Early and late recurrent cardiovascular events among high‐risk patients with an acute coronary syndrome: Meta‐analysis of phase III studies and implications on trial design

Abstract Background Despite low‐density lipoprotein cholesterol‐lowering therapies and other standard‐of‐care therapy, there remains a substantial residual atherosclerotic risk among patients with an acute coronary syndrome (ACS). This study aims to estimate the risk of early and late recurrent major adverse cardiovascular events (MACE) and address its implications on trial design. Methods A literature search was performed to collect phase III interventional trials on high‐risk ACS patients. Pooled event rates at 90 and 360 days were estimated by fitting random‐effects models using the DerSimonian–Laird method. Under the assumption of a total sample size of 10,000 and 1:1 allocation at a one‐sided alpha of 0.025 using the log‐rank test, the relationship between power and relative risk reduction (RRR) or absolute risk reduction (ARR) was explored for early versus late MACE endpoint. Results Seven trials representing 82,727 recent ACS patients were analyzed. The pooled rates of recurrent MACE were 4.1% and 8.3% at 90 and 360 days. Approximately 49% of events occurred within the first 90 days. Based on the estimated risks at 90 and 360 days, to attain 90% statistical power, a lower magnitude of RRR is required for late MACE than early MACE (22% vs. 30%), whereas a lower magnitude of ARR is required for early MACE than late MACE (1.2% vs. 1.8%). Conclusion The initial 90‐day window after ACS represents a vulnerable period for recurrent events. From a trial design perspective, determining a clinically important benefit by RRR versus ARR may influence the decision between early and late MACE as the study endpoint.

cular events (MACE) from recent landmark cardiovascular trials to better characterize the time course of the residual risk among patients who receive standard-of-care therapies following ACS. Furthermore, the incidence of early or late recurrent MACE events and its potential implications with respect to statistical power has never been quantitatively assessed. Accordingly, we also aimed to evaluate the relative statistical power of a primary MACE endpoint at these timepoints under a variety of assumptions. From a trial design standpoint, if most of the recurrent MACE events occur in the early post-ACS phase, a question naturally arises as to whether setting the primary endpoint at the early phase would yield a greater statistical power compared to setting the primary endpoint at the late phase when a consistent long-term treatment effect cannot be ascertained. To investigate this question, we compared the statistical power of a primary endpoint at 90 versus 360 days based on the event rates derived from the contemporary phase III randomized controlled trials.

| METHODS
A comprehensive literature search was performed in ClinicalTrials. gov to collect completed phase III interventional trials from inception to June 2019 targeting the study population of high-risk ACS patients. Full-text publications of the collected studies were retrieved for screening. Studies fulfilling the following criteria were included: (1) study participants had a recent ACS (up to 90 days from enrollment) and at least one additional cardiovascular risk factor (age ≥ 60 years, diabetes mellitus, multivessel coronary artery disease, polyvascular disease, prior history of myocardial infarction [MI], stroke, or coronary revascularization); (2) study participants received a background of optimal medical therapy including antiplatelet agents (aspirin or P2Y 12 inhibitor), angiotensin-converting enzyme inhibitor (ACEI), angiotensin receptor blocker (ARB), beta-blocker (BB), and statin; (3) the duration of follow-up was at least 12 months; and (4) the primary efficacy endpoint was MACE, defined as a composite of MI, stroke, or cardiovascular death. Two independent investigators searched and screened the qualified clinical trials, and disagreement was resolved by discussion and consensus. A flow chart depicting the selection of these trials is provided in Figure S1.  indicating a low, moderate, and high degree of heterogeneity, respectively). A sensitivity analysis using the leave-one-out approach was performed to test the robustness of estimated event rates at 90 and at 360 days. An independent investigator evaluated the risk of bias of included studies by the Cochrane Collaboration's tool. 5 The assessment of the risk of bias was validated by a second investigator.
To provide insights on trial design, the study assessed the statistical power of setting the primary endpoint as the early MACE versus late MACE based on estimated event rates pooled from these contemporary phase III trials. Under the assumption of a total sample size of 10,000 and 1:1 allocation at a one-sided alpha of 0.025 using the log-rank test, two separate approaches were applied to compare the statistical power associated with early MACE (90 days) versus late MACE (360 days) as the primary efficacy endpoint when prospectively designing an interventional trial. In the first approach, the relationship between power and hazard ratio (ranging from 0.99 to 0.70, corresponding to a relative risk reduction [RRR] from 1% to 30%) was explored. In the second approach, the relationship between power and absolute risk reduction (ARR; ranging from 0.1% to 3.0%) was explored.

| Summary of study characteristics
Seven trials, representing a total of 82,727 high-risk patients with recent ACS, were analyzed (Table S1). 6

| Absolute risk and relative percentage of early versus late MACE
The absolute risk of 90-and 360-day MACE ranged from 2.7% to 8.5% and 6.2% to 11.7%, respectively (Table S1 and  In the sensitivity analysis, the 90-day risk (ranging from 3.7% to 4.4%; Figure S3) and 360-day risk (ranging from 8.0% to 8.7%; Figure S4) were similar to the estimates from primary analysis.

| Statistical power of early versus late MACE as the primary endpoint
Based on pooled results from contemporary landmark trials, event rates of 4.1% at 90 days and 8.3% at 360 days were assumed in the comparator arm. With a total sample size of 10,000 and 1:1 allocation at a one-sided alpha of 0.025 using the log-rank test, two separate approaches were applied to compute the statistical power for early versus late MACE as the primary endpoint when prospectively designing an interventional trial.

| Relationship between power and HR
The relationship between power and hazard ratio (ranging from 0.99 to 0.70, corresponding to a RRR from 1% to 30%) was evaluated and is illustrated in Figure 4. When a predetermined HR was employed in the power calculation, greater statistical power was consistently achieved when late MACE was selected as the primary endpoint compared with early MACE. Specifically, with a hazard ratio (HR) of 0.80 (i.e., RRR of 20%) in favor of the intervention, the power would be 86% for the late MACE (corresponding to an event rate of 8.3% in the comparator arm vs. 6

| Relationship between power and ARR
The relationship between power and ARR (ranging from 0.1% to 3.0%) was also evaluated and is illustrated in Figure 5. When a predetermined ARR was employed in the power calculation, greater statistical power was consistently achieved when early MACE was selected as the primary endpoint compared with late MACE. Specifically, with an ARR of 1% in favor of the intervention, the power would be 77% for early MACE (corresponding to an event rate of 4.1% in the comparator arm vs. 3  The nonlinear trajectory of residual risk was also evident in the meta-analysis of statin trials from the Cholesterol Treatment Trialists' (CTT) Collaboration. 13 Specifically, the annual risk of major vascular events was greater in the first year (2.2% in the statin arm vs. 2.5% in the control arm) than in the subsequent 4 years (1.5%-1.6% in the statin arm vs. 1.8%-2.0% in the control arm). More importantly, the magnitude of RRR by statins was less pronounced in the first year (12%) than in the subsequent 4 years (22%-28%). The unattenuated early residual risk points to the unmet medical need for interventions that target atherothrombotic pathways unrelated to platelet activation or LDL-C (e.g., triglycerides, very-low-density lipoproteins [VLDL], and HDL-C), which have been extensively investigated in the past three decades. For instance, among statin-treated patients from the PROVE IT-TIMI 22 trial, a low triglyceride level was associated with a reduced cardiovascular risk independent of LDL-C levels. 14 Similarly, pharmacological reductions in the small, triglyceridedepleted VLDL were associated with a decreased atherosclerotic risk among participants in the JUPITER trial who had low LDL-C levels. 15 With respect to HDL-C, its inverse relationship with cardiovascular disease was supported by a meta-analysis of 20 statin trials with adjustment for statin potency and LDL-C levels. 16 17 Furthermore, Mendelian randomization studies did not demonstrate a causal relationship between genetically-altered plasma HDL-C levels and cardiovascular risk. [18][19][20][21][22] More research is warranted to unravel the underlying mechanisms and optimal preventive strategies for modifying the residual risk.
The first 90 days following an ACS episode marks not only the vulnerable window in which half of the MACE events occur, but also well as the importance of applying alternative approaches to assessing the treatment effect, such as restricted mean survival time. 23 The distinction of early residual risk has practical implications on trial design. If a novel intervention is hypothesized to primarily reduce early residual risk but not late residual risk (in which Kaplan-Meier curves separate before 90 days and stay parallel through 360 days), then the early MACE endpoint should be employed for sample size determination, because the early vulnerable period represents a more appropriate window for examining the mechanistic effect of such intervention. In this scenario, the proportionality of hazards assumption may be violated when the primary endpoint is set at 360 days. Conversely, if a novel intervention is hypothesized to reduce both early and residual risk with a constant HR over time, then the late MACE endpoint should be employed for sample size determination, as it accounts for both early and late events. Additional considerations should be made regarding the statistical power. This study demonstrates that the early MACE endpoint is associated with a greater power when ARR is employed to gauge the treatment effect, whereas the late MACE endpoint is associated with a greater power when RRR is employed to gauge the treatment effect. Therefore, a clinically important reduction in ARR or RRR should be predetermined, as the decision affects the statistical power and selection between early versus late MACE endpoint.

| LIMITATIONS
The cumulative incidence from the present analysis was estimated indirectly from the KM curves provided in the original publications.

| CONCLUSION
The initial 90-day window after an ACS event represents a vulnerable period for recurrent cardiovascular events, possibly attributable to atherothrombosis mechanisms not sufficiently attended by current optimal medical therapy. When designing an interventional trial, a clinically important reduction in absolute risk versus relative risk should be predetermined, as the decision affects the selection between early MACE versus late MACE as the primary endpoint.