Is surgery a never‐ending learning process?

Abstract

The concept of the learning curve is one of the most important issues in surgery and also one of the most overlooked. In the present issue of BJUI, Abboudi et al. 1 present an interesting review paper evaluating the concept of the learning curve in urological procedures. Specifically, the authors have conducted a methodologically consistent systematic review on the literature focused on the learning curve of some urological procedures, including mainly radical prostatectomy (RP), robot-assisted partial nephrectomy (RAPN) and percutaneous nephrolitotomy 1. Surprisingly, nothing was available for BPH treatments, which are among the most prevalent urological procedures. Most of the studies are focused on robot-assisted RP (RARP), but the available literature is of poor methodological quality, including mainly surgical series evaluating a limited number of surgeons, with a heterogeneous selection of outcomes from which to study the learning curve and a focus on short-term outcomes. Conversely, the literature on retropubic RP or laparoscopic RP is of higher quality, including a few very large multi-institutional studies encompassing the performances of several surgeons (reference nos. 24, 26, 29 and 30 in the review) and adopting sophisticated statistical methodology; however, the current interest for these procedures is quite limited, RARP being more commonly preferred. With the above-mentioned limitations in mind, what we have learnt is that RARP operating time plateaus after 50–200 cases, positive surgical margin (PSM) rates after 50–1600 cases, and continence and potency after 200 cases 1. Such data are only partially in line with the findings of a recent prospective Australian study 2, not included in the present systematic review, which evaluated the learning curve with RARP of a high-volume open surgeon (>3000 retropubic RPs performed before the study beginning). In that study, Thompson et al. 2 demonstrated that performances with RARP surpassed those with retropubic RP after ∼100 cases for sexual function scores and PSM rates in pT2 cancers, whereas ∼150 cases were needed to reach the same target with urinary function scores. Moreover, RARP performances kept on improving, with sexual and urinary scores plateauing after 600–700 and 700–800 cases, respectively. Similarly, with regard to PSMs, it was demonstrated that PSM rates in pT2 and pT3–4 cancers plateaued after 400–500 and 200–300 cases, respectively 2. Although improvement is likely, it is not clear how much these performances might improve with further extension of the caseload. Taken together, those data suggest that even with robotic assistance, a high volume of cases is strongly associated with improving oncological and functional outcomes after RARP. This is not an extraordinarily original concept, but implies that the daVinci platform, by itself, cannot guarantee excellent surgical quality and that the relevance of the surgeon is as high as ever. Limited data are available on other major robotic procedures, such as RAPN and robot-assisted radical cystectomy (RARC). Specifically, 20–75 cases are thought to be needed to observe a plateau in warm ischemia time (WIT) during RAPN, which is in line with our previous findings demonstrating a continuous decrease in WIT during the first 50 cases 3. Similarly, 20–30 cases are supposed to be needed to achieve acceptable operating times, lymph node yields and PSM rates after RARC; however, those findings do not take in account the burden of robotic experience achieved with RARP before embarking in RARC, which is clearly a major issue 4. Considering that the improvements in performances along the learning curve exceeded any effect sizes we might reasonably expect from a novel drug 5, it is clear that any attempt to centralise treatments for complex procedures in high-volume centres with high-volume surgeons should be attempted. Obviously, that is a very critical target, which is hard to achieve in many realities. In parallel,