Azithromycin for the Secondary Prevention of Coronary Heart Disease Events

Patients

To be eligible for inclusion in the trial, patients must have had a history of myocardial infarction more than 6 weeks before screening documented by electrocardiogram or by elevation of creatine kinase and have an IgG titer to C pneumoniae of 1:16 or greater by microimmunofluorescence. Those who met the inclusion criteria could not have had coronary artery bypass graft surgery or percutaneous coronary intervention in the preceding 6 months, did not require chronic antibiotic therapy and had not received antibiotics in the previous 3 months. Patients were enrolled at 271 centers from the United States, Canada, United Kingdom, Germany, France, Spain, Austria, India, and Argentina. Each of the participating sites received approval from an investigational review board and all participants in the trial provided written informed consent.

Treatment and Follow-up

The design of the WIZARD trial has been previously published.²⁴ The trial began on October 10, 1997, and follow-up concluded December 7, 2001. After a screening visit at which C pneumoniae antibody titers were measured, patients meeting eligibility criteria were randomly assigned to receive either azithromycin 600 mg or a matching placebo once daily for 3 days then once weekly for the subsequent 11 weeks. Sites were provided blinded medication in block sizes of 4, with an equal proportion of each treatment assignment. Sealed envelopes containing the treatment assignment were provided to each site, to be opened only in case of an emergency. The sponsor monitored the integrity of these envelopes at each monitoring visit. Patients, investigators, clinical site monitors, and the sponsor project team remained blinded through study completion.

Patients were seen in the clinic for collection of blood samples or contacted by telephone specifically at 6-week to 4-month intervals throughout the trial. Blood samples were analyzed for red blood cell indices, creatinine, and liver function.

Study End Points

A primary event was defined as death by any cause, recurrent myocardial infarction, coronary revascularization procedure (coronary artery bypass graft surgery or percutaneous coronary intervention), or hospitalization for angina, whichever came first. A committee blinded to treatment assignment reviewed detailed documentation of potential primary events for consistency with end point criteria defined in the protocol. Reinfarction was defined by the presence of 2 of the 3 following signs: the investigator's assessment of clinical signs and symptoms of ischemia, an elevation in levels of cardiac markers, specifically creatine kinase (CK) more than 2 times the upper limit of normal, CK-MB% at least 3%, troponin I or T more than 2 times the upper limit of normal, or new, diagnostic Q-waves in 2 contiguous leads. Hospitalization for angina was defined as exacerbation of the patient's usual symptoms that resulted in urgent hospitalization, supported by either electrocardiographic evidence of ischemia or an elevated level of a cardiac marker (troponin I or T, CK, or CK-MB) greater than normal but not diagnostic for myocardial infarction. Components of the primary end point were analyzed separately. Predefined secondary events included a noncoronary atherosclerotic event (stroke, transient ischemic attack, or intervention for peripheral vascular disease, whichever occurred first), cardiovascular death, and hospitalization for congestive heart failure.

Monitoring Guidelines

A data and safety monitoring board was responsible for monitoring for safety and efficacy. To preserve the overall type I error, an α-spending approach was used to determine interim monitoring boundaries. A conditional power calculation was used for determining the probability of a positive result. An independent statistical analysis center prepared data and safety monitoring board reports on the accumulating data.

Sample Size

The original sample-size calculations assumed that there would be a primary-event rate of 8% per year in the placebo group and that the treatment effect would be a 25% reduction in the hazard rate.²⁴ If these assumptions were met, the study with a total enrollment of 3300 patients and a total of 522 primary events would have 90% power to detect a treatment difference at a significance level of .05. The maximum study duration was originally estimated to be 36 months. During the course of the trial, information newly available provided support for the clinical utility of treatment effects of as low as 15%. As a consequence, after approximately 83% of the original targeted number of events had accrued, without unblinding the data, in April 2000, the protocol was amended to retain 90% power to detect a smaller treatment effect of 18.5% such that the study would then conclude when 1038 end points had accrued. To detect this effect within 12 to 24 months, approximately 4000 additional patients were enrolled.

Statistical Analyses

Data analyses were conducted using a modified intention to treat approach (all enrolled patients who received at least one dose of study drug). An evaluable patient assessment was also performed that included patients who met inclusion and exclusion criteria and took at least 80% of their assigned study drug with patients censored after use of an antibiotic from the tetracycline, macrolide, or quinolone classes. The primary analysis of the time to the primary event was performed using the χ² log-rank test. Components of the primary event and secondary end points were analyzed in a similar manner. Patients were censored at last contact or if all scheduled visits were completed at study end. For secondary end points without death, patients were censored at date of death. The Cox proportional hazards model was used to compute hazard ratios (HRs) and to consider possible confounding variables while the cumulative event curves were estimated by the product-limit (Kaplan-Meier) method. Risk reduction was calculated as 100% × (1 − HR). Annualized event rates are the number of events divided by person-years of follow-up. An assessment for departures from the proportional hazards assumption was prespecified in the protocol and formally tested by considering a Cox proportional hazards model with a time-dependent time-treatment interaction. In addition, a series of log-rank tests were performed on the observed data in which patients were censored at 6 months, 12 months, and every 6 months thereafter until 4 years. Statistical analyses were performed using SAS version 6.12 (SAS Institute Inc, Cary, NC). For binomial data, the Pearson χ² test was used.

Screening, Treatment, and Follow-up

Of the 11 451 patients who were screened, 7747 were randomized, 3538 of whom were randomized between October 10, 1997, and July 3, 1998, and 4209 between July 17, 2000, and July 22, 2001 (Figure 1). The primary reason for screening failure was an antibody titer to C pneumoniae of less than 1: 16 (19.5%). There was an equal distribution of patients by demographic variables including cardiac risk factors (Table 1). All but 25 randomized patients received the study drug. More than 90% of patients completed the 12 weeks of study treatment. The median time since the patient's most recent myocardial infarction was 2.6 years. The mean and median follow-up times were 25 months in the 1997-1998 group and 14 months in the 2000-2001 group. Forty-six percent of the patients were followed up for at least 42 months whereas only 35% were followed up for fewer than 12 months. Approximately 4.5% of patients chose not to continue through the December 7, 2001, close of the study.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 1. Study Flow Diagram

Analyses of Primary and Secondary Events

The annualized primary event rate for patients receiving placebo was 8.02%. Treatment with azithromycin was associated with a 7% nonsignificant reduction in the risk of the primary event (Figure 2) (95% CI, −5% to 17%; P = .23). There were no significant reductions in the risk for any components of the primary end point: death (8%), recurrent myocardial infarction (7%), revascularization procedures (5%), or hospitalizations for angina (−1%). Adjustment for cardiac risk factors including age, total cholesterol level, and history of smoking, hypertension, or diabetes, produced a similar result (RR, −6.5%; 95% CI, −5% to 17%; P = .27). A variety of other sensitivity analyses were performed on the primary end point (Figure 3). No difference in conclusions was seen after analyzing the data with nonadjudicated, investigator-specified end points (RR, 8%; 95% CI, −2% to 18%; P = .13) or including only evaluable patients (RR, 9%; 95% CI, −4% to 21%; P = .16). The time to other events related to the progression of atherosclerosis, including noncoronary disease (stroke, transient ischemic attack, peripheral vascular disease), coronary angiograms, and hospitalizations for congestive heart failure, was no different for either treatment regimen.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 2. Occurrence of Primary End Point

Kaplan-Meier estimates of the cumulative frequency of recurrent myocardial infarction, a coronary revascularization procedure, a hospitalization for angina, and death during follow-up postrandomization.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 3. Primary End Point Analyses

CI indicates confidence interval.

The proportional hazards assumption was examined because the length of the treatment period was 12 weeks and there was concern that the effect of treatment might be transient, a violation of this assumption. A Cox proportional hazards model that included a time × treatment interaction suggested an initial treatment effect that decreased with time for both the primary end point (P = .15) and the composite of death or recurrent myocardial infarction (P = .09). An analysis of the treatment effect through various observation periods is provided in Figure 4.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 4. Comparison of Hazard Ratios at Various Censoring Times

Hazard ratio for treatment effect with azithromycin at various times after randomization. Error bars indicate 95% confidence interval (CI).

C pneumoniae Titers

No significant association between C pneumoniae titers and treatment effect was observed (Figure 5). In addition, no association between titers to C pneumoniae and the likelihood of developing a primary event was seen within the group of patients randomized to either azithromycin or placebo. There was no difference between treatment groups at baseline or after baseline in the distribution of C pneumoniae titers. The risk reduction for the primary event was unchanged when the log of the baseline C pneumoniae titer was included in the Cox proportional hazards model.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 5. Treatment Effect by Risk Factors and Subgroups

MI indicates myocardial infarction. Box sizes in the plot reflect sample sizes.

Other Subpopulations

To investigate the effect of treatment within subpopulations, several different analyses were performed. Within groups of patients sharing specific cardiovascular risk factors, trends in reduction of risk attributed to treatment were most apparent in men, patients with diabetes, active smokers at baseline, and those with a history of hypercholesterolemia (Figure 5). In the subgroup of patients who had diabetes and smoked (n = 237), the annualized event rate for the primary end point was 14.6% for those receiving azithromycin vs 53% for those receiving placebo (HR, 0.47; 95% CI, 0.25-97).

In examining the treatment effect within the subsets of patients taking aspirin, statins, angiotensin-converting enzyme inhibitors, or calcium channel blockers (data not shown), no difference between the azithromycin and placebo groups was identified. In addition, an analysis assessing treatment effect by the total number of any one of these medications did not have a noticeable impact on study end point. No significant difference in treatment effect was observed when patient subpopulations were assessed by the time since their most recent myocardial infarction.

Safety

Adverse events related to study drug were reported by 13.2% of those randomized to azithromycin compared with 4.6% randomized to placebo. The most common adverse effects were related to the gastrointestinal tract, including diarrhea (8.1% vs 1.4%) and abdominal pain (2.2% vs 0.8%). Discontinuations from drug occurred in 1.6% and 0.4% of patients receiving azithromycin and placebo, respectively. There were fewer episodes of respiratory tract infection, bronchitis, pneumonia, or sinusitis reported during study drug dosing among patients taking azithromycin compared with those taking placebo (6.4% vs 9.4%, respectively; P<.001). Patients receiving azithromycin reported less use of other antibacterial drugs during the course of the study (32.7% vs 36.6%, P<.001).