Study design, materials and methods
There were 16 pairs of villages in the main trial. The study reported here is limited to those in the exercise arm whose intervention was completed after March 2017, with the first assessment post trial being made in April 2018, by which time all but 4 villages had completed the trial and field staff were available to examine maintenance. In the six villages reported here research paramedics from the main trial sought to interview each woman recruited to the trial, to have her complete validated questionnaires (previously used in the trial) on severity of urinary incontinence (Sandvik index), the extent to which she was bothered by symptoms (Urinary Distress Index v6 Short Form) and because of its important association, an assessment of depression (Center for Epidemiologic Studies-D). She was also asked to wear under her clothing, for 3 days, the ribbon belt devised for this study  and to tie a knot in a red ribbon each time she experienced urinary leakage during these 3 days. The change in number of knots tied was the primary outcome. All women were familiar with this three-day continence record (3DCR) which they had used 7 times in the main trial. Dryness was defined as no knot in a red ribbon during the 3 days of monitoring.
The research paramedic also asked the woman whether she had attended any group exercises since the end of the trial, and whether this still continued. The woman was asked further if she had continued to do the home exercises she had practiced during the trial, and if so, how frequently she was doing them at the time of the follow-up (maintenance study) interview. The extent of dryness at 12 months post trial was examined in relation to exercise post-trial in a multilevel logistic regression analysis allowing for clustering within villages and with adjustment for urinary leakage at recruitment, dryness achieved at the end of the trial, and for potential confounders ( age, parity and body mass index at the start of the trial).
The sample size for the follow-up study was determined pragmatically, reflecting the availability of field workers.
In the six villages reported here 96 women had taken part in the main study. Of these 80 were able to complete the maintenance study. Among the 16 who did not, 5 had died, 5 had moved away to live with family members, 3 could not be found and 3 were no longer intellectually able to collaborate. The 3-day urinary leakage (mean: SD) for the 80 women was 13.3 (3.8) at recruitment, 0.6 (0.8) at end of trial and 1.3 (2.69) at 12 months after the end of the intervention. Leakage measured by the 3DCR corresponded closely with severity of incontinence on the Sandvik index, with a Spearman’s rho of 0.84 at the end of trial and 0.98 at 12-month follow-up. These high correlations reflected the numbers achieving dryness. Among these 80 women, 45 were dry at the end of the main trial and 51 dry at 12 month follow-up. There was some change in the individuals with dryness on the 3DCR. Of the 45 dry at the end of trial, 28 (62%) were dry 12 months later. In addition, 23 of the 35 (66%) who had not achieved dryness by the end of the trial were dry 12 months later. Twelve were not dry at either point.
In three of the six villages the village paramedic had attempted to organise group exercises and 35 women reported that they had attended these, but most had attended only in the early weeks after the trial. Eight were still attending at least once/month at the time of the maintenance study. Only 9 women said that they had never done any of the home exercises since the end of the trial and a further 9 had given up by the time of the maintenance study. 78% were still carrying out the home exercises with 17 (22%) exercising at home most days. The likelihood of being dry at the time of the maintenance study was strongly related to the frequency of exercise (at home or in a group) at the time of the study with 82% of those exercising most days being dry, but only 33% of those reporting doing no exercise. In a regression analysis (table 2) it was confirmed that dryness at follow-up was independent of both urinary leakage at recruitment and dryness at the end of the trial, but was strongly related to the frequency of home exercises.
Interpretation of results
Education-only villages were not included here and the absolute change in incontinence at follow-up attributable to exercises during the trial cannot be estimated: some of the observed success in maintenance may have been due to education. Within women in six villages assigned to group exercise, it has been shown that continued use of the pelvic floor and mobility exercises taught and intensively practiced in the trial has been helpful in continuing to managing incontinence post-trial. The results suggest that motivating a woman to continue the exercises post intervention is essential for continued bladder control.