Spurious results in therapeutic drug monitoring research

Maximal correlation between measured blood concentration of a drug and an estimate of the area under the concentration-time curve (AUC) is widely used as criterion for the optimal blood sampling time-point in therapeutic drug monitoring (TDM) research. (More generally, the correlation between an estimate of AUC and a linear combination of several concentration measurements is considered, but the principles are the same.) This particular TDM research methodology is evaluated from a theoretical statistical perspective by considering a general nonspecific study. It is shown that the TDM research methodology produces spurious results because the optimal time-point is determined by irrelevant factors. Particularly, the sampling design is an important determinant. The sampling time-points are of course the only candidates for the optimal time-point, but they may also determine which candidate is optimal. In a special case, it is mathematically proven that any time-point except the first (trough level) can be made optimal by choosing the appropriate sampling design. This is probably true in all practical situations. The theoretical optimum is defined as the optimal time-point in the ideal theoretical sampling design where concentration measurements are made continuously in time. Hence, the theoretical optimum is independent of sampling designs, and the optimal time-point of a study is an approximation to the theoretical optimum. In a homogeneous study population, it can be proven, mathematically and under realistic assumptions, that the theoretical optimum is t(max). Particularly t(max), is the individual theoretical optimum. Heterogeneity of the study population can be an important determinant of the optimal time-point. In significantly heterogeneous study populations, the optimal time-point is usually extreme compared with the distribution of the individual optimal time-points in the population.