Efficiency in legal transplants: An essay in Comparative Law and Economics

Systematic review

Meta-analysis of delayed gastric emptying after pylorus-preserving versus pylorus-resecting pancreatoduodenectomy U. Klaiber , P. Probst , O. Strobel , C. W. Michalski, C. Dörr-Harim, M. K. Diener, M. W. Büchler and T. Hackert Department of General, Visceral and Transplantation Surgery, University of Heidelberg, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany Correspondence to: Professor T. Hackert (e-mail: [email protected])

Background: Delayed gastric emptying (DGE) is a frequent complication after pylorus-preserving

pancreatoduodenectomy. Recent studies have suggested that resection of the pylorus is associated with decreased rates of DGE. However, superiority of pylorus-resecting pancreatoduodenectomy was not shown in a recent RCT. This meta-analysis summarized evidence of the effectiveness and safety of pylorus-preserving compared with pylorus-resecting pancreatoduodenectomy. Methods: RCTs and non-randomized studies comparing outcomes of pylorus-preserving and pylorus-resecting pancreatoduodenectomy were searched systematically in MEDLINE, Web of Science and CENTRAL. Random-effects meta-analyses were performed and the results presented as weighted odds ratios (ORs) or mean differences with their corresponding 95 per cent confidence intervals. Subgroup analyses were performed to account for interstudy heterogeneity between RCTs and non-randomized studies. Results: Three RCTs and eight non-randomized studies with a total of 992 patients were included. Quantitative synthesis across all studies showed superiority for pylorus-resecting pancreatoduodenectomy regarding DGE (OR 2⋅71, 95 per cent c.i. 1⋅48 to 4⋅96; P = 0⋅001) and length of hospital stay (mean difference 3⋅26 (95 per cent c.i. −1⋅04 to 5⋅48) days; P = 0⋅004). Subgroup analyses including only RCTs showed no significant statistical differences between the two procedures regarding DGE, and for all other effectiveness and safety measures. Conclusion: Pylorus-resecting pancreatoduodenectomy is not superior to pylorus-preserving pancreatoduodenectomy for reducing DGE or other relevant complications. Presented to the World Pancreas Forum 2017, Berne, Switzerland, June 2017, and the annual conference of the German Society of General and Visceral Surgery, Dresden, Germany, September 2017 Paper accepted 30 October 2017 Published online in Wiley Online Library (www.bjs.co.uk). DOI: 10.1002/bjs.10771

Introduction

Since its introduction by Traverso and Longmire1 in 1978, pylorus-preserving partial pancreatoduodenectomy has become the standard treatment for benign and malignant pathologies of the pancreatic head and periampullary region. In contrast to the classical Kausch–Whipple procedure including distal gastrectomy, the entire stomach is preserved in pylorus-preserving pancreatoduodenectomy with the aim of preventing complications after gastrectomy such as dumping syndrome or jejunogastric reflux, and to improve postoperative digestion without constraining lymph node clearance. The justification for © 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

pylorus preservation has been confirmed in meta-analyses of RCTs2,3 , which showed that pylorus-preserving pancreatoduodenectomy and the Whipple procedure are equally effective with regard to mortality, morbidity and survival. However, delayed gastric emptying (DGE) occurs in up to 61 per cent of patients, and is still the most frequent complication after either procedure4 . DGE impairs patients’ well-being, prolongs duration of hospital stay and postpones adjuvant chemotherapy. With the aim of combining the benefits of pylorus-preserving pancreatoduodenectomy and the Whipple procedure, pylorus-resecting pancreatoduodenectomy (synonymously referred to as subtotal stomach-preserving pancreatoduodenectomy) BJS

U. Klaiber, P. Probst, O. Strobel, C. W. Michalski, C. Dörr-Harim, M. K. Diener et al.

was developed in the late 1990s. Since then, several studies have been performed to compare pylorus-preserving and pylorus-resecting pancreatoduodenectomy. However, data remain inconclusive regarding which procedure has a lower risk of DGE5 . In contrast to a previous RCT6 and existing meta-analyses5,7 – 10 , a recent blinded RCT11 including 188 patients did not show a significant difference between pylorus-preserving and pylorus-resecting pancreatoduodenectomy. This large RCT was designed and planned with the aim to overcome the limitations of previous studies with non-blinded trial designs and missing adherence to current standards in DGE definition according to the International Study Group of Pancreatic Surgery (ISGPS), and DGE treatment. As these data substantially increase the existing body of evidence, the aim of this systematic review and meta-analysis was to provide an update of the available evidence for the effectiveness and safety of pylorus-preserving versus pylorus-resecting pancreatoduodenectomy. Methods

This systematic review and meta-analysis is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement12 .

5 July 2017. In MEDLINE, the related citations function was used to search for additional relevant studies. Additionally, a hand search of the reference lists of relevant articles and related systematic reviews was performed. Furthermore, the clinical trial registries ClinicalTrials.gov, UMIN Clinical Trials Registry and the German Clinical Trials Register were searched to identify ongoing and unpublished trials. Register data were checked against the data published in the final publication. For each study included for review, the publication of the study protocol was sought and checked, if available.

Study selection Two independent investigators reviewed all records identified by the search. Trials meeting the following inclusion criteria were included for review: at least two intervention groups comparing postoperative outcomes, including DGE, of patients undergoing pylorus-preserving and pylorus-resecting pancreatoduodenectomy. There were no restrictions regarding patients’ age or underlying disease. Non-comparative studies were excluded. If the abstract suggested relevance, the full article was assessed for eligibility. Any disagreements between the two reviewers were discussed within the working group.

Data extraction Data sources and search strategy A systematic literature search was performed according to the recommendations of the Cochrane Collaboration13 . Searches were conducted to identify all published and unpublished RCTs or non-randomized studies referring to the effectiveness and safety of pylorus-preserving versus pylorus-resecting pancreatoduodenectomy. As pylorus-resecting pancreatoduodenectomy is not yet widely used, and because of the methodological challenges of RCTs of surgical interventions14 , a small number of RCTs involving only a few patients was anticipated. Thus, non-randomized studies were also eligible for inclusion. The following electronic bibliographical databases with relevance for surgical literature were surveyed: Cochrane Central Register of Controlled Trials (CENTRAL) from the Cochrane Library, MEDLINE (via PubMed) and Web of Science All Databases. The search strategy for MEDLINE, using a combination of medical subject headings (MeSH) terms and free text words, is shown in Appendix S1 (supporting information). The search strategy was adapted for each of the databases. The search was not restricted with regard to time period or language. The last electronic search was performed on © 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

For data extraction, specific forms were constructed and pilot-tested using representative studies. After finalization of the forms, data from included studies were extracted by two independent investigators. Any discrepancies were discussed to reach a consensus for each study. The following items were extracted: title of study, authors’ names and institutions, year of publication, journal, durations of recruitment and follow-up, study design and sample size. The baseline data extracted were age, sex and BMI of participants, previous diabetes mellitus, underlying disease, surgical procedures and perioperative management. In addition, all relevant outcomes parameters were extracted.

Outcomes The main outcome was the overall rate of DGE. Further outcomes were DGE severity as defined by the ISGPS, clinically relevant postoperative pancreatic fistula according to the currently updated definition of the ISGPS (grade B/C), postpancreatectomy haemorrhage, intra-abdominal fluid collection or abscess, bile leakage, leakage of duodenoenteric or gastroenteric anastomosis, lymphatic fistula, wound infection, pulmonary complications, mortality, www.bjs.co.uk

BJS

Delayed gastric emptying after pancreatoduodenectomy

reoperation, perioperative blood loss, duration of operation, length of hospital stay and quality of life. For each outcome, meta-analysis was performed when there were data from at least two studies.

differences in clinical variability arising from different surgical procedures. Funnel plots were created to evaluate the risk of publication bias19 . Results

Assessment of methodological quality of included studies The methodological quality of included RCTs and non-randomized studies was assessed by means of two appropriate critical appraisal tools. For RCTs, the Cochrane Collaboration tool for assessing risk of bias was used to evaluate random sequence generation, allocation concealment, blinding of participants, personnel and outcome assessment, incomplete outcome data, selective reporting and other sources of bias. For each domain, low, unclear or high risk of bias was assessed according to the Cochrane Handbook for Systematic Reviews of Interventions13 . To evaluate other sources of bias, register data, published study protocols, conflict of interest statements and sources of funding were assessed, if available15 . For non-randomized studies, ROBINS-I (Risk Of Bias In Non-randomized Studies – of Interventions) was used to evaluate bias resulting from confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes and selective reporting16 . Risk of bias was judged to be low, moderate, serious, critical or not assessable (no information) within each domain, and an overall risk-of-bias judgement was assessed for each study.

Statistical analysis Statistical analysis was performed with Review Manager (RevMan) version 5.3.3 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark). All relevant outcome parameters from the included studies were assessed for estimation of treatment effects. Dichotomous data were pooled to estimate odds ratios (ORs) and associated 95 per cent confidence intervals. For continuous data, mean differences and associated 95 per cent c.i. were reported. P < 0⋅050 was considered statistically significant. Heterogeneity among the studies was evaluated by the I 2 statistic, with 0–30 per cent representing no or low, 30–60 per cent moderate, 50–90 per cent substantial and 75–100 per cent considerable heterogeneity. For meta-analysis, a random-effects model was used to account for interstudy clinical diversity, such as variability in perioperative management17,18 . Subgroup analyses were conducted to allow for interstudy heterogeneity owing to the study design (RCTs versus non-randomized studies). Sensitivity analyses were performed to account for substantial © 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

Study selection The search yielded a total of 3976 records (Fig. 1). After screening the corresponding abstracts, 3963 records were excluded and full texts of the remaining 13 records were assessed for eligibility. Two of these 13 articles were excluded because they did not fulfil the inclusion criteria. Finally, three RCTs6,11,20 , two from Japan and one from Germany, involving a total of 418 patients, and eight non-randomized studies10,21 – 27 , six from Japan, one from China and one from Germany, involving 574 patients, were included in qualitative and quantitative syntheses.

Trial characteristics and study population Characteristics of the 11 studies included in qualitative and quantitative analyses are presented in Table 1. All trial results were published between 2007 and 2017. All study designs were single-centre, with sample sizes ranging from 33 to 112 patients in non-randomized studies and 100 to 188 patients in RCTs. Sample size calculation was based on a superiority trial design in all RCTs with DGE as primary endpoint. In non-randomized studies, no formal sample size calculation was performed but DGE was assessed in all studies. Two non-randomized studies were prospective21,25 , and the remainder were retrospective. All RCTs were registered in clinical trials registries, but only one group11 had registered the trial prospectively and published the corresponding study protocol29 . The two other RCTs6,20 had been registered retrospectively without publication of the study protocol. Across all studies, indications for surgery were benign and malignant diseases of the pancreatic head and periampullary region. Pancreatic cancer was found in one-quarter to one-half of the study participants (Table 2). Other main indications were intraductal papillary mucinous neoplasia and chronic pancreatitis. In some non-randomized studies, there were substantial intrastudy differences regarding tumour entities, with more patients suffering from pancreatic carcinoma in the pylorus-resecting group22,24,25 . As anticipated, baseline characteristics of patients in the two study groups were more closely comparable in RCTs than in non-randomized studies. As well as intrastudy imbalances in non-randomized studies, interstudy heterogeneity owing to different inclusion and exclusion criteria www.bjs.co.uk

BJS

Identification

U. Klaiber, P. Probst, O. Strobel, C. W. Michalski, C. Dörr-Harim, M. K. Diener et al.

Records identified through database searching n =5167

Additional records identified through other sources n=1

Screening

Records screened after duplicates removed n=3976

Records excluded n= 3963 Full-text articles assessed for eligibility n =13

Eligibility

Full-text articles excluded n=2 Definition of outcomes n=1 Long-term data of included RCT n=1

Included

Studies included in qualitative synthesis n=11

Fig. 1

Studies included in quantitative synthesis (meta-analysis) RCTs n=3 NRS n=8

PRISMA flow diagram showing selection of articles for review. NRS, non-randomized studies

was evident; for example, Matsumoto and colleagues20 excluded patients with suspected pancreatic cancer from study participation. Distribution of patients’ age, sex and BMI was comparable between the two study groups, although only two RCTs11,20 and two non-randomized studies21,23 provided data on BMI. Surgical procedures regarding pylorus-preserving and pylorus-resecting pancreatoduodenectomy were comparable between the studies except for the non-randomized study by Oida and co-workers27 , who performed a modified pylorus-resecting pancreatoduodenectomy with re-resection of the divided stomach depending on the angle from the choledochojejunostomy to the gastrojejunostomy. With respect to lymph node dissection and vascular resections, data were provided in six6,20,22 – 25 and six6,10,22 – 25 studies respectively. Kurahara et al.24 reported resection of additional visceral organs due to tumour infiltration, whereas Matsumoto and co-workers20 excluded patients having simultaneous liver resection. Regarding reconstruction procedures, pancreatojejunostomy was performed in all three RCTs6,11,20 and in six non-randomized studies10,21 – 23,25,26 , and pancreaticogastrostomy was used in two non-randomized

studies24,27 . The pancreatic anastomosis was performed under stenting or drainage of the pancreatic duct in four studies6,22,24,27 . With respect to the duodenoenteric or gastroenteric reconstruction route, antecolic duodenoenterostomy or gastroenterostomy was performed in seven studies6,10,11,21 – 23,26 . In contrast, a retrocolic route was reported by four groups20,24,25,27 . Furthermore, an additional Braun anastomosis was described in two non-randomized studies25,26 . Postoperative management, including nasogastric tube removal, start of oral intake and administration of prokinetic agents, differed considerably in individual studies. A definition of the primary endpoint DGE was given by all authors. The ISGPS definition30 was used in all RCTs and six non-randomized studies, whereas two groups25,26 adhered to the definition of van Berge Henegouwen et al.28 . Overall, considerable clinical heterogeneity, particularly differences in surgical procedures and perioperative management, had to be taken into consideration.

© 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

www.bjs.co.uk

Methodological quality of included studies Risk of bias in included RCTs, as assessed by the Cochrane Collaboration risk-of-bias tool, was low for the domains BJS

Delayed gastric emptying after pancreatoduodenectomy

Comparison and characteristics of RCTs and non-randomized studies comparing pylorus-preserving and pylorus-resecting pancreatoduodenectomy

Table 1

Reference

Study type

Sample size

Resection and reconstruction

Selected outcomes

Definition of DGE

11

RCT, single-centre

188

ppPD versus prPD PJ, antecolic DE/GE

DGE, CR-POPF, FC, mortality, blood loss, LOS: n.s.

ISGPS

20

RCT, single-centre

100

ppPD versus prPD PJ, retrocolic DE/GE

DGE, CR-POPF, FC, mortality, blood loss, LOS: n.s.

ISGPS

6

RCT, single-centre

130

ppPD versus prPD PJ, antecolic DE/GE

ISGPS

10

Retrospective NRS, single-centre

74

ppPD versus prPD

DGE: prPD+ CR-POPF, FC, mortality, blood loss, LOS: n.s. DGE, LOS: prPD+

ISGPS

Prospective NRS, single-centre

80

CR-POPF, FC, mortality, blood loss: n.s. DGE: prPD+

ISGPS

Retrospective NRS, single-centre

55

CR-POPF, FC: n.a. Mortality, blood loss, LOS: n.s. DGE: prPD+

ISGPS

Retrospective NRS, single-centre

89

Blood loss: ppPD+ CR-POPF: n.a. FC, mortality, LOS: n.s. DGE: prPD+

ISGPS

Retrospective NRS, single-centre

67

CR-POPF, mortality, blood loss, LOS: n.s. FC: n.a. DGE, LOS: prPD+

ISGPS

Retrospective NRS, single-centre

112

Mortality, blood loss: n.s. CR-POPF, FC: n.a. Blood loss: ppPD+

ISGPS

Prospective NRS, single-centre

64

Retrospective NRS, single-centre

33

21

22

23

27

24

25

26

PJ, antecolic DE/GE ppPD versus prPD PJ, antecolic DE/GE

ppPD versus prPD PJ, antecolic DE/GE

ppPD versus prPD PJ, antecolic DE/GE

ppPD versus Modified prPD PG, retrocolic DE/GE ppPD versus prPD PG, retrocolic DE/GE

ppPD versus prPD PJ, retrocolic DE/GE

ppPD versus prPD PJ, antecolic DE/GE

DGE, CR-POPF, FC, LOS: n.s. Mortality: n.a. Blood loss: ppPD+ DGE, FC, mortality, LOS: n.s. CR-POPF: n.a. DGE: prPD+ Mortality, blood loss, LOS: n.s. CR-POPF, FC: n.a.

Van Berge Henegouwen et al.28 Van Berge Henegouwen et al.28

DGE, delayed gastric emptying; ppPD, pylorus-preserving pancreatoduodenectomy; prPD, pylorus-resecting pancreatoduodenectomy; PJ, pancreatojejunostomy; CR-POPF, clinically relevant postoperative pancreatic fistula; FC, intra-abdominal fluid collection/abscess; LOS, length of hospital stay; n.s., no significant difference; ISGPS, International Study Group of Pancreatic Surgery; DE, duodenoenterostomy; GE, gastroenterostomy; +, superiority; NRS, non-randomized study; n.a., not assessed; PG, pancreaticogastrostomy.

random sequence generation, allocation concealment, incomplete outcome data, selective reporting and other sources of bias. Regarding blinding of participants, personnel and outcome assessment, risk of bias was low in one RCT11 (except for the surgeon), but high in the other two RCTs6,20 owing to non-blinded study designs (Fig. S1, supporting information). In comparison, overall risk of bias according to ROBINS-I was moderate in five non-randomized studies10,21,23 – 25 and serious in three22,26,27 (Table S1, supporting information). Owing primarily to the absence of methods to minimize confounding

and selection bias, none of the non-randomized studies was comparable to a well performed randomized trial. As a consequence, subgroup analyses were performed to allow for the differences in methodological quality between RCTs and non-randomized studies. Test for subgroup differences was significant for the outcomes DGE grades B/C and length of hospital stay. Regarding overall DGE rate, test for subgroup differences did not reach statistically significant difference; meta-analysis including only RCTs showed no significant difference between pylorus-preserving and pylorus-resecting pancreatoduodenectomy, whereas

© 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

www.bjs.co.uk

BJS

U. Klaiber, P. Probst, O. Strobel, C. W. Michalski, C. Dörr-Harim, M. K. Diener et al.

Table 2

Characteristics of the study populations RCTs

Mean age (years) Sex ratio (F : M) Mean BMI (kg/m2 )* Preoperative diabetes mellitus Histopathological diagnosis Pancreatic carcinoma IPMN Pancreatitis

Non-randomized studies

Total

ppPD (n = 209)

prPD (n = 209)

ppPD (n = 257)

prPD (n = 317)

ppPD (n = 466)

prPD (n = 526)

65 92 : 117 25 56 (26⋅8)

66 87 : 122 24 50 (23⋅9)

65 101 : 156 23 41 of 138† (29⋅7)

65 135 : 182 23 56 of 170† (32⋅9)

65 193 : 273 24 97 of 347 (28⋅0)

65 222 : 304 23 106 of 379 (28⋅0)

55 (26⋅3) 37 of 159‡ (23⋅3) 20 of 159‡ (12⋅6)

69 (33⋅0) 29 of 159‡ (18⋅2) 25 of 159‡ (15⋅7)

104 (40⋅5) 33 of 195§ (16⋅9) 7 of 195§ (3⋅6)

198 (62⋅5) 20 of 238§ (8⋅4) 7 of 238§ (2⋅9)

159 (34⋅1) 70 of 354 (20⋅8) 27 of 354 (7⋅6)

267 (50⋅8) 49 of 397 (12⋅3) 32 of 397 (8⋅1)

Values in parentheses are percentages. Data from *four studies11,20,21,23 , missing data in †four studies21,23,25,26 , ‡one study20 and §two studies10,27 . ppPD, pylorus-preserving pancreatoduodenectomy; prPD, pylorus-resecting pancreatoduodenectomy; IPMN, intraductal papillary mucinous neoplasia. Table 3

Summary statistics of meta-analyses comparing outcomes of pylorus-preserving and pylorus-resecting pancreatoduodenectomy Subgroup of RCTs only

Delayed gastric emptying Overall Grade A Grades B–C Clinically relevant postoperative pancreatic fistula Postpancreatectomy haemorrhage Intra-abdominal fluid collection or abscess Bile leakage Wound infection Pulmonary complication Mortality Reoperation Duration of operation (min) Perioperative blood loss (ml) Length of hospital stay (days)

All studies

Odds ratio

I2 (%)

P

Odds ratio

I2 (%)

P

1⋅60 (0⋅57, 4⋅47) 1⋅28 (0⋅29, 5⋅64) 1⋅30 (0⋅42, 4⋅02) 1⋅44 (0⋅83, 2⋅49) 0⋅90 (0⋅38, 2⋅13) 1⋅16 (0⋅64, 2⋅11) 0⋅13 (0⋅01, 2⋅67) 1⋅32 (0⋅56, 3⋅09) 0⋅60 (0⋅25, 1⋅46) 1⋅04 (0⋅21, 5⋅04) 1⋅30 (0⋅56, 3⋅01) −6⋅08 (−21⋅83, 9⋅66)* −145⋅02 (−450⋅87, 160⋅83)* 0⋅92 (−1⋅11, 2⋅96)*

69 47 61 0 0 9 n.a. 0 0 0 0 0 71 0

0⋅37 0⋅75 0⋅64 0⋅19 0⋅81 0⋅63 0⋅19 0⋅53 0⋅26 0⋅96 0⋅54 0⋅45 0⋅35 0⋅37

2⋅71 (1⋅48, 4⋅96) 0⋅58 (0⋅17, 1⋅95) 5⋅45 (1⋅78, 16⋅70) 1⋅19 (0⋅74, 1⋅89) 0⋅86 (0⋅41, 1⋅78) 1⋅23 (0⋅78, 1⋅92) 0⋅60 (0⋅03, 13⋅16) 0⋅98 (0⋅52, 1⋅84) 0⋅64 (0⋅29, 1⋅43) 0⋅74 (0⋅23, 2⋅41) 1⋅06 (0⋅51, 2⋅17) +3⋅11 (−15⋅30, 21⋅52)* −92⋅01 (−198⋅56, 14⋅55)* 3⋅26 (−1⋅04, 5⋅48)*

63 83 82 0 0 0 49 0 0 0 0 73 67 68

0⋅001 0⋅38 0⋅003 0⋅47 0⋅68 0⋅37 0⋅75 0⋅94 0⋅28 0⋅62 0⋅88 0⋅74 0⋅09 0⋅004

Values in parentheses are 95 per cent confidence intervals; *values are mean differences. n.a., Not applicable.

meta-analysis including both RCTs and non-randomized studies showed superiority for pylorus-resecting pancreatoduodenectomy.

Quantitative analysis of included studies Table 3 summarizes the statistical findings of meta-analyses comparing outcomes of pylorus-preserving and pylorusresecting pancreatoduodenectomy.

Delayed gastric emptying All studies investigated the occurrence of DGE, although two groups25,26 did not adhere to the ISGPS definition30 and thus did not provide data for DGE grades A, B and C. Regarding the overall rate of DGE, meta-analysis including both RCTs and non-randomized studies showed superiority for pylorus-resecting versus pylorus-preserving pancreatoduodenectomy (OR 2⋅71, 95 per cent c.i. 1⋅48 to 4⋅96; P = 0⋅001), with substantial statistical heterogeneity. © 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

However, there was no significant difference in subgroup analysis of RCTs (OR 1⋅60, 0⋅57 to 4⋅47; P = 0⋅37) (Fig. 2). With respect to DGE grade A, there was no significant difference between the two study groups in meta-analysis across all studies or in subgroup analysis (Table 3). In contrast, rates of DGE grade B/C were similar in meta-analysis of RCTs (OR 1⋅30, 0⋅42 to 4⋅02; P = 0⋅64), but not in meta-analysis including both RCTs and non-randomized studies, which showed superiority for the pylorus-resecting procedure (OR 5⋅45, 1⋅78 to 16⋅70; P = 0⋅003). However, there was considerable significant heterogeneity.

Morbidity and mortality Regarding clinically relevant postoperative pancreatic fistula (grade B/C), which was reported in all RCTs and three non-randomized studies10,23,24 , no significant difference between pylorus-preserving and pylorus-resecting pancreatoduodenectomy was found in either subgroup analysis including only RCTs (OR 1⋅44, 95 per cent c.i. 0⋅83 to www.bjs.co.uk

BJS

Delayed gastric emptying after pancreatoduodenectomy

DGE ppPD

prPD

RCT Hackert et al.11

24 of 95

29 of 93

14·5

0·75 (0·39, 1·41)

Kawai et al.6

11 of 64

3 of 66

9·4

4·36 (1·16, 16·44)

10 of 50

6 of 50

11·0

1·83 (0·61, 5·50)

45 of 209

38 of 209

34·9

1·60 (0·57, 4·47)

Reference

Matsumoto et al.20 Subtotal

Odds ratio

Weight (%)

Odds ratio

Heterogeneity: τ2 = 0·56; χ2 = 6·35, 2 d.f., P = 0·04; I2 = 69% Test for overall effect: Z = 0·89, P = 0·37 NRS Akizuki et al.25

3 of 34

3 of 30

7·4

0·87 (0·16, 4·68)

Fujii et al.23

9 of 33

3 of 56

9·0

6·63 (1·65, 26·67)

17 of 40

6 of 40

11·2

4·19 (1·44, 12·22)

6 of 12

3 of 21

7·5

6·00 (1·13, 31·73)

Kurahara et al.24

31 of 48

35 of 64

13·5

1·51 (0·70, 3·26)

Nanashima et al.22

27 of 28

16 of 27

5·5

18·56 (2·19, 157·53)

Oida et al.27

25 of 25

42 of 42

Zhou et al.10

17 of 37

6 of 37

11·1

4·39 (1·48, 13·03)

135 of 257

114 of 317

65·1

3·48 (1·82, 6·67)

Hackert et al.21 Hayashibe et al.26

Subtotal

Not estimable

Heterogeneity: τ2 = 0·32; χ2 = 10·72, 6 d.f., P = 0·10; I2 = 44% Test for overall effect: Z = 3·77, P < 0·001 Total

180 of 466

152 of 526

100·0

2·71 (1·48, 4·96)

Heterogeneity: τ2 = 0·55; χ2 = 24·61, 9 d.f., P = 0·003; I2 = 63% Test for overall effect: Z = 3·23, P = 0·001 Test for subgroup differences: χ2 = 1·58, 1 d.f., P = 0·21; I2 = 36·6%

0·01

0·1 Favours ppPD

1

10

100

Favours prPD

Forest plot comparing delayed gastric emptying (DGE) in pylorus-preserving (ppPD) and pylorus-resecting (prPD) pancreatoduodenectomy. A Mantel–Haenszel random-effects model was used for meta-analysis. Odds ratios are shown with 95 per cent confidence intervals. NRS, non-randomized study

Fig. 2

2⋅49; P = 0⋅19) or meta-analysis including both RCTs and non-randomized studies (OR 1⋅19, 0⋅74 to 1⋅89; P = 0⋅47). There was no significant heterogeneity (Table 3). The occurrence of haemorrhage after pancreatectomy and intra-abdominal fluid collection/abscess was assessed in all three RCTs and four non-randomized studies10,22,24,25 . There was no statistically significant difference between the two resection types for either of these outcomes in meta-analyses of RCTs (OR 0⋅90, 95 per cent c.i. 0⋅38 to 2⋅13, P = 0⋅81, and OR 1⋅16, 0⋅64 to 2⋅11, P = 0⋅63, respectively) and of all studies (OR 0⋅86, 0⋅41 to 1⋅78, P = 0⋅68, and OR 1⋅23, 0⋅78 to 1⋅92, P = 0⋅37, respectively). Heterogeneity was very low. Bile leakage was reported in only one RCT20 and one non-randomized study10 . As shown in the individual studies, meta-analysis yielded similar rates in both study groups (OR 0⋅60, 95 per cent c.i. 0⋅03 to 13⋅16; P = 0⋅75) with moderate heterogeneity. Leakage of the duodenoenteric or gastroenteric anastomosis and lymphatic fistula were reported in only one study each11,25 , so these two outcomes were excluded from meta-analysis.

Rates of wound infection, given in all RCTs and four non-randomized studies10,25 – 27 , and pulmonary complications, stated in two RCTs6,11 and two non-randomized studies25,27 , were comparable in meta-analysis including only RCTs and in meta-analysis across all studies, with no heterogeneity (Table 3). Data on mortality were reported in all but one study24 ; five groups10,20,23,26,27 reported no deaths. There was no significant difference between the two procedures in meta-analysis of either RCTs or all studies (Table 3); homogeneity was given in both analyses. Regarding reoperations, reported in all RCTs and four non-randomized studies10,21,22,25 , the two procedures yielded similar results in quantitative analysis including only RCTs and across all studies (Table 3). Statistical homogeneity was given for this outcome.

© 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

www.bjs.co.uk

Perioperative data and length of hospital stay Duration of surgery, perioperative blood loss and length of hospital stay were stated in all studies. However, the non-randomized study by Kurahara and colleagues24 was BJS

U. Klaiber, P. Probst, O. Strobel, C. W. Michalski, C. Dörr-Harim, M. K. Diener et al.

Publication bias

s.e. (log[odds ratio])

0

A funnel plot representative for the endpoint DGE is presented in Fig. 3. With only three RCTs, it is not possible to draw a conclusion regarding publication bias. However, for non-randomized studies, a possible publication bias was present, as almost all studies found an effect favouring pylorus-resecting pancreatoduodenectomy.

0·5

1·0

1·5

2·0 0·01

RCT Non-randomized study 0·1

1 Odds ratio

Discussion 10

100

Funnel plot for the endpoint delayed gastric emptying (DGE) in RCTs and non-randomized studies. Only six non-randomized studies are represented; that by Oida et al.27 is not shown because DGE was observed in all patients

Fig. 3

excluded from meta-analyses of continuous data as the data were insufficient. In addition, data on length of hospital stay from Akizuki et al.25 were excluded as patients with a prolonged stay because of postoperative chemotherapy were incorporated in these data. Regarding duration of operation and perioperative blood loss, there was no statistically significant difference between the two study groups (Table 3). I 2 values ranged from 0 per cent in meta-analysis of RCTs to 73 per cent in meta-analyses of both RCTs and non-randomized studies. Although length of hospital stay was significantly longer in the pylorus-preserving group in meta-analysis across all studies (mean difference 3⋅26 (95 per cent c.i. −1⋅04 to 5⋅48) days; P = 0⋅004), significance was not found in meta-analysis including only RCTs (mean difference 0⋅92 (−1⋅11 to 2⋅96) days; P = 0⋅37). Heterogeneity was substantial in meta-analysis across all studies, but absent in subgroup analysis of RCTs. Quality of life was assessed in two RCTs6,11 , both of which showed no significant difference in patients following pylorus-preserving versus pylorus-resecting pancreatoduodenectomy. Meta-analysis of data on quality of life was not performed because different questionnaires had been used. For different reconstruction techniques (antecolic versus retrocolic reconstruction, pancreaticojejunostomy versus pancreaticogastrostomy, exclusion of the Oida modification27 ), sensitivity analyses did not substantially alter any of the quantitative results. Exclusion of the RCT by Matsumoto and co-workers20 , which did not include patients with pancreatic cancer, did not affect the results significantly. © 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

This systematic review and meta-analysis provides an update of critically appraised and quantitative data on the effectiveness and safety of pylorus-preserving pancreatoduodenectomy compared with pylorus-resecting pancreatoduodenectomy. Quantitative synthesis across all included studies showed superiority for pylorus-resecting pancreatoduodenectomy regarding DGE and length of hospital stay. In contrast, subgroup analysis including only RCTs found no significant difference between the two procedures for either of the analysed outcomes. The methodological quality of RCTs was much better than that of non-randomized studies, which showed relevant sources of bias leading to overestimation of treatment effects. Thus, this meta-analysis of RCTs provides updated level I evidence that pylorus resection in partial pancreatoduodenectomy is not superior to pylorus-preserving pancreatoduodenectomy regarding DGE rate and other relevant outcomes. on outcomes of Since the first report26 pylorus-preserving compared with pylorus-resecting pancreatoduodenectomy, several non-randomized and randomized trials on this topic have been published. In 2014 and 2015, five meta-analyses5,7 – 10 summarizing the existing evidence were published; significantly reduced DGE rates were found in the pylorus-resecting group in all of these studies. This not only illustrates the high relevance of the topic but also reflects its controversy. Owing to limitations of the studies included in these meta-analyses, all authors concluded that further high-quality RCTs were needed to validate whether pylorus resection does reduce the rate of DGE. As additional results from a large blinded RCT are now available11 , the present updated meta-analysis was performed to summarize the existing evidence. In contrast to the RCT by Kawai and colleagues6 and the earlier meta-analyses5,7 – 10 , the latest RCT11 including 188 patients failed to show superiority of pylorus-resecting pancreatoduodenectomy regarding DGE and other relevant outcome parameters. The inclusion of these data in meta-analysis of existing RCTs yielded similar DGE rates in patients undergoing pylorus-preserving and www.bjs.co.uk

BJS

Delayed gastric emptying after pancreatoduodenectomy

pylorus-resecting pancreatoduodenectomy. Moreover, regarding other outcomes, the two resection procedures are also comparable. In contrast, meta-analyses across all studies (RCTs and non-randomized studies) showed superiority for pylorus-resecting pancreatoduodenectomy regarding DGE and length of hospital stay. As mentioned in previous meta-analyses including both RCTs and non-randomized studies7,9,10 , the methodological quality of the non-randomized clinical trials was poor and results of quantitative synthesis across all study designs must be looked at critically. Bearing a considerable risk of bias, these studies are prone to overestimate treatment effects, and the inclusion of non-randomized studies in meta-analysis can distort the results towards a statistically significant difference. Consequently, considering the differences in methodological quality of RCTs versus non-randomized studies, the results of the subgroup analyses including only RCTs are regarded as methodologically valid. The findings of the present meta-analysis of RCTs strengthen the data from the latest RCT11 and question evidence from previous meta-analyses. As long-term data on pylorus-preserving versus pylorus-resecting pancreatoduodenectomy are sparse, the risk of potential harm of pylorus resection, such as jejunogastric reflux and consecutive gastric ulcers or even cancer as a long-term complication, remains unclear. To date, there are only two RCTs20,31 providing long-term outcomes following pylorus-preserving and pylorus-resecting pancreatoduodenectomy. Both studies showed that long-term nutritional and diabetic status were comparable between the two procedures. Kawai et al.31 also reported similar quality of life after both procedures, and observed an increased incidence of peptic ulceration following pylorus resection, although this was not statistically significant. This trend should be addressed in further studies. Thus, based on the results of meta-analyses of RCTs showing that pylorus-resecting pancreatoduodenectomy is not superior to pylorus-preserving pancreatoduodenectomy, it may be concluded that, in general, pylorus-preserving pancreatoduodenectomy should remain the procedure of choice. However, whenever technically or oncologically required, pylorus resection represents the preferred alternative. The present meta-analysis has several limitations. Notably, there was considerable clinical variability in the studies, especially regarding operative procedures and perioperative management. In four studies6,22,24,27 the pancreatic anastomosis was performed under stenting or drainage of the pancreatic duct, but this was not done in the other studies. Regarding further differences in reconstruction techniques, sensitivity analyses showed that the

method of duodenoenteric or gastroenteric reconstruction (antecolic versus retrocolic) and pancreaticojejunostomy versus pancreaticogastrostomy did not affect the results substantially. This is in line with current data from well performed studies4,32,33 . With regard to perioperative and postoperative management, Fujii and co-workers23 reported that oral intake was started routinely on postoperative day 7, which is very late and not according to today’s standard34 . Akizuki et al.25 routinely placed jejunal tubes during surgery for postoperative commencement of enteral feeding, which differed from the peroral diet reported in other studies. Moreover, nasogastric tube management differed substantially between the studies. In most trials, nasogastric tubes were not removed before drainage was below 200–500 ml per day. In contrast, in the latest RCT11 nasogastric tubes were routinely removed at the end of the operation. These differences in perioperative and postoperative management may have great impact on DGE evaluation, even though the ISGPS definition was adhered to in all RCTs and in all but two non-randomized studies25,26 . A second shortcoming of this meta-analysis results from the above-mentioned limitations in the methodological quality of included studies. Methodological quality was poor in non-randomized studies, but the overall risk of bias was low in the three RCTs, and performance and detection bias must be considered in the two non-blinded RCTs6,20 . This is of importance as DGE cannot be regarded as an objective outcome parameter and, thus, its assessment may be prone to bias within an open-label study design. With eight studies included in this meta-analysis coming from Japan, the generalizability of the present results is limited. This bias may be explained by the origin of pylorus-resecting pancreatoduodenectomy in Japan associated with specific interest in scientific investigation of the method. It also suggests that surgeons from other countries such as the USA might not support pylorus-resecting pancreatoduodenectomy. The extent to which potential differences in patients’ constitution and BMI between Eastern and Western countries may play a role in differences in DGE rates cannot yet be evaluated. Non-blinded study designs and the considerable diversity in perioperative and postoperative management, among other possible reasons, may explain the variability in DGE rates between the studies. Kawai and colleagues6 reported a DGE incidence of 17⋅2 per cent in the pylorus-preserving group, compared with 4⋅5 per cent in the pylorus-resecting group. In the RCT by Matsumoto and co-workers20 , DGE rates were 20 per cent in patients undergoing pylorus-preserving pancreatoduodenectomy

© 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

www.bjs.co.uk

BJS

U. Klaiber, P. Probst, O. Strobel, C. W. Michalski, C. Dörr-Harim, M. K. Diener et al.

and 12 per cent in those having a pylorus-resecting procedure. In contrast, in the only blinded RCT, by Hackert et al.11 , overall DGE rates were higher and balanced between the two study groups (25⋅3 versus 31⋅2 per cent respectively). These data indicate a possible impact of detection bias, and all efforts should thus be taken to minimize bias in RCTs of surgical interventions, even though blinding of the surgeon is not possible35 . Disclosure

9

10

11

The authors declare no conflict of interest. References 1 Traverso LW, Longmire WP Jr. Preservation of the pylorus in pancreaticoduodenectomy. Surg Gynecol Obstet 1978; 146: 959–962. 2 Diener MK, Knaebel HP, Heukaufer C, Antes G, Buchler MW, Seiler CM. A systematic review and meta-analysis of pylorus-preserving versus classical pancreaticoduodenectomy for surgical treatment of periampullary and pancreatic carcinoma. Ann Surg 2007; 245: 187–200. 3 Huttner FJ, Fitzmaurice C, Schwarzer G, Seiler CM, Antes G, Buchler MW et al. Pylorus-preserving pancreaticoduodenectomy (pp Whipple) versus pancreaticoduodenectomy (classic Whipple) for surgical treatment of periampullary and pancreatic carcinoma. Cochrane Database Syst Rev 2016; (2)CD006053. 4 Eshuis WJ, van Eijck CH, Gerhards MF, Coene PP, de Hingh IH, Karsten TM et al. Antecolic versus retrocolic route of the gastroenteric anastomosis after pancreatoduodenectomy: a randomized controlled trial. Ann Surg 2014; 259: 45–51. 5 Chen QJ, He ZQ, Yang Y, Zhang YS, Chen XL, Yang HJ et al. Is there comparable morbidity in pylorus-preserving and pylorus-resecting pancreaticoduodenectomy? A meta-analysis. J Huazhong Univ Sci Technolog Med Sci 2015; 35: 793–800. 6 Kawai M, Tani M, Hirono S, Miyazawa M, Shimizu A, Uchiyama K et al. Pylorus ring resection reduces delayed gastric emptying in patients undergoing pancreatoduodenectomy: a prospective, randomized, controlled trial of pylorus-resecting versus pylorus-preserving pancreatoduodenectomy. Ann Surg 2011; 253: 495–501. 7 Hanna MM, Gadde R, Tamariz L, Allen C, Meizoso J, Sleeman D et al. Delayed gastric emptying after pancreaticoduodenectomy: is subtotal stomach preserving better or pylorus preserving? J Gastrointest Surg 2015; 19: 1542–1552. 8 Huang W, Xiong JJ, Wan MH, Szatmary P, Bharucha S, Gomatos I et al. Meta-analysis of subtotal stomach-preserving pancreaticoduodenectomy vs pylorus

© 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

12

13

14

15

16

17

18 19

20

21

preserving pancreaticoduodenectomy. World J Gastroenterol 2015; 21: 6361–6373. Wu W, Hong X, Fu L, Liu S, You L, Zhou L et al. The effect of pylorus removal on delayed gastric emptying after pancreaticoduodenectomy: a meta-analysis of 2599 patients. PLoS One 2014; 9: e108380. Zhou Y, Lin L, Wu L, Xu D, Li B. A case-matched comparison and meta-analysis comparing pylorus-resecting pancreaticoduodenectomy with pylorus-preserving pancreaticoduodenectomy for the incidence of postoperative delayed gastric emptying. HPB (Oxford) 2015; 17: 337–343. Hackert T, Probst P, Knebel P, Dorr-Harim C, Bruckner T, Klaiber U et al. Pylorus resection does not reduce delayed gastric emptying after partial pancreato-duodenectomy: a blinded randomized controlled trial (PROPP study, DRKS00004191). Ann Surg 2017; [Epub ahead of print]. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JPA et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ 2009; 339: b2700. Higgins JPT, Green S (eds). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration: Copenhagen, 2011. Farrokhyar F, Karanicolas PJ, Thoma A, Simunovic M, Bhandari M, Devereaux PJ et al. Randomized controlled trials of surgical interventions. Ann Surg 2010; 251: 409–416. Probst P, Knebel P, Grummich K, Tenckhoff S, Ulrich A, Buchler MW et al. Industry bias in randomized controlled trials in general and abdominal surgery: an empirical study. Ann Surg 2016; 264: 87–92. Sterne JA, Hernan MA, Reeves BC, Savovic J, Berkman ND, Viswanathan M et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ 2016; 355: i4919. Trikalinos TA, Balion CM, Coleman CI, Griffith L, Santaguida PL, Vandermeer B et al. Chapter 8: meta-analysis of test performance when there is a ‘gold standard’. J Gen Intern Med 2012; 27(Suppl 1): S56–S66. DerSimonian R, Laird N. Meta-analysis in clinical trials revisited. Contemp Clin Trials 2015; 45: 139–145. Macaskill P, Walter SD, Irwig L. A comparison of methods to detect publication bias in meta-analysis. Stat Med 2001; 20: 641–654. Matsumoto I, Shinzeki M, Asari S, Goto T, Shirakawa S, Ajiki T et al. A prospective randomized comparison between pylorus- and subtotal stomach-preserving pancreatoduodenectomy on postoperative delayed gastric emptying occurrence and long-term nutritional status. J Surg Oncol 2014; 109: 690–696. Hackert T, Hinz U, Hartwig W, Strobel O, Fritz S, Schneider L et al. Pylorus resection in partial pancreaticoduodenectomy: impact on delayed gastric emptying. Am J Surg 2013; 206: 296–299.

www.bjs.co.uk

BJS

Delayed gastric emptying after pancreatoduodenectomy

22 Nanashima A, Abo T, Sumida Y, Tobinaga S, Nonaka T, Takeshita H et al. Comparison of results between pylorus-preserving pancreaticoduodenectomy and subtotal stomach-preserving pancreaticoduodenectomy: report at a single cancer institute. Hepatogastroenterology 2013; 60: 1182–1188. 23 Fujii T, Kanda M, Kodera Y, Nagai S, Sahin TT, Hayashi M et al. Preservation of the pyloric ring has little value in surgery for pancreatic head cancer: a comparative study comparing three surgical procedures. Ann Surg Oncol 2012; 19: 176–183. 24 Kurahara H, Takao S, Shinchi H, Mataki Y, Maemura K, Sakoda M et al. Subtotal stomach-preserving pancreaticoduodenectomy (SSPPD) prevents postoperative delayed gastric emptying. J Surg Oncol 2010; 102: 615–619. 25 Akizuki E, Kimura Y, Nobuoka T, Imamura M, Nishidate T, Mizuguchi T et al. Prospective nonrandomized comparison between pylorus-preserving and subtotal stomach-preserving pancreaticoduodenectomy from the perspectives of DGE occurrence and postoperative digestive functions. J Gastrointest Surg 2008; 12: 1185–1192. 26 Hayashibe A, Kameyama M, Shinbo M, Makimoto S. The surgical procedure and clinical results of subtotal stomach preserving pancreaticoduodenectomy (SSPPD) in comparison with pylorus preserving pancreaticoduodenectomy (PPPD). J Surg Oncol 2007; 95: 106–109. 27 Oida T, Mimatsu K, Kano H, Kawasaki A, Kuboi Y, Fukino N et al. Preventing delayed gastric emptying in pancreaticogastrostomy by a modified subtotal-stomach-preserving pancreaticoduodenectomy: Oida modification. Hepatogastroenterology 2011; 58: 1384–1388. 28 van Berge Henegouwen MI, van Gulik TM, DeWit LT, Allema JH, Rauws EA, Obertop H et al. Delayed gastric emptying after standard pancreaticoduodenectomy versus pylorus-preserving pancreaticoduodenectomy: an analysis of

29

30

31

32

33

34

35

200 consecutive patients. J Am Coll Surg 1997; 185: 373–379. Hackert T, Bruckner T, Dorr-Harim C, Diener MK, Knebel P, Hartwig W et al. Pylorus resection or pylorus preservation in partial pancreatico-duodenectomy (PROPP study): study protocol for a randomized controlled trial. Trials 2013; 14: 44. Wente MN, Bassi C, Dervenis C, Fingerhut A, Gouma DJ, Izbicki JR et al. Delayed gastric emptying (DGE) after pancreatic surgery: a suggested definition by the International Study Group of Pancreatic Surgery (ISGPS). Surgery 2007; 142: 761–768. Kawai M, Tani M, Hirono S, Okada K, Miyazawa M, Yamaue H. Pylorus-resecting pancreaticoduodenectomy offers long-term outcomes similar to those of pylorus-preserving pancreaticoduodenectomy: results of a prospective study. World J Surg 2014; 38: 1476–1483. Keck T, Wellner UF, Bahra M, Klein F, Sick O, Niedergethmann M et al. Pancreatogastrostomy versus pancreatojejunostomy for RECOnstruction after PANCreatoduodenectomy (RECOPANC, DRKS 00000767): perioperative and long-term results of a multicenter randomized controlled trial. Ann Surg 2016; 263: 440–449. Hanna MM, Tamariz L, Gadde R, Allen C, Sleeman D, Livingstone A et al. Delayed gastric emptying after pylorus preserving pancreaticoduodenectomy – does gastrointestinal reconstruction technique matter? Am J Surg 2016; 211: 810–819. Balzano G, Zerbi A, Braga M, Rocchetti S, Beneduce AA, Di Carlo V. Fast-track recovery programme after pancreatico-duodenectomy reduces delayed gastric emptying. Br J Surg 2008; 95: 1387–1393. Probst P, Grummich K, Heger P, Zaschke S, Knebel P, Ulrich A et al. Blinding in randomized controlled trials in general and abdominal surgery: protocol for a systematic review and empirical study. Syst Rev 2016; 5: 48.

Supporting information

Additional supporting information can be found online in the supporting information tab for this article.

© 2018 BJS Society Ltd Published by John Wiley & Sons Ltd

www.bjs.co.uk

BJS