Surgery for constipation: systematic review and practice recommendations

This manuscript forms the final of seven that address the surgical management of chronic constipation (CC) in adults. The content coalesces results from the five systematic reviews that precede it and of the European Consensus process to derive graded practice recommendations (GPR).


Introduction
This manuscript forms the final of seven that address the surgical management of chronic constipation in adults. The content coalesces results from the five systematic reviews that precede it and of the European Consensus process to derive graded practice recommendations.

Methods
These have been fully explained in the 'Introduction and Methods' paper. Procedures have been grouped as: 1 Colonic resection, including total colectomy, subtotal and segmental colectomy (with some anastomotic variations for subtotal colectomy) by open and laparoscopic approaches; 2 Rectal suspension procedures, including forms of open and laparoscopic rectopexy; 3 Rectal excisional procedures, including stapled transanal rectal resection (STARR) and intra-anal Delorme's; 4 Rectovaginal reinforcement procedures, including trans-vaginal and trans-anal approaches with or without mesh; 5 Sacral nerve stimulation.
Results have been presented as follows: 1 Summary tables of results where these could be compared between classes of procedure based on homogeneous outcomes; 2 Graded practice recommendations. All prototype GPRs have been documented with consensus statistics and thence a clear indication of those that were upheld (found to be appropriate) by consensus. A final section addresses implications for future research. Note: consideration was given to summarizing all summary evidence statements in this manuscript however these are covered in each individual review and were omitted here for brevity.

Results
Study characteristics Table 1 repeats the information provided in the 'introduction and methods' paper on overall study characteristics by procedure. As previously noted, the overall quality of evidence was poor with 113/156 (72.4%) providing only level IV evidence. The best evidence was extracted for rectal excisional procedures where the majority of studies were level I or II.

Summary of systematic review data
In each of the five reviews, results were presented for perioperative variables, harms (post-operative complications and long-term adverse events), efficacy and prognostic factors. These data have been presented together below.

Perioperative variables
Data were available for nearly all procedure classes (except SNS) on operation duration and length of stay (Table 2, Figure 1). Not unsurprisingly, colectomy had the longest operative duration and length of stay. For the three classes of rectal procedure lengths of stay were similar, however duration of surgery was clearly longer for rectal suspension (rectopexy) and shortest for rectal excision -in effect for forms of stapled trans-anal resection (STARR).

Harms
There were large discrepancies in harm recording with selected outcomes being based on a priori knowledge of recognized harms for each class of procedure. Given considerable heterogeneity in reporting (covered in the individual reviews), it was only possible to summarize main harms semi-quantitatively (Table 3). A mortality rate of approximately 1/200 occurred after colectomy. Other procedures had no recorded mortality or a very low rate (rectovaginal reinforcement procedures: 1/1600). Colectomy was associated with substantial risks in the short and longterm, particularly in relation to small bowel obstruction and poor functional outcomes. Other procedures had generally fewer complications, including some where review data reflected concerns expressed widely in the international surgical community, notably mesh complications after rectopexy and chronic pain AE urgency after STARR.

Efficacy
Few variables could be analysed across procedure classes on the basis that, like harms, outcomes chosen tended to be bespoke to each procedure class. It was however possible to summarise global satisfaction ratings, i.e. the proportion of patients self-reporting a good or excellent outcome. Accepting the considerable limitations of such outcomes, data in Table 4 show that all procedures are almost equallywell received by patients with rates around 70-85% for all.

Patient selection
For most classes of procedure, some information could be obtained about prognostic baseline characteristics that might guide patient selection. In all instances, the level of evidence was poor with no formal stratified medicine studies and very few (if any) adequately powered post-hoc analyses of good quality cohort studies. Table 5 summarizes the broad phenotypes of patients that may most benefit from each procedure and some negative prognostic features.

Graded practice recommendations
A series of tables (Table 6 a-e) show all GPRs proposed by the clinical guideline group by main procedure class. The outcomes of the consensus process have been presented as median score (1)(2)(3)(4)(5)(6)(7)(8)(9) and by classification based on RAND-UCLA methodology: appropriate; uncertain and inappropriate. The reader is reminded that appropriateness is not directly extrapolated from the median score but rather the overall data distribution (see introduction and methods).

Discussion
This manuscript summarises the body of data from five systematic reviews and presents new graded practice recommendations.

Summary of systematic review data
The overall quality of data was poor with 113/156 (72.4%) of included studies providing only level IV evidence, thus greatly limiting the number and grade of summary evidence statements. This was a particular problem for colonic resection, rectal suspension procedures and sacral nerve stimulation, where nearly all data were derived from level IV studies. The limitations of such observational data are well acknowledged and are a source of concern when used as a basis for promoting procedures. For instance, colectomy for slow-transit constipation would, based on systematic review of 40 observational studies, appear to be an attractive prospect with 86% global satisfaction rate (the highest of any of the studied classes of procedure). However, recently published US retrospective cohort data on over 2000 patients [1] paint a very different picture of high complication rates and greater long-term post-procedural health utilization (ambulatory care, hospital admissions, radiology etc.) than before surgery. It is difficult to reconcile such disparity [2], and the increasing rates of colectomy for constipation in the US [1] also seem at odds with international opinion (that promotes extreme caution).
Sacral nerve stimulation also had generally supportive observational evidence based on seven included studies. However, subsequent randomised studies [3,4] directly contradict these data and most centres no longer offer SNS for the constipation indication. Perhaps the greatest area of academic contention in the pelvic floor community concerns the choice of procedure to address dynamic structural abnormalities of the pelvic floor that lead to prolapse and obstructed defaecation symptoms. The results presented here do little to help resolve this issue and certainly cannot help underpin a much needed treatment algorithm for such patients. In effect, all have similar global satisfaction ratings, similar lengths of stay and complication profiles that are to some extent procedure-specific. Based on reviewed indications, rectal suspension and excision procedures can be applied to patients with rectal intussusception and/or rectocoele and rectovaginal reinforcement procedures to rectocoele only. Aside from a generally longer operating time for rectopexy (and shorter for STARR), decision making for a patient with one or both of these abnormalities currently rests with personal views about the acceptability of certain complications and (possibly) surgeon enthusiasm for       type of approach and surgical instruments (flippantly whether the surgeon prefers basic surgical instruments, laparoscopy or staplers). With respect to complications, limited reporting prevented much discussion beyond the importance of counselling patients about established complications (covered in GPRs). However, it is tempting to speculate that future stratification might provide the opportunity to select patients for one or other procedure e.g. avoiding patients with certain prior phenotypic features or modifying risk. An example would be chronic pain development, where perhaps STARR should be relatively contra-indicated in patients with preceding evidence of pain syndromes (e.g. migraine, fibromyalgia or chronic back pain) or modified using one of a number of available agents to prevent sensitization during surgery e.g. pre-operative gabapentin or intra-operative ketamine [5]. At the very least the data provide the opportunity to appraise patients with the options and their complication profiles where more than one surgical option exists. Another difficulty with interpretation was that inclusion (in the review) necessarily reflected the availability of studies, in turn reflecting the tendency to publish studies of new techniques rather than well-established ones. Higher quality data were available for rectal excisional procedures due to several prospective cohort studies and small RCTs of the STARR procedure (and variations). It is well acknowledged that this body of data, including over 8000 patients, reflects a period of intense popularity for this procedure (nearly all published in the decade 2004-14) with (interestingly) no included papers arising from the final 18 months of the review period. The large numbers are also known to reflect industry investment in several data registries, two of which included over 2000 patients. Anecdotal evidence and expert opinion from international meetings is that the popularity for this procedure has waned (even in Italythe origin of the procedure and its main proponents). Such a peak and decline in popularity was not present for other procedures that were more evenly spread across the review period.

Graded practice recommendations
The clinical guidelines group developed a total of 100 'prototype' graded practice recommendations by taking forward summary evidence statements from the five reviews and combining these with expert opinion and a small number of RCTs (SNS only) published after the extraction data (22/02/2016). These statements covered patient selection, procedural considerations and patient counselling. The limitations in review evidence meant that only 59/100 prototype GPRs were directly derived from summary evidence (level II-IV; grades B-D) with the remainder, 41/100 derived by expert opinion only (level V; grade N). Of the 100 total, 85 were deemed 'appropriate' based on the independent scoring of 18 European experts and the remaining 15 were all deemed uncertain, i.e. none was considered inappropriate by the panel. This is a high level of consensus for a single round of questioning and suggests that there is reasonable European agreement as to selection of patients for each class of procedure, which procedure to perform and how to counsel the patient (often related to outlining potential harms). However, this does not signify unequivocal evidence of value for these recommendations and they do not represent minimum standards, but can act as a basis for further research and guideline development.
The 15 'uncertain' GPRs were spread across procedures with most in colectomy (n = 7) and least for rectal suspension (1) and SNS (0). The majority concerned patient selection (n = 8). Interestingly, only 5/ 15 (33.3%) related to prototype GPRs based only on expert opinion (level V, grade N). The remaining 10 included five where uncertainty by consensus accurately reflected uncertainty by grade (D) (33.3%), three with grade C summary evidence from the systematic reviews (20.0%) and two with grade B evidence (13.3%). There was thus no strong suggestion that grade weighed panelist opinion. The two grade B statements deemed uncertain both concerned rectal excision: first that 'rectocoele only' was an indication in terms of benefiting the patient; and secondly that significant complications such as sepsis, anastomotic dehiscence and bleeding can occur post-procedure in approximately 2% (1-4%) of patients. The panelist consensus on these two GPRs is surprising since both would seem to reflect widespread practice and knowledge, respectively. Overall, while it would be possible to have further rounds of consensus building among the European panel, the GPRs as stands are a good start to develop future clinical guidelines.

Future research recommendations
With the exception of rectal excision, there are clear deficiencies in the current evidence base for all surgical procedures directed at the management of chronic constipation in adults. This was particularly true in terms of availability of randomized controlled trials, where only four reviewed studies met criteria for level I evidence. The difficulties in conducting randomized trials for complex interventions such as surgery are well rehearsed [6], but their importance is exemplified by recent SNS RCTs [3,4] that directly contradict observational data. While it can be argued that sham surgery would be difficult to justify for patients with a chronic debilitating condition, it is disappointing that no level I evidence has been produced to compare classes of procedure where more than one is appropriate. Such comparison trials of different techniques may face problems of equipoise and interventional fidelity, and might need to overcome a speciality divide e.g. posterior repair vs transanal repair of rectocoele (the former performed largely by gynaecologists or urologists specializing in female patients and the latter by colorectal surgeons). An alternative is waiting-list designs where the wait time for surgery can be randomized and analysis-based on longitudinal outcomes before and after intervention [7]. An example of such a study is the CapaCiTY03 stepped-wedge randomised controlled study of laparoscopic ventral mesh rectopexy in adults with chronic constipation [8].
Accepting the difficulty in performing RCTs, there is still much opportunity to improve the evidence base by encouraging high quality observational studies. Prospective cohort studies could benefit from incorporating some of the scientific rigor of RCTs to limit obvious sources of bias e.g. by multicentre recruitment and use of blinded observers to collect outcomes. Awareness of reporting standards by authors and journals may in turn feed better protocol-driven research [9]. They should incorporate the few validated patient-reported outcome measures (PROMS) that are available e.g. PAC-QoL and PAC-SYM, internationally-accepted HR-QoL measures e.g. EQ-5D-5L and monitor harms in a systematic manner using established systems e.g. Clavien-Dindo [6]. They should also consider collecting health utilization data from patient information systems, the importance of which is illustrated by the Dudekula study [1] of colectomy.
The CCG make the following recommendations as research priorities: 1 Colonic resection: there is a need to determine prospectively and robustly the risks and benefits of this procedure. Considering its low incidence, a prospective cohort study across Europe (or internationally) is recommended. Observer-blinded outcomes (above) should be systematically recorded at regular intervals to 5 years. Standardised baseline phenotyping may permit determination of outcome predictors if numbers are large enough. Consideration could be given to a control group not undergoing surgery (although selection bias is acknowledged). All procedural variations could be evaluated although the main comparison of interest is now considered to be between more (total colectomy) and less radical (subtotal) laparoscopic resections. A double-blind RCT of this latter comparison might also be possible with international effort. 2 Rectal procedures for dynamic structural abnormalities of the pelvic floor. A UK RCT is underway to evaluate laparoscopic ventral mesh rectopexy [8]. A further RCT is however recommended to determine outcomes of repair of large rectocoele (in isolation), comparing posterior repair of the vagina vs transanal repair. It is acknowledged that this might require an expertise-based design [6,10] but it is an unanswered question for the indication of chronic constipation or obstructed defaecation. Systematic review data would also support a randomized comparison of STARR with rectopexy for patients with high-grade intussusception and rectocoele. However, expert opinion suggests that STARR is no longer popular. An alternative would be to perform a prospective cohort study (akin to colectomy) capturing all current practice. This could be performed internationally but might also be possible in a single country where all three main classes of procedure are still commonly utilized.

Conclusions
This manuscript concludes the series of seven, systematically detailing the outcomes of the main surgical procedures directed toward patients with chronic constipation. The current evidence base is poor and heavily reliant on low-quality observational data. On this basis, all procedures reviewed had generally positive (supportive) data. Several authors expressed concern that such data might not reflect the reality of clinical practice. While bias in such observational study designs is well recognized, it is possible that in surgical studies (usually performed by the proponents of the surgery) bias is both unidirectional (favouring the intervention) and powerful. Not only should this lead to a greater willingness to design and deliver high quality controlled trials, but also to an essential understanding that retrospective observational studies should be interpreted with caution. However the finding of widespread consensus for graded practice recommendations is encouraging. The stage is now set for recognised professional bodies worldwide e.g. Societies of Coloproctology/Colorectal surgery to build on this work by supporting the efforts of their membership to address future research recommendations and/or to help convert the recommendations documented in this series of papers into their clinical guidelines.