Background: The influencing factors on the prognosis of patients with testicular cancer (TC) remains unclear. We aimed to explore the clinical influencing factors of overall survival (OS) and cancer-specific survival (CSS) of Testicular Cancer (TC) patients and to build the nomograms.
Methods: Within the Surveillance, Epidemiology, and End Results (SEER) database, 7718 TC patients diagnosed between 2004 and 2018 were included and their clinical data were assessed. The data were randomly divided into training and validation cohorts in a ratio of 7:3 to construct the prediction models. Prognosis information including 3-, 5-, and 10-year OS and CSS rates were recorded. In this study, the Kaplan-Meier method was used to draw the survival curve and variables with P<0.05 in the univariate Cox proportional hazard regression model were included in the multivariate Cox proportional hazard regression model for analysis. The relative risk of death was described with the hazard ratio (HR) and the 95% confidence interval (CI). The differentiation of the model was evaluated by using ROC and calculating the AUC values at different time points. Based on the data of the prediction set, the Bootstrapping method was adopted, repeated sampling 1000 times, and calibration curves were drawn to evaluate the fit degree of the model. Based on the results of the Cox proportional risk regression model, we constructed and internally validated an applicable nomogram for predicting 3-year, 5-year and 10-year survival in patients with testicular cancer.
Results: A total of 7718 patients were included in the prognostic model. In the multivariate Cox regression model, older age at diagnosis, without surgery, regional lymph node involvement (N1, N2, N3 and NX), with chemotherapy and larger tumor size were independent predictors of poor prognosis factors for OS and CSS in TC patients. The C-indexes were 0.75 (95% confidence interval [CI]: 0.72- 0.78) for OS and 0.82 (95% CI: 0.79- 0.85) for CSS, indicating that the prognosis of testicular cancer patients had a good predictive effect. Based on the nomogram, in the verification group, the AUC values of the 3-, 5-, and 10-year OS were 0.799, 0.757; 0.725, and 0.827, 0.823, and 0.825 for 3, 5, and 10- years CSS, respectively. The differences in the OS and CSS between the real observation and the forecast were quite constant according to the calibration curves. The models were validated in the validation cohort, which also demonstrated the models had good reliability for prognostication.
Conclusions: Clinical factors including age at diagnosis, surgery conditions, N Stage, chemotherapy conditions and tumor size were good predictors for OS and CSS in TC. Our predictive nomogram may play an important role in predicting 3-, 5-, and 10-year outcomes in patients with TC.