OBJECTIVES: To evaluate the effectiveness, cost-effectiveness and cost-utility of surveillance of patients with cirrhosis [alcoholic liver disease (ALD)-, hepatitis B (HBV)- and C virus (HCV)-related], using periodic serum alpha-fetoprotein (AFP) testing and/or liver ultrasound examination, to detect hepatocellular carcinoma (HCC), followed by treatment with liver transplantation or resection, where appropriate. DATA SOURCES: Electronic databases were searched up to March 2006. REVIEW METHODS: A systematic review was carried out using standard methodological guidelines. A computerised decision-analytic model was then developed to compare various surveillance strategies. RESULTS: No studies were identified that met the criteria of the systematic review. Based on the assumptions used in the model, the most effective surveillance strategy uses a combination of AFP testing and ultrasound at 6-monthly intervals. Compared with no surveillance, this strategy is estimated to more than triple the number of people with operable HCC tumours at time of diagnosis, and almost halves the number of deaths from HCC. On all effectiveness measures and at both testing frequencies, AFP- and ultrasound-led surveillance strategies are very similar. This may be because test sensitivity was varied according to tumour size, which means that AFP testing is capable of identifying many more small tumours than ultrasound. The best available evidence suggests that AFP tests will detect approximately six times as many small tumours as ultrasound. Increasing the frequency of either test to 6-monthly intervals is more effective than performing combined testing on an annual basis. The undiscounted lifetime cost of the surveillance strategies, including all care and treatment costs, ranges from 40,300 pounds (annual AFP triage) to 42,900 pounds (6-monthly AFP and ultrasound). The equivalent discounted costs are 28,400 pounds and 30,400 pounds. Only a small proportion of these total costs results from the cost of the screening tests. However, screening test costs, and the cost of liver transplants and caring for people post-transplant, accounted for most of the incremental cost differences between alternative surveillance strategies. The results suggest that different surveillance strategies may provide the best value for money in patient groups of different cirrhosis aetiologies. The surveillance of people with HBV-related cirrhosis for HCC provides the best value for money, while surveillance in people with ALD-related cirrhosis provides the poorest value for money. In people with HBV-related cirrhosis, at an assumed maximum willingness to pay (WTP) for a quality-adjusted life-year (QALY) of 30,000 pounds, both the deterministic and probabilistic cost-utility analyses suggest the optimal surveillance strategy would be 6-monthly surveillance with the combination of AFP testing and ultrasound. In contrast, for those with ALD-related cirrhosis, annual screening with AFP as a triage test is the only surveillance strategy that is likely to be considered cost-effective at this WTP. The probabilistic analysis implies that the estimated benefits of a 6-monthly AFP triage strategy will only be worth the cost in those with ALD when society's WTP for a QALY exceeds around 40,000 pounds. For people with HCV-related cirrhosis, the model suggests that the most cost-effective surveillance strategy at a WTP threshold of 30,000 pounds/QALY would be surveillance with a 6-monthly AFP triage strategy. CONCLUSIONS: In a mixed-aetiology cohort, the most effective surveillance strategy is to screen each patient with AFP assay and ultrasound imaging on a 6-monthly basis. However, when costs are taken into account it is doubtful whether ultrasound should be routinely offered to those with blood AFP of less than 20 ng/ml, unless policy-makers are prepared to pay over 60,000 pounds per QALY for the benefits achieved. Furthermore, the cost-effectiveness of surveillance for HCC varies considerably depending on the aetiology of cirrhosis; it is much more likely to be cost-effective in those with HBV-related cirrhosis, and much less likely to be cost-effective in those with ALD-related cirrhosis. Further development of the model would help to enable refinement of an optimal screening strategy. Research into the use of contrast-enhanced ultrasound technology for HCC detection would also be valuable, as would research into the epidemiology and natural history of ALD-related cirrhosis. Studies are also needed to investigate the influence of cirrhosis aetiology on tumour AFP expression.