Geriatric assessment in hematology (GAH) scale for survival prognostication and early treatment discontinuation in older adults: Final results from the spanish hematogeriatric registry

Abstract Introduction: Comprehensive geriatric assessment (CGA) is the “gold standard” for determining frailty in older patients with hematological malignancies (HM), although its implementation may be challenging due to limited resources. The Geriatric Assessment in Hematology (GAH) scale has emerged as a simplified tool to assess health status in this population. We aimed to evaluate the prognostic impact of the GAH scale and to explore its correlation with the Cumulative Illnes Rating Scale (CIRS) comorbidity index and performance status (PS). Patients and methods: This multicenter, prospective, observational study included 285 patients aged ≥70 years with HM, recruited between 2018 and 2022, with a data cut-off in May 2025. Baseline GAH, ECOG PS, and CIRS scores were collected. The GAH score was calculated using the domain weights proposed in the RETROGAH study (range: 0-94 points). Survival analyses were performed in patients with complete follow-up data using the Kaplan-Meier method and log-rank test. Optimal GAH cut-offs for OS and early treatment discontinuation were determined using maximally selected rank statistics and ROC curve analysis with Youden's index. Results: The median age of patients was 79 years (IQR 74-83), and 53.3% were male. The most frequent diagnoses were non-Hodgkin lymphomas (36.8%) and myeloid neoplasms (MDS/AML, 29.8%) followed by multiple myeloma (MM) (18.9%), chronic lymphoproliferative disorders (7.7%), Hodgkin lymphomas (3.5%), and others (3.3%). While a majority of patients had a good PS (76.8% ECOG 0-1), the cohort was frail (median GAH 57, IQR 37-75) and comorbid (52% CIRS >6). After a median follow-up of 58.0 months (95% CI: 52.0-63.0), overall mortality was 64.8%. Baseline GAH score was a strong predictor of OS. A sequential optimal cut-off analysis identified thresholds of 22 and 81, stratifying the cohort into three distinct prognostic groups: fit (GAH ≤22; n=34), prefrail (GAH 23-81; n=184), and frail (GAH >81; n=26). This classification demonstrated a significant separation in OS (p=0.00085). Alternative cut-offs (17 and 45), previously derived from ROC analysis in this dataset, retained prognostic value (p=0.012). ECOG (0-1 vs ≥2; p=0.0002) and CIRS (>6 vs ≤6; p=0.016) were also significant OS predictors. However, the GAH scale appeared more effective in identifying hidden vulnerability. Among patients considered low-risk by ECOG, the GAH scale reclassified 77% into prefrail and 6.6% into frail categories. Similarly, 3.1% and 78% of patients with a low comorbidity burden were identified as being frail and prefrail, respectively by the GAH. Notably, 100% of GAH-defined fit patients had an ECOG PS of 0-1. GAH scores were not associated with chronological age (p=0.4) or type of HM (p=0.09). Finally, patients who discontinued therapy within the first 6 months had a higher median baseline GAH score compared to those who did not (64.0 vs 57.0, p=0.047), although, ROC curve analysis showed limited discriminatory capacity for this endpoint (Youden's Index 0.175). Conclusion: The GAH scale is a valuable prognostic tool for long-term survival in oler patients with HM. It provides added value in identifying frailty that might be underestimated by ECOG or CIRS. Our findings support integrating the GAH scale into routine clinical practice for more accurate risk stratification, especially in circumstances where CGA cannot be performed.