Reproducibility package for "Exact tipping points and time-resolved effects of PD-(L)1 plus BCG in high-risk NMIBC"

This repository contains the reproducibility and audit materials for the study “Exact tipping points and time-resolved effects of PD-(L)1 plus BCG in high-risk NMIBC.” The study is a secondary analysis of publicly available phase 3 randomized trial data from POTOMAC, CREST, and ALBAN. Individual patient-level survival data were reconstructed from published Kaplan–Meier curves and numbers at risk. Each trial was analysed separately, retaining its original efficacy endpoint. The repository is intended to provide a transparent and citable record of the reconstructed datasets, analysis specifications, statistical code, formal outputs, and audit materials used in the study. The primary time-resolved analysis framework included a 36-month analysis horizon, restricted mean survival time, interval-specific hazard ratios for 0–12, 12–24, and 24–36 months, landmark absolute treatment effects, and a within-trial later-to-early treatment-effect contrast. Sensitivity analyses included alternative RMST horizons, alternative prior specifications, and a two-period 0–12/12–36-month model. An exact integer early-censoring tipping-point analysis was added post hoc after examination of the primary results. It exhaustively enumerated feasible adverse reclassification of observations censored within 12 months and quantified the minimum patient count required to cross predefined inferential boundaries. This analysis is mechanism-agnostic and should not be interpreted as a bias-corrected estimate. The tipping-point analysis specification, source hashes, enumeration outputs, and boundary verification were frozen on 1 September 2026 before incorporation into the manuscript. This repository was created after completion of the analyses and does not constitute prospective preregistration. The materials explicitly distinguish analyses specified before formal analysis from analyses added post hoc. The reconstructed datasets are derived from publicly available aggregate survival information and do not contain original or identifiable individual patient data. The original trial reports and Kaplan–Meier figures remain available from the respective published sources.