The diagnosis of alcoholic hepatitis (AH) often requires a transjugular liver biopsy (TJLB), a procedure that is not always readily accessible. We analyzed plasma biomarkers to estimate the presence of histological features of AH among patients with clinical suspicion of AH. Using enzyme-linked immunosorbent assay, we tested M65 and M30 (circulating fragments of cytokeratin-18) and their respective fraction carried by microvesicles (MVs), CCL20 and TREM1. Leukocyte, platelet, and endothelial-derived MVs were quantified by way of flow cytometry. Test and validation cohorts prospectively included patients with clinical features of AH undergoing TJLB. In the test cohort, 46 of 83 (55%) patients showed histological features of AH. Age, bilirubin, INR, and creatinine (ABIC) score was B or C in 83%. Patients with histologically proven AH had higher levels of total and MV-bound M65 and total and MV-bound M30 and CCL20 than those without (P < 0.001 for all tests). Levels of TREM-1 and of subpopulations of MVs were not different between groups. M65 and M30 both had an area under the receiver operating characteristics curve of 0.84 to estimate the presence of AH. For M65, a cutoff of 2000 IU/L had a positive predictive value of 91%, whereas a cutoff of 641 IU/L had a negative predictive value of 88%. In the validation cohort, AH was histologically confirmed in 48 of 68 (71%) patients. ABIC score was B or C in 69% of patients. For M65, the above cutoffs had a diagnostic accuracy of 81%. Even better results were obtained in patients with suspicion of severe AH (ABIC B or C) in both cohorts.
Plasma levels of cytokeratin-18 fragments are reliable noninvasive markers of AH. Using the proposed cutoffs for M65, two thirds of TJLB can be avoided, which can be useful in centers where this technique is not readily available. (Hepatology 2017;66:555-563).