Early prediction of prognosis in advanced solid tumor patients using tumor growth rates with <i>g</i> score in early phase clinical trials.

Abstract

3149 Background: The primary objective of early phase clinical trials is to evaluate the safety of investigational drugs, which requires participants to have sufficient expected survival durations. Tumor growth rate using g scores, calculated using radiographic measurements and timing after treatment, is gaining attention as a potential tool for treatment efficacy assessment. This study aims to assess the utility of g scores before and after treatment in predicting prognosis, and explore suitable trial candidates for accelerating drug development in early phase clinical trials. Methods: We retrospectively reviewed patients who participated in early phase clinical trials after standard treatment at the Department of Advanced Medical Development, The Cancer Institute Hospital of Japanese Foundation for Cancer Research between January 2020 to December 2023. A mathematical exponential growth model was applied to estimate tumor growth rates ( g ) based on radiographic tumor measurements and interval time: f( t ) = exp ( g ⋅ t) , with pre- g scores derived from measurements before the clinical trial and post- g scores from measurements after trial initiation. Pre- g scores were calculated using trial baseline computed tomography (CT)s and the most recent CTs before trial enrollment, while post- g scores were calculated using baseline CTs and the first evaluated CTs after treatment. We defined dichotomized g score levels (high/low) using the time-dependent ROC curve procedure. We evaluated independent predictors for survival outcomes according to each g score and patient characteristics. Results: Of the 173 cases who participated in early phase clinical trials after standard treatment, 162 cases with evaluable CT scans before and after the clinical trial were included in this study. Median time to pre-trial CT was 29 days (range, 5–202), and median time to first post-treatment evaluation was 49.5 days (range, 16–87). Log-rank testing showed both high pre-. and post-. scores correlated to shorter overall survival (OS) compared to low-score groups (HR 2.16; 95% CI 1.22‐3.81; P = 0.0067, HR 2.68; 95%CI 1.84-3.90; P &lt; 0.001). Multivariate analysis showed both high pre- g and post- g scores were independent predictors of shorter OS (HR 2.06, 95% CI 1.13-3.75; P = 0.019, HR 3.80, 95% CI 2.44-5.90; P &lt; 0.001). Conclusions: This study is the first to incorporate pre-. scores as an independent prognostic factor and may serve as a valuable reference for patient enrollment in early phase clinical trials under late-line settings. Additionally, post-. scores were also identified as an independent prognostic factor across multiple cancer types in early phase clinical trials. These results indicate their potential use as surrogate endpoints to, which may help facilitate drug development.