Author + information
- Received August 6, 2007
- Revision received September 10, 2007
- Accepted September 17, 2007
- Published online January 29, 2008.
- ↵⁎Reprint requests and correspondence:
Dr. Uri Goldbourt, Neufeld Cardiac Research Institute, Sheba Medical Center, Tel Hashomer 52621, Israel.
Objectives This study was designed to evaluate the long-term cardiovascular benefit of bezafibrate therapy in coronary heart disease patients enrolled in the BIP (Bezafibrate Infarction Prevention) trial.
Background The BIP trial yielded a nonsignificant 7.3% reduction in the rate of major cardiac events after a mean follow-up period of 6.2 years, possibly owing to an increasing unbalanced usage of nonstudy lipid-lowering drugs (LLDs) during the course of the trial.
Methods The adjusted risk for the combined end point of cardiac death or nonfatal myocardial infarction during an extended mean 8.2-year follow-up period of the BIP trial was assessed in 3,090 patients allocated to the original bezafibrate (n = 1,548) and placebo (n = 1,542) groups of the trial.
Results During the extended follow-up period, nonstudy LLDs were administered to a significantly greater proportion of placebo-allocated patients (57%) than bezafibrate-allocated patients (53%; p = 0.02). Interaction-term analysis demonstrated that the benefit of bezafibrate therapy was pronounced (18% risk reduction; p = 0.03) without or before treatment with nonstudy LLDs initiated during follow-up and attenuated (hazard ratio 1.05; p = 0.85) after therapy with nonstudy LLDs initiated during the observation period. Consistent with these findings, treatment with bezafibrate was shown to be associated with a significant 17% risk reduction (p = 0.03) when study patients were censored from the analysis upon initiation of therapy with nonstudy LLDs.
Conclusions The data demonstrate that bezafibrate therapy in the BIP trial was associated with significant long-term cardiovascular protection that was attenuated by an unbalanced usage of nonstudy LLDs during the course of the trial.
The BIP (Bezafibrate Infarction Prevention) trial was designed to assess the effect of raising high-density lipoprotein cholesterol (HDL-C) and reducing triglycerides on cardiac risk in patients with established coronary heart disease (CHD), who exhibited total serum cholesterol in the “normal or slightly elevated” range (1). During the course of the study, HDL-C increased by 18% and triglycerides were reduced by 21% among patients treated with the study medication. However, despite these substantial lipid-modifying effects, bezafibrate therapy was associated with only a relatively small, statistically nonsignificant, 7.3% reduction in the study’s combined primary end point (fatal or nonfatal myocardial infarction [MI] or sudden cardiac death) (2). The difference in outcome in the BIP trial from other major randomized secondary prevention trials, in which fibrate therapy was associated with a more meaningful reduction in the risk of major cardiac events (3,4), may be attributed to patient selection or to increasing use of nonstudy lipid-lowering drugs (LLDs) among the placebo-allocated patients during the study. These factors may have attenuated the observed benefit of bezafibrate during the original course of the trial. Therefore, it is possible that a more pronounced benefit of the study medication may be identified after a longer period of observation and appropriate adjustment for nonstudy LLD use during follow-up.
We hypothesized that the lipid-modifying effects of bezafibrate therapy during the original double-blind phase of the BIP trial resulted in an additional reduction in cardiovascular events during a longer-term observation.
The BIP trial
The BIP trial evaluated the effect of bezafibrate versus placebo on major coronary events and mortality in CHD patients. Details of the study design have been previously published (1,2). Briefly, 3,090 male and female patients 45 to 74 years of age with a history of MI and/or angina and a lipid profile of serum total cholesterol between 180 and 250 mg/dl, low-density lipoprotein cholesterol (LDL-C) ≤180 mg/dl (≤160 mg/dl for patients <50 years), HDL-C ≤45 mg/dl, and triglycerides ≤300 mg/dl were randomized to bezafibrate 400 mg/day or placebo between May 1990 and January 1993 and followed up over a mean period of 6.2 years (median 6.2 years; interquartile range 5.3 to 6.8 years).
Extended follow-up phase
After discontinuation of the study medication, patients were observed for coronary events for an additional period, bringing the total follow-up time in the current analysis to a mean of 8.2 years (median 7.9 years; interquartile range 7.2 to 8.7 years). All cardiac events during the BIP trial and the extended follow-up period were monitored, and causes of death were established by the same independent committee of 3 experienced cardiologists.
The primary end point of the present analysis was the occurrence of cardiac death or nonfatal MI, whichever came first, during the total follow-up time (comprising the original double-blind phase and the extended follow-up phase). Secondary end points, for patients free of primary end point, included hospitalization for unstable angina, percutaneous transluminal coronary angioplasty, coronary artery bypass grafting, stroke, and death from any cause.
Baseline characteristics by treatment group were compared using the chi-square test for categoric parameters and Student t test for continuous variables. Survival curves were constructed by the Kaplan-Meier method, and the significance of the variation between them was assessed using a log-rank test.
To avoid possible bias created by the fact that a relatively high rate of nonstudy LLDs were administered to unequal proportions of patients in the 2 treatment arms, we used 2 alternative statistical methods to determine the benefit of bezafibrate therapy in patients who were treated solely with the study medication or before initiation of nonstudy LLD therapy during the course of the trial: 1) in the first method, patients who were treated with nonstudy LLDs were censored from the analysis upon initiation of those therapies; and 2) in the second method, multivariate models were further adjusted for time-dependent nonstudy LLD therapy, and the nonstudy LLD × bezafibrate interaction was used to identify the effect of the study medication before and after initiation of nonstudy lipid-lowering therapies.
Because significant baseline clinical differences existed between patients who were treated and untreated with nonstudy LLDs during follow-up, we included a propensity score (5) for receiving LLDs in the proportional hazards time-dependent models. The propensity score was calculated by running a saturated model based upon 25 baseline clinical and laboratory covariates thought to be associated with lipid-lowering therapy. To further validate our findings, all analyses that included a propensity score were repeated using multivariate Cox proportional hazards regression modeling. Adjusted Cox models included factors shown to be significantly associated with outcome (age, prior MI, history of diabetes mellitus, reported hypertension, smoking at the time of inclusion to the study, chronic obstructive pulmonary disease, and angina pectoris [AP] functional class).
The benefit of bezafibrate therapy was also assessed within prespecified risk groups (stratified by the presence or lack of the following: New York Heart Association functional class ≥2, AP functional class ≥2, history of MI, triglycerides ≥200 mg/dl, history of diabetes mellitus, HDL-C <40 mg/dl, and body mass index [BMI] >27 kg/m2). In these analyses, the prespecified covariates × treatment interaction terms were included in the models that were assessed using the 2 alternative statistical approaches described. The statistical software used for the analyses was SAS version 8.12 (SAS Institute, Cary, North Carolina).
The present study comprised patients included in the original publication (2): 1,548 patients allocated to the bezafibrate group and 1,542 patients allocated to the placebo group. Clinical characteristics, laboratory values, and medical therapy were similar in the 2 treatment groups and were described in the original publication (2). Mainly, nearly 80% of study patients had a history of MI, and 10% had treated diabetes mellitus. Beta-blockers were prescribed to nearly 40% of study patients, calcium-channel blockers to 50%, and angiotensin-converting enzyme inhibitors to 12%.
Nonstudy LLDs during the double-blind phase and the extended follow-up
Nonstudy LLDs comprised mainly statins and were administered to an increasing number of placebo-allocated patients during the double-blind phase of the trial, whereas after the BIP trial the rate of use of nonstudy LLDs increased substantially to more than 50% in both treatment groups (Fig. 1,Table 1). Notably, before and after blinded drug allocation, all types of nonstudy LLDs were administered to a greater proportion of placebo-allocated patients compared with patients allocated to bezafibrate (Table 1).
Patients in both treatment groups who received medical therapy with nonstudy LLDs during follow-up exhibited significant baseline clinical and laboratory differences from patients who were not treated with these medications during follow-up (Table 2), including a younger age, higher baseline serum levels of total cholesterol and LDL-C, and a lower frequency of medical therapy with beta-blockers. In addition, nonstudy LLDs were administered to a significantly lower proportion of bezafibrate-allocated patients who had a more advanced heart failure or AP functional class (Table 2).
Cardiac events during follow-up
During a mean follow-up period of 8.2 years, the combined end point of cardiac death or nonfatal MI occurred in 276 of the 1,548 study patients allocated to bezafibrate (17.8%) and in 313 of the 1,542 study patients allocated to placebo (20.3%; p = 0.09). The components of the combined end point consisted of, respectively, 125 (8.1%) and 144 (9.3%; p = 0.21) cardiac deaths and 177 (11.4%) and 203 (13.2%; p = 0.14) nonfatal MIs.
Survival curves, describing the probability of the occurrence of cardiac death or nonfatal MI in the 2 treatment groups (Fig. 2), demonstrated that after a total follow-up period of 9 years, the cumulative probability of the combined end point was 23.8% in the placebo group and 19.7% in the bezafibrate group, representing a 17.6% reduction in the rate of cardiac death or nonfatal MI (p = 0.03).
Similar trends, showing nonsignificantly lower cumulative event rates among patients allocated to the study medication, were displayed for the separate occurrence of cardiac death (22.6% reduction; p = 0.16), nonfatal MI (20.0% reduction; p = 0.10), and the need for coronary artery bypass grafting (12.6% reduction; p = 0.08) during follow-up, whereas the rates of hospitalization for unstable angina, coronary angioplasty, and noncardiac mortality during follow-up were similar in the 2 treatment groups (Table 3).
We attempted to adjust for the confounding effect of unbalanced nonstudy LLD usage during the course of the trial on bezafibrate efficacy by using 2 alternative statistical approaches (Table 4).
In the first method (model 1), study patients who received nonstudy LLDs during follow-up were censored from the analysis upon initiation of these medications. Using this methodology, bezafibrate therapy was shown to be associated with a significant 17% reduction in the risk of the combined end point of cardiac death or nonfatal MI compared with placebo therapy (p = 0.03). Consistent results were obtained when adjustment was made for time-dependent nonstudy LLD therapy in the proportional hazards modeling of the combined end point (model 2). This analysis demonstrated an overall 15% risk reduction (p = 0.06) with the study medication. Similar to the results that were obtained in model 1, interaction-term analysis of model 2 showed a significant 18% reduction in the risk of cardiac death or nonfatal MI with bezafibrate without or before therapy with nonstudy LLDs (p = 0.03), whereas the effect of the study medication was not significant (hazard ratio 1.05; p = 0.85) after therapy with nonstudy LLDs was initiated during follow-up (Table 4).
When the combined end point of cardiac death or nonfatal MI was analyzed within the prespecified risk subgroups (Table 5), the benefit of bezafibrate therapy was shown to be most prominent among patients with elevated triglycerides (≥200 mg/dl; 37% risk reduction) and increased BMI (>27 kg/m2; 27% risk reduction) and significantly attenuated among patients with advanced heart failure or ischemic symptoms. No significant difference in bezafibrate efficacy was shown in all other subgroups analyzed (Table 5).
In this extended follow-up study based on the intention-to-treat principle, we compared the occurrence of cardiac death or nonfatal MI in the 2 original treatment groups of the BIP trial. We have shown that lipid-modifying therapy, consisting of 400 mg bezafibrate per day, was associated with a significant long-term 18% reduction in the risk of cardiac death or nonfatal MI among study subjects who were not treated with concurrent nonstudy lipid-lowering therapies during follow-up or before nonrandomized treatment with these medications was initiated.
The VA-HIT (Veterans Affairs High-Density Lipoprotein Intervention Trial) (3) and the BIP trial (2) were 2 major randomized studies designed to determine the effect of fibrate therapy on the risk of nonfatal MI and cardiac mortality in CHD patients. Like gemfibrozil in the former study, bezafibrate in the BIP trial significantly increased HDL-C levels (by 18%) and lowered triglycerides (by 21%). However, unlike the results of the VA-HIT study, which demonstrated a significant 22% risk reduction with gemfibrozil (1,200 mg/day) during a median follow-up period of 5.1 years, the reduction in the cumulative probability of the primary end point of the BIP trial after a similar mean follow-up period of 6.2 years was of smaller magnitude (7.3%; p = 0.24).
Several possible explanations have been suggested to account for the negative findings in the BIP trial. First, LDL-C levels were in a moderately high range at baseline in the study, resulting in a relatively high frequency of treatment with adjunctive “off-trial” LDL-C–lowering therapies among placebo-allocated patients. The present study shows that nonstudy LLDs, mainly statins, were administered to an increasingly greater proportion of placebo-allocated patients during the double-blind phase in the BIP trial, possibly attenuating the effect of the study medication on outcome toward the end of the trial. Second, the BIP trial excluded several high-risk subgroups, including individuals with diabetes and higher levels of blood glucose and patients with components of the metabolic syndrome (6), in whom fibrate therapy has been suggested to be more effective (7–10). Thus, possibly owing to a higher use of off-study LLDs and less prevalent risk factors, the 5-year event rate in the placebo group in the BIP trial (13%) was relatively lower than in the VA-HIT study (22%).
In the present study, we attempted to address these 2 potential limitations of the BIP trial. First, in the present analysis, either patients were censored upon initiation of nonstudy LLDs or further adjustment was made for time-dependent medical therapy with nonstudy LLDs and its interaction with bezafibrate therapy. These 2 statistical approaches consistently demonstrated a significant long-term beneficial effect of bezafibrate therapy before additional off-trial lipid-lowering drugs were administered to an unequal proportion of patients in the 2 treatment groups. Second, the continued follow-up for fatal and nonfatal cardiac events, after discontinuation of the study medication, resulted in a cumulative event rate of 24% in the BIP trial over a mean follow-up period of 8.2 years, which is similar to the 22% event rate reported after 5 years of follow-up in the VA-HIT study (3).
The present findings are consistent with earlier studies that have shown that the benefit of fibrate therapy is more prominent among patients with components of the metabolic syndrome (7–10). However, the mechanisms underlying the findings regarding the lack of bezafibrate benefit among patients with symptoms suggestive of advanced ischemia and heart failure are not clear. Unbalanced administration of nonstudy LLDs occurred more frequently among study patients with these risk factors. Thus, a higher rate of cardiac events may have persisted in the subset of bezafibrate-allocated patients with symptoms of advanced heart failure and AP who received a lower proportion of nonstudy LLDs (and possibly other cardioprotective medications) than the respective patients in the placebo group (Table 2). Future prospective randomized trials are needed to evaluate bezafibrate efficacy in risk subgroups of CHD patients.
The present study is limited by the post hoc nature of the analysis. However, we believe that the consistent data collection regarding medical management and cardiac events after the double-blind phase of the trial allowed a comprehensive comparison of the clinical course of the original allocation groups over an extended follow-up period.
It should also be noted that the significant long-term effects of bezafibrate therapy were observed only after patients were censored from this analysis upon initiation of nonstudy LLDs or when interaction-term analysis was used to assess bezafibrate efficacy before other lipid-lowering therapies were administered. Therefore, further studies are needed to determine the incremental benefit of fibrates in the current era of widespread statin use for secondary prevention in CHD patients.
Data from major clinical trials support treatment of CHD patients with low HDL-C (11). The present findings suggest that the HDL-C–raising effect of bezafibrate therapy in CHD patients with dyslipidemia is associated with a significant long-term reduction in major cardiovascular events that extends beyond the period of active treatment with the drug.
- Abbreviations and Acronyms
- angina pectoris
- body mass index
- coronary heart disease
- high-density lipoprotein cholesterol
- low-density lipoprotein cholesterol
- lipid lowering drugs
- myocardial infarction
- Received August 6, 2007.
- Revision received September 10, 2007.
- Accepted September 17, 2007.
- American College of Cardiology Foundation
- Goldbourt U.,
- Behar S.,
- Reicher-Reiss H.,
- et al.,
- Bezafibrate Infarction Prevention Study Group
- BIP Study Group
- Rubins H.B.,
- Robins S.J.,
- Collins D.,
- et al.,
- Veterans Affairs High-Density Lipoprotein Cholesterol Intervention Trial Study Group
- Robins S.J.,
- Rubins H.B.,
- Faas F.H.,
- et al.
- Toth P.P.