Whether waist circumference provides clinically meaningful information not delivered by body-mass index regarding prediction of cause-specific death is uncertain.
We prospectively examined waist circumference (WC) and body-mass index (BMI) in relation to cause-specific death in 225,712 U.S. women and men. Cox regression was used to estimate relative risks and 95% confidence intervals (CI). Statistical analyses were conducted using SAS version 9.1.
During follow-up from 1996 through 2005, we documented 20,977 deaths. Increased WC consistently predicted risk of death due to any cause as well as major causes of death, including deaths from cancer, cardiovascular disease, and non-cancer/non-cardiovascular diseases, independent of BMI, age, sex, race/ethnicity, smoking status, and alcohol intake. When WC and BMI were mutually adjusted in a model, WC was related to 1.37 fold increased risk of death from any cancer and 1.82 fold increase risk of death from cardiovascular disease, comparing the highest versus lowest WC categories. Importantly, WC, but not BMI showed statistically significant positive associations with deaths from lung cancer and chronic respiratory disease. Participants in the highest versus lowest WC category had a relative risk of death from lung cancer of 1.77 (95% CI, 1.41 to 2.23) and of death from chronic respiratory disease of 2.77 (95% CI, 1.95 to 3.95). In contrast, subjects in the highest versus lowest BMI category had a relative risk of death from lung cancer of 0.94 (95% CI, 0.75 to 1.17) and of death from chronic respiratory disease of 1.18 (95% CI, 0.89 to 1.56).
Increased abdominal fat measured by WC was related to a higher risk of deaths from major specific causes, including deaths from lung cancer and chronic respiratory disease, independent of BMI.
Citation: Leitzmann MF, Moore SC, Koster A, Harris TB, Park Y, et al. (2011) Waist Circumference as Compared with Body-Mass Index in Predicting Mortality from Specific Causes. PLoS ONE 6(4): e18582. doi:10.1371/journal.pone.0018582
Editor: Vineet Gupta, University of Pittsburgh Medical Center, United States of America
Received: July 30, 2010; Accepted: March 12, 2011; Published: April 26, 2011
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: This research was supported by the Intramural Research Program of the National Institutes of Health (NIH), National Cancer Institute. AARP has not provided direct funds to the NIH-AARP Diet and Health Study. The AARP contribution has consisted of help in drawing the study sample and sponsoring Dr. Yikyung Park as a research fellow for two years. That funding went to the NCI (National Cancer Institute) but not to the NIH-AARP Diet and Health Study itself. AARP has also provided funds for the development of an online physical activity assessment tool, but that project is not part of the current study. In addition, AARP had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The term ‘AARP’ is not an acronym, but has been the legal name of the organization for the past 15 years. AARP was formerly known as the American Association of Retired Persons.
Competing interests: The authors have declared that no competing interests exist. Specifically, the NIH-AARP Diet and Health Study has no connection through AARP (the organization) to any potential contaminating influences through employment, consultancy, patents, products in development or marketed products.
Body-mass index (body weight in kilograms divided by the square of height in meters, BMI) is currently the most frequently used and widely accepted method to classify medical risk according to weight status. BMI is a useful measure of adiposity in young and middle-aged adults . However, an important limitation of the BMI is its inability to distinguish between fat mass and fat-free mass, which show opposing relations with health risk . In addition, the validity of BMI as an indicator of fatness in elderly persons is limited  because fat-free mass decreases with aging, even without a change in overall weight .
In contrast, waist circumference (WC) represents a measure of adiposity that takes into account the accumulation of abdominal fat. WC is simple to measure and interpret and it is highly correlated with visceral fat as assessed by computed tomography . WC has increased significantly in United States (U.S.) women and men in recent decades, a secular trend that mirrors the development of generalized adiposity in this country. The potentially greater relevance of WC than BMI for predicting adverse health conditions in the elderly is underscored by the observation that fat mass accumulates intra-abdominally with age . Older age groups have experienced a greater increase in abdominal obesity than younger age groups .
In a large, contemporary cohort of U.S. women and men, we examined the relations of general and abdominal adiposity to death from specific causes. Our study differs from most previous investigations in quantifying the dose-response associations according to both WC and BMI.
Materials and Methods
The NIH-AARP Diet and Health Study began in 1995–1996 when 566,407 AARP members aged 50 to 71 years and residing in one of six U.S. states (CA, FL, LA, NJ, NC, and PA) or two metropolitan areas (Atlanta, GA, and Detroit, MI) satisfactorily completed a baseline questionnaire on their medical history, current body weight, height, smoking habits, and diet . In 1996–1997, a second questionnaire requesting additional information on abdominal circumference was sent to baseline questionnaire respondents within six months and was returned by 59.5 percent of participants. The Special Studies Institutional Review Board of the U.S. National Cancer Institute accepted return of the questionnaires as an indication of voluntary participation, in lieu of signed consent.
Population for analysis
To incorporate the WC data collected on the second questionnaire, our analysis included the 234,182 potentially eligible participants who provided complete information regarding their WC on the second questionnaire. We excluded individuals with a BMI less than 18.5 kg/m2 (n = 2,720) because very low weight measured at older age is likely to reflect leanness as a consequence of chronic illness. We also excluded participants with a BMI greater than 60 (n = 65). We excluded subjects with missing information on smoking (n = 2,548). We also excluded persons with extreme values (beyond three inter-quartile ranges of the 75th and 25th percentiles) for WC (n = 3,137). The remaining analytical cohort comprised of 225,712 subjects.
Cohort members were followed-up from November 8, 1996 through December 31, 2005 by annual linkage of the cohort to the National Change of Address database maintained by the U.S. Postal Service and through processing of undeliverable mail, other address change update services, and directly from cohort members' notifications. For matching purposes, we have virtually complete data on first and last name, address history, gender, and date of birth. Social Security number is available for 85 percent of our cohort.
Vital status is ascertained by annual linkage to the Social Security Administration Death Master File in the U.S.  Verification of vital status and cause of death information is provided by follow-up searches of the National Death Index (NDI) Plus. We estimate that our mortality ascertainment is greater than 93 percent complete , . Cohort maintenance also involves periodic linkage to the eleven state cancer registries serving our cohort [eight baseline states plus three most common states of re-location (TX, AZ, and NV) during follow-up] .
Assessment of anthropometric variables
WC was assessed in the second questionnaire using a pictured instruction. Participants were requested to measure their waist with a tape measure one inch above the navel while standing and to report values to the nearest quarter inch. Participants for whom a tape measure was not available were asked to leave a blank response. Information on body weight and height was requested in the baseline questionnaire.
For the analyses of WC, participants were divided into the following categories: normal (women: less than 80 centimeters; men: less than 94 centimeters), action level 1 (women: 80 to 87 centimeters; men: 94 to 101 centimeters), and action level 2 (women: 88 centimeters or higher; men: 102 centimeters or higher). Those categories incorporate the WHO cut-points of 88 centimeters or higher for women and 102 centimeters or higher for men that define a large WC  and they correspond to the action levels proposed by Lean et al.  We also created a more extreme group of WC of 96 centimeters or higher for women and 118 centimeters or higher for men to investigate a broader range of WC in relation to death. The group of subjects with normal WC (less than 80 centimeters for women; less than 94 centimeters for men) served as the reference group.
Participants were divided into four BMI categories that correspond to the definitions of normal weight (18.5 to 24.9kg/m2), overweight (25.0 to 29.9 kg/m2), obesity class 1 (30.0 to 34.9 kg/m2), and obesity classes 2 or 3 (35 kg/m2 or greater) proposed by the World Health Organization (WHO) . The group of subjects with a BMI of 18.5 to 24.9 kg/m2 served as the reference group.
Our categorizations of WC and BMI resulted in similar fractions of the at-risk population in the lowest WC and BMI categories and in similar fractions of the at-risk population in the highest WC and BMI categories.
In large cohorts similar to ours, comparisons of self-reported and measured WCs showed correlation coefficients of 0.89 in women and 0.95 in men . The correlations between self-reported as compared with technician-measured BMI values have been found to be in the area of 0.94 for women and 0.92 for men .
Cox proportional hazards regression  with person-time as the time scale was used to estimate the relative risks and corresponding 95 percent confidence intervals of death. Follow-up time was calculated from the return date of the second questionnaire until death from any cause or the end of study in December 31, 2005, whichever occurred first. Tests of the proportional hazards assumptions for exposures and covariates included in our models indicated no departures. Tests of linear trend were conducted by modeling the median values of WC or BMI categories as a single continuous variable, the coefficient for which was evaluated using a Wald test.
The multivariate models were adjusted for age, gender, race/ethnicity, smoking status, and alcohol intake. We did not adjust for physical activity because physical activity is an important determinant of body weight. We considered education level as a potential confounder but did not retain that variable in the final model because it failed to substantially alter beta coefficients in our primary exposures. The analyses of WC were additionally adjusted for BMI, thereby producing a variable reflecting mainly abdominal fat depots. We tested for and found no effect modification by gender and therefore show results for women and men combined. All reported P values are based on two-sided hypothesis tests. Statistical analyses were conducted using SAS version 9.1 (SAS Institute, Cary, NC).
During 1,961,011 person-years of follow-up, we documented 20,977 deaths. The mean (SD) ages at study entry and end of follow-up were 62.9 (5.3) and 71.7 (5.3) years, respectively. For men, the minimum and maximum WC sizes were 69.2 and 136.5 cm, respectively. For women, the minimum and maximum WC sizes were 50.8 and 138.4 cm, respectively. WC was positively correlated with BMI (r = 0.72). Participants with large WC or those with high BMI tended to have lower education levels, they were less likely to currently smoke, and they consumed less alcohol than their lean counterparts (data not shown).
Risk of death from any cause increased monotonically with higher values of WC (Table 1). Using participants with normal WC as the reference group, women with WC of 96 centimeters or higher and men with WC of 118 centimeters or higher showed a nearly 70 percent increased risk of death from any cause (multivariate relative risk = 1.68; 95 percent confidence interval, 1.55 to 1.81; P for trend<0.0001). The relation of BMI to death from any cause was of similar magnitude as that with WC. As compared with normal BMI, the multivariate relative risk of death from any cause for obesity classes 2 or 3 was 1.68 (95 percent confidence interval, 1.57 to 1.79, P for trend<0.0001).
WC was significantly positively associated with death from cancer, most notably death from lung cancer (Table 2). The multivariate relative risk of lung cancer death comparing extreme categories of WC was 1.77 (95 percent confidence interval, 1.41 to 2.23, P for trend<0.0001). In contrast, the risk estimates relating BMI to death from lung cancer were below unity and the BMI category representing overweight showed a statistically significant inverse association (multivariate relative risk as compared with normal weight = 0.92; 95 percent confidence interval, 0.85 to 0.99). The P value for the test for trend was statistically non-significant at 0.16.
Both WC and BMI were significantly positively related to death from cardiovascular disease (Table 3). The positive relations of high versus low categories of WC to death from any cardiovascular disease, death from coronary heart disease, and death from other cardiovascular disease (including death from stroke) were 23 percent, 31 percent, and 9 percent weaker, respectively, than those with high versus low categories of BMI.
WC showed a monotonically positive association with death from non-cancer/non-cardiovascular disease, particularly death from chronic respiratory disease (Table 4). The multivariate relative risk of death from chronic respiratory disease comparing extreme categories of WC was 2.77 (95 percent confidence interval, 1.95 to 3.95; P for trend<0.0001). By comparison, BMI showed a J-shaped relation with death from chronic respiratory disease, with strong inverse associations for overweight and class 1 obesity and a weak, statistically non-significant positive relation for the combination of classes 2 and 3 obesity (multivariate relative risk compared with normal weight = 1.18; 95 percent confidence interval, 0.89 to 1.56; P for inverse trend<0.0001). The only non-cancer/non-cardiovascular death endpoint that showed no statistically significant positive relation with WC was death from injuries (multivariate relative risk = 1.14; 95 percent confidence interval, 0.70 to 1.87; P for trend = 0.94) (Table 5).
We examined the combined effects of WC and BMI on risk of death (Table 6). For this analysis, we collapsed the top two BMI categories. In the group of participants with BMI values between 25.0 and 29.9 kg/m2, high versus low WC predicted an increased risk of death from any cause and death from specific causes. Participants with BMI levels of 18.5 to 24.9 kg/m2 also showed positive relations of WC to risk of death from any cause and death from specific causes, but risk estimates failed to reach statistical significance for the highest WC category and were statistically significant only for the second to highest WC category. BMI levels of 30.0 kg/m2 or higher attenuated the positive relation of WC to death.
The primary finding of this large prospective investigation of over 225,000 women and men followed for nine years is that adiposity as assessed by WC was strongly and consistently related to death from a comprehensive list of death from major specific causes, including deaths from lung cancer and chronic respiratory disease, independent of BMI and other covariates. In contrast, adiposity as assessed by BMI showed inconsistent associations with specific causes of death, displaying positive relations to death from non-lung cancers, cardiovascular disease, and non-cancer/non-cardiovascular diseases, but inverse or null associations with deaths from lung cancer and chronic respiratory disease. This finding indicates that an increased amount of abdominal fat, but not general adiposity, represents a consistent predictor of premature death from major specific causes.
Our results have significant clinical implications because they suggest that WC represents a superior predictor than BMI of risk of deaths from lung cancer and chronic respiratory disease, two major causes of death. For example, obtaining a WC measurement in individuals already at increased risk for respiratory death (e.g., due to smoking or chronic obstructive respiratory disease) provides important information not delivered by BMI on a patient's risk of premature mortality.
Though speculative, one possible biologic reason for the heterogeneous associations of WC and BMI to mortality from lung cancer and chronic respiratory disease is that lung cancer mortality and chronic respiratory disease mortality may involve biologic pathways related to insulin resistance and related metabolic abnormalities, such as excess release of proinflammatory and prothrombotic factors , , which are more strongly related to abdominal adiposity than to overall excess body weight.
An alternative explanation for the divergent relations of WC and BMI to deaths from lung cancer and chronic respiratory disease is heterogeneity with respect to residual confounding by smoking. Current smoking is related to both lower BMI  and increased mortality from lung cancer and chronic respiratory disease. By comparison, current smoking is associated with visceral fat accumulation  and smoking cessation is related to increases in WC . Limiting the investigation to participants who never smoked helps resolve the issue of residual confounding by smoking with respect to both BMI and WC but our cohort lacked sufficient numbers of cases of mortality from lung cancer and chronic respiratory disease among persons who never smoked.
To the best of our knowledge, no previous study has examined the association between WC and risk of death from lung cancer. However, our results regarding WC are consistent with previous epidemiologic investigations that have focused on lung cancer incidence. For example, the Iowa Women's Health Study found a positive relation of WC (relative risk comparing extreme categories = 1.76; 95 percent confidence interval, 1.14 to 2.73) to lung cancer incidence . Similarly, the Women's Health Initiative observed positive relations of WC to lung cancer incidence among current smokers (relative risk = 1.56; 95 percent confidence interval, 0.91 to 2.69) and former smokers (relative risk = 1.50; 95 percent confidence interval, 0.98 to 2.31) .
Our findings regarding the relation of BMI to lung cancer death are strikingly consistent with recent data from a large pooled analysis of 57 prospective studies . In that analysis, each 5kg/m2 increment in BMI was associated with a relative risk of lung cancer death of 0.71 (95 percent confidence interval, 0.63 to 0.79) within the BMI range of 15 to 25 kg/m2 and was associated with a relative risk of lung cancer death of 0.98 (95 percent confidence interval, 0.88 to 1.09) within the BMI range of 25 to 50 kg/m2 .
Our results of a monotonically positive relation of WC to respiratory death are comparable to recent data from the European Prospective Investigation into Cancer and Nutrition (EPIC) Study . In that study, the relative risk of respiratory death comparing extreme quintiles of WC among women was 2.95 (95 percent confidence interval, 1.52 to 5.70) and the corresponding relative risk among men was 6.56 (95 percent confidence interval, 3.60 to 11.96).
Our results of a J-shaped association between BMI and death from chronic respiratory disease are comparable to the aforementioned pooled analysis of 57 prospective studies . Each 5kg/m2 increase in BMI was related to a relative risk of respiratory death of 0.31 (95 percent confidence interval, 0.28 to 0.35) within the BMI range of 15 to 25 kg/m2 and was related to a relative risk of respiratory death of 1.20 (95 percent confidence interval, 1.07 to 1.34) within the BMI range of 25 to 50 kg/m2 .
Notable strengths of our study include its prospective design, a large number of deaths from specific causes, a high follow-up rate, and detailed information on potentially confounding factors. Despite a number of advantageous aspects of our study, one potential limitation is a low response rate to the second questionnaire used to obtain WC information, which could have resulted in selection bias if, for example WC was preferentially missing for persons with high mortality risk. A further potential limitation is that WC, weight, and height were assessed using self-report, a method that is known to be imperfect. However, self-reported BMI values and WC measures have been found to be sufficiently precise for use in epidemiologic studies , . Self-reported height and weight vary systematically with BMI , but the importance of such bias is uncertain . Taken together, it is improbable that misclassification of self-reported WC, height, and weight fully accounted for our finding of markedly divergent relations of general and abdominal adiposity to risk for deaths from lung cancer and chronic respiratory disease. We did not evaluate the association between WC and mortality from chronic lung disease according to smoking status because that topic is beyond the scope of the present investigation. Future studies should examine that relation.
We conclude that increased abdominal fat, but not general adiposity is associated with elevated risk of a comprehensive list of death from major specific causes, including deaths from lung cancer and chronic respiratory disease. Based on our observational epidemiologic data, deaths from lung cancer and chronic respiratory disease may potentially be added to the list of causes of death directly related to adiposity.
We are indebted to the participants in the NIH-AARP Diet and Health Study for their outstanding cooperation. We thank Leslie Carroll at Information Management Services and Sigurd Hermansen and Kerry Grace Morrissey from Westat for data support, and Tawanda Roy at the Nutritional Epidemiology Branch for research assistance. Dr. Leitzmann had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Conceived and designed the experiments: MFL SCM TBH AS. Performed the experiments: MFL SCM AK YP. Analyzed the data: MFL SCM AK YP. Contributed reagents/materials/analysis tools: AH AS. Wrote the paper: MFL TBH AS.
- 1. Schreiner PJ, Terry JG, Evans GW, Hinson WH, Crouse JR 3rd, et al. (1996) Sex-specific associations of magnetic resonance imaging-derived intra-abdominal and subcutaneous fat areas with conventional anthropometric indices. The Atherosclerosis Risk in Communities Study. Am J Epidemiol 144: 335–345.
- 2. Seidell JC, Perusse L, Despres JP, Bouchard C (2001) Waist and hip circumferences have independent and opposite effects on cardiovascular disease risk factors: the Quebec Family Study. Am J Clin Nutr 74: 315–321.
- 3. Harris TB, Visser M, Everhart J, Cauley J, Tylavsky F, et al. (2000) Waist circumference and sagittal diameter reflect total body fat better than visceral fat in older men and women. The Health, Aging and Body Composition Study. Ann N Y Acad Sci 904: 462–473.
- 4. Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T, et al. (1996) How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups? Am J Epidemiol 143: 228–239.
- 5. Molarius A, Seidell JC (1998) Selection of anthropometric indicators for classification of abdominal fatness–a critical review. Int J Obes Relat Metab Disord 22: 719–727.
- 6. Enzi G, Gasparo M, Biondetti PR, Fiore D, Semisa M, et al. (1986) Subcutaneous and visceral fat distribution according to sex, age, and overweight, evaluated by computed tomography. Am J Clin Nutr 44: 739–746.
- 7. Okosun IS, Chandra KM, Boev A, Boltri JM, Choi ST, et al. (2004) Abdominal adiposity in U.S. adults: prevalence and trends, 1960–2000. Prev Med 39: 197–206.
- 8. Schatzkin A, Subar AF, Thompson FE, Harlan LC, Tangrea J, et al. (2001) Design and serendipity in establishing a large cohort with wide dietary intake distributions : the National Institutes of Health-American Association of Retired Persons Diet and Health Study. Am J Epidemiol 154: 1119–1125.
- 9. Hill ME (2001) Re: “Comparison of National Death Index and world wide web death searches”. Am J Epidemiol 153: 719.
- 10. Rich-Edwards JW, Corsano KA, Stampfer MJ (1994) Test of the National Death Index and Equifax Nationwide Death Search. American Journal of Epidemiology 140: 1016–1019.
- 11. Michaud DS, Midthune D, Hermansen S, Leitzmann MF, Harlan L, et al. (2005) Comparison of cancer registry case ascertainment with SEER estimates and self-reporting in a subset of the NIH-AARP Diet and Health Study. J Registry Management 32: 70–75.
- 12. WHO (2000) Obesity: Preventing and managing the global epidemic. WHO Consultation on Obesity. Geneva: World Health Organization. Technical report series 894 Technical report series 894.
- 13. Lean ME, Han TS, Morrison CE (1995) Waist circumference as a measure for indicating need for weight management. Bmj 311: 158–161.
- 14. WHO (1995) Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee. World Health Organ Tech Rep Ser 854: 1–452.
- 15. Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB, et al. (1990) Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1: 466–473.
- 16. McAdams MA, Van Dam RM, Hu FB (2007) Comparison of self-reported and measured BMI as correlates of disease markers in US adults. Obesity (Silver Spring) 15: 188–196.
- 17. Cox DR (1972) Regression models and lifetables. J R Stat Soc (B) 34: 187–220.
- 18. Mohamed-Ali V, Pinkney JH, Coppack SW (1998) Adipose tissue as an endocrine and paracrine organ. Int J Obes Relat Metab Disord 22: 1145–1158.
- 19. Wajchenberg BL (2000) Subcutaneous and visceral adipose tissue: their relation to the metabolic syndrome. Endocr Rev 21: 697–738.
- 20. Barrett-Connor E, Khaw KT (1989) Cigarette smoking and increased central adiposity. Ann Intern Med 111: 783–787.
- 21. Komiya H, Mori Y, Yokose T, Tajima N (2006) Smoking as a risk factor for visceral fat accumulation in Japanese men. Tohoku J Exp Med 208: 123–132.
- 22. Pisinger C, Jorgensen T (2007) Waist circumference and weight following smoking cessation in a general population: the Inter99 study. Prev Med 44: 290–295.
- 23. Olson JE, Yang P, Schmitz K, Vierkant RA, Cerhan JR, et al. (2002) Differential association of body mass index and fat distribution with three major histologic types of lung cancer: evidence from a cohort of older women. Am J Epidemiol 156: 606–615.
- 24. Kabat GC, Kim M, Hunt JR, Chlebowski RT, Rohan TE (2008) Body mass index and waist circumference in relation to lung cancer risk in the Women's Health Initiative. Am J Epidemiol 168: 158–169.
- 25. Whitlock G, Lewington S, Sherliker P, Clarke R, Emberson J, et al. (2009) Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet 373: 1083–1096.
- 26. Pischon T, Boeing H, Hoffmann K, Bergmann M, Schulze MB, et al. (2008) General and abdominal adiposity and risk of death in Europe. N Engl J Med 359: 2105–2120.
- 27. Plankey MW, Stevens J, Flegal KM, Rust PF (1997) Prediction equations do not eliminate systematic error in self-reported body mass index. Obes Res 5: 308–314.
- 28. Song YM, Sung J (2001) Body mass index and mortality: a twelve-year prospective study in Korea. Epidemiology 12: 173–179.