A simple predictive model with internal validation for assessment of stone-left after ureteroscopic lithotripsy in upper ureteral stones

Abstract

Background: Stone free rate in upper ureteral stones is not as high. We sought to identify easily accessible risk factors attributing to stones left in the ureteroscopy in the treatment of upper ureteral calculi, and to build a simple and reliable predictive model.

Methods: Patients treating only for upper ureteral stones in 2018 were retrospectively analyzed. Correlations between factors and the stone free rate were analyzed using bidirectional stepwise regression, curve fitting and binary logistic regression. Stone shape was judged by the gap between length and width in the two-dimensional section. A predictive nomogram model was built based on those selected variables (P<0.05). The area under the receiver operator characteristic curve (AUC) and calibration curve were used to access its discrimination and calibration. Decision curve analysis (DCA) was conducted to test the clinical usefulness.

Results: Totally, 275 patients with 284 stones were enrolled in this research. Bidirectional stepwise regression showed that stone length had a significant effect on stone free, instead of width or burden. Stone shapes were also found playing a big role. Curve fitting showed that quasi-circular stones had a high risk of retropulsion, and eventually led to stone left. Finally, stone length, shape, modality, and the distance of stones to the ureteropelvic junction were enrolled in the model. Among them, the distance of the stone to the ureteropelvic junction showed a noticeable impact on stone left. AUC was 0.803 (95% CI: 0.730-0.876), and the calibration curve showed good calibration of the model (concordance index, 0.792). DCA indicated the model added net benefit to patients.

Conclusions: The present predictive model based on those factors, stones length, shape, modality, and distance of the stone to the ureteropelvic junction was easy, reliable and useful.

Keywords: Ureteroscopy; nomogram; risk factors; stone free; ureteral calculi.

Figure 1

The relationship between stone length and the risk of retropulsion. (A) In all stones, a nonlinear relationship in all stones between length and retropulsion was observed after adjusting for operators and hydronephrosis. (B) In quasi-circular stones, a high incidence of retropulsion was observed after adjusting for operators and hydronephrosis.

Figure 2

ORs of factors on residual stones in binary logistic regression. CI, confidence interval; UPJ, ureteropelvic junction. URS, ureteroscopy; fURS, flexible ureteroscopy; ORs, odds ratio.

Figure 3

Nomogram model for predicting stone left based on length, shape, procedure, and distance from UPJ. Length, cm. Shape: 1, quasi-circular; 2, oval. Procedure: 1, semirigid ureteroscopy alone; 2, semirigid ureteroscopy with anti-retropulsive device; 3, flexible ureteroscopy. UPJ: 1, the distance of stone from UPJ ≤30 mm; 2, the distance of stone from UPJ 31–90 mm; 3, the distance of stone from UPJ >90 mm. UPJ, ureteropelvic junction.

Figure 4

Calibration and inspection of the model. (A) The area under the receiver operator characteristic curve (AUC) of the model. (B) Calibration curve of the nomogram model. Internal validation was performed using 1,000 bootstrap resamples, mean absolute error =0.019, mean squared error =0.00088. (C) Decision curve analysis for the nomogram model. The red line represents the nomogram model. The gray line represents the hypothesis that all patients had residual stones. The black line represents the hypothesis that no patients had residual stone.