Learning curve for pancreatoduodenectomy: can it be generalized?

Abstract

We hypothesized that a learning curve exists for pancreatoduodenectomy (PD) and that the factors that show a learning curve and the degree of improvement over time differ among surgeons. We analysed the outcomes of PD in 300 patients operated by two surgeons between 2001 and 2014; these comprised the first 150 patients operated by each surgeon. Patients operated by each surgeon were classified into three groups according to the chronological sequence of operation (50 patients per group). For surgeon A, the median operation time (496 versus 454 versus 418 min, P < 0.001) and the mean amount of red blood cell transfusion (3.76 versus 2.76 versus 1.26 packs, P < 0.001) showed a slight but progressive decrease from the first 50 cases to the second and the third 50 cases. For surgeon B, only the operation time, which decreased and ultimately plateaued at approximately case 100, was significantly different among the groups. There was no improvement in post-operative complications, re-operation rates, mortality, length of stay and re-admission rates over time for both surgeons. We demonstrated that a learning curve for PD exists with respect to various factors. However, the factors for which learning curves were exhibited and the degree of improvement over time were different between the two surgeons.