Antecolic versus retrocolic reconstruction after partial pancreaticoduodenectomy

Abstract

Background

Pancreatic cancer remains one of the five leading causes of cancer deaths in industrialized nations. For adenocarcinomas in the head of the gland and premalignant lesions, partial pancreaticoduodenectomy represents the standard treatment for resectable tumours. The gastro‐ or duodenojejunostomy after partial pancreaticoduodenectomy can be reestablished via either an antecolic or a retrocolic route. The debate about the more favourable technique for bowel reconstruction is ongoing.

Objectives

To compare the effectiveness and safety of antecolic and retrocolic gastro‐ or duodenojejunostomy after partial pancreaticoduodenectomy.

Search methods

We conducted a systematic literature search on 29 September 2015 to identify all randomised controlled trials in the Cochrane Central Register of Controlled Trials (CENTRAL), The Cochrane Library 2015, issue 9, MEDLINE (1946 to September 2015), and EMBASE (1974 to September 2015). We applied no language restrictions. We handsearched reference lists of identified trials to identify further relevant trials, and searched the trial registry clinicaltrials.gov for ongoing trials.

Selection criteria

We considered all randomised controlled trials that compared antecolic versus retrocolic reconstruction of bowel continuity after partial pancreaticoduodenectomy for any given indication to be eligible.

Data collection and analysis

Two review authors independently screened the identified references and extracted data from the included trials. The same two review authors independently assessed risk of bias of included trials, according to standard Cochrane methodology. We used a random‐effects model to pool the results of the individual trials in a meta‐analysis. We used odds ratios to compare binary outcomes and mean differences for continuous outcomes.

Main results

Of a total of 216 citations identified by the systematic literature search, we included six randomised controlled trials (reported in nine publications), with a total of 576 participants. We identified a moderate heterogeneity of methodological quality and risk of bias of the included trials. None of the pooled results for our main outcomes of interest showed significant differences: delayed gastric emptying (OR 0.60; 95% CI 0.31 to 1.18; P = 0.14), mortality (RD ‐0.01; 95% CI ‐0.03 to 0.02; P = 0.72), postoperative pancreatic fistula (OR 0.98; 95% CI 0.65 to 1.47; P = 0.92), postoperative haemorrhage (OR 0.79; 95% CI 0.38 to 1.65; P = 0.53), intra‐abdominal abscess (OR 0.93; 95% CI 0.52 to 1.67; P = 0.82), bile leakage (OR 0.89; 95% CI 0.36 to 2.15; P = 0.79), reoperation rate (OR 0.59; 95% CI 0.27 to 1.31; P = 0.20), and length of hospital stay (MD ‐0.67; 95%CI ‐2.85 to 1.51; P = 0.55). Furthermore, the perioperative outcomes duration of operation, intraoperative blood loss and time to NGT removal showed no relevant differences. Only one trial reported quality of life, on a subgroup of participants, also without a significant difference between the two groups at any time point. The overall quality of the evidence was only low to moderate, due to heterogeneity, some inconsistency and risk of bias in the included trials.

Authors' conclusions

There was low to moderate quality evidence suggesting no significant differences in morbidity, mortality, length of hospital stay, or quality of life between antecolic and retrocolic reconstruction routes for gastro‐ or duodenojejunostomy. Due to heterogeneity in definitions of the endpoints between trials, and differences in postoperative management, future research should be based on clearly defined endpoints and standardised perioperative management, to potentially elucidate differences between these two procedures. Novel strategies should be evaluated for prophylaxis and treatment of common complications, such as delayed gastric emptying.

Plain language summary

Comparison of bowel reconstruction routes after partial surgical removal of the pancreas and duodenum (first part of the small intestine)

Background

The pancreas is a digestive gland situated in the upper abdomen, which is also vital to normal control of blood sugar. Pancreatic cancer is one of the leading causes of cancer death in industrialized nations. The standard surgical treatment for cancer of the head of the gland and precancerous abnormalities is partial removal of the pancreas, together with the attached duodenum, known as a pancreaticoduodenectomy. Removal of the duodenum requires the restoration of the digestive pathway from the stomach to the rest of the gut. This can be accomplished by joining it to the jejunum (second part of the small intestine) either in front of (antecolic) or behind (retrocolic) the overlying large intestine (transverse colon).

Review question

It is unclear whether one of these two routes of reconstruction provides a benefit to the patient by reducing delayed gastric emptying (emptying of the stomach after ingestion of food); postoperative mortality (death); and other complications, such as pancreatic fistula (leakage of pancreatic juice), reoperation, perioperative measures (before, during, and after the operation), or length of hospital stay; and improving quality of life. Delayed gastric emptying was the primary outcome of this review because it is one of the most frequent complications after a pancreaticoduodenectomy; it can make it difficult to take anything by mouth and interferes with the patient’s quality of life, often resulting in a prolonged hospital stay and delay of further treatment.

Study characteristics

We included six randomised controlled trials (reported in nine publications), reporting data on a total of 576 adult participants, who underwent pancreaticoduodenectomy for any pancreatic disease. The evidence is current to September 2015.

Key results

We did not identify significant differences in delayed gastric emptying; postoperative mortality; postoperative pancreatic fistula, or other complications; reoperations; or length of hospital stay. Quality of life, only reported for a subset of participants in one trial, did not differ between the two groups. Our results do not suggest any relevant differences between antecolic and retrocolic reconstruction of the gastro‐ or duodenojejunostomy after partial pancreaticoduodenectomy.

Quality of evidence

The quality of the evidence was only low to moderate, due to clinical and statistical differences between individual trials, and risk of bias, due to shortcomings in the way the trials were conducted. Therefore, the results should be viewed with caution.

Summary of findings

for the main comparison.

Delayed gastric emptying following antecolic versus retrocolic reconstruction after partial pancreatoduodenectomy
Patient or population: Adults with any pancreatic diagnosis leading to an indication for elective (classical, pylorus‐preserving, pylorus‐resecting) partial pancreatoduodenectomy Settings: Inpatient treatment in Europe and Asia Intervention: Bowel continuity via an antecolic reconstruction Comparison: Bowel continuity via a retrocolic reconstruction
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Retrocolic reconstruction	Antecolic reconstruction
Delayed gastric emptying (all definitions)	43 per 100	35 per 100 (5 to 61)	OR 0.60 (0.31 to 1.18)	576 participants (6 trials)	⊕⊕⊝⊝ low1,2
Mortality	4 per 100	3 per 100	RD ‐0.01 (‐0.03 to 0.02)	576 participants (6 trials)	⊕⊕⊕⊝ moderate3
Pancreatic fistula	22 per 100	21 per 100 (5 to 38)	OR 0.98 (0.65 to 1.47)	576 participants (6 trials)	⊕⊕⊝⊝ low1,2
Postoperative haemorrhage	7 per 100	6 per 100 (5 to 6)	OR 0.79 (0.38 to 1.65)	530 participants (5 trials)	⊕⊕⊝⊝ low1,2
Intra‐abdominal fluid collection, abscess	12 per 100	12 per 100 (4 to 28)	OR 0.93 (0.52 to 1.67)	576 participants (6 trials)	⊕⊕⊝⊝ low1,2
Reoperations	8 per 100	4 per 100 (0 to 7)	OR 0.59 (0.27 to 1.31)	470 participants (4 trials)	⊕⊕⊝⊝ low2,4
Length of postoperative hospital stay	The mean length of postoperative hospital stay in the control groups ranged from 12 to 48 days	The mean length of postoperative hospital stay in the intervention groups ranged from 12 to 36 days	MD ‐0.67 (‐2.85 to 1.51)	576 participants (6 trials)	⊕⊕⊝⊝ low1,2,5
Quality of life	No significant differences in the EQ‐5D(TM) questionnaire, the 'European Organization for Research and Treatment of Cancer" QLQ‐C30 and PAN26 questionnaires and the 'Gastrointestinal Quality of Life Index' at 2, 4 and 12 weeks after the operation.	No significant differences in the EQ‐5D(TM) questionnaire, the 'European Organization for Research and Treatment of Cancer" QLQ‐C30 and PAN26 questionnaires and the 'Gastrointestinal Quality of Life Index' at 2, 4 and 12 weeks after the operation.	no quantitative synthesis	73 participants (1 trial)	⊕⊕⊝⊝ low6
*The basis for the assumed risk (e.g. the median control group risk across trials) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; MD: Mean Difference; OR: Odds Ratio; RD: Risk Difference [other abbreviations, eg. OR, etc]
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹potential risk of bias due to differing endpoint definitions, perioperative management, and methodological issues

²substantial heterogeneity and wide confidence intervals

³potential risk of bias due to methodological issues in some of the trials

⁴differing therapeutic measures in individual trials

⁵differing discharge policies between individual trials

⁶only evaluated in a subset of patients of one trial

Background

Pancreatic cancer was predicted to become the fourth leading cause of cancer deaths in 2014 in Europe (Malvezzi 2014), as well as in the United States (Muniraj 2013). The global age‐standardised incidence rate of pancreatic cancer was estimated to be around 5 per 100,000 men and 3.2 per 100,000 women, with great differences between highly developed countries and less developed countries (Globocan 2012). While mortality rates for other types of cancer are predicted to fall, the mortality rate for pancreatic ductal adenocarcinoma has been rising in recent years (Malvezzi 2014). The majority of pancreatic ductal adenocarcinomas are located in the head of the gland, and if the lesion is resectable, radical surgery is still the treatment of choice (Shaib 2007). Partial pancreaticoduodenectomy (pPD) is the standard treatment for tumours of the pancreatic head, benign precursor lesions such as intraductal mucinous neoplasia, or chronic pancreatitis that require a resective surgical approach (Schnelldorfer 2008). Due to improvements in imaging techniques and their increased use in health checks, an increasing number of asymptomatic lesions are discovered incidentally, leading to higher rates of pancreatic surgery (Plichta 2015; Spinelli 2004). Due to surgical expertise and standardised perioperative management, pancreatic surgery, which is technically challenging, can be performed in specialised centres with mortality rates of less than 5% (Büchler 2003; de Wilde 2012; McPhee 2007). However, morbidity, consisting typically of postoperative pancreatic fistula, post‐pancreatectomy haemorrhage, biliary leakage, intra‐abdominal abscess, and delayed gastric emptying, remains at a high level, up to 60% (Stojadinovic 2003).

A glossary of terms is provided in Appendix 1.

Description of the condition

One of the most frequent complications after pPD, with rates between 14% to 61% reported in the literature, is delayed gastric emptying (Kim 2005). Delayed gastric emptying impairs oral intake, interferes with the patient’s quality of life, and often results in prolonged hospitalisation and delay of further treatment (e.g. necessary adjuvant chemotherapy). Several modifications of surgical techniques have been used in an attempt to reduce the frequency of delayed gastric emptying, for instance, pylorus‐resecting versus pylorus‐preserving pPD, since the pylorus might have an influence on delayed gastric emptying, acting as a gatekeeper of the gastric outlet (Hackert 2013; Yang 2014). Billroth‐II versus Roux‐en‐Y reconstruction have also been compared, since the gastroenteric passage could be influenced by the passing of biliopancreatic juice (Shimoda 2013). Different perioperative treatment strategies have also been used to reduce the impact of delayed gastric emptying, e.g. erythromycin is supposed to improve delayed gastric emptying by acting as a motilin‐agonist (Yeo 1993), while somatostatin analogues decrease gastric motility in healthy patients (Kollmar 2008). The large variations in rates of delayed gastric emptying in different studies are attributable to the use of different definitions and assessment of delayed gastric emptying. In 2007, the International Study Group of Pancreatic Surgery (ISGPS) developed a consensus definition to provide an objective assessment of this condition, and comparability of future trial results (Wente 2007a). There is a reasonable assumption that the method of bowel reconstruction after pPD (antecolic versus retrocolic) has an impact on delayed gastric emptying.

Description of the intervention

The etiology of delayed gastric emptying has been attributed to multiple factors, including other intra‐abdominal complications, e.g. intra‐abdominal abscess or postoperative pancreatic fistula. Nevertheless, not all cases of delayed gastric emptying are related to other complications, and the route of reconstruction might be another influencing factor. Bowel continuity after pPD can either be established via an antecolic route (i.e. a jejunal loop is brought up anterior to the transverse colon for the gastroenteric anastomosis), or via a retrocolic route through a mesocolic window (i.e. a jejunal loop is brought up posterior to the transverse colon for the gastroenteric anastomosis). As suggested by several randomised and non‐randomised studies, the route of reconstruction and the angulation of the stomach and duodenum might influence the occurrence of delayed gastric emptying (Hartel 2005; Nikfarjam 2009; Tani 2006).

How the intervention might work

Several explanations have been provided for the potential difference between antecolic and retrocolic reconstruction: the mobility of the descending jejunal loop might be different, which supposedly decreases the risk of torsion or angulation. The different anatomic position between the gastroenteric anastomosis and the pancreaticojejunostomy might also have an effect on the gastrointestinal passage, e.g. caused by a small pancreaticojejunal anastomotic leak or a transient mild postoperative pancreatitis (Hartel 2005). Furthermore, the angulation of the stomach can be different between the two techniques, thus influencing the gastrointestinal passage.

Why it is important to do this review

Since delayed gastric emptying is still a frequent and clinically relevant problem after pPD, and the optimal reconstruction method to reduce its occurrence is still under debate; a recent systematic review including randomised and non‐randomised studies addressed this issue (Bell 2015). However, the analysis of the retrospective studies might introduce substantial bias in the light of a sufficient number of randomised controlled trials. Furthermore, one of the included trials (Chijiiwa 2009), represents a preliminary report on the same set of patients of another included trial (Imamura 2014), which is not addressed in the review. Therefore, a systematic review with clear methodology, that compares these two reconstruction techniques appears both feasible and important. The combined analysis of evidence from randomised controlled trials might be able to provide a conclusive answer to the ongoing debate.

Objectives

To compare the effectiveness and safety of antecolic and retrocolic gastro‐/duodenojejunostomy after partial pancreaticoduodenectomy (pPD).

Methods

Criteria for considering studies for this review

Types of studies

We only included randomised controlled trials (RCTs) in this review. We did not apply any restrictions to the language of the original report.

Types of participants

We included all RCTs considering adults with any pancreatic diagnosis (e.g. pancreatic carcinoma, chronic pancreatitis, cystic neoplasms of the pancreas, etc) leading to an indication for elective pPD.

Types of interventions

We only included trials that compared antecolic reconstruction with retrocolic reconstruction of bowel continuity after pPD. In cases of insufficient description of the surgical procedures involved, we contacted the trial authors for more details, to enable us to decide about including the trial.

Types of outcome measures

Primary outcomes

Delayed gastric emptying (preferably defined according to the International Study Group of Pancreatic Surgery (ISGPS) definition (Wente 2007a)).

Secondary outcomes

1. Clinically relevant delayed gastric emptying (ISGPS grade B/C)

2. Postoperative mortality (30‐day and in‐hospital mortality);

3. Postoperative pancreatic fistula (preferably defined according to the International Study Group of Pancreatic Fistula (ISGPF) definition (Bassi 2005));

4. Postoperative haemorrhage (preferably defined according to the ISGPS definition (Wente 2007b));

5. Bile leakage;

6. Intra‐abdominal fluid collection or abscess;

7. Reoperation rate;

8. Duration of operation;

9. Intraoperative blood loss;

10. Length of hospital stay;

11. Time to nasogastric tube (NGT) removal (in days) after surgery

12. Quality of life (considering all aspects e.g. physical and emotional quality of life, ability to eat, pain level, etc.).

Most of the outcomes we chose represent the major complications specific to pancreatic surgery. Duration of operation and length of hospital stay have been chosen as indirect measures of hospital costs. Quality of life represents the most patient‐relevant outcome.

Reporting of all outcomes listed here was not an inclusion criteria for the review. If a trial did not provide data on the primary outcome, we contacted the trial authors for clarification.

Search methods for identification of studies

Electronic searches

We conducted a systematic literature search to identify all published and unpublished RCTs that examined antecolic or retrocolic reconstruction after pPD. We designed the literature search to identify potential trials in all languages. We did not find any non‐English language papers that were relevant for the review.

We searched the following electronic databases for identification of potential trials:

• Cochrane Central Register of Controlled Trials (CENTRAL; The Cochrane Library 2015, issue 9; Appendix 2).

• MEDLINE (1966 to 29 September 2015; Appendix 3).

• EMBASE (1988 to 29 September 2015; Appendix 4).

We also searched the clinical trials register, ClinicalTrials.gov on 29 September 2015, to identify potential trials in the field that were unpublished or ongoing.

Searching other resources

We checked the reference lists of all primary studies and relevant review articles for additional references. We contacted authors of identified trials and asked them to identify other published and unpublished trials. Furthermore, we asked experts in the field for further relevant trials.

We searched PubMed on 29 September 2015 for errata or retractions from eligible trials.

Data collection and analysis

Selection of studies

We identified and excluded duplicates and multiple reports of the same trial, so that each trial, rather than each report, was the unit of interest in the review. Two review authors (FJH, RK) independently screened titles and abstracts of all potential studies, which were identified as a result of the search, and coded them as 'retrieve' (eligible or potentially eligible or unclear), or 'do not retrieve'. We retrieved the full text of the trial reports and publications deemed eligible, potentially eligible, or unclear, and the same two review authors independently screened the full text, identified trials for inclusion, and identified and recorded reasons for exclusion of the ineligible studies. They resolved any disagreements through discussion, or if required, in consultation with a third review author (MKD). We recorded the selection process in sufficient detail to complete a PRISMA flow diagram (Moher 2009), and 'Characteristics of excluded studies' table.

Data extraction and management

We used a standard data collection form, which had been piloted on at least one trial in the review, to gather trial characteristics and outcome data. Two review authors (FJH, RK) extracted these trial characteristics from included trials.

1. Methods: trial design, total duration of the trial, number of trial centres and location, trial setting, sample size calculation, withdrawals, publication date.

2. Participants: sample size, mean age, gender, diagnosis (underlying disease), inclusion criteria, exclusion criteria.

3. Interventions: intervention, comparison, concomitant medications, excluded medications, technical details of intervention.

4. Outcomes: primary and secondary outcomes specified and collected, time points reported, exact definitions of outcomes reported.

5. Notes: funding for trial, notable conflicts of interest of trial authors.

Two review authors (FJH, RK) independently extracted outcome data from included trials. If outcome data were reported in an unusable way, we noted this in the 'Characteristics of included studies' table. We resolved all disagreements by consensus, or by involving a third person (MKD). One review author (FJH) copied the data from the data collection form into the Review Manager file (RevMan 2014). We double‐checked that the data were entered correctly by comparing the trial reports with data in the systematic review. A second review author spot‐checked trial characteristics for accuracy against the trial report.

Assessment of risk of bias in included studies

Two review authors (FJH, RK) independently assessed risk of bias for each trial, using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). They resolved any disagreement by discussion or by involving a third review author (MKD). We assessed the risk of bias according to the following domains.

1. Random sequence generation.

2. Allocation concealment.

3. Blinding of participants and personnel.

4. Blinding of outcome assessment.

5. Incomplete outcome data.

6. Selective outcome reporting.

7. Other bias (e.g. baseline imbalance, early termination of the trial, funding bias, etc).

We graded each potential source of bias as high, low, or unclear, and we presented a quote from the trial report together with a justification for the judgement in the 'Risk of bias' table. We summarised the 'Risk of bias' judgements across trials for each of the domains listed. Where information on risk of bias related to unpublished data or correspondence with a trialist, we noted this in the 'Risk of bias' table.

When considering treatment effects, we took the risk of bias for the trials that contributed to that outcome into account.

Assesment of bias in conducting the systematic review

We conducted the review according to the published protocol, and reported any deviations in the 'Differences between protocol and review' section of the systematic review.

Measures of treatment effect

We analysed dichotomous data as odds ratios (ORs) and continuous data as mean differences (MDs) or standardised mean differences (SMDs). We ensured that higher scores for continuous outcomes had the same meaning for the particular outcome, we explained the direction to the reader, and we reported where the directions were reversed, if this was necessary. All analyses were conducted with RevMan 2014.

We only performed meta‐analyses where this was meaningful, i.e. if the treatments, participants, and the underlying clinical question were similar enough for pooling to make sense.

A common way that trialists indicate when they have skewed data is by reporting medians and interquartile ranges. When this was encountered, we used the methods described by Higgins 2011, Hozo 2005, or both, to calculate means and standard deviations. The decision to conduct quantitative synthesis with these data was based upon individual decisions for each outcome and are explained in the review.

Unit of analysis issues

When data were presented in different forms across included trials, or they were difficult to categorise, we dichotomised these data, if possible, for the purpose of analysis. We did not find any RCTs with non‐standard designs.

Dealing with missing data

We contacted investigators or trial sponsors in order to verify key trial characteristics and obtain missing numerical outcome data, where possible (e.g. when a trial was identified as abstract only).

Assessment of heterogeneity

We used I² statistics to measure statistical heterogeneity among the trials in each analysis.

We explored clinical heterogeneity by assessing differences in baseline data, definitions of outcome parameters, and operative or perioperative management, or both. We discussed clinical heterogeneity in the appropriate sections, and we considered the presence of strong clinical heterogeneity in the decision to conduct quantitative synthesis of data.

Assessment of reporting biases

We contacted trial authors to ask them to provide missing outcome data, if this was necessary.

We created a funnel plot and examined it for asymmetry to explore possible publication bias.

Data synthesis

Whenever sufficient data for a specific outcome were provided, and it made clinical and statistical sense to pool the results, we performed a meta‐analysis using a random‐effects model (DerSimonian 1986).

'Summary of findings' table

We created a 'Summary of findings' table using the following outcomes: delayed gastric emptying, postoperative mortality, postoperative pancreatic fistula, postoperative haemorrhage, intra‐abdominal fluid collection/abscess, reoperation rate, length of hospital stay and quality of life. These outcome parameters have been chosen because they represent the most important outcomes from a clinical point of view and could potentially be influenced by the different interventions. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to assess the quality of the body of evidence as it related to the trials that contributed data to the meta‐analyses for the prespecified outcomes. We used the methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and GRADEproGDT software (GRADEproGDT 2015). We justified all decisions to down‐ or up‐grade the quality of trials in footnotes, and provided comments to aid the reader's understanding of the review, where necessary. If there was any additional outcome information that we were unable to incorporate into the meta‐analyses, we considered and noted this in the comments. We stated if it supported or contradicted the information from the meta‐analyses.

Subgroup analysis and investigation of heterogeneity

We had planned to perform the following subgroup analyses, if sufficient data were available.

1. Antecolic versus retrocolic reconstruction after pylorus‐preserving pPD.

2. Antecolic versus retrocolic reconstruction after classical pPD (Whipple's procedure).

3. Antecolic versus retrocolic reconstruction for different underlying diagnoses (e.g. pancreatic adenocarcinoma, chronic pancreatitis, etc).

4. Antecolic versus retrocolic reconstruction with single‐loop reconstruction.

5. Antecolic versus retrocolic reconstruction with Roux‐en‐Y reconstruction.

We used the following outcome in subgroup analysis.

1. Delayed gastric emptying

Sensitivity analysis

We conducted the following sensitivity analysis, defined a priori, to assess the robustness of our conclusions:

1. only trials applying the ISGPS definition;

2. excluding studies that presented medians and ranges instead of means and standard deviations;

3. excluding trials with less than the average number of positive judgements in the risk of bias assessment; and

4. fixed‐effect meta‐analysis for the primary outcome: delayed gastric emptying.

Reaching conclusions

The conclusions from this systematic review are based only on findings from the quantitative or narrative synthesis of included trials for this review. We avoided making recommendations for practice and the 'Implications for research' section gives the reader a clear sense of where the focus of any future research in the area should be, and what the remaining uncertainties are.

Results

Results

Description of studies

See:Characteristics of included studies; Characteristics of excluded studies.

Results of the search

In total, we identified 216 titles and abstracts with the systematic literature search. After accounting for the exclusion process described below, the quantitative data of six trials reported in nine publications, with 576 analysed patients comparing antecolic (289 patients) and retrocolic (287 patients) reconstruction were found suitable for this review (Eshuis 2014; Gangavatiker 2011; Imamura 2014; Kurahara 2011; Tamandl 2014; Tani 2006). Another ongoing trial was identified through the search of the trial registry ClinicalTrials.gov, but no results were yet available, since the planned completion date is March 2017 (NCT01460550).

Included studies

Trials showed obvious discrepancy in sample sizes (range from 40 (Tani 2006), to 246 (Eshuis 2014)), but intra‐ and inter‐study population baseline characteristics revealed comparability.

We observed different technical approaches for the surgical intervention: one trial included only pylorus‐resecting Whipple's procedure (Kurahara 2011), three trials included only pylorus‐preserving Whipple's procedure (Imamura 2014; Tamandl 2014; Tani 2006), and two trials included both, classical and pylorus‐preserving Whipple's procedure (Eshuis 2014; Gangavatiker 2011). Reconstruction was performed by pancreaticojejunostomy in five trials and pancreaticogastrostomy in one trial (Kurahara 2011). All included trials described single‐loop reconstruction.

Different standards were applied for removing the nasogastric tube (NGT): In one trial, the NGT was removed in the operating room and was reinserted only if necessary (Tamandl 2014). In three trials, it was removed routinely if the gastric amount was below 500 mL per day (Imamura 2014; Kurahara 2011; Tani 2006). In one trial, the NGT was removed on or before the third postoperative day, or when daily output had fallen below 300 mL (Eshuis 2014). The NGT was removed if the output was less than 200 mL on two consecutive days in the last trial (Gangavatiker 2011).

Excluded studies

From the the total of 216 abstracts, we excluded 77 because they were duplicates. Of the remaining 139 abstracts, 128 were excluded (71 covered other topics and 57 were not RCTs). The study selection process is displayed in Figure 1.

1.

Study flow diagram.

Eleven publications were potentially eligible (Chijiiwa 2009 [reference listed under Imamura 2014]; Chijiiwa 2012 [reference listed under Imamura 2014]; Eshuis 2012 [reference listed under Eshuis 2014]; Eshuis 2014; Gangavatiker 2011; Imamura 2014; Kurahara 2011; Rebala 2012; Siripong 2012; Tamandl 2014; Tani 2006). The trial by Imamura 2014 presented the final results of both the preliminary report by Chijiiwa 2009 and Chijiwa's congress abstract in 2012, with overlapping patient samples across all three reports. Thus, we extracted only the data from Imamura 2014. Equally the congress abstract of Eshuis 2012 reported on the same trial population as the final report (Eshuis 2014) of which the data were finally extracted. We could not include the data of the trials by Rebala 2012 and Siripong 2012 in the quantitative analysis, since only congress abstracts were available and the presented data were not sufficient to estimate odds ratios for delayed gastric emptying. We contacted the trial authors for further information but did not get an answer from the authors of Rebala 2012; the authors of Siripong 2012 could not provide more data because the final results had not yet been published.

Risk of bias in included studies

Our assessment of risk of bias revealed a heterogenous picture of the design of the included trials. All included trials described a sample size calculation with a power of 80% and an α of 5%, but the underlying assumed incidence of delayed gastric emptying varied in both the retrocolic (30% to 50%) and the antecolic groups (10% to 15%), resulting in planned sample sizes of 91 to 20 patients per group (Eshuis 2014; Tamandl 2014; respectively). Three trials conducted planned interim analyses (Eshuis 2014; Kurahara 2011; Tani 2006). Investigators terminated two trials at that point due to significant inter‐group differences and ethical factors (Kurahara 2011; Tani 2006), whereas investigators in another trial increased the sample size from 91 to 126 patients per arm (Eshuis 2014).

Statistical analysis was performed according to the intention‐to‐treat principle in one trial (Eshuis 2014), and per‐protocol in another (Tamandl 2014); the other trial reports did not describe the statistical analysis models used.

We assessed publication bias by creating a funnel plot for the primary endpoint delayed gastric emptying (Figure 2). We only inspected the funnel plot for potential publication bias; we applied no formal tests since we included fewer than 10 trials in the analyses. There is some asymmetry in the funnel plot caused by the lack of smaller trials favouring retrocolic reconstruction.

2.

Funnel plot of comparison: 1 antecolic vs. retrocolic, outcome: 1.1 delayed gastric emptying (all definitions).

A summary of our risk of bias assessment is given in Figure 3 and Figure 4.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included trials.

4.

Risk of bias summary: review authors' judgements about each risk of bias item for each included trial.

Allocation

The randomisation process was adequately described in three trials as computer‐generated (Eshuis 2014; Gangavatiker 2011; Tani 2006); in one trial, a non‐random approach was used (Tamandl 2014); the remainder did not adequately specify the process of random sequence generation (Imamura 2014; Kurahara 2011).

Three trials described that allocation concealment was maintained by consecutive, sealed envelopes (Gangavatiker 2011; Imamura 2014; Kurahara 2011); central randomisation was performed intraoperatively in two trials (Eshuis 2014; Tani 2006). Thus, selection bias seemed to be low in five trials; in the last trial with the non‐random approach (allocation to group according to birth date), allocation concealment was obviously not sufficient (Tamandl 2014).

Blinding

In one trial, blinding of patients and outcome assessors was described in the protocol, but not in the paper. Blinding of patients was confirmed by the authors when we asked them directly. The authors wrote that "the treating physicians were usually not aware of the route of gastroenteric anastomosis, but formally, there was no blinding" (Eshuis 2014). None of the other trials accomplished blinding, although blinding of patients or outcome assessors would have been possible.

Incomplete outcome data

Two trials did not provide any information about length of follow‐up (Gangavatiker 2011; Kurahara 2011). Follow‐up varied in the other trials from one month after discharge (Eshuis 2014), to one year after surgery (Imamura 2014). In general, delayed gastric emptying appears in the first couple of days after the operation, so a long‐term follow‐up might not be critical; therefore, differences in follow‐up are not important to evaluate attrition bias in this trial. Drop‐outs occurred in some trials when allocation to the planned trial arm was not possible intraoperatively because of feasibility or anatomical reasons (Tamandl 2014), or after randomisation when two patients in each group suffered severe sepsis (Imamura 2014). In one trial, three patients died in the antecolic group and one in the retrocolic group (Gangavatiker 2011). In the trial by Tamandl 2014, attrition bias was judged as high, since data for several patients were not reported due to early discharge of the patients. All other trials reported data on drop‐outs sufficiently, so they could be judged as low risk.

Selective reporting

A published protocol with pre‐specified endpoints was only available for one trial; outcomes in the protocol and in the published report were compared and the risk of selective reporting was declared as low (Eshuis 2014). Two further trials were registered with open accessible clinical trial registries and sufficient information on the prespecified endpoints were available, so the risk for selective reporting was judged as low (Imamura 2014; Tamandl 2014). For the other trials, outcomes listed in the methods section of the article were compared with the results reported in the results section, and risk for reporting bias was judged as unclear.

Other potential sources of bias

In the trial by Eshuis 2014, the NGT management (described above) was a potential source of other bias with unclear risk, since it interfered with the definition of the primary endpoint 'delayed gastric emptying according to the ISGPS definition'. The retrospective use of the ISGPS definition of delayed gastric emptying represented another potential source of other bias in the trial by Gangavatiker 2011. In the trial by Tamandl 2014, the recruitment of patients was not in accordance with the planned sample size calculation; this was judged as an unclear risk of other bias. Finally, the risk for any other bias had to be judged as 'unclear' in the trials by Tani 2006 and Kurahara 2011, since these trials were terminated early, so no formal confrimatory analysis was possible.

Effects of interventions

See: Table 1

Primary outcome

Delayed gastric emptying

All six trials (N = 576; AC = 289; RC = 287) reported delayed gastric emptying; the rate did not differ significantly between the antecolic and retrocolic groups (OR 0.60; 95% CI 0.31 to 1.18; P = 0.14; I² = 57%; Analysis 1.1).

1.1. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 1 Delayed gastric emptying (all definitions).

Three trials used the ISGPS definition of delayed gastric emptying prospectively, whereas Gangavatiker 2011 and Tamandl 2014 applied this definition retrospectively, after they had first applied the Johns Hopkins Criteria (Yeo 1993). Tani 2006 defined delayed gastric emptying in a different way: (1) prolonged aspiration of 500 mL per day from a nasogastric tube left in place for 10 days (DGE10), (2) need for reinsertion of a nasogastric tube, (3) failure of unlimited oral intake by the 14th postoperative day (DGE14).

We performed a sensitivity analysis that only included trials that assessed delayed gastric emptying according to the ISGPS definition. There was no significant difference between the AC group (86/203 participants; 42%) and the RC group (98/205; 48%) when only trials with the ISGPS definition were included (OR 0.62; 95% CI 0.28 to 1.37; P = 0.24; I² = 56%; 3 trials; Analysis 1.2). There was also no significant difference seen in the rate of clinically relevant delayed gastric emptying (ISGPS Grade B/C) between the AC and the RC groups (OR 0.56; 95% CI 0.21 to 1.48; P = 0.24; I² = 46%; 3 trials; N = 408; AC = 203; RC = 205; Analysis 1.3).

1.2. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 2 Delayed gastric emptying (ISGPS definition).

1.3. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 3 Clinically relevant delayed gastric emptying (ISGPS grade B/C).

We performed a subgroup analysis for delayed gastric emptying that only included trials in which pylorus‐preserving pPD was performed (Imamura 2014; Tamandl 2014; Tani 2006). There was no significant difference between the AC and RC groups (OR 0.37; 95% CI 0.11 to 1.23; P = 0.10, I² = 55%; analysis not shown).

We could not perform the other planned subgroup analyses. No trials reported only classical Whipple's procedure and in the trials reporting both, there were no individual data for those subgroups reported (Eshuis 2014; Gangavatiker 2011). All included trials used single‐loop reconstruction, therefore, we could not compare single‐loop versus Roux‐en‐Y reconstruction. We could not perform the subgroup analysis for underlying diagnoses, since the trials did not provide these individual patient data.

Sensitivity analysis was performed that excluded trials with less than the average number of positive judgements in the risk of bias assessment. No significance difference was seen between the two reconstruction techniques (OR 0.97; 95% CI 0.64 to 1.48; P = 0.90; I² = 0%; analysis not shown; Eshuis 2014; Gangavatiker 2011; Imamura 2014)

Fixed effects sensitivity meta‐analysis also showed no significantly different overall effect estimate for delayed gastric emptying between the two reconstruction techniques (OR 0.73; 95% CI 0.51 to 1.05; P = 0.09; I² = 57%; 6 trials; analysis not shown).

We observed a statistical heterogeneity of 41% to 65% in the I² statistic; besides the different endpoint definitions, one reason might be that different drugs were administered, such as, octreotide and erythromycin. Somatostatin was not administered perioperative in most trials (Imamura 2014; Kurahara 2011; Tamandl 2014; Tani 2006); in others, it was administered for a soft pancreas or a non‐dilated pancreatic duct (Eshuis 2014; Gangavatiker 2011). Prokinetics were not applied routinely in most trials (Imamura 2014; Kurahara 2011; Tamandl 2014; Tani 2006), and on demand in the others (Eshuis 2014; Gangavatiker 2011).

Secondary outcomes

Postoperative mortality

Mortality was analysed using risk differences (RD) because both groups in one trial had zero events, which leads to a computational error when summarising these results as odds ratios (Kurahara 2011). Mortality rates were provided by all trials and showed similar ranges in the AC group (0% to 9.4% (Kurahara 2011; Tani 2006 and Gangavatiker 2011, respectively)) and the RC group (0% to 6.4% (Kurahara 2011 and Eshuis 2014, respectively)), with an mortality across all trials of 3.8%. There were no significant differences in mortality between the two groups (RD ‐0.01; 95% CI ‐0.03 to 0.02; P = 0.72; I² = 0%; Analysis 1.4; 6 trials).

1.4. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 4 Postoperative mortality.

Postoperative pancreatic fistula

All six trials reported fistula rates (N = 576; AC = 61/289; RC = 62/287). The fistula rate showed no significant difference between groups (OR 0.98; 95% CI 0.65 to 1.47; P = 0.92; I² = 0%; Analysis 1.5). There was also no significant difference in a subgroup of trials in which the ISGPS definition for pancreatic fistula was prospectively applied (OR 1.00; 95% CI 0.59 to 1.72; P = 0.99; I² = 20%; 3 trials; N = 408; AC = 52/203; RC = 52/205; Analysis 1.6). There was also no significant difference between AC and RC reconstruction in the same three trials in the clinically relevant (Grade B/C) postoperative pancreatic fistula rate (OR 0.85; 95% CI 0.49 to 1.47; P = 0.56; I² = 0%; Analysis 1.7).

1.5. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 5 Postoperative pancreatic fistula.

1.6. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 6 Postoperative pancreatic fistula ISGPF definition.

1.7. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 7 Clinically relevant postoperative pancreatic fistula ( ISGPF grade B/C).

Postoperative haemorrhage

Five trials addressed postoperative haemorrhage (N = 530; AC = 15/265; RC = 18/265). The pooled result was not significantly different (OR 0.79; 95% CI 0.38 to 1.65; P = 0.53; I² = 0%; Analysis 1.8).

1.8. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 8 Postoperative hemorrhage.

Bile leakage

Bile leakage was reported by all six included trials (N = 576; AC = 12/289; RC = 12/287). The risk of bile leakage did not differ significantly between the two groups (OR 0.89; 95% CI 0.36 to 2.15; P = 0.79; I² = 2%; Analysis 1.9), with an overall occurrence of 4.2% in both groups.

1.9. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 9 Bile leakage.

Intra‐abdominal fluid collection or abscess

Rates of intra‐abdominal fluid collection or abscess were reported by all six trials (N = 576; AC = 35/289; RC = 36/287). There was no significant difference between groups (OR 0.93; 95% CI 0.52 to 1.67; P = 0.82; I² = 9%; Analysis 1.10).

1.10. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 10 Intra‐abdominal fluid collection/abscess.

Reoperation rate

Four trials reported this outcome (N = 470; AC = 10/231; RC = 19/239). There was no statistically significant difference between groups (OR 0.59; 95% CI 0.27 to 1.31; P = 0.20; I² = 0%; Analysis 1.11).

1.11. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 11 Reoperation rate.

Duration of operation

Sufficient information on duration of the operation was provided in all six included trials (N = 580; AC = 291; RC = 289; 4 additional patients from the trial by Tamandl 2014, for which only intraoperative data were available). Mean operating time (in minutes) showed no significant difference between the groups (MD ‐7.75; 95% CI ‐20.46 to 4.96; P = 0.23; I² = 0%; Analysis 1.12). Mean operating time (± standard deviation) varied from 273 ± 40.7 to 558 ± 135 minutes in the AC group (Tamandl 2014 and Imamura 2014, respectively); and 280 ± 71.9 to 605 ± 150.3 minutes in the RC group (Tamandl 2014 and Kurahara 2011, respectively).

1.12. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 12 Duration of operation.

Intraoperative blood loss

Intraoperative blood loss was reported in five trials (N = 516; AC = 255; RC = 261) and showed a significant difference in favour of the RC group (MD 39.66; 95% CI 12.38 to 66.94; P = 0.004; I² = 0%; Analysis 1.13). However, a mean reduction in blood loss of about 40 mL does not seem to be clinically relevant during an operation with a mean blood loss of about 1150 mL reported in the included trials.

1.13. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 13 Intraoperative blood loss.

Length of postoperative hospital stay

All included trials reported length of hospital stay (in days; N = 576; AC = 289; RC = 287). Both groups showed similar results (MD ‐0.67; 95% CI ‐2.85 to 1.51; P = 0.55; I² = 44%; Analysis 1.14) with a mean length of hospital stay (± standard deviation) between 12 ± 9.6 and 28.7 ± 5.7 days in the AC group (Eshuis 2014 and Tani 2006, respectively) and 12 ± 8.2 and 47.7 ± 37.7 days in the RC group (Eshuis 2014 and Tani 2006, respectively). These large differences might be explained by the variations in discharge policies between the different countries and hospitals and the different periods of trial conduct.

1.14. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 14 Length of hospital stay.

The sensitivity analysis excluding trials that reported medians and ranges instead of means and standard deviations did not change the significance or the direction of the meta‐analysis (MD ‐3.85; 95% CI ‐10.64 to 2.94; P = 0.27; I² = 72%; analysis not shown).

Time to nasogastric tube (NGT) removal (in days) after surgery

The length of time before the NGT was removed postoperatively was reported in all six trials (N = 576; AC = 289; RC = 287) and did not differ significantly between the groups (MD 0.15; 95% CI ‐0.47 to 0.77; P = 0.63; I² = 45%; Analysis 1.15). However, the mean duration varied widely, from one to five days in the AC group and 1 to 19 days in the RC group.

1.15. Analysis.

Comparison 1 antecolic vs. retrocolic, Outcome 15 Length of postoperative NGT.

As mentioned above, different standards were applied to the removal of the nasogastric tube, which might explain the moderate heterogeneity in the I² statistic. Intertrial comparability cannot be guaranteed for this outcome.

Again, the sensitivity analysis excluding trials that reported medians and ranges instead of means and standard deviations did not change the significance or the direction of the meta‐analysis (MD ‐1.24; 95% CI ‐4.47 to 1.99; P = 0.45; I² = 69%; analysis not shown).

Quality of life

None of the main trial publications reported data on quality of life after surgery. However, quality of life data for a subset of the patients in the trial by Eshuis 2014, more specifically those who were recruited at the initiating centre of this trial, were reported in a separate report (Eshuis 2015 [reference listed under Eshuis 2014]; AC = 35; RC = 38). Quality of life was assessed by the EQ‐5D ^TM questionnaire, the European Organization for Research and Treatment of Cancer QLQ‐C30 and QLQ‐PAN26 questionnaires, and the Gastrointestinal Quality of Life Index preoperatively, and at two, four, and 12 weeks after surgery. No differences in quality of life between the AC and RC groups were assessed at any time point. However, patients with delayed gastric emptying had significantly worse scores compared to patients without delayed gastric emptying at two weeks after surgery.

When the ongoing RCT (NCT01460550) is completed, this meta‐analysis will be updated to include the results of this trial.

Discussion

Summary of main results

Over the last few years, there have been controversial discussions about reconstruction of bowel continuity after partial pancreaticoduodenectomy (pPD) via an antecolic or retrocolic route, and its potential influence on the incidence of delayed gastric emptying. Several randomised controlled trials and non‐randomised studies that examined reconstruction after classical, pylorus‐preserving and pylorus‐resecting Whipple have been performed, but results were inconclusive. Likewise, results of previously published meta‐analyses about the best type of reconstruction showed conflicting evidence (Ramia 2013; Su 2012; Zhou 2015). Thus, a current meta‐analysis was needed to include the three most recent RCTs, which had not been included in previous meta‐analyses. The qualitative and quantitative summary of existing results in this systematic review and meta‐analysis showed no evidence of superiority of one procedure over the other as indicated in the Table 1 for the main outcome parameters. Furthermore, clinically relevant delayed gastric emptying (defined as grade B/C according to the ISGPS definition) and a subgroup analysis, including only trials that used the ISGPS definition, resulted in no significant difference between the two groups either.

The type of reconstruction (antecolic (AC) versus retrocolic (RC)) also had no impact on mortality, further morbidity, length of hospital stay, or quality of life after pPD.

Overall completeness and applicability of evidence

The six included trials, that could be identified by the systematic literature search, provided comprehensive data on all of our predefined outcome parameters except of quality of life. Quality of life was only reported on a subset of patients from one trial (Eshuis 2014) and thus, the evidence for this outcome is by far weaker than for the other outcome parameters. Since the trials were conducted in different international regions, the results are applicable to a large population of patients undergoing pPD. Furthermore, we performed a subgroup analysis for pylorus‐preserving pPD and could not find a statistical difference for delayed gastric emptying in this subgroup, which strengthens the evidence for this subgroup. Equally, the sensitivity analyses did not change the overall results of our meta‐analyses for the primary outcome parameter.

Period and frequency of follow‐up was not described sufficiently and differed between trials, but for the primary endpoint, delayed gastric emptying, the immediate post‐operative period seems to be the most relevant, so that a severe attrition bias concerning delayed gastric emptying did not have to be suspected. However, long‐term differences between the two procedures may have been missed due to the short follow‐up in most of the trials. For instance, Imamura 2014 showed that postoperative weight gain was quicker and the retrocolic group recovered almost back to their preoperative weight after one year, whereas weight loss was prolonged in the antecolic group. Such long‐term differences might be crucial in the discussion about the optimal reconstruction technique and should be addressed in future trials.

Our findings are in line with those of the recently published multicentre RCT with the largest sample size, in which the incidence of delayed gastric emptying was examined according to the type of bowel reconstruction (Eshuis 2014). The results of this trial indicated that neither the overall delayed gastric emptying (AC versus RC: 61% versus 60%; P = 0.89), nor the clinically relevant delayed gastric emptying (AC versus RC: 34% versus 36%, resulting in an absolute risk difference of 2.1%) differed between the two trial groups. Since this trial had the largest sample size it also achieved the highest weight in the meta‐analysis. Thus, it had a strong influence on the results of our meta‐analysis. However, the NGT management policy, i.e. all patients received a NGT postoperatively for a median of three days, led to high rates in both trial groups, because of the interference with the ISGPS definition, which is based on the presence of a NGT after postoperative day three.

Several reports have shown that routine nasogastric drainage after pPD is not necessary and routine postoperative NGT does not represent the current standard in many centres that specialise in pancreatic surgery (Choi 2011; Fisher 2011; Robertson 2012). Therefore, the impact of the reconstruction method might be blurred by the excessive rates of delayed gastric emptying in the largest trial by Eshuis 2014.

Since all of the meta‐analyses of delayed gastric emptying (overall delayed gastric emptying, subgroup ISGPS definition of delayed gastric emptying, and clinically relevant delayed gastric emptying) showed lower rates in the antecolic route, the current meta‐analysis might still be underpowered to demonstrate a potential difference. However, in light of the current evidence, no relevant difference in delayed gastric emptying after antecolic and retrocolic reconstruction can be suggested.

Quality of the evidence

We included six trials from different countries with a total of 576 participants in this review. Interstudy results were inconsistent. On one hand, there were obvious sources of potential bias in the included trials, e.g. varying definitions of delayed gastric emptying, missing blinding of outcome assessors and patients, small sample sizes with two trials stopped at interim analyseis, and potential reporting bias. On the other hand, this systematic review and meta‐analysis has been performed with a clear methodology, thus creating a higher level of evidence through pooling of data from all available RCTs on this issue. Nevertheless, the overall quality of the evidence according to the GRADE criteria was only low to moderate, because it had to be downgraded due to potential risk of bias, some imprecision of the results and heterogeneity between the trials (Table 1).

Potential biases in the review process

Well‐conducted systematic reviews can allow generalisation of scientific findings and increase power and precision in estimating effects and risks (Mulrow 1994). They allow for a more objective appraisal of the evidence, which may lead to resolution of uncertainty and disagreement (Egger 1997). Obviously, the quality of a systematic review is directly depending on the quality of the reviewed studies. In spite of the sources of clinical and methodological heterogeneity already mentioned, inclusion criteria and baseline population characteristics were adequately comparable between included trials.

Even though, a comprehensive and systematic literature search of three large databases was conducted and reference lists of relevant publications were screened, some relevant trials might have been missed by the search. Concerning the data of the included trials, most publications presented sufficient information on the prespecified endpoints and further relevant information was achieved by contacting the respective authors. Data accuracy was secured by double checks of two independent reviewers.

For the continuous outcomes, some trials only reported medians and ranges or interquartile ranges. Means and standard deviations were calculated by the methods described by Higgins 2011 and Hozo 2005. Since reporting of medians and ranges in primary reports is suspicious of skewed data, there might remain some bias in the calculated means and standard deviations.

A further potential source of bias in the review process that has to be mentioned is the variation of the definition of delayed gastric emptying as the internationally accepted and scaled ISGPS definition was published in 2007 (Wente 2007a), when some of the included trials had already been started. The use of different definitions represent one restriction of the performed meta‐analysis, since it has been shown that applying different definitions of delayed gastric emptying on the same population introduces a large variation in the incidence of delayed gastric emptying, from 5.9% to 14.7% (Butturini 2006). Three trials prospectively used the definition of the most widely accepted consensus definition of the ISGPS, based on the duration of postoperative (naso)gastric drainage and return to solid food (Eshuis 2014; Imamura 2014; Kurahara 2011). Two trials retrospectively applied this ISGPS definition (Gangavatiker 2011; Tamandl 2014), after they had first applied the Johns Hopkins Criteria (Yeo 1993); another trial used a third definition of delayed gastric emptying similar to the Johns Hopkins criteria (Tani 2006). This heterogeneous definition of the primary endpoint may have caused some bias and the substantial heterogeneity in the meta‐analyses of delayed gastric emptying. However, a subgroup analysis was performed that only included trials primarily using the ISGPS definition prospectively. Moreover, perioperative management, in terms of the use, duration and indication of nasogastric tubes, influences the occurrence of delayed gastric emptying; in this case, inter‐study heterogeneity is obvious. Furthermore, prokinetics and somatostatin were applied as part of various standardised or non‐standardised therapeutic concepts in the different trials. Due to the lack of reporting of indication, duration, and dosage of these medications, an inter‐study heterogeneity must be suspected, which might decrease the external validity of our results.

Agreements and disagreements with other studies or reviews

One previously published meta‐analysis based on two RCTs and three non‐randomised studies favoured an antecolic reconstruction (Su 2012). Another meta‐analysis including four RCTs and seven non‐randomised studies came to the conclusion that the route of gastroenteric reconstruction with a benefit in delayed gastric emptying cannot currently be determined (Ramia 2013). Both studies were published before three RCTs, including the two largest RCTs, addressing this question were available (Eshuis 2014; Imamura 2014; Tamandl 2014). Two meta‐analyses on this issue were published in 2015 (Bell 2015; Zhou 2015). One of them included four non‐randomised studies (Bell 2015). Neither of them included the RCT by Kurahara 2011, both of them included the preliminary report of Chijiiwa 2009 and the final results of Imamura 2014 with their overlapping patient samples, which was not addressed in the review. The review including only RCTs concluded that the route of gastroenteric reconstruction had no impact on delayed gastric emptying, thus, our results agreed (Zhou 2015). On the other hand, Bell 2015 concluded that antecolic reconstruction was associated with a lower incidence of delayed gastric emptying, which might be because they included non‐randomised studies.

In summary, the results of previously published meta‐analyses are conflicting and their results must be interpreted with caution, due to the inclusion of non‐randomised trials and other potential sources of bias in the review process.

Authors' conclusions

Implications for practice.

In summary, we must recognise that on the basis of available evidence, there is no significant difference in delayed gastric emptying, overall morbidity, and mortality between antecolic and retrocolic reconstruction for bowel continuity. Delayed gastric emptying remains a burdensome and frequent complication after pancreaticoduodenectomy (PD), with its causative factors remaining still largely unknown. As shown by Qu and colleagues in a systematic review and meta‐analysis of the risk factors for delayed gastric emptying, preoperative diabetes, pancreatic fistulas, and postoperative complications were predictive risk factors for delayed gastric emptying (Qu 2013). As clinically relevant delayed gastric emptying often occurs in patients also suffering from other intra‐abdominal complications, such as postoperative pancreatic fistula, bile leakage, and abscess, prevention of these complications is necessary to reduce the incidence of delayed gastric emptying. Currently, gastroenteric reconstruction after PD might be routed according to the individual hospital standard and preference of the surgeon.

Implications for research.

Concerning the difference of 7% in the occurrence of delayed gastric emptying favouring antecolic reconstruction, the previous trials and even this strictly conducted meta‐analysis might have been underpowered to show a significant benefit. Furthermore, the differences in nasogastric tube (NGT) management might have influenced the results of this meta‐analysis. Therefore, a large‐scale RCT with clearly defined NGT management, according to modern treatment protocols, might finally elucidate the impact of antecolic versus retrocolic reconstruction. Future trials should also focus on long‐term follow‐up and patient‐relevant outcomes such as quality of life, to elucidate potential differences between these two procedures, which might have been missed by previous trials.

Furthermore, different surgical approaches to reduce the rates of delayed gastric emptying after PD should be assessed in future RCTs. One surgical variation that has been discussed in this respect is single‐loop versus dual‐loop reconstruction after PD. The results of a recent meta‐analysis did not show a benefit in terms of delayed gastric emptying or other postoperative complications; however, it did find a significantly prolonged operation time with dual‐loop reconstruction (Klaiber 2015). Another aspect, which might have an impact on delayed gastric emptying is pylorus‐resection versus pylorus‐preservation. We look forward to the results of the PROPP trial, the trial protocol of which was published in 2013 (Hackert 2013). Since it is unlikely that delayed gastric emptying is resolved only by a surgical solution, etiology, prevention, and especially therapeutic approaches in delayed gastric emptying require further multidisciplinary research.

It will be essential for all future trials considering delayed gastric emptying, to use strict definitions such as the ISGPS definition of delayed gastric emptying consistently. Core outcome sets (e.g. comet‐initative.org) might be a powerful tool to evaluate complex interventions, such as pancreaticoduodenectomy, in future trials and will make results of individual trials more comparable.

Appendices

Appendix 1. Glossary of terms

Abscess: A collection of pus that has built up within the body.

Adenocarcinoma: A type of cancer that can occur in several parts of the body with a glandular origin.

Adjuvant chemotherapy: Chemotherapy given after, and in addition to, resection of the tumour.

Anastomosis: Connection between two organs (e.g. stomach and small intestine), created by surgery.

Anterior: Situated in front of something.

Asymptomatic: Showing no clinical signs of disease or condition.

Benign: Tumours lacking the ability to invade neighbouring tissue or to metastasise.

Biliary: Anatomical term related to the bile secretion and ducts.

Chemotherapy: Medication administered to treat cancer.

Duodenum: The first part of the small intestine, connecting the stomach to the jejunum.

Erythromycin: An antibiotic increasing the motility of the intestine.

Etiology: Factors causing an illness.

Gastric: Related to the stomach.

Gastro‐/duodenojejunostomy: Surgical connection between the stomach or duodenum and the small intestine.

Gastroenteric: Connection between the stomach and the small intestine.

Haemorrhage: Bleeding.

Intra‐abdominal: Within the abdomen.

Intraductal: Within a duct.

Jejunum: Second part of the small intestine, connecting the duodenum to the ileum.

Mesocolic window: Artificial pathway through the mesocolon.

Mesocolon: Fatty tissue carrying the blood vessels that supply the colon.

Morbidity: Consequences of a procedure impairing the subject's health.

Mortality: A measure of the number of deaths in a given population.

Motilin‐agonist: Drug acting similar to an intestinal hormone, which promotes bowel movements.

Neoplasia: Tumour.

Pancreatic fistula: Opening between pancreas and another organ or space, allowing leakage of pancreatic secretions from the pancreatic gland.

Pancreaticoduodenectomy: Partial surgical removal of the pancreas, with the attached duodenum.

Pancreaticojejunal: Referring to the pancreatic remnant and the second part of the small intestine.

Pancreatitis: Inflammation of the pancreas.

Perioperative: In direct temporary context to surgery (before, during, and after).

Posterior: Situated behind something.

Premalignant: Precancerous.

Prokinetic: Stimulating the movements of the the oesophagus, stomach, and intestines.

Prophylaxis: Prevention.

Pylorus: Circular muscle building the connection between the stomach and the small intestine.

Resectable: Able to be removed by surgery.

Somatostatin analogues: Proteins that slow down the production of hormones, the emptying of stomach and bowel, and the release of hormones from the pancreas.

Torsion: Twisting of a structure (e.g. intestine).

Transverse colon: The middle part of the colon.

Appendix 2. CENTRAL search strategy

#1 MeSH descriptor: [Pancreaticoduodenectomy] explode all trees

#2 pancreaticoduodenectomy* or pancreatoduodenectom* (Word variations have been searched)

#3 pancreas* near/5 duodenectom* (Word variations have been searched)

#4 Whipple and pancrea* (Word variations have been searched)

#5 PPD and pancrea* (Word variations have been searched)

#6 pancrea* near/5 (surger* or operat* or resect*) (Word variations have been searched)

#7 #1 or #2 or #3 or #4 or #5 or #6

#8 antecolic or retrocolic or mesocolic (Word variations have been searched)

#9 #7 and #8

Appendix 3. MEDLINE search strategy

1946 to Present; via OVID

1 exp Pancreaticoduodenectomy/ or exp Pancreatectomy/

2 (pancreaticoduodenectomy* or pancreatoduodenectom*).ti,ab,kw.

3 (pancreas* adj5 duodenectom*).ti,ab,kw.

4 (Whipple and pancrea*).ti,ab,kw.

5 (PPD and pancrea*).ti,ab,kw.

6 (pancrea* adj5 (surger* or operat* or resect*)).ti,ab,kw.

7 or/1‐6

8 (antecolic or retrocolic or mesocolic).ti,ab,kw.

9 7 and 8

10 exp animals/ not humans/

11 9 not 10

12 remove duplicates from 11

Appendix 4. EMBASE search strategy

1974 to 2014 November 26

1 exp pancreaticoduodenectomy/

2 (pancreaticoduodenectom* or pancreatoduodenectom*).ti,ab,kw.

3 (pancrea* adj5 duodenectom*).ti,ab,kw.

4 (Whipple and pancrea*).ti,ab,kw.

5 (PPD and pancrea*).ti,ab,kw.

6 exp pancreas resection/ or exp pancreas surgery/

7 (pancrea* adj5 (surger* or operat* or resect*)).ti,ab,kw.

8 or/1‐7

9 (antecolic or retrocolic or mesocolic).ti,ab,kw.

10 8 and 9

11 exp animal/ not human/

12 10 not 11

13 remove duplicates from 12

Data and analyses

Comparison 1. antecolic vs. retrocolic.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Delayed gastric emptying (all definitions)	6	576	Odds Ratio (M‐H, Random, 95% CI)	0.60 [0.31, 1.18]
2 Delayed gastric emptying (ISGPS definition)	3	408	Odds Ratio (M‐H, Random, 95% CI)	0.62 [0.28, 1.37]
3 Clinically relevant delayed gastric emptying (ISGPS grade B/C)	3	408	Odds Ratio (M‐H, Random, 95% CI)	0.56 [0.21, 1.48]
4 Postoperative mortality	6	576	Risk Difference (M‐H, Random, 95% CI)	‐0.01 [‐0.03, 0.02]
5 Postoperative pancreatic fistula	6	576	Odds Ratio (M‐H, Random, 95% CI)	0.98 [0.65, 1.47]
6 Postoperative pancreatic fistula ISGPF definition	3	408	Odds Ratio (M‐H, Random, 95% CI)	1.00 [0.59, 1.72]
7 Clinically relevant postoperative pancreatic fistula ( ISGPF grade B/C)	3	408	Odds Ratio (M‐H, Random, 95% CI)	0.85 [0.49, 1.47]
8 Postoperative hemorrhage	5	530	Odds Ratio (M‐H, Random, 95% CI)	0.79 [0.38, 1.65]
9 Bile leakage	6	576	Odds Ratio (M‐H, Random, 95% CI)	0.89 [0.36, 2.15]
10 Intra‐abdominal fluid collection/abscess	6	576	Odds Ratio (M‐H, Random, 95% CI)	0.93 [0.52, 1.67]
11 Reoperation rate	4	470	Odds Ratio (M‐H, Random, 95% CI)	0.59 [0.27, 1.31]
12 Duration of operation	6	580	Mean Difference (IV, Random, 95% CI)	‐7.75 [‐20.46, 4.96]
13 Intraoperative blood loss	5	516	Mean Difference (IV, Random, 95% CI)	39.66 [12.38, 66.94]
14 Length of hospital stay	6	576	Mean Difference (IV, Random, 95% CI)	‐0.67 [‐2.85, 1.51]
15 Length of postoperative NGT	6	576	Mean Difference (IV, Random, 95% CI)	0.15 [‐0.47, 0.77]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Eshuis 2014.

Methods	Randomised controlled trial Exclusion after randomisation (total and per group): 6 (3 AC; 3 RC), due to total pancreatectomy Losses to follow‐up: no Intention‐to‐treat analysis: yes Description of sample size calculation: yes Duration of trial: 2009 to 2011
Participants	Number: 246 (AC: 121; RC: 125) Age: 65.4± 9.0 (mean in the AC group); 65.2± 10.3 (mean in the RC group) Sex ratio (M/F): 2.2 (AC), 1.2 (RC) Inclusion criteria: Patients ≥ 18 years scheduled to undergo PD Exclusion criteria: other surgical procedures besides PD Disease type: pancreatic adenocarcinoma (AC: 49; RC: 34), ampullary adenocarcinoma (AC: 20; RC: 21), distal bile duct adenocarcinoma (AC: 13; RC: 21), duodenal adenocarcinoma (AC: 6; RC: 2), other (pre)malignant lesions (AC: 8; RC: 10), chronic pancreatitis (AC: 2; RC: 7) other benign lesions (AC: 2; RC: 3) Equivalence of baseline characteristics: yes
Interventions	Antecolic reconstruction (AC): The anastomosis was positioned anterior to the transverse colon. No adjunctive measures to enhance gastric emptying, such as pyloric stretching, were taken. Retrocolic reconstruction (RC): The duodenal stump (or distal stomach) was brought down through a separate opening in the transverse mesocolon, at the left side of the middle colic artery. It was anastomosed end‐to‐side to the jejunum. The stomach was fixed to the mesocolon, to prevent herniation of the loop, thereby ensuring that the gastroenteric anastomosis was not positioned in the same abdominal compartment as the pancreaticojejunostomy and hepaticojejunostomy. Administration of prokinetics: only as needed Administration of somatostatin: For the indication of soft pancreatic tissue or a non‐dilated pancreatic duct. Standard management of NGT: All patients received an NGT. The NGT had to be removed before or on the third POD, or when daily output had fallen below 300 mL.
Outcomes	Primary outcome: postoperative incidence of delayed gastric emptying, according to the ISGPS consensus definition, and incidence of 'primary' delayed gastric emptying (delayed gastric emptying occurring in the absence of other intra‐abdominal complications). Secondary outcomes: morbidity, mortality, length of hospital stay, pancreatic fistula, haemorrhage, bile leakage, chylous ascites, intra‐abdominal abscess, wound infection, nonsurgical complications (pneumonia, other pulmonary complications, myocardial infarction, other cardiac complications, urinary tract infections, cerebrovascular accidents). Re‐laparotomies for any cause and hospital readmissions within 30 days of discharge
Notes	Country: The Netherlands The authors declared no conflicts of interest. No financial support had been taken.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	quote: 'computer‐generated random‐numbers, blocked randomisation stratified to trial centre'
Allocation concealment (selection bias)	Low risk	quote: 'central randomisation performed intraoperatively'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	quote: 'treating physicians were not blinded for the treatment allocation' comment: patients were blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	comment: no blinding of outcome assessors described
Incomplete outcome data (attrition bias) All outcomes	Low risk	comment: missing outcome data balanced in numbers across intervention groups, with same reasons for missing data across groups, transparently described in the patient flow chart.
Selective reporting (reporting bias)	Low risk	comment: trial registered at the Dutch primary register for clinical trials, trial protocol with pre‐specified endpoints available
Other bias	Low risk

Gangavatiker 2011.

Methods	Randomised controlled trial Exclusion after randomisation (total and per group): 4 died (3 AC; 1 RC) Losses to follow‐up: no Intention‐to‐treat analysis: not stated Description of sample size calculation: yes Duration of trial: September 2006 to November 2008
Participants	Number: 72 ‐ 4 (AC: 35 ‐ 3; RC: 37 ‐ 1) Age: 52.8± 11.6 (mean in the AC group); 50.8± 10.6 (mean in the RC group) Sex ratio (M/F): 2.2 (AC); 2.6 (RC) Inclusion criteria: Patients < 70 years with a good performance status who underwent a classical Whipple’s PD or a PPPD Exclusion criteria: Patients with metastatic and locally advanced disease, peptic ulcer disease, gastric outlet obstruction, tumour infiltration into the stomach, previous gastric surgery, and poorly controlled diabetes mellitus (long standing history of diabetes with poor glycaemic control (HbA1c > 7.5%) or systemic complications of diabetes) Disease type: carcinoma (periampullary, duodenal, and pancreatic head), neuroendocrine tumours, or chronic pancreatitis Equivalence of baseline characteristics: yes
Interventions	Antecolic reconstruction: the jejunal loop was brought up anterior to the transverse colon and anastomosed to the duodenum or stomach Retrocolic reconstruction: the jejunal loop was anastomosed to the duodenum or stomach through a separate mesocolic window on the left of the middle colic vessels. Administration of prokinetics: only as needed Administration of somatostatin: For the indication of soft pancreas and small ducts Standard management of NGT: The NGT was removed if the output was < 200 mL on two consecutive days.
Outcomes	Primary outcome: delayed gastric emptying, defined by the Johns Hopkins criteria; ISGPS criteria were applied retrospectively Secondary outcome: postoperative NGT, pancreatic fistula, mortality, haemorrhage, abscess, operation time, re‐operation rate, length of hospital stay
Notes	Country: India The authors declared no conflicts of interest. No funding declared.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	quote: 'computer‐generated random‐numbers'
Allocation concealment (selection bias)	Low risk	quote: 'sealed envelope technique'
Blinding of participants and personnel (performance bias) All outcomes	High risk	comment: blinding not described, probably not done
Blinding of outcome assessment (detection bias) All outcomes	High risk	comment: blinding not described, probably not done
Incomplete outcome data (attrition bias) All outcomes	Low risk	comment: transparent patient flow chart available, 3 patients died in the AC group, 1 patient died in the RC group; otherwise, all patients completed the trial and there were no further losses to follow‐up.
Selective reporting (reporting bias)	Unclear risk	no trial registration, no published protocol
Other bias	Unclear risk	ISGPS criteria applied retrospectively

Imamura 2014.

Methods	Randomised controlled trial Exclusion after randomisation (total and per group): 4 (2 AC; 2 RC) Losses to follow‐up: 45 patients (26 AC; 19 RC) lost to follow‐up after 1 year Intention‐to‐treat analysis: not stated Description of sample size calculation: yes Duration of trial: March 2005 to July 2011
Participants	Number: 120‐4 patients (58 AC; 58 RC) Age: 70.0 (36 to 86, median in the AC group); 69.0 (46 to 86, median in the RC group) Sex ratio (M/F): 1.6 (AC); 1.2 (RC) Inclusion criteria: PPPD Exclusion criteria: Patients who underwent pancreaticoduodenectomy with gastric resection, subtotal stomach‐preserving PD (SSPPD), additional hepatic resection and total pancreatectomy, old cerebral infarction, post spinal cord injury state, haemodialysis for chronic renal failure Disease type: pancreatic cancer (17 AC; 16 RC), bile duct cancer (17 AC; 20 RC), ampullary carcinoma (4 AC; 9 RC), duodenal cancer (2 AC; 0 RC), cystic tumour (IPMN/MCN, 11 AC; 6 RC), chronic pancreatitis (2 AC; 16 RC), benign bile duct disease (2 AC; 2 RC), others (3 AC; 2 RC) Equivalence of baseline characteristics: yes
Interventions	Antecolic reconstruction: the stomach and duodenum were brought down in a straight vertical manner anterior to the transverse colon. The anastomosis was performed at the caudal to the transverse colon. Retrocolic reconstruction: the left side of the transverse mesocolon (to the left of the middle colic vessels) was opened, and the stomach and duodenum were brought down in a straight, vertical manner. The retrocolic duodenojejunostomy was performed at the caudal side of the transverse mesocolon, and the gastric antrum was fixed to the transverse mesocolon with several sutures. A Braun anastomosis was added in both reconstruction procedures. Administration of prokinetics: no Administration of somatostatin: no Standard management of NGT: The NGT was routinely removed on POD 1 if the gastric amount was below 500 mL. If the patients vomited persistently, it was reinserted.
Outcomes	Primary outcome: Incidence of delayed gastric emptying defined by ISGPS criteria Secondary outcome: post‐operative complications besides delayed gastric emptying, evaluation of gastric emptying and nutritional status for 1 year after surgery
Notes	Country: Japan No conflicts of interest declared. Funded by the Japanese Ministry of Education, Culture, Sports, Science and Technology.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	quote: 'equal numbers of envelopes for antecolic and retrocolic were sequentially prepared' comment: sequence generation not described
Allocation concealment (selection bias)	Low risk	quote: 'sequentially prepared envelopes in a blinded fashion'
Blinding of participants and personnel (performance bias) All outcomes	High risk	no blinding performed as described in trial registry
Blinding of outcome assessment (detection bias) All outcomes	High risk	no blinding performed as described in trial registry
Incomplete outcome data (attrition bias) All outcomes	Low risk	comment: sufficient reporting of drop‐outs and losses in flow chart with reasons for exclusion
Selective reporting (reporting bias)	Unclear risk	comment: trial registered, National Clinical Database (University Hospital Medical Information Network Clinical Trials Registry as UMIN000001712),with sufficient information on pre‐specified endpoints. Definition of delayed gastric emptying changed during the trial (trial started 2005, ISGPS definition published 2007).
Other bias	Low risk

Kurahara 2011.

Methods	Randomised controlled trial Exclusion after randomisation (total and per group): none Losses to follow‐up: none Intention‐to‐treat analysis: not stated Description of sample size calculation: yes Duration of trial: May 2007 to June 2010
Participants	Number: 46 patients (24 AC; 22 RC) Age: 67.6±11.6 (mean in the AC group); 62.3±12.6 (mean in the RC group) Sex ratio (M/F): 2.0 (AC); 2.7 (RC) Inclusion criteria: SSPPD with pancreatogastrostomy for pancreatic head and periampullary tumours Exclusion criteria: classic PD or PPPD, pancreaticojejunostomy reconstruction, distant metastasis, or local unresectable tumours Disease type: pancreatic cancer (10 AC; 10 RC), bile duct cancer (7 AC; 4 RC), neuroendocrine tumour (3 AC; 2 RC), ampullary cancer (0 AC; 2 RC) duodenal cancer (1 AC; 1 RC) Equivalence of baseline characteristics: yes
Interventions	Antecolic reconstruction performed in the following sequence: end‐to‐side pancreatogastric anastomosis, end‐to‐side choledochojejunostomy, and end‐to‐side gastrojejunostomy by the antecolic route Retrocolic reconstruction performed in the following sequence: end‐to‐side pancreatogastric anastomosis using an internal stent, end‐to‐side gastrojejunostomy, and end‐to‐side choledochojejunostomy using an internal stent Administration of prokinetics: not stated Administration of somatostatin: not stated Standard management of NGT: The NGT was removed when drainage was < 500 mL/day, but it was reinserted in case of vomiting or nausea.
Outcomes	primary outcome: delayed gastric emptying (ISGPS definition) secondary outcome: rate of other postoperative complications, time‐to‐full resumption of diet, surgical mortality, readmission date, and length of postoperative hospital stay (LOS)
Notes	A 'Billroth‐1'‐type reconstruction reconstruction was compared to a 'Billroth‐2'‐type reconstruction Trial was terminated after planned interim analysis because of obviously lower delayed gastric emptying in AC group than in RC group. COI and sources of funding not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	comment: sequence generation not reported sufficiently
Allocation concealment (selection bias)	Low risk	quote: 'consecutive sealed envelopes containing random numbers'
Blinding of participants and personnel (performance bias) All outcomes	High risk	comment: no blinding described
Blinding of outcome assessment (detection bias) All outcomes	High risk	comment: no blinding described
Incomplete outcome data (attrition bias) All outcomes	Low risk	comment: All patients completed the trial and there were no losses to follow‐up
Selective reporting (reporting bias)	Unclear risk	comment: no trial registration; no protocol available
Other bias	Unclear risk	trial terminated early after interim analysis.

Tamandl 2014.

Methods	Randomised controlled trial Exclusion after randomisation (total and per group): 5 (2 AC; 3 RC, allocated treatment was not performed, due to surgeon's preference) Losses to follow up: 4 due to early discharge (2 AC; 2 RC) Intention‐to‐treat analyses: no; per‐protocol analysis used Description of sample size calculation: yes Duration of trial: April 2007 to November 2009
Participants	Number: 71 (38 AC; 33 RC) Age: 67.1 (55.7 to 75.3 median in the AC group); 65.4 (55.6 to 70.6 median in the RC group) Sex ratio (M/F): 0.9 (AC); 0.75 (RC) Inclusion criteria: adults between the ages of 18 and 90 years undergoing PPPD for cancer of either the pancreatic head, uncinate process, ampulla or distal common bile duct, or with a radiographically suspicious solid or cystic tumour requiring pancreaticoduodenectomy Exclusion criteria: distant metastases; locally unresectable tumours (arterial involvement or involvement of the portal vein or superior mesenteric vein); invasion of the stomach; prior surgical resection of the stomach or duodenum. Disease type: ductal adenocarcinoma (AC: 20; RC: 13), distal cholangiocarcinoma (AC: 1; RC: 1), ampullary cancer (AC: 4; RC: 5), IPMN (AC: 2; RC: 0), SCN/MCN (AC: 1; RC: 1), chronic pancreatitis (AC: 5; RC: 7), other (AC: 3; RC: 1) Equivalence of baseline characteristics: yes
Interventions	Antecolic reconstruction: antecolic end‐to‐side duodenojejunostomy Retrocolic reconstruction: retrocolic end‐to‐side duodenojejunostomy Administration of prokinetics: not given on a routine basis Administration of somatostatin: not reported Standard management of NGT: The NGT was removed in the operating room and was reinserted only in case of repeated vomiting or abdominal distension, or the inability to ingest food.
Outcomes	Primary outcome: delayed gastric emptying as defined by clinical criteria on POD 10. Delayed gastric emptying is defined as the NGT remaining in place beyond POD 10 and one of the following criteria: (1) emesis after NGT removal; (2) NGT reinsertion; (3) failure to progress with the diet; (4) use of prokinetics after POD 10. If the NGT was removed before day 10, two of the previous criteria had to be fulfilled to qualify for delayed gastric emptying. Secondary outcome: gastric emptying determined by the paracetamol absorption test; the kinetics of intestinal peptides (GLP‐1 and PYY) after ingestion of a test meal; postoperative length of hospital stay; morbidity and mortality.
Notes	Country: Austria The authors did not disclose any potential conflict of interest. Funding not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	quote: 'patients with an uneven birth date (e.g. February 1) were allocated to the antecolic arm. Patients who had an even birth date were treated with retrocolic reconstruction.' comment: non‐random approach
Allocation concealment (selection bias)	High risk	comment: obvious allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	comment: no blinding performed
Blinding of outcome assessment (detection bias) All outcomes	High risk	comment: no blinding performed
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	comment: transparent patient flow chart: 2 AC; 5 RC patients excluded intraoperatively due to technical feasibility or anatomical reasons; 2 AC; 2 RC patients excluded from analysis due to early discharge, before POD 10 (leaving 36 in the AC group and 28 in the RC group)
Selective reporting (reporting bias)	Low risk	comment: trial registry
Other bias	Unclear risk	comment: recruitment did not follow sample size calculation; ISGPS definition only applied retrospectively

Tani 2006.

Methods	Randomised controlled trial Exclusion after randomisation (total and per group): none Losses to follow‐up: none Intention‐to‐treat analysis: not stated Description of sample size calculation: yes Duration of trial: May 2002 to April 2004
Participants	Number: 40 (20 AC; 20 RC) Age: 63.1± 9.21 (mean in the AC group); 66.7± 12.2 (mean in the RC group) Sex ratio (M/F): 1.2 (AC); 1.0 (RC) Inclusion criteria: ppPD for periampullary and bile duct lesions Exclusion criteria: peptic ulcer, tumour infiltration in the stomach, metastasis into lymph nodes of the pre‐pylorus Disease type: pancreatic cancer (AC: 4; RC: 10), bile duct cancer (AC: 10; RC: 2), ampullary cancer (AC: 2; RC: 4), IPMN (AC: 2; RC: 3), solid‐pseudopapillary tumour (AC: 1; RC: 0), pancreatitis (AC: 1; RC: 1) Equivalence of baseline characteristics: yes
Interventions	Antecolic reconstruction Retrocolic reconstruction Administration of prokinetics: not given Administration of somatostatin: not given Standard management of NGT: The time for removal of the NGT was determined when drainage was < 500 mL/day.
Outcomes	Primary outcome: delayed gastric emptying defined as: (1) prolonged aspiration of 500 mL/day from a nasogastric tube left in place for 10 days (DGE10), (2) need for reinsertion of a nasogastric tube, or (3) failure of unlimited oral intake by the 14th POD (DGE14). Percentage of oral intake of solid foods was defined as the ratio between actual intake and provided diet. Secondary outcome: mortality and morbidity, including pancreatic fistula, intra‐abdominal haemorrhage, and intra‐abdominal abscess
Notes	Trial was terminated after the planned interim analysis, since the delayed gastric emptying rate was significantly lower in the antecolic group than in the retrocolic group. Country: Japan No COI statement. Funding not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	quote: 'computer‐generated random‐number pattern'
Allocation concealment (selection bias)	Unclear risk	comment: allocation concealment not described; intraoperative randomisation
Blinding of participants and personnel (performance bias) All outcomes	High risk	comment: no blinding described
Blinding of outcome assessment (detection bias) All outcomes	High risk	comment: no blinding described
Incomplete outcome data (attrition bias) All outcomes	Low risk	comment: All patients completed the trial and there were no losses to follow‐up; one patient died in the RC group due to gastrointestinal bleeding.
Selective reporting (reporting bias)	Unclear risk	comment: trial not registered; no published protocol
Other bias	Unclear risk	trial terminated at interim analysis

NGT = nasogastric tube

POD = postoperative day

PD = pancreaticoduodenectomy

ISGPS = International Study Group of Pancreatic Surgery

PPPD = pylorus preserving pancreaticoduodenectomy

IPMN = intraductal papillary mucinous neoplasm

MCN = mucinous cystic neoplasm

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Rebala 2012	Congress abstract only. Insufficient data available for inclusion in review. Authors contacted but no answer.
Siripong 2012	Congress abstract only. Insufficient data available for inclusion in review. Authors contacted but no data available, since final results have not yet been published.

Characteristics of ongoing studies [ordered by study ID]

NCT01460550.

Trial name or title	A prospective randomized controlled multicenter trial of delayed gastric emptying (DGE) after pancreaticoduodenectomy evaluating by gastrointestinal reconstruction
Methods	Randomised controlled trial
Participants	240 (planned)
Interventions	antecolic versus retrocolic gastro‐ or duodeno‐ jejunostomy in the pancreaticoduodenectomy
Outcomes	delayed gastric emptying, other postoperative morbidity
Starting date	November 2011
Contact information	Ippei Matsumoto, Kobe University, Kobe, Hyogo, Japan; email: imatsu@med.kobe‐u.ac.jp
Notes	estimated completion date: March 2017

Differences between protocol and review

The results for mortality have been pooled by using the risk difference as the summary measure instead of the odds ratio. This decision was made due to the well‐known computational problems in meta‐analyses concerning trials with zero events in both groups.

Clinically relevant delayed gastric emptying (ISGPS grade B/C) was included as a secondary outcome to assess potential differences between the two procedures.

The planned subgroup analyses for patients undergoing classical Whipple's procedure, single‐loop reconstruction, Roux‐en‐Y reconstruction and for underlying diseases could not be conducted, because sufficient data were not available.

Contributions of authors

Conceiving the review: FJH*, RK*
 Designing the review: FJH*, RK*, MKD
 Co‐ordinating the review: MKD, MWB
 Designing search strategies: FJH*, RK*, MKD
 Writing the review: FJH*, RK*, MKD
 Providing general advice on the review: MKD, AU, MWB
 Performing previous work that was the foundation of the current study: FJH*, MKD, AU, MWB

* Both authors contributed equally to this work.

Sources of support

Internal sources

• Department of General, Visceral, and Transplantation Surgery, University of Heidelberg, Germany.

  General support.

External sources

• No sources of support supplied

Declarations of interest

FH: none known.
 RK: none known.
 MKD: none known.
 MWB: none known.
 AU: none known.

New