A Longitudinal Study of Hypertension Risk Factors and Their Relation to Cardiovascular Disease

According to the report of the Seventh Joint National Committee on Hypertension, hypertension (HT) is now a major public health problem affecting &50 million individuals in the United States and &1 billion individuals worldwide, and, as the population ages, the prevalence of HT will increase even more unless broad and effective preventive measures are implemented.¹ Epidemiological studies conducted over the past decades have shown continuous, consistent, and independent relations of HT to cardiovascular disease (CVD).^2–5 Numerous cross-sectional and cohort studies have provided important information on HT prevalence and incidence and their correlates for white, black, and other ethnic populations, with fewer studies conducted in American Indians.^6–15 CVD is now the leading cause of death in American Indians, and HT is a strong predictor for the development of diabetes-associated complications, such as retinopathy and renal disease.^{3–6,16–18} The Strong Heart Study (SHS), initiated in 1988, is a longitudinal study of CVD in 13 American Indian tribes/communities in Arizona (AZ), North and South Dakota (ND/SD), and Oklahoma (OK).¹⁹ The data collected longitudinally from the 3 exams (1989–2001) in this population, which has a high rate of diabetes and CVD, provided an excellent opportunity to explore risk factors for the incidence of HT and the association between HT and CVD.

Methods

A total of 4549 American Indian men and women, aged 45 to 74 years, in 13 Indian tribes/communities in AZ, ND/SD, and OK, participated the SHS baseline examination from 1989 to 1991. The cohort was followed and reexamined in 1992–1994 and 1996–1999. A detailed description of the study design and methods of the SHS has been reported previously.¹⁹ Each follow-up examination included a personal interview and a physical examination. Blood was drawn at each examination after a 12-hour fast, and low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), triglycerides (TG), and fasting plasma glucose (FPG) were measured. Diabetic status was defined by the 2003 American Diabetes Association updated criteria²⁰ as: diabetes (DM), if FPG≥7.0 mmol/L (126 mg/dL) or receiving insulin or oral hyperglycemic treatment; impaired FPG (IFG), if 5.6≤FPG<7.0 (100 mg/dL≤FPG<126 mg/dL); and normal FPG (NFG), if FPG<5.6. A urine sample was taken for measurements of albumin and creatinine. Albuminuria was determined using the ratio of urinary albumin and creatinine (UACR): microalbuminuria if 30≤UACR<300; and macroalbuminuria if UACR≥300. Obesity status was defined as: normal weight, if body mass index (BMI) <25; overweight, if 25≤BMI<30; and obese, if BMI≥30.

Three measurements of systolic and diastolic blood pressures were taken on the right arm with an appropriately sized cuff using a Baum mercury sphygmomanometer (W. A. Baum Co) after the participant had been resting in a seated position for 5 minutes. The average of the second and third measurements was used as the blood pressure value for individual participants. According to the Seventh Joint National Committee on Hypertension criteria,¹ HT was defined as systolic blood pressure (SBP) ≥140 mm Hg, diastolic blood pressure (DBP) ≥90 mm Hg, or current use of antihypertensive medication; pre-HT if SBP was 120 to 139 mm Hg and DBP was 80 to 89 mm Hg; and normal, if SBP was <120 mm Hg and DBP was <80 mm Hg. HT control was defined as treatment with antihypertensive medication and a measured blood pressure of <140/90 mm Hg. The fatal or nonfatal CVD events until the end of 2001 for each participant were obtained using standard methods for identification and confirmation of events as described previously.^5,19 Leisure-time exercise and occupation-related physical activities were measured with the instrument developed by Kriska et al,²¹ which was modified for use with American Indians.⁶

All of the observers for the SHS were centrally trained and certified before each examination. Various standardized protocols (including those for blood pressure measurements) were used in training to make sure the procedures and methods that the certified observers would perform were consistent among the different sites at the 3 SHS centers.¹⁹ Indian Health Service Institutional Review Board, Institutional Review Boards of the participating institutions, and the participating tribes approved the study. Informed consent was obtained from all of the participants.

Statistical Analyses

Generalized linear models with unstructured covariance²² were used to adjust the related observations from 3 exams for each participant and to identify the risk factors of HT and the correlates of SBP or DBP. Cox proportional models with time-dependent covariates²³ were used to explore additional CVD risk factors other than those HT risk factors. In the generalized linear model for HT incidence, risk factor (independent) measurements from non-HT participants of an examination and HT status (outcome) from the next examination (averaged 4 years from the previous examination) were used. Therefore, the estimated HT incidences from the model were the 4-year cumulated HT incidences. In the generalized linear models for SBP and DBP, correlate (independent) measurements from nonhypertensive participants of an examination and SBP and DBP (dependent) measures from the next examination (except those participants who took antihypertensive medications before the next examination) were used. Therefore, the final models show the association of correlate measurements to the 4-year-later SBP and DBP measures. In the Cox proportional models, for each participant free of CVD at baseline, the CVD-free time (duration from the baseline examination to the time of the first diagnosis of CVD event, death, or the end of 2001) and CVD status were obtained from the SHS mortality and morbidity surveillance data. A 2-stage method was applied in the Cox proportional models with time-dependent covariates.²⁴ First, the mixed models²⁵ with random intercept and slope were applied to risk factor measurements obtained from all 3 of the examinations to estimate their values at 1 year before each CVD event or death. The 1-year lag is for studying the predictability of risk factors to CVD and possible confounding of the CVD event. These estimated risk factor measurements were then used as time-dependent covariates in Cox models. The time-dependent covariates were HT, DM, obesity, albuminuria, current alcohol drinking and smoking status, age, HDL, and LDL. Statistical significance was defined as P<0.05 for all of the tests.

Results

The estimated 4-year cumulated HT incidences by sex, center, and age groups are shown in Table 1. There were no sex differences in HT incidence among American Indians. For the same sex and age group, the HT incidence of American Indians in ND/SD and OK were not significantly different, but both of them were significantly &30% lower than that in AZ. For the same center and sex, there were significant age differences in HT incidence, with participants ≥65 years of age &38% higher than the aged 55 to 64 years group and &62% higher than the aged 45 to 54 years group. Those 55 to 64 years of age had marginally (P=0.051) higher HT incidence than those 45 to 54 years of age. For later comparison and discussion, the HT incidences for those 45 to 64 years of age are also shown in Table 1.

Table 2 shows comparisons of risks of HT among subgroups in each characterized group after adjusting for center, age, and sex. Prehypertensive American Indians had &3.5 times higher risk of developing HT than normotensive subjects. Similar assessments for BMI subgroups and other risk factors showed different risks as follows: obese versus normal, 1.9 times; overweight versus normal weight, 1.46; diabetic status (DM versus NFG, 2.3; DM versus IFG, 2.0), albuminuria (macroalbuminuria versus normal, 5.0; microalbuminuria versus normal, 2.1), and TGs [>2.24 mmol/L versus ≤2.24 mmol/L (>200 mg/dL versus ≤200 mg/dL), 1.35]. Insulin resistance [homeostasis model assessment insulin resistance=insulin*(FPG*0.05551)/22.5], as measured by the homeostasis model assessment, also showed a significant increasing risk of HT of &1.5 times with each increasing quartile. Those with a history of current alcohol consumption, a parental history of HT, a parental history of DM, or with higher insulin concentration had higher risk of HT than those without these factors. However, those who were current tobacco users had a lower risk of HT than those who were not, and there were no significant differences in the risk of HT among subgroups in Indian heritage (<25, 25 to 49, 50 to 74, and ≥75%), education (<12 years education and ≥12 years), quartile of physical activity, LDL [<3.36 mmol/L, 3.36 to 4.13 mmol/L, ≥4.14 mmol/L) (<130 mg/dL, 130 to 159, ≥160)] group.

From the generalized linear model for HT (Table 3), the risk factors in this model were selected by the stepwise selection method with 0.15 significant entry, and stay levels among all of the significant risk factors in Table 2 except TGs, pre-HT, obese/overweight, current alcohol drinking, DM, and macro/microalbuminuria are combined risk factors of HT after adjusting for center, age, and sex. Multivariate analyses of possible correlates of SBP or DBP are shown in the respective models. These 2 models included all of the variables in the HT model and those additional variables that were selected among the other significant risk factors except TGs in Table 2 by the stepwise selection method with the same criteria as above. During the averaged 4 years of follow-up, those with pre-HT would develop &10 mm Hg higher SBP and 4 mm Hg higher DBP than normotensive participants provided the other measures were the same. Higher obesity status predicted higher SBP and DBP. Those participants with older age or more severe albuminuria status developed higher SBP. However, older age was associated with lower DBP. DBP was not significantly affected by albuminuria status. Years of education and alcohol drinking were significantly and positively correlated with DBP but not SBP. Smoking was a significant and negative predictor of both SBP and DBP. The model for predicting HT is shown at the bottom of Table 3 and is obtained by using those significant variables (continuous version) in the HT incidence model.

The results (data not shown), obtained by adding a dummy variable to index the period of the second examination to the third examination in the generalized linear model for all of the participants in Table 3, showed that risk of developing HT was significantly increased &54% during the period of the second examination to the third examination compared with the period of the baseline to the second examination (odds ratio, 1.54; P<0.0001) in the SHS cohort participants after adjustments for all of the risk factors in the model.

For those participants free of CVD at baseline, Table 4 shows the association of HT and additional risk factors to CVD in the Cox models for male, female, or all participants. The Cox models included HT status, LDL, HDL, all of the variables in the HT model, and 2 additional variables, smoking and physical activity. HT status, LDL, HDL, and all of the variables in the HT model were forced into the Cox models, and additional smoking and physical activity were selected with 0.15 significant entry and stay levels from the other risk factors except TGs in Table 2. In Table 4, we only showed the part of the Cox models that related to HT status and the selected additional 2 risk factors, smoking and physical activity, to assess their association with CVD. Those significant HT risk factors, such as age, diabetic status, and macro/microalbuminuria, remained significant for CVD incidence in the models (data not shown). For all/male/female participants, after adjusting all of the other risk factors, uncontrolled, controlled, untreated hypertensive, and prehypertensive participants had hazard ratios of developing CVD of 2.77/2.68/2.84, 2.19/2.33/2.11, 1.81/1.77/1.86, and 1.74/1.84/1.69, respectively, compared with normotensive participants. Moreover, in all/female participants, uncontrolled hypertensive participants had significantly higher risk of developing CVD than the controlled hypertensive (P=0.0276/0.0342), untreated hypertensive (P=0.0018/0.0176), or prehypertensive participants (P≤0.0001/0.0002). However, in males, uncontrolled hypertensive participants had significantly higher risk of developing CVD than prehypertensive participants (P=0.0228), marginally higher than untreated hypertensive participants (P=0.0557), and not significantly higher than controlled hypertensive participants (P=0.4124). There were no significant differences in the risk of developing CVD between untreated hypertensive and prehypertensive males (P=0.8509)/females (P=0.5891), between untreated hypertensive and controlled hypertensive males (P=0.1669)/females (P=0.4873), or between prehypertensive and controlled hypertensive males (P=0.0732)/females (P=0.0953).

Discussion

The SHS data have provided an opportunity for prospective analysis of HT in American Indians and for the assessment of associations of HT to CVD incidence. The risk of developing HT is rising in American Indians from our study. This increasing tendency may be one of the reasons for the rising tide of CVD in American Indians,⁵ which coincides with national and world figures.^1,6–14 As shown in most HT studies, baseline blood pressure measurement is also a significant risk factor of HT in American Indians, especially for those prehypertensive American Indians, who had 3.21 times subsequent risk of being hypertensive when compared with normotensives.

Our results show that American Indians aged 65 to 74 years have significantly higher HT incidence than aged 55 to 64 and 45 to 54 years, even after adjusting for other risk factors. This finding coincides with most of the results reported in the literature for other ethnic groups.^8–15 SBP increases, but DBP decreases with age in American Indians. This means that the higher HT incidence in older American Indians was largely because of the increasing in SBP rather than in DBP. In fact, that percentage of hypertensive participants isolated by SBP of ≥140 mm Hg only is much more than that by DBP of ≥90 mm Hg only in SHS participants (64.5% versus 13.1% in the baseline examination, 74% versus 7.4% in the second examination, and 73.3% versus 7.5% in the third examination).

Center differences in HT incidence (Table 1) were diminished after adjusting for other factors (Table 3). Additional exploration shows that such differences were largely because of the UACR and BMI differences among the 3 centers, with ND/SD and OK American Indians having significantly lower UACR (P<0.0001) and BMI (P<0.0001) than AZ.

For either aged 45- to 64-year-old females or males, AZ American Indians have a higher average annual HT incidence than US whites and blacks (7.6% versus 3.2% and 6.6%, respectively in females; 7.8% versus 3.8% and 6.0% in males), whereas OK (5.6% in females and 5.8% in males) and ND/SD (5.8% in females and 6.0% in males) American Indians have a higher average annual HT incidence than US whites but similar rates with US blacks in the Atherosclerosis Risk in Communities and the Monitoring Trends and Determinants of Cardiovascular Disease studies.⁸ It has been noted that some ethnic differences in HT incidence may be explained by differences of HT risk factors, such as baseline blood pressures and obesity, among ethnic groups.⁸

There were no significant sex differences in HT incidence among American Indians, although males had significantly higher DBP than females. This may reflect that SBP of ≥140 mm Hg is the predominant form of HT rather than DBP of ≥90 mm Hg, and the males had similar SBP levels to females. Diminishing sex differences in older age groups within different ethnic groups was also reported in the literature.^8,13,26–30

Albuminuria is the most significant risk factor of HT in American Indians. Those macroalbuminuria and microalbuminuria participants had 3.47 and 1.72 times the risk, respectively, of developing HT compared with the normal albuminuria participants. Our results also show that UACR is significantly and positively associated with SBP but not with DBP. Albuminuria occurs mainly in the individuals with IFG and DM.^31–34 Similar to the reported results in most studies,^35–39 DM and obesity are also significant risk factors of HT in American Indians. These 2 risk factors should receive attention in future intervention programs, because their prevalence rates are very high in American Indians.^40–43

Our results also show that drinking alcohol increased the risk of being hypertensive and significantly increased DBP but not SBP. This positive association was also observed in other HT studies.^44,45 Smoking was not significantly and independently related to HT but was significantly and negatively related to both SBP and DBP in American Indians. Some studies in the literature report similar results, but some others do not.^8,10,46 The reason is unknown. Education was inversely related to DBP in our data. This may reflect various aspects of lifestyle, including diet and stress, that may affect the DBP.^9,15,47,48

Our results from the Cox models showed that HT is a strongly independent risk factor of CVD. After adjusting all of the other risk factors, those pretreated, untreated, controlled, and uncontrolled hypertensive participants had &1.74, 1.81, 2.19, and 2.77 times higher risk of developing CVD compared with normotensive participants, respectively. This is especially so for those uncontrolled hypertensive participants who had a significantly higher risk of developing CVD than all of the other groups. Our results show that there were no significant differences in the risk of developing CVD among those with pre-HT, untreated HT, or controlled HT. Additional tests show that although it may appear that those with untreated HT more closely resemble (with regard to CVD risk) those with pre-HT than those with controlled HT, the difference (the hazard ratio of untreated HT to pre-HT versus the hazard ratio of controlled HT to untreated HT) was not statistically significant (P=0.3922 for males, 0.9268 for females, and 0.5379 for all participants). This finding suggests that those with pre-HT should get blood pressure treatments as soon as possible to reduce their blood pressure to avoid the high risk of developing HT and CVD.

Perspectives

Our results show that the risk of developing HT is rising, the incidence of HT is high, and SBP of ≥140 mm Hg is the predominant form of HT in the SHS American Indian participants. Prehypertensive American Indians were at 3.2 times the risk of developing HT and at 1.74 times the risk of developing CVD as those normotensives. Age, DM, and macro/microalbuminuria were significant risk factors of both HT and CVD incidence. Being elderly, prehypertensive, obese/overweight, a current alcohol drinker, having DM, and having macro/microalbuminuria are combined significant HT risk factors. Therefore, intervention measures should emphasize the modifiable risk factors, such as blood pressure, obesity, plasma glucose, and albuminuria, to prevent HT, as well as CVD.

Footnotes