Level 1 versus experiential evidence for mitigating pancreatic fistula following pancreatoduodenectomy: Establishing a foundation for future progress.

Abstract

Clinically relevant postoperative pancreatic fistula remains the dominant problem following pancreatoduodenectomy. Level 1 evidence, accrued over 35 years, now includes dozens of randomized controlled trials and meta-analyses, which assess myriad mitigation techniques. Their conclusions are not generally reflected in actual adoption into clinical practice; meanwhile, experiential data might enable more comprehensive evaluation of mitigation. A conundrum arises: how can surgeons reconcile these contrasting sources to effect best practice? Level 1 evidence studies were collected from the International Study Group of Pancreatic Surgery Evidence Map of Pancreatic Surgery on 7 predominant fistula mitigation techniques. The Pancreas Fistula Study Group cohort provided an experiential comparison. Evidence sources were qualitatively evaluated with a novel "mitigation value proposition" construct. Randomized controlled trial cohorts and likewise-treated Pancreas Fistula Study Group patients were directly compared. Thirty-seven randomized controlled trials and 43 meta-analyses were analyzed. Randomized controlled trial patient inclusion yields were low, just 39.6%, and substantial discord was present among trial findings for 6 of 7 techniques. Risk stratification was absent from 22 of 37 trials. Meta-analysis conclusions were more uniform than those of randomized controlled trials. Rates of clinically relevant postoperative pancreatic fistula often differed significantly between patients in the Pancreas Fistula Study Group and those enrolled in randomized controlled trials. Experiential evidence generally agreed with level 1 evidence regarding stents, drains, and early drain removal but disagreed regarding pancreatogastrostomy and somatostatin analogs. Tension currently exists between level 1 data and experiential evidence. Significant drawbacks surface when assessing the full corpus of extant level 1 studies. Risk stratification addresses bias in experiential data and unlocks improved assessment of mitigation, thereby allowing nuanced, patient-tailored treatment. Lessons here exposed can direct improved trial design.