blood pressure increase is greater on ibuprofen
than it was on Lumiracoxib. I’m looking at the
estimated differences here on the right-hand portion of the screen. And really quite importantly are the types of drugs
you’re on for hypertension. So, one of the things about all
NSAIDs, including ibuprofen, is that it doesn’t just raise
blood pressure by itself because of a little
inhibition of natriuresis, it also interferes with the
anti-hypertensive effects of some of the classes of
the anti-hypertensive drugs. Most importantly are the
renin-angiotensin blocking drugs like the ACE inhibitors or the angiotensin receptor blockers, you see that the increase
in blood pressure is much more substantial in that subgroup than it is in patients, for example, taking a calcium channel
blocker or a diuretic, in which there doesn’t look
to be very much increase in blood pressure. So, I think we have to, when we look at the data from PRECISION and we look at the data from
NSAIDs in general as a class that is known to have this
increase in blood pressure, this has to be taken into consideration. If we could have the next slide, which I can’t see on the Webex, but I can see on YouTube again. So, hopefully that slide can be changed. – [Greg] Yeah, we’re there. – [William] Oh, now I, yeah, yeah, okay. I’m seeing the STABILITY trial. So, I’m not sure what you’re showing but I know that my next slide is a slide that evaluated Celecoxib
in the early days. I think the publication was in 2002. – [Greg] Yeah, we’re showing that. – [William] Thank you. I see it now on YouTube, thank you. And, so this was addressing, because the FDA, in fact, recognized that there was this potential when this new class of drugs known as COX-2 selective inhibitors
came out on the market that it would actually potentially
increase blood pressure in people who were
taking an ACE inhibitor. So we designed a study to take people who were on ACE inhibitors,
in this case Lisinopril, we standardized them and we found that 200 milligrams twice
a day compared to placebo, ’cause these were not arthritis patients, so we didn’t have that confounder of pain, caused some transient
increases in blood pressure relative to placebo but it
wasn’t all that bad looking. And this was with 400 milligrams a day, which would be the
maximal dose of this drug. And probably more in alignment with the 600 milligrams three times a day that was given on average
to patients in PRECISION. And just parenthetically, historically, the doses of ibuprofen used
in osteoarthritis trials and in rheumatoid arthritis
trials in the past was 800 milligrams three times a day. And there’s clearly a
dose-related increase in blood pressure, edema, heart
failure, and all that stuff with the NSAIDs,
particularly with ibuprofen. So if we go to the next slide, somebody mentioned an outlier analysis, I think it was Michael Weber, that we really do need to
understand individual responses when we’re looking at data that evaluate off-target effects. And in this situation, the
finding for this same study was a bit surprising. Hopefully you’re seeing the bar chart now. – [Greg] Yes, yes. – [William] It shows the outliers in which there’s bins of blood pressure and you’ll note that the main finding was not that there was
these huge increases in blood pressure on the drug at all, compared to placebo, but
actually a few less patients went down in blood
pressure, and a few more went up kind of in the five to 10 range. And that was the individual increases, and so, if you were to
characterize any drug in which there was ABPM or
clinical blood pressure data was taken precisely, as
Bill Cushman mentioned, you would have a different sense if the drug increased blood
pressure in all individuals by about three or four or
five millimeters as an outlier versus 10 or 15 or 20,
which would be more onerous for a population in a
shorter period of time. I think this is a very important part of the way the analyses
should be carried out when we’re looking at drugs effect, off-target effects on blood pressure. And if we could go to my last slide, just to make a comment about
two millimeters of mercury being a threshold for asking all sponsors to do cardiovascular outcome study, well, Tylenol does that. So Tylenol is a ubiquitous drug that is used over-the-counter throughout the entire world by millions and millions of people. And this small study was
done by a group in Europe who looked at acetaminophen
at doses of three grams a day, which is a pretty standard
osteoarthritis dose, or it might be taken for
a couple or three weeks if you have a pain syndrome, and it raised ambulatory blood pressure by about 2.7 millimeters
of mercury systolic, and about two millimeters
diastolic versus placebo. And it’s not thought,
most people don’t think that acetaminophen does this because nobody really
studied it thoroughly before, but this group did, and these, by the way, was in a population of people who were, had coronary disease. So, theoretically,
their risk might be more than a population at-large
who’s just taking acetaminophen or Tylenol for OA or just
general pain conditions. Do we have a possibility in this world of doing a cardiovascular outcome study with Tylenol or generic
acetaminophen, right now? Do we need to? Do we need to study any drug that has simply a blood pressure increase without other mechanistic concerns? NSAIDs are not a clean class of drugs. They do not increase cardiovascular harm only because they increase blood pressure. They do all kinds of other things, to tissue factor, to coagulation
profiles, and so forth. They cause the retention of salt and water that has other effects other than just increasing blood pressure. In contrast, there are other drugs which may do nothing but
increase blood pressure and not cause other
harmful, comorbid issues with regards to their off-target effects. So I think that in thinking
about the patient population, the target, the group of individuals for whom this off-target
set effect might be seen, we have to really strongly consider that. I mean, there are, as Dr. Blankfield said, numerous drugs that are registered, approved, and have been on the market for maybe 50 years or longer
that raise blood pressure and don’t have cardiovascular safety data. But I think that we understand that increases in blood pressure are a very clean and powerful surrogate all by themselves to
show that they increase, that increases cardiovascular events. Do we really need to perform a study that almost is gonna be impossible to do to prove that point? So I’ll leave it at that and open up for any other discussion. Thanks for listening. – Yeah, thanks, Dr. White. (audience clapping) And thanks for bearing with
us as we got your slides up. So I’ll go ahead and open
it up to the room around, comments and feedback on the
presentations that you saw regarding the PRECISION-ABPM data, but also the other trials and data that the three presenters put forth and implications for draft FDA guidance. So, Ellis? – [Ellis] I see, delay. Um, I’ve heard some discussion
about outcome studies, the need for outcome studies. So, I’d like to make an important point from FDA’s perspective. So if we know that a drug
increases systolic blood pressure by four and diastolic by two, for a drug, I’m gonna ask you guys
a rhetorical question, which is do we want a company
to do an outcome study to show that that increase translates into an increase in stroke. – I think the point has
already been made. (laughs) – [Ellis] Okay. But I did hear a lot of discussion, well, maybe it hasn’t been made. Okay. – I’ve heard a difference
of opinion this morning on that very question. I’ve heard a lot of people say that’s it’s a nuanced question, that we need to know more
about the populations, who’s at risk or not,
and much of the comments that I heard in the first session could be answered with more data, that we don’t have more data. So, I, I do think a blood
pressure of four and two is likely to translate to
increased cardiovascular risk, but that’s an assumption. And if we make that assumption, then it’s not fair to
go back and say well, we don’t know about the risk group. – [Ellis] Yeah, it’s, I
don’t even think it’s ethical to do a study and say listen, Mr. Jones, we’d like to figure out if
this drug could cause strokes, that’s the endpoint, is strokes, and we wanna know if it
increases your risk of a stroke. That’s not an ethical study. And, it’s okay if you’re saying well, we wanna see if it helps your arthritis, and by the way, we’ll look at safety– – So four, at four and
two there’s no equipoise? – [Ellis] Well, I don’t
know what the numbers are, but my point is that–
– Where is the equipoise? – [Ellis] Well, the equipoise can be if there’s a benefit that–
– Less than four. – [Ellis] Well, I said less than four. So if you’re saying oh, you wanna look at the effect on your osteoarthritis, and by the way, that’s fine. But to do a study purely to
see if it causes strokes, that’s not the kind of study
that we really wanna encourage. Bob, you’re saying? – [Bob] I think that we already know that increasing blood
pressure compared to nothing is bad for you. What the cutoff should
be, whether two is enough to reach that conclusion or
it needs to be more than that, I don’t know, but my view is
it’s a continuous function. We know higher blood pressure, I mean, all the epidemiologic
data shows that, higher is bad. It’s worse for you if you
already have underlying disease, it’s worse if you’re older,
it’s worse if a lot of things, but we sorta know that and
I’m not sure you need a study to do that. It’s worth remembering, the
studies that have been required for diabetic drugs were not because of blood pressure
anxieties, it was because, probably an erroneous assumption, that some of the early
drugs were bad for you in ways that we didn’t understand, so they all got that requirement and almost all of the
studies have been negative except for the ones that
have been positive lately. So, there’s been a lot of
attempt to reconsider that. But that wasn’t because of an anxiety about blood pressure effects particularly. I guess I think everybody knows that raising blood pressure
is probably, not chronically, is not good for you. And so I’m not sure you need a study. – Okay, so we’ll go to George and then back to this
table and then over there. – [William] I wanna just
expand what Bob said and view it from a clinician’s standpoint, because I get all these people
in the hypertension center because nobody can control them. So, one first of all has to appreciate what these drugs are
doing in the first place. So if you have pain, if you don’t think pain
raises blood pressure, we can do that experiment
on you and you can tell me. All right? And so, that’s a competing risk. So you need to control the pain and there’s a lot of other
immune, inflammatory processes which drug do raise blood pressure, but they actually balance out in terms of the benefit long-term. So I think to try to put a number on it is almost impossible if
you wanna do it accurately, and you can get truckloads of data. You may come closer but it’s
still not gonna be the same. I think the oncologists
actually got it right, and it’s rare that I
publicly praise oncologists, but I will say that I was
privileged to be part of a group invited by the oncologists to put together a consensus report, a guidance, in terms of blood pressure control for the VEGF inhibitors. And what we looked at
the data at that time, and this was 2011, but basically the essence of that came out, on top of understanding the mechanisms, was as long as you can keep
the blood pressure below 140, so know it’s gonna go up, how much it goes up is almost irrelevant, you can’t start therapy
unless your blood pressure is below 140. How you get there is your business, but you gotta stay below 140. That, to this day, still exists in terms of these protocols. And it seems to have worked out well, from a safety standpoint. So I’m just putting, keep it simple because it’s gonna become unwieldy if you try to make it too precise. – Okay, there’s a, back at this, state your name–
– Yep, Jeff with Bioclinica. I think I agree with
Dr. Bakris and Dr. White and I don’t think an observational study actually adds value. I think what we’ve been
doing so far in the industry is really trying to characterize the potential effect of the compound in the therapeutic population on the change in blood pressure. Because at the end of the day, it’s what is the clinician, he or she, going to have to do and
know about the compounds that they’re treating the patients with. So, questions might come up with, okay, I’ve got a drug for the
osteoarthritic population, my population is hypertensive, and do a focused assessment on that, more so than say a healthy population, because this is the population that the drug is gonna be exposed to and what’s that change? Likewise, if it’s gonna be a
short-term exposure to a drug, not a long-term exposure, you might wanna consider, okay, using ABPM, what’s my baseline, what’s my expected on treatment
and half-life of the drug, and what happens when I remove the drug? In fact, does the blood
pressure go back to baseline? Does it stay elevated? That because important
more in a longitudinal look of the impact of that
blood pressure change on that patient, that population. And you start to differentiate between the long-term
exposure to a compound and a short duration
exposure to the compound. So, I think we’re already
in the right direction and the goal is to really to characterize that change in blood pressure and then specifically look at
the therapeutic indication, and Dr. Bakris just explained. I think the oncology group really does have a good focus on things
because we’re talking about risk but we also have to look at benefit risk when we’re looking at these medications and the patient population. – Okay, great. I’ve got two comments over at this table. Uh, no. I think we’ve got another
mic coming your way. – [Sid] Thanks. Sid Wolfe, Health Research Group. Comments on both what Dr. Blankfield said and Dr. White said, in the context of what the question is. Is it appropriate to
look upon the findings of the PRECISION study to make the link, which if it is not the
complete explanation, it’s one between hypertension and increased cardiovascular events. So, 11 years before, 11 years before PRECISION was published there was a study
published on the same drug, not in people who had RA or osteoarthritis but in people who were
being experimented on, in the best sense, in this case, to prevent colorectal adenomas. And in this trial, A, a placebo
was used as the comparison, which a point Dr. Blankfield made, and B, they collected data on both blood pressure increments up, well, blood pressure increments period, and cardiovascular risks. And in 200 milligrams
twice a day with Celecoxib, they had a 2.6 highly
statistically significant increase in cardiovascular events and a two millimeter increase average versus placebo on the blood pressure. And when you went up to the higher doses, as in 400 twice a day,
the cardiovascular risk went up to three plus and the increment up and systolic blood
pressure went up to 2.9. So here is a older study, it is not on people with
osteoarthritis and so forth, and they have a nice correlation except for one point chose A, a dose-response increase,
dose-response cardiovascular risk with Celecoxib, and also
shows a dose-response increase in blood pressure. So I think for some of the reasons that Dr. Blankfield alluded
to, and maybe others, the idea of working on a guidance that
will help in the future identify increased blood pressure should be done versus placebo, and one of the slides Dr. White showed did show in certain circumstances increased blood pressure with Celecoxib. The other comment, which
is what Bob Temple made, is I think that we know from ever, Bob and I debated this about 25 years ago, in terms of placebo controlled
trials for hypertensives, how long they could be, but
I think that we agree fully, it is established that
increased blood pressure is a strong signal for
cardiovascular risks. And so back to Dr.
Blankfield’s slide or to, showing all the drugs with clearly, in most cases more than
just two systolic increases in blood pressure versus the placebo without any kind of black box warning. I think that the idea of putting a warning on the NSAIDs, particularly in a trial that had nothing to do with cardiovascular risks, the class and the VIGOR studies were done to try and get a better label on the gastrointestinal
risks, as you may remember, and yet, the finding
was in a different area and it had a profound
impact, as it should, on the labeling, black box
warnings on all these NSAIDs. So, I think that the challenge, I agree with Bob that it
is a well-established risk and on the ones where we
already have the data, the data are already there in terms of the blood pressure increase, why is there no black box warning? – Okay, thanks for that comment. I think there’s another one as well. – [Mark] Hi. Yeah, yeah, yeah I have
to put it a little closer. It’s Mark Short again. I think there’s two things. First, in terms of the two millimeters versus over one millimeter, there’s a question of PRECISION, not the study itself but
the statistical power, you need to establish that. And the meaningfulness of, if I can get enough data to show that there’s a one millimeter increase and the standard error is 4/10, or it’s 1.1 and standard error is 4/10 and I have another
study drug where it’s .9 and the standard error is .2, you know, is one meaningful and one not meaningful? I think it gets a little crazy when you start to talk
about these numbers, particularly, I think when
you look from the standpoint of a practicing physician. If I have a patient who I think is on, all the medications they’re
on are useful to that patient and their blood pressure
is say 160 over 100, I’m gonna treat that blood pressure. And if I have trouble getting
that blood pressure down, even with reasonably maximal treatment or, you know, anti-hypertensive treatments not causing serious adverse events with the patient, then I might ask myself could
it be one of these drugs and, let’s see what happens
if I stop one of these drugs or I look into an alternative and see what happens if I stop drug, what happens to the blood pressure, what happens to the other
things that they’re treating. And I can then make a decent
risk/benefit assessment. And for that, all I need to know is something in the
label that says this drug has been shown to increase blood pressure. And even whether it’s one millimeter or five millimeters on average, isn’t that important? Because of the variants, because
we have so many outliers. In Mayo we had an average increase of two millimeters of mercury and the guy’s blood
pressure is 160 over 100, what difference is two
millimeters gonna make? But maybe in this patient, it makes 10 millimeters
of mercury difference. Or maybe at this patient it’s zero. So I think there’s a real
danger in getting too hung up on these exact numbers. – Yeah, so great. So, we are a bit out of
time for this session but I do wanna turn back to our panelists, and Dr. White on the phone too to any lasting comments or feedback based on the discussion that we just had. A lot of input on thresholds and whether or not events,
cardiovascular events would be indicated for further collection or if further characterizing
the impact on blood pressure is sufficient, and then
questions about then what, and what does that necessarily
mean for the labeling. So I’ll turn the, Dr. Borer,
Blankfield, and then White for last comments. – Thank you. First of all, I think the
point was just made very well. The issue is not was it two millimeters, three millimeters, one millimeter. The issue is does the
drug raise blood pressure. If it does, then it’s really
up to the patient’s doctor to measure the blood pressure and see how high it goes
and treat the patient for it if it seems like that’s an
inappropriately high response. We know what the epidemiological data show and individual clinicians
have to make judgments based on those. The idea that every
drug has to be studied, every drug for blood pressure, for blood pressure
effects has to be studied against placebo is a wonderful idea and the FDA really wanted
that to be done for PRECISION, but it was impossible because, in fact, people with
arthritis were being studied and they would not tolerate
walking around with pain, so they had to be on something. The use of Tylenol was
suggested as a control, but Tylenol isn’t as
effective as the NSAID, so, couldn’t do it. So, you know, it’s good to be a purist and it’s nice to theoretically believe that placebo-controlled trials are better than non-placebo-controlled and I agree with that, except
you gotta be reasonable. If it can’t be done, it can’t be done. And in this case it couldn’t be done and we came up with
some good data, I think. So, those are the final
observations I would have. – The last gentleman, I agree
with all those comments. I agree that two and one is trivial and anything individual, I
agree that standard deviations sometimes are above and
greater than the number itself. I understand all that. I agree with the point
that raising blood pressure increases cardiovascular risk. I think everybody in the
room agrees with that. I agree that the starting
point in blood pressure, whether it’s controlled or not, and you add a couple of
points, I agree with all that. However, the bottom line must be that a drug like Vioxx,
a drug like Meridia, a drug like Bextra ought
not to get on the market without some warning to
physicians and patients regarding cardiovascular risk. – [Greg] Great, thanks. Dr. White? – Yeah, can you hear me?
– Yes. – [William] Okay, thank you. So, I’ve been listening to
this commentary closely. I really believe there
has to be some pragmatism with the development of drugs and if there are off-target
blood pressure effects. First of all, nobody’s been
talking about absolute risk and the population at-large has a risk. We understand that and I
understand Dr. Temple’s concerns in the population at-large, and the duration of the
drug, who’s going to get it. You know, if a drug raises blood pressure two millimeters of mercury
and most people who take it are 35 year olds who have fibromyalgia and don’t have any cardiovascular risks, the harm it’s going to
cause is vastly different from a 75 year old with osteoarthritis who has multiple risk factors. You’ll see the harm in a
much shorter period of time. To set up clinical trials
in young, healthy people who might be taking a drug which has an off-target blood pressure
increase is not practical. It just has to be a labeling
issue, if you ask me. And I think that we have
to take all these things into consideration throughout this day if we’re going to be
going outside of the idea of diagnosing hypertension with the drug versus what you do with the information that you’re going to get from the data in which you’re studying. So I think that’s just my final comment with regards to the issue. Thanks a lot. – Okay, great. I’d like to go ahead and
thank all of our presenters, panelists for a great discussion, and all of you for contributing to it. Thank you. So we’re gonna go, we’re gonna go ahead and
roll into our next session, which would be the last
session before the lunch. And so, as we’re changing the table, I’d just invite the next set of panelists and speakers to join us. Okay, so during our third session, we’re gonna turn to discussion regarding whether pressor risk intolerance among diverse development programs and how to look at that. As you’ve heard, as part of FDA’s overall benefit/risk assessment
of investigational drugs, FDA does acknowledge
that the draft guidance, in their draft on guidance
that increasing blood pressure can be acceptable or can
be managed satisfactorily in certain particular situations. In this session we’re gonna
dive a little bit deeper into the discussion on the
best regulatory approach to interpreting drugs potential
blood pressure effects. So, joining us in kicking
off that discussion with presentation is George
Bakris, Professor of Medicine and Direct of American Heart Association Comprehensive Hypertension Center at The University of Chicago. Following Dr. Bakris’ presentation, we’ll have panelists,
Vasilios Papademetriou is Professor of Medicine at the Georgetown University
School of Medicine and staff cardiologist
at the VA Medical Center here in Washington, DC. Dr. Brandon Atkins,
Senior Principal Scientist at Merck Research Laboratories; Philip Sager, Adjunct
Professor of Medicine at the Stanford University
School of Medicine; and then finally, Mitch Krucoff, Professor of Medicine and Cardiology at Duke University Medical Center. So, welcome to our
panelists and presenters and I’ll turn things over to Dr. Bakris. – Thank you very much. Actually, in some of the previous talks there’ve been great setups for this. And this is really a
continuum from what I can see, and certainly with regard to
the comments that I’ve made. So, the question is here. Is there a blood pressure increase concern applicable across all
development programs? Or should each program be individualized? And I think you’ve already
heard, very clearly, that different drugs
have different effects and also it depends upon the setting in which these drugs are being used as to the magnitude of the effect. And then we’ve also heard, consistently, that absolute risk is
probably more relevant here than relevant risk,
although if you’re coming up with a population, a generalizable kind of comment, then relative risk is
probably as important. So, I think we’ve already talked about this. Oh, okay. I think we’ve already talked about this in terms of a variety of
different classes of drugs and what their effects can be. I didn’t put testosterone
on there, but I apologize. It should be on there. But I think it’s important to keep in mind that there are a variety of settings, one of the big settings
that everybody I think kind of has in the back of their mind but only the hypertension
specialists really see this, are people with ADD, adult ADD. They’re on a number of these
sympathomimetic kinds of drugs and when you start messing
around with their blood pressure, then you have CNS symptoms
that you have to actually see. So it’s practical in
terms of the consequences of the patient, not just
the blood pressure changes. And again, it’s about chronicity. It is not about something
that’s gonna give you an effect for a week or two weeks, it’s something that you’re
gonna be taking lifelong. And let’s not forget, let’s not forget that these effects are also
affected by sodium intake, which we know is variable. And one of the other things
I didn’t put in there, but also as important, and has really become
more important recently, is quality of sleep. And many of these people with pain, many of these people with
psychiatric disorders have sleep disorders. I’m not talking about sleep
apnea, that’s a separate issue. I’m talking about people
that actually can’t sleep and trust me, that
contributes in a major way to blood pressure variability, which is far more predictive of stroke than any elevation in blood pressure. So I think that’s very important. So, I found the STABILITY trial. I’m not a cardiologist
and I’m not a lipid guy, although I play one on TV. I thought this study
was quite interesting, because it actually, the
focus of this was LpA2 in a double-blind placebo-controlled
15,000 patient study with coronary disease. And they looked at blood pressure
variability in this study and the primary outcome was MACE. So, that’s fine. I just wanna show you
the characteristics here because one of the things they did, and I apologize if you can’t see it well, but they actually looked at changes in blood pressure variability. And they looked at tertiles, much like what was suggested earlier, in terms of less than
7.1, 7.1 to almost 11, and then great than 11
millimeters of mercury and they looked at this in the context of all the different patients that fit. They’re not necessarily
obese but they’re overweight and they’re primarily
Caucasian, not African American. And I think, look at the
blood pressures at baseline. You’re talking a range between 128 to 134. So these are not wildly
hypertensive individuals. So I think that’s, to get
back to the point earlier about well, you have to be hypertensive to really see these effects. These are older people with
reasonable blood pressures and certainly diastolics that
are in a reasonable range. If we then look at what happened to the cumulative incidents
of the primary composite, MACE outcome, by tertiles, you can see, to the point made earlier
about two millimeters, that even the people that
were less than seven systolic, had a rise of less than seven systolic, still had significant
events, and this is one year, the data starts here, one year
after the trial was finished, in terms of long-term events. So you can see this
goes out, and of course, they wanna be dramatic so
they measure it in days. So it goes out to 1,200 days. So you can do the math and figure out, this is years of follow-up. But it’s pretty consistent, and it starts about a year after, so kind of two years after. So what you were seeing
from some of the older data, that’s for systolic and
then the bottom is diastolic and for sure, as far as diastolic, those that had greater
than seven millimeter rises that were sustained, really
had a relatively higher drama than the people that had less. But even people at less
than 4 1/2 had increases, it just was not as great. And this is long-term follow-up
as a result of this drug. So I think, you know, to
try and put a number on it as I said earlier, I
think is very difficult and we have truckloads
of epidemiologic data that will tell you two
millimeters of mercury is a bad thing and you can
increase cardiovascular risk by 14%. I mean, that’s a seminal
paper on over 50,000 patients. But again, I think, to be pragmatic, oh, this is just to show
it didn’t really matter whether you had diabetes or
not, pre-existing hypertension, and just basically if you’re
at the highest tertile, the risk was there, there was
no discerning variable there. So, the question comes up, should we have a specific plan for specific drugs in a specific program? And if so, what should that be? And I think it’s actually very difficult. I think each development program is unique and has its own unique risk tolerance, not just because of blood pressure but because of other factors. And I think that needs to be considered when you’re considering this. And, you know, fundamentally, we were talking about QT increases. You know, you could make
this reminiscent of that, but this is somewhat different anyway. So I think fundamentally, if you, again, I’m gonna reiterate this, I have to say, maybe taking a page out
of the oncology literature is a good way that regardless
of what your blood pressure is at baseline, if you see
increases over time, and especially if you’re hypertensive, you need to have the blood
pressure at least below 140 before you really can say
that you’ve reduced risk, because most of these people are gonna have higher pressures. A comment made earlier about
somebody walking around at 160, well, with all due respect,
that should’ve been treated way before any therapy was started. So, I think these are important points that have to be kept in mind
from a practical standpoint and I’ll just set the
discussion up with that. Thank you. (audience clapping) – Okay, go ahead. – George, thank you very much for an excellent overview of the topic. The changes in blood pressure
obviously have consequences and they have significance. As small changes in blood pressure, either systolic or diastolic
in large populations may shift the risk profile
of that population. But before we make definite decisions or recommendations
or guidelines, doesn’t matter how you wanna call them, we need to consider a number of things that go along with the
changes in the blood pressure. First, the population
that we are targeting and these changes have
different consequences in different patient population. In high-risk patients, small
changes in blood pressure may mean, may meaningful, in low-risk patients may
have no consequences. And I just wanted to bring
the MR systole, for example, that showed that for mild
to moderate hypertensives, five millimeter of changes
in diastolic pressure or 10 in systolic, we
needed to treat 850 patients for a year to see a change for one stroke. In the Heart study that
targeted diastolic pressure, changes of two millimeters,
if you remember the chief blood pressure
in the three groups in the Heart study was 81, 83, and 85 and had no consequences on outcome. So, in different populations
the changes in blood pressure may have different effects. Nevertheless, if we have
confirmed and definite data that changes in blood
pressure in a population that is targeted by the new medication are consistent significant, I think they’re meaningful
for the long run and we should take them into account. How much that blood pressure should be? In my opinion one or two
millimeters probably, within the range of error and probably we need confirmatory data before we consider them
as definite changes and as having an effect on outcomes. But if we confirm a chance
of increasing blood pressure then a level, even two
millimeters should be of concern, should be taken into account. – [Greg] Okay, great, thank you. Dr. Atkins? – Great, thanks. So if there’s a theme so far today, at least the one huge take
home that I have gotten so far is that blood pressure’s
extremely, extremely complicated. This obviously is not as
simple, and even QTc isn’t, but this is not the same situation as QTc, where arguably a certain
prolongation in QTc gives whatever patient who receives a drug an immediate risk of a
potentially lethal arrhythmia. And interestingly, the
current draft guidelines, which again, the FDA should be applauded for putting together,
really do outline this and I just wanna read line 117 to 123, ’cause I think it’s really
germane to what we’re discussing. It says, “Several factors
can influence the importance “of an effect on blood
pressure, including one, “the seriousness of the
condition being treated; “two, the effect of the
drug on the condition; “three, the underlying cardiovascular risk “in the patient population
most likely to use the drug; “four the availability of
other effective therapies “that do no raise blood pressure; “five, strategies that can be used “to mitigate the blood pressure effects; “and six, the anticipated duration “of treatment with the drug.” And then also in this room
today we’ve heard several others that aren’t listed here. And some of these, including
things that are addressed in the current guideline,
such as acute versus chronic, are not as straightforward
as they may sound. For example, if you’re developing a drug for an oncologic indication where lifespan may not be extremely
long, the potential risk of a long-term cardiovascular effect is really, really questionable in regards to the potential
benefit of the drug. So, all in all, I have
significant concerns about a one-size-fits-all approach to this obviously very complex issue. My concern is by going
with a one-size-fits-all, you may inadvertently
create significant barriers to the development of
arguably life-saving drugs that are otherwise safe
and clinically efficacious. And so obviously some thought, I really applaud again the group for coming together to address this, but I think it is a
very, very complex issue and I think every single
program really has their issues and it needs to be done
on a drug-by-drug basis and a case-by-case basis. A couple other things I just wanna mention because I don’t see them on
any place else on the agenda are some discussion about
de-risking these things in the pre-clinical context. I recognize that there is some literature that suggests that what
happens pre-clinical may not translate to clinical work. But there’s a large
amount of work that we do in drug development to
de-risk various issues. And does pre-clinical assessment play a role in determining
what the next steps should be? And then finally, one of these barriers that can be significant
are if we’re talking about de-risking, and the
current document suggests that these de-risking
should be done early, an early-stage program. We’re talking about de-risking a two millimeter systolic blood pressure, one millimeter diastolic blood
pressure in early programs. We’re talking about
exponentially increasing the size of those studies. A typical phase I study can be anywhere around 10 to 20 patients. By our kind of back of
the envelope calculations, looking at the variability of ambulatory blood pressure monitoring in studies that we conducted at Merck, we’re talking about
studies that are upwards of 100 or more patients. And so these are all the things
that need to be considered. Thanks for the time. – Great. Yeah, thanks for those comments. Dr. Sager? – You know, it’s always
great to be on a panel when there’s disagreements
and you can have some kind of hot discussion. I think we’re all in
extreme agreement here. (panelists laughing) George, I really enjoyed your talk and I think all the previous comments, I would add that I think
this is really different than the issue around QTc, which is both a risk that’s instantaneous and correlated to the
degree of QT prolongation and the Cmax of the drug that
we also are now well-aware that the degree of QT
prolongation is also, in terms of its risk, is drug dependent for drugs that effect
multiple ion channels. This is a very different situation in terms of chronic risk and
much more highly dependent on the underlying risk of the population, both cardiovascular risks,
their blood pressure. I think I’m in agreement
that we really need to have not one-size-fits-all,
but it be individualized for the individual drug, and of course, drugs also may exert benefits that counteract any risk around
blood pressure increases. So I think, thinking of this in terms of
is there an individual number, I don’t think there is. I think we really have to do it drug program by drug program, integrating all the different critical aspects. You know, there’s been some discussion about two or one millimeters of mercury, I think we really wanna get away from those kind of specific numbers. And also trying to do this in phase I, obviously is, Brandon has
said is totally unrealistic. This is something that
would have to be evaluated later in development. Thank you.
– Okay, great. Mitch? – Well, I guess, in some sense is the fierce agreement club
makes my job a little easier. So what I think is worth doing is stepping back to your
earlier understanding that the whole part of this discussion is focused on developing a guidance, whose major concerns are safety related. And at that level, I do
think that particularly respecting the fact that
Ellis, Norm, Bob, and Doug have already put this many hours into it that it’s not a slam dunk. However, I still think there is sort of a four-dimensional
road map approach, where we could think about the high, high, high, high risk corner and create some basic principles, that in fact would apply
across all programs, and then dichotomize or
individualize further down. So for instance, as we’ve
said a number of times, and I’m gonna ignore for
right now the modalities of measuring a one to two
millimeter change in a population and the technical side of that, we can get into that maybe later. I think we have to respect that even with ambulatory blood pressure, human liability with blood pressure, we’re taking snapshots. So, just to respect that
the measurement itself has its own technical features. The baseline substrates,
what are really the, so, I think maybe we could all agree, if you already have hypertension then further blood pressure elevation, you’re probably in a higher
risk category than if you don’t. And right down the list it’s been gone, so age, sex hasn’t come up, but actually there’s some interesting data that might be different, if
you already have heart failure, et cetera, et cetera. So I think there are certain categories of what is the patient’s
baseline substrate? What is our understanding of
what is normal and abnormal? That blood pressure
would provide a risk to, is something that we
could potentially create, what is a high, in this particular domain. The objectives of therapy
I think weigh in there too. So if you’re curing cancer, then even if you have high blood pressure and I think the guidelines that came out, control the blood pressure,
go with the drug, et cetera. But again, if you have a severe effect and you’re only gonna be
on the drug sporadically, or if you have a moderate degree of change but you’re gonna be on
the drug persistently, or if you have a modest effect or if you have a modest
effect and sporadic, you could see this as again, an area where you could say
what’s the high, high risk and long-term exposure, or a very high reaction would be too. The mechanistic rats nest,
I think we also really have to respect though, because
whether it effects vascular tone or whether you’re affecting
heart rate and cardiac output, which can actually be good
for you if it’s from exercise, whether it’s metabolic changes, if you’re burning out
the thyroid, you know, these are all really different ways of creating elevated blood pressure. And that takes me back to the fact that a lot of the data we’ve
looked at is epidemiologic, where correlation and
causation are not the same, and I think we gotta be
really careful about lumping. If my drug elevates heart rate and I take Bob’s approach and
I treat with some diuretics, I can probably make the
blood pressure normal, but ultimately the overdrive of my atrium puts me into atrial fibrillation,
I still have a stroke. So, I think we have to be careful that we’re on the right
track with mechanism. If the key recommendation is
if the blood pressure’s up, just treat the blood pressure and everything’s gonna be fine, I don’t think that’s a given. And I think actually, the
NSAIDs are another example where thrombotic complications, in fact, if you look a little deeper, how many of these are embolic strokes versus hemorrhagic strokes? How many of these are
hypertension-related strokes, as opposed to other mechanisms, I think is another question. And that comes to the final part, which I think we’ll also talk
more about hopefully later, which is statistically. I’m all in on population based statistics when it comes to public health questions and sort of getting cued in, but this is a particularly
interesting circumstance where I think we really
have to correlate that with much more of a
personalized type of statistic. So, people who on the drug
have blood pressure elevation and people who otherwise
seemed to be matched who on the drug don’t have elevation, or actually have lowered blood pressure, is telling us something
that we should not overlook by averaging it into a
population statistic. And insight into mechanism I think would be by best using
both statistical rubrics, not just one or the other. So the last part of this is ultimately what does it drive if you’re in the high, high, high risk zone, toward the guidance telling
you what you need to do next? And I think that’s where
is it an outcome study, is it a study where you randomize to treat any blood pressure,
and seeing if you make people and events go away, is it
something that could be done in a real world evidence
oriented progressive universe, in a post-market environment? But what is the barrier that comes with the signal, I think is another part,
a really important part of the guidance where
separating out the higher from the lower risk scenarios could lead to understanding
higher barriers versus lower barriers and what
they are and when they come. – Well, gee, Mitch, I thought that with everybody
on the panel agreeing with each other that this
was gonna end quickly, but you put forth what I think is a really good sort of, not a compromise but a sort of set of
principles for how to think. I think everybody on the
panel seemed to agree that a one-size-fits-all
is not gonna work, but maybe for certain
categories of the population that are in that high, high, high, then you could think of a sort of pan-drug development program that something that has to be done versus if you’re not in those high, high, high categories then that then there would be more of an individual drug
development program approach. But just to clarify really
quickly before I move on, ’cause I see even panelists here have questions about that for your, when you say high, high, high, that’s taking into account
the duration of the therapy, the potential effect
of the blood pressure, and the underlying risk characteristics of that population. Were you implying three
different dimensions of that high, high, high? – At least, yes.
– At least, okay. So, Dr. Sager? – Yeah so, Mitch, you brought
up this high, high, high maybe needing to do additional studies. You don’t feel like this could be handled through labeling and identifying
to practitioners risk and, you know, a need to
be aware of blood pressure and trying to control it? I mean, do you think a guidance should really mandate
additional studies here? Or do you think, again,
this would be handled on a very highly individualized basis, but usually this could be
handled through labeling. – Well, I won’t go down a rabbit hole but I will say that I think
fragmenting our thoughts about what’s real evidence
as requiring studies per se is wrong thinking for the
future of public health. I think continued evidence
and electronic evidence, analytic techniques in tracking
signals in the real world shouldn’t be setting up another
study and another study. But it should be a continuum of evidence and if we have a signal, we should have much
more facilitative ways, more efficient, more informative, and much less expensive to continue to track those specific signals and sort some of these things out. – So before I turn things
over to the audience, any other comments on this last proposal or
other things that came up during the presentations
from our panelists? – I think the changes in blood
pressure should be viewed in the context of risk and benefit. And obviously, we should estimate how
much benefit the patient is expecting from the
treatment, the new treatment, and how much risk that new
treatment is providing to him. And the risk of small
changes in blood pressure are sometimes negligible
in low-risk populations. You need to treat 10,000 patients before you can see a change in one stroke. One example I’m gonna bring
up is of the VA studies. When Ed Freis, my boss, did the first studies at the VA hospital, you only needed 180
patients to show benefit from treatment of severe
hypertension versus placebo. Today, we need 42,000 patients in ALLHAT to try to see if there is difference between the different drugs. So, the risk obviously of this population is much different and
we should consider that when we are considering the risk that’s associated with the new treatments or with the new drugs we
are bringing on board. And also we should also consider the availability of other drugs that can do the same job. Is the patient gonna benefit enough to make it possible for us to absolve them of small risk that we’re
adding to his therapy or not? And I think we should
be examining the changes in blood pressure in that context. – Okay, so I’ll turn to the audience. I think we had first over here, two at this table. (laughs) – [Woman] So, I think that
you all appear to agree because we did a bad job
wording this question. We made it a yes or no question. And had we made it a
what or a how question I think it would have been more clear that you all don’t agree. (laughs) But, so two thoughts. Number one, I think a lot of discussion is kind of bringing out perhaps that many of these issues
straddle FDA guidance versus clinical practice guidelines. And I think that would be helpful to kind of bring out a little bit more, but also a question
specifically to Dr. Bakris. You’re touting this oncology
model of doing things with their clinical practice guidelines, but I’m wondering for drugs, in some ways that’s an easy category of drugs because the benefit/risk is quite clear, but for drugs in which
benefit/risk is not as obvious, what would be your approach there? – Well, I would put the onus on you to tell me what you’re talking about. Because, for example, for example, if you have severe arthritis, it’s arguable most
rheumatologists would say that the NSAIDs are
giving you great benefit and they usually need more. If you’re talking about
psychiatric divisions, depression, et cetera,
clearly those drugs are, if it’s true depression, those drugs are really working well. Additionally I can tell you in the adult ADD patients that I have, if you start messing with
those drugs for blood pressure, you’re gonna be in deep
trouble very quickly from a patient’s perspective. So, I have not come across drugs that are not perceived, or actually are of benefit to the patient. And so, the reality is, to balance off, I realize that the perception and on the part of the public
is drama with oncology. You’re either dead or alive and
these are life-saving drugs. Well, I invite you to think
about cardiovascular risk. Those are life-saving drugs, by the way. So, I think that really
what we’re talking about is where’s the happy medium here. What can we tolerate, and again,
forget about acute effects. This is, I’m talking about
chronic, long-term effects. You’re gonna be on this
drug minimum six months and probably for years,
that’s what I’m talking about. And so those drugs need to have
some acceptable safety level and the reason I like the oncologists is they don’t care, well, don’t
take this the wrong way, they don’t necessarily care
about the mechanism of why, they just know that if you’re
above 140, it’s not good. And, in fact, they’re confident enough that their drug’s gonna
have you live much longer. So they want the pressures to be below 140 ’cause they know the totality of the data. Forget about nuances about 120, 140. They know if you’re below
140, that’s a good thing and you’re gonna do better
than if you’re above 140. Period. And I like that level of simplicity because everybody can understand it and you don’t get lost in the
bushes in terms of subtleties. I don’t know if that
answers your question. – Yeah, thank you.
– Sure. – Yeah, uh, next question. – [Karen] Great discussion,
and Dr. Krucoff, I was so glad that you actually brought up this other piece of information
about the heart rate because so many times, you know, the drug products that we review affect not only the blood
pressure but the heart rate and as you know, the
cardiovascular risk calculator that we have doesn’t have a
place for heart rate in there. And so while we have all
these experts together, if I could get further guidance about how heart rate would increase risk in addition to the blood pressure. – So, I’m gonna jump right in on that and I really, really, really–
– Thank you. – Wanna thank you for bringing that up. This has been a pet peeve
of mine for a long time and in the original guideline that I chaired for The Kidney Foundation on blood pressure management
and kidney disease, I actually put heart rate in there as something they needed to modify and my colleague said I was an idiot and what was I talking
about, it was irrelevant as far as kidney disease goes. And it is, but it’s not irrelevant as far as heart disease goes. Stevo Julius, to his credit,
has always maintained that increased sympathetic tone is a key underlying variable. And he did an analysis of the VALUE trial where he went back and looked
at tertiles of heart rate. So he took all the people
with blood pressures of less than 140 and then
looked at heart rates 60 to 69, 70 to 79, and then 80 or greater. And then he simply looked at mortality. And guess what? If you were above 80,
even though your pressure was less than 140, you
had higher mortality, and the lowest mortality was
at a heart rate in the 60s. And that was published in The
American Journal of Cardiology because apparently it was post hoc and not sexy enough to get
into the major journals. But I think the point is that heart rate needs to be looked at and certainly for the weight loss drugs and those that modify sympathetic tone, heart rate is a big deal and
I make it at least a part of my regimen to try to get that down. – So I would agree completely ’cause I think that we have to recognize that among the– – So, Mitch, go ahead. – One of the complexities of, Karen, that I think blood pressure brings us to, in fact, we could have this
same meeting about heart rate and that’s because they’re
end common pathways for so many human signals. I mean, normal physiology of exercise has been mentioned a couple of times. We elevate our heart
rate and blood pressure, it’s good for us the way it
happens, blah, blah, blah. But if we have a fever and an infection, we elevate heart rate and
can elevate blood pressure until we go into septic shock and then we drop blood pressure. And I think it was mentioned earlier, heart rate variability is another level of when you get to continuous
evaluation signaling. So I think we’re on a slippery slope because there probably quickly could become too much requirement for too much information, that gives us too much
buried individual information in population means of
very high-density sampling of information from human beings. On the other hand, we can’t ignore that blood pressure and heart rate that where we’re seeing
changes in evaluating a new molecular entity, are
also incredibly infected by everything else in our lives. Stress levels, diet, activity levels, et cetera, and diseases, vascular disease, hormonal diseases. So somewhere we gotta get smart and I think understanding
at an individual level who are the responders? Who are the patients exposed to a compound who show us today’s signal,
elevated blood pressure? And who are the folks who are exposed to that same compound who
don’t show us that response and get some much more
personalized insight? Unless we’re gonna go down the road where we’ve been before where barrier, we provide barriers to new
entities coming forward that get failed early on
for the wrong reasons. – That’s a great point, Mitch. And I’ll go to the rest
of the panelists too. And also sort of related
to the earlier point of what’s relevant for FDA guidance versus clinical practice
guidelines and treatment. But, Philip? – Yeah, we’re hear today
to discuss blood pressure but since heart rate has come up, clearly higher heart rates
on a population base, particularly older patients
may reflect more risk, but it may have to do with
underlying physiology. We’ve actually held two public meetings that have discussed, spent
sections of those meetings discussing the data that
supports or doesn’t support drug-induced heart rate increase as being a cardiovascular risk factor. You know, for these increases
under 10 beats per minute, the data, the determination
in both of those discussions was the data really didn’t support a risk. Now, we could come back
and have another meeting to discuss that but
drug-induced heart rate as a risk factor is, you
know, the consensus was in those two discussions,
which were DIA meetings a number of years ago,
seven or eight years ago, you know, was not, did not
really indicate a clear risk. I wouldn’t jump on that bandwagon, but today we are here for blood pressure. – Okay, so, okay, so, Vasilios, go ahead and we’ll get back to–
– Yeah, a comment on, on heart rate, okay– – [Karen] So, I only mentioned heart rate, not to increase barriers,
but just to get it out there, just to throw you guys a can of worms. But, that perhaps it might be reasonable to have a line in this
guidance for industry that addresses something about heart rate even though, in that it may impact risk. – I think you make Karen out to– – I think the data doesn’t support, the data, I think it has to be dated– – [Karen] You’re telling me it’s not but Dr. Bakris is telling me that it does. – I’m thinking at a guidance level, what you may wanna do is
account for the importance of understanding the mechanism
of blood pressure elevation. This is about blood pressure. But if it’s heart rate related, if every time you see a
blood pressure elevation you see a heart rate elevation, then I’d be careful about
using hydrochlorothiazide to treat the problem. – [Greg] Vasilios? – In that context I’d
probably agree with Mitch but the heart rate is definitely a marker of something going on with
the patient with hypertension. Increased heart rate is maybe due to underlying heart failure and that’s why the patient is going fast, so he’s higher risk. It maybe indicate increased
sympathetic outflow and that’s increased risk. The problem with using
heart rate as a measure or as an endpoint to treat our patients is that there’s no study and
there are no data indicating that reducing the heart
rate benefits the patient. Just remember the MRC study again, that compared diuretic to beta blocker. The beta blocker reduced
the heart rate much but the diuretic was better
in preventing outcomes. So we don’t have any solid evidence to support using the heart rate changes as a measure for our outcomes. – Okay. Okay, comment? (person speaking of mic) Okay. – [Bob] Lag, okay. The point I was going to make was just what you made. There are a lot of data about heart rate and I’ve spent 20 some odd
years producing some of them. We know that in patients with HFrEF, reducing heart rate to between 50 and 60 leads to optimal outcomes. We know that in people
with coronary disease, chronic stable coronary disease, reducing heart rate does nothing. So, you know, one of the key points made at the beginning of this conference was we have to define in our
guidance what we need to do about what we find. And I think that heart
rate, as Mitch said, maybe it’s too much. You know, blood pressure
seems to be the topic. We gotta figure out what to do about that. One point about mechanism. I gotta tell ya, I’ve been an
advisor to the FDA since 1977. Thank you. (panelists laughing) – So you’re as old as Bob? (panelists laughing) – [Bob] No, no, but he was
the one who asked me to do it. The point is that in all those years, in all those 41 years
of however many years, that it’s 42 years, I don’t know how the mechanism
of action of any drug. I know pharmacological effects of drugs. I know what’s associated with what. But mechanism? Very little information about that, so I’d be a little careful about talking about identifying mechanisms of action as a basis for guidance. – Okay, back table. Then I’ll go to you, yeah. (person speaking off mic) – Just leave the mics on. – Yeah. – It’s not working. – [Jeff] There ya go. I wanna go back to blood
pressure again as the topic and it doesn’t have to be
part of the guidance, per se. And it’s not one-size-fits-all, but we do know as an industry, a drug development industry is
a clinical practice industry and is a regulatory component, that pediatric and adolescent
drug development and care is a key consideration. And, that’s an area that again, doesn’t have to be
covered by the guidance. We can, in fact, do some
longitudinal studies, as Mitch spoke about, where it doesn’t have
to be additional money. We can have the ability
to follow these patients over a course of time. There is a study that was
done on 17 oncologic patients that were treated when they
were eight and nine years old. Now statistically it may
not be that significant, but you could see off of that data that they developed heart disease earlier. And you could watch the trend, and they passed away earlier
than what would be projected. So, I do think as an industry that there is a different
population management than there is when we look at adult and maybe taking Mitch’s component when we start to take a look at a quote/unquote “elderly”,
whatever that would be defined at and then an adult population
and then pediatric. And again, it doesn’t have
to be part of the guidance, but it is a consideration if
we’re talking about the safety and the effect that the blood
pressure impact on safety looking at a long-term exposure. So, just wanted to clarify that. – Okay, time for one last comment here. – [Bob] I think we all agree about drugs that have a huge
blood pressure effect. But, for drugs that have a small effect, I’d like to work backwards and try and bring up whether this guidance
is gonna be helpful or not. The FDA has a choice of
approving or not approving a drug and it doesn’t sound like we’re talking about not approving a drug simply because it has a small
blood pressure effect. It doesn’t sound like we’re talking about requiring a
cardiovascular outcome trial for all of these drugs. And, it sounds more as
though it’s about making sure that the effect is characterized properly and then that the appropriate
language goes into the label so that a physician prescribing the drug has some inkling that it might
be increasing blood pressure. But, let’s, let’s ask what happens after that. Let’s say drug X is approved, it produced a two millimeter increase in systolic blood pressure, one millimeter increase in
diastolic blood pressure, you’re gonna take it for years. What is the prescribing
physician supposed to do? My presumption, from what
everyone has been saying is check the blood pressure and if their blood pressure
is in the hypertensive range, you should treat it. But isn’t that true without
there being a guidance or any language in drug’s label? Shouldn’t you be doing that anyway? – Ellis, do you wanna? – [Ellis] Sorry. I’m sorry, I don’t know your name. – [Bob] Bob Kleiman from ERT. – [Ellis] Okay, well, you’re right. So we’re not talking about
whether this is the basis for an approval or a turndown. That’s correct, we’re not
talking about outcome studies. We are talking about how
to measure the changes, how to interpret the changes, and then what to put in the label. People around the room have said oh, well you can monitor blood pressure and people do it and you treat it. I mean, there are a lot
of people in their 20s who have ADD and they go to
psychiatrists once a year or once every six months to
get their drugs refilled. Most psychiatrists don’t
have a sphygmotonometer. They just don’t. They don’t check the blood pressure. Right? And there’s not a prominent
piece in the labeling that says hey, you
know, before you refill, you know, your Concerta,
oops, I used a brand name, I’m gonna go to jail.
(audience laughing) (laughs) Okay, I better stop. So, before you use your stimulant, check the blood pressure. I, myself, am on a number of
drugs for chronic conditions. Don’t worry, I’m not
gonna die any time soon, but some increase blood
pressure, some decrease. My doctor’s office is
six blocks from here. I haven’t seen her in more than a year. Nobody’s checkin’ my blood pressure. So, that’s why it’s important
to get something in the label to alert people that hey, you know, the drug actually could
affect the blood pressure and you need to watch it. – [Bob] Well, I don’t know
that it’s so clear that putting this information in the label will necessarily change
the behavior of physicians. And all you can do is try, that’s true. But I would ask the question: Should a physician, if
they see this in the label, be more inclined not to use this drug? Because otherwise they should
just be treating hypertension and checking blood pressure more often. Then you don’t need a guidance. – Okay, comment on this? – Give it a minute, John. – [William] Okay. I think one, a lot of what’s
been talked about today is almost irrelevant in the
clinical practice setting. Number one, you have almost no chance in picking up individual small changes with how inaccurate blood
pressure’s measured in the clinic. Okay. The inter-person standard
deviation is huge. It’s even sobering looking at ambulatory blood pressure monitoring, which we mistakenly believe gives everybody a stable baseline. It gives the group a stable baseline with people basically at the extremes having random changes
that are unpredictable. I do think though it is
worthwhile to figure out what these drugs do to blood pressure and warn physicians about it
without necessarily saying they can’t be used and
giving them some guidance. But in a clinical practice, if you’re trying to pick up three or four millimeters of mercury, you’re not gonna pick it up. You’re gonna be fooled by
randomness in both directions and it’s just gonna be
really, really hard. So I do think it’s worthwhile to get the information,
get the guidance and all, but I don’t think a lot of
this has much applicability to somebody in clinical practice outside of making broad recommendations. And if they see the blood
pressure’s not at target, maybe understanding that this
may be contributing to it. You may have to live with
it and intensify therapy. – Okay, great. Thanks for those comments. That will wrap us up for this session. This was a great discussion. We’re gonna go ahead and go to lunch and come back at one o’clock. There are a number of
restaurants in the area. If you’d like a list, see
the registration table. And we will continue
to discuss these issues in the afternoon and then there’s a sort of an open public session to bring out any other
additional comments or issues that we weren’t able to address. So thank you to the
panelist and presenters for this great session. Thank you. (audience clapping) – If you, if you wanna
change what’s going on there, no FDA policy is gonna do it. I talked to, I talked to, I know, but I’m just saying the one thing that I
found most influential– – University of Pennsylvania. Milton Pressler, Vice President, Clinical Development and Operations at Pfizer Global Product Development, and Frank Rockhold,
Professor of Biostatistics and Bioinformatics at Duke University. Okay, so our first
presentation to Dr. McDowell. – Good afternoon everyone and (laughs) thank you for being here. This is the first
session in the afternoon, I hope everyone had a great lunch. So, my talk is going to share with you some of the analysis data
we have done internally at the Division of
Cardiovascular Renal Product to evaluate whether or not there is a need to include placebo
controls in ABPM studies. Oh, disclaimer. So in the current draft guidance on the section (crackling drowns
out speaker) study design, there’s a small paragraph
talking about the control group. So, the guidance clearly states that, I mean, it is desirable
to have a placebo control in ABPM study. The concern raised here is that there could be a change
in blood pressure with time that masks the effect of the drug, make the placebo control more desirable. However, there’s still
a degree of uncertainty around this topic, and
even internally FDA, some people have different
opinions on what to recommend, so that’s why there’s a box below that. So a full discussion on
this topic is encouraged. So my colleagues and I, so we decided to look
at this issue further to assess the impact
of the placebo control using the ABPM database
that we have at FDA. So, this database
contains about 16 very old anti-hypertensive studies
between 1986 to 1994 and also six more contemporary ABPM study, most of them safety studies. So of these 22 studies, 11 studies includes a placebo control. This is what we are interested in here and is included in our analysis. So a little bit more
background about this project. So the objective is to evaluate the consistence of the change from baseline in placebo
arms between studies. So as I mentioned, the 11
studies with the placebo, we include the 11 studies
with the placebo arms for a total of 456
subjects in this database. And all the subjects
had baseline ABPM data and at least one post-baseline ABPM data. Some subjects had up to
three post-baseline visits. So the mean is, duration for, among most trials about six weeks. So we look at the mean
change from baseline for both systolic and
diastolic blood pressure, and we look at the blood
pressure within three windows, 24 hour, daytime, and nighttime, which are the conventional ABPM endpoints. So this slide just show you the hourly average blood pressure. On the left is systolic blood pressure, right is diastolic blood pressure and here I’m just using
the data from two studies as an example. And so, the black line indicates the hourly average blood pressure data from an anti-hypertensive study and the yellow line indicates data in a study in normal tensive populations. So we see in the study, I don’t
know if you can see clearly, but the solid line
indicates the baseline data and the dashed line indicates
the post-baseline data. So as you can see clearly from this graph is that the diurnal blood pressure pattern is very consistent over time for baseline and placebo patients. So the dashed line and the solid line they are almost overlapping to each other. And this is the case for
hypertensive population and also for the normal
tensive population. And, this slide shows
you the systolic average, average systolic blood
pressure across studies by the population. The baseline data show in black and the post-baseline data show in yellow. And as you can clearly see
that there’s no difference in average systolic blood pressure compared with baseline and placebo visit. The similar findings also found for diastolic blood pressure. So to evaluate the consistency for change from baseline across studies, and we calculate the mean
change from baseline, we think each study, and then
we illustrate the results here in this forest plot. And as you can see that across the study, the mean estimates are
centered around zero with 95% confident intervals across zero, which indicate that, they are seeing, I mean, they are seeing no difference between baseline and placebo visit. And this result is particular for change from baseline for 24 hour
systolic blood pressure and we found similar finding for using daytime and nighttime, and we do not see the difference between baseline and post-baseline data. Again, this is the result
for change from baseline for 24 hour diastolic blood pressure. The same pattern, you can see that all the estimates
are centered around zero and they seem to suggest
that there’s no difference between baseline and post-baseline visit. We also looked at some
subgroups of interest, including age, race, and sex. And subgroup results are
in general consistent with the overall results. However, you may note that
there is a numerical higher mean change from baseline among blacks, with an average about two
millimeter mercury increase. However, I want to point out that this subgroup is quite small and in general, we do not want
to over-interpret this type, I mean, this type of subgroup analysis unless we have a very good reason supporting the observed findings. So the next slide, so up to this point our data seemed to suggest
that it is not necessary to include placebo
control in ABPM studies. And this particular analysis show you that there’s actually a price to pay. So for example, if you
want to exclude a study with a four millimeter
increase in blood pressure, one would need about 30 subjects in a single arm ABPM study and more than 100 subjects is needed if you want to include
the placebo control. This is like more than doubling the size of a single arm studies,
nearly a fourfold increase. And so whether or not
to include placebo arm is a very important element to design efficient ABPM studies, which my colleague Dr. Johannesen is going to talk about that topic in more details in the following session. So here are like some points
I would like to bring up for your attention and perhaps just to facilitate some
discussion after this talk. First of all, findings of
a lack of a placebo effect is actually not new. I mean, this finding has been reported in some small anti-hypertensive
studies about 20 years ago. So our project with a much larger size confirmed this early findings, and in addition we also found that the lack of placebo
effects also are observed in the normal tensive population. So if you think about the
nature of the study design, which it makes sense that the
placebo effect’s less likely, especially for the drugs
that are taken orally, this type of ABPM study
has minimal interaction between the patient, clinician,
and treatment environment and we all know the advantage about ABPM compared to the in-office cuff measurement in terms of reducing
the white coat syndrome. And all this, I mean, everything
from the design aspect it seems somewhat supports the finding. So the third point I want to mention is that we know that some ABPM vendors, they routinely exclude the
first few hours of data because of the concern that there may be an early placebo effect in
the earliest part of ABPM that is because the subject
may have a transient reaction when you first put on the device. However, we do not know how
consistently this approach has been done across the vendors, and from our end we
note that this approach is not clearly described in the protocol or study report we receive. So we would like to hear from some of you to share the experience on this subject. But having saying all
this, we do not think that this early placebo
effect, if it exists, I mean, have a significant impact on 24 hour average or daytime, it’s just too trivial
for the average data. So the last but not least
points that I’ve been talking about placebo controlling
ABPM studies as a whole, but are there any different concerns regarding placebo effect
in an efficacy study compared to a safety study? And so, that’s just some of the points. And the last, so, so far
although we do not think that placebo control is
necessary for most of ABPM study, but we think there are some scenarios where placebo group is very helpful. And the first is whenever we want to target high risk patient population using blood pressure as
an inclusion criteria. For example, systolic blood
pressure greater than 160. And in this case, we have to
worry about the regression to the mean phenomena. That is that if we
selection subjects based on the extreme blood pressure, so from the tail end of the distribution and then this, I mean, the patient is going to,
their blood pressure is likely to be closer to the average
for the following measurement, so that’s why placebo,
including a placebo control is important here just
to minimize the risk of regression to the mean. And another approach in the design phase is that we should always have
a separate screening visit and you shouldn’t use
the baseline ABPM data to select your subjects. So the last, the second scenario is that whenever you are interested in using ABPM study to assess
a long-term effect of drug on blood pressure then the placebo control may be helpful here to
control seasonal change. However, we do not think
that this scenario applies to type of ABPM study we are interested in because for the safety ABPM study, usually we just try to quantify
the blood pressure effect when drugs reach a steady state, which is doesn’t take that long. We are talking about
study duration in weeks. And, next. So, in summary, in our data
shows that the diurnal pattern appears very consistent over time and we do not see a
difference in blood pressure between baseline visit and, I mean, between placebo visit and corresponding baseline visit. That’s it. Thank you.
– Great. (audience clapping) Thanks, Z. So, we’ll here from Raymond.
– Oh, great. So, knowing time is always of a premium, I had one slide I wanted to show and John Black took most of
my thunder just before lunch with respect to the impact of that, assuming it shows up here somewhere. So, there should be a fan plot
somewhere in that slide set. I can keep flipping through these, but I’m probably gonna
give somebody else’s talk if I keep doing this, so.
(panelists laughing) I may have have to do
the Mr. Rogers thing here if that doesn’t show up
in a couple of seconds. While we wait for this fan plot, let me prompt your memories to rewind about five years ago in March of 2014. A couple blocks away the
American College of Cardiology was holding its annual meeting here and the results from the
renal denervation study were released. If you looked just at the
hypertension denervation arm, you would’ve seen a 14 millimeter office blood pressure reduction
and a seven millimeter ambulatory systolic
blood pressure reduction and at the time, Medtronic
had parallel reviews going with FDA and CMS, and if you looked just at the denervation effectiveness, you would’ve put a little
tick in the plus column. What boogered up that
study was that the placebo, the sham control arm had a 12 millimeter office blood pressure
reduction and a five millimeter ambulatory systolic
blood pressure reduction. So one of the caveats I wanna make sure we all have somewhere
clearly laid out here is that sometimes that placebo
arm can uncover for you (sighs) perhaps off-protocol
phenomenon that occur that can taint your
placebo, taint, T-A-I-N-T, taint your placebo control group. So we learned a large lesson from that and had to build in
safeguards to make sure that any drug usage for blood pressure in a denervation, a device trial was carefully not only
monitored by our study centers but surveilled with
urine and blood screening for any hypertensive drugs. What I would’ve shown
you had the slide come up was just the raw data
for what a, you know, what ambulatory blood pressure
monitoring data looks like before and after in a placebo group. It’s one thing to see bar graphs and to see summary data with
nice standard deviations around a thing, it’s another
thing to take everybody’s 24 hour systolic blood pressure baseline and then measure it again at three months, or six weeks, or whatever. And that fan plot looks like this. You’ve got people that go
up and people that go down. The miracle is that usually the ups and the downs balance one another. But when you do consulting for industry, especially if they’re developing a drug for a non-blood pressure
related indication, and maybe someone noticed that there was some office pressures, but maybe this is a urologic drug and who knows how well
blood pressure are done in a urologic setting, or a derm setting, or even what you heard a little earlier about a psychiatric setting, you see this office signal and say okay, we should do an ambulatory study. And you do the study,
and maybe it comes out showing less than a millimeter
of change before and after. But then you have people that
wanna take apart the data and then they look at the fan plot and they pretty much freak
out when you see people that have 20 millimeters
systolic blood pressure increases in the placebo arm over a 24 hour period. So, my plea is that when we
begin to interpret the issues with respect to the need
for a placebo control arm, always keep in mind what kind of population we’re dealing with and the fact that there’s a certain amount of biologic variability in the phenomenon like the 24 hour blood pressure that you see really well when
you actually look carefully at the data, not just the
summary statistical data. There it is. Here it is here. This is what I’m talking about. You see people that go up
20 points and down 20 points in the sham control arm. So that’s why I find a
picture’s worth a thousand words and that kind of thing. So thank you for your attention. (audience clapping) – Okay, thanks. And there was your slide. (laughs) Milton? – Thank you. Well, I would say first of all, speaking for our company, we’ve operated under the view that using ABPM did not have
a significant placebo effect. And so, Dr. McDowell’s presentation doesn’t take us by surprise. That said, one always has to, whenever one uses the word lack you have to know the precision
around the use of that and we never felt that you
could exclude a millimeter, that there might be as
much as a millimeter or thereabouts effect. I would say that our view is also that we have some concern
about being prescriptive across all indications for use. A lot of the data that we are, that we’ve been examining today are in hypertension,
diabetes, and so forth, where one might be able
to have a placebo control. But as we went through PRECISION and we had people with
intolerable arthritic pain, we didn’t feel that a placebo was ethical, particularly for months of use. So, quoting Dr. Borer, “A placebo is good, “but sometimes can’t be done.” And, so that’s, I think for
us, placebo is very valuable trying to put context onto safety. But trying to measure small
differences in blood pressure, I think we almost need to
discuss with the agency case-by-case as to what use the
ABPM study is gonna be made. If you’re trying to quantify the blood pressure effect
in a chronic condition, the placebo helps to
reinforce your finding and that might be good to include. But if you use it across the board, now you’re creating a potential barrier to developing new treatments in considering all of the various domains that which companies are developing drugs. So. Good points.
– Okay. Great, thanks for those. And, Norman, I didn’t
introduce you at the beginning but for anyone in the
room who may not know you, Norman Stockbridge, Director of Division of Cardiovascular and
Renal Products at FDA. – So, our interest is in
trying to get some information on the blood pressure effects of drugs without it being anymore burdensome than it needs to be to get a
reasonable evaluation done. And, if there is no effect, no placebo effect that you can tell by looking at a large series of trials, large numbers of placebo or untreated subjects in trials, if you then do a study with a small number of subjects in it, you will sometimes see something that is, statistically does not look like zero. But, that’s a likely to
be a statistical artifact as it is to be anything related to your study
design if you’ve done a couple of sensible things. And, Z pointed them out, that there appears to be
a true seasonal difference that you’d need to compensate for if you’re looking at studies
run over many months, that is with baseline and on treatment, separated by months. And then, because there is day-to-day variability in an individual’s blood pressure, it is subject to regression to the mean. So if you screen your trial using your baseline blood pressure, if those are the same thing, even if you did ABPM and not cuff, if you screen based on that, you will see regression to the mean and that may look like a placebo effect. I don’t know what happened
in the small device trial that seems to show a treatment, a placebo effect. I don’t know whether
it was a design feature or a statistical fluke, but I wouldn’t use that to undermine a large series
of trials that seem to say there’s little or no
placebo effect with ABPM. – And just one post script,
the hypertension-3 trial was hundreds of patients, 500 and some I think were randomized, two to denervation, one to sham. So it was actually a very
large trial, much larger than any of the individual
trials you’ve seen here. And the lesson and the
point I was trying to make is that sometimes things
happen despite good protocols to a control group,
especially when you’re dealing with something that they can self-medicate outside the protocol. That was the only caveat
I was laying on the table. – Great, thanks for that. And I think we have Frank
Rockhold on the line too to chime in. – [Frank] Yes, thank you. I am Kevin Estrom. Kevin got called away on a
family emergency on Friday. I was not able to see the Webex. Apparently Webex is only good for people who are not actually at
Duke, so I couldn’t get in. But I was able to follow the
slides from the presenter. I didn’t unfortunately see Dr. Townsend’s (crackling
drowns out speaker) slide. Just a few comments. Just to go back to Dr.
Stockbridge’s comment, clearly regression of the
mean is a good utilization of placebo for any number of reasons and he highlighted some of them. So any kind of screening, whether it’s screening for
normal or hypertension patients, a placebo would be a good use. I guess you also need to be careful about is clearly you need to
use a lot more patients, but also recognize we’re not testing the
same hypothesis here, as they’re testing it between patients verus within patient hypothesis, so it shouldn’t surprise anybody that we’re gonna use a lot more patients. It’s not actually a
direct or fair comparison ’cause you’d get a lot more information about between patient
variability in the placebo study. I have a couple observations
which may be more, you maybe talked about
these earlier in the day. I’ve heard a discussion about looking at average blood pressure during the day. If there’s interest in
looking at the variability of the blood pressure during the day, rate of rise, say from
night to daytime, peaks, ’cause if I look at the
slide that was presented by Dr. McDowell with diurnal
variation and the hypertensive looks actually different
than a normal tensive. So if there’s a reason
to wanna pull that apart that may be another reason
to wanna include placebo in these studies. The other issue is these were six week
trials, as I understand it and whether or not if
you had looked at placebo in a longer term scenario, I would’ve been able to take
some of the things apart that Dr. Stockbridge was talking about. So those are the observations
I have from Dr. McDowell and listening to Dr. Townsend
and Stockbridge’s comments. – Great. So, a lot of really good
comments from the panelists. And before I turn to the audience, I’ll give you all a
chance to react or respond to what you’ve heard from
your fellow panelists or presenters. So, Z, any sort of reaction
to the last comment? – Yeah, I’ll answer his second
question first. (laughs) – About the daily variability? – No, no, I think the second question is about whether or not we
have a look at longer studies. I mean, in our study the
duration is about six weeks, but the longest trial is about 12 weeks. And at least up to 12 weeks
we do not see a difference between the baseline
and the placebo effects. And in terms of his,
the earlier questions, and we have not looked at, I mean, you’ve seen the variability
of the specific endpoint to assess the impact of a placebo control. I mean, for us we just try to use the conventional measurement to try to see whether or not
we can observe a difference between the baseline and placebo visit. So variability, we have not done that yet. – Okay, great. Any other comments from the panel? – Just one short one.
– Sure, go. – So just keep in mind
that when we manage people, we’re typically managing an
office-related blood pressure with all of its shortcomings. When we use ambulatory
blood pressure monitoring, we’re doing a very selective procedure that typically is not reimbursed except for the sole indication
of elevated blood pressure in the absence of hypertension. So, when we try and translate whatever recommendations we
make from this day forward, we always have to bear in mind that there are some limitations in terms of using this kind
of recommendation to actual, inpatient, in-office
encounters with patients because of the different
kinds of modalities that are being used. – [Greg] Good point. Norman? – I wanted to latch on
to something that Frank, and Ray said it too, and see whether or not those of you who live in the hypertension space have some better insights into this. My impression is that for
the vast majority of drugs that lower blood pressure, you will see a within day response that looks something, that tracks somehow the plasma level versus time response, but that if you looked
for eight or 12 weeks out, you don’t see that anymore at all. For most drugs, you get an effect on the blood pressure that even if it’s not a drug that has plasma
levels significantly high throughout the inter-dosing interval, the blood pressure effect is the same. The ABPM curve is shifted up, shifted down by some number of millimeters of mercury, but the time course of that has been lost. I think that’s true of most drug classes. Is that other people’s impression as well? – George? And then we’ll go over there next. So just to let you know since we were having issues this morning, with the microphones, there’s a delay for even the light to turn on. So you press it and you wait a little bit and then the light turns
on and then it turns on. So I know that those kinds of problems, but now that should be working. – [George] The good news
is by operant conditioning, we’ve come to this point.
(panelists laughing) Let me build on Norman’s point and then I wanna get back to Ray’s points because they’re actually
related, believe it or not. The data from SYMPLICITY-3,
and we’ve looked at this in great detail at (crackling
drowns out speaker) trying to explain this, has a lot to do with behavior
that you cannot control, independent of pharmacology. The word is known as adherence. And that also relates to some
of the things that you said and when I’m talking about adherence, I’m not talking about
they’re not taking the drug. They’re still taking the
drug, but by the way, their (crackling drowns
out speaker) changed, or oh, by the way, their
sleep behavior changed. Nobody’s accounting for these variables. They assume this is a canceling out event and it’s not, by any means. This is why in the studies
we’re doing today with ABPMs, we’re using DOT ABPMs. So we’re giving the drug
like the old psych units, with the drug being taken in front us. We’re actually seeing the
patient swallow the drugs and then we put the ABPM on. So we can actually see what’s going on and they’ve been advised
about salt and whatever. In that setting, in
the short-term setting, I think it is meaningful. But I think you have to also appreciate that these are the limitations of ABPM. You also have to appreciate
the underlying vascular biology because a lot of these people
are not gonna dip at night. If you look at 24 hour,
then that’s gonna give you a different thing than daytime ABPM. So there’s a gold mine
of information in there, but it becomes very murky
unless you tease it out in a way that I think is meaningful. These variables have to
be taken into account. – Okay. Point? – Mic.
– Okay. Yeah, at the, I think there were two
back there at that table. Okay. (person speaking off mic) Oh, there we go.
(banging on mic) – [Man] Figured it out. Norman and Bob will
probably remember this, but a number of years ago
the cardiorenal division did an analysis of ambulatory
blood pressure monitoring from the placebo groups
in a large number of conventional registration studies, thousands of patients, and you came to the conclusion back then that there is in fact no placebo effect when you use ambulatory
blood pressure monitoring, almost to a point where I
seem to recall Ray Lipicky, don’t know whether it
was a formal statement or an ad hoc comment at
one of the advisory boards saying that it may not even be necessary if you use ambulatory
blood pressure monitoring in a clinical trial to
have a placebo group because what the changes
in ABPM are meaningful. Now, what Dr. McDowell has shown is it’s still along those lines as well and I think most of us, as
George just pointed out, believe that for at least
over the short haul, a matter of weeks or of maybe
up to a couple of months, ABPM is gonna be very
consistent in a placebo group. The troubling point is the point that I brought up this morning and Ray, as (crackling drowns out
speaker) told me now, for a cohort, the
reproducibility is terrific. For individuals it’s all over the place. That’s a problem. But I’m sorry, Bob. I’m sure you have a
comment to make about that. – Done. – [Bob] And then I pause. Hello?
– Yes. – [Bob] Okay. No, I wasn’t gonna comment on that. I’m impressed by what you described and what Z described that
if there’s a placebo effect, it’s not very large and so on. And if what the study is designed to do is show the blood pressure
effect of the treatment, I’m perfectly plause, I
find it perfectly plausible that the small potential placebo
effect won’t obscure that. But what we’re talking about here is safety studies, to a degree, to find out whether the drug
has a very modest effect of two or three, like that. And, my bias is that, and even in some of the studies Z showed, there were effects of that size,
two millimeters of mercury, that could tag a drug as
increasing the blood pressure when it really didn’t, which doesn’t seem a
very smart risk to take. So I just wonder about that. But, if you’re trying to show that a drug has a six millimeter of mercury effect, this isn’t gonna obscure it. I’m not remotely worried about that. So maybe ABPM alone is fine there, but I’m still worried
about the safety things, where even a couple of
millimeters the wrong way for whatever reason
(crackling drowns out speaker) for having normal blood
pressure at the beginning. I don’t know what they do. I think we need to think about that too, ’cause the safety
situation seems different. – Of course, we don’t have
studies with three placebo groups in the same protocol to compare. But I suspect that if you did, they’d show as much variability within, as the between study variability is here. So, I don’t think that the
trend of one study placebo group to look a little lower and
a little higher in those, that’s just statistical artifact. That’s not, that’s probably got nothing
to do with the trial at all. – [Bob] Well, I just
think that whole thing oughta be looked at. Where a positive study is a two or 2 1/2 millimeter
difference between the two, are you somewhat likely to
get that as a spuriosity? More than if you had a concurrent placebo. ‘Cause that would be, that would be– – The question isn’t whether–
– Be very bad. – Whether a group might look like it’s got a two millimeter effect
when there really isn’t one. That’s as likely to be true
regardless of the group. Right?
– No, no. – And to take two
samples and subtract them isn’t reasonable. – [Bob] No, I don’t
know if it’s as likely. That depends on a lot of
the assumptions you make about the population
you put into this study. As Z said and others have said, if there’s any selection of the patients and it isn’t correct, for example, if you try to exclude anybody who was too high, so you pick people who are relatively low and they go back toward their baseline, then it could look like the
blood pressure increased, and it really wouldn’t be the drug, it would’ve been back to
where they were normally. I’m just saying you gotta worry about that because a spurious finding of
positive blood pressure effect is absolutely not what anybody wants. – But all of the studies
that Z talked about had blood pressure
selection criteria in them, but they were all based on
the cuff blood pressure, not on the ABPM.
– Right. And some of them showed a
change in the placebo group too. (crackling drowns out speaker) Well, sure. But in the case that I’m describing, you’re really mostly, first of all, if it went down you
might obscure an effect leaving that aside, but
if it went up, spuriously, you’d be then targeted as a
drug that raises blood pressure, which probably is not
what most people want. – [Greg] Okay. So, Mitch, do you wanna
weigh in on this? (laughs) There ya go. (tapping on microphone)
There, there ya go. – [Mitch] With patience. So, maybe I’m missing something but it seems to me we’re
talking about two things. One is, do you need a placebo because in a population you
might sub-select a group that (crackling drowns out speaker) from a safety point of view. So, randomizing might make
that a little more visible. The other, to me, as an
ex-sort of mechanistic guy or corp lab director is
do we have the ability, the sensitivity within the test to see the threshold confidently without (crackling drowns
out speaker) threshold. And for that, I guess, my
question to the panel would be well, so why not think about just doing, using the patient as their own control with an internal crossover
and just put ’em on the drug, have ’em on placebo, and
have ’em back on the drug, and if there’s really a very fine tune, ’cause we’re using the same
methodology, the same patient, and then a two millimeter change
would probably be visible, at least something you wouldn’t miss. – Vasilios and then that table back there. – [Vasilios] I think all these
are very important points and I wanted to elaborate on the point that Temple pointed out earlier. The ABPM is probably more
reliable than office pressures because it averages overall
about 70 pressures in 24 hours. There is variability of blood
pressure during 24 hours but there’s no variability longitudinally from visit to visit and that’s applicable for office pressure and
for 24 hour monitoring. So I think the variability that we see in different studies is not so much because of the technique, it’s because of the selection
bias of the studies, because of the design of the study. (crackling drowns out
speaker) to have a cutoff for inclusion criteria like
most of the denervation studies that require a systolic above 160. You are selecting patients
on their highest visit of blood pressure office and the next time they are
doomed to have a lower pressure, so you are going to see
variability in both, in the office, in the ABPM. If you look at the denervation studies that lately were designed with
more vigorous design plans, there is no placebo effect
or there’s no difference in the placebo or in the sham arm either in office or in ABPM. So, the design is probably more (crackling drowns out
speaker) variability, rather than the technique itself. The only difference or the
benefit I see from the ABPM is that it averages a
lot of blood pressures over a 24 hour period and is more reliable and more reproducible than the office, but it is subject to the same variability from visit to visit, and in that respect we
need to take into account and probably if we’re
gonna see small differences we should be doing more ABPMs and average them after intervention to see if there’s a real difference, or even comparing the two treatment arms, we should do more than one ABPM (crackling drowns out speaker) more than one after the intervention to average them as see if
there’s a real difference. – We have time for one last comment. Okay. (tapping on microphone) Okay, there we go.
– I got it right that time. A couple things. You know, you’re talking
about the intervention trials and with the renal denervation, ya have to, I hate to say this, put that on a separate shelf. There were challenges in that trial and all of them specific
to maintaining consistency in quality of the data collection. Sites were able to do all
kinds of things on their own with the ABPM, and what you need to do and what we had done historically if you’re going to look at hypertension is similar to what Dr.
Papademetriou would say and others was that it’s very controlled in the sense that a patient comes in between six and 10 o’clock in the morning. They’re dosed in that window. So you’re able to show a delta, you’re able to show that
change based upon response to the therapy. With the renal denervation trials, that was not held to the same level. You didn’t have to have X
number of readings per hour, so I’d like to put that on the shelf as a comparison when you’re trying to look at placebo effect or not. The other component when
you’re looking at trial design, that you wanna take a consideration of, is the fact of who you
wanna include and exclude, especially if you’re gonna
look at your 24 hour readings and your day and your nights, not that night shift
people are not good people to take a look at, but you
then start to see a change in the circadian rhythm and try and compare a night shift person to a regular day person will affect how you look at that data. When you start to take a
look at managing variability, you then wanna take a look
at the number of readings that you’re averaging per hour. There’s a difference between
what’s done in clinical care to get that vision of what that individual’s
blood pressure change is versus in a clinical trial, and how many readings
should be taken per hour to be able to draw a conclusion of what is that delta
change in blood pressure. And then I think you have to look at ABPM and other blood pressure
measurement techniques as complimentary, because as Dr. Townsend and others has said, ABPM is not something that is always available, at least within the
North American US market. So we do have to tie it to something and we do have to tie it
and improve methodology in taking office blood pressure and consider things
like home blood pressure in some scenarios based upon the compound. But again, when you’re looking at safety and we’re looking to show the potential of a blood pressure signal, we do have to be a bit more rigorous in how we go about assessing and defining what is a good 24 hour ABPM. It’s not just 70% valid readings. I wanna make sure I have X
number of readings per hour over the 24 hour, with exceptions so that I can draw that conclusion. Just putting an ABPM on
and getting a 24 hour mean, you could get a 24 hour mean and be missing eight hours of data if you don’t put in criteria that’s been selected and defined
as good 24 hour monitoring. Again, more with a goal of
being able to define a delta that might indicate a
blood pressure signal. And Tzu, to your question,
there is a blip in time where at least from the
industry perspective there were two articles that came out that said that first hour was potentially a technology response. I can speak for my own
group, we haven’t done it, except for that one, there
was one hypertension drug where they felt it, but
I don’t think any of us are really removing
that first hour anymore. – Okay, great. Thanks for that comment.
– Hey, Greg? Greg? – Frank, go ahead. – [Frank] Yes, just a
couple quick comments. My response to the point
that Bob Temple made that any selects in criteria
you put into the study, whether you’re looking at
normal or hypertensive patients, will generate some degree
a regress to the mean which will get attributed to the drug if you’re not using a placebo group. And also to point out, I
quite like Mitch’s idea of the crossover design. (crackling drowns out speaker) how long you wanna measure them. It could get to be a long study if you wanted to do the
three periods of the washout, but that is one way to maybe
meet both masters there. – Okay, great. So, I’d like to thank the
panelists and presenter for a wonderful session. Believe it or not, that’s
the end of our session one. We’re gonna run right into session two, so thanks to our panelists and
I’ll invite the next ones up. (audience clapping) – Yeah, and my thanks too. And I would like to ask
the session two panelists to come on up while I’m
introducing the session. So, you might think well,
we’ve been here a lot longer than what seems like one session, but there were four parts in dealing with some of the four key issues that we laid out at the
top that were raised in the draft guidance. You all, in discussing those issues, raised some issues that I
think are gonna be relevant for the remainder of the day too, as we focus on some key design issues, especially around ambulatory
blood pressure monitoring and other methodologic issues in designing pressor studies. So, right now for session two, we’re gonna focus squarely on ambulatory blood pressure monitoring. This has come up all day, including at the end of the last session. We wanna consider how
to efficiently design ambulatory blood pressure
monitoring studies, including some discussion of
the metrics for evaluation, what to do when studies are
complicated by missing data, and how sample sizes affect
the study’s statistical power, so some of the issues that
you’ve already touched on and a bit more. And as in our previous session, we’re gonna begin with a
presentation from an FDA colleague. Lars Johannesen is a clinical analyst in the Division of Cardiovascular
and Renal Products. We’re then gonna hear perspectives
from a set of panelists, Charles Benson, Senior Medical
Director of Clinical Pharm for Diabetes with Eli Lilly, and then Frank Rockhold,
Professor of Biostatistics and Bioinformatics at Duke again, Dr. Johannesen again, and I think that covers the group. Right? Okay, great. So, Lars, over to you. – There we go, all right. So thank you for the introduction. I’ll be providing some discussion on some of the science suggestions for efficient ABPM studies. And, the usual disclaimer. So, my presentation
consists of three parts. In the first part I’ll be discussing some key protocol features
I’ll be focusing on throughout the presentation and go through two recent ABPM studies and how contrast and compare
for these key protocol features and the key protocol
features I’ll be focusing on relates to the primary endpoint. For example, it was a 24 hour mean, a daytime mean, a nighttime mean, what placebo arm was included, the number of readings per hour, and definition for valid ABPM session, a recording was valid or not, what was the criteria that
triggered a repeat ABPM session. In the second part of my presentation I’ll be discussing how these
key protocol features impact the study size (crackling
drowns out speaker) increase when the true effect is
zero, in other words power and how the impact a false negative rate, in other words diffraction
studies have failed to exclude a four
millimeter mercury increase when the true effect is four. And again, I’m using four millimeter here just as an example
throughout my presentation and this part here will be assessed using a simulation study. And in the last part I’m gonna
put all the piece together and some suggestions for considerations when designing efficient ABPM studies. So, Z talked a little bit about this, we are just gonna repeat some key points. We have an ABPM research
database that consists of 22 studies conducted
between 1986 and 2017. Each of these studies have
less than 20 subjects per arm. Of the 22 studies, 11 of
them included a placebo arm for a total of 456 subjects, and the median study
duration in this database was approximately six weeks. Each study in our database
included between one and three post-baseline
visits and a median number of three measurements
per hour at nighttime and four measurements per
hour during the daytime. So, at the beginning of my presentation I just wanna contrast and compare two recent ABPM studies I
found in the literature, PRECISION-ABPM and the SYNERGY study. Both studies used a 24 hour
mean systolic ABPM measurement as a primary, one included a placebo (crackling drowns out speaker) morning. In terms of the number
of subjects per arm, it ranged 170 to 180. In the PRECISION there was accounting for dropouts in the study. In terms of the number
of measurements per hour, it ranged a little bit more
densely during the daytime of two measurements per hour, to three and four during the
nighttime with slight variation in the day and nighttime
(crackling drowns out speaker). I was unable to find a criteria
for the PRECISION study that triggered a repeat session, but in the SYNERGY it seemed
to be this 70% criteria that was proposed in a recent, by The European Society for Hypertension. I’ll be elaborating on
that a little bit later on in my presentation. So before I discuss the
simulation set up we did I just wanna show a few, show an example of what drug
effects you see in database because I need to simulate studies, I need to have a sense of
what drug effects we saw in our database. So what I’m showing here on the left part, I have, that button does not work, okay. I’ll try and point then. So, on the left part I
have time on the X axis and I have an hourly average of systolic blood pressure on the Y axis, with orange being post-baseline
and black being baseline. And the point I’m trying
to make in the left panel is that for this drug, the
delta on an hourly basis is approximately constant
over the 24 hours, whereas I have another
example on the right where I’m showing a, is an extreme example of a time-varying change where the change between the black baseline
and post-baseline in orange is minimal in the early
morning and late at night, compared to the parts during the day. Again, the most common pattern
was a constant-shift pattern, which is what I’m showing on
the left panel on the slide. With that in mind we set
up a simulation study to evaluate these key protocol features and how they impact the number of subjects required to exclude a four
millimeter of mercury effect or the numbers as they
fail to exclude an effect, and again I’m using four
millimeter of mercury as an exemplar. As noted before, the most
common pattern we saw of drug increases was a constant increase and for that reason I’m
simulating a drug increase in that manner. Lastly, these simulations were conducted with a systolic, diastolic,
and mean arterial pressure (crackling drowns out speaker)
in just the systolic pressure because of similar
findings across the board. So, in this simulation study I calculated three
different time averages, at the 24 average or
the conventional night and daytime average, which I’m showing on the right hand portion so I’m not gonna elaborate any further. All right, so we saw the
(crackling drowns out speaker) and Dr. McDowell’s presentation
in the previous section, but I just wanna pause a little bit and elaborate on the further
results and also methodology. So what I did was I
simulated a thousand studies, for example, of 20 subjects on active and 20 on placebo, or 20 on
active and zero on placebo and it repeated the exercise
over and over and over for a different number of subjects, and in both arms (crackling
drowns out speaker) of these similar studies excluded a four millimeter of mercury effect with no drug effect added. And then I plotted diffractional studies that managed to do this as a function of total number across both arms. And the point I’m trying to make here, acknowledging it’s a different hypothesis, that is a approximate
quadrupling of the sample size when comparing it with placebo and (crackling drowns out speaker). And again, this is for the 24 hour systolic average blood pressure. The same holds true for the
daytime and the nighttime again, approximately a quadrupling
of the sample size needed to maintain 80% power. And for reference I’m
including a standard deviation of the change from
baseline in our database on the right hand portion of this slide. (crackling drowns out speaker)
I wanna focus on later to determine if an ABPM session is valid or if it needs to be repeated. And what I’m showing on the
right hand portion of this slide is the, oh, it’s still there, are five criteria proposed by European Society For Hypertension. The point I wanna make here is that (crackling drowns out speaker) around the timing of the missing data, but more a number of measurements within either a 24 hour, within a day or nighttime, respectively. Additionally, the authors of this, of the paper with this, note that the recommendation
are not based on firm data. (crackling drowns out
speaker) simulation study to evaluate the impact
of the missing data, I wanted to explore what
missing data patterns look like in our database. So this is a fairly basic slide and I’m just gonna walk
through it carefully. On this slide I’m showing
missing data patterns for baseline to the left,
post-baseline the middle, and second post-baseline to the right. And each panel (crackling
drowns out speaker) have subjects rank, or
from least missing data to more missing data on the Y axis, with the same ordering between visits, and I have the two conventional periods of day and nighttime, color coded in the blue and
the green areas respectively. And then in each panel when
there’s a little black dot that represents a missing hour (crackling drowns out speaker) and when you look at the length of these missing data segments, they appear to be isolated. In other words, most of the missing data is one hour here and one hour there. Another point I need to make here is that like with most ABPM studies, there was a criteria
triggering a repeat session. Again here, I’m looking at data before the repeat was
taken and looking at data that triggered the
repeat for some subjects. And then other point is there are slightly more
missing data at night. So based on this observation, I wanted to evaluate the
impact of missing data on precision and accuracy
of the treatment effect in my simulations, and the way I did that
baseline simulation model, I dropped hourly data at random. In other words, if I’m
simulating dropping one hour, I might drop 1:00 p.m. at one visit and 1:00 a.m. another visit, and I repeat that over and over again, and then do it with two, do it with three, do it with four, and repeat
that, so on and so forth. That allows me to
evaluate how dropping data impacts accuracy, which is
on the chart on the left and precision on the chart on the right. So what we’re seeing here, and this was based on 24
hour average blood pressure with 100 subjects systolic. So what I’m showing on
the left hand portion is as we go from least missing
data to more missing data, when I’m simulating a true mean effect of four millimeter
mercury, there’s no bias in that estimated treatment effect (crackling drowns out
speaker) more and more data. However, when we look in
the right hand portion where I’m showing the standard errors of the precision of this
estimated mean effect, as we go from least missing
data to more missing data, that’s a loss in precision. So the key take away here
is that when you miss data, it’s not gonna bias your estimate but it’s gonna decrease your precision. And in both panels I’m showing a vertical (crackling drowns out speaker) I discussed earlier from the European Society of Hypertension. So I also looked at how this impacts power in inconclusive studies. So now I have the same data
but now I’m simulating studies without a drug effect in left panel and we can see there is an impact as we lose more and more
data, we start to have, we start to miss power, to lose power, that’s because of the loss of precision. And again, I’m starting in this 30% as I had previously discussed. On the other hand, when
I look at the fraction of inconclusive studies, that’s a study with a true mean effect of
four millimeter mercury when it did not exclude a millimeter
of mercury effect, we see the same pattern
but the other way around as we miss more and more data (crackling drowns out speaker) and more inconclusive studies. So the point I’m trying to make here today is that (coughs) sorry, under the conditions simulated that this 30% criteria
might be too conservative. The last protocol feature I wanna discuss relates to number of measurements per hour (crackling drowns out speaker) 24 hour and I’m gonna be
focusing on within hour. To assess that I took a
subset of our database, with all the intricate
here you had to have at least three measurements
per hour every 24 hours at baseline and post-baseline. From this subset of
patients I simulated studies with the three measurements per hour, two measurements per hour,
and one measurement per hour. And then I calculated
the standard deviation of change from baseline for
when you have one measurement, two measurements, and three measurements. And then I translated that into
number of subjects required to maintain an 80% power and
that’s what I’m showing here. For daytime on the left,
nighttime in the middle, and, sorry, daytime in the middle, and nighttime on the right. And I’m showing (crackling
drowns out speaker) in black to two measurements in orange as a slightly decreased number of subjects to maintain the 80% power, but has less of benefit
going from two in orange to three in blue. I just wanna synthesize all
the results of the simulation in terms of some suggestions
for consideration, in terms of the primary endpoint (crackling drowns out speaker)
metric seems reasonable. Again, the caveat is there is time-varying effects anticipated. Another metric might
need to be considered. In terms of placebo control, it’s not terribly surprising
but there’s a cost of the inclusion of placebo suggests not including a placebo. And as was discussed by Dr.
McDowell in a previous session, there might be some consideration (crackling drowns out speaker)
support inclusion of placebo, for example, the duration of study or if ABPM is used in screening data. This regression to the mean
was discussed previously. In terms of the threshold, I used four millimeter
mercury for my presentation. I only did that as an exemplar, and as we discussed earlier this morning, you would need to depend
on therapeutic area and target patient population. In terms of measurements per
hour, I did two per hour. And when it comes to validity criteria, under the conditions I simulated, 50% missing data seems
reasonable within the individual. Again, that would depend on what type of, if it has a time-varying effect, of the drug effect, missing data, a validity criteria that
minimizes missing data around peak effect might be warranted. (crackling drowns out speaker) thank my FDA colleagues for their input on this presentation. (audience clapping) – Thanks very much, Lars. Next we’re wanna hear
perspectives from our panelists, starting with Charles. – Well, thanks for the invitation. I want to start (crackling
drowns out speaker) I’m not actually a blood
pressure specialist but have been in early
phase drug development for almost 20 years, and that is where this potential blood pressure
study is being proposed to be used as an early phase. About half of those 20 years
I’ve spent working on QT, and although there are differences, I think there are reasonable similarities between QT and blood pressure that might be illustrative to
have a brief discussion on. And so it was mentioned this morning that blood pressure is not QT. One of the audience members said that QT is instantaneous and is not dependent upon patient characteristics
and that’s not correct. QT, just like blood pressure, does increase your risk of torsades, it does not mandatorily give you torsades once QT goes out to a certain amount. So for example, you
could have a QT of 500, like long QT syndrome patients, and (crackling drowns out
speaker) percent per year risk of torsades, because of course, there are also characteristics
which are necessary, such as hypokalemia, or hypomagnesemia, or structural heart defects
which also increase your risk. So, what we potentially
learned from working on QT for the last 20 years or so, was from a drug development standpoint was that the E14, which was written, was helpful actually
because they did come up with a cutoff of 10 milliseconds. Now certainly, we talked
for a very long time about 10 milliseconds and
whether that’s appropriate and whether or not there’s also risk at five milliseconds or three milliseconds or maybe even one millisecond if you extrapolate out across
millions in a population. However, we had to cut it off some place and we thought that 10 milliseconds was a reasonable place to cut it off because although you can extrapolate risk across millions of patients, you also have to extrapolate benefit across millions of patients as well and we hope the number
needed to treat for ABPM at the same point is
considerably less than millions and so you then would balance that cutoff from the benefits that you would see from a drug with potential issues. The other piece that we’re
still actually dealing with from a QT standpoint is that we learned that it’s more effective, rather than to have a
single clinical cutoff is to look at the totality of evidence across the potential for a signal, including both the non-clinical
and the early clinical, and even what we understand,
sometimes not so much, about mechanism of action
of any particular drug in that combination of these
(crackling drowns out speaker) that are operating characteristics from false positives and false negatives or positive/negative predictive value than we would have otherwise. And lastly, what we found was that if we’ve done a reasonable job in our discovery and chemists
in coming up with a compound (crackling drowns out
speaker) the prior probability going into whatever study you do is low and that influences the positive and negative predictive value. It’s fairly simple statistics, but if we have a low prior probability and even if we do a very good
study with 80% specificity, the false positive rate then actually eclipses the true positives by quite a bit and that’s what we saw with some of our early
thorough QT trials as well. And lastly, one I thought I’d touch on is we also learned that instead of just doing a single dose and crossover, we get a lot better data
coming out of these studies looking at dose-response,
or even more importantly constant (crackling drowns out speaker). And so I have a few slides
with a couple of more comments. I won’t spend too much time
on, they’re almost sort of stream of consciousness types of comment. If you could pull up the
slides it would be great. If not, then I might even– – [Mark] Here, just click along. – Oh, well, if I can do it
myself that’s even better. (crackling drowns out speaker) All right, so, good. So then I wanted to
start it and we’ve talked about a number of these things already, but trying to actually decide
what the signal is here, that certainly with Terfenadine and QT it really got us going that we really needed something
here to spur us forward, but whether or not the size of the signal how that we are now seeing
from a blood pressure problem is worth the additional study here I think is a reasonable question. So are we really missing these signals, especially when we put into account the non-clinical data that we’re getting from our dog studies or other
animal studies, for example. The next bit I want to touch on, in the guidance it says we should do a thorough blood pressure study and again, if you’re looking at things with very low prior probabilities, you might have problems
with false positives rather than true positives so for example, very well characterized antibodies, is that still going to need
a blood pressure study? No non-clinical signal whatsoever, very low risk populations, all of these things probably
should be taken into account. The question on my next bullet is about whether we could
use something similar to what we’re doing now with QT, that if we definitely do measure
blood pressure in phase I and maybe if we start doing triplicates, could we increase the power response and utilize the advantages
of pushing the dose in your phase I studies
from a dose-response to find small signals
that we could then use to trigger whether you
would do an additional study like an ABPM study, and that’s actually the way we’re doing it now at Lilly. So if we see something non-clinical or if we think that this is a mechanism (crackling drowns out speaker) issues, or if we see something in our
early phase clinical trials then we go off and do an
ABPM study from there. The next bullet is just sort of a warning. We wanna be sure that this
is an effective approach that we don’t want to repeat
some of the issues we had with thorough QT in that it turned out to be extremely (crackling
drowns out speaker) to a problem which probably did
not have a huge overall risk for the populations that we were studying. And so, because most drugs don’t work and you’re doing these in early phase, the cost can have a knock on effect and make it not very cost effective. We’ve already gone of the last bullet which has to do with the sum, negative and positive predictive values with a small prior probability. So, just a couple of responses to some of the questions
that were in the guidance. On of them had to do with what you should, (crackling drowns out speaker) from a, when this is done, first of all. Does it need to be done early? And then we’ve already made the point about this then exaggerates the cost. What the cutoff should be. At two to three millimeters, it’s certainly stated several times. However, looking at what we just saw from a sample size
standpoint to even four, including four millimeters takes a very large study
and can then result with still a very substantial
false positive rate here used to cut off 80% and if
you multiply that, again, times the prior probabilities, it can lead to very poor
positive predictive value. (crackling drowns out speaker) Is there a specific increase
across development programs that would be cause for concern? Or should each program
have its own threshold? And so, I was stating that perhaps we should try and do both. Again, similar to what we’ve done with QT, if we have a reasonable
screening criterion based upon our totality of evidence of what we have in the early
(crackling drowns out speaker) then we can use that to
then go off and say well, we need to identify and characterize this blood pressure response
further with an ABPM, which then that cutoff could be based upon all the things we
talked about this morning which has to do with the benefit/risk in the patient population and so on. (crackling drowns out speaker) it’s probably one of my weakest, and that’s whether or
not you have to do it in the patient population. Historically we’ve done well with healthy volunteers in phase I. Certainly there are some
caveats in this case. I think I would go back
with my original argument that if we push the dose and really exaggerate the pharmacology, many times we can see these signals in healthy volunteers, but
(crackling drowns out speaker) perhaps not my strongest. So from just a conclusion standpoint, blood pressure is certainly important but there’s lots of important
signals we try and find in early phase drug development. And so going and saying we have to do a dedicated blood pressure study, I think is very analogous to being told that we had to do a thorough QT trial (crackling drowns out speaker) and now we can just look at that within our current phase I
approaches, and it works well. Exploration by signals with non-clinical mechanisms
of action early phase has been reasonably
successful historically. We have to be careful with
unintended consequences, similar to the thorough QT with unknown study characteristics that might have poor
positive predictive values. Thanks. (audience clapping) – [Mark] Charles, thanks for
all the specific comments. So, um, Norm, you want
me to go to you next? Or you wanna–
– I’ll make only a couple of, a couple of comments.
– Okay. – Clearly, our intent is not to make this any more onerous than it needs to be to serve the purpose. And, it’s not clear you can get patients to do more than two ABPM, 24 hour ABPMs, so study designs that involve having
people do three or four probably aren’t gonna be well tolerated. We’re obviously thinking about how to make sure that any given ABPM doesn’t get thrown out
for arbitrary reasons and we’re trying to make sure that we’ve sort of thought
through the very reasons and we’ve trying to make sure that we’ve sort of thought
through what missing data, ’cause there’s a lot of
missing data possible with an ABPM recording. There presumably are gonna
be other things to learn. The application of
exposure-response modeling is appropriate to think about. And there may be other things that we’ll, that we’ll find along the way. So, you know, I think this will evolve, hopefully faster than our
thinking did on QT business. So, I’d like to get comments from the group here. – Thanks, Norm. And, Frank, if you can still hear us, any comments to add at this point? May have lost Frank. All right, well, in that case, let me open this up to
comments and questions from those of you who all are here. So, thank you all for teeing this up well. Please. – [Bob] Bob Kleiman. Bob Kleiman again. Professor Benson, (laughs) you seem to think that a
lot of very useful data can be gotten from doing
blood pressure assessments in early phase in healthy volunteers. I heard you say not always but that would be effective. But, this morning I heard someone say not to do it in healthy volunteers. Is there any consensus
for these two reasons (static drowns out speaker). – Are you asking me this question? I’m not an agency person.
(panelists laughing) So, yeah, right. So to clarify, I heard this morning that there are certainly, and I understand that
there are differences between an 80 year old and a 20 year old from a blood pressure standpoint. What I would like to have
further discussion on is whether we can get around
some of those drawbacks by exploring a full, as wide of a dose range as possible, ’cause that’s what we’ve done in the past with other potential safety signals. In phase I we push the
dose as far as we believe is reasonable but in order to cover those perhaps more rare events
or those smaller signals that you’ll see later down the line and I thought perhaps
the same thing could work in healthy volunteers in blood pressure. Don’t know. – [Mark] Um, Lars or Norm wanna comment? – Yeah, I think this was, this was a large focus of the meeting that we held in 2012 as well. And I suspect that unlike the QT business where what you really wanted to know was whether or not you were
blocking pERG channels. Here the system is more complicated and I think most of us think that you actually do need
to study the blood pressure, pressor effects in the target population, not in healthies, to get reasonable inferences
about what’s likely to be important. – So, just one more clarification. We would just use it as a screening tool, like we do with all of our phase I. So certainly phase I doesn’t tell you what’s gonna happen in your patients, but we think that that would
then give you the information to say now you should go off
and look in your patients to see what happens because
we’ve seen a signal. You can do it from your
non-clinical as well actually. – So just a question. I suppose it might work for
some mechanisms and not other because some mechanisms
might take a long time to sort of evolve whereas
some might lend them to more to short-term payout and I think that’s one
of the starkest contrasts to the QT business, is we’re
dealing with a biomarker that reacts fairly quickly
to changes in plasma level and might not necessarily be true of all blood pressure
mechanisms, I suspect. – [Charles] I agree, but
we’ve got plenty of biomarkers that we measure in phase I–
– True. – Which have delayed tachyphylaxis and other delayed responses. – Thanks, I know there are
a number of more comments in the room. I think we do have Frank on the line. Frank, can you hear me? – [Frank] Yeah. I think I was, Nicholas had me muted. So just a couple quick
comments, that’s it. Going to Johannesen’s simulation study, which is very interesting,
he was commenting on the amount of missing
data that you could tolerate and comparing to the European guidelines, and it’s very sensible
assumptions he made. I guess I have a question. You assumed that these
observations were missing at random and yet also pointed out that
they’re at greater likelihood they’re gonna be missing
in the middle of the night. So if I’d missed all of
my nighttime measurements, I could still qualify for 50% of having my expected
number of measurements. Did you simulate what
effect that would have? The sensitivity of the study? – So I did not simulate that directly but I did do the same just
looking at daytime average and nighttime average and again, it’s just an effect of
four millimeter constant throughout the day, then
the patterns look the same if you look at daytime compared to 24 hour and daytime average would be the same as you could lose all your nighttime data. It depends on the assumptions of what kind of drug effect
you’re operating with. Because if you have a time course too and you’re missing all the
data around a peak time then that’s pretty important. – And, Frank, any other comments? – [Frank] I guess I would add that, yeah, just that I would add
the same comment I had before and this probably was discussed earlier. We keep focusing on the mean
but I guess I’m wondering if the variability based
on the pharmacology, whatever the agent is
matters, during the day and the rate of rise of
change if that matters. And I think that’s the topic
for future work, I suppose. – I think it’s a very important point and was sort of alluded to this morning but I think it’d be worth
talking a little bit about, a little bit more about is
the influence of outliers on outcome. So again, from a QT standpoint, certainly the 10 milliseconds
doesn’t do anything but we’re looking for those patients that have significant outliers
greater than 60 milliseconds before we really start to run into risk and it certainly doesn’t
translate perfectly to blood pressure but likely, a two millimeter increase
for any particular patient doesn’t do anything, but if
there’s an outlier effect, you could choose 10, 20, whatever, that’s when you really
start getting into trouble from a stroke standpoint. And could we do something
similar from a blood pressure, where we again use some real, some sensitive cutoff to then go off and look for a number of outliers in your phase III population? – Okay. Thanks. Any further comments on this? Doug, did you have, then I’ll go around. – [Doug] Charles, I’m struck
by your parallels to the QT. I agree with you, I think
there’s some important lessons that we wanna make sure we capture there. You’re suggesting looking at
early phase trials intensely in a way that we’re thinking, we’re working with QT to do. I don’t know that that’s been
summarized systematically, where someone has gone
back and looked at drugs to determine whether early
phase studies like that are capable of detecting the kinds of mean blood pressure changes
we’re interested in here. Do you know about that? And the reason I’m asking
is my heuristics of course are captured by my own recollections. So, I was the cardiovascular
reviewer for, yes, Celecoxib. And the phase I studies looked wonderful. I don’t remember if there were some small, non-trivial effects that were identified, but there was no signal seen there at all. And so, I just wonder if we’ve looked at that systematically. ‘Cause it’d be a great thing if we have. – I agree. I think it needs to be done. Not that I know of and to my knowledge, no one’s doing, for example,
even triplicate measurements on blood pressure in phase I and that turned out to be a real boon from the standpoint of variability. You improve your, your accuracy quite a bit by just measuring just
a few times in a row. – [Mark] Okay. Gonna get some more comments in here. Go ahead.
– Okay. Yeah, and I think, Charles,
you’re right on target. I mean, we’re already doing blood pressure in the phase I trials. If you look at the schedule
of events, they’re there. And right now we’re seeing a trend and an increase in both
the SAD and the MAD for people being more rigorous or diligent in their blood pressure assessment. Their rigor and diligence has
been on the QT and the ECG, but if we’re looking to get
the biggest bang for the buck out of conducting a study,
you already got it in there and just be a bit more
rigorous in how you can do it. Now, it is a little bit different than how you build in that triplicate. It’s not always feasible in some cases and it has to do with clin/pharm unit and how you’re gonna collect
that triplicate data. Is somebody gonna write it
down or things like that, but there is definite, I think, value, but in the same light I absolutely think that our focus here on
the guidance in that area is taking a look at
the patient population. But we’re making a mistake
if we don’t get that data in the healthy volunteers
and standardize it in an environment which
allows us to do it. The other comment, Lars, with
that 50% number scares me a bit. Again, only because we don’t
wanna do this study twice, if you’re looking at a patient population. And I think we can be rigorous
in a controlled environment because the question that’s going to come, we’ve talked about it now,
is it two millimeters, is it four, if we do not do a good job of rigorous data collection so that I can look at hour
one, hour two, hour three and compare it within that
patient and across patients, and I think, Charles,
you might have said it, if I miss my nighttime
data, I still get 50%, or the gentleman on the phone. If I miss my nighttime, I still get 50%. What does that do to my analysis? So I think in this phase Ib, phase II, I think in many of the studies that myself and other
labs are involved with, we are getting a bit more rigorous to what is an acceptable hour. And in some cases you can’t
repeat, we understand that. You wanna get that first exposure because some drugs you
do see that response right off the bat. And so you have to be more flexible and it shouldn’t limit a patient’s
participation in a trial, but I do think we have
to take the best approach to getting quality data right off the bat. – No, no, I completely
agree with the points. What I was just trying
to make, the argument was that maybe if you’re looking
for a 24 hour average effect, the 70% is probably a
little bit too conservative, that was the point I was trying to make. I completely agree with you, and that was the other
point I was trying to make was that if you are
interested in time course then having a missing
data coordinate that says I want 20 during the day and
I want seven during the night, you could miss hours consecutive and then you would lose that completely. That was one of the
arguments I wanted to make. So I completely agree with
the point you’re making. – [Mark] I know there are a
number of additional comments. Going to the back first
and then over here. – [Daichi] Daichi Shimbo from Columbia. One question is the
distribution of daytime is actually much wider
than the distribution of nighttime blood pressure. So is that a fair comparison
of four millimeters of mercury during the awake period compared
to the nighttime period? I wonder if you consider
that in your power analysis? – I did not, I use the four throughout because in some of the, in my research database
most often I saw a constant during a 24 hours, I simulated
it under those conditions. Of course it would change the moment you put a time course into it. And again, this becomes a
little bit more complicated because a drug that
has an average increase over 24 hour of four isn’t’ the same as something that peaks at eight and goes down to two. So it gets very complicated very quickly. – [Man] Another thing I wanna highlight, it’s an interesting debate, the ESH guidelines are not inherent. They just got together and
decided what the minimums are. On one hand you could argue that since during a typical clinic visit you’re only taking three readings, though why isn’t on ABPM
the minimum number four? As opposed to 50%. Do you understand what I’m saying? – So this is where it gets, you’re talking about a number
of reading per hour, right? That’s another interesting
contrast to our QT, QT, drawing some parallels to. In the QT space you pay for replicates. We pay a different price
when they come in ABPM. I suppose we do a lot of readings, let’s just say 10 per hour, you’re gonna probably see people drop out or they don’t sleep or they
take the device off I suppose. So the cost for replicates
is a little bit different in the ABPM space and QT. I don’t know where or how
that operationally works out but it’s a little bit
different in terms of cost or where we see in other scenarios we’ve been drawing parallels to. – [Mark] Okay. (Mark laughs) There you go. Good. – [Milton] Milton Pressler, Pfizer. I certainly would like to
agree with Dr. Benson very much about utilizing the normal volunteers, but I’m not sure that it
will completely de-risk the questions we’re bringing up here, for a couple of reasons. One is, what do you do about drugs that are too toxic to
give to normal volunteers? Those are the ones often in oncology that we also have the most difficulty doing even thorough QT
studies because of the, trying to control the study. So that’s first point. And second, thank you Dr. Throckmorton for reminding us about Celecoxib. (audience laughing) I also just remind
everybody about Torcetrapib, our CETP inhibitor. That did not have any
blood pressure effects in normal volunteers,
that we could discern. It was only when we went into the patients that those blood pressure effects emerged. So, we have to be careful that in trying to make assumptions
from normal volunteers. – Yeah, point’s well taken
but we are doing patients much more often in phase
I, probably over half of my phase I studies now are in patients. Diabetes might be an exception and certainly in oncology it’s the rule. – I know we’ve got a number
of additional comments over here. – [Ellis] Yeah. Dr. Benson, I was
intrigued by your comments, many of which I hadn’t
really thought about, but now I’m thinking about them. So, I understand the parallels between a thorough blood pressure study and a thorough QT study but
if the thorough QT study was used to make some
regulatory decisions, some company made decisions
and killed various drugs because of signals, here is not, the idea isn’t to do a
thorough blood pressure study to make or break a drug, it’s really for labeling. So, what the FDA would like
to be able to do is say here’s a drug for osteoarthritis. Your typical 60 year old woman at the to be marketed
dose can expect an effect of X millimeters of mercury. It’s not helpful to say
that normal volunteers in their 20s, at a dose of eight times X had in fact a eight millimeter increase. So, it’s very different
from the thorough QT study. – So let me push back on that a little bit because the issue for us, even with QT, we were never told that we’re gonna have our drug rejected for QT. It just was gonna result in some labeling. But, labeling is everything and so if we can’t get a label that says our osteoporosis drug will
not impact blood pressure, we won’t go forward. And so therefore, we have
to know what study to design in order to avoid a label that says we’ve got a
blood pressure effect. And therefore now we’re
back to this question of how to design that study. Well, either we can do
an ABPM with 120 patients for every drug that we’ve got
going forward from now on, which is onerous and we’re
trying to avoid being onerous, or else we try and figure
out a different way to do it. And as I mentioned, I don’t
know whether what I mentioned about the healthies or doing
it in patients will work or not but I think it’s worth think about, at least thinking about that mindset that we’re gonna actually have to start excluding some
signal, not trying, (stammers) you could say well it’s just gonna be a little bit of a label,
but that’s not how my boss is gonna interpret it. – Got a lively discussion going here. There are a couple of
comments over on this side, and then I wanna go over to that table and back to the middle. Yeah, go ahead. – [Philip] Philip Sager, Stanford. Coming back to the healthy
volunteers versus patients, I agree with your comments totally about doing it in patients. There are examples, numerous examples, particularly drugs that may
affect sodium homeostasis that you would not pick
up in healthy volunteers. And just to remind people,
earlier was saw some data from a drug used for urinary incontinence where the patients had
really a minimal effect, much smaller, 1/4 of
what was seen actually in healthy volunteers. So there clearly are differences. And, Charles, I do understand that we all wanna have clean labels but I’m not sure this kind
of blood pressure labeling would translate into the
same kind of reaction that having the QT labeling would. I don’t feel this is as
big of an issue from, let’s say competitive
standpoint as maybe the QT is. – Well maybe ’cause you understand QT and you understand blood pressure but, actually knowing, my
understanding of both of those, I’d be much more worried
about a blood pressure effect than a QT effect. – One more comment here and then, clearly gonna generate
some good discussion. – [Preston] From a reality of day-to-day operationalizing this as far as healthy normals
versus to-be-targeted patients, I’m Preston Dunnmon, I’m one
of the clinical reviewers in Division of Cardiovascular
and Renal Products. I consult frequently for
other divisions around O & D on this very subject, and what I see is that the early phase studies, if you go look at the protocols, hypertension is excluded, any kind of cardiovascular
disease is excluded, et cetera, et cetera, and the patients are dramatically younger. So you got age knocked out and the people who are on
anti-hypertensives knocked out. And to extrapolate those people to how an older group of
people on anti-hypertensives are gonna behave when they get
exposed to a pressor effect, I think you can’t accurately do that, both as far as the mean
of the population effect as well as some fairly
dramatic outlier effects that can occur. – That’s why you study, if you saw something
you’d study it thoroughly, with either your ABPM inpatients or in your phase III population. Whether it would work, just
as a trigger I don’t know. So, you might be right. – I appreciate everyone’s patience. We’re gonna try to get
all these comments in. And please, go ahead over here. – [Man] This is more of a question. Since it was proposed to
run these studies more in the target population, would you then also control dietary,
particular salt intake? Would you also control for
routine exercise programs? Because some of these studies
are gonna last for some time. They are not just done over quickly. What are the criteria which
you think should be part of the control mechanisms in such studies? – Any comments? – I have no idea what to do. – If others have comments on that too, we’ve got a few minutes left. I’m gonna try to get
in as many as possible. Please go ahead. – [William] So, I just wanna
make sure I heard this right. This whole notion that if you see a signal in blood pressure you’re not
gonna develop a compound, for somebody outside the industry seems really kind of
out there and because, for example, in osteoarthritis
we probably wouldn’t have virtually anything on the
market if that were the case. So since these drugs are not
being pulled off the market and you’re only, few of ’em have warnings, maybe it’s just ’cause
I’m not in the industry but that strikes me as
nirvana seeking a compound that has no off-target
effects that in all, and for blood pressure, as a clinician, I worry a lot more about QT
than I will blood pressure because even though it may be infrequent, if it happens, it can be definitely lethal and compartmentalized to
that people with those risks. Whereas blood pressure, we know how to mitigate the risk of blood
pressure by controlling. So if I know a drug raises blood pressure, the patient comes to see me, and if they can’t get off of it, I intensify their treatment,
diet, lifestyle, whatever. But, it just strikes me
as really kinda tough to say I wouldn’t develop a compound just ’cause you raise the blood pressure two or three or four
millimeters of mercury if it’s highly beneficial. – The reality of our current situation is that drugs have to be
more and more effective and more safe in order
to get them paid for. And that’s just the way it is. And then we don’t need to
argue QT versus blood pressure but there’s also asymptomatic
blood pressure increases that you would worry
about that you don’t see in your doctor’s office or that you get over-diagnosed in the
doctor’s office as well, so it’s not an easy problem, either way. – [Vasilios] Yeah, a little comment on Dr. Benson’s question earlier, what kind of study will be needed to rule out increasing blood
pressure with your new drug before you market it. Let me propose another way
of assessing the effects of the new drugs on blood pressure rather than doing repeated ABPMs that are probably cumbersome,
patients don’t like them, and industry dislikes
them because of the cost. Recently you know there was a
very important study published by Dr. Cushman and
coworkers, the SPRINT study that use a unique way of
measuring blood pressure that was very carefully done. The patient seated in a quiet room, attended or unattended,
with a nurse present or not, and the patient waiting for five minutes with their legs uncrossed,
and the back supported, that’s the right way
of sitting the patient waiting to have his
blood pressure checked. And this study gave us
tremendously good results and impressive outcomes with
blood pressure redaction. This method of assessing blood pressure has been discussed and others like it, although they dispute
it, although they think it measures the blood pressure too low, others they don’t think so. We and others actually compared with other types of blood
pressure in the office, with ABPM, with home pressures, and we found that actually,
if it is well done, the SPRINT way of measuring blood pressure is equivalent to
ambulatory blood pressure. In fact, we just published
the first paper in JAHA with 146 patients and
actually the daytime ABPM was identical, although there
was variability with ABPM. So the SPRINT way of
measuring blood pressure I think is reliable, is
reproducible, is easier to do. And if you do it two or three times to exclude the visit-to-visit variability can give you probably
results equivalent to ABPM, easier to get, cheaper, and
more acceptable to the patients. – Didn’t they use a
replicate design in that? Or no, they did, right? Yeah. And, again, that’s analogous
to what we learned with QT, that we learned that
you had to do it right. Lay ’em down and keep ’em quiet. It was not easy but,
figure it out, eventually. – Yeah, other comments? A very lively discussion. So we are just about on
time, which is nice. (laughs) Later on in the day. We’re gonna take a break
for about 10 minutes or so and then reconvene. So I guess we’re a little, I guess we are running a little
bit late, but not too bad. So, maybe 2:55 aim to start again. And thank you all very much
for a stimulating discussion. (audience clapping) Ask our panelists, for
this session three panel to come on up, that’s Mitch Krucoff, Philip Sager, Michael Weber, and then William White’s
gonna be joining us by phone. While you all are coming up, let me introduce this last session. So we’ve covered a lot
of methodologic issues during the course of the day. In this session we wanna highlight some particular methodologic concerns and outstanding issues in the
realm of pressor study design. For this effort, for this session we’re gonna have two presenters to highlight two different issues, followed by some reactions from our panel, and then input from those
of you who are here with us in the room and joining online. The two presentations
are by Daichi Shimbo, who is Associate Professor of Medicine and the Ewig Clinical Scholar at the Columbia University Medical Center and Daichi is going to present on ambulatory blood pressure
monitoring and risk assessment. After that, we have a presentation from Rajnikanth Madabushi, who’s team lead for guidance and policy in the Office of Clinical
Pharmacology at CDER and Rajnikanth will present on clinical pharmacology considerations pertaining to pressor study design. So, we’ll get these discussions going now. If you’ll look at your agenda, the next session after this is a session on open audience feedback, which obviously might blend into this one, but I want you to be thinking
ahead for any questions or any topics you wanna
make sure are addressed. Before we wrap up today,
we will have some time to do those topics, even
if it doesn’t quite fit into the schedule of this session. But right now, let me turn to Daichi for the first presentation. – Good afternoon. Can you hear me in the back? – And if you try to speak kinda directly into the microphone it helps. – So I’m gonna give you
a five minute overview about is blood pressure and ambulatory blood pressure monitoring the right metric for assessing risk. I do have some disclosures. I don’t have any financial
conflicts of interest but I am currently vice-chair on an upcoming in press
scientific statement from the American Heart Association on blood pressure measurement. I’m also chair of a new policy statement on self-measured blood pressure and I’m a voting member on AAMI, which sets standards for
blood pressure measurement. I will say that I, my
opinions are mine only and do not reflect the
American Heart Association, the CDC, or the AMA. So I think you know this. There are two primary ways to
measure office blood pressure. This has been the
mainstay of the diagnosis and treatment of hypertension. There’s the auscultatory method and the oscillometric method. If I had more time I’d
probably bore you to tears about the differences between those two, but I will say that currently, the oscillometric method is now becoming probably the standard in clinical practice due to environmental concerns because of the auscultatory method which is typically in the past been done by the mercury column. I think the issue is that
this kind of method relies on making sure that you
have well-trained staff, and I only list some of these here but I think you all know this, which is that there are
multiple quality control issues about letting the air
out, there’s digit bias, failure to allow for five minutes of rest, patient position, and so forth, and something that I’m prone
to, which is expectation bias. As a cardiologist, I keep
taking that blood pressure until I get the value that I like. (audience laughing) I think the other issue is
that there’s a small number of readings that you typically
take in clinical practice, and it’s usually one or two, and even the best research studies, you’re taking at most
three and possibly four. And if you actually look
at recent guidelines, in order for you to get
really a good estimate of office blood pressure, you’re not only suppose
to take multiple readings at a single visit, but you’re also supposed to take it
across multiple visits. So, for example, the 2017 AHA guidelines on high blood pressure actually
mandates two plus readings at a single visit and having
it on two plus visits, and then you average
all of those readings. Data from my shop, there’s an investigator named Ian Kronish, actually showed that more confidence is gained by actually increasing
the number of visits rather than increasing the
number of readings per visit. And that’s particularly important because if you’re
looking at safety issues, really, can you rely on a single
office blood pressure visit with maybe two readings to estimate whether a drug is safe or not? The other thing I wanna tell you about is something about ecologic validity, particularly when it comes
to out-of-office monitoring. It’s very interesting, even when done under the
best of circumstances, you’re trying to do blood
pressure on a validated, standardized way in the office setting, but aren’t you really trying to estimate what the person’s blood pressure
is in the real environment? So the example I used to give is, or I always give is, if a FedEx driver has a clinic blood pressure of 110 over 60 but yet is lifting heavy
boxes during the entire day, sweating, and their blood
pressure is 200 over 100, I would argue to you that the
out-of-office blood pressure is actually more an
ecologic blood pressure than the well-conducted clinic. So in fact, we have two different methods of actually looking at
out-of-office monitoring. Of course, I’m focusing on ambulatory blood pressure monitoring, but there’s another one called home blood pressure monitoring. It’s an estimate of what we
call ecologic blood pressure or true blood pressure. It’s less reliant on clinical personnel, particularly ambulatory. Ambulatory method is
actually an automated method. Just keeps going off on a set frequency and you’re not reliant
on actually the patient or even clinical staff
to actually set that up. And importantly, there’s
more readings on ambulatory than there is on a single office visit. So this is the patient. You can see the circled
red is the systolic, the blue line is systolic
and the red line is diastolic and the circle is the office
blood pressure of this person. And then 24 hour ambulatory
monitoring was placed and you can see that you can get daytime, during the awake period,
there’s a blood pressure dip at night which you’re
aware about during sleep, and there’s a morning surge
when someone wakes up. And in fact, you can average
the daytime or awake readings to get average daytime. You can average the 24 hour readings, which is the readings
across a 24 hour period. You can discuss whether or not you’re gonna weight those averages during the awake and sleep period. And of course, averaging
the nighttime periods. Just to show you this is, you get it, that it’s outside, it’s ecologic, and in addition, you’re
taking many more readings that you would typically have obtained during a single office visit. This is just primary data. This is a meta-analysis. In the black box actually is estimates, the hazard ratios for, the
rows are different kinds of cardiovascular outcomes. You can see there’s mean 24 hour, mean daytime, and mean sleep. The black box is before you adjust for office blood pressure. You can all see very
consistent increased risk of cardiovascular events per increase in 24 hour, daytime, and then
nighttime blood pressure. The red box is after you adjust
for office blood pressure. It’s important to signify,
these are blood pressures that were done very well
in a research setting. And you can see here
that the hazard ratios are only slightly attenuated,
to signify that in fact, out-of-office monitoring,
in this case blood pressure, ambulatory blood pressure monitoring is a significant predictor
of cardiovascular events. There’s a lot of these
kind of meta-analysis. What’s very interesting
is in these meta-analyses, office no longer becomes
a significant predictor when you adjust for both
office and ambulatory. And that’s not just an
issue of multicollinearity because there’s actually
a poor correlation between office and any of these measures on ambulatory monitoring. We can also put people
into different categories. If you have high and low
clinic blood pressure and high and low ambulatory, I think you know these categories. If you’re high on both,
clinic and ambulatory, you’re what’s considered to
be sustained hypertension and if you’re normal on both, that’s considered to be
sustained normotension. And these categories that sort of cross are the discording categories. I think you all know white coat. White coat is high clinic
and normal ambulatory. And then masked hypertension
is normal clinic and high ambulatory. I’m only showing you
this, I don’t have time to go through all the
data, but you see the ones that are in black font. The data are pretty consistent that the ones in black font
are actually associated with the highest cardiovascular risk, and the ones in white font
are actually associated with the lowest cardiovascular risk. And I show you this, if that’s the case, do you see that ambulatory is actually the most important predictor
of cardiovascular events than clinic, because really,
the high-risk categories are determined merely by ambulatory. Now, I have to be thorough. I was asked to discuss multiple
metrics of blood pressure. The other out-of-office monitoring is actually self-measured blood pressure or home blood pressure monitoring. Slightly different. It’s not an automated,
it’s semi-automatic. The patient actually has to press it, measure their blood pressure
frequent times a day, and then get an estimate like we see here. This is a person that actually
measured a blood pressure two times a day. At each occasion took two readings and then did it across a two week period. So it is slightly
different than ambulatory, where in ambulatory you’re getting it over a 24 hour period. This is getting it over days to weeks. And, I don’t have data to show you, a lot of data to show you just
because of the time limit. The data mimicking home
cardiovascular events and office to cardiovascular events mimics the data on ambulatory, which is that home systolic
and home diastolic, both are associated with an increased risk of cardiovascular events but office is not a significant predictor
of cardiovascular events when you adjust for home blood pressure. So in this reason, at
least in the United States, we’re now beginning to embrace, actually for the routine
diagnosis and treatment of hypertension, it really
should really not only rest on clinic blood pressure but also on out-of-office monitoring. This is a level of, class
of recommendation one from the ACC/AHA guidelines in 2017. Now, I’m often asked which one is better for cardiovascular risk. We did a meta-analysis actually and we found that the data are too small to actually say that one
is superior to the other. I will say that most of
the body of evidence, if you just look at the evidence
across the past 40 years, actually heavily weights
toward ambulatory. It’s not that ambulatory’s
better than home, it’s just that there’s more evidence. In fact, that’s why for most reasons ambulatory is the defacto
reference standard, at least for blood pressure monitoring. So this is my last slide. I was asked to talk about what
are the evidence base gaps. I have to be careful, I have esteemed randomized
control trial scientists in the audience, and in fact, there are actually very few
randomized controlled trials where people are randomizing
to treating their office to a particular goal versus
treating their ambulatory to a specific goal. We just don’t have those outcome studies. However, I would argue to you that’s probably not the right question, whether treating ambulatory
blood pressure is good or bad, because I believe, personally, if you have high blood pressure, it’s bad and lower blood pressure is better. I think the better question is
what are the treatment goals, what are the thresholds by
which you have to treat to, and in fact, the thresholds
on ambulatory and home are virtually all dependent
on observational studies, they are not dependent on CVD events from randomized controlled trials. So that’s a huge evidence based gap. The second thing is,
I just highlighted is, yes, I’m here standing at an ambulatory blood
pressure monitoring conference, but really we don’t have empiric data to suggest that it’s superior to home, it’s just that there’s just more evidence with ambulatory compared to home. And then I’ve heard a lot of people talk about dipping or nighttime blood pressure. It’s an important debate
because there’s lots of evidence to suggest that nighttime
blood pressure during sleep is a significant predictor of outcome, independent of the awake blood pressure. The problem is that we don’t have many randomized controlled trials where what we call chronotherapy, or treating nighttime blood pressure, actually is better than
treating daytime blood pressure. Now there is one trial called
the MAPEC Study from Spain, but that’s really it. And in fact, I only note this because if this is an epiphenomena rather than a causal factor, then really do we have to
do ambulatory monitoring during the 24 hour period? It may be better just to do
it during the awake period, which is as you showed,
actually has less missing data. So I just wanted to throw that out there. Thank you. – Thank you. (audience clapping) And next is Raj. Thanks very much, Daichi.
– Sure. – Good afternoon everyone. Thank you very much for inviting and providing me with an opportunity to provide some thoughts on the clinical pharmacology
considerations here. Many of you might know clinical pharmacology’s a
multidisciplinary science that is concerned with translation of the relationship
between drugs and humans. It’s a very simplified definition. There are many sub-disciplines
under this feat, many of which are very
relevant for designing studies, conducting analysis, and in forming use. Some of these are pharmacology, which talks about the mechanism of action, pharmacokinetics, the time
course of blood levels, pharmacodynamics, the time
course of blood efforts, the drug effects as
well as the relationship between the drug exposures
as well as the effects and many others. For the purposes of today’s presentation I will focus on two aspects, which is the pharmacology
and exposure-response and I’ll provide some considerations. I’m happy that the
previous session happened because they are going
to be some of the themes that will be reflected here. First, most important is the concept of understanding the pharmacology of how these drugs manifest
their pressor effects. We have some understanding
of some of these mechanisms based on what we have seen in past. Broadly, I have put them
into three categories, but many of them do have mixed mechanisms. For starters there are
the central acting agents which are generally associated
with sympathomimetics, or your antidepressants, or stimulants, and also associated with some
of your weight loss drugs. The latter of those are also associated with some of the heart rate effects. The other is a classic salt or sodium or water retention phenomena, generally associated with the NSAIDs and some oral contraceptives which contain estrogens and progestins. And third is the category
which is involved with the nitric oxide activity, the reduction of it, or the
endothelin-1 production increase which we know the best examples
are the VEGF inhibitors, also associated with some of
the calcineurin inhibitors, as well as the glucocorticoids. And there are some which
exhibit mixed mechanisms, parts of all these three mechanisms. It’s very important to have
a priority understanding of what might be the mechanistic basis for some of these pressors
because I think they are going to help and form some of the
study design considerations, mainly with respect to study duration, or even choosing the study population, which has been the focus of
discussion earlier in the day. So, how is pharmacology important for figuring out study duration? There are some mechanisms
which evolve very quickly. In those situations, shorter duration studies may be possible. For example, for duloxetine,
there was a study which showed that as
early as around four days, the pressor effects could be demonstrated. It was almost like a QT kind of a study wherein it was titrated to to reach to the supratherapeutic dose and with every titration within four days, the study straight levels of
the blood pressure effects could be demonstrated. The other example is with the
tyrosine kinase inhibitor, sunitinib, wherein within one week, blood pressure effects were demonstrated, not only in normotensive patients but also in hypertensive patients, and it was maintained at
that particular level. So in those situations,
shorter duration studies can be envisioned. However, for some mechanisms,
this is not possible. The classic example is that of ibuprofen. This goes back, and while this
is around a 1987 publication where they did an eight day study with 2400 milligrams per
day, 800 milligrams TID, and they did 24 hours
blood pressure monitoring. At the end of the day, ibuprofen
did not show any difference compared to placebo in pressor effects. However, we heard from
the PRECISION trial, clearly, roughly around a four millimeter of mercury increase. So, some effects which are slow evolving, short duration studies
may not be possible. So, having some understanding
of how these pressor effects are manifested is going to be very useful to inform the study duration. Similar thought process
could also be put for identifying the study population. Those which are fast
evolving are those which are, those mechanisms which are prevalent in a particular population,
might be a population of choice because they would be more sensitive for manifesting these effects. The examples are those which act through the sympathomimetic pathways, wherein healthy volunteers
which do demonstrate this kind of a pathway are going
to be more discriminating or sensitive in demonstrating the pharmacological mechanism of action. We saw this for mirabegron, we
also saw this for duloxetine and many other stimulants wherein healthy volunteer
studies were possible, to be discriminating. However, for the ibuprofen situation where it is on the sodium retention, a healthy volunteer study
could not be discriminating because there are compensative pathways which are going to mask the signal, compared to those of a target population where these pathways are impaired, and as such, they are
going to amplify the signal and we are going to be
able to discriminate. So, pharmacology or the understanding of how these pressor effects manifest is going to be very useful to inform the study design elements. Now I’ll switch to the second topic, the concept of dose-response
or exposure-response and why it is important. Primarily we are looking
for small, modest effects, we are not looking for large effects which would be self-evident. So the evidentiary basis
for calling something as a potential pressor,
having supporting information is going to be very critical. For example, if a dose or a concentration response
relationship is demonstrated, that could be very powerful evidence that the drug in question is pressor, or you could call it as. I’m showing here an
example which we presented at one of the advisory committee
meetings for mirabegron for overactive bladder, this was also presented
earlier in the day. On the left, I’m showing
you three different doses, 50, 100, and 200 milligrams once a day. There is a clear dose
ordering or arranging and this was from a thorough QT study where moxifloxacin was
used as a positive control but was used as a control
or a negative control for blood pressure purposes. Clearly there is a rank ordering here. They also had collected
PK for the purposes of QT, concentration QT relationships and we took a look at it
and we were able to show a clear, positive slope. It is possible in some situations, especially for something
such as fast-evolving. In a sensitive population,
one would be able to utilize dose-response or
exposure-response relationships to be able to identify a drug, whether it demonstrates
pressor properties or not. However, there are some caveats. There are lot of learnings we have from the concentration QT experience, but not all of them are
directly translatable or applicable in the same fashion, maybe there are some principles. For example, when the
effects are slowly evolving, we cannot utilize the
one-to-one relationship with the concentration
such as entire time course to be able to derive a
relationship like we have in QT because the effects are very proximal and a time course could be utilized there. In this particular case
for slow evolving effects, maybe somebody measures off exposures, should be utilized to try to understand what is this nature of relationship, maybe it is in area under the
plasma concentration curve or an average concentration
at steady-state or a trough that could be utilized. There are a couple of advantages of having characterized an
exposure-response relationship for pressor effects. This would be very useful
for trying to project what might happen in a
supratherapeutic kind of a situation, specifically with respect
to drug interactions wherein if the drug in
question is a potential victim and there are going to be exposures beyond what is on an average that is seen at the clinically relevant doses, there might be a
possibility to extrapolate what might be the effects
in that situation. However, this might not be possible to extrapolate to unstudied
populations of interest. It’s still unclear, or there is enough of a
concern of the ability to extrapolate to different populations. It’s not sure whether we can use a healthy volunteer relationship and assume that that
exposure-response relationship will be similar or it will be steeper or it will be blunted. So, there is some uncertainty. However, having exposure-response does provide you with information whether the drug can
manifest pressor effects. So in summary, I think clinical
pharmacology considerations could be very useful in not
only designing the study, such as the study duration
or study population, but could also provide valuable evidence, whether a new molecular entity
has pressor effects or not in a sensitive and
discriminative population, and may have the ability to project to what might happen at a
supratherapeutic exposure, a concept that is similar to that of QT, QT assessment in some relationships. That’s a summary. I would like to thank
many people who helped me in coming up with these talks. Thank you.
– Thanks, Raj. (audience clapping) All right, so we’ve had two presentations on different methodologic
issues in pressor study design. I’d like to turn to our panel to kick off the discussion. Mitch, we’ll start with you. – Well, Mark, bear with me
but I’m a little troubled sitting here toward the end of the day because on the one hand I think we risk tipping into kind of flailing over two millimeters of mercury and at the same time I think
we’ve seen unequivocal evidence in a population base that
there’s a darn good reason we’re talking about this
guidance all together because at that level,
at a population base, we see a significant rise in events and meaningful clinical, bad events. The trouble is the gap in between because blood pressure,
as we mentioned earlier, like heart rate, is a signal that’s driven from multiple different
directions in a human being, one of which is drug exposure, and whether that correlates with the whole story and
the outcomes or part. If you get down to an individual level, we’ve heard a lot about
the technical challenges, in a single individual, really understanding whether this is a two millimeter of blood
pressure changes, that’s one. I think clinically as a clinician, if my patient’s pressure is
160 or 162, I don’t care, they’re both bad. And if it’s 120 or 122,
am I really gonna say he’s at higher risk? I don’t think so. And patient confusion
if the label says well, this drug’s been demonstrated
to have a 2.5 millimeter rise in blood pressure, is that
gonna be confusing or helpful? And I think it’s in that disconnect between what we know is
the population based reason that we’re here, which is
a public health rationale to talk about a safety
concern and a guidance toward that concern. On the flip side, also as
a backside of potentially debilitating a lot of innovation, and if it’s not really carefully crafted, leaving the sense that
here’s another Pandora’s box, here’s a door that opens
with a minimal signal, meaningful signal, but
that comes with what that follows that? More burden, more, and
I think we have heard we could be back to where we were at 2004 when Janet started Critical Path Program because rising costs and
dropping productivity were barrier-based approaches to our clinical evaluation of stuff. So, I think concentrating
on the high risk side, where really is the high,
high, high risk domain, is key number one to keeping this right. But I think the second part
maybe is, what worries me more is that a lot of the
conversation today, to me, and particularly when we get
to this part of the discussion, where do we go methodologically, is incredibly conservative. I mean, in the 21st century,
the tools that we need to take a signal that’s driven by multiple sources like blood pressure and figure out what the drug’s role and risk and benefit really are, we’re not talking about
the right toolkits. If you have to do multiple
perspective clinical trials and then flail the data into
extrapolated overextension, we’re in the wrong zone. The toolkits in the 21st
century for this kind of thing are mobile devices and
patient-centered approaches, and patient-empowered internet sites that have 450,000 patients in well-identified clinical syndromes, diabetes, Parkinson’s,
patients like me, et cetera, who are already putting their experiences and showing their
experiences with one another. We have Google and all
of these big data owners who use all their firepower
for marketing data. Well, why not tap some of that, and my erstwhile boss full-time is now my half-time boss,
but his other half-time is what if you put a public health app on those same resources
instead of marketing. And to what good could we use the ability that if in a sort of a standard,
pre-clinical evaluation a drug appears to raise a blood pressure by two millimeters of mercury, we have public health oriented resources that could really help us sort out, based on what and in whom
and who really is high risk, and if we need sub-sub-subsets, patients who have vascular
disease and hypertension and a history of an MI,
they’re really the high risk. Why do we talk about 100 of them as opposed to 100,000 of them, potentially taking a new blockbuster drug? So, I really look for
that as the frame shift of where do we tap on
behalf of the public health? And the ripple effects of that are, ’cause it was mentioned several times, pick your number but a half
to two-thirds of human beings in the United States
with high blood pressure are undetected. And frankly, in this country, our children by the
time they reach puberty are already fat and
hypertensive at an alarming rate compared to any other country
on the face of the earth. And we’re gonna start
treating them with pills, very early in life. How do we sort out lifestyle
versus pills versus, so there’s a very big
public health resource here that is essentially is untapped, and all we’re doing, based on
the discussion I’m hearing, is potentially opening a door that dumps the responsibilities
back on industry. Why don’t we think a little more broadly about how all of us and the whole question of undetected hypertension
plus drug safety risk could be sorted out by
thinking more widely about our ability to lock and
link in perspective trials that may have a safety signal into public health information resources that could help us sort out really what that safety signal means. – [Mark] Thanks very much, Mitch. Next is Phil. (laughs) – No, Mitch, you raise some great points. However, I do think the direction of having this kind of guidance, to me, makes a lot of sense, that this is an area where we really haven’t
paid close enough attention to blood pressure effects
in clinical development unless they’re pretty extreme. Having said that, we’ve
spent a lot of time kind of alluding to a
two millimeter increase. I think the guidance needs to really be kind of drug and patient specific and maybe it is a two millimeter
increase for some drugs in very high risk populations but a different threshold
in different populations. I don’t think there is any one set number. You know, it does, having a
careful evaluation makes sense. I think we talked a little
bit about healthy volunteers, patient populations, I
think there’s been a lot of data presented and discussions today that do focus on this needing to at least be definitively
done in a patient population, but it seems to me adding ambulatory blood
pressure monitoring in a phase II study is not really all that onerous or difficult, and it’s a study that
has to be done anyway. However, we’ve also
discussed healthy volunteers, phase I, and that is an area we’re, well I don’t think it’s definitive having, evaluating blood pressure
more carefully in phase I than it’s probably done now,
potentially in triplicate, having a standardized protocol for blood pressure assessments. Seeing a signal early
there could be very useful in designing the future
clinical development program, but typically the blood pressure
data we’d see in phase I is really hard to make anything of. And again, even for a drug that really affects blood pressure, you may not see it in healthy volunteers. But if you do, and potentially using
exposure-response modeling, Raj, you talked about, I mean, if there is a signal and
it does have a relationship that that might be helpful. Of course, there are
a number of mechanisms where you would not expect to see an exposure-response relationship because of delayed effects, so you’re not gonna have
a relationship to PK. But still I think, to me it makes me think that we probably should be looking somewhat more carefully in phase I, that’s really easy to do. It’s a very minor adjustment to how we currently do phase I trials. ‘Cause if we see a signal,
that could be helpful in terms of for the drug development. Thank you. – [Mark] Did you wanna respond quickly? – Yeah. So, you’re correct. Not every mechanism is going
to manifest that quickly. Even for slow-evolving drug effects, if they appropriate
exposure metric is available at the steady state of the
blood pressure effects, an evidence of a positive slope is going to be very powerful information to say that okay, there is
a drug related component, despite all the noise
and all the difficulty in having a precise measurement
that we are talking, so that’s going to be
useful no matter what. But the way one goes about planning for evaluating the
exposure-response will change from mechanism to mechanism. – [Mitch] Exactly, that’s right. – [Mark] Thank you. Michael? – Yes, well, first of all I do agree that if we’re looking for
modest or small changes in blood pressure then
placebo controlled studies with ambulatory blood pressure
monitoring are appropriate. And I don’t think anyone has expressed an opinion different from that. I do think though that
the way we manage this and label this problem has
to go directly to patients. I agree very much with
what Vasilios Papademetriou said earlier that well
measured blood pressures, away from the office or in the office can be a very good alternative. Anyone who’s done ambulatory
blood pressure monitoring on himself or herself knows that this is not a procedure
you want to more than twice, and even that is pushing it. It’s not comfortable, it’s not easy. The SPRINT method, I believe,
it should become the standard in everybody’s office, and
I feel that very strongly. It’s not hard to do, it’s
not an expensive device. It’s about a $500 device. It’s completely automated. Everyone in their office,
even in busy offices, there’s always an examining room where you can put a patient
for five or 10 minutes and let the machine do its work and give you a reliable measurement. In Canada this has been the standard now for a couple of years and I think it’ll soon
be the mandatory method for obtaining blood pressure
measurements in that country. And I think we should be
encouraging that here as well. I think this issue of some drugs causing
increases in blood pressure should be labeled and
just as the FDA now says about anti-hypertensive agents, that high blood pressure is associated with an increased risk
of cardiovascular events and that treatment of high blood pressure or reduction of high blood pressure reduces the probability of
strokes and heart attacks, we should have the
converse labeling for drugs where we think there is an
increase in blood pressure. We should simply say flat out increases in blood pressure are likely to increase the probability of strokes and heart attacks and
other cardiovascular events and anything that can be done
to prevent this from happening should be encouraged. We should label drugs that have blood pressure
raising potential, even NSAIDs, right on the
label where patients can see it that this drug can raise blood pressure and it is recommended that
you check your blood pressure on a regular basis, whatever
the wording would be. Home blood pressure monitors
now are easy to use, they’re reliable, and we
should strongly encourage that. Daichi, you showed that, you called it the
ecological blood pressure and I think that’s terrific. People don’t have to get obsessed about measuring their blood pressure but if they measure it
once or twice a week and maybe after a while
if it’s consistent, once a month or every couple
of months, that’s super. And if they see a trend they don’t like, they know to talk to the doctor and I think we can initiate that process through labeling and
education from the FDA. I think that is a big step forward. And I think by sharing
this problem with patients, we’re gonna make major progress. – Thanks very much, Michael. I think we still have
Dr. White on the phone. William, are you still with us? – [William] I’ve been with you all day, not just in spirit.
(panelists laughing) – Please go ahead. – [William] I just want you to know that this census has
dropped from about 122 to 80 on the webcast, I’ve been following that, but we’ve still got a pretty
good group listening in. I’ve been in pain for the last three hours because I haven’t been able to speak about many of the comments that were made that in a sense, for
having been in this field for almost 40 years,
and like Michael Weber, having done ambulatory monitoring in research and practice for I don’t know, 35 of those years, I started to sense a lot of reinvention of the
wheel in some of the comments. So what I decided I will do right now is focus primarily on the guidance that Norman asked for comments on in ways to make this as
non-onerous as possible in the appropriate situation. Keeping in mind that doing a study in order to evaluate whether a drug does or does not raise blood pressure into a clinically relevant level is different from clinical practice, it’s different from epidemiology, it’s different from pharmacology. It’s basically a simple test. So here’s my comment. Standardized clinic blood
pressures in this scenario are far better than home blood pressures. Home blood pressures are not reliable in clinical trials of short durations to evaluate for a drug effect. It’s less likely to pick up the signal. Of course, the standardized
clinical blood pressures have to be done properly. I don’t care if they’re
attended or they’re not, if they’re done properly
they’re going to relate very closely to daytime
ambulatory pressures. There’s a lot of data, a lot
that shows that to be the case. Second point, really placebo is not needed if you really just wanna
determine whether a drug at treatment versus baseline
raises blood pressure. What’s important to control
is the environment of the test on those two occasions
in short-term studies. If one day somebody’s laying
around watching TV all day long and the other day
they’re working actively, you’re going to have a very substantially
different blood pressure because of that, so the
condition of the behavior of the patient has to be controlled, and I’m not advocating
for inpatient studies for ambulatory monitoring. I think that’s a terrible idea because you completely reduce
or remove the circadian rhythm or circadian biology of
blood pressure by doing so. I’m in total agreement
with what Mike just said that two ambulatory blood pressure studies is all that people can really tolerate for retention into the study and if you wanna study a drug that’s got a typical half-life, you probably can do it in four weeks. Crossover, nice, but reduces
retention and compliance. Gotta worry about washout
and all that stuff. If you had to rule out a two millimeter difference
in blood pressure in an ambulatory monitoring study, you have a huge sample
size to contend with. I heard somebody from
I think Eli Lilly say that 120 patients for
every drug was onerous. Well this would be a hell of
a lot higher than 120 patients in order to rule out a two
millimeter effect of a drug versus baseline, or even versus placebo. And then what would
you do if you found out that 150 drugs on the
market raise blood pressure by two millimeters of mercury like Tylenol and low doses of NSAIDs,
and so forth and so on? What are we gonna do
about all those drugs? Do we label all of them? Do we even know how to label that? I don’t really think we have
an answer for that question. And finally, the idea
of healthy volunteers versus patients with the target disease, I am in agreement, healthy volunteer data will not be satisfactory
for a situation like this. If you know what the target is, that’s the study the
population should be utilizing and then if the population
is a low-risk population, not very likely to have hypertension, not very like to have diabetes, not very likely to be old, then study them I don’t disagree that
enhancing the population with those populations
that increase the chance of bringing out the signal are reasonable, but not if the patients are
never gonna take the drug. So I think that we have to
be thoughtful about that when requesting sponsors to increase this in that sub-population
when that’s the population who will never be treated by that drug. So those are some of my comments, not just from the last session, by the way both speakers
did a beautiful job, but it’s my comments from
the last three sessions because I finally got the stage. Thanks a lot.
(audience laughing) – All right, Dr. White. Thank you very much. Let me ask, any reactions to the comments that we’ve heard so far from
those of you on the panel? Very good discussion. Okay, I would like to open this up again to those of you in the room for comments on the topics we’ve talked
about here, keeping in mind that if there’s something
else you’d like to talk about, we do have a session coming
up right after this one. Sid, is that you? Okay, great. Yes. (tapping on microphone) – [Man] There we go. Just a small comment about
Dr. Shimbo’s discussion, and it seems to me that here we’re trying to determine what a drug does. So, is there any real difference between a drug that raises
someone’s office blood pressure from 130 to 140 and it raises their
ecological blood pressure from 120 to 130? Either way we know the
drug is having an effect. – Yeah, one thing I didn’t say and it was partly built into my question is the distribution of ambulatory is actually narrower than
the distribution of office. And I only mention that because that actually has implications
for power analyses, because again, a four millimeter
office change is different than a four millimeter
daytime blood pressure change in ambulatory. The second thing is, and I don’t know the literature on meds that
are non-antihypertensives, but antihypertensive drugs actually have differential effects on office compared to ambulatory, which is that there’s a
greater effect on office than on daytime ambulatory monitoring. So, I’m just bringing that up. There’s slightly different
kinds of oranges. (laughs) You understand what I’m saying? I have nothing else to say. – Thank you. (laughs) We’ll see. Other questions? Yeah. (laughs) – On the bottom. – So, back there first then
up here in the front, Robert. Yeah, go ahead. – [Chris] Hi, I’m Chris O’Connor, I’m actually a heart
failure clinical trialist so we actually, in our class for patients like to see two millimeters
of blood pressure elevation in our patients as a positive in a visit. (panelists laughing) But, I really like the
comments made by Dr. White and others about the SPRINT methodology. I mean, I think if we’re
thinking about the number of development programs that are out there for the whole wide range,
it’s going to be very hard to be embedding large sub-studies of ambulatory blood pressure
measurements of sample sizes that are gonna be sufficient. And so I think getting our
clinicians and our colleagues to use the SPRINT methodology, which we should be using in practice because it has afforded
such a great benefit, this is the way we should
be practicing medicine now, and we shouldn’t be saying listen, the SPRINT methodology
interrupts office flow, it’s too cumbersome to do, it’s probably the most
important thing we would do in a clinical visit with a patient. So, providing guidance
around the SPRINT methodology I think would be a pathway, a swim lane that we could start with potentially that would
not be onerous, I think, for drug development. – I have a comment.
– In my opinion. – Oh, yeah, do you wanna comment? – Bill Cushman’s in the
room so I’ll defer to him but you gotta be careful
about the SPRINT method because the SPRINT method is often thought to be an unattended approach. A person’s put in a room,
the coordinator leaves, pushes a button, it’s automated, measure three to five blood pressures, but in fact, it wasn’t done
in all patients in SPRINT, in all visits. And so, I know people often
say it’s a SPRINT method but we currently don’t know
actually if there’s a benefit to actually using the unattended approach where you’re putting someone in the room compared to if you were just standing in front of the patient and taking a good, standardized blood pressure measurement. And if you look at the
Canadian guidelines, they’re based on data
where they’re comparing a sort of a casual, crummy
clinic blood pressure to this unattended method. And this unattended
method typically is lower and the argument would be oh, gee, it eliminates white coat hypertension. The problem with it is
that in those studies, the comparator group wasn’t
blood pressure well done in a standardized approach,
just like in the AHA guidelines, and oftentimes participants
were not randomized. It was often the crummy blood
pressure that came first, followed by the unattended
office blood pressure. For that reason in the
Canadian guidelines, although I do like this approach, it’s actually a level D evidence, which is that they are recommending it, but it actually has very
poor evidence in its support. And I think that more work
has to be done in this area. I think there is promising data, (laughs) but I don’t think it’s there yet to actually say that it
should be implemented in clinical practice– – [Frederick] I’ll let Vasilios
respond first and then– – That’s just my response. – [Vasilios] I’m gonna give the microphone to Dr. Cushman next but I
have my own opinion here and I respectfully would
disagree with Dr. Shimbo because (coughs) this has
been discussed extensively (coughs) and researched appropriately. By now there are at least eight studies that compare attended and unattended and is identical, is the same. So, we don’t have any questions there whether the nurse was in the
room or outside the room, if there was any difference
in the blood pressure, they are just about the same. The SPRINT study also did an analysis of those centers who did
attended versus unattended, or they did mixture, whether
there was a difference in outcomes and there was
no difference in outcomes. So the SPRINT methodology,
if you follow the guidance of having the patient sit in a quiet room with their legs uncrossed
and their back supported and not talking during the
blood pressure measurements provides a very reproducible and reliable blood pressure measurement, either attended or unattended, and I think it should be adopted as the standard of care as Dr. Weber said. – All right, a couple of other comments before we finish this. Okay, three others before
we finish the session. We’ll go over here first. – [Bob] Okay, am I on? – Yeah. – [Bob] I want to very much
agree with Dr. Weber’s comments that labeling is the key to this and perhaps helps bridge the gap between the epidemiological
data and the individual data. – [Mark] Yeah, I agree. – [Bob] That if a drug
that raises blood pressure is identified by the FDA
and the FDA something that says this drug raises blood pressure, the next sentence is absolutely key. Some of the sentence, as worded now, is this drug is associated with increased cardiovascular risk is not nearly a strong enough statement. Santa Claus is associated with Christmas, turkeys are associated with Thanksgiving. We need a statement that conveys causality because we believe, I believe
everybody here believes that blood pressure is causally related to cardiovascular events. So if the next sentence is elevated blood pressure is known to increase cardiovascular events, heart attacks, and strokes, a statement that conveys causality and even another sentence that
patients and their physicians are advised to work together to manage their overall
cardiovascular risk. – Go ahead. – Yeah, I think it really
depends on the patient population and the degree of blood pressure. I think that as a blanket would
be probably not consistent with scientific data. – All right, two more comments before we wrap up this session. Go ahead. – [Man] Yeah, I think Dr.
Papademetriou said most of it, as far as SPRINT, I won’t
go into how it happened but everything was very standardized in the way the blood
pressures were measured and the most important thing I think was the waiting five minutes and taking multiple measurements and not interacting with the patient. Some sites did do it with
the staff person in the room and some did it with staff
person out of the room and I won’t get into why that happened, but any way we analyzed it
it didn’t seem to matter even though we didn’t do that randomly, but others, such as Dr.
Papademetriou have done studies looking at random order. And so I think the, and I can just say that
in large clinical trials it’s just never been considered
affordable to do ABPM on the entire population and
to have your enter criteria and your titration based on it, also for the reasons
that have been mentioned that you just can’t do
ABPM over and over again on an individual patient
and have them tolerate it. So, we were frustrated
over decades and decades of clinical practices and clinical trials and not being able to get
well-trained research nurses even to measure auscultatory blood pressures consistently correctly, as
well as they wanted to do it and so that’s the reason
in recent decades, such as in ACCORD and SPRINT and others, we have abandoned the
manual auscultatory method and gone to oscillometric method. – I just wanna quickly respond ’cause I don’t want there
to be misunderstanding. I think the blood pressure
measurement in SPRINT was very high quality. I think when we say the SPRINT method, I think some people
are thinking unattended while other people are thinking
oh, highly standardized. And I’m just saying that
I’m not convinced yet about the unattended piece, but I do think that the high quality piece
is extremely important, which was done in SPRINT. (laughs) – The technically
correct term, by the way, is not the SPRINT method, it’s the automated office blood pressure, that’s what it’s called. You don’t have to use the word SPRINT. – [Mark] (laughs) Good to know. And then a last comment for this session. – [Bob ] Yeah, along those
lines of consistency, of course using the replicate
or triplicate automated system with the proper time in between, we often overlook the simplest of that and that’s actually
selecting the right cuff. I think there does have to be a point that individuals should
have their arms measured in a clinical trial, because
not all cuffs are the same size and we know that there’s
a significant potential of getting an incorrect blood pressure for not using the right size cuff. So that, whether it be guidance or just in your clinical care
or in your clinical trial, making sure you use the
right size cuff is important. And the other component, going more to the clinical
trial side of things, has to do with again
trying to remove variables that can affect blood pressure. And I don’t know if it was discussed but there is a consideration, I think, when you take a look at
a patient population, not the healthies, of doing
an orthostatic assessment, either for screening or
at some point in the trial because the idea is to try to remove the potential other effects
that could effect blood pressure on the individual other than the drug. So, I don’t know what
percentage of the population has an orthostatic change, I don’t know if you’re just
looking at CNS indications. But is their value in implementing an orthostatic assessment,
maybe at screening? I don’t know, I would leave
it to Dr. Weber, Dr. White, Dr. Cushman, and Dr. Papademetriou
how that would affect it but I’m looking at more
from a clinical trial and addressing potential variability in what could change the blood pressure. We do know that in certain
parts of the population that we do see orthostatic changes. – All right, well I wanna thank
our presenters and panelists and all of you for a
wide-ranging discussion on some important methodologic issues. Thank you all very much. (audience clapping) And, for our last session, for topics that you haven’t
had a chance to discuss yet, we’re gonna have our FDA
colleagues back up here up front to bring up any remaining issue. I wanna turn this over to Greg. – Yeah, thanks, Mark. So this is the last session
for open audience feedback. As if we haven’t already done that throughout all of the sessions. This is the opportunity for you all to get any points or comments
or questions on the table before we adjourn. And I’d like to invite
some of the FDA leadership, Dr. Doug Throckmorton,
Deputy Center Director for Regulatory Programs at CDER FDA, in addition to those
that you’ve already met, Norm Stockbridge, Bob
Temple, and Ellis Unger to also join me up here. Okay, thank you. So we don’t have prepared remarks for this particular
session, although at the end Doug will help summarize some
takeaway points from the day. So let me go ahead and just
open it up to the audience for any additional comments, things that you heard that
you think are really important that you wanna make sure are on the record and are considered, or
questions or other comments that you might have. And I’ll also invite the
audience from online also to email questions to
[email protected] to go ahead. So, any thoughts? Oh, okay, great. That table back there, it’s
been an active table all day. We’ll go back to that one. – Can you here me now?
– Um. – No.
– Maybe speak up. – I’ll talk more loudly.
– Okay. – [Sid] (speaking off mic)
Health Research Group, I’m trying to put together a
lot of the things of the day, which I guess is what the session is about and just to quote from
one part of the guidance that has come up, it said, quote, “Small, sustained increases
in blood pressure, “two to three millimeters
of mercury, clinically,” that’s the end of the quote
and they cited the fact that there is logical evidence that this can cause a problem. I was interested in thinking again about the whole dose-response curve when Dr. Madabushi said in
his very nice presentation, I was just thinking
that (speaking off mic). I don’t care which population,
a number of these drugs (speaking off mic) that this drug produces (speaking off mic) Adding in the dose-response, because aside from (speaking off mic) that kind of problem could
be (speaking off mic) response. (speaking off mic) for
Celebrex and (speaking off mic) and all of the effects were studied. There was a dose-response curve, millimeters at a lower dose, 2.9 millimeters at a higher dose, and the higher dose correlated with the cardiovascular outcomes. So I think that when
taking into consideration and evaluating a new drug
that’s coming on the market, or an old drug that’s
already on the market, (speaking off mic) one could implement dose-responses. Dose-responses are, that
sort of cuts through a lot of the other variables which might affect blood
pressure other than the drug. Other parts, this meeting,
I think adhering to the idea (speaking off mic) had one of the things
he already had in there was what millimeter of
blood pressure increase would need that kind of
warning label for doctors and (speaking off mic) reduction, and once that label is there, discussion will take place
between doctor and patient. (speaking off mic) lower blood pressure. Most important is just the
guidance (speaking off mic). – Great, thank you. Thank you for your comments. I won’t call on any of the
FDA leadership up here, but if any of you ever wanna comment just wave your hand and
I’ll, um, okay, Bob. (laughs) – Well, I’m not 100% sure I
understand what Sid was saying, but it certainly is true that among other things you should know is the dose-response relationships for effect on blood pressure. And that’s important because
it’s gonna help you decide whether to increase the dose, whether the improvement in effectiveness is enough to justify it, and
it’ll make you more conscious of looking for that
when you raise the dose. So, there’s no question that kind of information
should be available. It’s also, I just wanna just support what Raj was saying before, it’s very, I mean, we never have
enough doses in our trials and I don’t expect to. But the concentration-response
relationship data which our clin/pharm people
almost always look at is a source of information about what you should be thinking about and what you should be looking for. I just wanna put in a plug for that. – Okay. Okay, over here. (man speaking off mic) Okay, Ellis? – Yes, so, you threw out the 160 number and someone about a half an hour ago also through out 160 as a
blood pressure of concern. He didn’t care whether it was 160 or 162, and likewise didn’t care
whether it was 120 or 122 and I think we all would agree. So maybe what we really do care about is the outlier population
and maybe you wanna know what percentage of patients
hit 140 at some point or hit 160 at some point,
because maybe that would be a more informative way
to look at the data. I just wanna point out, I’ll wait, one of the things I wanna say, but it’s not like we don’t
get blood pressure data. So, as I mentioned this morning, I mean, I’m looking at a drug
for a neurological indication, we have 44,332 measurements
of blood pressure in that development program. Here’s a psychiatry drug,
we have 33,443 measurements of blood pressure. Blood pressure’s being measured, but the problem is I think we’re getting, what you referred to as crummy, crummy office blood pressures and maybe if we instead try
to improve the methodologies so that we got good blood
pressure measurements that we could hang our hat on, realize for chronic use, a drug, typically we get data on a
thousand patients for a year. That’s a lot of data, right? So, it’s just a matter of
measuring blood pressure well, well enough that we can
make something out of it and explain okay, X
percent of people hit 160 and Y percent of people hit 140, and maybe that’s where we should
be going with all of this. – [Greg] Go ahead. – [William] Hi. I would just amend some of
the things you said, Ellis. First of all, I think we have lousy
blood pressure measurements because in most studies, particularly if it’s not
a cardiovascular drug, we’re not that interested in what the patient’s blood pressure
is if it isn’t 80 over 50 or 250 over 160. It’s purely, these
things are generally done just to make sure the patient is safe. That’s what goes in that category. – I don’t know. – [William] And then, the fact
is that we have lots of data and we could improve their measurement, but the, when we have lots of data, it’s also analyzed very, very poorly. They throw in these
integrated summary of safety and they don’t care what
the randomization is among studies, they don’t care differences when blood
pressure is measured, anything like that. I don’t think we’ve even
really tried to do that, and I do think we have a
tremendous amount of information that’s laying there, and
because we have so many patients and because so much of
it is placebo controlled, we have a lot of information
within patient variability. And the whole case that’s being made about ambulatory blood pressure monitoring that we can only do twice in a patient, which isn’t a good indication of, within subject variability,
is that we think that within subject
variability isn’t a problem. But I think it’s, isn’t there. I think there are lots of ways
in which we can demonstrate that a drug elevates blood pressure. And it doesn’t necessarily have to be ambulatory blood
pressure monitoring, we just need to do, need to
make a competent determination that the evidence points
in this direction, then we can put that on a label. And I think also, the tie
with separate comments since I have the microphone, is that, uh–
– And it’s working. – [William] Yes. What we label a drug–
– Can be removed though. – [William] Put a warning on a drug, say for Stevens-Johnson syndrome, we say this drug has been associated with or can cause Stevens-Johnson syndrome, but a patient that’s taking the drug, we know whether or not they
have Stevens-Johnson syndrome. But when we put a label on
a drug saying on the average it raises blood pressure
four millimeters of mercury, I don’t know whether it’s
raising blood pressure in this patient at all
or whether it’s raised by 10 millimeters of mercury, or whatever, and even if I know that, blood pressure in itself is not
the harm, it’s a risk factor that may or may not lead to
some morbid event going on. So that’s a couple degrees of separation that we’re dealing with. And so I think there’s an
element of false precision and false certainty that we’re trying to
build into the system. – Bob. – Maybe everybody already knows this, but I think what we’re trying to detect is the possibility that the
drug increases blood pressure and we said, we say oh, at least two millimeters
of mercury on average. It’s not ’cause everybody thinks that a two millimeter
increase from 120 to 122 is gonna kill anybody, but you know that if it increases it some, there’s gonna be a distribution. Some people will have more and
some people will have less, and that it will do that in populations at all levels of blood pressure. So, it tells you be on the lookout and that’s what we want everybody to do. What your threshold is gonna be depends on other risk factors,
the age of the patient, all those other things. It reminds you that
you’re supposed to look and I think that’s all
we’re really talking about. So we want to detect it, whether it should be an average
of two or average of three can be debated, that’s
not what’s critical. But, you’re gonna look at it
because going from 160 to 165 probably increases your
risk and we care about that. We have all kinds of data showing that risk is a continuous
measure, a continuous variable. It increases with increase
in blood pressure, we know that. It might take three or
four years to happen but for a chronically
used drug that matters and you’re supposed to worry about it. And I think that’s all
we’re talking about, detecting the drugs where
you have to pay attention. – [Greg] Uh, okay, back to the table. There we go. – [Man] I think that’s an
excellent point to take away and the other point is
that we talked about before is enhancing the quality
of the assessment itself is a key factor, again,
going to what data do we have to take a look at. I’d like to continue to support a term called complimentary
blood pressure assessment. If you take the ground floor of doing better office
blood pressure assessments, standardizing the process, identifying that potentially the
compound has a signal, then using a decision tree as
to what’s my next best method of defining that potential
blood pressure signal. Some things may be longitudinal because it’s gonna take time to see that. Some things are gonna
be better done by ABPM in two series to see
what’s my initial dose and is there a dose concentration. The more am I exposed to that
drug, do I see that signal? And that goes back to
the clinical pharmacology and what we understand
about the drug activity. So, I think it’s not one or the other. I think it’s again, doing
a better job at part one, which is getting good office assessment, and then taking that next step as to what am I learning about the drug. In that early paper
that we did at the CSRC, there was a decision tree
and it was something, again, that we adapted a little bit from Dr. White and Dr.
Pickering of when to use ABPM. So, I do wanna emphasize the fact that the concept should be complimentary. It’s not one or none, and again, to focus on the
quality of the collection. We have examples before
this guidance ever came up of companies doing a good job of defining a blood pressure signal. One of them is clearly,
there was a drug, Treximet, that had a blood pressure signal. There was a post-marketing requirement for them to assess the blood pressure. In this case it was better
to use home blood pressure because the patient was
taking the drug at home and they were able to assess it. They got some 50,000
blood pressure points, and at that point the
reviewer was able to say there’s no dangerous signal, but there should be
something put on the label that says this drug does
affect your blood pressure. So, I think that’s important. If we take a look at
Sunitinib and Axitinib, really good work done on
the blood pressure there, which defined the change
between cycle one, day one and cycle one, day 15, and
was there a continuous change in blood pressure or
increase in blood pressure. There wasn’t, but it did define that there was a blood pressure change. Benefit outweighed risk. But now the clinician, he or she knew, that there was going to be
a change in blood pressure. And I think that’s, as
Dr. Temple was saying, it’s kinda what we wanna
be able to tell people. So, I just wanted to
give those two examples and to emphasize the
fact that the component is complimentary assessment
of blood pressure. – Okay, go ahead. Not yet, okay. (man speaking of mic) – [Man] I’m hearing a
lot of agreement that we need to do a better job
in measuring blood pressure and we have been doing a lousy job. Interestingly enough, from here, examples of components with
small changes in blood pressure that were detected with the
lousy blood pressure monitoring. So we need to be very careful that in order to optimize
the blood pressure, that is critical, it’s very important that we don’t move the pendulum too much. I think there’s definitely room for ABPM. Is ABPM for all at all the times? I’m not sure about it. And a good example is what was shown here with the data from the
FDA with the placebo data. Based on that, there are
a couple of placebos, and I’m gonna be a little facetious, that require a labeling change because it changed the blood pressure by two millimeters of mercury. So, we need to be very careful
where we put the pendulum and not to ask too much. And hopefully learn within what we have and not an extra for, the solution is not thorough everything, thorough blood pressure, thorough renal, thorough liver, thorough CNS, et cetera. – Great. Anna and then in the back. – [Anna] There is something
that we also need to talk about. We don’t have interoperable data where we can understand the outcomes, in electronic medical records to be able to go back to data that is traceable, fully traceable, that we know that it was a change in a way that is still accurate. Then until we cannot get the outcomes of this type of measurements,
the final outcomes. It’s very difficult to
communicate at this point when we are trying to, we
have data in Cerner or APHIDS and we are trying to
transfer that information. We have a moving target because the box are
being fixed constantly, new functions are being added and then all of this makes
the data become unreachable, cannot be configured to
improve comprehension, then we need an interoperable
healthcare system. – Okay, thanks, Anna. In the back. – [Man] Yeah, I just wanna clarify where I think Bob was confused
about what I was saying and extend for 20 seconds
more about what he was saying. I think the reason I read
what was in the guidance was the FDA did have some discreet amount, didn’t say whether it was two or three, but I think that just as
there is uniform labeling now, which I think is appropriate
even though there are degrees of difference for all the NSAIDs, just basically says this essentially may increase the risk of
cardiovascular disease. It would be a binary thing if you met whatever the threshold
is, two or three or above, you would get a standardized label. ‘Cause I think what Bob was talking about and what I’m talking about
here (speaking off mic) it’s a public health effort
that will inform more people about something that’s well documented. (speaking off mic) There is at least two doses where you go up on both the blood pressure and (speaking off mic). I was not saying to stick in the label (speaking off mic) does or does not increase the risk, defer on (speaking off mic),
how old the patient is, but it’s, will precipitate
I think more often than not a discussion (speaking off mic). – Okay, other comments or questions? Okay, great. So, with that I’ll turn
things right over to Doug and the rest of you can stay up here as we’ll be wrapping up pretty shortly, to summarize some takeaway points that he came out with. – Yeah, thanks, Greg. And thanks to Duke for holding this. And thanks for letting me come. (Greg laughing) I’m not allowed to attend meetings of this kind very often anymore and it’s been a real pleasure, a tremendous learning opportunity for me. – [Greg] It’s not due to
past behavior at Duke events. (panelists laughing) – Look, there’s a lot that
you guys have talked about that I’m not gonna comment on. Should we be looking at blood pressure in ways that we’re not
today, in ways to improve it? If so, how we should do
that was this conversation about different kinds of
measures and things like that and when we should do it. Those are conversations I
think that we’re gonna listen, we’ve heard a lot of good
conversation about today and we’re gonna need to take back. From where I am, the thing
that I came to this meeting with an ear towards more than anything was what would this guidance, what would the ideas that
we had in the guidance mean for product development. And I think we had some
of the conversation about that this afternoon and
I really appreciated that. I think we do have to take
into account the lessons of QT prolongation, but I’d also say we have to take into account the lessons of drug-drug interactions and
the lessons of LFT measurement and all of the safety measurement things that are discreet, that we perform as a part of product development. And for each one of those, as they’ve been added on to
the drug development paradigm, we’ve had to work through how we did them, what we asked for in terms of guidance. We’re in the process of
putting out a guidance on a food effect measurement, how to decide whether or not
there’s a food effect in a drug during drug development. The guidance is highly detailed and in many ways dovetails
with some of the things that you guys are talking about and I’m gonna be watching
that one really carefully ’cause it’s, there is gonna
be an additional burden that’s gonna be placed on thing and we’ve tried
to be very thoughtful in the ways that we’ve laid that out to minimize that to
the extent that we can. So, I think we have to talk
about all those paradigms, whether it’s QT and the current concerns that you all have voiced in the past about impact on product development. I think we need to avoid those lessons to the extent we possibly can. But also LFT and all those other things. So that’s the first thing,
’cause as I think about them, what others have described
as false numeracy, the ability to quantitate these effects has led to overemphasis on them. And whether it’s QTs or it’s LFTs or it’s white blood cell counts, the ability to count them
and create thresholds has led to them being the focus of early product development
in a lot of ways. And it’s made the
thresholds that we’ve set in our guidances more
than usually important, because they’re sometimes misunderstood. The threshold that’s
described in the guidance that Ellis and the group
have been talking about is a threshold for determining
a pharmacologic property. So, we’re looking for a
way to identify or exclude a meaningful pharmacologic property in a product development. That’s different than saying you’re over that threshold,
you’re in big trouble, your product can’t go forward. And we’ve not always messaged that as carefully as we could have and I think we need to hear from industry about what we can do to make certain that that doesn’t happen again. We don’t want these
thresholds to be viewed as barriers that prevent
products from going forward, because they were never
intended to be that way. And there may be FDA reasons for that. There are also industry
reasons for how those guidances have been interpreted and
I hope that you all comment on ways that we can perhaps message those, the thresholds if they’re included so that we don’t have that problem. The second thing to think about is what the guidance is about altogether. It’s to guide medical product development and we know something about the kinds of information we need to
include in guidances like that. They need to have clarities of outcome, to be straightforward and easy to measure to the extent they can be, they need to be easy to measure, that is, within the
normal practice paradigm to the extent possible, and quantitative measures are preferred over subjective measures
just because they’re easy for all of us to assess when we’re looking to determine whether or
not the product exists. So, as we think about the guidance to the extent you’re
recommending changes to it, think about the goals of the guidance, the need to be as clear as possible to guide medical product development that is efficient and
transparent and scientific. In that sense, one-size-fits
all is not an insult, the way it was used I think in some sessions earlier on in the day. Its intent is to make it
transparent and common to have a straightforward,
single way to go forward and not something more than that. So, where do we go from here? I’m gonna take issue with Charles Benson, and I don’t see him around here so I can’t tell if I’m taking issue with a person that
can’t defend themselves, but anyway–
(person speaking off mic) Ah, all right. So, he–
(person speaking off mic) (laughs) Charles and I have
had wonderful discussions. One of the things he said
at the end of the morning was that from his perspective, medical product development was, drugs were required to become
more and more effective and more and more safe
to get on to the market. I don’t go back to the ’70s
or the ’80s, Bob, I’m sorry. I do go back to the ’90s, in
terms of my time in the agency and I just don’t see that. I see the natural progression
into product development. Product development begins with products that have high risk,
and can have high risk and have high benefits. And as the product area matures, there is a natural progression towards products that have
fewer adverse effects, a natural progression
towards understanding better those risks and making
them known, and labeling, and managing them, and mitigating them, and I see that playing out
in a variety of places. It played out in anti-hypertensives. This group knows that better than any. You look at the thiazide diuretics and things that were
available in the early ’70s for treating hypertension, you had to be hospitalized to
start some of those medicines. I mean, incredibly complicated, highly potentially
adverse reactions to them. We’ve gotten better. That transition I think occurs
in all therapeutic areas and I think it behooves all of us just to understand that’s a natural thing. Instead, what we have to do is make sure that we are, one, identifying those risks as
cleanly and clearly as possible and then finding ways
to manage them better. And here I think the
FDA is making progress and I’m looking at Peter Stein who’s the head of O&D, sitting back there, I think we’ve made some
progress in a few things. One, I think we’re better at systematically
identifying and evaluating all of the risks and the benefits. Through the benefit/risk
template we’ve got, I expect the reviewers to bring in and identify
all of the benefits and all of the risks. I wanna try to do away,
to the extent possible, with the undue focus on one benefit or the undue focus on one adverse effect and instead, the benefit/risk template is about a systematic looking
at the totality of the data and then transparently saying what you think about
the benefits and risks. But I think it’s a good thing. I think that’s the mechanism whereby a pharmacologic property like an elevation in blood pressure following a drug being administered gets integrated into the larger spectrum of benefits and risks of the product. We’re gonna make better
decisions when we have that. I also thing we’re doing
better at managing risks. The 2000s were a time
when we had to struggle to manage risks of identified products in the post-marketing space, and I think we’ve gotten better. I think we understand now
that we can’t just reach for the most restrictive
risk mitigation tools that we have available to us. Yes, we have restrictive tools available. Not allowing a product on the market is obviously the most
restrictive tool that we have, one of the most restrictive tools. We now understand those
things aren’t necessary, it’s not in the public health interest, it’s not the kind of way we should be approaching drug regulation and I think under Peter’s tutelage, I think with Janet in the position for, we’re doing better, looking at products in the full range of their benefits when we think about whether or not a product can be on the market. And so, I think these kinds
of pharmacologic properties will be viewed in that broader context in ways that they may not have been viewed back in the day when I
was a medical reviewer in cardiorenal. So, the last piece is alternative options, alternative approaches. I think that’s another strength of the cardiorenal division and I would say FDA’s approach in general. They are sincerely and genuinely open to alternative approaches. You all have raised a few of them that I think we need to talk through and understand better in
the days going forward. I’d really like to
understand ways to improve early blood pressure measurement in these clinical development products. The numbers that Ellis
gave are pretty striking. 35,000 blood pressures measured. I’m sure many of them are not
measured very thoughtfully. If there were ways to improve that and use that as a marker that
we could understand better, I think we’ve gotta be able to think about the possibility to
do those kinds of things. Critical Path was mentioned. It’s all about being willing
to question your assumptions. It’s all about being willing
to look at alternatives. Janet said it up in 2004, it is, that spirit still exists and I guarantee, to the extent an
alternative is demonstrated to be supported by data,
we’re absolutely interested in incorporating it into our
practice and review paradigm. So, you all have a very
challenging task, a data task, a challenge on you to
get some additional data, a better understanding
about how these products are being developed, and the kinds of data that we might be able to bring to bear. But, I see this as a positive, a positive opportunity
and I continue to believe that understanding this pharmacology, given the public health impact that it could potentially have, is valuable for all of us. Thanks for giving me
the opportunity to speak and safe travels for everybody. – Great, thanks, Doug.
(audience clapping) And, just as a quick wrap
up from my perspective, so the issue of blood
pressure effects of drugs during clinical development, obviously as we heard
to is not an easy issue and there is no clear slam dunk answer. But one thing that I would say and I think is clearly
evidenced by the folks from the FDA in the room today, is that the FDA is taking
the development of guidance that is useful and
practical very seriously and is thinking about
this very thoughtfully. We’ve heard a lot of consensus today around some issues and
a lot of suggestions, really good suggestions on
how guidance can be improved or how to better deal with measuring these blood pressure
effects and when to do it and how to do it across a
range of different populations. Very good insight and discussion that could be and will be, I’m sure, useful to the FDA as they
further consider the guidance. So, as for next steps on the Duke side, we’ll be making the slides from today as well as the wonderful video of all of us all day
long available online, and we’ll also be publishing a summary of today’s discussion as well. Before you go, I do have a list of folks that I would like to thank. First and foremost all of
you who are in the room today who contributed to the
discussion and stuck with us. A lot of times we get,
you know, a trickling out after the lunch and you
just sort of thin attendance by the last session, but that certainly did not happen today, and that’s a testament to
how important this issue is. So thank you very much for
your participation today. We did work very closely
with a lot of FDA colleagues in developing this event. Norman Stockbridge and the folks sitting at the panel up here, in
addition to Naomi Lowe, Meg Pease-Fye, Mona Fiuzat, Fred Senatore, Christine Garnett, and Z McDowell. Thank you very much for
all of your guidance in planning today’s
session and today’s event. I’d also like to thank those experts that we worked very closely
with outside of the FDA to further refine the discussion points and recruit speakers. Phil Sager, Billy White, Michael
Weber, and Mitch Krucoff. And then lastly I’d like
to thank and acknowledge the individuals at Duke who
helped in planning the event. They weren’t responsible
for the microphones, but (laughs) were responsible
for everything else, including Morgan Romine,
Nicholas Harrison, Sarah Supsiri, Elizabeth
Murphy, and Patty Green. So, thanks again for all of
you for today’s discussion and have a very nice evening. (audience clapping)

Leave a Reply

Your email address will not be published. Required fields are marked *