Tuesday, October 26, 2010: 10:15 AM
Grand Ballroom Centre (Sheraton Centre Toronto Hotel)
Bob J.H. van Kempen, MSc1, Bart S. Ferket, MD1, Rogier L.G. Nijhuis, MD, PhD2, Sandra Spronk, PhD1 and M.G. Myriam Hunink, MD, PhD1, (1)Erasmus MC, Rotterdam, Netherlands, (2)ZGT Hengelo, Hengelo, Netherlands
Purpose: methods of modeling the effect of statins in simulation studies vary amongst papers that have been published. In this abstract, we will illustrate the impact of using different modeling methods on the optimal decision. Method:
a previously developed and validated Monte Carlo-Markov model based on the Rotterdam Study, a cohort study of 6871 individuals aged 55 years and older with 7 years follow-up, was used. Life courses of 3501 participants with complete risk profiles on statin treatment vs. no statin treatment were simulated using six health states (well, coronary artery disease (CAD), stroke, both CAD and stroke, and death). Transition probabilities were based on 5-year risks predicted by Cox regression equations, including (amongst others) total and HDL cholesterol as covariates. We used three different methods to model the effect of statins on the incidence of CAD: (1) statins lower total cholesterol levels and increase HDL, which through the covariates in the Cox regression equations leads to a lower incidence of CAD; (2) statins decrease the incidence of CAD directly through a relative risk reduction (RRR), assumed to be the same for each individual; (3) the RRR with statin therapy on the incidence of CAD is made proportional to the absolute reduction in LDL-cholesterol levels, for each individual. Each of the three statin modeling alternatives was compared to the no statin strategy. Result:
in the 3501 subjects (mean age 69 ± 8.47, 39% men), lifeyears simulated for each of the three methods were: (1) 17.241, (2) 17.705 and (3) 17.709 years. At a willingness-to-pay of $ 50,000, net health benefits were (1) 9.67, (2) 9.87 and (3) 9.87. Figure 1 shows the probability that statin treatment is cost effective for each of the three methods, for varying willingness-to-pay thresholds. Conclusion:
the choice of modeling method of the effectiveness of a drug in simulation studies can potentially influence the optimal decision and the uncertainty associated with it.
Candidate for the Lee B. Lusted Student Prize Competition
