PAPER Associations between weight gain and incident ... - Nature

International Journal of Obesity (2002) 26, 58–64 ß 2002 Nature Publishing Group All rights reserved 0307–0565/02 $25.00 www.nature.com/ijo

PAPER Associations between weight gain and incident hypertension in a bi-ethnic cohort: the Atherosclerosis Risk in Communities Study Juhaeri1,2, J Stevens1,3*, LE Chambless4, HA Tyroler1, W Rosamond1, FJ Nieto5, P Schreiner6, DW Jones7 and D Arnett6 1

Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, North Carolina, USA; Department of Statistics and Primate Research Center, Bogor Agricultural University, Bogor, Indonesia; 3Department of Nutrition, School of Public Health, University of North Carolina, Chapel Hill, North Carolina, USA; 4Department of Biostatistics, School of Public Health, University of North Carolina, Chapel Hill, North Carolina, USA; 5Johns Hopkins School of Hygiene and Public Health, Baltimore, Maryland, USA; 6Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, USA; and 7Division of Hypertension and the Center for Excellence in Cardiovascular-Renal Research, The University of Mississippi Medical Center, Jackson, Mississippi, USA 2

OBJECTIVE: To examine associations between weight gain and changes in blood pressure and the incidence of hypertension in four ethnicity-gender groups. DESIGN: Longitudinal closed cohort studied over an average of 6 y. SUBJECTS: Total of 9309 white and African-American men and women 45 – 64 y of age who participated in the Atherosclerosis Risk in Communities (ARIC) Study. METHODS: Weight and blood pressure were measured at baseline and after an average of 3 and 6 y of follow-up. Proportional hazard models with weight gain as a time-dependent variable were used to examine the association between weight gain and changes in blood pressure and hypertension. Multivariate models were used with baseline SBP, DBP, age, BMI, height, WHR, smoking, physical activity, education, caloric intake, fat intake and study center as covariates. RESULTS: Weight gain was associated with increases in SBP and DBP in all groups. Hazard ratios for hypertension associated with 1 kg annual weight gain were 1.36 (95% CI, 1.29, 1.45) in white women, 1.12 (95% CI, 1.03, 1.21) in African-American women, 1.35 (95% CI, 1.27, 1.43) in white men and 1.43 (95% CI, 1.27,1.61) in African-American men. CONCLUSION: Weight gain was associated with increased blood pressure and increased incidence of hypertension. The association was weaker among African-American women compared to other ethnicity-gender groups. International Journal of Obesity (2002) 26, 58 – 64. DOI: 10.1038=sj=ijo=0801846 Keywords: weight change; incident hypertension; ethnicity; epidemiology; prospective study

Introduction Both cross-sectional1,2 and prospective3 – 6 studies have shown that obesity is associated with hypertension. Clinical trials also indicate that weight loss is associated with decreased blood pressure.7,8 In these clinical trials, partici-

*Correspondence: J Stevens, Department of Epidemiology, School of Public Health, CB# 7400 McGavran-Greenberg Hall, University of North Carolina Chapel Hill, Chapel Hill, NC 27599-7400, USA. E-mail: [email protected] Received 10 November 2000; revised 27 April 2001; accepted 6 July 2001

pants were limited to highly selected overweight subjects, and thus the generalizability is limited. Also, the populations studied were predominantly white, and results may not be applicable to other groups. Finally, the studies were conducted over relatively short time periods ( < 18 months) and the long-term effects (ie effects over a period of years) could not be evaluated. A limited number of population-based studies have examined the association of weight gain with hypertension. Kannel et al showed that weight gain was associated with the subsequent increase in the incidence of hypertension in 5127 white men and women aged 30 – 62 y participating in

Weight gain and hypertension Juhaeri et al

The Framingham Heart Study.4 Similar results were found in The Nurses Health Study cohort, a group of 82 473 predominantly white female nurses aged 30 – 55 y followed every 2 y from 1976 until 1992.6 In this study, long-term (12 – 50 y) and short-term (2 – 14 y) weight gain was associated with a subsequent increase in the incidence of hypertension.6 Using data from the same cohort with a later follow-up period (1993 – 1995), weight gain between 1989 and 1993 was associated with an increased risk of hypertension between 1993 and 1995 with a risk ratio of 1.20 (95% Confidence Interval (CI), 1.15, 1.24) for each 10 lb gain in weight.9 To the best of our knowledge, associations between weight gain and an increased blood pressure have been compared in African-American and white people only in a cohort from Evans County, GA, initially examined between 1960 and 1962.10 Daniels et al, showed that weight gain over a 7 y period was directly associated with increased systolic blood pressure (SBP) and diastolic blood pressure (DBP); however, the regression coefficients among white men and women and African-American men were twice as large as those among African-American women.10 Using data from the same cohort and the same follow-up period, Tyroler et al showed that weight gain was associated with a higher incidence of hypertension, and the effects of weight gain on the incidence and remission of hypertension were greater in white than in African-American people.11 The work presented here uses longitudinal data from the Atherosclerosis Risk in Communities (ARIC) cohort to examine associations between weight gain and the incidence rate of hypertension in African-American and white men and women, over a 6 y period. We also examined associations between gain in weight and gain in blood pressure over the same time period in this bi-racial cohort. There have been only a few studies that examined the association between weight gain and the incidence of hypertension, especially among African-American people. The unique contribution of this study to the literature is its ability to examine and to compare the associations in African-American and white people.

Materials and methods The ARIC study is a multicenter prospective study of atherosclerosis and cardiovascular disease.12 Baseline examinations were conducted between 1987 and 1990 in four US communities: Forsyth County, North Carolina (12% AfricanAmerican; 87.5% white); Jackson, Mississippi (100% African-American); the northwestern suburbs of Minneapolis, Minnesota (99.1% white); and Washington County, Maryland (98.9% white). Subjects participate in examinations with a planned interval of 3 y between examinations. We used data from examinations 1 (baseline), 2 and 3. Approximately 46% of persons who were eligible in Jackson and 65% in the other three communities were recruited into the cohort. The characteristics of non-responders were described in a previous report.13 Non-responders

59 were those who completed a home interview but not a clinic visit. People who refused an interview and did not have a clinic visit were not characterized in this study. The difference of the prevalence of self-reported hypertension between responders and non-responders varied by ethnicity and gender. Among white women, the prevalence of self-reported hypertension was higher among non-responders than among responders; whereas among African-American men, the prevalence was higher among responders.13 There was no difference in the prevalence of self-reported hypertension between responders and non-responders among white men and African-American women.13 The first cohort examination began in 1986 and finished in 1989. Of 15 792 participants who attended the baseline visit, 1233 (7.8%) were excluded because they did not complete both visits 2 and 3. We also excluded participants with ethnicities other than African-American or white (n ¼ 42), and African-Americans from the Washington County and Minneapolis field centers (n ¼ 49) because they were too small in number to allow ethnic and center-specific analyses. Participants who were hypertensive at baseline (defined as SBP of 140 mmHg, DBP of 90 mmHg, or self-reported history of antihypertensive drug use in the past 2 weeks; (n ¼ 4872) were then excluded. We also excluded participants who had missing information for SBP, DBP or weight at visit 1 or at both visits 2 and 3 (n ¼ 16). Those who had missing other pertinent variables at baseline (age, height, waist-to-hip ratio (WHR)), antihypertensive medication, total fat intake, total caloric intake, cigarette smoking, education and physical activity) were also excluded (n ¼ 241). Finally, we excluded participants with implausible weight or height data (n ¼ 20) detected by visual inspection of a series of plots. Our final sample size was 9319 (4164 white women, 972 African-American women, 3588 white men, and 595 African-American men).

Data collection Blood pressures were measured while participants were sitting and after 5 min rest using a standardized Hawksley random-zero sphygmomanometer. The first and fifth phase Korotkoff sounds are the criteria for SBP and DBP, respectively. Technicians conducted these measurements after completing training with Korotkoff sound tapes and double stethoscope. Training was targeted to reduce digit preference and systematic differences between technicians. For this study, we used the average of the second and third of three measurements. Weight was measured without shoes with participants wearing light-weight and non-constricting underwear and having an empty bladder. Weight was recorded to the nearest pound (lb; rounding down) and measured with the participants standing on a balance scale (Detecto, model no. 437). Pounds were then transformed into kilgrams (kg). Height was recorded to nearest centimeter (cm; rounding down) and measured without shoes using a metal centimeter ruler International Journal of Obesity


60 mounted to a wall with participants standing and looking straight ahead. Waist and hip circumferences were measured with anthropometric tape to the nearest centimeter with the subjects standing. The waist circumference was measured at the level of the umbilicus and hips at the level of the maximum protrusion of the gluteal muscles. Total usual nutrient intakes were calculated using a 66item food frequency questionnaire which was a modified version of the 62-item food frequency questionnaire developed by Willett et al.14,15 Ethnicity was assessed using selfidentification of race from a checklist. Physical activity was estimated based on type, duration and the intensity of sports during leisure time and categorized into three levels: light, moderate and heavy.16 Education was a categorical variable with three levels: less than high school, high school graduate, college or higher. Smoking was a categorical variable with three levels: current smoker, former smoker and never smoker. Smoking data were ascertained using two questions: ‘Have you ever smoked cigarettes?’ and ‘Do you now smoke cigarettes?’

Definitions of exposure and outcomes Weight gain was calculated for each participant as the difference in weight from visit 1 to visit 2 and from visit 1 to visit 3. Similarly, blood pressure gain was calculated as the difference in blood pressure from visit 1 to visit 2 and from visit 1 to visit 3. Incident hypertension was assessed at visits 2 and 3. Participants who reported taking antihypertensive drugs within 2 weeks preceding the two visits were classified as hypertensive. Participants not taking medication were classified as hypertensives if their DBP 90 mmHg and their DBP increased by at least 7 mmHg or their SBP 140 mmHg and their SBP increased by at least 11 mmHg from the preceding visit measurement. A minimum change in DBP or SBP was used to reduce misclassification of hypertensive status due to measurement error by reducing diagnoses of incident hypertension in participants with baseline values near the cut-off values and small increase in blood pressure. The criterion of 7 and 11 mmHg were set to approximate one standard deviation of the change in the respective variables from visit 1 to visit 2 and from visit 2 to visit 3. From visit 1 to visit 2, the standard deviations of DBP and SBP changes in this population were 7.2 and 11.3 mmHg, respectively. From visit 2 to visit 3, the corresponding standard deviations were 7.9 and 12.6 mmHg. We corrected for the measurement error in blood pressures by replacing the baseline blood pressures with the Stein estimates of the true values of these baseline blood pressures.17 Stein estimates of the blood pressures were calculated as the function of the observed values of the blood pressures and their repeatability coefficients. We used these Stein estimates to define SBP and DBP changes and the incidence of hypertension. These estimates were also used in the multivariate models as covariates. In these corrections International Journal of Obesity

for measurement error, we used repeatability coefficients of 0.75 and 0.62 for baseline SBP and DBP, respectively. These values were from a sample of 363 ARIC participants with repeat measurements 1 – 2 weeks apart during ARIC visit 3.

Statistical analysis Unadjusted and adjusted associations between weight gain and gain in blood pressure were examined using mixed model analyses,18,19 and blood pressure gain associated with one unit gain in weight (kg) were calculated. For hypertension, proportional hazard models with time-dependent weight gain were used to calculate adjusted and unadjusted hazard ratios with 95% CI.20 Several covariates were selected for inclusion in all models a priori: baseline age, study center, SBP, DBP, body-mass index (BMI), height, WHR, smoking status, educational level, total fat intake, total caloric intake and physical activity. Baseline SBP and DBP were included in models that examined gains in that measure. Analyses were stratified by ethnicity and gender. To examine the effect modifications by ethnicity and gender, interaction terms were tested in models where all ethnicity – gender groups were combined. Effect modifications by age, BMI, and pre-existing disease (cardiovascular diseases (CVD), diabetes or cancer) were also examined by testing interaction terms in the same models. Analyses were performed using the Statistical Analysis System (SAS) procedures PROC MIXED and PHREG for blood pressure gain and hypertension, respectively.21

Results Over a 6 y period, weight gain tended to be larger among women than men (Table 1). At baseline, African-American women had the highest mean BMI, followed by white men, African-American men, and white women. Mean SBP and DBP were higher in African-American than white participants of the same gender. White men had the highest mean fat and energy intakes. Table 2 shows the unadjusted and adjusted annual increase in blood pressure (mmHg=y) for one unit (kg=y) increment of annual weight gain by ethnicity and gender. Results are presented including and then excluding participants who took blood pressure medication at follow-up examinations. Over all ethnicity gender groups, there were 947 participants who reported taking blood pressure medication at either visit 2 or visit 3 (350 white women, 185 African-American women, 325 white men and 87 African-American men). Weight gain was associated with SBP gain in all groups. For example, among white men, after controlling for covariates, an increment of one unit (kg=y) in annual weight gain was associated with an annual gain of 0.67 mmHg= y in SBP. The associations were statistically different across ethnicity — gender groups (P < 0.001) with African-American women having a lower annual gain in SBP than seen in other groups. When those who reported blood pressure medication were excluded,


61 Table 1 Mean (standard deviation (s.d.)) and frequencies of selected baseline characteristics by ethnicity and gendera White women (n ¼ 4164) Weight gain, visits 1 – 2 (kg) Weight gain, visits 1 – 3 (kg) SBP (mmHg) DBP (mmHg) Age (y) 2 BMI (kg=m ) Height (cm) WHR Total fat (g) Total calories (kcal) Smoking status (%) current former never Educational level (%) < high school high school graduate > high school graduate Physical activity (%) low moderate high

3.0 6.6 111.2 67.6 53.2 25.8 162.2 0.88 55.0 1491

(10.1) (12.5) (12.8) (8.4) (5.6) (4.9) (6.0) (0.08) (23.4) (517)

African-American women (n ¼ 972) 3.4 6.0 116.5 73.0 52.0 29.5 163.4 0.88 54.6 1523

(10.8) (15.1) (11.8) (8.0) (5.6) (5.9) (6.0) (0.08) (23.9) (583)

White men (n ¼ 3588) 1.9 4.1 114.9 71.0 54.1 26.9 176.4 0.96 68.3 1798

(9.1) (11.5) (11.5) (8.3) (5.6) (3.7) (6.5) (0.05) (29.7) (647)

African-American men (n ¼ 595) 2.4 3.7 117.7 75.6 52.7 26.7 176.6 0.93 64.3 1794

(9.3) (11.9) (11.0) (7.8) (5.7) (4.3) (6.8) (0.06) (26.2) (665)

25.0 24.7 50.2

25.1 18.3 56.6

25.1 45.8 29.2

38.3 32.6 29.1

13.0 41.6 45.4

31.8 22.1 46.1

16.1 28.8 55.2

38.5 19.5 42.0

38.9 48.2 13.0

57.4 36.5 6.1

28.6 49.6 21.8

47.6 41.9 10.6

a

For continuous variables, F-test was used to test the mean difference ethnicity-gender groups, whereas for categorical variables we used chi-square test; the characteristics were different across ethnicity – gender group (P < 0.010).

Table 2

Mean annual increase in SBP and DBP (95% CI) for one unit (kg=y) increment of annual weight gain by ethnicity and gender White women

African-American women

White men

African-American men

SBP (mmHg=y) Participants with and without blood pressure medication at visit 2 or visit 3 combined unadjusted adjusteda Participants without blood pressure medication unadjusted adjusted

0.60 (0.54 – 0.66) 0.48 (0.42 – 0.54)

0.49 (0.37 – 0.61) 0.38 (0.26 – 0.50)

0.75 (0.69 – 0.81) 0.67 (0.61 – 0.73)

0.63 (0.47 – 0.79) 0.57 (0.39 – 0.75)

0.57 (0.51 – 0.63) 0.48 (0.42 – 0.54)

0.47 (0.35 – 0.59) 0.38 (0.24 – 0.52)

0.70 (0.64 – 0.76) 0.64 (0.58 – 0.70)

0.56 (0.38 – 0.74) 0.51 (0.33 – 0.69)

DBP (mmHg=y) Participants with and without blood pressure medication at visit 2 or visit 3 combined unadjusted adjusted Participants without blood pressure medication unadjusted adjusted

0.27 (0.23 – 0.31) 0.28 (0.24 – 0.32)

0.14 (0.06 – 0.22) 0.23 (0.15 – 0.31)

0.41 (0.37 – 0.45) 0.43 (0.39 – 0.47)

0.26 (0.14 – 0.38) 0.34 (0.22 – 0.46)

0.27 (0.23 – 0.31) 0.29 (0.25 – 0.33)

0.11 (0.03 – 0.19) 0.21 (0.13 – 0.29)

0.39 (0.35 – 0.43) 0.40 (0.36 – 0.44)

0.25 (0.13 – 0.37) 0.33 (0.21 – 0.45)

a Adjusted for baseline SBP (SBP change models), DBP (DBP change models), age, BMI, height, WHR, smoking, physical activity, education, total fat intake, total caloric intake and study center.

results tended to be the same in all groups. As with SBP, weight gain was associated with DBP gain. The associations differed across ethnicity — gender groups (P < 0.001). The difference might be due to the gender difference in both African-American and white people. There were 1839 incident cases of hypertension over a 6 y follow-up period (visit 1 – visit 3; Table 3). Of these, 972 cases occurred between visits 1 and 2, and 867 between visits 2 and 3. In all intervals, the cumulative incidence was higher in African-American than in white people (P ¼ 0.001).

Associations between weight gain and hypertension over visits 1 – 3 varied by ethnicity and gender (P < 0.05; Table 4). After controlling for the covariates, weight gain was associated with an increased incidence rate of hypertension in all groups. The hazard ratio, however, was smaller in AfricanAmerican women than in other groups. We found no effect modification by pre-existing disease. Although not statistically significant, the associations tended to be stronger in younger (age < 55 y) than in older groups (age 55 y) in white women, white men, and African-American men, but International Journal of Obesity


62 Table 3

Prevalence and cumulative incidence of hypertension by ethnicity and gendera

b

Prevalence (%) at visit 1 c Incidence (%) within visits 1 – 2 Incidence (%) within visits 2 – 3

White women


White men


26.2 (1580) 7.8 (322) 10.2 (361)

56.8 (1487) 17.4 (163) 20.2 (141)

28.6 (1541) 1.0 (393) 9.8 (284)

54.6 (887) 16.4 (94) 18.8 (81)

a

Number of cases in parentheses. Prevalent hypertension at visit 1 was excluded from the study. Incident hypertension between visits 1 and 2 was excluded from incidence measurement between visits 2 and 3.

b c

Table 4 Hazard Ratios (95% CI) of hypertension for one unit (kg=y) increment of annual weight gain by ethnicity and gender

Unadjusted a Adjusted

White women


White men


1.20 (1.15, 1.25) 1.36 (1.29, 1.45)

1.06 (0.98, 1.15) 1.12 (1.03, 1.21)

1.28 (1.21, 1.36) 1.35 (1.27, 1.43)

1.36 (1.21, 1.53) 1.43 (1.27, 1.61)

a Adjusted for baseline DBP, SBP, age, BMI, height, WHR, smoking, education, physical activity, total fat intake, total caloric intake and study center.

not in African-American women (results not shown). We found effect modification by BMI among African-American women (P-value of interaction term of 0.007), but not among other groups. Although significant, there was little difference in the association between weight gain and hypertension among obese (BMI 30 kg=m2) and non-obese African-American women (BMI < 30 kg=m2) with hazard ratios of 1.16 (95% CI, 1.03, 1.31) and 1.07 (95% CI, 0.94, 1.23) among obese and non-obese, respectively. When the level of risk at the reference level of the exposure varies among groups, the type of measures of effect used may result in different conclusions. To address this issue we examined the associations across groups using risk difference. Based on the annual weight change from baseline to visit 3, we categorized participants as follows: (1) weight loss, an annual weight change of 7 0.5 kg or lower; (2) stable weight (reference), an annual weight change between 7 0.5 and 0.5 kg; and (3) weight gain, an annual weight change of at least 0.5 kg. The risk difference of hypertension between weight gain and stable weight among African-American women was 2.8% (33.6 vs 30.8%). The corresponding risk differences among white women, white men and African-American men were 4.1% (18.2 vs 14.1%), 7.1% (23.0 vs 15.9%), and 12.1% (39.0 vs 27.6%), respectively. Therefore, in this study, both hazard ratio and risk difference resulted in the same conclusion, ie that the association between weight change and hypertension was weaker among African-American women than other groups.

Discussion In the ARIC cohort, weight gain was associated with SBP and DBP gains in all ethnicity — gender groups. A summary of five randomized clinical trials among white participants over short-term period (average 3 months) by MacMahon et al reported that a decrease in weight of 9.2 kg was associated International Journal of Obesity

with a decrease in SBP of 6.3 and a decrease in DBP of 3.1 mmHg.7 Another study with a longer follow-up (18 months) showed that, among men, a decrease in weight of 4.7 kg was associated with a decrease in SBP of 5.8 and a decrease in DBP of 6.4 mmHg.8 Our study showed that, over a 6 y period, a gain in weight of 1 kg=y was associated with an increase in SBP of 0.64 and an increase in DBP of 0.40 mmHg=y in white men. Theoretically, this would lead to a decrease in SBP of 5.9 mm Hg and a decrease in DBP of 3.7 mmHg for a 9.24 kg decrease in weight. Thus the shortterm results of the clinical trials were similar in magnitude to those seen in white men in this relatively long-term (6 y) observational study. In those clinical trials, however, they examined the association between weight loss and blood pressure loss, whereas in this study we examined the association between weight gain or loss and blood pressure change. We found that the association of weight gain with gains in SBP and DBP differed significantly across groups. Within ethnic groups, the impact of weight gain on blood pressure appeared larger in men than in women. African-American women had a weaker association between weight gain and gain in SBP than other ethnicity and gender groups, whereas for DBP, white men showed a stronger association than the other groups. After excluding participants on antihypertensive medications, changes in both SBP and DBP associated with weight change tended to be smaller in African-American women than in the other groups. This difference was smaller than that seen in the Evans County Study, which found associations between quetelet index gain and gains in both SBP and DBP were twice as great among white women, white men, and African-American men as in African-American women.10 This study did show a significantly smaller association between weight gain and hypertension among AfricanAmerican women compared to other groups. A smaller association among African-American people is partly due to


63 a much higher base incidence rate of hypertension among this group compared to others. A study that compared associations between weight gain and the incidence of hypertension in African-American and white people the Evans County Study, found that the effects of weight gain on the incidence of hypertension were greater in white than in African-American people; the effect of gender, however, was not examined in this study.11 Factors that may explain the differences are a different reactivity to hypertension risk factors, differences in blood pressure medication use and different psychosocial stressors.22 The fact that the prevalence of hypertension was higher in African-American than in white people at baseline may partly explain in the reduced effect of weight gain on hypertension observed in African-American women. As shown in Table 3, the prevalence of hypertension in both African-American men and women was approximately twice that of white men and women at baseline. However, the incidence of hypertension over the follow-up period remained approximately twice as large in African-American as in white people. Therefore, it does not appear that early incidence of hypertension (ie before baseline) resulted in lower incidence rates over the follow-up period. Nevertheless, we cannot rule out the possibility that more ‘weightsensitive’ African-American women may have developed hypertension prior to baseline and thus the effect of weight gain on the incidence of hypertension over the study period was reduced. We found effect modification by BMI only among AfricanAmerican women. There was a significant association among obese but not among non-obese African-American women. Although significant, the difference in hazard ratios between obese and non-obese African-American women was small. The roles of BMI on the association between weight gain and hypertension may be different in African-American women than in other groups. It is likely that variables that contribute to differences in the pathogenesis of hypertension in AfricanAmerican women were not assessed or controlled for here. Except for African-American women, the associations tended to be greater in younger ( < 55 y) than in older ( 55 y) groups. Among African-American women, the associations were almost the same in both age groups. Results in groups other than African-American women were similar to those from the Nurses’ Health Study.6 In the Nurses’ Health Study, participants were categorized as younger ( < 45 y of age), middle-aged (45 – 55 y), and older ( 55 y) women. This difference in the strength of the association between the younger and older groups may be because changes in weight may better reflect changes in fat mass in the younger than in the older participants. Many elderly individuals lose a substantial amount of lean body mass; consequently, variation in weight change in the older group is, to greater degree, attributable to variation in lean body mass and reduces the validity of weight change as a measure of adiposity change.23 The mechanisms by which weight gain is associated with hypertension are unclear. It has been hypothesized than fat

mass might play an important role in increasing blood pressure by raising cardiac output with relatively increased peripheral vascular resistance.24,25 It has also been suggested that increased sympathetic nervous system activity, insulin resistance and hyperinsulinemia, sodium retention and enhanced vascular reactivity are involved in the development of hypertension.26 Weight gain could impact hypertension through these mechanisms. More work needs to be done to improve our understanding of the ethnic and gender related trends seen here in the associations between weight gain and rises in blood pressure and the incidence of hypertension. Nevertheless, it is clear that, over a 6 y period in a community-based sample of middle-aged adults, weight gain was associated with increased SBP and DBP and an increased incidence of hypertension in all ethnicity-gender groups examined. Prevention of weight gain clearly is important to public health and the prevention of hypertension in African-American as well as white men and women.

Acknowledgements The ARIC Study was funded by contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC55020, N01-HC-55021 and N01-HC-55022 from the US National Heart, Lung, and Blood Institute. The authors thank the staff and participants in the ARIC study for their important contributions.

References 1 Croft J, Strogatz D, Keenan N, James S, Malarcher A, Garrett J. The independent effects of obesity and body fat distribution on blood pressure in black adults: the Pitt County study. Int J Obes Relat Metab Disord 1993; 17: 391 – 397. 2 Stamler R, Stamler J, Riedlinger W, Algera A, Roberts R. Weight and blood pressure: findings in hypertension screening of 1 million Americans. JAMA 1978; 240: 1607 – 1612. 3 Ascherio A, Rimm E, Giovannucci E, Colditz G, Rosner B, Willett W, Sacks S, Stampfer M. A prospective study of nutritional factors and hypertension among US men. Circulation 1992; 86: 1475 – 1484. 4 Kannel W, Brand N, Skinner J, Dawber T, McNamara P. The relation of adiposity to blood pressure and development of hypertension. Ann Intern Med 1967; 67: 48 – 59. 5 Cassano P, Segal M, Vokonas P, Weiss S. Body fat distribution, blood pressure, and hypertension: a prospective cohort study of men in the Normative Aging Study. Ann Epidemiol 1990; 1: 33 – 48. 6 Huang Z, Willett W, Manson J, Rosner B, Stampfer M, Speizer F, Colditz G. Body weight, weight change, and risk of hypertension in women. Ann Intern Med 1998; 128: 81 – 88. 7 MacMahon S, Cutler J, Brittain E, Higgins M. Obesity and hypertension: epidemiological and clinical issues. Eur Heart J 1987; 8(Suppl B): 57 – 70. 8 Stevens VJ, Corrigan S, Obarzanek E, Bernauer E, Cook NR, Hebert P, Mattfeld-Beman M, Oberman A, Sugars C, Dalcin A, Whelton P. Weight loss intervention in phase I of the Trials of Hypertension Prevention. Arch Intern Med 1993; 153: 849 – 858. 9 Field A, Byers T, Hunter D, Laird N, Manson J, Williamson D, Willett W, Colditz G. Weight cycling, weight gain, and risk of hypertension in women. Am J Epidemiol 1999; 150: 573 – 579.

International Journal of Obesity


64

10 Daniels S, Heiss G, Davis C, Hames C, Tyroler H. Race and sex differences in the correlates of blood pressure change. Hypertension 1988; 11: 249 – 265. 11 Tyroler H, Heyden S, Hames C. Weight and hypertension: Evans County studies of blacks and whites. In: Paul O (ed). Epidemiology and control of hypertension. Stratton Intercontinental Medical Book: New York; 1975. pp 177 – 204. 12 The Atherosclerosis Risk in Communities (ARIC) Study: Design and objectives. The ARIC Investigators. Am J Epidemiol 1989; 129: 687 – 702. 13 Jackson R, Chambless L, Yang K, Byrne T, Watson R, Folsom A, Shahar E, Kalsbeek W. Differences between respondents and nonrespondents in a multi-center community-based study vary by gender and ethnicity. J Clin Epidemiol 1996; 49: 1441 – 1446. 14 Shimakawa T, Sorlie P, Carpenter M, Dennis B, Tell G, Watson R, Williams O, ARIC Investigators. Dietary intake patterns and sociodemographic factors in the Atherosclerosis Risk in Communities (ARIC) Study. Prev Med 1994; 23: 769 – 780 15 Willett W, Sampson L, Stampfer M, Rosner B, Bain C, Witchi J, Hennekens C, Speizer F. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985; 122: 51 – 65. 16 Baecke J, Burema J, Frijters J. A short questionnaire for the measurement of habitual physical activity in epidemiologic studies. Am J Clin Nutr 1982; 36: 936 – 942.

International Journal of Obesity

17 Chambless L, Roeback J. Methods for assessing difference between groups in change when initial measurement is subject to intra-individual variation. Stat Med 1993; 12: 1213 – 1237. 18 Liang K, Zeger S. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73: 13 – 22. 19 Zeger S, Liang K-Y, Albert P. Models for longitudinal data: a generalized estimating equation approach. Biometrics 1988; 44: 1049 – 1060. 20 Kalbfleisch J, Prentice R. The statistical analysis of failure time data. John Wiley: New York; 1980. 21 SAS Institute I. SAS=STAT User’s Guide. SAS Institute Inc.: Cary, NC; 1997. 22 Blaustein M, Grim C. The pathogenesis of hypertension: blackwhite differences. In: Saunders E, Brest A (eds). Cardiovascular disease in blacks. FA Davis: Philadelphia, PA; 1999. pp 97 – 114. 23 Willett W. Anthropometric measures and body composition. In: Willett W (ed). Nutritional epidemiology. Oxford University Press: New York; 1998. pp 244 – 272. 24 Messerli F. Cardiovascular effects of obesity and hypertension. Lancet 1982; i: 1165 – 1168. 25 Siervogel R, Roche A, Chumlea W, Morris J, Webb P, Knittle J. Blood pressure, body composition and fat tissue cellularity in adults. Hypertension 1982; 4: 382 – 386. 26 Hsueh W, Buchanan T. Obesity and hypertension. Endocrinol Metab Clin N Am 1994; 23: 405 – 427.