Learning curve and patient outcomes in gallstone surgery: a national register-based cohort study

Abstract

Background

There is no consensus on, or a defined minimum number of surgical laparoscopic cholecystectomies required to be considered an independent surgeon. The aim of the present study was to assess how surgical outcomes improve with surgical volume.

Method

Data on laparoscopic cholecystectomies performed by surgical trainees and surgeons between 2006 and 2022 were retrieved from the Swedish national register for gallstone surgery (GallRiks). Surgeons who had not registered the number of procedures performed before first GallRiks entry were excluded. Comparisons regarding operation time and postoperatively observed complications were divided into strata of 50 performed procedures.

Results

Between 2006 and 2022, 139,010 cholecystectomies were performed by surgeons (n = 1664) who had not yet performed 2000 procedures. The mean operation time decreased from 111 min (95% confidence interval (CI) 111–112 min) to 75 min (95% CI 73–77 min) after 800 completed procedures. During the first 50 procedures, the incidence of postoperatively observed complications was 4.7 and 7.0%, respectively, if the cholecystectomy was performed with or without a registered assistant (p = 0.007). The difference in the incidence of complications for assisted and unassisted procedures remained statistically significant up to 250 interventions performed.

Conclusion

Improvement in surgical outcomes and shorter operation time were observed over a long course of the expected learning curve. Although the presence of an assistant had a significant impact on the outcome during the initial phase of the learning curve, we do not know to what extent the assistant took part in the procedure. In GallRiks, only the presence of an assistant is registered, not what the assistant did during the procedure.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s00464-026-13073-0.

Laparoscopic cholecystectomy, established as the gold standard for gallbladder surgery since the late 1980s, offers reduced hospital stays and lower complication rates, such as surgical site infections and incisional hernias, compared to open surgery [1, 2]. Despite these benefits, the procedure demands significant technical skills, and awareness of the risks of complications, such as bile duct injuries or bleeding [3]. Thus, proper training and understanding the learning curve are critical for ensuring patient safety and optimising outcomes [4]. The learning curve for general surgery trainees involves repeated practice, starting with supervised tasks and progressing to independent surgery. This curve typically features a steep initial phase of rapid improvement, followed by a slower rate of progress, eventually plateauing where further improvements are minimal [5–8].

No universally accepted standard exists for the minimum number of laparoscopic cholecystectomies required to achieve independent surgeon status or an acceptable complication rate during the learning phase. The Swedish Surgical Society recommends a minimum of 20 elective procedures for trainees, based on expert consensus [9]. Some studies suggest 20–30 procedures are needed for operation time and complication rates to stabilise, whilst others, like Williams et al. (2019), indicate 20 observations are required just to achieve a reproducible performance score amongst surgical trainees [10]. U.S. guidelines recommend at least 50 procedures during residency to ensure competence for routine and moderately complex cholecystectomy cases [11]. Higher surgical volumes are generally linked to better outcomes [12], with Swedish registry data reporting bile duct injury rates of 0.4% for elective and 0.5% for acute cholecystectomies [13–15]. Surgeons performing fewer than 10–20 procedures annually have higher complication rates compared to those performing 50–100 [16]. This study aims to explore the correlation between surgical volume and patient outcomes in a Swedish context.

Methods

The study was based on data from GallRiks (the Swedish registry for gallbladder surgery and endoscopic retrograde cholangiopancreatography (ERCP)). Ethical approval number 2022-04919-01. The registry was established in May 2005 and covers more than 90% of all cholecystectomies carried out in Sweden [17]. Variables registered in GallRiks include, amongst others, sex, age, operation indication, surgical technique, American Society of Anaesthesiologists (ASA) classification and body mass index (BMI). The timeline of each individual surgeon is traceable, even if transferring between different units in Sweden. Prior to a surgeon’s first entry, the number of procedures must be stated.

This was a retrospective observational cohort study based on prospectively collected data from GallRiks. Healthcare data were collected from the GallRiks register between 2006 and 2022 to investigate the relationship between surgical volume, operation time and patient outcomes for surgical trainees and surgeons. Abscesses included surgical site infections in the subcutaneous tissue as well as deep abscesses caused by contamination of the operative field. Bile duct injury was considered any tissue damage to the wall of any bile duct in the biliary tree detected during the procedure or postoperative period. The criterion for bleeding was any bleeding requiring blood transfusion or active intervention to stop the bleeding. Procedures were stratified by increments of 50 performed cholecystectomies to assess proficiency and outcomes. The number of procedures previously performed by the responsible surgeon was considered as exposure. This included procedures prior to the first GallRiks entry as well as all procedures included in the register. The formal position of the surgeon was not included in the analysis; only accumulated number of procedures was considered.

A multivariable logistic regression analysis was performed to study the association of the total accumulated procedure volume, the number of procedures performed in the year before each procedure, and the presence of an assistant as covariates. A multivariable linear regression analysis was performed with mean operation time as outcome and the same variables as covariates. The study included all surgeons and trainees who reported their prior procedure count at their first GallRiks registration. Exclusions applied to surgeons with 2000 or more performed cholecystectomies before data collection and those with unreported prior procedures. Both elective and acute cholecystectomies were included. The primary outcome was operation time (from first incision to last stitch, including intraoperative ERCP if performed). The secondary outcome was surgical complications. Independent factors included ASA physical status, supervised procedure status, acute cholecystitis, patient age, and procedure type (elective or acute). Procedures with multiple surgeons, including a specialist in surgery were classified as supervised. The assistant could be a trainee or a surgeon. Operation time was compared using a two-sample t-test, and surgical complications were analysed with a chi-square test. Classification trees, built with the chi-square automatic interaction detector (CHAID) algorithm, identified factors associated with surgical complications (abscess, bleeding or bile leakage) and operation time, grouping patients into distinct nodes based on similarity [18, 19].

Results

Between 2006 and 2022, 139,010 cholecystectomies were performed by 1664 surgeons who had not yet performed 2000 procedures. Patient characteristics are shown in Table 1. Evaluating the factors influencing operation time, the CHAID tree diagram identified surgical approach, acute cholecystectomy, surgical volume and the presence of an assistant as primary predictors. The first split was the surgical approach (laparoscopic cholecystectomy was faster with a mean operation time of 93 min). The second split showed that acute cholecystitis cases took longer time compared to elective cases. The third split showed that mean operation time decreased with increased surgical volume for both acute and elective cholecystectomies. (Online Appendix 1). Two histograms further show the relationship between surgical volume, operation time, and the use of assistants across varying levels of surgeon experience. The CHAID tree diagram showed that the laparoscopic approach was faster, as was an elective procedure with an assistant. Altogether, n = 6004 (4.3%) of the procedures were performed with an assistant present. The mean operation time decreased from 111 min (95% CI 111–112) to 75 min (95% CI 73–77) after 800 cholecystectomies (Fig. 1). Operation time decreased with increasing experience. This decline reflects a steeper initial learning phase (0–200 procedures, dropping from 100 to 85 min), followed by a slower improvement phase (200–800 procedures from 85 to 75 min). Beyond 800 procedures, operation time stabilised at 70–80 min indicating a plateau phase where further experience yielded minimal improvements. Variability in operation time as indicated by error bars was higher at lower surgical volumes and decreased with experience, suggesting greater consistency amongst experienced surgeons. During the first 50 procedures, the incidence of postoperatively observed complications was 4.7 and 7.0% respectively, if the cholecystectomy was performed with or without a registered assistant (p = 0.007). The difference in the complication rate for assisted and unassisted procedures remained statistically significant up to 250 procedures. The proportion of surgeries performed with or without an assistant across surgical volumes was also examined (Fig. 2). In a multivariable logistic regression analysis, the total accumulated volume of procedures (Odds Ratio [OR] 0.993, 95% confidence interval [CI] 0.991–0.994), the number of procedures performed the year before each procedure (OR 1.011, CI 1.009–1.013), presence of assisting non-specialist (OR 0.557, CI 0.427–0.727) and presence of assisting specialist (OR 0.719, CI 0.605–0.855) were significantly associated with the risk of postoperative complication (all p < 0.001). Similarly, in a multivariable linear regression analysis, the same covariates were significantly associated with operation time (all p < 0.001).

Table 1.

Patient characteristics

Fig. 1.

Mean operation time for gallstone surgery (cholecystectomy) decreases as a surgeon’s experience increases, measured by cumulative surgical volume (error bars: 95% CI)

Fig. 2.

Patient outcome (complication rate) by surgical volume and with or without assistance from another surgeon (error bars: 95% CI)

In subgroup analysis of procedures performed for acute cholecystitis without stratification for presence of an assistant, it was found that the complication rate decreased over time and that approximately 600 procedures were required to reach a complication rate not higher than 3.0%.

At lower surgical volumes (0–199 procedures), most surgeries were performed without an assistant with a complication rate ranging from 2.5 to 3.5% compared to 1 to 2% with an assistant. Assistant presence peaked at 350–399 procedures, reaching approximately 3%, nearly matching the proportion without an assistant around 3.5%. This peak corresponds to the slower improvement phase in operation time (200–500 procedures), suggesting that assistants may facilitate efficiency gains during this transitional period. Beyond 500 procedures, presence of assistant was around 1.5 to 2.5%, whilst the proportions without an assistant remained slightly higher at 2 to 3%, indicating that experienced surgeons may operate more independently once they reach peak efficiency.

Discussion

This population-based cohort study demonstrates that the learning curve for surgeons performing gallstone surgery can be quantified by operation time and complication rate. The results demonstrate a significant reduction in operation time, dropping from an average of 111 min to 75 min after 800 completed laparoscopic cholecystectomies. This indicates a prolonged learning curve, extending well beyond the initial training phase and into years of independent practice. Proficiency, as measured by operation time, stabilises only after a high volume of procedures. This contrasts with prior studies suggesting sufficient proficiency after 20–50 cases [20]. This discrepancy may reflect differences in how proficiency is defined—earlier studies often focussed on basic competence or complication rates, whereas this study emphasised efficiency (operation time) as a marker of mastery. The extended learning curve implies that surgical trainees and early—career surgeons may require more extensive experience than current Swedish surgical trainee guidelines (minimum of 20 procedures) anticipate reaching peak efficiency. The guideline recommendation of a minimum of 20 procedures excludes acute cholecystitis. As expected, the study highlights that operative time was longer in cases involving open cholecystectomy, acute cholecystitis, or the absence of an assistant. These findings align with clinical expectations, as open procedures and acute cases are inherently more complex, and the lack of assistance may slow workflow or increase the cognitive load on the primary surgeon [17]. The persistence of longer operation times in unassisted procedures up to 250 procedures suggests that the presence of a second surgeon or assistant plays a critical role in streamlining the procedure during the early-to-mid stages of a surgeon’s career. Not only the totally accumulated volume had an impact on the outcome, but also the intensity of training. A high number of procedures performed prior to the procedure significantly predicted the outcome, suggesting that concentrating training within a limited period has a greater impact on progress than spreading it over a longer period. The incidence of postoperative complications was notably higher during the first 50 procedures (4.7% with assistance vs. 7.0% without, p = 0.007), with a statistically significant difference persisting up to 250 procedures. The multivariable logistic regression analysis showed that the experience of the assistant influenced the outcome. The risk of complications was, however, greater in case an experienced colleague assisted. This is probably due to a selection bias, as more experienced surgeons may have been called for in case a complicated situation occurred. This underscores the importance of having an assistant/second surgeon during the early phases of a surgeon’s development. Laparoscopic cholecystectomy for acute cholecystitis is considered a procedure with longer learning curve than for planned laparoscopic cholecystectomy in case of uncomplicated gallstone disease. In the present cohort, it was found that 600 procedures were required to reach an acceptable complication rate in case of acute cholecystitis. Studies with more granulated data are needed to confirm this. The higher complication rate in unassisted procedures aligns with previous research linking lower surgical volume and lack of experience to adverse outcomes, such as bile duct injuries. Whilst the complication rate decline was less dramatic than the reduction in operative time, the sustained benefit of assistance up to 250 procedures suggests that independent practice should be deferred until a surgeon has substantial supervised experience. Given these findings, the Swedish Surgical Society’s recommendation of a minimum of 20 elective laparoscopic cholecystectomies for trainees appears insufficient. The U.S guideline of 50 procedures aligns more closely with the point at which complication rates begin to stabilise, though even this may not suffice for achieving efficiency, as operation time plateaus at 800 procedures. High-volume surgeons with about 50–100 procedures annually consistently show better outcomes in prior studies, and this study supports that trend, with complication rates and operative times improving with greater experience. However, the plateau at 800 procedures suggests that true mastery may require a decade or more of practice, especially in lower volume settings, which include most surgical departments in Sweden performing cholecystectomies. The considered high-volume centres in Sweden estimate around 50–80 cases per surgeon annually. Data from this study suggest that surgical trainees or surgeons with fewer than 50 laparoscopic cholecystectomy procedures should not perform these operations without an assistant. This raises questions regarding the safe minimum number of procedures annually per surgeon without affecting patient outcomes. These findings have significant implications for surgical education and healthcare policy. First, the prolonged learning curve suggests that training programmes should prioritise access to higher surgical volumes, potentially through longer uninterrupted rotation at a high-volume centre when learning cholecystectomies, starting with elective cases. This could include facilitating mandatory training time in simulation-based environments. The protective effect of having an assistant up to 250 procedures calls for structured mentorship programmes for early-career surgeons, ensuring they are not thrust into independent practice prematurely. Moreover, the correlation between volume and outcomes supports the designation of high-volume centres for complex cases, though this presents a challenge in balancing a broader general surgery concept against sub specialisation.

The study’s strength is its substantial sample size (139,010 procedures) and extensive observation period (2006–2022), both derived from a comprehensive national registry. This enhances the reliability and generalisability of the findings within Sweden. The use of classification trees to identify factors like acute cholecystitis and supervision adds nuance to the analysis. However, limitations include the retrospective design, which may miss unrecorded confounders. Potential sources of residual confounding are surgeon fatigue or patient comorbidities not captured in ASA scores.

Although the presence of an assistant had a significant impact on the outcome during the initial phase of the learning curve, we do not know to what extent the assistant took part in the procedure. In GallRiks, only the presence of an assistant is registered, not what the assistant did during the procedure. The criteria for complications did not unequivocally reflect unfavourable outcomes directly related to the surgical performance. Some of the complications outcomes may have been more related to patient factors or to anaesthesia. The variables registered in GallRiks are defined to focus on patient-related outcome, not the surgical success. This may have contaminated the results.

Furthermore, the inclusion of intraoperative ERCP for patients presenting with common bile duct stones contributed to a longer total operation time. Nevertheless, due to the substantial volume of procedures, this factor is not anticipated to materially impact the overall findings. Additionally, the focus on operation time as a surrogate for the learning curve might be a rough estimate of the needed surgical volume. Hence, patient outcomes in terms of complications should be considered with other outcomes like readmission rates and quality of life, providing a better overall representation of surgical success. However, in this study complication rate is considered the most clinically relevant outcome measure when determining the readiness for independent practice, but this is influenced by numerous factors other than the experience of the surgeon. Nevertheless, by defining a reduction of the complication rate to a level equal to that of procedures completed by surgeons unaccompanied by an assistant, 250 procedures are needed to acquire sufficient safety to give a general recommendation of the number of procedures needed to pass the learning curve.

Future studies should explore whether simulation training can shorten the learning curve, particularly for operative efficiency in cholecystectomies, since simulation training is not mandatory in the Swedish trainee curriculum. It is suggested that it be included in the trainee curriculum as something trainees need to practise regularly throughout their postgraduate education. It is already well established that practising a skill in simulations makes for more efficient performance [21–27]. Furthermore, investigating the impact of surgeon-specific factors like prior experience with related procedures and patient-specific factors, e.g., obesity or comorbidities, could refine the classification tree model. Comparative studies across countries with differing training structures and surgical volumes could clarify whether Sweden’s prolonged learning curve is unique.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open access funding provided by Karolinska Institute. All authors have no funding to declare.

Declarations

Disclosures

Nathalie Young, Lars Enochsson, Hans Pettersson, Peter Elbe, Ninos Oussi, and Gabriel Sandblom have no conflicts of interest or financial ties to disclose.