In Reply Dr Bhupathiraju and colleagues suggest that our findings are biased because of differences in follow-up time between cohorts and not adequately addressing confounding by restricting the analyses to never smokers younger than 70 years, without chronic disease at baseline, and excluding the first 5 years of follow-up.
We were aware of these potential pitfalls, and several sensitivity analyses in the article addressed such issues. First, the findings could not be attributed to differences in follow-up time, as shown in Figure 3 and eFigure 4 in the article. Second, an analysis restricted to never smokers without baseline disease still placed the BMI associated with the lowest mortality in the overweight range (26.1) for the most recent cohort (2003-2013) (eFigure 2 in the article). However, these sensitivity analyses have limited statistical power and further restriction would not provide usable results (eg, the additional suggested restrictions would reduce the sample size to 14 453 individuals [15% of original] and 63 deaths [1% of original] in the 2003-2013 cohort). Furthermore, we specifically wished to investigate the reported association between BMI and mortality in the general population and not in subgroups only. Also, the use of restrictions is not without pitfalls and has controversies of its own.1- 3
Afzal S, Nordestgaard BG. Body Mass Index and All-Cause Mortality—Reply. JAMA. 2016;316(9):992. doi:10.1001/jama.2016.10297