Do Different Methods of Modeling Statin Treatment Effectiveness Influence the Optimal Decision?

Purpose. Modeling studies that evaluate statin treatment for the prevention of cardiovascular disease (CVD) use different methods to model the effect of statins. The aim of this study was to evaluate the impact of using different modeling methods on the optimal decision found in such studies. Methods. We used a previously developed and validated Monte Carlo-Markov model based on the Rotterdam study (RISC model). The RISC model simulates coronary heart disease (CHD), stroke, cardiovascular death, and death due to other causes. Transition probabilities were based on 5-year risks predicted by Cox regression equations, including (among others) total and high-density lipoprotein (HDL) cholesterol as covariates. In a cost-effectiveness analysis of implementing the ATP-III guidelines, we evaluated the impact of using 3 different modeling methods of statin effectiveness: 1) through lipid level modification: statins lower total cholesterol and increase HDL cholesterol, which through the covariates in the Cox regression equations leads to a lower incidence of CHD and stroke events; 2) fixed risk reduction of CVD events: statins decrease the odds of CHD and stroke with an associated odds ratio that is assumed to be the same for each individual; 3) risk reduction of CVD events proportional to individual change in low-density lipoprotein (LDL) cholesterol: the relative risk reduction with statin therapy on the incidence of CHD and stroke was assumed to be proportional to the absolute reduction in LDL cholesterol levels for each individual. The probability that the ATP-III strategy was cost-effective, compared to usual care as observed in the Rotterdam study, was calculated for each of the 3 modeling methods for varying willingness-to-pay thresholds. Results. Incremental cost-effectiveness ratios for the ATP-III strategy compared with the reference strategy were (sic)56,642/quality-adjusted life year (QALY), (sic)21,369/QALY, and (sic)22,131/QALY for modeling methods 1, 2, and 3, respectively. At a willingness-to-pay threshold of (sic)50,000/QALY, the probability that the ATP-III strategy was cost-effective was about 40% for modeling method 1 and more than 90% for both methods 2 and 3. Differences in results between the modeling methods were sensitive to both the time horizon modeled and age distribution of the target population. Conclusions. Modeling the effect of statins on CVD through the modification of lipid levels produced different results and associated uncertainty than modeling it directly through a risk reduction of events. This was partly attributable to the modeled effect of cholesterol on the incidence of stroke.