To develop and validate nomograms for pre-treatment prediction of malignant histology (MH) and unfavorable pathology (UP) in patients with endophytic renal tumors (ERTs).
We retrospectively reviewed the clinical information of 3245 patients with ERTs accepted surgical treatment in our center. Eventually, 333 eligible patients were included and randomly enrolled into training and testing sets in a ratio of 7:3. We performed univariable and multivariable logistic regression analyses to determine the independent risk factors of MH and UP in the training set and developed the pathological diagnostic models of MH and UP. The optimal model was used to construct a nomogram for MH and UP. The area under the receiver operating characteristics (ROC) curves (AUC), calibration curves and decision curve analyses (DCA) were used to evaluate the predictive performance of models.
Overall, 172 patients with MH and 50 patients with UP were enrolled in the training set; and 74 patients with MH and 21 patients with UP were enrolled in the validation set. Sex, neutrophil-to-lymphocyte ratio (NLR), R score, N score and R.E.N.A.L. score were the independent predictors of MH; and BMI, NLR, tumor size and R score were the independent predictors of UP. Single-variable and multiple-variable models were constructed based on these independent predictors. Among these predictive models, the malignant histology-risk nomogram consisted of sex, NLR, R score and N score and the unfavorable pathology-risk nomogram consisted of BMI, NLR and R score performed an optimal predictive performance, which reflected in the highest AUC (0.842 and 0.808, respectively), the favorable calibration curves and the best clinical net benefit. In addition, if demographic characteristics and laboratory tests were excluded from the nomograms, only the components of the R.E.N.A.L. Nephrometry Score system were included to predict MH and UP, the AUC decreased to 0.781 and 0.660, respectively (P=0.001 and 0.013, respectively).
In our study, the pathological diagnostic models for predicting malignant and aggressive histological features for patients with ERTs showed outstanding predictive performance and convenience. The use of the models can greatly assist urologists in individualizing the management of their patients.