A predictive postoperative score of anastomotic leakage after subtotal esophagectomy : a retrospective study

Position du problème et objectif(s) de l’étude

Postoperative anastomotic leakage (AL) remains a major complication of esophagectomy for esophageal cancer, occurring in 5–40% of cases, and it dramatically impacts morbidity and mortality (1). Early diagnosis of AL is crucial to prevent subsequent complications and significantly improve outcomes. However, it remains a challenge, as conventional diagnostic techniques often lack sensitivity in the acute postoperative period (2–4). A reliable early predictive score based on clinical and biological features is critically needed and could enhance patient care by enabling timely intervention.

Matériel et méthodes

This single-center, retrospective study included all patients admitted to the intensive care unit (ICU) after Ivor-Lewis esophagectomy between January 2010 and October 2022 at a university hospital in Paris, France. A least absolute shrinkage and selection operator (LASSO) regression model was applied to clinically, surgically, and biologically relevant features to develop a predictive score on postoperative day 5. The model was validated using cross-validation, and its performance was assessed using the area under the receiver operating characteristic curve (AUC). Statistical analyses were performed using R software (version 4.3.1). This study was approved by the Research Ethics Committee in Anesthesia and Critical Care (IRB 00010254‐2023–038).

Résultats & Discussion

Among 191 screened patients, 187 were included in the final analysis, with 34 (18%) experiencing AL. The LASSO model identified 13 features predictive of AL, including diabetes, age, malnutrition, neoadjuvant treatment, hybrid surgery, and postoperative markers measured up to postoperative day 5 such as ARDS, septic shock, white blood cell count, C-reactive protein (CRP) concentration, and amylase levels in thoracic drains. The predictive model achieved an AUC of 0.77 (95% CI: [0.69–0.86]). An online tool was developed to facilitate clinical application at https://al-score.fr.

Conclusion

This study presents a novel machine learning-based predictive score for early AL detection following Ivor-Lewis esophagectomy. By integrating preoperative, intraoperative, and postoperative features, this model offers superior predictive accuracy compared to existing biomarkers. Prospective validation is required to confirm its clinical utility.

Auteurs

Victor HERVE (1) , Benjamin GLEMAIN (2), Amélie CAMBRIEL (1), Louise DODIN (1), Thibault VORON (3), Daniel EL KHOURY (1), Emmanuel PARDO (1), Natacha KAPANDJI (1), Thibaut GOBE (1), Nathanael LAPIDUS (2), Franck VERDONK (1) - (1)Aphp, Hôpital Saint Antoine, Dmu Dream, Department Of Anesthesiology And Intensive Care, Paris, France, (2)Sorbonne Université, Inserm, Institut Pierre-Louis D'épidémiologie Et De Santé Publique, Département De Santé Publique, Ap-Hp, Hôpital Saint-Antoine, Paris, France, (3)Sorbonne University And Department Of Digestive Surgery, Hôpital Saint-Antoine, Assistance Publique-Hôpitaux De Paris, Paris, France, Paris, France

Orateur(s)

Victor HERVE (Paris)