Regional and Racial Differences in Response to Antihypertensive Medication Use in a Randomized Controlled Trial of Men With Hypertension in the United States

Background  Stroke incidence and mortality rates are higher in the southeastern region of the United States, which is called the "Stroke Belt." We compared the response to antihypertensive medication use in patients from different US regions.

Methods  The short-term and 1-year efficacy of the antihypertensive medications hydrochlorothiazide, atenolol, diltiazem hydrochloride (sustained release), captopril, prazosin hydrochloride, and clonidine was compared by US region in a randomized controlled trial of 1105 men with hypertension from 15 US Veterans Affairs medical centers.

Results  Compared with patients outside the Stroke Belt, patients inside the Stroke Belt achieved significantly lower treatment success rates of diastolic blood pressure control at 1 year with hydrochlorothiazide (63% vs 41%), atenolol (62% vs 46%), captopril (60% vs 30%), and clonidine (69% vs 43%); there were no differences in treatment success rates with diltiazem (70% vs 71%) or prazosin (54% vs 53%). When controlling for race, patients inside the Stroke Belt had significantly lower treatment success rates with hydrochlorothiazide (P = .003) and clonidine (P = .003), and the lower success rate with atenolol approached significance (P = .15). Regardless of region, blacks were less likely than whites to achieve treatment success with atenolol (P = .02) or prazosin (P = .03) and more likely with diltiazem (P = .05). There was a trend for blacks residing inside the Stroke Belt to have a lower treatment success rate than other race-region groups when treated with captopril (P = .07). Many regional and racial differences in diet, lifestyle, and other characteristics were observed. After adjustment for these characteristics by regression analysis, the effect of residing inside the Stroke Belt remained for captopril (P = .01) and clonidine (P = .01) and approached significance for hydrochlorothiazide (P = .10).

Conclusions  Hypertension in patients residing inside the Stroke Belt responded less to the use of several antihypertensive medications and important differences were shown in a number of characteristics that may affect the control of blood pressure, compared with patients residing outside the Stroke Belt.

STROKE ACCOUNTS for 150,000 deaths annually and ranks third among all causes of death behind heart disease and cancer. In addition, there are approximately 500,000 new cases of stroke each year in the United States.¹ Some regions of the United States have significantly higher stroke incidence and mortality rates than other parts of the country.^2-4 Data from 1986^5-7 showed that there were 12 states with stroke mortality rates more than 10% above the mean rate for the rest of the United States: 10 of these states are in the southeastern region. This region has come to be known as the "Stroke Belt" because a higher stroke mortality risk in this region has persisted for more than 50 years despite major migrations in and out of the region and a 75% decrease in stroke mortality in the United States during that time.^3,8

Hypertension, one of the most important risk factors for stroke, is more prevalent and more severe in the Stroke Belt compared with other US regions.^3,9 Hypertension is recognized and treated in patients residing in the Stroke Belt as often as in patients in other regions, but it is less well controlled. Whether hypertension prevalence, severity, or control rates contribute to an increased risk in patients inside the Stroke Belt has not been proved. However, death rates from other complications associated with hypertension, such as myocardial infarction and renal failure, are also higher in the Stroke Belt.³

There are similar patterns for all-cause mortality, particularly in recent years. Although cardiovascular mortality has progressively decreased, especially since 1970, there is evidence that the decline did not persist in the 1990s and leveled out.¹⁰ Many factors have been proposed as contributors to the higher rates of stroke and other complications associated with hypertension in the Stroke Belt, but none have been clearly incriminated.⁷ Therefore, it is important to further examine why the risk of stroke and to some degree cardiovascular mortality is higher in the Stroke Belt so that appropriate public health or other prevention measures can be recommended. To increase our understanding of the Stroke Belt, we compared the antihypertensive efficacy of hydrochlorothiazide, atenolol, diltiazem hydrochloride (sustained release), captopril, prazosin hydrochloride, and clonidine in men with hypertension who resided inside and outside the Stroke Belt in the Department of Veterans Affairs Single-Drug Therapy for Hypertension Study.^11,12 Also, we report differences in patient characteristics according to region of residence. Since we have previously reported differences in blood pressure (BP) response to antihypertensive medication by race, we also examined the effect of residing inside the Stroke Belt by race.

The methods for this cooperative study were previously described in more detail.¹¹ Briefly, 1292 men from 15 US Veterans Affairs medical centers with untreated (placebo) diastolic BP (DBP) between 95 and 109 mm Hg on 2 consecutive visits were randomly assigned to receive hydrochlorothiazide (12.5-50.0 mg once daily), atenolol (25-100 mg once daily), sustained release diltiazem hydrochloride (60-180 mg twice daily), captopril (12.5-50.0 mg twice daily), prazosin hydrochloride (2-10 mg twice daily), clonidine (0.1-0.3 mg twice daily), or placebo. The placebo group was not included in this article (n = 187). All medications had 3 dose levels available for titration to a DBP goal of less than 90 mm Hg. If the DBP was controlled and the medication was tolerated, the patient entered a 1- to 2-year maintenance phase. The primary outcome measure of the trial was treatment success, defined as achieving the DBP goal during titration and maintaining a DBP of less than 95 mm Hg for 1 year.

As described previously,¹¹ demographics, medical history, and the results of physical examination were obtained prior to randomization. Blood pressure; heart rate; weight; height; urinary creatinine, sodium, and potassium excretion; serum creatinine, sodium, potassium, and fasting glucose levels; and plasma renin activity were measured prior to randomization while patients were taking the placebo. All but demographics and plasma renin activity were obtained periodically throughout the trial. The BP at each visit was the mean of 3 measurements (systolic and diastolic pressures were the first and fifth phases of Korotkoff sounds, respectively) obtained with a standard mercury sphygmomanometer while the patient was seated in a chair that supported the patient's back. Medication adherence rate was defined as the percentage of prescribed medications a patient apparently took based on pill count.

Patients from the 15 sites were classified according to their residence: (1) patients from Dallas and Houston, Tex; Jackson, Miss; Memphis, Tenn; St Louis, Mo; and Washington, DC, were classified as residing inside or at the periphery of the Stroke Belt; and (2) patients from Allen Park, Mich; Boston, Mass; East Orange, NJ; Manhattan, NY; Miami, Fla; Milwaukee, Wis; San Francisco, Calif; San Juan, Puerto Rico; and Topeka, Kan, were classified as residing outside the Stroke Belt.¹³

Actual income data were not collected as part of the original study. However, since stroke mortality rates are higher in areas with lower mean socioeconomic levels,¹⁴ we linked patient ZIP codes to US Census Bureau income data, which are reported by ZIP code, and used the median household income for the ZIP code area in which each patient resided as a proxy for actual income.

Baseline characteristics of the 4 groups of patients (whites residing inside the Stroke Belt, blacks residing inside the Stroke Belt, whites residing outside the Stroke Belt, and blacks residing outside the Stroke Belt) were compared using 2-way analysis of variance for continuous data and logistic regression for categorical data.

The mean change in DBP and systolic blood pressure (SBP) from baseline to the end of dose titration was calculated for each treatment group, and the mean values were compared using 2-way analysis of variance, which examined the effect of race, region, and their interaction on BP reduction.

Logistic regression analysis for each treatment group was performed to analyze the effect of race and region on the probability of treatment success at 1 year. Treatment success rate was defined as the percentage of randomized patients who achieved a DBP of less than 90 mm Hg at 2 consecutive visits during the dose titration period and who maintained a DBP of less than 95 mm Hg for 1 year. In addition, a second set of logistic regression analyses was performed to evaluate whether the effects of race and/or region persisted after adjustment for other patient covariates, including income; adherence to the medication regimen; age; baseline DBP; body mass index; alcohol consumption (percentage who consumed <1 drink per day or ≥ 1 drink per day); smoking status (smoker or nonsmoker); heart rate; 24-hour urinary potassium and sodium excretion; plasma renin activity; and serum potassium, creatinine, and fasting glucose levels.

All analyses were performed using SAS statistical software (version 6; SAS Institute, Cary, NC). The criterion for statistical significance was P≤.05, and all statistical tests were 2-sided. For the logistic regression analyses, P≤.20 was the criterion for variable inclusion in the model.

Table 1 displays baseline characteristics for each of the 4 groups. Independent of race, patients inside the Stroke Belt were younger and had lower serum potassium levels, lower plasma renin activity, and higher SBP, DBP, body mass index, heart rate, and urinary sodium excretion. Also, there was a trend for patients inside the Stroke Belt to have lower urinary potassium excretion and alcohol intake, a larger left ventricular mass, and higher serum creatinine levels. Independent of region, blacks were younger and had lower estimated household incomes, medication adherence rates (although pill counts were still above 90%), urinary sodium and potassium excretion, serum potassium and fasting glucose levels, alcohol intake, and plasma renin activity compared with whites. Also, blacks had higher DBPs, serum creatinine levels, and current smoking rates compared with whites. There was a trend for blacks inside the Stroke Belt to have a larger left ventricular mass and higher serum creatinine levels compared with the other 3 treatment groups.

Compared with patients outside the Stroke Belt, patients inside the Stroke Belt achieved significantly lower 1-year treatment success rates of DBP control with hydrochlorothiazide, atenolol, captopril, and clonidine, but there were no differences with diltiazem or prazosin (Figure 1). We previously defined a clinically significant difference in the 1-year treatment success rate as 15% or more.¹¹ Prazosin was at least 15% less successful in controlling DBP than diltiazem or clonidine in patients outside the Stroke Belt. In patients inside the Stroke Belt, diltiazem was at least 18% more successful than any other agent, and captopril was at least 15% less successful than atenolol, prazosin, or diltiazem.

The 1-year control rates of DBP in patients outside the Stroke Belt compared with patients inside the Stroke Belt for each medication according to race are shown in Figure 2. When controlling for race, patients inside the Stroke Belt achieved significantly lower treatment success rates with hydrochlorothiazide (P = .003) and clonidine (P = .003), and lower treatment success rates approached significance with atenolol (P = .15). Regardless of region, blacks taking atenolol were less likely (P = .02) to achieve treatment success and this decreased success appeared to be accentuated among blacks residing inside the Stroke Belt, although the interaction between race and region was not selected for the model. For patients taking captopril, an interaction between race and region approached significance (P = .07): blacks residing inside the Stroke Belt had the lowest likelihood of being classified as a treatment success, while the other 3 groups taking captopril showed similar levels of success. Regardless of region, blacks taking diltiazem were more likely (P = .05) and blacks taking prazosin were less likely (P = .03) to achieve treatment success compared with whites.

In blacks outside the Stroke Belt, diltiazem was at least 15% more successful than all other medications, and prazosin was at least 15% less successful than all other medications. For blacks inside the Stroke Belt, diltiazem was at least 28% more successful and captopril was at least 18% less successful than all other medications. In whites outside the Stroke Belt, there was less than 15% difference in the treatment success rate between any 2 medications. For whites inside the Stroke Belt, hydrochlorothiazide was at least 25% less successful than all other medications. The other medications had clinically similar (<15% difference) treatment success rates for whites inside the Stroke Belt.

Table 2 shows the significance of race and region on treatment success at 1 year for each treatment after adjusting for patient covariates. Variable selection for the models was stepwise, retaining variables with P<.20 but also keeping Stroke Belt status and race in the models. For patients taking hydrochlorothiazide, the negative effect of residing inside the Stroke Belt was weakened (P = .10), but black patients (P = .03), those who smoked (P = .02), and patients with higher baseline serum potassium levels (P = .01) were significantly more likely to achieve treatment success. Older age was almost a significant predictor of response to hydrochlorothiazide (P = .06). Covariate adjustment weakened the effect of race (P = .27) and residence in the Stroke Belt (P = .34) for atenolol, although a higher serum potassium level was associated with higher treatment success with atenolol (P = .03); smokers had lower treatment success with atenolol than nonsmokers (P = .08). For captopril, Stroke Belt residence (P = .01), black race (P = .02), and increasing age (P = .05) were negatively associated with treatment success. Negative effects of Stroke Belt residence for clonidine remained significant (P = .01). In addition, older age (P = .01) and lower urinary sodium excretion (P = .02) were associated with treatment success with clonidine. For diltiazem, black race showed a weakened positive effect (P = .14), although older age was associated with treatment success (P = .03). For prazosin, covariate adjustment eliminated the effect of race, although higher 24-hour urinary potassium excretion (P = .02) and higher medication adherence rate (by pill counts) (P = .02) were associated with higher treatment success with prazosin; there was a trend for higher serum creatinine levels to predict a higher treatment success rate with prazosin. In these analyses, household income, body mass index, alcohol intake, heart rate, or, as we have reported elsewhere,¹⁵ plasma renin activity did not show significant effects on 1-year treatment success rates for any of the medications in this trial.

Table 3 displays the mean change in DBP after dose titration by race, region, and treatment. Regardless of race, patients inside the Stroke Belt had less reduction in DBP compared with patients outside the Stroke Belt with hydrochlorothiazide and captopril, and there was a trend for patients inside the Stroke Belt to have less reduction with atenolol. Regardless of region, blacks had less reduction in DBP compared with whites with hydrochlorothiazide, atenolol, captopril, clonidine, and prazosin, and a greater reduction in DBP with hydrochlorothiazide and diltiazem. The race effect for clonidine approached significance (P = .07). For black patients who were taking atenolol and captopril, a smaller DBP decrement for patients residing inside the Stroke Belt approached significance compared with blacks outside the Stroke Belt.

In Table 4, the mean changes in SBP after dose titration are displayed by race, region, and treatment. Region did not have an effect on the decline in SBP for any treatment. Blacks showed lesser declines in SBP while taking atenolol, captopril, and prazosin and a greater decline while taking diltiazem compared with whites.

To our knowledge, this study is the first report of patients with hypertension exhibiting resistance to antihypertensive therapy, independent of baseline BP or race, related to region of residence in the United States. The most important finding in these analyses was that 4 of the 6 medications studied in this trial were less effective or tended to be less effective in treating hypertension for patients inside the Stroke Belt compared with patients outside the Stroke Belt. Blood pressure response to prazosin and diltiazem was not influenced by region of residence. In treating hypertension in blacks, atenolol and prazosin were less effective and diltiazem was more effective.

Important regional and racial differences in diet, lifestyle, and other patient characteristics also were observed. When controlling for race, patients inside the Stroke Belt had higher values for SBP, DBP, body mass index, heart rate, urinary sodium excretion, and serum creatinine and a larger left ventricular mass; and they had lower serum potassium levels, alcohol intake, and plasma renin activity. Blacks had higher DBPs, serum creatinine levels, and smoking rates; and lower household incomes, medication adherence rates, 24-hour urinary sodium and potassium excretion, serum potassium and fasting glucose levels, alcohol intake, and plasma renin activity. The values for serum potassium, plasma renin activity, and 24-hour urinary sodium and potassium excretion for the 405 blacks and whites inside the Stroke Belt in this cohort were very similar to the same values reported by our group¹⁶ from a cohort of 623 veterans from the southeastern United States observed approximately 10 years before the present cohort was examined. In that earlier cohort, we suggested that serum potassium levels may be a more stable reflection of potassium intake than potassium excretion from a single 24-hour urine sample. In addition, as observed in the earlier cohort, blacks in the present cohort had lower 24-hour urinary potassium excretion, serum potassium levels, and plasma renin activity compared with whites. These observations are likely a reflection of low long-term potassium intake, as low renin levels are increased to the levels of whites when potassium intake is supplemented during several months.¹⁷

When regression analysis was performed to analyze whether regional differences in the control of BP persisted after adjustment for these patient characteristics, the effect of residing inside the Stroke Belt was weakened, although the trend was still there for hydrochlorothiazide. The effects of hydrochlorothiazide on patients inside the Stroke Belt were weakened after adjustment for serum potassium levels, age, and baseline DBP. The effect of residing inside the Stroke Belt was no longer significant for atenolol; serum potassium levels and a history of smoking weakened the effect of residing inside the Stroke Belt. Cigarette smoking was previously demonstrated to reduce the antihypertensive effectiveness of β-blockers.¹⁸ Stroke Belt status and baseline DBP weakened the atenolol effect based on race. Residing inside the Stroke Belt was as important or more important a predictor for treatment success for captopril and clonidine after controlling for other characteristics. For prazosin, race effects were weakened by urinary potassium excretion, medication adherence rate, and baseline DBP.

A variety of factors, such as plasma renin activity, obesity, fasting glucose level, heart rate, alcohol intake (dichotomized at 1 drink per day because of the low prevalence of even moderate levels of drinking), and estimated household income did not predict a greater or lesser response to any of the study medications. However, lower urinary sodium excretion predicted higher treatment success rates with clonidine. Higher urinary potassium excretion and medication adherence rate only predicted higher treatment success rates in patients treated with prazosin. Age predicted higher treatment success rates with clonidine and diltiazem and lower treatment success rates with captopril and approached significance as a predictor of higher treatement success rates with hydrochlorothiazide. Smoking predicted higher treatment success rates with hydrochlorothiazide and lower treatment success rates with atenolol. Serum potassium levels strongly predicted treatment success with hydrochlorothiazide and atenolol.

We do not believe that the resistance to antihypertensive medications is the cause of the higher stroke mortality in the Stroke Belt, since the higher stroke mortality has been observed for more than 50 years, which includes an era prior to effective and widespread treatment of patients with hypertension. However, we suggest that environmental or other factors may contribute to both the higher stroke mortality and resistance to antihypertensive medications observed in patients residing inside the Stroke Belt.

From a practical management perspective, it can be expected that both black and white patients with hypertension in the Stroke Belt will often require more or different antihypertensive medications, more aggressive lifestyle changes, or both to reach recommended goals for BP¹⁰ compared with patients residing outside the Stroke Belt. Although diltiazem, a calcium antagonist, was equally effective in patients residing inside and outside the Stroke Belt, the preferred initial choice of antihypertensive medications should still be based on results from clinical trials designed to assess the effects of antihypertensive medication use on morbidity and mortality rates: diuretics and, to a lesser degree, β-blockers are preferred, although calcium antagonists have been shown recently to reduce strokes in patients with stages 2 and 3 isolated systolic hypertension,¹⁹ as previously demonstrated with diuretic-based therapy.²⁰ Resolution of whether there are differences in major cardiovascular outcomes between various classes of antihypertensive agents depends on the results of ongoing studies, such as the Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial (ALLHAT).²¹ However, until these studies are completed, once compelling indications for antihypertensive medications have been considered, predicted efficacy based on results from studies, such as the present report from the Department of Veterans Affairs Cooperative Study Group on Antihypertensive Agents, can help guide clinicians in prescribing effective antihypertensive therapy.

Although the geographic distribution of stroke mortality in the United States has changed somewhat in recent years, the southeastern United States has a higher stroke mortality than any other region of the country and presents a major public health challenge. Our study demonstrates important regional differences in a number of dietary, lifestyle, and other characteristics and resistance to single-drug antihypertensive therapy for patients with hypertension residing in this region.

Accepted for publication July 19, 1999.

Presented in part at and published as an abstract by the American Society of Hypertension 10th Annual Scientific Meeting, New York, NY, May 18, 1995 (Am J Hypertens 1995;8[4, pt 2]:34A).

Reprints: William C. Cushman, MD, Veterans Affairs Medical Center (111Q), 1030 Jefferson Ave, Memphis, TN 38104 (e-mail: William.Cushman@med.va.gov).