US Cannot Be Used to Predict the Presence or Severity of Hepatic Steatosis in Severely Obese Adolescents

Purpose: To evaluate the diagnostic accuracy of ultrasonography (US) for the assessment of hepatic steatosis in severely obese adolescents, with proton magnetic resonance (MR) spectroscopy as the reference standard, and to provide insight on the influence of prevalence on predictive values by calculating positive and negative posttest probabilities. Materials and Methods: This prospective study was institutional review board approved. All participants, and/or their legal representatives, gave written informed consent. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for the overall presence of steatosis and for the presence of substantial (moderate to severe) steatosis. Positive and negative posttest probabilities were calculated and plotted against prevalence. Results: A total of 104 children (47 male, 57 female) were prospectively included. Mean age was 14.5 years (range, 8.3-18.9 years) and mean age-adjusted standard deviation body mass index (BMI) score (BMI z score) was 3.3 (range, 2.6-4.1). The overall prevalence of hepatic steatosis was 46.2% (48 of 104). Sensitivity of US was 85.4% (41 of 48), specificity was 55.4% (31 of 56), PPV was 62.1% (41 of 66), and NPV was 81.6% (31 of 38). The prevalence of substantial steatosis was 15.4% (16 of 104), with US sensitivity of 75.0% (12 of 16) and specificity of 87.5% (77 of 88). PPV was 52.2% (12 of 23) and NPV was 95.1% (77 of 81). Plotting of posttest probabilities against prevalence for both disease degrees demonstrated how disease prevalence influences US accuracy. Conclusion: Positive US results in severely obese adolescents cannot be used to accurately predict the presence and severity of hepatic steatosis, and additional imaging is required. Negative US results exclude the presence of substantial steatosis with acceptable accuracy. Steatosis prevalence differs among specific populations, strongly influencing posttest probabilities. (C) RSNA, 2011