Prospective evaluation of Therapeutic Drug Monitoring of endoxifen: feasibility of observational and randomized trials.
Aurelia H.M. de Vries Schultink(1), Thomas P.C. Dorlo(1), Lisa Madlensky(2), John P. Pierce(2), Jos H. Beijnen(1,3), Alwin D.R. Huitema(1,4)
(1) Netherlands Cancer Institute, Amsterdam, the Netherlands (2) Moores Cancer Centre, San Francisco, USA, (3) Utrecht Institute for Pharmaceutical Sciences, Utrecht, the Netherlands, (4) University Medical Centre Utrecht, Utrecht, the Netherlands UIPS
Objectives Tamoxifen is an anti-estrogenic drug that has been used to treat estrogen receptor (ER)-positive breast cancer for decades. Five years of adjuvant treatment with tamoxifen lowers ER-positive breast cancer recurrence and mortality rates.1 Despite the proven established efficacy of tamoxifen, 25-30% of patients still experience recurrence within 10 years. Variability in response has been attributed to variability in pharmacokinetics (PK), more specifically to variability of endoxifen concentrations, the most important active metabolite of tamoxifen. Dose adjustments based on Therapeutic Drug Monitoring (TDM) of endoxifen as a strategy to improve survival in patients with ER-positive breast cancer are controversial, because the exposure-response relationship has not been well quantified. Additionally, the benefits of TDM have not been shown prospectively. Appropriately designed clinical trials are necessary to demonstrate the potential benefits of TDM. Therefore, clinical trial simulations using a parametric modelling approach based on large patient cohorts were performed. An observational design was simulated aiming to demonstrate an exposure-response relationship between endoxifen concentrations and breast cancer recurrence. Additionally, randomized controlled trials (RCTs) were simulated to support a design (including, power calculations, sampling size and follow-up time) evaluating the benefits of TDM over conventional dosing. These clinical trial simulations can help design appropriate trials and assess the validity of observational trials, in order to put previously conducted retrospective and prospective observational studies into the right perspective.2,3
Methods Data Recurrence free survival data and endoxifen concentrations were available from 1370 ER-positive breast cancer patients who participated in a previously conducted trial.2 These data were used to establish a parametric time-to-event model. In addition, TDM data of endoxifen from 658 breast cancer patients treated with tamoxifen in the adjuvant setting were available from the Netherlands Cancer Institute. These data were used to evaluate the effect of a TDM-based dose increment of tamoxifen from 20 mg/day to 40 mg/day on attainment of the steady state target concentration of 5.97 ng/mL. Proportions of patient below and above this target were used for the clinical trial simulations.
Time-to-event model Different time-to-event models (Gompertz, Weibull and exponential distribution hazard functions) were evaluated in NONMEM using the Laplacian estimation method. Tumor grade, tumor stage, menopausal status were evaluated as predictors of recurrence-free survival.
Simulations The patients to be included in the clinical trials were ER-positive breast cancer patients that initiate treatment with tamoxifen 20 mg/day. Tumor grade, stage and menopausal status proportions were sampled from a previously studied cohort.2 Firstly, simulations were conducted to evaluate the optimal design of an observational trial to evaluate the exposure-response relationship between endoxifen concentrations and breast cancer recurrence. Endoxifen concentrations were determined, though no dose adjustments were applied and only follow up was performed. Recurrence was compared between patients attaining and not attaining the endoxifen target concentration. Secondly, simulations were conducted to evaluate the optimal design of a RCT to determine the benefits of endoxifen TDM on breast cancer outcome. Two different RCT designs were evaluated. Design 1: patients initiating treatment with tamoxifen 20 mg/day were included and randomized (1:1) to either the control arm, where no TDM was applied and only follow up was performed, or to the intervention arm in which patients with endoxifen concentrations ≤5.97 received a dose increment to tamoxifen 40 mg/day. Design 2: endoxifen concentrations are determined and patients with low endoxifen concentrations are randomized (1:1) or (1:2) either to the control arm (no dose increment) or to the intervention arm, receiving a dose increment to 40 mg/day tamoxifen.
Power calculations Each design was simulated a 1000 times with varying numbers of patients. For each trial the hazard ratio between the intervention and the control arm was determined using a Cox proportional hazards model (mimicking the conventional practice of analysing clinical trial survival data). The Cox proportional hazards model accounted for different covariates. The power was determined by the percentage of trials with a significant difference in recurrence-free survival between the control and the intervention arm, with a p < 0.05. A sensitivity analysis was performed to investigate the effect of shorter follow up times. In addition, the uncertainty in the effect of the PK target on the hazard was evaluated by assuming a factor 2 increase or decrease of this effect on the hazard.
Results Parametric time-to-event model The baseline hazard was best described by a Weibull distribution model. A higher tumor grade or stage was associated with an increased risk of recurrence. In addition, postmenopausal patients had a decreased risk of recurrence compared to premenopausal patients. A random dropout was used to account for patients that were lost to follow up.
Simulations To demonstrate the exposure-response relationship with a power >0.8, an observational trial design including at least 1500 patients and an intended follow-up of 15 years is needed to find a hazard ratio of 0.71, assuming a 29% reduction in the hazard of recurrence for patients attaining the target >5.97 ng/ml endoxifen concentration compared to patients with lower endoxifen concentrations. In order to prospectively validate application of endoxifen TDM to improve breast cancer outcome (assuming the previously estimated HR of 0.71), using the second study design demonstrated to be most feasible. Design 2 needs 1600 patients per arm to demonstrate the same effect (power of 82.9%). For design 2, a three-fold more patients are needed to identify the 32.5% of patients with low endoxifen concentrations. In comparison to previously conducted trials, the retrospective analysis by Madlensky et al. had around 60% power to find an exposure-response relationship and the recently published prospective CYPTAM study had only around 30% power.2,3
Conclusions Currently, no prospective or retrospective trial with sufficient power and follow up has been performed to detect the proposed exposure-response relationship between endoxifen and breast cancer recurrence. Our clinical trial simulations and power calculations indicate that an observational or randomized trial in which only the patients with low endoxifen steady-state plasma concentrations are randomized could both be feasible, although the required sample size would require a multicenter trial and international collaboration. If such a trial would be initiated, follow-up of 15 years is necessary.
 Davies, C., Godwin, J. & Gray, R. Lancet 771–784 (2011)
 Madlensky, L. et al. Clin. Pharmacol. Ther. 89, 718–25 (2011).
 Beatriz Sanchez-Spitman, A. et al. J. Clin. Oncol. 36, 523–523 (2018).