Monday, October 24, 2011: 11:42 AM
Grand Ballroom EF (Hyatt Regency Chicago)
(MET) Quantitative Methods and Theoretical Developments

Candidate for the Lee B. Lusted Student Prize Competition

Julia F. Slejko, BA1, Patrick W. Sullivan, PhD2, Kavita V. Nair, PhD1, P. Michael Ho, MD, PhD3, Heather D. Anderson, PhD1 and Jonathan D. Campbell, PhD1, (1)University of Colorado School of Pharmacy, Aurora, CO, (2)Regis University, Denver, CO, (3)University of Colorado and US Department of Veterans Affairs, Denver, CO

Purpose: Because real-world patients may not exhibit the same level of medication adherence seen in clinical trials, the effectiveness of medications in routine practice may differ.  Cost-effectiveness analysis (CEA) models often do not incorporate adherence variation.  Furthermore, the Markovian assumption does not allow adherence history to affect future event probabilities.  We created a framework incorporating adherence history into a Markov model using the example of Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER).  

Method: Prescription claims records for primary prevention statin users were obtained using the IMS LifeLink Health Plan Claims Database.  Yearly adherence was measured as the proportion of days covered (PDC) for three years following statin initiation and was categorized as A0 (PDC=0), A1 (0<PDC≤.33), A2 (.33<PDC≤.66), or A3 (PDC>.66).  Yearly adherence transitions were incorporated into a Markov microsimulation using TreeAge software.  Tracker variables and global matrices stored adherence transitions which were used to adjust statin costs and subsequent probabilities of cardiovascular events over the patient’s lifetime.  Statin effectiveness was adjusted between 0% (level A0) and 100% (level A3) of trial-based risk reduction.  10,000 microsimulations were used to estimate incremental cost-effectiveness ratios (ICERs) as US dollars per quality-adjusted life-year (QALY).  The model was an extension of the authors’ previously published JUPITER CEA model in which adherence was not incorporated ("adherence-naïve"). 

Result: Among 27,862 new statin users, 58% began the first year of statin use in level A3, while 20% and 22% were in levels A2 and A1, respectively.  By year three, we found a significant decrease in adherence. 32% of patients were in level A3, 15% in A2, 20% in A1 and 33% in A0.  The model incorporating adherence resulted in an ICER of $23,459/QALY while the ICER of the adherence-naïve model was $11,127/QALY.  Patient subgroup analysis revealed that the ICER for patients beginning in level A1 was $52,214/QALY while the ICER for patients beginning in level A3 was $17,578/QALY.  The ICER for patients remaining in level A3 for three years was $8,347/QALY.

Conclusion: Patient-level simulations that include adherence behavior reveal value differences not seen in a cohort model based on the “average” patient.  In the interest of patient-centered outcomes research and personalized medicine, this approach adds insight to how patient subgroups may benefit from adherence-improving interventions.