Simulation-Based Assessment of the Impact of Non-Adherence on Endoxifen Target Attainment in Different Tamoxifen Dosing Strategies

Abstract

Tamoxifen is widely used in breast cancer treatment and minimum steady-state concentrations of its active metabolite endoxifen (CSS,min ENDX) above 5.97 ng/mL have been associated with favourable disease outcome. Yet, about 20% of patients do not reach target CSS,min ENDX applying conventional tamoxifen dosing. Moreover, 4–75% of patients are non-adherent, resulting in worse disease outcomes. Assuming complete adherence, we previously showed model-informed precision dosing (MIPD) to be superior to conventional and CYP2D6-guided dosing in minimising the proportion of patients with subtarget CSS,min ENDX. Given the high non-adherence rate in long-term tamoxifen therapy, this study investigated the impact of non-adherence on CSS,min ENDX target attainment in different dosing strategies. We show that MIPD allows to account for the expected level of non-adherence (here: up to 2 missed doses/week): increasing the MIPD target threshold from 5.97 ng/mL to 9 ng/mL (the lowest reported CSS,min ENDX in CYP2D6 normal metabolisers) as a safeguard resulted in the lowest interindividual variability and proportion of patients with subtarget CSS,min ENDX even in non-adherent patients. This is a significant improvement to conventional and CYP2D6-guided dosing. Adding a fixed increment to the originally selected dose is not recommended, since it inflates interindividual variability.

Generating a Precision Endoxifen Prediction Algorithm to Advance Personalized Tamoxifen Treatment in Patients with Breast Cancer

“…The results in Table 3 originate from studies that include patients at steady state and due to tamoxifen's long half-life there is little variation expected to occur throughout the day and in terms of timing of dosing. However, a major external source of endoxifen variability is adherence to the drug [93][94][95] which reduce the association between CYP2D6 and endoxifen concentrations and is thus an obvious factor that should be accounted for when associating CYP2D6 to clinical outcomes. The 30 studies summarized in Table 3 have applied distinct approaches for excluding non-adherent patients including the more classic approaches of self-reported compliance and pill-counts to using various pharmacokinetic cut offs based on tamoxifen concentrations.…”

“…Several approaches to building an integrated algorithm have been reported including using standard linear regression [67]. In recent years pharmacokinetic modeling approaches have been used to investigate the factors contributing to metabolite concentrations [94,97,[126][127][128]. These studies generally identified the same important factors but provide more mechanistic understanding of their effects on endoxifen concentration variability [97,127,128] and could be more precisely adapted into personalized dosing algorithms to achieve target endoxifen concentrations [94].…”

“…In recent years pharmacokinetic modeling approaches have been used to investigate the factors contributing to metabolite concentrations [94,97,[126][127][128]. These studies generally identified the same important factors but provide more mechanistic understanding of their effects on endoxifen concentration variability [97,127,128] and could be more precisely adapted into personalized dosing algorithms to achieve target endoxifen concentrations [94]. Finally, novel machine learning techniques have been utilized to improve model precision, though translation of these models into clinical practice may be challenging [129].…”

Computational Treatment Simulations to Assess the Need for Personalized Tamoxifen Dosing in Breast Cancer Patients of Different Biogeographical Groups

“…The suggested trial could also integrate routine CYP2D6 phenotyping using, e.g., a microdosed probe substance, such as yohimbine, and assess the correlation of CYP2D6 phenotype and C SS,min ENDX [ 55 ]. The CYP2D6 phenotype could then be used to guide initial dose selection followed by therapeutic drug monitoring-guided dose refinement as part of a previously proposed model-informed precision dosing framework [ 11 , 31 ].…”

“…This model describes not only the typical concentration–time profile but also several layers of variability around it. Once a reliable model has been developed, model-based simulations are a powerful tool to explore yet unstudied scenarios, such as differences in bioavailability, adherence, or sampling times [ 30 , 31 ]. The aim of this study was to apply the model to virtual breast cancer populations capturing the respective CYP2D6 allele frequencies, simulate tamoxifen standard dosing for each biogeographical group, and determine the frequency of patients with subtarget C SS,min ENDX .…”