The learning curves of major laparoscopic and robotic procedures in urology: a systematic review

**Background:** The concept of a learning curve, originally from aeronautics, describes the number of cases a surgical trainee needs to perform before their inexperience no longer affects outcomes. Defining these curves is critical for patient safety, training program design, and competency assessment. This systematic review aimed to provide an updated insight into the learning curves for major urological robotic and laparoscopic procedures, evaluate the methods used to analyze them, and offer recommendations for future studies.

**Methods:** The review was conducted in accordance with PRISMA guidelines and prospectively registered on PROSPERO (CRD42021251186). A systematic search of PubMed, EMBASE, and the Cochrane Library was performed from inception to December 2021, supplemented by grey literature searches. Two independent reviewers screened articles and extracted data using a predefined form. The Newcastle–Ottawa Scale was used for quality assessment. Due to substantial heterogeneity, a narrative synthesis was performed. The review included 97 eligible studies out of 3702 identified records.

**Key Results:** The included studies were predominantly retrospective (51 studies) or prospective (46 studies), with 55 studies mapping the learning curve of single surgeons. The number of procedures ranged from 20 to 27,348. For robot-assisted laparoscopic prostatectomy (RALP), the learning curve for operative time (OT) ranged from 10 to 250 cases, and for positive surgical margins (PSM) from 30 to 477 cases. For laparoscopic radical prostatectomy (LRP), OT learning curves ranged from 40 to 250 cases, and PSM from 150 to 350 cases. For robot-assisted partial nephrectomy (RAPN), the learning curve for warm ischemia time (WIT) was 4 to 150 cases, and for OT from 5 to 77 cases. For laparoscopic partial nephrectomy (LPN), the learning curve for WIT was 58 to 87 cases. For robot-assisted radical cystectomy (RARC), OT learning curves ranged from 10 to 50 cases. The learning curves for laparoscopic radical cystectomy (LRC) and robotic/laparoscopic retroperitoneal lymph node dissection (RPLND) remain undefined. Comparative studies showed that RALP and RAPN have shorter learning curves than their laparoscopic counterparts. For example, Sivaraman et al. found a learning curve of 100 cases for RALP versus 350 cases for LRP for PSM and biochemical recurrence. Hanzly et al. reported that the learning curve for WIT was reached between cases 29 and 58 for RAPN and between cases 58 and 87 for LPN.

**Clinical Implications:** The review underscores the considerable variation in outcome definitions and performance thresholds across studies, with poor reporting of confounders such as prior surgical experience. Operative time was the most commonly used metric, but its definition varied. The split-group method was most frequently used for analysis, but the cumulative sum (CUSUM) method is recommended for more precise characterization of learning curve phases and for continuous monitoring of trainee competency. The findings support the use of simulation training, mentorship, and gaining experience as a bedside assistant to reduce the learning curve before taking full control as lead surgeon. The review also highlights that RALP and RAPN have shorter learning curves for a range of metrics compared to their laparoscopic counterparts, suggesting that robot-assisted approaches may accelerate attainment of competency. Future studies should use multiple surgeons, large sample sizes, and standardized reporting to better define the currently undefined learning curves for LRC and RPLND.