Impact of the preoperative modified Glasgow Prognostic Score on disease outcome after radical cystectomy for urothelial carcinoma of the bladder.

To investigate the predictive and prognostic value of the preoperative modified Glasgow Prognostic Score (mGPS) in patients with urothelial carcinoma of the bladder (UCB) treated with radical cystectomy (RC). We conducted a retrospective analysis of an established multicenter database consisting of 4335 patients who were treated with RC±adjuvant chemotherapy for UCB between 1979 and 2012. The mGPS of each patient was calculated on the basis of preoperative serum C-reactive protein and albumin. Uni- and multivariable logistic and Cox regression analyses were performed. The discriminatory ability of the models was assessed by calculating the area under receiver operating characteristics curves (AUC) and concordance-indices (C-Index). The additional clinical net-benefit was assessed using the decision curve analysis (DCA). A mGPS of 0, 1, and 2 was observed in 3,158 (72.8%), 1,020 (23.5%), and 157 (3.6%) patients, respectively. On multivariable logistic regression analyses, mGPS of 1 or 2 were associated with an increased risk of pT3/4 disease at RC (OR 1.25, P=0.004 and OR 2.58, SP<0.001, respectively) and/or lymph node metastasis (OR 1.7, P<0.001 and OR 3.9, P<0.001, respectively). Addition of the mGPS to a predictive model based on preoperatively available variables improved its accuracy for prediction of lymph node metastasis (change of AUC +3.7%, P<0.001). On multivariable Cox regression analyses, mGPS of 1 or 2 remained associated with worse recurrence-free survival (HR 1.14, P=0.03 and HR 1.89 P<0.001, respectively), cancer-specific survival (HR 1.16, P=0.032 and HR 2.1, P<0.001, respectively) and overall survival (HR 1.5, P=0.007 and HR 1.92 P<0.001, respectively) compared to mGPS of 0. The additional discriminatory ability of the mGPS for prognosis of survival outcomes in separate models that included either established pre- or postoperative variables did not improve the C-Index by a prognostically relevant degree (change of C-Index <2% for all models). On DCA, the inclusion of the mGPS did not meaningfully improve the net-benefit for clinical decision-making regarding survival outcomes. We confirmed that an elevated mGPS is an independent risk factor for non-organ confined disease and poor survival outcomes in patients with UCB undergoing RC. However, the mGPS showed little value in improving the discriminatory ability of predictive and prognostic models that relied on either pre- or postoperative clinicopathological variables. The discriminatory ability of this biomarker in the age of immunotherapy warrants further evaluation.