Get access

Hepatic steatosis in diabetic patients does not predict adverse liver-related or cardiovascular outcomes



Michael A. Dunn, MD, UPMC Presbyterian, M2 C Wing, 200 Lothrop Street, Pittsburgh, PA 15213, USA

Tel: +1 412 692 2470

Fax: +1 412 647 9268



Background & Aims

Steatosis is a defining feature of nonalcoholic fatty liver disease (NAFLD). However, evidence that severity of steatosis can predict adverse outcomes in NAFLD or nonalcoholic steatohepatitis (NASH) is lacking. The aim of this study was to determine whether steatosis assessed by computed tomography (CT) imaging predicts adverse outcomes in diabetic patients at risk for NAFLD/NASH.


We studied deaths, liver-related and cardiovascular adverse outcomes in a 5-year retrospective observational cohort of 2343 type 2 diabetic patients in a large care network who had noncontrast CT imaging for clinical indications. We measured steatosis by subtraction of spleen from liver attenuation, a method that showed low sensitivity (0.417) and high specificity (0.882) compared with histopathological scoring. We evaluated outcomes prediction using multivariate Cox proportional hazards modelling of steatosis both as a categorical (≥30%) and continuous variable.


Steatosis ≥30% was present in 233 (9.9%) of the cohort at baseline. Over 5 years, there were 372 total deaths, 18 liver-related and 99 cardiovascular deaths, 48 liver transplants, 51 occurrences of hepatic encephalopathy, 41 hepatocellular carcinomas, 653 myocardial infarctions, 66 strokes, 180 occurrences of angina, 735 occurrences of arrhythmia and 772 occurrences of congestive heart failure. Steatosis had no predictive value for any adverse outcome. Patients with steatosis averaged 8 years younger than those without it. Age had a strong covariate influence on occurrence of total deaths, cardiovascular deaths, myocardial infarctions, arrhythmias and congestive heart failure.


Although steatosis on imaging is often the abnormality that triggers diagnosis and assessment of NAFLD/NASH, it lacks predictive value for adverse clinical outcomes.