Photo Credit: Elena Merkulova

The following is a summary of “External validation demonstrates machine learning models outperform human experts in prediction of objective and patient-reported overactive bladder treatment outcomes,” published in the September 2024 issue of Urology by Werneburg et al.

Predicting individual treatment responses for overactive bladder (OAB) remains a complex challenge in clinical practice. This study aimed to develop highly accurate machine-learning models to predict both objective and patient-reported outcomes following intravesical botulinum toxin (OBTX-A) injections. The models were rigorously validated using an external dataset comprising a markedly different patient cohort and dosing regimen, testing their robustness in a challenging clinical setting. The hypothesis was that these machine learning models would outperform human experts as well as top available algorithms in predicting treatment responses.

To achieve this, algorithms utilizing “operator splitting” were employed for their ability to maintain accuracy and efficiency, even with small and variably complete training datasets. These algorithms were trained on data from the ROSETTA trial cohort and validated against the ABC trial cohort, which included patients who underwent OBTX-A treatment. The performance of these models was evaluated by comparing the Area Under the Curve (AUC) with that of XGBoost with DART (a leading publicly available machine learning classifier), logistic regression with cross-validation, and predictions made by human experts.

In the validation set, the operator splitting neural network demonstrated superior performance with an AUC of 0.66, outperforming XGBoost (AUC 0.58), logistic regression (AUC 0.55), and human experts (AUC 0.47 – 0.53) in predicting clinical responder status. Similarly, the neural network showed greater accuracy in predicting patient-reported symptomatic improvement following OBTX-A, with an AUC of 0.64, again surpassing other algorithms and human experts (AUC 0.41 – 0.62).

In conclusion, the operator-splitting neural network not only outperformed human experts but also surpassed other machine learning models in predicting both objective and patient-reported outcomes for OBTX-A treatment in a challenging independent validation cohort. These findings suggest that clinical implementation of such machine learning models could significantly enhance patient counseling and optimize treatment selection for overactive bladder, leading to more personalized and effective care.

Source: sciencedirect.com/science/article/abs/pii/S0090429524007696