Cardiovascular outcome trials commonly face competing risks when non-CV death prevents observation of major adverse cardiovascular events (MACE). While Cox proportional hazards models treat competing events as independent censoring, Fine-Gray subdistribution hazard models explicitly handle competing risks, targeting different estimands. This simulation study using bivariate copula models systematically varies competing event rates (0.5%-5% annually), treatment effects on competing events (50% reduction to 50% increase), and correlation structures to compare these approaches. At competing event rates typical of CV outcome trials (~1% annually), Cox and Fine-Gray produce nearly identical hazard ratio estimates regardless of correlation strength or treatment effect direction. Substantial divergence occurs only with high competing rates and directionally discordant treatment effects, though neither estimator provides unbiased estimates of true marginal hazard ratios under these conditions. In typical CV trial settings with low competing event rates, Cox models remain appropriate for primary analysis due to superior interpretability. Pre-specified Cox models should not be abandoned for competing risk methods. Importantly, Fine-Gray models do not constitute proper sensitivity analyses to Cox models per ICH E9(R1), as they target different estimands rather than testing assumptions. As supplementary analysis, cumulative incidence using Aalen-Johansen estimator can provide transparency about competing risk impact. Under high competing-risk scenarios, alternative approaches such as inverse probability of censoring weighting, multiple imputation, or inclusion of all-cause mortality in primary endpoints warrant consideration.
Cardiovascular outcome trials (CVOTs) evaluate treatment effects on composite endpoints that typically include major adverse cardiovascular events (MACE-3: CV death, non-fatal myocardial infarction, or non-fatal stroke). When patients die from non-CV causes before experiencing the primary endpoint, these competing events prevent observation of the primary outcome and create a competing risk scenario requiring careful analytical consideration. In most CVOTs, non-CV death serves as the primary competing risk.
The standard analytical approach in most reported CVOTs uses Cox proportional hazards models (Cox, 1972) that treat non-CV death as independent censoring. Recent large-scale trials with semaglutide illustrate this approach: the SELECT trial evaluates MACE in patients with preexisting CV disease and overweight or obesity but without diabetes (Lincoff et al., 2023), while the FLOW trial assesses kidney outcomes in patients with type 2 diabetes and chronic kidney disease (Perkovic et al., 2024). However, the independent censoring assumption may be inappropriate when competing events are informatively related to the primary endpoint-for instance, when both CV and non-CV deaths share common risk factors such as age, comorbidity burden, or disease severity.
Journal editors and reviewers, particularly at high-impact venues such as The New England Journal of Medicine (NEJM), increasingly scrutinize the handling of competing risks in CV outcome trials. NEJM’s statistical reporting guidelines specifically address this issue in their “Considerations in Time-to-Event Analyses” (New England Journal of Medicine, 2022), stating that “when a substantial proportion of participants experience a competing event that precludes the occurrence of the primary outcome, alternative analytical approaches such as the Fine-Gray model may be considered.” The guidelines emphasize evaluating whether censoring by competing events might be informative. However, critical questions remain: What constitutes a “substantial proportion” of competing events? At what threshold does the choice between Cox and Fine-Gray models affect or alter trial conclusions? How does the dependency between competing events and primary events impact the result? Is Fine-Gray model the best model to handle competing risk? The NEJM guidelines do not provide quantitative benchmarks for these decisions, leaving investigators uncertain about when competing risk methods add value versus unnecessary complexity.
When competing events are present, two analytical approaches are commonly used, each targeting fundamentally different estimands. Table 1 summarizes key distinctions between the Cox proportional hazards model and the Fine-Gray subdistribution hazard model. Despite targeting at different estimands, both approaches are widely used in practice, and understanding when they yield concordant versus divergent numerical results has important practical implications for trial design, analysis planning, and interpretation. A key question is whether the choice between these methods affects trial conclusions in typical CVOT settings.
Competing events are treated as independent censoring
Competing events are explicitly incorporated through weighting in the subdistribution hazard
Proportional cause-specific hazards; independent censoring by competing events
Treatment effect on the instantaneous rate of the primary event among individuals who remain event-free
Treatment effect on the instantaneous rate of the primary event in the subdistribution (including those event-free or having experienced competing events)
The Cox proportional hazards model (Cox, 1972) is the most widely used approach in CVOTs, treating competing events as censored observations. This model estimates the instantaneous risk (hazard) of the primary CV event among patients still at risk. The hazard ratio quantifies the treatment effect on the rate of primary event occurrence, conditional on not yet having experienced either the primary or competing event. This estimand addresses: “Among patients who remain event-free, how does treatment affect the rate of primary CV events?” This interpretation is straightforward and aligns naturally with clinical reasoning about treatment mechanisms.
In the SOUL trial, the primary Cox analysis yields a hazard ratio of 0.86 (95% CI: 0.77-0.96) for MACE-3 (McGuire et al., 2025), with non-CV death occurring at approximately 1.4 per 100 patientyears.
The Fine-Gray subdistribution hazard model (Fine and Gray, 1999) is often recommended for explicitly handling competing risks. This approach models the subdistribution hazard, which relates directly to cumulative incidence functions. The subdistribution hazard ratio quantifies treat-ment effects on the subdistribution hazard-the instantaneous rate of primary event occurrence in the subdistribution, which includes both patients who remain at risk and those who have experienced competing events (with modified weights). Whil
This content is AI-processed based on open access ArXiv data.