Original Article
Duration of obesity and incident hypertension in adults from the Framingham Heart Study Stephanie K. Tanamas a, Evelyn Wong a,c, Kathryn Backholer a, Asnawi Abdullah b, Rory Wolfe c, Jan Barendregt d, and Anna Peeters a
Background: Previous studies exploring the association between obesity and hypertension generally used a single baseline measurement of obesity. The effect of accumulating excess adiposity over time on the risk of hypertension is uncertain. This study aimed to examine the relationship between duration of obesity and incident hypertension using the Framingham Heart Study. Methods: Two thousand, nine hundred and fifty-three participants aged 30–62 years without baseline hypertension were included. Blood pressure, height and weight were measured biennially. Duration of obesity was calculated. Time to incident hypertension was analysed using time-varying Cox proportional hazards regression with age as the time scale and censoring at time of death or end of follow-up. Results: Eighty percent of participants developed hypertension (median follow-up 15.9 years). A positive association between obesity duration and incident hypertension was observed in women. There was no longer an association when time-varying BMI was adjusted for (hazard ratio 0.95; (95% confidence interval 0.85– 1.05)). Conclusion: These findings suggest that the mechanism by which excess adiposity may increase blood pressure is primarily immediate and that long-term exposure to obesity does not further increase the risk of developing hypertension beyond the level of BMI attained. Keywords: hypertension, obesity, obesity duration
BACKGROUND
T
he health risks associated with excess bodyweight are generally well characterized. Obesity, measured by a range of anthropometric measures including BMI, has been shown to increase health risks, including hypertension, diabetes and disability [1–3]. One longitudinal study [4] of men examining the relationship between obesity and hypertension over the life course found that, compared with normal weight, overweight and obesity in early adulthood or middle age was associated with a higher risk of hypertension in later life. However, previous studies have generally relied on measurement of adiposity at one time-point, thereby not considering potential cumulative
542
www.jhypertension.com
effect of excess adiposity over time. As populations are becoming obese at younger ages, people are living longer with obesity. Being obese for longer has been shown to be associated with an increased risk of type 2 diabetes [5] and mortality [6]. Although two previous studies examining the relationship between duration of obesity and hypertension found no relationship [7,8], these were cross-sectional analyses of obese populations with retrospective selfreported onset of obesity. Thus, it remains unclear whether an association exists between obesity duration and hypertension. The present study aimed to determine the relationship between obesity duration and risk of developing hypertension using the Framingham Heart Study.
MATERIALS AND METHODS Study population The Framingham Heart Study (FHS), established in 1948, followed 5209 adults aged 30–62 years over approximately 60 years. Participants had repeat biennial examinations during which data on risk factors and health outcomes were collected. The current study included participants without hypertension at baseline (n ¼ 2953).
Duration of obesity Individuals without at least two consecutive obesity occurrences were assigned a value of zero for their obesity duration. For others, the beginning of their obesity duration interval was defined as the date of the first of the consecutive examinations at which they were obese. They were considered to be continuously obese until the first of at least two consecutive examinations at which they were nonobese, death or the end of follow-up. Duration of obesity Journal of Hypertension 2015, 33:542–545 a
Baker IDI Heart and Diabetes Institute, Melbourne, Victoria, Australia, bDepartment of Biostatistics and Population Health, Faculty of Public Health, University Muhammadiyah Aceh, Aceh, Indonesia, cDepartment of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria and dSchool of Population Health, University of Queensland, Herston, Queensland, Australia Correspondence to Stephanie K. Tanamas, Baker IDI Heart and Diabetes Institute, 99 Commercial Rd, Melbourne, VIC 3004, Australia Tel: +61 3 8532 1824; fax: +61 3 8532 1799; e-mail: [email protected] Received 21 April 2014 Revised 3 October 2014 Accepted 3 October 2014
J Hypertens 33:542–545 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI:10.1097/HJH.0000000000000441
Volume 33 Number 3 March 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Duration of obesity and incident hypertension
could increase incrementally at each examination and hence varied over time [5,6]. Those who were obese at baseline were included in the main analysis with obesity duration accumulating from examination one.
Measurement of BMI Height was measured to the nearest 0.25 inches and weight to the nearest 0.5 lbs. BMI was calculated as weight/height2 (kg/m2).
Measurement of hypertension Blood pressure was measured using a mercury column sphygmanometer and a standardized protocol [9]. The average of two measurements was recorded. Hypertension was defined as blood pressure more than 140/90 mmHg or the use of antihypertensive medication. An individual was considered to have incident hypertension if he developed hypertension during the follow-up. Time at risk was defined as time between baseline and the first examination at which the participant had hypertension, death or end of follow-up, whichever came first.
Measurement of covariates Information on marital status, smoking status, alcohol consumption, education, country of birth and menopause status were collected using questionnaires. Some studies have found higher odds of obesity in married men and women than their unmarried counterparts [10]. There is also evidence for differences in blood pressure by marital status [11]. Thus, marital status was included as a potential confounder. Country of birth was used as a surrogate measure for ethnicity and was included as a potential confounder given the evidence of ethnic differences in hypertension and obesity [12]. A person was defined as having type 2 diabetes if the participant was on insulin and/or an oral hypoglycaemic agent, or if their plasma glucose was 200 mg/dl at a given examination.
Imputation for missing values Values for BMI and smoking status were imputed if missing for two or less consecutive examinations. In the instance wherein only one examination was missed, the last observation was carried forward. In the instance that two consecutive examinations were missing, the first was imputed using the last observation carried forward and the second was imputed using the next observation carried backward. Where more than two consecutive examinations were missed, no imputation was performed and participants were censored from analysis at the occurrence of the first missing BMI or smoking status. Values for hypertension were imputed, by carrying the last observation forward, only if missing on isolated single examinations. Marital status and menopause status were imputed by carrying the last observation forward.
Statistical analysis The relationship between duration of obesity and incident hypertension was assumed to conform to a simple hazard ratio per 1 year of duration and estimation of the association with 95% confidence interval (CI) was based on a Cox Journal of Hypertension
proportional hazards regression using age as the time scale and cumulative duration of obesity as a time-varying covariate. Adjustments were made for the baseline variables of education, country of birth, alcohol consumption and for the time-varying variables of marital status, BMI, smoking, diabetes status and menopause status. Inclusion of BMI as a time-varying variable in the Cox model allowed us to account for any effect of current BMI. As alcohol consumption was collected only at eight out of 28 examinations, it was not included as a time-varying covariate, rather alcohol consumption at examination 2 (the first examination at which it was collected; considered as baseline alcohol consumption) was included. Analyses were stratified by sex. All analyses were performed using Stata statistical software package version 11.2 (StataCorp, College Station, Texas, USA).
Sensitivity analyses To test whether the association between obesity duration and incident hypertension might be influenced by imputing missing values, an analysis was performed including only participants without missing values of BMI or hypertension in any examination (n ¼ 1973). A sensitivity analysis was also performed to test the impact of uncertain obesity duration for those obese at baseline by analysing only those who were not obese at baseline (n ¼ 2768). Due to the large number of participants without obesity throughout the study, we also analysed the relationship in only those with obesity at baseline (n ¼ 185). The effect of defining duration using two consecutive occurrences of obesity (which was done to avoid potential misclassification of bodyweight due to measurement error or fluctuations between the borderline of the ‘overweight’ and ‘obese’ BMI categories) was tested by undertaking analyses with obesity duration defined using a simple count of single examinations with obesity.
RESULTS Of 2953 participants without hypertension at baseline, 2355 participants (80%) developed hypertension over the follow-up period (median follow-up time 15.9 years). Participant characteristics at baseline are described in Table 1. Briefly, the population comprised 43% men, with a mean baseline age of 43.1 and 41.2 years for men and women, respectively. The mean baseline BMI was 24.8 kg/m2 for men and 24.0 kg/m2 for women. Two hundred and eleven men (16.7%) and 379 women (22.4%) were ever obese during the follow-up [median duration of obesity, interquartile range (IQR) 14 (6–24) years for men and 14 (8–24) years for women]. There was a positive association between obesity duration and incident hypertension in women, adjusted for baseline alcohol consumption, education, country of birth, smoking, marital status, diabetes status and menopause status, although adjustment for current BMI shifted hazard ratios towards an absence of association (Table 2). The sensitivity analyses revealed similar findings to the main analysis when those with missing BMI or hypertension status at any examination were excluded, when stratified by baseline obesity status and when a simple count of www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
543
Tanamas et al. TABLE 1. Baseline characteristics of the study population
Age (years) Country of birth United States Not United States Marital status Single Married Widowed Divorced/separated Education Eighth grade Did not graduate high school Graduated high school Did not graduate college Graduated college Post graduate, business college, nursing/music/art school Smoking status Current smoker Not current smoker (ex-smoker and never smoker) Alcohol consumption (ounce/month) BMI (kg/m2) Diabetes status
Men (n ¼ 1261)
Women (n ¼ 1692)
43.1 8.5
41.2 7.8
1033 (81.9) 228 (18.1)
1432 (84.6) 260 (15.4)
57 1179 10 15
(4.5) (93.5) (0.8) (1.2)
194 1376 71 51
(11.5) (81.3) (4.2) (3.0)
309 172 347 114 137 130
(25.6) (14.2) (28.7) (9.4) (11.3) (10.8)
360 192 546 134 115 283
(22.1) (11.8) (33.5) (8.2) (7.1) (17.4)
993 (79.4) 258 (20.6)
828 (49.1) 857 (50.9)
7 (1–29) 24.8 3.2 19 (1.5)
1 (0–8) 24.0 3.7 17 (1.0)
Data presented as mean SD, n (%) or median (25th percentile–75th percentile).
obesity (single occurrences) was used to calculate obesity duration.
DISCUSSION In this analysis of almost 3000 adults without hypertension at baseline, there was no relationship between duration of obesity and incident hypertension independent of current BMI. Few studies have examined the role of obesity duration in the development of hypertension. In a previous
cross-sectional examination of obese individuals who were admitted to hospital for obesity-related health problems, duration of obesity was significantly associated with glucose tolerance but not with arterial hypertension [8]. Notably, obesity duration in their population was calculated using self-reported weight and height at various ages and was prone to recall bias. Similar findings were reported by another cross-sectional study of 200 obese participants wherein duration of obesity, based on self-report, was not associated with either type 2 diabetes or hypertension [7]. Our findings using longitudinal data with a long period of follow-up and measured height and weight were in line with these previous findings. Although duration of obesity was associated with an increased risk of incident hypertension, this relationship was attenuated once adjusted for current BMI. This suggests that being obese for longer is not a significant predictor for the development of hypertension, above and beyond current BMI. Our study was limited by a relatively small number of participants who developed obesity over the period of follow-up [n ¼ 590 (20%)]. Although this may be related to our calculation of obesity duration erring on the side of underassessment of having a measurable duration of obesity (because of requirement for obesity in at least two consecutive examinations), our findings were not changed when we used the simple count of single examinations with obesity [number who developed obesity ¼ 765 (26%)]. As it was not possible to include any potential obesity duration prior to the commencement of the study, total obesity duration in our study may be underestimated. Nevertheless, similar results were found when those who were obese at baseline were excluded. There are suggestions that other measures of obesity, such as waist circumference and waistto-hip ratio (WHR), are better predictors of health outcomes. However, current evidence, at least for hypertension, suggests that the relationship with BMI, waist circumference and WHR is similar [13]. Nevertheless, the
TABLE 2. Relationship between duration of obesity and incident hypertension Model 1 HR (95% CI)
Model 2 P
Total population 1.17 (1.10–1.25)
Duration of obesity and incident hypertension in adults from the Framingham Heart Study Stephanie K. Tanamas a, Evelyn Wong a,c, Kathryn Backholer a, Asnawi Abdullah b, Rory Wolfe c, Jan Barendregt d, and Anna Peeters a
Background: Previous studies exploring the association between obesity and hypertension generally used a single baseline measurement of obesity. The effect of accumulating excess adiposity over time on the risk of hypertension is uncertain. This study aimed to examine the relationship between duration of obesity and incident hypertension using the Framingham Heart Study. Methods: Two thousand, nine hundred and fifty-three participants aged 30–62 years without baseline hypertension were included. Blood pressure, height and weight were measured biennially. Duration of obesity was calculated. Time to incident hypertension was analysed using time-varying Cox proportional hazards regression with age as the time scale and censoring at time of death or end of follow-up. Results: Eighty percent of participants developed hypertension (median follow-up 15.9 years). A positive association between obesity duration and incident hypertension was observed in women. There was no longer an association when time-varying BMI was adjusted for (hazard ratio 0.95; (95% confidence interval 0.85– 1.05)). Conclusion: These findings suggest that the mechanism by which excess adiposity may increase blood pressure is primarily immediate and that long-term exposure to obesity does not further increase the risk of developing hypertension beyond the level of BMI attained. Keywords: hypertension, obesity, obesity duration
BACKGROUND
T
he health risks associated with excess bodyweight are generally well characterized. Obesity, measured by a range of anthropometric measures including BMI, has been shown to increase health risks, including hypertension, diabetes and disability [1–3]. One longitudinal study [4] of men examining the relationship between obesity and hypertension over the life course found that, compared with normal weight, overweight and obesity in early adulthood or middle age was associated with a higher risk of hypertension in later life. However, previous studies have generally relied on measurement of adiposity at one time-point, thereby not considering potential cumulative
542
www.jhypertension.com
effect of excess adiposity over time. As populations are becoming obese at younger ages, people are living longer with obesity. Being obese for longer has been shown to be associated with an increased risk of type 2 diabetes [5] and mortality [6]. Although two previous studies examining the relationship between duration of obesity and hypertension found no relationship [7,8], these were cross-sectional analyses of obese populations with retrospective selfreported onset of obesity. Thus, it remains unclear whether an association exists between obesity duration and hypertension. The present study aimed to determine the relationship between obesity duration and risk of developing hypertension using the Framingham Heart Study.
MATERIALS AND METHODS Study population The Framingham Heart Study (FHS), established in 1948, followed 5209 adults aged 30–62 years over approximately 60 years. Participants had repeat biennial examinations during which data on risk factors and health outcomes were collected. The current study included participants without hypertension at baseline (n ¼ 2953).
Duration of obesity Individuals without at least two consecutive obesity occurrences were assigned a value of zero for their obesity duration. For others, the beginning of their obesity duration interval was defined as the date of the first of the consecutive examinations at which they were obese. They were considered to be continuously obese until the first of at least two consecutive examinations at which they were nonobese, death or the end of follow-up. Duration of obesity Journal of Hypertension 2015, 33:542–545 a
Baker IDI Heart and Diabetes Institute, Melbourne, Victoria, Australia, bDepartment of Biostatistics and Population Health, Faculty of Public Health, University Muhammadiyah Aceh, Aceh, Indonesia, cDepartment of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria and dSchool of Population Health, University of Queensland, Herston, Queensland, Australia Correspondence to Stephanie K. Tanamas, Baker IDI Heart and Diabetes Institute, 99 Commercial Rd, Melbourne, VIC 3004, Australia Tel: +61 3 8532 1824; fax: +61 3 8532 1799; e-mail: [email protected] Received 21 April 2014 Revised 3 October 2014 Accepted 3 October 2014
J Hypertens 33:542–545 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI:10.1097/HJH.0000000000000441
Volume 33 Number 3 March 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Duration of obesity and incident hypertension
could increase incrementally at each examination and hence varied over time [5,6]. Those who were obese at baseline were included in the main analysis with obesity duration accumulating from examination one.
Measurement of BMI Height was measured to the nearest 0.25 inches and weight to the nearest 0.5 lbs. BMI was calculated as weight/height2 (kg/m2).
Measurement of hypertension Blood pressure was measured using a mercury column sphygmanometer and a standardized protocol [9]. The average of two measurements was recorded. Hypertension was defined as blood pressure more than 140/90 mmHg or the use of antihypertensive medication. An individual was considered to have incident hypertension if he developed hypertension during the follow-up. Time at risk was defined as time between baseline and the first examination at which the participant had hypertension, death or end of follow-up, whichever came first.
Measurement of covariates Information on marital status, smoking status, alcohol consumption, education, country of birth and menopause status were collected using questionnaires. Some studies have found higher odds of obesity in married men and women than their unmarried counterparts [10]. There is also evidence for differences in blood pressure by marital status [11]. Thus, marital status was included as a potential confounder. Country of birth was used as a surrogate measure for ethnicity and was included as a potential confounder given the evidence of ethnic differences in hypertension and obesity [12]. A person was defined as having type 2 diabetes if the participant was on insulin and/or an oral hypoglycaemic agent, or if their plasma glucose was 200 mg/dl at a given examination.
Imputation for missing values Values for BMI and smoking status were imputed if missing for two or less consecutive examinations. In the instance wherein only one examination was missed, the last observation was carried forward. In the instance that two consecutive examinations were missing, the first was imputed using the last observation carried forward and the second was imputed using the next observation carried backward. Where more than two consecutive examinations were missed, no imputation was performed and participants were censored from analysis at the occurrence of the first missing BMI or smoking status. Values for hypertension were imputed, by carrying the last observation forward, only if missing on isolated single examinations. Marital status and menopause status were imputed by carrying the last observation forward.
Statistical analysis The relationship between duration of obesity and incident hypertension was assumed to conform to a simple hazard ratio per 1 year of duration and estimation of the association with 95% confidence interval (CI) was based on a Cox Journal of Hypertension
proportional hazards regression using age as the time scale and cumulative duration of obesity as a time-varying covariate. Adjustments were made for the baseline variables of education, country of birth, alcohol consumption and for the time-varying variables of marital status, BMI, smoking, diabetes status and menopause status. Inclusion of BMI as a time-varying variable in the Cox model allowed us to account for any effect of current BMI. As alcohol consumption was collected only at eight out of 28 examinations, it was not included as a time-varying covariate, rather alcohol consumption at examination 2 (the first examination at which it was collected; considered as baseline alcohol consumption) was included. Analyses were stratified by sex. All analyses were performed using Stata statistical software package version 11.2 (StataCorp, College Station, Texas, USA).
Sensitivity analyses To test whether the association between obesity duration and incident hypertension might be influenced by imputing missing values, an analysis was performed including only participants without missing values of BMI or hypertension in any examination (n ¼ 1973). A sensitivity analysis was also performed to test the impact of uncertain obesity duration for those obese at baseline by analysing only those who were not obese at baseline (n ¼ 2768). Due to the large number of participants without obesity throughout the study, we also analysed the relationship in only those with obesity at baseline (n ¼ 185). The effect of defining duration using two consecutive occurrences of obesity (which was done to avoid potential misclassification of bodyweight due to measurement error or fluctuations between the borderline of the ‘overweight’ and ‘obese’ BMI categories) was tested by undertaking analyses with obesity duration defined using a simple count of single examinations with obesity.
RESULTS Of 2953 participants without hypertension at baseline, 2355 participants (80%) developed hypertension over the follow-up period (median follow-up time 15.9 years). Participant characteristics at baseline are described in Table 1. Briefly, the population comprised 43% men, with a mean baseline age of 43.1 and 41.2 years for men and women, respectively. The mean baseline BMI was 24.8 kg/m2 for men and 24.0 kg/m2 for women. Two hundred and eleven men (16.7%) and 379 women (22.4%) were ever obese during the follow-up [median duration of obesity, interquartile range (IQR) 14 (6–24) years for men and 14 (8–24) years for women]. There was a positive association between obesity duration and incident hypertension in women, adjusted for baseline alcohol consumption, education, country of birth, smoking, marital status, diabetes status and menopause status, although adjustment for current BMI shifted hazard ratios towards an absence of association (Table 2). The sensitivity analyses revealed similar findings to the main analysis when those with missing BMI or hypertension status at any examination were excluded, when stratified by baseline obesity status and when a simple count of www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
543
Tanamas et al. TABLE 1. Baseline characteristics of the study population
Age (years) Country of birth United States Not United States Marital status Single Married Widowed Divorced/separated Education Eighth grade Did not graduate high school Graduated high school Did not graduate college Graduated college Post graduate, business college, nursing/music/art school Smoking status Current smoker Not current smoker (ex-smoker and never smoker) Alcohol consumption (ounce/month) BMI (kg/m2) Diabetes status
Men (n ¼ 1261)
Women (n ¼ 1692)
43.1 8.5
41.2 7.8
1033 (81.9) 228 (18.1)
1432 (84.6) 260 (15.4)
57 1179 10 15
(4.5) (93.5) (0.8) (1.2)
194 1376 71 51
(11.5) (81.3) (4.2) (3.0)
309 172 347 114 137 130
(25.6) (14.2) (28.7) (9.4) (11.3) (10.8)
360 192 546 134 115 283
(22.1) (11.8) (33.5) (8.2) (7.1) (17.4)
993 (79.4) 258 (20.6)
828 (49.1) 857 (50.9)
7 (1–29) 24.8 3.2 19 (1.5)
1 (0–8) 24.0 3.7 17 (1.0)
Data presented as mean SD, n (%) or median (25th percentile–75th percentile).
obesity (single occurrences) was used to calculate obesity duration.
DISCUSSION In this analysis of almost 3000 adults without hypertension at baseline, there was no relationship between duration of obesity and incident hypertension independent of current BMI. Few studies have examined the role of obesity duration in the development of hypertension. In a previous
cross-sectional examination of obese individuals who were admitted to hospital for obesity-related health problems, duration of obesity was significantly associated with glucose tolerance but not with arterial hypertension [8]. Notably, obesity duration in their population was calculated using self-reported weight and height at various ages and was prone to recall bias. Similar findings were reported by another cross-sectional study of 200 obese participants wherein duration of obesity, based on self-report, was not associated with either type 2 diabetes or hypertension [7]. Our findings using longitudinal data with a long period of follow-up and measured height and weight were in line with these previous findings. Although duration of obesity was associated with an increased risk of incident hypertension, this relationship was attenuated once adjusted for current BMI. This suggests that being obese for longer is not a significant predictor for the development of hypertension, above and beyond current BMI. Our study was limited by a relatively small number of participants who developed obesity over the period of follow-up [n ¼ 590 (20%)]. Although this may be related to our calculation of obesity duration erring on the side of underassessment of having a measurable duration of obesity (because of requirement for obesity in at least two consecutive examinations), our findings were not changed when we used the simple count of single examinations with obesity [number who developed obesity ¼ 765 (26%)]. As it was not possible to include any potential obesity duration prior to the commencement of the study, total obesity duration in our study may be underestimated. Nevertheless, similar results were found when those who were obese at baseline were excluded. There are suggestions that other measures of obesity, such as waist circumference and waistto-hip ratio (WHR), are better predictors of health outcomes. However, current evidence, at least for hypertension, suggests that the relationship with BMI, waist circumference and WHR is similar [13]. Nevertheless, the
TABLE 2. Relationship between duration of obesity and incident hypertension Model 1 HR (95% CI)
Model 2 P
Total population 1.17 (1.10–1.25)
