The SPARC Score: A Multifactorial Outcome Prediction Model for Patients Undergoing Radical Cystectomy for Bladder Cancer

While multiple independent clinicopathological variables are associated with the outcome of radical cystectomy for bladder cancer, limited prediction tools exist to facilitate individualized risk assessment. We developed the SPARC (Survival Prediction After Radical Cystectomy) score, a prediction model for bladder cancer specific survival after radical cystectomy.We evaluated 2,403 patients who underwent radical cystectomy without neoadjuvant therapy at our institution between 1980 and 2008 with pathological re-review of all specimens. Of these patients 1,776 with nonmetastatic urothelial carcinoma were identified for analysis. A multivariate model was developed using stepwise selection to determine variables associated with cancer specific survival. We created a scoring system based on the β coefficients of this model.Median followup after radical cystectomy in patients alive at last followup was 10.5 years (IQR 7.3, 15.3), during which time 610 had died of bladder cancer. In addition to pathological tumor stage, nodal status, multifocality and lymphovascular invasion, the patient specific factors of Charlson comorbidity index, Eastern Cooperative Oncology Group (ECOG) performance status, current smoking, preoperative hydronephrosis and receipt of adjuvant chemotherapy were significantly associated with the risk of bladder cancer death. We used cumulative scores of these variables to stratify patients into risk groups with 95%, 80%, 60%, 38% and 23% 5-year cancer specific survival from the lowest to the highest risk group, respectively (p<0.0001). The concordance index of this model was 0.75.We present a model to individualize the estimation of cancer specific survival after radical cystectomy. Pending external validation, these data may be used for patient counseling, specifically in regard to recommendations for adjuvant therapy and surveillance frequency, as well as for clinical trial development.