Validation of a prognostic scoring system for brain metastases with synchronous detection of primary cancers at initial consultation

Yohei Yamamoto¹Gentoku Saito¹Tomona Maetani¹Yurika Terasawa¹Naoki Kato¹Takuya Ishii¹Yasuharu Akasaki²Yuichi Murayama²Toshihide Tanaka^2*

• ¹Jikei University Daisan Hospital, Tokyo, Japan
• ²Jikei University School of Medicine, Minato, Japan

Objective: To develop and validate a simple 0–7 point prognostic score for patients with synchronous brain metastases (s-BM) at initial consultation and to provide an exploratory comparison with metachronous BM (m-BM). Methods: We retrospectively analyzed 297 patients with BM (2014–2022): s-BM n=64 and m-BM n=233. The score uses five pre-treatment factors available at time-zero: age <70 y (1 point), KPS ≥70 (2 points), absence of extracranial metastases (ECM) (2 points), presence of a lung mass on imaging (1 point), and absence of a digestive-tract mass (1 point). Patients were stratified as Score A (0), B (1–4), and C (5–7). Discrimination and parsimony were assessed by Harrell’s C-index and AIC; proportional-hazards (PH) assumptions by global Schoenfeld test; time-dependent AUCs at 6/12 months were calculated; bootstrap (B=500) provided optimism-corrected C-indices. (Synchronous was defined as ≤30 days between first ascertainment of primary tumor and BM; m-BM >30 days.) Results: Median OS appeared similar for s-BM and m-BM in this cohort (5.2 vs 6.4 months; exploratory). In s-BM, KPS ≥70 and absence of ECM were independent predictors of longer survival. The proposed score produced stepwise separation across A/B/C. It showed the highest C-index (0.690) and lowest AIC (375.57) versus RPA, GPA, SIR, and BSBM, with no PH violations (global p=0.983). AUCs were 0.779 (6 mo) and 0.795 (12 mo); bootstrap-corrected C-index ≈0.691. A coefficient-weighted variant yielded a similar C-index (0.694) but a higher AIC (381.9). Conclusions: A five-factor, 1–2 point score usable at initial consultation provides superior discrimination and parsimony in s-BM compared with established systems. External, multicenter validation is warranted.