The aim of this longitudinal study is to examine the relationship between weight loss from maximum body weight, body mass index (BMI), and mortality in a nationally representative sample of men and women.
Longitudinal cohort study.
In all, 6117 whites, blacks, and Mexican-Americans 50 years and over at baseline who survived at least 3 years of follow-up, from the Third National Health and Nutrition Examination Survey Linked Mortality Files (1988–1994 with passive mortality follow-up through 2000), were included.
Measured body weight and self-reported maximum body weight obtained at baseline. Weight loss (maximum body weight minus baseline weight) was categorized as <5%, 5–<15%, and ⩾15%. Maximum BMI (reported maximum weight (kg)/measured baseline height (m)2) was categorized as healthy weight (18.5–24.9), overweight (25.0–29.9), and obese (⩾30.0).
In all, 1602 deaths were identified. After adjusting for age, race, smoking, health status, and preexisting illness, overweight men with weight loss of 15% or more, overweight women with weight loss of 5–<15%, and women in all BMI categories with weight loss of 15% or more were at increased risk of death from all causes compared with those in the same BMI category who lost <5%; hazard ratios ranged from 1.46 to 2.70. Weight loss of 5–<15% reduced risk of death from cardiovascular diseases among obese men.
Weight loss of 15% or more from maximum body weight is associated with increased risk of death from all causes among overweight men and among women regardless of maximum BMI.
One of the puzzles of obesity research is that numerous observational studies of the relationship between weight loss and mortality, including some based on large nationally representative samples, have found that the risk of dying increases, rather than decreases, with increasing weight loss.1, 2, 3, 4, 5, 6, 7, 8, 9, 10 This finding appears paradoxical given that obesity is associated with excess all-cause mortality (largely because of increased death rates for cardiovascular disease, cancer, and diabetes) and that the intentional loss of excess weight typically results in improved risk factor profiles for disorders associated with obesity including cardiovascular disease and diabetes.11, 12, 13, 14 Many of the observational studies have not distinguished between intentional and unintentional weight loss, a limitation that may have resulted in underestimation of the protective effect of weight loss because unintentional weight loss is often associated with unrecognized disease or health problems or with more severe disease. However, even studies that have focused on intent of weight loss have not produced consistent results.15, 16, 17, 18, 19, 20 Other criticisms of observational studies of weight loss include the failure of some to control for important confounding variables such as cigarette smoking, alcohol consumption, health status, and preexisting illness and disease, the inability to distinguish between loss of lean body mass and loss of adipose body mass, measurement of body weight at only one or two points in time, and lack of information about study participants' lifetime weight change trajectories. A number of clinical trials are underway, but it appears that even these have significant limitations.21, 22, 23 Two recent papers carefully examine the evidence for and against the assertion that ‘weight loss causes increased mortality’ and discuss possible biologic pathways as well as gaps in our knowledge of the causal pathways through which obesity operates.21, 24
Despite their shortcomings, observational studies continue to be an important tool for monitoring the association between weight loss and mortality in the general population. Given that the prevalence of overweight and obesity is increasing in the United States, that dieting (even among those who are not overweight) is common, and that the risk factor profiles of overweight and obese persons appear to be changing, examination of the associations between weight loss and mortality in recent population samples is desirable.25 The goal of this study is to contribute to this effort by examining the relationships among weight loss, body mass index (BMI), and mortality in the Third National Health and Nutrition Examination Survey (NHANES III) Linked Mortality File, which has a baseline period of 1988–1994 and mortality follow-up through 2000, a time period coincident with the recent increases in obesity in the United States.
Subjects and data collection
The NHANES III is an extensive interview and examination study of a representative sample of noninstitutionalized Americans aged 2 months and older that was conducted from 1988–1994. The study was approved by the appropriate institutional review authorities. Details of the survey methods have been described elsewhere.26 The National Center for Health Statistics has linked eligible NHANES III respondents aged 17 years and over at their interview to the National Death Index to obtain mortality information (derived from death certificates) for respondents who died in the United States during 1988–2000. Length of follow-up for survivors ranges from 6 to 12 years. This analysis uses the restricted-use NHANES III Linked Mortality File.27
The analyses were restricted to the 7327 non-Hispanic white, non-Hispanic black, and Mexican-American men and women who were 50 years and older at the NHANES III interview. Persons of other race-ethnicity groups were omitted because of the small number of respondents. Four respondents lacked some information needed for linkage and thus were ineligible for mortality follow-up. Excluded from all analyses were 252 respondents with missing values on the maximum weight, baseline weight, or height variables, six respondents whose maximum BMI was <18.5, and 192 respondents with missing values on baseline covariates. Respondents whose measured baseline weight was greater than their reported maximum weight by 10% or more (n=49) were also excluded. As weight loss could be the result of a serious illness that may not have been diagnosed, an additional 707 respondents who died within 3 years of their baseline exam were excluded. The final analytic sample included 6117 (2888 men and 3229 women).
Measured body weight was obtained at the NHANES III baseline examination (1988–1994). Self-report of maximum body weight was determined at baseline by asking participants ‘Up to the present time, what is the most you have ever weighed? (FEMALES): Do not include any times when you were pregnant.’ For respondents reporting a maximum weight less than their measured baseline weight, the measured baseline weight was considered their maximum weight. Percent weight loss was calculated as ((maximum weight−baseline weight)/maximum weight) × 100 and categorized as <5%, 5–<15%, and ⩾15%. Maximum BMI was calculated as reported maximum weight (kg)/measured baseline height (m)2 and categorized as healthy weight (18.5–24.9), overweight (25.0–29.9), and obese (⩾30.0).28
The coverage period for mortality data collection was from the date of the NHANES III examination through 31 December 2000. There were 1602 deaths (835 men and 767 women). Cardiovascular disease deaths include deaths with ICD-9 codes 390–459 and ICD-10 codes I00–I99. Noncardiovascular disease deaths include all other deaths.
Information on baseline age, race-ethnicity, cigarette smoking status, respondent-assessed health status, and history of chronic conditions was obtained at the NHANES III interview. Respondent-assessed health status was based on the question, ‘Would you say your health in general is excellent, very good, good, fair, or poor?’ Presence of preexisting health conditions related to excess weight (heart disease, stroke, diabetes, or hypertension) was based on self-report of a doctors’ diagnosis of the condition, any mention of recent blood pressure medication use, or elevated blood pressure measurement at the baseline exam (systolic ⩾140 mm Hg or diastolic ⩾90 mm Hg). Presence of preexisting health conditions related to weight loss (cancer, chronic bronchitis, or emphysema) was based on self-report of a doctor’s diagnosis of these conditions.
All analyses were carried out using Survey Data Analysis (SUDAAN), version 10.0. 0 (RTI International, Research Triangle Park, NC, USA) so that the sample weights and complex survey design could be incorporated.29 Differences in risk factor means or percentages among persons in the three BMI categories and the three weight loss categories were tested using the CROSSTABS and DESCRIPT procedures. Estimates of the hazard ratio of dying associated with percentage of weight lost were derived from Cox proportional hazards regression models with age as the time scale and left truncation on age at interview using the SURVIVAL procedure. Separate models were run for the six gender-maximum BMI (healthy weight, overweight, obese) strata. Although there was no statistically significant interaction between weight loss and gender, separate models were run for men and women because the earlier study of NHANES I Epidemiologic Follow-up Study (NHEFS) participants by Pamuk et al. found that the relationship between weight loss and mortality differed by gender.1, 4 Models were stratified by maximum BMI category so that the effect of weight loss could be assessed within each BMI category. Persons with little or no weight loss (<5%) served as the reference category. All models were adjusted for age at interview in single years and race-ethnicity (white, black, Mexican-American). Risk-adjusted models also included cigarette smoking status (current, former, and never), respondent-assessed health status (excellent, very good, or good; fair or poor), and history of health conditions related to excess weight and/or to weight loss. Three dummy variables were included in the models to adjust for preexisting health conditions: one variable for a history only of heart disease, stroke, hypertension, or diabetes, one variable for a history only of cancer, chronic bronchitis, or emphysema, and one variable for a history of conditions from both groups. To control for weight loss that may have resulted from serious illness not diagnosed at baseline, risk estimates were derived after excluding respondents who died within 3 years of the baseline examination. As exclusion of respondents who died within 3 years of baseline may not have adequately controlled for risk of death because of preexisting illness, the analyses of all-cause mortality risks were repeated after excluding persons in fair or poor health and after excluding persons with preexisting conditions (except hypertension). As the smoking status covariates included in the model may not have adequately adjusted for confounding, the analyses for women were repeated after excluding current and former smokers and those for men were repeated after excluding current smokers (small numbers did not allow exclusion of former smokers for men). Cause-specific hazard ratio estimates (cardiovascular and noncardiovascular) were not obtained for healthy weight persons because of small numbers of deaths. Tests of significance of trends across the three weight loss categories were conducted by treating the three weight loss levels as ordered categories. Physical activity and alcohol consumption were included as covariates in initial models but omitted from final models because they were not statistically significant and their omission did not affect the estimates of relative risk. The validity of the proportional hazards assumption was confirmed using both graphical and time-dependent covariate approaches.
Baseline characteristics of the men and women in the analytic cohort are shown in Table 1 and their mean weight, BMI, and weight loss are shown in Table 2 according to maximum BMI and percentage of weight lost. Regardless of BMI category, compared to persons with <5% weight loss, those with ⩾15% weight loss were more likely to be older and current smokers, to report poorer health status, and to have lower baseline BMI and weight. The baseline characteristics of persons with a weight loss of 5–<15% are similar to those of persons with a weight loss of <5%, except that those with a weight loss of 5–<15% have lower baseline BMI and weight than those with a weight loss of <5%. About 59% of men and women in the analytic cohort had lost 5% or more of their maximum weight at the baseline examination.
The hazard ratios of death from all causes, cardiovascular diseases, and noncardiovascular diseases associated with weight loss by level of maximum BMI among men and women are shown in Tables 3 and 4. For both men and women, risk of death generally was attenuated when models included the risk factors and preexisting illness indicator variables. Healthy weight men who lose 5% or more of their maximum weight do not have a statistically significantly elevated risk of dying compared with men who lose <5% of their maximum weight. Overweight men who lose 15% or more of their maximum weight have a statistically significantly increased risk of death from all causes; this results from a significantly increased risk of noncardiovascular disease death and an elevated risk of cardiovascular disease death. Obese men who lost 15% or more of their maximum weight have a statistically significantly elevated hazard ratio of dying from noncardiovascular disease. Obese men with a weight loss of 5–<15% have a statistically significantly reduced risk of dying from cardiovascular disease. Among women, those who were overweight and had a weight loss of 5–<15% have a statistically significantly elevated risk of death from all causes, which results from a significantly increased risk of cardiovascular disease death. Regardless of BMI category, women with weight loss of 15% or more have a statistically significantly increased risk of death from all causes. Obese women with a weight loss of 15% or more have a statistically significantly increased risk of death from both cardiovascular diseases and from noncardiovascular diseases; overweight women with weight loss of 15% or more have elevated risks of cardiovascular and noncardiovascular disease death that are not statistically significant.
The hazard ratio estimates for all-cause mortality obtained from the various subanalyses generally were consistent with those obtained from the primary analyses presented in Tables 3 and 4 (data not shown). The all-cause mortality hazard ratio estimates (and their significance levels) for men who were former or never smokers, for men who reported being in excellent, very good, or good health, and for men with no chronic conditions (except hypertension) generally were consistent with those shown in Table 3 for all men; however, the hazard ratios for overweight men with weight loss of 15% or more who were in excellent, very good, or good health or had no chronic conditions were not statistically significantly elevated, whereas the hazard ratio for obese men with weight loss of 15% or more who were in excellent, very good, or good health was statistically significantly elevated. The all-cause mortality hazard ratio estimates for healthy weight and obese women who never smoked, or who reported being in excellent, very good, or good health, or who had no chronic conditions (except hypertension) were similar or higher than those shown in Table 3 for all women; and as for all women, the risks for these subgroups of healthy weight and obese women who lost 15% or more of their maximum weight were statistically significantly elevated. However, though the hazard ratios for overweight women who never smoked or who had no chronic diseases (except hypertension) were similar to those observed for all women, those for overweight women in excellent, very good or good health were attenuated and not significantly elevated. Models of linear trend provide evidence of a dose–response relationship between weight loss and all-cause mortality for men and women whose maximum BMI was 25.0 or higher.
In this multi-ethnic national cohort of men and women, weight loss of 15% or more was associated with an increased risk of death from all causes among overweight men and among women regardless of maximum BMI. Weight loss of 15% or more also was associated with an increased risk of death from noncardiovascular diseases among overweight and obese men and from both cardiovascular and noncardiovascular diseases among obese women. Moderate weight loss (5–<15%) was associated with increased risk of death from all causes and from cardiovascular diseases among overweight women. Moderate weight loss reduced risk of death from cardiovascular diseases among obese men. Results from subanalyses of never or former smokers, respondents in excellent, very good, or good health, and respondents with no chronic conditions generally were consistent with those obtained from the primary analyses.
Our finding of increased mortality among respondents with extreme weight loss is consistent with the results of other observational studies. Of particular interest is the comparison of the results of this study with the results from previous studies using the NHEFS because both are large national samples and because the NHEFS baseline period of 1971–1975 and its follow-up through 1987 precedes the recent ‘obesity epidemic’, whereas the baseline period of the NHANES III (1988–1994) and its follow-up period through 2000 overlaps it.1, 4 Not surprisingly, the prevalence of overweight and obesity in the NHANES III sample is higher than in the NHEFS sample; however, the percentage that have lost 15% or more of their maximum body weight is lower. The hazard ratios reported from NHEFS tend to be somewhat higher than corresponding risks from our NHANES III study and more of them were statistically significant. Both our study and the NHEFS study found increased risks of death from all causes to be associated with weight losses of 15% or more among overweight men and among women in all three BMI categories and with weight losses of 5–<15% among overweight women. Our study and the NHEFS study also reported increased risk of death from cardiovascular disease for overweight women with weight losses of 5–<15% and obese women with weight losses of 15% or more and increased risks of death from noncardiovascular disease for overweight men with weight losses of 15% or more. Both studies also reported a significantly reduced risk of death from cardiovascular disease among obese men who lost 5–<15% of their maximum weight. It should be noted that the NHEFS may have greater power to detect elevated or reduced risk of death than NHANES III because the NHEFS has a longer follow-up period and hence, more deaths.
The reasons for the associations between weight loss and mortality reported from numerous observational studies are not well understood because of gaps in our knowledge of the causal pathways through which obesity operates. Arguments for and against the validity of the observed associations have been made.18, 21 The possibility exists that the associations reported from observational studies may be due, at least in part, to bias resulting from various limitations inherent in the observational studies. In this study, a limitation is that maximum weight was reported, not measured and though self-reported past weight is strongly correlated with past measured weight, recall of maximum weight has not been validated.30 Thus, maximum weight may be subject to recall bias, digit preference, and other sources of bias and error. Further, some women may have reported their pregnancy weight as their maximum weight. Additionally, no information was available about the age at which the maximum weight occurred or about how long the maximum weight was sustained. No information was available about whether weight loss was intentional or unintentional, how weight was lost, whether weight loss was due to loss of adipose or lean body mass, or about fluctuations in weight before and after the baseline examination. No information was available about weight gain so respondents who gained weight before baseline are included in the reference groups (as their maximum body weight is equal to their baseline weight). As some studies have found weight gain to be associated with increased mortality risk, the inclusion in the reference groups of respondents who gained weight may have biased the results toward the null.8, 10 As height decreases with age, some misclassification of maximum BMI may have occurred because of the use of baseline height in the calculation of maximum BMI. Although we were unable to distinguish between unintentional and intentional weight loss in our analyses, we attempted to control for unintentional weight loss by excluding respondents who died within 3 years of follow-up and by including indicator variables in the models for preexisting health conditions associated with weight loss and weight gain and for health status. These measures may not fully control for the effect of occult illness on mortality, though subanalyses that examined risks among never and former smokers, respondents in excellent, very good, or good health, and in respondents with no chronic diseases produced hazard ratios that generally were similar to those obtained from the main analyses for all respondents. We were unable to control for moderation of the association between weight loss and mortality because of intentional weight loss. Some types of intentional weight loss may be harmful; some types (such as weight loss because of improved physical fitness) may be beneficial. This study included only middle-aged and elderly adults; results may have differed if younger adults also had been included because bodyweight tends to decline in old age and because the likelihood that weight loss is unintentional increases with age and unintentional weight loss is associated with adverse traits.31, 32 Further limitations of this study include possible selection bias arising from missing data on baseline risk variables or other exclusions and measurement error, though this should be minimal given the standardized nature of the NHANES III examination.
The health effects of weight loss continue to be an important area of public health research given that the prevalence of overweight and obesity is increasing in the United States, that dieting is a common practice, and that physicians generally recommend weight loss for overweight persons who have associated medical conditions and for obese persons. Despite their limitations, observational studies have an important function in the continuing efforts to understand the associations between weight loss, BMI, and mortality. Updating earlier observational studies using more recent data bases is important because of the changing prevalence of the condition and apparent changes in the risk profiles for cardiovascular disease of the overweight and obese populations. The study reported here has the advantage of being based on a nationally representative sample that included a large number of deaths and of having a baseline period of 1988–1994 and mortality follow-up through 2000, a time period coincident with a portion of the ongoing obesity epidemic. The association between weight loss and subsequent mortality, after adjustment for various risk factors and after taking multiple steps to control for possible confounding because of preexisting illness, has now been reported across multiple time periods.
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention or the National Heart, Lung, and Blood Institute.
Pamuk ER, Williamson DF, Madans J, Serdula MK, Kleinman JC, Byers T . Weight loss and mortality in a national cohort of adults, 1971–1987. Am J Epidemiol 1992; 136: 686–697.
Williamson DF, Pamuk ER . The association between weight loss and increased longevity: a review of the evidence. Ann Intern Med 1993; 119: 731–736.
Andres R, Muller DC, Sorkin JD . Long-term effects of change in body weight on all-cause mortality: a review. Ann Intern Med 1993; 119: 737–743.
Pamuk ER, Williamson DF, Serdula MK, Madans J, Byers TE . Weight loss and subsequent death in a cohort of US adults. Ann Intern Med 1993; 119: 744–748.
Lee IM, Paffenbarger Jr RS . Is weight loss hazardous? Nutr Rev 1996; 54: S116–S124.
Mikkelsen KL, Heitmann BL, Keiding N, Sorensen TI . Independent effects of stable and changing body weight on total mortality. Epidemiology 1999; 10: 671–678.
Allison DB, Zannolli R, Faith MS, Heo M, Pietrobelli A, VanItallie TB et al. 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.
Wannamethee SG, Shaper AG, Walker M . Weight change, weight fluctuation, and mortality. Arch Intern Med 2002; 162: 2575–2580.
Breeze E, Clarke R, Shipley MJ, Marmot MG, Fletcher AE . Cause-specific mortality in old age in relation to body mass index in middle age and in old age: follow-up of the Whitehall cohort of male civil servants. Int J Epidemiol 2006; 35: 169–178.
Myrskyla M, Chang VW . Weight change, initial BMI, and mortality among middle- and older-aged adults. Epidemiology 2009; 20: 840–848.
Flegal KM, Graubard BI, Williamson DF, Gail MH . Excess deaths associated with underweight, overweight, and obesity. JAMA 2005; 293: 1861–1867.
National Task Force on the Prevention Treatment of Obesity. Overweight, obesity, and health risk. Arch Int Med 2000; 160: 898–904.
Pi-Sunyer FX . A review of long-term studies evaluating the efficacy of weight loss in ameliorating disorders associated with obesity. Clin Ther 1996; 18: 1006–1035.
Goldstein DJ . Beneficial health effects of modest weight loss. Int J Obes Relat Metab Disord 1992; 16: 397–415.
Yaari S, Goldbourt U . Voluntary and involuntary weight loss: associations with long term mortality in 9228 middle-aged adult and elderly men. Am J Epidemiol 1998; 148: 546–555.
Williamson DF, Pamuk E, Thun M, Flanders D, Byers T, Heath C . Prospective study of intentional weight loss and mortality in never-smoking overweight white US women aged 40–64 years. Am J Epidemiol 1995; 141: 1128–1141.
Williamson DF, Pamuk E, Thun M, Flanders D, Byers T, Heath C . Prospective study of intentional weight loss and mortality in overweight white men aged 40–64 years. Am J Epidemiol 1999; 149: 491–503.
Williamson DF, Thompson TJ, Thun M, Flanders D, Pamuk E, Byers T . Intentional weight loss and mortality among overweight individuals with diabetes. Diabetes Care 2000; 23: 1499–1504.
Gregg EW, Gerzoff RB, Thompson TJ, Williamson DF . Intentional weight loss and death in overweight and obese US adults 35 years of age and older. Ann Intern Med 2003; 138: 383–389.
Sorensen TI, Rissanen A, Korkeila M, Kaprio J . Intention to lose weight, weight changes, and 18-y mortality in overweight individuals without co-morbidities. PLoS Med 2005; 2: e171.
Sorensen TI . Weight loss causes increased mortality: pros. Obes Rev 2003; 4: 3–7.
Courcoulas AP, Flum DR . Filling in the gaps in bariatric surgical research. JAMA 2005; 294: 1957–1960. [Erratum, JAMA 2005; 294:2848].
Sjöström L, Narbro K, Sjöström CD, Karason K, Larsson B, Wedel H et al. Effects of bariatric surgery on mortality in Swedish obese subjects. N Eng J Med 2007; 357: 741–752.
Yang D, Fontaine KR, Wang C, Allison DB . Weight loss causes increased mortality: cons. Obes Rev 2003; 4: 9–16.
Gregg EW, Cheng YJ, Cadwell BL, Imperatore G, Williams DE, Flegal KM et al. Secular trends in cardiovascular disease risk factors according to body mass index in US adults. JAMA 2005; 293: 1868–1874.
National Center for Health Statistics. Plan and operation of the Third National Health and Nutrition Examination Survey, 1988–94. National Center for Health Statistics: Hyattsville, MD, 1994. Vital Health Stat 1 (32), 1994.
National Center for Health Statistics. The Third National Health and Nutrition Examination Survey (NHANES III) Linked Mortality File: Matching Methodology. http://www.cdc.gov/nchs/data/datalinkage/matching_methodology_nhanes3_final.pdf (Accessed on 21 May 2008).
Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults, the evidence report. Publisher: National Institutes of Health: Bethesda, MD, 1998. NIH Publication no. 98-4083. Available from: http://www.nhlbi.nih.gov/guidelines/obesity/ob_gdlns.htm.
Research Triangle Institute. SUDAAN Language Manual, Release 10.0. Research Triangle Institute: Research Triangle Park, NC, 2008.
Perry GS, Byers TE, Mokdad AH, Serdula MK, Williamson DF . The validity of self-reports of past body weights by US adults. Epidemiology 1995; 6: 61–66.
Newman AB, Yamez D, Harris T, Duxbury A, Enright PL, Fried LB . Weight change in old age and its association with mortality. J Am Geriatr Soc 2001; 49: 1309–1318.
Wannamathee SG, Shaper G, Whincup PH, Walker M . Characteristics of older men who lose weight intentionally or unintentionally. Am J Epidemiol 2000; 151: 667–675.
We thank Elsie Pamuk and Katherine Flegal for their insightful comments and suggestions during preparation of this manuscript.
The authors declare no conflict of interest.
About this article
Cite this article
Ingram, D., Mussolino, M. Weight loss from maximum body weight and mortality: the Third National Health and Nutrition Examination Survey Linked Mortality File. Int J Obes 34, 1044–1050 (2010). https://doi.org/10.1038/ijo.2010.41
- follow-up studies
- longitudinal studies
- proportional hazards models
Bulletin of the National Research Centre (2019)
Trajectories of body mass index before the diagnosis of cardiovascular disease: a latent class trajectory analysis
European Journal of Epidemiology (2016)
Current Diabetes Reports (2014)
Monash Bioethics Review (2014)