Design

The review included a total of 32 reports of 11 trials: five parallel‐arm randomized controlled trials (RCTs; Brown 2006b; Dowd 1996; Phelan 2012; Subak 2005; Subak 2009), four randomized cross‐over trials (Hashim 2008; Manonai 2006; Swithinbank 2005; Wells 2011), and one quasi‐randomised trial that used health record numbers as the basis for assigning people to interventions (Bryant 2002). We also identified one unpublished RCT with limited information (Miller 2007).

Participants

The included trials involved a total of 5974 participants, who were predominantly female (5954 women and 20 men). The average age (it was unclear if this was a mean or median) of the participants in the included trials ranged from 49 to 58 years, except for two trials with means of 62.7 years (Hashim 2008), and 70.25 years (Dowd 1996).

Sample size varied across trials. The majority of included trials had 60 or fewer participants (seven trials), however, two trials had more than 1000 participants and a further two trials had more than 100 participants Brown 2006b; Phelan 2012; Subak 2009; Swithinbank 2005).

In four trials, all trial participants had UI at baseline:

• Dowd 1996: 58 women with UI;

• Subak 2005: 48 women with SUI (6%), stress predominant MUI (40%), UUI (11%) and urgency predominant MUI (43%);

• Subak 2009: 338 women with SUI (8%), stress predominant MUI (25%), UUI (18%) and urgency predominant MUI (48%);

• Swithinbank 2005 : 110 women with USI (57%) and IDO (43%).

Four trials included adults with OAB, leading to UUI in some of the trial participants:

• Bryant 2002: 9 men and 86 women with symptoms of urgency, frequency and/or UUI; 83% had UUI at baseline;

• Hashim 2008: 11 men and 13 women with OAB; 29% had UUI at baseline;

• Miller 2007: 60 women with OAB;

• Wells 2011: 14 women with OAB, with or without UI.

One trial included women with urogenital atrophy, leading to UI in some of the participants:

• Manonai 2006: 42 women with urogenital atrophy; 61% had SUI and 19% had UUI at baseline.

The other trials that contributed the largest numbers of participants were sub‐studies of large diabetes trials of intensive weight loss programmes, namely the DPP (Diabetes Prevention Program; Brown 2006b), which focused on the prevention of diabetes, and Look AHEAD (Action For Health in Diabetes; Phelan 2012), which evaluated cardiovascular morbidity and mortality among individuals with type 2 diabetes. Not all of the trial participants in these trials had UI at baseline, but reported prevalence of UI at follow‐up:

• Brown 2006b: 2191 women in a diabetes trial; no baseline measures of UI;

• Phelan 2012: 2994 women in a diabetes trial; 27% had weekly UI at baseline.

Interventions

Outcome

Quality of reporting of outcomes was generally poor. Not all specified outcomes were reported. Where reported, outcomes were reported using diverse measures, which made the results difficult to interpret. Meta‐analysis was performed only for cure and improvement rates and UI prevalence from the weight loss interventions. All other outcomes were summarised narratively.

Description of studies

We identified four trials involving a total of 4701 participants that compared intensive lifestyle weight loss interventions with control or no active interventions in relation to incontinence (Brown 2006b; Phelan 2012; Subak 2005; Subak 2009). All participants in two of the included trials had UI (Subak 2005; Subak 2009). The other two trials were sub‐studies of large diabetes trials (Brown 2006b; Phelan 2012), DPP and Look AHEAD, respectively, and contributed 4315 (92%) of the trial participants to the analysis. These trials did not recruit participants specifically with UI and therefore not all the participants had it. We extracted outcome data on cure and improvement (based on quantification of symptoms) from a subgroup of people with UI in the Look AHEAD trial (N = 738; Phelan 2012), and data on prevalence of UI at follow‐up from the whole (sub‐)study (N = 2994). The trial authors reported the one‐year results of a four‐year intensive weight loss programme; the trialists planned that follow‐up of this trial would run until 2014. The only relevant outcome from the DPP trial was prevalence at follow‐up of UI from the whole (sub‐)study (N = 1321; Brown 2006b); the proportion of people with UI at baseline was unknown (not reported).

All participants in the included trials were female. The weight loss groups were given a reduced‐calorie diet and increased physical activity according to a structured and supervised protocol. The comparison groups received:

• no intervention (waiting list; Subak 2005);

• a structured education programme on weight loss (Subak 2009);

• diabetes support and education (Phelan 2012); or

• a placebo drug (Brown 2006b).

Duration of the interventions varied across trials. The intensive intervention phase lasted for:

• three months (Subak 2005);

• six months followed by a further randomisation in the intervention group (not the control group) to motivation‐based or skill‐based maintenance programmes for an additional 12 months (Subak 2009);

• six months with monthly follow‐up thereafter for an average of 2.8 years (Brown 2006b); or

• 12 months (Phelan 2012).

All four trials reported that women allocated to the intervention group achieved a statistically significant decrease in body weight from baseline compared with those in the control group.

Primary outcomes

Improvement rates based on women's perception (self‐report) were reported in one trial (Subak 2009). The results showed that at six months women in the intervention group were more likely to report improvement than those in the control group at six months (163/214 (76%) versus 49/90 (54%), risk ratio (RR) 1.4, 95% confidence interval (CI) 1.14 to 1.71; Analysis 1.1), 12 months (298 women in analysis, 75% versus 68%, RR not estimable, reported P value 0.2) and 18 months (291 women in analysis, 75% versus 62%, RR not estimable, reported P value 0.02) after randomisation (Analysis 1.2). The reported P values suggest that the differences were statistically significant at six months and 18 months. No information was available on self‐reported cure.

The intervention group also reported that incontinence had less adverse impact on their lives (median Incontinence Impact Questionnaire scores, 40 women in analysis, 37 versus 89, P value 0.01) and was less distressing (median Urogenital Distress Inventory scores, 40 women in analysis, 104 versus 195, P value < 0.0001) compared with the control group in one trial that reported these outcomes (Subak 2005; Analysis 1.3).

Adverse effects appeared to be relatively uncommon, with one trial reporting that the intervention had 'few side effects' (Subak 2005).

Secondary outcomes

Three trials reported cure and improvement rates based on quantification of symptoms (rather than women's perception; Subak 2005; Subak 2009; Phelan 2012; Analysis 1.4; Analysis 1.5; Analysis 1.6; Analysis 1.7). Depending on the trial, length of follow‐up and type of incontinence, cure rates for the intervention group ranged from 7% to 35% and improvement rates ranged from 37% to 64%. In the control group cure rates ranged from0% to 32% of women, while improvement ranged from 0% to 62%.

In general the intervention group had higher rates in terms of both cure and improvement compared with the control group when stress and urgency UI symptoms were considered together ('all UI'). For improvement rates, the difference between the groups was statistically significant at three months (7/19 (37%) versus 0/21 (0%), RR 16.50, 95% CI 1.01 to 270.78; Analysis 1.6), six months (88/214 (41%) versus 20/90 (22%), RR 1.85, 95% CI 1.22 to 2.81; Analysis 1.6) and 12 months (234/583 (40%) versus 146/449 (32%), RR 1.21, 95% CI 1.02 to 1.44; Analysis 1.6), although the effect was attenuated over time and was no longer statistically significant at 18 months (91/197 (46%) versus 36/90 (40%), RR 1.15, 95% CI 0.86 to 1.55; Analysis 1.6). The difference for cure rates did not reach statistical significance (Analysis 1.4).

Looking at different types of UI at each outcome time point, the Subak 2009 trial showed a similar pattern with a tendency towards greater improvement in the intervention group than in the control group among the subgroup of women who reported stress symptoms at six, 12 and 18 months (P values 0.01, 0.01 and 0.92 , respectively) or urgency symptoms at six, 12 and 18 months (P values 0.04, 0.07 and 0.03, respectively; Analysis 1.7). In the same trial, cure rates by type of UI also favoured the intervention group for both SUI (P value 0.004) and UUI (P value 0.02) at six months, but no further follow‐up was available (Analysis 1.5).

The prevalence of weekly (or more frequent) UI of any type (stress or urgency) was lower in the intervention group than in the control group in the two sub‐studies of diabetes trials with a follow‐up of between one and 2.8 years (Analysis 1.8). According to adjusted odds ratios (ORs) reported by trial authors, the intervention was associated with a statistically significant reduction in the odds of having UI by around 20% to 24% compared with the control group (in Phelan 2012, adjusted OR 0.80, 95% CI 0.65 to 0.98; in Brown 2006b, adjusted OR 0.76, 95% CI 0.61 to 0.95). The prevalence of weekly SUI was also lower in the intervention group compared with the control in both trials (in Phelan 2012, adjusted OR 0.73, 95% CI 0.55 to 0.96, Analysis 1.9; in Brown 2006b, adjusted OR 0.80, 95% CI 0.64 to 1.01) but no such difference was apparent for UUI (Analysis 1.9). The trial authors suggest that the reduction in the prevalence of overall weekly incontinence may be due to differences in weekly SUI.

Compared with women in the control group, those in the intervention group had a greater percentage reduction from baseline in weekly incontinence episodes over the period of three to 18 months regardless of type of UI (all, stress or urgency; Analysis 1.10; Analysis 1.11). Differences between the groups for all UI and SUI episodes were reported to be statistically significant at three, six and 12 months but no longer significant at 18 months. The difference for UUI was not statistically significant at any point in time after three months.

General health‐related quality of life was measured only in one small trial with 40 participants using SF‐36 ( (Subak 2005; Analysis 1.3). The median SF‐36 Physical Component Score favoured the intervention group (55 versus 47, P value 0.003) but there was no significant difference between the groups in the Mental Component Score of the same instrument (48 versus 51, P value 0.09).

Description of study

We identified only one small trial that assessed the effect of dietary factors on UI (Manonai 2006). The trial used a randomized cross‐over design and compared a soy‐rich diet with a control (soy‐free) diet in 42 women who experienced at least one of urinary or genital symptoms owing to urogenital atrophy. At baseline around 61% and 63% of women in the intervention and control groups, respectively, had SUI episodes and 19% and 11%, respectively, had UUI episodes. Partcipants underwent two two‐week treatment periods in random order with two four‐week washout periods before and between treatments. The trial authors found compliance to the diet to be satisfactory on the basis of the elevation of serum levels of daidzein and genistein during the soy‐rich diet period. Outcome data were available for 36 women who completed the trial. As data subgrouped by incontinence status were not available, the extracted data were from the whole study.

Primary outcomes

The trial did not address self‐reported cure and improvement rates, condition‐specific quality of life and adverse effects.

Secondary outcomes

The trial did not address cure and improvement rates based on quantification of symptoms, number of UI episodes and generic quality of life. The available data suggest that the percentage of women with UUI episodes in the control group increased from baseline during the control diet period (N = 36, from 11% to 22%, P value not reported; Analysis 2.1). Correspondingly, symptom scores (mean, SD) of UUI significantly increased during the control diet period (N = 36, from 0.14 (0.35) to 0.25 (0.50), P value < 0.05; Analysis 2.2), although the difference was small.

Description of studies

We identified three trials that examined altering the level of fluid intake (Dowd 1996; Hashim 2008; Swithinbank 2005).

One RCT allocated 58 women with UI to one of three groups that increased fluid intake by 500 ( Dowd 1996). The trial provided a five‐week programme with randomisation in the second week. The trial reported that adherence to the fluid manipulation was poor, which made results difficult to interpret.

Another randomized cross‐over trial with 84 women with UI reported outcome data for the 69 women (39 with USI and 30 with IDO) who completed the trial (Swithinbank 2005). The trial lasted four weeks. In the first week participants drank normally (week 1, baseline) and in the second week drank normally, but only caffeine‐free fluids (week 2, caffeine‐free baseline). Participants were then randomized in the order in which they either increased fluids to 3 litres daily, or decreased fluids to 750 ml daily, in the third and fourth weeks while maintaining caffeine restriction (i.e. only drinking caffeine‐free fluids). Adherence to fluid intake protocols seemed fair, with a mean fluid intake of 1639 ml for Week 1, 1630 ml for Week 2, 2673 ml for the week of increasing fluids and 872 ml for the week of decreasing fluids.

In another cross‐over trial with 24 participants (11 men and 13 women) with OAB (Hashim 2008), only seven (29%) participants had UUI at baseline. Participants were randomized into two groups and asked to either increase or decrease their fluid intake from baseline. As outcome data specific to a subgroup of people with UI were not reported separately, the extracted data applied to the whole study. Group I was asked to drink at < 25% of baseline for four days, followed by two days' normal drinking, four days' at < 50%, two days' normal drinking, four days at > 25%, two days' normal drinking, and then four days at > 50%. Group II did the reverse. The trial reported that participants had difficulty in either increasing or decreasing fluids by 50%.

Primary outcomes

One cross‐over trial assessed quality of life using the Bristol Female Lower Urinary Tract Symptoms questionnaire (Swithinbank 2005). Quality of life improved when fluid intake was decreased compared with baseline in women with USI (N = 39, P value < 0.003) or IDO (N = 30, P value < 0.003) but the women reported no significant difference in the impact of incontinence symptoms on their daily life before and after treatment (no further data were available).

Regarding adverse effects, the same cross‐over trial reported that, with decreasing fluids, ‘side effects such as constipation and thirst were troublesome’ (Swithinbank 2005). Another cross‐over trial with 24 participants reported that adverse events observed were mild and tolerable: four participants felt thirsty and two had headaches, constipation or concentrated urine when fluid intake was decreased by 50% from baseline; and one had headache when intake was reduced by 25% (Hashim 2008). 

No information was available regarding self‐reported cure or improvement.

Secondary outcomes

The number of daily incontinent episodes was reported by three trials that used different measures. A four‐week cross‐over trial stratified results by type of UI at baseline (Swithinbank 2005; Analysis 3.1). Among 39 women with USI, the week of decreasing fluid intake (with caffeine restriction) was associated with a statistically significant reduction in the median number of daily incontinent episodes compared with the week of increasing fluid intake (with caffeine restriction; 0.5 versus 0.7, P value 0.006). Daily incontinent episodes after decreasing fluid intake (with caffeine restriction) were also statistically significantly fewer compared with the baseline week when participants drank normally (week 1, 0.5 versus 1.6, P value 0.006), but there was no significant difference when compared with the caffeine‐free baseline week in which participants maintained a similar fluid intake from baseline, but substituted caffeine‐free drinks for caffeine‐containing drinks (week 2, 0.5 versus 0.8, P value 1.000). The week of increasing fluid intake (no caffeine) did not differ significantly from the caffeine‐free baseline week (week 2, 0.7 versus 0.8, P value 0.426) in terms of daily incontinent episodes. For 30 women with IDO, drinking less fluid (no caffeine) had no statistically significant effect on daily incontinent episodes (0.5 versus 0.6, P value not significant) when compared with the caffeine‐free baseline, but drinking more fluid (no caffeine) resulted in a significant worsening (increase) of the symptom (1.1 versus 0.6, P value < 0.003).

In the other cross‐over trial participants were asked to increase or decrease their fluid intake by 25% and 50% from baseline in random order (Hashim 2008). There was no statistically significant difference in the mean number of daily incontinent episodes between the baseline period and each of the fluid manipulation periods, although it should be noted that only seven (29%) of the 24 participants had UUI at baseline (Analysis 3.2).

The Dowd 1996 trial randomized participants to three groups (maintain, increase or decrease fluid), but the authors reported that adherence to the fluid manipulation protocol was poor and that results were inconclusive (Analysis 3.3).

Description of studies

We identified three trials that assessed the effects of a reduction in caffeine intake on incontinence (Bryant 2002; Miller 2007; Wells 2011).

In Bryant 2002, 95 participants (86 women, 9 men) with OAB (83% had UUI at baseline) were randomized by use of health record numbers (quasi‐randomised) to caffeine reduction education or control (continued caffeine intake). In addition, both groups received bladder training and were followed up for four weeks. Caffeine intake in the intervention group was reduced significantly from baseline compared with the control group (58% versus 11%, P value < 0 000.1)

Wells 2011 was a randomized cross‐over trial in which 14 women with OAB (with or without UI) underwent two two‐week periods of caffeinated or caffeine‐free fluids intake with a 14‐day washout period between treatments. It was a feasibility trial and identified only in abstract form. Data were available for the 11 women who completed the trial. Two participants did not comply with caffeine substitution.

The third trial, Miller 2007, was an unpublished two‐arm RCT that evaluated the effect of restricting 'irritating' beverages (caffeinated or non‐caffeinated). In this trial around 60 women with OAB (it was unclear if some or all participants were incontinent) were asked to substitute 'irritating' beverages with milk or water, but to maintain a similar volume of fluid from baseline. The request to stop drinking irritating beverages was associated with an improvement in OAB symptoms, but the trial author noted that the findings were confounded by a significant reduction in overall fluid intake in the intervention group from baseline (email communication from the trial author to the Cochrane Incontinence Group). No further information was available regarding this trial. The remainder of this section therefore focuses on the first two trials.

In all three trials, outcome data were extracted from the whole study, as data specific to the UI subgroup of the trial population were not reported separately.

Primary outcomes

The Wells 2011 trial reported condition‐specific quality of life using ICIQ Overactive Bladder (ICIQ‐OAB) and ICIQ Overactive Bladder Symptoms Quality of Life (ICIQ‐OABqol) questionnaires among 11 of the 14 women who completed the trial. Overall, women had lower (better) scores during the period of caffeine substitution (when drinking caffeine‐free fluids) compared with the caffeine exposure period, but the difference in total scores for the ICIQ‐OABqol was not statistically significant (mean 54 versus 68, P value 0.065; Analysis 4.1). No information was available regarding self‐reported cure and improvement, or adverse effects.

Secondary outcomes

There was no evidence of a difference in incontinence episode frequency between the caffeine reduction and caffeine exposure groups. The Bryant 2002 trial reported a mean difference of 0.2 episodes per day (Analysis 4.2: mean difference (MD) ‐0.20, CI ‐1.02 to 0.62), whereas the Wells 2011 trial reported 'no difference' with no numerical data provided. No information was available for cure and improvement rates based on quantification of symptoms or generic quality of life.