External validation of the VOID predictive model for sacral neuromodulation in non-obstructive urinary retention and/or slow urinary stream.

PURPOSE: A multivariable predictive model for sacral neuromodulation (SNM) success in non-obstructive urinary retention and/or slow urinary stream (NOUR/SS), named VOID, was recently developed and internally validated. We aimed to externally validate this model.

METHODS

This retrospective multicenter study included all patients undergoing a two-stage SNM for NOUR/SS between 2018 and 2025 across 7 academic centers. The primary outcome was defined as the permanent implantation rate. Secondary outcomes included changes in bladder voiding parameters. Model performance was assessed through discrimination (area under the curve [AUC]), calibration (graphical assessment), and overall performance (Brier score).

RESULTS

Of 239 patients included, 152 (63.6%) underwent permanent implantation. The model showed fair discrimination (AUC = 0.627) and poor overall accuracy (Brier score = 0.221). Calibration analysis indicated imperfect agreement between predicted and observed probabilities. Among the original predictors, only female sex remained independently associated with implantation (OR 2.08, 95% CI 1.14-3.79; p = 0.017), while age ≤ 52 years showed a non-significant trend (p = 0.074). Maximal urethral closure pressure ≥ 70 cmH₂O and absence of neurological disease affecting the lower motor neuron were not associated with the outcome. The retrospective design and heterogeneity across centers represent the main limitations.

CONCLUSIONS

The VOID model demonstrated only fair discrimination and suboptimal calibration, precluding immediate clinical use without prior recalibration. Beyond methodological limitations, these findings suggest that underperformance may reflect the inherent constraints of a purely clinicourodynamic paradigm, underscoring the need for conceptually refined models integrating broader determinants of SNM response.