2.1.1. Pharmacokinetic Parameters
For monoclonal antibodies, particularly those that demonstrate time-dependent PK, the selection of the appropriate exposure metric to use for E-R analyses is critical. An E-R analysis simulation for nivolumab, for example, tested 3 different exposure metrics and resulted in different E-R conclusions. The exposure metrics used were: (1) average concentration at steady state (Cavgss), (2) average concentration at cycle 1 (Cavg1st-dose), and (3) trough concentration at cycle 1 (Cmin1st-dose).12 When Cavgss was used as the exposure metric there was a steep E-R relationship with a hazards ratio (HR) of 0.92 and 0.14 between quartiles 1 and 4, and the case-matched control arm, respectively. However, when Cavg1st-dose and Cmin1st-dose were used the apparent E-R relationship was flat. The flat E-R relationship is consistent with the lack of dose-response relationship derived from the randomized dose-ranging trial for nivolumab.13 The observed inconsistency in E-R relationship between the different exposure metrics used (Cavgss, Cavg1st-dose, and Cmin1st-dose) can be attributed to the dependence of Cavgss on concentrations from later post-treatment time points. Patients with improving disease status have reduced drug clearance, so at later time points there is an apparent correlation between exposure and response. Using exposure metrics derived from earlier time points reduces the change in disease status influencing PK, and allows for more accurate assessment of the impact of treatment exposure on clinical response.14
In addition to using exposure metrics derived from early time points consideration should be placed on whether observed or model-derived exposure metrics are used in the E-R analysis. In the E-R analysis for trastuzumab emtansine (T-DM1) the results of Cox proportional-hazards modeling (CPH) were not consistent between model-derived and observed exposure metrics. After adjusting for baseline risk factors in the analyses for both metrics the model-predicted Cmin1st-dose and area under the curve at cycle 1 (AUC1st-dose) were significantly associated with OS though the observed Cmin1st-dose and AUC1st-dose were not.15 Due to the limited understanding of this discrepancy a strong recommendation could not be made regarding the selection of observed or model-derived metrics, and it would be prudent for an E-R analysis to examine both. Upon examining both metrics if either of them is significantly associated with OS further investigation into the population PK model and E-R analysis is warranted.