Predicting who will succeed in reducing incontinence by learning to preemptively contract the pelvic floor muscles at the moment of leakage provocation

Abstract

Objective: The long term goal is to predict an individual woman`s odds that she will significantly decrease her urine leakage at the one-month post Knack-instruction visit. The Knack maneuver emphasizes "quick skill" (rather than strength training) to preempt a moment of expected urine leakage by contracting the pelvic floor muscles (Miller et al, 1998).

Methods: Sample: 59 women with mixed incontinence. Inclusion criteria: self-reported leakage on coughing, at least one episode of urine leakage on 3-day diary. Procedures: Baseline: Data was gathered on potential predictor variables of response to Knack-instruction: age, race, education, parity, BMI, POPQ, urethral closure profile at rest (MUCP) and during a Kegel (KUCP), hypermobility on ultrasound, incontinence severity by questionnaire, and a single self-efficacy question of "How confident are you that you can do things that will avoid urine loss?" We taught the Knack maneuver as "contract/hold/cough" using digital palpation and ultrasound visual feedback, and provided a handout about using the Knack to suppress unwanted urge sensations and stress-type incontinence. Follow-up: Response was evaluated at 1-month, with "responder" determined as 50% improvement on at least 2 of the following 3 measures: -leakage episodes on diary, -leakage volume on the quantified standing stress test (Miller 1998), -self report of improvement scaled 0 - 100%. Statistical analysis: All predictor variables were screened to determine which variables related significantly (p<.05) to our success/fail outcome. Those which were significant, were considered for inclusion in a multivariable logistic regression model.

Results: 35 women (59%) were responders and 24 women (41%) were non-responders. The most parsimonious logistic regression model which emerged (p < .05) retained the variables of self-efficacy, age, genital hiatus, and MUCP. In a new sample, the model is estimated to correctly classify 78.9% of total cases, to correctly classify 91.4% of actual responders, and to correctly classify 59.1% of actual non-responders. This sensitivity and specificity is based on using a cut value of .45 (to reduce the odds of a case wrongly classified as non-responder). Alternative cut values could be chosen to improve classification of non-responders (specificity), if desired.

Conclusion: Results provide a preliminary predictive model of response to Knack therapy. We are currently testing and refining the model in a larger sample of women.