Pelvic floor muscle training (PFMT) is a first-line treatment for urinary incontinence (UI) in adult women. Clinicians have proposed a large array of variations and add-ons to support PFMT, but their underlying evidence is sometimes unclear. Furthermore, the sheer amount of research, as well as the complexity of its clinical questions, make it difficult to appreciate where the evidence is solid versus where it is shaky. To address these issues, our interdisciplinary team performed a comprehensive systematic review and constructed interactive online evidence maps.

Our 6 data sources comprised 3 completed systematic reviews, the PCORI website, ClinicalTrials.gov, and an updated literature search. For the literature search, an experienced information specialist searched PubMed, EMBASE/Medline, and PsycINFO from December 4, 2017 to July 5, 2019. We only included English-language randomized trials on one of 4 comparison types: the effect of PFMT, the effect of variations on PFMT, the effect of add-ons, and alternatives to PFMT. At least 90% of those enrolled had to be adult women with UI. Included studies had to report either cure rates, data on the frequency of UI events, or quality of life data using one of four specific QOL instruments. The minimum follow-up was 4 weeks, and we required that the study report data on at least 10 patients per group at follow-up who represented at least 50% of those enrolled. We excluded studies of women who were pregnant, or had neurogenic lower urinary tract dysfunction, cognitive impairment, surgically-treated UI, were seen in acute care settings, or at least 10% had received ancillary UI treatments. We also excluded studies in which the researchers’ method of administering PFMT varied by type of UI.

A team of 3 systematic reviewers screened abstracts and full articles (at least two screeners per article, with disagreements resolved by consensus). Within each of the four types of treatment comparisons, we created sub-comparisons comprised of studies making similar comparisons (for example, within the variations category, one subcategory was group PFMT vs individual PFMT). For each outcome, we computed the effect size and its standard error (odds ratio for cure, and Hedges’ g for the other two outcomes). We performed random-effects meta-analyses using the method of DerSimonian and Laird. We rated the strength of evidence using a modification of the system developed by the AHRQ Evidence-based Practice Centers. For statistically non-significant differences, we evaluated the confidence interval against the minimal important difference in order to determine whether a conclusion of equivalence was appropriate.

Seventy-three RCTs met our inclusion criteria (62 from previous reviews, and 11 from the search update), and we also included 13 unpublished trials from ClinicalTrials.gov.

We created two maps using Tableau: a Research Volume (RV) map, and a Treatment Effects (TE) map. The RV map (Figure 1) uses a bar chart to display the number of women with outcome data (vertical axis) for each of the 4 overall categories (horizontal axis).  The top of the chart shows the total number of treatment comparisons as well as the number of RCTs. Five interactive filters along the left side of the map (type of UI, age, number of weeks since the start of treatment, publication year, and region of the world) allow the user to redraw the chart according to their specific interests (filters can be used in any combination). Hovering over a bar provides additional details such as the top 2 comparisons in that category (by number of women), and clicking on a bar yields counts for each subcategory, per-study Ns, and hyperlinks to trial abstracts.

The TE map (Figure 2) is a bubble chart displaying meta-analytic results in each of the four comparison categories (one tab per category). For each tab, the bubbles are arranged in a grid in which each column is an outcome and each row is a specific treatment comparison. Bubble color indicates the direction of effect (with gray indicating insufficient data), and for non-gray bubbles, the bubble size is linearly related to the meta-analytic point estimate. An asterisk in a cell indicates that there is no evidence on that outcome for that comparison. Hovering over a bubble provides the overall conclusion for that bubble, and clicking on a bubble gives more details including the meta-analytic summary estimate and confidence interval, a forest plot, and hyperlinks to trial abstracts. The first tab has 2 interactive filters down the left side (type of UI and age), whereas the other 3 tabs only have a type-of-UI filter. Using the filters redraws the bubbles accordingly. Note that the TE map only includes comparisons investigated by at least 2 studies; data for single-study comparisons are accessible by clicking a box below the filters.

Evidence is often voluminous and confusing. Carefully constructed evidence maps can efficiently use bars, bubbles, and colors to accelerate users’ understanding of the evidence. Interactive features, such as filters, hovers and clicks, allow users to tailor the display to their clinical interests and desired detail.