 Thank you for the kind introduction Dr. Donovan. It's a great pleasure to be here and thank you to the organizers for inviting me So I'm going to talk to you today about our work on dairy products and incident type 2 diabetes So the context why am I interested in this as a physician in diabetes and my training both in diabetes and public health? I'm deeply interested in etiology. I want to understand why disease happens and and The context is that type 2 diabetes is High in prevalence as this map many of you will be familiar is from the International Diabetes Federation Diabetes Atlas of which I'm a member of that Atlas committee The prevalence of diabetes is high and here's the US of course these hot spots with more than 20 million people with diabetes already in the world and The sad or the depressing fact is that it seems to be unrelenting the numbers are still going up These are latest estimates that be released at the end of 2013 and the numbers estimated of people with diabetes is going up Just in a matter of years. You can see there from 382 million estimated to have diabetes to nearly 600 million people this is serious stuff and this is Aburdened to individuals to their carers and to society and note here that no region of the world is exempt So here's North America and the Caribbean and other areas of the world where the prevalence is set to rise or estimated to rise so the search for solutions and modifiable factors is on and As part of that endeavor you will be aware I hope of the lifestyle intervention studies that provided pretty Convincing and I would say unequivocal evidence that Lifestyle interventions of both diet and physical activity and together do have an impact on Delaying the onset of diabetes or in in preventing it completely And this was a meta analysis in the British Medical Journal by some colleagues of ours in Leicester and they Meta analyzed that the overall effect was a reduction by 50 percent by lifestyle intervention So then it begs the question if we know that lifestyles work that better diets work What's the big deal? Why don't we just go achieve this and the point is in free-living individuals outside of the clinical trials context It's a different story So this was a review now 10 years ago by the World Health Organization and note here I do not want you to look for all the detail But the point to note here is that if you look at the convincing evidence column There is not one dietary factor that shows up in free-living individuals in the appraisal of evidence on factors that could be Instrumental in the prevention of diabetes So in the real world we have an issue of what diets to recommend now clearly in the past 10 years progress has been made which is great, which is encouraging and of particular interest of course is Dairy products as we've been discussing all morning And I will not go into any of this detail because we've had some excellent lectures already on why we might be interested in dairy products For chronic disease and in this case, I'm addressing type 2 diabetes One that has not been referred to so very much so far I will highlight is the issue about the saturated fat which is in dairy products. Is it somehow fundamentally different the old Conventional notion of all saturated fat is bad has adverse health effects Does that hold for dairy fat and that is a question as Dr. Donovan said in my introduction That I'm interested in biomarkers Particularly I'm investigating that with biomarkers of saturated fat intake currently and we have just finished Very large study investigating exactly that So my interest and our group's interest in in this whole area was sparked off when This systematic review and meta-analysis was published by Elwood and colleagues from Wales in 2010 and What was interesting about this was that high-fat dairy was found not to be associated with risk of cardiovascular disease type 2 diabetes and mortality and What was even more is that for total dairy intake which they couldn't differentiate between the high and low fat There were inverse associations for each of these outcomes So we wanted to take this further and see if we could contribute further to the literature because at this point the Number of studies in this review were few and it was at the very start of beginning to appraise This this area of science. So our aim was to examine the association between both the amount and the type of dairy product intake and the risk of new onset or clinically incident type 2 diabetes We had two specific objectives one was to use heterogeneity of dietary exposure across different countries using conventional methods and Secondly, we also wanted to use more detailed dietary assessment methods in in this case a food diary Another thing in the kind introduction Dr. Donovan said that is my interest is to endeavor to use more precise methods Going beyond the food frequency questionnaire wherever possible both in terms of self-report and in terms of biomarkers So these were two linked objectives that we tried to address So to address objective one This work is published in the American Journal of Clinical Nutrition just about 18 months ago, so and Yvonne Schlauss was a PhD student then who came to work with me from Utrecht in the Netherlands And now she's a postdoc. So she did a fantastic piece of work here So for this work, we worked with the interact study Which is a large study funded by the European Union under the framework 6 FB 6 And the for those who are interested in methods. We did publish the methods in Diabetes aloja and This study is based within the European prospective investigation into cancer study the epic study and of the 10 countries that is the epic Europe study Interact included eight of them Including both northern countries all the way down to the south The design we have used here is a nested case cohort study within epic Europe, and that is the name of the study epic interact Therefore, this is a prospective study nested in a large prospective cohort The distinguishing features are that it is large It's about half a million people of whom 340,000 were eligible for this work by virtue of having Blood samples available and detailed dietary information Because the purpose of interact as the name might be alluding to or suggesting there is to study the interaction Between lifestyles and genetic factors. So we did need the blood for the DNA Hence the eligible sample. We have long follow-up of nearly 12 years We had nearly 4 million person years of follow-up and we assembled what we believe is the world's largest data set of Incident cases of type 2 diabetes. So these are people who at baseline did not have diabetes They did not have prevalent diabetes and we followed them forward in time So here we have exposure heterogeneity by virtue of these different countries across Europe So drilling down into the method of the study If you imagine this bigger circle as the entire cohort that had the eligible sample of 340,000 people from that we selected Not selected we ascertained Type 2 incident diabetes cases in the entire cohort So that is all the cases occurring and verified as well So unless a case that be ascertained was verified by external information We did not count it as a case of diabetes We then selected a sub cohort which effectively serves as a controlled group Our exposure of course is dairy product intake and here are some important bits of information We use a case cohort analysis which was suitable for the case cohort design because you will note That there's a small number of cases that appear in the sub cohort But that is by design because it's totally random So it also brings in some future incident cases just a small number We use the residual method where we adjusted the dairy intake for energy intake But we also ran a standard multivariable model We did a random effects meta-analysis across the eight countries And we adjusted the analysis for a range of potential important confounding factors an issue as you know That was discussed earlier in one of the talks We also did a range of sensitivity analysis and robustness analysis to test out whether our findings were Stood up to further challenges So on to the results We set out to look for heterogeneity in exposure and we got it So this is data in men and in women and what this shows is that across the countries of Europe spelled out here There are wide variations in the types of dairy product intakes So in the UK milk intake is high for example in men and women in France no surprises the intake of yogurt and cheese intake is high and Other variations to go with it. Okay, so on to the main results So these are forest plots that show the results for the association between total dairy products and the risk of incident diabetes by each country and then pooled across the entire sample and Note that these comparisons are for the highest fifth of intake versus the lowest fifth of intake But we've also examined p-values for trend and we've also done those response analyses in a separate approach You will note that our model presented here is our if you like the maximally adjusted model which includes a whole host of factors that are considered relevant to this association because they may distort the findings and As well as the standard things you would expect in terms of risk factors for diabetes we have also included a range of dietary factors and In these sensitivity analyses, we had even further dietary factors including for instance sugar-sweetened beverages Which are known to be related with diabetes and certain other factors too so what this shows is that in the pooled sample there was a null association between total dairy products intake and incident diabetes Moving on to milk intake. We also found an overall null association There is a hint that there's a slight increase of 10% but of course that is a Sort of false perception in the sense that the 95% confidence intervals do overlap unity and therefore that's a null association for milk intake We examined yogurt and thick fermented milk intake together because in this study We were unable to separate the two because of the high thick fermented milk intake particularly in places like the Netherlands and Overall while there was a tendency towards a reduced risk It was just overlapping unity. So it was not significant Cheese intake was an interesting one where there was a 12% reduction in diabetes incidence The people trend across the quintiles of intake was significant, but in the overall Q5 to Q5 comparison. This was not significant However a priori when we also looked at the combined fermented dairy products intake now we combine yogurt cheese and thick fermented milk We found a significant Association in the inverse direction with a 12% reduced risk of incident diabetes here In this work, we were unable to separate out High fat and low fat intakes because of the nature of how the data were collected with dietary questionnaires food frequency Questionnaires in these studies did not permit us to do that So conclusions from this objective from the first piece of work Acknowledging the strengths and limitations, which I don't have time to go into but very happy to discuss There was no association for total dairy product intake but there was an inverse association for fermented dairy products and We felt that this deserved further interrogation and hence why we moved into our second objective For this I'm grateful to Laura O'Connor Who's my postdoctoral fellow who worked on this study in now the epic Norfolk study Which is again part of the same study, but in the UK component of the epic Europe study We have collected much more detailed dietary data, which I will show you in a minute So here we have used the food diary and this was a prospective food diary which collected information on habitual diets in dynamic time recording for seven days You're not meant to be able to see the detail there, but I think you'll get the idea that As opposed to a food frequency questionnaire where you have the pre-specified list some people only tick boxes They cannot give you any detail in the diary. We have specific information on the amounts the portion sizes The brand names the type of food they ate even if it's tea or coffee where the milk was added and so on So very detailed information and we can differentiate between specific subtypes So following a similar design as in the larger study that I showed you this was a smaller study but this is a trade-off between detail of information and The size of the study so here again similarly we have incident type two cases ascertained which were also verified and We ended up with just under 800 Cases and 4,000 sub cohort I'm taking a similar approach. We applied a coax regression with the specific Design for for the case cohort study. We use the residual method but we also repeated with the standard multivariable method and similar approach to confounding and to sensitivity analysis So on to some results. It's a bit of a busy slide. I want you to only focus on the black Lines not the grayed out lines only to say that they just show Across thirds of the distribution of dairy intake. They show the amounts Of intake the range of intake. So we're not going to focus on those For total dairy intake. There was a null association with incident diabetes and when we looked at high fat and low fat dairy intake now here we were able to Get great granularity in our data by defining the cutoff of high fat and low fat at a threshold of three point nine percent of Fat content of whole milk, which is the content of fat in whole milk in the UK So we use that threshold of three point nine percent to Distinguish between high fat and low fat dairy and for either of these we did not see a Significant association in this more adjusted model for the low fat dairy There was an association in model one which was less adjusted, but clearly confounding was playing a part So on to sub types of dairy intake again, please focus not on the gray, but only on the black and For milk intake. We saw an overall null association This really jumped out at us though yoghurt intake. We found quite a substantial Decreased risk association. We have a hazard ratio of point seven two which shows a twenty eight percent Reduction in the incidence of diabetes here and there was a significant trend across the turtiles For cheese intake. It was non-significant For total fermented dairy product intake similarly non-significant But here's the result for low fat fermented dairy products a 24% reduction in the hazard of diabetes at this point I should add that yogurt intake contributed to Some 87% of the low fat fermented dairy products. So in this population in the UK In the 90s, which is when the baseline of this data collection was all the yogurt Fell into the low fat fermented dairy product category So our conclusions from this again acknowledging the strengths and limitations I'd like to think we have more strength than limitations, but as always there are limitations We found that low fat fermented dairy product in particular yogurt intake was associated with a decreased risk of diabetes other dairy types were not associated and Translating this into a sellable message if you like or a public health message. We worked out that in the UK looking at the Serving size or the pot size in the UK of 125 grams. I looked this up to make it US relevant I think that's just over four ounces, I believe About four and a half of these per week Basically equated to that risk reduction that I just showed you on the previous slide So to put this into context what have others shown so this was what started us off into this Research area when we saw this at that point this was based on five studies and around 7,000 type 2 diabetes cases since then there have been three other systematic reviews and meta-analyses and As you can see the numbers have been increasing This study had about double the number of type 2 cases. They didn't report it But if you counted up that's about 13 to 14 thousand cases and what's remarkable is that across all of these systematic reviews? There is a Uncanally similar risk reduction for total dairy products Which is interesting that it's picked up on meta-analyses, but not on when individual studies are examined The key findings for sub types this one did not look examined sub types But each of these other three have shown again a consistency of finding which is very reassuring so in this review low-fat dairy and yogurt intakes were associated inversely in This one by Gau et al again low-fat dairy cheese and yogurt were associated inversely and in this latest one Low-fat dairy and cheese were so fermented dairy products were associated and our work from interact was included in these two latter ones here So I'd like to end by offering you an interpretation of these findings We believe that this shows that dairy products do have a role in the prevention of diabetes There are specifically important inverse associations for low-fat fermented dairy products and yogurt is a key contributor to this inverse association We know that nutritional epidemiology has limitations But the findings are robust and strong in different settings and with different methods The so we know that also association is not the same as causality But trials of diets I would offer to you are not feasible for hard health endpoints Yes, sure. One can study intermediate endpoints But you're not going to put people on different diets and wait x years till they develop some chronic illness And we have to rely on other criteria. There's a whole set set of criteria including the credibility and the robustness of finding One caveat which I'd like to repeat from the talk of Professor Tremblay earlier is dairy products should be considered with an Overall healthy diet is not a magic bullet on its own because some of the headlines we got from our research were You know if we had a whole front page of the male online male daily mail newspaper Which was kind of suggesting just eat yogurt and you're cured of diabetes. You'll never get diabetes So I think we have to be careful in causal inference and how we interpret and put this in a wider context So this is my last slide and I want to Obviously point out that we need to understand mechanisms work that we're very interested in is to understand the dietary behaviors to move from research to practice just because we come and give talks and show that We have good studies that that show that yogurt or dairy products are Inversely related with chronic disease outcomes doesn't mean the population adopts that we have a center in Cambridge on Which is called the center of diet and activity research and there we are specifically studying the barriers and facilitators of dietary intakes with my colleagues We have to place them in wider context as I said and the final point really is that currently most of the dietary guidelines do allude to dairy products as a whole But I think perhaps we now need to start moving into subtypes To acknowledge my colleagues without whom this work would not have been possible. Thank you very much for your attention I need to and apprentice here This is tremendous work and I'd really like to congratulate you and your colleagues and just mention It's a shame that Sheila Bingham who designed all those dietary surveys Isn't alive and able to see some of the results that are coming from this now I find this very very compelling evidence and what you haven't done at the end there is go back to your Your early slides where you and propagate up what an effect a 28% reduction in diabetes would be between the This is potentially a very important effect What I'd like to draw out to you if I may is that although you you had on the one hand the Daily Mail Saying that you know yogurt would save the world Why you did and I'm sure you're aware of it have a count of you from NHS choices Which accused you of not having proven the thing? So the question I have is one of their criticisms was that your recording of the exposure seven days dietary intakes that many years ago was not adequate But surely I'm right in suggesting that that would tend towards a null result Absolutely, so you absolutely right that would tend to attenuate the associations rather than falsely inflate them This is a common problem the entire body of research that I showed you with those four separate systematic reviews Is based on the food frequency questionnaire measured once pretty much largely Except for a couple of American cohorts as we know the nurses health study where there were some repeat measures But largely it's based on a single FFQ so moving on to the diary is a huge leap forward We have lots of research currently ongoing in my group and we are beginning to look at repeat measures And they're also trying to combine approaches of the food frequency questionnaire together with the diary Plus bio markers, and I think that will be very strong Indeed to do so. So this is not at all a finished piece of work. This is just a little taster And other things are become