No Incremental Value of mGPS Beyond CGA for Six-Month Survival in Older Advanced Cancer Patients

Introduction: Accurate prognostication is essential to guide treatment decisions and care planning in older patients with cancer. The modified Glasgow Prognostic Score (mGPS), an inflammation-based biomarker combining C-reactive protein and albumin, is recommended by the European Society for Medical Oncology for prognostic assessment. However, its added value beyond comprehensive geriatric assessment (CGA) in vulnerable older patients remains unclear. This study evaluated whether mGPS improves six-month survival prediction.

Methods: In this retrospective study, patients aged ≥70 years with stage IV cancer were identified from registries at the Kortrijk Cancer Centre (Belgium) from November 2012 onwards. All patients underwent Geriatric 8 (G8) screening; vulnerable patients (G8 ≤14/17) received a CGA. Prognostic models were constructed using variables identified in univariable survival analyses and retained as independent predictors in multivariable Cox models on a complete-case cohort. The incremental prognostic value of mGPS was evaluated by comparing AUROC between models with and without mGPS using DeLong’s test.

Results: In vulnerable stage IV patients (n = 308), the baseline CGA model, based on complete-case multivariable analysis (n = 129), demonstrated moderate discrimination for six-month survival (AUROC = 0.661). Addition of mGPS did not significantly improve prognostic performance (AUROC = 0.661 vs 0.662; Δ = +0.001; DeLong p = 0.8851). Similar findings were observed at 3-, 9-, and 12-month time points (AUROC base vs +mGPS: 0.672 vs 0.703, 0.717 vs 0.712, and 0.726 vs 0.724, respectively; all DeLong p > 0.05). These findings were consistent across secondary and exploratory analyses, including models incorporating ECOG and G8, and in non-vulnerable and haematological cohorts.

Discussion: mGPS did not demonstrate meaningful incremental predictive value beyond the CGA, as reflected by minimal and non-significant improvements in AUROC. This may suggest that the prognostic information captured by mGPS largely overlaps with CGA-derived prognostic variables and is already encompassed within prior G8-based stratification.