Learning curves in minimally invasive pancreatic surgery: a systematic review

Abstract

Background: The learning curve of new surgical procedures has implications for the education, evaluation and subsequent adoption. There is currently no standardised surgical training for those willing to make their first attempts at minimally invasive pancreatic surgery. This study aims to ascertain the learning curve in minimally invasive pancreatic surgery.

Methods: A systematic search of PubMed, Embase and Web of Science was performed up to March 2021. Studies investigating the number of cases needed to achieve author-declared competency in minimally invasive pancreatic surgery were included.

Results: In total, 31 original studies fulfilled the inclusion criteria with 2682 patient outcomes being analysed. From these studies, the median learning curve for distal pancreatectomy was reported to have been achieved in 17 cases (10-30) and 23.5 cases (7-40) for laparoscopic and robotic approach respectively. The median learning curve for pancreaticoduodenectomy was reported to have been achieved at 30 cases (4-60) and 36.5 cases (20-80) for a laparoscopic and robotic approach respectively. Mean operative times and estimated blood loss improved in all four surgical procedural groups. Heterogeneity was demonstrated when factoring in the level of surgeon's experience and patient's demographic.

Conclusions: There is currently no gold standard in the evaluation of a learning curve. As a result, derivations are difficult to utilise clinically. Existing literature can serve as a guide for current trainees. More work needs to be done to standardise learning curve assessment in a patient-centred manner.

Keywords: Laparoscopic; Learning curve; Minimally invasive surgery; Pancreatic surgery; Robotic; Training.

Fig. 1

PRISMA flow diagram of study selection

Fig. 2

Individual learning curve plots for different modalities of DP and PD

Fig. 3

A Changes in operative time before and after the learning curve was achieved in LDP (n = 3, independent t-test, p = 0.307) and RDP (n = 2, independent t-test, p = 0.180). B Changes in operative time before and after the learning curve was achieved in LPD (n = 9, independent t-test p = 0.108) and RPD (n = 3, independent t-test, p = 0.095)

Fig. 4

A Changes in estimated blood loss before and after the learning curve was achieved in LDP (n = 2, independent t-test, p < 0.05) and RDP (n = 2, independent t-test, p = 0.210). B Changes in estimated blood loss before and after the learning curve was achieved in LPD (n = 7, independent t-test, p < 0.05) and RPD (n = 4, independent t-test, p = 0.154). * = p < 0.05