Background & Aims The prevalence of fatty liver underscores the need for non-invasive characterization of steatosis, such as the ultrasound based controlled attenuation parameter (CAP). Despite good diagnostic accuracy, clinical use of CAP is limited due to uncertainty regarding optimal cut-offs and the influence of covariates. We therefore conducted an individual patient data meta-analysis. Methods A review of the literature identified studies containing histology verified CAP data (M probe, vibration controlled transient elastography with FibroScan®) for grading of steatosis (S0–S3). Receiver operating characteristic analysis after correcting for center effects was used as well as mixed models to test the impact of covariates on CAP. The primary outcome was establishing CAP cut-offs for distinguishing steatosis grades. Results Data from 19/21 eligible papers were provided, comprising 3830/3968 (97%) of patients. Considering data overlap and exclusion criteria, 2735 patients were included in the final analysis (37% hepatitis B, 36% hepatitis C, 20% NAFLD/NASH, 7% other). Steatosis distribution was 51%/27%/16%/6% for S0/S1/S2/S3. CAP values in dB/m (95% CI) were influenced by several covariates with an estimated shift of 10 (4.5–17) for NAFLD/NASH patients, 10 (3.5–16) for diabetics and 4.4 (3.8–5.0) per BMI unit. Areas under the curves were 0.823 (0.809–0.837) and 0.865 (0.850–0.880) respectively. Optimal cut-offs were 248 (237–261) and 268 (257–284) for those above S0 and S1 respectively. Conclusions CAP provides a standardized non-invasive measure of hepatic steatosis. Prevalence, etiology, diabetes, and BMI deserve consideration when interpreting CAP. Longitudinal data are needed to demonstrate how CAP relates to clinical outcomes. Lay summary There is an increase in fatty liver for patients with chronic liver disease, linked to the epidemic of the obesity. Invasive liver biopsies are considered the best means of diagnosing fatty liver. The ultrasound based controlled attenuation parameter (CAP) can be used instead, but factors such as the underlying disease, BMI and diabetes must be taken into account. Registration: Prospero CRD42015027238.
All Science Journal Classification (ASJC) codes