On 2017 Apr 02, Katherine Flegal commented:

Kim et al Kim SH, 2016 provide a valuable review of the complexities surrounding issues of obesity and cardiovascular disease. They describe the AARP 10-year follow-up study of 527,265 respondents (Adams et al NEJM 2006 Adams KF, 2006) as showing that overweight was associated with higher mortality than normal weight. They then contrast these findings to those of the meta-analysis by Flegal et al (JAMA 2013 Flegal KM, 2013), attributing the apparent differences to a difference in the normal weight BMI category used in the two studies. However, the findings of these two studies are not as different as they may appear. The AARP article used the high normal BMI of 23-24.9 as the reference category and divided overweight into low (BMI 25-26.4), medium (BMI 26.5-27.9), and high (BMI 28.0-29.9) overweight. For men, the multivariate-adjusted HRs were 0.95 (95% CI 0.91–0.98) for low overweight. 0.95 (95% CI 0.92–0.98) for medium overweight and 1.00 (95% CI 0.96–1.04) for high overweight. The corresponding values for women were 1.00 (95% CI 0.94–1.07), 1.06 (95% CI 0.99–1.12) and 1.07 (95% CI 1.01–1.14). Thus even with a narrower reference BMI category, these findings from the full sample in the AARP study do not suggest excess mortality in the full overweight category (BMI 25-<30). Similarly, the ALLHAT study (Shah et al 2014, J Clin Hypertens Shah RV, 2014) cited by Kim et al found an HR of 0.96 (0.76–1.23) for overweight relative to BMI 22-24.9), again suggesting that a narrower reference category is not the explanation for findings of lower risk in overweight than in normal weight. The finding of higher mortality in overweight categories in the AARP study is only seen after the authors switched to a subgroup of 111,181 respondents and used reported past weight at age 50 instead of baseline weight. This special analysis with reported past weight rather than baseline weight, rather than the use of a narrower reference category, is the source of the apparent disagreement between the AARP study and the Flegal meta-analysis. Results from selective analyses of subgroups should be interpreted with caution.

On 2017 Apr 02, Katherine Flegal commented:

Kim et al Kim SH, 2016 provide a valuable review of the complexities surrounding issues of obesity and cardiovascular disease. They describe the AARP 10-year follow-up study of 527,265 respondents (Adams et al NEJM 2006 Adams KF, 2006) as showing that overweight was associated with higher mortality than normal weight. They then contrast these findings to those of the meta-analysis by Flegal et al (JAMA 2013 Flegal KM, 2013), attributing the apparent differences to a difference in the normal weight BMI category used in the two studies. However, the findings of these two studies are not as different as they may appear. The AARP article used the high normal BMI of 23-24.9 as the reference category and divided overweight into low (BMI 25-26.4), medium (BMI 26.5-27.9), and high (BMI 28.0-29.9) overweight. For men, the multivariate-adjusted HRs were 0.95 (95% CI 0.91–0.98) for low overweight. 0.95 (95% CI 0.92–0.98) for medium overweight and 1.00 (95% CI 0.96–1.04) for high overweight. The corresponding values for women were 1.00 (95% CI 0.94–1.07), 1.06 (95% CI 0.99–1.12) and 1.07 (95% CI 1.01–1.14). Thus even with a narrower reference BMI category, these findings from the full sample in the AARP study do not suggest excess mortality in the full overweight category (BMI 25-<30). Similarly, the ALLHAT study (Shah et al 2014, J Clin Hypertens Shah RV, 2014) cited by Kim et al found an HR of 0.96 (0.76–1.23) for overweight relative to BMI 22-24.9), again suggesting that a narrower reference category is not the explanation for findings of lower risk in overweight than in normal weight. The finding of higher mortality in overweight categories in the AARP study is only seen after the authors switched to a subgroup of 111,181 respondents and used reported past weight at age 50 instead of baseline weight. This special analysis with reported past weight rather than baseline weight, rather than the use of a narrower reference category, is the source of the apparent disagreement between the AARP study and the Flegal meta-analysis. Results from selective analyses of subgroups should be interpreted with caution.