The study sought to examine which biomarkers have the best predictive capabilities for future alcohol-related liver cirrhosis (ARLC) in a general population setting. This population-based cohort study includes approximately 35% of the inhabitants of Stockholm County who had left a blood sample at an outpatient visit in primary care or occupational health screening from 1985 to 1996. All subjects with a blood sample measurement of alanine aminotransferase and aspartate aminotransferase (AST) were included, exclusions were made for persons with known liver disease. We ascertained incident ARLC by linkage to Swedish national health registers between to the end of 2011. Associations between biomarkers and incident ARLC were analyzed with Cox regression models and discrimination was assessed using C-statistics. In all, 537,230 adult subjects were included. The mean age was 45 years and 53% were men. During a mean follow-up of 19.0 years, 2725 (0.51%) subjects developed ARLC. The biomarkers with the highest discrimination (C-index) for incident ARLC at 5 years were: AST (0.89), mean corpuscular volume (0.88), and γ-glutamyltransferase (0.81). Scoring systems including Fibrosis-4 (0.86) and the AST/alanine aminotransferase ratio (0.81) performed similarly well. The negative predictive value for ARLC was generally high (∼99.6%) across biomarkers, using routine clinical cutoffs to identify pathological values. However, positive predictive values were generally low (0.6%-15.9%). Biomarkers commonly used in primary care settings are highly associated with incident ARLC in the general population. Elevation of these commonly available biomarkers should prompt consideration of further investigation of a possible high level of alcohol consumption.
Read full abstract