The Calculation of a Confidence Interval on the Absolute Estimated Benefit for an Individual Patient

Abstract

Physicians need a method to predict the individualized absolute therapeutic benefit before deciding which therapy to prescribe to a given patient and the confidence intervals around this estimate. We have derived a method to predict the absolute individual therapeutic benefit in a previous work. In this paper, we present a Monte Carlo simulation to estimate the bias of prediction for an individual with certain characteristics and use a bootstrap method to compute its confidence intervals. Because the bootstrap approach does not depend upon the parametric assumption for the distribution of the prediction, it can be applied to situations where the parametric distribution is unknown. Over 35,000 cases of subjects at risk of cardiovascular events were available for analysis. Our results show the 95% confidence intervals for each individual. In a clinical setting, the use of this approach makes it possible to predict the absolute therapeutic benefit for each patient (the quantity of individual benefit) with sufficient precision.