Renal function after radical nephrectomy: development and validation of predictive models in Japanese patients.

Objectives To develop and validate predictive models for postoperative estimated glomerular filtration rate and risk of chronic kidney disease after radical nephrectomy in Japanese patients. Methods The present retrospective study included a development cohort of 209 patients without preoperative chronic kidney disease who underwent radical nephrectomy between 1994 and 2008, and were followed up for longer than 3 years, and a validation cohort of 144 similar such patients. Univariate and multivariate linear regression or logistic regression analyses were carried out to identify the independent predictors of estimated glomerular filtration rate or chronic kidney disease 3 years after radical nephrectomy. Incorporating all independent predictors, predictive models for postoperative renal function were developed and externally validated. Results Age, the presence of diabetes mellitus, and preoperative estimated glomerular filtration rate were independent predictors of both postoperative estimated glomerular filtration rate and chronic kidney disease. A formula for predicting the postoperative estimated glomerular filtration rate and a nomogram for predicting the risk of postoperative chronic kidney disease were developed. The adjusted R2 of the formula and area under the receiver operating characteristic curves of the nomogram were 0.446 and 0.865 in the development cohort, and 0.396 and 0.787 in the validation cohort, respectively. Conclusions We developed and validated novel predictive models for the postoperative renal function 3 years after radical nephrectomy in Japanese patients.