Plain language summary
Lateral pararectal versus transrectal stoma placement for prevention of parastomal herniation
A parastomal hernia is defined as an incisional hernia related to a stoma and belongs to the most common stoma-related complications. Many factors concerning the operative technique that are considered to influence the incidence of parastomal herniation have been investigated. However, it remains unclear whether the enterostomy should be placed through or lateral to the rectus abdominis muscle in order to prevent parastomal herniation and other important stoma complications for patients.
Nine retrospective cohort studies with a total of 761 participants met the inclusion criteria. All included studies reported results for the primary outcome (parastomal herniation), and one study also reported data on one of the secondary outcomes (stomal prolapse). None of the included studies compared the two interventions with regard to other secondary outcomes. There was neither a significant difference in terms of the risk for parastomal herniation nor with regard to the occurrence of stomal prolapse.
In summary, the quality of the identified evidence is too poor to allow a robust conclusion regarding the objectives of the review. This highlights the need for randomized trials to evaluate the effectiveness of the lateral pararectal versus the transrectal approach in preventing parastomal herniation and other stoma-related and patient-important morbidity in people requiring enterostomy placement.
Pararectal latéral par rapport à une mise en place de stomie transrectale pour la prévention d’hernies parastomales
Une hernie parastomale est définie comme une hernie incisionnelle liée à une stomie et est la complication la plus courante relative à la stomie. De nombreux facteurs concernant les techniques opératoires, qui sont considérés comme ayant une influence sur l'incidence d’hernies parastomales, ont été étudiés. Cependant, on ignore si l'enterostomie devrait être placée via ou latéralement au muscle grand droit de l’abdomen afin de prévenir une hernie parastomale et d'autres complications importantes de la stomie chez les patients.
Neuf études de cohortes rétrospectives portant sur un total de 761 participants remplissaient les critères d'inclusion. Toutes les études incluses rapportaient des résultats pour le critère de jugement principal (hernie parastomale) et une étude a également rapporté des données sur l'un des critères de jugement secondaires (prolapsus de la stomie). Aucune des études incluses ne comparaient les deux interventions concernant les autres critères de jugement secondaires. Il n'y avait aucune différence significative en termes de risque d’hernie parastomale, ni en en termes de survenue de prolapsus de la stomie.
En résumé, la qualité des preuves identifiées est trop faible pour permettre une conclusion solide concernant les objectifs de la revue. Ce qui souligne le besoin d'essais randomisés pour évaluer l'efficacité de la pararectal latérale versus l'approche transrectale dans la prévention d’hernie parastomale, les complications relatives à la stomie, la morbidité chez les patients nécessitant une mise en place d’enterostomie.
Notes de traduction
Traduit par: French Cochrane Centre 14th January, 2014
Traduction financée par: Minist�re Fran�ais des Affaires sociales et de la Sant�, Instituts de Recherche en Sant� du Canada, Minist�re de la Sant� et des Services Sociaux du Qu�bec, Fonds de recherche du Qu�bec Sant� et Institut National d'Excellence en Sant� et en Services Sociaux
Description of the condition
A parastomal hernia is an incisional hernia related to a stoma (Pearl 1989). Parastomal herniation is one of the most common stoma-related complications (Carne 2003). On physical examination, a parastomal hernia is diagnosed if bulging restricted to the peristomal area occurs during the Valsalva maneuver and a fascial defect can be palpated (Cingi 2006; Sjodahl 1988). On a computed tomography (CT) scan, a parastomal hernia can be described as "a loop of intestine or any abdominal organ, as well as preperitoneal fat, protruding through the defect alongside the ostomy" (Cingi 2006).
The incidence of parastomal hernias varies greatly between studies, but most likely ranges from 30% to 50% depending on the length of follow-up and on the type of enterostomy (Israelsson 2008). The consequences and implications of parastomal hernias are highly relevant for patients. Many patients complain of pain, suffer from an impaired body image, and have increasing difficulties to apply the stoma bag properly leading to leakage and skin irritation (Martin 1996; Ripoche 2011). In rare cases, bowel incarceration occurs requiring emergency surgery (Cuthbertson 1977). In addition, stomal herniation necessitating reoperation and hospitalization has a negative socioeconomic impact, especially since recurrence rates after surgical correction are high (Israelsson 2008; Tekkis 1999), ranging up to 50% within an average of six months after operative repair (Ripoche 2011).
Description of the intervention
Two interventions were compared, placement of the stoma lateral to versus through the rectus abdominis muscle (lateral pararectal versus transrectal).
Lately, mesh reinforcement of the stoma has been advocated to lower the incidence of parastomal herniation. Two recent systematic reviews of three RCTs (Hammond 2008; Janes 2004; Janes 2009; Serra-Aracil 2009) with a total of 129 patients have approached the question, whether prophylactic implantation of prosthetic mesh concurrently with the index operation could reduce the occurrence of parastomal herniation (Shabbir 2012; Wijeyekoon 2010). Meta-analysis showed that mesh reinforcement was associated with a reduction in parastomal herniation as compared to conventional stoma formation with no excess morbidity.
How the intervention might work
Lateral pararectal stoma placement may have a different incidence of parastomal herniation in comparison to transrectal stoma formation. Since the transrectal stoma is surrounded by the thick muscle layers of the rectus abdominis muscle, one could expect a tighter fit and a reduced incidence of parastomal herniation due to the muscle contractions. On the other hand, one could argue that preserving the integrity of the rectus abdominis muscle and its sheath, as it is the case in lateral pararectal stomas, minimizes anterior abdominal wall disruption and consequently reduces the risk of parastomal hernia (Evans 2011; Stephenson 2010).
Why it is important to do this review
Parastomal herniation is one of the most common complications after stoma placement. Since parastomal hernias often have a negative impact on the patient’s health and quality of life, the operative techniques should be optimized in order to prevent this postoperative complication. The need for prevention becomes even more relevant in the light of the high recurrence rates after operative correction of parastomal hernia, ranging from 50% to 76% after aponeurotic repair and from 24% to 86% after relocation, respectively (Israelsson 2008). Patients with a permanent stoma, for example after abdomino-perineal resection for low rectal cancer, especially suffer from the morbidity caused by parastomal herniation. Many surgical approaches have been attempted and propagated. However, there is still no clinical evidence from RCTs on which is the best location site for a stoma. We therefore assumed clinical equipoise in regard to the question where a stoma should be located in relation to the rectus abdominis muscle.
Two studies showed a significant reduction in the incidence of parastomal herniation if the stoma was sited through the rectus abdominis muscle versus a location lateral to the muscle (Eldrup 1982; Sjodahl 1988). One prospective uncontrolled cohort study including 72 consecutive, unselected patients described a novel anatomical approach to stoma formation, the lateral rectus abdominis positioned stoma (LRAPS), which involves minimal anterior abdominal wall disruption. The authors presented parastomal herniation rates of 5% (3/62) and 10% (4/41) at a follow-up of one year and two years, respectively (Evans 2011; Stephenson 2010), which is consistent with a relevant reduction in parastomal hernia rate when compared to data from the literature (Robertson 2005).
In view of these controversies, a systematic review of the available non-randomized studies (NRS) is considered justified to efficiently integrate existing information and provide data for rational decision making, as well as to gain insight into treatment effects and potential adverse outcomes.
Another reason for conducting this systematic review was to evaluate the strengths and weaknesses of the available NRS in order to examine the case for undertaking randomized trials. This means that the findings of our review could be applied to shape the design of prospective RCTs (Reeves 2011).
To assess if there is a significant difference regarding the incidence of parastomal herniation and other stomal complications, such as ileus and stenosis, in lateral pararectal versus transrectal stoma placement.
Furthermore, it has to be determined whether a lower parastomal herniation rate goes along with an increased incidence of ileus and stenosis.
Description of studies
See: Characteristics of included studies; Characteristics of ongoing studies.
Results of the search
As recommended by the PRISMA statement (Liberati 2009), a study flow diagram (Figure 4) summarizes the study selection process.
A total of 2831 references were retrieved by the electronic searches (first search run October 2012) of The Cochrane Library (CDSR 29; DARE 7; CENTRAL 15; NHSEED 3), MEDLINE (877), EMBASE (1297), Web of Science (558), and LILACS (45); 1319 duplicates and 1499 clearly irrelevant references, identified by reading titles and abstracts, were excluded. The 13 remaining references were retrieved for further assessment; three (two of them were reports on the same study) of the 13 references from the electronic database search as well as the nine references identified by checking the reference lists of relevant studies and reviews were excluded after reading the full texts (see: Excluded studies) for any of the following reasons (number of studies in brackets):
only transrectal stomata (without a lateral pararectal control group) (n = 2);
only lateral pararectal stomata (without a transrectal control group) (n = 1);
stomal site in relation to the rectus abdominis muscle unclear (n = 8).
The remaining 10 references relating to nine non-randomized studies (NRS) fulfilled our inclusion criteria (see: Included studies).
A total of 46 trials were identified by searching five different trial registers (ClinicalTrials.gov: 22; HKUCTR: 1; UKCTG: 3; UKRNPD: 5; ICTRP: 15). Except for one ongoing single-center RCT (see: Ongoing studies), none of the trials were relevant with regard to our review objective.
Nine retrospective NRS met the inclusion criteria. In order to determine the study type and distinguish between case series and cohort studies, we applied the criteria recently proposed by Dekkers et al (Dekkers 2012). According to their proposal a cohort study is defined by the following main features:
- sampling is based on exposure;
- the occurrence of outcomes is assessed during a specified follow-up period;
- exposure can be a risk factor, a disease, or an intervention;
- absolute risk (or rate) calculation is possible;
- if a comparison group is included, relative risk can be calculated.
All included studies showed these features and were consequently classified as cohort studies.
For more details on the included studies, see Characteristics of included studies.
No trials were eligible for inclusion in this section. See: Characteristics of excluded studies.
Risk of bias in included studies
Reeves et al. hypothesized that the risk of bias in NRS depends on the specific study features rather than on the design labels (such as cohort or case-control study) (Reeves 2011). Thus, we attempted to assess the study features that have a major impact on the risk of bias using the Newcastle-Ottawa Scale (NOS) for cohort studies. See Figure 2; Figure 3; Table 1. Furthermore, we analyzed risk of bias in the usual domains:
In NRS, groups are unlikely to be comparable due to the lack of concealed randomization within the allocation process. Consequently, NRS are more susceptible to selection bias as compared to RCTs, which is regarded as the main difference between RCTs and NRS (Reeves 2011). With regard to selection bias, we checked the representativeness of the intervention cohort (lateral pararectal stoma placement) and the comparability of the intervention (lateral pararectal) and control (transrectal) cohorts. The majority of the included studies were at unclear risk of selection bias since the authors did not sufficiently describe the derivation of the intervention cohort and whether the intervention and control cohorts were drawn from the same community.
We considered the risk of performance bias to be relatively low in the included studies because the surgical procedure (lateral pararectal versus transrectal stoma placement) was ascertained reliably either by reviewing secure records, such as operative reports and patient charts, or by CT or clinical examination, or both. In two radiological studies the position of the stoma in relation to the rectus abdominis muscle was assessed by CT scanning (Cingi 2006; Williams 1990). In another study the intervention was reassessed by CT examination if it remained unclear after clinical evaluation (Ortiz 1994). In one study the stomal site in relation to the rectus muscle was determined by clinical examination, through a surgeon and a staff nurse at the stoma therapy service (Sjodahl 1988). In the remaining studies the intervention must have been ascertained by chart review (though this was not always explicitly stated) (Eldrup 1982; Pilgrim 2010; von Smitten 1986) or review of a standard operative 'pro forma' (Leong 1994; Londono-Schimmer 1994).
Similarly, risk of detection bias was estimated to be low in most of the included studies as the outcome (parastomal herniation) was assessed either by clinical examination or CT scanning, or both, or by reviewing reliable patient records. In three studies, the outcome was assessed by physical examination and CT scanning (Cingi 2006; Ortiz 1994; Williams 1990). In two Scandinavian studies from the 1980s, parastomal herniation was assessed only by clinical examination (Sjodahl 1988; von Smitten 1986). In the most current included study, the outcome was ascertained by physical examination including "digital examination through the stomal opening" "by enterostomal nurse specialists" (Pilgrim 2010). In both studies from St. Mark's Hospital in London, UK, the outcomes (parastomal herniation and stomal prolapse) were assessed by reviewing standard operative pro formas (Leong 1994; Londono-Schimmer 1994). In the oldest of the included studies, a Danish study including patients from two Copenhagen county hospitals, all eligible living patients were screened for parastomal hernia by physical examination, whereas all eligible dead patients were assessed for the outcome by chart review (Eldrup 1982).
Incomplete outcome data
All included studies were at high risk of presenting incomplete outcome data (attrition bias). Applying the NOS, we identified that five of the nine included studies showed follow-up rates < 80% (Cingi 2006; Eldrup 1982; Leong 1994; Londono-Schimmer 1994; Williams 1990). The remaining three studies either did not present any numbers of the patients examined for eligibility or lost to follow-up (Ortiz 1994; Pilgrim 2010; Sjodahl 1988), or the numbers given seemed unreliable from a clinician's standpoint (von Smitten 1986).
Selective outcome reporting bias
The common lack of the use of a study protocol in NRS leads to a higher risk of potential selective outcome reporting bias for NRS as compared to RCTs (Reeves 2011). Features of RCTs to prevent reporting bias (for example a pre-specified protocol, ethical approval including progress and final reports, and the CONSORT statement) were missing in all included studies. Comparing the information given in the methods and results sections regarding outcomes to be assessed and the outcomes assessed, we did not discover any discrepancies. Hence, we concluded that the likelihood of selective outcome reporting remained unclear.
Other potential sources of bias
One study was supported by a grant, but no further information was provided (Ortiz 1994). This could potentially introduce additional bias.
Effects of interventions
See: Summary of findings for the main comparison Lateral pararectal versus transrectal enterostomy placement for parastomal herniation (prevention)
All included studies reported results for the primary outcome (parastomal herniation), and one study also reported data on one of the secondary outcomes (stomal prolapse) (Leong 1994). None of the included studies compared the two interventions with regard to other secondary outcomes.
There was neither a significant difference in terms of the risk for parastomal herniation (RR 1.29; 95% CI 0.79 to 2.1) (Analysis 1.1) nor with regard to the occurrence of stomal prolapse (RR 1.23; 95% CI 0.39 to 3.85) (Analysis 1.2). An I² value of 65% indicated substantial statistical heterogeneity in the meta-analysis (Analysis 1.1). The I² statistic describes the percentage of total variation across studies that is caused by heterogeneity rather than by chance (Higgins 2011). Both P values (0.31 and 0.73, respectively) were interpreted as a finding of uncertainty. The confidence intervals were very wide in both analyses. In Analysis 1.1, this could be the result of the small number of studies combined in the meta-analysis as well as the substantial heterogeneity. The wide confidence interval in Analysis 1.2 was likely to be caused by the small sample size of the only included study.
In summary, there was inconclusive evidence of a preventive effect of the compared interventions with regard to parastomal herniation.
Parastomal herniation is one of the most common stoma complications and causes significant stoma-related morbidity, especially in patients with permanent enterostomies, for example patients with low rectal cancer who had to undergo abdomino-perineal resection. The patient relevance of this complication is high because parastomal herniation often results in further morbidity, such as skin irritation and maceration due to leakage. Moreover, it can even necessitate emergency surgery for strangulated bowel. A recent retrospective study from France, which included 782 stoma patients, showed that after a mean follow-up of 10.5 years the prevalence of parastomal herniation was 25.6% (n = 202). Of the 202 patients with parastomal hernia, more than one third (35%) complained of pain and almost one third (28%) suffered from leakage due to difficulties in fitting the stomal appliance. More than half of the patients (n = 114) underwent operative repair, but again in half of these patients (n = 57) parastomal herniation recurred within six months (Ripoche 2011). The high recurrence and morbidity rates after hernia repair make the situation of the affected patients even more precarious and underline the need for effective preventive measures.
With regard to the etiology of parastomal herniation, one has to differentiate between patient-related and surgical or technical risk factors for parastomal herniation. Ripoche et al. studied several patient-related potential risk factors such as age, gender, past medical history (ventral hernia, diabetes, corticosteroid therapy), and diagnosed condition requiring stoma formation (for example malignancy, inflammatory bowel disease, diverticulosis). The only patient factor which was significantly associated with parastomal herniation in a multivariate analysis was age > 60 years at the time of stoma placement (Ripoche 2011), which is supported by data from other studies investigating long term hernia rate and risk factors (Hong 2013; Mylonakis 2001; Pilgrim 2010). The influence of operative factors is also much debated. Trephine size, closure of the lateral space, stomal fixation to the fascia, intra- versus extraperitoneal route, and location in relation to the rectus abdominis muscle may have an influence on the occurrence of parastomal herniation (Carne 2003; Hotouras 2013). Furthermore, type of enterostomy (ileostomy versus colostomy, end versus loop ostomy), operative setting (emergency versus elective surgery), and whether a stoma therapist marked the site of the stoma preoperatively seem to have an impact on the risk of parastomal herniation.
Regardless of the fact that there was no evidence from randomized trials, the surgical dogma has emerged that a stoma should be placed through the rectus abdominis rather than lateral to the muscle in order to prevent parastomal herniation. This credo has been persistently repeated in the surgical literature, usually without any scientific evidence (Park 1999; Pearl 1985). But how could this surgical dictum arise? And why does it still persist despite the lack of evidence? In the mid 1930s Gabriel and Lloyd-Davies recommended transrectal colostomy formation for prevention of parastomal herniation (Gabriel 1935), which was approved by Turnbull 20 years later (Turnbull 1958). In the mid 1970s a study from St. Mark's Hospital in London, UK, was published reporting data on 227 patients undergoing surgery for rectal adenocarcinoma. In 35 of the included patients the colostomy was created transrectally and, after a follow-up of one to six years, six of the 35 patients were diagnosed with a parastomal hernia (Marks 1975). Thus, on the basis of their data, the authors concluded that the transrectal route cannot generally be recommended. One explanation why surgeons continued to adhere to the thesis that the stomata should be placed transrectally could be the assumption that pulling the stoma through the rectus abdominis muscle provides a more snug fit since the muscle contracts around the stomal aperture. On the other hand, the same thinking could lead to the concern that there is a higher risk of stoma-related stenosis and ileus due to tight muscle contractions around the enterostomy.
Recently, Stephenson described a novel approach to stoma formation, the lateral rectus abdominis positioned stoma (LRAPS) (Stephenson 2010), being convinced that preservation of the rectus abdominis muscle and its sheath could prevent parastomal herniation. The new technique implies that the anterior and posterior rectus sheath are only divided horizontally (not vertically) and that the rectus muscle is not incised but separated from its sheath by sharp and blunt dissection and is then retracted medially. After the bowel is delivered, any defect of the medial margins of the incisions in the anterior and posterior sheath is closed with interrupted absorbable sutures. Stephenson presented data from 29 consecutive, unselected patients who received a LRAPS. After a mean follow-up of 13 (range 7 to 18) months no parastomal hernia had been detected. In 2011 Evans published a further follow-up of the LRAPS (Evans 2011); at one year of follow-up the authors observed parastomal hernias in 5% of their patients with a LRAPS (3/62), and, at two years, the parastomal hernia rate had increased to 10% (4/41). Although the follow-up duration was still rather short, the reported parastomal herniation rates are lower than what is usually reported in the literature (Israelsson 2008; Robertson 2005).
Another emerging surgical trend is the use of prophylactic mesh for prevention of parastomal herniation. Recently, two systematic reviews of three randomized trials that included 129 patients were conducted (Shabbir 2012; Wijeyekoon 2010). Meta-analysis showed that mesh reinforcement of stomas was associated with a decreased incidence of parastomal herniation (RR 0.23; 95% CI 0.06 to 0.81; P = 0.02) as well as with a reduced proportion of patients with parastomal hernias requiring surgical repair (Wijeyekoon 2010). Though the results of these three small RCTs (Hammond 2008; Janes 2004; Janes 2009; Serra-Aracil 2009) showed a beneficial effect of mesh reinforcement with regard to prevention of parastomal herniation, with no excess morbidity, scepticism regarding the safety of placing prosthetic mesh adjacent to open bowel still seems justified. Furthermore, the overall follow-up period was relatively short (range 12 to 57 months). Hence, it remains unclear if the preventive treatment effect of prophylactic parastomal mesh reinforcement persists and is safe enough in the long term. In addition, the review authors concluded that there is a lack of evidence with regard to the preferable type of mesh and the optimal site for mesh placement; and that the role of prophylactic mesh in the emergency surgery setting is still obscure. In summary, further evidence from (multicenter) randomized trials is needed to clarify the role of prophylactic mesh placement for prevention of parastomal hernia.
On the premise that surgical prevention of parastomal herniation without the use of mesh is worthy of consideration, we aimed to collect all existing evidence regarding the question whether a stoma should be placed through or lateral to the rectus abdominis muscle. The main results of our reviewing process are summarized in the Summary of main results.
In conclusion, there is only very limited evidence from a few non-randomized cohort studies with high risk of bias regarding our review question. This lack of evidence gives rise to collective uncertainty within the surgical community as to which treatment is preferable, lateral pararectal or transrectal stoma formation. Thus, we assume 'clinical equipoise', which should be the ethical basis for the initiation of randomized trials comparing lateral pararectal to transrectal stoma placement. In the light of the complete absence of evidence from randomized trials the authors of this systematic review (JH, FH, PK, SP) designed PATRASTOM, a single-center RCT investigating the review objective (Hardt 2012), which started recruiting in April 2012.
Summary of main results
Currently, there are no RCTs assessing our review objective. However, there is one ongoing single-center pilot RCT comparing lateral pararectal to transrectal stoma placement that is scheduled for completion in 2014 (Hardt 2012). Due to the lack of RCT evidence, we reviewed the existing observational studies addressing our review objective. All included studies reported our primary outcome (parastomal hernia), and one study (Leong 1994), that included 145 patients, additionally reported one of our secondary outcomes (stomal prolapse). Seven of the nine included cohort studies, with a total of 491 participants, did not find a statistically significant difference in parastomal herniation rates after lateral pararectal compared to transrectal stoma formation. In contrast, the remaining two studies (Eldrup 1982; Sjodahl 1988) found a significant reduction of parastomal herniation if the stoma was pulled out through the belly of the rectus abdominis muscle. Eldrup presented data on the prevalence of paracolostomy hernia in 140 patients who had undergone permanent end-sigmoidostomy placement, either transrectally (n = 77) or lateral to the rectus abdominis in the left iliac fossa (n = 63) (Eldrup 1982). The second study, a Swedish study including 130 patients attending the stoma therapy service at a university hospital in Linköping, showed an almost 10-fold reduced risk of parastomal herniation after transrectal stoma placement compared to lateral pararectal stoma siting (Sjodahl 1988). Pooling the data of all included studies could not provide conclusive evidence (see Effects of interventions). Reasons for the widely differing estimates of the intervention effect (that is heterogeneity) across studies could be the lack of standardization of the surgical procedure as well as the absence of a uniform definition and detection method for parastomal hernia.
Also see Summary of findings for the main comparison.
Overall completeness and applicability of evidence
Our review aimed to compare lateral pararectal to transrectal stoma placement for prevention of parastomal herniation and other complications that are important to the patient. This was only partly accomplished. All included studies reported parastomal herniation, our primary outcome, but only one of them presented data on stomal prolapse. None of them reported any other of our secondary outcomes. Furthermore, we could only conduct one of the planned subgroup analyses and none of the sensitivity analyses because the required data were not reported in the included studies. Thus, the evidence with regard to our review objective is very limited and the identified studies could not address all of the objectives of the review. Consequently, no recommendations can be formulated on the basis of such incomplete and inconclusive evidence. The studies identified are all retrospective cohort studies and at high risk of bias, especially with regard to selection bias and incomplete outcome data (attrition bias). Since the participants were not randomly assigned to the two intervention groups, the risk of selection bias is high. Moreover, it mostly remained unclear whether the individual participants were representative of the population. Hence, external validity and generalizability of the results remain unclear, which greatly restricts the applicability of the evidence identified.
Quality of the evidence
Nine retrospective cohort studies with a total of 761 participants were identified and included in this systematic review. For key methodological features and limitations of the studies please see the Characteristics of included studies. Confidence in the estimates of the treatment effect is limited in respect to risk of bias, publication bias, imprecision, inconsistency, and indirectness. All these issues are included in the GRADE system and are summarized in the Summary of findings for the main comparison. Overall risk of bias in the included observational studies is high, especially due to possible selection bias (no randomization, unclear representativeness and comparability of the two cohorts) and high risk of attrition bias. It remains unclear whether publication bias affected the estimates of treatment effect, but it cannot be excluded because observational studies reporting higher rates of parastomal herniation and other stoma-related complications might be less likely to be published than studies presenting lower complication rates. This may be a particular concern in surgery. Since the 95% confidence interval around the estimate of effect includes both no effect and appreciable benefit or appreciable harm, and because the total number of events is less than 300, we assume that there is considerable imprecision. Moreover, the quality of evidence is diminished due to inconsistency because of the widely differing estimates of the intervention effect (that is heterogeneity) across studies, for which we failed to identify a plausible explanation. However, we did not find any evidence of indirectness with respect to the target population, intervention, and outcome of interest.
Hence, the quality of evidence is downgraded by one level, from low to very low, due to high risk of bias, imprecision, and inconsistency.
In summary, the body of evidence identified does not allow a robust conclusion regarding the objectives of the review.
Potential biases in the review process
Although we followed a strict protocol consistent with the Methodological standards for the conduct of new Cochrane Intervention Reviews, Version 2.1, 8 December 2011 (MECIR) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) statements (Stroup 2000), and applied a comprehensive search strategy, this review is still prone to bias. For example, publication bias cannot be excluded in our review because NRS are not consistently indexed in trial registers (Reeves 2011). Consequently, it must be taken into consideration that we may have missed completed or ongoing NRS. Moreover, we did not screen conference proceedings, which could potentially lead to further publication bias.
Agreements and disagreements with other studies or reviews
To our knowledge, this is the first systematic review addressing this particular review question. However, several narrative reviews have discussed the objective of our review (Carne 2003; Hotouras 2013; Israelsson 2008; Martin 1996; Pearl 1989).
Contributions of authors
Design, data extraction, analysis, interpretation, and drafting of review: JH, FH.
Design, resolution of discrepancies, interpretation, drafting, and supervision of review: JM, PK, SP.
Design and execution of search strategies, search documentation, and drafting of review: MM.
Differences between protocol and review
We undertook a broader search approach than stated in the protocol, searching for all types of non-randomized studies, because study design labels are not used consistently by authors and are thus not indexed reliably by bibliographic databases. This applies especially for non-randomized studies. We proved this by adding a specificity search filter (Fraser 2006) to our subject search in MEDLINE, which resulted in exclusion of relevant references mentioned in the protocol. Therefore, we decided to search without any restriction by publication or study type.
In order to further enhance the overall recall of the search and to minimize language bias, we searched two more databases than stated in the review protocol: Science Citation Index Expanded and LILACS.
The German Clinical Trials Register, that is mentioned in the protocol, is included in the International Clinical Trials Research Platform (ICTRP) and was thus not searched separately. Three trial registers (HKUCTR, UKCTG, and UKRNPD),which were not mentioned in our protocol, were searched since they are not included in the ICTRP.
We also concretized the inclusion criteria for studies. Studies had to compare lateral pararectal versus transrectal enterostomy placement with regard to the incidence of parastomal herniation.
Only one of the planned subgroup analyses could be conducted (see Subgroup analysis and investigation of heterogeneity and Analysis 2.1). None of the planned sensitivity analyses were conducted (see Sensitivity analysis).