¹Department of Health Economics and Health Promotion Sciences, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan

²PL Tokyo Health Care Center, Tokyo, Japan

Correspondence to: J S Lee, Department of Health Economics and Health Promotion Sciences, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. Email: jslee-tky@umin.ac.jp

OBJECTIVE: The purpose of the present study was to determine the effects of weight variability on cardiovascular risk factors (CRF) based on a large sample of community-resident Japanese males and females.

METHOD: A total of 3564 men and 1955 women, all Japanese, aged 30-69 y in the baseline year (1987), were followed-up for up to 10 y (end-point in 1996). Height, body weight, systolic and diastolic blood pressure, fasting serum total cholesterol, triglyceride and fasting plasma glucose were measured as an annual health check-up. At least six times in 10 y, body mass index (BMI) mean was calculated as an index of the BMI level of each subject (BMI mean). Direction and magnitude of the change in a subject's BMI was determined by a regression slope of BMI values over time (BMI slope). BMI fluctuation was defined as the root mean square error (BMI RMSE) of a regression line. The slopes of the five CRF were calculated using each regression equation over time.

RESULTS: The BMI slope strongly correlated to each CRF slope independent of baseline age, baseline CRF value, smoking habit, BMI mean and BMI RMSE. BMI RMSE did not correlate to any CRF slopes.

CONCLUSION: This study indicates that weight gain and weight loss have a much greater effect on CRF change than does weight fluctuation in Japanese middle-aged men and women.

International Journal of Obesity (2001) 25, 1063-1067

body mass index; cardiovascular risk factor; body weight change; body weight fluctuation

Introduction

Overweight has long been a concern of health care researchers because it has well-established health hazards such as greater risk of cardiovascular disease, hypertension, diabetes and dyslipidemia, as well as its association with premature mortality.^1,2,3 Weight reduction in overweight individuals has beneficial health effects, including improvements in blood pressure,⁴ glucose tolerance,⁵ insulin resistance,⁶ and lipid levels.^7,8 These associations have followed the public health recommendation that persons whose weight is substantially above an average weight-for-height should lose weight.

However, concern has been raised about the possible negative effects of weight change and weight fluctuation. Most people who lose weight often regain it.⁹ Such variability in body weight is a common phenomenon,¹⁰ but the long-term health consequences of weight variability are not well understood. It is important to examine the health implications of weight variability.

Studies of the effects of weight variability on morbidity and mortality have produced controversial results. In the Framingham study,¹¹ subjects with highly variable body weights have higher total mortality and higher coronary heart disease mortality and morbidity. However, the Gothenburg prospective studies¹² found no relation between body weight variability and mortality.

If there is a positive relation between weight variability and cardiovascular disease mortality, weight variability presumably increases the level of cardiovascular risk factors (CRF) as well. However, little information is available concerning the effects of weight variability on CRF and controversies exist.^{11,13,14,15,16} It is important to examine the effects of body weight variability on the CRF in a variety of populations with different cardiovascular disease mortality.

The purpose of the present study was to determine the effects of weight variability on CRF based on a large sample of community-residing Japanese males and females.

Methods

Subjects

The study population was selected from the 23 716 men and women who underwent annual periodic health examinations in 1987 at the PL Tokyo Health Care Center in Tokyo, Japan. This center is a periodic health examination facility mainly for employees and their family members referred from over 350 employment-based health insurance societies. Insured persons are mostly residents of the Tokyo metropolitan area. The study subjects consisted of 8157 people (5509 men and 2648 women), aged 30-69 y at the baseline year (1987) and had health check-ups at least six times during 10 y up to 1996.

From the physician's diagnosis, a total of 32% of the registries had heart disease, cerebrovascular disease, hypertension, hyperlipidemia or diabetes mellitus, and they had already received medical treatment in this study period (10 y). Because these treatments would affect the weight and CRF changes, we excluded these subjects from the analysis.

Measurements

Body mass index (BMI) variability indices: Weight and height were obtained from the annual health examination records during the follow-up periods. Weight and height were measured in the morning fasting state to the nearest 0.1 kg and 0.1 cm respectively, while the subjects were in light clothing and without shoes. BMI (weight (kg) over height (m)²) was used as the relative weight. BMI mean was calculated as an index for the BMI level of each subject during the follow-up periods (BMI mean). Using a simple linear regression model in which each subject's BMI value was a dependent variable and the examination years independent ones, we divided BMI variables into two components as follows: the slope coefficient of this model representing an individual's BMI change trend of direction and magnitude (BMI slope) and the standard deviation around this slope, root mean square error, representing the BMI fluctuation magnitude (BMI RMSE).

CRF: Systolic (SBP) and diastolic blood pressure (DBP), fasting serum total cholesterol (TC) and serum triglyceride (TG), fasting plasma glucose (FPG) were obtained from each health examination record. SBP and DBP were measured using a standard mercury sphygmomanometer after the subject had sat at rest for 30 min. Blood was taken from the anterior cubital vein in the fasting state and TC, TG and FPG were analyzed by enzymatic methods with an automatic analyzer (Hitachi 7450 Automatic Analyzer). Quality control for blood testing was performed every day using pooled standard blood samples. Coefficients of variation (CV) of TC, TG and FPG in this measurement system were 0.98, 1.59 and 1.23 respectively.

The slopes for these five CRF (CRF slope) on each individual were calculated using a simple linear regression equation over time, respectively.

In addition, we collected information about present illness, past history of illness and smoking habit (at baseline) from the medical records.

Statistical analysis

In order to describe our study subjects, we compared the data at the baseline and endpoint using the paired t-test. To compare the subjects with and without present illness, Student's t-test was used. To examine the associations between the BMI variables and changes in each CRF (CRF slope), we calculated Pearson's correlation coefficients. To assess which of the BMI variables were most strongly related to each CRF change, we used a multiple linear regression analysis. In this analysis, each CRF slope was used as the dependent variable, and the independent variables were the baseline value of each CRF, baseline age, smoking habits, BMI mean, BMI slope, and BMI RMSE.

Results

Characteristics of the subjects

The subjects were 30 to 69-y-old at baseline year (1987) and free from heart disease, cerebrovascular disease, hypertension, hyperlipidemia or diabetes mellitus in the study periods. When we compared our study subjects with excluded subjects, the latter were significantly older, heavier in their body weight and had higher CRF levels (P<0.001) at baseline (data not shown). There was no significant difference in the rate of smoking habit at baseline.

As a result, 3564 men and 1955 women were analyzed in our study. All of them had at least six health examinations up to the end-point. The mean values of each examination at the baseline and end-point are shown in Table 1. At the end-point, weight and BMI were significantly (P<0.001) heavier than the baseline values in both males and females. SBP, DBP and FPG became significantly (P<0.001) higher at the endpoint than the baseline, too. Compared to the baseline, TC and TG were significantly (P<0.001) lower at the end-point in males, however no significant changes occurred in females.

Table 2 shows the BMI variability indices in males and females. There were no significant differences between males and females in each value.

Correlations of BMI variability indices with CRF changes

The BMI slope strongly correlated with the slope of each CRF in both males and females (Table 3). This indicated that those with BMI gain would have CRF increase and vice versa. Correlation coefficients between BMI mean or BMI RMSE and the slope of each CRF were significant in males except for that between FPG slope and BMI RMSE. However, in females the only significant correlation was seen between BMI mean and SBP slope. The positive correlation between baseline age and SBP slope in males and females indicated that those who were older at the baseline tended to have an increase in SBP. The negative correlation between age and TC and TG slope in males and females indicated that those younger at the baseline tended to have an increase in TC and TG. The baseline CRF values had negative correlation with each CRF slope, while the baseline smoking habit had almost no correlation with the CRF slopes.

Multivariate analysis

Table 4 shows the results of the multiple regression analysis between BMI variability indices and each CRF slope adjusted by the baseline CRF value, age, smoking habits and the other BMI variability indices. The BMI slope significantly correlated with the slope of each CRF independently, while no correlation was found between BMI RMSE and each CRF slope in both males and females. BMI mean correlated to SBP, DBP, TC and FPG slope in males only.

Furthermore, we re-performed the multiple regression analysis after dividing the subjects to the younger age group (age 30-49 y) and older age group (age 50-69 y), because correlation between BMI variability indices and CRF slopes might be different between younger and older subjects and age might not be linearly associated with CRF change (Table 5). After adjustment for baseline age, baseline value of each CRF, smoking habit and the other BMI variability indices, the results were the same as in Table 4 and did not differ between younger and older age groups.

Discussion

The most important finding of this study was that BMI slope itself strongly correlated with the five CRF slopes in males and females after adjustment for baseline CRF value, age, smoking habits and other BMI variability indices. This indicated that weight gain is associated with an increase in the level of CRF and weight loss with a favorable improvement in CRF. However, BMI RMSE, which we chose as an index of weight fluctuation, correlated to none of the CRF slopes.

This finding was consistent with results of Taylor et al¹³ in a community-based sample of men and women, Jeffery et al¹⁴ in obese men and women, and also our earlier report¹⁵ for white collar middle-aged Japanese male workers. However, Lissner et al¹⁶ found that weight variability decreased glucose tolerance in the Baltimore study. These controversies may come from the differences of the magnitude of the weight change in the study subjects, and differences of weight variation indices they used. Also, the correlation between weight variation indices and CRF change might be different by the age of the study population. Therefore, we conducted the multivariate analysis after dividing the study subjects into the younger and older groups and the results were the same between younger and older groups. Correlation between BMI slope and CRF slope were not affected by the age of this study population.

BMI mean represented the measure of whether the subjects were lean or obese during the follow-up periods. In our result, correlation coefficient between BMI mean and CRF slope was small or not significant after adjustment for BMI slope and other variables. Because body weight and body weight change during the follow-up period was relatively small and body weight gain or loss was voluntary in our study population, BMI slope was more important than BMI mean as a BMI variability index which correlated with CRF changes.

In this study, we used BMI RMSE as an index reflecting the intra-individual body weight fluctuation. Intra-personal coefficient of variation (CV) of BMI was used to describe weight fluctuation in Lissner et al.^11,16,17 Problems in using weight fluctuation indices have been reported,^18,19 especially that a person who has a small and steady weight gain over a long period of time could have a similar CV to a person with a large weight gains and losses resulting in no overall weight gain. Itoh et al²⁰ compared interrelationships among different body weight variability indices and their relationships with CRF. They reported that CV of BMI showed a significant relationship with CRF, but weight change category (cycler, gainer or noncycler), number of weight change episodes and RMSE of BMI did not show any relationship with CRF. The CV of BMI implies mainly simple weight gainers, and CV does not present weight fluctuation. Because CV and slope are correlated, one cannot discriminate between a nonlinear slope effect and instability of changes by CV itself. In contrast, BMI RMSE could be a more sensitive measure of instability, because it had no correlation with BMI slope.¹⁵ In this study, BMI RMSE, which we chose as an index of weight fluctuation, related to none of the CRF slopes. This result was also consistent with those of former study results^13,15 which also found no weight fluctuation effect on any CRF variables measured in their population.

Unfavorable health effects of weight fluctuation have been attributed to the body fat mass gain, a decrease in resting energy expenditure, or an increase in abdominal fat.²¹ However, according to the review articles^22,23 concerning the effects of weight variability, negative effects of weight fluctuation would be overcome by the positive effects of weight reduction. We also support the potential benefits of moderate weight loss in overweight persons.

Most studies concerning the effects of longitudinal weight changes used self-reported body weight values or past weight loss episode by questionnaire, and were not free from information bias. The merits of our study were that we used actual measurements of the weight variables because these data were based on the physical examination in health check-ups of the study subjects. In addition, we could exclude those with heart disease, hypertension, hyperlipidemia or diabetes mellitus by the physician's medical records. These disease conditions and treatment might affect the weight and CRF changes during the follow-up.

Earlier large-scale studies did not control the effect of smoking habits,^11,14,16,17 yet smoking habits might influence CRF slope via the effects on BMI change. We analyzed the effect of smoking habit at the baseline on CRF changes and no correlation was found (the smoking rates of our population reflected a Japanese national health and nutrition survey²⁴). We could not obtain the longitudinal change of smoking habit during the study period, however it is unusual for middle aged-men and women to quit smoking in Japan.

In summary, our results suggest the different effects of weight change and fluctuation and we conclude that weight gain or loss has a much greater effect on CRF change than weight fluctuation. The idea of diet-related weight fluctuation hazards are presently premature and losing weight remains an important public health recommendation for the health of overweight individuals. Studies should also be conducted on subjects of different ethnicity and age, as well as on different magnitudes of weight change.

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

2 Higgins M, Kannel W, Garrison R, Pinsky J, Stokes J III. Hazards of obesity¾the Framingham experience. Acta Med Scand 1988; 723: (Suppl) 23-26,

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

4 Reisin E, Abel R, Modan M, Silverberg DS, Eliahou HE, Modan B. Effect of weight loss without salt restriction on the reduction of blood pressure in overweight hypertensive patients. New Engl J Med 1978; 298: 1-6, MEDLINE

5 Amatruda JM, Richeson JF, Weles SL, Brodows RG, Lockwood DH. The safety and efficacy of a controlled low-energy ('very-low-calorie') diet in the treatment of non-insulin-dependent diabetes and obesity. Arch Intern Med 1998; 148: 873-877,

6 DeFronzo RA. Lilly lecture 1987. The triumvirate: -cell, muscle, liver. A collusion responsible for NIDDM. Diabetes 1988; 37: 667-687, MEDLINE

7 Wolf RN, Grundy SM. Influence of weight reduction on plasma lipoproteins in obese patients. Arteriosclerosis 1983; 3: 160-169, MEDLINE

8 Wood PD, Stefanick ML, Dreon DM, Frey-Hewitt B, Garay SC, Williams PT, Superko HR, Fortmann SP, Albers JJ, Vranizan KM. Changes in plasma lipids and lipoproteins in overweight men during weight loss through dieting as compared with exercise. New Engl J Med 1988; 319: 1173-1179, MEDLINE

9 Stunkard AJ, Penick SB. Behavior modification in the treatment of obesity. The problem of maintaining weight loss. Arch Gen Psychiat 1979; 36: 801-806, MEDLINE

10 Brownell KD, Jeffery RW. Improving long-term weight loss: pushing the limits of treatment. Behav Ther 1987; 18: 353-374,

11 Lissner L, Odell PM, D'Agostino RB, Atokes J III, Kreger BE, Belanger AJ, Brownell KD. Variability of body weight and health outcomes in the Framingham population. New Engl J Med 1991; 324: 1839-1844, MEDLINE

12 Lissner L, Bengtsson C, Lapidus L, Larsson B, Bengtsson B, Brownell K. Body weight variability and mortality in the Gothenburg Prospective Studies of men and women. In: Bjorntorp P, Rossner S (eds). Obesity in Europe 1988: Proceedings of the First European Congress on Obesity. Libbey: London, 1989, pp 55-60.

13 Taylor CB, Jatulis DE, Fortmann SP, Kraemer HC. Weight variability effects: a prospective analysis from the Stanford Five-City Project. Am J Epidemiol 1995; 141: 461-465, MEDLINE

14 Jeffery RW, Wing RR, French SA. Weight cycling and cardiovascular risk factors in obese men and women. Am J Clin Nutr 1992; 5: 641-644,

15 Lee JS, Kawakubo K, Miyamoto K, Sasaki M. Body weight change and fluctuation effects on coronary risk factors in the Japanese middle aged men. Jap J Public Health 1999; 46: 373-379, (in Japanese).

16 Lissner L, Andres R, Muller DC, Shimokata H. Body weight variability in men: metabolic rate, health and longevity. Int J Obes 1990; 14: 373-383, MEDLINE

17 Stevens J, Lissner L. Body weight variability and mortality in the Charleston Heart Study. Int J Obes 1990; 14: 385-386, MEDLINE

18 Wing RR. Weight cycling in humans: overview of the literature. Ann Behav Med 1992; 14: 113-119,

19 French SA, Jeffery RW, Folsom AR, Williamson DE, Byers T. Weight variability in a population-based sample for older women: reliability and intercorrelation of measures. Int J Obes Relat Metab Disord 1995; 19: 22-29, MEDLINE

20 Itoh T, Horie S, Takahashi K, Okubo T. An evaluation of various indices of body weight changes and their relationship with coronary risk factors. Int J Obes Relat Metab Disord 1996; 20: 1089-1096, MEDLINE

21 van der Kooy K, Leenen R, Seidell JC, Deurenberg P, Hautvast JG. Effect of a weight cycle on visceral fat accumulation. Am J Clin Nutr 1993; 58: 853-857, MEDLINE

22 Muls E, Kempen K, Vansant G, Saris W. Is weight cycling detrimental to health? A review of the literature in humans. Int J Obes Relat Metab Disord 1995; 19: (Suppl 3) S46-S50, MEDLINE

23 Jeffery RW. Does weight cycling present a health risk? Am J Clin Nutr 1996; 63: (Suppl) 452S-455S, MEDLINE

24 The Ministry of Health and Welfare. Annual report on health and welfare 1996-1997. Japan International Corporation of Welfare Services: Tokyo, 1998,

Table 1 Ten year-transition from baseline year (1987) to end-point (1996)

Table 2 BMI variability indices in males and females

Table 3 Bivariate correlation between CRF slopes and BMI variability indices

Table 4 The association between BMI variables and each CRF slope by the multiple linear regression analyses

Table 5 The association between BMI variables and each CRF slope by the multiple linear regression analyses