Factors affecting inter-individual variability in endoxifen concentrations in patients with breast cancer: results from the prospective TOTAM trial

Abstract

Introduction: Endoxifen-the principal metabolite of tamoxifen-is subject to a high inter-individual variability in serum concentration. Numerous attempts have been made to explain this, but thus far only with limited success. By applying predictive modeling, we aimed to identify factors that determine the inter-individual variability. Our purpose was to develop a prediction model for endoxifen concentrations, as a strategy to individualize tamoxifen treatment by model-informed dosing in order to prevent subtherapeutic exposure (endoxifen < 16 nmol/L) and thus potential failure of therapy.

Methods: Tamoxifen pharmacokinetics with demographic and pharmacogenetic data of 303 participants of the prospective TOTAM study were used. The inter-individual variability in endoxifen was analyzed according to multiple regression techniques in combination with multiple imputations to adjust for missing data and bootstrapping to adjust for the over-optimism of parameter estimates used for internal model validation.

Results: Key predictors of endoxifen concentration were CYP2D6 genotype, age and weight, explaining altogether an average-based optimism corrected 57% (95% CI 0.49-0.64) of the inter-individual variability. CYP2D6 genotype explained 54% of the variability. The remaining 3% could be explained by age and weight. Predictors of risk for subtherapeutic endoxifen (< 16 nmol/L) were CYP2D6 genotype and age. The model showed an optimism-corrected discrimination of 90% (95% CI 0.86-0.95) and sensitivity and specificity of 66% and 98%, respectively. Consecutively, there is a high probability of misclassifying patients with subtherapeutic endoxifen concentrations based on the prediction rule.

Conclusion: The inter-individual variability of endoxifen concentration could largely be explained by CYP2D6 genotype and for a small proportion by age and weight. The model showed a sensitivity and specificity of 66 and 98%, respectively, indicating a high probability of (misclassification) error for the patients with subtherapeutic endoxifen concentrations (< 16 nmol/L). The remaining unexplained inter-individual variability is still high and therefore model-informed tamoxifen dosing should be accompanied by therapeutic drug monitoring.

Keywords: Early breast cancer; Endoxifen; Predictive modeling; Tamoxifen; Therapeutic drug monitoring.