Time dependent outcomes modeling in a real-world analysis of the molecular tumor board at University Cancer Center Hamburg (2016-2022)

INTRODUCTION: Molecular tumor boards (MTBs) integrate clinical, pathological, and molecular data to prioritize therapy options. Because MTB-guided treatment starts at varying time points after MTB discussion, vulnerabilities to immortal time bias occur if treatment delay is ignored. METHODS: Retrospective survival analysis of 949 consecutive individuals discussed at the University Cancer Center Hamburg MTB (2016-2022). Index date (t0) was the first MTB discussion. Therapy initiation was modeled as a time-dependent exposure using Mantel-Byar methods and visualized with Simon-Makuch plots. Therapy strata were priority 1 targeted, priority 2 targeted, and non-targeted therapy. RESULTS: Among solid tumors (n = 854), targeted therapy was implemented in 24% and was off-label in 51%. ORR/DCR were 30.8%/61% for priority 1, 11.4%/36.4% for priority 2, and 37.5%/52.1% for non-targeted therapy (P = .017). The progression-free survival ratio (MTB-directed therapy vs immediately preceding line) exceeded 1.3 in 63% of targeted therapy recipients. From the MTB date, median survivalMTB was 16.5 months (priority 1), 15.5 months (priority 2), and 11.0 months (non-targeted). In Mantel-Byar analysis, progression-free survivalSimonMakuch favored priority 1 (P = .002) with 1-year rates of 28%, 12%, and 15% for priority 1, priority 2, and non-targeted therapy, respectively. One-year overall survivalSimonMakuch was 52%, 34%, and 42%. In a time-dependent Cox model, priority 1 was associated with lower risk of death versus no targeted post-MTB therapy (hazard ratio 0.80, 95% confidence interval 0.64-0.99, P = .038). CONCLUSIONS: After correction for immortal time bias, implementation of the top-priority MTB recommendation was associated with improved disease control and survival signals, although confounding and selection effects remain key limitations.