Analytical approach:
A Markov cohort model was used to extrapolate the long-term costs and benefits of the interventions beyond the short follow-up of the clinical trials. The authors stated that the perspective was that of the UK NHS.
Effectiveness data:
The authors employed several published sources to inform the model. The baseline transition probabilities were from a recent German prospective, multicentre, observational study and from the Global Registry of Acute Coronary Events (GRACE) for the UK. The clopidogrel transition probabilities were from two pivotal trials of this drug for STEMI: CLARITY-TIMI 28 and COMMIT/CCS-2. Both of these were randomised, double-blind, placebo-controlled trials that compared the addition of clopidogrel to aspirin alone or with fibrinolytics against the addition of placebo. The model was run separately for each trial. Other data were from a primary analysis of data from the Nottingham Heart Attack Register (NHAR).
Monetary benefit and utility valuations:
The utility values and their uncertainty for the stroke health states were from a meta-analysis (Tengs, et al. 2007, see 'Other Publications of Related Interest' below for bibliographic details). The Harvard CEA Registry of utilities was used for the other health states.
Measure of benefit:
The main summary benefit measure was quality-adjusted life-years. A secondary summary measure was life-years, and the number of cases of myocardial infarction and stroke were reported. The QALYs were discounted at a rate of 1.5%.
Cost data:
The costs for drug treatments were from the British National Formulary. The health state costs were from a patient-level resource use publication (Palmer, et al. 2002, see 'Other Publications of Related Interest' below for bibliographic details). The cost of stroke was from a relevant UK study. All costs were expressed in 2006 UK pounds sterling (£) and they were discounted at a rate of 6%.
Analysis of uncertainty:
Two base-case analyses were undertaken, one for each trial (CLARITY-TIMI 28 and COMMIT/CCS-2). In each case, a univariate sensitivity analysis was carried out to investigate the individual impact of key parameters on the results. Probabilistic sensitivity analyses were conducted, using Monte Carlo simulation, with all the effectiveness, utility, and cost parameters assigned distributions. The results of the probabilistic sensitivity analyses were presented in cost-effectiveness acceptability curves.