Diagnosing clinical subsets of autoimmune liver diseases based on a multivariable model

Publisher Website
Google Scholar
Abstract
Diagnosing autoimmune hepatitis (AIH), primary biliary cirrhosis (PBC), primary sclerosing cholangitis, and other autoimmune liver diseases remains an imperfect process. We need a more accurate, evidence-based diagnostic system. We conducted a national survey and identified 988 cases of liver disease which did not satisfy the inclusion criteria for any liver disease of known etiology. We expected these cases to include autoimmune liver disease (AILD) and its variant forms. We selected 269 prototype cases for which histological re-evaluation of liver biopsy by independent expert hepatopathologists and the original diagnosis coincided. We did a multiple logistic regression analysis to determine explanatory variables that would distinguish cases of AIH and PBC from those of non-AIH and non-PBC, respectively. We constructed a multivariable diagnostic formula that gave AIH and PBC disease probabilities and validated it in a study of an additional 371 cases (validation group). Based on the results of the statistical analysis, we selected three laboratory tests and four histological features as independent variables correlated to the diagnosis of both AIH and PBC. For the validation group, assuming that the original diagnosis was correct, the sensitivity and specificity for AIH were 86.3% and 92.4%, respectively. For PBC the sensitivity and specificity were 82.5% and 63.7%, respectively. A detailed analysis of inconsistent cases showed that the diagnosis based on the formula had given the correct diagnosis, for either AIH or PBC, except for 5 cases (1.3%) in which disease probability was low for both. A seven-variable formula based on three laboratory tests and four histological features gives significant information for the diagnosis of AILD.