The optimal choice for pancreatic anastomosis after pancreaticoduodenectomy: A network meta-analysis of randomized control trials

Accepted Manuscript The optimal choice for pancreatic anastomosis after pancreaticoduodenectomy: A network meta-analysis of randomized control trials Weidong Wang, Zhaohui Zhang, Chichang Gu, Qingbo Liu, Zhiqiang Liang, Wei He, Jianping Chen, Jiaming Lai PII:

S1743-9191(18)30684-8

DOI:

10.1016/j.ijsu.2018.04.005

Reference:

IJSU 4570

To appear in:

International Journal of Surgery

Received Date: 14 October 2017 Revised Date:

3 March 2018

Accepted Date: 2 April 2018

Please cite this article as: Wang W, Zhang Z, Gu C, Liu Q, Liang Z, He W, Chen J, Lai J, The optimal choice for pancreatic anastomosis after pancreaticoduodenectomy: A network meta-analysis of randomized control trials, International Journal of Surgery (2018), doi: 10.1016/j.ijsu.2018.04.005. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT The optimal choice for pancreatic anastomosis after pancreaticoduodenectomy: a network meta-analysis of randomized

RI PT

control trials

Weidong Wang，MD,1# Zhaohui Zhang, MD,2# Chichang Gu, MD,1

Qingbo Liu,

MD,2 Zhiqiang Liang, MD, 1 Wei He, MD,

Jiaming Lai,

PhD2*

1

Jianping Chen, MD,1

SC

1 Second Department of General Surgery, Shunde Hospital, Southern Medical University

M AN U

2 Department of pancreatico-Biliary Surgery, First Affiliated Hospital, Sun Yat-Sen

University, Guangzhou 510080, China.

TE D

#Contributed equally.

*Corresponding author: Jiaming Lai. Department of pancreatico-Biliary Surgery,

EP

First Affiliated Hospital, Sun Yat-Sen University, Guangzhou 510080, China. E-mail:

AC C

[email protected]

Short Title: pancreatic anastomosis after pancreaticoduodenectomy

ACCEPTED MANUSCRIPT The optimal choice for pancreatic anastomosis after pancreaticoduodenectomy: a network meta-analysis of randomized

RI PT

control trials

Abstract

A

number

pancreatic

including

anastomosis

methods

for

pancreaticogastrostomy(PG),

M AN U

pancreaticoduodenectomy

of

SC

Background:

duct-to-mucosa pancreaticojejunostomy(duct-to-mucosa PJ), invagination

pancreaticojejunostomy(invagination

PJ)

and

binding

TE D

pancreaticojejunostomy(BPJ), but the optimal choice remains unclear. We

performed a network meta-analysis to synthesize direct and indirect evidence

identify

the

optimal

choice

for

pancreatic

anastomosis

after

EP

to

AC C

pancreaticoduodenectomy

Methods: We searched the Embase, PubMed and Cochrane library databases

for randomized control trials. The relative risk (RR) and its 95% confidence

interval (CI) were calculated. The primary outcome is postoperative pancreatic

fistula (POPF).

1

ACCEPTED MANUSCRIPT Result: In total, 16 RCT studies, including a total of 2396 patients, met our

criteria. The results showed that PG is not superior to invagination PJ (RR 0.70

RI PT

95%CI: 0.35-1.39) and duct-to-mucosa PJ (RR 0.58 95%CI: 0.30-1.10)

according to the ISGPS definition. Furthermore PG cannot reduce the POPF

SC

rates than invagination PJ (RR 0.51 95%CI: 0.2-1.21) and duct-to-mucosa PJ

M AN U

(RR 0.46 95%CI: 0.16-1.14) according to the soft pancreatic texture. BPJ

might reduce the incidence of POPF than duct-to-mucosa PJ (RR 0.00 95%CI:

0.00-0.04), invagination PJ (RR 0.00 95%CI: 0.00-0.03), PG (RR 0.00 95%CI:

TE D

0.00-0.03), but the results have major limitations with only one RCT reported

BPJ and different definition of POPF.

EP

Conclusion: There are no significant differences among BPJ, duct-to-mucosa

AC C

PJ, invagination PJ and PG in the prevention of POPF, overall morbidity,

mortality and DGE. However, further randomized controlled trials should be

undertaken to ascertain these findings, especially for BPJ. Key words: pancreatic anastomosis, pancreaticoduodenectomy, pancreatic fistula, network meta-analysis

2

ACCEPTED MANUSCRIPT Introduction Pancreaticoduodenectomy (PD) is a traditional surgical procedure in the treatment of benign or malignant disease in the pancreatic head and

RI PT

periampullary regions [1]. Although advances in technique and instrument, there was substantial risk for postoperative pancreatic fistula (POPF), which is the main reason of morbidity after PD and hospitalization extended [2]. In

SC

addition, POPF was closely associated with increased mortality rate. In order

have

been

proposed,

such

M AN U

to decreasing the risk of POPF, a number of pancreatic anastomosis methods as

pancreaticogastrostomy

(PG)

and

pancreaticojejunostomy (PJ). The PJ was further divided into duct-to-mucosa pancreaticojejunostomy (PJ), invagination PJ and binding PJ (BPJ).

TE D

The efficacy and safety of these methods has been heated discussed for several decades, which produced over 10 randomized control trials (RCTs).

EP

However, the results across studies varied significantly. For example, Berger et al [3]. reported that the invagination PJ could reduce the POPF rate than

AC C

duct-to-mucosa PJ, whereas another RCT observed opposite results [4]. There were also several meta-analysis evaluating the outcomes of different pancreatic anastomosis methods [5]. However, the results were limited to the two interventions because only head-to-head comparisons were performed in traditional meta-analysis. In addition, the authors did not consider various types of PJ, which may affect the consequences largely. Therefore, the

3

ACCEPTED MANUSCRIPT question “optimal choice in performing a pancreatic anastomosis” remains unanswered. Network meta-analysis is an extension of traditional meta-analysis, which

RI PT

allows for combining direct and indirect evidence, even the interventions has not been compared directly. Network meta-analysis increased the precision of estimations as it contains direct and indirect comparisons. Furthermore,

SC

network meta-analysis provides relative ranks for multiple interventions based

M AN U

on their efficient probabilities. The objective of this study was to derive evidence-based guidelines for pancreatic anastomosis in PD by performing a network meta-analysis and to provide relative rankings of these methods

Search strategy

TE D

Method

EP

The agreement for the systemic review was based on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses)

AC C

extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions [6]. By the end of March 2017, we searched the Embase, PubMed and Cochrane library databases using the key words

‘pancreatoduodenectomy,

pancreaticoduodenectomy,

pancreaticojejunostomy, pancreaticogastrostomy, pancreatic fistula and fistula risk’ without limitations on data or language. We also conducted a manual

4

ACCEPTED MANUSCRIPT search on the reference lists of articles that have been published and complemented the literature searches by perusing previous meta-analyses.

RI PT

Study selection The selection criteria of studies are as follows: (i) Design: randomized controlled trials that has been published as abstracts, peer- reviewed articles

(ii)

Intervention

and

control:

technique

of

M AN U

pancreaticoduodenectomy;

SC

or letters that writing to the editors; (ii) Population: patients received

pancreaticojejunal anastomosis: duct-to-mucosa PJ, invagination PJ, PG and binding PJ; (iii) Outcomes: using the POPF as primary outcome, and secondary outcome measures included overall morbidity, overall mortality and

TE D

delayed gastric emptying (DGE). Studies that included the patients with metastatic hepatocellular carcinoma, total pancreatectomy, or any operation

EP

other than PD were excluded.

AC C

Data extraction

Two investigators independently inspected the complete manuscripts of

qualified studies and excerpted information from them to build a database, covering publication data (year of publication, the name of first author, and the studied population country), compared treatment programs and the number of patients who has been allocated to each group, the interested events’ number in each group, and results. Any differences in relation to the data extraction 5

ACCEPTED MANUSCRIPT were solved by supplemental investigators. Absent data in eligible studies would be found by the authors through e-mail demand.

RI PT

Risk of bias assessment

Two reviewers would evaluate the selected articles’ quality. Explanations

SC

for Cochrane Summary of Findings Table of GRADE system was made by

M AN U

software GRADEprofiler (version 3.6, http://www. gradeworkinggroup.org/).

Statistical analysis

TE D

Pairwise meta-analysis (head-to-head) was performed by using Stata SE 12.0 (StataCorp LP, College Station, TX). The statistical heterogeneity across studies was quantified by I2 statistic, with I2 values >50% indicating significant

EP

heterogeneity. Pooled estimates were calculated in random-effects model

AC C

regardless of heterogeneity. For POPF, the relative risk (RR) was used and a RR<1.0 indicates the first intervention is better.

Network meta-analysis (indirect and mixed) was performed by using the

Aggregate Data Drug Information System (ADDIS 1.16.8) under the Bayesian framework. This method takes the direct and indirect evidence of two treatments in a combined analysis into account. Indirect comparison of a pair

6

ACCEPTED MANUSCRIPT of treatments needs a general reference. The non-informative priors was adopted and the Markov chain Monte Carlo method was used to estimate the fitted each Bayesian random-effects model. There were 4 chains with 20,000

RI PT

iteration, 10 thin interval in each chain. The Gelmane-Rubin approach and trace plots were used to confirm the convergence of the models.

SC

Assumption for network meta-analysis was assessed using node-splitting

M AN U

approach, which evaluates the discrepancies between indirect and direct evidence. When the node split model was P > 0.05, the consistency model was employed or it was finished. If node-splitting analysis could not be established, the outcomes of inconsistency and consistency analysis would be

TE D

presented at the same time.

EP

We ranked the outcomes of each intervention according to their probabilities for each outcome. The ranking probability charts were also

AC C

generated to find the best adjuvant therapy. Further sensitivity analysis was performed by removing studies that caused significant heterogeneity in pairwise meta-analysis. Additional network meta-analysis was performed for subgroups in studies using International Study Group of Pancreatic Surgery (ISGPS) definition [7], studies involving patients with soft pancreas.

Results 7

ACCEPTED MANUSCRIPT

Study Characteristics and Quality. Figure 1 shows the flow of study selection. Based on the former criteria,

RI PT

we determined 2761 relevant citations through the literature search and selection, and eventually 16 RCT [3, 4, 8-21] researches including 2396 patients were contained in this meta-analysis (Figure 1). These 16 researches

SC

(Table 1) showed 4 techniques of pancreaticojejunal anastomosis for patients

M AN U

after PD including: duct-to-mucosa PJ, invagination PJ, PG and binding PJ. Figure 2 shows the network of direct comparison for PD. Only one RCT study reported the BPJ. All of studies reported information about POPF overall. The definitions of POPF varied among the different trials (Table 1). Eleven studies

TE D

applied the ISGPS definition of POPF. The outcomes of the random-effects

EP

network meta-analysis were outlined in Figure 3 and Figure 4.

According to the relationships between each strategy and GRADE system,

AC C

we assessed the quality of the evidence through relative direct comparisons. The evidence demonstrated that seven evidences have moderate or high quality. However, the other nine comparisons revealed low or very low quality (Table 1).

Primary outcomes

8

ACCEPTED MANUSCRIPT All the sixteen RCTs reported the POPF rates. The overall POPF rate after PD was18.5%(137/740) in duct-to-mucosa PJ group, 17.5%(153/874) in the invagination PJ group, 17.3%(117/676) in the PG group and 0%(0/106) in the

RI PT

BPJ group. After pooled estimation, the network meta-analysis revealed that compared with other techniques of pancreaticojejunal anastomosis, BPJ could be an effective strategy to significantly reduce the incidence of POPF and was

SC

shown to be much better than duct-to-mucosa PJ (RR 0.00 95%CI: 0.00-0.04),

M AN U

invagination PJ (RR 0.00 95%CI: 0.00-0.03), PG (RR 0.00 95%CI: 0.00-0.03). (Figure 3)

Eleven RCTs reported the POPF rate according to the ISGPS definition.

TE D

The overall POPF rate after PD was 21.9%(124/565) in duct-to-mucosa PJ group, 22.7%(117/516) in the invagination PJ group and 17.9%(108/603) in the

EP

PG group. But no significant different were found among the three groups, indicating that PG cannot decrease the POPF rate than invagination PJ(RR

AC C

0.70 95%CI: 0.35-1.39) and duct-to-mucosa PJ(RR 0.58 95%CI: 0.30-1.10) according to the ISGPS definition. (Figure 3)

Owing to pancreatic texture is a significant factor for POPF, so we performed a subgroup analysis. Eight RCT studies were considering. The POPF rate was 18.2%(83/457) in the PG with soft pancreas, While that in the invagination PJ group and in the duct-to-mucosa PJ group was 22.5%(99/439) 9

ACCEPTED MANUSCRIPT and 20.6%(109/529), And No significant different were found among the three groups, indicating that PG cannot decrease the POPF rate than invagination PJ (RR 0.51 95%CI: 0.2-1.21) and duct-to-mucosa PJ (RR 0.46 95%CI:

RI PT

0.16-1.14) according to the soft pancreatic texture. (Figure 3)

Secondary outcomes

SC

Overall morbidity

M AN U

Fourteen RCTs reported the morbidity rates. The overall postoperative morbidity rate among the fourteen RCTs was 50%(267/534) in duct-to-mucosa PJ group, 48.4%( 423/874) in the invagination PJ group, 52.7%(235/446) in the PG group and 24.5%(26/106) in the BPJ group. Although the BPJ group

TE D

revealed a lower morbidity rate, there was no significant difference among the four groups. (Figure 4)

EP

Overall mortality

Thirteen RCTs reported the mortality rates. The overall postoperative

AC C

mortality rate among the thirteen RCTs was 3.5%(22/626) in duct-to-mucosa PJ group, 3.1%(23/745) in the invagination PJ group, 4.3%(24/550) in the PG group and 2.8%(3/106) in the BPJ group. There was no significant difference among the four groups. (Figure 4) Delayed gastric emptying In most of the RCTs, postoperative DGE was defined as gastric stasis for more than a week. Among the sixteen RCTs, only one did not report the 10

ACCEPTED MANUSCRIPT postoperative DGE rate. The overall postoperative DGE rate a was 17.4%(112/643) in duct-to-mucosa PJ group, 7.5%(54/724) in the invagination PJ group, 23.7%(149/628) in the PG group and 3.8%(4/106) in the BPJ group.

difference among the four groups. (Figure 4) Node split of network meta-analysis

RI PT

Although the BPJ group revealed a lower DGE rate, there was no significant

SC

Pair wise meta-analysis did not suggest inconsistency in direct

M AN U

comparison. The node-splitting models suggested that there was no statistical difference between direct and indirect comparisons (P > 0.05) (Table 2).

Discussion

TE D

The best anastomotic technique in pancreas surgery has been an area of controversy for a long time and raised considerable debate among experts and

EP

investigators [22]. Our research intended to find the optimal pancreatic anastomosis after PD. The principal findings of this study are as follows: (1)

AC C

the BPJ did not have advantage in cutting down the incidence of overall morbidity, mortality and DGE, while it might largely decrease the incidence of POPF rate than invagination PJ, invagination PJ and PG after PD. (2) the PG is not superior to duct-to-mucosa PJ, invagination PJ in the prevention of POPF. PJ is the most common method to restore pancreatic anastomosis after PD [23]. And the anastomosis between pancreatic stump and the jejunum is 11

ACCEPTED MANUSCRIPT usually conducted with duct-to-mucosa PJ or invagination PJ. An anastomosis with duct-to-mucosa is more beneficial for recovery than with invagination, and jejunal serosa can prevent the pancreatic remnant from Intestinal juice.

RI PT

However, a potential space between the jejunal serosa and pancreatic stump may cause the accumulation of pancreatic juice because of accessory pancreatic duct [24]. Furthermore, it’s a difficult method for duct-to-mucosa

SC

anastomosis to dealing with a soft, friable, and fatty pancreas with a small duct

M AN U

[24]. Hence, the invagination PJ is more suitable for managing a soft pancreas and decreasing the risk of PF owing to the accessory pancreatic duct. However, the invagination PJ may lead to ischemia of pancreatic stump[10]. In 2009，Berger et al [3] conducted an RCT including 197 patients at two

TE D

institutions underwent PD, and the results revealed that there were 23 fistulas (24%) in the duct to mucosa cohort and 12 fistulas (12%) in the invagination

that

the

EP

cohort (p < 0.05), while in 2016, another RCT reported by Bai X et al [4] found clinically

relevant

POPF

rate

in

the

patients

treated

AC C

with invagination vs duct-to-mucosa anastomoses was were 17.6% vs 3.1%, respectively (p = 0.004). Thus, a meta-analysis including seven RCTs was conducted by Shuisheng Zhang et al [25] and found that the difference in POPF rate between the duct-to-mucosa and invagination PD was not significant (RR = 1.03, 95% CI = 0.76-1.39, P = 0.86). In our study, we also reached a more reliable result that duct-to-mucosa and invagination pancreaticojejunostomy techniques after pancreaticoduodenectomy were 12

ACCEPTED MANUSCRIPT comparable in terms of postoperative pancreatic fistula through the network meta-analysis. In addition, we also carried out subgroup analysis based on the POPF definition proposed by ISGPS and soft pancreas. The results still

RI PT

showed no significant difference between duct-to-mucosa and invagination PJ. However, the POPF rate is still nearly to 20% in duct-to-mucosa and

SC

invagination PJ after PD.

M AN U

Since 2002, a novel method of pancreatic anastomosis, binding PJ (BPJ), was carried out by Peng et al, and several case reports indicated that BPJ may associated with lower incidence of POPF [26-28]. Furthermore, In 2007, Peng et al. [19] presented a prospective randomized study on 217 patients

TE D

comparing BPJ and invagination PJ, and found that BPJ significantly decreased postoperative complication and pancreaticojejunostomy leakage

EP

rates (0% vs 7.2%; P<0.005). It is hypothesized that POPF may result from a needle unintentionally penetrating a pancreatic duct, or a laceration of fragile

AC C

pancreatic parenchyma in the processes of suturing or tying a knot. And the minor pancreatic juice gradually results in severe POPF because of auto-digestion around the anastomoses [19]. Thus, more studies have been presented to assess the safety and efficacy of BPJ after PD. A French prospective study including forty-five consecutive patients with soft pancreas and non-dilated main pancreatic duct showed that four patients (8.9%) developed a POPF and binding pancreaticojejunostomy according to Peng is a 13

ACCEPTED MANUSCRIPT safe and secure technique that improves the rate of pancreatic fistula, especially in case of soft texture of the pancreas remnant [29]. Similarly, a retrospective study involving 21 patients who received BPJ from 2006 to 2011

RI PT

reported that a total 2 patient (9.5%) developed clinical POPF [30]. However, another prospective and dual-institution study analyzed 121 patients underwent BPJ or duct-to-mucosa PJ, and found that BPJ was not significantly

SC

related with the occurrence of POPF, and instead increased the length of

M AN U

hospital stay [31]. Furthermore, a recent meta-analysis including one RCT and four observational studies revealed that no significant superiority was found in BPJ

group

regarding

the

incidence

of

POPF

than

conventional

pancreaticojejunostomy [32]. These conflicting conclusions may be explained

TE D

as early learning curve results.

Although our results show that BPJ might reduce the POPF rate

EP

comparing with invagination, invagination PJ and PG, It is hardly to conclude BPJ is superiority than other techniques. Only one RCT reported BPJ and the

AC C

results showed a decreased pancreatic fistula rate (0% vs 7.2%), which might influence the statistical results. In addition, the definition of POPF did not follow the ISGPF definition in Peng’s research, and this might cause heterogeneity. Furthermore, there are no significant difference in the incidence of overall morbidity, mortality and DGE. Thus, more RCT studies are needed to prove the efficiency of BPJ.

14

ACCEPTED MANUSCRIPT Pancreaticogastrostomy has been gaining favor in recent years. In 1995,

Yeo et al [18] found the incidence of pancreatic fistula was similar for the

RI PT

pancreaticogastrostomy (12.3%) and pancreaticojejunostomy (11.1%) groups,

and the data did not support the hypothesis that pancreaticogastrostomy is

SC

safer than pancreaticojejunostomy. Another RCT study by Bassi et al [7]

M AN U

showed that comparing with PJ, PG did not show any significant differences in

the overall postoperative complication rate (29% in PG, 39% in PJ, P = not

significant) or incidence of pancreatic fistula (13% in PG and 16% in PJ, P =

TE D

not significant). However, in 2013, A Belgian RCT study including 167 patients

to receive PJ and 162 to receive PG, and showed PG is more efficient

reducing

the

incidence

of

EP

than pancreaticojejunostomy in

AC C

postoperative pancreatic fistula (OR 2.86, 95% CI 1.38-6.17; p=0.002) [21].

Furthermore, in a recent meta-analysis by Han Qin et al [33] comparing PG

versus PJ after PD, 10 RCTs representing 1629 patients (826 PG, 803 PJ)

were included, and PG was associated with significantly less POPF when

compared to PJ (11.2% vs 18.7%; OR = 0.53; 95% CI 0.38–0.75; P = 0.0003).

15

ACCEPTED MANUSCRIPT However, the authors did not take different type of PJ into consideration, which

may influence the results significantly. So in our network meta-analysis, we

RI PT

found that PG is not superior to duct-to-mucosa PJ and invagination PJ in the

prevention of POPF. In addition, we revealed that no significant different were

SC

found among the three groups according to the ISGPS definition and soft

M AN U

pancreatic texture. Thus, PG might not superior to duct-to-mucosa PJ,

invagination PJ in the prevention of POPF.

The present meta-analysis has some strengths. Firstly, this network

TE D

meta-analysis includes sixteen RCT conducted in different centers in different

countries, including PG, duct-to-mucosa PJ, invagination PJ and BPJ, which

EP

provides a relatively high-level of evidence. Secondly, this network

AC C

meta-analysis is stable as proven by subgroups. However, this analysis also

has some limitations. Firstly, only one RCT reported BPJ and the results

showed a decreased pancreatic fistula rate (0% vs 7.2%). Secondly, several

known risk factor to POPF such as diameter of the main pancreatic duct, stent

16

ACCEPTED MANUSCRIPT use and octreotide use were unable to be explored in subgroup analysis due to

deficient information in the included trials.

RI PT

Conclusion Our study showed that PG is not superior to duct-to-mucosa PJ and

SC

invagination PJ in the prevention of POPF according to the ISGPS definition

M AN U

and pancreatic texture. BPJ is comparable to PJ and PG in terms of overall

morbidity, mortality and DGE, and it might decrease the incidence of POPF.

However，the results have major limitations with only one RCT reported BPJ

TE D

and different definition of POPF. Thus there are no significant differences

among BPJ, duct-to-mucosa PJ, invagination PJ and PG，further randomized

AC C

for BPJ.

EP

controlled trials should be undertaken to ascertain these findings, especially

17

ACCEPTED MANUSCRIPT Refence 1.

Schafer M, Mullhaupt B, Clavien PA, Evidence-based pancreatic head resection for pancreatic cancer and chronic pancreatitis, Ann. Surg. 236

RI PT

(2002) 137–148. 2. Wang S, Wang X, Dai H, et al., The effect of intraoperative fluid volume administration on pancreatic fistulas after pancreaticoduodenectomy, J.

AC,

Howard

TJ,

Kennedy

EP,

et

al.,

Does

type

of

M AN U

3. Berger

SC

Invest. Surg.27 (2014) 88–94.

pancreaticojejunostomy after pancreaticoduodenectomy decrease rate of pancreatic fistula? A randomized, prospective, dualinstitution trial, J Am Coll Surg. 208 (2009) 738-49.

TE D

4. BaiX, ZhangQ, GaoS, et al., Duct-to-mucosa vs invagination for pancreaticojejunostomy after pancreaticoduodenectomy: a prospective,

EP

randomized controlled trial from a single surgeon, J Am Coll Surg. 222 (2016) 10-8.

B,

AC C

5. Menahem

Guittet

Pancreaticogastrostomy

L, is

Mulliri superior

A, to

Alves

A,

Lubrano

J,

pancreaticojejunostomy

for

prevention of pancreatic fistula after pancreaticoduodenectomy: an updated meta-analysis of randomized controlled trials, Ann. Surg. 261 (2015) 882-7.

18

ACCEPTED MANUSCRIPT 6. Moher D, Liberati A, Tetzlaff J, Altman DG, Preferred reporting items for systematic

reviews

and

meta-analyses:

the

PRISMA

statement,

International journal of surgery 8 (2010) 336–341.

RI PT

7. Bassi C, Dervenis C, Butturini G, et al., Postoperative pancreatic fistula: an international study group (ISGPF) definition, Surgery 138 (2005) 8–13.

8. Bassi C, Falconi M, Molinari E, et al., Duct-to-mucosa versus end-to-side

SC

pancreaticojejunostomy reconstruction after pancreaticoduodenectomy:

M AN U

results of a prospective randomized trial, Surgery 134 (2003) 766-771. 9. Chou FF, Sheen-Chen SM, Chen YS, et al., Postoperative morbidity and mortality of pancreaticoduodenectomy for periampullary cancer, Eur J Surg. 162 (1996) 477-481.

duct

TE D

10. El Nakeeb A, El Hemaly M, Askr W, et al., Comparative study between to

mucosa

and

invagination

pancreaticojejunostomy

after

EP

pancreaticoduodenectomy: a prospective randomized study, Int J Surg. 16 (2015) 1-6.

AC C

11. Han JM, Wang XB, Quan ZF, et al, Duct-to-mucosa anastomosis and incidence of pancreatic fistula following pancreaticoduodenectomy, J Med Postgrad. 22 (2009) 961-964.

12. Langrehr JM, Bahra M, Jacob D, et al., Prospective randomized comparison

between

a

new

mattress

technique

and

Cattell

(duct-to-mucosa) pancreaticojejunostomy for pancreatic resection, World J Surg. 29 (2005) 1111-1119. 19

ACCEPTED MANUSCRIPT 13. J. Grendar, J.F. Ouellet, F.R. Sutherland, et al., In search of the best reconstructive

technique

after

pancreaticoduodenectomy:

pancreaticojejunostomy versus pancreaticogastrostomy, Can. J. Surg. J.

RI PT

Can. de Chir. 58(3) (2015) 154-159. 14. A. El Nakeeb, E. Hamdy, A.M. Sultan, et al., Isolated Roux loop pancreaticojejunostomy

versus

pancreaticogastrostomy

after

SC

pancreaticoduodenectomy: a prospective randomized study. HPB official J.

M AN U

Int. Hepato Pancreato Biliary Assoc, 16 (8)(2014)713-722.

15. J.Figueras, L.Sabater, P.Planellas, et al, Randomized clinical trial of pancreaticogastrostomy versus pancreaticojejunostomy on the rate and severity of pancreatic fistula after pancreaticoduodenectomy, Br. J. Surg.

TE D

100 (12) (2013) 1597-1605.

16. U.F. Wellner, O. Sick, M. Olschewski, et al., Randomized controlled trial

comparing

pancreatogastrostomy

versus

pan-

EP

single-center

creaticojejunostomy after partial pancreatoduodenectomy, J. Gastrointest.

AC C

Surg. official J. Soc. Surg. Alimentary Tract 16(9) (2012) 1686-1695. 17. L. Fernandez-Cruz, R. Cosa, L. Blanco, M.A. Lopez-Boado, et al., Pancreatogastrostomy with gastric partition after pylorus-preserving pancreatoduodenectomy

versus

conventional

pancreatojejunostomy:

prospective randomized study, Ann. Surg. 248 (6) (2008) 930-938.

20

a

ACCEPTED MANUSCRIPT 18. C.J. Yeo, J.L. Cameron, M.M. Maher, et al., A prospective randomized trial of pancreaticogastrostomy versus pancreaticojejunostomy after pancreaticoduodenectomy, Ann Surg. 222 (4) (1995) 580.

RI PT

19. Peng SY, Wang JW, Lau WY, et al., Conventional versus binding pancreaticojejunostomy after pancreaticoduodenectomy: a prospective randomized trial, Ann Surg. 245 (2007) 92-8.

SC

20. T. Keck, U.F. Wellner, M. Bahra, et al., Pancreatogastrostomy versus

M AN U

pancreatojejunostomy for reconstruction after pancreatoduodenectomy: perioperative and long-term results of a multicenter randomized controlled trial, Ann Surg. 263(3) (2016) 440-449.

21. B. Topal, S. Fieuws, R. Aerts, et al, Pancreaticojejunostomy versus

TE D

pancreaticogastrostomy reconstruction after pancreaticoduodenectomy for pancreatic or periampullary tumours: a multicentre randomised trial,

EP

Lancet Oncol. 14 (7) (2013) 655-662. 22. Søreide K, Labori K J, Risk factors and preventive strategies for pancreatic

fistula

after

pancreatic

surgery:

a

AC C

post-operative

comprehensive review, Scandinavian Journal of Gastroenterology 51(10) (2016) 1147–1154.

23. Fisher WE, Chai C, Hodges SE, et al., Effect of BioGlue on the incidence of pancreatic fistula following pancreas resection, J Gastrointest Surg. 12 (2008) 882–890.

21

ACCEPTED MANUSCRIPT 24. Marcus SG, Cohen H, Ranson SH, Optimal management of the pancreatic remnant after pancreaticoduodenectomy, Ann. Surg. 222 (1995) 635-45.

RI PT

25. Zhang, S., Lan, Z., Zhang, J., et al., Duct-to-mucosa versus invagination pancreaticojejunostomy after pancreaticoduodenectomy: a meta-analysis, Oncotarget 8(28) (2017) 46449–46460.

SC

26. Peng S, Mou Y, Cai X, et al., Binding pancreaticojejunostomy is a new

M AN U

technique to minimize leakage, Am J Surg.183 (2002) 283-285. 27. Peng SY, Mou YP, Liu YB, et al., Binding pancre- aticojejunostomy: 150 consecutive cases without leakage, J Gastrointest Surg. 7 (2003) 898-900. 28. Peng SY, Wang JW, Li JT, et al., Binding pancreaticojejunostomy-a safe

29. Buc

E,

TE D

and reliable anastomosis procedure, HPB (Oxford) 6 (2004) 154–160. Flamein

R,

after

C,

et

al.,

Peng’s

pancreaticoduodenectomy:

a

binding French

EP

pancreaticojejunostomy

Golffier

prospective study, J Gastrointest Surg. 14 (2010) 705–710.

AC C

30. Kim JM, Hong JB, Shin WY, et al, Preliminary results of binding pancreaticojejunostomy, Korean J Hepatobiliary Pancreat Surg. 18 (2014) 21–25.

31. Casadei

R,

Ricci

pancreaticojejunostomy

C,

Silvestri

after

S,

et

al.,

Peng’s

pancreaticoduodenectomy.

An

binding Italian,

prospective, dual-institution study, Pancreatology 13 (2013) 305–309.

22

ACCEPTED MANUSCRIPT 32. Zhang X, Dong X, Liu P, et al., Binding versus Conventional Pancreaticojejunostomy in Preventing Postoperative Pancreatic Fistula: A Systematic Review and Meta-Analysis, Dig Surg. 34 (4) (2017) 265-280.

pancreaticojejunostomy

on

pancreatic

RI PT

33. Qin H, Luo L, Zhu Z, et al., Pancreaticogastrostomy has advantages over fistula

after

pancreaticoduodenectomy: A meta-analysis of randomized controlled

EP

Figure legends

TE D

M AN U

SC

trials, Int J Surg. 36 (2016) 18-24.

Fig. 1. Flow diagram of search strategy and study selection

AC C

Fig. 2. Network of the comparisons for the Bayesian network meta-analysis. The size of the nodes is proportional to the number of patients (in parentheses) to receive the treatment. The width of the lines is proportional to the number of trials (beside the line) comparing the connected treatments. Fig. 3. Pooled hazard ratios for POPF Fig. 4. Pooled hazard ratios for overall morbidity, overall mortality and DGE

23

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

24

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

25

ACCEPTED MANUSCRIPT Overall -0.05 (-0.56, 0.50) 0.42 (-0.17, 1.00) 0.37 (-0.20, 0.97) 0.55 (-0.10, 1.20) 0.35 (-0.33, 1.04) -0.19 (-0.85, 0.46) 0.67 (-0.19, 1.59) 0.78 (-0.13, 1.83) 0.12 (-0.61, 0.87) 0.02 (-0.46, 0.48) -0.05 (-0.45, 0.36) -0.07 (-0.41, 0.28) -0.02 (-0.86, 0.89) 0.04 (-0.85, 0.94) 0.07 (-0.88, 0.94) -0.02 (-0.62, 0.57) -0.37 (-0.99, 0.30) -0.35 (-0.97, 0.36)

AC C

EP

TE D

M AN U

SC

RI PT

Table 2:Results of node-splitting models for the test of difference between direct and indirect effect in the analysis Outcome Name Direct effect Idirect effect PJ-DtoM, PJ-Inv 0.09 (-0.52, 0.79) -0.42 (-1.61, 0.62) PF PG-Inv, PJ-DtoM 0.62 (-0.13, 1.43) 0.10 (-1.02, 1.09) PG-Inv, PJ-Inv 0.19 (-0.65, 0.95) 0.70 (-0.27, 1.85) PG-Inv, PJ-DtoM 0.61 (-0.17, 1.51) 0.38 (-0.95, 1.59) PG-Inv, PJ-Inv 0.28 (-0.70, 1.15) 0.47 (-0.61, 1.74) PF(ISGPS definition) PJ-DtoM, PJ-Inv -0.13 (-0.92, 0.76) -0.35 (-1.70, 0.82) PG-Inv, PJ-DtoM 0.51 (-0.50, 1.63) 1.02 (-0.78, 2.63) PG-Inv, PJ-Inv 1.06 (-0.44, 2.63) 0.54 (-0.75, 2.00) PF( soft pancreatic texture) PJ-DtoM, PJ-Inv 0.08 (-0.76, 0.90) 0.65 (-1.34, 2.28) PG-Inv, PJ-DtoM 0.51 (-0.18, 1.25) -0.31 (-0.92, 0.25) PG-Inv, PJ-Inv -0.24 (-0.71, 0.16) 0.56 (-0.19, 1.39) Overall morbidity PJ-DtoM, PJ-Inv 0.08 (-0.29, 0.44) -0.75 (-1.58, 0.10) PG-Inv, PJ-DtoM -0.03 (-1.17, 1.13) 0.07 (-1.56, 1.63) PG-Inv, PJ-Inv 0.09 (-1.12, 1.24) -0.02 (-1.60, 1.53) Overall mortality PJ-DtoM, PJ-Inv 0.00 (-1.13, 1.06) 0.08 (-1.65, 1.70) PG-Inv, PJ-DtoM 0.09 (-0.65, 0.90) -0.39 (-1.71, 0.84) Delayed gastric emptying PG-Inv, PJ-Inv -0.50 (-1.48, 0.31) -0.08 (-1.14, 1.16) PJ-DtoM, PJ-Inv -0.18 (-1.01, 0.78) -0.60 (-1.86, 0.47)

26

P-value 0.42 0.4 0.38 0.74 0.78 0.75 0.58 0.56 0.5 0.07 0.08 0.07 0.93 0.96 0.96 0.43 0.53 0.51

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT Highlights A network meta-analysis is performed to synthesize direct and indirect

RI PT

evidence to identify the optimal choice for pancreatic anastomosis after

pancreaticoduodenectomy.

SC

PG is not superior to duct-to-mucosa PJ and invagination PJ in the

M AN U

prevention of POPF according to the ISGPS definition and pancreatic

texture.

no significant differences among BPJ, duct-to-mucosa PJ, invagination PJ

TE D

and PG in the prevention of POPF, overall morbidity, mortality and DGE.

However, further randomized controlled trials should be undertaken to

AC C

EP

ascertain these findings, especially for BPJ.