Doxazosin for the management of hypertension: implications of the findings of the ALLHAT trial

Views regarding the choice of first-line antihypertensive medication have been polarized for many years.1,2 The conservative view—unchanged since the introduction of “stepped care” in 1977—has been to support the unconditional use of diuretics or β-blockers on the grounds that, until recently, these were the only drugs evaluated in long-term morbidity and mortality trials.1 An alternative approach—unsupported by long-term morbidity or mortality trial data—has advocated the “tailoring” of all classes of antihypertensive medication to suit the demography, comorbid disease status, and risk profile of individual patients.2 The rationale for this latter approach was largely based on the heterogeneous nature of hypertensive patients and the differential effects of the drug classes on various surrogate end points such as regression of left ventricular hypertrophy and proteinuria, or the presence of comorbid conditions such as diabetes mellitus.3 Despite the introduction of the “newer” antihypertensive agents almost 30 years ago, the first trial to compare three of these drug classes (α-blocker, angiotensin converting enzyme (ACE) inhibitor, calcium channel blocker) with “conventional” diuretic therapy, the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT),4 only began recruitment in 1994 and is due to report in 2002. In early 2000, the α-blocker limb of ALLHAT was curtailed prematurely by the investigators due to concern about a doubling of additional treatment or hospitalization for heart failure in the patients randomized to receive the α-blocker doxazosin when compared with those randomized to receive the diuretic chlorthalidone.5 This decision to discontinue the doxazosin limb in ALLHAT has generated variable degrees of concern among physicians and their patients with regard to the continued use of α-blockers (particularly doxazosin) for the treatment of hypertension. Thus, it is important to consider the basis for the decision taken by the ALLHAT investigators and the outstanding questions that must be answered before a pragmatic conclusion can be reached about the role of α-blockers in the treatment of hypertension. During follow up in ALLHAT, systolic blood pressure (BP) levels were about 3 mm Hg higher among those patients randomized to doxazosin compared with those randomized to chlorthalidone. Surprisingly, there was no difference in the primary end point despite seemingly inferior BP control in the patients randomized to doxazosin. However, the combined cardiovascular disease end point—including heart failure, stroke, angina, and coronary revascularizations—was significantly higher in those randomized to doxazosin compared with those randomized to chlorthalidone. This difference was largely due to the increased rate of hospitalization for, or treatment of, heart failure. Of interest, despite this reported increase in the incidence of heart failure in the doxazosin group, which occurred very early after randomization, there was no significant increase in mortality or in the rate of the primary end point of the study (fatal coronary heart disease and nonfatal myocardial infarction) in those randomized to receive doxazosin. The publication of the effects of doxazosin in ALLHAT5 was followed by a range of responses, including one editorial that raised the question as to “whether doxazosin should continue to be used as add-on antihypertensive therapy.”6 Unfortunately, the ALLHAT data, as presented and published, cannot help to evaluate the role of α-blockers as second- and third-line therapy and, meanwhile, the ALLHAT investigators have not suggested that those who are taking open-label doxazosin as second, third, or coexistent medication in ALLHAT should stop doing so. The report comparing doxazosin with chlorthalidone in ALLHAT discusses possible explanations for the findings.5 One critical issue was the higher systolic BP among those who had been randomized to doxazosin compared with those randomized to chlorthalidone. On the basis of the other trial or epidemiologic data, the authors suggest that although these BP differences may explain the differences in stroke rate and some other events, they were insufficient to explain the higher rate of treatment or hospitalization for heart failure in the cohort randomized to doxazosin treatment. It is important to note that the 3 mm Hg difference in mean BP was an “average” effect and may understate larger BP differences in some patients. Moreover, the impact on cardiovascular events of small BP differences in high-risk patients may be larger than expected, as suggested by the benefits observed in the diabetic subgroup of the Hypertension Optimal Treatment (HOT) trial.7 Similarly, in the United Kingdom Prospective Diabetes Study (UKPDS),8 a treatment-induced 10 mm Hg reduction in systolic BP, albeit potentially confounded by differential drug usage, was associated with a 56% reduction in heart failure. Extrapolating these data to similarly high-risk patients in ALLHAT suggests that the 3 mm Hg difference in systolic BP between the doxazosin- and chlorthalidone-treated patients could potentially explain most of the difference in heart failure between these two patient groups. The BP differences between groups notwithstanding, it is perhaps not surprising that a drug (chlorthalidone) that causes a natriuresis and diuresis, and that is known to treat or to potentially mask heart failure symptoms,9 would result in fewer treatments or hospitalizations for heart failure when compared with a drug such as doxazosin, which has the potential to cause sodium retention and peripheral edema. What is perhaps surprising, although possibly a function of type II error, is that death rates from heart failure were not different in the two groups, and there was no increase in overall mortality among those randomized to doxazosin despite an 8% incidence of heart failure at 4 years. This is remarkable, given the usually dramatic effect of heart failure on mortality. Another unusual feature of the “heart failure” end point in ALLHAT was the remarkably high rates reported—much higher than those observed in other trials of similarly high-risk patients.10 Furthermore, the curves for heart failure incidence in the doxazosin and chlorthalidone groups separated within weeks after randomization; thereafter there was little further separation. Hence the effects apparently occurred early but, unusually, were not cumulative or progressive. This finding would be compatible with the withdrawal of prior therapy (eg, diuretics or ACE inhibitors) that had controlled occult heart failure before randomization. Treatment with doxazosin thereafter may have unmasked occult heart failure, whereas diuretic treatment would have been more likely to maintain control of the heart failure signs and symptoms. Before policy decisions can be made regarding the future use of α-blockers for managing hypertension, further information is clearly required, including: What antihypertensive therapy was in use by those in the doxazosin and chlorthalidone groups before entry into ALLHAT, and particularly among those who developed heart failure? What was the demography of those who did and did not develop heart failure, by randomized group? What were the compliance rates and drug doses used among those who did and did not develop heart failure, by randomized group? Was there a dose-response relationship between use of doxazosin and risk of heart failure? Was there a history of coronary heart disease or heart failure in those who did and did not develop heart failure, by randomized group? What other treatments were being taken during the trial by those who did and did not develop heart failure, by randomized group? What were the pretrial and on-treatment BPs of those who did and did not develop heart failure, by randomized group? What were the heart failure rates among those using α-blockers as additional or concomitant therapy in ALLHAT, compared with those using other agents as additional or concomitant therapy? What were the heart failure rates among those with equally well controlled BP levels, by randomized group? Clearly any “on-treatment” analyses are subject to the criticism that the advantages of randomization are fore gone. In addition such post hoc, potentially underpowered subgroup analyses are prone to generating spurious associations. However, in light of the aforementioned concerns about some unusual features of “heart failure” associated with doxazosin in this study, a better understanding of the mechanisms underpinning the ALLHAT findings is important. Such an insight will only come after more complete analyses of the data, as suggested above. Only then will it be possible to define the role of α-blockers for the management of hypertension.