Hypothesis / aims of study
When all the conservative options (physiotherapy and medication) have failed to help patients with overactive bladder (OAB), therapeutic choices may include intra detrusor botulinum toxin injections (BTX) or Sacral Neuromodulation (SNM) [1].
While certain scenarios may favor one option over the other—such as patient preferences or concurrent fecal complaints—deciding between them can remain a daunting task for both caregivers and patients alike.
In such instances, having a reliable predictive tool could significantly aid in treatment decision-making, offering tailored guidance based on individualized factors.
Our study aims to fill this crucial gap by investigating a comprehensive array of clinical and urodynamic parameters to identify predictors of favorable outcomes following BTX injections. By constructing a dynamic nomogram, our study endeavors to equip clinicians with a powerful tool for personalized treatment decision-making in refractory OAB.
Study design, materials and methods
We retrospectively enrolled patients who underwent intradetrusor injections of 100U BTX between January 2017 and December 2022. Inclusion criteria comprised presentation with idiopathic OAB resistant to first-line treatments, absence of contraindications to BTX, prior pre-injection urodynamic assessment, and capability to complete a voiding diary.
Clinical Variables Extraction:
For each patient, clinical data including age, BMI, ethnicity, smoking status, history of pelvic surgery, diabetes, and types of complaints were collected. Additionally, urodynamic data encompassing B1, maximum bladder capacity, maximum detrusor pressure (Pdetmax), detrusor pressure at maximum flow (PdetQmax), maximum flow (Qmax), post-void residual (PVR), and presence of detrusor overactivity were extracted.
Endpoints:
Efficacy of BTX was evaluated using a composite endpoint consisting of subjective patient satisfaction and an objective criterion derived from a voiding diary at 6 weeks post-treatment. Improvement exceeding 50% in the main complaint constituted treatment success (definition of success is based on the one used for SNM: >50% improvement of the main symptom).
Statistical Analysis:
Continuous variables were summarized using mean and standard deviation for normal distributions or median and interquartile range for non-normal distributions. Categorical variables were expressed as percentages.
T-student tests were applied to normal variables, Brunner-Munzel tests to non-parametric continuous variables, and Chi² tests to categorical variables.
Nomogram Construction:
Variable selection for the generalized linear regression model (GLM) was iterative, aiming to optimize the area under the ROC curve (AUROC). Each variable was individually introduced into the model, followed by 50,000 shuffle split iterations to train and evaluate the GLM. Variables contributing to the best AUROC were progressively added to the model until further additions did not enhance performance. The dataset was divided into two subsets: a training set (80%) and a validation set (20%).
Results
Population Characteristics and Outcomes:
120 patients were included in the study.
The mean age of the cohort was 60 years (SD ±19), with a mean BMI of 28 kg/m² (SD ±7).
Out of the patients, 89 (74,16%) demonstrated significant improvements based on voiding diaries. Table 1 describes the patients’ characteristics.
Modified Stepwise Selection, Nomogram Construction and Development of Prediction Tool:
A GLM utilizing a stepwise approach allowed to retain nine variables: sex, history of pelvic surgery, diabetes, overactive bladder, B1, bladder capacity, maximum flow rate (Qmax), detrusor pressure at maximum flow (PdetQmax), and post-void residual (PVR) (Table 2). The median area under the ROC curve (AUC) was 0.84 (CI 0.55-1).
The resulting nomogram was constructed based on the coefficients of the GLM model (Figure 1) (using a Jupyter notebook in Python 3).
Interpretation of results
This nomogram represents the first predictive tool incorporating both clinical and urodynamic parameters, offering valuable insights for personalized treatment decision-making in refractory OAB. It includes nine variables, including sex, diabetes, type of complaint, detrusor overactivity, B1, bladder capacity, peak flow, PdetQmax, and post-micturition residual.
In our study, we investigated factors influencing the response to BTX treatment at 6 weeks, revealing a notable success rate of 75.42%, consistent with existing literature [2]. Sex, complaints of overactive bladder, urodynamic parameters, including bladder capacity and urinary flow rate, emerged as a significant predictor, aligning with previous findings and literature [3].
In the realm of urodynamic parameters, a PdetQmax exceeding 110 cm H2O, elevated bladder outlet obstruction index (BOOI), along with diminished compliance (less than 10 ml/cmH2O) are recognized as predictors of failure. Our investigation did not identify these factors. It is important to note that studies typically suffer from low statistical power and are conducted on small cohorts. Moreover, to the best of our knowledge, research findings frequently rely on mono- or bivariate analyses and seldom consider clinical and urodynamic data in conjunction.
The resulting model exhibited a notable area under the ROC curve of 0.84, indicating its robust discriminatory capacity.
The 95% confidence interval for this area under the ROC curve is 0.55 to 1, suggesting some potential variability in model performance. However, the minimum value of 0.55 still indicates a discrimination capacity superior to random chance (0.5).
Validation of the nomogram on an independent population would be crucial to establish its reliability and generalizability.