Abstract
Background: We evaluated the life-time health outcomes and direct costs of first-line rituximab, fludarabine and cyclosphosphamide (R-FC) treatment for chronic lymphocytic leukemia (CLL) patients in US clinical practice, using long-term data from a retrospective cohort comparison (Tam et al., 2008). Additionally, prognostic factors were examined for association with treatment outcomes.
Methods: A Cox analysis was conducted to assess for potential heterogeneity and treatment association with baseline prognostic factors. Baseline prognostic factors included: age, gender, beta-2 microglobulin (β2M) and Rai stage. Different lengths of follow-up in FC (1995–2007, n= 108) and R-FC (1999–2007, n=300) treated patient cohorts were also incorporated into the analysis. In the cost-effectiveness model, patients were assumed to be in one of three health states; PFS, Progressed or Death. The best parametric fit (Weibull) was used to extrapolate PFS beyond the end of the cohort follow-up period to a 30 year life-time horizon. The number of patients in each treatment arm that died while in PFS was based on the maximum of either the observed rate of death or background mortality. Because median overall survival had not been reached, a Markov process was constructed to model the transition from the progressed health state to death. Given the non significant difference in post progression survival by treatment (R-FC or FC), patients transitioning from progression to death were modeled as a single population with mean time to death (Kaplan-Meier) converted to a monthly probability of dying. This approach is conservative in that treatment benefit is exclusively a function of time spent in PFS. To account for quality of life and estimate the Quality Adjusted Life Years (QALYs), the predicted time in each health state was weighted using CLL utility scores (Hancock et. al. 2002). Direct costs were estimated using Medicare reimbursed rates, MS-DRGs for CLL and published drug prices, and include the cost of administration and adverse events. Costs (in USD) and QALYs were both discounted at 3% per annum.
Results: Prognostic factors were evenly distributed between treatment groups. In univariate Cox models, age, Rai stage and β2M were confirmed as prognostic factors. For β2M, the hazard ratio (HR) was 2.41 (1.72–3.38) ≥2x upper limit normal (N) compared to <2N. Similar significant increases were observed in the elderly (>70 years) and patients with Rai C stage. The treatment effect of R-FC versus FC adjusted for β2m, Rai and age (HR 0.54 (0.38–0.77), was broadly similar to univariate estimate (HR 0.57 (0.40–0.81). Compared to FC, R-FC was estimated to generate an additional 2.19 years in mean life expectancy and an additional 2.53 years of PFS. After adjusting for health-related quality of life, the estimated incremental QALYs for R-FC compared to FC was 1.82 years. Assuming a shorter time horizon of 15 years, R-FC generated an additional 1.41 years in mean life expectancy and an additional 2.04 years of PFS versus FC. Total direct costs were higher for R-FC by $22,503 per patient, which was partially offset by a reduction in total medication and monitoring costs incurred in the progressed health state. The incremental cost-effectiveness ratio was $12,382 per QALY gained for R-FC. The results of the sensitivity analysis provided reassurance that the assumptions made were acceptable and that the results held under most plausible assumptions.
Conclusion: The treatment benefit of R-FC over FC in this CLL observational cohort is not affected by prognostic factors. R-FC patients experienced longer PFS which translated into a considerable increase in life expectancy at an acceptable cost to the US healthcare system.
Neighborhood Disadvantage and Active, Disabled and Total Life Expectancy Among Community-living Older Persons
