American Journal of Epidemiology Vol. 152, No. 4 : 324-333
Copyright © 2000 by The Johns Hopkins University School of Hygiene and Public Health
ORIGINAL CONTRIBUTIONS |
Associations of Weight Change and Weight Variability with Cardiovascular and All-Cause Mortality in the Chicago Western Electric Company Study
From the Department of Preventive Medicine, Northwestern University Medical School, Chicago, IL.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Few studies of associations between weight loss or weight fluctuations and mortality have been sufficiently long term to permit exclusion of early deaths for a portion of follow-up long enough to eliminate likely effects of illness-related weight loss. This study examined associations of the variation (standard deviation and standard deviation about the trend (slope)) and trend (weight loss or weight gain) in body mass index (weight (kg)/height (m2) between 1958 and 1966 (minimum of five measurements) with subsequent 25-year mortality among 1,281 men originally aged 40–56 years from the Chicago Western Electric Company Study. In multivariate Cox regression models that included two slope variables representing weight loss and weight gain and each variability measure separately, weight loss and weight gain were significantly related to 15-year mortality but weight variability was not. Relative risks for cardiovascular disease mortality were 1.25 (95% confidence interval (CI): 1.09, 1.45) and 1.14 (95% CI: 0.97, 1.33), respectively, for weight loss and weight gain slopes larger by 0.12 kg/m2 per year; corresponding relative risks for all-cause mortality were 1.23 (95% CI: 1.10, 1.38) and 1.15 (95% CI: 1.03, 1.29), respectively. For follow-up years 16–25, none of these weight variables were significantly related to mortality. These results indicate that an association between weight loss and mortality may not persist beyond 15 years, and that weight variability may not be related to mortality independently of weight loss or weight gain. Am J Epidemiol 2000;152:324–33.
body mass index; body weight; body weight changes; cardiovascular diseases; mortality; weight gain; weight loss
Abbreviations: BMI, body mass index; CI, confidence interval; ICD-8, International Classification of Diseases, Eighth Revision.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Associations between weight fluctuation/variability and increased mortality from cardiovascular diseases and from all causes have been observed in several studies (1
–5). However, two studies did not find increased mortality with greater weight variability (6, 7). Increased risk with weight loss or an inverse association with weight change has also been found in many studies (2–5, 7–24), although several studies have shown no increased risk with weight loss and/or increased risk associated with weight gain or weight change (9, 10, 12, 24–32). However, it is unclear whether weight loss or weight variability is causally related to increased mortality, since underlying causes of weight loss or weight change are usually not determined in prospective epidemiologic studies. Thus, it is generally unknown whether weight changes are intentional or unintentional and whether changes in weight are an antecedent of illness or a consequence of it.
One approach used to control for unintentional illness-related weight loss is exclusion of early deaths (2, 4–6, 16–19, 21–23, 33). However, the length of follow-up for which early deaths have been excluded has exceeded 5 years in only one study (22). In that study, which excluded deaths occurring during the first 8 years, there was still an association between weight loss and increased cardiovascular disease mortality in some subgroups. In a previous report from the Chicago Western Electric Company Study on associations of body mass index (BMI) and other measures of adiposity with coronary heart disease mortality (34), we found U-shaped relations with mortality through the first 14 years of follow-up for each measure of adiposity. However, for follow-up years 15–22, mortality was progressively higher with greater baseline adiposity. Hence, exclusion of deaths occurring in the first few years of follow-up may not be adequate to remove effects of illness on weight.
Few studies of weight change and/or weight variability have reported on associations with mortality for 
20 years of follow-up (1, 6, 16, 27). Hence, there is a paucity of data on the long term associations of weight change with mortality.
In this study, we examined associations of variations and trends in BMI over 8 years with subsequent 25-year mortality from cardiovascular causes and all causes in the Chicago Western Electric Company Study. Because weight loss and weight variability could be consequences of preexisting disease, rather than antecedents, we also examined associations separately for deaths occurring in the first 15 years of follow-up and deaths occurring in years 16–25.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Study population
Participants in the Chicago Western Electric Company Study were randomly sampled from the population of 5,397 men who had been employed at the company's Hawthorne works in Chicago, Illinois, for at least 2 years and who were aged 40–55 years in 1957. Sampling procedures have been reported elsewhere (35
). Of 3,102 randomly selected men who were invited to participate, 2,080 agreed, and together with a pilot group of 27 men, they were initially examined between October 1957 and December 1958. Those men who continued to participate in the study were reexamined annually through 1966. Data used in these analyses were taken from the baseline examination and the first seven or eight annual reexaminations. Of the 2,107 men initially examined, 826 were excluded from these analyses for the following reasons: they had coronary heart disease (n = 44) or cancer (n = 13) at baseline; they did not undergo examination in 1965 or 1966 (n = 354); they experienced incident coronary heart disease (n = 144) or cancer (n = 19) between baseline and 1965–1966; they had no weight measurement for both 1965 and 1966 (n = 3); they had fewer than five weight measurements from baseline to 1965–1966 (n = 15); or they changed their smoking status (from nonsmoker to smoker (n = 26) or from smoker to nonsmoker (n = 208)) between 1958 and 1965–1966. We excluded men with smoking status changes because such changes could confound associations of BMI variables with mortality (36–38). For example, because of the impact of smoking cessation on subsequent risk, weight gain following smoking cessation might have had a different association with mortality than weight gain in someone who continued to smoke. These exclusions left 1,281 men aged 48–66 years in 1965–1966 who were available for these analyses.
Variables
BMI (weight (kg)/height (m)2) was computed for each weight measurement. The slope of the regression line relating each man's BMI values to the time from baseline was computed to assess his yearly rate of change. Two additional slope variables representing weight loss and weight gain, respectively, were also computed to account for nonlinear associations between slope and mortality. The weight loss slope was defined as the computed slope if it was less than or equal to zero, and zero otherwise. The weight gain slope was defined as the computed slope when it was greater than zero, and zero otherwise. The standard deviation of BMI and the standard deviation about the regression, which assesses variation about the trend in BMI over the period, were computed as measures of weight fluctuation or variation. The average of all body mass measurements between 1958 and 1965–1966 was used to assess overall level of BMI. Other variables included in the analyses were age at the time of the last weight measurement in 1965 or 1966 and number of cigarettes smoked per day, which was the average of all values reported between baseline and 1965–1966.
Vital status was determined periodically by mailed questionnaire and telephone interview through the 25th anniversary of the initial examination, when vital status was known for all 2,107 men. Subsequently, vital status was ascertained through the National Death Index. Death certificates were obtained for all decedents and coded with regard to underlying cause of death. These analyses utilized mortality follow-up through December 31, 1990, which corresponds to an average of 25 years from the examination in 1965 or 1966. Underlying cause of death was classified according to the International Classification of Diseases, Eighth Revision (ICD-8), adapted for use in the United States (39). The mortality endpoints were death from all causes and death due to cardiovascular diseases (ICD-8 codes 400.0–444.9).
Statistical procedures
Cox proportional hazards regression (40) was used to assess associations of BMI variability and trends with mortality. We first divided each variability measure into quintiles. To categorize slope, we divided the nonpositive slopes at the median and the positive slopes into tertiles, again creating five categories. Relative risks were then computed for each category or quintile, with adjustment for age and number of cigarettes smoked per day. The purpose of these analyses was to assess the shapes of the relations of these variables with mortality.
We next included each measure of variability or the two slope variables as continuous variables in Cox models, along with age and cigarettes/day. We then fitted models that included the two slope variables and either the standard deviation of BMI or the standard deviation about the regression, along with age, cigarettes/day, and average BMI. We did not adjust for blood pressure or total cholesterol in these analyses, because the associations between adiposity measures and mortality may be partially mediated by risk factors related to body size and obesity (33).
Because weight loss and weight variability could be a consequence of preexisting disease rather than an antecedent, we examined associations with 25-year mortality and then with deaths occurring in the first 15 years of follow-up and deaths occurring during years 16–25 of follow-up. We did this to determine whether associations were similar across these two follow-up periods, and thus whether associations might reflect effects of illness on weight rather than a true effect on mortality of weight change or weight variability. This division of the follow-up period was based on a previous report which suggested that it may be necessary to exclude deaths for as many as 15 years before a positive association between adiposity and mortality emerges (34).
We also compared associations of these measures with mortality for 685 men aged 48–55 years and 596 men aged 56–66 years and for 644 nonsmokers and 637 smokers to assess the consistency of findings by age and smoking status.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Descriptive statistics
Between 1958 and 1965–1966, the average BMI for these men was 25.8 (table 1), and the average within-person standard deviation was 0.75, which was slightly larger than the average within-person standard deviation about the regression. Among men who lost weight, the average BMI decrease was 0.13 kg/m2 per year, while among men who gained weight, the increase averaged 0.15 kg/m2 per year.
|
Interrelations among BMI measures
The correlation between the slope and the standard deviation of BMI was -0.77 for men with negative slopes and 0.75 for men with positive slopes, but it was only 0.08 overall (table 2). These high correlations in the subgroups with positive or negative slopes made it difficult to separate associations with mortality for the standard deviation and the weight loss and weight gain slopes. The overall correlation between the slope of BMI and the standard deviation about the regression was –0.13, with correlations of –0.42 and 0.24 for men with negative and positive slopes, respectively.
|
BMI variability and mortality
Tables 3 and 4 give age- and smoking-adjusted relative risks of all-cause and cardiovascular disease mortality by quintile for both measures of variability (standard deviation of BMI and standard deviation about the regression) for all 25 years of follow-up, the first 15 years of follow-up, and years 16–25 of follow-up, with the second quintile being used as the referent category. The tables also give the relative risks for a difference of +1 standard deviation for each measure from Cox models that included each measure as a continuous variable, plus age and cigarettes/day.
|
|
For 25-year all-cause mortality, there were increases in risk in the fifth quintile for both measures of variability, with relative risks for the first four quintiles generally being close to 1.0. For 25-year cardiovascular disease mortality, relative risks were also generally close to 1.0 for the first four quintiles but showed greater increases in risk in the fifth quintile, compared with results for all-cause mortality. Both measures were significantly related to each endpoint in the continuous variable Cox models, with relative risks for a one-standard-deviation difference ranging from 1.09 to 1.20.
For the first 15 years of follow-up, risk generally increased progressively from the first quintile through the fifth quintile for both endpoints for the standard deviation about the regression (table 4). This measure was significantly related to cardiovascular disease mortality for both years 1–15 and years 16–25, with little or no difference in relative risks being seen for the two follow-up periods for either endpoint. For the standard deviation of BMI (table 3), the association with 15-year cardiovascular disease mortality was J-shaped, and the risk of all-cause mortality generally increased progressively as the standard deviation increased. This measure was significantly related to both endpoints during the first 15 years of follow-up and to cardiovascular mortality in years 16–25. There was no clear pattern for the association with all-cause mortality for years 16–25.
Trends in BMI and mortality
Table 5 gives age- and smoking-adjusted relative risks by category for the slope of BMI, as well as the relative risks for the two slope variables entered as continuous variables in Cox models. In these analyses, the referent category was the first tertile for men with positive slopes. In the continuous variable Cox models, the relative risks were calculated for a difference of +1 standard deviation in the weight gain slope and a difference of -1 standard deviation in the weight loss slope (a larger weight loss).
|
Relative risks for both endpoints for 25- and 15-year mortality were generally highest among men who had the greatest decreases in BMI from 1958 to 1965–1966. However, for follow-up years 16–25, relative risks for both endpoints were less than 1.0 or close to 1.0 for men who lost weight. In the continuous variable Cox models, the weight loss slope was significantly related to both endpoints for 25- and 15-year follow-up, with relative risks ranging from 1.13 to 1.30 for a slope more negative by 0.12 kg/m2 per year. The weight loss slope was not significantly related to either endpoint for years 16–25. Regression coefficients for all-cause mortality differed significantly for years 1–15 and years 16–25 (p < 0.05).
For men who gained weight, relative risks for 25-year all-cause mortality differed little across positive slope tertiles, while the relative risk for cardiovascular disease mortality was increased for men in the third tertile. For 15-year follow-up, relative risks for both endpoints were largest among men who had the greatest increases in BMI, with values similar to those observed for the men who lost the most weight. However, in years 16–25, men with increases in the top tertile had relative risks for both endpoints that were less than 1.0. While the weight gain slope showed a positive association with each endpoint for both 25- and 15-year follow-up, only the association with 15-year all-cause mortality was significant. Regression coefficients for the weight gain slope also differed significantly between years 1–15 and years 16–25 for all-cause mortality (p < 0.05). Regression coefficients did not differ significantly between men aged 48–55 years and men aged 56–66 years or between smokers and nonsmokers for the two slope variables or the standard deviation of BMI for either endpoint in any follow-up period. For example, in the two age groups, with 15-year follow-up, relative risks for a slope more negative by 0.12 kg/m2 per year were 1.23 (95 percent confidence interval (CI): 0.96, 1.57) and 1.26 (95 percent CI: 1.11, 1.41), respectively, for all-cause mortality and 1.47 (95 percent CI: 1.07, 2.01) and 1.27 (95 percent CI: 1.09, 1.48), respectively, for cardiovascular disease mortality. In smokers and nonsmokers, respectively, relative risks were 1.28 (95 percent CI: 1.13, 1.45) and 1.28 (95 percent CI: 1.03, 1.59) for all-cause mortality and 1.35 (95 percent CI: 1.16, 1.57) and 1.30 (95 percent CI: 0.99, 1.72) for cardiovascular disease mortality. While regression coefficients for the weight loss slope did not differ significantly by age in follow-up years 16–25 for either endpoint, the relative risk for cardiovascular disease mortality among men aged 48–55 years was close to 1.0, i.e., 0.98 (95 percent CI: 0.67, 1.45), while the relative risk among men aged 56–66 years was 1.19 (95 percent CI: 1.02, 1.40), a value significantly different from 1.0. The regression coefficient for the standard deviation about the regression for 16- to 25-year cardiovascular disease mortality was significantly larger in men aged 48–55 years than in men aged 56–66 years, with relative risks for a difference of 0.34 kg/m2 of 1.42 (95 percent CI: 1.19, 1.71) and 1.03 (95 percent CI: 0.86, 1.25), respectively. The regression coefficients for this variable did not differ significantly for either endpoint for 25- or 15-year follow-up.
Multivariable associations with mortality
Table 6 gives relative risks for associations of BMI variables with mortality for each follow-up period from Cox models that included either the standard deviation about the regression (model 1) or the standard deviation of BMI (model 2), the two slope variables, and average BMI, age, and cigarettes/day. The weight loss slope was significantly related to both endpoints for both 15- and 25-year follow-up in the analyses that included the standard deviation about the regression. However, this variable was significantly related only to 15-year mortality if the standard deviation of BMI was included instead. The weight gain slope was also significantly related to all-cause mortality in 15-year follow-up in model 1. Regression coefficients for both slope variables differed significantly for years 1–15 and years 16–25 for all-cause mortality in both model 1 and model 2 (p < 0.05).
|
In these analyses, neither measure of variability was significantly related to either endpoint in any follow-up period. However, if slope was entered into the model as a single variable instead of two variables representing weight loss and weight gain, the standard deviation of BMI was significantly related to 25-year cardiovascular disease mortality, with a relative risk of 1.13 (95 percent CI: 1.03, 1.24), and to 15-year all-cause and cardiovascular disease mortality, with relative risks of 1.14 (95 percent CI: 1.04, 1.26) and 1.19 (95 percent CI: 1.05, 1.34), respectively, while the standard deviation about the regression was significantly related to 25-year cardiovascular mortality, with a relative risk of 1.10 (95 percent CI: 1.00, 1.21).
Table 7 gives multivariable model results for the standard deviation about the regression and the two slope variables for 15-year and 16- to 25-year follow-up among men aged 48–55 and 56–66 years and among smokers and nonsmokers. Associations were adjusted for age and average BMI, as well as cigarettes/day for the age subgroups. Regression coefficients for each variable did not differ significantly by age for 15-year follow-up or by smoking status for either follow-up period. For years 16–25, the regression coefficients for the standard deviation about the regression differed significantly by age for both endpoints, and the coefficients for the weight loss slope differed by age for cardiovascular disease mortality. In these analyses, mortality increased with increasing variability for the younger men and with greater weight loss in the older men. While the standard deviation about the regression was significantly related to both endpoints in the younger men for this follow-up period, the weight loss slope was not significantly related to either endpoint in the older men. The standard deviation of BMI was also significantly related to both endpoints for years 16–25 in comparable multivariable models in the younger men (data not shown).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
In these analyses of associations of variations and trends in BMI over the period 1958–1966 with subsequent 25-year mortality from all causes and cardiovascular diseases in the Chicago Western Electric Company Study, negative and positive trends in BMI, reflecting weight loss or weight gain between 1958 and 1965–1966, showed significant independent associations with mortality for all 1,281 men. However, the association between weight gain and mortality was present only in the first 15 years of follow-up, and any association between weight loss and mortality in years 16–25 was present only among men aged 56–66 years. While both the standard deviation of BMI and the standard deviation about the regression or trend of BMI over the period had significant associations with mortality in each follow-up period when examined separately in models that included only age and cigarettes/day, they were not significantly related in models that included slope variables representing weight loss and weight gain for all 1,281 men. However, both measures of variability had significant independent associations with 16- to 25-year mortality in the men aged 48–55 years. Associations with mortality did not differ significantly between smokers and nonsmokers in any period of follow-up.
Several studies have found weight variability or weight fluctuation to be associated with increased risk for all-cause mortality and cardiovascular disease morbidity or mortality (1–5). In a previous report from the Chicago Western Electric Company Study (1), men who had large fluctuations in weight between the ages of 20 and 40 years, where fluctuation was defined as a loss of at least 10 percent of body weight in one 5-year period and a gain of at least 10 percent in another 5-year period, had twice the 25-year risk of coronary heart disease mortality as men who reported no substantial change in weight. In the Framingham Heart Study (2), the coefficient of variation of BMI, defined on the basis of recalled weight at age 25 and weight at the first eight examinations, was significantly related to 18-year mortality from all causes and coronary heart disease in both men and women, with relative risks ranging from 1.27 to 1.93 for men or women in the highest tertile compared with those in the lowest. In these analyses, deaths occurring in the first 4 years of follow-up were excluded. In the Multiple Risk Factor Intervention Trial (3), the standard deviation of weight calculated from weights measured during clinic visits over a 6- to 7-year period was significantly related to 3.8-year mortality from all causes and cardiovascular diseases, with relative risks of 1.64 and 1.85, respectively, for men in the fourth quartile compared with those in the first. In the Gothenburg prospective studies of men and women (4), the coefficient of variation of BMI, based on three weights measured over a 10- or 11-year period, was significantly related to 15-year mortality from all causes and coronary heart disease in men and 13-year all-cause mortality in women. In the Honolulu Heart Study (5), the standard deviation about the regression of BMI, based on three weights measured over a 6-year period, was significantly related to 14.5-year mortality from all causes, cardiovascular diseases, and causes other than cardiovascular disease or cancer, with relative risks of 1.25, 1.41, and 1.53, respectively, for men in the fifth quintile compared with those in the first. In these analyses, deaths occurring in the first 5 years of follow-up were excluded.
In contrast to these findings, in the Charleston Heart Study (6) the coefficient of variation of BMI, defined on the basis of weight at age 25 and weights in 1960 and 1963, was not significantly related to 23-year all-cause mortality in Black men, White men, Black women, or White women. In these analyses, deaths occurring in the first 2 years of follow-up were excluded. In the Baltimore Longitudinal Study of Aging (7), the standard deviation about the regression of BMI, based on three weight measurements, was not significantly related to 16.5-year all-cause or coronary heart disease mortality in men.
The rationale for using the standard deviation about the regression as the measure of weight variability rather than the coefficient of variation or standard deviation in this study and in two other studies (5, 7) was to remove from the measure of weight variability that portion of variability due to systematic gain or loss in weight. In this study, the correlation between the standard deviation of BMI and the slope was 0.08 when the direction of the trend in weight was ignored. However, the correlation was –0.77 in men who lost weight (negative slope) over the period and 0.75 in men who gained weight (positive slope). These high correlations with weight loss and weight gain make it difficult to assess whether an association of variability with increased mortality is a true association or rather reflects associations of weight loss and/or weight gain with mortality. The use of the standard deviation about the regression did not entirely eliminate the problem of distinguishing between variability and the trend in weight over time, since the correlation with the slope was –0.42 in men who lost weight and 0.24 in men who gained weight. While these values are substantially smaller than those for the standard deviation of BMI, they nonetheless indicate some overlap between this measure of variability and the trend in weight.
In the Honolulu Heart Study (5), the standard deviation about the regression was related to all-cause and cardiovascular disease mortality in a multivariate model that included the trend in weight. In the present analyses, the standard deviation of BMI was significantly related to 25-year cardiovascular mortality and to 15-year all-cause and cardiovascular disease mortality if the slope was included as a single continuous variable in the model. However, with inclusion of slope variables that took into account the direction of the trend, neither measure of variability was significantly related to either endpoint.
In the Framingham Heart Study (2), the slope of BMI had a significant inverse association with mortality from all causes and coronary heart disease in both men and women in models that included the average and coefficient of variation of BMI. Similarly, in the Gothenburg studies (4), change in BMI was also inversely related to all-cause mortality in men and women. Among men in the Multiple Risk Factor Intervention Trial (3), the trend in weight over the 6- to 7-year study period was not included as a covariate in the analyses. It is unclear whether the results of these three studies would have been altered if the analytical models had included terms taking into account both positive and negative trends in weight. However, it seems likely that the strength of the associations of the variability of BMI with mortality would have been reduced in such analyses. It is also unclear how such an analytical approach might have affected the results from the Honolulu Heart Study (5), since only a single variable for the slope appears to have been included in the analyses involving variation in BMI. In that study, weight loss was also associated with increased mortality.
Increased risk of mortality with weight loss or an inverse association with weight change has been found in many studies (2–5, 7–24), although a number of studies have also shown no increased risk with weight loss and/or increased risk associated with weight gain or weight change (9, 10, 12, –32). Studies of weight change can generally be divided into four groups based on the methodology used to define weight change or the time period over which weight change is measured. These four groups include studies which calculate weight change as: 1) the difference in weight between two study examinations (3–5, 8, 12, 14–17, 19, 23–25); 2) the difference between current weight and a recalled weight in young adulthood, e.g., age 18 or age 25 (1, 9–11, 13, 24, 26–32); 3) the difference between current weight and highest lifetime weight (12, 18, 20–22); and 4) the trend of weight over a series of examinations (2, 7).
The associations of weight change with mortality vary across these different types of studies. Studies that have examined the association using the difference between current weight and the highest lifetime weight have consistently shown weight loss to be associated with increased risk of mortality from both all causes (12, 18, 21, 22) and cardiovascular diseases (18, 2021–22). In contrast to these findings, only a handful of the studies defining weight change as the difference between current weight and weight during young adulthood have shown increased risk with weight loss (9–11, 13). Most of these studies have shown no association between weight loss and mortality (30, 31) or a positive association between weight gain or change and mortality (24, 27–32). Furthermore, two of the studies showing increased risk with weight loss also found weight gain to be associated with increased risk (9, 10). Studies defining weight change as the difference in weight between two examinations are relatively consistent in showing weight loss to be associated with increased risk (3–5, 8, 14–17, 19, 23, 24). Only a few of these studies have shown weight gain to be associated with increased risk (15, 17, 24, 25). The two studies which examined an association between trends in weight and mortality showed inverse associations (2, 7), although only the results in the Framingham Heart Study (2) were statistically significant. In the present study, a negative trend in weight was associated with higher mortality. However, in follow-up years 16–25 this association was present only for cardiovascular disease mortality in the men aged 56–66 years.
It is unclear whether weight loss is causally related to increased mortality, however, since underlying reasons for weight loss or weight change are usually not determined in these types of studies. Therefore, it is generally unknown whether weight changes are intentional or unintentional, and thus whether changes in weight are an antecedent of illness or a consequence of it. Only three studies have divided weight loss into categories of intentional and unintentional loss (23, 41, 42). In the first of these studies (23), intentional weight loss among overweight women aged 40–64 years with obesity-related conditions was generally associated with decreased mortality. However, among women with no preexisting illness, the association was equivocal. In the Iowa Women's Health Study, a study of women aged 50–69 years (41), having one or more intentional weight loss episodes of 20 pounds (9.1 kg) was not significantly associated with higher total or cardiovascular disease mortality risk in comparison with never losing 20 pounds. However, having one or more unintentional weight loss episodes was associated with a 26–57 percent higher total mortality risk and a 51–114 percent higher cardiovascular disease mortality risk in comparison with never losing 20 pounds. In the third study (42), among overweight men aged 40–64 years, neither intentional nor unintentional weight loss of 20 pounds was associated with total or cardiovascular disease mortality, either in those with existing health conditions or in those without them.
Differences in the results of the various weight change studies may also be attributable to the influences of illness-related weight loss. In particular, weight change over a relatively short period of time, e.g., between two examinations, is more likely to be influenced by illness-related weight loss than is weight change since young adulthood. Similarly, change in weight from the highest lifetime weight would also appear to be more likely to be influenced by illness-related weight loss. Hence, it is perhaps not surprising that studies using weight change since young adulthood are the studies showing increased risk with weight gain and little or no increased risk with weight loss, whereas the other types of studies show an increased risk associated with weight loss.
One approach used to control for illness-related weight loss is exclusion of early deaths (2, 4–6, 16–19, 21–23, 33). However, the length of follow-up for which early deaths have been excluded has exceeded 5 years in only one study (22), and in that study, which excluded deaths occurring in the first 8 years, there was still an association between weight loss and increased cardiovascular disease mortality among men and women with body mass indices of 26–29. In a previous report from the present study (34), we found U-shaped relations with coronary heart disease mortality through the first 14 years of follow-up for BMI and other measures of adiposity. However, for years 15–22, mortality was progressively higher with higher baseline adiposity. In the present analyses, weight loss was not associated with increased risk in years 16–25 among men aged 48–55. These results suggest that the effects of illness-related weight loss may persist far beyond the exclusion periods used in most epidemiologic studies.
In conclusion, these results indicate that the increased risk of mortality associated with weight loss and/or weight fluctuation/variability which is observed in many studies may not persist beyond 15 years, and that weight variability may not be related to mortality independently of weight loss and weight gain.
|
ACKNOWLEDGMENTS |
|---|
This work was supported by grant HL21010 from the National Heart, Lung, and Blood Institute.
The authors thank the many physicians who participated in the annual examinations of this cohort (35). In addition, they thank Dan Garside for his technical assistance with the study data.
|
NOTES |
|---|
Reprint requests to Dr. Alan R. Dyer, Department of Preventive Medicine, Northwestern University Medical School, 680 North Lake Shore Drive, Suite 1102, Chicago, IL 60611-4402 (e-mail: adyer{at}nwu.edu).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
-
Hamm P, Shekelle RB, Stamler J. Large fluctuations in body weight during young adulthood and twenty-five-year risk of coronary death in men. Am J Epidemiol 1989;129:312–18.
[Abstract/Free Full Text] - Lissner L, Odell PM, D'Agostino RB, et al. Variability of body weight and health outcomes in the Framingham population. N Engl J Med 1991;324:1839–44.[Abstract]
-
Blair SN, Shaten J, Brownell K, et al. Body weight change, all cause-mortality, and cause-specific mortality in the Multiple Risk Factor Intervention Trial. Ann Intern Med 1993;119:749–57.
[Abstract/Free Full Text] - Lissner L, Bengtsson C, Lapidus L, et al. Body weight variability and mortality in the Gothenburg prospective studies of men and women. In: Bjorntorp P, Rossner S, eds. Obesity in Europe '88: proceedings of the First European Congress on Obesity. London, United Kingdom: John Libbey and Company Ltd, 1989:55–60.
-
Iribarren C, Sharp DS, Burchfiel CM, et al. Association of weight loss and weight fluctuation with mortality among Japanese American men. N Engl J Med 1995;333:686–92.
[Abstract/Free Full Text] - Stevens J, Lissner L. Body weight variability and mortality in the Charleston Heart Study. (Letter). Int J Obes 1990;14:385–6.[Web of Science][Medline]
- Lissner L, Andres R, Muller DC, et al. Body weight variability in men: metabolic rate, health and longevity. Int J Obes 1990;14:373–83.[Web of Science][Medline]
- Hammond EC, Garfinkel L. Coronary heart disease, stroke, and aortic aneurysm. Arch Environ Health 1969;19:167–82.[Web of Science]
- Avons P, Ducimetiere P, Rakotovo R. Weight and mortality. (Letter). Lancet 1983;1:1104.
- Rhoads GG, Kagan A. The relation of coronary disease, stroke, and mortality to weight in youth and in middle age. Lancet 1983;1:492–5.[Web of Science][Medline]
- Paffenbarger RS Jr, Hyde RT, Wing AL, et al. Physical activity, all-cause mortality, and longevity of college alumni. N Engl J Med 1986;314:605–13.[Abstract]
-
Sidney S, Friedman GD, Siegelaub AB. Thinness and mortality. Am J Public Health 1987;77:317–22.
[Abstract/Free Full Text] - Wilcosky T, Hyde J, Anderson JJ, et al. Obesity and mortality in the Lipid Research Clinics Program Follow-up Study. J Clin Epidemiol 1988;43:743–52.
-
Harris T, Cook F, Garrison R, et al. Body mass index and mortality among nonsmoking older persons: The Framingham Heart Study. JAMA 1988;259:1520–4.
[Abstract/Free Full Text] -
Wannamethee G, Shaper AG. Weight change, perceived health status and mortality in middle-aged British men. Postgrad Med J 1990;66:910–13.
[Abstract/Free Full Text] - Deeg DJ, Miles TP, Van Zonneveld RJ, et al. Weight change, survival time and cause of death in Dutch elderly. Arch Gerontol Geriatr 1990;10:97–111.[Web of Science][Medline]
-
Lee I-M, Paffenbarger RS Jr. Change in body weight and longevity. JAMA 1992;268:2045–9.
[Abstract/Free Full Text] -
Pamuk ER, Williamson DF, Madans J, et al. Weight loss and mortality in a national cohort of adults, 1971–1987. Am J Epidemiol 1992;136:686–97.
[Abstract/Free Full Text] -
Higgins M, D'Agostino R, Kannel W, et al. Benefits and adverse effects of weight loss: observations from the Framingham Study. Ann Intern Med 1993;119:758–63.
[Abstract/Free Full Text] -
Harris TB, Ballard-Barbasch R, Madans J, et al. Overweight, weight loss, and risk of coronary heart disease in older women: The NHANES I Epidemiologic Follow-up Study. Am J Epidemiol 1993;137:1318–27.
[Abstract/Free Full Text] - Rumpel C, Harris TB, Madans J. Modification of the relationship between the Quetelet index and mortality by weight-loss history among older women. Ann Epidemiol 1993;3:343–50.[Medline]
-
Pamuk ER, Williamson DF, Serdula MK, et al. Weight loss and subsequent death in a cohort of U.S. adults. Ann Intern Med 1993;119:744–8.
[Abstract/Free Full Text] -
Williamson DF, Pamuk E, Thun M, et al. Prospective study of intentional weight loss and mortality in never-smoking overweight US white women aged 40–64 years. Am J Epidemiol 1995;141:1128–41.
[Abstract/Free Full Text] -
Galanis D, Harris T, Sharp DS, et al. Relative weight, weight change, and risk of coronary heart disease in the Honolulu Heart Program. Am J Epidemiol 1998;147:379–86.
[Abstract/Free Full Text] -
Noppa H. Body weight change in relation to incidence of ischemic heart disease and change in risk factors for ischemic heart disease. Am J Epidemiol 1980;111:693–704.
[Abstract/Free Full Text] - Schroll M. A longitudinal epidemiological survey of relative weight at age 25, 50, and 60 in the Glostrup population of men and women born in 1914. Dan Med Bull 1981;28:106–16.[Web of Science][Medline]
-
Hubert HH, Feinleib M, McNamara PM, et al. Obesity as an independent risk factor for cardiovascular disease: a 26-year follow-up of participants in the Framingham Heart Study. Circulation 1983:67:968–77.
[Abstract/Free Full Text] - Manson JE, Colditz GA, Stampfer MJ, et al. A prospective study of obesity and risk of coronary heart disease in women. N Engl J Med 1990;322:882–9.[Abstract]
-
Rimm EB, Stampfer MJ, Giovannucci E, et al. Body size and fat distribution as predictors of coronary heart disease among middle-aged and older US men. Am J Epidemiol 1995;141:1117–27.
[Abstract/Free Full Text] -
Willett WC, Manson JE, Stampfer MJ, et al. Weight, weight change, and coronary heart disease in women: risk within the "normal" weight range. JAMA 1995;273:461–5.
[Abstract/Free Full Text] -
Manson JE, Willett WC, Stampfer MJ, et al. Body weight and mortality in women. N Engl J Med 1995;333:677–85.
[Abstract/Free Full Text] -
Rexrode KM, Hennekens CH, Willett WC, et al. A prospective study of body mass index, weight change, and risk of stroke in women. JAMA 1997;277:1539–45.
[Abstract/Free Full Text] -
Manson JE, Stampfer MJ, Hennekens CH, et al. Body weight and longevity: a reassessment. JAMA 1987;257:353–8.
[Abstract/Free Full Text] - Spataro JA, Dyer AR, Stamler J, et al. Measures of adiposity and coronary heart disease mortality in the Chicago Western Electric Company Study. J Clin Epidemiol 1996;49:849–57.[Web of Science][Medline]
-
Paul O, Lepper MH, Phelan WH. A longitudinal study of coronary heart disease. Circulation 1963;28:20–31.
[Abstract/Free Full Text] - Stamler J, Rains-Clearman D, Kenz-Litzow, et al. Relation of smoking at baseline and during years 1–6 to food and nutrient intakes and weight in the special intervention and usual care groups in the Multiple Risk Factor Intervention Trial. Am J Clin Nutr 1997;65(suppl):374s–402s.
-
Froom P, Kristal-Boneh E, Melamed S, et al. Smoking cessation and body mass index of occupationally active men: The Israeli CORDIS Study. Am J Public Health 1999;89:718–22.
[Abstract/Free Full Text] - Office of the Surgeon General, US Public Health Service. The health benefits of smoking cessation: a report of the Surgeon General, 1990. Washington, DC: US Department of Health and Human Services, 1990.
- National Center for Health Statistics. International classification of diseases, Eighth Revision, adapted for use in the United States. Vol 1. Washington, DC: US Public Health Service, 1967:49–522.
- Cox DR. Regression models and life tables. J R Stat Soc Ser (B) 1972;34:187–202.
-
French SA, Folsom AR, Jeffery RW, et al. Prospective study of intentionality of weight loss and mortality in older women: The Iowa Women's Health Study. Am J Epidemiol 1999;149:504–14.
[Abstract/Free Full Text] -
Williamson DF, Pamuk E, Thun M, et al. Prospective study of intentional weight loss and mortality in overweight white men aged 40–64 years. Am J Epidemiol 1999;149:491–503.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. E. Field, S. Malspeis, and W. C. Willett Weight Cycling and Mortality Among Middle-aged or Older Women Arch Intern Med, May 11, 2009; 169(9): 881 - 886. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. M. Corrada, C. H. Kawas, F. Mozaffar, and A. Paganini-Hill Association of Body Mass Index and Weight Change with All-Cause Mortality in the Elderly Am. J. Epidemiol., May 15, 2006; 163(10): 938 - 949. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Breeze, R. Clarke, M. J Shipley, M. G Marmot, and A. E Fletcher 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., February 1, 2006; 35(1): 169 - 178. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S G. Wannamethee, A G. Shaper, and M. Walker Overweight and obesity and weight change in middle aged men: impact on cardiovascular disease and diabetes J Epidemiol Community Health, February 1, 2005; 59(2): 134 - 139. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. L. Daviglus, K. Liu, L. L. Yan, A. Pirzada, D. B. Garside, L. Schiffer, A. R. Dyer, P. Greenland, and J. Stamler Body Mass Index in Middle Age and Health-Related Quality of Life in Older Age: The Chicago Heart Association Detection Project in Industry Study Arch Intern Med, November 10, 2003; 163(20): 2448 - 2455. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. G. Wannamethee, A. G. Shaper, and M. Walker Weight Change, Weight Fluctuation, and Mortality Arch Intern Med, December 9, 2002; 162(22): 2575 - 2580. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Tilling, J. A. C. Sterne, and M. Szklo Estimating the Effect of Cardiovascular Risk Factors on All-Cause Mortality and Incidence of Coronary Heart Disease Using G-Estimation: The Atherosclerosis Risk in Communities Study Am. J. Epidemiol., April 15, 2002; 155(8): 710 - 718. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. N. Singh and X. Wang Simulation Study of the Effect of the Early Mortality Exclusion on Confounding of the Exposure-Mortality Relation by Preexisting Disease Am. J. Epidemiol., November 15, 2001; 154(10): 963 - 971. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Other Articles Noted Evid. Based Nurs., October 1, 2001; 4(4): E1 - 11. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||