Wednesday, May 18, 2016

Dr. Andrew Stokes On That New Excess Weight is Healthier Study

Last week saw the publication of yet another study looking to quantify risk with weight, and like some studies have found in the past, the conclusion was that some excess weight might in fact be beneficial to health. From my perspective, the question of "what weight is healthiest" is flawed in and of itself. As I tend to prattle on, the goal for all of us, regardless of our weights, should be to stack our personal decks as much as is realistically possible by living with the healthiest life that we can honestly enjoy. These studies, whatever their findings, tend to emphasize that scales can usefully measure health. I don't wholly agree, but that said, I invited Andrew Stokes, a researcher with a special interest in the association between weight and mortality, to give us his take on the new research.
A new study published in JAMA finds that the BMI value associated with the lowest risk of dying shifted from the normal to overweight range between 1976-1978 and 2003-2013. The study is short on explanations for this puzzling finding, but does offer up the possibility that people with overweight and obesity are receiving better treatment and thus living longer than they did in earlier decades.

An alternative explanation for these findings is a flawed research design. Although some of the coverage, including on NPR and on Ted Kyle’s obesity blog, pointed out the potential pitfalls of using BMI as a proxy measure of body fat, they miss a more fundamental problem with this study, a problem that is shared by many other studies of the health consequences of obesity.

The problem, which I discussed in a previous post on Yoni’s blog, is the reliance on weight assessed at a single point in time. Disregarding weight history is a problem because people who have long maintained a weight in the normal range are mixed together with those who formerly had overweight or obese and lost weight. Although some of the weight loss in the latter group is healthy weight loss, much of it is tied to conditions such as heart disease, cancer, COPD and other ailments. Because of this—as several recent studies have shown — including the weight losers as part of the normal weight category obscures the substantial benefits associated with maintaining a normal body weight.

In the JAMA study, attempts were made to address this bias (referred to by epidemiologists as confounding by illness) but the measures taken were ad hoc and incomplete, leaving plenty of room for the bias to creep into the estimates. Unfortunately, there is no way of knowing from the data presented how the composition of the normal weight category changed across the cohorts studied. It could be that as a result of differences in age composition or smoking status across cohorts, that a larger fraction of the normal weight group in the later cohort once had overweight or obesity, which could explain the apparent finding of the nadir of the BMI mortality relationship increasing across cohorts.

The substantial discrepancy in length of follow-up across cohorts may have also compromised the comparison (median length of follow-up was 19.8, 11.0 and 4.6 years in the 1976-1978, 1991-1994 and 2003-2013 cohorts, respectively). Length of follow-up has been shown to be a significant effect modifier of the association between excess weight and mortality, with shorter duration follow-up often leading to greatly attenuated effects. The findings of the JAMA study are consistent with bias due to duration effects, with the risks getting progressively weaker between the first cohort in which follow-up was greatest and the last cohort in which follow-up was shortest. Although the authors acknowledge the possibility and present several tables and sensitivity analyses aimed at testing for duration effects, the results are not entirely convincing.

The problems discussed above raise doubts about how much we should take away from the new JAMA study. Although it is possible that the nadir of the BMI-mortality curve has shifted over time, this study is far from conclusive on the matter. A more likely explanation behind the striking pattern is that it is a spurious result of a flawed study design.

Andrew Stokes is an Assistant Professor in the Department of Global Health at Boston University. His research is focused on the causes and consequences of the global obesity epidemic and developing novel approaches to combating obesity at the population level through interventions that target aspects of the social and physical environment. You can also follow him on Twitter.