We updated the 2011 Cochrane review comparing different approaches to pelvic floor muscle training (PFMT) [1] to treat female urinary incontinence (UI) for two reasons. First, to address ongoing uncertainties as a scoping search suggested many more potentially eligible randomised controlled trials (RCTs) were published. Second, neither the previous review nor the most up-to-date alternative [2] applied ‘Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) to gauge evidence certainty to underpin clinical decision-making. Therefore, we aimed to complete a systematic review and meta-analysis, using contemporary Cochrane methods, of RCTs comparing different approaches to PFMT to increase the power, accuracy, and certainty in effect estimates. We investigated differences in exercise type, dose, and intervention delivery.

Systematic review methods were according to the Cochrane Handbook for Systematic Reviews of Interventions (version 6.4). The previous review was, as per Cochrane practice, the published protocol for the update [1].

Eligible trials were RCTs (excluding cross-over RCTs) in females with UI. Trials with pregnant or postpartum participants or those with neurological conditions were excluded. To investigate different pelvic floor muscle (PFM) exercise types, eligible RCTs compared coordinated (voluntary PFM contraction with other body movement, e.g. squats), or functional (voluntary PFM contraction within activities of daily living, e.g. the Knack), or indirect (exercise to improve PFM function without voluntary PFM contraction) or combined (indirect with direct) training versus direct PFMT (repeated, isolated, voluntary PFM contractions). To investigate exercise dose, eligible trials had the same exercise type and delivery method in both trial arms but with differences in exercise dose (e.g. more versus fewer training sessions per week). For exercise intervention delivery eligible trials compared different methods of delivery/supervision (e.g. more versus less in-person clinic supervision).

The Cochrane Incontinence Specialised Registry was searched on 27 September 2023, with no date or language limits. The register comprises studies identified from: the Cochrane Central Register of Controlled Trials, MEDLINE, CINAHL and handsearching. Study records were screened in Covidence. Each review stage was completed by two independent reviewers with any disagreements resolved through discussion: title and abstract screening, full text screening, data extraction onto a template revised from the previous review to include Consensus on Exercise Reporting Template PFMT (CERT-PFMT) items [3], and risk of bias (RoB) assessment using Cochrane ROB tool v1. All records were re-screened and evaluated regardless of inclusion or exclusion from the previous review. Two reviewers (JHS, MSP) cross-checked all RoB assessments, and completed the GRADE certainty of evidence ratings. Decision-rules were documented to ensure consistency.

The primary outcome was incontinence or lower urinary tract symptom specific quality of life (QoL) at the primary endpoint (as defined by trialists), measured using any instrument rated A or A+ by the 7th ICI based on psychometric properties [2]. We contacted study authors if data were collected but missing, incomplete, or reported in unusable format. The secondary outcomes (incontinence episode frequency, incontinence symptom severity, patient-reported improvement, patient-reported satisfaction, and adverse events) are not reported here. 

A standardised mean difference (SMD) with inverse variance weighted method was used in RevMan Web 2023 for meta-analysis. Thresholds suggested by Cohen (1988) were used for interpretation: a small (> 0.2 to <0.5), moderate (> 0.5 to < 0.8), and large (> 0.8) effect. Effect sizes under 0.2 were considered unimportant even if statistically significant.

Data were pooled in subgroups only; subgroups organised data by intervention and comparator. If within subgroup heterogeneity was substantial (I2 > 50%) we conducted a sensitivity analysis (low versus higher risk of selection bias, and attrition bias). If the between subgroup heterogeneity was substantial, we narratively summarised plausible explanations.

After removing duplicates, we screened 2385 records and excluded 2172 based on title/abstract. From 213 full texts retrieved, 87 were excluded, most commonly for an ineligible comparison. The remaining full texts represented 126 trials (some with more than one full text): 64 included trials, 40 eligible ongoing trials, and 22 trials awaiting classification (i.e. missing information precludes eligibility decision). Trials included 4972 participants: previous review included 21 trials, 1490 participants [1]. 

Sixty-one RCTs were parallel designs, and three were cluster RCTs. Sample sizes ranged from 11 to 362. Nine trials recruited > 50 participants per trial arm. The nine larger trials contained 42% of participants (2090/4972) but three of them reported no outcome of interest or usable data. 

No trial was conducted in a low-income country. Seven were completed in lower-middle income countries but 3 reported no outcome of interest or usable data. 

Overall RoB rating considered selection, attrition, reporting and other bias with five at low risk overall, six at high risk, and the remainder at unclear risk. Regarding selection and attrition bias, on which sensitivity analysis was based, there were 19 low and six high risk, and 22 low and 27 high risk trials respectively. Risk of bias rating did not consider blinding because all outcomes of interest in the review were patient-reported.

UI diagnoses were stress UI (n=36), stress predominant mixed UI (n=10), stress or mixed UI (n=6), stress, urgency, or mixed UI (n=7), and undefined “urinary incontinence” (n=4). Trial participants were typically aged from 45 to 65 years and parous, with no prior incontinence treatment or pelvic surgeries, or other appreciable pelvic floor dysfunction. 

Trials compared exercise type (27 trials; 3 subgroups), dose (11 trials, 4 subgroups) and intervention delivery (26 trials; 5 subgroups with data, and one without any usable data). Correct voluntary PFM contraction was confirmed for all women (35 trials), in one trial arm (five trials), or not mentioned. 

In addition to reporting only QoL here, we do not report findings from subgroups containing a single, small trial; in those instances, the number of downgrades of evidence quality precluded a certainty of evidence statement. 

Summary of findings is presented in Table 1.

The number of trials, and participants, has trebled. Progress is being made toward addressing the highest priority uncertainty in incontinence research—what is the optimal PFMT protocol—identified by Buckley and colleagues using a James Lind Alliance approach in 2010. However, the specific uncertainties they mentioned—training frequency and duration, and the optimal training for different patterns of UI—remain. Too few trials investigate exercise dose or recruit females with diagnoses other than stress or stress predominant UI.  

There is now some moderate certainty evidence to support clinical decisions about PFMT and its delivery in mid-age and older women. There is probably not support for indirect training approaches, alone or in combination with direct PFMT. Adding a resistance device to PFMT probably adds no benefit. There is probably no important difference in incontinence QoL outcome between individual and group supervision of PFMT; a correct contraction was confirmed prior to group supervision in 3 of 6 trials (209/280 women, 75%). Using e-health for delivery of PFMT is probably better than written instructions only.   

To increase evidence certainty for clinical decision-making, in addition to the usual improvements in trial size, methods and reporting, we need to address important uncertainties such as those identified by Buckley and colleagues.

Of the many methods of training that appear to be used in practice, direct PFMT is probably the intervention of choice. In-person supervision can probably be offered individually or in groups after confirmation of a correct PFM contraction. If supervision is not-in-person then e-health is probably better than written instruction.

Hay-Smith EJC, Herderschee R, Dumoulin C, Herbison GP. Comparisons of approaches to pelvic floor muscle training for urinary incontinence in women. Cochrane Database of Systematic Reviews 2011, Issue 12. Art. No.: CD009508. DOI: 10.1002/14651858.CD009508.
Cardozo, L, Rovner, E, Wagg, A, Wein, A, Abrams, P. (Eds) Incontinence 7th Edition (2023). ICI-ICS. International Continence Society, Bristol UK, ISBN: 978-0-9569607-4-0
Slade, S. C., Morris, M. E., Frawley, H., & Hay-Smith, J. (2021). Comprehensive reporting of pelvic floor muscle training for urinary incontinence: CERT-PFMT. Physiotherapy, 112, 103-112.

Continence 12S (2024) 101581
DOI: 10.1016/j.cont.2024.101581