Author + information
- Steven E. Nissen, MD* ()
- ↵*Reprint requests and correspondence:
Dr. Steven E. Nissen, Cleveland Clinic Foundation, Cardiovascular Medicine, 9500 Euclid Avenue, Desk F-15, Cleveland, Ohio 44195-0001
This issue of the Journalcontains 2 papers that provide interesting insights from 2 large randomized lipid-lowering trials that addressed relatively similar scientific questions. In the PROVE IT–TIMI 22 (Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis In Myocardial Infarction 22) trial, investigators compared atorvastatin (80 mg) with pravastatin (40 mg) for patients randomly assigned to treatment after an acute coronary syndrome (ACS) (1). In the IDEAL (Incremental Decrease in End Points Through Aggressive Lipid Lowering) trial (2), the 80-mg dosage of atorvastatin was compared with 20 to 40 mg of simvastatin in patients with a prior myocardial infarction (MI). Both studies were performed well and contributed significantly to the contemporary understanding of the benefits of intensive reduction in cholesterol in high-risk secondary prevention patients. As a consequence of these studies and others, the most recent update to the National Cholesterol Education Program recommends a target level of low-density lipoprotein cholesterol of 70 mg/dl for high-risk patients (3).
Both studies pre-specified analysis of efficacy based upon the time to first cardiovascular event, a time-honored approach for comparing survival curves for alternative interventions. In such analyses, cardiovascular events occurring subsequent to the first event are ignored or “censored” and therefore do not contribute to the statistical analysis. Now, both sets of investigators are reporting exploratory analyses of these 2 trials using an alternative approach in which the previously censored events are included in the analysis (1,2). However, there are important differences between the approaches used by the PROVE IT–TIMI 22 and IDEAL investigators for their respective re-analyses. In the PROVE IT–TIMI 22 trial re-analysis, Murphy et al. (1) examined exactly the same types of events included in the original paper. In the IDEAL study re-analysis, Tikkanen et al. (2) sought to examine a broader composite end point that included many additional categories of cardiovascular events not included in the original publication.
The rationale for these exploratory analyses is clearly stated. As articulated by the 2 groups of investigators, atherosclerosis is a lifelong disease, and patients continue to face risks even after they suffer an initial cardiovascular event. Because the disease targeted by lipid-lowering therapy, atherosclerosis, is a pan-vascular process, involving nearly every major blood vessel, the IDEAL study investigators included end points not originally used as a component of the primary efficacy analysis, such as stroke, hospitalization for unstable angina, revascularization or heart failure, and peripheral vascular disease. For both the PROVE IT–TIMI 22 and IDEAL re-analyses, the investigators had similar and worthy goals: to more fully describe the effects of intensive lipid-lowering therapy on the total burden of ischemic disease in cardiovascular patients.
The findings of these 2 re-analyses are interesting, although perhaps not surprising. When additional events are considered, the number of events prevented by more intensive lipid-lowering therapy was substantially greater. For the PROVE IT–TIMI 22 trial, the authors calculate a needed-to-treat number of 14, which is impressively small (1). Although not calculated by the authors, the apparent needed-to-treat number for the IDEAL study using the broader definition of morbidity-mortality events was also quite small (2); in that study, there was a 17% overall reduction in risk, and the total number of events was very large, namely, 4,295. Taken together, these post-hoc analyses provide further support for the “lower is better” concept in the management of low-density lipoprotein cholesterol in high-risk patients.
Although these were post-hoc exploratory analyses, an obvious question is whether such approaches should be used more commonly in prospective randomized clinical trials. There are 2 principal issues. First, should a broader composite of morbid events be used in major cardiovascular outcomes trials? Proponents of the use of broader end points contend that events such as hospitalization for unstable angina, heart failure, or revascularization represent important adverse experiences for patients. Hospitalization and invasive procedures are unpleasant and result in considerable anxiety, discomfort, and cost. Similarly, peripheral vascular events can induce suffering and markedly impair quality of life. If the goal of medical therapy is to relieve suffering, why should such events be routinely excluded from primary efficacy analyses?
A further argument in favor of the use of broader end points is the contemporary reality that cardiovascular event rates have decreased dramatically in recent years. In the PROVE IT–TIMI 22 trial, ≈25% of patients suffered an event during 30 months of follow-up. However, with the development of better therapies, including intensive statin administration and dual antiplatelet therapy, clinical trials under design involving ACS patients postulate substantially lower event rates. As a consequence of lower event rate estimates, there has been a literal explosion in sample sizes for more recent ACS or secondary prevention trials, now frequently approaching or exceeding 20,000 patients. The costs of such studies are daunting and may now result in a failure of sponsors to be willing to expend the resources necessary to answer important scientific questions.
Despite these rationales for use of broader end points, there are compelling counterarguments. So-called “hard” outcomes such as death, MI, and stroke are irrevocable events with permanent consequences, whereas hospitalization or revascularization may not have enduring consequences. Indeed, some fraction of these “soft” events may be considered “false alarms.” For example, in some patients hospitalized for angina, the cardiac enzymes are never elevated, no procedures are undertaken, and the patient is discharged relatively promptly with no permanent consequences. Patients admitted for congestive heart failure may actually have an exacerbation of chronic lung disease.
We must also consider the reliability of such end points. Most major outcomes trials are conducted globally, not infrequently including patients from industrialized nations in North America and Europe, but also involving emerging or less developed countries in Asia, Eastern Europe, and elsewhere. In such nations, the indications for hospitalization may differ greatly from those in North America or Europe. Indeed, even within Europe, different health care delivery models may result in very divergent hospitalization rates for identical presenting symptoms. Accordingly, the region of the world contributing the events can heavily influence interpretation of studies using soft end points.
There is also a major problem in adjudicating soft events. Most modern trials rely upon centralized, blinded adjudication committees to review source material and determine whether an investigator-reported event should be included in the final analysis. Hard events, such a MI or stroke, are relatively easy to adjudicate. Source materials such as electrocardiograms or computed tomography of the brain can be objectively evaluated, and the event classified relatively reliably. However, for many of the soft events, adjudication is difficult and inherently less reliable. Was a hospitalization truly caused by incident congestive heart failure or pneumonia? Is the onset of “intermittent claudication” a correct or incorrect diagnosis in a patient with symptoms? In global trials, the availability of source material may be compromised by local standards of care. In the U.S., for example, angiography is more often employed in the diagnosis of patients with suspected coronary ischemia. In Europe, and particularly in developing countries, such resources are used more sparingly.
There remains another major risk in the application of broad cardiovascular end points. Not all therapies have an equal effect on all types of adverse cardiovascular outcomes. An infamous example comes from the diabetes literature in the form of the PROACTIVE (Prospective Pioglitazone Clinical Trial in Macrovascular Events) trial, which compared pioglitazone to placebo for patients with type II diabetes mellitus (4). The principal investigator was a vascular surgeon, and therefore the study included events such as leg revascularization and amputation in the broad composite outcome. This composite end point failed to achieve statistical significance (hazard ratio: 0.90, p = 0.095). However, the hard end point of all-cause mortality, MI, and stroke showed a hazard ratio of 0.84 (p = 0.027). By including nonmodifiable events, the investigators missed an opportunity to advance scientific understanding of the effects of a diabetes therapy.
Because of such concerns, a broad composite outcome such as the end points used in this re-analysis of the IDEAL trial is not yet viewed as acceptable as a primary efficacy measure for contemporary trials, particularly by the regulatory community. To my knowledge, the U.S. Food and Drug Administration has not granted approval or modified the label of products on the basis of results of a trial using an end point as broad as that employed in the IDEAL study re-analysis. Such analyses are useful as secondary end points, but should not substitute for hard cardiovascular outcomes as the primary efficacy measure.
What about using the totality of cardiovascular event, including all adverse cardiovascular outcomes, not just the first post-randomization event, as employed by the authors in both of these papers (1,2)? Again, there are cogent arguments in favor of such an approach, as nicely articulated by both sets of authors. However, there are also compelling counterarguments. In coronary disease, the best predictor of a cardiovascular event is the temporal proximity of the patient to a previous event. Stated another way, the event rate after an ACS event is very high is the first few hours or days after the initial event, then declines gradually over the next 1 to 2 years. Accordingly, the patients who suffer an initial event in a clinical trial do not face a similar risk as patients who have not yet suffered a morbid event—their risk is much higher. Accordingly, the apparent benefit of a more effective therapy is greatly magnified by counting subsequent events.
As a result of this phenomenon, counting recurrent events overemphasizes the contribution of a few patients who have an early event in a clinical trial. The conventional approach, the time to first event, allows each patient to contribute an equal weight to the analysis. All patients can be considered as having an event or not having an event. However, when multiple events are considered, patients with recurrent events begin to dominate the analysis.
In summary, the findings reported in this issue of the Journalby these 2 groups of investigators are useful and contribute to our understanding of the totality of the benefits of intensive lipid-lowering therapy (1,2). In my view, however, they do not provide compelling evidence for the use of such methods in prospectively designed trials. The risks of overstating the benefits of any studied therapy are considerable, and the reliability of efficacy analyses may be compromised. Some adjustment in the use of the “hardest” end points (stroke, MI, and cardiovascular death) must be considered because of the explosion of sample size in contemporary trials and the recognition that some soft end points contribute to the overall morbidity of this disease. Hospitalization for unstable angina or revascularization has been included as a component of the primary efficacy measure in some recent trials and has generally yielded reliable results. Nonetheless, the most objective standard remains time-to-event analysis for cardiovascular death, MI, and stroke.
Dr. Nissen has received research support to perform clinical trials through the Cleveland Clinic Coordinating Center for Clinical Research from Pfizer, Astra Zeneca, Novartis, Roche, Daiichi-Sankyo, Takeda, Sanofi-Aventis, Resverlogix, and Eli Lilly. He consults for many pharmaceutical companies, but requires them to donate all honoraria or consulting fees directly to charity so that he receives neither income nor a tax deduction.
↵* Editorials published in the Journal of the American College of Cardiologyreflect the views of the authors and do not necessarily represent the views of JACCor the American College of Cardiology.
- American College of Cardiology Foundation
- Murphy S.A.,
- Cannon C.P.,
- Wiviott S.D.,
- McCabe C.H.,
- Braunwald E.
- Tikkanen M.J.,
- Szarek M.,
- Fayyad R.,
- et al.,
- IDEAL Investigators
- Grundy S.M.,
- Cleeman J.I.,
- Merz C.N.,
- et al.,
- Coordinating Committee of the National Cholesterol Education Program
- Dormandy J.A.,
- Charbonnel B.,
- Eckland D.J.,
- et al.,
- PROactive Investigators