Lower hypertension risk in Mexico City than in San Antonio

Abstract

Background

We examined the effects of blood pressure (BP), weight, and weight gain on hypertension risk in two similar ethnic origin populations, subjects in Mexico City and Mexican Americans in San Antonio.

Methods

The Mexico City Diabetes Study and San Antonio Heart Study are population-based, epidemiologic studies with identical survey protocols. Incident hypertension (BP ≥140/90 mm Hg or current antihypertensive treatment) was analyzed in subjects aged 35 to 64 years of Mexican ethnicity living in low-income neighborhoods (n = 1467 in Mexico City, n = 628 in San Antonio).

Results

In Mexico City, 10.6% of men and 13.1% of women developed hypertension in a 6.5-year period; in San Antonio, 28.6% and 28.7% in a 7.5-year period, respectively. Poisson regression analysis demonstrated a greater hypertension risk in San Antonio for both men (risk ratio [RR] = 1.75, 95% confidence interval [CI]: 1.19–2.56) and women (RR = 1.40, 95% CI: 1.05–1.86). In a multiple linear regression analysis, systolic BP change was associated with weight gain in Mexico City (P < .001 in men and women) and San Antonio (P = .045 in men, and P = .027 in women) independently of age, BP, obesity, alcohol consumption, cigarette smoking, diabetes, and antihypertensive treatment. These covariates did not fully explain greater increments of systolic BP in San Antonio than in Mexico City (P < .001 in men and women).

Conclusions

Hypertension risk is lower in Mexico City than in San Antonio. Systolic BP increases with weight gain, independently of other determinants of hypertension. Am J Hypertens 2005; 18:385–391 © 2005 American Journal of Hypertension, Ltd.

Hypertension, obesity, incidence, ethnicity

In San Antonio, hypertension risk is relatively lower in Mexican Americans than in non-Hispanic whites despite greater type 2 diabetes and obesity rates.¹ On the other hand, subjects in Mexico City have a lower prevalence of hypertension than Mexican Americans in San Antonio, Texas^2,3 in addition to less type 2 diabetes⁴ and overall adiposity.^2,3 In this study, we compared the incidence of hypertension between these two populations to help to understand prevalence differences of hypertension.

Because of their similar ethnic origin,^5,6 subjects in Mexico City and Mexican Americans in San Antonio also provide an optimal ground to sort the effects of environmental determinants of hypertension.^7–10 In middle-aged residents of low-income neighborhoods, we determined the risk for hypertension associated with blood pressure (BP) and weight, and the impact of weight gain on BP change independently of the effects of other determinants of hypertension.

Methods

Study population

The San Antonio Heart Study (SAHS) is a population-based, epidemiologic study on diabetes and cardiovascular disease in Mexican Americans and non-Hispanic whites of San Antonio, Texas.^1–4 Households from low-, middle-, and high-income neighborhoods were randomly sampled. A total of 5158 men and nonpregnant women aged 25 to 64 years were initially enrolled in two phases (response rate 65.3%). Phase 1 participants completed a 7- to 8-year examination between October 1987 and November 1990¹¹; phase 2 participants, between October 1991 and October 1996.¹²

An analogous study, the Mexico City Diabetes Study (MCDS), used the same sampling method among residents in low-income neighborhoods of Mexico City between November 1989 and February 1990. Initial enrollment was 2282 men and nonpregnant women aged 35 to 64 years (response rate 68.5%).³ Participants returned for a follow-up examination 6 to 7 years later.⁴ In both studies, survey protocols were approved by local institutional review boards. All subjects gave written informed consent.

In the SAHS, only Mexican Americans aged 35 to 64 years from low-income neighborhoods were eligible for comparisons with participants from the MCDS because of the known associations of hypertension with age, ethnic origin, and socioeconomic status.³ The median of educational attainment was 6 years in the MCDS and 7 years in the SAHS. In the MCDS, baseline prevalence of hypertension was 17.1% in men and 18.1% in women; in the SAHS, 26.7% and 24.7%, respectively.

Normotensive subjects at baseline, 1861 participants in the MCDS and 850 Mexican Americans in the SAHS, were suitable for analysis. Mortality was similar in both cites: 67/1634 in the MCDS and 35/832 in the SAHS (3.5% v 3.2% after the adjustment for age and sex, P = .634). Among participants alive at follow-up, incident hypertension was ascertained in 1467 (81.8%) subjects in the MCDS and 628 (77.1%) Mexican Americans in the SAHS. In both studies, subjects who returned for a follow-up examination had similar baseline body mass index (BMI), waist circumference, triglycerides, HDL-cholesterol, systolic and diastolic BPs, and type 2 diabetes to those who did not return. Phone interviews were also administered at follow-up. In the MCDS, self-reported incident hypertension was similar in subjects who returned for a follow-up visit and in those who did not (21.5% v 19.0%, P = .381). However, self-reported incident hypertension was more prevalent in SAHS participants who returned for a follow-up visit than in those who did not (22.3% v 9.6%, P < .001).

Definition of variables and outcomes

In addition to using identical protocols, both studies had standardized and joint training for medical staff. Educational attainment and current cigarette smoking, alcohol consumption, and antihypertensive treatment were self-reported. Educational attainment was used as a dichotomous variable (≥8 v <8 years of education), because <8 years of education was associated with a higher prevalence of hypertension among men.³ Waist circumference was measured at the level of the umbilicus. Systolic and diastolic BPs (first and fifth phases of Korotkoff sounds, respectively) were measured to the nearest even digit with the participant in the sitting position, and reported as the mean of the second and third BP readings. Blood samples were collected after a 12- to 14-h fast. A 75-g oral glucose load (Orangedex; Custom Laboratories, Baltimore, MD) was administered to assess glucose tolerance status. Plasma glucose and serum lipids were measured with an Abbott Bichromatic Analyzer (South Pasadena, CA) in the Division of Clinical Epidemiology in San Antonio.^2,3,11

We used the 1999 World Health Organization definition of type 2 diabetes (fasting glucose ≥7.0 mmol/L or 2-h glucose ≥11.1 mmol/L).¹³ Subjects on oral antidiabetic medications or those on insulin plus BMI >27 kg/m² and age at onset >30 years were also consider to have type 2 diabetes. Hypertension was defined as BP ≥140/90 mm Hg or current treatment with antihypertensive medications.¹⁴

Statistical methods

Using the SAS statistical software system (SAS Institute Inc., Cary, NC), we compared continuous variables by one-way analysis of covariance and dichotomous variables by logistic regression analysis or χ² statistic. We used Spearman correlations to examine the relationship between continuous variables, Poisson regression analysis to estimate hypertension risk ratios (RR) between populations, and multiple linear regression analysis to assess the effect of determinants of hypertension on BP changes. All probability values were two-sided.

We observed interactions (P < .3) between population and incident hypertension for the following variables: waist circumference (P = .238), diastolic BP (P = .231), cigarette smoking (P = .137), and alcohol consumption (P = .066) in men, and systolic BP (P = .126), triglycerides (P = .266) and alcohol consumption (P = .186) in women.

Results

In both populations, incident hypertension was associated with older age and higher baseline BP in men and women and greater baseline BMI in women (Table 1). Among participants with incident hypertension, self-reported antihypertensive treatment was comparable in Mexico City and San Antonio (in men 32.8% v 23.7%, P = .267; in women, 54.5% v 53.1%, P = .825), but was more frequent in women than in men (P = .014 in Mexico City and P < .001 in San Antonio). Women had greater BMI than men (P = .048 in San Antonio and P < .001 in Mexico City), but lower systolic and diastolic BPs (P < .001 for all comparisons). In comparisons with subjects in Mexico City, men and women in San Antonio had greater BMI and systolic BP (P < .001 for all comparisons), and more diabetes (in men, P = .051; in women, P = .001). Women in San Antonio had also greater diastolic BP (P < .001), but men did not (P = .428).

Table 1

Age-adjusted baseline characteristics by city, sex, and incident hypertension

	Mexico City			San Antonio
Incident Hypertension	No	Yes	P	No	Yes	P
Men
n	542	64	—	147	59	—
Age (y)*	45.5 ± 0.35	49.9 ± 1.00	<.001	49.0 ± 0.66	53.1 ± 1.05	.001
BMI (kg/m2)	26.8 ± 0.17	26.8 ± 0.48	.896	28.8 ± 0.32	28.6 ± 0.51	.719
Waist circumference (cm)	93.6 ± 0.41	92.6 ± 1.20	.375	96.3 ± 0.97	97.8 ± 1.55	.457
Triglycerides (mmol/L)	2.78 ± 0.08	2.72 ± 0.24	.893	2.19 ± 0.16	2.03 ± 0.26	.419
Type 2 diabetes (%)	8.5%	14.3%	.144	12.8%	19.1%	.158
SBP (mm Hg)	114.1 ± 0.46	121.9 ± 1.32	<.001	117.8 ± 0.87	124.0 ± 1.40	<.001
DBP (mm Hg)	72.2 ± 0.34	77.7 ± 0.99	<.001	72.3 ± 0.65	75.9 ± 1.05	.002
Women
n	748	113	—	301	121	—
Age (y)*	45.3 ± 0.28	50.4 ± 0.73	<.001	47.4 ± 0.45	52.1 ± 0.71	<.001
BMI (kg/m2)	28.3 ± 0.18	29.9 ± 0.47	.002	29.3 ± 0.29	31.3 ± 0.46	<.001
Waist circumference (cm)	97.2 ± 0.47	100.8 ± 1.22	.015	88.1 ± 0.91	93.6 ± 1.41	<.001
Triglycerides (mmol/L)	1.97 ± 0.05	2.37 ± 0.13	.007	1.67 ± 0.08	1.84 ± 0.13	.152
Type 2 diabetes, (%)	10.4%	18.8%	.017	15.1%	21.3%	.059
SBP (mm Hg)	109.3 ± 0.38	120.5 ± 0.99	<.001	112.9 ± 0.60	121.8 ± 0.96	<.001
DBP (mm Hg)	68.6 ± 0.28	73.4 ± 0.73	<.001	69.7 ± 0.44	74.4 ± 0.71	<.001

BMI = body mass index; DBP = diastolic blood pressure; SBP = systolic blood pressure.

Data are n, means ± SE, or %.

*

Values not adjusted for age.

Incident hypertension was diagnosed in 10.6% of men and 13.1% of women from Mexico City, and in 28.6% of men and 28.7% of women from San Antonio. The incidence of hypertension per 100 person-years of observation was 1.64 (95% CI: 1.24–2.04) in men and 2.05 (95% CI: 1.68–2.42) in women from Mexico City, and 3.54 (95% CI: 2.63–4.46) in men and 3.35 (95% CI: 2.72–3.98) in women from San Antonio. Men and women had comparable age-adjusted hypertension risk by Poisson regression analysis: Mexico City, RR = 0.80 (95% CI: 0.58–1.08); San Antonio, RR = 0.99 (95% CI: 0.71–1.37).

Follow-up examination rate was greater in the Mexico City than in San Antonio (78.8% v 73.5%, P = .002). After stratification for age and diabetes status, follow-up rates were statistically significant only in nondiabetic subjects aged 35 to 44 years (Fig. 1A). However, men and women of all age groups had a greater risk for hypertension in San Antonio (Fig. 1B).

(A) Follow-up examination rates by city, diabetes status, and age group. Number of subjects for each category is indicated. No statistical differences were observed in the 45- to 54-year-old group (P = .821) and in the 55- to 64-year-old group (P = .312). In the 35- to 54-year-old group, more nondiabetic subjects in San Antonio missed the follow-up examination (P < .001). (B) Incidence of hypertension per 100 years of observation by city, sex, and age group. Sex-adjusted RR for hypertension was greater in San Antonio than in Mexico City for all age groups: RR in the 35- to 44-year-old group = 1.74 (95% CI: 1.10–2.71), RR in the 45- to 54-year-old group = 1.53 (95% CI: 1.04–2.25), and RR in the 55- to 64-year-old group = 1.41 (95% CI: 1.02–1.95).

Figure 1.

Baseline BMI predicted hypertension only in women independently of age and type 2 diabetes (Table 2). However, baseline BP explained the risk for future hypertension associated with baseline BMI in women from San Antonio, and partially the risk in women from Mexico City. In comparisons between San Antonio and Mexico City, age- and education-adjusted hypertension risk was greater in San Antonio: 75% more among men (RR = 1.75, 95% CI: 1.19–2.56) and 40% among women (RR = 1.40, 95% CI: 1.05–1.86). Additional adjustment for type 2 diabetes, BMI, and BP values partially explained differences in hypertension risk between populations although less in men RR = 1.47 (95% CI: 0.98–2.20) than in women RR = 1.12 (95% CI: 0.83–1.51). In comparisons between men and women, age-adjusted hypertension risk was similar in both cities, Mexico City (RR = 0.80, 95% CI: 0.58–1.08) and San Antonio (RR = 0.99, 95% CI: 0.71–1.37).

Table 2

Poisson regression analysis with incidence of hypertension as the dependent variable

	San Antonio				Mexico City
	Model 1		Model 2		Model 1		Model 2
	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
Men
Baseline age (10-yr interval)	1.51	(1.08–2.15)	1.37	(0.96–1.98)	1.70	(1.26–2.29)	1.52	(1.11–2.07)
Baseline BMI (5-kg/m2 interval)	0.93	(0.69–1.23)	0.87	(0.64–1.16)	0.98	(0.69–1.38)	0.74	(0.51–1.06)
Baseline type 2 diabetes (yes vs. no)	1.52	(0.79–2.75)	1.49	(0.79–2.67)	1.60	(0.78–3.01)	1.53	(0.75–2.86)
Baseline SBP (10-mm Hg interval)	—		1.44	(1.06–1.99)	—		1.53	(1.14–2.06)
Baseline DBP (5-mm Hg interval)	—		1.21	(0.98–1.51)	—		1.35	(1.11–1.67)
Women
Baseline age (10-yr interval)	1.70	(1.33–2.20)	1.42	(1.09–1.86)	1.85	(1.46–2.36)	1.41	(1.07–1.85)
Baseline BMI (5-kg/m2 interval)	1.17	(1.00–1.37)	1.00	(0.84–1.18)	1.30	(1.07–1.55)	1.17	(0.96–1.41)
Baseline type 2 diabetes (yes vs. no)	1.27	(0.81–1.94)	1.19	(0.77–1.87)	1.57	(0.98–2.43)	1.37	(0.86–2.14)
Baseline SBP (10-mm Hg interval)	—		1.58	(1.27–1.97)	—		2.01	(1.64–2.48)
Baseline DBP (5-mm Hg interval)	—		1.13	(0.98–1.30)	—		1.02	(0.88–1.18)

CI = confidence interval; RR = risk ratio; other abbreviations as in Table 1.

Model 1 includes baseline age, BMI, and type 2 diabetes as independent variables; Model 2, variables in Model 1 plus SBP and DBP.

Baseline systolic BP correlated positively with age and baseline BMI (Table 3). Systolic BP change correlated positively with age and BMI change and negatively with baseline systolic BP. Baseline BMI did not correlate with systolic BP change, but correlated negatively with BMI change in men (Mexico City: r = −0.09, P < .05; San Antonio: r = −0.18, P < .05) and women (Mexico City: r = −0.10, P < .010; San Antonio: r = −0.19, P < .001).

Table 3

Spearman correlations between age, baseline SBP, baseline BMI, SBP change, and BMI change

	Baseline SBP		SBP Change at Follow-up
	Mexico City	San Antonio	Mexico City	San Antonio
Men
Baseline age	0.24 ‡	0.15 †	0.07	0.14 *
Baseline SBP	—	—	−0.30 ‡	−0.25 ‡
Baseline BMI	0.28 ‡	0.13 *	−0.05	−0.11
BMI change at follow-up	−0.09 *	−0.09	0.12 †	0.16 *
Women
Baseline age	0.33 ‡	0.28 ‡	0.11 ‡	0.13 †
Baseline SBP	—	—	−0.20 ‡	−0.20 ‡
Baseline BMI	0.22 ‡	0.36 ‡	0.04	−0.08
BMI change at follow-up	−0.12 ‡	−0.08	0.11 †	0.13 †

Abbreviations as in Tables 1 and 2.

*

P < .05;

†

P < .01;

‡

P < .001.

In women, age-adjusted BMI change increased more in San Antonio than in Mexico City (1.28 ± 0.13 v 0.71 ± 0.09 kg/m², P < .001), but increased similarly in men from either city (0.37 ± 0.18 v 0.42 ± 0.10 kg/m², P = .634). For comparable weight gain, annual increments of systolic BP were greater in San Antonio than in Mexico City (Fig. 2), and greater in women than in men (P < .001 in Mexico City and P = .001 in San Antonio).

Age-adjusted annual change in systolic BP (mm Hg/year) by annual change in BMI (tertiles). Mexico City versus San Antonio in men (1st tertile: P = .009, 2nd tertile: P = .068, and 3rd tertile: P < .001) and women (1st tertile: P < .001, 2nd tertile: P < .001, and 3rd tertile: P < .001). Annual change in SBP in Mexico City (P for trend in men = .025, and P for trend in women < .001) and San Antonio (P for trend in men = .011, and P for trend in women = .003).

Figure 2.

To further examine the relationship between systolic BP change and BMI change, we used a multiple linear regression analysis with annual change of systolic BP as the dependent variable (Table 4). We included age, systolic and diastolic BPs, alcohol consumption, cigarette smoking, type 2 diabetes, and BMI at baseline, antihypertensive treatment at follow-up, and annual change of BMI as independent variables. Annual change of systolic BP was independently associated with annual change of BMI. After accounting for the effects of all those covariates plus diabetes incidence and educational attainment, annual change of systolic BP (mm Hg/year) was greater in San Antonio than in Mexico City for both men (β = 0.82, 95% CI: 0.45–1.07) and women (β = 0.76, 95% CI: 0.41–1.22).

Table 4

Multiple linear regression analysis with annual change of SBP as the dependent variable

	San Antonio *		Mexico City *
	β †	95% CI	β	95% CI
Men
Baseline age (× 10-yr interval)	0.42	(−0.01–0.85)	0.40	(0.20–0.60)
Baseline BMI (× 5-kg/m2 interval)	−0.02	(−0.41–0.36)	0.06	(−0.17–0.29)
Annual change in BMI (× 1-kg/m2 interval)	1.02	(0.02–2.02)	1.16	(0.72–1.60)
Women
Baseline age (× 10-yr interval)	0.29	(−0.03–0.60)	0.61	(0.43–0.80)
Baseline BMI (× 5-kg/m2 interval)	0.08	(−0.12–0.29)	0.22	(0.07–0.38)
Annual change in BMI (× 1-kg/m2 interval)	0.65	(0.07–1.22)	1.04	(0.67–1.41)

Abbreviations as in Tables 1–3.

*

Multiple linear regression analysis included age, SBP, DBP, alcohol consumption, cigarette smoking, type 2 diabetes, and BMI at baseline, and BMI change and treatment with hypertensive medications at follow-up as independent variables;

†

β indicates covariate-adjusted regression coefficients (mm Hg/yr).

Discussion

The risk for hypertension is greater in Mexican Americans from San Antonio than in subjects from Mexico City with differences even after adjusting for the effects of obesity, BP, and diabetes status. Similarly, annual change of systolic BP is greater in San Antonio than in Mexico City despite adjustment for BP, alcohol consumption, cigarette smoking, diabetes status, diabetes incidence, obesity, weight change, and antihypertensive treatment. Other determinants may be also important such as diet and salt consumption,¹⁵ physical activity,¹⁶ insulin resistance,¹⁷ inflammation,¹⁸ arterial stiffness,¹⁹ and altitude.²⁰ Fat consumption and physical activity are less favorable in San Antonio than in Mexico City.² However, they are not included in the analysis, because food frequency and physical activity questionnaires are available only at the follow-up examination in the SAHS.²

Baseline systolic BP is a strong predictor of hypertension,⁷ but baseline systolic BP correlates negatively with systolic BP change. This apparent contradiction may be due to a “regression of the mean” effect: the higher the systolic BP at baseline, the lower the future increment of systolic BP. A similar effect may be responsible for the negative correlations between baseline BMI and BMI change. In any case, systolic BP change is positively associated with BMI change^8,21 independently of baseline systolic BP and baseline BMI. In randomized control trials, weight loss has already shown to be associated with a delay/prevention of hypertension.²² Potential mechanisms for the effect of BMI change on systolic BP change are resistance to insulin²³ or leptin,²⁴ activation of the sympathetic nervous system and renin-angiotensin system,²⁵ or arterial stiffness.²⁶

In our cohorts of middle-aged subjects, men and women have comparable incidence of hypertension despite differences in BP and obesity. Men have greater BP at baseline; women, more obesity and weight gain. Vasan et al⁷ have described an independent association between baseline BMI and incident hypertension but without looking at sex differences. Wilsgaard et al⁸ have reported an independent association between baseline BMI and incident hypertension only in women. We describe also an association between baseline BMI and incident hypertension only in women. However, hypertension risk related to baseline BMI is explained, at least in part, by baseline BP. Sex-specific differences in hypertension risk may be conditioned by the stronger relationship between BP and insulin resistance in women than in men.²⁷

Age is also a predictor of hypertension, and is associated with systolic BP change independently of the effects of BP, obesity, and glucose tolerance status. Age-related hypertension risk may be mediated by insulin resistance²⁸ and arterial stiffness,²⁹ and has been described absent in populations with very low salt or fat consumption or with no association between BMI and age.^15,30

Finally, this study has several strengths. MCDS and SAHS follow similar sampling methods and survey protocols, and have comparable recruitment rates. Eligible participants have similar socioeconomic status, ethnic ancestry, age, and treatment rates with antihypertensive medications. Nevertheless, there are also some limitations. Hypertension status at follow-up is missing in one-fifth of the eligible participants. Follow-up rates are lower in young nondiabetic subjects from the SAHS than in those from the MCDS. Self-reported incident hypertension by phone interview is more prevalent in SAHS participants who returned for a follow-up examination than in those who did not. In spite of these difficulties, bias remains unlikely. In both cities, subjects with and without follow-up examination have similar baseline characteristics. In addition, the incidence of hypertension is higher in Mexico City than in San Antonio for men and women of all age groups.

In summary, our finding of a lower hypertension risk in Mexico City than in San Antonio is likely to be for reasons other than the effect of obesity, diabetes, and BP. Baseline systolic BP is a strong predictor of hypertension, and a significant determinant of the hypertension risk associated with baseline BMI among women. Weight change, an essential target for the prevention of hypertension,²² is positively associated with systolic BP change. Although secular trends in hypertension are conflicting in the US,^31,32 worsening obesity³³ may hinder the efforts on the prevention, treatment, and control of hypertension.