Cost, context, and confounding: rethinking real-world data generation and interpretation in advanced NSCLC

As immunotherapy becomes entrenched in the frontline management of advanced NSCLC, post-progression treatment decisions face increasing uncertainty. A recent JAMA Network Open study using Flatiron Health data suggested that rechallenging with pembrolizumab may offer survival benefit after chemo immunotherapy progression. However, these findings raise fundamental questions about how real-world data (RWD) should be interpreted and operationalized in health economics.

In this viewpoint, we critically analyze the methodological and interpretive assumptions behind retrospective RWD-based cost-effectiveness modeling. We argue that unadjusted survival metrics absent stratification for PD-L1 status, tumor burden, or clinical performance may artificially elevate ICER/QALY estimates and misguide reimbursement decisions. Using the Velcheti et al. study as a case anchor, we highlight how structural biases in RWD can lead to premature or inflated conclusions about the value of immunotherapy rechallenge. We also propose a visual framework to distinguish between plausible clinical signals and policy-grade evidence.

Health systems and stakeholders must apply greater rigor when translating RWD into economic models. We advocate for analytic transparency, pre-registered sensitivity analyses, and adjustment for patient-level variables before using RWD to inform value-based oncology.

RWD are powerful hypothesis generators, not substitutes for controlled evidence. Without methodological discipline, their policy application risks becoming speculative rather than strategic.