To develop and validate the prognostic performance of models estimating risks of persistent stress and urgency urinary incontinence (SUI/UUI) 1 year after midurethral sling (MUS) surgery (with and without behavioral therapy) in women with mixed urinary incontinence (MUI).

Data collected from women enrolled in a randomized trial comparing combined behavioral/pelvic floor muscle therapy with MUS to MUS alone for MUI were used for prognostic modeling. Participants had moderate or severe MUI symptoms for at least 3 months, and at least 1 SUI and 1 UUI episode on a 3-day bladder diary. Four outcomes were modeled at 1 year after surgery. A composite treatment failure outcome defined as: (1) not meeting the MID on the Urogenital Distress Inventory (UDI)-total score or (2) not achieving ≥70% improvement in mean daily incontinence episodes (IEs) on bladder diary, or (3) undergoing additional treatment for any lower urinary tract symptom (LUTS). Three additional outcomes were any additional treatment for LUTS, UDI-total score, and total IEs. Prognostic models used penalized logistic and linear regression with least absolute shrinkage and selection operator (LASSO) with baseline clinical characteristics, patient-reported outcome measures and urodynamic variables. Variable selection and internal validation were performed using 5-fold cross validation and uncertainty was assessed via bootstrap.

The derivation and internal validation cohort consisted of 379 women. At 1 year, 30% met criteria for composite treatment failure; mean (SD) UDI-total score was 37.3 (47.9), median (IQR) total IEs was 0 (0-1) and 15% required additional treatment for LUTS. The final models consisted of 11 variables: race, ethnicity, menopausal status, prior OAB medication use, insurance type, patient global impression of severity, comorbidity index, detrusor overactivity and SUI on cystometrogram, baseline detrusor pressure, and treatment assignment. The composite treatment failure model discriminated between those with and without the composite outcome (c-statistic [equivalent to the area under the receiver operator characteristic curve] =0.65; 95%CI: 0.53-0.75, Brier=0.20). Similar performance was seen for the model predicting additional treatment (c-statistic=0.70; 95%CI: 0.50-0.77, Brier=0.11). Calibration for the composite treatment failure model showed reasonable calibration when predicted probabilities were less than or equal to 50% but over-predicted risk when >50%. Similarly, the additional treatment model showed reasonable calibration when predicted probabilities were less than or equal to 30% but over-predicted risk when > 30%. The UDI-total score model was accurate within 35 points (mean absolute error=33.2, R2=0.08). The total IEs model was accurate within one IE (mean absolute error=0.97, R2=0.16). The addition of urodynamic variables improved the prognostic performance of models predicting total IEs (MAE 1.05 to 0.97) and need for additional LUTS treatment (c-statistic 0.63 to 0.70).

Among women undergoing MUS surgery for mixed urinary incontinence, the models provide opportunity to generate individualized estimates of risk of composite treatment failure when risk is less than or equal to 50%, additional treatment for LUTS when risk is less than or equal to 30%, UDI-total score within 35 points, and total IEs within 1 IE. Urodynamic variables are a valuable addition when combined with other variables in the total IE model and need for additional LUTS treatment model.

These models may provide individualized estimates for women with MUI of risk for persistent SUI/UUI and need for additional LUTS treatment after surgery with MUS. Urodynamic variables add new prognostic information to patient characteristics alone.