Principal Results of the Controlled Onset Verapamil Investigation of Cardiovascular End Points (CONVINCE) Trial FREE

The background and scientific rationale, inclusion and exclusion criteria, baseline characteristics, and early blood pressure control data regarding the participants for the Controlled Onset Verapamil Investigation of Cardiovascular End Points (CONVINCE) Trial have been previously published.^1- 2 CONVINCE was planned in 1994 as a large, simple trial³ to assess the equivalence of controlled-onset extended-release (COER) verapamil and standard therapy in preventing cardiovascular disease–related events. It was also the first prospective study of the timing of acute myocardial infarction (MI), cardiovascular event–related death, and stroke—all of which have their highest incidence during the early morning hours (6 AM to noon).^4- 5 Planned mean follow-up was 5 years; however the trial was stopped 2 years early by the sponsor for commercial reasons.

Participants were randomized from 661 clinical sites in 15 countries.² All had signed informed consent, were 55 years or older, and had at least 1 other established risk factor (eg, diabetes or cigarette smoker) for cardiovascular disease, in addition to hypertension.¹

CONVINCE was a randomized, double-blind, active-controlled, multicenter, international clinical trial (Figure 1). One group initially received COER verapamil (Covera-HS in the United States, Chronovera in other countries; Pharmacia Corp, Peapack, NJ), which exerts its major antihypertensive effect 6 to 12 hours after administration.^6- 7 The active-control group began with either hydrochlorothiazide or atenolol. The choice was made by the investigator prior to randomization, based on which treatment the investigator thought would be more suitable for the individual participant, should the participant be randomized to the atenolol or hydrochlorothiazide group.

The primary objective was to compare the 2 regimens in preventing acute MI, stroke, or cardiovascular disease–related death.¹ Major secondary outcomes included (1) an expanded cardiovascular disease end point, which included hospitalization for angina, cardiac revascularization or transplant, heart failure, transient ischemic attack or carotid endarterectomy, accelerated or malignant hypertension, or renal failure in addition to the primary outcome; (2) all-cause mortality; (3) cancer; (4) hospitalization for bleeding (excluding hemorrhagic stroke); and (5) incidence of primary end points occurring between 6 AM and noon.¹

The design of CONVINCE assumed 2024 participants would develop a primary end point, which provided 84% power for detecting a 14% difference between regimens at a 2-sided significance level of .05. Estimates of noncompliance to COER verapamil (7.5 in first year; 3% each subsequent year) and lost to follow-up rates (1% per year) were incorporated in the sample size and power calculations. It was estimated that 15 000 participants would have to be enrolled to obtain the 2024 events over a 5-year mean follow-up. The sample size was increased to 16 600 (with a target of 2246 events) when it became apparent early in the study that the withdrawal rate from study medication was greater than initially assumed. Equivalence bounds for the hazard ratio (HR) were prespecified as 0.86 to 1.16.¹

The randomization schedule was stratified by site and atenolol or hydrochlorothiazide choice in successive permuted blocks of 2, 4, or 6, selected randomly. Schedules were prepared by the statistical center at the Division of Biostatistics at the University of Minnesota, Minneapolis, and provided only to a contract research organization (Parexel International, Waltham, Mass), which used them to implement a transtelephonic interactive voice response system, whereby participants could be randomized by calling a toll-free number.¹

Participants received 2 bottles for initial treatment (step 1). One bottle contained tablets, 1 to be taken at bedtime, of either placebo or 180 mg of COER verapamil, which provides plasma levels of verapamil that peak⁸ at about the same time as the early morning increase in blood pressure, pulse rate, and risk of cardiovascular disease–related events.^4- 5 The other bottle contained tablets, 1 to be taken each morning, of placebo or 50 mg of atenolol, or 12.5 mg of hydrochlorothiazide. The dose of step 1 medication was doubled if systolic blood pressure remained at or above 140 mm Hg and/or diastolic blood pressure remained at or above 90 mm Hg. If blood pressure still was not controlled after increasing the dose, 12.5 mg of hydrochlorothiazide could be added to either the initial dose of either atenolol or COER verapamil. Or 50 mg of atenolol could be added to the initial dose of hydrochlorothiazide (step 2 treatment). The added drug could also be doubled in dose if needed. All step 1 and 2 drugs were blinded. Any additional antihypertensive agent (except a nondihydropyridine calcium antagonist, thiazide diuretic, or β-blocker) could be added as a step 3 medication (nonblinded) if needed. An angiotensin-converting enzyme inhibitor was recommended (but not mandated). The study physician could change doses or medications at any time during follow-up.

All study medication for a participant was obtained by using the interactive voice response system, which provided the appropriate bottles of medication based on the step and dose of treatment the participant was currently taking. Each bottle used in the study was assigned a unique number to maintain blinding.

Participants were seen at least semiannually for blood pressure measurements, treatment dispensing, and end point surveillance. On-site data verification was performed at least annually by the contract research organization. An independent data and safety monitoring board met semiannually to review accumulating data. Confidence intervals (CIs) based on the Lan-DeMets version⁹ of the O'Brien-Fleming group sequential boundaries were used as guidelines for early termination.¹⁰ The data and safety monitoring board met 8 times and recommended continuation of the trial after each meeting. All analyses were performed independently of the sponsor by the statistical center at the Division of Biostatistics at the University of Minnesota. All study investigators and the study sponsor were blinded to all between-treatment comparisons until completion of end point data collection and review.

The sponsor closed the study 2 years earlier than originally planned for commercial reasons. A common calendar date of December 31, 2000, was chosen through which all participants would be followed up. Clinical sites were asked to verify the end point status (primary end point and survival status) of each participant as of this date. In addition, in the United States, vital status was determined through the National Death Index for 263 participants whose vital status was unknown through other sources at the end of the trial. Four previously unknown decedents and 15 individuals with previously unknown cause of death were identified as having died from cardiovascular disease.

All possible primary end points were reviewed by an end points committee that was blinded to treatment assignment. This committee consisted of 7 experts in cardiovascular medicine. Two members of the committee independently reviewed documentation provided by the clinical sites for each event to assess whether the event met prespecified criteria. When the 2 reviewers did not agree, the full committee reviewed and adjudicated the case. If documentation was insufficient to confirm (or reject) the event, the site clinician's diagnosis was used. For deaths found only through the National Death Index, the coded cause of death provided was used. Of the 729 participants with a primary event included in the analysis, the end point committee confirmed 651 (89%).

Acute MI required 2 of the following 3 conditions (1) symptoms compatible with acute MI (eg, chest pain) lasting longer than 15 minutes; (2) electrocardiographic changes (new persistent ST-segment elevation or pathological Q waves in 2 contiguous leads); or (3) increased cardiac enzymes (more than twice the upper limit of normal). A diagnosis of stroke required the presence of focal neurological deficit lasting longer than 24 hours. Imaging studies were not required to document a stroke. Any death thought to be compatible with coronary heart disease (eg, heart failure, sudden death) or cardiovascular disease was counted as a cardiovascular disease–related death.

Time to event methods (Cox proportional hazards model and Kaplan-Meier curves) were used to compare outcomes for participants randomly assigned COER verapamil with those assigned atenolol or hydrochlorothiazide. Analyses were by modified intent to treat (modified by the exclusion of 2 sites with data integrity concerns), unless otherwise specified, and were stratified by the choice of standard of care and geographic region (United States, Canada, Western Europe, and other [Eastern Europe, Mexico, Israel, or Brazil]) in which the participant's clinical site was located. Analyses of primary and secondary events considered censoring due to losses to follow-up (ie, participants for whom the primary event status was unknown on the closing date), noncardiovascular disease–related deaths (as appropriate), and the closing date of the study. Losses were censored at the date the primary event status was last known (either the date provided by the site during the closeout process; or the date of the last follow-up visit). The HR for COER verapamil vs atenolol or hydrochlorothiazide was estimated from a stratified Cox model with a binary indicator (COER verapamil vs atenolol or hydrochlorothiazide) as the sole covariate. The proportional hazards assumption was tested by including an interaction term between the randomized treatment indicator and log-transformed follow-up time. Heterogeneity of HR for selected prespecified baseline subgroups was assessed by inclusion of an interaction term (randomized treatment by subgroup variable) in the Cox model. To determine whether HRs for the primary end point varied according to time of day of the event, a competing risk analysis was performed for 4 time intervals: midnight to 6 AM; 6 AM to noon; noon to 6 PM; and 6 PM to midnight.¹¹ This analysis allows simultaneous estimation of the HRs for each time interval. The homogeneity of HRs across intervals was tested using a χ²₃ statistic. For the secondary outcome of primary events that occurred between 6 AM and noon, events with other times or unknown times were censored at the date of the event. P values are 2-tailed.

Blood pressure changes from baseline were compared between the 2 treatment groups using the t test. All analyses were performed using SAS statistical software (Version 8.0, SAS Institute Inc, Cary, NC).

Between September 1996 and December 1998, 16 602 participants were randomized. Participants from 2 sites (n = 126; 62 randomized to COER verapamil) were excluded because of data integrity concerns. Thus, our report is based on 16 476 randomized participants (Figure 1). The mean age was 66 years; 56% of participants were women.² Most participants (84%) had previously been prescribed antihypertensive drugs; hydrochlorothiazide was chosen over atenolol for 7617 (46.2%) participants; 49% had 2 or more risk factors (Table 1).²

Participants were followed up for at least 2 years and a maximum of 4.25 years; the median follow-up was 3 years. Among the participants who were alive at the time of each visit, attendance was 92% for the COER verapamil group compared with 93% for the atenolol or hydrochlorothiazide group at 12 months; 86% for both groups at 24 months; and 79% for the COER verapamil group compared with 80% for the atenolol or hydrochlorothiazide group at 36 months. On the closing date of the trial, the primary end point status was unknown for 570 participants assigned to the COER verapamil group compared with 563 assigned to atenolol or hydrochlorothiazide (P = .52 by log-rank test for time to lost to follow-up). Corresponding percentages for unknown vital status were 1.5% for COER verapamil and 1.6% for the atenolol or hydrochlorothiazide group (P = .67). In both treatment groups, those lost to follow-up were older; had higher baseline blood pressure levels; were more likely to smoke cigarettes; and have a history of MI, stroke, or transient ischemic attacks compared with participants followed up through the closing date.

Median time receiving blinded treatment was 2.2 years for both the COER verapamil group and the atenolol or hydrochlorothiazide group. By the close of the trial, 39.4% of COER verapamil group and 39.7% of the atenolol or hydrochlorothiazide group had discontinued blinded study medication. The proportion was identical for both groups at 1 year (27%). Reasons for withdrawal were available for 78% of those withdrawing in each treatment group. Participants assigned COER verapamil withdrew more often due to adverse signs or symptoms compared with those assigned atenolol or hydrochlorothiazide (P = .02); the most common reason was constipation (216 in the COER verapamil group compared with 28 in the atenolol or hydrochlorothiazide group). However, fewer participants assigned COER verapamil (n = 115) withdrew because of poor blood pressure control (treatment failures) compared with those assigned atenolol or hydrochlorothiazide (n = 207) (P<.001 by log-rank).

Both regimens lowered blood pressure significantly.² Averaged over the entire follow-up period, systolic blood pressure was reduced by 13.6 mm Hg and diastolic blood blood pressure by 7.8 mm Hg from baseline in the COER verapamil group. In the atenolol or hydrochlorothiazide group, systolic blood pressure was reduced by 13.5 mm Hg and diastolic by 7.1 mm Hg. The mean differences in blood pressure change (COER verapamil minus atenolol or hydrochlorothiazide) were small for systolic blood pressure (0.06 mm Hg; 95% CI, −0.44 to 0.56 mm Hg) and diastolic blood pressure (0.67 mm Hg; 95% CI, 0.38-0.95 mm Hg). At the last follow-up visit attended, a systolic blood pressure of less than 140 mm Hg and diastolic blood pressure of less than 90 mm Hg was achieved in 65.5% of the COER verapamil group and 65.9% of the atenolol or hydrochlorothiazide group.

During follow-up, the prescribed antihypertensive regimen was similar across randomized groups (Figure 2). In both groups, the proportion of participants prescribed more than 1 antihypertensive medication increased with longer follow-up.

At the last visit, 28.4% of the COER verapamil group and 26.1% of the atenolol or hydrochlorothiazide group were taking only the initially assigned medication. The percentages of participants taking step 2 or step 3 treatment or nonblinded medication only were similar between treatment groups. For step 2, 15.5% of the COER verapamil group were receiving treatment compared with 16.1% of the atenolol or hydrochlorothiazide group; step 3, 16.7 vs 18.2%; and no blinded medication, 39.4% vs 39.7%, respectively. Since β-blockers, by design, were not used with COER verapamil, the 2 treatment groups differed more with respect to blinded β-blocker use (0% of the COER verapamil group vs 43% of the atenolol or hydrochlorothiazide group) than with diuretic use (26% vs 44%, respectively).

There were 364 first primary events in the group randomized to COER verapamil compared with 365 among the atenolol or hydrochlorothiazide group (Table 2). Figure 3 shows the Kaplan-Meier curves for the primary end point. Among those assigned COER verapamil, 1.51% of participants had an event after 12 months of follow-up, 3.12% after 24 months, and 4.94% after 36 months. Among those assigned atenolol or hydrochlorothiazide, 1.69% of participants had an event after 12 months of follow-up, 3.10% after 24 months, and 4.73% after 36 months. The HR for COER verapamil group compared with the atenolol or hydrochlorothiazide group was 1.02 (95% CI, 0.88-1.18; P = .77). There was no evidence that this HR varied over follow-up (proportional hazards P = .36).

Two supplemental on-treatment analyses were performed. In 1 analysis, follow-up was censored 30 days after blinded study medication was discontinued (220 events among participants in the COER verapamil group compared with 217 events in the atenolol or hydrochlorothiazide group; HR, 1.05 [95% CI, 0.87-1.26]; P = .63). A second analysis included only those participants who were still taking blinded study medication 1 year after randomization. The result of this analysis, which only included events after the first year (171 events among participants in the COER verapamil group compared with 157 in the atenolol or hydrochlorothiazide group), yielded a HR of 1.14 (95% CI, 0.91-1.41; P = .25). After adjusting for age, sex, and risk factors at entry, the HRs were 1.06 (95% CI, 0.88-1.28) using the first on-treatment analysis method and 1.12 (95% CI, 0.90-1.40) using the second on-treatment analysis method.

For fatal or nonfatal MI, the HR was 0.82 (95% CI, 0.65-1.03; P = .09); fatal or nonfatal stroke, 1.15 (95% CI, 0.90-1.48; P = .26); and cardiovascular disease–related death, 1.09 (95% CI, 0.87-1.37; P = .47). Among participants in the COER verapamil group, there were 152 cardiovascular disease–related deaths (24 MIs, 18 strokes, and 110 had other cardiovascular disease). There were 143 cardiovascular disease–related deaths among participants in the atenolol or hydrochlorothiazide group (22 MIs, 21 strokes, and 100 had other cardiovascular disease).

The HR for the primary end point or cardiovascular-related hospitalization was 1.05 (95% CI, 0.95-1.16; P = .31) and 1.08 (95% CI, 0.93-1.26; P = .32) for death. Hospitalization for heart failure, a component of the secondary cardiovascular disease end point, was 30% higher with COER verapamil compared with atenolol or hydrochlorothiazide (HR, 1.30; 95% CI, 1.00-1.69; P = .05). More participants assigned COER verapamil (n = 118; 1.4%) than atenolol or hydrochlorothiazide (n = 79; 1.0%) died or were hospitalized for bleeding unrelated to stroke (HR, 1.54; 95% CI, 1.15-2.04; P = .003). Six participants in each group died of hemorrhage not related to stroke (HR, 1.02 [95% CI, 0.33-3.17]; P = .97). Cancer incidence did not vary by treatment group (310 participants reported cancer in the COER verapamil group vs 299 in the atenolol or hydrochlorothiazide group; HR, 1.06 [95% CI, 0.91-1.24]; P = .46). About 14% of participants in each treatment group were hospitalized at least once during follow-up (P = .62).

For the primary end point, treatment HRs did not vary significantly by time of day of the event (P = .43). In each treatment group, more participants had primary events between 6 AM and noon than any other 6-hour period (Figure 4). The HR for the 6 AM to noon events was 1.15 (95% CI, 0.86-1.53; P = .34). For an additional analysis that censored participant follow-up 30 days after blinded medication was discontinued, the HR for events occurring between 6 AM and noon was 1.19 (95% CI, 0.84-1.70).

When participants were grouped according to baseline characteristics, there was no evidence of a treatment by subgroup interaction for any of the predefined baseline subgroups (Table 3).

The results of this study indicate thatCOER verapamil is not equivalent to atenolol or hydrochlorothiazide in preventing cardiovascular disease–related events. The upper bound of the 95% CI for the primary end point (1.18) slightly exceeded the prespecified boundary (1.16) for equivalence of COER verapamil and atenolol or hydrochlorothiazide.^1,12 All HRs involving efficacy were close to 1.0, including the primary end point (1.02), the major secondary combined cardiovascular disease end point (1.05), and mortality (1.08). Similarly, although the HRs increased slightly in the on-treatment analyses for the primary end point, they remained close to 1.0 (1.06 and 1.12). The relative risks for the primary end point did not differ significantly across multiple subgroups (Table 3), including choice of atenolol or hydrochlorothiazide, suggesting that there was no particular baseline characteristic associated with a different outcome overall. Further analyses by choice of atenolol or hydrochlorothiazide that take into account use of treatment steps 2 and 3 are planned. The HR was also close to unity for all of the prespecified secondary cardiovascular disease–related events, with the exception of heart failure. This was expected because both diuretics and β-blockers are usually recommended for heart failure,¹³ whereas verapamil has been associated with an increased risk of heart failure in previous studies.^14- 15

The treatment regimens showed some minor and statistically nonsignificant differences in the incidence of each component of the primary end point. The incidence of acute MI was about 18% lower with COER verapamil (P = .09) than with the atenolol or hydrochlorothiazide group; this benefit was offset by a 15% higher risk of stroke (P = .26). Although quite possibly due to chance, these trends are consistent with COER verapamil's ability to inhibit platelet aggregation,^16- 18 leading to both reduced MI incidence and more bleeding. Such a mechanism has been postulated to explain aspirin's tendency to increase hemorrhagic stroke, which is outweighed by a protective effect on MI.¹⁹ These trends are opposite to those seen in the Nordic Diltiazem (NORDIL) study, which used the nondihydropyridine calcium antagonist diltiazem.²⁰ A similar contrast is observed when comparing the results of CONVINCE and a recent meta-analysis.²¹

The prospectively gathered data of Figure 4 confirm previous epidemiological observations concerning the increased risk of cardiovascular events in the early morning hours,^4- 5 but do not support the concept of chronotherapeutics.²² However, our analyses of this secondary outcome were limited by the smaller than expected number of events, and may be further confounded by nonadherence and use of multiple drugs in each treatment group to control blood pressure.

The findings of CONVINCE are subject to many limitations. The decision to terminate CONVINCE prematurely did not derive from a recommendation of the data and safety monitoring board,^23- 26 nor from review of external data from other clinical trials.^27- 28 In this respect, the study is flawed. When stopped, the results were still inconclusive with respect to the prespecified equivalence bounds. Less than a third of the planned number of events were observed, which affects the width of the 95% CIs, and limits the ability to make conclusions about any results (which may not be similar) with longer follow-up. More participants than expected discontinued blinded medication, which pushes the overall result toward the null. Approximately 7% of participants were lost to follow-up for the primary end point. In both treatment groups, participants who were lost to follow-up had baseline risk factors that placed them at a higher risk of cardiovascular disease–related events than participants who continued follow-up. Finally, the effectiveness of atenolol or hydrochlorothiazide in the population studied might differ from that in previous clinical trials establishing the efficacy of atenolol or hydrochlorothiazide, which is a common problem in interpreting the results of equivalence trials.

In summary, CONVINCE was unable to demonstrate equivalence of a COER verapamil–based antihypertensive regimen and a regimen beginning with a diuretic or β-blocker. When considered in the context of other trials of calcium antagonists, including the larger Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT),²⁹ which found that the calcium channel blocker amlodipine was not superior to the diuretic chlorthalidone, in reducing the rate of coronary heart disease or stroke and was associated with a higher rate of heart failure, these data indicate that the effectiveness of calcium channel blocker therapy in reducing cardiovascular disease–related morbidity and mortality is similar but not better than diuretic or β-blocker treatment. These data support the recommendation of the Sixth Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure³⁰ for low-dose diuretic (or possibly β-blocker) therapy for hypertensive patients who have no specific indication for another antihypertensive drug.