Diuretic Versus α-Blocker as First-Step Antihypertensive Therapy

Study Participants

The rationale and design of ALLHAT have been presented in detail elsewhere.^1,3 Recruitment was designed to generate a sample reflecting patients actually seen in the kind of practice settings where most care is provided.⁴ Men and women ≥55 years of age with systolic blood pressure (SBP) ≥140 mm Hg or diastolic blood pressure (DBP) ≥90 mm Hg or on medications for hypertension and an additional risk factor for CHD were eligible to participate in ALLHAT. The risk factors could include previous MI or stroke, left ventricular hypertrophy by electrocardiogram or echocardiogram, previously diagnosed type 2 diabetes, current cigarette smoking, low HDL cholesterol level, or certain other manifestations of CVD.^3,5 Patients with HF or known low ejection fraction were excluded. There were 15 255 participants randomized to chlorthalidone and 9061 participants randomized to doxazosin from February 1994 to January 1998 in 623 centers in the United States, Canada, Puerto Rico, and the US Virgin Islands. All participants signed an informed consent form, and all centers received Institutional Review Board approval.

Treatment

Randomization was stratified by center and blocked over time to maintain the ratio of 1.7:1. The BP treatment goal was SBP/DBP <140/90 mm Hg. Dosages for doxazosin were 2, 4, or 8 mg/d and for chlorthalidone were 12.5, 12.5 (sham titration), or 25 mg/d. Blinded 1 mg and 12.5 mg does of doxazosin and chlorthalidone, respectively, were used during the first week to minimize postural hypotension associated with doxazosin. Participants meeting the BP goal at maximal tolerated doses of these agents could receive open-label drugs, including study-provided atenolol, reserpine, clonidine, and/or hydralazine. However, use of open-label medications from 1 of the masked classes of drugs was to be avoided unless SBP was >160 mm Hg and/or DBP was >100 mm Hg after maximum tolerated titration of drugs from each of the 3 steps or a compelling indication, such as HF, arose. Open-label drugs of the same classes as study agents could be added to half-maximal dose; these were not supplied by the study. Detection of a serum potassium value <3.5 mmol/L required addition of potassium supplements. After initial monthly titration visits, protocol visits were prescribed every 3 months for the first year and every 4 months thereafter.

Outcomes

The primary outcome was the combined incidence of fatal CHD or nonfatal MI. The prespecified secondary end points included all-cause mortality, combined CHD (fatal CHD, nonfatal MI, hospitalized angina, and coronary revascularization procedures), stroke, and combined CVD. The outcomes added here (696 combined CHD, 939 combined CVD, and 662 deaths) occurred before February 15, 2000, but for which complete information was not available for inclusion in the preliminary report. Details of event ascertainment have been reported elsewhere.^1,3 In brief, study clinical outcomes were based on clinic investigator reports, combined with the collection of death certificates and hospital discharge summaries. This report also presents data on end-stage renal disease and quality-of-life outcomes. Quality of life was ascertained at baseline and every 2 years by asking the participant “In general, would you say your health is excellent, very good, good, fair, or poor?” and by requesting that the participant rate his/her health on a 0 to 100 scale.

Statistical Analysis

Data were analyzed according to participants’ treatment assignments. Comparability of baseline characteristics and outcomes were assessed by χ² tests for categorical variables and standard normal (z) tests for continuous variables. Cumulative event rates were calculated by the Kaplan-Meier procedure. Although rates are presented only through 4 years, both the log-rank test and the Cox proportional-hazards (PH) model incorporated the participant’s entire trial experience to evaluate differences between cumulative event curves and to obtain 2-sided probability values. Only the PH results are presented, because probability values were essentially identical. Hazard ratios (hereafter called relative risks, RRs) and 95% confidence intervals (CIs) were obtained from the PH model.⁶ For the primary end point, given the expected event rate, treatment crossovers, and losses to follow-up, ALLHAT had an 83% power to detect a 16% reduction between chlorthalidone and doxazosin at a 2-sided α=0.0178 (z=2.37) to account for the 3 original comparisons (equivalent to a 98.2% CI).⁷ A 95% CI may be converted to a 98.2% CI by multiplying the upper limit and dividing the lower limit by RR^(0.41/z), where z is the value of the test statistic for the RR estimate. The PH assumption was examined by using log-log plots and testing a treatment-by-time (time-dependent) interaction term; if it was violated, RR estimates from a 2×2 table were used.⁶ Subgroups were examined by testing for treatment-covariate interaction with the PH model at a value of P<0.05.

Participant Characteristics

Of the 24 316 participants, at the end of follow-up there were 481 (3.2%) lost to follow-up in the chlorthalidone group and 447 (4.9%) in the doxazosin group (Figure 1). By February 15, 2000, 9232 years of additional follow-up were included in this analysis compared with the original report.² Person-years of observation were 95% of those expected and were similar in both arms of the trial. Table 1 demonstrates the similarity of the chlorthalidone and doxazosin groups at baseline, both of which comprised an almost equal percentage of women and men, more than one third blacks, and 16% Hispanics. Mean age was 67 years, and mean BP at baseline was 146/84 mm Hg. Of note was the finding that nearly 90% of the study participants had received antihypertensive therapy before randomization.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. ALLHAT profile.

Medication Adherence and Course in Treatment

Of those who attended a study visit (4836/6059, or 80%) at 4 years of follow-up, 78% of the chlorthalidone group and 71% of the doxazosin group had remained on their assigned treatment. Most of that difference occurred by the end of the first year. A detailed description of drug therapy has been reported.⁸ Exclusive use of assigned medication observed at the 4-year follow-up visit was 34% for chlorthalidone and 23% for doxazosin, whereas dual medication use was 35% for chlorthalidone and 37% for doxazosin (≈17% to 18% of the chlorthalidone group and 18% of the doxazosin group were on atenolol). Conversely, 23% of the doxazosin participants were prescribed a diuretic, and 4% of chlorthalidone participants were prescribed an α-blocker. About 19% of those in the chlorthalidone group were prescribed 3 or more medications to lower their BP, compared with 27% of participants in the doxazosin group. In both groups, unspecified refusals were the most common reason (more than one third) for discontinuing the blinded study medication.

Intermediate Outcomes

Mean BP at randomization was 146/84 mm Hg for both groups, and after the first year, it was 137/79 mm Hg for the chlorthalidone group and 140/80 mm Hg for the doxazosin group (Table 2). Starting at 24 months, the systolic difference narrowed to 2 mm Hg, with concomitant narrowing of differences in the percentage at goal. Nevertheless, more chlorthalidone than doxazosin subjects were at goal pressure at their 4-year follow-up visit (63% vs 58%).

Serum cholesterol, initially 216 mg/dL for the chlorthalidone group and 215 mg/dL for the doxazosin group, fell to 197 and 187 mg/dL, respectively, at 4 years (P<0.001). About 25% of each group had participated in the lipid-lowering trial. Serum potassium, initially 4.3 mmol/L in the chlorthalidone group and 4.4 mmol/L in the doxazosin group, was 4.1 and 4.4 mmol/L at 4 years, respectively (P<0.001). Fasting serum glucose, initially 123 mg/dL in the chlorthalidone group and 122 mg/dL in the doxazosin group, was 125 and 117 mg/dL, respectively, at the fourth-year visit (P<0.001).

Baseline serum creatinine was 1.0 mg/dL in the 2 groups. After 4 years of follow-up, creatinine level rose to 1.2 mg/dL in the chlorthalidone group and to 1.1 mg/dL in the doxazosin group (P<0.001). Serum creatinine doubled during treatment in 0.8% of the chlorthalidone and in 0.5% of the doxazosin participants (P=0.02).

Responses to the question about general health status were similar between the chlorthalidone and doxazosin groups at baseline and at years 2 and 4. About 75% reported excellent, very good, or good health status both at baseline and subsequently. Responses to the question about health status as a continuous value (with 100 being “perfect health”) at baseline were 73.9 in both groups, and at year 4, 73.5 and 73.1 for the chlorthalidone and doxazosin groups, respectively.

Primary and Secondary End Points

Table 3 and Figure 2 include 431 and 265 additional combined CHD events, 584 and 355 additional combined CVD events, and 407 and 255 more deaths from all causes than previously reported for the chlorthalidone and doxazosin groups, respectively. There were also 126 and 93 additional HF cases in the chlorthalidone and doxazosin groups, respectively. Altogether, 98.4% (4020/4085) and 98.8% (2670/2703) of the events were confirmed by appropriate documentation (hospital record and/or death certificate).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Kaplan-Meier estimates for CHD (primary outcome), all-cause mortality, combined CHD, combined CVD, congestive HF, and stroke.

There was no significant difference between the chlorthalidone and doxazosin groups for the primary CHD end point of combined nonfatal MI and CHD death (Figure 2a), all-cause mortality (Figure 2b), or combined CHD (Figure 2c). Highly significant (P<0.001) increases in combined CVD (Figure 2d), HF (Figure 2e), and stroke events (Figure 2f) were experienced by participants randomized to doxazosin compared with the chlorthalidone group. Specifically, participants randomized to doxazosin experienced a 26% higher risk of stroke, an 80% higher risk of HF, a 12% higher risk of coronary revascularization, and a 13% higher risk of hospitalized or treated angina. Thus, the combined CVD risk was 20% higher (4-year rates of 28.6 vs 25.1/100) among participants randomized to doxazosin compared with chlorthalidone. About half of the 20% was accounted for by the increased risk of HF. With a more restrictive outcome measure for HF (only hospitalized or fatal cases; Table 3), the RR for doxazosin compared with chlorthalidone was 1.66 (95% CI, 1.46 to 1.89).

To elucidate the nature of the PH assumption for the HF outcomes analyses, we used a 2-time-interval (baseline to year 1 and year 1 to end of study) analysis wherein RRs were constant within each interval.⁶ For all HF, participants assigned to doxazosin compared with those assigned to chlorthalidone had a 259% increased risk during year 1, which diminished to a 36% increased risk thereafter; for hospitalized/fatal HF, there was a 222% increased risk during year 1 and a 30% increased risk thereafter. These increased risks were statistically significant in both intervals.

The HF findings were generally reflected in all predefined subgroups, including black and nonblack participants, men and women, diabetic and nondiabetic persons, those older and younger than 65, and in post hoc defined groups, including antihypertensive drug treatment status at entry and prior CHD. (Figure 3 displays similarity across subgroups for combined CVD and HF). In subjects who had or had not been regularly taking antihypertensive medications before enrollment in ALLHAT, the respective RRs of combined CVD for doxazosin compared with chlorthalidone were each significant at 1.18 and 1.39, respectively. The test for interaction was not significant (P=0.13). For HF, RRs were 1.87 and 1.54 for those treated and untreated, respectively. Again, there was no evidence of an interaction (P=0.41). Although the 95% CI for the untreated subjects was 0.98 to 2.42, the intervals for the treated and untreated groups were broadly overlapping.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. RRs and 95% CIs for doxazosin/chlorthalidone comparisons in subgroups.

Noncardiovascular end points occurred at similar rates in the 2 groups. Cancer incidence did not differ significantly between chlorthalidone and doxazosin subjects (7.37 vs 6.53 per 100; P=0.11). End-stage renal failure, as defined by initiation of dialysis, transplantation, or renal disease death, occurred in 1.1% of both groups.

Perspective

Hypertension is a major contributor to cardiovascular morbidity and morality (CVD). Although it has long been known that pharmacologic intervention can significantly reduce the incidence of CVD, the availability of newer antihypertensive agents has raised the possibility that they might offer greater cardioprotection than traditional diuretic-based therapy. To test that hypothesis, ALLHAT was designed to compare representatives of 3 newer classes of antihypertensive agents with chlorthalidone, the standard used in earlier placebo-controlled trials. That portion of the study comparing doxazosin, an α-blocking agent, with chlorthalidone, a thiazide-type diuretic, was terminated ahead of schedule after finding a substantial excess of cardiovascular events in the doxazosin group and determining that continuation of the study was unlikely to produce evidence of significant benefit for doxazosin. In this final and more complete report, we confirm the initial results showing an excess of 80% for HF and of 26% for stroke in the doxazosin group. These final results now provide further evidence that neither the 2 to 3 mm Hg higher SBP nor the unmasking of latent HF at entry to the study for participants previously treated with a diuretic bore the major responsibility for the additional HF burden of participants randomized to doxazosin. It is likely that these results apply to other α-blockers. In view of superior BP control, greater tolerability, and a significant reduction in cardiovascular events, we believe that these results provide further direct evidence that antihypertensive therapy should routinely be initiated with a thiazide-type diuretic.