Prognostic value of existing scoring systems for early mortality in phase I immunotherapy trials: a head-to-head real-world cohort study.

BACKGROUND: Patients enrolled in phase I oncology clinical trials are expected to survive more than 3 months, yet up to 15% may die before this threshold. Several prognostic scores have been developed to identify patients unlikely to derive benefit from trial participation. We aimed to compare the prognostic performance of five described scores. METHODS: This retrospective study included 844 patients treated with immunotherapy in phase I trials at Clínica Universidad de Navarra (2016-2023). Prognostic scores (RMH, GRIm, LIPI, PIPO, ROPRO) were calculated according to their original definitions. Discriminative performance was assessed using Harrell's C-index, 3-month AUC, and time-dependent AUC. RESULTS: Median OS was 9.08 months. All scores demonstrated statistically significant prognostic stratification. C-index values were: RMH 0.58, GRIm 0.55, LIPI 0.66, PIPO 0.63, and ROPRO 0.69. AUC at 3 months was highest for ROPRO (0.76). However, no score identified a subgroup with a median OS below 3 months. CONCLUSIONS: All five scores stratified prognosis but failed to select patients with OS below 3 months. These findings challenge their use as rigid exclusion criteria and support their role as stratification factors. New models integrating dynamic clinical or molecular biomarkers are needed to improve patient selection in phase I immunotherapy trials.