A nomogram model for determining optimal patients for local therapy in metastatic prostate cancer: a SEER database-based study

Abstract

Abstract Background Numerous studies have shown that local therapy can improve long-term survival in patients with metastatic prostate cancer. However, it is unclear which patients are the potential beneficiaries. Methods We obtained information on prostate cancer patients from the Surveillance, Epidemiology, and End Results database and divided eligible patients into the local treatment group and non-local treatment group. Propensity score matching (PSM) was used to reduce the influence of confounding factors. In the matched local treatment (LT) group, if the median overall survival time (OS) was longer than the Nonlocal treatment (NLT) group, it was defined as a benefit group, otherwise, it was a non-benefit group. Then, univariate and multivariate logistic regression were used to screen out predictors associated with benefits, and a nomogram model was constructed based on these factors. The accuracy and clinical value of the models were assessed through calibration plots and decision curve analysis. Results The study enrolled 7255 eligible patients, and after PSM, each component included 1923 patients. After matching, the median OS was still higher in the LT group than in the NLT group [42 (95% confidence interval: 39–45) months vs 40 (95% confidence interval: 38–42) months, p = 0.03]. The independent predictors associated with benefit were age, PSA, Gleason score, T stage, N stage, and M stage. The nomogram model has high accuracy and clinical application value in both the training set (C-index = 0.725) and the validation set (C-index = 0.664). Conclusions The nomogram model we constructed can help clinicians identify patients with potential benefits from LT and formulate a reasonable treatment plan.