Advertisement

Sign up for our daily newsletter

Advertisement

Mortality attributable to obesity: Mortality attributable to obesity among middle-aged adults in the united states

The effect of age on the association between body-mass index and mortality.

William Thompson
Sunday, May 17, 2020
Advertisement
  • Kushi, P. Gregg, E.

  • These estimates are potentially very sensitive to minor differences. Skip Nav Destination Article Navigation.

  • In the target population, the respective risks were 1.

  • Estimating deaths attributable to obesity in the United States. Another consideration is whether the estimate of interest is all deaths attributable to overweight and obesity or only premature deaths occurring at ages of less than 75 years.

MeSH terms

A review of adjusted estimators of attributable risk. View all jobs. InAllison et al.

We then repeated the calculations by using the identical relative risks but applied the partially adjusted method and treated the population ogesity a single age-sex group. A lower prevalence of mortality attributable to obesity in the derivation cohort increased the bias from the partially adjusted method; a higher prevalence decreased the bias table 5. Oxford University Press is a department of the University of Oxford. FDA talk paper T, April 29, This paper has focused on how estimates of the number of deaths attributable to a risk factor can be biased if the estimates are calculated improperly or if the relative risks are not estimated accurately. Google Scholar. This method is in general biased when there is confounding of the exposure-disease association because of the use of an attributable risk formula appropriate for unadjusted relative risks only.

  • FDA approves dexfenfluramine to treat obesity.

  • For example, the total number of deaths attributable to obesity has been invoked in connection with new drug approvals 6. Effect of differences in relative risk in exposure categories between the derivation sample and the target population.

  • Article Google Scholar Price, G.

  • This equation is appropriate when there is no confounding.

N Engl J Med ; : TABLE 5. Correspondence to Dr. We used the weighted-sum method to calculate mortqlity correct number of deaths due to overweight and obesity in the hypothetical situation in which the relative risks were identical for all age-sex subgroups in the US population, as shown in table 1 confounding-only example. JAMA ; : — For example, in the study by Calle et al.

Attributable fractions were calculated and used mortality attributable to obesity estimate mortapity deaths. McGinnis and Foege 5 subsequently clarified that their estimate ofdeaths per year referred to all aspects of diet and physical activity and not specifically to obesity. It is well established in the statistical literature that using an adjusted relative risk estimate with an attributable fraction formula for an unadjusted relative risk will not in general give unbiased estimates when there is confounding TABLE 2. Katherine M. Body weight and coronary heart disease mortality: an analysis in relation to age and smoking habit.

Publication types

We calculated a total ofdeaths due to overweight and obesity table 3. Mass media, 5 wttributable 8 scholarly journals, surgery center columbia - 11 and pharmaceutical handouts 12 have citeddeaths per year in the United States as being attributable to obesity, a number that may have been adapted from an analysis of precursors of premature death in the United States for13 attributingdeaths to "overnutrition. Their data also support the use of a single recommended range of body weight throughout life. We consider two general equations to calculate PAF, as described by Rockhill et al. An evidence-based assessment of federal guidelines for overweight and obesity as they apply to elderly persons.

Dietz, E. N Engl J Med ; : 1 —7. An increase in the efficacy and availability of such interventions would reduce HRs associated with obesity. However, this may be offset by mortality attributable to obesity fact that when most of the cohort studies used were initiated, there were fewer intervention strategies to reduce risk factors associated with obesity and fewer medical therapies for postponing death from obesity-related diseases. However, if the original data from the various sources are available, bootstrap or jackknife methods 17 can be applied to obtain standard errors from which confidence intervals can be constructed.

FDA talk paper T, April 29, One set did not vary by age-sex group representing possible confounding but no interaction. In attdibutable, Allison et al. These three older age groups were selected because they account for over half of deaths in adults. The method used by Allison et al. In this paper, we investigate the possible magnitude and direction of the bias in estimates of deaths attributable to obesity when confounding and effect modification are not adequately accounted for.

Publication types

Google Scholar Pollack, C. Cleeman, D. Article Google Scholar Oliver, J.

Kuo, K. Chang Authors Neil K. Article Google Scholar Calle, E. Article Google Scholar U.

Article Google Scholar Serdula, M. Williamson, and M. Obesity: Preventing and Managing the Global Epidemic. This cohort is nationally representative of the US civilian noninstitutionalized population and includes all 14, persons who completed the medical examinations at baseline Article Google Scholar Couzin, J.

  • Smith, J. Sign in to download free article PDFs Sign in to access your subscriptions Sign in to your personal account.

  • Estimates of numbers of deaths attributable to overweight and obesity may be surprisingly unreliable, particularly if the relative risks are based on small numbers of outcomes or if the sample sizes for estimating exposure to risk factors are not large.

  • Related articles in Web of Science Google Scholar.

  • New issue alert. Passaro, R.

  • It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide. Am J Epidemiol.

However, when there is effect modification, such differences affect the degree obesity bias from the partially adjusted method. Barry I. The relative proportions of the elderly, the death rates for the nonobese, and the prevalence of obesity in many cohorts are likely to be lower than those in the US population as a whole, and the relative risk may also be higher because of exclusions. We allowed the proportion in the derivation cohort to increase or decrease by three percentage points, to 6. Dire warnings about obesity rely on slippery statistic. N Engl J Med. The relative risk estimates in a derivation cohort may not be perfectly identical to the true relative risks in a population.

We calculated a total ofdeaths due to overweight and obesity mortality attributable to obesity 3. Considerable evidence suggests that the effects of obesity on mortality differ strongly by age 8 — Our numerical examples, using plausible values for the US population, suggest that the Allison et al. In addition, these estimates appear to be sensitive to minor differences in relative risks between a derivation cohort and the target population. The biases were greater for older people and for those with higher BMIs.

References

Gaziano, I. Accessed March 11, J Chronic Dis. Cubbin, S. Carroll, M.

  • The HR for a nonobese person also not in the reference category relative to someone in the reference category is q.

  • Kassirer JP, Angell M. Existing estimates of the number of deaths attributable to overweight and obesity 7 were calculated by using a method likely to produce biased estimates when the effects of obesity vary by age or other characteristics.

  • Weight categories were normal range Krueger, P.

It is often stated that obesity is a major cause of death in the United States, accounting for as many asdeaths per year and rivaling smoking as a public health threat 1 — 3. TABLE 4. This method is in general biased when there is confounding of the exposure-disease association because of the use of an attributable risk formula appropriate for unadjusted relative risks only. The bias in the partially adjusted method can be seen by comparing the values for the partially adjusted method in table 4 with the values for both weighted sums. When there are subgroups within the population e.

Body-mass index and mortality in a prospective cohort of U. Effect of age on excess mortality in obesity. This equation is appropriate when there is no confounding. Am J Epidemiol.

Publications

In the target population, the respective risks were 1. In another approach, referred to by Benichou 13 as the weighted-sum method, unadjusted relative risks are calculated separately for obwsity subgroup in the derivation cohort and used in equation 1 or 2, along with information on the prevalence of obesity and the number of deaths within each subgroup, to calculate the number of deaths attributable to obesity in each subgroup. In this study, the authors investigated the possible magnitude and direction of bias in estimating deaths attributable to obesity when such a method is used.

Am J Public Health. This Issue. McDowell, C. We calculated a total ofdeaths due to overweight and obesity table 3. Idler, E. Pfahlberg, and O. Woo, J.

ALSO READ: Nice Guidelines Cg43 Obesity Facts

N Engl J Med ; : Corrections were made for the reputed regression-dilution bias by using the average BMI during the decade before follow-up as predictor. View Metrics. A review of adjusted estimators of attributable risk. Looking for your next opportunity? One set did not vary by age-sex group representing possible confounding but no interaction.

We report a detailed analysis aimed obesity calculating the annual aytributable of deaths attributable to obesity. Chideya, C. Korn, E. The overall pattern was similar to HRs derived for all subjects. Because of the distributive property of PAF, this equation can also be used to calculate a level-specific attributable fraction for any given level of exposure, relative to a fixed reference category, by including only that level of exposure in the numerator.

Additional bias resulted from slight differences between the derivation cohort and the target population. InAllison et al. BMI category No. Arch Intern Med. For convenience, the percentage of the population in each sex-age group was calculated from the sample weights for the Third National Health and Nutrition Examination Survey sample, but note that the target population for the National Health and Nutrition Examination Survey is the civilian noninstitutionalized population rather than the total population.

Petitti DB. TABLE 2. Article Google Scholar Woo, J. Ballard-Barbash, A. This can be expressed as follows:.

Peeters, J. Article Google Scholar Poortinga, W. Issue Section:. Citing articles via Web of Science Second, because persons go from being alive to dead over some time frame, use of relative risk RR estimates from studies without adjustment for time can bias results though the bias may be small.

Cross-sectional analysis: precursors of premature death in the United States. A constant hazard rate implies an exponential survival distribution. The Tecumseh Study: design, progress, and perspectives. An estimateddeaths per year are due to the obesity epidemic Article Google Scholar Gregg, E.

Kawas, F. In both of the above situations, if relative risks and deaths attributable to overweight and obesity are calculated separately for each subgroup the weighted-sum methodthe sum of deaths attributable to overweight and obesity will be mortality attributable to obesity. Obesify articles in Web of Science Google Scholar. Arch Intern Med. Criteria used to select the data sets included 1 US source; 2 public availability or availability via extraction from published reports ie, HRs for BMI categories ; 3 not derived predominantly from ill, high-risk, or elderly subjects; and 4 well-documented characteristics. To estimate the number of deaths due to obesity, these relative risk estimates were then combined with data from the target population the US population on the prevalence of obesity and the number of deaths.

For example, the total number of deaths attributable to obesity has been invoked in attrjbutable with new drug approvals 6. When there is confounding only, such differences sc obesity surgery center columbia sc have no effect on the bias from the partially adjusted method. Estimates of deaths attributable to obesity can be biased if confounding and effect modification are not properly taken into account or if the relative risks are not estimated accurately. Certain types of differences between derivation cohort and population appear more likely than others.

Dire warnings about obesity rely on slippery statistic. These numbers are then summed over subgroups to obtain the total number of deaths in the population attributable to obesity. The equation used by Allison et al. Am J Public Health.

The relative attibutable estimates in a derivation cohort may not be perfectly identical to the true relative risks in a population. Stat Methods Med Res. InMcGinnis and Foege 4 estimated thatdeaths a year in the United States were due to poor diet and physical inactivity, making these combined factors the second leading modifiable factors, after smoking, contributing to death. The equation used by Allison et al. To calculate deaths attributable to overweight and obesity, the attributable fraction and attributable deaths can be calculated for each level of overweight and obesity and the results added together. Efron B, Tibshirani RJ. These explorations suggest that when the mortality relative risk is estimated with a single age-adjusted relative risk, overestimation of deaths attributable to overweight and obesity in the US population is more likely than underestimation.

ALSO READ: Proteobacteria And Bacteroidetes And Obesity

The first equation is. Arch Intern Med. Obesity this paper, we investigate the possible magnitude and direction of the bias in estimates of deaths attributable to obesity when confounding and effect modification are not adequately accounted for. Advanced Search. Citing articles via Web of Science This method is in general biased when there is confounding of the exposure-disease association because of the use of an attributable risk formula appropriate for unadjusted relative risks only.

We recommend that sensitivity analyses be included in future studies. In the target population, the obewity risks were 1. Abstract Previously reported estimates of deaths attributable to obesity in the United States have been based on a method that only partially adjusts for confounding and does not allow for effect modification. National Center for Health Statistics. However, the effect of eliminating smokers from the data set does not seem to be a lowering of risk in the very lean nor a lowering of the BMI mortality curve nadir, but rather a slight increase in obesity hazard relative to average-weight persons. Gaziano, I. A Systematic Review.

  • We used the weighted-sum method to calculate the correct number of deaths due to overweight and obesity in the hypothetical situation in which the relative risks were identical for all age-sex subgroups in the US population, as shown in table 1 confounding-only example.

  • Article Contents Abstract. Am J Epidemiol ; : —

  • Article Google Scholar Janssen, I.

  • In effect, this approach uses a fully saturated model that can account for both confounding and interaction because it allows for different relative risks within each subgroup.

  • Rights and permissions Reprints and Permissions. Sempos, C.

Google Scholar Visscher, T. With respect to whether residual confounding or effect modification is likely, the literature is divided. Spelsberg, J. Google Scholar. Overall values are fairly consistent with a mean estimate ofannual deaths attributable to overweight or obesity range, ,

N Engl J Med ; : attributable —7. These numbers are then summed over subgroups to obtain the total number of deaths in the population attributable to obesity. The weighted-sum method is not affected by these differences between the derivation cohort and the target population. To calculate deaths attributable to obesity, the total number of deaths is multiplied by the population attributable fraction PAFwhich may be interpreted as the proportion of deaths attributable to obesity. Barry I.

We recommend ogesity sensitivity analyses be included in future studies. To estimate the number of deaths due to obesity, these relative risk estimates were then combined with data from the target population the US population on the prevalence of obesity and the number of deaths. US Food and Drug Administration. The hypothetical true values are shown as WS1 weighted sum 1, for confounding only and WS2 weighted sum 2, for effect modification.

Pierson, T. We allowed the proportion in the derivation cohort to increase or decrease by three percentage points, to 6. January 1, Hummer, and J. Rights and permissions Reprints and Permissions.

Article Google Scholar Manson, J. The Framingham Heart Study, initiated in to assess prospectively cardiovascular disease risk factors among a two-thirds sample of the residents of Framingham, Mass, 23 consisted of persons response rate, Am J Public Health. Corrada, M.

We follow the Allison et al. Assuming the population tp and characteristics are stable ie, that death rates and mortality attributable to obesity rates are maintaining the population at equilibriumthe hazard rate, averaged across all members of the population at any 1 point in time, must be constant over time. Sign in to save your search Sign in to your personal account. In addition, these estimates appear to be sensitive to minor differences in relative risks between a derivation cohort and the target population. We have no way of knowing whether residual confounding or effect modification is the more likely explanation. Lipsitz, and E.

Table 2. The Nurses' Health Study 25 NHS was established inwhenfemale registered nurses 30 to mortality attributable to obesity years of mkrtality completed questionnaires on medical history, height and weight, and health behavior. However, if the original data from the various sources are available, bootstrap or jackknife methods 17 can be applied to obtain standard errors from which confidence intervals can be constructed. Hypothetical examples are based on US population data and published relative risks. Gronniger, J.

N Engl J Med ; : 1 —7. A lower prevalence of obesity in the derivation cohort increased attributabke bias from the partially adjusted method; a higher prevalence decreased the bias table 5. However, when there is effect modification interactionthat is, differential effects of obesity by age or other factors, yet further bias can arise from the partially adjusted method when the derivation cohort differs from the target population. These explorations suggest that when the mortality relative risk is estimated with a single age-adjusted relative risk, overestimation of deaths attributable to overweight and obesity in the US population is more likely than underestimation. The number of deaths attributable to obesity in the target population is then calculated by multiplying the total number of deaths by the attributable fraction.

Although those hazard ratios were adjusted for three factors—age, sex, morfality smoking—here we consider only age and sex because mortality data are tabulated by age and sex. It is often stated that obesity is a major cause of death in the United States, accounting for as many asdeaths per year and rivaling smoking as a public health threat 1 — 3. Using average BMI as predictor increased the estimate to JAMA ; : —8. Johannes L, Stecklow S. We selected this second set of values to show a decline with relative risks by age, consistent with the literature on this topic. Death rates were higher at older ages than at younger ages and were lower for women than for men.

Grundy, and C. Hollenbeck, and M. Alternately, these negative values may represent random fluctuations because of sampling variation.

Another consideration is whether the estimate of interest is all deaths attributable to overweight and obesity atributable only premature deaths occurring at ages of less than 75 years. Abstract Objective: To assess whether a recent qttributable that found a relatively small number of excess deaths attributable to obesity may have underestimated by not correcting for statistical biases. We recommend that sensitivity analyses be included in future studies. The degree of bias has not often been quantified, however. Even if the relative risks within age-sex subgroups are identical between the derivation cohort and the target population, the derivation cohort may differ from the target population in terms of the distribution of subgroups, the prevalence of obesity, or the absolute risk of disease within the reference BMI category. Thus, the bias due to applying this method in the US population may increase over time as a higher proportion of deaths occur in older people and as obesity increases.

  • Estimating deaths attributable to obesity in the United States. Google Scholar Visscher, T.

  • In another approach, referred to mortality attributable to obesity Benichou 13 as the weighted-sum method, unadjusted relative risks are calculated separately for each subgroup in the derivation cohort and used in equation 1 or 2, along with information on the prevalence of obesity and the number of deaths within each subgroup, to calculate the number of deaths attributable to obesity in each subgroup. Article Navigation.

  • Accessed February 18, Early deaths were not excluded as has sometimes been advocated, 38 in part because there is no published proof or formal statistical justification for merits of this technique.

  • Google Scholar. In the US population, age and sex are associated with both obesity and mortality and thus confound the obesity-mortality relation.

  • Ho, A. Childhood obesity: future directions and research priorities.

Stat Methods Med Res ; 10 : — This equation is appropriate when there is no confounding. The relative risk estimates in a derivation cohort may not be perfectly identical to the true relative risks in a population. JAMA ; : —8. The two sets of hypothetical relative risks within subgroups that we used are shown in table 1.

In our numerical examples, a combination of these characteristics, including an overestimation of the relative risk by 0. Efron B, Tibshirani RJ. The relative risks associated with obesity may be different in epidemiologic cohort studies than in the general US population In the article by Allison et al.

The bias in the partially adjusted method can be seen by comparing the values for obesity partially adjusted method attributale table 4 with the values for both weighted sums. However, when there is effect modification interactionthat is, differential effects of obesity by age or other factors, yet further bias can arise from the partially adjusted method when the derivation cohort differs from the target population. Another set arrived at empirically varied by age-sex group representing effect modification or interaction but resulted in the same final adjusted relative risks adjusted across subgroups by using the Mantel-Haenszel method.

Article Google Scholar Jemal, A. Google Scholar Manson, J. The risk of mortality increased with increasing BMI at all ages and for all categories of death. Stampfer, G.

Durazo-Arvizu, D. If the relative risk estimates from the attributabl cohort are slightly inaccurate, however, neither a single adjusted mortality attributable to obesity estimate nor the weighted-sum method will give the correct results. They expressed HRs as age-adjusted and adjusted for numerous potential confounders eg, smoking, menopausal status, oral contraceptive and postmenopausal hormone use, and parental history of myocardial infarction before age Previously reported estimates of deaths attributable to obesity in the United States have been based on a method that only partially adjusts for confounding and does not allow for effect modification. Hennekens, and F. The number of deaths attributable to obesity in the target population is then calculated by multiplying the total number of deaths by the attributable fraction.

InMcGinnis and Foege 4 estimated thatdeaths a year in the United States were due to poor diet and physical inactivity, making these combined factors the second leading modifiable factors, after smoking, contributing to death. New issue alert. Arch Intern Med. We recommend that sensitivity analyses be included in future studies. For convenience, the percentage of the population in each sex-age group was calculated from the sample weights for the Third National Health and Nutrition Examination Survey sample, but note that the target population for the National Health and Nutrition Examination Survey is the civilian noninstitutionalized population rather than the total population.

Although those hazard ratios were adjusted for three factors—age, sex, obesity smoking—here we consider only age and sex because mortality data are tabulated by age and sex. We recommend that sensitivity analyses be included in future studies. These three older age groups were selected because they account for over half of deaths in adults. JAMA ; : —8. We also considered the effects of minor differences in relative risk between the derivation cohort and the target population. In the article by Allison et al. Objective: To assess whether a recent study that found a relatively small number of excess deaths attributable to obesity may have underestimated by not correcting for statistical biases.

Proceedings of the American Statistical Association Sesquicentennial Allison et al. We estimate h and q from each data source Table 1. Laird, W.

Annual deaths attributable to obesity in the United States. For example, the number of deaths attributable to obesity can be calculated separately for each age group mortaality age-specific values for relative risk, the prevalence of obesity, and the number of deaths and then added together across age groups. Although those hazard ratios were adjusted for three factors—age, sex, and smoking—here we consider only age and sex because mortality data are tabulated by age and sex. Nonetheless, this estimate has often been interpreted as representing the number of deaths caused by obesity 12 and is also often cited to motivate increased efforts to treat and control obesity. View Metrics.

If you have questions concerning secton three of this report, contact the Division of Health Promotion at Rodriguez, K. Mozaffar, and A. Williams, K. The Tecumseh Study: design, progress, and perspectives. Looking for your next opportunity?

ALSO READ: Roupas Gestante Obesity

Mokdad, A. Even if the relative risks within age-sex subgroups are identical between the derivation mortality attributable to obesity and the target population, the derivation cohort may differ from the target population in terms of the distribution of subgroups, the prevalence of obesity, or the absolute risk of disease within the reference BMI category. Alternately, had we relied only on studies in which the BMI-mortality association increases in a monotonic not "U-shaped" manner 34 and set either the threshold for overweight and the reference category lower, the number of attributable deaths would have increased substantially. The effect of age on the association between body mass index and mortality. Kalmijn, S.

This equation is appropriate when there is no confounding. The hypothetical true values are shown as WS1 weighted sum 1, for confounding only and WS2 weighted sum 2, for effect modification. InAllison et al. N Engl J Med ; :

Tecumseh Community Health Study. The investigation of body weight and all-cause mortality reported by Manson et al 34 that provided the data herein was based on information aboutwomen without diagnosed cardiovascular disease or cancer in who reported height and weight. A direct association was observed between BMI and mortality among women who had never smoked. InMcGinnis and Foege 4 estimated thatdeaths a year in the United States were due to poor diet and physical inactivity, making these combined factors the second leading modifiable factors, after smoking, contributing to death. Brody, L.

FlegalKatherine M. As our hypothetical relative risks, we used the relative risk hazard attributahle estimates from the Allison et al. The older age groups make up a small proportion of the population but account for disproportionate numbers of deaths. Effects of interaction, confounding and observational error on attributable risk estimation.

Article Google Scholar Sempos, C. This cohort is nationally representative of the US civilian noninstitutionalized population and includes all 14, persons who completed the medical examinations at baseline Cite Cite Katherine M. We recommend that sensitivity analyses be included in future studies. Geneva, Switzerland: World Health Organization;

Hu, M. Article Google Scholar Mortality attributable to obesity, J. Their data also support the use of a single recommended range of body weight throughout life. Body weight and coronary heart disease mortality: an analysis in relation to age and smoking habit. Krieger, N. Early deaths were not excluded as has sometimes been advocated, 38 in part because there is no published proof or formal statistical justification for merits of this technique. After controlling for preexisting disease, the mean annual number of obesity-attributable deaths was estimated to be, based on CPS1 data andbased on NHS data.

Sidebar1?
Sidebar2?