Personalized Oncology: Feasibility of Evaluating Treatment Effects for Individual Patients

The effectiveness of personalized oncology treatments ultimately depends on whether outcomes can be causally attributed to the treatment. Advances in precision oncology have improved molecular profiling of individuals, and tailored therapies have led to more effective treatments for select patient groups. However, treatment responses still vary among individuals. As cancer is a heterogeneous and dynamic disease with varying treatment outcomes across different molecular types and resistance mechanisms, it requires customized approaches to identify cause-and-effect relationships. N-of-1 trials, or single-subject clinical trials, are designed to evaluate individual treatment effects. Several works have described different causal frameworks to identify treatment effects in N-of-1 trials, yet whether these approaches can be extended to single-cancer patient settings remains unclear. To explore this possibility, a longitudinal dataset from a single metastatic cancer patient with adaptively chosen treatments was considered. The dataset consisted of a detailed treatment plan as well as biomarker and lesion measurements recorded over time. After data processing, a treatment period with sufficient data points to conduct causal inference was selected. Under this setting, a causal framework was applied to define an estimand, identify causal relationships and assumptions, and calculate an individual-specific treatment effect using a time-varying g-formula. Through this application, we illustrate explicitly when and how causal treatment effects can be estimated in single-patient oncology settings. Our findings not only demonstrate the feasibility of applying causal methods in a single-cancer patient setting but also offer a blueprint for using causal methods across a broader spectrum of cancer types in individualized settings.