Our aim was to demonstrate how a synthetically generated sample of individuals could be used to populate clinical models. Our synthetic sample in this case was generated to match the risk factor variables in the 4S study sample1, using similar distributions and correlation structure compared with the general population (as reported by the Health Survey for England).
A synthetic population of 2,222 people was generated that matched the placebo group characteristics on binary variables (gender, smoking status, diabetes) and continuous factors (age, BMI, systolic BP, total cholesterol: HDL cholesterol ratio, cigarettes and alcohol consumption)2. For more information on how this was achieved please click here.
A stochastic Markov model was then created in R using the Framingham points-based system3 for CVD events and calibrated to match the event outcomes in the 4S placebo group. Reduction in cholesterol with the PCSK9i evolocumab was taken from the FOURIER study4, in which patients were already being treated with simvastatin. Reduction in LDL was 61% in the evolocumab plus simvastatin group and 25% in the simvastatin monotherapy group.
We modelled 1,000 simulations for three treatment arms over 5.4 years. The sample was randomly allocated to either receive simvastatin 20-40 mg daily, evolocumab 420 mg every 4 weeks, or placebo, at a 1:1:1 ratio. At each cycle individuals could experience no change, CVD events (myocardial infarction or ischaemic stroke) or death from these events, non-CVD death, or treatment discontinuation. Figure 1 shows the model cycle.
Myocardial infarctions (MI), ischaemic strokes (IS) and CVD deaths were all lower in the evolocumab group compared with the statin and placebo groups. Differences were significant for IS events and CVD deaths, but not for MI events.
With reference to the placebo group outcomes we found a 28% reduction in MI mortality in the statin group, and 49% reduction in MI mortality in the evolocumab group, a 45% reduction in IS mortality in the statin group, and 70% reduction in IS mortality in the evolocumab group. No differences were found in non-CVD-related deaths between treatment arms and placebo group.