ABSTRACT
Until now we have assumed that the effect of exposure on the response is instantaneous. However, this assumption is rarely true, although sometimes reasonable as an approximation. In pharmacometrics, in most cases, the PD change lags behind the PK change that caused it, a phenomenon known as hysteresis. In this case, it is important to investigate and to model the temporal relationship between PK and PD. Ignoring hysteresis may induce bias in the estimation of exposure-response relationships. Even when hysteresis does not cause significant bias, if not dealt with properly, it may increase the residuals, hence reduce the accuracy/power of statistical inference. Often the temporal relationship itself may also be important, e.g., to predict the time of onset of an adverse event or the efficacy response to treatment. Csajka and Verotta (2006) give a comprehensive review of PKPD modeling , in particular, hysteresis in PKPD relationships. There are mainly two ways to model the temporal relationship between the exposure and response. One assumes that the response model is instantaneous, but it depends on a latent “exposure”, which follows a dynamic model that links to the exposure we observe. The other is based on a dynamic response model describing the natural course of the disease process, e.g., tumor growth. The exposure affects the disease process by changing its natural course.
