Merck recently announced that Tredaptive, its compound of extended-release niacin and laropiprant, failed to meet the primary-endpoint goal in the HPS2-THRIVE trial, a comparison with statin therapy alone. Dr. John Ryan has now asked Dr. William E. Boden, lead investigator of the AIM-HIGH trial, for his perspective on existing evidence about niacin and other HDL-modifying therapies. A related discussion between Dr. Boden and CardioExchange’s Dr. Harlan M. Krumholz will follow next week.
Ryan: Many practitioners are losing confidence in niacin and other HDL-modifying therapies. Others agree with you that, as you said to theheart.org, “we just have not designed the right trial.” Why haven’t the right trials been conducted, and what aspects of the published trials would you change?
Boden: In designing AIM-HIGH, the NHLBI Ethics Review Panel proscribed a statin washout as being “unethical,” so we were obligated to enroll patients with established coronary disease and low HDL levels who already had a class I indication for a statin. We designed AIM-HIGH with the expectation that our starting baseline LDL level would be about 100–102 mg/dL, but we wound up with 71 mg/dL in 94% of patients who had been on statins for at least 1 year (75% of enrollees) or at least 5 years (40% of enrollees). In short, these patients were too well treated for too long on a statin to discern an incremental clinical benefit of niacin, and we enrolled far too few statin-naive patients.
The placebo tablets (each 500 mg) were spiked with 50 mg of crystalline niacin to impart a mild cutaneous flush and mask the identity of treatment. That meant patients taking 2000 mg daily (four 500-mg placebo tabs) were taking 200 mg of immediate-release niacin. This could explain why the mean on-treatment HDL increased from a baseline of 35 mg/dL to 38 mg/dL in the placebo group. Some have therefore opined that AIM-HIGH was a trial of high-dose vs. low-dose niacin, and was not really placebo controlled.
We also permitted active uptitration of LDL after randomization, by either increasing the dose of simvastatin and/or adding ezetimibe to achieve and maintain an aggressive on-treatment range of LDL (40–80 mg/dL). Mean on-treatment LDL was 62 mg/dL at 3 years—arguably the best LDL control in any randomized controlled trial of lipid-altering therapy.
Lastly, by design, we excluded ACS and acute-MI patients (admittedly, patients with a higher incidence of CV events) because we were concerned that most would undergo early PCI, which would then alter the natural history of the disease and could confound our results. So, for many reasons, in our zeal to be scientifically careful and pristine, we actually outsmarted ourselves by manipulating too many variables. This is what I meant by our not yet having designed “the right trial.”
Neither HPS-2 THRIVE nor the two failed CETP-inhibitor trials specifically targeted patients with low baseline HDL levels, like we did in VA-HIT and AIM-HIGH. These “all-comers” designs presumably seek to test the intervention in the broadest population of patients to make it more generalizable. But given the inverse and curvilinear relationship between HDL and incident CV-event rates, the steepest part of the mortality curve occurs with HDLs below 30 mg/dL, whereas a “normal” baseline HDL level of 50 mg/dL (the population mean) is on the flat part of the event curve. Increasing the HDL level by 20% (from 50 to 60 mg/dL) simply may not have much impact on events, if the risk is already too low at baseline.
Both AIM-HIGH and HPS2-THRIVE had design flaws that, in retrospect, I believe plausibly explain the clinical findings. It’s easy to worship at the altar of evidence-based medicine when you lack a gray scale. Two negative outcomes trials with niacin in studies with clear design limitations should give us some pause. The patients to whom the results apply likewise must be considered. None of the results in any way apply to ACS or acute-MI patients, as these were notable exclusions in both trials.
Ryan: The clinical role of laropiprant, a prostaglandin inhibitor, is largely unproven. Might laropiprant have obscured the clinical effects of niacin?
Boden: In HPS2-THRIVE, concomitant use of laropiprant might very well have caused unforeseen off-target effects negating the clinical benefits of extended-release niacin. With a prostaglandin inhibitor, I would worry about either bleeding or thrombosis. Merck obviously banked on the fact that this duo was a clean-drug combo, and this is a serious study-design flaw in my view. It’s also why the FDA declined to approve Tredaptive in 2008 when it came up for a panel review. Merck should never have conducted a 25,000-patient trial and spent a gazillion dollars (or euros) on an unproven drug combination that had a propensity for off-target effects that might negate niacin’s benefit. I think this was an ill-advised decision to seek regulatory approval for a drug combo before safety and efficacy had first been established in a phase III study.
Ryan: 2012 was a bad year for HDL. After the ILLUMINATE trial of torcetrapib, negative results with dalcetrapib were published in the dal-OUTCOMES trial. These findings have cast doubt on the “HDL-raising hypothesis” of additional benefit on top of statin therapy. At what stage do we just move on from HDL as a therapeutic target rather than continually reserving judgment until we get results from the next study — which is now REVEAL (HPS3-TIMI55)?
Boden: It has been a bad 12 to 18 months for HDL-raising therapies. We have no “pure” (isolated) HDL-raising therapeutic intervention. All available agents affect multiple lipid parameters, so isolating a treatment effect to one lipoprotein may be difficult, if not impossible. Some have argued that because CETP inhibitors may create large HDL particles, the result could be dysfunctional HDL, which might explain the negative/neutral effects seen in ILLUMINATE and dal-OUTCOMES. Also, both were “all-comers” designs. Expecting to raise the HDL from 50 mg/dL to 60 or even 70 mg/dL may not translate into clinical CV-event reduction in the super-normal range.
Regarding the “HDL hypothesis,” VA-HIT proved that one could lower CHD death/MI and death/MI/stroke by 22% and 29%, respectively, with gemfibrozil versus placebo in the absence of any LDL change from baseline (111 mg/dL) to 5 years of follow-up (113 mg/dL). HDL increased from 32 to 34 mg/dL, and triglycerides dropped from 164 to 120 mg/dL. So a 6% increase in HDL and a 31% decrease in triglycerides were associated with significant clinical benefit without any change in LDL — but also in the absence of statin therapy. So I think the HDL hypothesis has been upheld, but in the contemporary era of optimal medical therapy, where patients may take 6 to 8 concomitant medications, it’s becoming harder to demonstrate incremental benefit when you are layering the 7th or 8th drug on top of so many other effective agents.
What are your thoughts about HPS2-THRIVE? And don’t forget to stay tuned for the discussion between Boden and Krumholz next week.