The ALLHAT Lipid Lowering Trial—Less Is Less

The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial Lipid Lowering Trial (ALLHAT-LLT),¹ published in this issue of THE JOURNAL, is the second-largest lipid-lowering trial yet reported. This randomized trial, a subset of the larger ALLHAT hypertension study,² followed up 10 355 patients (49% women) who had hypertension and were 55 years or older for a mean of 4.8 years. Patients were randomly assigned to receive 40 mg of pravastatin (open label) vs "usual care" in 513 clinical centers, the majority of which were community-based. These centers were a subset of the 623 centers that treated patients for the larger (N = 33 357) ALLHAT hypertension study.² Despite a 28% reduction in low-density lipoprotein cholesterol (LDL-C) from baseline in the pravastatin treatment group, the trial failed to show a benefit in either the primary outcome (all-cause mortality) or the key secondary outcome (combined nonfatal myocardial infarction and fatal coronary heart disease).¹ How can this be? Why didn't ALLHAT-LLT work?

Several practical issues and fundamental factors influenced this trial. First, the groups were evenly matched in virtually all respects, except that the usual care group had a statistically significant but clinically insignificant larger number of patients with preexisting coronary heart disease. More important, by year 6 of the study, only 70.3% of patients in the treatment group were still taking the protocol-specified 40 mg of pravastatin, whereas 28.5% of the usual care group was receiving a lipid-lowering drug (26.1% received a statin). Lack of adherence and crossover rates increased during the trial, the latter possibly related to publication of the 4S study (Scandinavian Simvastatin Survival Study), the first large statin prevention trial, in 1994, soon after ALLHAT started.³ Although the reduction in total cholesterol in the ALLHAT-LLT treatment group was comparable to that seen in other statin trials,^{4 - 7} the reported decrease in LDL-C was not based on the total sample, only on a 10% random subset of participants in the pravastatin group and 5% of participants in the usual care group. Moreover, because of reduced adherence and the large degree of crossover, the difference in total cholesterol and the apparent difference in LDL-C between the 2 groups were modest and substantially less than the differences reported in other statin trials (as clearly shown for total cholesterol in Figure 5 of the ALLHAT-LLT article).¹ The difference in percentage changes in total cholesterol between the pravastatin and usual care groups (9.6% for ALLHAT-LLT) is markedly lower than the mean difference for all trials, even when ALLHAT-LLT is included (18.5% for all trials). Six of the 8 other trials had at least twice the total cholesterol differential between groups compared with that seen in ALLHAT-LLT.

Second, certain design features also must be considered to understand the study results. The initial plan had been to recruit 20 000 patients.⁸ However, for a variety of practical reasons, presumably partly related to the difficulty of finding patients who were not already being treated with statins and whose physicians were willing to have them randomized to usual care, ALLHAT-LLT enrolled half this amount. As more aggressive lipid lowering became the standard of care after publication of the results of statin trials of the 1990s, the remaining group available for recruitment into a "usual care" controlled trial was smaller and had lower cardiovascular risk. The authors indicate that the population recruited would have given them 84% power to detect a 20% mortality reduction. The original sample size was estimated to have 80% power (although the earlier methods article⁸ notes 85.5% power) to detect a 12.5% reduction in mortality. Thus, the sample size provided adequate power for detecting a 20% decrease in mortality, which the authors suggest is comparable to 4S.³ However, 4S was quite different: patients were at much higher risk, they had a much higher adherence rate at the end of the trial, there was a much lower crossover rate, and the treatment group experienced much greater lipid lowering and greater difference in lipid reduction between the groups.³ Thus, the revised ALLHAT sample size required a larger reduction in mortality than originally planned, despite a diminished lipid difference between the groups. In retrospect, this plan seems overly optimistic.

Third, the open-label nature of the study created an added problem. Although this design feature apparently was necessary for practical reasons, because ALLHAT-LLT was embedded in the larger hypertension study, the open-label nature of the study likely both reduced adherence and increased crossover rates. The largely community-based investigators may have been more motivated to treat usual care patients based on their perception of the individual patient's risks and need for treatment. Because crossover occurred in nearly 30% of these patients by 6 years, it seems likely that a large proportion of higher-risk individuals in usual care were being treated with statins.

Fourth, it is not clear precisely how the patients were chosen for LDL-C testing. Did all of those randomly selected actually have testing, or could those returning for testing have been more motivated to be tested because of their concurrent adherence to therapy? The number of patients undergoing LDL-C testing (5% of the usual care group and 10% of the treatment group) was small, unlike other trials in which nearly all patients had such testing. If adherent patients were more likely to return for LDL-C testing, the differential could be even smaller than the 16.7% reported, further explaining the similar outcomes observed in the treatment and usual care groups. Furthermore, diet, weight, and physical activity data were not obtained. It is possible that the usual care group could have been more motivated to undertake lifestyle changes because they or their physicians knew that they were not receiving a statin.

Physicians might be tempted to conclude that this large study demonstrates that statins do not work; however, it is well known that they do.^{9 - 10} Rather, it appears that statins may be less effective in the primary care setting in which they were used in ALLHAT-LLT. Statins have been proven efficacious in a wide array of primary and secondary prevention randomized, blinded, controlled trials. ALLHAT-LLT shows that the effectiveness may be limited in a setting that more closely mirrors clinical practice. Further evidence of possible reduced effectiveness in clinical practice was shown in 2 recent studies that documented very poor persistence of use of statins as soon as 6 months after the initiation of therapy.^{11 - 12} These studies, combined with ALLHAT-LLT, point to the need for improved adherence and understanding reasons for nonadherence.

ALLHAT-LLT also suggests that LDL-C lowering remains central to the benefit produced by statin therapy for the prevention of coronary heart disease events. Much has been made of the "pleiotropic" effects of statins, suggesting that such effects might be non–dose related. Although other data are inconsistent with this point,¹³ ALLHAT-LLT indirectly suggests that robust LDL-C lowering is necessary to achieve robust risk lowering. Statins of greater potency, at higher doses, or, most important, with greater adherence, appear to lead to greater efficacy in controlled trials. In reciprocal fashion, this study confirms those observations. Thus, ALLHAT-LLT should not be viewed as a negative trial for statin use, or for pravastatin specifically.

Three large trials separately^{4 - 6} and together¹⁴ confirm the ability of pravastatin to prevent coronary heart disease events. In addition, findings from the large PROSPER trial¹⁵ demonstrate the benefits of pravastatin, 40 mg, in higher-risk patients (N = 5804) aged 70 to 82 years. Those investigators reported a 19% relative risk reduction (12.2% event rate in placebo vs 10.1% in treatment group) for the end point of cardiac death or nonfatal myocardial infarction. Relevant to ALLHAT-LLT, the event rate reduction in PROSPER was less that that seen in many other statin trials, possibly the result of a smaller LDL difference between groups, although larger than that shown in ALLHAT. The important analysis presented in Figure 5 in the ALLHAT-LLT report¹ clearly shows that this trial is not an outlier but is consistent with previous trial findings that less cholesterol lowering produces less benefit in clinical outcomes. Without a robust difference in total cholesterol and LDL-C between the treatment and control groups, particularly in a relatively low-risk cohort, outcomes will be similar. Although this may come as no surprise, it does reaffim the importance of cholesterol lowering in the benefits of statins.

If less is less, is it necessarily true that more is more? Thus far, not for certain. However, a series of studies over the last several years,^{13 ,16 - 19} and the recently published ARBITER (Arterial Biology for the Investigation of the Treatment Effects of Reducing Cholesterol) trial,²⁰ support the conclusion that more aggressive lipid therapy leads to greater benefit. Confirmation likely will come over the next few years, when several large randomized controlled outcome trials testing the "lower is better" hypothesis reach completion.

What then are the lessons of ALLHAT-LLT for the practicing clinician? Effective prevention must rest on a firm foundation of evidence. The evidence that statins work is substantial, as firmly concluded in the latest cholesterol guidelines from the National Cholesterol Education Program.²¹ Inevitably, effective prevention also rests on application in clinical practice. Here the evidence is more equivocal. Adherence to treatment in routine practice settings is essential to realize the translation of efficacious therapy into effective practice. Until routine practice becomes closer to the conditions achieved in randomized clinical trials, there will continue to be a gap between optimal care based on the best knowledge and actual care in clinical practice.²² Perhaps one of the most important factors contributing to this gap is "clinical inertia" as described by Phillips et al,²³ and the important point that understanding is not sufficient to produce action.²² Patients, caregivers, and the health care system need to work together to improve effectiveness of treatments. This is not a small challenge. With continued progress on this front, improved focus on implementation of guidelines, implementation of care plans, and implementation of behavior change must remain central to the delivery of better health care.