Is pancreaticogastrostomy safer than pancreaticojejunostomy after pancreaticoduodenectomy? A meta-regression analysis of randomized clinical trials.

OBJECTIVE: To evaluate the clinically relevant POPF rate between Pancreatogastrostomy (PG) and pancreaticojejunostomy (PJ) after pancreaticoduodenectomy (PD). To evaluate the confounding factors affecting meta-analytic results.

METHODS: A systematic literature search of randomized clinical trials (RCTs) comparing PG to PJ with an International Study Group of Pancreatic Fistula (ISGPF) definition of postoperative pancreatic fistula (POPF). Risk difference (RD) and number needed to treat or harm (NNT and NNH) were used. Fixed and random-effect models were applied. Impact of confounding covariates on the meta-analytic results was evaluated using meta-regression analysis, reporting β coefficient ± standard error (SE).

RESULTS: Seven RCTs were identified involving 1184 patients: 603 PG and 581 PJ. RD in the fixed model of clinically relevant POPFs suggested that PG was superior to PJ (RD-0.07; 95% CI: -0.11 to -0.03) with an NNT of 14 (95% CI: 9 to 33). In random model, PG was not superior to PJ (RD-0.06; 95% CI: -0.13 to 0.01) with an NNT of 17 and a possibility of harm in some cases (NNH = 100). Meta-regression suggested that the increase in the proportion of "soft pancreas" in the PG arm corresponded to a more positive value of RD (β = 0.47 ± 0.19; P value: 0.045 ± 0.003).

CONCLUSION: A PG could be slightly superior to PJ in the prevention of clinically relevant POPF. The presence of high risk pancreatic remnant remains the main limitation of PG.