Robotic versus open pancreatoduodenectomy: a propensity score-matched analysis based on factors predictive of postoperative pancreatic fistula

Background Improvement in morbidity of pancreatoduodenectomy (PD) largely depends on the reduction in the incidence of clinically relevant (CR) postoperative pancreatic fistula (POPF). Methods After internal validation of the clinical risk score (CRS) of POPF, and identification of other predictive factors for POPF, robotic (RPD), and open (OPD) PDs were stratified into risk categories and matched by propensity scores. The primary endpoint of this study was incidence of CR-POPF. Secondary endpoints were 90-day morbidity and mortality, and sample size calculation for randomized controlled trials (RCT). Results No patient undergoing RPD was classified at negligible risk for POPF, and no CR-POPF occurred in 7 RPD at low risk. The matching process identified 48 and 11 pairs at intermediate and high risk for POPF, respectively. In the intermediate-risk group, RPD was associated with higher rates of CR-POPF (31.3% vs 12.5%) (p = 0.0026), with equivalent incidence of grade C POPF. In the high-risk group, CR-POPF occurred frequently, but in similar percentages, after either procedures. Starting from an unadjusted point estimate of the effect size of 1.71 (0.91-3.21), the pair-matched odds ratio for CR-POPF after RPD was 2.80 (1.01-7.78) for the intermediate-risk group, and 0.20 (0.01-4.17) for the high-risk group. Overall morbidity and mortality were equivalent in matched study groups. Sample size calculation for a non-inferiority RCT demonstrated that a total of 31,669 PDs would be required to randomize 682 patients at intermediate risk and 1852 patients at high risk. Conclusions In patients at intermediate risk, RPD is associated with higher rates of CR-POPF. Incidence of grade C POPF is similar in RPD and OPD, making overall morbidity and mortality also equivalent. A RCT, with risk stratification for POPF, would require an enormous number of patients. Implementation of an international registry could be the next step in the assessment of RPD.