True and false interindividual differences in the physiological response to an intervention.

Abstract

What is the topic of this review? In 'personalized medicine', various plots and analyses are purported to quantify individual differences in intervention response, identify responders/non-responders and explore response moderators or mediators. What advances does it highlight? We highlight the impact of within-subject random variation, which is inevitable even with 'gold-standard' measurement tools/protocols and sometimes so substantial that it explains all apparent individual response differences. True individual response differences are quantified only by comparing the SDs of changes between intervention and comparator arms. When these SDs are similar, true individual response differences are clinically unimportant and further analysis unwarranted. Within the 'hot topic' of personalized medicine, we scrutinize common approaches for presenting and quantifying individual differences in the physiological response to an intervention. First, we explain how popular plots used to present individual differences in response are contaminated by random within-subject variation and the regression to the mean artefact. Using a simulated data set of blood pressure measurements, we show that large individual differences in physiological response can be suggested by some plots and analyses, even when the true magnitude of response is exactly the same in all individuals. Second, we present the appropriate designs and analysis approaches for quantifying the true interindividual variation in physiological response. It is imperative to include a comparator arm/condition (or derive information from a prior relevant repeatability study) to quantify true interindividual differences in response. The most important statistic is the SD of changes in the intervention arm, which should be compared with the same SD in the comparator arm or from a prior repeatability study in the same population conducted over the same duration as the particular intervention. Only if the difference between these SDs is clinically relevant is it logical to go on to explore any moderators or mediators of the intervention effect that might explain the individual response. To date, very few researchers have compared these SDs before making claims about individual differences in physiological response and their importance to personalized medicine.