Effect of estradiol as a continuous variable on breast cancer survival by menopausal status: a cohort study in China

High levels of circulating estradiol (E2) are associated with increased risk of breast cancer, whereas its relationship with breast cancer prognosis is still unclear. We evaluated the effect of E2 concentration on survival endpoints among 8766 breast cancer cases diagnosed between 2005 and 2017 from the Tianjin Breast Cancer Cases Cohort. Levels of serum E2 were measured in pre-menopausal and post-menopausal women. Multivariable-adjusted Cox proportional hazards models were used to estimate hazard ratios (HR) and 95% confidence intervals (95% CI) between quartile of E2 levels and overall survival (OS) and progression-free survival (PFS) of breast cancer. The penalized spline was then used to test for non-linear relationships between E2 (continuous variable) and survival endpoints. 612 deaths and 982 progressions occurred over follow-up through 2017. Compared to women in the quartile 3, the highest quartile of E2 was associated with reduced risk of both PFS in pre-menopausal women (HR 1.79, 95% CI 1.17–2.75, P = 0.008) and OS in post-menopausal women (HR 1.35, 95% CI 1.04–1.74, P = 0.023). OS and PFS in pre-menopausal women exhibited a nonlinear relation (“L-shaped” and “U-shaped”, respectively) with E2 levels. However, there was a linear relationship in post-menopausal women. Moreover, patients with estrogen receptor-negative (ER-negative) breast cancer showed a “U-shaped” relationship with OS and PFS in pre-menopausal women. Pre-menopausal breast cancer patients have a plateau stage of prognosis at the intermediate concentrations of E2, whereas post-menopausal patients have no apparent threshold, and ER status may have an impact on this relationship.


Introduction
Breast cancer is the sixth leading cause of death in China [1,2] and the fifth leading cause of death worldwide [3]. Most breast tumors are estrogen-dependent, and medications to inhibit or block estrogen are the mainstay of treatment for women with estrogen receptor (ER)-positive breast cancer [4]. As one of the most biologically active estrogens, estradiol (E2) behaves differently in different menopausal status [5]. In pre-menopausal women, estradiol is mainly produced by granulosa cells of follicles, which fluctuate during the Junxian Li and Chenyang Li have contributed equally to this work. female menstrual cycle [6]. The serum concentrations of E2 are subject to slight fluctuations after menopause [7]. There is an increasing interest in the role of E2 levels for breast cancer initiation [8,9], progression, and prognosis. Several lines of epidemiologic evidence have indicated a role of pre-menopausal E2 in breast cancer etiology [9,10]. Among post-menopausal women, a large-scale prospective cohort has firmly established high E2 levels as being significantly associated with increased breast cancer risk [11]. With regard to prognosis, few studies have explored the relationship between E2 concentrations and prognosis among pre-menopausal women newly diagnosed with breast cancer. In post-menopausal patients, studies to date have offered conflicting insights on the association between E2 levels and clinical outcomes of breast cancer [12][13][14]. There is likewise little evidence for a dose-response relationship between serum E2 and prognosis of breast cancer in premenopausal and post-menopausal patients.
Therefore, the present study aimed to investigate the association between preoperative E2 levels and breast cancer death, recurrence, and metastasis among Chinese pre-menopausal and post-menopausal patients.

Study population and design
The Tianjin Breast Cancer Cases Cohort (TBCCC) was launched in 2004 and last updated by the end of 2017 from Tianjin Medical University Cancer Institute and Hospital. TBCCC aimed to support studies on breast cancer survival, disease progression, quality of life in Chinese breast cancer patients, and strove to achieve precise medical care [15,16]. Patients were diagnosed as breast cancer for the first time through clinical-pathological examination after admission to Tianjin Medical University Cancer Institute and Hospital. We collected the demographic and epidemiological data of patients through a structured questionnaire, and abstracted clinicopathological and therapeutic information from medical records, and attained vital status once a year by telephone. Data of demographic characteristics (age at diagnosis, height, weight, marital status, etc.), reproductive factors (menopausal status, breastfeeding, ovariectomy, etc.), lifestyle habits (smoking status, drinking, negative events, etc.), and family history were collected by trained interviewers with face-to-face questionnaire interview. Moreover, hospital information system (HIS) and Tianjin Cancer Death Registry System were used to confirm self-reported information of vital status, such as recurrence, metastasis, and death. This study was approved by the research ethics board of the Tianjin Medical University Cancer Institute and Hospital.
All patients were diagnosed with stage I-III invasive breast cancer, never received menopausal hormone therapy, and not diagnosed with other cancer previously. All patients had written informed consent in TBCCC. The final analytic sample included 8766 patients who donated a blood sample and had complete data on preoperative estradiol (E2) levels.

Measures of E2 levels and covariates
Fasting, early follicular blood samples were obtained from each patient at baseline within 5 days after menstruation cessation. Circulating E2 measurements were completed by using Mindray CL-6000i chemiluminescence analyzer detection system (China). Blood samples were detected within 24 h since the completion of collection.
The questionnaire was used to obtain baseline information on age at diagnosis and body mass index (BMI). HIS electronic medical record was used to obtain information on histological grade, pT, pN, stage, ER, progesterone receptor (PR), human epidermal growth factor receptor-2 (HER2), and treatment regimen. BMI was calculated as kilograms per square meter of height (kg/m 2 ). Underweight was defined as BMI < 18.5 kg/m 2 , normal as BMI of 18.5-23.9, overweight as BMI of 24-27.9, and obesity as BMI ≥ 28 kg/m 2 [17]. The ER, PR, and HER2 expressions were detected by the immunohistochemistry (IHC) method. The tumor was considered ER-positive or PR-positive if there were ≥ 1% nuclear-stained malignant cells. For HER2, the stained tumor cell nuclear and nuclear staining intensity was semiquantitatively evaluated using a grade between 0 and 3 +. Results of 0, 1 +, or 2 +/FISH negative were indicated as HER2 negative, whereas 2 +/FISH positive or 3 + were positive [18].

Statistical analysis
Categorical variables are presented as frequencies (n) and percentages (%) and continuous variables as mean (SD) for normally distributed data or median (IQR) for non-normal data. Characteristics of the two different subgroups were compared using χ 2 analysis, the t test or Mann-Whitney test, as appropriate. Overall survival (OS) was measured from time at diagnosis to death, regardless of cause, and progression-free survival (PFS), to survival without evidence of relapse, progression, or death. Multivariableadjusted Cox proportional hazards models were used to evaluate associations between serum E2 levels and mortality or disease progression risk. Models were adjusted for age at diagnosis, BMI, pT, pN, stage, histological grade, ER, PR, HER2 status, chemotherapy, radiotherapy, and endocrine treatment. The selection of the final model stratified by menopausal status (pre-menopause and post-menopause) was performed using a backward step 1 3 down-selection process. The penalized spline (P-spline) in the Cox model allows a nonlinear relationship of E2 levels with the logarithm (ln hazard ratio [19]) of disease progression or mortality, estimated from the Cox regression model adjusted for significant covariates in univariable analysis (Supplementary Tables S1, S2). We used the dfmacox (degrees of freedom in multivariate additive Cox models) function in smoothHR to obtain the optimal number of degrees of freedom in the extended Cox-type additive multivariate analysis [20]. The lowest risk point was used as the reference value. Performance of E2 concentrations was plotted by ER status using hazard ratios (HR) and 95% confidence intervals (CI). Linearity and nonlinearity were inspected visually using Martingale residual plots and examined statistically using likelihood ratio tests [21]. The assumption of proportional hazards was examined by visual inspection of residuals plots on time and using the Schoenfeld residuals test. Statistical significance was set as P < 0.05 in a two-tailed test. All statistical analyses were performed using R version 4.1.0 (R Foundation) and SAS 9.4 (SAS Institute Inc).

Prognostic effects of E2 levels on breast cancer survival
The association of the serum E2 levels with breast cancer survival, including overall survival (OS) and PFS, is presented in Fig. 1  IQR inter-quartile range, ER estrogen receptor, PR progesterone receptor, HER-2 Human epidermal growth factor receptor-2

Dose-response relationship between E2 and breast cancer survival
The dose-response relationships between E2 and death or progression were analyzed in patients before and after menopause, adjusting for the same confounding factors (Fig. 2). When E2 level was a continuous variable attempting to account for non-linear association, the relationship between survival endpoints and E2 was significantly nonlinear in pre-menopausal women. E2 levels were prognostic of OS in a J-shaped pattern (nonlinear P = 0.039); risk was flat to a threshold (~ 250 ng/L), then increased gradually ( Fig. 2A). E2 levels were prognostic of PFS in a U-shaped pattern (nonlinear P = 0.013); risk was higher at the extreme (300 ng/L) but lower at intermediate values (> 50 to < 150 ng/L) (Fig. 2B). The plot of the Fig. 2A, B showed a J-shaped and U-shaped relationship between log relative hazard and E2, indicating that lower or higher levels of E2 were associated with a higher hazard of death or progression, with the lowest risk of 128.7 ng/L and 103.3 ng/L, respectively. When the E2 concentration was above 300 ng/L, the risk of PFS gradually became flat, whereas OS was still elevated. As for postmenopausal women, there was a linear relationship, and the linearity P values were 0.105 and 0.158, respectively. The plot showed that the risk of death or progression increased gradually with increasing E2 concentration, both tending to be stable after 30 ng/L (Fig. 2C, D).

ER status of patients modified the prognostic effects of E2 levels on breast cancer survival
When we analyzed the ER status, both ER-positive and ER-negative demonstrated inconsistent fluctuation curves (Fig. 3). Patient ER status modified the prognostic association of E2 levels and OS or PFS in the pre-menopausal women. Among ER-positive patients, E2 levels were associated with OS or PFS in a linear pattern (linear OS: P = 0.568; PFS: P = 0.412), whereas among ER-negative patients, E2 levels were associated with OS or PFS in a U-shaped pattern (nonlinear OS: P = 0.0006; PFS: P = 0.001) (Fig. 3A, Linearity P=0.158 B). Compared with ER-negative patients, the risk of disease progression or death was higher in ER-positive breast cancer patients when E2 concentration was 100 ng/L-300 ng/L. In post-menopausal women, there was a positive linear correlation between E2 and death or progression in ER-positive and ER-negative breast cancer (Fig. 3C, D). ER status did not modify the prognostic association between E2 levels with OS or PFS in the post-menopausal women. Within 10 ng/L, the progression or death risk for ER-positive was approximately the same as that for ER-negative. As for above 10 ng/L, the risk for ER-positive was higher than ER-negative.

Discussion
In this population-based cohort study of 8766 patients with breast cancer, E2 concentrations were prognostic of OS or PFS. In pre-menopausal women, the J (gradually increasing) and U (decrease first and then increase) shapes of these associations were generally nonlinear and modified by patients' ER status. Among ER-positive patients, lower E2 levels were associated with lower risk of death or progression, whereas among ER-negative patients, moderate levels of E2 were associated with lower risk of death or progression. Conversely, in post-menopausal women, the concentration of E2 was linearly proportional to death or progression and not modified by ER status.
In pre-menopausal women, there are few literatures about the relationship between E2 and breast cancer prognosis due to the inaccessibility of timed samples. As for the dose-response relationship between E2 and survival endpoints, Patients with an intermediate level of E2 had a better prognosis. Estrogens can bind to ERs and play a role in breast cancer progression [22]. It remains unclear how well circulating E2 levels cooperated with hormone receptor signaling in the breast. The evidence suggests that E2 levels in circulation are positively correlated with that in the breast tissue [23][24][25]. From previously published results, increased E2 may reduce the risk for breast cancer-specific mortality among participants with ER-negative tumors (HR 0.16, 95% CI 0.05-0.63), but not among ER-positive tumors [12]. In our stratified analyses, ER-negative was U-shaped, while ER-positive increased linearly before menopause. The possible reason is that estrogen metabolites will cause cumulative genotoxic insult in ER-positive tumor [26]. In the absence of ER status breast cancer, low or excessive E2 levels could affect the function of the uterus or ovaries, thereby affecting the secretion of E2 or other hormones [27,28], and indirectly affecting the prognosis of breast cancer. Micheli et al. have reported that high plasma testosterone strongly predicts poorer prognosis in breast cancer patients [29]. Low or high E2 concentration is just a marker of a general metabolic imbalance that provides a milieu in which cancer cell proliferation in the breast or another site is facilitated [30]. Further, compared with ER-negative, the risk of disease progression or death was higher in ER-positive breast cancer when E2 concentration was 100 ng/L-300 ng/L, while there was an opposite relationship when it is less than 100 ng/L or greater than 300 ng/L. The high expression of estrogenresponsive genes (PGR, GREB1, TFF1, and PDZK1) may be at a higher or lower level of E2, which affects the prognosis of breast cancer [31,32].
In post-menopausal women, our finding that higher E2 levels were associated with moderately worse OS and PFS mirrors prior studies [12,14,33]. Lønning et al. reported that patients with higher estrogen levels had shorter time to progression [34]. However, in another study of 358 women diagnosed with stage I-IIIA breast cancer, E2 level in the higher 50th percentile was not associated with a higher risk of mortality (HR 1.01, 95% CI 0.66-1.53) [13]. The reason for this result may be the modeling of E2 levels using categories [6]. As for the dose-response relationship between E2 and breast cancer prognosis, E2 levels had an increasing linear association with the prognosis of breast cancer and similar patterns in ER-positive and ER-negative patients. When the E2 concentration was above 20 ng/L, ER-positive patients had a slightly higher risk of death or progression than ER-negative patients. Estrogens can bind to ERs and play a role in breast cancer progression [22,35]. Preoperative E2 levels without hormone treatment degrade the prognosis of ER-positive patients compared with ER-negative patients. The possible reason for this result may be the data distribution. Most post-menopausal E2 concentrations are concentrated within 20 ng/L, but there is still a lack of data for higher concentrations of E2 (only about 10%), which may be associated with a better prognosis for ER-negative breast cancer when the concentration is high.
There are several strengths of this study. To the best of our knowledge, this is the first study to quantify the increased risk of all-cause mortality or disease progression for a specific E2 concentration among pre-menopausal breast cancer patients in China. In contrast to prior studies, our study models E2 levels as continuous variables using splines, which can contribute to keeping continuities in mortality risk between exposure concentration that enhances statistical power and is often easy to justify biologically and interpret clinically. The main strengths are the large sample size and hospital-based, timed sample, and high follow-up rates. The large sample size offers sufficient statistical power to evaluate the prognostic role of E2 levels by ER-specific subgroups. The use of clinically acquired preoperative E2 levels, coupled with menopausal status, makes the integration of clinico-pathological variables into dividing distinct risk profiles and offers the opportunity for personalized treatment.
Despite several strengths, there are several limitations. The number of patients with progression or death is small due to the generally good prognosis of breast cancer, even over an extended follow-up. Therefore, further multicenter prospective studies with longer-term follow-up will be needed to resolve these limitations. In addition, experimental studies to investigate the mechanism of the prognostic role of E2 in future studies are needed.

Conclusion
In summary, E2 levels were prognostic of all-cause mortality or progression in breast cancer, and the strength and shape of these prognostic associations varied in pre-menopausal and post-menopausal patients and were modified by ER status. Measurements of E2 levels can be used to accurately identify high-risk profiles, and have a prompting effect on patients' outcomes clinically.