THE IMPACT OF TUMOR KINETIC GROWTH MODELS ON THE COST-EFFECTIVENESS OF PERIODIC SURVEILLANCE FOR HEPATOCELLULAR CARCINOMA

Sunday, October 24, 2010
Sheraton Hall E/F (Sheraton Centre Toronto Hotel)
Haku Ishida, M.D.1, John Wong, MD2, Isao Sakaida, M.D.3 and Yuji Inoue, M.D.1, (1)Yamaguchi University Hospital, Ube, Japan, (2)Tufts Medical Center, Boston, MA, (3)Yamaguchi University, School of Medicine, Ube, Japan

Purpose: To evaluate the effect of alternative tumor kinetic growth models on the cost-effectiveness of periodic surveillance of hepatocellular carcinoma (HCC).

Method: We considered four tumor kinetic models: declining growth rate (Gompertzian), constant growth rate (exponential), initial 12 month tumor latency followed by constant growth, and a combination of these 3.  The Monte Carlo simulation considered two strategies for 55-year-old patients with HCV-related Child’s A cirrhosis: periodic surveillance and no surveillance HCCs were grouped by diameter: small (≤3cm), medium (3-5cm) and large (>5cm) and received treatments as observed in actual practice except for small HCC, where options included hepatic resection, local ablation therapy or liver transplantation. Other data on diagnostic performance of surveillance tests, HCC incidence, tumor volume doubling time, HCC mortality and hepatic decompensation were literature-based. Direct costs of treatments and testing were obtained from health insurance claims.

Result: Surveillance yielded substantially different mean HCC size depending on which tumor growth model was used with the combined model yielding one most similar to literature observations. Progression from small to medium, and medium to large HCC took more than twice as long with the slower Gompertzian growth model than the exponential model.  This slower growth and the lower quality of life from living with HCC limited the benefit of surveillance to 0.09 quality-adjusted life years (QALYs) compared with no surveillance.  In contrast, with an exponential growth model, surveillance increased QALYs by 0.79 QALYs.  In a latent exponential model, surveillance increased QALYs by 0.76, and in a combined model QALYs increased by 0.53.  The incremental cost-effectiveness ratios between surveillance and no surveillance varied considerably depending on the underlying growth model with ratios ranging from US$34,300 per QALY gained with an  exponential model to $258,400 per QALY gained with a Gompertzian model.  The latent exponential and the combined model yield ICERs of $39,600 and $50,500 per QALY gained, respectively.

Conclusion: The choice of tumor kinetic growth model substantially alters the estimated incremental cost-effectiveness of surveillance for HCC, suggesting that structural sensitivity analyses should be performed with alternative models.  Additional research on methods to select the most appropriate tumor kinetic model should be performed.