Interpreting the AIM-HIGH Trial Results: Insights and Implications for Clinical Practice

On May 26, 2011, the National Heart, Lung, and Blood Institute (NHLBI) announced publicly that the Atherothrombosis Intervention in Metabolic Syndromes with Low HDL/High Triglycerides and Impact on Global Health Outcomes (AIM-HIGH) Trial was being terminated due to futility, approximately 18 months earlier than planned, because of the inability to demonstrate that combination dyslipidemic therapy with extended-release niacin (ERN) and simvastatin therapy was superior to LDL-C reduction therapy alone in improving long-term clinical outcomes among patients with established atherosclerotic cardiovascular (principally non-acute coronary, cerebrovascular, and peripheral arterial) disease who had residually low levels of HDL-C. Since the formal trial results were presented as a Late-Breaking Clinical Trial at the AHA on 1/15/11 and published online in the New England Journal of Medicine, multiple opinion leaders and experts have commented openly about many features of the trial design, statistical methodology, study findings, and implications for clinical practice. We will discuss these issues in detail, and attempt to provide an appropriate clinical context within which these data and findings can be appropriately interpreted.

Rationale and Design

AIM-HIGH was undertaken to address an important scientific question in contemporary dyslipidemia therapeutics, namely, that despite significant reductions in LDL-C and cardiovascular (CV) events with statins, residual risk for subsequent events in coronary heart disease (CHD) patients remains unacceptably high, and may be attributable to low levels of HDL-C and/or elevated triglycerides (TG). For example, in the PROVE-IT Trial of acute coronary syndrome patients who were randomized to intensive LDL-C reduction therapy with high-dose atorvastatin (80 mg daily) vs. standard LDL-C reduction therapy with moderate-dose pravastatin (40 mg daily), the cumulative 2-year composite CV event rate was still 22.6% despite a mean on-treatment LDL-C level of 62 mg/dL in the patients receiving high-dose atporvastatin. Such data suggested that one possible candidate for this appreciable residual risk might be low levels of HDL-C and/or elevated TG.

AIM-HIGH was a randomized, placebo-controlled clinical trial in patients with a history of atherosclerotic CV disease and atherogenic dyslipidemia (low HDL-C, high triglycerides). We hypothesized that raising HDL-C with extended-release niacin (ERN) would reduce the risk of CV events (i.e., the time to first event for the primary composite endpoint of CHD death, non-fatal MI, ischemic stroke, hospitalization for ACS, and symptom-driven coronary or cerebral revascularization) among patients treated with intensive simvastatin ± ezetimibe therapy 10 mg daily, as needed, to achieve an LDL-C level of 40–80 mg/dL in both arms. Following a 4-8 week open-label run-in with simvastatin 40 mg daily and rapid uptitration of progressively increasing dosages of ERN (500 mg daily during week one; 1000 mg daily during week two; 1500 mg/daily during week three; 2000 mg daily during week four), participants tolerating at least 1500 mg daily were randomly assigned in a double-blind (1:1 randomization scheme) to ERN or matching placebo. In this event-driven trial, we projected that 800 adjudicated primary events during a 2.5 to 7 year (mean: 4.6 year) follow-up would provide 85% power to detect a relative 25% treatment difference between the ERN and placebo groups. Follow-up was scheduled to conclude in December, 2012. A total of 3,414 men and women, mean age 64 years, were recruited from 92 enrolling centers across the U.S. and Canada and followed for an average of 36 months. At entry, 3,196 patients (94%) were taking a statin, with a mean baseline LDL-C of 71 mg/dL, HDL-C was 35 mg/dL, and triglyceride (TG) level was 161 mg/dL; by contrast, only 218 patients (6%) were statin-naive at trial entry. In this small cohort, mean baseline LDL-C was 125 mg/dL, HDL-C was 33 mg/dL, and TG was 215 mg/dL.

During a 36-month follow-up, compared to placebo, ERN raised mean HDL-C by 25% (to 42 mg/dL), lowered TG by 29% (to 122 mg/dL), while LDL-C further declined from 74 mg/dL to 62 mg/dL. As noted above, the trial was stopped at a formal interim analysis due to lack of efficacy of ERN. The primary endpoint occurred in 282 ERN-treated subjects (16.4%) as compared with 274 placebo-treated patients (16.2%); HR 1.02, 95%CI, 0.87-1.21); P=0.80.

Among the many questions posed over the past several months since the original NHLBI press release in late May, most have focused on why the preliminary trial results have been negative (neutral) for hypothesized clinical superiority of combination dyslipidemic therapy. Some have questioned why the study was discontinued early after a mean 36-month follow-up and not allowed to continue (for perhaps an additional 5 or more years) in order to potentially discern a later clinical therapeutic benefit. Others have suggested that these results merely confirm what many lipidologists and trialists have suggested for years; that is, if LDL-C is consistently reduced to low levels with statin therapy, HDL-C or other lipid fractions become largely irrelevant clinically. Still others have postulated that the AIM-HIGH results were “predictable”, based on research which shows that niacin, while raising serum concentrations of HDL (HDL-C), may not favorably affect HDL particles (HDL-P), but may act principally on favorably altering LDL particles (LDL-P). While the main study findings have been recently published and will be discussed below, detailed quantitative assessments of lipid subfractions, particle size and concentration, and on-treatment analyses of both LDL-C and HDL-C have yet to be initiated.

Until all of these uncertainties in AIM-HIGH can be carefully and comprehensively analyzed, the following questions might be posed as a prelude to these more definitive, upcoming analyses:

  1. Was the fundamental “HDL Hypothesis” as configured in AIM-HIGH wrong?
  2. Was the therapeutic intervention wrong?
  3. Was the trial execution wrong?
  4. Were the expectations of clinical benefit for ERN and simvastatin wrong?

The HDL Hypothesis

For more than 4 decades, there is an abundance of robust epidemiological evidence that supports the observation that low levels of HDL-C and elevated levels of LDL-C are independently predictive of the risk of developing CHD in both men and women. Numerous observational studies have demonstrated that low levels of HDL-C are associated with increased CV events. Thus, the lack of clinical benefit of extended-release niacin in patients with established CHD and low levels of baseline HDL-C is disappointing, particularly since previous niacin studies have shown apparent benefits in both surrogate outcome measures (improvements in carotid intimal media thickness [ARBITER-6/HALTS), regression of angiographic coronary artery stenoses [FATS; HATS]) and in clinical outcomes. In the original placebo-controlled Coronary Drug Project, published in 1975, high-dose immediate-release niacin (3,000 mg/day) was associated with a significant 14% reduction in CHD death or MI, a 26% reduction in non-fatal MI alone, and a 21% reduction in stroke/transient ischemic attacks—event rate reductions that are comparable to those achieved in contemporary placebo-controlled statin trials. In 1999, the VA-HIT Trial demonstrated convincingly that, in 2,531 male veterans with low levels of HDL-FC randomized to gemfibrozil vs. placebo (without a statin) demonstrated a 22% reduction in CHD death or nonfatal MI during a 5.1 year mean follow-up, while the combined incidence of CHD death/MI/stroke was reduced significantly by 24%--importantly in the absence of any LDL-C reduction (mean baseline LDL-C = 111 mg/dL; mean LDL-C at 5.1 years = 113 mg/dL). Seemingly, these data would support the so-called “HDL Hypothesis” that raising low levels of HDL-C (by 6%, or 2 mg/dL) from 32 mg/dL at baseline to 34 mg/dL at follow-up and lowering TG levels (by 31%) from 160 mg/dL at baseline to 115 mg/dL at follow-up, was associated with significant clinical event reduction. However, these relatively small changes in HDL-C do not fit well with the changes in HDL-C that might be required to provide such an effect. Some investigators have offered the alternative explanation that the changes in TG and non-HDL-C might be responsible for the observed changes in clinical events in VA-HIT. As noted above, baseline LDL-C in this trial, which pre-dated widespread statin use, was 111 mg/dL, as compared with 71 mg/dL in AIM-HIGH among those receiving a statin at trial entry. This ~40 mg/dL between-trial difference in baseline LDL-C is consistent with the significant impact statins have made on reducing both elevated LDL-C levels and cardiovascular risk.

When VA-HIT was designed in the early 1990’s, there was considerable discussion and debate about whether a combination of gemfibrozil + niacin would be a more robust HDL-C raising and TG lowering strategy than gemfibrozil alone. However, the concern was raised that the anticipated LDL-C reduction with niacin might confound the interpretation of trial results if both LDL-C and TG reduction, as well as HDL-C raising, reduced clinical events. Hence, the fact that VA-HIT demonstrated clinical benefit from gemfibrozil independent of LDL reduction should be interpreted as supporting the “HDL Hypothesis” clinically.

Beyond the important role of statins in secondary prevention, the AIM-HIGH Trial findings further underscore the profound, continuing evolutions in medical therapy that have occurred in CHD patients over the past several decades with other “disease-modifying interventions” such as anti-platelet therapy, beta-blockers in post-MI patients, and inhibitors of the renin-angiotensin system, all of which improve outcomes and were widely used in the present study. Aggregate effects of contemporary optimal medical therapy, particularly if it incorporates intensive statin therapy and achieved LDL-C levels in the low 60 mg/dL range, may have further reduced incident cardiovascular event rates during follow-up. This may make it difficult to demonstrate incremental clinical benefit with additional therapeutic lipid interventions such as niacin, fibrates, and cholesterol ester transfer protein (CETP) inhibitors. Nevertheless, trials are ongoing with niacin and CETP inhibitors in patients with stable cardiovascular disease. A large, international trial comparing simvastatin plus niacin/laropiprant versus simvastatin alone in CHD patients (Heart Protection Study-2) is in the follow-up phase after completing enrollment. This is a 25,000+ patient study that is scheduled to end in 2013. Unlike the present study, enrolled patients were not pre-selected on the basis of a particular lipid profile such as low baseline HDL-C.

Therapeutic Intervention Utilized

Data from VA-HIT have shown that gemfibrozil increased HDL particle number (HDL-P) more than HDL-C. By contrast, niacin significantly raises HDL-C but may not concomitantly increase HDL-P, while it modestly reduces LDL-C and appreciably reduces LDL particle number (LDL-P). Recent studies have shown that, after adjusting for LDL-P, HDL-C may be less strongly related to risk than HDL-P. While this could be a plausible interpretation of the AIM-HIGH Trial results from a laboratory scientist’s perspective, this remains speculative until the LDL and HDL particle size/composition analyses within AIM-HIGH are completed and reported. Thus, it remains unclear whether therapeutic interventions designed to increase HDL-C and lower TG levels (or modulate HDL-P and LDL-P levels) in the setting of intensive LDL reduction therapy with statins will result in meaningful amelioration of residual CV risk in statin-treated patients.

Trial Design and Execution

AIM-HIGH enrolled eligible subjects from among 92 U.S. clinical sites. Centers were selected because of investigator experience and prior performance in randomized trials as well as physician skill in lipid management, particularly the use of ERN. The trial was modeled to test a superiority hypothesis that the combination of ERN + simvastatin would result in a 25% incremental reduction in the primary outcome measure (time to first event) for the composite of CHD death, nonfatal MI, ischemic stroke, hospitalization for ACS, and symptom-driven coronary or cerebral revascularization. Based on the results of an earlier study showing a significant (>50%) relative risk reduction with niacin and simvastatin (HATS Trial) in CHD patients with low HDL-C, we projected the need to accrue 800 adjudicated primary events during a 5-6 year follow-up (estimated mean follow-up of 4 years) in order to have sufficient power to detect a 25% treatment effect with combination dyslipidemic therapy vs. statin monotherapy.

Much attention has been directed toward the postulated effect size difference (the 25% relative risk reduction between niacin- and placebo-treated patients), with many criticisms that this was an over-zealous estimate. The data to support the 25% effect size difference emanates from three sources:

  1. Pooled data from the Helsinki Heart Study, VA-HIT, WOSCOPS, and AFCAPS/TexCAPS indicate that, for every 1 mg/dL increase in HDL-C, there is an approximately 3% reduction in CHD death alone. We had postulated an absolute increase of approximately 8 mg/dL in HDL-C with ERN in AIM-HIGH; thus, one would estimate that, based on the above, this would result in an approximate 24% reduction in CHD death alone.
  2. A meta-analysis of 23 prospective, randomized controlled trials of statins and other dyslipidemic agents indicated that the effects of niacin, when added to a statin, would result in an additive clinical event reduction attributable to both HDL-C raising and additional LDL-C reduction. A meta-regression analysis yielded an expected reduction in CHD death and MI of approximately 36% with combination statin + niacin therapy.
  3. Outcomes data obtained from the CHARISMA Trial, a prospective, randomized, placebo-controlled study of dual antiplatelet therapy (clopidogrel + aspirin) in stable CHD patients without acute coronary syndrome (similar to the AIM-HIGH study population), published in 2006, revealed a 28-month composite event rate for CHD death, MI, stroke, and hospitalization for unstable angina of 17.9%; this was the same composite endpoint as the AIM-HIGH initial primary endpoint. Over an anticipated mean 54-month follow-up for the AIM-HIGH participants, this would have projected to an event rate of 34.5%. Importantly, CHARISMA patients were not pre-selected on the basis of low baseline HDL-C, suggesting that in AIM-HIGH the event rate might have been projected to be higher than 35%. Thus, there is no reason to conclude that the projected effect size for the AIM-HIGH population was overly optimistic.

In addition to the anticipated effects of ERN on raising HDL-C, and lowering both TG and LDL-C, we designed AIM-HIGH with an aggressive on-treatment LDL-C target of 40-80 mg/dL, in part because of the continued evolution in clinical practice that has supported lower levels of LDL-C in high-risk patients with metabolic syndrome and atherogenic dyslipidemia, as were targeted for enrollment in the trial. Yet, in the aftermath of the AIM-HIGH published results, many have asked why the investigators went to such great lengths to permit both the uptitration of simvastatin (up to 80 mg/day, if needed) and the co-administration of ezetimibe, if the objective of the trial was to elucidate the effect of ERN on raising HDL-C and reducing CV events? We recognized that not all patients would be able to achieve and maintain the target on-treatment LDL-C range of 40-80 mg/dL, even if the maximum does of simvastatin 80 mg/day was used. Thus, ezetimibe 10 mg daily was made available to all study subjects in both arms of the trial, as needed, to achieve and maintain the low levels of LDL-C that we believe represented optimal treatment—and is consistent with existing, guideline-driven therapy of high-risk patients with metabolic syndrome that AIM-HIGH targeted.

By design, we did not mandate a statin washout in patients prior to obtaining a qualifying fasting lipid panel at trial entry. Because all enrolled patients had to qualify on the basis of established CV disease and would have already met a Class I indication for statin therapy, we felt that requiring patients already treated with statins to omit guideline-driven therapy for even 4 weeks could have posed a clinical risk to patients, and an ethical conundrum for both patients and physicians. Nevertheless, our a priori expectation was that we would recruit and enroll high-risk patients with baseline LDL-C levels in the 95-105 mg/dL range. As noted above, patients already on statins who were randomized to AIM-HIGH (94% of the total) had a mean baseline LDL-C level of 71 mg/dL, in contrast to the 6% of statin-naïve patients, whose mean baseline LDL-C level was 125 mg/dL—an almost 45 mg/dL difference. Again, by comparison to the pre-statin VA-HIT, where the baseline LDL-C was 111 mg/dL, the baseline LDL-C difference was 40 mg/dL. In addition to the very well-controlled LDL-C levels in AIM-HIGH patients at baseline, the levels of baseline non-HDL-C (mean: 108 mg/dL) and apoB (mean: 81 mg/dL) were likewise very low at baseline. Hence, the patients enrolled in AIM-HIGH exhibited excellent lipid control at baseline, which reflected the proficiency and dedication of the trial investigators in optimizing lipid treatment and secondary prevention. In retrospect, it may well be that the inclusion of such very well-treated patients with such low levels of baseline LDL-C, non-HDL-C, and apoB played an important role in mitigating much of the long-term residual risk we sought to demonstrate with ERN. Importantly, 75% of the statin-treated patients in AIM-HIGH at baseline had been taking a statin for at least 1 year, and 40% of patients had been taking a statin for 5 or more years. Because of the long-standing treatment with statins and the concomitant use of aggressive secondary prevention (or disease-modifying therapies), it may have been difficult to discern incremental clinical benefit with ERN.

Expectations of Clinical Benefit with ERN + Simvastatin

Clearly, our projection of a 25% relative risk reduction in our primary endpoint was with combination dyslipidemic therapy could not be substantiated during the 36-month mean follow-up when the trial was terminated due to futility (based on the DSMB review if the first 511 adjudicated primary events). Based on this analysis of primary endpoints, the statistical probability of crossing the pre-defined upper efficacy boundary (hazard ratio < 0.75) was < 0.0001, meaning that there was a less than 1:10,000 chance that the trial, if continued as planned, would ever show a positive therapeutic outcome. The neutral primary outcome for combination dyslipidemic therapy may reflect the fact that the enrolled population was not as high-risk as we had hypothesized (e.g., by design, much higher risk patients with ACS or acute MI within the preceding 4 weeks were excluded, and only a small percentage of enrolled patients were statin-naïve at entry), or could possibly be the result of such aggressive LDL-C lowering with both simvastatin and ezetimibe, as noted above. With achieved on-treatment LDL-C values in the low-mid 60’s, it may simply have been impossible to further demonstrate incremental clinical benefit with HDL-C raising therapy.

Lastly, what could explain the lack of clinical benefit viewed from the perspective of the practicing clinical cardiologist, who is often faced with such patients who have low levels of HDL-C, but varying levels of LDL-C? As hypothesized above, it is possible that long-standing, aggressive LDL-C reduction therapy with statins may deplete the soluble constituents of the large eccentric lipid cores of vulnerable coronary plaques and, by so doing, may convert such vulnerable plaques destined for rupture (with associated sudden cardiac death, MI, or ACS) to stable, quiescent plaques, where the risk for such plaque ruptures is significantly reduced. This remains speculation at present, but could provide a plausible explanation for the findings we observed.

Clinical Implications

How should clinicians interpret the results of AIM-HIGH and, as a corollary, are there subsets of patients for whom niacin should (or should not) continue to be administered? Recognizing that the study, by design, included only those patients with established, stable, non-acute atherosclerotic cardiovascular disease, the results of our trial only apply to the types of patients we enrolled, and should not be generalized to the broader subpopulations we excluded (such as those with acute MI, ACS, or those likely to require myocardial revascularization in the subsequent 4-8 weeks after trial enrollment). However, for those stable, non-acute CHD patients with residually low levels of HDL-C who are able to achieve and maintain optimally low levels of LDL-C on a statin, the results of AIM-HIGH do not support the use of ERN to further reduce clinical risk and improve cardiovascular outcomes. Data derived from several, prospective, observational registries suggest that only about 15-20% of all treated high-risk CHD patients are able to achieve and maintain the level of very low LDL-C that we achieved in AIM-HIGH; as such, our trial results may directly apply only to this subset of CHD patients with dyslipidemia.

Thus, we believe that, until the results of HPS-2/THRIVE have been published, physicians should be prudent and cautious in applying the AIM-HIGH results to patients with mixed dyslipidemia, or in whom agents such as statins may either be contraindicated or associated with unacceptable side effects—and for which niacin would be an appropriate therapeutic alternative. The AIM-HIGH Trial did answer an important and compelling scientific question that had been heretofore unanswered, namely, that if residually low levels of HDL-C persisted in the face of optimally low levels of LDL-C on statin therapy, would extended-release niacin provide incremental clinical benefit in stable non-acute patients with established atherosclerotic CHD. Whether there is a subsequent role for niacin in other higher-risk patient subsets will have to await future prospective study.

Clinical Topics: Cardiovascular Care Team, Diabetes and Cardiometabolic Disease, Dyslipidemia, Vascular Medicine, Atherosclerotic Disease (CAD/PAD), Hypertriglyceridemia, Lipid Metabolism, Statins

Keywords: Dyslipidemias, Global Health, Hypertriglyceridemia, Medical Futility, Metabolic Syndrome, National Heart, Lung, and Blood Institute (U.S.), Peripheral Arterial Disease, Peripheral Vascular Diseases, Simvastatin, Triglycerides


< Back to Listings