Risk assessment using a novel score to predict anastomotic leak and major complications after oesophageal resection

Background Oesophagectomy is associated with significant morbidity and mortality. A simple score to define a patient's risk of developing major complications would be beneficial. Methods Patients who underwent upper gastrointestinal resections with an oesophageal anastomosis between 2005 and 2010 were reviewed and formed the development dataset with resections performed in 2011 forming a prospective validation dataset. The association between post-operative C-reactive protein (CRP), white cell count (WCC) and albumin levels with anastomotic leak (AL) or major complication including death using the Clavien–Dindo (CD) classification were analysed by receiver operating characteristic curves. After multivariate analysis, from the development dataset, these factors were combined to create a novel score which was subsequently tested on the validation dataset. Results Two hundred fifty-eight patients were assessed to develop the score. Sixty-three patients (25%) developed a major complication, and there were seven (2.7%) in-patient deaths. Twenty-six (10%) patients were diagnosed with AL at median post-operative day 7 (range: 5–15). CRP (p = 0.002), WCC (p < 0.0001) and albumin (p = 0.001) were predictors of AL. Combining these markers improved prediction of AL (NUn score > 10: sensitivity 95%, specificity 49%, diagnostic accuracy 0.801 (95% confidence interval: 0.692–0.909, p < 0.0001)). The validation dataset confirmed these findings (NUn score > 10: sensitivity 100%, specificity 57%, diagnostic accuracy 0.879 (95% CI 0.763–0.994, p = 0.014)) and a major complication or death (NUn > 10: sensitivity 89%, specificity 63%, diagnostic accuracy 0.856 (95% CI 0.709–1, p = 0.001)). Conclusions Blood-borne markers of the systemic inflammatory response are predictors of AL and major complications after oesophageal resection. When combined they may categorise a patient's risk of developing a serious complication with higher sensitivity and specificit