International Journal of Obesity
SEARCH     advanced search my account e-alerts subscribe register
Journal home
Advance online publication
Current issue
Press releases
For authors
For referees
Contact editorial office
About the journal
For librarians
Contact Springer Nature
Customer services
Site features
NPG Subject areas
Access material from all our publications in your subject area:
Biotechnology Biotechnology
Cancer Cancer
Chemistry Chemistry
Dentistry Dentistry
Development Development
Drug Discovery Drug Discovery
Earth Sciences Earth Sciences
Evolution & Ecology Evolution & Ecology
Genetics Genetics
Immunology Immunology
Materials Materials Science
Medical Research Medical Research
Microbiology Microbiology
Molecular Cell Biology Molecular Cell Biology
Neuroscience Neuroscience
Pharmacology Pharmacology
Physics Physics
Browse all publications
March 2002, Volume 26, Number 3, Pages 410-416
Table of contents    Previous  Article  Next   [PDF]
Differential associations of body mass index and adiposity with all-cause mortality among men in the first and second National Health and Nutrition Examination Surveys (NHANES I and NHANES II) follow-up studies
D B Allison1, S K Zhu2, M Plankey2, M S Faith2 and M Heo2

1Department of Biostatistics and Clinical Nutrition Research Center, University of Alabama at Birmingham, Birmingham, Alabama, USA

2New York Obesity Research Center, St Luke's/Roosevelt Hospital, Institute of Human Nutrition, Columbia University College of Physicians and Surgeons, New York, USA

Correspondence to: D B Allison, Department of Biostatistics, Ryals Public Health Bldg 327, 1665 University Blvd, University of Alabama at Birmingham, UAB Station, Birmingham, AL 35294-0022, USA. E-mail:


OBJECTIVE: The frequently observed U-shaped relationship between body mass index (BMI; kg/m2) and mortality rate may be due to the opposing effects of fat mass (FM) and fat-free mass (FFM) components of BMI on mortality rate. The purpose is to test the hypothesis stated above.

DESIGN: Longitudinal prospective cohort studies. The mortality follow-up of the first and second National Health and Nutrition Examination Surveys (NHANES I and NHANES II).

SUBJECTS: A total of 10 169 male subjects aged 25-75 who participated in NHANES I and II were selected for analyses. Follow-up continued until 1992. The mean follow-up time was 14.6 y for NHANES I and 12.9 y for NHANES II. Ninety-eight percent of the participants were successfully followed representing a total of 3722 deaths.

MEASUREMENTS: Subscapular and triceps skinfolds thickness were used as FM indicators, whereas upper arm circumference was used as a FFM indicator. The Cox proportional hazards model tested the relationships of BMI, FM and FFM with all-cause mortality adjusting for age, smoking status, race and education levels.

RESULTS: BMI had a U-shaped relationship with mortality, with a nadir of approximately 27 kg/m2. However, when indicators of FM and FFM were added to the model, the relationship between BMI and mortality became more nearly monotonic increasing. Moreover, the relationship between FM indicator and mortality was monotonic increasing and the relationship between FFM indicator and mortality was monotonic decreasing.

CONCLUSION: These results support the hypothesis that the apparently deleterious effects of marked thinness may be due to low FFM and that, over the observed range of the data, marked leanness (as opposed to thinness) has beneficial effects.

International Journal of Obesity (2002) 26, 410-416. DOI: 10.1038/sj/ijo/0801925


BMI; adiposity; all-cause mortality; longitudinal cohort study


It is well established that obesity is associated with increased mortality rate.1 However, the cause of the frequently observed U- or J-shaped relationship between body mass index (BMI) and mortality rate remains poorly understood.2

It has been contended that the elevated mortality rate with low BMI is due to confounding by smoking and preexisting disease.3 While some studies support this hypothesis,4,5 most studies continue to find a U- or J-shaped curve between BMI and mortality rate even after attempting to control for smoking and pre-existing subclinical or occult disease.1,6 This puzzling finding seems inconsistent with the clinical observation that weight loss improves (short-term) health7 and experimental data that laboratory animals subjected to marked caloric restriction live substantially longer than animals fed ad libitum.8 Allison et al9 proposed an additional explanation for the elevated mortality seen with low BMI. Specifically, BMI was noted to be the summation of fat mass (FM)/meters2 and fat-free mass (FFM)/meters2. To the extent that FM and FFM have opposite effects on mortality, simulation studies demonstrated that a U- or J-shaped relationship between BMI and mortality rate could occur.9 With low BMIs, this might be due to elevated mortality due to insufficient FFM.

Several studies offer limited information about the potential validity of the hypothesis10,11,12 and have generally supported the expectation that measures of adiposity have a more nearly monotonic increasing relationship with mortality rate than do measures of total body mass or BMI. However, these studies were limited in sample size, body composition measurements (often relying on poorly validated measures), and the thoroughness of the statistical analysis (eg not assessing conditional effects, using inappropriate tests etc).

Three additional studies have addressed the relationship between adiposity per se, BMI and mortality rate.13,14,15 Lee et al16 examined associations of FM, FFM and fitness with all-cause mortality rate at mean follow-up of 8 y in 21 925 men in an observational study. Unfortunately, Blair et al13 did not assess the conditional effects of FM and FFM or of FM and BMI and therefore did not eliminate the confounding among these measures that may have existed due to their close correlation. Kalmijn et al14 examined the independent associations of BMI, waist-to-hip ratio and skinfold thickness with all-cause mortality rate among 3741 elderly Japanese-American men. Following subjects for a mean of 4.5 y, investigators found inverse and direct associations of BMI and waist-to-hip ratio, respectively, with mortality rate. Whereas increasing BMI was associated with reduced mortality rate, increasing waist-to-hip ratio was associated with increased mortality rate. Heitmann et al15 studied 22 y mortality as a function of BMI and body fat among 787 Swedish men followed from age 60 to 82 with body composition assessed by total body potassium counting. Results were consistent with those hypothesized by Allison et al.9 Specifically, there was a U-shaped relationship between BMI and mortality rate, a monotonic increasing relationship between adiposity and mortality rate, and a monotonic decreasing relationship between FFM and mortality rate. This study strongly supports the hypotheses that the U-shaped relationship between BMI and mortality rate may be due to the fact that BMI is a composite of FM and FFM. However, its relatively small sample size and the advanced age (60-82 y) of the subjects suggest the importance of additional studies in larger samples of the broader population.

To address this, we used data from the National Health and Nutrition Examination Surveys (NHANES) I and II, which consist of representative samples of the US civilian non-institutionalized population. In each survey, height, weight, subscapular and triceps skinfolds were measured and mortality data collected. Collectively, these data sets yield over 10 000 adult male subjects, making this the largest study to date to address the differential effects of adiposity and BMI on mortality rate and the first to do so in a nationally representative sample. We will report on women in a separate paper given the substantial sex differences in adiposity,17 body fat distribution,18 and mortality rate.19


Study sample

Subjects were from two sources: the National Health and Nutrition Examination Survey I (NHANES I), and the National Health and Nutrition Examination Survey II (NHANES II). NHANES I and II were multi-purpose cross-sectional surveys of the civilian non-institutionalized population of the US conducted from 1971 to 1980 by the National Center for Health Statistics using a stratified multi-stage clustered probability sampling design. Details about these two surveys are well described elsewhere.20,21,22,23 In the present study we restricted analyses to male subjects aged 25-75 y in NHANES I (5820 NHANES I) and those aged 30-75 y in NHANES II (4349 NHANES II). The vital status follow-up was conducted and all-cause mortality data were compiled in 1992 for both NHANES I and II, which consisted of 2464 and 1258 deaths in NHANES I and II, respectively. Survival time from baseline to date of death was calculated. Right-sided censoring was used for the 170 subjects lost to follow-up.


The surveys covered many aspects relevant to health, nutrition and living conditions as well as anthropometrics. The following covariates were selected for analyses in this study: baseline age, height, weight, triceps and subscapular skinfolds, arm circumference, smoking status, vital status, race and education level. Height, weight and anthropometric measurements using a standardized protocol were performed in both NHANES I and NHANES II.22,23 The smoking variable was defined as current, unknown and never smokers. Race was categorized as white, black and other. Education level was divided into three categories, ie attaining less than 8 y of education, 8-12 y, and more than 12 y. BMI was calculated as kg/m2. The operational definition of fat mass indicator (FMI) is the sum of the Z-score of triceps skinfolds thickness and the Z-score of subscapular skinfolds thickness. Z-score was defined as a deviation from the sample mean value in sample standard deviation units. The upper arm circumference was used as a fat-free mass indicator (FFMI). Although the upper arm circumference includes muscle, bone and adipose tissue, ie it is a composite of FM and FFM, in the presence of FM indicators in the same model, ie conditional on FM indicators, upper arm circumference should become largely representative of fat-free arm tissue. The choice of the prediction variables depends on the hypotheses to be tested.

Statistical analysis

We used Cox proportional hazards model for the survival analysis. Statistical comparisons of baseline characteristics of subjects between NHANES I and II were made by the chi2 test and Student's t-test. All analyses used SAS statistical software (SAS version 7, SAS Institute Inc., Cary, NC, USA). Two-tailed (alpha=0.05) tests of significance were used.

Outcome variable: Time from baseline to all-cause death and right censoring for 170 subjects with unknown vital status in NHANES I were used as the endpoint. Subjects dying early were not excluded for reasons discussed elsewhere.24,25,26

Prediction variables: BMI, FMI, FFMI, BMI2 and FMI2.

Main covariates: Age, cohort (NHANES I vs II) and smoking status. These covariates were always included in modeling.

Sensitivity covariates: Race and education level. These covariates were added to the main models below to assess the sensitivity of the results.

In terms of modeling we considered four main models¾models 1, 2, 3 and 4. In model 1, the prediction variables were BMI and BMI2 to test for the U-shaped relationships between BMI and all-cause mortality that have been previously observed. In model 2, the prediction variables were BMI, BMI2 and FFMI. This model was to test the hypotheses of decreasing (ie protective) effect of FFMI on all-cause mortality rate in the presence of BMI and BMI2. Model 3 tested the hypotheses of increasing (ie harmful) effect of FMI and decreasing (ie protective) effect of FFMI on all-cause mortality rate conditional on a given BMI. Model 4 tested the shape of the relationship between FMI and all-cause mortality by adding FMI2 into model 3. Sensitivity analyses were followed by adding the 'sensitivity covariates' above in four models.


Descriptive statistics

The baseline characteristics of subjects who participated in NHANES I and II are shown in Table 1. There was a statistically significant difference in age between the two cohorts. Weight, BMI, subscapular and triceps skinfolds were greater for participants in NHANES II compared to NHANES I; however there was no statistically significant difference in height or arm circumference. There were also statistically significant differences in mean survival time, smoking status, vital status (including number of all-cause deaths), race and educational levels between NHANES I and II.

The correlations among BMI, FMI and FFMI are shown in Table 2. The correlation coefficient was highest between BMI and FFMI and lowest between FMI and FFMI.

Primary analyses

Table 3 shows the hazard ratios (HRs) for all-cause mortality for each of the four Cox proportional models performed. In model 1, there was a significant U-shaped relationship between BMI and all-cause mortality after adjusting for age, smoking status and cohort, but without adjusting for FMI and FFMI. The estimated nadir of the BMI curve was 27.3 kg/m2. After adding FFMI into the model (model 2), the U-shaped relationship between BMI and all-cause mortality remained significant, but the nadir of BMI was reduced to 19.5 kg/m2 which corresponds to the 4.7th percentile in NHANES I and II combined. FFMI showed a significantly negative (ie protective) relationship with all-cause mortality rate (HR=0.928, 95% CI 0.911, 0.945, P<0.0001). Figure 1 illustrates these relationships by plotting the hazard ratios against an abscissa in Z-score units to standardize across different the metrics of the independent variables.

In model 3, after adjusting for both FMI and FFMI, the relationship between BMI and all-cause mortality and the nadir of BMI were similar to model 2. In addition, there was a positive relationship between FMI and all-cause mortality rate (HR=1.033, 95% CI (1.005, 1.063), P=0.0213), whereas the relationship between FFMI and all-cause mortality remained significantly negative (ie protective; HR=0.923, 95% CI (0.906, 0.941), P<0.0001). After adding FMI2 into model (model 4), the relationships of FMI, FFMI and BMI with all-cause mortality rate remained similar to the Cox model 3. Moreover, the relationship between FMI2 and all-cause mortality rate was not significant (HR=0.999, 95% CI (0.993, 1.005), P=0.7744), which indicated a monotonic increasing relationship between FMI and all-cause mortality rate. Current smokers showed a significantly higher rate of all-cause mortality when compared to non-smokers in all four models.

Sensitivity analyses

As shown in Table 4, after adding race and education levels to the models, the relationships of BMI, BMI2, FMI, FMI2, FFMI, age, smoking status and cohort with all-cause mortality rate did not change.

When we omitted subjects those with unknown smoking status, the results were virtually unchanged compared with those using all subjects (data not shown).

We also evaluated the potential effect of multicollinearity between BMI, FMI and FFMI on our results. Standard diagnostic tests did reveal modest multicollinearity. Although this might have yielded larger standard errors, the parameter estimates remain unbiased. Furthermore, given that the key estimates were generally statistically significant, the multicollinearity did not seem to eliminate our ability to draw reasonably confident conclusions from our results. This was confirmed with sensitivity analysis where two orthogonal principal component scores representing 'FM' and 'FFM' were added to the Cox models. These models yielded similar results: all-cause mortality decreased with FFMI and increased with FMI, suggesting multicollinearity contributed negligibly to these analyses.


In this study, we observed that the relationship between BMI and all-cause mortality was U-shaped with the nadir of BMI being approximately 27 kg/m2. However, after adjusting for FFMI alone or together with FMI, the nadir decreased to 19.5 and 20.5 kg/m2, respectively. Moreover, unlike BMI, FMI tended to be monotonically increasing with respect to mortality rate, whereas the association between all-cause mortality rate and FFMI was monotonically decreasing. These results are consistent with other studies,9 which have hypothesized that the consistently demonstrated U-shaped relationship between BMI and all-cause mortality may be the result of the inverse independent effect of FMI and FFMI on mortality. We have shown in these analyses using crude, surrogate indicators of FM and FFM that this hypothesis may be valid. Whether FM and FFM can be more precisely parceled out of BMI using more accurate and precise measurements of adiposity such as dual-energy X-ray absorptiometry (DEXA) or magnetic resonance requires further investigation. It should be noted that this study did not measure changes in FM and FFM indicators in relation to mortality, which addresses a different question. There is reason to suspect that changes in these respective indicators may have different associations with mortality. An analysis of the Tecumseh Community Health Study and Framingham Heart Study concluded that weight loss is associated with increased mortality rate and fat loss is associated with decreased mortality rate.27

The strengths of this present study are several-fold. Pooling these nationally representative samples enhances statistical power, the precision and external validity of the results. The high-quality standardization of the anthropometric measurements in these cohorts also minimizes measurement error and potential biases. Additionally, the simultaneous use of indicators of FM and FFM in the survival models allows for the potential assessment of each effect independent of the other.

Given the mean differences that exist between the two cohorts utilized in terms of baseline body mass indicators, it is reasonable to inquire how this might have affected results. Significant differences (in means) of variables between the two cohorts will not affect the estimated effects of the independent variables when data from the two cohorts are pooled together, if a term for cohort is included in the model as we did and there is no interaction between cohort and the independent variables. Although not reported in detail, those interactions were tested and found not to be statistically significant. Therefore, for this particular study, the significant differences in means do not seem to influence the results. Therefore, pooling these two nationally representative samples seems well justified.

The present study demonstrated a differential association of BMI and adiposity with all-cause mortality rate. These results, in conjunction with recent reports,13,14,15 should stimulate use of more rigorous body composition measurements in epidemiologic studies. Additionally, and perhaps most importantly, these studies may have implications for the development of clinical and public health guidelines pertaining to the weight and body composition associated with optimal health.28

The results of this study, however, should be interpreted taking into account several study limitations. Although the inclusion of skinfold measurements expands the underlying role of BMI on all-cause mortality, estimates of FM using these measurements are not optimal.29 Newer methods such as DEXA, for example, provide more accurate estimates of total body fat29,30 and would be preferable. However, these methods are relatively expensive and may be difficult to use in large-scale epidemiologic research. Until such studies are available, the anthropometric measurements may offer safe, inexpensive and practical proxies of subcutaneous fat.31

This study did not have measurements of regional adiposity. Waist circumference, for example, as a measure of visceral adiposity, has been shown to be more strongly associated with obesity-related comorbidities and mortality rate than subcutaneous adiposity.32,33 Additionally, intra-abdominal FM measured by magnetic resonance or computed tomography imaging techniques may more accurately measure visceral FM.34

Since this was an observational study, causal pathways underlying the observed associations cannot be inferred with confidence. We cannot rule out the possibility, for example, that underlying wasting disease partially accounted for the relationship between FFM and mortality rate. We believe our findings here are provocative and future studies should replicate these finding using more accurate body composition measurements. Future studies using NHANES III and IV data are expected to provide more precise answers to the research questions posed herein. NHANES III made use of bioimpedance analysis and, to our knowledge, NHANES IV will make use of DEXA. Finally, despite the practical challenges, studies making use of regional adiposity may yield novel insights into true body composition-mortality relationships.


We are grateful to David F Williamson for his helpful comments. This research was supported in part by NIH grants AG17166, DK51716 and DK26687, and unrestricted grants from Merck Pharmaceuticals, Tanita Incorporated Mead-Johnson and the Sugar Association.


1 Troiano RP, Frongillo EA Jr, Sobal J, Levitsky DA. The relationship of body weight and mortality: a quantitative analysis of combined information from existing studies. Int J Obes Relat Metab Disord 1996; 20: 63-75. MEDLINE

2 Allison DB, Saunders SE. Obesity in North America: an overview. Med Clin N Am 2000; 84: 305-332. MEDLINE

3 Manson JE, Stampfer MJ, Hennekens CH, Willett WC. Body weight and longevity: a reassessment. JAMA 1987; 257: 353-358. MEDLINE

4 Manson JE, Willett WC, Stampfer MJ, Colditz GA, Hunter DJ, Hankinson SE, Hennekens CH, Speizer FE. Body weight and mortality among women. New Engl J Med 1995; 333: 677-685. MEDLINE

5 Stevens J, Cai J, Pamuk ER, Williamson DF, Thun MJ, Wood JL. The effect of age on the association between body-mass index and mortality. New Engl J Med 1998; 338: 1-7. MEDLINE

6 The BMI in Diverse Populations Collaborative Group. Effect of smoking on the body mass index-mortality relation: empirical evidence from 15 studies. Am J Epidemiol 1999; 150: 1297-1308. MEDLINE

7 Goldstein DJ. Beneficial effects of modest weight loss. Int J Obes Relat Metab Disord 1992; 16: 397-415. MEDLINE

8 McCarter RJ. Role of caloric restriction in the prolongation of life. Clin Geriatr Med 1995; 11: 553-565. MEDLINE

9 Allison DB, Faith MS, Heo M, Kotler DP. Hypothesis concerning the U-shaped relation between body mass index and mortality. Am J Epidemiol 1997; 146: 339-349. MEDLINE

10 Comstock GW, Kendrick MA, Livesay VT. Subcutaneous fatness and mortality. Am J Epidemiol 1966; 83: 548-563. MEDLINE

11 Keys A, Taylor HL, Blackburn H, Brozek J, Anderson T, Simonson E. Mortality and coronary heart disease among men studied for 23 y. Arch Intern Med 1971; 128: 201-214. MEDLINE

12 Menotti A, Descovich GC, Lanti M, Spagnolo A, Dormi A, Seccareccia F. Indexes of obesity and all-causes mortality in Italian epidemiological data. Prev Med 1993; 22: 293-303. Article MEDLINE

13 Blair SN, Brodney S. Effects of physical inactivity and obesity on morbidity and mortality: current evidence and research issues. Med Sci Sports Exerc 1999; 31: (Suppl) 646-662. MEDLINE

14 Kalmijn S, Curb JD, Rodriguez BL, Yano K, Abbott RD. The association of body weight and anthropometry with mortality in elderly men: the Honolulu Heart Program. Int J Obes Relat Metab Disord 1999; 23: 395-402. MEDLINE

15 Heitmann BL, Erikson H, Ellsinger BM, Mikkelsen KL, Larsson B. Mortality associated with body fat, fat-free mass and body mass index among 60-year-old Swedish men¾a 22-year follow-up. The study of men born in 1913. Int J Obes Relat Metab Disord 2000; 24: 33-37. MEDLINE

16 Lee CD, Blair SN, Jackson AS. Cardiorespiratory fitness, body composition, and all-cause and cardiovascular disease mortality in men. Am J Clin Nutr 1999; 69: 373-380. MEDLINE

17 Goran MI, Allison DB, Poehlman ET. Issues relating to normalization of body fat content in men and women. Int J Obes Relat Metab Disord 1995; 19: 638-643. MEDLINE

18 Vague J. Les Obesities Etude Biometrique. Biol Med 1947; 36: 1.

19 Tuljapurkar S, Li N, Boe C. A universal pattern of mortality decline in the G7 countries. Nature 2000; 405: 789-792. Article MEDLINE

20 Cornoni-Huntley J, Harris T, Everett D, Albanes D, Micozzi MS, Miles TP, Feldman JJ. An overview of body weight of older persons, including the impact on mortality. The National Health and Nutrition Examination Survey I¾Epidemiologic Follow-up Study. J Clin Epidemiol 1991; 44: 743-753. MEDLINE

21 Freedman D, Williamson D, Croft J, Byers T. Relation of body fat distribution to ischemic heart disease. The NHANES I Epidemiologic Follow-up Study. Am J Epidemiol 1995; 142: 53-63. MEDLINE

22 Department of Health and Human Services NCHS. Plan and operation of the second National Health and Nutrition Examination Survey 1976-1980. Vital and Health Statistics Series 1, Vol 15 DHHS Publication No. (PHS) 81-1317. US Government Printing Office: Washington, DC, 1981.

23 Department of Health and Human Services NCHS. Plan and operation of the NHANES I Epidemiologic Followup Study, 1987. Vital and Health Statistics Series 1, Vol 27, DHHS publication no. (PHS). 92-1303, National Center for Health Statistics, 1992.

24 Allison DB, Heo M, Flanders DW, Faith MS, Williamson DF. Examination of 'early mortality exclusion' as an approach to control for confounding by occult disease in epidemiologic studies of mortality risk factors. Am J Epidemiol 1997; 146: 672-680. MEDLINE

25 Allison DB, Faith MS, Heo M, Townsend-Butterworth D, Wiliamson DF. Meta-analysis of the effect of excluding early deaths on the estimated relationship between body mass index and mortality. Obes Res 1999; 7: 342-354. MEDLINE

26 Allison DB, Heo M, Flanders DW, Faith MS, Carpenter KM, Williamson DF. Simulation study of the effects of excluding early deaths on risk factor-mortality analyses in the presence of confounding due to occult disease: the example of body mass index. Ann Epidemiol 1999; 9: 132-142. MEDLINE

27 Allison DB, Zannolli R, Faith MS, Heo M, Pietrobelli A, Vanltallie TB, Pi-Sunyer FX, Heymsfield SB. Weight loss increases and fat loss decreases all-cause mortality rate: results from two independent cohort studies. Int J Obes Relat Metab Disord 1999; 23: 603-611. MEDLINE

28 National Heart Lung and Blood Institute. Clinical guidelines for the identification, evaluation, and treatment of overweight and obesity in adults. The evidence report. Obes Res 1998; 6: (Suppl 2) 51-209.

29 Heymsfield SB, Wang ZM, Baumgartner RN, Ross R. Human body composition: advances in models and methods. A Rev Nutr 1997; 17: 527-558.

30 Heymsfield SB, Allison DB, Heshka S, Pierson RN Jr. Assessment of human body composition. In: Allison DB (ed). Handbook of assessment methods for eating behaviors and weight-related behaviors Sage: Thousand Oaks, CA, 1995, 515-560.

31 Norton K. Anthropometric estimation of body fat. In: Norton K, Olds T (eds.). Anthropometrica Sydney: UNSW Press, 1996, 172-198.

32 Couillard C, Gagnon J, Bergeron J, Leon AS, Rao DC, Skinner JS, Wilmore JH, Despres JP, Bouchard C. Contribution of body fatness and adipose tissue distribution to the age variation in plasma steroid hormone concentrations in men: the Heritage Family Study. J Clin Endocrinol Metab 2000; 85: 1026-1031. MEDLINE

33 Pi-Sunyer FX. Medical hazards of obesity. Ann Intern Med 1993; 119: 655-660. MEDLINE

34 Mitsiopoulos N, Baumgartner RN, Heymsfield SB, Lyons W, Gallagher D, Ross R. Cadaver validation of skeletal muscle measurement by magnetic resonance imaging and computerized tomography. J Appl Physiol 1998; 85: 115-122. MEDLINE


Figure 1 Hazard ratios (HRs) relative to Z-scores of zero. HR for BMI is based on model 1 in Table 3. HRs for both FFMI and FMI are based on model 4 in Table 3.


Table 1 Baseline characteristics of male participants of NHANES I (1971-1975) and NHANES II (1976-1980)

Table 2 Correlation among BMIa, FMIb and FFMIc

Table 3 Hazard ratios for all-cause mortality, male subjects aged 25-75, NHANES I and NHANES II combined

Table 4 Hazard ratios for all-cause mortality additionally adjusted for race and education levels, male subjects aged 25-75, NHANES I and NHANES II combined

Received 6 February 2001; revised 9 October 2001; accepted 18 October 2001
March 2002, Volume 26, Number 3, Pages 410-416
Table of contents    Previous  Article  Next    [PDF]