QUARTERLY FOCUS ISSUE: PREVENTION/OUTCOMES: CLINICAL RESEARCH: LIPID-LOWERING AND TOTAL CARDIOVASCULAR BURDEN
Total Cardiovascular Disease Burden: Comparing Intensive With Moderate Statin TherapyInsights From the IDEAL (Incremental Decrease in End Points Through Aggressive Lipid Lowering) Trial
Matti J. Tikkanen, MD, PhD*,*,
Michael Szarek, PhD ,
Rana Fayyad, PhD,
Ingar Holme, PhD ,
Nilo B. Cater, MD,
Ole Faergeman, MD, DMSc ,
John J.P. Kastelein, MD, PhD||,
Anders G. Olsson, MD, PhD#,
Mogens Lytken Larsen, MD, DMSc,
Christina Lindahl, MD**,
Terje R. Pedersen, MD, PhD for the IDEAL Investigators
* Department of Medicine, Division of Cardiology, Helsinki University Central Hospital, Helsinki, Finland
Pfizer Inc., New York, New York
Centre of Preventive Medicine, Ullevål University Hospital, Oslo, Norway
Department of Medicine-Cardiology A, Århus University Hospital, Århus, Denmark
|| Department of Vascular Medicine, Academic Hospital Amsterdam, Amsterdam, the Netherlands
# Department of Internal Medicine, Faculty of Health Sciences, University Hospital, Linköping, Sweden
** Pfizer Sweden, Sollentuna, Sweden
Manuscript received April 30, 2009;
revised manuscript received August 17, 2009,
accepted August 25, 2009.
* Reprint requests and correspondence: Dr. Matti J. Tikkanen, Department of Medicine, Division of Cardiology, Helsinki University Central Hospital, 00290 Helsinki, Finland (Email: matti.j.tikkanen{at}helsinki.fi).
 |
Abstract
|
|---|
Objectives: This post-hoc analysis of the IDEAL (Incremental Decrease in End Points Through Aggressive Lipid Lowering) trial was designed to assess the comparative treatment efficacy of high-dose atorvastatin and usual-dose simvastatin for the prevention of events subsequent to the first event, using the Wei, Lin, and Weissfeld method.
Background: Time-to-first-event analysis of data is frequently utilized to provide efficacy outcome information in coronary heart disease prevention trials. However, during the course of such long-term trials, a large number of events occur subsequent to the first event, the analysis of which will be precluded by this approach.
Methods: The Wei, Lin, and Weissfeld method allows the analysis of repeated occurrence of events of the same type or of entirely different natures. It regards the recurrence times as multivariate event (failure) times, and models the marginal (individual) distribution for each event with the Cox proportional hazards model.
Results: In the IDEAL trial, compared with patients taking simvastatin 20 to 40 mg daily, patients receiving atorvastatin 80 mg daily had their relative risk of a first cardiovascular event reduced by 17% (p < 0.0001), of a second by 24% (p < 0.0001), of a third by 19% (p = 0.035), of a fourth by 24% (p = 0.058), and of a fifth by 28% (p = 0.117).
Conclusions: Our results indicate that intensive statin therapy continues to be more effective than standard statin therapy, even beyond the first event, and suggest that clinicians should not hesitate to prescribe high-dose statin therapy for patients experiencing multiple recurrent cardiovascular events.
Key Words: statins • cardiovascular events • WLW method
|
Abbreviations and Acronyms
| | CHD = coronary heart disease | | CV = cardiovascular | | CVD = cardiovascular disease | | MI = myocardial infarction | | WLW = Wei, Lin, and Weissfeld |
|
The conventional type of time-to-first-event analysis of trial data has proven useful in providing important outcome information concerning coronary heart disease (CHD) prevention by lowering cholesterol. This type of analysis is straightforward and simple, and allows comparisons with other trials. Unique and well-defined outcome variables, such as the composite of the "hard end points" nonfatal myocardial infarction (MI) plus CHD death, are usually defined as primary end points. However, the classical analysis of time-to-first hard end points has 2 shortcomings. First, during the course of long-term trials, a large number of events occur subsequent to the first, the analysis of which, despite their clinical and health economics importance, will be precluded by the time-to-first-event approach. Second, recent developments have provided marked improvements in the treatment of cardiovascular (CV) atherosclerotic disease, changing the scope and relevance of atherosclerosis prevention trials: patients may be hospitalized for acute coronary syndrome and treated with coronary angioplasty or thrombolysis before progressing to MI. Thus, the proportion of hard outcome variables has become smaller compared with the number of other first events, such as revascularization procedures, hospitalizations for unstable angina pectoris or congestive heart failure, and peripheral artery disease. This "changing face of CV risk" (1) has led to the situation where an increasing proportion of outcome variables are classified as secondary end points. Therefore, to gain maximal information from clinical trials, it has become important to analyze more inclusive end points containing a number of different types of events in addition to the conventional primary outcome variables, and to analyze time to events occurring after the first event.
In a post-hoc analysis of a major statin trial, the IDEAL (Incremental Decrease in End Points Through Aggressive Lipid Lowering) trial, we aimed to assess the comparative treatment efficacy of 2 statin regimens for the events beyond the first, using the broadest secondary end point (any cardiovascular disease [CVD]), consisting of the primary end point (CHD death, nonfatal MI, and resuscitated cardiac arrest) plus the following: stroke, revascularization, hospitalization for unstable angina pectoris, hospitalization for congestive heart failure, and peripheral artery disease. In this type of study, participants may experience after the first event any of the outcome variables noted in the preceding text, and nonfatal events may occur in any order. We employed Cox's proportional hazards model to estimate treatment hazard ratios separately for the time to first, second, third, fourth, and fifth events.
|
Methods
|
|---|
Patient population.
The IDEAL study randomly allocated 8,888 patients with a history of confirmed acute MI according to the prospective, randomized, open-label, blinded end point design (2); full design details and end point results were published earlier (3,4). In short, patients were men (80.8%) and women <80 years of age who qualified for statin therapy. Exclusion criteria comprised any known contraindication to statin therapy, liver enzymes >2 times the upper limit of normal, and using other lipid-lowering drugs. Randomization was carried out through a central interactive voice-response system with equal allocation to either atorvastatin 80 mg or simvastatin 20 mg daily within each center. There was no wash-out or run-in period. Simvastatin-treated patients could be titrated to 40 mg daily if plasma cholesterol at 24 weeks was at least 5.0 mmol/l. Except for such cases, lipid levels were not revealed to study personnel during the study. High-dose atorvastatin compared with usual-dose simvastatin reduced the first major coronary events (major cardiovascular event: fatal and nonfatal MI or resuscitated cardiac arrest; primary end point) by 11% (p = 0.07), and any CV event (major cardiovascular event plus stroke, hospitalization for unstable angina pectoris, coronary revascularization procedures, peripheral vascular disease, and hospitalization for nonfatal congestive heart failure) by 16% (p < 0.001).
Statistical analysis.
In 1989, Wei, Lin, and Weissfeld proposed a marginal approach (the WLW method) to the analysis of multiple event data, based on the Cox proportional hazards model (5). This method is used in the analysis of recurrent events, namely, the repeated occurrence of the same type of event or the occurrence of events of entirely different natures (6). The WLW analysis regards the recurrence times as multivariate event (failure) times, and models the marginal (individual) distribution for each event with the Cox proportional hazards model, allowing for dependence among the events. In this method, the individual subject is simultaneously at risk for all events; this preserves the randomization, and permits valid treatment effect estimation for recurrences past the first event (7). This is an advantage of the WLW approach. For example, if 5 recurrences are being analyzed, a person who experiences only 1 recurrence is considered to be at risk not only for a second recurrence but also for a third, fourth, and fifth recurrence. In contrast to other methods (8), the WLW procedure allows for valid analysis of treatment effects for recurrences subsequent to the first event due to the preservation of randomization (7). The WLW procedure can be applied to the model also for terminating events such as death or dropping out of the study (7,9). The WLW method has been used for analysis treatment trials of chronic diseases, such as human immunodeficiency virus/acquired immunodeficiency syndrome (10,11), recurrent malignancies in the urinary bladder (5,12), and rejection episodes in kidney transplant patients (13), but not, to our knowledge, in trials of CHD prevention.
The great majority of events in the IDEAL trial were nonfatal. We conducted a post-hoc time-to-event analysis to estimate the treatment hazard ratio separately for the time to the first, second, third, fourth, and fifth CV events. In addition, because treatment effect was tested on multiple event times, the WLW method was used to perform an omnibus test that controlled for type I error, to compare treatment effect across the 5 events. Times to events past the first event were computed from the time of randomization, and all subjects were included in the analysis, regardless of whether they had a first event or not. In addition, for our analysis of any CV event, CV deaths were a component of the end point. The CV events were censored at the time of death for subjects who died of a non-CV cause. So, for example, if a subject had a nonfatal MI on day 50 and died a non-CV death on day 55, the subject's first event will be day 50, and the second, third, fourth, and fifth events will be censored on day 55. Adjustments were made for age and sex for the WLW analysis, and SAS version 8.0 software (SAS Institute, Cary, North Carolina) was used.
|
Results
|
|---|
Of the 8,888 patients randomized in the IDEAL study (Fig. 1), 2,546 experienced a first any CV event, 1,048 experienced a second event, 416 a third event, 192 a fourth, and 93 a fifth event (Table 1). As shown in Figure 2, patients randomly assigned to atorvastatin 80 mg/day, compared with patients randomly assigned to simvastatin 20 to 40 mg/day, had a reduction in the relative risk of a first event by 17% (p < 0.0001), of a second event by 24% (p < 0.0001), of a third event by 19% (p = 0.035), of a fourth event by 24% (p = 0.058), and of a fifth event by 28% (p = 0.117). The omnibus test was also applied for the first through fifth events, showing that patients randomly allocated to atorvastatin compared with patients allocated to simvastatin had a relative risk reduction of 17% (p < 0.001) across all 5 events.
View larger version (11K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2 Effects of Atorvastatin 80 mg Daily Versus Simvastatin 20 to 40 mg Daily on Any Cardiovascular Event
Effects are a composite of the following: fatal or nonfatal myocardial infarction, resuscitated cardiac arrest, stroke, hospitalization for unstable angina pectoris, coronary revascularization procedures, peripheral vascular disease, and hospitalization for nonfatal congestive heart failure. *Adjusted for sex and age at baseline. CI = confidence interval.
|
|
The main types of the first 2,546 events were coronary revascularization (32%), nonfatal MI (18%), and hospitalization for unstable angina (14%). The corresponding percentages were 42%, 15%, and 9% for the second 1,048 events; and 32%, 15%, 11% for the 416 third events—suggesting that no major changes in the event profile occurred in the first through third events with sufficient numbers of events for comparing percentage distribution.
|
Discussion
|
|---|
Using the WLW method, we compared the efficacy of intensive and standard statin treatment in the prevention of these repeated occurrences of events and showed that intensive therapy increased relative risk reductions significantly for the second and third events. Risk reductions were even greater for the fourth and fifth events but lost significance, probably due to smaller event numbers. We chose this approach because treatment effects on these recurrent CHD manifestations are of great interest for both clinicians and patients, and they add significantly to disease and health economic burden.
This analysis provides new insights into the treatment of patients experiencing repeated occurrences of CVD. Often the question arises whether to continue high-dose statin therapy if such has been prescribed. The treating physician suspiciously considers the increasing possibility of drug–drug interactions, and the patient may feel that it might be prudent to use a smaller statin dose. Our results seem to indicate that, especially for such patients, intensive statin therapy is preferable to standard therapy.
A number of statistical methods have been developed to enable comparative analyses of treatment effects on multiple types of disease manifestations, which may occur repeatedly in one form or another, or may result in death (6). The WLW method is particularly useful for chronic diseases that may present with several types of first events, and in which patients may undergo each of several types of event more than once, the occurrence of 1 type of event not precluding that of others (12,13). In this analysis, all subjects (in this case 8,888 persons) are at risk from the start of the trial, and thus randomization is maintained. This differs from approaches that consider subjects at risk for a fifth event only after the recurrence of a fourth event (8). Although such an approach appears to make more sense, it does not allow valid estimation of multiple recurrent events as randomization is lost. The WLW approach considers all patients to be at risk for all events, and can be applied to modeling also for terminating events such as death or being removed from the study for any reason (7,9).
The WLW method has been used for comparative analysis of antiretroviral treatments for acquired immunodeficiency syndrome (11). This disease may typically present as a variety of different opportunistic infections or as the onset of various types of malignancies, and the patients may experience recurrences of any of these in no predictable order, or the follow-up may be terminated by death related to acquired immunodeficiency syndrome. In a similar vein, CVDs may present as any kind of CV event that may repeatedly occur in one form or another, and in no predictable order. We are not aware of WLW analyses of previous CHD prevention trials, but the method has been used for examining multiple biopsies per patient for the rejection analysis in a trial comparing tacrolimus and cyclosporine regimens in cardiac transplantation (14).
Study limitations.
The following limitations of our analysis must be considered. The analyses were based on original treatment group assignments so that switchovers in therapy and discontinuations were not taken into account. Also, our analysis does not provide insight into the interrelationship among event times, nor how prior events affect the risk of subsequent events. Thus, the WLW method does not take into account dependences that are probably present between recurrences: the severity of the second event may be influenced by the type of the first event, for example, a severe MI may have different consequences than a timely coronary revascularization.
|
Conclusions
|
|---|
A statistical method that, to our knowledge, has not been previously used for analysis of CV prevention trials was employed for the analysis of the IDEAL study cohort. We explored the possibility that new information concerning treatment efficacy of intensive statin therapy compared with standard statin therapy could be gained by analysis of repeated occurrences (after the first event) of CV events of different kinds. The method allows analysis of not only time to first CV event, but also time to second, third, fourth, and fifth events. Our results indicate that intensive statin therapy continued to be more effective than standard statin therapy even beyond the first event.
|
Appendix
|
|---|
For a supplementary table on the nth end points and their components, please see the online verison of this article.
|
Acknowledgments
|
|---|
The authors thank Dr. Paul Lane, PhD, of UBC Scientific Solutions, Ltd. for editorial support, formatting of the document for submission to the Journal, and preparation of Figure 1.
|
Footnotes
|
|---|
Continuing Medical Education (CME) is available for this article.
This study was funded by Pfizer, Inc., New York, New York. Dr. Tikkanen receives speaking and consulting honoraria from Pfizer, Merck, Orion, and AstraZeneca, and a research grant from Pfizer and Takeda. Dr. Holme receives Steering Committee honoraria from Pfizer. Dr. Faergeman receives honoraria and research grants from Pfizer, Merck, Sharpe & Dohme, and AstraZeneca. Dr. Kastelein receives consulting and lecture fees from Pfizer, AstraZeneca, Merck, and Schering-Plough. Dr. Olsson has consulting agreements with AstraZeneca, KaroBio, Merck, Sanofi-Aventis, Pfizer, and Roche. Dr. Larsen receives consultancy fees and research grants from Pfizer, Merck, Sharpe & Dohme, and AstraZeneca. Dr. Pedersen receives consultation fees and honoraria from Pfizer, Merck, Merck AG, and AstraZeneca, and research grants and Steering Committee fees from Pfizer and Merck. Drs. Cater and Fayyad are employees of Pfizer, Inc. Drs. Szarek and Lindahl were employees of Pfizer at the time of the data analysis and drafting of the paper. Dr. Szarek is currently an employee at ImClone Systems LLC.
|
References
|
|---|
1. Grundy SM. The changing face of cardiovascular risk J Am Coll Cardiol 2005;46:173-175.[Free Full Text]2. Hansson L, Hedner T, Dahlof B. Prospective Randomized Open Blinded End-point (PROBE) study: a novel design for intervention trials Blood Press 1992;1:113-119.[Medline] 3. Pedersen TR, Faergeman O, Kastelein JP, et al. Design and baseline characteristics of incremental decrease in end points through aggressive lipid lowering study (IDEAL) Am J Cardiol 2004;94:720-724.[CrossRef][Web of Science][Medline] 4. Pedersen TP, Faergeman O, Kastelein JP, et al. High-dose atorvastatin vs. usual-dose simvastatin for secondary prevention after myocardial infarction. The IDEAL study: a randomized controlled trial. JAMA 2005;294:2437-2445.[Abstract/Free Full Text] 5. Wei LJ, Lin DY, Weissfeld L. Regression analysis of multivariate incomplete failure time data by modeling marginal distributions J Am Stat Assoc 1989;84:1065-1073.[CrossRef][Web of Science] 6. Wei LJ, Glidden DV. An overview of statistical methods for multiple failure time data in clinical trials Stat Med 1997;16:833-839.[CrossRef][Web of Science][Medline] 7. Ghosh D. Methods for analysis of multiple events in the presence of death Control Clin Trials 2000;21:115-126.[CrossRef][Web of Science][Medline] 8. Prentice RL, Williams BJ, Peterson AV. On the regression analysis of multivariate failure time data Biometrika 1981;68:373-379.[Abstract/Free Full Text] 9. Li QH, Lagakos SW. Use of the Wei-Lin-Weissfeld method for the analysis of recurring and a terminating event Stat Med 1997;16:925-940.[CrossRef][Web of Science][Medline] 10. Finkelstein DM, Schoenfeld DA, Stamenovic E. Analysis of multiple failure time data from an AIDS clinical trial Stat Med 1997;16:951-961.[CrossRef][Web of Science][Medline] 11. Walker AS, Babiker AG, Darbyshire JH. Analysis of multivariate failure-time data from HIV clinical trials Control Clin Trials 2000;21:75-93.[CrossRef][Web of Science][Medline] 12. Metcalfe C, Thompson SG. Wei, Lin and Weissfeld's marginal analysis of multivariate failure time data: should it be applied to a recurrent events outcome? Stat Meth Med Res 2007;16:103-122.[Abstract/Free Full Text] 13. Cook RJ, Lawless JF. Marginal analysis of recurrent events and a terminating event Stat Med 1997;16:911-924.[CrossRef][Web of Science][Medline] 14. Taylor DO, Barr ML, Radovancevic B, et al. A randomized, multicenter comparison of tacrolimus and cyclosporine immunosuppressive regimens in cardiac transplantation: decreased hyperlipidemia and hypertension with tacrolimus J Heart Lung Transplant 1999;18:336-345.[CrossRef][Web of Science][Medline]
Related Articles
-
High-Dose Statin, Not So IDEAL?
- Eddie Vos
J. Am. Coll. Cardiol. 2010 55: 2516-2517.
[Full Text]
[PDF]
-
Cardiovascular Outcomes in Randomized Trials: Should Time to First Event for "Hard" End Points Remain the Standard Approach?
- Steven E. Nissen
J. Am. Coll. Cardiol. 2009 54: 2363-2365.
[Full Text]
[PDF]
-
Inside This Issue
J. Am. Coll. Cardiol. 2009 54: A35.
[Full Text]
[PDF]
This article has been cited by other articles:
|
|
|
|
 
D. Preiss, S. R. K. Seshasai, P. Welsh, S. A. Murphy, J. E. Ho, D. D. Waters, D. A. DeMicco, P. Barter, C. P. Cannon, M. S. Sabatine, et al.
Risk of Incident Diabetes With Intensive-Dose Compared With Moderate-Dose Statin Therapy: A Meta-analysis
JAMA,
June 22, 2011;
305(24):
2556 - 2564.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. A. Hlatky and P. A. Heidenreich
The Year in Epidemiology, Health Services Research, and Outcomes Research
J. Am. Coll. Cardiol.,
May 10, 2011;
57(19):
1859 - 1866.
[Full Text]
[PDF]
|
|
|
|
|
|
|
E. Vos
High-Dose Statin, Not So IDEAL?
J. Am. Coll. Cardiol.,
June 1, 2010;
55(22):
2516 - 2517.
[Full Text]
[PDF]
|
|
|
|
|
|
|
S. E. Nissen
Cardiovascular Outcomes in Randomized Trials: Should Time to First Event for "Hard" End Points Remain the Standard Approach?
J. Am. Coll. Cardiol.,
December 15, 2009;
54(25):
2363 - 2365.
[Full Text]
[PDF]
|
|
|
|
for this article
Top
Abstract
Methods