1. INTRODUCTION
While contemporary drug development in oncology strives to deliver novel therapies to patients rapidly, it is also important to optimize dosing regimens to improve patient-centered care. Doses selected for pivotal trials may be efficacious doses, but not necessarily optimal to minimizing toxicity and maximizing clinical efficacy for all patients. Exposure-response (E-R) analysis is an approach that is used to support dose selection by characterizing the relationship between drug concentrations, efficacy, and safety. A variety of E-R analyses have supported dose labeling of many approved oncology drugs.1 Among the oncology therapies, however, additional complexity has been observed in characterizing E-R relationships for monoclonal antibodies. Specifically, prognostic factors can impact both pharmacokinetics (PK) and efficacy. This may result in a correlation between exposure and outcome that does not represent a causal E-R relationship and therefore, may not provide a useful basis for dose recommendations. This was exemplified by the HELOISE trial (NCT01450696) of trastuzumab, which was conducted as part of a post-marketing requirement. Following the phase 3 trial ToGA (NCT01041404), trastuzumab was approved in combination with chemotherapy for first-line treatment of HER2-positive advanced gastric cancer. An E-R analysis, however, found that the patients in the lowest exposure quartile had an overall survival (OS) approximately 8 months shorter than those with higher exposures.2 This suggested that trastuzumab exposure in this low-exposure subgroup may survival benefit, and supported the requirement of conducting a post-marketing trial for a higher dose. For this requirement, in the HELOISE trial, a higher trastuzumab dose was compared with the labeled dose in a population with similar prognostic factors as the low-exposure subgroup of the ToGA trial. Despite reliably increasing exposure, the higher dose did not improve OS in patients.3 This discrepancy between the results of the E-R analysis and the HELOISE trial indicates confounding in E-R analyses of monoclonal antibodies at a single dose level in oncology.
In addition to the potential confounding factors for E-R analyses in oncology, there have been reports of time-dependent changes in the PK that require additional considerations. Monoclonal antibodies that target B-cell receptors, such as rituximab, have been reported to exhibit time-dependent decrease in clearance (CL) owing to target mediated drug disposition (TMDD).4-6 Time-dependent PK has also been observed for checkpoint inhibitors nivolumab, pembrolizumab, durvalumab, and avelumab. Across these molecules the range of CL decrease over time was 17% to 32%.7-10Best overall response was included as a covariate on CL in the time-dependent population PK models of nivolumab and pembrolizumab.7, 9 The time-dependent population PK model of durvalumab included time-varying albumin and tumor size as covariates on CL.8 While the model for avelumab did not incorporate response or time-varying biomarkers as covariates on CL, visual inspection of change in CL over time demonstrated a larger reduction in CL in responders than in nonresponders.10Overall, the decrease in CL over time in these molecules corresponded to changes in patients’ prognoses based on their responses to treatments over time. This observation may be attributed to changes in catabolic degradation of the monoclonal antibodies as a result of changing disease status.11 The changing drug CL and patient prognostic factors over time could potentially confound E-R analyses.
In this review we will discuss key considerations in interpreting E-R relationships and mitigation strategies to address the confounding effects in E-R analyses in oncology.
2. EXPOSURE-RESPONSE (E-R) ANALYSIS CONSIDERATIONS IN ONCOLOGY
In order to address confounding factors in E-R analyses for monoclonal antibodies in oncology several key determinants need to be considered. In this review we have described the importance of selecting the appropriate drug exposure metric for E-R analyses and have summarized three main approaches in oncology to address confounding in E-R analyses: CPH and case-matching analysis, tumor growth inhibition overall survival (TGI-OS) modeling, and clinical studies with multiple dose levels (Figure 1, Table 1 ). In addition to describing these approaches we will discuss their strengths and limitations. The current review will focus on E-R analyses for exposure-survival relationships.