Assessing Variability in Systemic Drug Exposure

Some drugs (e.g. intravenously administered medications with well-known pharmacokinetics) have relatively consistent and predictable dose-exposure relationships. The blood levels of such drugs can be estimated reasonably well and measuring them may not provide additional valuable information. Here, TDM is of limited relevance, except perhaps for cases of medication non-compliance or failure of organs involved in drug metabolism and excretion. For many orally administered medications, on the other hand, numerous additional variables such as bioavailability and first-pass metabolism widen the range of possible blood level concentrations and make predicting systemic drug exposure from the dose much more difficult. Thus, the earliest evidence in support of TDM for a new oral anticancer drug would be derived from observation of a large range of not otherwise predictable blood drug levels between patients on the same treatment regimen (Figure 1, Stage 1).
There are several ways of measuring the systemic exposure to a drug. Examples include comprehensive sampling of numerous time points to determine area under the concentration (AUC) time curves as well as parsimonious or limited sampling strategies such as trough (Cmin) levels, peak levels, or a combination thereof. For routine monitoring of oral drugs that are taken once or twice daily in an outpatient setting, measuring the drug concentration in a single sample collected prior to the next dose (i.e. trough level monitoring) is often the only practical option. Even during phase 2 and phase 3 trials, blood sampling is restricted and, if possible, sparse sampling strategies (e.g. trough concentrations) should be used to study PK and PK/PD. We will therefore only focus on trough level monitoring.
The relationship between a trough level and the systemic exposure, as determined by area under the concentration (AUC) time curves, can often be obtained from phase 1 and phase 2 clinical studies.8–10 Although most early-phase trials collect the data to derive this correlation, it may not be explicitly reported. Using phase 1 and 2 study data, one can get a reasonable idea if trough level monitoring could be used as a proxy for the more comprehensive AUC analysis. It is important to keep in mind that a clinical drug development PK study is usually much more controlled in terms of drug intake and sample collection than routine patient care and that parameters obtained in such studies may not translate to real world patients. Therefore, there may be added value from assessing the relationship between trough levels and AUC in a patient care setting. Factors that may confound the relationship between Cminand AUC include PK drug-drug interactions, alterations in PK as a result of (auto-)induction, and inhibition of metabolizing enzymes and transporters, as well as inaccuracies in determining the triad of time of drug intake, time of sample collection, and half-life of a drug. All components of this triad are relevant for an accurate assessment of the systemic exposure and all may differ from patient to patient.11 Nevertheless, it is intriguing that less-than-perfect correlation between trough levels and systemic exposure can still be useful for assessing systemic exposure. Indeed, even for some of the most monitored drugs that utilize trough levels, such as cyclosporine and tacrolimus, the correlation coefficient between trough levels and systemic exposure is in the 0.7-0.8 range.12,13
For effective TDM, in addition to being able to measure systemic exposure (e.g. trough levels), one must also be able to predict how changes in dosing will change the drug exposure. Consequently, the dose-exposure relationships must be well-characterized. To this end, serial sampling of drug exposure in the same individual over time provides crucial information and should be incorporated into precision dosing studies whenever possible.14 First, it enables evaluation of intra-individual exposure variability over time. This helps estimate how well the systemic exposure can be predicted from dose alterations. Second, it improves estimation of the total systemic exposure over the course of treatment. Finally, it allows for determination of additional parameters, such as the maximum or minimum blood drug concentrations, which may also be relevant for predicting drug efficacy, resistance, and toxicity.
The inter- and intra-individual PK variability and the strength of correlation between trough levels and AUC are important considerations for calculation of sample sizes in clinical studies. These parameters should guide not only the number of study participants but also the number of samples per individual as well as the sampling frequency.