Applying clinical trial results in clinical decision-making: Impact of a patien's baseline risk on the relative effect of an intervention strategy.

Applying clinical trial results in clinical decision-making: Impact of a patient's baseline risk on the relative effect of an intervention strategy.

Typically, clinical trials report a single relative effect measure for large groups of patients. Individual patient characteristics may, however, influence the relative response to treatment. For translation of an overall trial result to the expected absolute treatment effect for an individual patient, possible differences in the relative treatment effect between individual patients should be considered. Analyses in subgroups are often performed to assess differences in the relative effect of a treatment between groups of patients, however, there are several major limitations to these analyses. For one, trials are usually underpowered for subgroup analyses. Performing multiple subgroup analyses will increase the risk of chance findings. This may lead to both over- and underestimation of the treatment effect for a certain group of patients. Secondly, subgroup analyses can only study one characteristic at a time, while overall patient risk is determined by multiple patient characteristics. Relative effect of a treatment may also be dependent on these patient characteristics. In that case, an association between the relative risk reduction and a patient's baseline (untreated) risk can be expected. In the proposed study, the aim is to describe and show a methodology to assess whether relative treatment effect is dependent on baseline risk. Individual patient data from trials provide an opportunity to assess relative treatment effect heterogeneity of patients with varying baseline risks (i.e. risk without the allocated trial intervention). This aids in treatment decisions for individual patients.The RE-LY study data will be used as an example of the proposed methodology to assess effect modification of treatment effect by baseline risk for stroke or systemic embolism, or for major bleeding. Four trials studying the effect of ACE inhibitors will be used as an additional example of the methodology in this manuscript. Baseline risk (risk without the allocated intervention) for a clinical outcome is estimated with a risk model composed of multiple prognostic factors, preferably with an existing risk score. If this is not possible due to e.g. missing variables in the studies, a new risk score may be developed. We will then develop a new model using the linear predictor of the model (which determines the individual's baseline risk) and treatment allocation as predictors in this model. The potential effect modification of the relative treatment effect by baseline risk for the outcome is assessed by adding an interaction term between the two variables to the model, and then comparing whether this model is better in predicting the outcome than a model without this interaction term.