Diagnostic Accuracy of Noninvasive Fibrosis Scores in a Population of Individuals With a Low Prevalence of Fibrosis

Background & Aims Noninvasive scoring systems for fibrosis are increasingly used in the clinic and in research because of their ease of use, accessibility, and low cost. However, test performance characteristics were established in groups of patients with a high prevalence of advanced fibrosis; little is known about diagnostic accuracy in low-risk populations. Methods In a cross-sectional study, 922 members of a general ambulatory population in Hong Kong (randomly selected; 18–70 years old) underwent clinical assessment from May 2008 through December 2010. All participants completed a standard questionnaire that collected information on age, sex, and history of smoking and alcohol use. Results of fasting blood tests and transient elastography were used as the reference standard to identify patients with advanced fibrosis. We assessed performance characteristics of 3 noninvasive fibrosis scoring systems: the nonalcoholic fatty liver disease fibrosis scoring system, the Fibrosis-4 scoring system, and aspartate transaminase to platelet ratio index, using standard thresholds. To calculate diagnostic test characteristics, we constructed a 2-by-2 table with the presence or absence of advanced fibrosis according to the transient elastography reading against the presence or absence of advanced fibrosis according to the scoring systems. Area under the receiver operating curve was calculated to assess overall diagnostic accuracy. Results Of the 922 individuals evaluated by transient elastography, 749 had a valid reading and 15 had advanced fibrosis (2%). The specificity of noninvasive scores in detection of advanced fibrosis approximated 100% (95% confidence interval [CI], 99%–100%), with a negative predictive value of 98% (95% CI, 97%–99%) for all systems. However, the scoring systems detected fibrosis with a low level of sensitivity, ranging from 7% (95% CI, 0%–32%) to 13% (95% CI, 2%–40%). Positive predictive values ranged from 50% (95% CI, 7%–93%) to 67% (95% CI, 9%–99%). Their negative likelihood ratios ranged from 0.87 (95% CI, 0.71%–1.06%) to 0.93 (95% CI, 0.82%–1.07%); positive likelihood ratios were uninformative because of the small number of people with positive scores. Conclusions In low-risk populations, negative results from noninvasive scoring systems reliably exclude advanced fibrosis, without requirements for further tests. Positive test results are often a false-positive result and should prompt further testing.