Predictors of Ultrasound Failure to Detect Hepatocellular Carcinoma

Current guidelines recommend ultrasound (US) for hepatocellular carcinoma (HCC) surveillance in cirrhosis. We assess predictors of decreased US sensitivity for detecting HCC. At a single center in the United States, all HCC patients evaluated for liver transplantation (LT) received an abdominal US. From 2007-2015, consecutive patients presenting for untreated lesions found on computed tomography (CT) or magnetic resonance imaging (MRI) within 3 months of US were compared with US findings. Multivariate logistic regression models compared US sensitivities by patient characteristics. Of 1007 patients completing LT evaluation, 47.5% had indeterminate or previously treated nodules and were excluded; 10.4% had imaging that was too far apart or nondiagnostic. Median Model for End-Stage Liver Disease (MELD) of the cohort (n= 352) was 11 (interquartile range [IQR], 9-14), median body mass index (BMI) was 28 kg/m² (IQR, 25-32 kg/m² ), 39% had received locoregional therapy, and 10% had moderate/large ascites. Per-patient sensitivity of US compared with CT/MRI was 0.82 (95% confidence interval, 0.76-0.86). Patients with BMI ≥ 30 kg/m² had a US sensitivity of 0.76 versus 0.87 for BMI < 30 kg/m² (P = 0.01). MELD and ascites did not affect sensitivity. US sensitivity was decreased in patients with nonalcoholic steatohepatitis (NASH) versus other etiologies (0.59 versus 0.84; P = 0.02). Relative to CT/MRI, US is significantly less sensitive in patients with NASH or BMI ≥ 30 kg/m² . Further study is necessary to examine the added value of cross-sectional imaging for patients with NASH or obesity.