Basic versus biofeedback-mediated intensive pelvic floor muscle training for women with urinary incontinence: the OPAL RCT

Possible mechanisms for how biofeedback might improve continence

Biofeedback acts as a strong and specific training stimulus, which has the potential to markedly intensify a training programme. This effect is evidenced in the biofeedback and general exercise literature. 6 In this report, the exercise being referred to is PFME. In the context of supervised PFMT for treatment of UI, biofeedback contributes to:

• teaching women the precise, correct muscle contraction in terms of technique, force and timing, and assisting the therapist in prescribing the correct exercise dose to ensure that appropriate physiological responses occur in the muscles

• informing adjustments to the exercise prescribed by the therapist (modulating)

• highlighting to women improvements in their muscle function (encouraging)7

• increasing the quality of the interaction with the therapist (via biofeedback used in ways congruent with health behaviour change theory and practice)

• change in women’s behaviour relating to home exercise, as biofeedback has been classified as a behaviour change technique (BCT) in a taxonomy of BCTs. 8

These features have the potential to increase a woman’s confidence (self-efficacy) in doing PFMT; self-efficacy for PFMT is a determinant of intention to adhere9 and adherence itself. 10,11 For PFMT to have an effect on UI, sufficient exercise must be carried out over a long enough period to strengthen (hypertrophy) the muscle. 12 Thus, maximising adherence to exercise is key to an intensive PFMT intervention. Intensive PFMT, which fosters PFMT self-efficacy, can potentially be achieved through the addition of biofeedback, leading to improvement in the quality and quantity of PFMT performed by women, and to more improvement in the severity of their UI than a less intensive programme.

Evidence for the effectiveness of biofeedback for female urinary incontinence

Both a Cochrane systematic review of biofeedback-assisted PFMT for women with UI7 and a Health Technology Assessment-funded systematic review of non-surgical treatments for women with SUI13 found that biofeedback-assisted PFMT appeared to offer benefit over basic PFMT alone. However, the Cochrane review,7 with its more restricted scope, conducted a more detailed analysis of the biofeedback trials and found that this effect may be confounded by the greater amount of health professional contact received by women in the biofeedback groups.

Rationale for research

Based on current evidence, it is not clear if the apparent benefit of biofeedback can be attributed to the biofeedback or to some other variable, such as additional health professional contact. Some of the trials in the Cochrane review7 had the same PFMT programme in both arms of the trial and the same amount of health professional contact. The results from these trials still favoured biofeedback, although the difference between groups was not statistically significant. A common problem in all these trials was the failure to clearly state the purpose of biofeedback, or to describe the intervention protocol. Thus, it was not clear if biofeedback could, theoretically or in practice, change the clinical effectiveness of the PFMT. Currently, the routine use of biofeedback as part of PFMT is not recommended and should be considered only for women who are unable to contract their pelvic floor muscles. 1 Despite this, we know that therapists use biofeedback in their clinical practice as an adjunct to a supervised PFMT programme when treating female UI. 14

Thus, a robust comparison of biofeedback-mediated intensive PFMT compared with basic PFMT, in which both groups have the same basic PFMT programme and the same amount of health professional contact, is imperative to establish whether or not biofeedback does add value and improve incontinence outcomes.

Aims and objectives

The overall aim of the Optimal Pelvic floor muscle training for Adherence Long-term (OPAL) trial was to carry out a randomised controlled trial (RCT) to evaluate a theoretically based biofeedback-intensified PFMT intervention to answer the primary research question: is biofeedback-mediated intensive PFMT, compared with basic PFMT, a clinically effective and cost-effective treatment for women with SUI or MUI (stress and urgency)?

The trial comprised the following components:

• Development of biofeedback-mediated intensive PFMT and basic PFMT interventions based on health behaviour theory designed to treat SUI and MUI.

• A RCT in which the use of biofeedback as an adjunct to basic PFMT (biofeedback PFMT) was compared with PFMT alone (basic PFMT), in terms of clinical effectiveness and cost-effectiveness.

• A process evaluation to identify and investigate the possible mediating factors that affect the effectiveness of the interventions (including fidelity to intervention delivery and uptake), how these mediating factors influence effectiveness and whether or not the factors differ between randomised groups.

• A longitudinal qualitative case study to investigate women’s experiences of the trial interventions, to identify the barriers and facilitators that affect adherence, to explain the process through which they influence adherence and to identify whether or not these differ between randomised groups.

Development of the OPAL trial interventions

The OPAL trial interventions (including biofeedback-mediated intensive PFMT and basic PFMT protocols) were developed by Jean Hay-Smith, Doctor of Philosophy (PhD) (University of Otago, Dunedin, New Zealand); Sarah G Dean, PhD (University of Exeter, Exeter, UK); Doreen McClurg, PhD [Nursing, Midwifery and Allied Health Professions Research Unit (NMAHP RU) Research Unit, Glasgow Caledonian University (GCU), Glasgow, UK]; Suzanne Hagen, PhD (NMAHP RU, GCU, Glasgow UK); Carol Bugge, PhD (University of Stirling, Stirling, UK); and Joanne Booth, PhD (GCU, Glasgow, UK).

The OPAL trial interventions are essentially behavioural intervention (i.e. an intervention focusing on the adoption and maintenance of a PFMT programme, with or without use of biofeedback). PFMT adherence is problematic and often decreases over time. 15 PFMT interventions typically include a number of ‘active’ ingredients that ‘interact’ (e.g. in the OPAL trial, the exercise and biofeedback ingredients are presumed to interact to intensify PFMT) to produce an effect that is potentially greater than the sum of the parts. Thus, the OPAL interventions had two core characteristics (behavioural difficulty and interacting components) of a ‘complex’ intervention. 16 A further element of complexity was the tailoring components within the interventions, which allowed flexibility in delivery; these components have been called ‘optional’ in the intervention protocols.

Four sources were used to develop the intervention:

1. theory consistent with our objective of changing health behaviour: the information motivation behavioural skills (IMB) model17

2. existing qualitative evidence about women’s experiences of PFMT and why they did or did not adopt the behaviour to undertake PFMT18

3. the explicit use of BCTs to support adoption and maintenance of the behaviour8

4. the existing PFMT intervention from the Pelvic Organ Prolapse PhysiotherapY (POPPY) trial,19 which provided the outline for what therapists considered ‘basic’ PFMT in UK clinical practice.

Once developed, the intervention protocols to be delivered were detailed in the intervention manual, which was given to each therapist. Separate protocols were available for biofeedback-mediated intensive PFMT (referred to hereafter as biofeedback PFMT) and basic PFMT. In practice, the interventions were delivered over six appointments with a trial therapist, which the participant attended. Details of the desired content of each appointment were provided in the intervention manual and in checklist format in the therapist assessment form (TAF), completed by the therapist at each appointment. This is described in more detail in Chapter 3.

Both trial groups (biofeedback PFMT and basic PFMT)

In both groups, participants completed the basic PFMT protocol over a 16-week period:22 six individual face-to-face appointments with a therapist were offered at weeks 0, 1, 3, 6, 10 and 15, based on a previous trial of PFMT. 19 At the first appointment the therapist performed a vaginal examination, including visual inspection and digital assessment of the pelvic floor muscles (the Oxford Classification,23 and the International Continence Society method24).

During this assessment, all participants were taught how to contract and relax their pelvic floor muscles and how to pre-contract their pelvic floor muscles prior to (and in expectation of) increases in abdominal pressure, such as coughing and sneezing (‘the Knack’25).

Based on the findings of the assessment, the therapist and participant agreed a PFMT programme, tailored according to the woman’s ability and lifestyle, which progressed over time. The PFMT programme focused on strength and endurance, aiming for three sets of contractions per day, progressed by increasing the number of repetitions, hold duration and body position modification on progression (i.e. lying, sitting, standing, squatting).

As per local practice, therapists provided participants with information to support good bladder management and methods of dealing with urgency and frequency. Therapists also used the BCTs that had been mapped to, and detailed in, the intervention protocols to encourage participants’ adherence to the PFMT programme. Participants recorded the PFMT they performed in a diary [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)].

No EMG biofeedback equipment was used during appointments for the basic PFMT group. Verbal feedback based on digital vaginal palpation was, however, permitted.

Biofeedback PFMT

In addition to the basic PFMT protocol provided to both groups [see Both trial groups (biofeedback PFMT and basic PFMT)], women randomised to the biofeedback PFMT group received a biofeedback protocol during appointments and were provided with a biofeedback unit to use at home. Biofeedback units were provided by the trial office and the same type of unit was used during appointments and at home [NeuroTrac^® Simplex single-channel EMG biofeedback unit (Verity Medical Ltd, Romsey, UK)]. The biofeedback unit displayed participants’ pelvic floor muscle activity, providing audio and visual feedback on muscle contraction and relaxation. Participants were instructed how to insert, use and clean the probe, operate the unit and interpret the biofeedback information. Therapists set the parameters of the biofeedback unit and agreed with participants the frequency for use at home. The units stored data on home usage and participants also recorded their home use (and PFMT practice) in a diary.

Primary outcome

The primary outcome was the ICIQ-UI SF score at 24 months. This met with the commissioning brief from the funder, which highlighted symptom severity as an important outcome and 2 years as the minimum duration of follow-up. The ICIQ-UI SF26 is a four-item questionnaire (total score ranging from 0 to 21, with higher scores indicating greater severity), covering frequency of UI (never = 0, once a week = 1, two or three times a week = 2, once a day = 3, several times a day = 4, all the time = 5), amount of leakage (none = 0, small = 2, moderate = 4, large = 6), overall impact of UI (not at all = 0, a great deal = 10) and a diagnostic question to identify type of UI. Responses to the first three questions are summed to give the total score. The maximum score is 21; the higher the score, the more severe the UI, with a score of ≥ 13 rated as severe. 27

Secondary outcomes

Economic-related outcomes

The primary health economic outcome measure of cost-effectiveness was incremental cost per quality-adjusted life-year (QALY) at 24 months, based on responses to the EQ-5D-3L. This incremental cost-effectiveness ratio (ICER) was defined by the difference in cost between two trial interventions, divided by the difference in their effect.

Secondary outcomes were (1) the cost of resources used by participants, including the use of primary (GP services) and secondary (outpatient visits, inpatient stay, surgical interventions for incontinence) care services, and further referral for subsequent additional specialist management; (2) the personal costs to the participants, including costs of travelling to appointments and work/social restrictions; and (3) QALYs derived using the ICIQ-LUTSqol responses.

Full cost-effectiveness methods are described in detail in Chapter 4.

Data collection

Resource utilisation was measured over four time points (baseline and 6, 12 and 24 months) using TAFs and participant-completed questionnaires, which included health-care utilisation questions. The second year costs and benefits were discounted using the National Institute for Health and Care Excellence (NICE) recommended 3.5%. 39 Intervention resource use included the number of appointments with the therapists as captured on the TAF. Although the first appointment was expected to last for 1 hour and each follow-up appointment to last for 30 minutes, data on the length of each appointment were also recorded on the TAF. The biofeedback PFMT group had the additional resource of biofeedback in the clinic and a portable biofeedback unit for home use. Information on the biofeedback units, such as type of units, type and number of electrodes, internal probes, laptops, software and licences for laptops, printers and cables, was sourced from the trial office.

Further resource use relating to UI was recorded retrospectively for every woman in the trial. These data were collected at 6, 12 and 24 months, using participant-completed questionnaires and included the use of primary (GP services) and secondary (outpatient appointments, inpatient stay and surgical interventions for incontinence) health care. Health service costs refer to those incurred directly by the NHS, such as related surgery, additional appointments and procedures. As physiotherapist appointments were part of the trial interventions, it was probable that participant-reported appointments in the 6-month questionnaire included some or all of the trial-related appointments and would introduce the chance of double-counting number and cost. Therefore, the numbers of trial appointments and participant-reported incontinence-related appointments were compared to establish a clear pattern in the latter, in an attempt to exclude any already captured in the intervention costing. It was assumed that a one-appointment difference in participant-reported appointments was most probably due to patient recall bias. It was assumed that these participants had no additional appointments, other than the trial appointments, and incurred no additional cost. Those women who reported more than one appointment in addition to their trial appointments incurred the cost of the difference in participant-reported and trial appointments. Resource use incurred at personal cost to the participants (such as purchase of pads, medication) was collected using the participant-reported 24-month questionnaire.

Intervention costs

The cost of the biofeedback unit was calculated on a per-participant basis and was estimated based on the following assumptions: the cost of biofeedback units and equipment (laptops, printers, cables) obtained from the trial office was spread out over 5 years, reflecting the expected lifespan of all equipment (Amanda Tombs, Verity Medical Ltd, 2018, personal communication). It was assumed that, on average, three women would use the unit every year to allow for the 16-week intervention period and time for units to be returned to the centres. The cost of consumables (two packets of electrodes and a probe) was also included. As the biofeedback units and consumables were purchased across several years, the Hospital and Community Health Service (HCHS) index40 was used to bring all costs to the same year. The cost of the biofeedback unit per user was then annuitised, using an annual rate of 3.5% over its lifespan of 5 years. This produced a per-participant unit cost of biofeedback of £37.40 for the 16-week intervention period (see Appendix 9 for details on the costing).

Medication costs

The incontinence care pathway suggests that, before women progress to the severity level, when surgery is needed, treatment with medications for UI and overactive bladder is recommended. 10 Incontinence medications have a very low unit cost; Mirabegron (Betmiga; Astellas Pharma Inc., Tokyo, Japan) and Solifenacin (Vesicare; Astellas Pharma Inc.) are the most frequently reported and have the highest unit cost (see Appendix 10). As participant-reported medication data were available only at distinct time points (from 6-, 12- and 24-month follow-up questionnaires), an assumption was made that participants started their treatment halfway between the follow-up points.

Unit costs

Unit costs and prices were obtained using published estimates from the British National Formulary for medications,41 NHS Reference Costs 2016/1742 for secondary care resource use and the Personal Social Services Research Unit’s Unit Costs of Health and Social Care 201740 for primary care resource use, as outlined in Table 15.

Estimation of participant costs

For each area of resource use, estimates of resource utilisation were multiplied by unit costs to derive total costs for each item of resource use and each participant. These data were averaged to provide estimates of the average cost per participant for each item of resource use. These costs were summed to produce a total cost for each participant and an average total cost per participant in each intervention group.

Participant- and companion-incurred costs and indirect costs

Personal costs to the participants (such as costs of travelling to appointments and work/social restrictions) were also investigated. Participant resource utilisation comprised three main elements: (1) self-purchased health care; (2) travel costs for making return appointment(s) to receive NHS health care (such as petrol, public transport and parking); and (3) time costs of travelling and attending NHS health care (such as time involved away from usual activities or work). The self-purchased health-care items included in the investigations were services, medications and pads paid for by participants for the treatment or management of UI symptoms, and time and travel costs related to time spent travelling to, and attending, hospital or primary care providers in relation to UI. Estimation of travel costs included information from participants about the number of appointments with, for example, their GP or physiotherapist (estimated from the health-care utilisation questions at the various follow-up questionnaire time points) and the unit cost of making a return journey to each type of health-care provider (from the participant time and travel cost questions collected at 24 months). The cost of participant time was estimated in a similar manner. The participant was asked how long they had spent travelling to, and attending, their previous appointment with each type of health-care provider. Information was sought on the activity they would have been undertaking (e.g. paid work, leisure, housework) had they not attended the health-care provider. They were also asked if they were accompanied by a friend or a relative and their time and travel costs were also incorporated into the analysis. These data were presented in their natural units (e.g. hours) and also costed using standard economic conventions, using the Department for Transport44 estimates for the value of work and leisure time. These unit time costs were combined with the number of health-care contacts derived from the health-care utilisation questions, to elicit a total time and travel cost from a participant perspective. Details of unit costs applied to the various activities are included in Table 16.

Quality-of-life measures

A generic instrument, the EQ-5D-3L, was used to measure the participants’ quality of life. The EQ-5D-3L was completed at baseline and at 6, 12 and 24 months post randomisation. The responses to the EQ-5D-3L questionnaires were valued using UK general population tariffs, based on the time trade-off technique to generate a utility score for every participant in the trial. 47 QALYs were calculated by combining utility with health state duration, using an area beneath the curve approach, assuming linear interpolation of utility between time points. Quality-of-life data were also collected using items from the condition-specific tool ICIQ-LUTSqol for comparison. This tool was included as it was anticipated that the generic measure (EQ-5D-3L) may not be sensitive enough to pick up the continence-related factors that influence quality of life for these participants, such as impact on personal relationship and sleep. 48 The ICIQ-LUTSqol is a 21-item, condition-specific measure of health for urinary incontinent participants, developed in a UK population similar to that in this trial. This instrument was revised into a five-dimensional health state classification amenable to valuation, using items selected using psychometric evidence. Forty-nine states were valued using standard gamble by a representative sample of patients with UI attending UK hospital outpatient clinics. These data were converted into a utility index using a published algorithm6 and used as an alternative utility value for condition-specific QALY calculation.

Data analysis

The economic analysis was undertaken on the ITT principle. All components of costs were described with the appropriate descriptive statistics: mean and SD for continuous and count outcomes; numbers and percentages for dichotomous and categorical outcomes (e.g. numbers reporting problems on EQ-5D-3L). All analyses were conducted using Stata^® version 14.1 software (StataCorp LP, College Station, TX, USA). Investigations were carried out for skewed cost data (i.e. a small proportion of participants incurring very high costs), using GLMs to test alternative model specifications for appropriate fit to the data. The GLM allowed for heteroscedasticity by selecting and specifying an appropriate distributional family and link function for the data. This family offered alternative specifications to reflect the relationship between the mean and variance of the estimates under consideration. 49,50 Two diagnostic actions were performed to identify the most appropriate distributional family: (1) a modified Park test and (2) the Akaike information criterion were consulted. The best model fit for analysing cost data appeared to be the gamma distribution with an identity link. The Gaussian distribution with an identity link was chosen for the QALY analysis. The appropriate link to characterise the relation between the linear combination of predictors and the prediction was selected based on the results from the Pearson correlation, Pregibon link and the modified Hosmer–Lemeshow tests (see Appendix 12).

Both cost and QALY difference analyses were adjusted for baseline prognostic factors (all of which were minimisation covariates plus the type of therapist and baseline EQ-5D-3L utility score):

• centre number

• age in years

• type of UI

• UI severity

• type of therapist (physiotherapist/nurse)

• baseline EQ-5D-3L utility score.

The first five factors are in line with the clinical effectiveness analyses and the baseline EQ-5D-3L was included for the economic analysis. Standard parametric tests for differences in costs were performed, with the robustness of the parametric tests confirmed using bias-corrected, non-parametric bootstrapping. 51

Missing data

Missing data are a frequent problem in economic evaluations undertaken in a RCT setting. There are several possible methods that can be employed to account for such missing data, including mean or multiple imputation. Multiple imputation analysis was conducted for the base-case analysis, as > 5% of the quality-of-life and cost data needed were missing for the primary analysis. Components of cost data were imputed, based on linear regression models that were adjusted for minimisation variables, baseline utility and intervention allocation group. Missing utility values were imputed using predictive mean matching (the mean of five nearest values). Chained equations were used for the imputations; 20 imputations were considered sufficient to generate stable and reliable estimates for analysis. Imputed data were then analysed using the appropriately specified GLMs described in Data analysis.

Incremental cost per quality-adjusted life-year gained

Incremental cost-effectiveness ratios were computed using the imputed data. Differences in costs were based on resource utilisation costs over the 24-month follow-up period. The difference in effectiveness was expressed in terms of QALYs at 24-month follow-up. The economic results were reported using a cost–utility analysis framework, with the ICER representing the incremental cost per QALY gained. The point estimate of the ICER was calculated as:

in which Ci and Cj are the mean costs among participants in the biofeedback PFMT and basic PFMT group, respectively. Similarly, Ei and Ej are the mean QALYs in the biofeedback PFMT and basic PFMT groups, respectively. The ICER was assessed against the NICE-recommended cost-effectiveness threshold of £20,000–30,000 per QALY gained.

Measures of variance for NHS costs and QALYs were derived using non-parametric bootstrapping. From the results of the bootstrapping, cost-effectiveness acceptability curves (CEACs) were created. CEACs are used to display the inherent uncertainty surrounding cost-effectiveness at various threshold values for society’s willingness to pay (WTP) for an additional QALY. CEACs present results when the analysis follows a net benefit approach. This approach utilises a straightforward rearrangement of the cost-effectiveness decision rule used when calculating ICERs to create the net monetary benefit (NMB) for each bootstrapped iteration at increasing values of WTP per QALY:

in which λ represents a decision-maker’s WTP for incontinence avoided or a QALY gained. If the above expression holds true for a given iteration and threshold WTP value (λ), then the intervention is considered cost-effective for that iteration. As society’s WTP is unknown, the NMB will be calculated for a number of possible λ values, including the usual £20,000–30,000 range often adopted by policy-makers in the NHS. 39 A balance sheet approach was used to report the costs and QALYs of women who are or are not incontinent.

Sensitivity and subgroup analyses

Sensitivity analyses were performed to gauge the impact of varying key assumptions and/or parameter values in the base-case analysis. Sensitivity analyses in relation to the assumptions made in the cost of the intervention were performed. The base-case analysis utilised number of minutes recorded by therapists for each appointment, to estimate the average cost per intervention. The first sensitivity analysis performed was based on the recommended time of appointments of 1 hour for the first appointment and 30 minutes for the following five appointments. 1 The base-case analysis in terms of utilities was adjusted for baseline EQ-5D-3L values to account for variability that may be present among the intervention groups. An unadjusted analysis was also performed as a sensitivity analysis to highlight the importance of this base-case assumption. An analysis exploring the impact of changing the discount rate used for second-year costs and QALYs in accordance with NICE best-practice recommendations, varying the discount rate from 0% to 6% per annum, was undertaken. 1 A subgroup analysis similar to that described in the statistical analysis plan was undertaken. This was based on:

• type of incontinence (SUI or MUI)

• age (< 50/≥ 50 years)

• UI severity (ICIQ-UI SF score of < 13 or ≥ 13).

A subgroup analysis on the type of therapist was not performed, as almost all participants were seen by a physiotherapist.

Sensitivity analysis was also performed to assess the impact of the assumption that data were MNAR. Several methods have been proposed in the statistical literature to conduct analyses under MNAR. 52 The analysis was conducted using pattern mixture models to capture how the distribution of missing values could differ from the conditional distribution based on the observed data. Several scenarios were considered. It was supposed that women who did not complete the questionnaires were in a poorer health state and possibly incurred higher costs. We assumed that the health state could have been 10% lower and costs would have been 10% higher than the missing at random setting. These assumptions were applied to two different settings, that (1) they were missing at the same value for both groups and (2) they were missing for one group only each time. In total, seven scenarios were considered.

Missing data

Data on intervention resource use and costs were missing for participants who withdrew consent, but their baseline data were still usable (five in the biofeedback PFMT group and two in the basic PFMT group). Details of missing participant-reported data on resource use in the primary and secondary care trial intervention are reported in Table 17. The proportions of missing data for cost and utility outcomes were equally distributed between the two groups. Complete total cost data were available for only 78 participants: 40 participants (13%) in the biofeedback PFMT group and 38 participants (13%) in the basic PFMT group. Utility data were available for 174 (biofeedback PFMT) and 158 (basic PFMT) participants. A summary of missing data for each cost variable, first year and total QALY is presented in Table 17 (see Appendices 13 and 14).

As described in Methods, missing data are a significant issue for the calculation of complete-case cost and QALY pairs. This is a common issue in cost-effectiveness analyses alongside trials, in which data can be missing at an individual and item non-response level. It was therefore necessary to impute to address the significant missing data in the base-case cost-effectiveness analysis.

Resource use

Table 18 presents the resource use associated with the intervention and follow-up care in the biofeedback PFMT and basic PFMT groups.

Intervention

Overall, 264 out of 300 (88%) participants randomised to biofeedback PFMT received a home unit. Sixteen (44%) of those participants who did not get a biofeedback unit did not receive any intervention within the trial, five (14%) withdrew consent, seven (19%) withdrew after their first appointment, four (11%) received basic PFMT in error and the reason why the four (11%) remaining participants did not receive a home unit was unclear. One participant in the basic PFMT group was given a home biofeedback PFMT unit in error. Participants in both trial groups attended an average of four out of their six scheduled appointments, but the overall time spent at appointments was, on average, 26 minutes longer (95% CI 16.21 to 35.42 minutes) for those participants in the biofeedback PFMT group.

Follow-up care

Overall, the two trial groups had very similar levels of resource use in the follow-up period. Full details of all follow-up care resource use reported in the trial are provided in Table 18. The numbers of general practice doctor and nurse appointments were similar for both groups. The differences for both types of appointments were not statistically significant.

Although the biofeedback PFMT group had higher resource use for UI symptom management in secondary care (visits to hospital doctor, nurse or physiotherapist, and inpatient stay) than the basic PFMT group, the differences in resource use were not statistically significantly different. Furthermore, there was no significant difference in the number of surgical interventions (tension-free vaginal tape, bulking agents) or medications.

Intervention and follow-up costs

The differences in costs mirror the differences reported in resource use. Details of the cost results are reported in Table 19. As expected, the cost of biofeedback units was £33.76 (95% CI £31.99 to £35.52) higher in the biofeedback PFMT group. The cost of the appointments was also £53.12 (95% CI £34.20 to £72.05) higher in the biofeedback PFMT group. Both of these costs were statistically significantly higher. Therefore, the total cost of the within-trial intervention was significantly higher (£86.43, 95% CI £67.09 to £105.76) in the biofeedback PFMT group than in the basic PFMT group. The remaining cost differences were not statistically significantly different between the groups. Overall, the total cost was higher in the biofeedback PFMT group, but not statistically significant (difference was £120.97, 95% CI –£409.08 to £651.02). These findings are based on a complete-case analysis and it is notable that full-cost profile data were available for only 40 out of 300 (13%) participants and 38 out of 300 (13%) participants in the biofeedback PFMT and basic PFMT groups, respectively.

Generic and condition-specific quality-of-life measures

Results of EQ-5D-3L scores and adjusted differences between the groups are presented in Table 20. Participants randomised to the biofeedback PFMT group reported significantly lower EQ-5D-3L utility scores at baseline (mean difference biofeedback PFMT vs. basic PFMT –0.048, 95% CI –0.085 to –0.01). None of the follow-up scores was statistically significant between groups. The biofeedback PFMT group had lower QALYs (1.571) overall than the basic PFMT group (1.619) at 2 years, driven by the EQ-5D-3L imbalance in the baseline score. When adjusted for baseline differences, the incremental QALY (biofeedback PFMT vs. basic PFMT) was –0.041 (95% CI –0.121 to 0.039), which was not statistically significant.

The proportion of trial participants who reported problems in the domains of the EQ-5D-3L were equally distributed between the two groups (Figure 5). Pain/discomfort and anxiety/depression was the dimensions associated with most problems. Details of the number of participants are reported (see Appendix 15).

FIGURE 5.

Proportion of participants experiencing any problems in each of the EQ-5D-3L domains.

The mean ICIQ-LUTSqol scores at baseline and at each follow-up time point, as well as the differences between the two trial groups, are presented in Table 21. None of the differences in ICIQ-LUTSqol was statistically significant across the groups.

The proportion of participants in each trial group who reported any issues on each domain of the ICIQ-LUTSqol utility measure are reported in Figure 6 (see Appendix 16). The problems reported by the highest proportion of participants were with physical activities, such as travelling, walking, running (between 79% and 97%), followed by emotional distress associated with their incontinence (between 60% and 83%) and limitations associated with their job or normal daily activities in and outside their household (between 53% and 84%), over the different time periods. The largest reduction in the proportion of participants reporting problems was observed between baseline and 6 months, and this occurred in both trial groups. The proportion of participants who reported issues remained stable for the rest of the follow-up (see Figure 6).

FIGURE 6.

Proportion of participants experiencing any problems in each of the ICIQ-LUTSqol domains by group.

Costs incurred by and quality-adjusted life-years of cured and incontinent participants

At 24 months, the total cost of those participants who were reported as cured according to the primary clinical outcome (ICIQ-UI SF score) was £1032 (SD £942). The mean total cost for those who were still incontinent at the end of the follow-up period was £1203 (SD £1333). The mean total QALY for participants who were cured was 1.85 (SD 0.17), whereas it was 1.57 (SD 0.48) for those who were still incontinent. These results are based on only those who had complete data.

Costs directly incurred by participants and carers and indirect costs of time and productivity lost due to urinary incontinence

There were no statistically significant differences in personal costs to the participants between the two groups (Table 22). The number of participants who reported the need to take full days off work because of incontinence was low (n = 16). Among those who required time off from usual activities in the follow-up period, the average number of days off in the basic PFMT group was 12.5 and the average number of days off in the biofeedback PFMT group was 10. The differences between groups were not statistically significant. The cost of self-purchased incontinence products (pads) constituted the highest annual cost to the participants: £806 for the biofeedback PFMT group and £862 for the basic PFMT group; however, the difference between the groups was not statistically significant. Only five participants reported self-purchasing other incontinence products, such as biofeedback units, vaginal cones or pessaries, in the follow-up period. The time and travel cost to the participant and carer (if applicable) of attending outpatient appointments was the second highest cost (on average, £200 for the biofeedback PFMT group and £303 for the basic PFMT group). Only very few participants (six in the basic PFMT group and eight in the biofeedback PFMT group) reported use of private secondary care (appointments to private doctors, nurses and physiotherapists). The total participant perspective cost difference (biofeedback PFMT vs. basic PFMT) was –£74.21 (95% CI –£201.21 to £52.78), but this was not statistically significant.

Base-case cost-effectiveness analysis

As described in Missing data, missing data were a significant issue for the calculation of complete-case cost and QALY pairs. This is a common issue in cost-effectiveness analyses alongside trials, in which data can be missing at an individual and item non-response level. Therefore, the remaining results (base-case and all sensitivity analyses) refer to analyses conducted on the multiply imputed data set, following best practice economic evaluation methodology. Results of the base-case analysis based on the reported intervention time indicated that, on average, the biofeedback PFMT group was £50 more costly than and had almost equivalent (+0.0009) QALYs to the basic PFMT group. The QALY differences are exceptionally small over a 2-year time horizon and equate to only 8 hours or 0.33 days improvement in quality of life over the trial time horizon. The base-case ICER was £56,617. The probability that biofeedback PFMT would be cost-effective was 48% and 49% at £20,000 and £30,000 WTP for a QALY thresholds, respectively (Table 23). The probability of being cost-effective remains unchanged at much higher WTP thresholds, as illustrated by the CEAC (Figure 7). The uncertainty is further illustrated by the scatterplots of incremental costs and incremental QALYs (Figure 8).

FIGURE 7.

Cost-effectiveness acceptability curves: therapist-reported duration of appointment (imputed data set).

FIGURE 8.

Scatterplot of incremental costs and QALYs for biofeedback PFMT vs. basic PFMT: therapist-reported duration of appointment (imputed data set).

Sensitivity analysis

Recommended appointment time was based on the assumptions that there was no difference in the duration of the appointments and that both biofeedback PFMT and basic PFMT appointments were similar (1 hour for the first appointment and 30 minutes for each subsequent appointment). The results indicate that, on average, the biofeedback PFMT group’s costs were £23 higher than the costs of basic PFMT group, but the biofeedback PFMT group had 0.004 fewer QALYs than the basic PFMT group (Table 24). Biofeedback PFMT group was dominated by the basic PFMT group, as it cost more and was less effective. The probability that biofeedback PFMT was cost-effective was 42% at the £20,000 WTP threshold and 43% at the £30,000 WTP threshold (Figures 9 and 10).

FIGURE 9.

Cost-effectiveness acceptability curves: recommended appointment time (3.5 hours for six appointments).

FIGURE 10.

Scatterplot of incremental costs and QALYs for biofeedback PFMT vs. basic PFMT: recommended appointment time (3.5 hours for six appointments).

Subgroup analysis

The results of the sensitivity analyses exploring the cost-effectiveness of biofeedback PFMT compared with basic PFMT for different subgroups are detailed in Table 25. The results of those participants with SUI were similar to those of the base-case analysis (for sensitivity analyses CEACs and ICER scatterplots, see Appendix 17, Figures 13–32).

Type of incontinence

On average, the biofeedback PFMT intervention cost £152 more and was more effective than basic PFMT. The ICER was £57,936 and the probability that biofeedback PFMT would be considered to be cost-effective was 39% at society’s £20,000 WTP threshold and 44% at the £30,000 threshold. The results of those participants with MUI differed from the base-case results. On average, the biofeedback PFMT intervention cost less and was more effective than basic PFMT. Biofeedback PFMT dominated basic PFMT. The probability that biofeedback PFMT would be considered to be cost-effective was 53% at society’s £20,000 WTP threshold and 52% at the £30,000 threshold.

Age

The results of those participants aged < 50 years were also different from the base-case analysis. On average, the biofeedback PFMT intervention cost more and was less effective than basic PFMT. Biofeedback PFMT was dominated by basic PFMT. The probability that biofeedback PFMT would be considered to be cost-effective was 11% at society’s £20,000 WTP threshold and 12% at the £30,000 threshold. For participants aged ≥ 50 years, biofeedback PFMT cost, on average, £82 more than basic PFMT, and the difference in QALYs was 0.058 (vs. 0.0009 in the base-case analysis). The ICER was £1417. The probability that biofeedback PFMT would be considered to be cost-effective was 92% at society’s £20,000 WTP threshold and 93% at the £30,000 threshold.

Severity of incontinence

The results for those participants with a lower-severity ICIQ-UI SF score (< 13) indicated that, on average, biofeedback PFMT cost £29 more and was 0.026 QALYs more effective than basic PFMT. The ICER was £1397. The probability that biofeedback PFMT would be considered to be cost-effective was 81% at society’s £20,000 WTP threshold and 80% at the £30,000 threshold. The results for those participants with a higher-severity ICIQ-UI SF score (≥ 13) indicated that, on average, biofeedback PFMT cost £59 more and was 0.014 QALYs less effective than basic PFMT. The biofeedback PFMT group was dominated by the basic PFMT group. The probability that biofeedback PFMT would be considered to be cost-effective was 11% society’s £20,000 WTP threshold and 12% at the £30,000 threshold.

Quality-of-life scores

The results of the analysis using the ICIQ-LUTSqol were similar to those of the base-case analysis. On average, biofeedback PFMT cost £50 more and was more effective than basic PFMT. The ICER was £177,668 and the probability that biofeedback PFMT would be considered to be cost-effective was 54% at society’s £20,000 WTP threshold and 57% at the £30,000 threshold.

Discounting and unadjusted costs and quality-adjusted life-years

The results of the discounting and unadjusted analyses differ from the base-case analysis. On average, biofeedback PFMT cost more and was less effective than basic PFMT; therefore, biofeedback PFMT was dominated, although the differences were not statistically significant. The change in results is mainly because the baseline imbalance was maintained, as the data were not adjusted to take into account this difference.

Assuming that data were missing not at random

As shown in Table 25 (also see Appendix 17), this assumption has an impact on the results, in which the missing quality-of-life data are 10% lower in the biofeedback PFMT group only and the QALY difference is statistically significantly lower –0.060 (95% CI –0.117 to –0.003) for the biofeedback PFMT group. Biofeedback PFMT is dominated by basic PFMT as it costs more and is less effective, and the probability that biofeedback PFMT would be considered to be cost-effective was 0.4% at society’s £20,000 WTP for a QALY threshold and 0.4% at the £30,000 WTP threshold. The results are reversed when we assume a reduction of 10% in the missing of quality-of-life data in the basic PFMT group only, as the QALY is statistically significantly higher (0.060, 95% CI –0.04 to –0.115) for the biofeedback PFMT group. The ICER for the biofeedback PFMT group is £843, as it costs more and is more effective, and the probability that biofeedback PFMT would be considered to be cost-effective was 99% at society’s £20,000 WTP for a QALY threshold and 99% at the £30,000 WTP threshold. The rest of the scenario results were the same as those reported when it was assumed that data were missing at random.

Appointment checklists

A protocolised checklist (basic PFMT group or biofeedback PFMT group) was completed by all therapists for each appointment (for examples of core and optional components for each group, see Appendix 18). These checklists were embedded in the TAF completed at each appointment [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)]. Each checklist contained the core and optional intervention components for the appointment, and varied by appointment number (1–6) and by group. For each appointment, the biofeedback PFMT checklists had all the core basic PFMT components and additional biofeedback-related components. Therapists ticked yes or no for each component to allow assessment of fidelity. At the end of the tick-box section, an open-text space enabled therapists to explain the clinical reasoning for any omissions or additions to tailor the intervention based on a participant’s needs.

Appointment audio-recordings

Audio-recordings were planned to obtain a deliberately heterogeneous purposive sample of 100 appointments from across both basic PFMT and biofeedback PFMT groups; from the different appointment time point (1–6); centres and UI type and severity; and different therapist type (physiotherapist/nurse). Challenges with arranging recordings and the increase in the number of centres meant that we subsequently focused sampling on the first and last appointments, as a result of our a priori hypothesis that treatment delivery may be more intensive and concentrated in these appointments. The first appointment was scheduled to be longer (1 hour) and involved assessment (including assessment of correct contraction), education and teaching for participants on how to properly perform PFMEs, use of clinic biofeedback and teaching use of home biofeedback (for the biofeedback PFMT group only) and explanation of any recommended changes to lifestyle. However, as occurs in clinical practice, some of these elements could be deferred to the second appointment based on the individual needs of the woman. The last appointment (30 minutes scheduled) was important because supervised therapy was ending and participants were being given important information to allow them to self-care, such as instructions regarding the maintenance PFMT dose. Appointments 2–5 were also 30 minutes in duration and component content was mostly similar (i.e. repeated) across these four appointments.

If participants had signed on their consent form that they were willing to have an appointment audio-recorded, the researcher purposively selected participants and telephoned them to ask if they were willing to have a specific appointment recorded, answer any questions they had and obtain verbal consent. Therapists were then informed which participants and appointments to record, using a centre-specific password-protected digital audio-recorder; these were returned to the researcher at regular intervals for downloading data. Recordings were anonymised and transcribed verbatim.

Therapist interviews

Semistructured telephone interviews were undertaken with staff at the end of their centre’s participation in the delivery of the trial interventions. Initially, we planned to interview at least one therapist from each centre; however, for various reasons this was not possible and a decision was taken by the process evaluation team to widen recruitment to include at least one person from each centre, but extending the sample to administrative support staff and nurses involved only in recruiting women. In the first instance, the interview researcher e-mailed the therapist(s) at each centre, informing them about the interview purpose and inviting them to participate. Their reply, with a suitable date and time, was accepted as consent to be interviewed. A topic guide was developed from the literature, from issues arising during delivery of the trial interventions and from case study participant interviews. The resultant broad categories of questions were delivery of the interventions to participants; response of participants; maintenance of PFMT; and aspects of theory underlying the intervention. In these categories, the interviewer deliberately sought therapist perspectives on their protocol adherence and participants’ PFMT adherence. Interviews were audio-recorded, anonymised and transcribed verbatim then uploaded to NVivo 11 (QSR International, Warrington UK) for coding.

PFMT exercise diaries

Exercise diaries were provided to each participant at appointments 1–5, to be completed as a record of PFMT undertaken at home and returned at the following appointment. Therapists wrote the agreed PFMT (and biofeedback use) to be followed (the ‘dose’) in the diary and participants were asked to sign their agreement (a BCT called ‘commitment’) to undertake home PFMT and, if allocated, home biofeedback PFMT. At the following appointment, exercise diaries were collected, progress reviewed and a new diary issued. Diaries provided space for participants to record, on a daily basis, the number of sets of exercises undertaken and, if allocated, their use of biofeedback PFMT, along with any free-text comments about their homework.

Appointment checklists

The number of available checklists (i.e. a checklist for all appointments attended) for participants in both groups was compared with the potential number. In each appointment, the number of components checked ‘yes’ were summarised descriptively (means, modes and interquartile ranges) to report the extent to which therapists delivered the core and optional components of each appointment in each group. Therapist free-text comments about intervention omissions, additions and tailoring were coded using a framework developed following content analysis of a 10% representative sample of appointments.

Appointment audio-recordings

A quantitative coding scheme was developed based on the appointment checklists. The coding scheme for each appointment contained explicit guidance for assessing whether or not the protocol components (core and optional) were delivered (yes/no), followed by an assessment of the quality of delivery. The reliability of the coding scheme was checked by comparing the codes independently assigned by three members of the research team to a sample of audio-recordings. Following discussion of differences in coding, the agreed coding scheme was finalised. Using both the audio-recordings and the transcribed audio-recordings, the coding was completed and entered into IBM SPSS Statistics 25 (IBM Corporation, Armonk, NY, USA) for data management. Coded data were summarised descriptively by appointment and by group.

Therapist interviews

Using a general inductive and descriptive qualitative approach, analysis began with familiarisation with the data. A thematic framework56 was developed by one researcher from initial coding of 10 interviews; a second researcher then independently coded four transcripts, and a third researcher cross-checked and agreed the final set of codes. This coding framework was then applied across the data set. Findings are reported in alignment with the sequence of issues discussed in the interviews, with a focus on the therapists’ perspectives on the barriers to and facilitators of intervention fidelity, and participants’ engagement and adherence, to demonstrate variation over time.

PFMT exercise diaries

Content analysis was used to summarise free-text entries made by participants, checking for similarities and differences between groups and appointments. This analysis took place prior to the main trial results being known. Subsequent to knowing the results, we examined the returned diaries for completion of the therapist- and participant-signed PFMT agreement. This BCT was a core component in both trial interventions and was summarised by appointment and by group as a proportion of those attending each appointment.

Appointment checklists

The potential maximum number of checklists was 3600 (600 participants, six appointments each). The number of available checklists by group and by appointment is summarised in Table 26, and does not necessarily reflect how well the checklists were completed by the therapist. Missing checklists include participants who withdrew, participants who did not attend a specific appointment, non-submission of a checklist by a therapist (some forgot to complete it) or a missing TAF. There were no clear differences between groups in the proportion of available checklists, either by specific appointment or over time. There was a steadily decreasing trend from appointment 1 (91% submitted) to 6 (60% submitted).

On the checklists, therapists ticked ‘yes’ or ‘no’ to delivery of each core component for each appointment [for examples, see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)]. Table 27 summarises the number of core components ticked ‘yes’, by group and by appointment. Between the two groups, there were no observed differences in the number of basic core components (i.e. components that were the same in both groups and ticked ‘yes’ by therapists per appointment). Therapists most often ticked either a full complement of core components or just one less than a full complement; this was the case for every appointment apart from appointment 5 in the biofeedback PFMT group, when the mode was only 15 of a possible 18 ticks. The interquartile ranges reveal a skewed data distribution in favour of a high number of basic PFMT core components being ticked ‘yes’ for all appointments and for both groups (see Table 27, columns 2 and 3).

In the biofeedback PFMT group only, therapists most often ticked ‘yes’ to either all the biofeedback core components or one less than a full complement. In all but the last biofeedback PFMT appointment, at least 75% of the checklists had more than half of the core components ticked ‘yes’ (see Table 27, column 4). When combining the basic and biofeedback core components (see Table 27, column 5), it can be seen that the biofeedback PFMT group received more ‘yes’ ticks overall than the basic PFMT group. For example, in appointment 1, the median number of ticks was 18 for the basic PFMT group (out of a possible 19, see Table 27, column 2) and for the biofeedback PFMT group, the median was 26 (out of a possible 28, see Table 27, column 5). This pattern was similar for all appointments, indicating that, based on therapist reporting, the biofeedback PFMT group did have more core components delivered during appointments, consistent with the intervention protocol.

A similar summary for the optional intervention components [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)] revealed a very different pattern: very few were ticked ‘yes’ in any appointment, for either group (Table 28). The mode is consistently zero for all appointments in both groups; the interquartile range is much larger and in only appointment 6 does the median reach half the number of available basic optional components. The optional component pattern is consistent across the appointments and between the groups for the basic PFMT components. The biofeedback PFMT group did receive some additional biofeedback-specific optional components, but relatively few compared with those available on the checklist (see Table 28, column 3).

Appointment audio-recordings

As described in Methods, collecting the full intended sample of audio-recordings was challenging because therapists were unfamiliar with equipment, declined to do their quota because of staff shortages or forgot to record or identify the patient/appointment number, or because participants failed to attend the selected appointment or withdrew from the trial. Table 29 demonstrates our decision to focus on collecting data from appointments 1–6, as appointments 2–5 were designed to repeat intervention content from previous appointments, and, although appointment 2 in the biofeedback group could be when the device was first introduced, it was not feasible to know this in advance when arranging the appointments to be audio-recorded. Although we achieved more recordings in appointments 1 and 6, recruiting to appointment 1 remained difficult, but we achieved a reasonably balanced sample across groups (see Table 29). Overall, the number of audio-recordings made was 88% of the original target.

The number of basic PFMT and biofeedback PFMT core components that were heard being delivered are summarised in Table 30. We agreed that some components were unlikely to be heard and these were not coded; hence, the potential range of components differs from the checklist summaries.

Therapists were heard to use fewer core components, overall, than the number that were potentially audible for coding. This was the case for both groups and across all appointments. The biofeedback PFMT group received more core components (basic and biofeedback combined, see Table 30, column 5) than the basic PFMT group. No participant in the biofeedback PFMT group started biofeedback for the first time in session 2.

A similar summary of audio-recording data was not undertaken for the optional intervention components, given that we knew from the checklist data that therapists were not reporting the use of many of these optional components.

Therapist interviews

We were able to recruit from 21 of 23 centres, obtaining 30 interviews from 26 physiotherapists, one nurse continence specialist (total of 27 therapists delivering the interventions), two nurses and one administrator involved in a variety of tasks, including participant recruitment, obtaining consent and dealing with IT issues.

We report therapist perceptions of recruitment and delivering the trial overall, their perceptions of both basic and biofeedback PFMT interventions and any differences between them. We pay particular attention to therapists’ perspectives on (1) their protocol fidelity and (2) participants’ intervention adherence.

Participant diaries

Although the number of returned diaries dropped from appointment 2 to 6, in much the same way appointment attendance dropped (see Table 30), there were no apparent differences in the pattern between the two groups (Figure 11). Nor were there apparent differences in the proportion that were signed by participant and therapist (a BCT called ‘commitment’).

FIGURE 11.

Proportion of exercise diaries signed (by participant and therapist) and returned by appointment and by group (in respect to the total number attending that appointment in that group).

In free-text diary entries from participants, the most frequent reasons for being unable to exercise were time (biofeedback PFMT = 13 comments, basic PFMT = 33 comments), forgetting (biofeedback PFMT = 7 comments, basic PFMT = 24 comments) and other physical health reasons (biofeedback PFMT = 25 comments, basic PFMT = 13 comments), with menstruation featuring more frequently in the biofeedback PFMT (14 comments) than the basic PFMT group (6 comments) (participants were not expected to use biofeedback during this time).

Sampling and recruitment

Forty randomised women (20 in each group) were purposively sampled for variance in centre type, women’s type of UI and therapist type. Each recruited woman was one case. Women were asked to consent to the case study specifically (having already consented to take part in the trial). The women were given an additional invitation letter [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)] and patient information leaflet [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)]. Women who remained interested were contacted by telephone approximately 1 week later to ask if they would like to participate. Written consent was obtained at the time of the first interview [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)].

Case study data collection

Data were collected by a series of semistructured interviews [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)]. Each interview had a specific focus:

• Baseline pre-treatment interviews (face to face) explored the woman’s experience of UI, the social contexts within which she experienced UI and her expectations of treatment.

• A 6-month post-treatment interview (face to face) explored the woman’s experience of the trial intervention, her adherence to therapy appointments and the prescribed programme, factors that affected that adherence and her perceptions of treatment outcome.

• 12- and 24-month interviews (telephone) explored, at each time point, the woman’s experience of UI post intervention, the intervention, factors that influence ongoing PFMT adherence and treatment outcome.

Interview data were, with consent, collected using a password-protected audio digital recorder. Interview audio-recordings were anonymised, transcribed verbatim and entered into NVivo software to support analysis.

Case study data analysis

Analysis was guided by the OPAL trial protocol [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)] and the OPAL qualitative study and process evaluation analysis plan [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)]. Three different researchers have worked on the OPAL case study (Anne Taylor, Aileen Grant and Marija Kovandzic), alongside the responsible grant holders (Carol Bugge, Jean Hay-Smith and Sarah Dean). By the nature of qualitative analysis, each analyst had a different approach to data analysis. This was encouraged by the grant holders, within the confines of the protocol, to maximise the insights into the data. Sources that were drawn on to support that analysis included Yin,60,61 Alvesson and Sköldberg,62 Grant et al.,63 Kovandžić et al.,64 Stake65 and Ritchie et al. 66

Overall, analysis was iterative with data collection. Analysis occurred on four interacting levels to facilitate within- and cross-case comparisons.

Management and governance

Ethics approval for the case study was gained within the main trial approvals (see Chapter 2).

The case study and process evaluation team had a management group with the required mix of clinical, qualitative, quantitative and theoretical skills and experience. The group met regularly to discuss the research management and emerging findings. The case study was carried out at a separate academic institution to the main trial. The case study team participated in trial meetings to understand how the trial was progressing, but the case study and process evaluation team meetings were closed. Data were not shared from the case study and process evaluation group with the main trial group until the final PMG meeting in September 2018.

Sample

Forty women, 20 per group, were recruited to the case study, as planned. Twenty-five women completed all four interviews, but, owing to the technical problems with the audio-recorder, a full data set was available for only 24 women (10 biofeedback PFMT and 14 basic PFMT). The total data set consisted of 125 interviews, including 24 complete cases (96 interviews). The total number of minutes of recorded interviews per case ranged from 15 minutes to 126 minutes, with a total of 2856 minutes of recorded interview data. There were 40 baseline interviews (20 biofeedback PFMT and 20 basic PFMT), 32 interviews at 6 months (16 biofeedback PFMT and 16 basic PFMT), 28 interviews at 12 months (13 biofeedback PFMT and 15 basic PFMT) and 25 interviews at 24 months (11 biofeedback PFMT and 14 basic PFMT).

The age of women in the case study ranged from 20 to 76 years, with both the biofeedback PFMT and the basic PFMT groups including women with a wide age range (Table 31). In the main trial, women ranged in age from 20 to 83 years (22–83 years in the biofeedback PFMT group and 20–78 years in the basic PFMT group); thus, the women in the case study were comparable in age to the main trial sample. From the total case study sample, 11 women had SUI and 29 MUI; the proportions were similar within groups. Six women in the sample were treated in community clinics, 16 in university hospitals and 18 in district general hospitals; again, there were similar proportions in the groups. The vast majority of women were treated by physiotherapists (n = 36) and four women were treated by nurses.

Women’s adherence to the interventions

Women’s adherence to the interventions was analysed in two phases: active treatment and maintenance. ‘Active treatment’ refers to the time when women were attending appointments and receiving the OPAL trial interventions delivered by a trained therapist. It is the proxy for shorter-term adherence – the uptake and adoption phase of PFMT – including women’s attendance at appointments, receiving biofeedback-mediated PFMT or basic PFMT in the clinic and then undertaking their prescribed programme (biofeedback PFMT or basic PFMT) at home between appointments. The ‘maintenance’ phase is when long-term adherence is demonstrated and is the period of time after the active treatment has ended when women were asked to continue PFMT themselves at home without therapist supervision, including relapse management, up to their final follow-up at 24 months.

Table 32 shows examples of the variation in women’s adherence to PFMT. These examples illustrate that there are no obvious group differences in adherence in the case study sample in terms of the frequency with which they undertook biofeedback PFMT or basic PFMT.

Facilitators of adherence during the active treatment phase

There was greater similarity than difference in facilitators of adherence in the active treatment phase when the trial groups were compared. Two key themes, among the many that were identified, focused on UI symptoms and factors related to the OPAL trial therapist.

Urinary incontinence symptoms acted as a facilitator in several ways. One way was through the mechanism of women wanting to eliminate or reduce their UI, so that they could get on with their lives and improve their quality of life:

Well I’m hoping that it’ll help the leaking and it’ll, it might never stop, but it won’t be as bad as it’s been . . . that’s what I’m hoping.

Yeah. Is there a goal; do you have, like, a personal goal that you would like?

Just that really, just . . . to stop the leaking, maybe be able to go back to yoga and not feel like I’m worrying about leaking or whatever.

I’m not that old that I, I’m ready to kind of hang up my dancing shoes.

Case 26, 0M, basic

Women also wanted to prevent a deterioration in their UI symptoms and to avoid surgery. Seeing an improvement in UI during the active treatment phase motivated women to adhere because they felt that their treatment, and their skill to undertake the exercise, was working:

Doing the exercises [was most helpful about treatment] and noticing that there was a change, do you know what I mean? And then realising myself that that was, there had been a change . . .

Case 24, 6M, basic

For women from both groups who had a break in their regular biofeedback PFMT or basic PFMT practice, a deterioration in symptoms (after a period of improvement) provided proof of PFMT effectiveness and acted as a facilitator to use the skills that they had learned to overcome the symptomatic deterioration.

Many women from both groups talked at length about the positive impact of the therapist. Women talked about their therapist as an important and credible source of information, as a motivator and as someone who taught them the exercise, lifestyle and behavioural skills needed to undertake biofeedback PFMT or basic PFMT (in line with theoretical model underlying the interventions). 17 All of these factors influenced adherence in the active treatment phase in both groups. However, possibly the most important element of the interventions in each trial group was the instruction on how to perform PFMEs, given by the therapist during the vaginal examination (digital assessment). Given the sensitivity of the topic, vaginal examination was not easy to talk about during the research interviews and, consequently, not an easy finding to capture in the analysis. Yet there was a consistent observation of the importance of the therapist-mediated vaginal feedback as being one of two distinctive and valuable forms of vaginal feedback in PFMT (therapist mediated and EMG mediated). The findings from the case study point to the therapist-mediated feedback as being the priority and as one of the most important therapist-related facilitators in gaining confidence in PFMT skills and adhering to treatment.

The quotation below provides an illustration of the difficulties of articulating experience of PFME instructions during the vaginal examination, as well as the importance of these instructions, which included feedback (as exemplified in the quotation, a part of the feedback loop was the act of the therapist feeling the difference in muscular activity during the examination):

That was quite good actually, having somebody there, and I think when you’re doing exercises and then being able tae feel that it was working, do you know that way when you would get your assessment . . . and you did have to do them, the exercises, and she could feel the, the difference [ . . . ] I felt [ . . . ] that was good, u-uuh, just to know that you were doing it properly [ . . . ] ’cause you do those exercises and you really don’t know one way or another if you are doing it right.

Case 30, 6M, biofeedback

Another important therapist-related factor that had an impact on adherence was the rapport created between the therapist and the woman. The conditions for creating rapport require further analysis. It is possible that the above-mentioned therapist-mediated vaginal feedback plays a role, but at this point of analysis it is certain that having dedicated space and time (secured by OPAL trial intervention design) to build understanding and trust through repeated appointments with the same therapist acted as a motivator to adhere to the treatment, if not being a therapeutic agent on its own:

And it’s very motivating . . . you know, seeing someone who’s interested in you, who wants to help you is terribly motivating . . . ‘cause otherwise you’re just on your own, ‘cause you don’t chat to your friends about it . . . the only person I’ve ever really spoken to about this [UI] is [OPAL therapist] and the nurse specialist.

Case 32, 6M, biofeedback

[ . . . ] it was good having a one to one with someone who kind of constantly, you were able to talk to about your symptoms and how to improve it and I think just knowing that em that they were there and they were able to tell you, you know, ‘if you work on this, it will improve’ and I think that was a big help, even right at the end there, it was a really good for her to tell me, the physio[therapist] to tell me, like, what exercises it’s best for you to do, what’s not good for you to do and if you keep going wi’ this it’s going to continue to improve, I think that will help me. [ . . . ] you know, psychologically, even if it wasn’t physically, you know, I mean it eventually will be physically, but em, you know, even psychologically I think that was good.

Case 15, 6M, basic

Other therapist-related facilitators included education provided by the therapist, being treated by an accommodating and skilful therapist, being treated by a therapist who adjusted the treatment protocol based on individual needs and feeling accountable to the therapist:

I think it was the, the, eh what’s the, what’s the best way to describe it, the actual having to report back to [therapist], because then you knew, you know, you can’t, well you can’t just sort of, you know, sit there and say ‘right, OK I didn’t do it,’ and she would know herself when we did the sort of, the few, even, you know, not the internal examination, when we did the actual work, you know, when she was there and she could tell from my posture, you know, if I was doing it right or not, she was like ‘right, you’re slacking’ . . .

Case 3, 24M, basic

Beyond the symptom- and therapist-related facilitators summarised above, women identified other facilitators of adherence that included the following:

• Service structure, framing and physical environment. Having regular appointments; ease and flexibility of making appointments; feeling positive about the physical environment of the treatment facilities; feeling that the intervention was within the framework of womanhood; or the woman finding the treatment as a whole a novelty were all facilitators of adherence.

• Ownership and agency in the intervention. In terms of ownership and agency, the following were seen as facilitators of adherence: the woman feeling accountable to herself; having a new focus on herself through starting a new lifestyle routine; developing a sense of embodiment and empowerment as starting to feel (and control) her own pelvic floor muscles; and through their own personal characteristics, such as ‘strong will’, commitment and altruism:Case 20, 24M, basic

  I was so determined though, I mean the thing is you’ve got to want to, to help yourself I think [ . . . ] you know, it’s just not going to, just taking a note of what somebody says to do, you’ve got to want to do it as well [ . . . ] you’ve got to want to, you’ve got to need to do it as well, you know.

• Support from relevant others. Their partner, participation in the trial and a sense of accountability to the trial team were all facilitators of adherence.

• The intervention. In terms of the intervention, the following were seen as facilitators of adherence: making use of the exercise diary; a woman gaining knowledge about her anatomy, UI and PFMT; learning the skills to undertake biofeedback PFMT or basic PFMT; learning routines for biofeedback PFMT or basic PFMT (such as keeping the biofeedback unit next to her bed) or doing PFMT while watching TV; and learning that PFMT exercise was easy and could be performed at any place and any time:Case 13, 6M, biofeedbackCase 34, 12M, basic

  . . . so the education was eh the principal thing, when you learn how to do and why it’s wrong, what is wrong . . . and then you can do your, do good for your body.

  . . . it’s probably the easiest form o’ exercise you could do, I mean you don’t even need tae go tae a gym, it’s so easy.

There were facilitators that were specific to biofeedback. Some women reported liking the biofeedback device and having confidence that, by using biofeedback PFMT, they were more likely to achieve symptomatic improvement than if they were doing PFMT alone. In developing the OPAL trial intervention, the research team hypothesised that visualisation of the pelvic floor muscle contraction via biofeedback PFMT would support self-efficacy for performing the correct contraction, leading to improved adherence and better outcomes. Some women in the biofeedback PFMT group did report that visualisation was important for them for two main reasons: (1) they could see if they were doing the pelvic floor muscle contraction correctly and (2) they could see improvement in their pelvic floor muscle contraction ability over time. Women valued the opportunity to be able to discuss the visualised contraction with their therapist.

Other features of biofeedback PFMT that women valued were biofeedback supporting women being competitive with themselves; having a new ‘toy’ to play with; the physical presence of the unit acting as a reminder; getting instruction from the biofeedback device in terms of counting of repetitions and pace of PFMT; and an awareness that the data on the biofeedback device would be looked at by and discussed with the therapist:

I thought it was quite positive that when you were actually using it you could see, and I think it did make you try, it definitely made me try harder, and also I felt that I was doing it for longer, like it, you know, a 10-second hold I think when you haven’t got the biofeedback is probably, in reality, an 8-second hold, because you count quite quickly . . . whereas with the biofeedback I felt that you were doing it properly and I was definitely trying harder because I was seeing it and I was thinking ‘right, I want’ it’s that sort of slightly competitive side to human nature, you’re thinking ‘right, I want to get, I want to get it higher’.

Case 8, 6M, biofeedback

In summary, although some group differences were noted, there were more similarities in facilitators of adherence than differences. There were many facilitators of adherence in the active treatment phase, with being motivated to improve symptoms and the effect of the therapist being clear facilitators in both groups.

Barriers to intervention adherence during active treatment phase

There were more similarities across barriers than there were differences between the groups. Time and contextual factors in a woman’s life (such as daily routines) were two of the themes that could be seen to act as barriers to adherence.

Women talked about having a lack of time for themselves; hence, finding time for appointments and to exercise was difficult. Women reported a lack of time to attend appointments in general and frequent appointments in particular, to focus on practising PFMT, either with or without biofeedback; biofeedback was even more time-demanding and, as such, a potentially greater barrier in the biofeedback group:

I don’t know who supplied the physio[therapist] with the dates, but she kinda had a calendar, at the end o’ my appointment she could tell me the time frame when I was due back . . . and I would look at my diary and was like ‘oh, that’s only like 2 weeks’ time’, so I don’t know, maybe even once every 5, 6 weeks or something, em but that, again, that’s just because I’m a working mum and I don’t always have the child care, so em it wasn’t always easy for me to, to get the kids watched, . . .

Case 16, 6M, biofeedback

Lack of time was compounded by having a generally busy life that included being a working mother, having unpredictable work patterns and going on holiday. For several women, their UI, and its treatment, was not a priority given the array of other things that were competing for their time. Illness – theirs, or in family members – was a particular barrier to adherence:

Em . . . most of the time I’m OK now, as I say I still do my pelvic floor exercise at the moment, eh it’s not always OK, but [most of them are?], em [sighs] that’s nothing to do wi’ the machine [?] that I dropped out [of treatment], I took, mum took no’ well and I took really bad depression and I would’nae get out the bed.

Case 17, 6M, biofeedback

Other contextual factors that acted to diminish adherence included not having a routine (or hook) for doing PFMT, lack of privacy at home, lack of support from their partner and simply forgetting (in the array of other things to do).

Several other barriers could also be identified, these included the following:

• A lack of sufficient, or sufficiently quick, improvement in the UI symptoms. This led to a drop in motivation to adhere. Despite this drop in motivation, many women were still inclined to continue treatment.

• There were service factors that acted as a barrier to adherence, some of which were OPAL trial-specific and others which were not. These included general transport or parking problems; the treatment facilities being unappealing or inappropriate; a lack of suitable appointments; and, although uncommon, a feeling that staff were being difficult, ‘one of the receptionists at the [clinic] is an absolutely [sic] dragon’ (Case 14, 6M, basic). Women also reported anxiety and/or embarrassment about having to undergo vaginal examination; being put off by other trial requirements, such as completion of the bladder diaries; not being randomised to their preferred group; and, in a few cases, there were indications of rapport and/or satisfaction with the therapist not being as good as for the majority of other women. Some women commented that there was repetition between the visits and that this made them unsure about the value of continued appointments:Case 15, 6M, basic

  Yeah, there seemed to be quite a lot, you know, I seemed to have a lot of appointments, em . . . my husband’s going ‘oh you’re not going there again, what are you going for, what on earth are you going for this time?’ . . . Em maybe the odd time I did feel a bit like that ‘cause I felt, u-uuh, at times I thought ‘oh God here, we’re just going to talk about exercises’ [ . . . ] the odd time I did feel ‘gosh, maybe that was a bit of a waste of time’ [slight laugh] . . .

• A few women in both groups felt that they did not learn the correct PFMT technique in the clinic or how and when it should be applied, which could manifest as a lack of confidence about doing the exercises correctly. It is possible that this lack of confidence is linked to not receiving enough therapist-mediated vaginal feedback:Case 26, 6M, basic

  I thought there would be, I thought there would maybe be more em involved in helping support you doing the actual . . . exercises; not that they were difficult or anything like that, I just, I, I think I just felt, you know, you get told what to do, you’re advised about what, how to do them, they don’t, [sighs] I’ve only once been checked to make sure I was doing them right, em so my feeling kinda was am I doing these right? Are they really effective?, and it was a bit hit and miss I felt . . . to how well I was doing; . . .

There were some barriers that were specific to the biofeedback PFMT group. Some women found the biofeedback device intrusive or painful to use and others found it inconvenient (e.g. having to set it up, or to clean it):

I found it intrusive and painful to be honest [ . . . ] if I had of [sic] found it less uncomfortable it possibly would have made me notice what I was doing more, but I, I just couldn’t put up [with] the, the pain of it, so I couldn’t be bothered with it.

Case 5, 24M, biofeedback

Women reported that they needed to find even more time to undertake PFMT supported by biofeedback. Some women also reported embarrassment and a lack of privacy about using biofeedback:

I think it was quite a good idea, but I don’t think it worked for me, for my personal circumstances, I found it too footery [fiddly] to do, and I just found it quite difficult to have that kind of privacy . . . just to do it, because I found it easier if I was lying down in the bedroom but then, you know, the kids were always like in and out, running around and obviously I didn’t want them to see it, and I just felt it took quite a lot of time and I just felt I didn’t really have the privacy to do it properly, em, so I don’t think it really worked for me, I felt it was too footery; but on the other hand I think it had lots of advantages, ‘cause I think it was quite useful to see, to see what was actually happening.

Case 8, 24M, biofeedback

Other issues with the biofeedback included one woman reporting that she got thrush from using the biofeedback unit; the biofeedback unit could be framed as externalising the movement of the pelvic floor muscles and a distraction to embodiment of PFMT; and practical problems with the biofeedback unit that hampered ability to use it:

I thought I was doing super, then one day it died and it, I knew it had a brand new battery so that shouldn’t have happened . . . it died, so I rang them up and I took it in and we got a new battery, then I came back and it happened again, it kept doing weird things, and then I bought batteries up the road in the end, so, . . . And then I realised that by looking at the machine I was distracted from doing the exercises.

Case 32, 6M, biofeedback

In summary, there were more similarities than differences in barriers to adherence in the active treatment phase; there were also additional barriers in the biofeedback PFMT group. A lack of time and many contextual factors were the key barriers to adherence to biofeedback PFMT and basic PFMT.

Facilitators of women’s adherence in the maintenance phase

None of the women in the biofeedback PFMT group reported using biofeedback after the end of treatment in the trial. None of the interviewed women reported buying biofeedback equipment; some therapists did give women the probe to keep and use, yet none of the women reported using it, even though some reported intention to use it. Thus, the data below relate to women, from both groups, undertaking basic PFMT in the maintenance phase.

Women in both groups reported a change in their adherence from the active treatment phase. PFMT maintenance was not consistent over time in either group and there were no differences (from qualitative comparison) between the groups in their adherence. The inconsistency in adherence between women can be seen in Table 32, in which, at the extremes, some women undertook PFMT in a regular and daily manner, whereas others did not do PFMT at all. In between these extremes were women who undertook PFMT with varying degrees of regularity. As well as the inconsistency between different women, there were fluctuations in adherence for individual women over the time period with, for example, other health concerns taking over and diminishing adherence at some points in time.

Many of the facilitators that applied when women were in the active phase of treatment also applied in the maintenance phase.

Similar to the active treatment phase, women’s desire to lessen UI symptoms supported adherence to PFMT in the maintenance phase. If women perceived symptom deterioration or recurrence and associated this with PFMT as a mechanism to improve symptoms, adherence was facilitated. The interpretation of the data would suggest that symptoms may only act as a prompt to undertake PFMT in the maintenance phase if the woman perceived that there was an improvement in symptoms as a consequence of doing PFMT during the active treatment:

Not really no [been doing PFMT], but quite often in the last week, ‘cause I’ve noticed a difference that’s why I’ve sort of started to try and do it again, ‘cause I have noticed a difference in not doing it . . .

Case 8, 6M, biofeedback

Oh yes, I always will [exercise] now, that’s it . . . that’s it, because I know it, I know it’s, you know how much it’s helped.

Case 20, 24M, basic

There were multiple factors that seemed to influence women’s confidence (self-efficacy) to continue, or feel able to restart, PFMT in the maintenance phase. Many women reported feeling that they had good levels of knowledge and skill to undertake PFMT correctly. Beliefs in their skills and knowledge could be attributed to women feeling they had mastery of PFMT; having memories of the support they received from the therapist during active treatment; recalling information imparted by the therapist; using the resources given by the therapist (such as information leaflets); keeping a record of PFMT like an exercise diary; and recalling the sense of hope given during treatment and the control they gained:

I don’t feel like I need to go back and see a doctor or, you know, see a nurse or anything, I feel like if it got bad again I could, you know, I’ve got these exercises to fall back on.

Case 27, 24M, biofeedback

I remember the girl who, or the nurse that, the lady, you know the . . . pelvic floor . . . in [location], and I remember her, she was very good, gave me a lot of confidence in myself, you know and . . . it was really good, she was very, very helpful, and I can remember, I can remember the improvement, . . .

Case 20, 24M, basic

In the biofeedback PFMT group some women related having good skills and knowledge of PFMT directly to biofeedback during active treatment. Women in the basic group also felt that they had good skills and knowledge of PFMT acquired from teaching by, and feedback from, therapists. Therefore, biofeedback was not a necessary prerequisite for having skills and knowledge for PFMT maintenance:

[I remember] learning to use the machine properly . . . knowing I was doing it right and . . . yeah, and just generally being made more aware of the muscles that you need to squeeze and . . . when you’re, and you know you do one at a time and then you hold them all . . . so yeah . . . being taught how to do pelvic floor . . . muscle training . . . yeah, being taught that properly, yeah, . . . made a big difference.

Case 23, 24M, biofeedback

Other factors that facilitated adherence in the maintenance phase included the following:

• A supportive home environment.

• Establishing the intervention as part of life: being able to find time for themselves; helpful work patterns (e.g. working from home or time spent commuting); making use of available time to do exercises (e.g. sitting and waiting time such as while commuting, or watching television).

• Development of routines or ‘hooks’ that continued to supported continued PFMT. In the example below the woman refers to keeping note of her exercising (like in an exercise diary):Case 24, 12M, basic

  I don’t think so, I mean I know what to do, I mean it’s . . . I, I, aye, I know what tae dae and I know what I should be doing but it’s just the getting intae it, so . . . maybe I should start up a wee book kinda thing again, I done that the last time and I was dain’ wee [unclear word] like when, how many I had done, do you know, how many kinda exercises I’d done that day, and eh, do you know what I mean, I think I should start that, when it’s doon in black and white sometimes that kinda, kinda motivates you oan . . .

• Ownership and agency in the intervention – motivation and determination; cognitive features, such as remembering to do PFMT; and framing PFMT as ‘control’ (e.g. PFMT as a tool to control her body, or PFMT and the control over UI it gives being the only part of her life she can regain control of):Case 17, 24M, biofeedback

  Oh probably [doing PFMT] daily, ‘cause I do sort o’ try to keep it going . . . ‘cause I’ve got to keep control o’ something, I can’nae control everything else [that I’ve got?] . . . [I was more?] conscious of it then, but, as I said, it’s one thing I’m sort of trying to keep control of . . . so I’ll try and keep that bit going.

• Trial-specific factors – research interviews acting as a trigger to undertake PFMT, interviews providing a space for reflection on a woman’s own PFMT practice, attending the 6-month pelvic floor assessment and wanting to demonstrate that the therapist had done an excellent job. These factors occurred in both groups.

• Women in both groups reported using an application on their phone that supported doing/remembering to do PFMT:Case 19, 24M, basic

  I [got] the squeeze App on my phone and that was really good . . . you know, it helps you, you obviously train yourself to hold for longer [kind of thing], that was good.

In summary, adherence did change in the maintenance phase from the active treatment phase. There was considerable variance, in individual women and between women, in adherence in the longer term. Many of the facilitators that supported women in adhering in the active treatment phase continued to facilitate adherence in the maintenance phase.

Barriers to women’s adherence in the maintenance phase

One barrier to adherence that was unique to the maintenance phase was the loss of therapist support, and accountability to the therapist, when the active treatment phase ended. Some women felt ‘alone’ in their efforts to improve their UI. Others expressed the view that, because they were no longer accountable to the therapist, there was no longer that prompt to exercise. Other women said that they got out of the habit of writing their exercises down (as they would have done in the exercise diary during active treatment):

I thought, you know when the nurse did it with me, you know, did it, that helped me a lot, really it did, if I could keep going to the physiotherapist and if she kept checking me, because I think, you think you’re doing it right and then I could be doing it wrong and that, you know what I mean, I mightn’t be feeling . . . Yeah, I mean I would have liked then to be able to phone up, you know, the physio[therapist] and say ‘look, can I have another appointment?’, rather than the length of time between each, and then of course it stopped for so many months . . .

Case 6, 12M, basic

Otherwise, the main barriers for women in maintaining PFMT, whether allocated to the biofeedback PFMT or basic PFMT group, were similar to those found in the active treatment phase. First, some women’s UI had improved to such an extent that they had no symptoms to act as a reminder to exercise:

. . . as I said my symptoms have reduced so there’s not so much of a physical reminder that ‘oh, I need to do them’ [PFMT].

Case 28, 12M, biofeedback

The second key barrier was the loss of motivation or loss of the habit of doing PFMT due to life events taking over, even if this was contrary to the intent they had at the end of active treatment. Women spoke of various contextual factors in their lives that prevented them from maintaining a PFMT regime, such as having too many other things to do, work commitments or work changes getting in the way, or more generally feeling that they lacked support. Commonly, women talked of having other non-UI health problems that overshadowed their focus on UI and/or on their attention being more on the needs of others (commonly immediate family). In keeping with the active treatment phase, the findings suggest an interaction between a lack of time (e.g. as shown below, women not having time for themselves) and the multiple other contextual factors that get in the way of life:

Well I’ve had a lot o’ other health issues so it’s kinda been, that’s [PFMT] been the least o’ my worries [UI] tae be honest wi’ yae [slight laugh].

Case 10, 24M, basic

. . . it really is down to me . . . I expect you hear that from a lot of women . . . And it’s very hard to put yourself at the top of your own time agenda . . .

As women . . .

Yeah, yeah [talks about her husband exercising every day no matter what] . . . So, but with me something seems to come up, [then it’s?] all my stuff goes to pot on my own agenda . . ., I suppose it’s just, [he’s] not easily distracted but there are more pressures on me . . . and I think it’s probably the same for women generally.

Other factors that acted as barriers to adherence in the maintenance phase were as follows:

• Not establishing PFMT as part of life. Some women found maintaining a PFMT exercise programme to be difficult because they had no routine in life generally, or for PFMT specifically, or that routine changed (e.g. going on holiday).

• Not feeling confident in their PFMT technique and when to use it after treatment had stopped. This seemed to manifest as a lack of confidence in (1) their ability to undertake PFMT generally or (2) how to get restarted after a break in PFMT. Various reasons can be identified for this lack of confidence in the maintenance phase: forgetting what they were taught, not feeling that PFMT was going to work, not perceiving that their UI was caused by pelvic floor weakness (it was caused by something else) or because they had not seen symptomatic improvement during active treatment. However, there was a stronger pattern for women to feel confident about continuing PFMT than not having the confidence to continue.

• Ownership and agency in PFMT. Some women talked about a lack of motivation and willpower, they talked about forgetting to exercise (sometimes or always), and PFMT lost the novelty factor and priority over time.

In summary, large-scale systematic differences between the biofeedback PFMT and basic PFMT groups in barriers to PFMT maintenance were not evident from the data set. Key barriers to maintaining PFMT lay in loss of support following the active treatment phase and busy lives.

Women’s urinary incontinence outcomes in the short and long term

The case study did not set out to explore outcome, but women discussed outcome as part of their experience. Given the longitudinal case study design, and the core aim of the trial, it was useful to consider women’s views of outcome in this chapter in relation to UI symptoms. However, interviewed women reported outcomes that were considerably broader than UI symptoms alone. For example, the women talked about changes they made to their lifestyle, changes to their feelings about UI and about a myriad of things they had learned from being part of the trial. These additional outcomes will be documented in more detail in future publications.

At 24 months (when the primary outcome was measured in the trial) there was no obvious difference between the groups in UI severity from qualitative comparison; rather, there were women in both groups with varying outcomes (Table 33).

In both the biofeedback PFMT and basic PFMT groups there were more women talking about positive outcomes in relation to their UI symptoms at 24 months than there were talking about poor outcomes (i.e. from baseline it seemed as if women tended to be better than they were before they entered the trial). This information, however, needs to be considered with caution, as qualitative studies do not aim to statistically generalise:

I was just going to say well no, thank you for the opportunity because I’ve seen a massive, you know, improvement and because I’ve got a prolapse and obviously, I’m quite young, I’m only 38, it was making me sort of anxious about [?] and you know, everything has improved, my bladder control and my prolapse symptoms have improved, I’m not getting as many em, I used to get sort of quite a lot of dragging sort of tummy ache [muscle, or little,?] and I don’t get that any more, so, you know, and I know that that is definitely all down to the trial, if I wouldn’t have been involved in that, then I know that I’d still be having the problems and still be anxious, you know, if I went out walking or if I went, went running, or to the gym or whatever, so, so I’d like to say thank you to you guys as well.

Case 28, 24M, biofeedback

In terms of short-term outcomes, in both groups there was a pattern that suggested that women were likely to have better UI outcomes at 6 months (immediately post-active treatment phase) than at 24 months. For example, case 13 (biofeedback PFMT) reported symptomatic improvement at 6 and 12 months, but at 24 months reported that her symptoms were the same or a little worse than when she started the trial. There was, however, variance between individuals. For example, for case 32 (biofeedback PFMT), there was no improvement noted at 6 months and at 24 months her symptoms were worse than when she started the trial. There were other cases when improvements occurred beyond 6 months (i.e. 6 months was not the best outcome point). For example, case 36 (basic PFMT) reported good improvement at 6 months, further improvement at 12 months and yet further improvement at 24 months.

In summary, there were no obvious differences in UI outcome between the trial groups.

Theoretical propositions

Two theoretical propositions and one rival explanation were considered. The theoretical propositions were driven by the theory that supported the hypothesised mechanism of action (propositions 1 and 2) and one rival explanation that arose from analysis of the data (proposition 3).

Strengths and limitations of the case study

A key strength of the case study, and qualitative research linked to trials in general, is that it facilitated the voice of those whom the intervention aimed to help to be heard and represented. The longitudinal nature of the case study, with detailed follow-up at the same time points as the trial, and the purposeful searching for expansion on emerging ideas at subsequent interviews, allowed consideration of women’s expressions of adherence to PFMT over time. Studies of long-term adherence in UI are rare (only one other longitudinal study,68 with women who have UI, has been identified), but are important as reduced adherence is a common explanation for why treatment effect is not sustained over time. 69 The two-tailed case study design offers a robust, qualitative means of comparison that supports the comparison in the trial.

The process evaluation and case study drew on a contemporary published framework in further developing the work and developing the analysis plan. 63 That framework proposes multiple candidate approaches to understanding various features of the trial and its effects. However, one weakness was that data were not gathered on all the candidate approaches. 63 However, data were gathered on several candidate approaches that were central to the research questions, such as maintenance. Another potential weakness of the case study in relation to the trial is that the interviews may have acted as a co-intervention to promote adherence, for example women reflected that they undertook PFMT because they knew that an interview was coming up. However, the case study recruited women from both the biofeedback PFMT and basic PFMT groups and any effect of the interviews on adherence potentially occurred equally in both groups.

Comparison of findings to existing literature

The evidence from the case study is consistent with the trial finding that biofeedback PFMT did not improve UI outcomes more than basic PFMT. Insights from the case study are helpful in explaining the main trial finding. The qualitative data demonstrated that biofeedback could work as anticipated, with women reporting the benefits of being able to visualise the contraction and know that they were doing the contraction correctly, alongside their learning in partnership with the therapist. However, women in the basic PFMT group also had confidence in their ability to undertake PFMT. For these women, this was based on learning in partnership with the therapist. A possible conclusion is, therefore, that biofeedback does not need to be added to a strong basic PFMT programme in order for women to achieve self-efficacy for and adherence to PFMT; good therapist input can also provide self-efficacy and adherence.

Aspects that are central to this conclusion are that the OPAL trial basic PFMT (and biofeedback PFMT) programmes allowed sufficient time with therapists to support a treatment effect;70 both interventions were based on BCTs;8 both demonstrated that some women achieved self-efficacy for PFMT;71 both groups received therapist-mediated vaginal feedback; and one group also received biofeedback. 7 Although it is possible that if biofeedback PFMT had been compared with a less robust basic PFMT programme there would have be been a difference between groups, it would then have been difficult to reach conclusions about the effectiveness of adding biofeedback because of other confounding variables. 7 Our conclusions therefore support a finding that, if all other aspects of PFMT are kept equal, the addition of biofeedback may not lead to a greater improvement in continence outcomes.

Another possible explanation for why biofeedback PFMT was not more effective than the basic PFMT is that, although women in the biofeedback PFMT group did identify features of biofeedback as facilitators of adherence, they also identified features of biofeedback as barriers to adherence. One tentative hypothesis here is that the facilitators and barriers simply cancel one another out. However, this needs further analysis.

Case study findings demonstrated that women generally reported positive experiences of the OPAL trial interventions. Women were positive about learning to do PFMT (with or without biofeedback), which is consistent with a previous qualitative synthesis. 18 Women were also very positive about the therapists. Women talked about the therapist in ways that suggested that the therapist was seen as a credible source of information, a motivator and as someone who could support the learning of the necessary behavioural skills, all of which supports the theory underlying the development of the interventions (IMB17). Furthermore, in keeping with previous suggestions, rapport between woman and therapist was seen as a factor in supporting adherence. 18

There was a trend identified in the case study data for women to perceive that their UI was better at 2 years than it was when they started the trial. This was not the case for all women. It was, however, an important finding in the context of the worldwide evidence that UI negatively affects women’s day-to-day lives (see, for example, Bradway,72 Delarmelindo Rde et al. 73 and Hamid et al. 74). Although it is possible that the improvement described by women is not linked to the interventions, the evidence suggests that women did perceive a link between the intervention they received, their adherence to PFMT and their positive outcome. This link will be explored in more detail in further analysis.

Adherence to PFMT did change over time, but not differently between the allocated groups. A key reason for including the case study alongside the trial was the recognition of the influence of context on the effectiveness of complex interventions. 67,75 It is now widely recognised that context interacts, modifies, shapes and constrains the intervention and implementation. 76 This study chose to investigate the influence of context in-depth from the participants’ perspectives (rather than also exploring the problem, trial and organisational contexts), because it was believed these would be the most important factors to shape the interventions and influence their effectiveness. It is important to understand the dynamic relationship between context, implementation and intervention to define what was implemented and understand how works in certain contexts. It is now no longer enough to say what works, we need to explore what works, for whom and in what context. 53 It was clear that many varied personal contextual factors influenced adherence. The longitudinal nature of the study was important in highlighting that for all women, context and implementation were dynamic and life events got in the way. In addition, many women put the needs of others before themselves. We need to carry out a nuanced analysis to explore the characteristics of these women to understand the various ways women may overcome these events. Previous research supports the links between life taking over and inconsistent adherence in UI. 57,77 These findings suggest that when delivering a PFMT intervention, or in future research, consideration should be given to helping women balance the multiple contextual factors in ways that may support their engagement with PFMT and their re-engagement with PFMT after a break.

Urinary incontinence symptoms were an important factor in adherence at the outset, and continued to be so in the long term. Symptoms influenced adherence in a number of ways. Women adhered to rid themselves of symptoms, but, conversely, when symptoms were no longer present, the trigger to exercise was no longer there and some women then stopped exercising. Women had to perceive change and believe that it was linked to treatment to maintain adherence in the longer term. This finding is consistent with other studies. 18 It is an important feature of care delivery for therapists to keep the connection between PFMT and symptomatic improvement at the forefront of women’s minds.

Robustness of primary outcome analysis

The OPAL trial was large enough to detect a meaningful difference between groups if one existed, with a sample size far exceeding the largest previous trial of its type (n = 238, PFMT plus biofeedback and biofeedback vs. PFMT alone). 79 There was limited information about the minimal clinically important difference of the ICIQ-UI SF on which to base the sample size calculation at the outset of this trial; however, a recent publication indicated that a 3-point difference was clinically important, supporting our original assumption. 28 In addition, the observed SD of the primary outcome was 5 points, which was lower than our original assumption of 10 points, again pointing to the trial having sufficient statistical power.

In sensitivity analyses we showed that our primary outcome analysis was robust to the effects of missing data (tested under assumptions of missingness at random and not at random) and treatment non-compliance.

Further analyses of the primary outcome

In further analysis of the primary outcome, we also examined whether or not specific subgroups benefited from biofeedback PFMT compared with basic PFMT. The mechanism for the effect of PFMT in urgency UI differs from that for SUI; therefore, we investigated whether or not there was a different treatment effect for those women with pure SUI and those with a mix of SUI and urgency UI. There may be reasons why older, more so than younger, women might do less well with the addition of biofeedback, such as older women being less likely to engage with the biofeedback technology. The varying levels of severity of UI might be due to different underlying problems, which could respond differently to the addition of biofeedback. Finally, we explored whether or not women treated by different types of therapists might have different responses to the interventions. There was no evidence to suggest that women who differed with respect to type of UI (SUI vs. MUI), age (< 50 years vs. ≥ 50 years), UI severity (mild/moderate vs. severe) or therapist type (physiotherapist vs. non-physiotherapist) differentially benefited from either intervention type.

Strengths

A major strength of the OPAL trial was its mixed-methods design, including a RCT, with a process evaluation and longitudinal case study alongside. This offered the potential for a much deeper understanding of the intervention mechanisms and trial results from different perspectives.

The trial itself was conducted and reported following CONSORT recommendations [URL: www.consort-statement.org/ (accessed 31 July 2019)]. We used an automated computer randomisation application developed by CHaRT, accessed remotely by centre staff, to assign group allocation, thus minimising the risk of selection bias. The analyses were conducted on an ITT basis and in accordance with a predefined statistical analysis plan [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/117103/#/ (accessed 29 July 2019)], agreed with the TSC. We chose a well-validated and commonly used instrument as our primary outcome measure (ICIQ-UI SF) and a long duration of follow-up.

Participants were recruited from various outpatient and community settings, adding to the generalisability of the findings. At baseline, the two groups were highly comparable, indicating that the allocation process implemented in the trial had been successful. Our follow-up questionnaire response rates were high, indicating a low risk of attrition bias, with rates of 74% at 6 months, 84% at 12 months and 78% at 24 months. The lower response rate at 6 months is probably because of a less intensive process for contacting non-responders at this time point, which was not the primary end point. Such high retention rates limited the impact of missing data on our findings and this was confirmed by the sensitivity analyses, which showed no impact on the primary outcome analysis findings of various assumptions about missing responses.

All elements of the trial delivery were standardised, as far as possible, to minimise the risk of performance bias. Therapists in trial centres received training in all aspect of the interventions, including using the biofeedback equipment, delivering the PFMT protocol and incorporating the core and optional BCTs. This information was also provided to all therapists in a detailed intervention manual.

Limitations

Reasons for exclusion of women from the trial were kept to a minimum; however, one reason for women being ineligible was being unable to contract their pelvic floor muscles. Such women were excluded based on the NICE guideline, which recommends that women who cannot actively contract their pelvic floor muscles should be offered biofeedback. 1 If we had included these women, half would have been allocated to basic PFMT and therefore denied biofeedback. This means, however, that conclusions cannot be drawn from the trial about this subpopulation of women with UI.

It was not possible to blind participants as to which group they were in because of the nature of the interventions, and for the same reason it was not possible to blind treating therapists. The risk of detection bias was, therefore, high, given that participants were the main source of outcome data. The pelvic floor assessment at 6 months was, however, carried out by a clinician who was blind to the participant’s group and, in keeping with other outcomes, the finding was no evidence of a difference between groups.

Approximately one-third of women in both groups attended all six of their scheduled therapy appointments (109/295, biofeedback PFMT and 106/298, basic PFMT). Women commented in interviews on the lack of time to attend frequent appointments. This would reflect the situation in clinical practice, in which the ‘did not attend’ rate for PFMT appointments is high.

Another potential limitation is that biofeedback equipment is advancing rapidly and biofeedback units are now available that incorporate bluetooth and smartphone technology. These new features, not widely available at the time of developing this trial, may mean that therapists and women would now be less constrained in using such devices. Thus, there may be potential for an enhanced effect of biofeedback delivered using newer devices. These advancements may also have overcome some of the technical issues associated with biofeedback reported by therapists and women in the OPAL trial.