Cumulative sum analysis after the initial learning curve for laparoscopic pancreaticoduodenectomy: Does one continue to learn?

McKean, Jordan A; Hamilton, Jakob S; Tushoski-Aleman, Gerik; Thompson, Grace; Herremans, Kelly; Han, Song; Nassour, Ibrahim; Behrns, Kevin E et al. · Surgery · 2025

retrospective_cohort · Level III

Where this comes from

Abstract

In this study, the cumulative sum analysis was used to examine the learning curve of laparoscopic and open pancreaticoduodenectomy, excluding the first 100 cases. We hypothesized cumulative sum analysis would demonstrate continued evolution after the first learning curve. Cumulative sum analysis is an effective tool to understand the evolution of an individual surgeon or surgical procedure over an experience. The analysis represents a continuous summation of the differences between individual data points of a performance variable and the variable's overall mean. Most studies use operative time and analyze the early experience (ie, first 100 cases). This is a retrospective, single-institution study of consecutive patients from 2012 to 2023. The learning curve was evaluated not only for operative time, but also conversion rate, postoperative pancreatic fistula rate, blood loss, and length of stay using the cumulative sum method. This study included 381 patients with 324 (85%) planned laparoscopic and only 57 (15%) planned open surgeries. Of the laparoscopic cases, 29% were converted to open. The median operative time was 277 minutes, median estimated blood loss was 350 mL, and median length of stay was 8 days. Learning curves are separated into 3 phases: initial learning phase, consolidation phase, and proficiency phase. Operative time (P ≤ .00001), estimated blood loss (P ≤ .00001), length of stay (P ≤ .00001), and postoperative pancreatic fistula rate (P ≤ .0001) demonstrated improved outcomes across all 3 phases. Improved results continued to evolve, and proficiency was observed between 313 and 373 cases for operative time, 159 and 294 cases for estimated blood loss, 155 and 237 cases for length of stay, and 171 and 250 for postoperative pancreatic fistula. The conversion rate declined from phase 1 to 2 (P = .0104) but then increased from phase 2 to 3 (P = .0013). In a large experience of laparoscopic pancreaticoduodenectomies in which cumulative sum analysis is applied, we demonstrate that mastery continues to evolve past the initial learning curve.

Medical subject headings