What is the best surveillance for hepatocellular carcinoma in chronic carriers of hepatitis B?

Many serum markers and screening methods have been proposed to detect hepatocellular carcinoma at a treatable stage, but only 2-AFP and US-are in clinical use. A Cochrane systematic review on screening for hepatocellular carcinoma in the HBsAg+ population was published in 2003 and updated May 2004. Our literature search did not find any subsequent relevant trials. The Cochrane review included 2 randomized control trials. The larger trial was performed in Shanghai, China and included 18,816 HBsAg+ patients aged 35 to 55 years. Subjects were recruited from their place of employment and randomized to either AFP/US every 6 months (n=9373) or to no screening (n=9443). Fifty-one hepatocellular carcinomas were diagnosed in the control group and 86 in screened group. Screened subjects had a significantly higher percentage of tumors that were less than 5 cm at the time of diagnosis and a higher number of patients who underwent resection. While the 5-year survival for those with hepatocellular carcinoma in the screened group was higher, the disease-specific mortality rate was not statistically different between the 2 groups. Additional data became available in 2002. The original study authors claimed the new data showed a statistically significant disease-specific mortality rate ratio of 0.63, favoring the screened group. However, the Cochrane group performed their own analysis on the same data and determined that no statistically significant difference in the disease-specific mortality rates existed between the 2 groups. Therefore, it is not clear whether these new data definitively demonstrate that screening provides any benefit. The other randomized control trial took place in Toronto, and included 1069 patients, 71% of whom were of Asian ancestry. Subjects had AFP testing every 6 months, and half were randomly assigned to have US performed every 6 months. Eight of the 11 incident tumors would have been diagnosed based on AFP levels alone, and 3 would have been missed with US alone. The authors conclude that for AFP, sensitivity=64.3% and specificity=91.4%; for US, sensitivity=78.8% and specificity=93.8%. However, their study was too small to determine if AFP/US is superior to AFP for hepatocellular carcinoma screening in a HBsAg+ population. They estimate that detecting such a difference would take a sample size of 10,000 or more. Both studies have important flaws. Neither study applied a reference standard test (such as a computed tomography scan or magnetic resonance imaging) to both study arms. Carcinomas may have been undetected by either AFP or US. Without knowing the real prevalence of hepatocellular carcinoma, the true sensitivity and specificity for AFP, US, and AFP/US in these studies cannot be determined. Both studies included prevalent tumors (tumors diagnosed during the very first screening cycle) in their analysis. Approximately 20% of detected carcinomas in both studies were present at the start of the studies and did not represent newly incident tumors detected by regular screening. Both of these trials would be improved if they started with cohorts known to be disease-free at baseline. Additionally, the Shanghai study randomized patients in clusters. The only English-language report of this study did not describe whether adjustments for this were made in analysis; failing to do so could overestimate the benefit of screening.