Evolution of renal function after ureteroscopy in solitary kidney

Flexible Ureteroscopy (FURS) in patients with a solitary kidney (SK) has been sparsely studied, especially its safety. Available literature suggests that the renal function remains stable after FURS with low evidence. We aimed to assess and predict the postoperative renal function after FURS in patients with SK, using Machine Learning (ML) models.

A dualcenter retrospective study was conducted between 2013 and 2023, including patients undergoing FURS in solitary kidneys. Demographic, peri-operative data, pre- and postoperative serum creatinine and estimated Glomerular Filtration Rate(eGFR) were collected. Multivariate analysis intended to identify risk factors for renal function decline. After data splitting, 12 ML models were trained and tested to predict the 3 months postoperative serum creatinine.

A total of 87 patients were included. 84% were treated for urinary stones, 11% for upper tract urothelial carcinoma(UTUC), and 4% for ureteral stricture. With a mean follow-up of 32 months, the serum creatinine remained stable at 3-month and long-term follow-ups (- 0.4 mmol/L, p = 0.77 and - 2 mmol/L, p = 0.57), despite a transient early elevation (+ 14.4 mmol/L, p = 0.00069). Multiple linear regression identified six significant predictors of postoperative renal function modification: enterocystoplasty, renal agenesis, Candida albicans in preoperative urine culture, elevated preoperative serum creatinine, and preoperative JJ stent for renal colic or obstructive pyelonephritis. After training, the "Stacking" Ensemble model presented the lowest mean absolute error between pre- and postoperative(3months) serum creatinine (4% vs 17% without ML) in the test group.

FURS seems safe in patients with SK, with transient renal function effects. Accurate predictive models could improve patient management, helping to prevent renal impairment, with better patient selection.