Abstract
Objective: To evaluate the overall and step-specific learning curves of extraperitoneal single-port robot-assisted radical prostatectomy (SP-RARP) using the da Vinci SP system during the initial adoption phase. Methods: This retrospective observational study included consecutive patients with localized prostate cancer who underwent extraperitoneal SP-RARP at Saitama Medical University International Medical Center between April 2025 and March 2026. All procedures were performed by a single surgeon. The operation was subdivided into nine predefined surgical steps. Learning curves were constructed using console time, step-specific operative time, and estimated blood loss. Cumulative sum analysis and segmented regression were used to identify breakpoints. Results: A total of 31 consecutive patients were included. The median age was 72.8 y (IQR, 68.5–75.6 y), the median body mass index was 23.5 kg/m2 (IQR, 21.3–24.9 kg/m2), and the median prostate-specific antigen level was 10.6 ng/mL (IQR, 6.5–16.2 ng/mL). According to National Comprehensive Cancer Network criteria, 19 patients (61.3%) had biopsy grade group 4 or 5, and 19 (61.3%) had high- or very-high-risk disease. The median console time was 162 min (IQR, 153–184 min), with a breakpoint at approximately cases 16–17, whereas the estimated blood loss stabilized earlier at approximately cases 3–4. Step-specific analyses demonstrated heterogeneous maturation. Early stabilization was observed in the endopelvic fascia incision, dorsal venous complex control, and posterior reconstruction; intermediate stabilization in the vesicourethral anastomosis, anterior reconstruction, rectum-prostate dissection, and vas deferens/seminal vesicle dissection; and delayed stabilization in the development of the Retzius space. Bladder neck dissection showed no clear breakpoint. Conclusion: Extraperitoneal SP-RARP improved progressively during the initial adoption phase, but technical maturation was not uniform across surgical steps. Step-specific analysis provided a more detailed understanding of procedural learning than overall operative metrics alone did.