JMIR Publications

ABSTRACT

Background:

Diagnosing the underlying causes of non-neurogenic male lower urinary tract symptoms (LUTS) associated with bladder outlet obstruction (BOO) is challenging. Video-urodynamic studies (VUDS) and pressure-flow studies (PFS) are both invasive diagnostic methods for BOO. VUDS, in particular, can more precisely differentiate etiologies of male BOO, such as benign prostatic obstruction (BPO), primary bladder neck obstruction (PBNO), dysfunctional voiding (DV), potentially outperforming PFS.

Objective:

The invasive nature of these examinations highlights the need for developing non-invasive predictive models to facilitate BOO diagnosis and reduce the necessity for invasive procedures.

Methods:

A retrospective study was conducted on a cohort of male patients with medication-refractory, non-neurogenic LUTS suspected of BOO who underwent VUDS from 2001 to 2022. Two predictive models for BOO were developed: one based on the International Continence Society's definition (ICS-BOO) and the other on VUDS diagnoses (VBOO). The patient cohort was randomly split into training and test sets for analysis. Six machine learning algorithms, including logistic regression, were utilized for model development. In the model development process, we first performed development validation using repeated 5-fold cross-validation on the training set. Subsequently, we conducted test validation to assess the model's performance on an independent test set. Both models were implemented as paper-based nomograms and integrated into a web-based AI prediction tool to aid clinical decision-making.

Results:

In the cohort of 307 patients, 82 (26.7%) met the ICS-BOO criteria, while 252 (82.1%) were diagnosed with VBOO. The ICS-BOO prediction model showed an AUC of 0.74±0.09 and accuracy of 0.76±0.04 in development validation, and 0.86 AUC with 0.77 accuracy in test validation. The VBOO prediction model yielded an AUC of 0.71±0.06 and accuracy of 0.77±0.06 internally, with an 0.72 AUC and 0.76 accuracy externally. When the predictions from both models are applied to the same patient, their combined insights can significantly enhance clinical decision-making and simplify the diagnostic pathway. Six non-invasive parameters were adopted for dual-model predictions: IPSS-Voiding score, maximum urine flow rate, voided volume, total prostate volume, prostatic urethral angle, and intravesical prostatic protrusion. By the dual-model prediction approach, if both models positively predict BOO, suggesting all cases were actually caused by either medication-refractory PBNO or BPO, surgical intervention may be considered, thus VUDS might be unnecessary for about one-third (32.6%) of patients. Conversely, when prediction for ICS-BOO is negative but VBOO is positive, indicating varied etiology, VUDS is advised rather than PFS for precise diagnosis and guiding subsequent therapy, accurately identifying 51.1% of patients for VUDS.

Conclusions:

Two machine learning models predicting ICS-BOO and VBOO, based on six non-invasive clinical parameters, both demonstrate commendable discrimination performance. By the dual-model prediction approach, when both models predict positively, VUDS may be avoided, assisting in male BOO diagnosis and reducing the need for such invasive procedures.