Specificity of unenhanced CT for non-invasive diagnosis of hepatic steatosis: implications for the investigation of the natural history of incidental steatosis.

To determine a highly specific liver attenuation threshold at unenhanced CT for biopsy-proven moderate to severe hepatic steatosis (≥30% at histology). 315 asymptomatic adults (mean age ± SD, 31.5 ± 10.1 years; 207 men, 108 women) underwent same-day unenhanced liver CT and ultrasound-guided liver biopsy. Blinded to biopsy results, CT liver attenuation was measured using standard region-of-interest methodology. Multiple linear regression analysis was used to assess the relationship of CT liver attenuation with patient age, gender, BMI, CT system, and hepatic fat and iron content. Thirty-nine subjects had moderate to severe steatosis and 276 had mild or no steatosis. A liver attenuation threshold of 48 HU was 100% specific (276/276) for moderate to severe steatosis, with no false-positives. Sensitivity, PPV and NPV at this HU threshold was 53.8%, 100% and 93.9%. Hepatic fat content was the overwhelming determinant of liver attenuation values, but CT system (P < 0.001), and hepatic iron (P = 0.035) also had a statistically significant independent association. Unenhanced CT liver attenuation alone is highly specific for moderate to severe hepatic steatosis, allowing for confident non-invasive identification of large retrospective/prospective cohorts for natural history evaluation of incidental non-alcoholic fatty liver disease. Low sensitivity, however, precludes effective population screening at this threshold. • Unenhanced CT liver attenuation is highly specific for diagnosing moderate/severe hepatic steatosis. • Unenhanced CT can identify large cohorts for epidemiological studies of incidental steatosis. • Unenhanced CT is not, however, effective for population screening for hepatic steatosis.