Screening for liver cancer: the rush to judgment.

In 2005, a professional society issued a level I recommendation to use ultrasonography to screen the 3 million Americans with cirrhosis for hepatocellular carcinoma (HCC) every 6 months. This designation was based on a large randomized trial from China that reported a reduction in HCC deaths from screening in carriers of hepatitis B surface antigen. However, besides the difference in population, this trial did not report all deaths, excluded patients after randomization, and would almost certainly have found no significant difference if the cluster randomization had been accounted for in the analysis. Misplaced confidence in the Chinese trial has led many writers to accept the effectiveness of HCC screening as proven, making it more difficult to conduct the high-quality randomized trials that are needed to ensure optimal patient care.