Urinary incontinence (UI) is a common, distressing condition affecting a third of women (1). Current national guidelines (2) recommend a minimum of 3 months’ pelvic floor muscle training (PFMT) for women with UI. The addition of electromyography (EMG) biofeedback may support women’s exercise motivation and performance, potentially improving their continence compared with PFMT alone, but may have greater associated costs (3). This study aimed to determine the effectiveness and cost-effectiveness of EMG biofeedback-assisted PFMT (biofeedback PFMT) compared to PFMT alone (PFMT) for female stress or mixed UI (SUI, MUI).

A multicentre randomised controlled trial compared biofeedback PFMT with PFMT in terms of UI severity at 2 years. A mixed-methods process evaluation and longitudinal qualitative case study were carried out in parallel (not reported here). Participants were recruited between February 2014 and July 2016 at 23 community and outpatient centres. Women 18 or older with a new diagnosis of SUI or MUI, who were able to contract their pelvic floor muscles, were eligible. Excluded were women who had received PFMT within the past year, had urgency UI or prolapse > stage II, were pregnant or <6 months postnatal, undergoing treatment for pelvic cancer, had cognitive impairment, a neurological disease, or a known nickel intolerance, or were participating in other UI research.  

Participants were randomly assigned (1:1) via remote computer allocation (minimised by type of UI, centre, age <50/≥50 years, and UI severity) to the biofeedback PFMT or PFMT group. All participants were offered six therapist appointments over 16 weeks and received an individually tailored PFMT programme. The biofeedback PFMT group additionally received biofeedback during appointments and were given a biofeedback unit for home use. Recognised Behaviour Change Techniques (BCTs) were incorporated into the PFMT protocols delivered for both groups. Outcome data were collected via questionnaire at 6 months, 1 and 2 years. Primary outcome: International Consultation on Incontinence Questionnaire UI Short Form (ICIQ-UI SF) score at 2 years. Secondary outcomes included: ICIQ-UI SF at 6 months and 1 year, global impression of improvement (PGI-I), uptake of other UI treatment, and pelvic floor muscle function (blinded assessment at 6 months). Primary health economic outcome: incremental cost per quality-adjusted-life-year (QALY) gained at 2 years based on EQ-5D-3L.Participants completed diaries to record their home PFMT and biofeedback use.  Blinding of participants, treating therapists and researchers to group allocation was not possible. 

The primary analysis was by intention-to-treat, with participants’ observed data analysed according to their randomised group. Group differences in ICIQ-UI SF scores at 2 years were assessed using a linear mixed model adjusting for minimisation variables, therapist type (physiotherapist/other therapist) and baseline score. Centre was fitted as a random effect. Secondary outcomes were analysed similarly, using appropriate generalised linear models (GLMs). 

A sample size of 468 (234 per group) was needed to detect a difference of 3 points on the ICIQ-UI SF between the groups with 90% power and 5% significance level, assuming a standard deviation of 10. We aimed to randomise 600 women to allow for over 20% dropout.  Sensitivity analyses of the primary outcome measure were conducted to examine the effect of missing data and non-compliance.  Exploratory subgroup analyses by age (<50/≥50 years), UI type and severity (ICIQ-UI SF score <13/≥ 13), and type of therapist used a stricter 1% significance level.  Cost-effectiveness analyses used GLMs, adjusting for minimisation variables and baseline EQ-5D-3L scores. Sensitivity analyses gauged the impact of varying key assumptions and/or parameter values in the base case analysis.

600 women were randomised, 300 per group. The participant mean age was 47.7 years (SD 11.5), and 61.3% had MUI. The trial group characteristics were well-balanced at baseline. Adherence (fully explored in the process evaluation) was comparable between groups in terms of appointment attendance and undertaking the home programme.

Questionnaire response rates at 6 months, 1 and 2 years were 74%, 83%, 77% in the biofeedback PFMT group and 74%, 85%, 79% in the PFMT group. Primary outcome data were available for 460 participants (225 biofeedback PFMT and 235 PFMT). There was no significant difference between the groups in the ICIQ-UI SF score at 2 years (mean for biofeedback PFMT 8.2 (SD 5.1) and PFMT 8.5 (SD 4.9); mean difference -0.09, 95% CI -0.92 to 0.75) (see Table). The results of planned sensitivity analyses produced very similar findings under different assumptions about non-compliance with the intervention and missing data (results not shown). There were no significant subgroup interactions: the treatment effect between groups did not appear to differ by age, UI type, UI severity or therapist type (results not shown). There were no differences between groups in the ICIQ-UI SF score at 6 months (mean difference 0.39, 95% CI -0.33 to 1.12) and 1 year (mean difference 0.57, 95% CI -0.17 to 1.31) (see Table).

Participant reported improvement: At 2 years there was no significant difference between groups in the PGI-I score (OR 1.14, 95% CI 0.75 to 1.72), with 41% (biofeedback PFMT) and 38% (PFMT) reporting they were “very much better” or “much better”.

Exercise adherence: At 2 years, the proportion exercising at least once a week was 52.0% in the biofeedback PFMT group and 46.3% in the PFMT group (OR 1.20 95% 0.83 to 1.74, from post-hoc analysis).

Pelvic floor muscle contraction strength: There was no difference between the groups in the Oxford Score for slow contraction strength at 6 months (OR 1.28, 95% CI 0.86 to 1.89, p=0.22), with 43.8% (biofeedback PFMT) and 39.7% (PFMT) scoring 4 or 5 out of 6.

Uptake of further treatment: By 2 years, similar proportions of women had received surgery for UI (12.3% biofeedback PFMT, 9.3% PFMT, OR 1.25, 95% CI 0.35 to 4.46). Uptake of further non-surgical care/treatment for UI was also comparable between groups (81.7% biofeedback PFMT, 79.7% PFMT, OR 1.35, 95% CI 0.54 to 3.41).

Cost-effectiveness: Biofeedback PFMT was not significantly more expensive than PFMT alone and did not generate significantly higher QALYs.  The incremental cost-effectiveness ratio (ICER) was £56,617.  The probability that biofeedback PFMT would be cost-effective was 48% and 49% at £20,000 and £30,000 thresholds for willingness to pay (WTP) for a QALY.

We found no evidence of a difference between biofeedback PFMT and PFMT in UI severity at 2 years, nor any difference in secondary outcomes.  The ICER indicated that Biofeedback PFMT would not be considered cost-effective at society’s WTP upper threshold of £30,000. Symptoms of UI were no better for women who had received biofeedback. Around half of participants were exercising weekly at 2 years regardless of trial group.

There is no evidence from this trial that the addition of EMG biofeedback to a robust protocolised programme of PFMT including BCTs offers benefit in terms of long-term continence outcomes. Other methods of encouraging long-term practice of PFMT to maximise its benefits should be investigated.

Abrams P, Cardozo L, Wagg A, Wein A.  Incontinence, 6th Edition. Health Publication Ltd 2017.
National Institute for Health and Clinical Excellence. Urinary incontinence: the management of urinary incontinence in women. 2013 RCOG Press.
Imamura M, Abrams P, Bain C, Buckley B, Cardozo L, Cody J, Cook J, Eustice S, Glazener CM, Grant A, Hay-Smith J. Systematic review and economic modelling of the effectiveness and cost-effectiveness of non-surgical treatments for women with stress urinary incontinence. Health Technol Assess 2010.