Beyond Binary Mortality Endpoints

Objective: To evaluate the T-MOD PD (Timing–Mechanism–Opportunity–Disease) framework for classifying mortality after pancreatoduodenectomy (PD) , focusing on preventability and rescue opportunities. Background: Pancreatoduodenectomy mortality (2–10%) is reported as binary endpoints, obscuring distinctions between surgery-related and systemic causes and limiting targeted quality improvement. Methods: Retrospective analysis of 1,727 consecutive PDs (2014–2024) at a high-volume centre. All 58 deaths (3.4%) were independently adjudicated across four axes: Timing (T1–T3), Mechanism (M1–M3), Opportunity (O1–O4), and Disease (D1–D3). Deaths were categorized as surgery-attributable potentially preventable (SAPM), surgery-attributable non-preventable (SANPM), or non-surgery-attributable mortality (NSAM). Results: Striking phenotypic clustering emerged: 38% of deaths (22/58) converged in T1M1O1D1 (early, surgery-related, strategically modifiable, resectable disease), with 19% (11/58) in T2M1O1D1. Together, these postoperative pancreatic fistula (POPF)-driven phenotypes accounted for 57% of mortality. Domain analysis revealed 66% early deaths (T1), 74% surgery-related mechanisms (M1), and critically, 72% potentially modifiable opportunities (O1: 72%; O2: 22%; O3: 2%; O4: 3%). Overall, 74% were classified as SAPM. Timeline reconstruction identified median 72-hour delays in recognition and escalation despite warning signs at postoperative day 5–6, with 22-day median interval from clinically relevant POPF diagnosis to death, suggesting 35–40% preventability in the dominant phenotype. Perfect interobserver agreement was achieved (κ=1.0). Conclusions: T-MOD PD provides reproducible mortality phenotyping revealing 72% of deaths are potentially modifiable, with dominant phenotypes sharing POPF pathways and identifiable rescue delays, enabling phenotype-specific quality improvement.