Background: The sensitivity of current upper limit of normal (ULN) of serum alanine aminotransferase (ALT) levels for detecting chronic liver disease has been challenged recently. Aim: To identify modulating factors for serum ALT levels and to refine its ULN threshold. Methods: We enrolled 34 346 consecutive subjects who completed the health check-up at Taipei Veterans General Hospital from 2002 to 2009. ULN was set for healthy ALT level to the 95th percentile of the reference healthy population. Results: A group of 21 282 subjects were used as a training set to define an ULN with the highest sensitivity; afterwards, this ULN was validated in another set of 13 064 subjects. A reference healthy population was selected from the training set after excluding subjects with any abnormalities in independent risk factors associated with elevated serum ALT level (>40 IU/L) by multivariate analysis like body mass index, waist circumference, glucose, cholesterol, high-density lipoprotein-cholesterol, triglyceride, hepatitis B virus surface antigen, anti-hepatitis C virus antibody and fatty liver. The new ULN of serum ALT level defined as the 95% percentile in the healthy population were 21 IU/L and 17 IU/L for men and women respectively. These cut-off values had the highest Youden's index and areas under the corresponding receiver operating curves among four widely applied thresholds in both the training and validation sets. Conclusions: The suggested threshold of upper limit of normal provides better discrimination between healthy and unhealthy status. Viral hepatitis, metabolic syndrome and fatty liver are the major risk factors of elevated serum alanine aminotransferase levels.
ASJC Scopus subject areas