The Risk Predictive Values of ACG Classification in a Cohort of Colon Ischemia: Refining the Definition of Mild Disease

Abstract

Introduction: Although most cases of colon ischemia (CI) are mild and self-limiting, when severe it implies high mortality rates. We aimed to evaluate the risk predictive value of the classification proposed by American College of Gastroenterology (ACG) guidelines (2015), created to provide a management algorithm. Methods: A retrospective multicenter study was conducted on adult patients with definite CI (clinical, endoscopic, pathologic and culture criteria), between 2013 and 2016. Data was collected on clinical presentation, including comorbidities, organ failure, management and outcome. Cases were evaluated according to ACG risk factors (RF) (gender, systolic blood pressure, heart rate, abdominal pain without rectal bleeding, BUN, Hb, LDH, serum sodium, WBC). Cases were classified as mild (0 RF), moderate (1-3 RF) and severe (>3 RF or any of the following: peritoneal signs, pneumatosis or portal venous gas, gangrene, pan-colonic or isolated right-colon ischemia). Results: 349 cases with the clinical diagnosis of IC were analyzed. 193 met inclusion criteria of definitive diagnosis (63% females; mean age 72 years (±13). ACG classification of mild, moderate and severe disease was attributed respectively to 21% of patients (0 intra-hospital deaths), 45% (2 deaths) and 34% (12 deaths). The number of ACG RF was: 40% with 0 RF, 8% with 1, 9% with 2, 15% with 3, 16% with 4, 8% with 15%, 4% with 6 and 1% with 7. No patient with 0 or 1 RF died. Only 1 patient with 2 RF died. The remaining 13 deaths were verified with at least 3 RF. The univariate analysis revealed a statistical correlation between RF and intra-hospital or 30-day mortality as well as the need for surgery (P<0.001, P<0.001, and P=0.006, respectively). ACG classification presented high predictive accuracy for in-hospital and 1-month mortality with an AUROC of 0.78 and 0.79 (P<0.001), respectively. For a cutoff of 2 ACG RF, the sensibility (SE) for death was 100%, specificity (SP) was 52%, with a positive predictive value (PPV) of 14% and negative predictive value (NPV) of 100%. For 3 ACG RF the results were: SE 93%, SP 61%, PPV 16% and NPV 99%. Three or more RF had an OR of 20.2 (95%CI: 2.59-158) for intra-hospital and 18.4 (95% CI: 2.34-144) for 1-month mortality. Conclusion: No patient in this cohort with less than 2 ACG RF died, suggesting that mild disease of ACG classification may include 0 and 1 RF without changing prognosis. Short-term mortality risk increases significantly in patients with at least 3 ACG RF.Figure: ROC curve comparing the sensitivity of prediction of mortality at 1 month using the ACG classification, ACG classification adjusting mild disease until 1 RF and the number of ACG RF.