On 2015 Apr 25, Arnaud Chiolero MD PhD commented:

Probably the best and more exhaustive review addressing the issue of overweight and mortality. Many health scientists have difficulties to accept the idea that a BMI in the overweight category is not associated with a higher mortality. To move beyond this controversy, studies assessing the effect of interventions to reduce BMI will be needed (see e.g. Chiolero Epidemiology 2015). Meanwhile, it seems difficult to deny these observations.

On 2015 Apr 18, John Spitzer commented:

The article in JAMA by Flegal et al is a summary of the literature. It includes 97 articles with 2.88 million individuals and more than 270,000 deaths. The conclusions described as "misleading" by Munafò et al (below) are simply a summary of the results that arise from the data and the statistical methods. Munafò et al do not present any specific criticisms of the methods of the Flegal et al article nor do they assert that the summary of the literature is incorrect. The single study by Lawlor, described by Munafò et al as "large," based its results on 5,337 never-smoking participants, and it was one of the articles included in the review by Flegal et al. The majority of those articles do not show smoking and pre-existing illness as important causes of bias. Opinions about "detrimental" public health messages arising from scientific findings should be distinguished from scientific criticisms of those findings. In a commentary on reactions to the Flegal et al article the editors of Nature stated: "It is risky to oversimplify science for the sake of a clear public-health message." http://www.ncbi.nlm.nih.gov/pubmed/23936910

On 2015 Apr 02, Marcus Munafò commented:

Flegal and colleagues suggest that being overweight may decrease risk of all-cause mortality, and being slightly obese confers no increased risk compared to being of normal weight [1]. We think these conclusions are misleading and may give a detrimental public health message, particularly given the publicity they attracted.

Despite the authors’ suggestion that there is “little support for the suggestion that smoking and pre-existing illness are important causes of bias”, these factors are not consistently controlled for within the contributing studies [1]. It has been demonstrated in two large prospective studies that the increased risk of death from being overweight or obese can be masked when smoking is not adequately controlled for, and when sufficient initial years of follow up are not excluded [2]. Smoking is associated with both lower body mass index (BMI) and increased risk of death, while underlying illness can lower BMI prior to death. Furthermore, it is widely appreciated that conventional statistical approaches to “control” for such factors are inadequate. Support for a positive causal effect of BMI on mortality comes from studies less subject to the biases associated with BMI measured in later life because they used BMI in adolescence or offspring BMI as the exposure [3, 4]. These measures are suitable proxies since they are strongly associated with an individual’s BMI in middle age but are not substantially affected by reverse causality and are less confounded by smoking. Higher BMI in adolescence was associated with greater risk of all-cause mortality in middle age in a study of over 200,000 individuals [3]. Stronger inferred associations were observed with offspring BMI and all-cause and cardiovascular mortality than with own BMI, suggesting that positive associations of BMI and mortality may in fact be underestimated in conventional observational studies [4]. In addition, this study demonstrated that the commonly observed inverse association between BMI and death from respiratory and other diseases may also be due to such biases.

Given the importance of the public health messages regarding “healthy” weight, we feel that further research is needed in this area using innovative research methods to overcome potential biases; further conventional observational studies will simply recapitulate the biases inherent in all such investigations. One such approach would be Mendelian randomisation, using obesity-related genetic variants, such as those in FTO. Mendelian randomisation methods, when applied correctly, are free from confounding by environmental factors and are not affected by reverse causality [5].

Marcus Munafò, Amy Taylor and George Davey Smith

1. Flegal, K.M., et al., Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA, 2013. 309(1): p. 71-82.
2. Lawlor, D.A., et al., Reverse causality and confounding and the associations of overweight and obesity with mortality. Obesity (Silver Spring), 2006. 14(12): p. 2294-304.
3. Bjorge, T., et al., Body mass index in adolescence in relation to cause-specific mortality: a follow-up of 230,000 Norwegian adolescents. Am J Epidemiol, 2008. 168(1): p. 30-7.
4. Davey Smith, G., et al., The association between BMI and mortality using offspring BMI as an indicator of own BMI: large intergenerational mortality study. BMJ, 2009. 339: p. b5043.
5. Davey Smith, G. and S. Ebrahim, 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol, 2003. 32(1): p. 1-22.

On 2015 Apr 02, Marcus Munafò commented:

Flegal and colleagues suggest that being overweight may decrease risk of all-cause mortality, and being slightly obese confers no increased risk compared to being of normal weight [1]. We think these conclusions are misleading and may give a detrimental public health message, particularly given the publicity they attracted.

Despite the authors’ suggestion that there is “little support for the suggestion that smoking and pre-existing illness are important causes of bias”, these factors are not consistently controlled for within the contributing studies [1]. It has been demonstrated in two large prospective studies that the increased risk of death from being overweight or obese can be masked when smoking is not adequately controlled for, and when sufficient initial years of follow up are not excluded [2]. Smoking is associated with both lower body mass index (BMI) and increased risk of death, while underlying illness can lower BMI prior to death. Furthermore, it is widely appreciated that conventional statistical approaches to “control” for such factors are inadequate. Support for a positive causal effect of BMI on mortality comes from studies less subject to the biases associated with BMI measured in later life because they used BMI in adolescence or offspring BMI as the exposure [3, 4]. These measures are suitable proxies since they are strongly associated with an individual’s BMI in middle age but are not substantially affected by reverse causality and are less confounded by smoking. Higher BMI in adolescence was associated with greater risk of all-cause mortality in middle age in a study of over 200,000 individuals [3]. Stronger inferred associations were observed with offspring BMI and all-cause and cardiovascular mortality than with own BMI, suggesting that positive associations of BMI and mortality may in fact be underestimated in conventional observational studies [4]. In addition, this study demonstrated that the commonly observed inverse association between BMI and death from respiratory and other diseases may also be due to such biases.

Given the importance of the public health messages regarding “healthy” weight, we feel that further research is needed in this area using innovative research methods to overcome potential biases; further conventional observational studies will simply recapitulate the biases inherent in all such investigations. One such approach would be Mendelian randomisation, using obesity-related genetic variants, such as those in FTO. Mendelian randomisation methods, when applied correctly, are free from confounding by environmental factors and are not affected by reverse causality [5].

Marcus Munafò, Amy Taylor and George Davey Smith

1. Flegal, K.M., et al., Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA, 2013. 309(1): p. 71-82.
2. Lawlor, D.A., et al., Reverse causality and confounding and the associations of overweight and obesity with mortality. Obesity (Silver Spring), 2006. 14(12): p. 2294-304.
3. Bjorge, T., et al., Body mass index in adolescence in relation to cause-specific mortality: a follow-up of 230,000 Norwegian adolescents. Am J Epidemiol, 2008. 168(1): p. 30-7.
4. Davey Smith, G., et al., The association between BMI and mortality using offspring BMI as an indicator of own BMI: large intergenerational mortality study. BMJ, 2009. 339: p. b5043.
5. Davey Smith, G. and S. Ebrahim, 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol, 2003. 32(1): p. 1-22.

On 2015 Apr 18, John Spitzer commented:

The article in JAMA by Flegal et al is a summary of the literature. It includes 97 articles with 2.88 million individuals and more than 270,000 deaths. The conclusions described as "misleading" by Munafò et al (below) are simply a summary of the results that arise from the data and the statistical methods. Munafò et al do not present any specific criticisms of the methods of the Flegal et al article nor do they assert that the summary of the literature is incorrect. The single study by Lawlor, described by Munafò et al as "large," based its results on 5,337 never-smoking participants, and it was one of the articles included in the review by Flegal et al. The majority of those articles do not show smoking and pre-existing illness as important causes of bias. Opinions about "detrimental" public health messages arising from scientific findings should be distinguished from scientific criticisms of those findings. In a commentary on reactions to the Flegal et al article the editors of Nature stated: "It is risky to oversimplify science for the sake of a clear public-health message." http://www.ncbi.nlm.nih.gov/pubmed/23936910

On 2015 Apr 25, Arnaud Chiolero MD PhD commented:

Probably the best and more exhaustive review addressing the issue of overweight and mortality. Many health scientists have difficulties to accept the idea that a BMI in the overweight category is not associated with a higher mortality. To move beyond this controversy, studies assessing the effect of interventions to reduce BMI will be needed (see e.g. Chiolero Epidemiology 2015). Meanwhile, it seems difficult to deny these observations.