Portal de Periódicos da CAPES

Size isn't everything - ALLHAT in perspective

2003; Lippincott Williams & Wilkins; Volume: 21; Issue: 3 Linguagem: Inglês

10.1097/00004872-200303000-00003

1473-5598

Gordon T. McInnes,

Pharmaceutical Practices and Patient Outcomes

The clinical trial evidence which established the cardiovascular protective effects of drug treatment for hypertension comes mainly from studies with therapy based on thiazide diuretics [1]. However, prevention of cardiac events were less than expected [2], perhaps as a consequence of the adverse metabolic effects of diuretics [3–7]. Newer antihypertensive drugs should avoid these metabolic complications and might provide advantages beyond blood pressure reduction. Comparative outcome trials with ACE inhibitors [8,9] and calcium channel blockers [10,11] have failed to demonstrate convincing advantages over diuretics. Individual trials were underpowered and hence uninformative. Even meta-analysis of the accumulated data could not exclude a clinically relevant difference in cause-specific outcomes including cardiac events [12]. A bigger trial was needed and ALLHAT offered the welcome prospect of a definitive comparison between diuretic-based therapy and treatment based on newer drugs [13]. ALLHAT (Antihypertensive and Lipid Lowering to prevent Heart Attack Trial), the largest ever randomized trial of antihypertensives, was designed specifically to have sufficient power to examine cardiac outcome (fatal and non-fatal myocardial infarction) as the primary end point. The trial randomized 42 418 patients aged 55 years or older with mild to moderate hypertension and at least one additional cardiovascular risk factor to double-blind therapy based on chlorthalidone 12.5–25 mg daily (n = 15 255) or one of three other antihypertensive treatments: amlodipine 2.5–10 mg daily (n = 9048), lisinopril 10–40 mg daily (n = 9054) or doxazosin 1–8 mg daily (n = 9061). The alpha-blocker arm was stopped prematurely, leaving 33 357 participants who were followed up for a mean of 4.9 years on diuretic, calcium channel blocker or ACE inhibitor. The primary outcome occurred in 2956 participants with no difference between the rates with the reference drug chlorthalidone (11.5%) and amlodipine (11.5%) or lisinopril (11.4%). Relative risks (RR) were 0.98 [95% confidence interval (CI) 0.90–1.07] for amlodipine and 0.99 (95% CI 0.91–1.08) for lisinopril. The investigators prespecified four secondary end points – all cause mortality, fatal and non-fatal stroke, combined coronary heart disease (CHD) (primary outcome plus coronary revascularization or angina with hospitalization) and combined cardiovascular disease (CVD) (combined CHD plus stroke, treated angina without hospitalization, heart failure and peripheral arterial disease). All cause mortality and combined CHD did not differ between the groups. However, lisinopril was associated with higher rates of combined CVD (RR 1.10, 95% CI 1.05–1.16) and stroke (RR 1.15, 95% CI 1.02–1.30). Heart failure was reported significantly more often in patients randomized to amlodipine (RR 1.38, 95% CI 1.25–1.52) and lisinopril (RR 1.19, 95% CI 1.07–1.31). The authors concluded that thiazide-type diuretics are superior to newer drugs in preventing CVD and are less expensive, and that these drugs should be the preferred first-line antihypertensive therapy. The findings from such a large trial are likely to be hugely influential but require careful critical scrutiny before changes in clinical practice are adopted. The ALLHAT results were achieved only at enormous cost. Was the price worthwhile? Size is not everything. A large sample size is only relevant if participants continue on randomized therapy. The study report makes it very difficult to assess this critical information with certainty. By year 5, it appears that 28–39% of participants had discontinued randomized study drugs. At least 5–10% provided no drug data at year 5 and such individuals have a high probability of protocol deviation suggesting that the true continuation with randomized therapy was 50–60%. In addition, 13–17% of randomized patients were taking a comparator drug and a further 7–9% were taking a comparator drug without randomized therapy or a drug from the same class. Thus 20–25% and probably more (if those with no drug data are included) of patients randomized to a particular drug may have been taking a comparator drug. The consequence of such partial and complete crossover of therapy is a gross underestimation of the true differences between treatments. If at the end of treatment only 50% of patients randomized to drug A (or B) are taking drug A (or B) without drug B (or A) and 25% of the drug randomized to drug B (or A) are also taking drug A (or B), it would be hardly surprising if no difference between treatments A and B were observed. Conservatively, the effect would be that the observed difference would underestimate the true differences between therapies by 50%. Critically, the true 95% CI would also increase, making it likely that an advantage of 10–15% (clinically relevant) for the primary outcome in favour of each of the newer drugs cannot be excluded. The result is uncertainty. A fundamental requirement for comparisons of outcomes with antihypertensive agents is that blood pressure in each group should be the same or similar. This was not the case in ALLHAT. Five-year systolic blood pressure were significantly higher in the amlodipine (0.8 mmHg, P = 0.03) and lisinopril (2 mmHg, P < 0.001) groups compared with chlorthalidone, and five year diastolic blood pressure was significantly lower with amlodipine (0.8 mmHg, P < 0.001). Differences in systolic blood pressure early in the trial was even greater. Although the authors are dismissive of the influence of these blood pressure differences, metaregression analysis of data from outcome trials suggests that minor differences can have major influences on outcome [14]. This may be particularly important in high-risk populations [15, 16] such as studied in ALLHAT. The difference in systolic blood pressure control between the lisinopril and chlorthalidone groups could certainly readily explain the observed excess risk of stroke in patients randomized to ACE inhibitor. ALLHAT reported an excess of heart failure in the amlodipine and lisinopril groups compared with chlorthalidone. Together with the excess of stroke, heart failure were entirely responsible for the higher rate of combined CVD in the lisinopril group. The diagnosis of heart failure was determined by the investigator with only a tiny minority of events undergoing independent adjudication. It is noteworthy that the relative risk of hospitalized heart failure was much less extreme, 1.35 for amlodipine and 1.10 for lisinopril, and heart failure mortality was similar in the treatment groups: 1.1% for chlorthalidone, 1.3% for amlodipine (P = 0.36) and 1.1% for lisinopril (P = 0.92). Thus, differences between the groups for heart failure and combined CVD depend on data which must be considered unreliable. Concerns about the conduct of ALLHAT were first raised with publication of the results of the prematurely discontinued doxazosin arm of the trial [17]. The main reason for discontinuation was a higher incidence of combined CVD (a secondary end point) driven mainly by a highly significant excess of heart failure on doxazosin (RR 2.04, 95% CI 1.79–2.32), although mortality was not increased in the doxazosin group (RR 1.03, 95% CI 0.90–1.15). Whether this finding truly reflected doxazosin-induced cardiac dysfunction remains uncertain since events were largely unadjudicated, and other causes of fluid retention (eg. discontinuation of diuretics or a side effect of alpha-blockade) cannot be excluded with confidence. The other reason given for premature termination of the doxazosin arm was futility. With essentially equal rates in the two treatment arms for the primary CHD outcome (RR 1.03, 95% CI 0.90–1.17), a beneficial effort of doxazosin at the scheduled trial termination was highly unlikely, based on conditional power calculations. As in the other arms of the trial, there was a high rate of discontinuation from trial therapy, crossover of therapies and inequality of systolic blood pressure control in favour of chlorthalidone. The decision to prematurely discontinue the doxazosin arm is difficult to justify and has led to uncertainty. The best and probably the only opportunity to assess the potential value of the alpha-blockade in prevention of CHD was lost. To allow wide application of the results of a clinical trial, the population studied should be heterogeneous and reflect the range of patients likely to be treated. However, strict commitment to this philosophy can lead to results which are difficult to interpret, particularly if usual clinical flexibility in prescribing is denied. In ALLHAT, this stacked the odds against the ACE inhibitor arm. The significantly less good blood pressure control in lisinopril-treated patients than in those given chlorthalidone was largely because one-third of the study population comprised black Americans, in whom ACE inhibitors are generally less effective [18]. The pre-specified second line therapy was a beta-blocker, usually considered a less than ideal accompaniment for an ACE inhibitor [18]. An alternative conclusion from the results of the lisinopril versus chlorthalidone comparison is that ACE inhibitor-based therapy provides benefits equivalent to that of a thiazide-based regimen even if handicapped by use in an inappropriate population and with inappropriate concomitant therapy, and despite less good blood pressure control. ALLHAT was a prodigious undertaking not only in size but also in the clinically relevant questions addressed. Was the effort worthwhile? More than 10 years ago in this Journal, we concluded that ‘low-dose thiazides are among the most effective and well-tolerated treatments of hypertension, and they are also the simplest, safest and least expensive. There is every reason for their continued use as first-line treatment of hypertension’ [19]. The findings of ALLHAT add little to this view. Clinical trials should be informative and sadly ALLHAT joins the category of uninformative trials. There are too many imponderables to allow the conclusion that thiazide-type diuretics are superior to newer drugs in preventing cardiovascular events. The findings are insufficiently reliable. ALLHAT simply addressed too many questions and failed to deliver on any. This should be a warning and reminder that clinical trials should be simple and the outcomes robust. No trial (even one as grand as ALLHAT) can provide all the answers. ALLHAT has merely raised more questions.