Application of Economic Analyses in U.S. Managed Care Formulary Decisions: A Private Payer's Experience

BACKGROUND: Promoting use of pharmacoeconomic models by formulary reviewers is a goal of the Academy of Managed Care Pharmacy (AMCP) Format for Formulary Submissions, but relatively few decision makers use such models, and many doubt that they provide meaningful input. OBJECTIVES: To demonstrate how sophisticated disease-based pharmacoeconomic models can aid formulary decision makers when long-term outcomes data are lacking. METHODS: The Center for Outcomes Research (CORE) Diabetes Model (CDM), a published, validated Markov pharmacoeconomic model that projects clinical and economic endpoints, was used to model the cost-effectiveness of exenatide, a new injectable antidiabetic agent that enhances glucose-dependent insulin secretion, in a standard cohort of type 2 diabetes patients (mean body mass index [BMI] = 27.5 ± 3 kg/m2), compared with a modified obese cohort (mean BMI = 35 ± 3 kg/m2) that was otherwise demographically identical at baseline to the standard cohort. The standard cohort was assumed to maintain baseline weight during treatment, and the modified obese cohort was assumed to experience weight loss of approximately 9% (mean = 3 kg/m2), with corresponding improvements in blood pressure, low density lipoprotein cholesterol, and triglycerides. We selected a 30-year time horizon because it was the time interval during which the CDM predicted most of the subjects would have died, and the costs obtained thus reasonably projected lifetime total direct medical costs for these cohorts. While treatment options certainly will change over a 30-year period, our goal was to estimate the incremental effect of exenatide over other available therapies. RESULTS: The model predicted reduced long-term treatment costs in obese patients, driven by an 11% decrease in cardiovascular disease burden and derived from the presumed weight loss. The incremental cost-effectiveness ratio (ICER) for adding exenatide over 3 years was $35,000/quality-adjusted life-year (QALY). Using a 30-year horizon, ICER values were $13,000/QALY versus insulin, $32,000 versus generic glyburide, and $16,000 versus no additional treatment. Exenatide dominated pioglitazone. By comparison, the 30-year ICER for exenatide versus insulin in the nonobese cohort was $33,000. These results were presented to the pharmacy and therapeutics (PandT) committee and influenced its decision to add exenatide to the drug formulary. While our modeling assumed certain patient characteristics (e.g., obesity, need of further A1c reduction at baseline, motivation to lose weight), the PandT committee imposed only a step-therapy requirement to try either metformin or a sulfonylurea before trying exenatide and did adopt a nonspecific requirement for physician reauthorization of refills before the fourth pharmacy claim for exenatide. CONCLUSIONS: Disease-based pharmacoeconomic models may help third party payers project costs and be particularly useful when only data from short-term clinical trials are available. In the present case, the pharmacy staff of a health plan used a pharmacoeconomic model for drug treatment of type 2 diabetes provided by the manufacturer as part of the AMCP Format dossier process to project cost outcomes for exenatide, adjunct injectable therapy for patients taking metformin and/or sulfonylurea. The PandT committee approved the drug for inclusion in the drug formulary based in part on the results of the pharmacoeconomic model produced from the cost inputs entered into the model by the health plan pharmacists.