点击显示 收起
1 From the Centre for Preventive Medicine, Unit for Dietary Studies, Glostrup University Hospital, University of Copenhagen, and the Danish Epidemiology Science Centre at the Institute of Preventive Medicine, Copenhagen, University Hospital (BLH), and the Department of Physiology, University of Odense, Denmark (LG).
2 Supported by The Danish Research Council (FREJA), the Danish Health Insurance Foundation, The Wedell-Wedellsborgs Foundation, and the Danish National Research Foundation. 3 Reprints not available. Address correspondence to BL Heitmann, Institute of Preventive Medicine, Copenhagen University Hospital, Øster Farimagsgade 5A, opg 23A 1399 København K, Denmark. E-mail: blh{at}ipm.hosp.dk.
ABSTRACT
Background: Weight loss may be associated with unfavorable changes in body composition not compensated for by subsequent weight gain.
Objective: We examined the composition of weight change in relation to obesity, previous weight changes, weight-loss attempts, and physical activity.
Design: Part of the Danish MONICA (Monitoring Trends in Cardiovascular Disease) project, this was a longitudinal population study of changes in weight and body composition, with examinations in 1982–1983, 1987–1988, and 1993–1994. A total of 1236 men and 1200 women aged 35, 45, 55, or 65 y in 1987–1988 participated. Changes in fat and fat-free mass were measured by bioelectrical impedance.
Results: Before adjustment for age-related changes, fat-free mass made up 41% of weight lost and 24% of weight gained in men. In women, loss of fat-free mass (35%) was more than double that of gains (15%). After adjustment, the fractions of weight gained as fat-free mass were not significantly different from the fractions lost. These fractions were independent of age, obesity, and weight changes in the previous 5 y; successful weight-loss attempts; and physical activity. Independent of age and degree of obesity, weight changes were associated with greater changes in fat-free mass in men than in women.
Conclusions: These data do not support the theory that weight loss or weight cycling may lead to an unfavorable body composition, nor do they provide a biological explanation for why long-term weight loss is often unsuccessful. However, the metabolic and health consequences of weight change may differ in men and women.
Key Words: Body composition • obesity • weight change • longitudinal population study • Denmark • fat-free mass • weight cycling • Danish MONICA project
INTRODUCTION
Most studies report that weight loss is associated with increased mortality (1). In addition, most available studies suggest that repeated weight loss and gain—so-called weight cycling—carries an independent health risk (2–4). One hypothesis is that weight loss is associated with an unfavorable change in body composition that is not compensated for by a subsequent gain in weight, ie, that weight cycling results in a higher ratio of body fat to fat-free mass. Such changes are not picked up by measures of relative overweight, such as the body mass index (BMI), and may help explain why both weight cycling and weight loss seem to be related to increased mortality. Several studies examined this question in experimental and observational settings (5–10). These studies generally found no evidence of a difference in gain and loss of fat-free mass with weight change, but they had limitations: the authors did not specifically address the question of disproportionate loss and gain of lean mass (9); data were restricted to small, often highly physically active groups (7, 8); data were cross sectional in origin (5, 7, 8); or there was insufficient statistical information to interpret their data (10). We used cross-sectional data in a previous study to examine the composition of body weight differences in large groups of apparently healthy subjects of the same height and showed that fractions of fat-free mass per differences in body weight (in kg) were 72% and 65% in women and men, respectively (5). We suggested that these fractions closely resemble the composition of weight change in comparable groups of subjects.
In the present study we examined the composition of weight changes in 2436 Danish adults between examinations in 1987–1988 and 1993–1994. Because many factors—age, degree of obesity, previous weight changes, successful weight-loss attempts, and physical activity during leisure—may influence the composition of the weight lost or gained, we investigated whether weight loss and gain are accompanied by different relative changes in body composition in groups of subjects characterized by differences in these factors.
SUBJECTS AND METHODS
Subjects
The study was part of the Danish MONICA-I (Monitoring Trends in Cardiovascular Disease) project, an international study conducted under the auspices of the World Health Organization to monitor trends in and determinants of mortality from cardiovascular disease (11). The first survey was carried out in 1982–1983. An age-stratified sample of subjects (n = 4807) aged 30, 40, 50, and 60 y was selected at random from the Central Person Register from citizens living in 11 municipalities in the western part of Copenhagen County. Subjects not born in Denmark were excluded (n = 226). The remaining 4581 subjects were found to be reasonably representative of the total Danish population with respect to sex, age, educational level, occupation, and housing, but persons employed in agriculture, horticulture, or fishery and self-employed and unskilled workers were slightly underrepresented (12). Of the 4581 subjects, 3608 (79%) were given a general health examination, including measurement of height and weight. All of these subjects who were still living were invited to 2 reexaminations, one performed between December 1987 and November 1988 (13) and another between May 1993 and November 1994 (14). A total of 2987 subjects (83%) aged 35, 45, 55, or 65 y were examined in 1987–1988, and 2656 subjects (74%) aged 41, 51, 61, and 71 y were examined in 1993–1994. Height and weight were measured at all 3 examinations, but body composition was measured only in the latter 2. The population groups and measures were described in detail elsewhere (13, 14). Complete data from all 3 examinations were obtained for 2436 subjects. All subjects gave written informed consent. The project was approved by the Ethics Committee for Copenhagen County and is in accordance with the Helsinki Declaration of 1975 as revised in 1983.
Anthropometric data
All anthropometric measurements were made in accordance with World Health Organization standards (15). Height was measured to the nearest 0.5 cm while the subjects were standing without shoes, with their heels together, and with their heads in the horizontal Frankfurter plane. Body weight was measured to the nearest 0.1 kg on a Seca scale (Copenhagen), with the subjects wearing only hospital underwear.
Measurements of electrical impedance
We used a BIA-103 body-composition analyzer (RJL Systems, Detroit) to measure electrical impedance, following the manufacturer's instructions. Measurements were taken with a tetrapolar electrode arrangement while the subjects were lying relaxed on a couch. The electrodes were placed on the dorsal surfaces of the right hand and foot at the distal metacarpals and metatarsals, respectively, and between the distal prominence of the radius and the ulna at the wrist and the medial and lateral malleoli at the ankle. In a previous study we used a subset of our population to develop an equation for estimating body fat from impedance, with measurements of total body water and potassium as a reference (16):
RESULTS
Nonparticipation
The nonparticipants were older than the participants (P < 0.0001). After adjustment for these age differences, height, weight, and BMI did not differ significantly at baseline or follow-up between participants and nonparticipants. Nonparticipation was described in detail previously (12, 13).
Changes in body composition related to age
Mean (±SD) characteristics of subjects from the 3 examinations are shown in Table 1 by sex for the total group and by weight status (ie, stable weight, weight loss >3 kg, and weight gain >3 kg) between the 1987–1988 and 1993–1994 examinations.
View this table:
TABLE 1 . Subject characteristics and body-composition indexes estimated by electrical impedance for all subjects and for weight-stable, weight-loss, and weight-gain groups1
Between the 1987–1988 and 1993–1994 examinations, 156 men and 125 women lost >3 kg and 480 men and 486 women gained >3 kg. In 1987–1988 BMI and body fat were greater in men and women who subsequently lost weight than in those who gained weight, but fat-free mass did not differ significantly between the sexes.
Changes in body fat and fat-free mass in subjects whose weight was stable during the 6-y period reflect the expected loss in fat-free mass and gain in body fat related to age. Before adjustment for these age-related changes in body composition, which are independent of weight change, the fraction of body weight lost as fat-free mass was greater than that gained as fat-free mass (men, 41% compared with 24%; women, 35% compared with 15%). However, after adjustment for age-related changes in body composition, the fractions of body weight lost and gained as fat-free mass were not significantly different, and these fractions made up 33% of the body weight change in men and 25% of the weight change in women. The difference between men and women was significant across all weight groups (Table 2).
View this table:
TABLE 2 . Relative change in fat-free mass as a fraction of body weight in the 3 weight groups1
Subjects who lost weight were older and subjects who gained weight were younger than the average (Table 1). Age-related changes increased with age in both men and women. Before the adjustment for the age-related changes in body composition, the fraction of fat-free mass lost with weight loss was greater than the fraction gained with weight gain, and this difference increased with age (Figure 1). However, after adjustment, fractions of fat-free mass gained or lost with weight changes were independent of age for both men and women, as indicated by intercepts that were not significantly different and estimated slopes that were not significantly different from zero (Figure 2). Differences between men and women remained, however, and were independent of age (P = 0.01). The interaction of sex and age was significant both before and after adjustment (P < 0.0001).
FIGURE 1. . Mean (±SD) change () in fat-free mass (FFM) as a percentage of the change in body weight (BW) among men (solid lines) and women (dotted lines) aged 35, 45, 55, or 65 y who gained (•) or lost () >3 kg BW between the 1987–1988 and 1993–1994 examinations, before adjustment for age-related changes in FFM. The changes in FFM were significantly different with weight loss and weight gain in both sexes (P < 0.003, ANOVA) and between sexes (P < 0.0001, ANOVA).
FIGURE 2. . Mean (±SD) change () in fat-free mass (FFM) as a percentage of the change in body weight (BW) in men (solid lines) and women (dotted lines) aged 35, 45, 55, or 65 y who gained (•) or lost () >3 kg BW between the 1987–1988 and 1993–1994 examinations, after adjustment for age-related changes in FFM. There were no significant within-sex differences between groups, but the differences between men and women were significant (P < 0.0001, ANOVA).
Changes in body composition related to degree of obesity
Men and women were grouped according to quartiles of percentage body fat in 1987–1988. Before adjustment for age-related changes in body composition, the fraction of fat-free mass lost with weight loss was greater than the fraction gained with weight gain, and this difference was greater in the lean than in the obese persons (all P < 0.003; data not shown). After adjustment, fractions of fat-free mass gained or lost with weight changes were not significantly different between the lean and the obese persons, as indicated by intercepts that were not significantly different and estimated slopes that were not significantly different from zero. Data are given for men and women separately in Figure 3. Differences between men and women remained and were independent of percentage body fat (P = 0.01). The interaction of sex and percentage body fat was significant both before and after adjustment (P < 0.0001).
FIGURE 3. . Mean (±SD) change () in fat-free mass (FFM) as a percentage of the change in body weight (BW), by percentage body fat adjusted for age, in men (solid lines) and women (dotted lines) who gained (•) or lost () >3 kg BW between the 1987–1988 and 1993–1994 examinations, after adjustment for age-related changes in FFM. There were no significant within-sex differences between groups, but the differences between men and women were significant (P < 0.0001, ANOVA).
Changes in body composition related to amount of physical activity
The analyses showed that physical activity did not significantly influence the fractions of body weight changes subsequently gained or lost as fat-free mass, but that fractions of fat-free mass lost with weight loss were greater than fractions gained with weight gain before adjustment for age-related changes in body composition (both P < 0.0005; data not shown). After adjustment, the fractions of fat-free mass gained or lost with weight changes were not significantly different for active and inactive subjects (Figure 4), as indicated by intercepts that were not significantly different and estimated slopes that were not significantly different from zero. However, the differences between men and women remained significant and were independent of physical activity levels. The interaction of sex and physical activity was significant both before and after adjustment (P < 0.0001).
FIGURE 4. . Mean (±SD) change () in fat-free mass (FFM) as a percentage of the change in body weight (BW), by physical activity in men (solid lines) and women (dotted lines) who gained (•) or lost () >3 kg BW between the 1987–1988 and 1993–1994 examinations, adjustment for age-related changes in FFM. There were no significant within-sex differences between groups, but the differences between men and women were significant (P < 0.0001, ANOVA).
Changes in body composition related to weight changes in the previous 5 y and successful weight-loss attempts
Weight changes in the previous 5 y did not significantly influence the fractions of body weight changes subsequently gained or lost as fat-free mass, but the fractions of fat-free mass lost with weight loss were greater than the fractions gained with weight gain before the adjustment for age-related changes in body composition (both P < 0.002). After adjustment, the fractions of fat-free mass gained or lost with weight changes were not significantly different for subjects with a weight gain or loss in the previous 5 y (Figure 5), as indicated by intercepts that were not significantly different and estimated slopes that were not significantly different from zero. The interaction of sex and weight change in the previous 5 y was significant both before and after adjustment (P < 0.0001).
FIGURE 5. . Mean (±SD) change () in fat-free mass (FFM) as a percentage of the change in body weight (BW), by weight change in the 5 y before the 1987–1988 examination (±>1 kg) in men (solid lines) and women (dotted lines) who gained (•) or lost () >3 kg BW between the 1987–1988 and 1993–1994 examinations. Adjustment was made for age-related changes in FFM. There were no significant within-sex differences between groups, but the differences between men and women were significant (P < 0.0001, ANOVA).
The fraction of fat-free mass lost with weight loss was not significantly dependent on successful weight-loss attempts, although the fractions of fat-free mass lost with weight loss tended to be smaller for intentional losses (Table 3). The interaction of sex and successful weight loss was significant both before and after adjustment (P < 0.01). However, differences between men and women remained and were independent of weight changes in the previous 5 y and of successful weight-loss attempts (both P < 0.0001).
View this table:
TABLE 3 . Relative mean loss of fat-free mass as a fraction of body weight in subjects who lost >3 kg between the examinations in 1987–1988 and 1993–1994 with or without a previous intentional weight loss of >5 kg1
DISCUSSION
The present study showed that the weight-gaining groups were younger and the weight-losing groups were older than average, in agreement with the findings of other studies (13, 20). Hence, those losing weight would be expected to lose relatively more fat-free mass as a simple consequence of being older. Indeed, it is well known that changes in body composition that are independent of weight changes vary with age (9, 21), suggesting that a substantial reorganization of fat and fat-free mass occurs with aging. The present study agrees with other studies in showing that age-related losses of fat-free mass, independent of weight changes, increase with age. However, changes in body composition occurring with weight changes were constant in persons aged 35–70 y, as suggested by the fact that fat-free mass did not make up a greater fraction of weight loss than of weight gain when age-related changes in body composition were adjusted for. These findings are in general agreement with results obtained from experimental and cross-sectional population studies (5–10), even though the very elderly may show a disproportionally large loss of fat-free mass in relation to weight loss (18). Data from the present study on the composition of body weight changes (Table 2) are almost exactly the same as our cross-sectional data on weight differences (5). Of particular interest in the present study is the remarkably small variation among subjects in the composition of the weight changes, with CVs of only 3–4%.
We found that weight changes were associated with more unfavorable relative changes in fat-free mass in men than in women, even after considering age-related changes in fat-free mass and the differences in percentage body fat between men and women. This finding agrees with that of our cross-sectional study (5). However, the finding is surprising because it was assumed that such sex differences depended on differences in body fatness and because most, but not all (21), studies failed to show such differences (22–24). Experimental studies have found that physical activity decreases losses of fat-free mass (20, 25). Hence, differences in physical activity provided one speculative explanation for the differences between the sexes. However, in the present study more women than men reported being inactive, and changes in fat-free mass were not significantly different between those who were physically active and those who were inactive at baseline (data not shown).
Our results imply that in middle-aged adults, metabolic and other health consequences of weight change are dependent on sex. Few studies have had the aim or the statistical power to explore health differences between men and women related to BMI or changes in BMI. However, crude data from 1.7 million Norwegian adults suggest that compared with middle-aged women, middle-aged men have a greater mortality both at a high and at a low BMI (26). At later ages these differences seem to disappear.
The finding that the change in fat-free mass relative to the change in body weight was unrelated to percentage body fat was also surprising, because others suggested that during weight change more fat is lost or gained by obese than by lean persons (9, 21, 24, 27). Although the possibility cannot be excluded that body-composition changes in subjects who are very lean differ from those of who are very obese, the results of the present study, conducted on a large sample of adult Danes with substantial variation in body composition, provide no support for this.
Weight loss may be voluntary or involuntary, and the composition of the weight lost may depend on this. We addressed this question by examining whether a different fraction of weight lost was fat-free mass in subjects who reported successful weight-loss attempts than in those who did not. Although we found no significant difference once age-related changes in body composition were accounted for, the data showed that less fat-free mass was lost by subjects with a history of successful weight-loss attempts than by subjects without such a history. The subjects did not provide information on whether the successful weight-loss attempts were the consequences of diet, physical activity, drugs, or a combination of these; hence, we cannot explore this further.
We anticipated that the fraction of fat-free mass lost during a loss of body weight would be larger than the fraction gained with weight gain. The present study provided no support for this when age-related changes in body composition that are independent of weight change were accounted for. Hence, the results of the present study, which measured body-composition changes over time in a large and representative sample of adult Danes with a high response rate does not seem to provide a biological explanation for why long-term weight loss is often unsuccessful. Nor do the results offer support for the theory that weight cycling results in a low percentage of fat-free mass, which may be associated with increased mortality (28).
Some limitations to the present study should be noted. First, it may be argued that impedance measurements are not sufficiently precise to detect small changes in body composition. However, several studies have documented the reliability and validity of this method of measuring body composition (29). In addition, the measured difference in fractions of fat-free mass gained and lost were small. Therefore, it is not likely that we overlooked small significant differences. Furthermore, several studies have shown that impedance performs well in measuring body composition at the group level in epidemiologic surveys but not as well when measuring individuals in clinical settings (29). Furthermore, the particular equation used for computing fat and fat-free mass in the present study was developed specifically for the present population sample (16). When we reexamined the data, we found the fractions of weight change made up by fat-free mass to be independent of the size of the weight change (data not shown). Hence, measurement error did not seem to be a problem in the present study.
Second, we cannot exclude the possibility that the measurement of leisure-time physical activity may have been subject to misclassification and was too crude to differentiate between sedentary and more active individuals. However, the questionnaire was found to have high validity when compared with fitness and other measures of activity (17), has been used for >30 y in Scandinavia, and was shown to be a strong predictor of risk factors for all causes of death and total mortality (30, 31). Although we cannot exclude the possibility that the crude measure of physical activity was partly responsible for the lack of difference in results between the active and inactive subjects, the present data do not provide evidence that differences in physical activity are responsible for the sex difference.
A third limitation of the present study may be the lack of information about the time course of the weight changes. Rapid weight changes may involve a relatively larger change in fat-free mass than do gradual weight changes (25); these 2 types of weight changes were not analyzed separately in the present study. Hence, any possible metabolic differences between adapted and transitional changes cannot be ascertained. Indeed, whether men experience more rapid weight changes than do women is unknown, but the possibility cannot be excluded that the sex differences may relate to this. On the other hand, during the 6 y between the examinations in 1987–1988 and 1993–1994, the women generally reported more weight-loss attempts that resulted in a weight loss of >5 kg than did the men, suggesting that the women may have experienced more periods of rapid weight loss than did the men.
Finally, the reliability of self-reported information on intentionality of weight loss may be limited, as pointed out by French et al (32). Furthermore, it is likely that in a population sample such as that in the present study, subjects tend to cycle in weight above and below an average level with an unknown periodicity, which may not be equivalent to the 6-y weight changes reported here. Hence, the measurement time points may be considered "snapshots," catching some on their way up in weight and others on their way down. Indeed, surveys done at different times are likely to catch different people in weight-gaining and -losing groups, as indicated by the fact that twice as many subjects who reported weight gains in the 5 y before the study fell into the weight-loss group than into the weight-gain group at subsequent examinations. In other words, regression toward the mean may be part of the observed result. However, the fact that weight changes in the previous 5 y did not influence the composition of the subsequent weight changes does not support the importance of regression toward the mean.
The present study, in which there was a high response rate from a large and representative sample of the Danish population, showed no significant within-sex difference in the fraction of weight lost or gained as fat-free mass. The hypothesis that weight cycling may result in adverse health effects by causing a relative loss of fat-free mass cannot be supported by these data. On the other hand, even after taking into account percentage body fat and age-related changes in fat-free mass, weight changes were found to be associated with a more unfavorable relative change in fat-free mass in men than in women, suggesting that the metabolic and health consequences of weight change may be dependent on sex to some degree.
ACKNOWLEDGMENTS
We thank statisticians Michael Gamborg and Stine Segel for assistance with the statistical analyses.
REFERENCES