The Prognostic Values of Androgen Receptor in Breast Cancer: A Hospital-Based Retrospective and Propensity Score–Matched Cohort

This study was approved by the Ethics Committee of the Fudan University Shanghai Cancer Center (Shanghai, China; ID: 050432-4-1911D, 1905202-7) and conducted in accordance with the Declaration of Helsinki. This study was a retrospective cohort study, and thus informed consent from patients was not needed.

Records of breast cancer patients from 2013 to 2016 in Fudan University Shanghai Cancer Center were retrieved. The patients fulfilled the following inclusion criteria: (1) primary breast cancer without metastatic disease; (2) AR assessment by immunohistochemistry (IHC) methods in full sections; (3) did not receive neoadjuvant treatment, including chemotherapy, radiotherapy, and endocrine therapy; and (4) at least 1-year follow-up available. IHC for ER, PR, HER2, and AR (1:200 dilution; ab133273, Abcam) was automatically performed by the BenchMax XT IHC/ISH Staining Module (Roche) in the Department of Pathology at Fudan University Shanghai Cancer Center according to the standard protocol. A breast pathologist assessed the IHC results, and a senior breast pathologist reviewed the results.

Pathologic stages were classified according to the American Joint Committee on Cancer staging manual, 8th edition.¹⁰  HER2 positivity was defined as 3+ by IHC or fluorescence in situ hybridization positivity. ER, PR, and AR were considered positive when the section showed 1% staining or more; Ki-67 was defined as high if the section showed 15% staining or more.

The outcomes were overall survival (OS) and disease-free survival (DFS). OS was calculated from the day of patient diagnosis to the last day of follow-up or death, and DFS was defined as the day from diagnosis to the day the event (death, recurrence, or metastasis) occurred.

The data were analyzed using SPSS 21.0 software. Categorical variables are expressed using frequency, and continuous data are expressed using the mean and SD. Survival was estimated using the Kaplan-Meier method. Univariate and multivariate analyses were performed using Cox proportional hazards regression, and the hazard ratio (HR) with its 95% CI was calculated. All important factors that might influence DFS and OS were considered for multivariate analysis, including pathologic tumor size (pT), pathologic node (pN), ER, PR, HER2, Ki-67, age, and AR. The χ² test or Fisher exact test was used to analyze the relationships between variables and AR. The PSM method was used to develop a cohort of patients with similar characteristics. Matching was performed using a 1:1 matching protocol with a caliper width equal to 0.2 of the SD of the logit of the propensity score. P < .05 was considered significant.

Our study included 5765 patients, among whom 58.13% (3351) were AR positive. AR positivity was common in those 65 years of age or older (278; 63.08%), in those with invasive lobular breast carcinoma (82; 62.60%), and in pT1 tumors (2022; 60.00%). AR was positively associated with ER and PR expression (Table 1). The median duration of follow-up was 3.67 years (interquartile range, 3.09–4.30; mean ± SD, 3.67 ± 0.85; range, 1.00–6.90).

A PSM cohort was developed to balance the baseline characteristics of ER, PR, HER2, Ki-67, pT, and pN. The PSM cohort included 3978 patients (1989 patients in each cohort), and no significant differences were found in baseline characteristics between AR⁺ and AR⁻ patients (Table 1).

The 5-year DFS of AR⁺ breast cancer in the entire cohort was 95.73% (3208 of 3351), compared with 92.21% (2226 of 2414) in AR⁻ breast cancer. Additionally, a higher 5-year OS rate was found in AR⁺ breast cancer than in AR⁻ breast cancer (98.84% versus 95.94%). Univariate analysis showed that a positive AR status was associated with longer OS (HR, 0.36; 95% CI, 0.25–0.52; P < .001) and DFS (HR, 0.60; 95% CI, 0.49–0.75; P < .001). After adjusting for age, pT, pN, ER, PR, HER2, and Ki-67, AR positivity was still associated with longer OS (HR, 0.44; 95% CI, 0.30–0.64; P < .001) and DFS (HR, 0.72; 95% CI, 0.57–0.90; P = .004).

In the PSM cohort, univariate analysis showed that AR positivity was associated with longer OS (HR, 0.47; 95% CI, 0.30–0.75; P = .001) and DFS (HR, 0.74; 95% CI, 0.57–0.97; P = .03). AR positivity was associated with longer OS (HR, 0.45; 95% CI, 0.28–0.71; P = .001) and DFS (HR, 0.72; 95% CI, 0.55–0.94; P = .01) after adjusting for age, pT, pN, ER, PR, HER2, and Ki-67 (Figure 1, A and B).

In our study, we did not observe any impact of AR status on OS and DFS in luminal A, luminal B, or HER2⁺ breast cancer patients, and AR positivity was associated with longer OS in triple-negative breast cancer (TNBC) (Tables 2 and 3; Supplemental Figure 1 [see the supplemental digital content containing 1 figure and 2 tables at https://meridian.allenpress.com/aplm in the September 2023 table of contents]).

There was a significant interaction across ER, PR, and AR (P < .001) in terms of DFS, but no significant interaction between AR and HER2 was observed (P = .44). AR⁺ was associated with superior DFS in ER⁺ (HR, 0.63; 95% CI, 0.44–0.89; P = .01) and PR⁺ (HR, 0.60; 95% CI, 0.39–0.90; P = .01) patients but not in ER⁻ (HR, 0.95; 95% CI, 0.63–1.43; P = .80) or PR⁻ (HR, 0.87; 95% CI, 0.62–1.24; P = .44) patients. AR⁺ was associated with superior OS in ER⁻ (HR, 0.35; 95% CI, 0.17–0.75; P = .004) and PR⁻ (HR, 0.40; 95% CI, 0.22–0.73; P = .001) patients but not in ER⁺ (HR, 0.58; 95% CI, 0.32–1.05; P = .07) or PR⁺ (HR, 0.62; 95% CI, 0.30–1.26; P = .18) patients. Multivariate Cox analyses confirmed that AR⁺ was associated with superior DFS in ER⁺ (HR, 0.62; 95% CI, 0.44–0.89; P = .01) and PR⁺ (HR, 0.58; 95% CI, 0.39–0.88; P = .01) patients, and AR⁺ was associated with superior OS in ER⁻ (HR, 0.29; 95% CI, 0.14–0.63; P = .002) or PR⁻ (HR, 0.37; 95% CI, 0.20–0.68; P = .001) patients (Tables 2 and 3).

We further analyzed the prognostic role of AR in combination with ER and PR. AR⁺ was associated with superior DFS (HR, 0.60; 95% CI, 0.39–0.90; P = .01) but not with superior OS (HR, 0.62; 95% CI, 0.30–1.62; P = .19) in ER⁺PR⁺ patients. In ER⁻PR⁻ patients, AR⁺ was associated with superior OS (HR, 0.35; 95% CI, 0.17–0.75; P = .005) but not with superior DFS (HR, 0.95; 95% CI, 0.63–1.43; P = .80). AR⁺ was associated with higher DFS (HR, 0.58; 95% CI, 0.39–0.88; P = .01) in ER⁺PR⁺ patients and superior OS (HR, 0.29; 95% CI, 0.14–0.63; P = .002) in ER⁻PR⁻ patients in the multivariate Cox analysis. However, in ER⁺PR⁻ patients, AR⁺ was not associated with superior DFS (HR, 0.71; 95% CI, 0.37–1.36; P = .30) or OS (HR, 0.50; 95% CI, 0.18–1.44; P = .20) (Tables 2 and 3).

When evaluating the combinations of AR and ER in the PSM cohort, ER⁺AR⁺ was associated with the longest DFS and OS among the different ER/AR combinations (Figure 1, C and D). PR⁺AR⁺ was also associated with the longest DFS and OS among different combinations of AR and PR statuses (Figure 1, E and F). The patients were classified into 6 categories based on ER, PR, and AR status. ER⁺PR⁺AR⁺ patients had the highest 5-year DFS rate (2293 of 2368; 96.83%), and ER⁻PR⁻AR⁻ patients had the lowest 5-year DFS rate (846 of 939; 90.10%) (Figure 1, G and H). Kaplan-Meier analysis showed that ER⁺PR⁺AR⁺ patients had the longest DFS and OS (Tables 4 and 5).

AR failed to predict prognosis in terms of DFS of those patients who received anthracycline (HR, 0.43; 95% CI, 0.16–1.15; P = .09), taxanes (HR, 0.69; 95% CI, 0.33–1.43; P = .31), anthracycline plus taxanes (HR, 0.76; 95% CI, 0.52–1.11; P = .15), trastuzumab (HR, 0.92; 95% CI, 0.51–1.67; P = .79), radiotherapy (HR, 0.72; 95% CI, 0.48–1.09; P = .12), tamoxifen (HR, 1.14; 95% CI, 0.49–2.67; P = .76), toremifene (HR, 0.34; 95% CI, 0.11–1.04; P = .06), letrozole (HR, 0.43; 95% CI, 0.17–1.11; P = .08), exemestane (HR, 1.48; 95% CI, 0.61–3.57; P = .38), and anastrozole (HR, 0.53; 95% CI, 0.23–1.21; P = .13) (Supplemental Table 1).

ER/PR/AR combinations could not predict the treatment effects for adjuvant trastuzumab but could be used for the selection of adjuvant chemotherapy and endocrine therapy. The shortest survival was found in ER⁺PR⁻AR⁻ patients receiving toremifene (HR, 9.77; 95% CI, 2.22–43.1; P = .001) (Figure 2, A), ER⁺PR⁻AR⁺ patients receiving exemestane (HR, 5.57; 95% CI, 1.57–19.72; P = .01) (Figure 2, B), ER⁺PR⁺AR⁻ patients receiving anthracycline (HR, 8.29; 95% CI, 1.02–67.59; P = .05) (Figure 2, C), and ER⁻PR⁻AR⁺ patients receiving taxanes (HR, 4.84; 95% CI, 1.28–18.37; P = .02) (Figure 2, D) among different ER/PR/AR combinations. ER⁺PR⁻AR⁻, ER⁻PR⁻AR⁺, and ER⁻PR⁻AR⁻ were associated with the shortest survival among those who received radiotherapy and anthracycline plus taxanes (Figure 2, E; Table 6) among the different combinations of ER, PR, and AR.

The prognosis did not differ across different chemotherapy and endocrine regimens for different ER/PR/AR combinations, except for ER⁺PR⁺AR⁻. We found that ER⁺PR⁺AR⁻ patients treated with anthracycline plus taxanes gained more survival benefits (HR, 0.36; 95% CI, 0.16–0.81; P = .01), and those who were treated with anastrozole (HR, 3.74; 95% CI, 1.44–9.71; P = .01) and letrozole (HR, 3.32; 95% CI, 1.31–8.47; P = .01) exhibited shortened survival in terms of DFS (Supplemental Table 2).

Among the AR⁺ patients who received endocrine therapy in the entire cohort, toremifene was not associated with longer DFS than tamoxifen (HR, 0.73; 95% CI, 0.23–2.33; P = .60). For AR⁻ breast cancer, tamoxifen was associated with longer DFS than toremifene (HR, 0.41; 95% CI, 0.19–0.89; P = .03) (Figure 2, F). In the PSM cohort, tamoxifen was associated with longer DFS than toremifene (HR, 0.40; 95% CI, 0.19–0.87; P = .02) for AR⁻ breast cancer (Supplemental Table 2). This difference was still significant after adjusting for age, pT, pN, ER, PR, HER2, and Ki-67 (HR, 0.41; 95% CI, 0.19–0.92; P = .03).

In our study, we confirmed that AR is an independent prognostic factor for breast cancer patients. Meanwhile, AR could divide ER⁺PR⁺ and ER⁻PR⁻ patients into 2 separate categories. Six categories of breast cancer patients were identified according to ER, PR, and AR status. This 6-classification system can identify patients with the best prognosis in the adjuvant setting, and might be used to predict chemotherapy and endocrine treatment effects in the adjuvant setting.

The prognostic role of AR in breast cancer is still under debate. Hwang et al¹¹  found that AR is an independent prognostic factor for both OS (HR, 0.59; 95% CI, 0.38–0.90) and DFS (HR, 0.43; 95% CI, 0.27–0.67), which was also confirmed by meta-analysis.^6,12  However, other studies have shown that AR expression is not an independent prognostic factor.^13,14  This difference might be due to the broader molecular profile, which might influence the functions of AR signaling and clinical outcome.⁶  Many studies have shown that AR is associated with a small tumor size and ER and PR expression, which might influence the prognosis of breast cancer patients.^15,16  PSM methods could ensure that important factors are similar in different groups in a real-world setting, which might bias the research findings. Our study is the first cohort study to use PSM methods to ensure that the baseline characteristics are balanced between AR⁺ and AR⁻ breast cancer patients, and finds that AR is an independent prognostic biomarker for breast cancer.

The prognostic role of AR in ER⁺ breast cancer has been established in individual studies.^12,17–20  Kensler et al¹⁷  found that AR expression was associated with a 47% reduction in ER⁺ cancers (HR, 0.53; 95% CI, 0.41–0.69) after a 7-year follow-up. A pooled analysis of 8 studies showed that AR was associated with improved 3-year OS (OR 0.45; 95% CI, 0.29–0.69).¹²  AR activity is mediated by binding to an ER half site 3′ to the FOXA1 and ERα-binding sites.²¹  AR potently inhibits ERα transactivation activity and 17β-estradiol–stimulated growth of breast cancer cells.²²  Agonist activation of AR alters the genomic distribution of ER and essential coactivators (p300, SRC-3), resulting in the repression of ER-regulated cell cycle genes and the upregulation of AR target genes, including known tumor suppressors.²³  This effect might explain why AR is associated with longer survival in ER⁺ breast cancer patients. We further analyzed the prognostic roles of AR in PR⁺, ER⁺PR⁺, and ER⁺PR⁻ breast cancer and found that AR was associated with increased DFS but not OS in PR⁺ and ER⁺PR⁺ patients. Activated PR directly interacts with ER, resulting in a widespread gain of ER-binding events in vitro and in vivo,²³  which might explain why AR is associated with longer survival in ER⁺PR⁺ breast cancer patients but not in ER⁺PR⁻ breast cancer patients.

The prognostic role of AR in ER⁻ breast cancer is inconsistent. AR expression is associated with a 62% increase in breast cancer mortality for ER⁻ cancers (HR, 1.62; 95% CI, 1.18–2.22).¹⁷  Other retrospective studies¹⁸  and meta-analyses¹²  do not confirm this finding. In our study, AR was associated with increased OS but not DFS in ER⁻ and ER⁻PR⁻ breast cancer patients. Although substantial evidence shows that AR can promote the survival of TNBC cells,²⁴  the prognostic roles of AR in TNBC are still under investigation. Several studies have found that AR is significantly associated with prolonged survival.^25,26  Bleach et al²⁷  found that AR is associated with reduced progression-free survival in TNBC. However, a meta-analysis of 4914 TNBC patients from 27 studies showed no correlation between AR and patients’ DFS and OS.²⁸  We found that AR was associated with increased OS, but not DFS, in PR⁻, ER⁻PR⁻, and TNBC patients, which suggests that AR might affect the prognosis of breast cancer patients without the influence of ER and PR. The role of AR for TNBC in in vitro studies and clinical studies is different, but no proper explanation could be given for this phenomenon. Further studies are needed to classify the role of AR in ER⁻PR⁻ breast cancer, especially in TNBC.

Hickey et al²³  found that AR agonists might downregulate the expression levels of ER and PR. GSE27473 data by Al Saleh et al²⁹  showed that ER silencing induces the upregulation of AR at the mRNA level. AR activity is mediated by ER, and AR might also inhibit ER activity.^21,22  This finding implies that there might be a significant interaction among ER, PR, and AR, which was confirmed by our study. Gong et al³⁰  showed that AR⁻ER⁻ patients have significantly worse OS and DFS than patients who have tumors with other combinations of AR/ER status. In our study, not only ER⁺AR⁺ but also PR⁺AR⁺ was associated with the longest DFS and OS, and ER⁻AR⁻ or PR⁻AR⁻ was associated with the shortest DFS and OS. Previous studies have confirmed that ER⁺AR⁺ is associated with the longest DFS and OS, and we first found that PR⁺AR⁺ is also associated with the longest DFS and OS. In our study, there was a significant interaction among ER, PR, and AR in DFS. We considered ER, PR, and AR together. Six categories were established, among which ER⁺PR⁺AR⁺ was associated with the longest survival and ER⁻PR⁻AR⁻ was associated with the shortest survival. Lakis et al³¹  classified breast cancer patients into luminal (ER⁺/PR^±/AR^±), molecular apocrine (ER⁻/PgR⁻/AR⁺), and hormone receptor–negative carcinomas (HR⁻, [ER⁻/PR⁻/AR⁻]). This type of classification might not work in Chinese patients, as survival differs across different types of ER⁺/PR^±/AR^± breast cancer patients. AR could distinguish the prognosis of ER⁺PR⁺ and ER⁻PR⁻, but not ER⁺PR⁻, breast cancer patients.

Akashi et al³²  observed a higher pathologic complete response in AR⁺ HER2⁺ early breast cancer patients in the neoadjuvant setting. In HER2-enriched metastatic breast cancer, AR is an independent prognostic factor for favorable PFS and OS and can predict the efficacy of first-line trastuzumab treatment.³³  Few studies have assessed the treatment predictive role of AR in the adjuvant setting. ER⁺ breast cancer patients benefit from tamoxifen regardless of AR expression.³⁴  The treatment effect does not significantly differ between letrozole and tamoxifen for AR⁺ (HR, 0.63; 95% CI, 0.44–0.75; P = .003) and AR⁻ tumors (HR, 0.39; 95% CI, 0.21–0.72; P = .002).⁷  AR negativity predicts early treatment failure with aromatase inhibitor but not tamoxifen.⁹  However, AR positivity fails to predict the treatment effects for adjuvant chemotherapy, endocrine therapy, and trastuzumab. This finding might be due to many other factors in the real-world setting that may affect the results. When considering ER, PR, and AR together, the 6-category system could be used for adjuvant treatment selection. Toremifene shows the worst response for ER⁺PR⁻AR⁻ breast cancer patients in terms of DFS, and exemestane shows the worst response for ER⁺PR⁻AR⁺ patients. Meanwhile, tamoxifen is associated with longer DFS than toremifene in ER⁺AR⁻ breast cancer. For ER⁺AR⁺ breast cancer, tamoxifen and toremifene might have similar survival benefits. Although we conducted PSM to ensure that each group had similar characteristics, this finding should be confirmed in a randomized setting. AR could not be considered a treatment predictive biomarker, but it has value when combined with ER and PR. Routine assessment of AR in the breast should be considered.

The threshold for AR positivity differs across different studies. The definition of AR positivity has been 10% or higher, 5%, or 1% staining in tumor cells. Many studies have used 10%, and we found that AR 10% or higher was associated with longer survival than negative AR. There were some other studies that used 1%. Several studies found AR 1% or higher was associated with longer survival than negative AR. In our study, we confirmed that AR 1% or higher was associated with longer survival than negative AR. Both AR 1% or higher and AR 10% or higher were associated with longer survival, and we thus used the smaller value.

This study is the largest study to comprehensively assesses the prognostic and treatment predictive roles of AR in the adjuvant setting for breast cancer. We found that PR might interact with AR and affect the prognosis of breast cancer patients. Thus, we considered ER, PR, and AR together and classified breast cancer patients into 6 categories. Furthermore, we found that AR can distinguish the prognosis of ER⁺PR⁺ and ER⁻PR⁻ but not ER⁺PR⁻ breast cancer patients. However, this study has several limitations. First, all data were collected retrospectively based on real-world data, so there might have been selection bias and attribution bias during data collection. Second, only the role of AR in the adjuvant setting was analyzed; its role in the neoadjuvant setting was not analyzed.

In our study, we found that AR 1% or higher could predict the prognosis of breast cancer and distinguish ER⁺PR⁺ and ER⁻PR⁻ breast cancer. Routine assessment of AR in the breast should be considered in the future. However, cutoff points for IHC receptor expression (≥1%, ≥5%, or ≥10%) were different across studies, and thus a well-recognized cutoff point should be proposed. Although we classified breast cancer patients into 6 groups and identified different prognoses for different groups, further studies are needed. For example, the ER⁺PR⁺AR⁺ group is considered to have the best prognosis, and the ER⁻PR⁻AR⁻ group is considered to have the worst survival. However, for any specific group, which treatment is optimal has not been well explored. In our study, we found that prognoses did not differ across different chemotherapy and endocrine therapy regimens for different ER/PR/AR combinations, except for ER⁺PR⁺AR⁻. We found that patients treated with anthracycline plus taxanes gained more survival benefits (HR, 0.36; 95% CI, 0.16–0.81; P = .01), whereas those who were treated with anastrozole (HR, 3.74; 95% CI, 1.44–9.71; P = .01) and letrozole (HR, 3.32; 95% CI, 1.31–8.47; P = .01) had shorter survival in terms of DFS. However, the sample size was small, and the results might be biased because of real-world data. Randomized controlled trials are needed for further confirmation. In the future, the roles of AR in breast cancer in adjuvant and neoadjuvant settings should be further researched.

AR is an independent prognostic biomarker for breast cancer patients, and patients with AR positivity are associated with longer DFS and OS. ER⁺PR⁺AR⁺ breast cancer patients had the longest survival, and ER⁻PR⁻AR⁻ breast cancer patients had the shortest survival. The worst survival was found in ER⁺PR⁻AR⁻ patients receiving toremifene, ER⁺PR⁻AR⁺ patients receiving exemestane, ER⁺PR⁺AR⁻ patients receiving anthracycline, and ER⁻PR⁻AR⁺ patients receiving taxanes. ER⁺PR⁻AR⁻, ER⁻PR⁻AR⁺, and ER⁻PR⁻AR⁻ patients were associated with the worst survival among those who received radiotherapy and anthracycline plus taxanes. Furthermore, we found that patients treated with anthracycline plus taxanes gained more survival benefits, and those who were treated with anastrozole and letrozole had shorter survival in terms of DFS for ER⁺PR⁺AR⁻ patients. AR in combination with ER and PR could predict the prognosis and treatment effects for chemotherapy, endocrine therapy, and radiotherapy in the adjuvant setting. In the future, routine assessment of AR in the breast should be considered.