November 17th, 2014

IMPROVE-IT Meets Endpoint and Demonstrates Real But Modest Clinical Benefit for Ezetimibe

After all the waiting and all the controversy it turned out to be pretty simple. The IMPROVE-IT trial did what it set out to do and reached its primary endpoint. The benefit wasn’t very big or impressive but it will be enough to put to rest concerns that ezetimibe might have been an expensive placebo or that LDL might not be a reliable surrogate endpoint. The IMPROVE-IT results will also provide comfort to companies developing the next generation of cholesterol drugs, since their approval may have depended on validation of LDL as a surrogate endpoint.

The  Improved Reduction of Outcomes: Vytorin Efficacy International Trial, presented Monday morning at the American Heart Association meeting in Chicago, randomized 18,144 high-risk patients within 10 days of an acute coronary event to either ezetimibe or placebo on top of a statin. These patients had LDL levels between 50-125 mg/dL, or between 50-100 mg/dL if on a prior cholesterol drug. Patients were followed for an average of six years. The trial ended after there were 5,250 primary endpoint events (CV death, MI, hospital admission for unstable angina, coronary revascularization more than a month after randomization, or stroke).

Primary endpoint events occurred in 34.7% of the control group versus 32.7% of the treatment group, representing a 6.4% reduction in risk (HR 0.936, CI 0.887-0.988, p=0.016). The investigators calculated that 50 patients would need to be treated for seven years to prevent one event. There was no difference between the groups in overall deaths, coronary deaths, or cardiovascular deaths, but there were significant reductions in MI (13%, p =0.002), stroke (14%, p=0.052), and ischemic stroke (21%, p=0.008). The effect was significant across subgroups, except for diabetics, who had a larger benefit than nondiabetics.

The results came about even though 42% of patients in each arm discontinued taking the study drug before the end of the trial. A secondary analysis, presenting the results of only the patients who stayed on the study drug, is scheduled for presentation on Tuesday.

As expected, LDL was significantly different throughout the trial; median levels were  69.9 mg/dL in the control group versus 53.2 mg/dL in the treatment group. No safety issues were found during the trial and there were no significant differences in cancer or muscle-related events.

In his presentation of the results, Chris Cannon said the trial provided a positive answer to three questions that have been hanging for more than a decade:

  • Whether lowering LDL with the non-statin ezetimibe can lower cardiac events?
  • Is it worthwhile to take an already low LDL cholesterol (65 mg/dL) even lower (50 mg/dL)?
  • Is ezetimibe safe?

Cannon added that the trial “reaffirms the LDL hypothesis, that reducing LDL prevents cardiovascular events” and that the results should be incorporated in future guidelines.

A Modest Proposal

Despite its history of divisive controversy, the discussion about IMPROVE-IT fell within a fairly narrow range of views. Most experts agreed that the trial results were trustworthy and genuine, but that the benefit seen in the trial was “modest.” Lipid experts in particular expressed relief that the trial did not invalidate the lipid hypothesis.

The result, said James Stein, “makes perfect sense and fits with the physiology just presented last week in the NEJM. Any drug that lowers LDL-C safely, given for a long enough time, will reduce cardiovascular disease risk. The key issues are extent of LDL-C lowering, duration, and baseline level of CVD risk. I am very glad to see these long overdue results as they give us another tool to help prevent heart attacks and strokes. They also help refocus us on the physiology — bad cholesterol — and away from the statin class of drugs as the only treatment for heart disease. It really is about the risk factors. And I can’t emphasize it enough. Anything that lowers LDL-C safely, when used for a long enough amount of time, will reduce CVD risk.”

Sekar Kathiresan, who was the senior author of the NEJM paper cited by Stein, was even more enthusiastic: “These results are a home-run for patients. In human genetics studies for heart-attack risk, LDL has always emerged as the key causal risk factor. And now, we have two medicine options for patients — statins and ezetimibe — proven in clinical trials to lower LDL as well as heart attacks. The question for LDL lowering and cardiovascular disease prevention now moves from ‘how low’ to ‘how long’.”

Kathiresan also argued that the NNT (number needed to treat) of 50 was “outstanding for a drug that basically has no side effects.” But Steven Nissen, a vocal critic of ezetimibe in the past, pointed out that the NNT occurred over seven years, so the yearly NNT was actually 350.

But Nissen also said that the trial result was trustworthy and that the trial was well conducted, but he emphasized the modest effect and asked “when have you ever seen a risk reduction as small as 6%?” He noted that the trial was powered to detect a 9% difference but got only a 6% reduction.

Nissen said he thought the results were strong enough to support the approval of the PCSK9 inhibitors for the limited indications of familial hypercholesterolemia and for those who are truly statin intolerant.

At the AHA press conference, Neil Stone, the chairman of the AHA/ACC cardiovascular guidelines committee, said that ezetimibe should now be viewed as a proven therapy. The results, he said, expand the options for additional proven lipid lowering therapies that have been shown to add incremental benefit.

Stone, Cannon, and others warned that the results of IMPROVE-IT should not be applied to lower-risk groups using ezetimibe for primary prevention. Cardiologist Melissa Walton-Shirley noted that most patients who now take ezetimibe are, in fact, low-risk patients taking it for primary prevention. Cardiologist Karol Watson pointed out that “the NNT goes up with low-risk patients but the NNH (number needed to harm) remains constant.”

Cannon said that the long length of the trial was at least partly a reflection of the excellent state-of-the-art treatment received by study patients. This was “good for the patients but bad” for those who wanted results sooner.

Cannon also pointed out that the 42% drop-out rate over the seven years of the trial worked out to 7% per year, which is actually better than other similar trials in the past.

For more of our AHA.14 coverage of late-breaking clinical trials, interviews with the authors of the most important research, and member perspectives on the most interesting presentations at the meeting, check out our AHA.14 Headquarters.


11 Responses to “IMPROVE-IT Meets Endpoint and Demonstrates Real But Modest Clinical Benefit for Ezetimibe”

  1. Sergio Kaiser, M.D., MSc, PhD, F.A.C.C. says:

    So far, LDL-lowering drugs have not contradicted genetic studies, which is not the case for HDL-raising drugs. Whether “it’s not how low you make it, it’s how (long) you make it low”…, well, maybe it is rather a combination of both!
    Statins do seem to have a pleiotropic effect which may help to achieve clinical endpoints by acting on inflammatory pathways, as Jupiter has suggested. A persistent LDL reduction in the long term would further avoid heart attacks and strokes…
    PCSK9 inhibitors have been shown to be very well tolerated and can push LDL to even lower values, never before reached by pharmacological interventions. After IMPROVE-IT, however, It will not be surprising if the two ongoing PCSK9 phase III studies in high risk patients may as well have to be extended…

  2. Antonio H. Reis, Ph.D says:

    I have several concerns about IMPROVE-IT conclusions:
    a) All IMPROVE-IT patients had low or very low LDL concentrations.
    b) The absolute risk reductions was only 2%; Such level of risk reduction indicates that the focusing on LDL lowering is excessive and not justified, and might lead to misdirect research on the many real causes of cardiovascular disease (CVD);
    c) By focusing on LDL lowering, and considering the “question for LDL lowering and cardiovascular disease prevention now moves from ‘how low’ to ‘how long’”, one may think that at that risk reduction rate, CVD would still exist at ZERO LDL concentration. With this kind o reasoning one cannot explain why at the opposite side, people with very high LDL levels never get CVD.
    d) Whenever statistically studying association of events it is never possible to distinguish between causal association and “proxy” association. Therefore, I believe that mention to that “ LDL has always emerged as the key causal risk factor “ is non-scientific and not justified.
    e) I do not share Cannon’s satisfaction about the IMPROVE-IT, and consider risky that “that the results should be incorporated in future guidelines”.
    f) Finally, I am also concerned that most of the cardiologists are so much focused on the LDL paradigm that kept their eyes closed to many other possible causes of CVD.

  3. It is pleasing to finally know ezetimibe is effective, although in a small way. Most important, that LDL-C lowering remains viable in general. And, no less significant, that nonstatin drugs can do good.

    Still, there are some statistical issues raised by Dr Sanjay Kaul about the study reaching its primary end point, and the clinical significance of the findings.

    To put IMPROVE-IT in perspective, the benefit reported applied only to the ACS population studied, not to the large populations in which ezetimibe is being used: for primary prevention and stable ischemic heart disease. More, the effects of ezetimibe were on the order of adding just 1 portion of either a vegetable, a fruit or fiber daily, and any of these interventions would occur with greater certainly in all populations, at a fraction of the cost, with perhaps some gustatory pleasure.

    Of course, if patients were to follow a Mediterranean Diet or 30 min of moderately intensive activity daily, the return would be massive in comparison: nearly absolute certainty of reducing CV events and mortality by well over 80%, in addition to prevention of developing T2D and dementia.

    Now there is some good material for shared decision-making!
    Richard Kones MD

  4. In the comment below, “occur” should read “produce similar results”


  5. Guy De Gent, MD,FACC says:

    I am sorry, but Dr Cannon is too embedded with the drug companies.A 2% ARR in yet again a combined endpoint trial ( a major problem in all recent cardiology trials). I agree with Dr. Reis that this reduction does not conclusive show causal association for LDL in atherosclerosis in this trial. HR almost equals 1 with the upper CI essentially 1, suggesting there is a reasonable chance Ho is true.Yet again, all we try to do is bring a drug to the market that now hundreds of thousand of people have to take with zero benefit.Trials like this are a pure travesty, filling the accounts of the drug companies. I would hardly call this science.In addition because of poor external validity of most of these trials, in the real world this 2%, certainly will almost be 0%.
    As a rule of thumb trials with a 2% ARR with several combined endpoints are really not worth looking at in the real world setting. I am always amazed that none of the presenters at the large symposia even know that the p- value is probably not small enough in this trial. Let alone that they really know what the p- value means. I suggest Dr Cannon look at the p- value dance :

  6. Uffe Ravnskov, MD, PhD says:

    The authors of IMPROVE-IT claim that better results may arrive if we lower LDL-cholesterol even more. Have they forgotten the results achieved by Al-Mallah et al.(1)? They found that LDL-cholesterol in patients with acute CHD measured within the first 24 hours of admission was lower than normal, not higher. They decided therefore to lower the patients LDL-C even more.

    The result? I cite:

    “At three years, patients with admission LDL 105 mg/dL (14.8% vs. 7.1%, p = 0.005, odds ratio 2.3, 95% confidence interval 1.3–4.1). This continued to be significant after adjusting for gender, race, prior myocardial infarction, hypertension, diabetes, lipid lowering therapy, prior aspirin, beta-blockers, diastolic blood pressure, and admission HDL level (OR 2.0, 95% CI 1.1–3.5, p = 0.02).”

    Obviously the authors of IMPROVE-IT have forgotten the result from ENHANCE as well, a trial with the same design as was used in IMPROVE-IT, but including individuals with familial hypercholesterolemi only (2). Although LDL-C was 36 % higher in the simvastatin group the outcome was better than in the ezetimibe/simvastatin group.

    1. Al-Mallah MH et al. Cardiol J 2009;16:227-33
    2. Kastelein JJ et al. N Engl J Med. 2008;358:1431-43.

  7. As expected, the relatively modest ASCVD benefit (6% relative and 2% absolute) in IMPROVE-IT has brought comments from cholesterol skeptics. The reduction is actually quite impressive, however, for many reasons, including: (1) over 1/3 of the events occurred within the first year after the qualifying ACS episode, and such events are likely unrelated to lipid levels (HR appears to be ~0.9 after 1 year), (2) the relationship between LDL-C and ASCVD is log-linear, so ezetimibe would likely have a much greater benefit if used at an LDL-C above 70 mg/dL, where ezetimibe is actually used in clinical practice, (3) the control group was much more likely to take a higher simvastatin dose, artifactually blunting the observed ezetimibe benefit and (4) ezetimibe is a relatively weak LDL-C lowering agent, the most important result of IMPROVE-IT being a reaffirmation of the LDL hypothesis.
    Incorporation of IMPROVE-IT into guidelines, as suggested by Dr. Cannon, simply means to extract ourselves from the “statin-only” paradigm of the 2013 ACC/AHA guidelines and to revert to the “lower is better” paradigm of all other lipid guidelines before and since. Many of us actively partner with Pharma in many ways, but that partnership doesn’t invalidate the LDL paradigm. It just means we need to be vigilant in remaining objective about Pharma-friendly data and statements.
    Yes, Dr. Reis, there are other possible causes of CVD and these must continue to be studied, but no, clinical use of the LDL paradigm doesn’t stop that research.
    To answer Dr. Ravnskov: yes, LDL-C is decreased temporarily by an ACS episode, resulting in the “hockey-stick” LDL-C curves during the trial, but the ASCVD benefit of ezetimibe occurred long after that effect subsided. Importantly, the results of the small, short-term ENHANCE trial (biased by pre-study statin use) were proven clinically irrelevant by IMPROVE-IT, not vice versa.
    Of course, ezetimibe has never been seriously advocated to replace a healthy diet, exercise or a statin, but IMPROVE-IT results warrant its use as an adjunct to these when LDL-C remains high (although sadly this was not tested directly).

  8. Carlos Cuneo, MD says:

    With these small differences between simvastatin and simvastatin/ezetimibe, should we use the combination only in our diabetic patients after a ACS ?

  9. Have we really proven that the LDL theory is correct and the lower the LDL, the fewer heart attacks occur? I think we are seeing what we are looking for, not what we are looking at.

    Prior studies of statin therapy have not clearly shown that the subjects with the lowest LDL avoid the heart attacks while those with higher LDLs have the heart attacks. Indeed in his atorvastatin/coronary calcium progression study, Raggi demonstrated that while there was a strong relationship between calcified plaque progression and MI incidence, there was no relationship between treatment LDL levels and subsequent MI.

    Statins reduce production of LDL but potentiate intestinal reabsorption, while at the same time reduce inflammation. They should be a home run however the heart protection study demonstrates a paltry 25% reduction in MI and an even lower reduction in coronary death. If the LDL theory is correct, why are these numbers not more robust.

    Zetia reduces reabsorption of cholesterol from the gut. This likely results in a reduction in the average age of LDL particles in the blood. Perhaps it is the younger, likely less atherogenic, plaque present as a result of zetia that reduces events rather than the reduced LDL level.

  10. The relationship between levels of atherogenic lipoproteins and ASCVD may seem “paltry” at times, but it is statistically and clinically significant, especially in settings where cholesterol lowering is widely acknowledged to be appropriate. Science is always evolving, but yes, the LDL hypothesis that “lower is better” is truly proven as well as anything in medical science. No other set of ASCVD risk factors has such robust and broad clinically applicable evidence! Many publications establish this, including several from the Cholesterol Treatment Trialists (CTT) group. Non-HDL-C, LDL-P and/or apo B are even stronger factors than LDL-C, and LDL particle size may also play a role.
    To me, the main message of IMPROVE-IT is that we do not need to limit the use of ezetimibe to the narrow range of the study subjects (e.g. ACS patients with DM2). Knowing that lower IS better can and should guide clinical judgement in treating high cholesterol in various high-risk patients.
    Yes, if one corrects for coronary artery calcium, cholesterol levels may drop out of the prediction equation, but that is because they work by means of their effects on atherosclerosis.
    There are reciprocal changes in cholesterol absorption and synthesis (and in PCSK9 levels) with statin and ezetimibe treatment, but they are relatively minor.
    Much has been written about anti-inflammatory and other so-called “pleiotropic” effects of statins, but there is relatively little evidence so far that these are clinically important, and IMPROVE-IT was able to show benefit in their absence.

  11. Uffe Ravnskov, MD, PhD says:

    If high total and/or LDL cholesterol is the causal factor of atherosclerosis, there should be an association between the lipid values and the degree of atherosclerosis, but this is not so. When I published my review about that issue all studies of unselected people have found lack of an association. Moreover, almost all angiographic trials have found lack of exposure-response, both for total, LDL, HDL and VLDL cholesterol, triglycerides, apolipoprotein B, apolipoprotein A1, and small, dense LDL‐cholesterol (1) People with low lipid values become just as atherosclerotic as people with high values.

    1. Ravnskov U. Is atherosclerosis caused by high cholesterol? QJM 2002; 95: 397-403.