Lars Enochsson

BACKGROUND Virtual reality (VR) simulators record objective metrics such as task time, path length and errors, and programs across surgical specialties increasingly consider using them for progression and credentialing. Evidence that VR training improves operative performance does not establish that an individual's simulator score predicts how that person will operate on patients. OBJECTIVE This systematic review aimed to determine whether objective VR simulator metrics predict externally assessed operating room (OR) performance of surgeons and surgical trainees across surgical specialties. METHODS We searched PubMed/MEDLINE, Scopus and Web of Science from inception to 15 September 2026 and searched citations. Eligible English-language studies enrolled surgeons or surgical trainees in any surgical specialty and reported an individual-level association between a VR simulator metric and rated or timed performance during surgery on human patients, or a patient outcome of that surgery. Records excluded under an earlier general surgery-only definition were rescreened. Risk of bias was appraised with the Quality In Prognosis Studies (QUIPS) tool, and findings were synthesized narratively following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020. The review was not prospectively registered. RESULTS Of 2902 records screened and 170 full texts assessed, 23 studies with 360 analyzed participants were included (one study did not report its sample size): seven in general surgery, seven in urology, five in ophthalmology, two in gynecology, one in otolaryngology and one of unknown specialty. Associations were predominantly supportive in 13 studies, mixed in four, null or inverse in five and equivocal in one. Four of five ophthalmology studies were supportive: Eyesi scores correlated with masked ratings of cataract surgery (r=0.65; P=.003), and in one program first-year scores correlated with third-year operative ratings (r=0.81; P=.002). In urology, the strongest correlations came from samples that pooled trainees and experts (rho=0.81; P<.001), whereas correlations within experts were weak (rho=0.37; P=.18) and absent in the study of unknown specialty (r=-0.03). Two prostatectomy studies related VR suturing metrics to continence, in both directions. In general surgery and gynecology, associations were weak, inconsistent or inverse. Risk of bias was moderate in 11 studies and high in 12, and nine were appraised from abstracts alone. CONCLUSIONS VR metrics correlate with operative performance most clearly when samples span a wide range of experience, when the same instrument scores both performances and when the simulated task reproduces the operation. These conditions overstate its ability to discriminate among trainees at a similar stage. The evidence is small, mostly single-center and largely concurrent. A stand-alone VR score should not be used for high-stakes competency or credentialing decisions; VR assessment belongs within multimodal, longitudinal assessment. CLINICALTRIAL The review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement; the completed checklist is provided in Multimedia Appendix 1. It was not registered prospectively with the International Prospective Register of Systematic Reviews (PROSPERO), and no separate protocol was prepared.

BACKGROUND General surgery residency and specialty-training programs face record application volumes, evolving assessment policies (including the exclusion of the US Medical Licensing Examination [USMLE] Step 1 numeric scores, and the rapid emergence of large language models (LLMs) used by both applicants and reviewers. Artificial intelligence (AI) tools have been proposed for applicant screening, personal statement evaluation, letter-of-recommendation analysis, and detection of AI-generated text, raising both efficiency promises and fairness accountability requirements. OBJECTIVE This narrative review synthesizes peer-reviewed evidence published since 2020 on AI applications in general surgery residency selection and to derive implementation recommendations for general surgery programs. METHODS We conducted a targeted, proportionate narrative evidence review of literature from 2020 onward, using keyword concepts spanning artificial intelligence, machine learning, and residency or specialty-training selection, supplemented by hand-searching of reference lists. General surgery was the primary focus with supplementary inclusion of related surgical specialties. Two reviewers independently screened records for relevance and extracted data on sample size, AI method, human comparator, concordance and agreement metrics, predictive performance, selection-criteria weights, and fairness findings. The review was not registered and was not designed as a systematic review. RESULTS Fifteen studies applying AI or machine learning methods to a residency- or fellowship-selection task met inclusion criteria; a seventeenth human-only holistic-review study is discussed as contextual comparator. Five studies provided direct general-surgery evidence and ten addressed adjacent surgical or medical specialties (otolaryngology, orthopedic surgery, plastic surgery, obstetrics and gynecology, and emergency medicine). Three findings were convergent across the evidence base. First, AI-based applicant screening can shift cohort composition relative to human review, with only 7% overlap between AI and program-director selections in the single available general-surgery comparison. Second, large language models achieve only slight agreement with human reviewers when scoring personal statements (weighted kappa 0.12-0.18) and tend to undervalue implicit leadership and resilience narratives. Third, human reviewers cannot reliably distinguish AI-generated from human-written personal statements and letters of recommendation, a pattern replicated across general surgery and adjacent specialties, and perceived authorship influenced interview-offer decisions in at least one study. CONCLUSIONS Direct evidence on AI in general surgery residency selection is limited but consistently shows that AI alters selection cohorts, produces low agreement with humans on personal statement scoring, and inherits historical decision patterns. ACGME-accredited programs should treat AI as decision support behind program director accountability, with local validation, prespecified fairness audits, screen-in-only deployment, documented data governance, and post-match outcome monitoring.

Background: Endoscopic retrograde cholangiopancreatography (ERCP) for common bile duct stone (CBDS) removal can be performed under varying levels of sedation. Deeper sedation has been associated with more successful cannulation, but little is known about the effectiveness of ERCP, measured as rate of retained CBDS, depending on the level of sedation. Methods: In this observational population-based cohort study data on 121 252 ERCP procedures were retrieved from the Swedish Register for Gallstone Surgery and ERCP (Gallriks). Adjusted odds ratios (OR) for retained CBDS 12 months after CBDS removal were estimated. Retained CBDS was defined as reintervention with ERCP finding a CBDS. Results: General endotracheal anaesthesia (GEA) was associated with 27% lower odds for retained CBDS compared to light endoscopist-directed conscious sedation (EDCS) after adjusting for confounding factors. Adjusted OR for GEA was 0.73 (CI 0.66–0.80), for anaesthetist-directed propofol sedation (monitored anaesthesia care) (ADPS (MAC)) 0.84 (CI 0.76–0.93), and for endoscopist-directed propofol sedation (EDPS) 0.83 (CI 0.72–0.95). Pairwise comparison revealed a 13% (OR 0.87 CI 0.77–0.99) reduction of odds in favour of GEA compared to ADPS (MAC), but no significant difference compared to EDPS. Conclusion: An association was observed between deeper sedation and lower odds of retained CBDS compared to EDCS, with a marginal benefit of GEA over ADPS (MAC). Safety outcomes were not assessed, and findings are subject to unmeasured confounding why no causal conclusion can be made.

Background: Acute cholecystitis is a common complication of gallstone disease. Although early laparoscopic cholecystectomy is recommended, some patients do not undergo early surgery and remain at risk of recurrent disease. This study investigated whether early cholecystectomy for recurrent cholecystitis is associated with higher complication rates versus first-time cholecystitis. Methods: A retrospective population-based cohort study was conducted using data from the Swedish Registry of Gallstone Surgery. Patients undergoing early cholecystectomy for acute cholecystitis in Sweden between 1 January 2006, and 31 December 2020, were included. Patients with recurrent cholecystitis were compared to those with a first episode. The primary outcome was the total 30-day complication rate. Secondary outcomes included open surgery, prolonged surgery (≥ 120 minutes), bile duct injury, and specific complications such as intestinal injury, bleeding, reoperation, abscess, and 30-day mortality. Multivariable logistic regression was used to calculate odds ratios (OR), adjusting for age, sex, American Society of Anesthesiologists (ASA) grade, and time from admission to surgery as confounders. Results: Among 34 925 patients, 3384 had recurrent cholecystitis and 31 541 had first-time cholecystitis. The recurrent cholecystitis group had a higher complication rate (20.2 versus 13.8%) and an increased risk of bile duct injury (OR 2.44; 95% confidence interval (c.i.) 1.67 to 3.56), intestinal perforation (OR 2.54; 95% c.i. 1.51 to 4.25), prolonged surgery (OR 1.64; 95% c.i. 1.53 to 1.67), and open surgery (OR 1.76; 95% c.i. 1.64 to 1.92). However, patients with recurrent cholecystitis were older and had a higher ASA grade. Conclusion: Early cholecystectomy for recurrent cholecystitis is associated with increased complication rates compared with first-time cholecystitis. These findings support early surgical intervention during the first episode to reduce the risk of adverse outcomes associated with recurrent disease.

Background and aims: Whereas the value of endoscopic retrograde cholangiopancreatography (ERCP) training in clinical practice is well known, the impact on stress markers and performance in a virtual reality (VR) simulator is not. The primary aim of the study was to see how the number of clinical ERCPs performed during a 1-year period influenced VR-ERCP performance. A secondary aim was to compare differences in salivary stress marker levels, between the first and final simulator attempts. Methods: Thirty-one endoscopists completed three VR-ERCP procedures of increasing difficulty. The times taken to complete the different steps of the procedures were recorded. Saliva chromogranin A, cortisol, and α-amylase were measured before and after each phase of the cystic leakage procedure. Participants then did 1 year of clinical ERCP training at their respective centers. The remaining cohort (26/31) was divided into two subgroups according to their level of clinical training. They then completed the same VR-ERCP procedures. Differences in time before and after each phase as well as stress marker levels during the cystic leakage procedure were assessed. Results: Those with >100 ERCPs of clinical training improved times to completion of all 15 phases in the VR-ERCP procedures (p < 0.05) in contrast to the group with 20–50 ERCPs who only improved in 11/15. Differences in increases in salivary stress marker levels of chromogranin A before and after each phase of the cystic leakage procedure, adjusted for number of ERCPs, showed significant reductions in four of the five phases measured. Conclusion: Clinical ERCP training enhances subsequent performance in terms of time to completion in a VR-ERCP simulator. Additional intended use of simulators could be used as a benchmark for clinical progress. Saliva markers may be feasible to use in measuring stress reactions in a training setting.

Background: Previous research has suggested that female surgeons achieve better outcomes in various procedures, but the underlying reasons remain unclear. This study aimed to explore potential explanations for observed gender disparities in cholecystectomy outcomes. Methods: A follow-up analysis was conducted using a previously reported population-based cohort from the Swedish Registry of Gallstone Surgery. All patients who underwent cholecystectomy in Sweden between 2007 and 2019 were included, with a subgroup (2017–2019) analysed for case mix. The association between surgeon gender and outcomes was assessed, considering caseload, case mix, team composition, and surgeon–patient gender dynamics. Results: The analysis included 150 509 cholecystectomies performed by 2555 surgeons (849 women, 1706 men). Female surgeons had fewer complications across all experience categories. The gender disparity in bile duct injuries was most notable among male surgeons with < 3 years of experience (risk ratio 1.77, 95% confidence interval 1.30 to 2.40). After adjusting for case mix, male surgeons had higher risks of surgical complications (risk ratio 1.28, 95% confidence interval 1.12 to 1.45), bile duct injuries (risk ratio 1.63, 95% confidence interval 1.01 to 2.62), and total complications (risk ratio 1.10, 95% confidence interval 1.03 to 1.19), but shorter operating times. Teams with only male surgeons had the highest complication rates. Women experienced the best outcomes when treated by female surgeons. Conclusion: Female surgeons were associated with more favourable cholecystectomy outcomes. These differences persisted after accounting for caseload, case mix, and team composition. Inexperienced male surgeons had a higher risk of bile duct injury, underscoring the importance of rigorous mentorship during the early years of practice.

Background: Stress during the early ERCP learning curve may interfere with acquisition of skills during training. The purpose of this study was to compare stress biomarkers in the saliva of trainees before and after familiarisation with ERCP exercises on a virtual simulator.Methods: Altogether 26 endoscopists under training, 14 women and 12 men, completed the three phases of this study: Phase 1. Three different ERCP procedures were performed on the simulator. Saliva for α-amylase (sAA), Chromogranin A (sCgA), and Cortisol (sC) were collected before (baseline), halfway through the exercise (ex.), and 10 min after completion of the exercise (comp.); Phase 2. A three-week familiarisation period where at least 30 different cases were performed on the virtual ERCP simulator; and Phase 3. Identical to Phase 1 where saliva samples were once again collected at baseline, during, and after the exercise. Percentage differences in biomarker levels between baseline and exercise (Diff_ex) and between baseline and completion (Diff_comp) during Phase 1 and Phase 3 were calculated for each stress marker.Results: Mean % changes, Diff_ex and Diff_comp, were significantly positive (p < 0.05) for all markers in both Phase 1 and Phase 3. Diff_ex in Phase 1 was significantly greater than Diff_ex in Phase 3 (p < 0.05) for sAA and sCgA. Diff_comp for sAA in Phase 1 was significantly greater than Diff_comp in Phase 3 (p < 0.05). No significant differences were found in sC concentration between Phases 1 and 3.Conclusion: This study shows that familiarisation with the ERCP simulator greatly reduced stress as measured by the three saliva stress biomarkers used with sAA being the best. It also suggests that familiarisation with an ERCP simulator might reduce stress in the clinical setting.

Background: There is a lack of evidence regarding the effectiveness of virtual simulators as a means to acquire hands-on exposure to endoscopic retrograde cholangiopancreatography (ERCP). The present study aimed to assess the outcome and construct validity of virtual ERCP when training on the GI II Mentor simulator.Methods: A group of seven experienced endoscopists were compared with 31 novices. After a short introduction, they were requested to carry out three virtual ERCP procedures: diagnosing and removing a common bile duct (CBD) stone; diagnosing and taking brush cytology from a hilar stenosis; and, finally, diagnosing and treating a cystic leakage with a BD stent. For each task, the total time required to complete the task, time required to correctly view the papilla, total time of irradiation, time to deep cannulation, time to define diagnosis, time to complete sphincterotomy, and time to complete the respective intervention were measured. Cannulation of the BD, correct diagnosis, sphincterotomy, and time to complete intervention were assessed by an assessor blinded to the status of the endoscopist who performed the virtual ERCP.Results: The time required to visualize the papilla and to cannulate deeply when removing the BD stone was significantly shorter for the experts (both p < 0.05). The time to visualize the papilla, cannulate deeply, reach a diagnosis, complete sphincterotomy, and complete the intervention was significantly shorter for the experts when managing cystic leakage (all p < 0.05). In diagnosing and taking brush cytology from a hilar stenosis, there was only a trend toward the experts needing less time for the deep cannulation of the BD (p = 0.077).Conclusion: The performance differed between experts and novices, especially in the management of cystic leakage. This corroborates the construct validity of the GI II Mentor simulator.