DIAGNOSING ALCOHOLIC LIVER DISEASE: A PILOT STUDY FOR NICE

Purpose:   This work was undertaken as a pilot study for the recently formed NICE diagnostic appraisal committee. Non-invasive liver tests (NILTs) for alcoholic liver disease (ALD) were chosen; these tests have the potential to reduce both the costs and risk of mortality associated with liver biopsy. Four NILTs were evaluated: three blood tests (FibroTest, FibroMAX and the ELF test) and one transient elastography (FibroScan).

Method: Systematic reviews were undertaken and complimented by specialist expert input to estimate parameters within a mathematical model. Key parameters included the sensitivity and specificity of each NILT compared with biopsy; the sensitivity of biopsy; the mortality associated with biopsy; and the cost and QALY implications for combinations of diagnostic test outcomes (true positives, false positives etc) and patient abstinence status (abstinent or not). The model was constructed to assess the cost-effectiveness of strategies using a NILT followed by a confirmatory biopsy where cirrhosis was indicated and strategies relying on the diagnosis of the NILT alone compared with biopsying all patients.

Result: The most cost-effective intervention was markedly influenced by the assumed rates of abstinence for each strategy. Abstinence is the mainstay of treatment for all patients with ALD and can greatly increase life expectancy. Studies report a higher abstinence rate for patients whose biopsy diagnosed cirrhosis compared with those where it did not. If these rates are appropriate to NILT-only strategies, then tests with low specificity produce the greatest health benefit due to an increase in abstinence amongst the false positives. A key uncertainty is whether abstinence rates following a diagnosis made using a NILT alone would be influenced by patient or physician knowledge of low specificity.  The cost-effectiveness of a NILT followed by a biopsy strategy is currently uncertain with scenario analyses reporting favourable and unfavourable values. These results are dependent on assumptions regarding sensitivities and specificities for each diagnostic test, and on the cost and utility estimates from treating diseases unrelated to ALD detected on biopsy, which have relatively wide confidence intervals.

Conclusion: The cost-effectiveness of NILT-only strategies are determined by the abstinence rates assumed in those diagnosed as having cirrhosis, many of whom will be false positives. A NILT used prior to a biopsy may be cost-effective. Further data are required before definitive conclusions can be determined.