Major Outcomes in High-Risk Hypertensive Patients Randomized to Angiotensin-Converting Enzyme Inhibitor or Calcium Channel Blocker vs Diuretic:  The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) FREE

Treatment and complications among the 50 to 60 million people in the United States with hypertension are estimated to cost $37 billion annually, with antihypertensive drug costs alone accounting for an estimated $15.5 billion per year.¹ Antihypertensive drug therapy substantially reduces the risk of hypertension-related morbidity and mortality.^2- 6 However, the optimal choice for initial pharmacotherapy of hypertension is uncertain.⁷

Earlier clinical trials documented the benefit of lowering blood pressure (BP) using primarily thiazide diuretics or β-blockers.^2- 3,8 After these studies, several newer classes of antihypertensive agents (ie, angiotensin-converting enzyme [ACE] inhibitors, calcium channel blockers [CCBs], α-adrenergic blockers, and more recently angiotensin-receptor blockers) became available. Over the past decade, major placebo-controlled trials have documented that ACE inhibitors and CCBs reduce cardiovascular events in individuals with hypertension.^9- 11 However, their relative value compared with older, less expensive agents remains unclear. There has been considerable uncertainty regarding effects of some classes of antihypertensive drugs on risk of coronary heart disease (CHD).^6,12- 16 The relative benefit of various agents in high-risk hypertensive subgroups such as older patients, black patients, and patients with diabetes also needed to be established.¹⁷

The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), a randomized, double-blind, multicenter clinical trial sponsored by the National Heart, Lung, and Blood Institute, was designed to determine whether the occurrence of fatal CHD or nonfatal myocardial infarction is lower for high-risk patients with hypertension treated with a CCB (represented by amlodipine), an ACE inhibitor (represented by lisinopril), or an α-blocker (represented by doxazosin), each compared with diuretic treatment (represented by chlorthalidone).¹⁸ Chlorthalidone was found to be superior to doxazosin and was previously reported after early termination of the doxazosin arm of the trial.^19- 20 Secondary outcomes included all-cause mortality, stroke, and other cardiovascular disease (CVD) events. A lipid-lowering subtrial was designed to determine whether lowering cholesterol with 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitor (pravastatin) compared with usual care reduced all-cause mortality in a moderately hypercholesterolemic subset of ALLHAT participants.^18,21 To evaluate differences in CVD effects of the various first-step drugs, ALLHAT was designed with a large sample size (9000-15 000 participants/intervention arm) and long follow-up (4-8 years). This study presents results of the amlodipine and lisinopril vs chlorthalidone comparisons on major CVD outcomes.

The rationale and design of ALLHAT have been presented elsewhere.¹⁸ Participants were men and women aged 55 years or older who had stage 1 or stage 2 hypertension with at least 1 additional risk factor for CHD events.^18,22 The risk factors included previous (>6 months) myocardial infarction or stroke, left ventricular hypertrophy demonstrated by electrocardiography or echocardiography, history of type 2 diabetes, current cigarette smoking, high-density lipoprotein cholesterol of less than 35 mg/dL (<0.91 mmol/L), or documentation of other atherosclerotic CVD. Individuals with a history of hospitalized or treated symptomatic heart failure (HF) and/or known left ventricular ejection fraction of less than 35% were excluded.

Unless the drug regimen had to be tapered for safety reasons, individuals continued any prior antihypertensive medications until they received randomized study drug, at which point they stopped taking all previous medications. Treatment with the study drug was initiated the day after randomization. By telephone, participants were randomly assigned to chlorthalidone, amlodipine, or lisinopril in a ratio of 1.7:1:1. The concealed randomization scheme was generated by computer, implemented at the clinical trials center, stratified by center and blocked in random block sizes of 5 or 9 to maintain balance. Participants (n = 33 357) were recruited at 623 centers in the United States, Canada, Puerto Rico, and the US Virgin Islands between February 1994 and January 1998. (The original reported number of 625 sites changed because 2 sites and their patients with poor documentation of informed consent were excluded.²⁰) All participants gave written informed consent, and all centers obtained institutional review board approval. Follow-up visits were at 1 month; 3, 6, 9, and 12 months; and every 4 months thereafter. The range of possible follow-up was 3 years 8 months to 8 years 1 month. The closeout phase began on October 1, 2001, and ended on March 31, 2002.

Trained observers using standardized techniques measured BPs during the trial.²⁰ Visit BP was the average of 2 seated measurements. Goal BP in each randomized group was less than 140/90 mm Hg achieved by titrating the assigned study drug (step 1) and adding open-label agents (step 2 or 3) when necessary. The choice of step 2 drugs (atenolol, clonidine, or reserpine) was at the physician's discretion. Nonpharmacologic approaches to treatment of hypertension were recommended according to national guidelines.^4,23 Step 1 drugs were encapsulated and identical in appearance so that the identity of each agent was double-masked at each dosage level. Dosages were 12.5, 12.5 (sham titration), and 25 mg/d for chlorthalidone; 2.5, 5, and 10 mg/d for amlodipine; and 10, 20, and 40 mg/d for lisinopril. Doses of study-supplied open-label step 2 drugs were 25 to 100 mg/d of atenolol; 0.05 to 0.2 mg/d of reserpine; or 0.1 to 0.3 mg twice a day of clonidine; step 3 was 25 to 100 mg twice a day of hydralazine. Other drugs, including low doses of open-label step 1 drug classes, were permitted if clinically indicated.^18,20

The primary outcome was fatal CHD or nonfatal myocardial infarction combined.¹⁸ Four major prespecified secondary outcomes were all-cause mortality, fatal and nonfatal stroke, combined CHD (the primary outcome, coronary revascularization, hospitalized angina), and combined CVD (combined CHD, stroke, other treated angina, HF [fatal, hospitalized, or treated nonhospitalized], and peripheral arterial disease). Coronary revascularization included coronary artery bypass graft, percutaneous angioplasty, insertion of stents, and atherectomy. Individual components of the combined outcomes were prespecified and examined, as were other secondary outcomes including cancer, incident electrocardiographic left ventricular hypertrophy, end-stage renal disease (ESRD) (dialysis, renal transplant, or death), and slope of the reciprocal of longitudinal serum creatinine measurements. Change in estimated glomerular filtration rate^24- 25 was examined post hoc.

Study outcomes were assessed at follow-up visits and reported to the clinical trials center.¹⁸ Hospitalized outcomes were primarily based on clinic investigator reports, and copies of death certificates and hospital discharge summaries were requested. Among all combined CVD events that resulted in deaths, hospitalizations, or both, the proportion with documentation (ie, a death certificate or a hospital discharge summary) was 99% in all 3 treatment groups. In addition, searches for outcomes were accomplished through the Center for Medicare and Medicaid Services, the Department of Veterans Affairs, the National Death Index, and the Social Security Administration databases. A death was ascertained by clinic report or by match with the aforementioned databases plus a confirmatory death certificate. A death pending confirmation is one found using databases but for which a confirmatory death certificate has not yet been obtained. Medical reviewers from the clinical trials center verified the physician-assigned diagnoses of outcomes using death certificates and hospital discharge summaries. More detailed information was collected on a random (10%) subset of CHD and stroke events to validate the procedure of using physician diagnoses.¹⁸ When a large excess of HF became evident in the doxazosin arm, a 1-time sample of HF hospitalizations was reviewed by the ALLHAT Endpoints Subcommittee. Agreement rates between the subcommittee and clinic investigators were 90% (155/172) for the primary outcome, 85% (33/39) for HF hospitalizations,²⁶ and 84% (129/153) for stroke, and were similar in all treatment groups.

Two major safety outcomes, angioedema and hospitalization for gastrointestinal bleeding, were prespecified. Occurrence of gastrointestinal bleeding was ascertained from Center for Medicare and Medicaid Services and Department of Veterans Affairs hospitalization databases, representing 74% of ALLHAT participants (persons ≥65 years, Department of Veterans Affairs participants, or both).²⁷ Angioedema was ascertained using a solicited event question on a serious adverse event form.

To maximize statistical power, 1.7 times as many participants were assigned to the diuretic group as to each of the other 3 groups.¹⁸ Given the achieved sample size and expected event rate, treatment crossovers, and losses to follow-up, ALLHAT had 83% power to detect a 16% reduction in risk of the primary outcome between chlorthalidone and each other group at a 2-sided α = .0178 (z = 2.37) to account for the 3 original comparisons.²⁸ Data were analyzed according to participants' randomized treatment assignments regardless of their subsequent medications (intent-to-treat analysis). Cumulative event rates were calculated using the Kaplan-Meier method. Although rates are presented only through 6 years, both the log-rank test and Cox proportional hazards regression model incorporated the participant's entire trial experience to evaluate differences between cumulative event curves and to obtain 2-sided P values. Only the Cox proportional hazard regression results are presented, because P values were essentially identical. Hazard ratios (relative risks [RRs]) and 95% confidence intervals (CIs) were obtained from the Cox proportional hazards regression model.²⁹ For consistency with α = .0178, 95% CIs may be converted to 98.2% limits 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 Cox proportional hazards regression model assumption was examined by using log-log plots and testing a treatment × time (time-dependent) interaction term; if it was violated, the RR estimate from a 2-by-2 table was used.²⁹ Heterogeneity of effects in prespecified subgroups, (1) men and women, (2) participants less than 65 and 65 years or older, (3) black and nonblack participants, and (4) diabetic and nondiabetic participants, and the post hoc subgroups presence or absence of CHD at baseline, was examined by testing for treatment-covariate interaction with the Cox proportional hazards regression model by using P<.05. SAS version 8.0 (SAS Institute, Cary, NC) and STATA version 7 (Stata Corp, College Station, Tex) were used for statistical analyses.

A National Heart, Lung, and Blood Institute–appointed data and safety monitoring board met at least annually to review the accumulating data and to monitor for safety and efficacy. The Lan-DeMets version of the O'Brien-Fleming group sequential boundaries was used to assess treatment group differences, and conditional power was used to assess futility.^30- 31

Table 1 presents baseline characteristics for the 33 357 participants in the chlorthalidone, amlodipine, and lisinopril treatment groups. The mean age was 67 years; 47% were women, 35% were black, 19% were Hispanic, and 36% were diabetic. There were nearly identical distributions of baseline factors in the 3 treatment groups.²²

Figure 1 shows the number of participants randomized and followed up to the time of closeout. In all 3 treatment groups, the mean (SD) length of follow-up was 4.9 years (1.4 years), and 99% of expected person-years were observed. The maximum range of follow-up was 8.0, 7.9, and 8.1 years in the chlorthalidone, amlodipine, and lisinopril groups, respectively. At trial closeout, 419 (2.7%) of the chlorthalidone group, 258 (2.8%) of the amlodipine group, and 276 (3.0%) of the lisinopril group had unknown vital status. Among participants with unknown vital status, the distributions of most baseline factors were similar among the 3 treatment groups, but participants assigned to lisinopril were less likely to be black and more likely to be women, have untreated hypertension, evidence of CHD or atherosclerotic CVD, and a lower mean serum glucose.

Visit adherence decreased over time from about 92% at 1 year to 84% to 87% at 5 years in all 3 treatment groups (Table 2). Among participants in the chlorthalidone group who were contacted in the clinic or by telephone within 12 months of annual scheduled visits, 87.1% were taking chlorthalidone or another diuretic at 1 year, decreasing to 80.5% at 5 years; 67.5% (n = 4387) were taking a diuretic without a CCB or an ACE inhibitor; and 13.2% were taking a diuretic with a CCB (5.8% [n = 399]) or an ACE inhibitor (9.3% [n = 641]). Only 9.0% were taking either a CCB (5.8% [n = 399]) or an ACE inhibitor (5.6% [n = 385]) without a diuretic at 5 years.

Among participants in the amlodipine group, 87.6% were taking amlodipine or another CCB at 1 year, decreasing to 80.4% at 5 years; and 63.8% (n = 2502) were taking a CCB alone without a diuretic. Another 16.6% were taking a CCB with a diuretic, and only 6.9% were taking a diuretic without a CCB. Among participants in the lisinopril group, 82.4% were taking lisinopril or another ACE inhibitor at 1 year, decreasing to 72.6% at 5 years; 56.9% (n = 2143) were taking an ACE inhibitor alone without a diuretic; and 15.7% were taking an ACE inhibitor with a diuretic at 5 years. About 8.5% were taking a diuretic without an ACE inhibitor.

The most common reasons for not taking step 1 medication at 5 years in the chlorthalidone, amlodipine, and lisinopril groups were unspecified refusals (41.4% [n = 775], 40.5% [n = 443], and 37.9% [n = 552], respectively) and symptomatic adverse effects (15.0% [n = 282], 16.4% [n = 180], and 18.1% [n = 264], respectively). Elevated BP (4.5% [n = 84], 3.5% [n = 38], and 9.0% [n = 131]) or other adverse effects such as abnormal laboratory values (3.8% [n = 71], 1.6% [n = 17], and 2.3% [n = 34]) were other reasons given for discontinuation of step 1 medications. Among participants with available medication information at 1 year, 26.7%, 25.9%, and 32.6% of those assigned to chlorthalidone, amlodipine, and lisinopril, respectively, were taking a step 2 or step 3 drug. At 5 years, the corresponding percentages were 40.7%, 39.5%, and 43.0%, respectively. Usage patterns of specific step 2 drugs were similar among groups. Participants could be taking more than 1 step-up drug. At 1 year, 40.0% (n = 4645), 44.0% (n = 3017), and 43.8% (n = 2764) of participants assigned to chlorthalidone, amlodipine, and lisinopril, respectively, still taking their blinded medication were receiving the maximal study dose. At 5 years, the percentages were 56.9% (n = 2629), 65.7% (n = 1856), and 60.3% (n = 1391), respectively.

Given the large sample size in ALLHAT, almost all differences in follow-up BP and biochemical measurements were statistically significant (Table 3 and Table 4a). Mean seated BP at randomization was about 146/84 mm Hg in all 3 groups, with 90% of participants reporting current antihypertensive drug treatment (Table 1). Follow-up BPs in all 3 groups are shown in Table 3 and Figure 2.

Mean total serum cholesterol levels at baseline and 4 years follow-up are shown in Table 4. At 4 years, about 35% to 36% of participants in all 3 groups reported taking lipid-lowering drugs, largely statins, some as a result of participation in the ALLHAT lipid trial. Mean serum potassium levels at baseline and follow-up are also shown; about 8% of the chlorthalidone group were receiving potassium supplementation at 5 years compared with 4% in the amlodipine group and 2% in the lisinopril group. Among individuals classified as nondiabetic at baseline, with baseline fasting serum glucose less than 126 mg/dL (7.0 mmol/L), incidence of diabetes (fasting serum glucose, ≥126 mg/dL [7.0 mmol/L]) at 4 years was 11.6%, 9.8%, and 8.1%, respectively.

Mean estimated glomerular filtration rate at baseline was about 78 mL/min per 1.73 m² in all groups. At 4 years, it was 70.0, 75.1, and 70.7 mL/min per 1.73 m² in the chlorthalidone, amlodipine, and lisinopril groups, respectively. The slopes of the reciprocal of serum creatinine over time were virtually identical in the chlorthalidone and lisinopril groups (–0.018 and –0.019 dL/mg per year), whereas the decline in the amlodipine slope (–0.012 dL/mg per year) was less than that of the chlorthalidone slope (P<.001).

Amlodipine vs Chlorthalidone. No significant difference was observed between amlodipine and chlorthalidone for the primary outcome (RR, 0.98; 95% CI, 0.90-1.07) or for the secondary outcomes of all-cause mortality, combined CHD, stroke, combined CVD, angina, coronary revascularization, peripheral arterial disease, cancer, or ESRD (Table 5, Figure 3, and Figure 4). The amlodipine group had a 38% higher risk of HF (P<.001) with a 6-year absolute risk difference of 2.5% and a 35% higher risk of hospitalized/fatal HF (P<.001). The treatment effects for all outcomes were consistent across the predefined subgroups (Figure 5) and by absence or presence of CHD at baseline. Cause-specific mortality rates (except for unintentional injuries/suicides/homicides in amlodipine compared with chlorthalidone, not a prespecified hypothesis) were similar for the 2 groups (Table 6).

Lisinopril vs Chlorthalidone. No significant difference was observed between lisinopril and chlorthalidone for the primary outcome (RR, 0.99; 95% CI, 0.91-1.08) or for the secondary outcomes of all-cause mortality, combined CHD, peripheral arterial disease, cancer, or ESRD (Table 5, Figure 3 and Figure 4). Cause-specific mortality rates were also similar in the 2 groups (Table 6). The lisinopril group had a 15% higher risk for stroke (P = .02) and a 10% higher risk of combined CVD (P<.001), with a 6-year absolute risk difference for combined CVD of 2.4%. Included in this analysis was a 19% higher risk of HF (P<.001), a 10% higher risk of hospitalized/fatal HF (P = .11), an 11% higher risk of hospitalized/treated angina (P = .01), and a 10% higher risk of coronary revascularization (P = .05). The treatment effects for all outcomes were consistent across subgroups by sex, diabetic status (Figure 6), and baseline CHD status. For combined CHD, there was a significant differential effect by age (P = .01 for interaction) with RRs (lisinopril vs chlorthalidone) of 0.94 for those less than 65 years vs 1.11 in those 65 years or older. However, when age was modeled as a continuous variable, there was no significant interaction. For stroke and combined CVD, there was a significant differential effect by race (P = .01 and P = .04 for interaction, respectively). The RRs (lisinopril vs chlorthalidone) were 1.40 (95% CI, 1.17-1.68) and 1.00 (95% CI, 0.85-1.17) for stroke and 1.19 (95% CI, 1.09-1.30) and 1.06 (95% CI, 1.00-1.13) for combined CVD in blacks and nonblacks, respectively.

The mean follow-up systolic BP for all participants was 2 mm Hg higher in the lisinopril group than the chlorthalidone group, 4 mm Hg higher in blacks, and 3 mm Hg higher in those 65 years or older. Adjustment for follow-up BP as time-dependent covariates in a Cox proportional hazards regression model slightly reduced the RRs for stroke (1.15 to 1.12) and HF (1.20 to 1.17) overall and in the black subgroup (stroke, 1.40 to 1.35; and HF, 1.32 to 1.26), but the results remained statistically significant.

Six-year rates of hospitalization for gastrointestinal bleeding, available only in Medicare and Department of Veterans Affairs participants, occurred in 8.8%, 8.0%, and 9.6% participants in the chlorthalidone, amlodipine, and lisinopril treatment groups, respectively, with no significant differences (Table 5). Angioedema occurred in 8 of 15 255 (0.1%), 3 of 9048 (<0.1%), and 38 of 9054 (0.4%) persons in the chlorthalidone, amlodipine, and lisinopril treatment groups, respectively. Significant differences were seen for the lisinopril vs chlorthalidone comparison overall (P<.001), in blacks (2 of 5369 [<0.1%] for chlorthalidone, 23 of 3210 [0.7%] for lisinopril; P<.001), and in nonblacks (6 of 9886 [0.1%] for chlorthalidone, 15 of 5844 for lisinopril [0.3%]; P = .002). The only death from angioedema was in the lisinopril group.

Neither amlodipine (representing CCBs, particularly dihydropyridine [DHP]–CCBs) nor lisinopril (representing ACE inhibitors) was superior to chlorthalidone (representing thiazide-type diuretics) in preventing major coronary events or in increasing survival. Chlorthalidone was superior to amlodipine (by about 25%) in preventing HF, overall, and for hospitalized or fatal cases, although it did not differ from amlodipine in overall CVD prevention. Chlorthalidone was superior to lisinopril in lowering BP and in preventing aggregate cardiovascular events, principally stroke, HF, angina, and coronary revascularization. ALLHAT previously reported that chlorthalidone was superior to doxazosin (representing α-blockers) in reducing BP and preventing cardiovascular events, particularly HF.^19- 20

It is not surprising that no significant differences in CHD and stroke rates were found between chlorthalidone and amlodipine-based therapy in ALLHAT. In the Systolic Hypertension in the Elderly Program and the Systolic Hypertension in Europe trial, in which a thiazide-like diuretic (chlorthalidone) or a DHP-CCB was compared with a placebo, major CHD events were reduced by 27% and 30%, and stroke by 37% and 42%, respectively.^8- 9 More direct evidence comes from 2 large active-controlled trials that compared DHP-CCB and traditional first-step drugs. The Swedish Trial in Old Patients with Hypertension-2 and the International Nifedipine GITS (long-acting gastrointestinal formulation) Study: Intervention as a Goal in Hypertension Treatment (INSIGHT), found no significant differences for major CHD or stroke rates between the treatment groups.^32- 33 Some of these individual trials have had limited power to evaluate differences between treatments.³⁴ In meta-analyses of 5 positive-controlled trials, which included both DHP-CCB and non–DHP-CCB trials, there were trends that favored CCB-based therapy for stroke and traditional treatment for CHD, with no difference for all-cause mortality.^13- 14 However, ALLHAT observed approximately the same number of strokes and nearly twice as many CHD events as all 5 trials combined, which suggests that the aggregate of the evidence would indicate no difference between CCB-based treatment and diuretic-based treatment for these outcomes.

The amlodipine vs chlorthalidone findings for HF reinforce previous trial results. In the diuretic-based Systolic Hypertension in the Elderly Program, active therapy reduced HF occurrence by 49% compared with placebo (P<.001), although in the DHP-CCB–based Systolic Hypertension in Europe trial, it was reduced by 29% (not statistically significant).^9,35 In the INSIGHT trial, HF was approximately twice as frequent in the CCB vs the diuretic arm.³³ The previously cited meta-analyses reported a higher rate of HF with CCB-based treatment than traditional regimens, with no difference in RR for DHPs compared with non–DHPs.^13- 14

A body of literature based on observational studies and secondary CHD prevention trials of short-acting CCBs has suggested that CCBs, especially DHP-CCBs, may increase the risk of cancer, gastrointestinal bleeding, and all-cause mortality.^14,36- 37 The results of ALLHAT do not support these findings. In fact, the mortality from noncardiovascular causes was significantly lower in the CCB group (Table 6).

There were no significant differences in the incidence of ESRD between chlorthalidone and amlodipine, consistent with findings from the INSIGHT trial.³³ Comparison of the reciprocal serum creatinine slopes suggested a slower decline in kidney function in the amlodipine group. However, this finding requires cautious interpretation because studies assessing glomerular filtration rate more directly have shown a hemodynamically mediated acute increase in glomerular filtration rate followed by a more rapid rate of decline with chronic therapy using amlodipine and other CCBs.^38- 40

Comparison of the lisinopril and chlorthalidone groups revealed better drug tolerance and BP control with chlorthalidone. Angioedema, a rare but potentially serious adverse effect of ACE inhibitor use, occurred 4 times more frequently in participants randomized to lisinopril than in those randomized to chlorthalidone. Cholesterol levels, the prevalence of hypokalemia (serum potassium <3.5 mEq/L), and new diabetes (fasting glucose ≥126 mg/dL [≥7.0 mmol/L]) were higher in the chlorthalidone than the other groups following 2 and 4 years of follow-up. Overall, these metabolic differences did not translate into more cardiovascular events or into higher all-cause mortality in the chlorthalidone group compared with the other 2 groups.

The ALLHAT findings for some major outcomes are consistent with predictions from placebo-controlled trials involving ACE inhibitors and diuretics. Specifically, for ACE inhibitor and diuretic trials, respectively, the reductions in CHD rates were 20% and 18%, and for all-cause mortality, 16% and 10%.¹³ The 10% greater rate of combined CVD in the lisinopril than in the chlorthalidone group was due to increased occurrences of stroke, HF, angina, and coronary revascularization. Results for some of these outcomes may seem surprising, because of reports of beneficial effects of ACE inhibitors on surrogate markers of atherosclerosis and reductions in vascular and renal events in individuals with HF, diabetes, kidney disease, and cerebrovascular disease in placebo-controlled trials.^41- 43 However, the finding in ALLHAT that HF incidence was lower in the diuretic vs the ACE inhibitor group is also consistent with previous reports. In the Systolic Hypertension in the Elderly Program trial (chlorthalidone vs placebo), there was a 49% decrease in the development of HF, whereas in the Studies of Left Ventricular Dysfunction Prevention (enalapril vs placebo) and Heart Outcomes Prevention Evaluation trials (ramipril vs placebo), there were only 20% and 23% reductions, respectively.^8,10,44 In published meta-analyses of placebo-controlled trials, the reductions in rates for stroke with ACE inhibitor and diuretics were 30% and 34%, translating into nearly equivalent results.^3,13 The 15% relative increase in stroke incidence for lisinopril compared with chlorthalidone treatment in ALLHAT must be considered in the context of heterogeneity of the results by race. The Swedish Trial in Old Patients with Hypertension-2 trial, which compared ACE inhibitors with conventional treatment (diuretics and/or β-blockers), showed no significant differences in CHD, stroke, HF, or all-cause mortality.³² Although these findings are somewhat different from the experience in ALLHAT, consideration needs to be given to respective confidence limits, population differences (especially race), and study designs (open vs double-blind).

No substantial differences in incidence of ESRD, glomerular filtration rate, or reciprocal creatinine slopes were noted for the lisinopril vs chlorthalidone comparisons. The ALLHAT study population was selected for high CVD risk and had a baseline mean creatinine of only 1.0 mg/dL (88.4 µmol/L). More detailed analyses of high renal risk subgroups (ie, diabetic, renal-impaired, and black patients) will be the subject of subsequent reports.

Analyses of RRs for stroke and HF adjusted for follow-up BP suggest that the 2-mm Hg systolic BP difference overall (4 mm Hg in black patients) between the lisinopril and chlorthalidone groups only partially accounts for the observed CVD event difference. However, such analyses are limited by the infrequency and imprecision of BP measurements for individual participants and regression dilution, which underestimates CVD risk associated with BP differences based on single-visit (or even visit-averaged) measurements.⁴⁵ Such modeling is also unable to account for differences among individuals due to other unmeasured or poorly represented risk factors; thus, participants who lower their BP by a given amount with one drug may not be comparable to those who lower their BP by the same magnitude with another drug.

Using an external standard of pooled results of long-term hypertension treatment trials and observational studies (10-12 mm Hg systolic BP difference associated with 38% stroke reduction), a 2- to 3-mm Hg difference in BP might account for a 6% to 12% difference in stroke rates.^45- 46 This is consistent with the observed 15% difference for stroke overall but not with the difference seen in black patients (13%-16% expected, 40% observed). For the HF outcome, trial results in isolated systolic hypertension suggest that a 3-mm Hg higher systolic BP could explain a 10% to 20% increase in risk.^8,47 The forgoing ignores the absence of a diastolic BP difference in ALLHAT; however, the relationship of diastolic pressure and CVD events in elderly persons who often have increased pulse pressure is not entirely clear.⁴⁸

The primary and secondary outcome results for the amlodipine vs chlorthalidone group comparisons were consistent for all subgroups of participants: older and younger, men and women, black and nonblack, diabetic and nondiabetic. For the lisinopril vs chlorthalidone comparisons, results were generally consistent by age, sex, and diabetic status. Thus, for the important diabetic population, lisinopril appeared to have no special advantage (and amlodipine no particular detrimental effect) for most CVD and renal outcomes when compared with chlorthalidone. In fact, chlorthalidone was superior to lisinopril for several CVD outcomes and superior to amlodipine for HF in both diabetic and nondiabetic participants. The consistency of the ALLHAT findings across multiple patient subgroups provides confidence in the ability to generalize the findings to most patients with hypertension.

In the lisinopril vs chlorthalidone comparisons, there were 2 outcomes with significant interactions. The greater differences observed in black vs nonblack patients for combined CVD and stroke, along with a similar trend for HF and lesser BP lowering with lisinopril, are in accord with the multiple reports of poorer BP response with ACE inhibitor in black patients.^49- 51 They are also consistent with reports of lesser effects of ACE inhibitors in secondary prevention of HF in this population,^52- 53 although these findings have been recently questioned.⁵⁴ The differential responses for disease outcomes parallel the lesser response in the black subgroup for BP, although the differences in outcomes are not substantially reduced by statistically adjusting for systolic BP.

Although subordinate to safety and efficacy, the cost of drugs and medical care for the individual and society is a factor that should be considered in the selection of antihypertensives. One of the stated objectives of ALLHAT was to answer the question, "Are newer types of antihypertensive agents, which are currently more costly, as good or better than diuretics in reducing CHD incidence and progression?"¹⁸ Consideration of drug cost could have a major impact on the nation's health care expenditures. Based on previous data that showed that diuretic use declined from 56% to 27% of antihypertensive prescriptions between 1982 and 1992, the health care system would have saved $3.1 billion in estimated cost of antihypertensive drugs had the pattern of prescriptions for treatment of hypertension remained at the 1982 level.⁵⁵ Further economic analyses based on the results of ALLHAT are under way.

The strengths of ALLHAT include its randomized double-blind design, statistical power to detect clinically meaningful differences in CVD outcomes of interest, diverse population with adequate representation from subgroups of special interest in the treatment of hypertension, and varied practice-based settings. In addition, the agents that were directly compared represent 3 of the most commonly used newer classes of antihypertensives vs the best studied of the older classes.

Some limitations are worth noting. After ALLHAT was designed, newer agents have been or may soon be released (eg, angiotensin-receptor blockers, selective aldosterone antagonists), which were not evaluated. Although clinical centers were blinded to the regimen and urged to achieve recommended BP goals, equivalent BP reduction was not fully achieved in the treatment groups. Furthermore, because diuretics, ACE inhibitors, CCBs, and α-blockers were evaluated in the trial, the agents available for step-up led to a somewhat artificial regimen (use of sympatholytics rather than diuretics and CCBs) of step-up drugs in the ACE inhibitor group. This may have contributed to the higher BPs in the ACE inhibitor group, especially in the black subgroup. However, mean follow-up BPs were well below 140/90 mm Hg in all treatment groups. Although ALLHAT did not compare a β-blocker to chlorthalidone, previous trials have suggested equivalence⁴⁵ or even inferiority³ for major CVD events.

The ALLHAT results apply directly to chlorthalidone, amlodipine, and lisinopril. Combined with evidence from other trials, we infer that the findings also broadly apply to the drug classes (or subclass in the case of the dihydropyridine CCBs) that the study drugs represent. The evidence base for selection of antihypertensive agents has been markedly strengthened by the addition of ALLHAT.

In conclusion, the results of ALLHAT indicate that thiazide-type diuretics should be considered first for pharmacologic therapy in patients with hypertension. They are unsurpassed in lowering BP, reducing clinical events, and tolerability, and they are less costly. For patients who cannot take a diuretic (which should be an unusual circumstance), first-step therapy with CCBs and ACE inhibitors could be considered with due regard for their higher risk of 1 or more major manifestations of CVD. Since a large proportion of participants required more than 1 drug to control their BP, it is reasonable to infer that a diuretic be included in all multidrug regimens, if possible. Although diuretics already play a key role in most antihypertensive treatment recommendations, the findings of ALLHAT should be carefully evaluated by those responsible for clinical guidelines and be widely applied in patient care.