Background:
Pharmacoeconomic models in cardiovascular disease (CVD) are typically developed on the basis of an assumed relative risk reduction (RRR) for new treatments which is in turn applied to estimates of baseline risk in patients subject to standard of care medications for coronary heart disease. However, baseline risk of CVD varies by relevant patient characteristics (e.g., age, gender, event history) and additional risk factors (e.g., smoking, diabetes, chronic kidney disease, lipids, blood pressure, obesity) which can be applied across diverse populations and settings. This study aimed to find and adapt a prognostic equation to adjust baseline risk of major adverse cardiovascular events (MACE=CV death, non fatal MI, Stroke), in a general pharmacoeconomic model, based on population risk factor prevalence.
Methods:
In order to find an equation that includes all relevant risk factors, a literature search was conducted to update a review (published 2008) of 70 pre-2005 primary cardiovascular risk equations. Once an equation was selected, a relative risk was calculated from the ratio of estimated one-year CV risks of the modeled population versus the population in the selected equation. An initial assumption of independence of risk factors was made. The resulting ratio is adapted as a risk multiplier to estimate risk when modeling specific trial populations or subpopulations.
Results:
None of the original 70 equations included all of the six factors of interest, and only two contained kidney disease. In addition to updates and modifications of the 70 equations, four new studies were identified, including the UK-based QRISK2 which was the only prognostic equation that uses all six factors. QRISK2 was developed using over 16 million person-years’ worth of observations with 140,000 cardiovascular events, and was validated against the UK THIN population. The adapted QRISK2 equation estimates the first-year CV risk for a population adjusted for the six additional risk factors to be 7.1%. However the subset with diabetes and/or kidney disease, which makes up about 43% of the population, has an estimated risk of about 11%. If a treatment is assumed to reduce CV risk by 15% in both the entire and subset populations, then the number needed to treat to avoid one event is reduced from 93 for the entire population to 61 for the diabetes and/or renal impairment subset.
Conclusions:
Estimating the ratio of absolute risks for specific populations and subsets using QRISK2 can help a model to predict the value of targeting relevant subpopulations for a new treatment. When this risk adjustment is included in a lifetime model, additional benefits from secondary prevention, and increased likelihood that a treatment is cost effective in a UK context, can also be estimated. For non-UK populations, calibration of QRISK2 is needed or an alternative equation must be used.