Competing risks and cancer-specific mortality: why it matters

In studies with multiple endpoints, such as cancerspecific mortality and noncancer-specific mortality, conventional statistical approaches evaluate these endpoints using a separate Kaplan-Meier analysis without considering whether these endpoints are competing events for each other. In our recent study, we examined the influence of increasing age on lung cancer-specific and noncancerspecific mortality among patients with early-stage, nonsmall cell lung cancer (NSCLC) using a competing risk analysis [1]. We found that both the 5-year lung cancerspecific and noncancer-specific cumulative incidence of death increase with age. We also reported a higher incidence of noncancer-specific mortality compared with lung cancer-specific mortality among patients who were ≥75 years of age that lasted up to 2 years postresection. These findings highlight the necessity of accounting for noncancer-specific mortality as a competing event when assessing cancer-specific mortality in elderly patients. Our article sheds light on two related concepts—competing risks and the statistical challenges that arise from competing risks in the elderly patient population. Clinical research that involves time-to-event endpoints, such as all-cause mortality, conventionally uses the Kaplan-Meier approach that includes censoring patients at the end of the follow-up period. If the event of interest is cancer-specific mortality, a patient may die due to causes unrelated to the disease; such events are termed “competing risk events.” Clinical research also may focus on non-mortality related outcomes such as incidence of disease recurrence, second primary cancer, and treatment success. In these incidences, death without observation of the event of interest will be considered a competing risk event. Competing risk events may substantially alter the probability of occurrence of the event of interest or even preclude its onset. Statistical methods have been developed to assess time-to-event outcomes in the presence of competing risks (competing risk approaches). With a growing elderly population, the recognition and proper examination of competing risks within the intersection of geriatric and oncologic research is more important than ever. A higher proportion of elderly patients consequently increases the incidence of diseases that are attributable to aging and frailty, thus making the cohort of elderly patients highly susceptible to competing risk events. A review of 50 clinical studies published in high-impact journals that were focused on the aging or multimorbid population found competing risk issues in 70% of those studies [2]. The conventional Kaplan-Meier framework assumes that censoring is uninformative since, without competing risk events, the estimates from both the naïve KaplanLetter to the Editor