Learning curve of robotic-assisted transabdominal preperitoneal inguinal hernia repair (r-TAPP): a scoping review of CUSUM-based studies

Robotic-assisted transabdominal preperitoneal repair (r-TAPP) is increasingly adopted for inguinal hernia repair, but the number of procedures required to achieve stable operative performance remains unclear. This scoping review synthesised CUSUM-based learning curve evidence to characterise reported operative-time stabilisation thresholds and their methodological determinants. PubMed, Embase, Scopus, Cochrane Library, and Google Scholar were searched for studies published between January 2000 and December 2025 reporting CUSUM-based learning curve analyses of r-TAPP or SPTAPP using Da Vinci systems. The review followed PRISMA-ScR guidance and was registered with the Open Science Framework. The primary outcome was the CUSUM inflection point, summarised descriptively using median, samplesize-weighted mean, range, and non-parametric bootstrap 95% confidence intervals. Operative time trends were synthesised narratively without pooling. Seven studies, including eight independent learning curves and 1,361 patients, were included. CUSUM inflection points ranged from 12 to 138 procedures (median 32, 95% CI 13–43; weighted mean 65, 95% CI 22–108). Sensitivity analysis excluding the single series using a composite efficiency-plus-safety CUSUM yielded a median of 29 procedures (95% CI 13–35) and weighted mean of 28 (95% CI 19–36). Both estimates are valid within their methodological contexts. Operative time decreased after stabilisation in all studies by approximately 7–24 min. Complications were predominantly minor; intraoperative events were rare, and reported conversions were absent. Safety comparisons between phases were underpowered. Available evidence is compatible with r-TAPP operative-time stabilisation after approximately 25–35 procedures in structured training settings. This estimate is hypothesis-generating, not a credentialing benchmark, given heterogeneous CUSUM definitions and very low-certainty evidence.