P APER OF THE 21ST ANNUAL ESA MEETING “Surgeons’ Intuition” Versus “Prognostic Models” Predicting the Risk of Liver Resections Olivier Farges, MD, PhD, ∗ Eric Vibert, MD, PhD,† Cyril Cosse, BSc,‡ Franc ¸ois Ren´ e Pruvot, MD,§ Yves Patrice Le Treut, MD,¶ Olivier Scatton, MD,‖ Christophe Laurent, MD, ∗∗ Jean Yves Mabrut, MD,†† Jean-Mac Regimbeau, MD,‡ Mustapha Adham, MD,‡‡ Bruno Falissard, MD, PhD,§§ and Emmanuel Boleslawski, MD§ Objectives: Analyze surgeons’ anticipation of the risk of hepatectomy. Background: Risk prediction enables adequate counseling and improves safety. Models are available that predict postoperative morbidity and length of stay (LOS), but their performance is ill-defined. Surgeons’ ability to predict these endpoints is unknown. Methods: This prospectively designed, multicenter trial included all adult pa- tients undergoing elective hepatectomy. Primary endpoints were 90-day mor- bidity and mortality and LOS. Explanatory variables included (i) “surgeons’ intuition” (surgeons’ anticipation) of the difficulty of the procedure, postop- erative morbidity, and LOS and (ii) “prognostic models” (models based on objective clinic-biological variables) available at the time of anticipation. The performance of “surgeons’ intuition” and “prognostic models” was assessed by area under the receiver operating characteristic curve and its accuracy by the diagnostic odd ratios. Results: Between October 2012 and September 2013, 946 patients operated on in hepato-pancreatico-biliary units in 9 teaching hospitals by 26 surgeons were enrolled. Mortality, morbidity, and median LOS were 3.3%, 49.4%, and 8 days, respectively. Preoperative surgeons’ intuition of difficulty correlated with actual difficulty (Kendall τ = 0.97; P = 0.0001) but not with morbidity (Kendall τ = 0.01; P = 0.0006) or LOS (Kendall τ = 0.10; P = 0.004). Morbidity was predicted accurately in 38.8% of patients and underestimated in 38.2%. Anticipation of LOS was accurate (±2 days) in 30.0% and un- derestimated in 47.1%. The accuracies and performance of preoperative and postoperative “surgeons’ intuition” were not different and were not different between centers or surgeons’ experience. The accuracy of “prognostic mod- els” was significantly greater than that of anticipations and not improved by adding “anticipations” to the model. From the ∗ Department of Hepatobiliopancreatic Surgery and Liver Transplantation, Hˆ opital Beaujon, Assistance Publique Hˆ opitaux de Paris, Universit´ e Paris 7, Clichy, France; †Department of Hepatobiliopancreatic Surgery and Liver Trans- plantation Hˆ opital Paul Brousse, Villejuif, France; ‡Department of Methodol- ogy and Statistics, Hˆ opital Nord, Amiens, France; §Department of Hepatobil- iopancreatic Surgery and Liver Transplantation, Hˆ opital Claude Huriez, Lille, France; ¶Department of Hepatobiliopancreatic Surgery and Liver Transplanta- tion, Hˆ opital de la Conception, Marseilles, France; ‖Department of Hepatobil- iopancreatic Surgery and Liver Transplantation, Hˆ opital Saint Antoine, Paris, France; ∗∗ Department of Hepatobiliopancreatic Surgery and Liver Transplan- tation, Hˆ opital Saint Andr´ e, Bordeaux, France; ††Department of Hepatobil- iopancreatic Surgery and Liver Transplantation, Hˆ opital Croix Rousse, Lyon, France; ‡‡Department of Hepatobiliopancreatic Surgery, Hˆ opital Edouard Her- riot, Lyon France; and §§Department of Methodology and Statistics, INSERM U669, Paris, France. Disclosure: This study was funded by a Programme Hospitalier de Recherche Clin- ique grant (PHRC National 2011, AOM 11060) from the French Ministry of Health and by the Association de Chirurgie Hepato-Biliaire et de Transplanta- tion H´ epatique (ACHBT). The authors declare no conflicts of interest. Reprints: Olivier Farges, MD, PhD, Department of HPB Surgery, Hˆ opital Beaujon, 100 bld du General Leclerc, F-92118 Clichy, France. E-mail: olivier.farges@bjn.aphp.fr. Copyright C  2014 by Lippincott Williams & Wilkins ISSN: 0003-4932/14/26005-0923 DOI: 10.1097/SLA.0000000000000961 Conclusions: Surgeons should be aware of the limited accuracy of their intu- ition. Keywords: anticipation, intuition, length of stay, liver resection, morbidity, predictive models, risk prediction (Ann Surg 2014;260:923–930) E valuation of the benefit and risk of a given medical intervention is central to the process of care. In the context of surgery, and in particular major surgery such as liver resection, the evaluation of risk is crucial. It is required to help the surgeon make an informed decision to provide accurate counseling to patients as part of shared decision making and to share information with other medical team members. 1 There is evidence that these factors, along with openly addressing the risk of a procedure and improving the accuracy of its prediction, increase patients’ safety. 2 Another reason for carefully analyzing the risk associated with liver resection is that there may be alternative therapeutic options that are less invasive and increasingly effective. 3,4 Finally, risk associated with a procedure has been used more recently to compare performance and outcome among different units. 5 The most frequently used endpoints to assess the risk of a sur- gical procedure are mortality and morbidity. These should be assessed at 90 days rather than as in-hospital events. 6 Length of stay (LOS) has been evaluated less frequently as it is considered a surrogate or intermediate outcome. However, there is increasing demand from pa- tients for less invasive procedures with reduced hospital stay, a need to initiate adjuvant chemotherapy rapidly when the indication is liver metastasis, and concerns about readmission rates. 7 Several prognostic models have been designed to predict these endpoints, but in the context of liver surgery 8 or of non–cardiac surgery in general, 9 most have not been validated and their perfor- mance is low. Surgeons’ own ability to predict these endpoints has been evaluated less frequently. This study was designed to compare “surgeons’ intuition” and “prognostic models” in predicting the risk associated with liver resection. METHODS Study Design and Participants This was a prospectively designed, registered (clinicaltrials.gov NCT01715402), multicenter, observational study investigating sur- geons’ accuracy in predicting the postoperative course of elective liver resection. It was deemed noninterventional by the ethics committees of participating institutions, approved by the Comit´ e Consultatif sur le Traitement de l’Information en mati` ere de Recherche dans le do- maine de la Sant´ e (CCTIRS), and by the agency for data protection, and was funded by the French Ministry of Health. Informed consent was obtained from participating surgeons and patients. Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Annals of Surgery  Volume 260, Number 5, November 2014 www.annalsofsurgery.com | 923