(Hypertension. 1997;30:1511-1516.)
© 1997 American Heart Association, Inc.
Articles |
Relationship Between Blood Pressure and Body Mass Index in Lean Populations
From the Department of Preventive Medicine and Epidemiology (J.S.K., A.L., A.E.L., R.S.C.), Loyola University Stritch School of Medicine, Maywood, Ill; the Department of Preventive and Social Medicine (M.C.A.), University College Hospital, Ibadan, Oyo State, Nigeria; the Department of Physiology (J.M.), University of Zimbabwe, Harare, Zimbabwe; and the Tropical Metabolism Research Unit (T.F., R.W.), University of the West Indies, Mona, Jamaica.
Correspondence and reprint requests to Jay S. Kaufman, PhD, Department of Research Planning and Evaluation, Carolinas Medical Center, PO Box 32861, Charlotte, NC 28232-2861. E-mail jkaufma{at}orion.it.luc.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Abstract Associations between body mass index (BMI) and blood pressure (BP) have been consistently observed, but remain poorly understood. One unresolved question is whether there is a linear relationship across the entire BMI range. We investigated this question among 11 235 adult men and women from seven low-BMI populations in Africa and the Caribbean. We used kernel smoothing and multivariate linear and spline regression modeling to examine gender differences in the relationship and to test for a threshold. Age-adjusted slopes of BP on BMI were uniformly higher in men than women, with pooled slope ratios of 2.00 and 2.20 for systolic and diastolic BPs, respectively. Men displayed no evidence of age modification or nonlinearity in the relationship, and the age-adjusted slope of systolic BP on BMI was 0.90 (95% confidence interval [CI], 0.76 to 1.04). Women demonstrated both age modification and nonlinearity. For both younger (<45 years) and older (45+ years) women, the optimal change point for a single threshold model was found to be 21 kg/m.2 Slopes of systolic BP on BMI above this threshold were positive and significant: 0.68 (95% CI, 0.54 to 0.81) and 0.53 (95% CI, 0.29 to 0.76) for younger and older women, respectively. Slopes below the threshold were essentially zero for both groups of women, and the difference between the slopes above and below the threshold was significant for younger women (P=.019). In summary, we observed a threshold at 21 kg/m2 in the relationship between BMI and BP for women but not for men. This contributes to the effort to identify the mechanisms that underlie this relationship and how they differ by gender.
Key Words: blood pressure • body mass index • Africa • epidemiology
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The association between measures of body mass and blood pressure has been extensively documented, usually with body mass index (kg/m2) as the measure of relative weight.1 Despite the consistency with which this correlation is observed, mechanistic explanations for the phenomenon are still being debated, and no biological model of the process has been established.2 Epidemiological investigations have been hindered by several factors. The association, however consistent, is rather modest in magnitude, and large sample sizes are therefore required in order to make estimates with any degree of certainty.3 Furthermore, the use of BMI has been questioned because percent body fat, absolute fat mass, and body-fat distribution, or other relevant biological quantities, may not be linearly related to BMI across the entire range of possible values or across different population subgroups.4–6 Finally, the effect of treatment in many populations truncates the distribution of blood pressures and therefore reduces the correlation that would be observed between BMI and blood pressure in untreated settings.
Although the question is often framed in terms of "obesity," defined by BMI cutpoints, correlations between BMI and blood pressure have been observed even in very lean populations, including groups in Africa,7,8 Asia,9,10 and South America.11 Large studies of low-BMI groups in Asia appear to show a monotonic relationship between BMI and blood pressure among groups for whom treatment and other potential confounders are rare.12,13 These studies also show a consistently steeper slope of blood pressure with BMI for men than for women. These observations present both an opportunity and a challenge for biological inference. If body fat is the operative characteristic, for example, then why should women, who have a higher percentage of body fat than men at any given BMI value, have a shallower slope of blood pressure with BMI?
Interpreting the blood pressure–BMI relationship is further complicated by the suggestion from some studies of a threshold effect below which there appears to be no correlation between the variables.14 Some of these studies have relatively small samples and therefore limited power to detect a true relationship.15 It is not surprising, therefore, that the threshold is suggested most commonly for women, who tend to exhibit weaker correlations between BMI and blood pressure in all studies. Some authors have suggested that for women in unindustrialized settings there is no identifiable association between blood pressure and BMI, even at levels that would be considered obese (ie, 30 kg/m2).16,17 Even in studies with substantial statistical power, however, analyses are seldom conducted that explore the consistency of the blood pressure–BMI correlation across the range of BMI values. Despite significant overall slopes or linear correlations, therefore, thresholds are not directly contradicted by any of the published data of which we are aware.
We sought to explore nonlinearity (eg, thresholds) in the relationship between blood pressure and BMI in a pooled sample of individuals from low-BMI populations in Africa and the Caribbean. Specifically, we wished to replicate the finding of Bunker et al14 of a threshold BMI below which there is no association between BMI and blood pressure, in contrast to other studies that propose a roughly linear pattern of association. Additional goals of the study were to assess any relationship between mean blood pressure in a population and the magnitude of association between BMI and blood pressure, to compare the magnitude of association for men and women, and to use these observed patterns to further biological inferences concerning the mechanisms through which BMI is predictive of blood pressure.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We examined the age-adjusted relationship between blood pressure and BMI in seven low-BMI populations of adult men and women in Africa and the Caribbean. Five of the seven sites (Urban Nigeria, Rural Cameroon, Urban Cameroon, Zimbabwe, and Jamaica) were included in ICSHIB,18 and data were collected in these sites in accordance with a standardized protocol.19 Although part of the same overall collaborative project, the remaining two sites (Rural Nigeria, Urban Nigeria) comprise subjects recruited in a surveillance of two communities as part of the NAMS, an investigation of the rate and causes of adult mortality in African communities.20 Anthropometric values were collected in a uniform fashion in all seven sites, as described previously.21,22 Blood pressures collected as part of the ICSHIB protocol were calculated from the mean of the second and third of three manual readings using a standard sphygmomanometer, whereas measures of blood pressure in the NAMS were calculated as the mean of two semiautomatic readings using an Omron portable monitor (HEM 402C). The comparability of the semiautomatic and manual readings has been validated in a separate substudy.23 Antihypertensive treatment was rarely reported (<3% of participants) and believed in most cases to be intermittent or ineffectual. For this reason, no exclusions or adjustments were made for those reporting treatment.
Multivariate linear regression was used to estimate age-adjusted slopes of blood pressure with BMI. Age-adjusted means for blood pressure and BMI were computed using direct standardization based on the proportions in each age decile in the combined superpopulation. Although it has been the most common analytical strategy, we avoided categorization of BMI because of the consequent loss of power and the bias associated with choice of categorization points.24 Rather, we depicted the sex- and age-specific associations between blood pressure and BMI using "lowess" kernel smoothing (bandwidth=0.5) in each sex by age-category strata.25 To avoid misinterpretation of erratic behavior in regions with sparse data, plots were constructed from these analyses by using values from the 5th to 95th percentiles of BMI in each gender group. Finally, the threshold hypothesis was tested using splined linear regression with a single change point.24 The location of this singular knot was selected by searching iteratively over the range of BMI values to minimize the covariate-adjusted sum of square errors. All analyses were conducted using Stata Statistical Software.26
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Mean ages, age-adjusted means, and 95% confidence intervals for BMI, SBP, and DBP are displayed in Table 1
for 11 235 subjects from seven sites.
The ages of the subjects ranged from 15 to 101 years, with mean ages in
each site ranging from 40 to 46 years. Age-adjusted mean BMIs ranged
from 20 to 28 kg/m.2 Age-adjusted mean SBP and DBP
values ranged from 114 to 128 and 70 to 80 mm Hg,
respectively.
|
The linear slopes of blood pressure with BMI, adjusted for age, for SBP
and DBP by site and gender are shown in Table 2. All sites had positive slope
estimates, and all but one were significantly different from zero at
the P<.05 level with the available sample sizes.
Age-adjusted slopes for SBP ranged from 0.40 to 1.70 for men and 0.15
to 1.04 for women. For DBP, the ranges were 0.48 to 1.24 for men and
0.33 to 0.77 for women. In every site, slopes were steeper for men than
women, with the male/female ratio ranging from 1.06 to 2.98 for SBP and
1.40 to 1.86 for DBP. There was a positive correlation between
age-adjusted slope of BMI and mean blood pressure in a population. For
SBP, for example, the coefficient was equal to 0.52 (Fig 1). Mean SBPs by continuous BMI, age
category, and gender are shown in Figs 2 and 3 (analyses for DBP were
similar and were omitted for the sake of brevity). For men, there was a
roughly linear association between SBP and BMI in all age groups, with
a slope that is approximately constant at nearly 1.0 across the age
range, shifted only by a constant. For women, on the other hand, the
younger age categories resemble the plots for men, albeit with a more
modest slope and a flattening out at lower BMI values; women aged 45
years and above had a more distinct leveling or even possible upturn at
BMI values lower than about 21 kg/m.2 On the basis
of these analyses, we determined that we could pool the age
strata for men and adjust for age, since ANCOVA assumptions apply (ie,
parallel lines in each age strata). For women, the distinct shapes in
younger versus older women suggest an age modification in the blood
pressure–BMI relationship that warrants stratification of the women
into two groups at age 45.
|
|
.
|
|
To further clarify the potential threshold in the relationship between blood pressure and BMI among women, we used iterative searching with a spline regression model with a single variable knot. Running the possible knot locations from the 5th to the 95th percentile of BMI values for each of the three groups (men, younger women, and older women) led to a minimum age-adjusted residual sum of squares with the knot at a BMI value of 21.0 kg/m2 for both groups of women. For men, the criterion was met at the extreme BMI value, further indicating that no change point was indicated in that group.
Estimates from linear regression for men and spline regression with a
knot at 21.0 for both groups of women are shown in Table 3. Differences between younger and older
women appear to be less dramatic once age has been adjusted for within
each category. Both age groups show a significant slope above the
threshold and essentially no slope below the threshold (Fig 4). When the two groups of women were
combined, we found a slope estimate below the threshold of essentially
zero (ß=-0.21; 95% CI, -0.63 to 0.21) and an estimate above the
threshold of ß=0.57 (95% CI, 0.44 to 0.70). These slopes are also
significantly different (P=.002). Table 3 also shows results
for a trimmed regression in which the most extreme values (upper and
lower 5%) are omitted. This restriction increases the magnitude of the
estimates both above and below the knot and accentuates the statistical
difference between the slopes.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We have demonstrated nonlinearity (threshold) in the relationship between blood pressure and BMI for women in a pooled population of individuals from low-BMI populations in Africa and the Caribbean. The proposal by Bunker et al14 of a BMI threshold at approximately BMI=21 was confirmed in this study, although only for women. In the case of men, we observed nothing that would contradict published suggestions of a continually decreasing mean blood pressure at lower BMI values.12 Furthermore, we have documented a consistent pattern of steeper BMI slopes for men than for women, with a pooled value for the male/female slope ratio of roughly 2.0. Some of this effect can be explained by the nonlinearity in the relationship for women, which underestimates a slope that is restricted to the linear form. When we consider the pooled slope value above the threshold point for women, the male/female ratio reduces to 0.90/0.57=1.58, for example.
While the outcome we observed is relatively straightforward on a descriptive level, its interpretation is less so. The pathophysiological significance of relative weight has never been well defined. In addition to greater fat stores and its associated direct metabolic consequences, persons with higher BMI values consume more sodium27,28 and engage in less physical activity.29 To make matters more complex, changes in body composition and fat distribution are not linear with BMI and vary by gender. A recent study by our group using bioelectric impedance analysis demonstrated a curvilinear relationship between BMI and percent body fat in these same populations.30 Of course women have twice the percent body fat at a given BMI than men and appear to regulate related hormones, like leptin, at a different level.31
A number of metabolic consequences of obesity have been proposed as the blood pressure–elevating mechanism.2 Increasing weight has been shown to increase salt retention;32,33 and insulin resistance is proposed by some to be a cause of hypertension; adipose tissue produces substantial amounts of AGT, and we recently documented a correlation between BMI and AGT, and between blood pressure and AGT, independent of BMI in Nigerian and Jamaican population samples.34,35 Physical activity can have a substantial effect on blood pressure.36 We have recently shown a strong inverse correlation between energy expended in physical activity and fat stores in this same Nigerian population using the doubly labeled water technique.37 BMI should therefore be viewed conceptually as a proxy for other causal exposures; whether it is diet, hormone changes, physical activity, or other factors that link increasing relative weight to rises in blood pressure cannot be determined at the present. It is clear from the data in this study, however, that obesity, per se, is not required for this association to be manifest, since for men at least BMI values <21 kg/m2 appear linearly related to blood pressure.
It would seem most reasonable to view the BMI–blood pressure relationship over the broadest range as sigmoid in shape. At the lowest extreme, blood pressure cannot approach zero, but must asymptote above a SBP of 80 to 100 mm Hg. By the same token, the cardiovascular system cannot sustain pressures much greater than 200 mm Hg for long periods of time. From that perspective, it would appear that for women, but not men, the lower range of that sigmoid curve is reached at a BMI of roughly 21. For men, on the other hand, the curve appears to have undergone a leftward shift and this flattening of the slope is not seen. A leftward shift would also be consistent with the steeper slope among men. Whether these gender differences are related to sexual dimorphism in body composition cannot be stated. At the same time, whether the patterns we observed are unique to this population, or even this sample, can only be determined by comparative studies. Our relatively large sample size and replication in multiple distinct samples lend consistency to the data, however.
In summary, we have investigated in further detail the relationship between BMI and blood pressure at the lower ranges of relative weight. A diminution of the slope between these two variables is apparent for women but not men. At the same time, this slope is only half as steep for women. The underlying mechanistic processes that link changes in relative weight to its physiological consequences for blood pressure regulation deserve to be better studied. Modeling this relationship with greater precision remains an important challenge for epidemiologists. Studies that measure body composition directly and provide information on physiological intermediates should be particularly useful.
|
Selected Abbreviations and Acronyms |
|---|
|
|
Acknowledgments |
|---|
Support for this work was provided by National Institutes of Health grant No. HL 45508. We acknowledge the International Collaborative Study of Hypertension in Blacks (ICSHIB) and Nigerian Adult Mortality (NAMS) coinvestigators who participated in the collection of data described in this manuscript.
Received April 11, 1997; first decision April 30, 1997; accepted June 13, 1997.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1. Stamler J. Epidemiologic findings on body mass and blood pressure in adults. Ann Epidemiol. 1991;1:347–362.[Medline] [Order article via Infotrieve]
2.
Hall JE. Renal and cardiovascular
mechanisms of hypertension in obesity. Hypertension. 1994;23:381–394.
3. Dyer AR, Elliot P, for the INTERSALT Cooperative Research Group. The INTERSALT study: relations of body mass index to blood pressure. J Hum Hypertens. 1989;3:299–308.[Medline] [Order article via Infotrieve]
4. Roche AF, Siervogel RM. Measures of body composition: their relationship to blood pressure and use in epidemiologic research. Ann Epidemiol. 1991;1:313–320.[Medline] [Order article via Infotrieve]
5.
Spiegelman D, Israel RG, Bouchard C, Willett WC.
Absolute fat mass, percent body fat, and body-fat distribution: which
is the real determinant of blood pressure and serum glucose?
Am J Clin Nutr. 1992;55:1033–1044.
6.
Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris
T, Heymsfield SB. How useful is body mass index for comparison of body
fatness across age, sex and ethnic groups. Am J
Epidemiol. 1996;143:228–239.
7. Poulter NR, Khaw KT, Mugambi M, Peart WS, Rose G, Sever P. Blood pressure patterns in relation to age, weight and urinary electrolytes in three Kenyan communities. Trans R Soc Trop Med Hyg. 1985;79:389–392.[Medline] [Order article via Infotrieve]
8.
Swai AB, McLarty DG, Kitange HM, Kilima PM, Tatalla S,
Keen N, Chuwa LM, Alberti KGMM. Low prevalence of risk factors for
coronary heart disease in rural Tanzania. Int J
Epidemiol. 1993;22:651–659.
9. Gupter R, Guptha S, Gupta VP, Prakash H. Prevalence and determinants of hypertension in the urban population of Jaipur in western India. J Hypertens. 1995;13:1193–1200.[Medline] [Order article via Infotrieve]
10. Folsom AR, Li Y, Rao X, Cen R, Zhang K, Liu X, He L, Irving S, Dennis BH. Body mass, fat distribution and cardiovascular risk factors in a lean population of south China. J Clin Epidemiol. 1994;47:173–181.[Medline] [Order article via Infotrieve]
11. Mancilha-Carvalho JJ, de Oliveire R, Esposito RJ. Blood pressure and electrolyte excretion in the Yanomamo Indians, an isolated population. J Hum Hypertens. 1989;3:309–314.[Medline] [Order article via Infotrieve]
12. Jones DW, Kim JS, Andrew ME, Kim SJ, Hong YP. Body mass index and blood pressure in Korean men and women: the Korean National Blood Pressure Survey. J Hypertens. 1994;12:1433–1437.[Medline] [Order article via Infotrieve]
13.
He J, Klag MJ, Whelton PK, Chen JY, Quian MC, He GQ.
Body mass and blood pressure in a lean population in southwestern
China. Am J Epidemiol. 1994;139:380–389.
14.
Bunker CH, Ukoli FA, Matthews KA, Kriska AM, Huston SL,
Kuller LH. Weight threshold and blood pressure in a lean black
population. Hypertension. 1995;26:616–623.
15.
Kaufman JS, Owoaje EE, James SA, Rotimi C, Cooper RS.
The determinants of hypertension in West Africa: Contribution of
anthropometric and dietary factors to urban-rural and socio-economic
gradients. Am J Epidemiol. 1996;143:1203–1218.
16. Walker AR, Walker BF, Manetsi B, Tsotetsi NG, Walker AJ. Obesity in black women in Soweto, South Africa: minimal effects on hypertension, hyperlipidaemia and hyperglycaemia. J R Soc Health. 1990;110:101–103.[Medline] [Order article via Infotrieve]
17. Walker ARP, Walker BF, Manetsi B, Molefe O, Walker AJ, Vorster HH. Obesity in indigent elderly rural African women: effects on hypertension, hyperlipidaemia and hyperglycaemia. Int J Vitam Nutr Res. 1990;61:244–250.
18.
Cooper R, Rotimi C, Ataman S, McGee D, Osotimehin B,
Kadiri S, Muna W, Kingue S, Fraser H, Forrester T, Bennett F, Wilks R.
Hypertension prevalence in seven populations of African origin.
Am J Public Health. 1997;87:160–168.
19. Ataman SL, Cooper R, Rotimi C, McGee D, Osotimehin B, Kadiri S, Kingue S, Muna W, Fraser H, Forrester T, Wilks R. Standardization of blood pressure measurement in an international comparative study. J Clin Epidemiol. 1996;49:869–877.[Medline] [Order article via Infotrieve]
20. Asuzu MC, Johnson OO, Owoaje EE, Rotimi CN, Kaufman JS, Cooper RS. Questions on adult mortality. World Health Forum. 1996;17:373–376.[Medline] [Order article via Infotrieve]
21. Rotimi CN, Cooper RS, Ataman SL, Osotimehin B, Kadiri S, Muna W, Kingue S, Fraser H, McGee D. Distribution of anthropometric variables and the prevalence of obesity in populations of West African origin: the International Collaborative Study on Hypertension in Blacks (ICSHIB). Obesity Res. 1995;3(suppl 2):95S–105S.
22. Kaufman JS, Durazo-Arvizu RA, Rotimi CN, McGee DL, Cooper RS. Obesity and hypertension prevalence in populations of African origin: results from the International Collaborative Study of Hypertension in Blacks. Epidemiology. 1996;7:398–405.[Medline] [Order article via Infotrieve]
23. Cooper RS, Puras A, Tracy JA, Kaufman JS, Ordunez PO, Mufunda J, Sparks H. Evaluation of an electronic blood pressure device for epidemiologic studies. Blood Press Monit. 1997;2:35–40.[Medline] [Order article via Infotrieve]
24. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology. 1995;6:356–365.[Medline] [Order article via Infotrieve]
25. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc. 1979;74:829–836.
26. StataCorp. Stata Statistical Software: Release 4.0. College Station, TX: Stat Corporation; 1995.
27. Sanchez-Castello CP, Warrender S, Whitehead TP, James WPT. An assessment of the sources of dietary salt in a British population. Clin Sci. 1987;72:95–102.[Medline] [Order article via Infotrieve]
28. Stamler J. The INTERSALT Study: background, methods, findings, and implications. Am J Clin Nutr. 1997;65(suppl):626S–642S.
29. Nelson L, Jennings GL, Esler MD, Korner PI. Effect of changing levels of physical activity on blood pressure and haemodynamics in essential hypertension. Lancet. 1986;2:473–476.[Medline] [Order article via Infotrieve]
30.
Luke A, Durazo-Arvizu R, Rotimi C, Prewitt TE,
Forrester T, Wilks R, Ogunbiyi OJ, Schoeller DA, McGee DA, Cooper RS.
Relationship between body mass index and body fat in black populations
from Nigeria, Jamaica and the United States. Am J
Epidemiol. 1997;145:620–628.
31. Rotimi CN, Luke A, Li Z, Compton JA, Bowsher RR, Cooper RS. Familial aggregation of plasma leptin levels in African-American families. Genet Epidemiol. 1997;14:255–264.[Medline] [Order article via Infotrieve]
32. Rocchini AP, Key J, Bondie D, Chico R, Moorehead C, Katch V, Martin M. The effect of weight loss on the sensitivity of blood pressure to sodium on obese adolescents. N Engl J Med. 1989;321:580–585.[Abstract]
33. Cooper RS, Van Horn L, Liu K, Trevisan M, Nanas S, Stamler J. A randomized trial on the effect of decreased dietary sodium intake on blood pressure in adolescents. J Hypertens. 1984;2:361–366.[Medline] [Order article via Infotrieve]
34. Forrester T, Cooper R, Bennet F, McFarlane-Anderson N, Puras A, Rotimi C, Durazo R, Tewksbury, Ward R. Angiotensinogen and blood pressure among blacks: Findings from a community survey in Jamaica. J Hypertens. 1996;14:315–321.[Medline] [Order article via Infotrieve]
35.
Rotimi CN, Cooper RS, Ogunbiyi O, Morrison L, Ladipo M,
Tewksbury D, Ward R. Hypertension, serum angiotensinogen,
and molecular variants of the angiotensinogen gene among
Nigerians. Circulation. 1997;95:2348–2350.
36. Arroll B, Beaglehole R. Does physical activity lower blood pressure? A critical review of the clinical trials. J Clin Epidemiol. 1992;45:439–447.[Medline] [Order article via Infotrieve]
37. Luke A, Rotimi C, Forreser T, Wilks R, Ogunbiyi OJ, Schoeller DA, Cooper RS. Obesity and energy expenditure among populations of the African Diaspora (abstract). 16th International Congress of Nutrition, Montreal, Canada, July 27-August 1, 1997.
This article has been cited by other articles:
 |
|
 |
|
  G. Livshits and L. M. Gerber Familial Factors of Blood Pressure and Adiposity Covariation Hypertension, March 1, 2001; 37(3): 928 - 935. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. SIFFERT, P. FORSTER, K.-H. JÖCKEL, D. A. MVERE, B. BRINKMANN, C. NABER, R. CROOKES, A. D. P. HEYNS, J. T. EPPLEN, J. FRIDEY, et al. Worldwide Ethnic Distribution of the G Protein {beta}3 Subunit 825T Allele and Its Association with Obesity in Caucasian, Chinese, and Black African Individuals J. Am. Soc. Nephrol., September 1, 1999; 10(9): 1921 - 1930. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||