Confounding mitigation for the exposure-response relationship of
bevacizumab in colorectal cancer patients
Abstract
Aims. The exposure-response relationship of bevacizumab may be
confounded by various factors, i.e. baseline characteristics,
time-dependent target engagement and recursive relationships between
exposure and response. This work aimed at investigating the
exposure-response relationships of bevacizumab in mCRC patients while
mitigating potential sources of bias. Methods. Bevacizumab
pharmacokinetics was described using target-mediated drug disposition
(TMDD) modeling. The relationships between target kinetics, and
progression-free (PFS) and overall (OS) survivals were assessed using
joint pharmacokinetic and parametric hazard function models. Both
potential biases due to prognostic-driven and response-driven of the
concentration-effect relationship were mitigated. These models were used
to evaluate the effect of increased antigen target levels and clearance,
as well as intensified dosing regimen, on survival. Results. Estimated
target-mediated pharmacokinetic parameters in 130 assessed patients
were: baseline target levels (R0=8.4 nM), steady-state dissociation
constant (KSS=10 nM) and antibody-target complexes elimination constant
(kint=0.52 day-1). Distribution of R0 was significantly associated with
an increased baseline CEA and circulating VEGF levels, and the presence
of extra-hepatic metastases. Unbound target levels (R) significantly
influenced both progression and death hazard functions. Increased R0 or
CL values led to decreased bevacizumab unbound concentrations, increased
R levels, and shortened PFS and OS, whereas increasing bevacizumab dose
led to decreased R and longer survival. Conclusion. This study is the
first to show the relationship between bevacizumab concentrations,
target involvement and clinical efficacy by mitigating potential sources
of bias. Most of target amount may be tumoral in mCRC. A more in-depth
description of this relationship should be made in future studies.