The complex mechanism of brain control of the bladder is not yet well understood. We used a therapeutic probe to investigate brain changes associated with therapy-induced changes in urgency urinary incontinence (UUI) frequency. We performed functional magnetic resonance imaging (MRI) during an urgency simulation task both before and 6-8 weeks after onabotulinumtoxinA treatment for UUI. We selected onabotulinumtoxinA on the premise that it is a peripheral treatment without central effects and therefore the brain changes associated with successful treatment should be relatively simple to interpret, reflecting response to changes in bladder function and not because of a direct effect on the brain.

We recruited 22 women over 60 years of age with non-neurogenic UUI and >5 leaks per week, who were scheduled to have onabotulinumtoxinA treatment. Each participant underwent clinical and physical examination, urinary incontinence evaluation, and a 3-day bladder diary. Prior to treatment, participants underwent MRI investigation, including structural brain imaging and concurrent functional MRI evaluation with an urgency-simulating protocol. This MRI protocol was repeated 6-8 weeks after onabotulinumtoxinA injection when therapeutic bladder responses had been achieved. The functional urgency simulating protocol was performed with a full bladder (strong urge to void signaled by the participant): 20 ml fluid was infused per 8Fr catheter over 12 seconds and withdrawn 20 seconds later over a 12 second period. This was repeated 4 times. ‘Activity’ was defined as blood oxygenation level dependent (BOLD) signal during the fluid withdrawal phases subtracted from that during the fluid infusion phases. 
Activity was evaluated individually at the first level (production of voxel-wise t-statistic maps for each individual) at baseline and post-therapy and compared at the second level (paired t-test of pre- vs post- in each individual combined to form a group average t-map) (SPM12, Wellcome trust, UK), and then evaluated by response rate; responders were classed as those with reduction in number of UUI episodes >50%). After voxel-wise calculations, we used the SPM12 small volume correction tool to assess our a priori  selected regions of interest (ROI): medial prefrontal cortex; dorsal anterior cingulate cortex/supplementary motor area; and right insula). Finally, we used a pattern recognition leave-one-out classification (PRoNTo[1]) with the baseline functional images to initially assess the ability to sort the participants into responders and non-responders for identifying predictive patterns for treatment response.

We analyzed 20 women; 2 were excluded for technical reasons. Ages were 60-84 years (mean (SD)=70.2(6.2)). Comparing all participants pre- and post- treatment, we found decreased activity following treatment in the right insula, posterior cingulate, right superior and right inferior parietal lobules, left inferior frontal gyrus, precentral gyrus, precuneus, caudate and right and left middle frontal gyri (p<0.05, uncorrected; 16 voxel threshold). We found increased activity following treatment in the left inferior parietal lobule, hippocampus, left posterior cingulate and culmen (p<0.05 uncorrected; 16 voxel threshold).
When sub-analysis was performed by response rate, we found decreased activation following treatment in responders (n=13) in the right inferior frontal and supramarginal gyri, right inferior and superior parietal lobules (Figure 1) as well as in our a priori ROI, the right insula (MNI [38 16 6]; r=10mm; cluster-level p<0.026, FWE corrected). We found no changes in non-responders from pre- to post-treatment. 
Using a leave-one-out cross-validation, we developed a support vector machine (SVM) classifier model to differentiate by response using the baseline functional images. We found an accuracy of 95%: 17/18 images were classified correctly.

There were significant preliminary changes in brain activity in response to urgency simulation after onabotulinumtoxinA therapy. The initial analysis of all participants showed many areas often found in continence investigation of the brain, including the insula which is found in the working model of continence control. 
We did further sub-analysis by response to assess if these changes were a function of reduction in UUI. fMRI imaging showed that activity significantly decreased in the insula, specifically within our a priori selected volume of interest (p<0.05 FWE corrected), after onabotulinumtoxinA therapy. This reduction appears to be driven by responders to therapy as it is more significant (whole brain analysis: P<0.001, uncorrected) once non-responders are grouped separately. This could signify that in those who respond, there is less afferent sensation from the bladder reaching the insula. No other ROI reached this level of significance in either the responders or non-responders. 
The use of PRoNTo as a pattern recognition tool is the first step towards a new way of using and understanding imaging data. Our aim was to find patterns within the imaging data that correspond to treatment response. The high accuracy rate of the SVM classifier to differentiate responders from non-responders is particularly encouraging. This is the first step in identifying patterns of brain activity which might predict therapeutic response, which, when probed further, might highlight important brain areas suggestive of different phenotypes of UUI. When applied to changes in response to therapy, this could also provide insight into more complex changes in brain function in response to urgency.

There are changes in brain activity in response to onabotulinumtoxinA therapy for UUI. One a priori selected ROI, the right insula, showed a significant reduction in activity after successful therapy which may reflect the reduction of bladder afferent sensation. Exploratory work using pattern recognition software suggests the potential for future analysis to identify functional brain activity patterns important in both predicting the success of therapy and changes associated with therapy, potentially highlighting further pathways involved in the continence mechanism.

Neuroinformatics. 2013 Jul;11(3):319-37