Blumgart anastomosis reduces the incidence of pancreatic fistula after pancreaticoduodenectomy: a systematic review and meta-analysis

Postoperative pancreatic fistula (POPF) is the most serious complication after pancreaticoduodenectomy (PD). Recently, Blumgart anastomosis (BA) has been found to have some advantages in terms of decreasing POPF compared with other pancreaticojejunostomy (PJ) using either the duct-to-mucosa or invagination approach. Therefore, the aim of this study was to examine the safety and effectiveness of BA versus non-Blumgart anastomosis after PD. The PubMed, EMBASE, Web of Science and the Cochrane Central Library were systematically searched for studies published from January 2000 to March 2020. One RCT and ten retrospective comparative studies were included with 2412 patients, of whom 1155 (47.9%) underwent BA and 1257 (52.1%) underwent non-Blumgart anastomosis. BA was associated with significantly lower rates of grade B/C POPF (OR 0.38, 0.22 to 0.65; P = 0.004) than non-Blumgart anastomosis. Additionally, in the subgroup analysis, the grade B/C POPF was also reduced in BA group than the Kakita anastomosis group. There was no significant difference regarding grade B/C POPF in terms of soft pancreatic texture between the BA and non-Blumgart anastomosis groups. In conclusion, BA after PD was associated with a decreased risk of grade B/C POPF. Therefore, BA seems to be a valuable PJ to reduce POPF comparing with non-Blumgart anastomosis.

Methodological quality of included studies. The quality assessment score of the included studies is shown in Table 1. The quality of only one RCT study was assessed using the Cochrane Collaboration Handbook 57 . The RCT trial 30 clearly reported allocation concealment methods, withdrawals, dropouts and losses to follow-up, while not describing any kind of blinding; so, we deemed it to carry an unclear risk. The methodological quality of the included non-RCT studies was evaluated as described by McKay and colleagues 58 . Results of the meta-analysis and subgroup analysis. BA versus non-Blumgart anastomosis. Primary outcomes. The forest plots of the primary outcomes are shown in Fig. 2. All included studies reported POPF (grade B or C), while only 4 studies reported grade A or biochemical leak POPF. Therefore, we only summarized and reported the rate of grade B/C POPF. Although some degree of heterogeneity was present among these studies (I 2 = 76 per cent), the use of the random-effects model did not change the result. The BA group was associated with significantly lower rates of POPF (grade B/C) (OR 0.38, 0.22 to 0.65; P = 0.004) and POPF (grade B/C) using 2017 ISGPF definition (OR 0.58, 0.39 to 0.87; P = 0.008) than non-Blumgart group. However, there was no difference in the rate of POPF (grade B/C) in soft pancreatic texture and grade C POPF between the two groups.
Secondary outcomes. The pooled results of the secondary outcomes of BA group versus non-Blumgart group are summarized in Table 3. In the study of Kojima 34 , conventional PJ was divided into the CWA and KA groups. The duration of the operation was significantly longer as result of the additional operation including abdominal lavage and covering the wound and drain with dressing materials; therefore, it was removed from the sensitivity analysis. In addition, the intraoperative blood loss and postoperative hospital stay were reported in the study of Kojima in the CWA and KA groups. In summary, BA were associated with significantly lower rates of overall postoperative haemorrhage (OR 0. 48 Table 4 and Appendix 1. BA was associated with significantly lower rates of POPF (grade B/C) (OR 0.43, 0.21 to 0.76; P = 0.004), grade C POPF (OR 0.24, 0.06 to 0.89; P = 0.03) and reoperation (OR 0.41, 0.18 to 0.90; P = 0.03), as well as shorter postoperative hospital stay (WMD − 9.80, − 15.19 to − 4.14; P = 0.0004) than invagination PJ. However, major morbidity and mortality were comparable between the two approaches.

Publication bias.
To examine any publication bias in the included studies, a funnel plot was constructed using the Review Manager 5.3. The funnel plot based on grade B/C POPF is shown in Fig. 3. The funnel plot was asymmetric; therefore publication bias might exist.

Discussion
Until now, the optimal reconstruction technique for PJ after PD has remained controversial 59 . This systematic review and meta-analysis not only made a comparison between BA and non-Blumgart PJ, but it also compared BA with CWA, KA and invagination PJ. This study suggested that the rates of grade B/C POPF, morbidity and postoperative haemorrhage were significantly lower in the BA group than in the non-Blumgart group. Therefore, BA appeared to be a safe, feasible and effective PJ technique compared to non-Blumgart PJ. According to the previous reports, there are a number of plausible explanations for why BA was superior to a non-Blumgart anastomosis procedure in reducing the POPF rate. First, BA reduces tangential tension and shear force at the pancreatic stump via the use of the transpancreatic U-sutures. Second, BA maintains the pancreatic stump with a sufficient blood supply by interrupted mattress U-sutures. Furthermore, BA guarantees a tensionfree approximation between the posterior and anterior seromuscular jejunum and excellent visualization of the Table 1. Study characteristics. BMI, Body Mass Index; PV, portal vein; SMV, superior mesenteric vein; O, octreotide; MPD(N/D), main pancreatic duct(Non-dilated/dilated); BA, Blumgart anastomosis; CWA, Cattell-Warren anastomosis; KA, Kakita anastomosis; NA, Data not available; BMDPP, benign and malignant disease of the pancreatic head and the periampullary region; PD, pancreaticoduodenectomy; PPPD, pyloruspreserving pancreatoduodenectomy; SSPPD, subtotal stomach-preserving pancreatoduodenectomy; S, Select; Ex, external stent; In, internal stent; RCT, Randomized Controlled Trial; EA, Embedded anastomosis; DtoM, duct-to-mucosa anastomosis. # Data was recorded as Mean ± SD or median (range). *Randomized clinical trials (RCTs) were scored according to the RoB 2.0 of the Cochrane Collaboration; the method of McKay and colleagues was used for non-randomized studies.

Author
Year Country Design Group  28 reported that BA was incomplete and resulted in an unstable covering of pancreas stump that is prone to evoke POPF when joining a thin jejunum and a thick pancreas. To further achieve improvement, accumulated modifications of Blumgart anastomosis were proposed, including utilization of one suture for the anterior and posterior wall 19 , knot-tying on the ventral wall of the jejunum 28,30 , the use of closed drains and dressing materials to cover the wound and drains 34 , and a wide U-shape suture 31 that minimized the space between the knots. Recently, Hirono et al. 30 suggested that pancreatico-enteric anastomosis should use as few sutures as possible, taking care to not tie the suture too tightly and thus maintaining blood flow in the pancreatic stump. The definition and classification of ISGPF was used in all the included studies. However, the ISGPF was updated in several studies, and the POPF grade A was called a "biochemical leak" because it has no significance in clinical practice. However, the definitions of grade B/C POPF are not very different between the 2005 and 2017 ISGPF. In addition, all included studies reported grade B or C POPF, while only 4 studies reported all POPF (including grade A or biochemical leak, grade B and grade C). Therefore, in the analysis of postoperative outcomes following PD, the present study mainly focused on grade B/C POPF 60 . In the present meta-analysis, the BA group had a lower rate of grade B/C POPF (8.3% vs 22.4%, P = 0.0004) than the non-Blumgart group, which was similar to the result of a previous study 39 . The incidences of grade B/C POPF after BA ranged from 0 to 30.8% as has been described in previous case series studies ( Table 5). One of the important factors that affected the development of POPF was pancreatic texture. For soft pancreatic texture, the incidence of POPF (grade B/C) was lower in the BA group than in the conventional PJ group (27.3% versus 41.2%), although there was no statistically significant difference (OR 0.46, 0.14 to 1.53; P = 0.21).Therefore, it is possible that a soft pancreas led to a high incidence of pancreatic fistula, regardless of which way the PJ anastomosis was used.
Previous studies have suggested that POPF was the main cause for intra-abdominal abscess, postoperative haemorrhage and DGE after PD 2 . Thus, to some extent, it is clear that once the incidence of POPF decreases, perhaps postoperative morbidity would significantly decline. Our analyses indicated that the rates of intraabdominal abscess and postoperative haemorrhage were significantly lower in the BA group (9.1% vs 16.5%, P < 0.0001), which was mainly due to the absence of dead space between the pancreatic cut surface and the jejunal wall in the U-suture technique group 30 . According to the results of the current meta-analysis, BA might significantly minimize the rate of reoperation (3.0% vs 4.9%, p = 0.005). The incidence of reoperation mainly resulted from severe complications including POPF (grade B/C), bleeding, and abscess formation. Therefore, the rate of overall postoperative morbidity and mortality in the BA group were 23.7% and 0.9%, respectively, less than in previous studies. At the same time, because of the decrease in complications, postoperative hospital stays were also reduced. The subgroup analysis that focused specifically on clinical trials comparing Blumgart anastomosis with other types of PJ anastomosis still favoured the advantages of BA. www.nature.com/scientificreports/ There were some limitations in our meta-analysis that should be acknowledged. First, most included studies were retrospective before-after studies that inevitably led to selection bias. Second, the Blumgart technique was slightly different among studies with several modifications. Third, there was probably publication bias in the current study, mainly due to the unpublished studies with negative results.

Conclusions.
In conclusion, compared with non-Blumgart PJ, BA was safer and more effective after PD with a lower incidence of grade B/C POPF, comparable operative time and intraoperative blood loss, lower morbidity and a shorter postoperative hospital stay. However, before recommending widespread use, it is necessary to  Inclusion and exclusion criteria. The studies were included based on the following criteria: English language articles published in peer-reviewed journals; human studies; studies with at least the primary outcome mentioned; only comparative clinical trials with full-text descriptions; clear documentation of the PJ technique and where multiple studies came from the same institute and/or authors, either the higher quality study or the more recent publication was included in the analysis. The following studies were excluded: abstracts, letters, editorials, expert opinions, case reports, reviews, trial protocols, and studies related to comparing BA with PG.   Data extraction and quality assessment. Data were extracted independently by two reviewers using standard forms and were cross-checked. Inconsistencies were resolved through discussion until consensus was reached, or a third reviewer would take part in the discussion. The RCT was assessed according to the Cochrane Collaboration Handbook 57 . The scoring system included the following criteria: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of the results assessment, incomplete data of the results, selective reporting, and other sources of bias. Observational studies were assessed as described by McKay and colleagues 58 , including assessment of data collection (prospective versus retrospective), assignment to BA or non-Blumgart PJ group by means other than the surgeon's preference, and an explicit definition of POPF (studies were given a score of 1 for each of these areas, giving a total score of 1-4). Continuous variables were presented as the mean with corresponding standard deviations to be pooled in the meta-analysis. When  www.nature.com/scientificreports/ the trials had reported medians and ranges instead of means and standard deviations, the estimation methods were used basing on the literature 66,67 . Quantitative data was extracted from the selected studies, including population characteristics (age, gender, BMI), intraoperative conditions (type of anastomosis, pancreatic texture, mean main pancreatic diameter, operative time and intraoperative blood loss) and postoperative parameters (POPF(grade B/C), DGE, intra-abdominal abscess, bile leakage, wound infection, morbidity, mortality, reoperation, duration of drainage and postoperative hospital stay) in each study.
Statistical analysis. Data analyses were performed using Review Manager 5 software (The Cochrane Collaboration, Oxford, UK). Heterogeneity was evaluated by means of the χ2 test, with P ≤ 0.10 considered to represent a significant difference. I 2 values were used for the evaluation of statistical heterogeneity; an I 2 value of 50% or more indicated the presence of heterogeneity 68 . Initially, a fixed-effects model was used to synthesize all data. With regard to outcomes when significant heterogeneity existed across studies, sensitivity analysis was performed by sequentially omitting each study to test the influence of an individual study on pooled data. However, if there was evidence of heterogeneity among the included studies, random-effects analysis according to DerSimonian and Laird 69 was used. Clinical heterogeneity could be explained by different definitions of outcome parameters, and variability of interventions and perioperative management. The result of meta-analysis was presented as WMD or OR with 95%confidence intervals (CI). Data analysis was performed by comparing BA versus non-Blumgart PJ (including CWA, KA and invagination PJ). Funnel plots were constructed to evaluate potential publication bias, based on the grade B/C POPF.