A probabilistic Markov model was constructed to evaluate the cost-effectiveness of the two treatments in a hypothetical cohort of 1,000 postmenopausal women with hormone receptor-positive early breast cancer. The health states were reported and the time horizon was 25 years. The authors reported that the perspective was that of the US health care system.
The effectiveness data were mainly derived from a five-year completed treatment trial; the 'Arimidex', Tamoxifen Alone or in Combination (ATAC) trial. These data were augmented with data from the published literature. A Weibull survival curve was used to extrapolate the recurrence rate beyond the end of the clinical trial. The main clinical parameters included the rates of local or regional and distant recurrence, death due to breast cancer or other causes, and treatment adverse events.
Monetary benefit and utility valuations:
The utility estimates were obtained from a published cross-sectional study of 44 women with a mean age of 67.5 years and with early breast cancer. This study used the chained standard gamble method to estimate patients’ preferences.
Measure of benefit:
The measures of benefit were quality-adjusted life-years (QALYs) and life-years gained (LYG). These benefits were discounted at an annual rate of 3%.
The cost categories included the costs of diagnosis and treatment, follow-up and monitoring, adverse events, and death. The resource use data were mainly derived from published literature and augmented by data derived from structured interviews with nine US oncologists. The cost data were mainly from national sources and published literature, and, where this information was not available, the data were based on the expert opinion of the oncologist panel. All costs were reported in US dollars ($) for the price year 2002 to 2003 and they were discounted at an annual rate of 3%.
Analysis of uncertainty:
The parameter uncertainty was investigated using one-way sensitivity analysis on the key model parameters and assumptions. The results were presented in a tornado chart. Probabilistic sensitivity analysis was also completed, with all the parameters in the model being assigned prior probability distributions. Incremental cost-effectiveness ratios were calculated and cost-effectiveness acceptability curves were generated.