Surgery for constipation: systematic review and practice recommendations

To assess the outcomes of recto‐vaginal reinforcement procedures in adults with chronic constipation.


Introduction Background and procedural variations
Chronic constipation is related to an inability to evacuate the rectum in over half of all adults presenting for specialist advice. This phenomenon, which may be variably described as obstructed defaecation or rectal evacuation disorder (and many other terms), is characterised by excessive straining, the feeling of incomplete evacuation, post-defaecatory seepage and often mucous discharge, and pelvic pain [1]. In some of these patients there is clinical and radiological (usually proctographic) evidence of a dynamic structural abnormality leading to physical impediment to emptying during defaecation. By far away, the way most common abnormalities are rectocoele and/or intussusception.
Traditionally a rectocoele may be considered either a bulge into the vagina, giving vaginal prolapse symptoms, or the sensation of a lump or mass. Also, a herniation of the rectum into the vagina preventing complete evacuation may lead to 'trapping' of faeces in the 'pocket' and subsequent incomplete evacuation (ballooning may also lead to loss of vector forces along the anorectal axis). This process can lead to dissatisfaction with emptying and repeated visits to the toilet, postdefaecatory soiling and the need to apply pressure to the posterior vaginal wall or perineum to splint the rectocoele and maximise emptying. Many women will also 'digitate' or manually assist emptying using a finger either in the rectocoele via the anus to aid with complete emptying or in the vagina to splint against the posterior vagina to prevent trapping and direct evacuatory forces down along the axis of the anus.
It follows that strategies have been employed to anatomically correct a rectocele by reinforcing the barrier between the rectum and vagina (rectovaginal septum). Access for rectovaginal reinforcement varies with three main approaches described: the posterior vagina (posterior repair [2]), the perineum (transperineal repair [3]), or via the anus (transanal repair [4,5]). Vaginal repairs all involve an incision in the posterior wall either longitudinally to open the entire length of the rectocoele or transversely to produce a broad based flap and expose the entirety of the rectocoele, or a combination of the two incisions in the shape of an inverted 'T'. In most repairs the redundant rectal wall is plicated outside the bowel wall. The vaginal wall is then reconstructed with resection of any excess vaginal mucosa. Many repairs include some degree of approximation of the levator ani and pelvic side wall muscles to formally reinforce the reconstruction of the rectovaginal septum, although with the reputed higher risk of dyspareunia. More recently attempts have been made to localise specific defects in the rectovaginal septum. This has led to 'site specific' repairs [6,7], where the individual defects are repaired before the vaginal wall is closed. All of these types of repairs may be augmented by mesh reinforcement, usually using one of a variety of collagen meshes.
The rectovaginal septum may be entered through a transperineal route. A transverse or curved incision is made in the perineal body towards the vagina and the dissection extended anterior to the sphincter complex to gain access to the rectovaginal septum. Having gained exposure, a repair is made in the same way as in a transvaginal approach. The potential advantage to this route is that it may be combined with a sphincteroplasty in those women with a deficient sphincter and a concurrent rectocoele.
Finally, a rectocoele can be considered as a redundant pocket of rectum, rather than a weakness in the rectovaginal septum leading to herniation of the rectum. A transanal repair addresses the rectal redundancy with either an anterior Delorme's style repair (Sarle's repair [5]) or with a sutured pexy of the anterior rectal wall (Block repair [4]). In the former repair, the mucosa is dissected free from the rectal muscle and excised. Following this, the rectal muscle coat is plicated longitudinally to obliterate the rectocoele pocket and the mucosa is re-approximated to close the defect. In the Block repair, full thickness longitudinal sutures are placed anteriorly to draw together the redundant anterior rectal pocket and close the rectocoele. It is evident that these procedures necessitate a degree of rectal wall excision or suspension respectively.

Scope
The overall purpose of the CapaCiTY review series is to assess the efficacy and harms of procedures for chronic constipation in adults. The focus is therefore directed to a population presenting with symptoms of obstructed defecation, rather than women with only prolapse, urinary symptoms or both (see overview and methods paper). One group of procedures considered beyond the scope of this systematic review included those where excision is effected without reinforcement, e.g. rectal excision only, such as stapled transanal resection of the rectum (STARR). As is noted above, anterior Delorme's or Sarle's procedures are included, which include a degree of resection, but also have an significant element of reinforcement with muscle wall plication (targeted at the rectovaginal plane), which is not present in pure resection. This noted, a small number of studies have specifically focused stapled resection to the obliteration of a rectocele by anterior deployment of a stapler to exact mucosal excision (much like stapled haemorrhoidopexy). In a sense, these reinforce the RV septum by tightening tissue and have been included for completion in this review. Further, most compare this approach with one of the other approaches also covered in this review. The separate review covering rectal excision includes numerous studies of STARR in which the device is used to excise circumferential full-thickness excision and the reviews overlap in only one study [8]. Circumferential Delorme's procedures are excluded from this review as their focus is not on reinforcement of the rectovaginal septum but rather on general resection for prolapse (covered in rectal suspension review). Studies where outcomes could not be segregated by eligible procedure were also excluded, due to a mixed patient population with internal and external rectal prolapse, mixed indications including numerous pelvic floor abnormalities or limited postoperative outcomes.

Previous reviews
There have been no systematic reviews of the results of rectocoele repair for the treatment of obstructed defaecation or constipation. There has been a systematic Cochrane review (with updates) on the surgical management of pelvic organ prolapse in women [9][10][11][12][13], which includes review of posterior vaginal repair for rectocoele. This review included seven randomised trials on women with posterior vaginal wall prolapse [3,[14][15][16][17][18][19], three trials comparing vaginal and transanal approaches [3,14,18]. In addition, another trial provided data for women with rectocele undergoing posterior repair with and without absorbable mesh [16]. There has been one systematic review of the use of biologics in urogynaecological repairs [20]. In addition, the International Continence Society, 5th International Consultation on Incontinence, Paris, February 2012, pages 1411-1414 [21], covers posterior repair. Neither this, nor the Cochrane process focussed on the specific management of patients with constipation.

Summary of search results and study quality
The search yielded a total of 72 citations for full text review (Fig. 1). From these, 44 articles published between 1990 and 2016 contributed 43 studies to the systematic review (one cohort was spread across two publications with a focus on perioperative morbidity [22] and clinical outcomes [23], respectively). Data on outcomes were provided for a total of 3346 patients (study mean: 78, range 13-307) ( Table 1). Specific exclusions after full-text review (and after exclusion of five non-English language publications) included: four studies where the population sample was confirmed to be <20 patients; eight studies where follow up was less than 12 months; two studies of out of scope procedures; four studies where data were considered as duplicate; three studies where outcomes could not be segregated by eligible procedure; and two studies with no relevant outcome data.
The quality of studies varied. The 43 included studies included three randomised controlled trials (RCTs) and 40 observational studies. The former included one good quality RCT (level IB) with a low level of susceptibility to bias and two with less well described methodology (level IIB). The 40 observational studies included eight good quality cohort studies with low susceptibility to bias (level IIB). Other studies were a mix of prospective and retrospective case series. Mean study follow-up was 2.1 years (range 0.7-6.2 years); 19 studies originated from European centres, 12 from the USA and 12 from other countries. The studies covered the full spectrum of types of repair and are shown in Fig. 2.

Perioperative data
The 43 studies included 62 patient cohorts receiving one of eight procedures ( Table 2). Perioperative data were reported inconsistently by studies, with 24 (39%) cohorts reporting average procedural duration and 29 (47%) reporting average length of stay (Tables 2 and 3). There was considerable variation in these measures when comparing procedures: design heterogeneity, small numbers of studies and large range of values precludes any clear pattern. The overall average duration of procedures was 67 (range 20-169) minutes, and the overall average length of stay was 3.9 (range 1-15) days.

Summary evidence statements: perioperative data
• The average duration of procedures was about one hour, although this is inconsistently reported and ranged widely between studies from 20 to 169 min (level IV).
• The average length of stay was about 4 days although this is inconsistently reported and ranged widely between studies from 1 to 15 days (level IV).
• There was inadequate evidence to determine variations in procedural duration or length of stay by type of procedure (level IV).

Perioperative complications
Four measures of harm were reported by a majority of studies within the review: overall complications, bleeding, fistulation and haematoma/sepsis. There was considerable heterogeneity in surgical morbidity, reported as overall procedural complication rates; with individual study rates varying from 0% to 61% (see Table 2). This heterogeneity may have reflected (for example) differing inclusion, procedural content, context of care, or thresholds or conventions for recording complications. Random effects meta-analysis found the overall complication rate to be 11.5% (95% CI: 7.2-16.6%), I 2 = 87% (Fig. 3). Overall complication rates varied within and between procedures without evidence to favour one or more procedures. Although variable, the bleeding complication rate was generally low being reported as zero in 50% of cohorts of procedures. Random effects meta-analysis found the pooled bleed rate to be 2.0% (95% CI: 0.7-3.6%), I 2 = 68%. The reported fistulation rate was consistently low (0% in 80% of cohorts). Random effects meta-analysis found the fistulation rate to be 0.0% (95% CI: 0.0-0.1%), I 2 = 0%. The rate of haematoma or sepsis varied between studies but was generally low (0% in 56% of cohorts). Random effects meta-analysis found the pooled rate of haematoma or sepsis to be 0.9% (95% CI: 0.2-2.0%), I 2 = 54% (Fig. 4). Although there was some evidence of variation by procedure, sub-group findings are based on selective reporting of relatively few and small studies and should be viewed with caution.

Long term adverse outcomes
Measures of long term harm were poorly reported: urgency (19%) and anal stenosis (13%). Although 37 cohorts (60%) made some assessment of dyspareunia, these assessments were not made in an adequately comparable fashion. Wide variations between studies may reflect inconsistency in the studies when reporting improvement, deterioration or de novo dyspareunia. Rates of mesh erosion were as high as 30% but were only reported in two studies [40,46]. This adverse outcome is devastating for the female sufferer and is rightly receiving extensive media coverage, focus by some regulatory bodies (e.g. Scottish Government) and much interest by surgeons themselves. It is thus unfortunate that in the narrowly defined context of this systematic review, evidence regarding this outcome was of poor quality.

Summary evidence statements: harms
• Evidence is drawn from observational studies and comparisons. Most comparisons featured considerable heterogeneity, which may have multiple causes (level IV).
• Overall procedural complication rates ranged from 0% to 61%. However, these complications typically occurred in about 7-17% of procedures (level IV).
• Fistulation was a very rare complication, not occurring at all in most studies (level IV).
• Mesh erosion was a common complication, but only reported in two studies (level IV).
• The occurrence of haematoma or sepsis was typically low at around 0-2% (level IV).
• Considering measures of harm, there was insufficient evidence to prefer one type of procedure over another (level IV).
• Long term adverse outcomes were poorly reported (level IV).
• Dyspareunia was reported too inconsistently to make meaningful comparisons between procedures, although it may be central to patient decision making (level IV).
• Only two procedure-related deaths were reported for 3209 patients included in studies (level IV).

Efficacy
Most studies used ad hoc questionnaire assessment of symptoms (n = 27) to assess efficacy; other tools used included Cleveland Clinic Constipation score (n = 2), Wexner constipation score (n = 3), obstructed defaecation score (n = 4) with single studies using the following: PAC QOL; Short Form-36 (SF-36); St Mark's       (Table 4). Further, some studies also reported individual symptoms. No study reported acquiring data objectively using personnel not involved in the surgical care of the patient or data collection blind to intervention status (RCTs were not observer-blinded). Data were too inconsistently reported to usefully analyse efficacy outcomes either in their natural units or after standardisation. Instead a binary response of global improvement (yes/no) was derived for each cohort of patients. Reported in 76% of procedures, random effects meta-analysis found global improvement to be 72.8% (95% CI: 66.8-78.3%), I 2 = 86% (Fig. 5); again there was considerable heterogeneity between findings. Given the crudeness and imprecision of estimates, as well as small numbers of studies/patients for some procedures, there is insufficient evidence that particular procedures provide higher levels of improvement.
For individual symptoms, changes in percentage of patients experiencing each symptom were not reported for a majority of studies. Straining was reported for 45% of total procedures; incomplete emptying for 47%; vaginal digitation for 50%. Other symptoms were less frequently reported. All symptom measures featured considerable heterogeneity and selective reporting by (generally) small studies limiting any comparison between procedures. These caveats accepted, the overall The aim of reinforcing the rectovaginal septum is to restore normal anatomy which theoretically will enable normal evacuation. Given the underlying aim of surgery is to correct anatomy, an assessment of anatomical recurrence is also important (although this is necessarily only a surrogate of clinical outcome). Studies variously and inconsistently reported clinical, radiological and symptom recurrence (Table 5). Clinical recurrence was reported for 44% of cohorts (Fig. 6). The overall clinical recurrence rate was 17.1% (95% CI: 11.7-23.3%), I 2 = 89%, with individual findings for procedures varying from 0% to 55%. There is a suggestion that site specific TVR features a higher clinical recurrence rate that other procedures, based on four studies and 349 patients, although this may be a chance finding given the selective reporting by studies (Fig. 6). One study showed that the results of a site specific repair are further compromised by the concurrent use of collagen mesh [15].

Summary evidence statements: efficacy
• Data on efficacy were inconsistently measured and findings heterogeneous, making estimates tentative and imprecise (level IV).
• Although inconsistent, assessments of patient global improvement typically suggest a good outcome in about 67-78% of patients (level IV). • Findings for global improvement, derived from global satisfaction rating scales, provide insufficient evidence to prefer one type of procedure over another (level IV).
• Other patient symptom scores were inconsistently reported (level IV).
• Approximately 30-50% of patients may experience reduced symptoms of straining, incomplete emptying or reduced vaginal digitation (level IV).
• Anatomical recurrence (as judged by a variety of measures) occurred in approximately 17% patients at mean follow up of 23.4 months (range 12-74) (level IV).

Patient selection
Patient selection is generally perceived by experts as important when choosing a surgical approach. There was a vast variation in the preoperative investigation of patients in the studies. Most studies included evacuation imaging (33 out of the total 43) where defaecation proctography was used in all but one (which used isotope imaging) [36]. There was selective use of anal ultrasound, anorectal physiological assessment, with colonic transit studies being recorded in 12 studies. There was however little correlation between the results of surgery and preoperative investigation results. Key features include the presence of a symptomatic  Table 5 Recurrence rates by type of operation rectocoele greater than 2 cm on proctography with evidence of contrast trapping at the end of maximal evacuation. Where graded, generally rectocoeles operated upon were grade 2 or 3. Only one study measured the rectocoele clinically with POPQ assessment [37]. Other than medical co-morbidity precluding surgery there were very few consistent exclusion criteria in any of the studies, although several excluded those women with slow transit constipation. Furthermore, in the few studies that related preoperative assessment to outcome, none could show an association between baseline symptoms or size of rectocoele and functional outcome. The need to digitate to assist evacuation did not appear to predict outcome following surgery, but may predict the need to digitate postoperatively.

Summary evidence statements: patient selection
• Although patient selection is perceived as vital in predicting outcome it was inconsistently documented (level IV).
• There was no evidence to support better outcomes based on selection of patients with a particular size or grade of rectocoele (level IV).

Conclusions
A systematic review of evidence for the perioperative and long terms benefits and harms of recto-vaginal reinforcement procedures to treat symptoms of constipation identified only one high quality study. Two further randomized controlled trials were identified although these were small and had methodological limitations. The evidence base was therefore characterised almost exclusively by observational studies of variable and often uncertain methodological quality. Future studies should provide robust and comparative evidence for clinicians to support patient decision making, both in terms of the incremental benefits and harms of procedures. Greater understanding is required of the mediating effects of prognostic factors particularly preoperative definition of both functional and radiological parameters that impact upon treatment success.