Endoxifen Concentration Is Associated with Recurrence-Free Survival in Hormone-Sensitive Breast Cancer Patients
Article information
Abstract
Purpose
The metabolism of tamoxifen is influenced by various cytochrome p450 enzymes, including CYP2D6 and CYP2C19, leading to variations in the levels of endoxifen, even with the same tamoxifen dose. However, the clinical significance of endoxifen for the prognosis of breast cancer patients remains controversial. This study aimed to elucidate the relevance of endoxifen level to recurrence-free survival censored with tamoxifen discontinuation (RFSt), representing the RFS for tamoxifen itself, of breast cancer patients and determine a suitable cutoff for prognostication.
Materials and Methods
The study included 478 breast cancer patients. Tamoxifen and its metabolites, including endoxifen, were measured using liquid chromatography-tandem mass spectrometry. An optimal cutoff was determined with maximally selected rank statistics. Survival analysis and Cox regression were conducted based on this cutoff.
Results
An endoxifen level of 21.00 ng/mL was the optimal cutoff for prognostication. Survival analysis revealed a statistically significant difference in RFSt between the low endoxifen group (≤ 21.00 ng/mL) and the high endoxifen group (> 21.00 ng/mL) (log-rank test, p=0.032). The 10-year probability of RFSt was 83.2% (95% confidence interval [CI], 77.0 to 89.9) and 88.3% (95% CI, 83.3 to 93.5) in the low and high endoxifen groups, respectively. Multivariable Cox proportional hazards regression indicated endoxifen concentration as a significant factor associated with prognosis.
Conclusion
Endoxifen could serve as a marker for appropriate tamoxifen treatment with a cutoff of 21.00 ng/mL. Based on this cutoff, therapeutic drug monitoring would benefit patients displaying suboptimal endoxifen concentrations.
Introduction
Endoxifen is the most important active metabolite of tamoxifen, a drug used for treatment of hormone-sensitive breast cancer patients. While estrogen stimulates proliferation of estrogen receptor (ER)–positive breast cancer cells, tamoxifen and its metabolites render an antagonistic effect to suppress cancer cell proliferation by inhibiting expression of various genes [1]. However, even with the suppressive effect of tamoxifen on hormone-sensitive breast cancer, one-third of breast cancer patients treated with tamoxifen will experience relapse [2].
Cytochrome P450 (CYP) 2D6 is the major enzyme involved in the metabolism of tamoxifen. There have been multiple studies with conflicting results regarding the prognostic value of CYP2D6 genotype [3,4]. Although early studies demonstrated a positive association between CYP2D6 genotype and clinical outcome [5,6], subsequent studies reached a consensus that CYP2D6 genotype poorly correlates with the prognosis of breast cancer patients [7-9]. Hence, attempts were made to predict the clinical outcome based on endoxifen concentration rather than CYP2D6 genotype. Therapeutic drug monitoring (TDM) of endoxifen concentration could be beneficial for breast cancer patients if endoxifen concentration correlates with prognosis. However, there is no consensus on the cutoff for endoxifen to distinguish prognosis, with multiple studies proposing different cutoff values [10-14]. While a previous study also attempted to elucidate the prognostic significance of endoxifen using various previously reported cutoffs ranging from 3.3 to 10.3 ng/mL, none of these cutoffs were significant in the survival analysis [14].
One of the reasons for these conflicting results could be largely varying endoxifen concentrations across different populations of various studies. In a previous study incorporating a subset of the present cohort [15], the mean endoxifen concentration of patients treated with 20 mg/day tamoxifen was 25.00 ng/mL (66.93 nM). However, regarding the mean endoxifen concentration of patients treated with 20 mg/day, studies conducted in the Netherlands (the TOTAM study) [16] and Australia (the TADE study) [17], which mostly involved Caucasians, reported concentrations of 12.06 ng/mL (32.30 nM) and 9.63 ng/mL (25.80 nM), respectively. Meanwhile, a previous study, comprising Vietnamese and Filipino patients treated with 20 mg/day tamoxifen, reported a mean endoxifen concentration of 35.40 ng/mL (94.78 nM) [18]. The main reason for higher endoxifen concentrations and cutoff in our study population could be the different distributions of metabolic activity in the Korean population compared to other ethnicities. It has been reported that the frequency of poor metabolizers is significantly lower in the Korean population compared to Caucasians and Africans [19]. In addition, it has been suggested that using a relatively non-selective liquid chromatography-tandem mass spectrometry (LC-MS/MS) method could result in an endoxifen level twice as higher than that measured with a highly selective LC-MS/MS method [20].
Regardless of what contributes to the distribution of endoxifen concentration, it is likely that an appropriate endoxifen cutoff should be determined for each specific population. Considering the high endoxifen concentration in our previous study [15], we hypothesized that a higher cutoff might lead to superior predictability regarding the prognosis of hormone-sensitive breast cancer patients receiving tamoxifen treatment. In this study, we explored an endoxifen cutoff for prognostication and evaluated this presumed cutoff along with other cutoffs suggested in previous studies.
Materials and Methods
1. Study population and procedure
This study was a single-center observational study conducted at Samsung Medical Center, Seoul, Korea. The patients were enrolled from December 2009 to January 2016. The study population consisted of female patients diagnosed with breast cancer and receiving adjuvant tamoxifen treatment, which arrived at a total of 573 patients fully encompassing the 281 patients from our previous study [15]. Of note, the initial adjuvant endocrine therapy for all patients started with tamoxifen, not an aromatase inhibitor. Refer to our previous publication for detailed inclusion and exclusion criteria for the study population [15]. In this study, patients diagnosed with carcinoma in situ (with or without microinvasion) or those with poor compliance were additionally excluded. Poor compliance was determined based on patient self-report of low adherence or an endoxifen level lower than the limit of detection. In case poor compliance was identified, the information on the date of discontinuation or irregular administration of tamoxifen was also collected. In case of follow-up loss and tamoxifen treatment interruption due to various reasons, the corresponding cases were right censored. Additionally, those who received an escalated tamoxifen dose were right censored at the time of dose escalation. Clinicopathological information of age at diagnosis, pathologic findings, bilaterality, ER and progesterone receptor (PR) status, ERBB2 status, pathologic stage, tamoxifen dose, and other treatment modalities was collected. Menopausal status was not collected since we could not confidently determine the corresponding information from the medical records. The pathologic stage of each patient was classified according to the 7th edition of the American Joint Committee on Cancer (AJCC) staging manual. In patients with synchronous bilateral breast cancer, the higher pathologic stage of either side was used. ER and PR were considered positive if at least 1% of tumor cells were positive for nuclear staining. The ERBB2 status was considered positive for grade 3+ in immunohistochemical (IHC) staining or grade 2+ in IHC staining with positive in situ hybridization, but was otherwise considered negative. The in vivo endoxifen concentration was considered to be at a steady-state level if venous sampling was conducted at least eight weeks after the initiation of tamoxifen treatment. Extended hormone therapy was defined as having been treated with tamoxifen or aromatase inhibitor after finishing 5 years of tamoxifen administration. Concentrations of tamoxifen and its metabolites, which include (Z)-endoxifen, N-desmethyltamoxifen, and (Z)-4-hydroxytamoxifen, were measured from plasma with LC-MS/MS using an API 4000 tandem mass spectrometer (AB Sciex, Foster City, CA) in tandem with an Agilent Technologies Series 1200 HPLC system (Agilent Technologies, Santa Clara, CA). Details regarding the measurement procedure are described in our previous publication [15].
2. Statistical analysis
The primary endpoint in this study was recurrence-free survival (RFS) since the purpose of using tamoxifen is to suppress cancer cell proliferation and prevent relapse. To determine an optimal cutoff for endoxifen concentration in predicting RFS, maximally selected rank statistics was utilized. Using the presumed optimal cutoff, patients were divided into low endoxifen and high endoxifen groups. In reference to a previous study, RFS censored following tamoxifen discontinuation (RFSt) and complete RFS (RFSc) were evaluated [9]. RFSt represents the RFS for tamoxifen per se, and all patients with tamoxifen treatment interruption were right censored at the corresponding time point for RFSt regardless of the reason [9]. In addition, if tamoxifen dose was increased, these patients were also right censored at the time point of dose escalation. On the other hand, if the patients changed the tamoxifen dose or switched to an aromatase inhibitor, these patients were not censored for RFSc and the full period of hormone therapy was taken into account [9]. Notably, the period of observation after tamoxifen discontinuation could potentially confound the results. Since this study focused on the association between endoxifen concentration and prognosis, RFSt was mainly evaluated [9]. Regarding the baseline characteristics of the two groups, comparisons were conducted with t test and chi-square test as appropriate. Survival analysis was conducted with the Kaplan-Meier estimator, and the statistical significance of the difference between the two groups was determined with a log-rank test. Along with the presumed cutoff, endoxifen cutoffs suggested in previous studies [10-14] were evaluated with our cohort. The evaluated cutoffs from the literature were as follows: (1) 3.36 ng/mL (9.00 nM) [10], (2) 5.29 ng/mL (14.15 nM) [11], (3) 5.60 ng/mL (15.00 nM) [12], (4) 5.97 ng/mL (15.98 nM) [13], and (5) 10.30 ng/mL (27.58 nM) [14].
Cox proportional hazards regression was used to investigate other clinicopathological factors associated with increased risk of breast cancer relapse. Backward elimination was used to select the variables and build the model. The proportional hazards assumption was verified using scaled Schoenfeld residuals.
All statistical analyses were carried out with R ver. 4.3.0 (R Foundation for Statistical Computing, Vienna, Austria) with a two-sided significance level of p=0.05; plots were illustrated with ggplot2 3.4.2, maximally selected rank statistics was implemented with maxstat 0.7-25, and survival analysis and Cox proportional hazards regression analysis were performed with survival 3.5-5 and survminer 0.4.9 in R 4.3.0 (R Foundation for Statistical Computing).
Results
1. Demographics
From December 2009 to January 2016, 573 breast cancer patients were enrolled at Samsung Medical Center, Korea. Among the 573 patients, 95 subjects were excluded, and 478 patients were analyzed in this study (Fig. 1). The 478 eligible subjects had a median follow-up period of 98.5 months (interquartile range [IQR], 57.0 to 120.8 months) with respect to RFSt. The median duration from tamoxifen treatment initiation to sample collection was 5.0 months (IQR, 4.5 to 6.0 months). The subjects were divided into two groups based on an endoxifen cutoff of 21.00 ng/mL (56.22 nM), the determining method of which will be described in the following section. The clinical characteristics of the subjects are listed in Table 1, and there were no statistically significant differences between any characteristics of the two groups except for luteinizing hormone-releasing hormone (LHRH) agonist treatment (χ2 test, p=0.029). There were no patients who underwent bilateral oophorectomy during the observed period.
2. Prognostic significance of endoxifen concentration
A presumed endoxifen cutoff of 21.00 ng/mL (56.22 nM) was defined with maximally selected rank statistics, which exhibited the highest standardized log-rank statistic value. The histogram based on this cutoff (21.00 ng/mL) is presented in Fig. 2A. However, as the threshold determined with the highest log-rank statistics does not necessarily guarantee a statistically significant difference in the primary endpoint, RFSt in our study, further investigation of this presumed cutoff was carried out with survival analysis.

(A) Distribution of endoxifen concentration in our study population. Kaplan-Meier curve of RFSt (B) and RFSc (C) probability of breast cancer patients according to an endoxifen cutoff of 21.00 ng/mL. RFSc, complete recurrence-free survival; RFSt, recurrence-free survival censored following tamoxifen discontinuation.
Of the 478 subjects, 56 (11.7%) experienced relapse before October 2022. In the low endoxifen group (≤ 21.00 ng/mL), 29 of 149 patients (19.5%) experienced relapse, compared to 27 of 329 patients (8.2%) in the high endoxifen group (> 21.00 ng/mL). An endoxifen cutoff of 21.00 ng/mL was able to distinguish the prognosis represented by RFSt with statistical significance (log-rank test, p=0.032) (Fig. 2B). The 10-year probability of RFSt was 83.2% (95% confidence interval [CI], 77.0 to 89.9) and 88.3% (95% CI, 83.3 to 93.5) in the low and high endoxifen groups, respectively. On the other hand, when RFSc was evaluated with the same endoxifen cutoff, the difference was not statistically significant (log-rank test, p=0.127) (Fig. 2C). The 10-year probability of RFSc was 86.2% (95% CI, 81.1 to 91.6) and 89.2% (95% CI, 84.8 to 93.8) in the low and high endoxifen groups, respectively.
When the cutoffs for endoxifen suggested in previous literature were applied to our study population, relatively few patients were assigned to the low endoxifen group, which limited the evaluation of these cutoffs. As applying the cutoff of 5.97 ng/mL already resulted in fewer than five subjects being assigned to the low endoxifen group, cutoffs lower than 5.97 ng/mL were not further evaluated. For the survival analysis curve illustrated using cutoffs of 5.97 ng/mL and 10.30 ng/mL, refer to Fig. 3. There was a statistically significant difference in RFSt according to the cutoff of 5.97 ng/mL (log-rank test, p=0.047) (Fig. 3A), but not in RFSc (log-rank test, p=0.280) (Fig. 3B). However, the cutoff of 10.30 ng/mL failed to exhibit statistically significant differences in both RFSt (log-rank test, p=0.632) (Fig. 3C) and RFSc (log-rank test, p=0.516) (Fig. 3D).

Kaplan-Meier curve of RFSt (A, C) and RFSc (B, D) probability of breast cancer patients according to an endoxifen cutoff of 5.97 ng/mL (15.98 nM) (A, B) and 10.30 ng/mL (27.58 nM) (C, D). RFSc, complete recurrence-free survival; RFSt, recurrence-free survival censored following tamoxifen discontinuation.
3. Cox proportional hazards regression
Using backward elimination, the variables selected in the final model were endoxifen, ER status, PR status, pathologic stage, adjuvant trastuzumab treatment, LHRH agonist treatment, and extended hormone therapy. Results of the univariable and multivariable Cox proportional hazards regression model are listed in Table 2. In the univariable analysis, a significantly different hazard ratio (HR) of RFSt was associated with low endoxifen level (≤ 21.00 ng/mL; HR, 1.78; 95% CI, 1.04 to 3.04; p=0.034). In addition, ER status, PR status, and adjuvant trastuzumab treatment were significantly associated with RFSt in the univariable analysis. In the multivariable analysis, endoxifen concentration was still a significant factor associated with RFSt (HR, 1.78; 95% CI, 1.03 to 3.08; p=0.038). Besides, ER status (HR, 4.53; 95% CI, 1.07 to 19.26; p=0.041), PR status (HR, 4.92; 95% CI, 1.87 to 12.93; p=0.001), adjuvant trastuzumab treatment (HR, 3.00; 95% CI, 1.26 to 7.15; p=0.013), LHRH agonist treatment (HR, 2.45; 95% CI, 1.23 to 4.86; p=0.011), and extended hormone therapy (HR, 0.45; 95% CI, 0.23 to 0.85; p=0.015) were significantly associated with RFSt in the multivariable analysis.
Discussion
CYP2D6 is the key enzyme of tamoxifen metabolism, and a new CYP2D6 locus was recently discovered as the principal determinant of endoxifen concentrations [21]. Still, enzyme activity determined with CYP2D6 genotyping has limited value in estimating the efficacy of tamoxifen since tamoxifen metabolism is affected by various factors such as age, weight, concurrently used drugs, and activity of other enzymes including CYP2C9, CYP2C19, CYP3A4, CYP3A5, and CYP2B6 [21-23], complicating the prediction of endoxifen concentration. Currently, the current National Comprehensive Cancer Network (NCCN) guideline does not recommend CYP2D6 genotyping in breast cancer patients [24]. While the Clinical Pharmacogenetics Implementation Consortium (CPIC) Guideline suggests dose escalation to 40 mg/day in CYP2D6 intermediate metabolizers and poor metabolizers [25], it was insufficient to achieve the endoxifen concentration observed in normal metabolizers [26]. Considering this situation, the current CYP2D6-based dose adjustment has limited value, and endoxifen-based dose adjustment seems more plausible. However, regarding the clinical significance of endoxifen concentration, there is no consensus on the prognostic value of endoxifen concentration [26], with various studies suggesting different endoxifen cutoff values [10-14]. Furthermore, varying endoxifen distribution in different populations [15-19] as well as discrepant endoxifen levels depending on the method of measurement [20] warrant a specific endoxifen cutoff for different contexts.
Even though tamoxifen is widely used for breast cancer patients, research on its efficacy is challenging owing to reasons such as (1) plans for pregnancy, (2) side effects such as depression and liver enzyme elevation, (3) poor compliance and failure to complete the 5-year course, (4) switch to an aromatase inhibitor, (5) extended tamoxifen treatment for 10 years, and (6) extended aromatase inhibitor treatment after finishing 5-year tamoxifen treatment. While concurrent drugs such as selective serotonin reuptake inhibitors are known to affect the metabolism of tamoxifen, the results of our study would not be affected by such drugs as we measured the concentration of endoxifen itself, the most potent active metabolite of the original drug.
Recently, the feasibility of increasing endoxifen concentration through TDM has been proposed [16,27]. To our knowledge, this study is the first to report the clinical relevance of endoxifen concentration and prognosis of breast cancer in the Korean population. Our study recommends an endoxifen concentration greater than 21.00 ng/mL considering its association with RFSt, which is a novel cutoff higher than those described in previous reports. Accomplishing this goal would be plausible using TDM as higher tamoxifen dose results in elevated endoxifen concentration [17]. Both inter-individual differences in endoxifen concentration and varying characteristics depending on the duration of hormone therapy substantiate the need for a tailored approach to tamoxifen dose adjustment.
While LHRH agonist treatment status differed significantly between the low and high endoxifen groups, endoxifen concentration remained a significant predictor in multivariable Cox regression adjusted for other clinically important variables. Regarding the various treatment modalities, low endoxifen concentration, adjuvant trastuzumab treatment, and LHRH agonist treatment were associated with higher HR, whereas hormone therapy extension was associated with lower HR in the multivariable Cox regression model for RFSt. We suspect that the higher risk in trastuzumab-treated patients would be associated with the relatively aggressive nature of ERBB2-positive breast cancer compared to ERBB2-negative breast cancer [28]. Besides, we acknowledge that the higher risk associated with LHRH agonist treatment is rather counterintuitive. Still, we suspect that varying age distribution and adjuvant chemotherapy treatment may have contributed to the poor prognosis observed in patients receiving LHRH agonist treatment. In our study population, patients treated with LHRH agonist had lower age (median, 41.2; IQR, 39.2 to 44.2) compared to those who did not receive LHRH agonist treatment (median, 44.2; IQR, 40.4 to 47.0; Wilcoxon test, p < 0.001). Thus, the poor prognosis in patients on LHRH agonist treatment could be due to the relatively poor prognosis in young breast cancer [29,30]. Moreover, although pathologic stage did not differ significantly between patients treated with and without LHRH agonist, a higher proportion of patients treated with LHRH agonist were on adjuvant chemotherapy (65.3%) compared to those who were not (38.8%; χ2 test, p < 0.001). This implies that varying clinical characteristics between the two groups, which would have led to different treatment choices, as well as the administration of adjuvant chemotherapy itself may also have contributed to the poor prognosis in patients treated with LHRH agonist. We would not address the prognostic effect of variables other than endoxifen concentration in further detail as it falls beyond the scope of our study.
In this study, we established a novel cutoff for endoxifen (21.00 ng/mL). As we had suspected, previously suggested cutoffs were unable to distinguish prognosis in our study population. Nevertheless, adopting the novel cutoff of our study in a different setting should be done with caution. We necessitated the need for a higher endoxifen cutoff because endoxifen concentration could vary across different populations [15-19] and measurement methods [20]. This logic also applies to other potential groups that plan to utilize our study in predicting the prognosis of breast cancer patients. Here, we would like to emphasize the significance of establishing a population-specific and institute-specific cutoff for endoxifen. Further validation would be required to determine the generalizability of our new endoxifen cutoff, as our study is the first to validate the prognostic significance of such a high endoxifen cutoff.
There are some limitations in this study to declare. First, the number of samples drawn from each patient was limited. Due to this limitation, there might be some noise in our data. For instance, if a patient adds a new medication affecting tamoxifen metabolism after measuring endoxifen concentration, this change would not be reflected in the data. Although we had no means to identify prescriptions from other hospitals, a meticulous review of our medical records was conducted to exclude medications from our institute affecting the CYP2D6 metabolism. Second, while menopausal status is an important factor that affects the prognosis of breast cancer patients, our study did not account for menopausal status since we couldn’t confidently obtain the relevant information from the medical records. Third, only a small number of patients received tamoxifen monotherapy. While other treatment modalities could potentially confound the effect of tamoxifen treatment, we addressed this through multivariable Cox proportional hazards regression, which showed that endoxifen concentration is still a significant factor associated with RFSt even with other clinicopathological factors adjusted.
In summary, an endoxifen cutoff (21.00 ng/mL) higher than previously suggested thresholds could be utilized in prognostication of breast cancer patients. Individualized tamoxifen dose adjustment would be beneficial for breast cancer patients to improve their prognosis by elevating the endoxifen concentration.
Notes
Ethical Statement
This study was approved by the Institutional Review Board of Samsung Medical Center (IRB No. 2011-04-030-001). Informed consent was obtained from each participant.
Author Contributions
Conceived and designed the analysis: Lee B, Lee JE, Lee SY.
Collected the data: Lee JE, Lee SY.
Contributed data or analysis tools: Nam SJ, Kim SW, Yu J, Chae BJ, Lee SK, Ryu JM, Lee JE, Lee SY.
Performed the analysis: Lee B, Lee SY.
Wrote the paper: Lee B, Lee SY.
Conflicts of Interest
Conflict of interest relevant to this article was not reported.
Acknowledgements
This research was supported by a grant of the Korea Health Technology R&D Project administered through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (HI14C3418). The funder had no role in the study design, analysis, and manuscript preparation and publication.