BMC Cancer · Aug 7, 2026

Development and internal validation of the ABC score for overall survival prediction in cancer patients aged 80 years and older

By Nascimento GHB, de Alcantara Sousa LV, Cubero DIG et al.

BACKGROUND: Individuals aged over 80 years (the "oldest-old") represent the fastest-growing demographic in clinical oncology. While the Age-Adjusted Charlson Comorbidity Index (ACCI) is a traditional benchmark for multimorbidity, its predictive accuracy in the ultra-elderly may be limited by competing risks of death and the physiological burden of active malignancy. OBJECTIVE: To evaluate the prognostic utility of the ACCI and to develop and internally validate a streamlined stratification tool, the ABC Score, for octogenarian cancer patients. METHODS: We conducted a retrospective cohort study of 203 patients aged ≥ 80 years, first evaluated between 2022 and 2024 at a Brazilian public institution. Overall survival (OS) was analysed using Cox proportional-hazards models. Variables significant at p < 0.20 in univariate screening were entered into a multivariable model with backward elimination to derive a points-based score from the β coefficients. Discrimination was assessed by Harrell's C-index and internally validated by bootstrap resampling with optimism correction; calibration was assessed at 6, 12, and 24 months. RESULTS: Over a median follow-up of 30.5 months, 80 deaths (39.4%) occurred among 203 patients. The ACCI was significant in univariate analysis (HR = 1.28, p < 0.001) but lost independent prognostic value after adjustment (p = 0.30). Independent predictors were ECOG performance status, clinical stage, and chronic pulmonary disease. The ABC Score had an apparent C-index of 0.778 and an optimism-corrected C-index of 0.771 (95% CI 0.721-0.825) after bootstrap validation, with excellent calibration at 6, 12, and 24 months. Three risk tiers stratified survival (log-rank p < 0.001). CONCLUSION: The ABC Score is a promising, pragmatic exploratory tool that prioritises functional and respiratory reserve over aggregate comorbidity counts; it requires external, multicentre validation before clinical use.

doi.org/10.1186/s12885-026-16627-6Open accessPubMed 42983708Biomarkers