ORIGINAL COMMUNICATION Obesity, Weight Change and Risks for Hypertension, Diabetes and Hypercholesterolemia in Japanese Men

Home , Hypercholesterolemia

European Journal of Clinical Nutrition (2002) 56, 601–607 ß 2002 Nature Publishing Group All rights reserved 0954–3007/02 $25.00 www.nature.com/ejcn ORIGINAL COMMUNICATION Obesity, weight change and risks for hypertension, diabetes and hypercholesterolemia in Japanese men

K Ishikawa-Takata1*, T Ohta2, K Moritaki3, T Gotou3 and S Inoue4

1Division of Health Promotion and Exercise, National Institute of Health and Nutrition, Tokyo, Japan; 2Chubu National Hospital, Japan; 3Human Resources Department, NTN Corporation, Japan; and 4Kyoritsu Women’s University, Japan

Objective: To examine the effects of the degree of body mass index and weight gain as the risks for hypertension, hypercholesterolemia and diabetes in Japanese men, and to compare that to the corresponding effects in a Caucasian population. Design: Prospective cohort study. Setting: Cohort of all male employees at a company. Subjects: A total of 4737 male employees followed until retirement or for 4 y (longest term 4 y). Results: Increase of body mass index beyond 22 kg=m2 related to an increased risk for hypertension even after being adjusted for possible confounding factors such as age, smoking status, alcohol intake, family history and baseline value of systolic blood pressure. The risks greatly increased in subjects with a body mass index above 27 kg=m2 for hypertension, and 29 kg=m2 for diabetes and hypercholesterolemia. Weight gain (more than 2 kg) was strongly related to an increased risk for hypertension and hypercholesterolemia after being adjusted for possible confounding factors. However, weight gain was not related to diabetes, and weight loss did not decrease the risks for any of the three diseases. Conclusions: Among Japanese, the degrees of body mass index associated with risks for hypertension, diabetes and hypercholesterolemia were lower than those in Caucasians. The risks for hypertension and hypercholesterolemia were strongly associated with weight gain in a Japanese male population who showed a low prevalence of severe obesity, and the risks were similar to or somewhat higher than those in a Caucasian population with a high prevalence of severe obesity. Sponsorship: Health Science Research Grant by the Ministry of Health and Welfare of Japan. European Journal of Clinical Nutrition (2002) 56, 601 – 607. doi:10.1038=sj.ejcn.1601364

Keywords: body mass index; weight gain; hypertension; diabetes; hypercholesterolemia

Introduction (relative risk, more than three times), and those of CHD and Obesity is often associated with type 2 diabetes, hyperten- hypertension are also increased (relative risk, two to three sion and hyperlipidemia, which consequently result in cor- times; WHO, 1998). Although there are several studies onary heart disease (CHD). According to a report of the examining the relationship between obesity and the inci- World Health Organization (WHO), risks for type 2 diabetes dence of these diseases, most of them have examined Cau- and dyslipidaemia are greatly increased in obese subjects casians (Huang et al, 1998; Chan et al, 1994; Colditz et al, 1990; Helmrich et al, 1991; Manson et al, 1992; Rimm et al, 1995). Since Asian populations, including Japanese, have a *Correspondence: K Ishikawa-Takata, Division of Health Promotion and much lower body mass index (BMI) than Caucasians (Flegal Exercise, National Institute of Health and Nutrition, 1-23-1 Toyama, et al, 1998; Liu et al, 1999; Ministry of Health and Welfare in Shinjuku, Tokyo 162-8636, Japan. Japan, 1996), an underlying question is how data on the E-mail: [email protected] Guarantors: K Ishikawa-Takata and T Ohta. relationship between BMI and the incidence of these dis- Contributors: KI-T was the lead investigator and wrote the paper. TO eases, and on the effect of weight gain on these diseases was the medical supervisor. KM and TG managed measurements and in Asian populations, compare to corresponding data for followed the subjects. SI participated in drafting and editing the Caucasians. paper. All investigators were involved in the study design. Received 1 March 2001; revised 8 October 2001; Several cross-sectional studies have indicated that the accepted 17 October 2001 relationship between an increase in BMI and mortality or Obesity, weight change and disease risk K Ishikawa-Takata et al 602 morbidity in Japanese was similar to that in Caucasians (Liu demia (Japan Arteriosclerosis Society, 1997). Overall, 3994, et al, 1999; Todoroki et al, 1994; Tokunaga et al, 1991; 4385 and 2995 men were analyzed for risk of hypertension, Nakayama et al, 1997). However, there is only one prospec- diabetes and hypercholesterolemia, respectively. tive study, to our knowledge, that has examined the effect of an increase in BMI on the incidence of hypertension and type 2 diabetes in Asian populations (Tai et al, 1992). Physical checkup Furthermore, the effect of weight gain during adulthood The annual physical checkup consisted of physical examina- has not been examined among Japanese, who have a histori- tions, anthropometric measurements (height and body cally low prevalence of severe obesity. weight), and measurements of blood pressure, fasting The present study was the first to prospectively examine plasma glucose and lipids. BMI was calculated by weight in the effects of an increase in BMI and weight gain on the kilograms divided by the height squared in meters. Resting incidence of hypertension, type 2 diabetes and hypercholes- blood pressure was obtained in duplicate after at least a 5 min terolemia in Japanese men initially without any of these rest using a sphygmomanometer. Venous blood was drawn abnormalities. for measurements after an overnight fast, and analyzed at a central laboratory (Mitsubishi Kagaku Bio-Clinical Labora- tories Inc., Tokyo, Japan). Each subject filled out a question- Subjects and methods naire of medical history family history and health-related Subjects behaviors such as physical activity, smoking and daily alco- Participants were male employees of a company with a head hol consumption. Physicians and nurses employed by the office, 36 branches, and 12 factories all over Japan. In 1994, company conducted all measurements. 8823 employees of that company (7560 male, 1263 female) took regular physical check-ups. Among the total of 7560 male employees, 14% (n ¼ 1058) were office workers, and the other 86% (n ¼ 6502) were factory workers. Sixty-one people Criteria of hypertension, diabetes and with a history of ischemic heart disease, cancer or cerebro- hypercholesterolemia vascular accident were excluded. In addition, we excluded Hypertension, diabetes and hypercholesterolemia were diag- 1066 men who worked alternate day and night shifts because nosed at the annual physical checkup. Hypertension was they showed a higher prevalence of comorbidity than day- defined by the criteria of JNC VI (systolic blood pressure time workers, as well as 1686 men with data missing from 140 and=or diastolic blood pressure 90 mmHg) or by the their baseline physical checkup. We also excluded 10 men use of antihypertensive medication (The Joint National with BMI of more than 35 kg=m2 at baseline because of their Committee on Prevention, Detection, Evaluation, and Treat- small number. This left 4737 qualified subjects, all of whom ment of High Blood Pressure, 1997). Diabetes was defined by underwent a regular annual physical checkup and were the criteria of the Japan Diabetes Association (fasting blood followed until retirement (1 – 4 y) or up to 1998, whichever glucose 126 mg=dl) or by taking medication for diabetes came first. (Japan Diabetic Society, 1999). Hypercholesterolemia was In an analysis of the incidence of hypertension, we defined by the criteria of the Japan Arteriosclerosis Society excluded a further 733 men who already had an above (total plasma cholesterol 220 mg=dl) or by taking medica- normal blood pressure at baseline among the 4737 men. tion for hyperlipidemia (Japan Arteriosclerosis Society, The normal range for blood pressure was defined as systolic 1997). < 130 mmHg and diastolic < 85 mmHg according to the criteria of the 6th report of the Joint National Committee of Detection and Evaluation for Hypertension (JNC VI), or Statistical analysis subjects not taking medication for hypertension (The Joint Person-years for each man for hypertension, diabetes or National Committee on Prevention, Detection, Evaluation hypercholesterolemia were estimated from the day of his and Treatment of High Blood Pressure, 1997). In an analysis first regular physical checkup in 1994 (baseline) to the day of the incidence of diabetes, we excluded 352 men who when one of these diseases was diagnosed, to his retirement already had above-normal fasting plasma glucose at baseline. year, or to 1998, whichever came first. Subjects were classi- The normal range for fasting plasma glucose was defined as fied into underweight (BMI < 18.5 kg=m2), normal (BMI < 110 mg=dl according to the criteria of the Japan Diabetic < 25 kg=m2), obesity I (BMI 25 kg=m2 and < 30 kg=m2), Society, or those not taking medication for diabetes (Japan and obesity II (BMI 30 kg=m2 and < 35 kg=m2) at baseline Diabetic Society, 1999). In the analysis of incidence of using the criteria of the Japan Society for the Study of hypercholesterolemia, we excluded another 1742 men who Obesity (Matsuzawa et al, 2000). Normal and obesity I sub- already had an above-normal total plasma cholesterol at jects were further categorized into 12 BMI groups using a baseline. The normal range for total plasma cholesterol was whole-number cut-off point. In addition, to examine the defined as < 200 mg=dl by the criteria of Japan Arteriosclero- effect of changes in body weight, men were classified into sis Society, or persons not taking medication for hyperlipi- three levels: weight change within 2 kg; and weight loss or

European Journal of Clinical Nutrition Obesity, weight change and disease risk K Ishikawa-Takata et al 603 gain of more than 2 kg as designated in the previous study Table 1 Baseline characteristics of the subjects (Huang et al, 1998). Mean s.d. (Min – max) Incidence rates were obtained by dividing the number of cases by person-years in each classiﬁed category (BMI or Age (y) 37.6Æ 10.4 (18 – 59) weight change). Body weight (kg) 64.2Æ 8.9 (42 – 102) Height (cm) 168.6Æ 6.2 (147 – 190) Multivariate Cox proportional hazards regression analysis BMI (kg=m2) 22.6Æ 2.8 (14.9 – 34.4) was used to adjust simultaneously for age (1 y categories), Systolic blood pressure (mmHg) 122.0Æ 13.9 (80.0 – 192.0) smoking status (never, past, current), alcohol intake (never, Diastolic blood pressure (mmHg) 72.8Æ 11.8 (32.0 – 124.0) less than three times per week, more than three times per Fasting blood glucose (mg=dl) 92.2Æ 18.4 (50 – 496) Total cholesterol (mg=dl) 190.7Æ 34.3 (93 – 413) week), family history (yes or no), and baseline measurements Current smoker (%) 61.7 (systolic blood pressure, fasting plasma glucose, or total Current drinker (%) 72.9 plasma cholesterol). In addition, relative risk was also Habitual physical exercise (%) 39.5 adjusted for BMI at baseline to examine the effect of body Family history of hypertension (%) 3.4 Family history of diabetes mellitus (%) 2.3 weight changes independent of the initial BMI. BMI, body mass index. n ¼ 4737.

Results Table 1 illustrates baseline characteristics according to BMI category at the start of this study. During a 4 y follow-up subjects with BMI of more than 27 kg=m2, even after adjust- period, newly diagnosed cases numbered 662 (46.7 per 1000 ing for possible confounding factors. person-years), 242 (14.7 per 1000 person-years), and 251 The age-adjusted risk for diabetes was not different (22.6 per 1000 person-years) for hypertension, diabetes and between subjects with BMI ranging from 18.5 to 27 kg=m2 hypercholesterolemia, respectively. and those with BMI < 18.5 kg=m2 (Table 3). However, sub- The age-adjusted risk for hypertension increased as BMI jects with BMI between 27 and 29 kg=m2 had a two-fold increased (Table 2). The risk in subjects with a BMI above higher relative risk than subjects with BMI < 18.5 kg=m2, and 20 kg=m2 was significantly higher than subjects with a BMI the risk was dramatically increased in subjects with BMI of below 18.5 kg=m2. The risk more than doubled with BMIs more than 29 kg=m2. After adjusting for the possible con- above 21 kg=m2, more than tripled with BMIs above founding factors, the multivariate risk was only slightly 24 kg=m2, and increased four-fold when BMI exceeded attenuated, but the higher risk remained significant in sub- 27 kg=m2. After adjustments for age, smoking status, alcohol jects with BMI above 29 kg=m2. intake, family history of hypertension and baseline value of The age-adjusted risk for hypercholesterolemia increased systolic blood pressure, the relative risk was only slightly as BMI increased (Table 4). The risks proved to be two-fold attenuated and remained significant in all subjects with BMIs higher in subjects with BMI above 21 kg=m2, and more than of more than 22 kg=m2. It showed a much higher risk for three times higher in subjects with BMI above 25 kg=m2. The

Table 2 Relative risks for hypertension according to the level of obesity

Relative risk Category Person- Cases of Rate per 1000 of BMI years hypertension person-years Age-adjusted (95% CI)a Multivariate (95% CI)b

< 18.5 878 16 18.2 1 (reference) 1 (reference) 18.5 522 14 26.8 1.63 (0.79 – 3.34) 1.75 (0.85 – 3.59) 19 1503 49 32.6 1.70 (0.97 – 3.00) 1.40 (0.80 – 2.47) 20 1932 65 33.6 1.82 (1.05 – 3.15) 1.46 (0.84 – 2.53) 21 2207 87 39.4 2.11 (1.23 – 3.59) 1.58 (0.92 – 2.69) 22 2055 99 48.2 2.43 (1.43 – 4.12) 2.03 (1.20 – 3.45) 23 1752 97 55.4 2.80 (1.65 – 4.76) 2.18 (1.28 – 3.71) 24 1216 79 65.0 3.21 (1.88 – 5.51) 2.26 (1.32 – 3.88) 25 831 57 68.6 3.39 (1.94 – 5.91) 2.31 (1.32 – 4.03) 26 611 39 63.8 3.25 (1.81 – 5.82) 2.20 (1.23 – 3.94) 27 259 24 92.7 4.66 (2.47 – 8.78) 3.40 (1.80 – 6.42) 28 223 20 89.7 4.90 (2.54 – 9.47) 2.89 (1.49 – 5.60) 29 83 8 96.4 4.60 (1.96 – 10.75) 2.19 (0.93 – 5.14) 30 and < 35 111 8 72.1 4.56 (1.95 – 10.68) 2.34 (1.00 – 5.50)

BMI, body mass index. aAdjusted for age (1 y categories). bAdjusted for age (1 y categories), smoking status (never, past, current), alcohol intake (never, less than three times per week, more than three times per week), family history, and baseline value of systolic blood pressure.

European Journal of Clinical Nutrition Obesity, weight change and disease risk K Ishikawa-Takata et al 604 Table 3 Relative risks for diabetes mellitus according to the level of obesity

Relative risk Category Person- Cases of Rate per 1000 of BMI years diabetes mellitus person-years Age-adjusted (95% CI)a Multivariate (95% CI)b

< 18.5 900 8 8.9 1 (reference) 1 (reference) 18.5 553 6 10.8 1.44 (0.50 – 4.17) 1.39 (0.48 – 4.02) 19 1564 20 12.7 1.32 (0.58 – 3.01) 1.35 (0.59 – 3.08) 20 2049 25 12.2 1.37 (0.62 – 3.04) 1.36 (0.61 – 3.02) 21 2443 23 9.4 1.01 (0.45 – 2.26) 0.97 (0.43 – 2.18) 22 2394 26 10.8 1.07 (0.48 – 2.36) 0.98 (0.44 – 2.16) 23 2092 35 16.7 1.58 (0.73 – 3.40) 1.47 (0.68 – 3.19) 24 1596 26 16.2 1.47 (0.66 – 3.25) 1.32 (0.59 – 2.94) 25 1102 24 21.7 2.03 (0.91 – 4.52) 1.71 (0.76 – 3.82) 26 752 14 18.6 1.79 (0.75 – 4.28) 1.52 (0.63 – 3.64) 27 371 11 29.6 2.61 (1.05 – 6.51) 1.96 (0.78 – 4.91) 28 352 8 22.7 2.39 (0.89 – 6.38) 1.85 (0.69 – 4.95) 29 139 8 57.5 6.33 (2.37 – 16.87) 5.16 (1.92 – 13.80) 30 and < 35 149 8 53.6 6.62 (2.38 – 17.65) 5.25 (1.96 – 14.04)

Table 4 Relative risks for hypercholesterolemia according to the level of obesity

Relative risk Category Person- Cases of Rate per 1000 of BMI years hyperchole-sterolemia person-years Age-adjusted (95% CI)a Multivariate (95% CI)b

< 18.5 790 8 10.1 1 (reference) 1 (reference) 18.5 475 6 12.6 1.32 (0.46 – 3.82) 1.13 (0.39 – 3.26) 19 1319 21 15.9 1.54 (0.68 – 3.49) 1.28 (0.56 – 2.90) 20 1686 30 17.7 1.80 (0.82 – 3.93) 1.47 (0.67 – 3.22) 21 1756 39 22.2 2.19 (1.02 – 4.68) 1.48 (0.69 – 3.18) 22 1653 45 27.2 2.67 (1.26 – 5.68) 1.62 (0.76 – 3.46) 23 1226 29 23.6 2.29 (1.04 – 5.03) 1.39 (0.63 – 3.05) 24 846 24 28.3 2.73 (1.22 – 6.08) 1.58 (0.71 – 3.53) 25 523 19 36.3 3.48 (1.52 – 7.97) 1.94 (0.85 – 4.46) 26 365 8 21.9 2.06 (0.77 – 5.49) 1.04 (0.39 – 2.79) 27 177 9 50.8 4.85 (1.87 – 12.60) 2.68 (1.02 – 7.01) 28 164 5 30.4 3.00 (0.98 – 9.19) 1.31 (0.42 – 4.02) 29 39 4 102.5 10.13 (3.04 – 33.66) 4.06 (1.21 – 13.60) 30 and < 35 59 4 67.7 6.40 (1.92 – 21.29) 2.44 (0.73 – 8.16)

BMI, body mass index. aAdjusted for age (1-y categories). bAdjusted for age (1-y categories), smoking status (never, past, current), alcohol intake (never, less than three times per week, more than three times per week), family history, and baseline value of fasting total cholesterol.

risks dramatically increased in subjects with BMI of more related to increased risk for hypertension and hypercholes- than 29 kg=m2. Although the risk was attenuated after multi- terolemia. The age-adjusted risk for these diseases in subjects variate adjustment, it remained significant in subjects with gaining more than 2 kg of body weight increased 1.26 times BMI above 29 kg=m2. for hypertension and 2.17 times for hypercholesterolemia The relative risks for hypertension, diabetes and hyperch- compared to subjects maintaining stable weight. The risk olesterolemia as classified by weight change are shown in was not altered after adjustment of baseline BMI in 1994. Table 5. Only 1.9% of subjects decreased their body weight After adjustments for age, BMI, smoking status and other by more than 5 kg, and 8.5% of subjects increased their body possible confounding factors, this relative risk was slightly weight by more than 5 kg during the 4 y follow-up period. attenuated, but still significantly increased in subjects gain- When we consider the group with stable weight (weight ing more than 2 kg. In contrast, the risk for diabetes was change 2 kg) as a referent, weight losses were not related not related to weight gain after adjusting for possible to the risk for these diseases. In contrast, weight gains were confounding factors.

European Journal of Clinical Nutrition Obesity, weight change and disease risk K Ishikawa-Takata et al 605 Table 5 Relative risks for hypertension, diabetes mellitus, and hypercholesterolemia according to weight change

Weight change since 1994 to endpoint

Loss > 2.0 kg Loss or gain 2.0 kg Gain > 2.0 kg

Hypertension Person-years 1629 8712 3828 Cases of hypertension 73 404 185 Rate per 1000 person-years 44.8 46.3 48.3 Relative risk: Age-adjusted (95% CI)a 0.96 (0.75 – 1.23) 1.00 (reference) 1.26 (1.05 – 1.50) Age, BMI-adjusted (95% CI)b 0.90 (0.70 – 1.16) 1.00 (reference) 1.26 (1.05 – 1.50) Multivariate (95% CI)c 0.87 (0.67 – 1.12) 1.00 (reference) 1.23 (1.03 – 1.47)

Diabetes mellitus Person-years 1895 10008 4551 Cases of diabetes mellitus 35 144 63 Rate per 1000 person-years 18.4 14.3 13.8 Relative risk: Age-adjusted (95% CI)a 1.27 (0.88 – 1.84) 1.00 (reference) 1.18 (0.87 – 1.59) Age, BMI-adjusted (95% CI)b 1.16 (0.80 – 1.68) 1.00 (reference) 1.15 (0.85 – 1.56) Multivariate (95% CI)c 1.16 (0.80 – 1.69) 1.00 (reference) 1.14 (0.85 – 1.54)

Hypercholesterolemia Person-years 1224 6746 3106 Cases of hypercholesterolemia 24 118 109 Rate per 1000 person-years 19.6 17.5 35.0 Relative risk: Age-adjusted (95% CI)a 1.08 (0.70 – 1.68) 1.00 (reference) 2.17 (1.66 – 2.83) Age, BMI-adjusted (95% CI)b 0.94 (0.61 – 1.47) 1.00 (reference) 2.08 (1.59 – 2.72) Multivariate (95% CI)c 0.94 (0.60 – 1.47) 1.00 (reference) 2.08 (1.59 – 2.73)

BMI, body mass index. aAdjusted for age (1 y categories). bAdjusted for age (1 y categories) and BMI at 1994 (continuous). cAdjusted for age (1 y categories), BMI at 1994 (continuous), smoking status (never, past, current), alcohol intake (never, less than three times per week, more than three times per week), family history, and baseline value of systolic blood pressure, fasting blood glucose, or total cholesterol.

Discussion the Study of Obesity (Matsuzawa et al, 2000), both of which In the present study, we found that an increase in BMI even were considered a low prevalence of obesity by the WHO in the normal-weight and obesity I groups was associated classification (WHO, 1998). Although excluded from the with an increased risk for hypertension and hypercholester- analysis, there were only 10 subjects with BMI above olemia. This remained true even after adjustments for several 35 kg=m2. The proportion of severe obesity was very much confounding factors. The risk was greatly increased in sub- smaller in our subjects than in the US population (Flegal jects with BMI of more than 27 kg=m2 for hypertension, and et al, 1998), as noted in the results of a National Nutrition more than 29 kg=m2 for diabetes and hypercholesterolemia. Survey in Japan (Ministry of Health and Welfare in Japan, Although weight gain (more than 2 kg) was related to an 1996). increased risk for hypertension and hypercholesterolemia According to the report of WHO, the risk for hypertension after adjustments for several confounding factors, it was was moderately increased (relative risk, two to three times) in not related to the risk for diabetes. Weight loss was not subjects with a BMI of more than 30 kg=m2 (WHO, 1998). related to risks for any of the three diseases. However, other studies found a two to three times higher risk The distribution of BMI in the present study was similar to in men with BMI more than 24 or 26 kg=m2 in US and that in the National Nutrition Survey and National Cross- Chinese populations (Willett et al, 1999; Huang et al, 1998; sectional Surveys on Circulatory Disorders in Japanese (Min- Tai et al, 1992). The present results indicated that the age- istry of Health and Welfare in Japan, 1996; Liu et al, 1999). adjusted risk for hypertension increased consecutively in Thus, our BMI distribution results can be considered repre- subjects with BMI more than 20 kg=m2: it became three sentative of the Japanese population. Among the subjects, times higher in those with BMI more than 24 kg=m2, and 11.6 – 16.6% were classified as obesity I (BMI 25 to four times higher in those with BMI above 27 kg=m2. < 30 kg=m2), and only 0.6 – 1.0% as obesity II (BMI 30 to Although the risk was attenuated after adjusting for several < 35 kg=m2) according to the criteria of the Japan Society for confounding factors, it remained significant in subjects with

European Journal of Clinical Nutrition Obesity, weight change and disease risk K Ishikawa-Takata et al 606 BMI above 22 kg=m2. Thus, we may presume that the In the present study, the effect of weight gain on the risk increased risk for hypertension in Japanese seems to require of diabetes was not significant. In Chan’s study, that risk for a lower BMI compared with that in Caucasian populations. subjects who gained 4.5 – 13.6 kg was only 1.7 times higher The WHO reports indicated that the risk for diabetes compared with that of the subjects who showed a weight increased three times in subjects with BMI more than change of less than 4.5 kg. Among our subjects, only 8.5% 30 kg=m2 (WHO, 1998). Willet et al (1999) reported that increased their body weight more than 5 kg. Then great gain the risk for diabetes was four times higher in men with of body weight or higher initial BMI might affect the increase BMI of 26 kg=m2 than those with BMI less than 21 kg=m2, of risk for diabetes. and that in men with BMI more than 29 kg=m2 the risk We found no study that examined the effect of weight was greatly increased. Tai et al (1992) found that the risk changes on the risk of hypercholesterolemia. The present for diabetes in Chinese men with BMI above 25.27 kg=m2 study indicated that gaining weight also significantly was 2.87 times higher than those with BMI below increased the risk for hypercholesterolemia (over two-fold 21.26 kg=m2. In the present study, the risk for diabetes in more than stable weight). The risk was higher than that for subjects with BMI less than 29 kg=m2 was not significant. hypertension and was not altered even after adjustment for However, it increased dramatically in subjects with BMI of initial body weight or other confounding factors. Therefore, more than 29 kg=m2. That dramatic increase is similar to we assume that weight gain is one of the major risk factor for the results of Willet et al (1999). On the other hand, the hypercholesterolemia. present study showed that risk for diabetes in subjects with The effect of weight gain on the risks discussed was BMI less than 29 kg=m2 is lower than that for Caucasians independent of the degree of BMI at the baseline for hyper- and Chinese. Although the reason for this difference was tension and hypercholesterolemia. These results suggest that unclear, differences in lifestyle or race might be important the effect of weight gains among Japanese who show a low factors. prevalence of severe obesity was similar to or somewhat To our knowledge, there is no study examining the effect higher than that in the US population, which shows a high of the obesity on the incidence of hypercholesterolemia. prevalence. Age-adjusted risk for hypercholesterolemia in subjects with We found that weight loss did not affect any of the three BMI from 25 to 29 kg=m2 was similar to the risk for dyslipi- diseases (1.9% of subjects decreased their body weight more demia in subjects with BMI more than 30 kg=m2 as shown in than 5 kg, but most lost less than 5 kg). Huang et al (1998) the 1998 report of WHO. In the present study, the risk for reported that a loss of more than 5 kg of weight was related subjects with BMI from 25 to 29 kg=m2 was not significant to a reduced risk for hypertension. We consider that the after adjustment for several confounding factors. That lack of reason for detecting no adverse association between weight significance was probably due to a correction of cholesterol loss and the risk for diseases in our subjects was the very at baseline by multivariate adjustment. However, the risk in small weight loss. subjects with BMI above 29 kg=m2 remained significant even In summary, the results suggested that it is beneficial for after adjustment for confounding factors, and it remained Japanese who show a low prevalence of severe obesity to more than four times higher. keep BMI within 18.5 and 25 kg=m2, and to maintain a stable There is a possibility that overweight younger and older body weight (within 2 kg change) for the prevention of adults have different risks for these diseases. Since the age hypertension, diabetes and hypercholesterolemia. distribution of our subjects was wide (18 – 59 y), we examined the interaction between age and BMI for evidence of risk of these diseases. The results showed that only hypertension had a positive interaction of age and BMI, suggesting References that overweight in older subjects posed more of a risk for Chan JM, Stampfer MJ, Rimm EB, Willett WC & Colditz GA (1994): Obesity, fat distribution, and weight gain as risk factors for clinical that disease than in younger subjects. On the other hand, we diabetes in men. Diabetes Care 17, 961 – 969. found no interaction between age and BMI on the risk for Colditz GA, Willett WC, Stampfer MJ, Manson JE, Hennekens CH, diabetes and hypercholesterolemia. Arky RA & Speizer FE (1990): Weight as a risk factor for clinical Recent studies described the effect of changes in body diabetes in women. Am. J. Epidemiol. 132,501– 513. Flegal KM, Carroll MD, Kuczmarski RJ & Johnson CL (1998): Over- weight on the incidence of obesity-related diseases (Huang weight and obesity in the United States: prevalence and trends, et al, 1998; Chan et al, 1994). As for the risk of hypertension, 1960 – 1994. Int. J. Obes. Relat. Metab. Disord. 22,39– 47. Huang et al (1998) indicated that, among women 30 – 55 y of Helmrich SP, Ragland DR, Leung RW & Paffenbarger RS (1991): age, the relative risk increased about 1.3 times in subjects Physical activity and reduced occurrence of non-insulin-dependent diabetes mellitus. New Engl. J. Med. 325, 147 – 152. with body weight gains of 2.1 – 4.9 kg compared with that in Huang Z, Willett WC, Manson JE, Rosener B, Stampfer MJ, Speizer FE subjects whose body weight changes were less than 2 kg & Colditz GA (1998): Body weight, weight change, and risk for during the long-term (12 – 50 y) and medium-term (2 – hypertension in women. Ann. Intern. Med. 128,81– 88. 14 y). The effects of weight gain on the incidence of hyper- Japan Arteriosclerosis Society (1997): Guideline for diagnosis and treatment of hyperlipidemia. TJ. Jap. Atheroscler. Soc. 25,1– 34. tension in our study was similar to those in Huang et al Japan Diabetic Society (1999): Diagnosis. In Guidelines for Treatment (1998). of Diabetes, ed. Japan Diabetic Society, pp 12 – 17. Tokyo: Bunkodo.

European Journal of Clinical Nutrition Obesity, weight change and disease risk K Ishikawa-Takata et al 607 Liu L, Choudhury SR, Okayama A, Hayakawa T, Kita Y & Ueshima H Tai T, Chuang L, Wu H & Chen C (1992): Association of body build (1999): Changes in body mass index and its relationships to other with non-insulin-dependent diabetes mellitus and hypertension cardiovascular risk factors among Japanese population: results among Chinese adults: a 4-year follow-up study. Int. J. Epidemiol. from the 1980 and 1990 National Cardiovascular Surveys in 21, 511 – 517. Japan. J. Epidemiol. 9, 163 – 174. The Joint National Committee on Prevention, Detection, Evaluation, Manson JE, Nathan DM, Krolewski AS, Stampfer MJ, Willett WC & and Treatment of High Blood Pressure (1997): The sixth report of Hennekens CH (1992): A prospective study of exercise and inci- The Joint National Committee on Prevention, Detection, Evalua- dence of diabetes among US male physicians. J.A.M.A. 268,63– tion, and Treatment of High Blood Pressure. Arch. Intern. Med. 157, 67. 2413 – 2446. Matsuzawa Y, Inoue S, Ikeda Y, Sakata T, Saitou Y, Satou Y, Sirai K, Todoroki I, Shinchi K, Kono S & Imanishi K (1994): Lifestyle and Ohno M, Miyazaki S, Tokunaga M, Fukagawa K, Yamanouchi K & glucose tolerance: a cross-sectional study of Japanese men. Ann. Nakamura T (2000): New diagnosis and classiﬁcation of obesity. Epidemiol. 4,363– 368. J. Jap. Soc. Study Obes. 6,18– 28. Tokunaga K, Matsuzawa Y, Kotani K, Keno Y, Kobatake T, Fujioka & Ministry of Health and Welfare in Japan (1996): National Nutrition Tarui S (1991): Ideal body weight estimated from the body mass Survey in 1994. Tokyo: Daiichi Shuppan. index with the lowest morbidity. Int. J. Obes. 15,1– 5. Nakayama K, Koyohara Y, Kato I, Iwamoto H, Ueda K & Fujishima M. WHO (1998): The health consequences of overweight and obesity in (1997): Effect of body mass index on morbidity and mortality in a adults and children. In Obesity, Preventing and Managing the Global general Japanese population — The Hisayama Study. Jpn. J. Geriat. Epidemic,pp43– 72. Geneva: World Health Organization. 34, 935 – 941. Willett WC, Dietz WH & Colditz GA (1999): Guidelines for healthy Rimm EB, Stampfer MJ, Giovannucci E, Ascherio A, Spiegelman D, weight. New Engl. J. Med. 341, 427 – 434. Colditz GA & Willett WC (1995): Body size and fat distribution as predictors of coronary heart disease among middle-aged and older US men. Am. J. Epidemiol. 141, 1117 – 1127.

European Journal of Clinical Nutrition