Probabilistic Assessment of Composite Urodynamic Success After Posterior Urethral Valve Ablation: A Monte-Carlo Simulation and Binormal ROC Analysis

Abstract

Background. Surgical techniques for posterior urethral valve (PUV) ablation are compared by their mean postoperative urodynamic values. Clinical success, however, is defined by thresholds: a patient either does or does not reach an acceptable residual volume, flow rate and outlet resistance. Whether a technique delivers success for an individual child depends on the dispersion of its results as much as on their mean, and a mean-based ranking can therefore mislead.

Objective. To estimate the probability that an individual patient meets a composite urodynamic success criterion under each of three ablation techniques, and to determine the discriminating capacity and optimal cut-off of the urethral resistance coefficient.

Methods. Postoperative distributions of post-void residual volume, mean flow rate and urethral resistance coefficient reported for 142 boys (endoscopic resection n = 32; polyethylene valvulotome n = 49; metallic valvulotome n = 61) were used as inputs to a Monte-Carlo simulation of 200 000 virtual patients per technique. Composite success was defined as residual volume ≤ 15 ml and mean flow rate ≥ 10 ml/s and resistance coefficient ≤ 0.12, all three simultaneously. The simulation was repeated with an inter-variable correlation of 0 and 0.30. Discrimination between the obstructed and relieved states was quantified by binormal ROC analysis with Youden-optimal cut-offs.

Results. Simulated composite success probabilities were 39.8 % for endoscopic resection, 97.4 % for the polyethylene valvulotome and 82.8 % for the metallic valvulotome; introducing correlation changed these by less than 4.3 percentage points. The ranking differs from that obtained by comparing means, in which the metallic valvulotome performs best on every individual variable. The discrepancy is attributable entirely to the dispersion of postoperative flow rate in the metallic group (15.36 ± 5.60 ml/s), which reduced the marginal probability of reaching 10 ml/s to 83.1 % despite the highest mean. Binormal ROC analysis of the resistance coefficient gave areas under the curve of 0.967, 1.000 and 0.996 with Youden-optimal cut-offs of 0.192, 0.146 and 0.100 respectively.

Conclusions. Probabilistic modelling identifies variability, rather than the mean, as the factor limiting composite success after metallic valvulotome ablation. The dispersion of postoperative flow rate in that group is disproportionate to the other groups and requires verification against the primary records. Comparative studies reporting only means may rank techniques in an order that does not reflect the probability of success for an individual child.