Related Experiment Video
Updated: Aug 25, 2026

Constructing and Visualizing Models using Mime-based Machine-learning Framework
Published on: July 22, 2025
Myeloma Elderly Prognostic Score (MEPS): A Proposed Prognostic Score Integrating Age, Renal Function, Performance
Eirini Katodritou1, Bruna Velosa Ferreira2, Efstathios Kastritis3
1Hematology-Oncology, Theagenio Cancer Hospital, Thessaloniki, Greece.
None:
Established disease-centered prognostic systems may be less tailored to elderly/non-transplant-eligible (NTE) patients with multiple myeloma (MM), in whom host-related factors such as age, performance status, and renal function strongly influence all-cause outcomes. We developed and evaluated the Myeloma Elderly Prognostic Score (MEPS), a pragmatic four-factor score integrating age ≥ 75 years, renal function < 40 by eGFR or available creatinine clearance (CrCl), ECOG performance status ≥ 2, and ultra-high-risk cytogenetics, using baseline registry data from 881 newly diagnosed NTE MM patients from Greece, with OS analyses performed in the 878 patients with evaluable survival follow-up, and externally evaluated it in an independent Portuguese cohort. In the Greek OS-evaluable cohort, 485 deaths were observed. Median OS by MEPS group was 96, 65, 39, and 15.3 months for scores 0, 1, 2, and ≥ 3, respectively. In a baseline score-level Cox model, each ordinal MEPS category increase was associated with inferior OS (HR 1.74, 95% CI 1.58-1.92; p < 0.001). MEPS showed higher OS discrimination than R2-ISS in the Greek cohort (C-index 0.655 vs. 0.602; bootstrap delta 0.053, 95% CI 0.021-0.085). The Portuguese validation dataset included 300 patients, of whom 290 were evaluable for the primary OS analysis, with 151 deaths. Median OS was 59.6 months (95% CI 52.4-75.7), and MEPS again showed higher discrimination than R2-ISS (C-index 0.729 vs. 0.647; bootstrap delta 0.082, 95% CI 0.041-0.127). MEPS provides a simple all-cause prognostic tool for elderly/NTE MM, complementing biologically oriented staging with routinely available host-related variables.