🤔ANOTHER ONE!? Study Claims Red Meat Linked to Diabetes. Let’s digest, again… This meta-analysis included 31 cohorts from regions across the globe and used individual participant level data to determine Hazards Ratios (HR) for the associations between unprocessed red meat, processed red meat, and poultry. Here I will focus on the unprocessed red meat findings since I think they’re of greatest physiological interest. 🚨Findings 👉Unprocessed red meat consumption was associated with a HR of 1.10 for T2D per 100g red meat intake. That’s small. 👉There was a large degree of heterogeneity across studies. 🚨Visualizing Heterogeneity You don’t need a professional degree to see this. If a given row has a horizontal black line that crosses the vertical black line perpendicular to 1.0, the results of that study were not significant. So, for example, if you look at Europe as a region, there are 9 studies included. Only 1 of 9 found a significant association between unprocessed red meat and diabetes. And, if a given black square is to the left of the vertical solid black line, it means there was a trend towards a negative association between meat intake and diabetes, e.g. the SUN in European group. 🚨“But it’s a Meta” Of course, the immediate counterpoint is “but it’s a meta-analysis.” Otherwise put, one could argue that, while there is heterogeneity and many studies might have negative findings, the aggregate of studies is more likely to reveal the truth. However, mixing different studies on distinct populations also sacrifices specificity and has the potential to muddy the waters. This is an implicit limitation of meta-analyses. So, while it’s possible a meta can help provide clarity where there are conflicting data, it can also obscure “truth” by the process of lumping different studies with different methods and different confounders etc. My opinion stated simply: Metas don’t necessarily represent a great truth… they can also be used to reinforce a deception… just with more perceived academic muscle 🚨Why I Remain Skeptical The triad of (i) a small effect size + (ii) large heterogeneity + (iii) clear confounders Should already raise a red flag. Of course, there’s a potential contribution of healthy user bias, including BMI as one manifestion of healthy user bias, etc. Importantly, the study does analysis with and without adjustments for BMI, with adjustment for BMI lowering the effect size. This consistent with the idea there is healthy user bias – not at all a surprise. And it’s important to acknowledge that these biasing factors can never by fully and confidently be accounted for, as the paper itself admits… 🗣️“[T]he observed associations could therefore reflect the bias away from the null owing to … relatively unhealthy lifestyles associated with meat consumption.” 🚨BY REGION Compounding on this, it’s worth noting the effects were attenuated in regions one might expect to have relatively better overall metabolic health. Specifically, the HR for the association between unprocessed red meat intake and T2D went from small to smaller, at a meager 1.07, when looking at European region vs America. And, in Eastern Mediterranean, although only 2 studies were included, there was clearly no effect with one of the two studies trending to a negative association. What I gather from this is further evidence that healthy user bias could be entirely to blame for the net effect seen in the meta. This is possible, when considering all the above factors further framed by… 🚨BIOLOGICAL PLAUSIBILITY Furthermore, I again reinforce the important for scrutinizing biological plausibility. The paper provides some attempt at presenting possible physiological explanations, e.g. the “high intake of animal proteins” may increase risk of T2D or gesturing at TMAO, but I didn’t find it particularly compelling. While honest (and I give points for that), repetition of statements like “no definitive effects have been reported,” and “the exact mechanism is yet to be establish,” etc. reinforces the point that the mechanistic explanations remain extremely shaky, despite an abundance of interest and resources spent on studying this topic. 🚨CONCLUSIONS 🤔Arguments that red meat causes diabetes stand on shaky data, at best. 🤔The observed associations may be driven by confounders. Statistical “adjustments” cannot be assumed to fully account for these. 🤔Even if there is an effect, it’s tiny. By comparison to the HR of 1.10 (with 1.0 representing no increased risk) reported in this study, the Relative Risk for diabetes according to BMI > 30 vs BMI <30 is ~4.44 (Fung TT JAMA IM) All this is not to claim a causal relationship is "impossible," nor it is a commentary on other relationships re red meat consumption. This thread is specific to the claims at hand regarding Red Meat and Diabetes. Bottom line, when it comes to diabetes prevention, red meat elimination is not the low hanging fruit @SukhSandhu @EricTopol @guardian @itvnews @theproof @PlantChompers @KenDBerryMD @SBakerMD @MikhailaFuller Also, I see a lot of people "jumping" on this one, and it's trending on @X. In choosing what source you want to listen to, consider who is evaluating the data and who is just amplifying the title and abstract and/or citing vanity metrics like the sample size because it might be consistent with their world view. Again, I'm open to the possibility there could be a relationship between red meat and diabetes... but this study and the recent Haem iron study seem to mostly hype feeding the incentive structure of the media and popular health narratives than compelling science. That's my take, for now.