Ability of Noninvasive Scoring Systems to Identify Individuals in the Population at Risk for Severe Liver Disease

Abstract

Noninvasive scoring systems are used to identify persons with advanced liver fibrosis. We investigated the ability of scoring systems to identify individuals in the general population at risk for future liver-related events. We collected data from the Swedish Apolipoprotein Mortality Risk cohort on persons 35 to 79 years old who had blood samples collected from 1985 through 1996. We collected APRI (n= 127,302), BARD (n= 75,303), FIB-4 (n= 126,941), Forns (n= 122,419), and the nonalcoholic fatty liver disease (NAFLD) fibrosis scores (NFS, n= 13,160). We ascertained incident cases of cirrhosis or complications by linking Swedish health data registers. Cox regression was used to estimate hazard ratios (HRs) for severe liver disease at 5, 10, and a maximum follow-up time of 27 years. The predictive ability of the scores was evaluated using area under the receiver operating characteristic (AUROC) curve and C-statistics analyses. Our specific aims were to investigate the predictive capabilities of scoring systems for fatal and nonfatal liver disease, determine whichscoring system has the highest level of accuracy, and investigate the predictive abilities of the scoring systems in persons with a higher probability of NAFLD at baseline. A similar proportion of individuals evaluated by each scoring system developed cirrhosis or complications thereof (1.0%-1.4%). The incidence of any outcome was increased in intermediate- and high-risk groups compared with low-risk groups, with HRs at 10 years in the high-risk group ranging from 1.67 for the BARD score to 45.9 for the APRI score. The predictive abilities of all scoring systems decreased with time and were higher in men. All scoring systems were more accurate in persons with risk factors for NAFLD at baseline, with AUROCs reaching 0.83. Higher scores from noninvasive scoring systems to evaluate fibrosis are associated with an increased risk of cirrhosis in a general population, but their predictive ability is modest. Performance was better when patientswere followed for shorter time periods and in persons with a higher risk of NAFLD, with AUROC values reaching 0.83. New scoring systems are needed to evaluate risk of fibrosis in the general population and in primary care.