Limitations of standard scoring systems in predicting PCNL outcomes for horseshoe kidney patients.

OBJECTIVE

This study aims to evaluate and compare the predictive performance of three widely used nephrolithometry scoring systems—S.T.O.N.E., CROES, and Guy’s Stone Score—in determining stone-free outcomes following percutaneous nephrolithotomy (PCNL) in patients with horseshoe kidney (HSK), a population in which the utility of these models remains uncertain.

METHODS

In this retrospective single-center study, 64 adult patients with HSK who underwent unilateral PCNL between January 2017 and December 2024 were included. Preoperative non-contrast CT images were used to assess S.T.O.N.E., CROES, and Guy’s scores by two blinded endourologists. The primary outcome was stone-free rate (SFR), defined as complete clearance or residual fragments < 4 mm at 1-month postoperative CT. Predictive performance was evaluated via univariate/multivariate logistic regression and ROC curve analysis.

RESULTS

The overall SFR was 65.6%. All three scoring systems were significantly associated with SFR in univariate analysis. However, in multivariate analysis, only the CROES score remained an independent predictor of SFR (OR: 1.07; 95% CI: 1.02–1.13; p = 0.005). ROC analysis demonstrated the highest predictive accuracy for the CROES score (AUC: 0.81), followed by the S.T.O.N.E. score (AUC: 0.71) and Guy’s score (AUC: 0.55).

CONCLUSION

Among the evaluated scoring systems, the CROES nomogram achieved the best predictive performance; however, all three tools were developed for standard renal anatomy and demonstrated limited applicability in horseshoe kidney. These findings highlight the need for scoring models specifically adapted to this population.