Effects of Torcetrapib in Patients at High Risk for Coronary Events

Study Design

We conducted a prospective, randomized, multicenter, double-blind clinical trial, using a centralized randomization strategy with a block size of four. The trial was designed by an independent steering committee in collaboration with the sponsor, Pfizer. Data were collected by PharmaNet Development Group and analyzed independently by the Statistical Data Analysis Center at the University of Wisconsin, Madison; representatives of these organizations vouch for the completeness and veracity of the data and the analyses, respectively. An institutional review board at each center approved the protocol, and patients provided written informed consent. The original protocol was amended on November 28, 2006, a change that had not yet received institutional review board approval at the time of the trial's termination. The amendment included the addition of hospitalization for unstable angina to the primary outcome to increase the number of events and thus increase the statistical power to reject the null hypothesis.

Men and women between the ages of 45 and 75 years were eligible to participate in the study if they had a history of cardiovascular disease (including myocardial infarction, stroke, acute coronary syndrome, unstable angina, peripheral vascular disease, and cardiac revascularization) 30 days to 5 years before screening. Patients with type 2 diabetes without previous cardiovascular disease who met American Diabetes Association criteria or were receiving hypoglycemic agents were also eligible. Patients were excluded if they had evidence of an unstable medical condition, a life expectancy of less than 5 years, or an LDL cholesterol level of less than 100 mg per deciliter (2.6 mmol per liter) if the patient was not receiving a lipid-altering drug. Patients were also excluded if they had had a cardiovascular event during the run-in period or uncontrolled hypertension (defined as a systolic blood pressure of >140 mm Hg or a diastolic blood pressure of >90 mm Hg) or if the LDL cholesterol target level had not been reached at the end of the run-in period.

During a run-in period of 4 to 10 weeks, patients underwent lifestyle counseling and also received atorvastatin titrated (if needed) at 2-week intervals to achieve an LDL cholesterol level of less than 100 mg per deciliter. A variability tolerance of +15 mg per deciliter (0.4 mmol per liter) was allowed. Patients whose LDL level met the target were randomly assigned to receive either atorvastatin (at a dose established during the run-in period) plus 60 mg of torcetrapib or atorvastatin plus placebo. After termination of the trial, the sponsor maintained support for independent oversight during the trial's closeout (which involved four of the academic authors) with continuing support from the Statistical Data Analysis Center, which performed the primary statistical analyses.

Study Outcomes

Patients' visits were scheduled at 1, 3, 6, 9, and 12 months after randomization. Thereafter, patients were to be seen twice yearly. An increase or decrease in the dose of atorvastatin was allowed at the 12-month visit on the basis of predefined LDL cholesterol levels. The level of HDL cholesterol was determined through enzyme analysis with the use of polyethylene glycol–modified cholesterol esterase, cholesterol oxidase, and dextran sulfate to generate peroxide that was measured calorimetrically. Total cholesterol and triglyceride levels were determined by standard enzymatic techniques. LDL cholesterol was quantified by the Friedewald formula, except when the triglyceride level was more than 400 mg per deciliter (4.5 mmol per liter), in which case the level was measured by direct beta quantification. Apolipoproteins A-I and B-100 were measured by means of immunoturbidimetric assay. High-sensitivity C-reactive protein (CRP) was measured with the use of a particle-enhanced immunoturbidimetric assay. All measurements were performed at a central laboratory (MDS). Blood pressure was measured in triplicate at each visit with the use of a standard calibrated measuring device with the patient seated after 5 minutes of rest. The first reading was not used, and the latter two readings were averaged for the visit measurement. The estimated glomerular filtration rate was calculated as described previously.¹⁵

The QT interval was calculated with the use of Bazett's correction. A decision to measure aldosterone was made after termination of the trial after investigators observed a pattern of change in serum electrolytes and blood pressure. Stored serum samples that had been obtained from patients at baseline and at 3 months were used, with measurements made by means of liquid chromatography–tandem mass spectrometry (Mayo Central Laboratory for Clinic Trials).

Efficacy Measures

The primary outcome was the time to the first occurrence of a major cardiovascular event, a composite that included four components: death from coronary heart disease (defined as fatal myocardial infarction excluding procedure-related events, fatal heart failure, sudden cardiac death, or other cardiac death), nonfatal myocardial infarction (excluding procedure-related events), stroke, and hospitalization for unstable angina.

Secondary outcomes were the time to the first occurrence of each individual component of the primary outcome, the time to death from any cause, and the change from baseline in LDL and HDL cholesterol levels. Tertiary outcomes included further breakdowns in categories and composites of secondary outcomes.

Members of a central committee who were unaware of study-group assignments adjudicated potential outcomes as reported by the investigators. Adjudicated outcomes are not included in totals of adverse events and serious adverse events.

Statistical Analysis

The original design assumed a sample size of 13,000 patients (6500 per treatment group) to yield 551 primary outcomes in the atorvastatin-only group (8.48%) and 433 in the group receiving both atorvastatin and torcetrapib (6.66%) after an average of 4.5 years of follow-up — in other words, an absolute reduction of 1.82 percentage points or a relative reduction of 21% in the cumulative incidence. This number of patients would provide a statistical power of 90% to detect a treatment difference in the primary efficacy analysis at the two-sided level of 0.05 with the use of a log-rank test. A total of 15,067 patients underwent randomization, with overenrollment owing to an increase in screening activity after notification of an upcoming closure in enrollment at each site. On the basis of the modified primary outcome in the amended protocol and on the increased number of patients, the number of primary outcomes at an average follow-up of 4.5 years was expected to be 1820.

The prespecified safety-monitoring boundary (a P value <0.01, unadjusted for multiple comparisons) was based on a log-rank test for death from any cause. An efficacy-monitoring boundary (to be implemented after approximately 50% of the expected number of events had occurred in the primary outcome) was prespecified, but the requisite event count was never reached.

All treatment comparisons were performed with the use of an intention-to-treat analysis. All data were censored for the primary analyses on December 2, 2006, when the trial was terminated. Events occurring after that date, in the period between termination of the study and the end of data collection, are also reported. These events were captured either at a final visit after the discontinuation of a study drug (active surveillance) or as a result of instructions to patients to report serious adverse events (passive surveillance). The last adjudicated outcome reported during this observation period occurred on July 15, 2007.

P values for continuous and ordered categorical data were computed with the use of a nonparametric Wilcoxon test. Pearson's chi-square test (without continuity correction) was used for dichotomous and unordered categorical data. The log-rank test was used for time-to-event analyses. Post hoc exploratory analyses were also performed; only descriptive statistics were used to identify patterns of association, since these analyses were not inferential in nature. No adjustments have been made for multiple comparisons.

Patients

Figure 1. Figure 1. Enrollment and Outcomes.

The numbers of deaths shown in the figure do not include those reported after the discontinuation of administration of a study drug on December 2, 2006.

Table 1. Table 1. Demographic and Clinical Characteristics of the Patients.

Between August 23, 2004, and December 28, 2005, a total of 15,067 patients underwent randomization at 260 centers in seven countries. Of these patients, 7534 were assigned to receive atorvastatin plus placebo (atorvastatin-only group), and 7533 were assigned to receive torcetrapib plus atorvastatin (torcetrapib group) (Figure 1). At the end of the study on December 2, 2006, the median follow-up in each group was 550 days. Earlier discontinuation of treatment had occurred in 831 patients in the atorvastatin-only group (11.0%) and in 1008 patients in the torcetrapib group (13.4%). Higher rates of discontinuation owing to nonfatal adverse events in the torcetrapib group were associated mainly with a higher frequency of hypertension, nonspecific gastrointestinal symptoms, and headache. Follow-up was 99.7% complete, with 20 patients in the atorvastatin-only group and 19 patients in the torcetrapib group who were not followed until December 2, 2006. Baseline demographic and clinical characteristics of the two groups are presented in Table 1. Patients with a history of diabetes but no evidence of cardiovascular disease at study entry represented 18.8% of the atorvastatin-only group and 17.9% of the torcetrapib group (data not shown).

Baseline and Follow-up Laboratory and Clinical Assessments

Table 2. Table 2. Changes from Baseline at 3 Months and 12 Months in Selected Measures.

Changes in lipids were evident within the first month after randomization (Fig. 1 of the Supplementary Appendix, available with the full text of this article at www.nejm.org). At 1 year, there were significant differences (P<0.001) between the torcetrapib group and the atorvastatin-only group. In the atorvastatin-only group, all lipid changes were minimal during the study; in the torcetrapib group, these changes included an increase of 72.1% in the HDL cholesterol, a decrease of 24.9% in LDL cholesterol, and a decrease of 9% in triglycerides (Table 2). Apolipoprotein measurements at 3 months reflected the changes in lipids. The mean increase in systolic blood pressure from baseline to month 12 was 5.4 mm Hg in the torcetrapib group, as compared with 0.9 mm Hg in the atorvastatin-only group (P<0.001). The change in high-sensitivity C-reactive protein from baseline to month 3 differed significantly between the two groups (P=0.01), but the magnitude of the difference (0.04 mg per liter) was small.

Electrolyte levels were similar in the two groups at baseline (Table 1). At 12 months, there was a mean decrease in potassium of 0.08 mmol per liter in the torcetrapib group, as compared with an increase of 0.06 mmol per liter in the atorvastatin-only group (P<0.001). After 12 months, 2.3% of patients in the torcetrapib group and 0.6% of patients in the atorvastatin-only group had potassium levels of less than 3.5 mmol per liter (P<0.001). In the torcetrapib group, as compared with the atorvastatin-only group, there were greater increases in levels of sodium (1.39 mmol per liter and 0.78 mmol per liter, respectively) and bicarbonate (2.28 mmol per liter and 1.93 mmol per liter, respectively; P<0.001).

At 12 months, the estimated glomerular filtration rate increased by 0.8 ml per minute per 1.73 m² of body-surface area in the torcetrapib group but decreased by 0.3 ml per minute per 1.73 m² in the atorvastatin-only group (P<0.001). The median change from baseline to month 12 in the QT interval was an increase of 3.3 msec in the torcetrapib group and a decrease of 0.3 msec in the atorvastatin-only group (P<0.001).

Post hoc measurements of aldosterone were performed for all patients for whom stored samples from both baseline and 3 months were available. One percent of samples submitted to the laboratory were of insufficient volume to permit analysis. Analysis was performed on baseline samples obtained from 6745 patients in the atorvastatin-only group (90.0%) and 6662 patients in the torcetrapib group (88.5%) and on samples obtained at 3 months for 6664 patients (88.4%) and 6562 patients (87.1%), respectively. Most analyzed samples (56.5% in the atorvastatin-only group and 52.7% in the torcetrapib group) had aldosterone levels below the lower limit of quantification for the test used (<4 ng per deciliter for samples with sufficient volume for undiluted testing and <8 ng per deciliter for samples with insufficient volume requiring dilution). Because of this, a direct comparison of median or mean values was not possible with these data. However, it was possible to calculate values in the 85th percentile and above, since this calculation depended only on the values of 8 ng per deciliter or more for which data were complete. At baseline, the 85th, 90th, and 95th percentiles were 8.5, 10.0, and 13.0 ng per deciliter, respectively, in the atorvastatin-only group and 8.3, 10.0, and 13.0 ng per deciliter, respectively, in the torcetrapib group (P=0.21). At 3 months, these percentiles were 8.6, 10.0, and 13.0 ng per deciliter in the atorvastatin-only group and 9.5, 11.0, and 14.0 ng per deciliter in the torcetrapib group (P<0.001). (The P values are Wilcoxon comparisons performed after truncating the data below 8 ng per deciliter.)

Study Outcomes

Figure 2. Figure 2. Kaplan–Meier Curves for Death from Any Cause and for the Primary Composite Outcome.

Panel A shows the between-group comparison of patients who died from any cause during the study: 59 patients in the atorvastatin-only group and 93 patients in the torcetrapib group. Panel B shows the between-group comparison of patients who had the primary composite outcome: 373 patients in the atorvastatin-only group and 464 patients in the torcetrapib group. The primary outcome was the time to the first occurrence of a major cardiovascular event, a composite that included four components: death from coronary heart disease, nonfatal myocardial infarction (excluding procedure-related events), stroke, and hospitalization for unstable angina. Analyses in both panels were censored on December 2, 2006.

Table 3. Table 3. Estimated Hazard Ratios for Protocol-Specified Cardiovascular Outcomes.

Figures 2A and 2B show Kaplan–Meier curves for death from any cause and the primary composite outcome. The hazard ratio for the primary outcome — major cardiovascular events — was 1.25 in the torcetrapib group, as compared with the atorvastatin-only group (95% confidence interval [CI], 1.09 to 1.44; P=0.001) (Table 3). The hazard ratio estimates for the individual components of the composite outcome ranged from 1.35 for hospitalization for unstable angina (P=0.001) to 1.08 for stroke (P=0.74).

Table 4. Table 4. Causes of Death.

At study termination, there were 93 deaths in the torcetrapib group and 59 in the atorvastatin-only group, for a hazard ratio of 1.58 in the torcetrapib group (95% CI, 1.14 to 2.19; P=0.006). The adjudicated causes of death are shown in Table 4. There was no significant interaction between study-group assignment and cause of death (P=0.18). In the torcetrapib group, as compared with the atorvastatin-only group, there was an increased risk of death from both cardiovascular causes (49 vs. 35) and noncardiovascular causes (40 vs. 20). No single cause of death explained the increased cardiovascular risks. For death from noncardiovascular causes, more patients in the torcetrapib group than in the atorvastatin-only group died from cancer (24 vs. 14) and infection (9 vs. 0). The primary sites of fatal cancers were similar in the two groups (Table 1 of the Supplementary Appendix). Seven of the nine deaths from infection were in patients with diabetes.

Numbers of reported major cardiovascular events and deaths occurring after the termination of the trial were similar in the two groups: 38 major cardiovascular events in each group, with 14 deaths in the torcetrapib group and 20 deaths in the atorvastatin-only group.

Adverse Events

Adverse events were reported in 86.6% of patients in the torcetrapib group and in 83.3% of patients in the atorvastatin-only group (P<0.001). Among events that were significantly more frequent in the torcetrapib group than in the atorvastatin-only group were reported hypertension, which occurred in 1411 patients (18.7%) and 564 patients (7.5%, P<0.001), respectively; peripheral edema, in 467 (6.2%) and 353 (4.7%, P<0.001); angina pectoris, in 451 (6.0%) and 360 (4.8%, P=0.001); dyspnea, in 313 (4.2%) and 243 (3.2%, P=0.003); and headache, in 412 (5.5%) and 296 (3.9%, P<0.001). Serious adverse events were reported more frequently in the torcetrapib group than in the atorvastatin-only group (16.4% vs. 15.0%, P=0.02) (Table 2 of the Supplementary Appendix). Reported neoplasms (128 in the torcetrapib group and 136 in the atorvastatin-only group) and infections or infestations (182 and 177) were reported with similar frequencies in the two groups.

Post Hoc Exploratory Analyses

Table 5. Table 5. Relationship between Changes from Baseline to 1 Month in Key Measurements and Death from Any Cause or from Coronary Heart Disease and the Primary Outcome among 7533 Patients Who Received Torcetrapib.

Post-randomization changes in selected measurements in the torcetrapib group were examined for their relationship to major cardiovascular events and death from any cause and from coronary heart disease (Table 5). The numbers and rates for these outcomes are given for subgroups whose change in the indicated measure from baseline to month 1 (or to month 3 for apolipoprotein A-I) was at or below the study-group median, as compared with above the median. The earliest time points were chosen to capture the maximum amount of information available before death. For death from any cause, higher rates were observed in association with greater decreases in potassium and greater increases in bicarbonate. For major cardiovascular events, lower rates were apparent in those with greater increases in HDL cholesterol and apolipoprotein A-I and for those who had smaller decreases in potassium and increases in bicarbonate. Paradoxically, there was an increased risk of death and major cardiovascular events in patients whose increase in systolic blood pressure was less than the median. It should be emphasized that the results shown in Table 5 were both exploratory and post hoc.

The increased mortality associated with the use of torcetrapib included increased risks of death from both cardiovascular and noncardiovascular causes. There was also a significant increase in the risk of major cardiovascular events in the torcetrapib-treated group. The question arises: By what mechanism did torcetrapib cause harm?

Clinical trials such as ours are not designed to elucidate mechanisms of either benefit or harm associated with the use of a drug. However, they may provide clues that have the potential to inform future research. To this end, we conducted a series of exploratory post hoc analyses in an attempt to gain some insight into what might have occurred.

There are at least two possible explanations for the observation of increased mortality and morbidity associated with the use of torcetrapib in our study: an off-target effect of torcetrapib, unrelated to CETP inhibition, and an adverse effect of CETP inhibition per se, with the possible generation of dysfunctional or even proatherogenic HDL cholesterol.

A known off-target effect of torcetrapib is an increase in blood pressure. At 12 months in our study, systolic blood pressure increased by a mean of 5.4 mm Hg in the torcetrapib group from baseline, a greater effect than had been observed in earlier studies of shorter duration^9,10 but consistent with the longer phase 3 imaging trials.^11,13 The relationship between changes in blood pressure and clinical outcome in the torcetrapib group was counterintuitive, with an apparent increased risk of death in patients whose increase in systolic blood pressure was less than the median. However, it appeared that an increase in blood pressure above the median identified a group with lower baseline blood pressure levels, making it difficult to interpret the relationship without further analysis. The observed reduction in potassium and increases in sodium and bicarbonate in the torcetrapib group, as compared with the atorvastatin-only group, raised the possibility that the increase in blood pressure may have been a manifestation of mineralocorticoid excess. This proposition gained further support from post hoc findings of an increase in aldosterone levels in the torcetrapib group. The mechanism by which torcetrapib may have increased aldosterone levels is unknown.

Although cardiovascular events caused by a torcetrapib-induced increase in aldosterone is one possible explanation for the observed adverse outcomes, it does not rule out other unknown off-target effects of the agent. Nor does it rule out the possibility that CETP inhibition per se may have adverse effects. It has been suggested that the inhibition of CETP may generate HDL particles that are nonfunctional or even proatherogenic.¹⁶ Our study does not address the issue of how torcetrapib has a functional effect on HDL particles, although it was interesting to note that in the torcetrapib group, rates of cardiovascular events and death from coronary heart disease were lower in those whose increase in HDL cholesterol or apolipoprotein A-I was greater than the median. However, it must be emphasized that these post hoc observations are merely suggestive and do not rule out HDL dysfunctionality, nor do they rule out the possibility that other unknown effects of CETP inhibition may have contributed to a mechanism-related adverse outcome.

In conclusion, our study neither validates nor invalidates the hypothesis that raising levels of HDL cholesterol by the inhibition of CETP may be cardioprotective. Thus, the possibility that the inhibition of CETP may be beneficial will remain hypothetical until it is put to the test in a trial with a CETP inhibitor that does not share the off-target pharmacologic effects of torcetrapib.

Committee members of the ILLUMINATE trial are as follows: Steering Committee — P.J. Barter, Heart Research Institute, Sydney (chair); M. Caulfield, the Royal London School of Medicine, London; M. Eriksson, Karolinska University Hospital, Huddinge, Stockholm; S. Grundy, University of Texas Southwestern Medical Center, Dallas; J. Kastelein, Academic Medical Center, Amsterdam; M. Komajda, University Pierre et Marie Curie, Paris; J. Lopez-Sendon, Hospital Universitario La Paz, Madrid; L. Mosca, Columbia University, New York; J.-C. Tardif, Montreal Heart Institute, Montreal; D. Waters, San Francisco General Hospital, San Francisco. Clinical Outcome Adjudication Committee — B. O'Neil, Queen Elizabeth II Health Sciences Centre, Halifax, NS, Canada (chair); C. O'Connor, Duke University Medical Center, Durham, NC; U. Sechtem, Robert Bosch Krankenhaus, Stuttgart, Germany; J.-M. LaBlanche, Hôpital Cardiologique, Lille, France; F. Welty, Beth Israel Deaconess Medical Center, Boston; J.D. Easton, Brown University, Providence, RI; M.M. Brown, University College, London; I.E. Silverman, the Stroke Center at Hartford Hospital, Hartford, CT.