Fiona Godlee: Closing comments and open discussion on the way forward

Fiona Godlee: Closing comments and open discussion on the way forward


I’ll be brief with some overall thoughts first. First, to Swiss Re, thank you for having this
conference. I will say that when you showed your map of
the places and partners you have, I didn’t find an institute that focuses on nutrition,
and Tufts is one of the world’s leading institutes that focus on nutrition science and policy. I’ve mentioned to several, we’ve been working
with John Hancock, blue chip life insurance company, for almost two years helping make
all of their life insurance clients live healthier lives to save their bottom line through a
really formal collaboration. Happy to tell you more about that, and we’d
love to add Swiss Re because as you said, “Our interests are actually truly aligned.” If you care about nutrition, you need to have
an institution that’s serious about nutrition. So that’s a first comment. I think the second comment is I really enjoyed
all of the discussion. I talked to Fiona this morning. I’m going to pass on all my time to summarize
what we could all talk about big picture, and hopefully we’ll leave that to the questions,
to use all my time to kind of give a counterpoint to John Ioannidis’ talk last night. I thought it was really good to have his perspective
and good to talk about limitations to data, but in the themes of this conference, we always
wanted a point, counterpoint, and the counterpoint was just missing. And so I’m going to use my time, which is
50 minutes. That’s great. I’m going to use my time … Maybe you could
change it to my time, so I can have a sense, but- 13 minutes. Okay, great. To really give a counterpoint I think it’s
important. I think it’s very important to highlight limitations
of the data, and I think that what we need to do is highlight those limitations in a
way that understands what are the bounds of the realistic limitations, what direction
are those limitations in? What’s the bounds of reasonable uncertainty,
not what’s the most extreme possible hyperbolic uncertainty, but really what are the reasonable
bounds. And so I think that’s what I want to talk
about. So, I’m going to talk about kind of two or
three different things. The first thing I want to talk about is this
idea that dietary questionnaires and food frequency questionnaires in particular are
hopelessly flawed. I mean I heard that term thrown around a lot,
so I just want to show people … and this is borrowed directly from Walter Willett’s
nutritional epidemiology class. I want to show people how questionnaires work. How many people here never drink milk? All right, so never. So look around so everybody can see. All right, how many people never drink milk? How many people drink milk let’s say once
or twice a week? How many people drink milk several times per
week? How many people drink milk one glass a day? And how many people drink milk more than one
glass a day? That’s a food frequency questionnaire. That’s what it does. And the folks in the middle were a little
bit mixed about, right? Whether you said one to two servings per week,
you probably were a little unsure. Should I say one to two? Should I say one per week? Should I say more? But those at the beginning that said zero,
and those at the end that said they drink it every day, they are correctly classified,
right? That’s all a food frequency questionnaire
does is it gives a relative ranking of people to compare the high versus the low. We don’t often care about the middle categories. That’s the squishy categories, and this has
been written about. And again, Walter and others have documented
this. We care about the extremes. It works very, very well. If you want to know exactly what the population
is consuming 1.2 or 1.3 or 1.4 servings, it doesn’t do that. That’s not the point. The point is to rank participants, and it
does it extremely well. So to dismiss and entire class of science
because you don’t understand it I think is troublesome. I want to then talk to you a little bit about
how we assess the validity of dietary questionnaires, and so let’s look a this example of Omega
3 intake, which is where I started my research. We have self reported Omega 3 intake. That’s not the truth, right? The self reported intake is not the truth. The truth we want is the true habitual intake,
but we don’t know what that is. So how do we understand the relationship between
self reported Omega 3’s and true habitual intake? Well, what we do is we measure some other
marker such as blood Omega 3’s. Blood Omega 3’s are also not the true habitual
intake. They’re influenced by metabolism. They’re short term. There’s a few weeks of intake. They’re also not the true intake. So what’s really nice about measuring the
self reported Omega 3 intake and measuring the blood Omega 3 bio marker intake, they’re
both related somehow to the truth, which we don’t know, but they’re only related to each
other if they’re related to the truth. In other words, their errors are uncorrelated. The errors in self reported intake, there
are errors there and biases there, are totally uncorrelated with the errors in blood measures. If I put another dietary questionnaire up
there, there would be correlated errors. But blood bio markers and self reported Omega
3’s, the errors are uncorrelated. So they’re only related to each other in so
much as they’re related to the truth. We have looked, and many have looked at correlations
between self reported Omega 3 intake and blood bio markers in the cardio vascular heath study
which I reported, and it’s about .6. that’s about the correlation you get with
a food frequency questionnaire and blood Omega 3 bio marker in most studies. So what does that tell us about the relationships
to the truth? This is a very simple mathematical equation
when the errors are then correlated. This is just basic epidemiology that I feel
like some of the nay sayers have not really studied. So what is the correlation? Does any … Walter, you can’t say the number,
but what is the relationship of each of these to the truth if their correlated with each
other about .6? It’s actually simple math. The correlation is multiplicative to equal
.64, so they’re each correlated about .8 with the truth. The must be. They have to be, or they wouldn’t be correlated
with each other. So these types of correlations which you see,
and they’ve been validated again and again with adipose tissue, vitamin E bio markers,
other bio markers, correlations are .5, .6 between a bio marker and a diet questionnaire
indicate that the correlation with the truth is higher and pretty reasonable. Are these perfect? Absolutely not. Are they hopelessly flawed and give us no
information? Of course that’s not true. The final thing that’s important about this
is the nature of prospective versus retrospective observational studies. If there’s error here in a prospective study
where people haven’t developed disease yet, the error is random. The changes, the errors, the misreported memory
isn’t related to this future risk of disease usually, especially if the disease is years
away. So what happens when Matias talked
about this … It’s sort of technical speak. What happens is misclassification. It’s random error and it biases toward the
no. So it makes the effect sizes smaller than
we’d otherwise see. And so you always have to think about that
our effect sizes may be under estimated in nutritional epidemiology because of this error
in prospective studies. It’s crucial. It’s just basic kind of epidemiology to understand. Now, I did talk about briefly that there have
been conflicting results of observational studies versus supplement trials. Vitamin C, vitamin E, thiamin, many of the
supplement trials conflicted considerably. Now, that particular conflict I think has
been explained by this reductionist approach that doesn’t work for vitamins. It doesn’t mean all observational studies
are wrong. And so I want to give you one example. Many people including John, who I’ve seen
talk about the limitations of observational studies use the example of hormone replacement. People probably heard about the hormone replacement
trial, the Women’s Health Initiative, that showed that hormone replacement did not reduce
the risk of coronary heart disease like the observational studies. This was a big deal about 10 years ago. This is the actual results from the Women’s
Health Initiative, the randomized trials, and the prospective cohort studies here. So the Women’s Health Initiative in follow
up, not in their primary analysis, they stratified by time since menopause. Women that were close to menopause actually
experienced significant benefit with a significant interaction. Women close since menopause experienced a
significant benefit in both mortality and heart disease. This is from the Women’s Health Initiative,
not as popularized as the main results. In fact in prospective of courts
these observational studies, all the women in these studies are close to menopause. They’re women who are starting menopause. So the results for total mortality and heart
disease, if you stratify it by time since menopause, are actually very similar. And there’s very good biology to explain this
that I won’t go into the time. But even if you ignore this discrepancy for
heart disease, even if you say, “Well, okay. There was a discrepancy for heart disease,”
look at the other primary outcomes in the Women’s Health Initiative. This is the primary outcomes and results of
observational cohort studies. Stroke, venus throm embolism, breast cancer,
colorectal cancer, hip fracture, very, very similar findings in the Women’s Health Initiative
study. Yet, people who are skeptics just point out
the heart disease and ignore the rest. So people who study bias has bias in reporting
the bias. So it’s really striking, and it’s really quite
amazing to me. This is not just me. We have to look at this systematically, and
so this is a Cochrane review. They identified 1,583 meta analyses of 228
medical conditions, not just nutrition, all of conditions. They directly looked at what’s the effect
in the trials versus the observational studies and looked at the ratio of the risk to see
is there systematic bias. Not are there differences, they’re always
may be differences. Is there systematic bias? In the relative risk of ratio of effect sizes
in all observational studies is 1.08, and for prospective cohorts, which are better
than retrospective studies, is 1.04, not statistically significant. No evidence for systematic bias when comparing
observational studies to trials. This is, how did we come up with this list? I didn’t mention that before. I showed this list, and several people thought
it was helpful. I want to show you how we went through that,
and this paper was published last year, so I won’t go into all the details. We went through all of the evidence. These are the Bradford Hill criteria. We looked at strength, consistency, temporality,
coherence, specificity, analogy. All the things that have been sort of forgotten
about in some of these conversations, and that’s why we picked those dietary factors. Just a handful of dietary factors, 10 linked
to cardiovascular disease and diabetes, many, many more that we rejected. And importantly we also wanted to validate
our final relative risk estimates. I’m not going to go into all the details,
but what we did is we compared our relative risk estimates that we identified from meta
analyses of prospective cohort studies, which look at one food at a time. So a meta analysis of fruits and heart disease,
a meta analysis of nuts and heart disease, a meta analysis of fish and heart disease,
and we compared it to diet pattern studies because diet pattern studies look at the whole
picture. First we compared it to prospective cohort
studies of diet patterns here. And here’s the observed relative risks in
the quintiles of the diet pattern studies, and here’s what we calculated based on our
individual relative risks from individual foods looked at one at a time. Then we looked at randomized controlled feeding
trials looking at effects of diet patterns on blood pressure and LDL. Based on meta regression, we estimated from
those randomized controlled feeding trials how fruits, vegetables, nuts and seeds, and
so forth, affect blood pressure and LDL and how that would be predicted here to affect
heart disease. Then we looked at the observed relative risks
for these components almost the same. Then we looked at the predimed
study, a randomized clinical trial, or at least partially randomized clinical trial,
evaluating the effects of a dietary pattern. You can see the incredible concordance of
results whether we compare trials of events, trials of surrogate outcomes, or cohort studies
of diet patterns compared. So to just dismiss this whole field as not
knowing anything is just really quite problematic. I think that when we think about evidence
synthesis we want to know the true effect of anything in science, in this case nutrition. We don’t know the true effect. We’re never going to know the true effect. We have to look at a range of studies, and
so we can’t know how any of these relate to the true effect. So what we do try to understand is we don’t
know this, right. This is what we want to know, but we don’t
know this. Similar to what I showed you for food frequency
questionnaires, what we can study is this. We can study how these different study designs,
which generally but not always, have uncorrelated errors. The errors in an in vitro study are not going
to be the same as a predictive modeling or as a prospective cohort study in general. We have to look at all of these lines of evidence
and how they correlate with each other, when there’s concordance, similar effect sizes,
similar effect direction, similar uncertainty. The only reason why there would generally
be a concordance across all of these these studies if they’re all linked to the truth
in the same way. That’s why to my friend Gary Taubs,
you had lots of good arguments about this, I hope this is my best 15 minute ability to
say why we need to look at all the evidence when we try to understand nutrition. That’s really I think for me one of the take
home messages is that we have real science here. We don’t have all the answers, but we shouldn’t
just be dismissing science out of hand because of sort of theoretical concerns about bias
or measurement error. We can actually quantify and understand the
directions and size of those effects. I really look forward to continuing discussions,
so thank you very much. Thank you. Thank you very much, Dari. Hand me over now to Salim Yusuf. Thank you. I’ve got 13 minutes, not 11 minutes here. I’ll come in one minute under time, 15 not
14. I have no links to the food industry, but
in small print I have actually said perhaps I should. Now, I’m going to speak on two things. One is the conventional thinking on diet and
health has problems. We think of linearity, like increase sodium,
increased blood pressure, increased CBD, and increased mortality, or saturated fats increase
LDL, CHD mortality. We’ve really not tested the entire causal
chain, and I’m going to take blood pressure and salt and show you whether or not it holds. The second thing is uni dimensionality. It assumes there’s no other effect of that
on any other part of the human biology. The third is that it applies to all populations
and that it is linear. Again I’m going to show you it’s not linear
and that extreme changes are feasible because as Dari said the epidemiology is you take
the quintile versus the bottom quintile. Nothing wrong with it, but you can’t assume
you’re going to see that in real life. That diet is the most incredibly complex exposure
that we know, much more than any drug, much more than the complexity of drug, it contains
hundreds of thousands of active ingredients. Each active ingredient affects numerous biological
pathways, so I think a given pattern of diet may be affecting 10,000 pathways. For us to imagine we can decipher that out
and control for that by multi variant whatever regression analysis is just smoking something. The next one is there are numerous interactions,
diet diet interactions, gene diet interactions, microbe diet interactions, metabolite metabolite
interactions. Then we have non linear associations, so the
net effect of diet on health cannot be reliably assessed by individual surrogate outcomes. So you show lowering sodium reduces blood
pressure, now you can’t assume the rest will happen. Now, this is the first thing on linearity. These are data on 100,000 people of blood
pressure and sodium intake measure. You will see that the higher the sodium intake,
the blood pressure really is a big change. For one gram change is nearly four millimeters. When you come to lower levels under three
grams of sodium intake, it’s tiny. And in fact the trials show exactly this. Hypertension in hypertensives, the effect
is relatively important. In the others, it’s much smaller. So, you can’t sort of draw a straight line
and say it’s going to save lives here. Sorry it’s going to save lives here at the
top end and save the same similar amount of lives down there, because lots of things are
going on that you can’t control. Now, the other thing is this U shaped curve
on physiology. This is something classical epidemiology nutrition
is ignored. This is from chapter one of Walt’s superb
textbook that I refer to all the time, and that is too high is bad. Too low is bad. It’s different for external type toxins, tobacco,
air pollution, whatever, arsenic, whatever. The right level is zero, but food isn’t arsenic,
or maybe it is. Hemoglobin, it’s a U shaped curve. Thyroid, it’s a U shaped curve. it’s a U shaped curved. Everything physiological is U shaped curves. Again, we can’t draw a line all the way down. Let me show you this. This is sodium, and on the horizontal axis
is plotted sodium intake. We followed 30,000 people for five years,
and we saw a U shaped relationship with the data of the … This is pretty tricky. It’s not working. Okay, maybe my anti salt people are sabotaging
that. You’ll see that if it goes below around four
and a half, the risk goes up about four and a half, it goes up as well. Now, when we first published this, there was
all kinds of criticism including reverse causality. Since then, there have been 17 studies, and
using different methods of sodium assessment. 14 of them showed the same thing. None actually shows a sodium level intake
below three is associated with a lower event rate. Some are pretty flat. The second thing is to prove non-linearity,
you need thousands of events. So many small studies like the to had a few hundred events, 200, 300, 400 events. You can’t prove non-linearity. You don’t have the power. Now, this is a summary by Graudal to show
the effects of lowering sodium, but it increases renin. It increases aldosterone, epinephrine, norepinephrine,
triglycerides, so to say sodium is the only thing, blood pressure’s the only thing that
matters, and none of the other biological effects matter, it- And none of the other biological effects matter
is a bit short sighted but that’s what the field has been. Then what we did in another paper where we
did a meta regression analysis of three studies on 140,000 people followed for many years,
nearly 10,000 events. We divided people into hypertensives on the
left side and non hypertensives. The blue line is the change in blood pressure
with changes in sodium and you will see that it’s a fairly steep relationship. The right side, which is non hypertensive,
58,000 people, pretty flat. There’s a small significant increase. Then we plotted the effect on outcome. In the left side if your sodiums are over
four you get an increased mortality, but if it is less than three it also goes up. So there’s a dissociation between the effects
of blood pressure and the effects on events. On the right side you’re not seeing any association,
they go paradoxically the opposite way. So hold on a minute, how does the story of
sodium, blood pressure, CBD, mortality fit? We need hard data. Now, we said okay individual data has problems. So can we be clever? Now the PURE study had an interesting design
because we were interested in how communities change, so we sampled by communities. There’s 1000 communities in PURE right now
and the strength of doing a community level analysis that reduces random management error
because not everybody in the community on the same day eats Chinese food, or the second
it minimizes regression dilution bias and it minimizes reverse causality. It has limitations. So the two methods give you the same results. You could say well there’s something that
approximates the truth. We call that the hybrid method. The traditional problem with ecological analysis,
you don’t have detailed covalence in the individual level. That’s overcome with this analysis. So just to go through, in the end we decided
to look at larger communities for clinical events, smaller communities for blood pressure
because blood pressure is available in everybody and we found that when we did the individual
level slope you’re getting a blood pressure change of 2.11 per gram. When we did the community you get about a
30% steeper slope. So the method is valid. And then we looked at clinical events and
what we found is … So this is regressing communities versus events and what you’re
seeing is, just like the individual data, over about 4.5, if your community had a value
over 4.5 there is an increase in CBD. If your community was below 4 it’s an inverse
association and in between it’s pretty flat. So the U shape curve, we’re seeing the two
methods. We then said what about communities with lots
of sodium and countries with less sodium? So China has a ton of sodium. The rest of the world, like us, have lower
levels. So China has a mean level of 5.58 … I think,
Swiss Re you should invest in better pointers. And rest of the world has about 4.5. And you’ll see in China there’s a steep positive
association. The rest of the world it’s inverse, but the
rest of the world is here, most of China is here. So it makes a great deal of sense to do sodium
production in China based on the epidemiology, makes no sense to make our lives miserable. So that’s one. So this debate has been going on. So six wise people sat in a room. It was like electing the Pope. It’s somewhere in South Carolina and they
couldn’t come out till they came up with a consensus. They first said a randomized trial was needed. Second they said “In whom can you do the randomized
trial to meet the WHO guidelines of two grams or less?” They said “What about military recruits? They conscript, they’ll do whatever we tell
them to.” Well, they were too young, they don’t die. “What about old people who are in an old age
home? Well, they’ve got too many other problems
so you can’t have them.” So they landed on prisoners to be randomized
through randomized prisons. So when this came out I told one of these
newspapers “Well, these people have really done us a service. First they said a randomized trial is needed. Second they said you could only achieve the
targets if you imprisoned everybody which means in the unlikely event that the study
is positive, the U.S. will expand it’s prison population so that 300,000,000 people will
be living in prisons.” This is the dystopian dream of the current
guidelines on salt. That’s the point I wanted to get to you and
these good scientists including Paul Whelton one side, Rob Califf on the other side came
up with this brilliant idea which I think is great. So now let’s come back to evaluating the effects
on diet on health outcomes. Most interventions including diet can only
have moderate effects, 10%, 20%, 30%. So if you see something that’s reducing mortality
by 50%, that is coo coo, it’s for the birds. It’s too good to be true. Reality is most things have moderate effects. Second, your errors have to be a fraction
of the difference. So you need a methodology. If you’re looking for a 25% reduction, your
error should be a third or a fourth. That is tough. Second, you have systematic errors which is
biases like the ones that Dary mentioned, that’s hard in epidemiology. And there are interactions, much more so in
nutrition than in any other field. All the interactions I talked about, you need
thousands and tens of thousands of events. Now, the analysis I showed you of the community
level has 10,000 events. That’s what you need to get that. The second thing, therefore is you need large
trials. There’s absolutely no data, you need trials
if you can do them. There are limitations of trials. No doubt. You don’t get your quintile one versus quintile
four, you’re going to get more moderate but more feasible differences. So the effects also will be moderate. You can’t do 30 year trials, that’s where
epidemiology trumps you. But you can do five year trials, ten year
trials. And randomization, however, is the only way
and the word is only, to reliably know that you’ve controlled all possible confounders
and biases. You need to power it for moderate effects. So randomized trial of a few thousand people
for common events is no good. You need 10,000 events, not 10,000 people. That’s what you need. And obviously there’s the issue of feasibility
especially long term trial, and a common jargon, especially in the United States, they have
to be expensive. They needn’t be expensive. I’ve done large trials of 20,000 people at
2,000,000 pounds. So it can be done. Now it’s not diet, you may need two or three
or five times that but it can be done. So in epidemiology … Now in trials you can
actually make your trials more efficient by doing cluster randomization. Factorial designs where you get two or three
answers for the price of one. So if you’re going to spend 20,000,000 dollars
you might as well get three answers rather than one. And you can obviously do clever designs with
supplementation. Like for instance for salt one of my colleagues
Martin O’Donnell said “Let’s do our best lower salt as much as we can, but then let’s supplement
some people with extra salt. So you’ll get that contrast. So you don’t worry about the contrast.” We haven’t yet done it, but that’s the idea. So large clinical trials with clinical outcomes
are essential before we promolgate things or inflict it on the entire innocent population
of the United States or Europe or India without informed consent. That’s what’s happening with these dietary
guidelines at the moment. So here is my hierarchy of evidence. I would say when feasible, large long term
trials on clinical outcomes are necessary. It’s not often feasible and they have their
limitation. So I’d say large long term prospective observational
cohorts with standardized methods. This is where the superb work from Boston
has actually improved the field. Extensive covariates, broad populations. You can’t just study a high SES people with
one profession, that’s a starting point but you need to expand it to everybody else because
there is a non linearity issue. But then we need to also think about hybrid
designs at the very beginning by multilevel sampling as I showed you. Now that was fortuitous because we did that
for other reasons; I was interested in urbanization. The large long term prospective observational
studies like the best ones that have been done like EPIC or the health professional
studies that output them. And you should do small and medium sized randomized
trials on multiple, not a single outcome, multiple biological outcomes to give you a
clue. And then you need to triangulate all this
and see do they fit in? If they don’t fit in, hang on. And large case control studies. So I think, this is my last slide. First we should encourage people who only
do observational epidemiology to user expertise to design and work with people who do large
trials. Reverse, people in large scale clinical trials
should understand both the strengths and the limitations of epidemiology. The other thing is something that is heresy
to this group. Given that the food industry and the markets
are huge and they’re several times larger than the pharmaceutical market, a one to 2%
investment of the annual revenues in research other than food formulations is appropriate,
you know, and we need to do large RCTs. We need to do smart observational studies
where they simultaneously look at metabolomics, proteomics, gene environment interactions
and gut bio. And instead of discouraging and criticizing
those who collaborate with industry we should say “You guys are good guys.” And let’s encourage those collaborations between
academia, governments and industry but of course with transparent mechanisms to ensure
integrity of data, minimize competing interests and the results are out. Let’s not criticize the people who try to
do them. Money for this kind of thing is incredibly
scarce. So where do you go? You go to the bank. And the bank, in these cases, is the food
industry. Thank you very much. So we have … How many minutes do we have? Probably 30 or so. Really welcome your questions and we’re looking
forward now, forward, forward, forward. There’s Steve at the back. If the microphone could go to Stephen, if
you could introduce yourself and then the microphones going to go to the back and then
the gentleman there with his hand up with a beard, excuse me, and then the lady there
and then we’ll come to the next three. We’ll do three at a time, if that’s alright. Thank you. Excellent presentations on two aspects of
the difficulties of recording what people eat and how to analyze their effects. I’d like to offer a different perspective
in terms of energy intake and dietary records and there is a long history of studies reporting
caloric intakes based on dietary records. And there’s a seminal paper by Walter Mertz
and the group from the USDA in Beltsville where they actually objectively measured people’s
energy expenditure after having them very honestly record their caloric intakes prior
to being quote incarcerated for 60 days and they reported that
typically for adults who have no reason to under report their calories, it wasn’t an
obesity study, they were missing about a third of the calories that they had to have been
consuming. Most recently the Stanford diet fit published
in Jama back in the end of February. They, with a straight face reported that people
weighing 90kg after a year of following either a very high carbohydrate diet or a very low
carbohydrate diet burned about 1950 calories per day, weighing 90kg. So that comes out at about 23 calories per
kilogram. And we know from multiple human studies that
for free living ambulatory but non athletic adults the expenditures are between 30 and
35 calories per kilogram. Briefly if you may, your question. The point is if we’re looking at energy balance
and their reporting caloric intakes that are obviously absurd, shouldn’t we hold the authors
to a higher standard? Whether it’s the reviewers or whether it’s
the editors. Or at least if you’re doing an energy balance
study and you say “Oh gee this is what they ate and this is what we found.” It’s sort of like saying “We flew 70% of the
way to Hawaii from California, six, seven, eight times and we never found it there.” Okay. Thank you very much indeed. That’s question one. Gentleman there, do you have a microphone? And then Christine. Yes I’ve got a question as to- Say who you are, sorry. Yeah Richard Morris is my name. I should declare my bias, I’m a type II diabetic
who went on a low carb diet and remissed his disease. But I want to just make a comment. I heard something yesterday somebody said
in the panel about what science can we trust, somebody made the comment that an absence
of correlation is something we should consider. I just want to ask the panel in the context
of the PURE study, the absence of the correlation between saturated fat and cardiovascular disease
flies in the face of a lot of dogma. And so I’m interested in the comment, the
absence of correlation between saturated fat intake and cardiovascular disease. Okay. I’ll put that- Flies in the face of a lot of our assumptions
so I would be curious about that comment. Thank you. And the question there. I’ve been again … Jane- Sorry, say again who you are. Sorry, Jane Coliss. I’ve been a human guinea pig for twin studies,
pre Tim Spector, sadly. And I’ve seen every single trend going and
it did start off where they were getting success with basically a low carb normal fat diet
which meant, in reality, people were then following a normal diet instead of over eating. But I find now that people are so confused,
they go to doctor Google and professor daily blurb and get the latest trends, then they
ask the doctor about it within their five minutes of wanting to know all about it and
they come out knowing nothing. I’ve met many people who don’t even know what
a carbohydrate is. So the confusion should start with the patients
who actually just people, free range, looking after their own diet in best way possible
without any medical intervention. It’s something that, if you go as a diabetic
to a doctor, it’s the only disease where you’re told you will lose your eyesight, lost the
kidney function and lose your limbs if you do not do what we say. And I think a bit of kindness wouldn’t go
amiss. Thank you for that. Thank you. Thank you for those comments. We’ll turn back to Dary and Saleem if you’re
thinking while they’re responding to those three points, I would really like to hear
from you about where this debate should go now, okay? Some of the conversations we’ve had about
specifics, I’d like to slightly spend the last half hour, if we can, just taking it
to next steps. So Dary and Saleem, we’ve got diet question
misleading and shockingly misleading. We’ve got the PURE study which Saleem will
… And then we’ve got this idea of confusion in the public and ignorance and how can we
address that. So I’ll be super reamed for time, so diet
questioners are well known not to measure total energy well, that’s not the purpose. The purpose is to measure diet composition. The point of measuring energy is to adjust
for it. If you want to measure energy you should look
at BMI or use doubly labeled water. So that’s not their intent and again there’s
a whole chapter in Walter’s textbook on this. For the question on PURE, I’ll let Saleem
give more comments but there were 16 core studies before PURE showing the same thing
PURE confirmed in a multinational group the same thing that saturated fats neutral for
cardiovascular disease. And then I think lastly I think the issue
of trust is really, really important and this gets back to appropriately and reasonably
and sort of soberly assessing strengths, limitations of studies, discussing them versus just hyperbole
about fake science. And in the current era where fake news has
become a political weapon and antiscience has become a political weapon, I mean this
is really happening right under our feet, right? Portions of the world are turning antiscience,
fake news to undermine credibility. We’re only feeding into that narrative with
hyperbolic comments about limitations of science. And so science has challenges but to dismiss
the field is worsening trust in the public and there’s nowhere good to go from that. Thank you. Saleem. Well, I won’t comment on the first one. What Dary said is fine. On the PURE, he’s right. Thank you for saying that Dary. The totality of the epidemiological studies
support PURE, so PURE shouldn’t be a surprise. The reason it is a surprise to some people
is selected bits of the information have been unduly emphasized and Zoe mentioned the hupa
met analysis and ten out of 11 comparisons were non significant. The one that was significant was a weird collection
of multiple CBD events that included arrhythmia, angina, surgery, this, that and the other. And some weren’t even cardiovascular events,
they called it cardiovascular. So I think when you read … Most people don’t
read the whole paper, they read the abstract of things. When we read the paper you find out that that’s
the case. The totality of evidence from the randomized
trials is well support- Totality of evidence from the randomized trials
as well support what peer showed and the observation study showed. Lastly, I think even I was brought up with
the low-fat hypothesis. I worked at NIH. And so anything that came in that contradicted
through proved. You know, so we all start with the belief
system. I believe the biggest conflict of interest
is you inherited belief system in this field. It’s less so whether you collaborated with
industry or not because you can declare it. You can be transparent about it but the belief
system, is one that you can deal with it. I know in the session earlier on today, which
I had to go to sleep, my jet lag caught up on me, people were talking still about extreme
sodium low rate with no data. It makes no sense. At least let’s get the data. So my suggestion is we need to be transparent,
unfortunately in the belief system it is harder to know what the transparency is. Yeah. I just want to make one brief comment. I didn’t know we were supposed to declare
our conflicts. My conflicts are in the paper, all my research
is supported by government or the Gates Foundations. I have given talks for a range of companies
and received honoraria and I’m an advisor to a day-two Elysium Health and Omada Health
so please look at the conflicts. I forgot to mention that. Thank you very much. I mean the point about unsaturated fat is
the evidence is now looking pretty good but the guidance hasn’t shifted. I think that doesn’t seem to be an enormous
mea culpa from the scientific community that we got it so wrong. I just find that does surprise me. I think reason, if I may comment on that,
Fiona, is we got brainwashed by a very questionable study called The Seven Countries Study many
years ago and it was ingrained in our DNA. And generations of us were brought up with
that. The guidelines committee in the US and the
US guidelines committee influence what happens at the WHO And elsewhere. Had people who were the disciples of the people
who did that, did The Seven Countries study, somebody said you have to wait for the guidelines
committee to die before you can change the guidelines. Yeah. But, Salim, maybe one outcome of this meeting
would be, for this meeting, I don’t know if anyone else would say this, to say that that’s
gone now. The science has changed on that. I mean, am I right, Salim? Am I right, Dariush? Is anyone here going to say? I’m going to ask Gary because he’s got to be honest. Not quite. It seems to me that should be an outcome of
some sort from this meeting. Gary, briefly… Yeah, I’m going to say I
agree on this. Just introduce yourself,
Gary, for the cameras. But it’s transformed from, and even when we
talk about saturated fat being neutral compared to what? So that argument can still transform now the
thinking is it’s neutral compared to carbohydrates but it’s not neutral compared to polyunsaturated
fats and vegetable oils, therefore that implies that either saturated fat is delutarious or
the polyunsaturated fats and vegetable oils are beneficial to our health and so the argument
hasn’t gotten away, it’s transformed. And there’s still, or there still should be
plenty of controversy about the second issue and randomized control trials would come in
real hand date if we could do them and there were fees along other issues. Okay. Thank you. Because of all the hands, I’m going to just
go to the audience, but you’re going to have to keep your comments really short. So people we haven’t heard from, if I may,
gentleman there in the blue shirt with the glasses, the lady there with the blue dress,
so I’m going to sound like Robin Day here, and I think the gentleman there with the glasses
on his head with the blue shirt. So, quickly, if you may just go… Robert Verkerk, Alliance for Natural Health
International, I’m just wondering if we look at the classifications that we look at for
micronutrients where we think of essential, semi-essential, and non-essential they can
all be important. If we apply that thinking to macronutrients,
could we not rethink as much for the research community as well as for the public and policy
makers as carbohydrates are semi-essential macronutrients rather than essential macronutrients. Very interesting. Thank you. The lady there, could you say who you are,
please? Anna Peters from Australia. Thanks very much. I guess I just wanted to get a reflection
in terms of going forward around most people in the world probably don’t want to spend
their lives thinking about their diets and their probably not looking for the perfect
diet so I agree that we need to understand the physiology and I think these diets are
really important but I’d love a reflection on for the majority of people who basically
eat too much unhealthy food. What is a way forward to help them shift that
back to healthy diets? Now, we thank you, Anna, for that. Yeah. Norman Temple, Athabasca University, Canada. I want to come to this issue of foods being
macronutrients. Daruish put up his slide earlier and also
yesterday with this vertical arrow, it doesn’t have words like carbohydrates and fatenay,
only foods. Healthier foods at the top, unhealthy foods
at the bottom, and neutral foods in the middle. He said it first thing yesterday morning,
it was then totally ignored and pretty much this meeting was hijacked by the benefits
of low-carbohydrate diets which I think is severely misleading because it totally ignores
food quality. To talk about amount of carbohydrates is utterly
meaningless. It’s what are the foods that you’re eating
so cabbage is healthy, potatoes are healthy, whole grain cereals are healthy, sugar and
white bread are unhealthy. To say low-carbohydrates or high carbohydrates
or lots of fats is an utterly meaningless and severely misleading concept. Thank you. We might have a doubt about the potatoes but
I think they’re- three other people we haven’t heard from, if I may, the lady here, the lady
here, and I’m going to say, just because it’s David, David here. Sorry, the lady in the back there. Starting with you. In the- fine yes, go ahead. My name is Georgia Idema, psychiatrists from
the united states. Id like future research to question the assumptions
common to many of the presentations given in the conference which is that plant foods
and fiber are necessary and beneficial for human health and that red mean endangers health,
assumptions for which I find no evidence in the scientific literature. In fact, there’s plenty of evidence for the
contrary if you look for it. There are no human clinical trials demonstrating
that simply removing animal foods from the diet without also removing processed foods
provides and health benefits. We need, I believe, an RCT comparing a whole
foods plant-based diet to a whole foods numerous diet. I’m hoping there’s the intellectual curiosity
and the funding available to begin to ask more, I think, fresher fundamental questions
about food and health that go beyond dietary patterns to food biochemistry and even beyond
metabolic disorders to other aspect of health, such as metal health. Thank you. Perfect. Thank you very much. I’m behind you. Copenhagen, Denmark. We’ve talked a lot today how to terat or reverse,
and yesterday, diabetes, how to treat and maintain weightless, but we haven’t really
talked about the primary drivers of weight. We haven’t talked about what makes the normal
weight individual, the normal weight child go from being normal weight moving into overweight
and becoming obese. And in fact, we know very little about this. We have no primary prevention studies for
obesity among adults and we have one primary prevention study for obesity among children. So if you ask what should be the next steps
we should take, also for the industry and also for insurance, I think we should put
our bets in primary prevention. Thank you very much. David. Thank you. David, I’m with the GPM Royal College, clinical
expert in diabetes. I would like to ask the panel how you would
give hope to small guys in research. So, I’m a GP. We’ve got amazing data. I have data going back twenty years. When I see all that you’re doing, it seems
so far away and above me and yet I like to join in research and I’d like to give hope
to GP’s and people in primary care to do research because we do see trends and we can make changes
and then maybe we could publish things. But could you give me some hope and some ideas
of how GP’s could join in? Thank you. Fantastic. I know there a lot of other people wanting
to speak but my clock’s slightly different to this but we’ve got a remaining ten minutes
so I can remind you what those questions were and then you’re thoughts on them so we had
the business about micro and macronutrients so we could consider carbohydrates a semi-essential
macro/micronutrient. What about how can we shift people from unhealthy
diets to healthy diets, you know, what behavioral things or policies? What about food quality? It’s not actually food as a macronutrients
but the food quality is the big issue. Red meats, what about red versus processed
meats and then can we have this randomized trial of whole food plant-based versus whole
food omnivore food plan? How we would we resource that, might be a
good question. And then the prevention, I think we all agree
how to stop people gaining weight through life course. And then the small researcher, we would not
call you that, David, but you modestly call yourself a small primary-case based researcher. How can they find hope in the future. Can I address a couple of them, I think Dariush
is always more qualified than me in addressing everything else, so Flattery will do you great. I think, first, I think we should get away
from nutrients. We don’t go into a supermarket and say, “give
me X grams of saturated fats or carbs.” We eat foods and I think moving towards food,
which I think you’ve said, Anita has said, it’s probably the way to go. I mean, there is a role in nutrients but that
shouldn’t be the basis of our recommendations. I think I would apply that to my current macronutrients
as a whole. I would love to see a trial that was, I think
you raised of whole foods of different kinds, I do want to say that in our data, and this
is from four studies we haven’t fully published it, the protective affect of plant-based food
is not seen innate in Indian particular in South Asia. And this four big studies we haven’t published
it yet, soy and intra heart, we didn’t publish it. Intra stroke, we haven’t published yet. We’ve seen it again in pure. And I think vegetables are not vegetables. How they’re cooked matters. How they’re grown matters. Whether there is insecticides and pesticides
residues on them matter. What they’re eaten with matters. So I think it’s over-simplistic just to say
plant-based versus animal based. It’s much more complex than that. I don’t have the answer, I’m just pointing
out the problems. So those are my two comments and we need to
do trials and, you know, there is a big industry that produces animal foods and a big industry
that produces plant-based food and we’ve got to tap into them along with governments to
fund these trials. Yes. It’s a bit like speed dating to try to remember
all the questions but I think that the first question, I think that micronutrients and
the RDA’s and all these limits in essentiality is for clinical deficient deceases and that’s
where it works and it shouldn’t be applied or extended to macronutrients that would be
a problem because most macronutrients are not essential so I don’t think that would
be the approach. I agree with the comment on low-carb versus
low-fat being the wrong question. Overwhelming evidence shows that high or low-fat
diets, on average, have not relationship with obesity, diabetes, cancer, any diseases and
same with high or low-carb. There is place for “pure low-carb” if you
want to lose weight rapidly, if you have severe instinctual type two diabetes to get it in
control. Short term, I think really going low-carb
makes sense but after that it should be low starch and sugar and so I really hope, you
know, we move away from low-carb because the low-carb folks are making the same mistakes
as the low-fat folks. It’s over-simplification. And so we really should be talking about starch
and sugar, not low-carb. I think the question on meats, I don’t think
I heard, unless I missed it, any of the scientists I know say that we should be eating a plant-based
diet, that that’s the healthier approach. What we’ve talked about is the combination
of foods that are rich in phytochemicals, can be supplemented with fish, with dairy,
with other foods. The public and popular press talk about plant-based
diets and vegan/vegetarian diets but there’s no- you know, fries and a coke are vegan,
right? So you can have a really horrible vegan diet
and you can have a healthy non-vegan diet. I think that’s clear but I would really push
back at the couple statements that have been made today and yesterday that there’s no evidence
that meats are harmful. That’s just flat out wrong, right? There’s plenty of observational studies that
show processed meats in particular are harmful and if you look at unprocessed red meats versus
processed red meats, they have very similar associations with lifestyles so the confounding
structure for people that eat hamburger, which is an unprocessed red meat or a steak, is
not that different from someone who eats porchetta or salami and in fact people that eat processed
low-fat deli meats, like low-fat turkey, low-fat roast beef, you know, low-fat bologna tend
to be healthier people and yet even thought their confounding structures are similar,
unprocessed red meat is neutral for heart disease, neutral for stroke, and has a very
modest association with diabetes. Processed red meats have very strong associations
with stroke predominately, mostly explainable by the blood pressure affect if you do back-of-the-envelope
calculations. They have very strong associations with diabetes
and they have very strong associations with heart disease. In addition, heme iron as I mentioned a couple
times, could be a common denominator for diabetes. People with hemochromatosis, a genetic disorder,
get type two diabetes and when you bleed them phlebotomy to treat their hemochromatosis,
their type two diabetes improves. Women who have higher ferritin, which is generally
thought of as a marker of health, higher iron source, are at higher risk for gestational
diabetes. En vitro studies show that heme iron alters
pancreatic beta cell function. So to say there’s no evidence that red meats
are harmful is just incorrect. And, I guess, lastly, the comment of our primary
prevention, there haven’t been large trials. It’s very, very difficult to do a large trial
of primary prevention of weight loss but there’ve been plenty of longterm twenty year observational
studies that we and others have published which give a wealth of information that is
now being confirmed for randomized trials on yogurt being good, cheese being neutral,
white bread being bad, meats being bad for longterm weight gain and so on. So, we have evidence on primary prevention
and i fully agree with the importance of primary prevention. Did I forget one question? Researching primary care. Oh, thank you. I forgot that one question. So I think this was addressed yesterday with
a question and I think I agree with this. So, we have to be cautious not to use the
N of one experience or even the N of a thousand experiences in clinical practice as evidence
itself because that gets us back to the days of anecdotal clinical experience and we’ve
learned that anecdotal clinical experience is not in and of itself, its the worst kind
of observational epidemiology, right? Because it’s your own biases and your own
observations and who comes back and so forth. But, it’s essential, essential we have physicians
there in the room where clinicians and dieticians who are designing studies, working with patients,
have insights, have hypotheses when studies are being designed and implemented and in
the United States, President Obama launch PCORI, which is the patient centered outcomes
research institute, which have several billion dollars a year to do this kind of research
where you must have clinicians and patients from the beginning of study design through
to the end. Okay. And so I think PCORI funding is open globally,
like all US funding, so any clinician interested in being involved, PCORI funding is a really
strong Thank you. Salim, briefly if you would. I think, David, your question, there’s no
doubt you can do trials of vitamin supplementation or simple drugs, like asprin or something
simple in general practice, you can do very good trials. The issue comes complex trials and complex
interventions and diet is usually a complex intervention. I would look then at two steps. Step one, let’s find out what actually we
need to alter. Maybe the trials in primary care can be implementation
strategy trials, Strategy A versus Strategy B. Another idea where we can do good research,
it’s not necessarily trials, is certain standardized data collections. Like on of the biggest data sets on blood
pressure in the UK comes from a general practice sort of data base. So superb associations with the right
events. They had millions of people in it. So observational studies could be done, especially
in the UK where you have good vital registration systems in central databases. You can do simply randomized trials and you
can do implementation studies. But even, not even I, I would be careful of
jumping into doing trial of a complex intervention like diet before I knew that it works and
I’ve done a lot of trials. So, what you can do is a lot. I’m really sorry, but that’s the end the questions. Sorry, first of all, thank you to Dariush. Thank you to Salim for that.

One comment

Leave a Reply

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