Surgery for constipation: systematic review and practice recommendations

To assess the outcomes of colectomy in adults with chronic constipation (CC).


Introduction Background and procedural variations
The concept of resection of the colon to treat constipation originates more than a century ago [1]. Modern surgical approaches mostly continue in the historical practice of removing the whole colon with anastomosis of the terminal ileum to the upper rectum (or very distal sigmoid) usually at the level of the sacral promontory. Usually termed colectomy and ileorectal anastomosis (CIRA), the procedure is also sometimes described as colectomy with ileoproctostomy. Total colectomy is not favoured by all surgeons and other less radical colonic resections may also be employed. The simplest variation is to perform a subtotal colectomy and ileosigmoid anastomosis (SCISA) but an increasingly popular choice is subtotal colectomy with sparing of the caecum and thence caecorectal anastomosis (SCCRA). Since this is not a common procedure in routine colorectal surgical practice, it merits some background description. First described by Ogilvie (1931), retention of the ileocaecal junction has the theoretical advantage of preservation of absorptive functions (bile, vitamin B12 and electrolytes) and thus perhaps reduced diarrhoea. No standard technique exists for creating a CRA. The general principle involves colonic mobilization followed by ligation of all vascular pedicles except the ileocolic branches. In the technique proposed by Lillehei and Wangensteen (1955) a 180°rotation of the remaining mesentery from the right to the left is performed to place the caecum in the left iliac fossa, with apex cephalad. During the rotation the remaining mesocolon passes over the aorta, and it is sutured to the mesorectum and to the third portion of the duodenum to avoid internal hernia or intestinal obstruction, which may complicate such technique. Deloyers (1963) proposed a variation of this technique in which there is a craniocaudal rotation of the caecum to allow a pelvic isoperistaltic CRA (IPSCCRA) but this required a retroileal tunnel and 180°torsion of the vascular pedicle which may result in ischemia or venous stasis. These difficulties have in part be mitigated by development of an antiperistaltic end-to-end caecorectal anastomosis (attributed to Sarli [2]) (APSCCRA) which avoids the vascular problems due to the torsion of the pedicle, obviates the need to tailor the caecum and lowers the risk of intestinal obstruction due to the rotation of the mesocolon in front of the aorta.
These procedures have been variously employed for chronic constipation with or without ileal pouch formation in small numbers of patients and usually as a salvage after failed colectomy; 2 Subtotal colectomy and modification of the rectal reservoir (modified Duhamel procedure; Jinling procedure); 3 Colonic exclusion and ileorectal anastomosis i.e. without resection.

Previous reviews
Narrative reviews focused on the outcome of colectomy for constipation have been published in 1996 [10], 1999 [11], and 2006 [12]. No previous systematic review was identified.

Summary of search results and study quality
The search yielded a total of 85 manuscripts for full text review (Fig. 1). From these, 40 articles published between 1988 and 2015 contributed to the systematic review, providing data on outcomes in a total of 2045 patients (range 20-144 patients per study) based on 39 defined patient cohorts (Table 1). A US nationwide dataset derived from hospital episode statistics was also included covering 2377 procedures coded as colectomy for constipation indications [19]. Specific exclusions after full-text review (and after exclusion of non-English language publications: n = 10) included 27 studies where the population sample was confirmed to be less than 20 patients, four studies of out of scope procedures, one study where data were considered duplicate [13], one where outcomes could not be segregated by eligible procedure [14], and one where data for multiple clinical indications for colectomy were merged [15]. The general quality of studies was poor due to inadequate description of methods. The 40 included studies comprised: a single poor quality randomized trial (uncertain or high risk of bias in most domains) [20] (Oxford level IIB); one good quality prospective [21] and one retrospective cohort study [22] (level IIB); and 37 level IV studies (comprising 14 poor quality cohort studies, i.e. 'case comparison studies'; one poor quality case-control study with non-consecutive controls; eight prospective case series; and 14 retrospective case series). A general problem was the lack of prospectively defined follow up intervals. Patient follow up ranged from 12 months to 11 years (median 47 months) but this clearly varied greatly for individual patients within studies without defined follow up periods. Eleven studies derived from US centres, 11 from European centres, nine from Chinese centres with the remaining nine spilt across five countries.

Perioperative data
Perioperative data were reported by 37 studies ( Table 2). Reporting of procedure duration was inconsistent but mean procedural duration ranged from approximately 2-4 h. Within this variation were trends of shorter operating times for open vs laparoscopic procedures (e.g. colectomy and ileorectal anastomosis (CIRA), median open: 167 min vs median laparoscopic: 210 min), as well as for subtotal procedures: median 135 min. The average length of stay (LOS) reported was 10.4 days, ranging from 7.0 to 15.5 days duration. However laparoscopic procedures consistently reported shorter lengths of stay e.g. the median LOS for open CIRA was 10.6 days compared to 8.1 days for laparoscopic CIRA. This evidence is supported by individual cohort comparisons [34,55] and in the single RCT where mean LOS was reduced from 9.7 to 7.6 days with laparoscopy [20].
Summary evidence statements: perioperative data 1 Length of stay after colectomy for constipation is 7-15 days, even in the modern era (level IV). 2 Laparoscopic surgery may be associated with longer operating times and modest decreases in length of stay (from 10 to 8 days), however there is considerable variation between studies (level IV).

Perioperative complications
Presented meta-analyses showed considerable heterogeneity of complications, not explained by procedure or age of publication. The attentiveness to harm recording and the duration of recording were inconsistent and studies limited only to laparoscopic procedures are characterized by small numbers limiting scope for comparison with open procedures and no adjustment has been attempted for potential differences in the populations recruited into individual studies. Consequently estimates of harm provided are necessarily tentative. Surgical morbidity remains a concern for all types of colectomy with total complication rates. A random effects meta-analysis estimated total complications to be 24.4% (95%CI: 17.8-31.7%); I 2 = 88.1% (Fig. 2), although findings were heterogeneous including individual study rates from 7% to 54% (Table 2). Aside from the incidence of anastomotic leaks and other more general complications (high even in some recent series from expert centres [56] and including six fatalities in 1568 patients: 0.4%) the incidence of prolonged post-operative ileus (POI) and early adhesional small bowel obstruction (SBO) are known to be disproportionally high for patients undergoing colectomy for slow-transit constipation when compared to other indications [17].
A random effects meta-analysis estimated early postop POI/SBO to be 9.7% (95%CI: 5.7-14.6%); I 2 = 87.9% (Fig. 3), although findings were heterogeneous including individual study rates from 0% to 33%. Current findings are inconclusive as to whether laparoscopic or open surgery are safer: there is some suggestion however that the POI rate is lower in newer studies. Rates of further surgical intervention for POI in the perioperative period were similarly heterogeneous: 2.7% (95%CI: 1.0-5.0%) I 2 = 64.3%, including study rates from 0% to 15%.
These data have been put into a broader context by the recent (2015) US national database study of 2377 colectomies for constipation between 1998-2011 [19], providing 60% of all subjects within the review. This study based on registered health episodes re-affirms a high rate of perioperative complications in a national sample (42.7% patients during index hospitalisation 30day period), with the main contribution (27%) coming from "intestinal obstruction, ileus, nausea & vomiting".

Long-term adverse outcomes
Long-term rates of SBO reported by studies were heterogeneous: 15.2%, (95%CI: 10.2-20.9%) I 2 = 85.5%, including study rates from 0% to 71% (Table 3; Fig. 4a). Re-operation rates (principally for SBO but also other severe ongoing functional  problems) were similarly heterogeneous 13.3%, (95%CI: 8.6-18.7%) I 2 = 87.7%, including study rates from 0% to 45% (Fig. 4b). Particular to colectomy for constipation is the concept that laparoscopy might reduce the well-established high incidence of post-operative SBO. The review provided only limited data from small studies comparing open with laparoscopic procedures, although SBO rates appeared much lower. Ho et al. [34] found that early adhesion formation leading to bowel obstruction was more frequent in patients undergoing laparoscopically assisted colectomy (29%) compared to open (13%). A larger series of 124 patients also showed no differences in post-operative morbidity between approaches [55]. Conversely, a low quality case-control study of 15 laparoscopic vs 15 open subtotal colectomy with antiperistaltic CRA showed that bowel obstruction rates were halved (from 13.3 to 6.7%) in the laparoscopic groups [54]. The follow up in these studies (12-20 months) was generally shorter than the average (47 months) although very high rates of SBO were reported by a study of exclusively open CIRA with 12 months follow up [52]. Finally, while the results for laparoscopic approach offer some optimism based on the small numbers of patients in these studies, no differences were observed in complication rates between open and laparoscopic procedures in the US nationwide survey of 2377 colectomies [19]. However, the most revealing conclusions can be drawn from further analysis of US national database study in which longitudinal data were recorded on 166 patients recorded on State Inpatient Databases of Florida and California (2005)(2006)(2007)(2008)(2009)(2010)(2011). These data agreed with the whole national dataset (n = 2377) in confirming high perioperative (30-day) complication and re-admission rates, but also showed that resource utilisation in the form of emergency department visits, hospitalisation and surgical intervention remained high in the following 1 year. Excluding the colectomy itself, these 166 patients had a total of 2355 encounters, which included 1494 emergency department visits and 861 hospitalisations by 149 and 144 patients, respectively. Among the 1494 emergency department visits, the 674 that occurred postoperatively were shared across 119 (72%) patients; among the 861 hospitalisations, 488 occurred after colectomy and affected 110 (66%) patients. A breakdown of the motivation for these attendances reveals the well-described issue of ongoing abdominal pain, which as well as other gastrointestinal symptoms and postoperative complications, increased after colectomy.

Efficacy
Measurement of outcome was inconsistent, including variable use of validated and un-validated scoring instruments for symptoms e.g. Cleveland Clinic Constipation Score or quality of life (QoL), GI quality of life, individual symptom reporting and global 'success' or 'satisfaction' ratings (GSR) obtained via a variety of methods (where 'satisfied' or 'very satisfied', 'good', 'very good' and 'excellent' were interpreted as positive outcomes).
No study documented that data were acquired objectively by using personnel not involved in the surgical care of the patient. Only one study documented that collection of data was blind to intervention status [20] and this RCT only blinded observers for 4 days while presenting follow up data to 32 months. Average  (Table 4; Fig. 5). Again study findings are heterogeneous, with no clear advantage for any particular procedure or surgical approach. However, such levels of satisfaction can be related to marked changes in bowel frequency (generally from a mean of once per week to three times per day in the 14 studies reporting both variables) (Table 5), and where recorded (three studies only), marked changes in summative symptom scores e.g. the Cleveland Clinic Constipation score reduced from a mean of > 20 points pre-operatively (indicative of severe constipation) to approx. 2-3 points (low normal range) post-operatively. Individual symptom outcomes highlighted the well documented problems of diarrhoea: 9.8% (95%CI: 4.7-16.4%), I 2 = 76.9% (Fig. 6); and incontinence: 7.4%, (95%CI: 2.2-14.7%), I 2 = 90.8% following colectomy, ongoing or recurrent constipation: 18.2%, (95%CI: 9.3-29.2%), I 2 = 91.4%; persistent (or worsened) abdominal pain: 39.3%, (95%CI: 28.8-50.1%), I 2 = 89.0%; and bloating 23.9%, (95%CI: 11.9-38.1%), I 2 = 92.7%. Poor functional outcomes contributed to further resection or permanent stoma: median 5% (range 0-28%) patients when reported (by only seven studies; data not shown).
Meta-analyses of efficacy outcomes featured considerable heterogeneity, not explained by procedure or age of publication. Given the different duration of studies and variable follow up within studies there is also the potential for time-confounding. Studies limited only to laparoscopic procedures are characterized by small numbers limiting scope for comparison with open procedures and no adjustment has been attempted for potential differences in the populations recruited into individual studies. Consequently efficacy estimates are tentative.
Accepting the caveat that only a minority of studies reported functional variables, several observations can be made regarding functional outcomes in studies of less radical colonic resections (Tables 5b and c) compared to those for CIRA (Table 5a). The general premise of such procedures is to reduce the risk of longterm diarrhoea and incontinence and this concept is in part supported by data that, accepting small study numbers and heterogeneity, suggest potential to reduce rates of diarrhoea for segmental and subtotal resections (Fig. 6). However, this was at the cost of increased ongoing or recurrent constipation (median 8.7% for CIRA compared to 26.8% for more conservative resections). The latter has proved a particular problem for segmental resections (right or left hemicolectomy) with generally poor results compared to colectomy mainly due to unresolved constipation requiring further intervention (Table 5c). De Graaf et al. [28] used segmental transit (radio-opaque marker) methodology to select patients for partial left sided colectomy or subtotal colectomy. Whilst results as a whole were disappointing, the study concluded that in terms of complications and functional outcome, there was little difference between procedures, and that a more limited resection was therefore a reasonable option in this selected group. You et al. [35] reported the use of left, right or subtotal colectomy based on segmental transit time measurements with excellent results. Further, in the three cases where constipation recurred following segmental resection, a subtotal colectomy was undertaken successfully at a later date. This experience was not however repeated by Lundin et al., [44] when recurrent constipation was experienced by 46% patients despite transitguided resection. Thus, while in the laparoscopic era where there is a greater theoretical advantage of not meeting the technical challenges of mobilizing both colonic flexures laparoscopically, the tailoring of segmental resections using these specialist investigations of transit is inconsistently supported by published data. Further, the tests required to accurately determine resection level e.g. isotope scintigraphy have limited availability.
Subtotal resection with ileosigmoid anastomosis is generally considered less effective than ileorectal anastomosis based on several relatively small case series mixing both procedures (Table 4d). Contemporary data on subtotal resections with CRA come mainly from a few institutions in Italy and China. Conclusions from these studies vary. For example, Li et al. [57] demonstrated good results for both isoperistaltic CRA and CIRA. Feng et al. [22] compared isoperistaltic CRA with SCISA. Surgical safety outcomes and length of stay were similar but patients were more satisfied after ileosigmoid anastomosis mainly due to ongoing constipation in caecorectal group. However patients experienced slightly less  diarrhoea and incontinence after caecorectal anastomosis. Jiang et al. [50] compared antiperistaltic CRA with ileorectal anastomosis. Again there were no differences in post-operative course, however patients undergoing caecorectal anastomosis had less diarrhoea, higher postoperative quality of life (not recorded pre-operatively) and overall reported GSR (88 vs 65%). Laparoscopic surgery has the theoretical advantages of better cosmesis (especially in young women) and perhaps lower long-term complication rates (see above). Such factors have not however yet translated into improved functional outcomes mainly because these have not yet been the focus of comparative studies. Ho et al. [34] found no difference in GSR between open and laparoscopic CIRA (96 vs 100%). In the case-control study of Marchesi et al. [54], despite the halving of SBO rates (from 13.3 to 6.7%) in the laparoscopic groups, long-term functional outcomes and GI quality of life were very similar.   Table 5 Functional outcomes by procedure.

Author
Year Procedure

Patient selection
While clinical experience suggests careful patient selection for procedures is important, few studies systematically addressed this issue [12]. Main findings from studies that stratified outcomes based on baseline phenotype are included in Table 6. These studies provide some information on clinical characteristics but more so on results of specialist physiological testing. Pikarsky et al. [39] studied whether colectomy can be performed in elderly patients (defined 65-80 years in their series). Although overall success was diminished on the older age group (64% vs 95%, P = 0.01), the authors concluded that the results were acceptable and that the procedure was safe based on no increase in observed morbidity. The question of whether the presence of severe psychological problems adversely influences outcome has been discussed by studies that noted both poor outcomes and a number of post-operative psychological problems including suicide [23]. Others have made post-hoc correlations between prior psychiatric disease and poor outcome [31]. This factor was only addressed as a stated aim by Hasegawa et al. [37], who  reported a statistically significant prejudicial influence of 'severe psychological disorder'.
Outcomes of colectomy are improved by selection of patients with proof of slow colonic transit. Although it could be argued that other factors may have also influenced outcomes (e.g. mix of surgical approaches, surgical technique and equipment), this statement is corroborated by comparing outcome data from an era when specialist investigations of transit were variably applied [23][24][25]37] with subsequent studies that always performed transit studies and used these as a selection criteria. Most contemporary studies also evaluated anorectal physiology especially in relation to the diagnosis of a combined slow-transit and defaecatory disorder phenotype. The management of this patient group remains contentious. Bernini et al. [36] in a study of 106 patients demonstrated that despite preoperative biofeedback training, patients with non-relaxing pelvic floor (n = 16) had significantly higher rates of recurrent defaecatory difficulty (38 vs 4%), and lower rates of satisfaction after colectomy (56 vs 78%). However, three other studies (Table 6) found little effect on functional outcome or complication rates when functional or structural defects were addressed prior to colectomy. These studies included the contemporary Cleveland Clinic experience of 144 patients where obstructed defaecation (n = 41) had no influence on outcome from laparoscopic or open colectomy [56].
It is generally accepted that some patients with slow colonic transit also manifest upper GI symptoms (especially nausea and vomiting). Abnormalities of oesophageal, gastric and small bowel function can be demonstrated in a proportion of patients by a variety of methods [18]. Ghosh et al. [32] showed that the high proportion of patients undergoing colectomy who subsequently developed SBO episodes (71% with 42% requiring surgery in their series) were more likely to have non-colonic visceral and autonomic nervous system abnormalities on post-operative testing. This observation has been considerably strengthened by the prospective cohort study of Redmond et al. [21]. A significant fall in long-term success rate (to 10 years) as a result of persistent constipation, abdominal pain and distension) was observed in patients defined as having a generalized intestinal disorder (GID) on the basis of having both upper and lower GI dysmotility using a battery of intraluminal tests. Successful outcome was observed in only 12.5% patients with GID vs 90%: without.
Summary evidence statements: patient selection 1 Outcomes of colectomy may be poorer in patients with significant psychological disorder (level IV).
2 Outcomes of colectomy may be improved by selection of patients with definitive proof of slow colonic transit (level IV). 3 Outcomes of colectomy are inconsistently influenced by concomitant rectal evacuation disorder although data suggest that structural and functional defecation disorders, if evident, should be treated prior to colectomy (level IV). 4 Outcomes of colectomy may be prejudiced by preoperative evidence of upper gastrointestinal dysmotility (level IV).

Conclusions
A systematic review of evidence for the perioperative and long-term benefits and harms of colectomy identified no high quality studies. The evidence base is characterised by observational studies of variable and often uncertain methodological quality. Current data suggest a balance of harms against efficacy with evidence that outcomes are at best variable. Future studies should provide high quality evidence for clinicians to support patient decision making, both in terms of the incremental benefits and harms of colectomy and in understanding the effects of prognostic factors upon treatment success.