BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participantsBMJ 2016; 353 doi: https://doi.org/10.1136/bmj.i2156 (Published 04 May 2016) Cite this as: BMJ 2016;353:i2156
All rapid responses
Recent evidence from long term observations of hundreds of thousands of women, in 10 European Countries, clearly demonstrated that the use of oral contraceptives reduced mortality by roughly 10%. 
Despite all thromboembolism risks, apparently.
Newer oral contraceptives increase womens' overall survival.
In comparison, drastically reducing obesity by 5 BMI points would only reduce mortality by 5%, from 1.05 to 1. 
 doi: 10.1186/s12916-015-0484-3.
Competing interests: No competing interests
Re: BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participants
I am grateful to the authors for responding (26May) to my comments of 8 May. May I, briefly, take up some points?
1. I agree that we should study the causes of preventable diseases so that we can try to prevent them. But, if you prevent heart disease in a given person and allow him to live longer till he dies a lingering death, from, say, malignancy, he (she) may not thank you.
You could justifiably claim that you, the epidemiologist merely reported and recommended population-level causes and prevention. However, you could, indeed ought to warn those who use your findings, that they should look at the individual patient before dishing out treatment.
2. You could, indeed should, warn doctors that the BMI is a rough tool, that fat being much lighter than bone and muscle, the BMI reflects largely, the welcome burden of bone and muscle rather than of adiposity. Please consider how easily research findings get accepted as valuable for producing guidelines by bodies such as NICE. Then the guidelines become straitjackets in the hands of those who decide whether you are a good doctor. Then doctors decide to delegate THEIR CLINICAL RESPONSIBILITY to the individual patient to the algorithm driven nurse or nursing assistant. In this process the patient becomes a tick-in-a-box.
3. You could warn insurance companies that when they blithely use such tools (as, say, through QRISK) to determine premiums, they will be acting wrongly and unfairly.
4. You acknowledge that the study did not report results stratified by "ethnicity" ... not possible to conduct detailed sub-group analysis.
This is one of the deficiencies of meta-analyses.
The very term "ethnicity" is meaningless. It is not " nationality". Thousands who were born in Danzig but moved (or were moved from ) to Germany: would you call them Germans by ethnicity, or Prussians by ethnicity or Polish by ethnicity? Elsewhere in the world these little difficulties are magnified thousands of times.
You mention in your paper, " various ethnic groups....including Asians, Caucasians".
Since Asia starts in Turkey (even Turkey straddles Europe and Asia) and sweeps Eastwards embracing China and Japan - one might even say the US state of Hawaii - the definition is very loose.
Asians include Mongols, Semites (Arabs and Jews) and Caucasians. Also, mixtures and small populations of aborigines. Population scientists ought to define their terms. Blumenbach did a good job, I think.
Competing interests: No competing interests
Re: BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participants
RE: "Higher mortality in "normal weight" adults: should the BMI range for normal weight be redefined?" by Professor Wang et al
We thank Professor Wang and colleagues for their comments on our paper on BMI and all-cause mortality. Wang et al state that our data does not support our conclusion that the difference in the results between our meta-analysis (1) and that of Flegal et al (2) may have been due to confounding by smoking and illness. We disagree with this statement for several reasons outlined below. In addition to confounding by smoking and existing illness there were several other differences between the two meta-analyses (described in detail in our original paper) that could account for the differences in the results. Below are the main issues that we think explain the difference in the results between our study and the study by Flegal et al:
1) Exclusion of a large number of relevant cohort studies
We have included more than 130 additional studies compared to the Flegal analysis (2), approximately 40 of which had been excluded for not using the WHO cut-off points for BMI and 53 studies which had been published after the publication of the work by Flegal et al., and a few additional studies that either may have been missed by their literature search or been excluded for other reasons. Altogether we had more than 10 times the number of deaths and participants (3.74 million deaths and 30.3 million participants among all participants vs. 270 000 deaths and 2.88 million participants) and more than twice the number of studies (230 studies vs. 97 studies) included compared to the Flegal analysis. Even our analysis of never smokers had almost 10 million participants and 738000 deaths which is 3-4 times the number of deaths and participants as the previous meta-analysis (2) and several times the number of deaths and participants as previous pooled analyses of never smokers (3;4). Our meta-analysis therefore provides a more comprehensive assessment of the available evidence which allows us to draw more robust conclusions.
2) Confounding by smoking and prevalent disease
Because there were quite large differences in the shape of the dose-response relationship between BMI and all cause mortality when the analyses were conducted among never smokers compared to all participants, current, former or ever smokers, with a higher nadir of the dose-response curve among all participants, and in the subgroups of current, former and ever smokers than among never smokers (the nadir was at a BMI of 25 in the former groups vs. 23-24 among never smokers, Figure 2) we disagree with the conclusion of Wang et al that confounding by smoking is not part of the explanation for the difference between our results and the results by Flegal et al. For example, the summary RR from the nonlinear dose-response analysis for a BMI of 27.5 compared to a BMI of 23 was 0.98 (0.96-1.01) for all participants, 0.99 (0.92-1.06) for current smokers, 1.00 (0.93-1.08) for former smokers, and 0.98 (0.93-1.04) for ever smokers, compared to 1.07 (1.01-1.14) among never smokers, and 1.11 (1.05-1.18) among healthy never smokers (Table 2). Although Wang et al claim that "the lower risk in the overweight group could not be explained by the confounding effect of smoking and pre-existing illness" our results show the contrary, that indeed there is an increase in risk in the overweight never smokers and healthy never smokers. This increase is not observed in the analyses of all participants and smokers in our publication, while in the analysis of Flegal et al there was an inverse association in the overweight group, suggesting the importance of confounding from smoking and prevalent disease in addition to their exclusion of a large number of available studies.
Further, in subgroup analyses among all participants, there was significant heterogeneity when studies were stratified for adjustment for cigarettes smoked per day, with stronger associations among studies with such adjustment (summary RR per 5 unit increase=1.17 (1.15-1.18) compared to studies without such adjustment (summary RR=1.03 (1.02-1.05)) (Supplementary Table F). We have also conducted additional nonlinear dose-response analyses among all participants by adjustment for number of cigarettes smoked per day which show a nadir of the dose-response curve is at a BMI of 24 among studies with such adjustment and between 25-27 among studies without such adjustment (Figure 1 and 2 below). There was also some indication of stronger associations among studies with adjustment for smoking status, and years since quitting smoking compared to studies without such adjustments (Supplementary Table F).
Flegal et al did not analyse underweight subjects and also did not conduct analyses for finer categories within the normal weight BMI range and therefore we are not able to compare our results for the underweight and low normal weight with them. However, our analyses also show that the increased risk in low normal weight and underweight subjects is much more pronounced in all participants, and in the subgroups of current, former and ever smokers than among never smokers and healthy never smokers. For example, the summary RR for a BMI of 16 compared to 23 was 1.83 (1.77-1.91) for all participants, 2.08 (1.84-2.34) for current smokers, 2.15 (1.91-2.42) for former smokers, and 1.81 (1.66-1.97) for ever smokers, compared to 1.66 (1.51-1.83) for never smokers and 1.31 (1.19-1.44) for healthy never smokers. In addition, the summary RRs for a BMI of 20 versus 23 were 1.15 (1.13-1.17) for all participants, 1.19 (1.13-1.25) for current smokers, 1.20 (1.14-1.26) for former smokers, 1.16 (1.12-1.20) for ever smokers, compared to 1.10 (1.05-1.14) for never smokers and 1.03 (0.99-1.07) for healthy never smokers. This suggests that the increased risk observed in subjects with a BMI of 20 vs. 23 may be largely explained by confounding from smoking and illness.
Many of the very large cohort studies excluded by Flegal et al for using more refined BMI categorisations provided results stratified by smoking status and prevalent disease and for many of these cohorts the results from the overall analysis differed considerably when analyses were restricted to never smokers and participants without prevalent disease (5-11). Flegal et al did not conduct analyses stratified by smoking status, adjustment for smoking or by whether prevalent disease had been excluded, but merely stated in the discussion that "adjustments or exclusions for these factors had little or no effect". Our findings provide strong evidence to the contrary and supports our conclusion that smoking and prevalent disease are important confounding factors that probably has influenced the results both in the meta-analysis of Flegal et al as well as in our meta-analysis of all participants by driving the nadir of the dose-response curve upwards as well as exaggerating the increased risk in the low normal weight and underweight categories. Our findings are also consistent with a large pooled analysis of individual participant data from the National Cancer Institute Consortium (3) and similar to the results from the Prospective Studies Collaboration (4).
Therefore we disagree with the statement by Wang et al that "the data could not support this explanation" and rather conclude that smoking and existing disease contributes to the U-shaped association observed in all participants.
3) Confounding by prediagnostic weight loss
We also acknowledged that there was some increase in risk among underweight and low normal weight subjects. However, undiagnosed disease and prevalent conditions for which data had not been collected in the studies may still have confounded those results and could explain the increased mortality with a low BMI. We therefore conducted additional analyses stratified by duration of follow-up because prediagnostic weight loss would have less and less influence on the results the longer the duration of follow-up is, because the early follow-up would tend to account for a smaller and smaller proportion of the total duration of follow-up with increasing duration of follow-up. As shown in Figure 3 and Table E of our paper, the increased risk in never smokers with a low normal BMI or who were underweight was substantially reduced among studies with longer durations of follow-up compared to studies with shorter durations of follow-up. For example, among never smokers the summary RR for a BMI of 20, 17.5, and 16 compared to 23 was 0.99 (0.94-1.03), 1.06 (0.97-1.16) and 1.16 (1.04-1.30), respectively, among studies with ≥20 years follow-up compared to summary RRs of 1.07 (1.00-1.14), 1.24 (1.10-1.41), and 1.43 (1.21-1.70) for studies with <10 years of follow-up. In addition, the summary RR for overweight (BMI=27.5) was 1.17 (1.09-1.26) for ≥20 years follow-up compared to 1.06 (0.98-1.15) for studies with <10 years follow-up. Even in the analysis of all participants there is a change in the shape of the dose-response relationship from a U-shaped association among studies with short durations of follow-up to a J-shaped association among studies with longer durations of follow-up (Table G). Our findings when stratified by duration of follow-up are similar to those of the NCI Cohort Consortium (3), and suggests the importance of prediagnostic weight loss on the results, although we cannot entirely exclude that weight changes after baseline also could contribute to this observation.
4) Study quality issues
As shown in our subgroup analyses of all participants, there was heterogeneity when studies were stratified by study quality, with stronger associations among studies with a study quality score of 7-9 than among those with a score of 4-6 (Supplementary Figure K and L, Supplementary Table F and I). The nadir of the curve was also higher (27.5) among studies with a study quality score of 4-6 than among studies with a study quality score of 7-9 (24-25).
We also only partly agree with the statement that the difference in the reference category between our meta-analysis and the one by Flegal et al is the explanation for the difference in the results. Because we used different methods (nonlinear models) which allowed us to include all studies with 3 or more categories of BMI we estimated the midpoints for each category when not provided in the article. For studies that used the WHO cut-off points for BMI the estimated midpoint for the normal weight category would be 21.7 ((18.5+24.9)/2=21.7) or close to 22. Given that the relative risks among never smokers and healthy never smokers with a BMI of 22 vs. 23 are 1.01 and 1.00 this would practically have no or minimal influence on the relative risk estimates among never smokers in Table 2 of our article. Among all participants and among smokers, the summary RRs may have been slightly more inverse in the overweight range (BMI=27.5) if we had used a BMI of 22 as the reference point instead of a BMI of 23, as the RR for a BMI of 22 vs. 23 was 1.03-1.04. For example for all participants the summary RR would have been 0.95 instead of 0.98 for a BMI of 27.5 if using a BMI of 22 rather than 23 as the reference (0.98/1.03=0.95).
We stated in our paper that " The current results support these recommendations [WHO recommendations] but suggest that the lowest mortality is observed with a BMI of 22-24 (depending on whether prevalent disease is excluded or not), although we cannot entirely rule out the possibility that this might be a slight overestimate if the increased risk observed among people with a BMI of 20 is non-causal, as indicated by the studies with longer durations of follow-up." Taking all the potential biases into account (including the influence from prediagnostic weight loss), our findings suggest that there is little evidence of increased risk in the low normal weight categories.
The study by Yi et al was included in our meta-analysis and we agree that there may be differences in the association between BMI and mortality by age, sex, or ethnicity as we also found some indication of this in our analyses. We provided results stratified by sex in Table D, and for African American subjects in the supplementary appendix (Figure D) and results stratified by geographic location in Table H (which also included Asian studies which would largely include Asian ethnicities). Because the number of studies was relatively small when stratified by geographic location we were not able to reliably conduct further analyses within these strata by whether prevalent disease was excluded or stratified by duration of follow-up (which would be important to exclude confounding by prevalent disease and prediagnostic weight loss). We also showed that there were differences in the association when studies were stratified by age with stronger associations in younger than older populations (Table J), and this is consistent with other large studies (3).
The study by Afzal et al (12) is limited by the fact that the cohort studies included in their analysis differed substantially by the length of follow-up, with 19.8, 11.0 and 4.6 years of follow-up in the first (oldest), second and third (newest) study, respectively. Therefore it is possible that bias due to prediagnostic weight loss may have influenced the studies to different degrees (e.g. more bias in shorter than longer studies) and that this could be at least part of the explanation for the change in the nadir of the curve over time. For example looking at Table G in our supplement, the nadir of the dose-response curve among all participants is reduced from 27.5 among studies with <5 years of follow-up compared to 23-24 among studies with longer durations of follow-up (≥20 or ≥25 years) (equal to a 3.5 to 4.5 point difference in BMI). Although we agree that a change in the nadir of the curve over time is possible, for example because of better treatment and management of risk factors among overweight and obese subjects or due to changes in the prevalence of other modifying risk factors, more studies are needed before a firm conclusion can be made with regard to this question. Especially, studies with longer durations of follow-up that are less influenced by prediagnostic weight loss are needed.
In addition, an increasing number of studies also report increased risk of the incidence of a number of chronic conditions (13-18) even within the high end of the normal BMI range, thus any revisions to the current recommendations need to take these findings into consideration given the current epidemic of overweight and obesity worldwide (19).
RE: "BMI, adiposity, "overweight", whatever they mean" by Dr. Anand
Epidemiology is the science of occurence and determinants of diseases. Epidemiologists should work on topics that are important for contemporary public health. Because overweight, obesity and chronic non-communicable diseases account for a large and increasing part of the burden of disease worldwide (20), many of which are to a large degree preventable with a healthy lifestyle, it is only natural that increasing resources are being spent on identifying the causes of as well as on prevention of these conditions.
We never refused in our study that populations are mixed, but because many of the studies did not report results stratified by ethnicity it was not always possible to conduct detailed subgroup analyses for many of the ethnic groups Dr. Anand mentions. We did report results separately for the few studies available among African American subjects and stratified by geographic location (which could be considered a crude, although imperfect indicator of ethnicity, at least for some areas) (Table H). Pooled analyses have also reported on the relationship between BMI and mortality in various ethnic groups, including Asians (21), Asian Americans (22), African Americans (23) and Caucasian s (3) with largely similar findings as in our study.
Rather than laying epidemiology to rest, we think more resources should be spent on epidemiological research to identify the causes of chronic non-communicable diseases to increase the possibilities for prevention. If people kept falling off a cliff it would be a lot wiser and make a lot more sense to put up a fence on top of the cliff rather than increasing the number of ambulances picking up the dead and wounded at the bottom.
RE: "Better mechanistic understanding of what drives the obesity paradox should serve, is required" by Dr. Koh
We thank Dr. Koh for his insights and agree with many of the points he mentions. Our results can be generalised to mainly free-living populations, and not to specific patient populations like persons with diabetes and heart failure as studies in such patient groups were excluded from our analysis. Therefore our results cannot address survival or a potential obesity paradox in such patient groups.
We agree that currently there is not strong evidence supporting lower cut-off points for obesity among Asians based on the available BMI and mortality data.
RE: "Contribution (confounding) of clear-cut diseases due to obesity should be excluded before concluding on all other causes of mortality" by Dr. Gupta
We thank Dr. Gupta for his interest in our study. Because overweight and obesity contributes to the development of diabetes, CVD, cancer and other chronic conditions that can result in death, exclusion or adjustment for incident cases or deaths from such conditions is not appropriate practice and is therefore not recommended (24). All-cause mortality is indeed a heterogenous outcome and includes causes of death 1) where overweight and obesity is detrimental (cancer, cardiovascular disease, diabetes etc.), 2) where both underweight and excess weight may be detrimental (infections), 3) where obesity may be protective (suicide), and 4) where there may be no relationship with weight. However, for both the individual and for the society and public health, all-cause mortality is of major importance as it provides an overall measure of the risk with overweight and obesity. We agree with Dr. Gupta that the most prevalent conditions for which overweight and obesity are risk factors (cardiovascular disease, cancer, diabetes) probably account for much of the increased risk observed for all-cause mortality. We also agree that it will be interesting to investigate BMI and specific causes of death, and also stated this in the second last paragraph: "Any further studies should investigate in more detail the association between BMI and other adiposity measures and specific causes of death, including less common diseases contributing to all cause mortality, and take into account the important methodological issues that have been highlighted in the current meta-analysis."
RE: "ECDF: maintaining a metabolically healthy profile for the obese" by Dr. Wang
We agree with Dr. Wang that there are a number of complications related to overweight and obesity and that it is possible that some subgroups of the population that may have less derangements with regard to metabolic health may have a lower risk than persons with such risk factors. Nevertheless, there is evidence that even metabolically healthy overweight and obesity compared to normal weight is associated with increased risk of cardiovascular disease, diabetes and mortality (25-29), so while interventions should be made to manage metabolic health to reduce the risk associated with overweight and obesity, we should also be careful not to trivialize adverse effects of metabolically healthy overweight and obesity.
Correction of error in Table 1.
We have discovered three minor errors in the published manuscript, which do not alter any of the results in any way, but which we nevertheless would like to correct. We regret these errors.
The correct affiliation for Serena Tonstad is 4 Section of Preventive Cardiology, Department of Endocrinology, Morbid Obesity and Preventive Medicine, Oslo University Hospital Ulleval, Oslo, Norway
The number of participants among never smokers in Table 1 should be >9 976 077 rather than >9 981 558.
There should also have been a footnote to Table 1 reading as follows with regard to the number of studies (risk estimates) by study size: "The size of the individual studies contributing to larger combined studies was not always available and for such studies the total number of participants for all the studies combined are used, although each individual study is counted as a separate study and will have contributed to a lower number of participants than the combined study in total."
1. Aune D, Sen A, Prasad M et al. BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participants. BMJ 2016;353:i2156.
2. Flegal KM, Kit BK, Orpana H, Graubard BI. Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA 2013;309:71-82.
3. Berrington de Gonzalez A, Hartge P, Cerhan JR et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med 2010;363:2211-9.
4. Whitlock G, Lewington S, Sherliker P et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet 2009;373:1083-96.
5. Engeland A, Bjorge T, Selmer RM, Tverdal A. Height and body mass index in relation to total mortality. Epidemiology 2003;14:293-9.
6. Adams KF, Schatzkin A, Harris TB et al. Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old. N Engl J Med 2006;355:763-78.
7. Jee SH, Sull JW, Park J et al. Body-mass index and mortality in Korean men and women. N Engl J Med 2006;355:779-87.
8. Freedman DM, Ron E, Ballard-Barbash R, Doody MM, Linet MS. Body mass index and all-cause mortality in a nationwide US cohort. Int J Obes (Lond) 2006;30:822-9.
9. Pischon T, Boeing H, Hoffmann K et al. General and abdominal adiposity and risk of death in Europe. N Engl J Med 2008;359:2105-20.
10. Zhang X, Shu XO, Chow WH et al. Body mass index at various ages and mortality in Chinese women: impact of potential methodological biases. Int J Obes (Lond) 2008;32:1130-6.
11. Boggs DA, Rosenberg L, Cozier YC et al. General and abdominal obesity and risk of death among black women. N Engl J Med 2011;365:901-8.
12. Afzal S, Tybjaerg-Hansen A, Jensen GB, Nordestgaard BG. Change in Body Mass Index Associated With Lowest Mortality in Denmark, 1976-2013. JAMA 2016;315:1989-96.
13. World Cancer Research Fund/American Insitute for Cancer Research. Food, Nutrition, Physical Activity and the Prevention of Cancer: a Global Perspective. Washington DC: AICR, 2007.
14. Field AE, Coakley EH, Must A et al. Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Arch Intern Med 2001;161:1581-6.
15. Aune D, Navarro Rosenblatt DA, Chan DS et al. Anthropometric factors and endometrial cancer risk: a systematic review and dose-response meta-analysis of prospective studies. Ann Oncol 2015;26:1635-48.
16. Aune D, Sen A, Norat T et al. Body Mass Index, Abdominal Fatness and Heart Failure Incidence and Mortality: A Systematic Review and Dose-Response Meta-Analysis of Prospective Studies. Circulation 2016;133:639-49.
17. Aune D, Norat T, Vatten LJ. Body mass index and the risk of gout: a systematic review and dose-response meta-analysis of prospective studies. Eur J Nutr 2014;53:1591-601.
18. Aune D, Norat T, Vatten LJ. Body mass index, abdominal fatness and the risk of gallbladder disease. Eur J Epidemiol 2015;30:1009-19.
19. NCD Risk Factor Collaboration. Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19.2 million participants. Lancet 2016;387:1377-96.
20. GBD 2013 Risk Factors Collaborators: Forouzanfar MH, Alexander L, Anderson HR et al. Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks in 188 countries, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015;386:2287-323.
21. Zheng W, McLerran DF, Rolland B et al. Association between body-mass index and risk of death in more than 1 million Asians. N Engl J Med 2011;364:719-29.
22. Park Y, Wang S, Kitahara CM et al. Body mass index and risk of death in Asian Americans. Am J Public Health 2014;104:520-5.
23. Cohen SS, Park Y, Signorello LB et al. A Pooled Analysis of Body Mass Index and Mortality among African Americans. PLoS One 2014;9:e111980.
24. Rothman, K. J., Greenland, S., and Lash, T. L. Modern Epidemiology. 2008. Philadelphia: Wolters Kluwer/ Lippincott Williams & Wilkins.
Ref Type: Generic
25. Roberson LL, Aneni EC, Maziak W et al. Beyond BMI: The "Metabolically healthy obese" phenotype & its association with clinical/subclinical cardiovascular disease and all-cause mortality -- a systematic review. BMC Public Health 2014;14:14.
26. Morkedal B, Vatten LJ, Romundstad PR, Laugsand LE, Janszky I. Risk of Myocardial Infarction and Heart Failure Among Metabolically Healthy But Obese Individuals: HUNT (Nord-Trondelag Health Study), Norway. J Am Coll Cardiol 2014;63:1071-8.
27. Kramer CK, Zinman B, Retnakaran R. Are metabolically healthy overweight and obesity benign conditions?: A systematic review and meta-analysis. Ann Intern Med 2013;159:758-69.
28. Fan J, Song Y, Chen Y, Hui R, Zhang W. Combined effect of obesity and cardio-metabolic abnormality on the risk of cardiovascular disease: a meta-analysis of prospective cohort studies. Int J Cardiol 2013;168:4761-8.
29. Lotta LA, Abbasi A, Sharp SJ et al. Definitions of Metabolic Health and Risk of Future Type 2 Diabetes in BMI Categories: A Systematic Review and Network Meta-analysis. Diabetes Care 2015;38:2177-87.
Lars J. Vatten1
1 Department of Public Health and General Practice, Faculty of Medicine, Norwegian University of Science and Technology, Trondheim, Norway
2 Department of Epidemiology and Biostatistics, Imperial College, London, UK
3 Department of Community Medicine, Postgraduate Institute of Medical Sciences,
Rohtak, Haryana, India
4 Section of Preventive Cardiology, Department of Endocrinology, Morbid Obesity
and Preventive Medicine, Oslo University Hospital Ulleval, Oslo, Norway
Competing interests: No competing interests
In this systematic review and meta-analysis, Aune and colleagues concluded that their analysis “provides strong evidence that overweight and obesity increases the risk of all-cause mortality” . The conclusion that overweight associated with increased mortality contradicted a previous meta-analysis , in which Flegal and colleagues reported that overweight (BMI: 25-29.9) was associated with lower mortality relative to normal weight (BMI:18.5-24.9) . Aune and colleagues explained the seemingly contradictory conclusions as the previous meta-analysis being prone to confounding by smoking and existing illness . However, it appears that their data could not support this explanation.
To illustrate the relation between BMI and mortality from their published data, we generated Figure 1 for never smokers and Figure 2 for healthy never smokers . We included a narrower BMI range in those figures to focus on the comparison between normal weight (BMI: 18.5-24.9) and overweight (BMI: 25-29.9). Among never smokers, the relative risk (RR) was 1.01 for BMI of 25 and 1.07 for BMI of 27.5, with BMI of 23 as the reference. The RR for BMI of 27.5 was equivalent to that for BMI of 20.7, indicating that normal weight adults with BMI of 18.5-20.7 had a higher risk than overweight adults with BMI of 25-27.5. Therefore, the lower risk in the overweight group could not be explained by the confounding effect of smoking. Even in healthy never smokers, those with BMI of 25 still had a lower risk (1.03) than those in the lower normal weight range (BMI: 18.5-20) (Figure 2), suggesting that the lower risk associated with overweight could not be fully explained by the confounding effect of smoking and pre-existing illness.
There was a key methodological difference between two meta-analyses. Flegal and colleagues used the WHO BMI categories with normal weight (18.5-24.9) as the reference while Aune and colleagues used BMI of 23, which was about the nadir of the BMI-mortality curve, as the reference. It would be interesting to know if both meta-analyses would have reached the same conclusion had they used the same reference.
Aune and colleagues further suggested that their results support the WHO recommendation of the normal BMI of 18.5-24.9. On the contrary, as shown in Figures 1 and 2, BMI of 18.5-20 marked a substantially increase in mortality, and the risk decreased steeply with increasing BMI. Therefore, instead of debating about whether overweight is associated with higher or lower mortality relative to normal weight, we should shift our focus to evaluating the appropriateness of the current WHO BMI normal weight range. As large amount of data have been accumulated, researchers are in a good position to reassess and redefine normal BMI values. Based on large Korean data, Yi and colleagues  found that optimal BMI ranges associated with minimal risk of death varied with age and sex, and were generally higher than the current normal weight (BMI of 18.5-24.9), suggesting that the WHO normal BMI range is questionable for Korean adults. In another study of three Danish cohorts, Afzal and colleagues found that BMI associated with the lowest mortality increased by 3.3 from 1976-1978 (BMI: 23.7) to 2003-2013 (BMI: 27.0) . Therefore, it is possible that the current normal BMI range based on evidence from several decades ago may be no longer applicable to contemporary adults.
Although there were over 9 million never smokers included in this meta-analysis, the nonlinear curves were established based on a very limited number of data points available in the original articles. Since a large amount of relevant data have already been collected, sharing de-identified original data with individual observations is critical for updating BMI values for normal weight.
1. Aune D, Sen A, Prasad M, et al. BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participants. BMJ 2016;353:i2156 doi: 10.1136/bmj.i2156.
2. Flegal KM, Kit BK, Orpana H, Graubard BI. 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):71-82 doi: 10.1001/jama.2012.113905.
3. Yi SW, Ohrr H, Shin SA, Yi JJ. Sex-age-specific association of body mass index with all-cause mortality among 12.8 million Korean adults: a prospective cohort study. Int J Epidemiol 2015 doi: 10.1093/ije/dyv138.
4. Afzal S, Tybjaerg-Hansen A, Jensen GB, Nordestgaard BG. Change in Body Mass Index Associated With Lowest Mortality in Denmark, 1976-2013. JAMA 2016;315(18):1989-96 doi: 10.1001/jama.2016.4666.
Competing interests: No competing interests
This is a combined plea to the " epidemiologists", statisticians, funders.
There was a time, when I was young and epidemiology meant something. It was the epidemiology of smallpox, cholera and other acute infectious diseases. When you (rather I and others engaged in public health MEDICINE) went out to see a patient reportedly returned from a small pox area. You were perhaps not born then.
But now? The epidemiology of non-infectious disease. The killer cardiovasular disease.
We all have to die some time. The planet is said to be creaking under the burgeoning population of fat, adipose, obese, high BMI people.
The population is described as Asian, Chinese (as if China had migrated to another continent) South Asian, European, Caucasian, Black.
These researchers refuse to acknowledge that TODAY, the population of England is very much " mixed". It may be relatively pure in parts of Norway. It certainly is a mixture in London, in Sicily, in Gibraltar, in Malta.
Pakistanis? Indians? If you have read history, you might know that people of these countries are largely mixed and also have discrete populations of "aboriginal" tribes who have not been totally assimilated.
The Iranians, largely Caucasian, have Semitic input.
The man or woman on your examination couch is very likely to have some genes which might not manifest themselves in his melanin, his features, his hair. This patient may know some of his/her ancestry. Or, might not. Or, might not share the information with you.
I request that this epidemiogly be laid to rest. The workers should be retrained to work in clinical medicine. We are said to be short of doctors in paediatrics, in A and E, in general practice.
Competing interests: I like medical time and public funds to be spent wisely.
Aune et al (1) report a meta-analysis of 230 prospective studies with more than 3.74 million deaths among more than 30.3 million participants, providing further evidence that adiposity [measured by body mass index (BMI)] increases the risk of premature death. The lowest mortality was observed in the BMI range 20-22 among studies of never smokers with longer durations of follow-up (≥20 and ≥25 years). The findings show the importance of smoking and comorbidity in confounding the association between BMI and mortality and contributing to the apparent paradox of a U shaped association. Song et al (2) used an interesting strategy to try to find out how weight trajectories from age 5 to 50 years influence all cause and cause specific mortality among adults over 60 years of age. The investigators studied associations between changes in somatotypes over time and mortality outcomes, using data from two large US prospective cohort studies. The authors found that people who reported remaining lean throughout life had the lowest mortality and that those who reported being heavy as children and who remained heavy or gained further weight had the highest mortality. Gaining weight from childhood to age 50 was associated with increased mortality compared with people who reported remaining lean. Weight gain was more strongly associated with cardiovascular than all cause mortality, and the effect was more pronounced among never smokers than ever smokers.
Recently, Zhao et al (3) investigated the association between BMI and death risk among patients with diabetes mellitus in a prospective cohort study. They observed that a U-shaped association of BMI with all-cause mortality risk among black and white patients with type 2 diabetes mellitus. A significantly increased risk of all-cause mortality was observed among blacks with BMI <30 kg/m2 and ≥35 kg/m2 and among whites with BMI <25 kg/m2 and ≥40 kg/m2 compared with patients with BMI of 30 to 34.9 kg/m2. Khalid et al (4) investigated the prognostic impact of pre-morbid obesity in patients with heart failure (HF) in the ARIC study. They observed that patients who were overweight or obese before HF development have lower mortality after HF diagnosis compared with normal BMI patients. Thus, weight loss due to advanced HF may not completely explain the protective effect of higher BMI in HF patients.
Obesity has reached epidemic proportions and complications related to obesity contribute substantially to health care costs and mortality. Obesity, particularly when accompanied by an excess of visceral/ectopic fat, is a major risk factor for diseases ranging from insulin resistance, type 2 diabetes, nonalcoholic fatty liver disease, and cardiovascular disease. The epidemic proportion reached by obesity has made these conditions a global problem in human health. (5,6) In addition, adipose depots surrounding the heart, such as epicardial and perivascular adipose may also exert important roles in the pathogenesis of cardiovascular disease beyond the contribution of visceral fat due to their close anatomic relationships with vascular structures and myocardium. (7) Many experimental and clinical studies have demonstrated the pathophysiological links between visceral, epicardial and perivascular adipose and atherosclerotic cardiovascular disease. (5,6)
Thus, it is surprising that obese and overweight patients with established type 2 diabetes seem to have a better prognosis than lean patients. Indeed, a growing number of studies have described a paradoxically longer survival (the obesity paradox) among overweight and obese adults. (8) Can obesity really confer protection in patients with type 2 diabetes? The “obesity paradox” is on the basis of findings from observational studies, which have 3 principal threats to validity. The first and second are chance and unmeasured factors (confounders) separately associated with exposure and outcome can create the false impression that exposure and outcome are causally related. In regards to the obesity paradox, a potential confounder could be muscle and bone mass wasting, a feature of sarcopenia that is an independent predictor of frailty and subsequent mortality. It does not seem both are limitations. The third category of threats to study validity is bias. Bias can occur in participant selection for a study sample (selection bias), data ascertainment (information bias), or both. (9)
Indeed, many issues regarding assessment/management of obesity remain to be addressed. The first issue is that obesity is measured by various means, such as BMI, waist circumference, waist-to-hip ratio, or by assessing visceral adiposity by imaging techniques. This means that the diagnosis of obesity could be influenced by the index used. It is already well established that there are considerable variations in the waist circumference and visceral adipose tissue at any given BMI value. (10,11) The second issue is that although WHO proposed lower BMI cut-off points for obesity among Asians, (12) cohort studies about the relationships between BMI and mortality do not seem to consistently support the need for lower BMI thresholds for the Asian population. (13-15) Finally, within the Asian population itself, there may be differences in body composition/adipose tissue distribution.
What are the clinical and scientific implications of understanding the obesity paradox? A causal link between obesity and better survival would have intriguing implications. Some studies have demonstrated biologically-plausible mechanisms to explain, however, there are few. To understand obesity’s physiological consequences, further work is warranted in both experimental and human studies, with the latter collecting more detailed metabolic and physiological data. Ultimately, a better mechanistic understanding of what drives the obesity paradox should serve as the basis for postulating whether an “optimal” BMI exists for patients and whether interventions are warranted to maintain or achieve this BMI. (5-7)
Funding: None, Disclosures: None
1. Aune D, Sen A, Prasad M, et al. Body mass index and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 prospective studies with 3.74 million deaths among 30.3 million participants. BMJ 2016;353:i1256.
2. Song M, Hu FB, Wu K, et al. Trajectory of body shape in early and middle life and all cause and cause specific mortality: results from two prospective US cohort studies. BMJ 2016;353:i2195.
3. Zhao W, Katzmarzyk PT, Horswell R, et al. Body mass index and the risk of all-cause mortality in patients with type 2 diabetes mellitus. Circulation 2014;130:2143–51.
4. Khalid U, Ather S, Bavishi C, et al. Pre-morbid body mass index and mortality after incident heart failure: the ARIC study. J Am Coll Cardiol 2014;64:2743–9.
5. Lim S, Després J-P, Koh KK. Prevention of atherosclerosis in overweight/obese patients: in need of novel multi-targeted approaches. Cir J 2011;75:1019-27.
6. Kim SH, Després J-P, Koh KK. Obesity and cardiovascular disease: friend or foe? Eur Heart J 2015 Dec 18. pii: ehv509. [Epub ahead of print] Review.
7. Lee H-Y, Després J-P, Koh KK. Perivascular adipose tissue in the pathogenesis of cardiovascular disease. Atherosclerosis 2013;230:177-84.
8. Flegal KM, Kit BK, Orpana H, Graubard BI. Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA. 2013;309:71–82.
9. Wang TJ. The obesity paradox in heart failure: weighing the evidence. J Am Coll Cardiol. 2014;64:2750-2.
10. Després JP. Excess visceral adipose tissue/ectopic fat the missing link in the obesity paradox? J Am Coll Cardiol 2011;57:1887-9.
11. Nazare JA, Smith J, Borel AL, et al; INSPIRE ME IAA Investigators. Usefulness of measuring both body mass index and waist circumference for the estimation of visceral adiposity and related cardiometabolic risk profile (from the INSPIRE ME IAA study). Am J Cardiol 2015;115:307-15.
12. WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004;363:157-63.
13. Jee SH, Sull JW, Park J, et al. Body-mass index and mortality in Korean men and women. N Engl J Med 2006;355:779-87.
14. Gu D, He J, Duan X, et al. Body weight and mortality among men and women in China. JAMA 2006;295:776-83.
15. Zheng W, McLerran DF, Rolland B, et al. Association between body-mass index and risk of death in more than 1 million Asians. N Engl J Med 2011;364:719-29.
Competing interests: No competing interests
Contribution (confounding) of clear-cut diseases due to obesity should be excluded before concluding on all other causes of mortality:
All-cause mortality is nothing but a crude death rate. It is known that obese people are predisposed to diabetes, cardiovascular diseases or metabolic syndrome, cancer, etc.. These serious diseases which can make obese people prone to death should be excluded from analysis or adjusted or are equal in all BMI groups, so that inference on other causes of mortality can be reliably conferred on obesity. Probably the diseases of which obesity is a clear-cut risk factor are contributing to major chunk of all-cause mortality.
Competing interests: No competing interests
The systematic review and meta-analysis bring novel insights on the detrimental impact of overweight and obesity to the scientific community. The association of body mass index (BMI) and mortality has been newly defined in terms of quantitative curve graphs. Indeed, accumulating evidence shows that overweight and obesity have negative effects on the health of the musculoskeletal system. Overweight and obese people are at a higher risk of developing lumbar spinal stenosis according to a large Swedish prospective cohort analysis . Moreover, obesity is an independent risk factor for venous thromboembolic events after spinal fusion, which includes pulmonary emboli and deep venous thromboses . As well, obesity contributes somewhat to low back pain and intervertebral disc degeneration (IDD) .
In general, the common ill effects of obesity include insulin resistance, hyperglycemia, hypertension, and dyslipidemia . Notwithstanding obesity brings various harms to the health of the population; we should note that there is a subgroup of obese individuals free from the ill effects. Based on a 20-year follow-up of young adults, the metabolically healthy tips have been unraveled . On the other hand, we should not only fully understand the novel vision of BMI and mortality, but propose practicable strategies for the fat to keep metabolically healthy and minimize the mortality rate. Accordingly, we propose Education, Cession of smoking, Dietary pattern adjustment, and Fitness (ECDF) for the people with overweight and obesity.
Education is to spread the harmful effects of overweight and obesity among the public. As a spinal surgeon, few people are fully aware of the concept of BMI and the standard for judging overweight and obesity during my practice in out-patient clinics. Therefore, there is a still a long way to go to improve the perception of the public.
In addition to the direct harmful effects to cardiorespiratory system, smoking is definitely harmful for the development and treatment outcome of spinal diseases. Smoking is linked with the severity and quality of life of spondyloarthritis in a dose-dependent manner [7 8]. Furthermore, smoking causes DNA damage and promotes vertebral bone degeneration and IDD . Smoking is an independent predictor of reoperation after lumbar laminectomy . Therefore, it is of critical importance for the obese to quit smoking due to their increased risks for spinal diseases per se.
Dietary pattern adjustment means replacing the Western dietary pattern with a prudent dietary pattern. The updated Dietary Guidelines for Americans  suggest replacing intake of saturated fats with unsaturated fats. The hallmarks of a Western dietary pattern are rich in saturated fats, margarines and oil with low amount of dietary fiber; whereas the characteristics of a prudent dietary pattern are to the contrary. Prudent dietary pattern corresponds to low dietary inflammation index and is beneficial for health .
The unraveled metabolically healthy tips consist of high cardiorespiratory fitness through physical exercise, as well as “6 instruments to obesity at risk”, which is a criterion for the fat . The 6 instruments are as follows: (a) triglycerides≥1.7mmol/L, (b) HDL cholesterol<1.04mmol/L (male), 1.29mmol/L (female), (c) LDL cholesterol≥4.1mmol/L, (d) systolic blood pressure≥130 mmHg or diastolic blood pressure≥85 mm Hg, (e) fasting glucose≥5.6mmol/L, (f) HOMA-IR≥2.5.
In fact, as the editor-in-chief of WeChat platform as "Spine Truth" (jzglyl), we have released several science popular essays through our platform in Feb, 2016, the Spring Festival of China.
1. Aune D, Sen A, Prasad M, et al. BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participants. BMJ 2016;353:i2156 doi: http://dx.doi.org/10.1136/bmj.i2156[published Online First: Epub Date]|.
2. Knutsson B, Sanden B, Sjoden G, et al. Body Mass Index and Risk for Clinical Lumbar Spinal Stenosis: A Cohort Study. Spine 2015;40(18):1451-6 doi: 10.1097/BRS.0000000000001038[published Online First: Epub Date]|.
3. Goz V, McCarthy I, Weinreb JH, et al. Venous Thromboembolic Events After Spinal Fusion: Which Patients Are at High Risk? The Journal of bone and joint surgery American volume 2014;96(11):936-42 doi: 10.2106/JBJS.L.01602[published Online First: Epub Date]|.
4. Dario AB, Ferreira ML, Refshauge KM, et al. The relationship between obesity, low back pain, and lumbar disc degeneration when genetics and the environment are considered: a systematic review of twin studies. The spine journal : official journal of the North American Spine Society 2015;15(5):1106-17 doi: 10.1016/j.spinee.2015.02.001[published Online First: Epub Date]|.
5. Brochu M, Tchernof A, Dionne IJ, et al. What are the physical characteristics associated with a normal metabolic profile despite a high level of obesity in postmenopausal women? The Journal of clinical endocrinology and metabolism 2001;86(3):1020-5 doi: 10.1210/jcem.86.3.7365[published Online First: Epub Date]|.
6. Fung MD, Canning KL, Mirdamadi P, et al. Lifestyle and weight predictors of a healthy overweight profile over a 20-year follow-up. Obesity 2015;23(6):1320-5 doi: 10.1002/oby.21087[published Online First: Epub Date]|.
7. Poddubnyy D, Haibel H, Listing J, et al. Cigarette smoking has a dose-dependent impact on progression of structural damage in the spine in patients with axial spondyloarthritis: results from the GErman SPondyloarthritis Inception Cohort (GESPIC). Annals of the rheumatic diseases 2013;72(8):1430-2 doi: 10.1136/annrheumdis-2012-203148[published Online First: Epub Date]|.
8. Wendling D, Prati C. Spondyloarthritis and smoking: towards a new insight into the disease. Expert review of clinical immunology 2013;9(6):511-6 doi: 10.1586/eci.13.35[published Online First: Epub Date]|.
9. Nasto LA, Ngo K, Leme AS, et al. Investigating the role of DNA damage in tobacco smoking-induced spine degeneration. The spine journal : official journal of the North American Spine Society 2014;14(3):416-23 doi: 10.1016/j.spinee.2013.08.034[published Online First: Epub Date]|.
10. Bydon M, Macki M, De la Garza-Ramos R, et al. Smoking as an independent predictor of reoperation after lumbar laminectomy: a study of 500 cases. Journal of neurosurgery Spine 2015;22(3):288-93 doi: 10.3171/2014.10.SPINE14186[published Online First: Epub Date]|.
11. Frank AP, Clegg DJ. Dietary Guidelines for Americans-Eat Less Fat. Jama 2016;315(17):1919 doi: 10.1001/jama.2016.0972[published Online First: Epub Date]|.
12. Johns DJ, Lindroos AK, Jebb SA, et al. Dietary patterns, cardiometabolic risk factors, and the incidence of cardiovascular disease in severe obesity. Obesity 2015;23(5):1063-70 doi: 10.1002/oby.20920[published Online First: Epub Date]|.
Competing interests: Supported by National Natural Science Foundation of China (No. 81270028 and 81572182).