Analytical approach:
A within-trial analysis was conducted, with the efficacy, resource use, and productivity losses all estimated from one clinical trial. The time horizon was the follow-up period, which was a median of 4.8 years. A Markov model was also constructed to simulate the ongoing risks of MI, revascularisation, and death, over a lifetime horizon. The data were from a variety of sources, based on a model used in Denmark, Norway, and Sweden (Lindgren, et al. 2007, see 'Other Publications of Related Interest' below for bibliographic details). The cycle length was one year and all patients were switched to simvastatin after five years. The age of the cohort at the start of the model was 62 years, the proportion of women was 30%, the proportion of participants with diabetes was 25%, and the baseline low-density lipoprotein cholesterol (LDL-C) level was 2.75 mmol/L (106 mg/dL). The authors stated that a Canadian societal perspective was taken for both analyses.
Effectiveness data:
The effectiveness data were from a prospective, randomised, open-label, blinded-end-point trial, called the Incremental Decrease in End-Points Through Aggressive Lipid-Lowering (IDEAL) trial. This followed-up 8,888 patients with a previous MI, who were from Northern Europe, over a median of 4.8 years. For the within-trial analysis, the end point-related events, such as stroke, heart failure, and cardiac transplant, from the clinical trial, were aggregated to produce the mean number of events per patient, for each arm. For the Markov model, the baseline risks of MI, revascularisation, and death with simvastatin were estimated by applying a Weibull regression calculation to the data from patients who had experienced the relevant event with simvastatin. The relative risk ratio of MI and revascularisation, with atorvastatin compared with simvastatin, were based on the trial and a published paper. It was assumed that the risk of death was the same for patients in the health state after revascularisation as in the health state at risk and it was assumed to be the same in both treatment arms. The main effectiveness outcome was the LDL-C level.
Monetary benefit and utility valuations:
The utility values for the Markov model were based on a published Swedish population survey that used the European Quality of life (EQ-5D) questionnaire. They were adjusted using a utility decrement from a cross-sectional survey, using the EQ-5D, of patients hospitalised at coronary care units in Sweden. An additional decrement was applied after an MI or a revascularisation. The values were assumed to return to at-risk levels in the second year after an event.
Measure of benefit:
In the within-trial analysis, the primary measure of benefit was the number of events avoided (including stroke, transient ischaemic attack, nonfatal MI, heart failure, angina pectoris, cardiac arrest, percutaneous coronary intervention, coronary artery bypass graft, cardiac transplant, other cardiothoracic procedures, and other vascular procedures) and these were not discounted. In the Markov model, the primary measure of benefit was quality-adjusted life-years (QALYs) and the secondary measure of benefit was life-years gained (LYG); both of these were discounted at a rate of 5% per annum.
Cost data:
The cost categories included hospitalisation, medications, and productivity lost, and these were collected from the IDEAL trial. In the within-trial analysis, the resource use and employment status were based on patient-level data collected during the trial. The unit costs were from standard sources. The cost of cardiothoracic procedures was assumed to be 1.39 times the cost of a bypass graft, while the cost of other vascular procedures was assumed to be the same as a percutaneous coronary intervention. These costs were not discounted.
In the Markov model, the costs for the at-risk health state included all the costs of a MI or revascularisation, while productivity losses were assumed to be zero. The resource use and productivity losses (compared with baseline) for the post-MI and post-revascularisation health states were based on data from the trial. The annual costs were assumed to return to those for the at-risk health state in the fourth year following an MI and the fifth year following a revascularisation; the cost of the event itself was included in the first-year after the event. These costs were discounted at a rate of 5% per annum.
The costs were reported in Canadian dollars (CAD) and the price year was 2006. Any cost adjustments were based on the Health and Personal Care component of the Consumer Price Index.
Analysis of uncertainty:
The uncertainty was explored using one-way, threshold, and probabilistic sensitivity analyses. In the within-trial analysis, the parameter uncertainty was explored, using threshold analysis of the acquisition costs of simvastatin, when the incremental costs of the two treatment arms were equal. Bootstrapping of the patient-level data, using angular transformation, was also conducted, with 1,000 samples, to produce confidence intervals. In the Markov model, the parameter uncertainty was explored using one-way sensitivity analysis and non-parametric bootstrapping of the patient-level data, with 1,000 simulations. The results were presented on a cost-effectiveness plane and a cost-effectiveness acceptability curve.