Clinical Pharmacology – Chemical
2019 PharmSci 360
To date, exposure-response (E-R) analysis has been well recognized and frequently utilized in support of regulatory approval. The quantitative relationship between drug exposure and response with regard to efficacy and safety is critical in justifying or confirming a dosing regimen in the context of the overall risk/benefit balance. Frequently, E-R analyses performed in oncology were conducted using time-independent models such as logistic regression and Cox regression models. This type of models uses a summary-level exposure measure such as average trough concentration without accounting for the time-varying exposure due to dose modifications. However, dose reduction and interruption driven by AEs are quite common in oncology clinical trials. Dynamic adjustments in dose can lead to changes in systemic exposure throughout the treatment duration, causing bias in the results generated from time-independent exposure efficacy analyses. Patients generally have a higher drug concentration in the earlier cycles compared to that in the later cycles. The longer the treatment duration, the higher chance there is for dose reduction and dose interruption. Potentially, patients who benefit from the treatment stay longer in the trial, and may therefore appear to have a lower PK concentration. On the other hand, patients who progress early would discontinue from the trial early with less opportunity for dose reduction or interruption. These patients would appear to have a higher PK, causing a spurious reverse relationship between PK and response. In this program, we will present a more appropriate E-R analysis approach, which takes into consideration the entire concentration-time profile of the drug, i.e., to associate the exposure at different time intervals with the corresponding efficacy RECIST endpoints such as overall response and progression free survival in a longitudinal fashion. The application of this time-dependent E-R analysis approach could potentially address the aforementioned limitation of time-independent analysis approach using average exposure measure. Using ceritinib, an oral ALK inhibitor approved for ALK-positive NSCLC, as an example, analyses based on phase 3 data illustrate that the dynamic E-R relationship in clinical trials with long duration could be better analyzed by using time-dependent models such as repeated-measures logistic regression model and time-dependent cox regression model. By using time-varying PK and the paired response outcome at each assessment interval, richer information at each time interval can be fully utilized for better and more accurate characterization of drug effect on efficacy outcome.