Non-Proportional Hazards in Network Meta-Analysis: A Five-Step Governance Protocol for Time-to-Event HTA Submissions

The proportional hazards (PH) assumption underpins the standard hazard ratio approach to network meta-analysis (NMA) of time-to-event outcomes. When that assumption fails — as increasingly observed in oncology with immunotherapy and combination regimens — evidence teams must switch to non-PH NMA methods or risk producing biased comparative effectiveness estimates. This playbook sets out a five-step governance protocol covering PH diagnostics, model selection, sensitivity analysis, result...