Results

2472 studies were identified. 86 studies met inclusion criteria after full text review (PRISMA diagram S3).

Characteristics of included studies:

Studies varied in quality and methodology. One RCT and non-randomised studies including retrospective reviews, case-control, service evaluations and questionnaire studies were included (Table S1).
For measurement of our primary outcome, eight studies used a validated questionnaire (7, 12, 17, 18, 27, 28, 29, 30) and four studies did not use a validated method of symptom assessment, relying on description alone (18, 31, 32, 33). It was possible to analyse data for an asymptomatic group of women after index OASI for two studies (Figure 3D) (27, 28).
Risk of bias:
All ten studies contributing data to meta-analysis for the primary outcome were at high risk of bias for at least one domain (Figure 1). One multicentre RCT contributed data to the analysis; however only 14.7% of 222 women randomised had a third-degree tear at the time of the index birth and the other 85.3% women were recruited after a forceps birth and had an anal sphincter defect identified on endoanal ultrasound, which was considered a surrogate marker for OASI. 71% women randomised in this study had an intact perineum at index birth (28). One prospective cohort study to 6 months follow up was included (7). Other studies were confounded by inclusion of women symptomatic of AI or were biased by retrospective reporting of outcomes after treatment allocation, failure to account for important factors such as parity or to analyse by the intention-to-treat approach, (e.g. by excluding emergency cesareans, or including elective and emergency cesarean data together). Studies were also biased by retrospective recall of AI symptoms from patients, different populations included (e.g. patients with OASI diagnosed sonographically, exclusion of women with OASI recurrence, sphincter injury not classified by Sultan criteria), use of non-validated symptom assessment tools (e.g. patient description), selective outcome reporting, use of multiple analyses, use of counselling pathways advising CS for different groups, (e.g. patients with symptoms, 4th degree tear or EAUS defects), and methodological limitations (e.g. retrospective service evaluation, follow-up questionnaire with limited response rate).

Data heterogeneity:

A wide range of methods and protocols for counselling mode of subsequent birth were described based on patient symptoms alone, endoanal ultrasound, anorectal manometry, and findings on 3D transperineal ultrasound. Seven studies used EAUS to counsel women regarding mode of subsequent birth (18, 28, 29, 32, 33, 34, 35) and four used ARMS (18, 29, 32, 33) (S4). One study used 3D-TPUS but did not contribute data to meta-analysis (11).
Recurrence of OASI
49 studies reported recurrence of OASI in women undergoing subsequent vaginal birth (S5). These included 10 national or state database studies (4, 12, 36, 37, 38, 39, 40, 41, 42, 43, 44), 3 US state databases (36, 38, 40) and 1 Canadian state database (44). Three studies were multicentre studies (27, 28, 45) and 32 studies were single centre studies (11, 17, 29, 30, 31, 32, 33, 34, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69). 29 studies took place in a university or teaching hospital setting, (11, 17, 27, 28, 30, 32, 33, 34, 45, 46, 47, 49, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 65, 67, 68, 69) and six were in a district general hospital setting (29, 31, 48, 50, 64, 66).
The mean index OASI rate was 3.31% and mean recurrence rate was 6.6% across all studies (7413 cases: range 1.9-25%, S5). Index OASI rates were higher in teaching hospitals (6.13% n=32) compared to population-based studies (4.34% n=11) and district general hospitals (3.2% n=6). Recurrent OASI rates were higher in district general hospitals (11.74% range 6.6-25% n=6) but similar in teaching hospitals (6.02% range 1.9%-13.4% n=32) and population-based studies (5.59% range 2.1-7.2% n=11). A summary of rOASI rates provided by population-based studies (Figure 2) and all studies (S5) is described.
Incidence of AI after subsequent vaginal birth (VB) and subsequent CS after OASI
The incidence of AI after a subsequent birth after OASI was reported by twenty-two studies. 12 studies (one RCT and 11 non-randomised studies) were suitable for meta-analysis as they included data for both CS and VB in a suitable form for pooling, and their characteristics are described (S4, Table S1) (7, 12, 18, 27, 28, 29, 31, 32, 33, 34). Some participants from one study (70) were included in a later publication (29) therefore we excluded the earlier study from meta-analyses. Two studies described satisfaction with MOB choice (17, 34), two studies described regret with MOB choice (27, 30) and two studies described quality of life after subsequent birth (7, 12). 11 studies were not suitable for meta-analysis (11, 13, 17, 46, 65, 67, 69, 71, 72, 73, 74) as they either did not contain a comparison group or did not contribute data in an interpretable format (and either were not able to provide this or did not respond on contact with the corresponding author).
There was no evidence of a difference in deterioration in AI (new or worsening symptoms) after subsequent VB compared to CS after OASI across all follow up time periods studied (Figure 3A: p=0.74; CI 0.72-1.19, 9 studies- 1 RCT 8 non-RCT, 2104 participants). There was no evidence of a difference for short-term symptoms at 3-24 months (Figure 3B: p=0.71; CI 0.65-1.72, 7 studies, 569 participants) or long-term symptoms at ≥5 years (Figure 3C: p=0.39; CI 0.65-1.18, 2 studies, 1536 participants) between groups. There was no evidence of a difference in AI between VB and CS in studies who recruited only asymptomatic women (Figure 3D: p=0.43; CI 0.71-2.20, 2 studies, 220 participants).
The total incidence of AI after subsequent birth after OASI may be higher after CS than VB; the wide CI is consistent with little effect or harm (Figure 4: p=0.03; 95% CI 1.08-3.47; 6 studies, 2253 participants). These studies include women who were symptomatic pre-subsequent birth. Symptomatic women were more likely to have subsequent CS than VB across the studies. The presence of pre-existing AI symptoms is therefore a significant confounder.
The incidence of incontinence of solid or liquid stool, flatal incontinence and fecal urgency does not differ between women who have a subsequent CS vs VB (S6). There is no evidence of a difference in levels of satisfaction (17, 34) and regret (27, 30) (S7); regret is inversely proportional to satisfaction (30). There was no evidence of a difference in QOL overall (S7); one study demonstrated higher levels of ‘incontinence impact’ and ‘physical limitations’ in women having CS vs VB (7). Reduction in QOL is directly proportional to patient symptoms of AI after OASI (70).
In a study not suitable for meta-analysis, patients subsequently delivering by CS had a significantly lower Vaiezy score compared with subsequent VB, and those following MOB recommendations also demonstrated significantly less AI (65). Another study demonstrated higher Wexner scores overall in women delivering by CS vs VB, however there was no difference in average pre- and post- birth scores in the CS and VB groups (17). These studies demonstrate the difficulties with interpreting results from individual sites, the confounder of a higher symptomatic population in the CS group and widespread practice variation.

Other outcomes

No studies reported on symptoms of irritable bowel syndrome, obstructive defecation, blood loss or length of stay.
Anal incontinence and subsequent birth by any mode
Twelve studies have investigated whether there is a difference in AI in women who have a subsequent birth compared to women who have not had a subsequent birth after OASI. Ten studies were suitable for meta-analysis (S8A) (13, 31, 55, 73, 75, 76, 77, 78, 79, 80). There is no evidence of a difference in AI in women who have a subsequent birth by any mode vs those who have not (S8A: p=0.9; CI 0.71-1.34, 10 studies 970 participants). One study looked at QOL with a subsequent birth vs no subsequent birth (S8B). There was no difference in QOL outcomes for women with a subsequent birth by any mode compared to those who did not (55). A previous review meta-analysed this outcome in 2016 (35). Only one further study has been added since this time (7).

Anal incontinence pre- and post- subsequent birth

The incidence of AI in the same patients pre- and post- subsequent birth by any mode does not differ between a mixed symptomatic and asymptomatic patient group across all time periods studied (S9: p=0.31; CI 0.51-1.25, 13 studies, 5496 participants) (7, 12, 13, 18, 31, 32, 33, 34, 70, 74, 76, 81, 82).