Title:Time-to-Event Modeling with Pseudo-Observations in Federated Settings

Abstract:In multi-center clinical studies, concerns about patient privacy often prohibit pooling individual-level time-to-event data. We propose a non-iterative, one-shot federated framework using distributed pseudo-observations, derived from a sequentially updated Kaplan-Meier estimator and fitted with renewable generalized linear models. This framework enables the estimation of survival probabilities at specified landmark times and accommodates both time-invariant and time-varying covariate effects. To capture site-level heterogeneity, we introduce a soft-thresholding debiasing procedure that adaptively shrinks local estimates toward the global fit. Through extensive simulations across varying event rates and site-size distributions, our method demonstrates performance comparable to pooled Cox and the one-shot Optimal Distributed Aggregation (ODAC) models, with added flexibility to capture non-proportional hazards. Applied to pediatric obesity data from the Chicago Area Patient-Centered Outcomes Research Network (CAPriCORN), which comprises four different sites and includes a total of 45,865 patients. The federated pseudo value regression model produced estimates of both time-constant and time-varying hazard ratios that closely aligned with those obtained from the pooled analysis, demonstrating its utility as a robust and privacy-preserving alternative for collaborative survival research. To further address potential heterogeneity across sites, we applied a covariate-wise debiasing algorithm, enabling site-level adjustments while preserving consistency with the global model.