The Impact of Methodological Approach on Cost Findings in Comparison of Epoetin Alfa with Darbepoetin Alfa

Background/Objective: There is interest in longer acting therapies with extended dosing schedules as a mechanism to improve patient compliance and/or decrease health resource utilization (in the case of therapies administered in physician offices/clinics). Based on randomized controlled trial (RCTs) data, the duration of clinical benefit (DCB) of Epoetin alfa (EA) ranges from 2–7 days depending on dose and darbepoetin alfa (DA) ranges from 7–21 days. Comparison of weekly cost between DA and EA in claims data analysis may not be valid if their estimated DCBs are not taken into account. Thus the objective of this study was to use a case study to examine the costs of epoetin alfa (EA) and darbepoetin alfa (DA), 2 erythropoiesis stimulating agents (ESAs), for cancer patients with anemia due to cancer, chemotherapy or radiation treatment.

Methods: The MarketScan Research Database, a large national claims database, was used to build EA and DA treatment episodes among patients with a cancer diagnosis on at least 2 medical or facility claims, with no diagnoses indicating chronic renal disease, and with at least 1 EA or DA medical or pharmacy claim. Claims were from patients who were continuously enrolled for benefits during the 6-month period from January-June 2005. Base-case average weekly costs were calculated, with the episode of care (EOC) being the number of days from first date of service for an ESA claim to the last date of service for an ESA claim or until there was a 42-day or longer gap in ESA therapy. Each episode was augmented with an estimated duration of clinical benefit calculated from the last dose administered in the EOC. DCB estimates were based on the results of randomized controlled trials: EA doses <35,000 units =2 days and >=35,000 units =7 days; DA doses <=100 mcg =7 days, 101–299 mcg =14 days, and >=300 mcg =21 days. Cost was reimbursed amount observed in the claims database.

Results: Analysis-of-variance (ANOVA) models estimated an adjusted average weekly cost controlling for difference in clinical/demographic differences across EA and DA users. Younger patients, those with more comorbidities, and those with advanced cancer, were found to have increased likelihood of DA use (all P<0.001). After accounting for DCB, the average weekly dose across all episodes was 43,113 units for EA and 94 mcg for DA; the average weekly cost of DA was significantly lower than that of EA ($561 vs. $983, P<0.001). Sensitivity tests on patient selection criteria, length of follow-up period, and definitions of average weekly cost and dose had minimal impact on findings, while changes were more dramatic when different DCB rules were applied. DA had lower costs than EA in all sensitivity analyses except when single dose episodes were excluded and no DCB adjustment was incorporated. With ANOVA adjustment to equalize the population characteristics, estimated average weekly cost was $647 for DA and $833 for EA in the base case..

Conclusion: Failure to adjust for differences in patient demographic and clinical characteristics between users of EA and DA can result in misleading comparisons. Cost calculations are sensitive to incorporating the assumption of duration of clinical benefit and the decision to drop and/or retain single dose episodes. To reduce the risk of potential bias, these factors should be taken into account when using retrospective claims data to conduct cost comparisons between agents that have significant differences in dosing schedule.