Personalized cardiovascular disease prevention by applying individualized prediction of treatment effects

Large-scale randomized clinical trials have established the efficacy of cholesterol-lowering, blood pressure-lowering, and anti-platelet therapy to prevent cardiovascular diseases. A challenge for clinicians is to apply group-level evidence from these trials to individual patients. Trials typically report a single treatment effect estimate which is the average effect of all participants, comprising patients who respond poorly, intermediately, and well. Clinicians would preferably make patient-tailored treatment decisions. Therefore, one would require an estimate of an individual patient's response to therapy. Although not yet widely recognized, trials contain this type of information. In this paper, we show how available information from landmark trials can be translated to an individual 'treatment score' through the use of multivariable therapeutic prediction models. These models provide an individual estimate of the absolute risk reduction in cardiovascular events given the specific combination of multiple clinical characteristics of a patient under care. Based on this individualized treatment estimate and metrics such as the individual number-needed-to-treat, clinicians together with their patients can decide whether drug treatment or what treatment intensity is worthwhile. Selective treatment of those who can anticipate the greatest benefit and the least harm on an individualized basis could reduce the number of unnecessary treatments and healthcare costs beyond that currently achievable by subgroup analyses based on single patient characteristics.