Introduction

A preoperative nomogram to predict complications following radical nephroureterectomy (RNU) has previously been created. This nomogram incorporated 5 variables (patient age, race, ECOG performance status, CKD stage, and Charlson Comorbidity Index [CCI]) with an area under the curve of 72%. We externally validate this nomogram with a distinct international RNU patient population.

Methods

Amongst 610 RNU patients treated at 7 academic medical centers between 2005 and 2015, 382 (63%) had complete data with all variables reported as the initial nomogram. Logistic regression determined the association between preoperative variables and post-RNU complications. Nomogram validation was performed by analyzing the area under the receiver operating characteristics (AUC-ROC) curve. A calibration plot compared the nomogram-predicted probability of complications with the observed rate of complications within 30 days.

Results

225 men and 157 women with a median age of 71 were included. 85% of the patients were Caucasian, 18% has an ECOG ≥ 2, 25% had a CCI score >5 and 52% had baseline chronic kidney disease (CKD) ≥ stage III. Overall, 93 patients (24%) experienced a complication, including 31 (8%) with Clavien grade ≥ III. The performance of the nomogram was evaluated using two methods. Discrimination between individual patients was assessed by analyzing the AUC-ROC curve, which was 67.0% (95% CI 60.3%-73.7%). (Figure 1) A calibration plot compared the performance of the ideal nomogram (indicated by the dotted line), whereas the solid line represents the performance of this specific nomogram. There was a slight underestimation of complications for patients with high nomogram-predicted probabilities. (Figure 2)

Conclusions

External validation of a preoperative RNU complications nomogram noted an AUC-ROC curve of 67% with underestimation of complications for higher predicted probabilities. These observations may be a result of a lower complication rate observed in the validation versus original cohort (24% vs. 38%).

Funding

None