33 CRISCO-UGIS: COMPOSITE RISK SCORE TO PREDICT POST-OPERATIVE COMPLICATIONS FOR UPPER GASTROINTESTINAL CANCER SURGERY

Abstract Gastroesophagectomy for cancer is associated with high risk of morbidity and mortality. Careful patient selection for surgery is essential for optimizing outcomes. The purpose of this study was to develop a reliable composite risk calculator to inform pre-operative decision-making using readily available patient and tumour characteristics. Methods Patients undergoing curative-intent resection for gastroesophageal cancer from January 2010 to December 2018 were identified from a prospectively collected database. Data were verified and collected from patients’ medical records. Multiple logistic regression modeling identified pre-operative variables associated with significantly elevated risk of moderate to severe post-operative complications (Clavien-Dindo score ≥ 2) within 30 days of surgery and a nomogram was developed to form the composite risk score (CRiSCo-UGIS). Internal model validation was performed. Results Among 520 patients, independent pre-operative risk factors for complications were: female sex, age-adjusted Charlson comorbidity index ≥5, albumin <35 g/L, proximal tumour location (esophagus and gastroesophageal junction) and reduced performance status (Eastern Cooperative Oncology Group performance status 2–3) (Table 1). CRiSCo-UGIS had a good model fit (Hosmer-Lemeshow p = 0.47, area under receiver operating characteristic curve = 0.69). The overall observed-to-expected ratio was 1.0 [95% confidence interval 0.9–1.1]. Patients were divided into low (0), moderate (1–2) and high (≥3 points) risk categories. The observed risk of complications was: 42% in low-risk, 53% in moderate-risk and 80% in high-risk. Conclusion CRiSCo-UGIS can reliably predict post-operative complications for gastroesophageal cancer resection based on readily available pre-operative characteristics. This composite score can be used to guide patient selection for surgery and enhance informed consent discussions.