Entinostat plus exemestane for hormone receptor-positive advanced breast cancer: long-term overall survival results from the EOC103A3101 randomized controlled trial

Abstract

The addition of entinostat, a histone deacetylase inhibitor, to exemestane has demonstrated improved progression-free survival (PFS) in hormone receptor-positive (HR + ), human epidermal growth factor receptor 2-negative (HER2-) advanced breast cancer (ABC) following endocrine therapy failure in the EOC103A3101 randomized trial (NCT03538171). This study reports long-term overall survival (OS) results. In this study, 354 patients with HR + /HER2- ABC that had recurred or progressed after endocrine therapy were randomized (2:1) to receive entinostat plus exemestane (n = 235) or placebo plus exemestane (n = 119). After a median follow-up of 31.8 months, the median OS was 38.6 months (95% confidence interval [CI]: 29.6–44.0) in the entinostat-exemestane group versus 29.2 months (95% CI: 22.5–41.6) in the placebo-exemestane group (hazard ratio [HR] 0.83, 95% CI: 0.62–1.10, P = 0.184). The entinostat-exemestane group demonstrated more pronounced OS benefits when adjusting subsequent antitumor therapies such as cyclin-dependent kinase 4/6 (CDK4/6) inhibitors, or the combination of endocrine therapy, CDK4/6 inhibitors, and chemotherapy. The sensitivity analysis based on the per-protocol set showed more significant OS benefits. The median time to post-discontinuation chemotherapy and chemotherapy-free survival in the entinostat-exemestane group were 11.7 and 9.3 months, respectively, compared to 6.8 and 5.6 months in the placebo-exemestane group (both P < 0.05). Grade ≥3 treatment-emergent adverse events occurred in 65.5% of patients in the entinostat-exemestane group, without new safety signals observed. In conclusion, among patients with HR + /HER2- ABC after endocrine therapy failure, entinostat plus exemestane was associated with a numerically longer OS, with an acceptable safety profile.

Introduction

Hormone receptor-positive (HR + )/human epidermal growth factor receptor 2-negative (HER2-) advanced breast cancer (ABC) accounts for approximately 70% of breast cancer cases and is associated with a significant disease burden.^1,2 Despite advancements in first-line treatment, which typically includes a combination of cyclin-dependent kinase 4/6 (CDK4/6) inhibitors and endocrine therapy,³ resistance remains a prevalent clinical challenge. Quite a number of patients may experience primary or secondary resistance to endocrine therapy.⁴ Approximately 12.5% of patients suffered disease progression within six months despite treatment.⁵ Post-progression management lacks a universally accepted standard of care. For patients with PI3K and ESR1 wild-type ABC, options include CDK4/6 inhibitor rechallenge or the addition of mTOR inhibitors. However, findings from the PALMIRA⁶ and PACE⁷ trials indicate limited efficacy for continued use of the same CDK4/6 inhibitors, whereas the MAINTAIN⁸ and postMONARCH⁹ trials suggest that switching to a different CDK4/6 inhibitor may provide modest progression-free survival (PFS) benefits, although no clear overall survival (OS) advantage has been demonstrated. While everolimus, an mTOR inhibitor, has shown improved outcomes in patients who progress after aromatase inhibitor therapy,¹⁰ real-world studies report a median PFS of only six months in patients previously treated with CDK4/6 inhibitors.¹¹ Despite the emergence of other novel therapies such as elacestrant and capivasertib, clinical benefits remain modest and often depend on biomarker-defined subgroups, limiting their broad applicability.

Epigenetic dysregulation plays a pivotal role in driving endocrine resistance, with histone deacetylase (HDAC) implicated in tumor growth, gene expression regulation, and treatment resistance.¹² HDAC inhibitors have emerged as promising agents for HR + /HER2- breast cancer. The ACE trial previously demonstrated that the combination of the HDAC inhibitor chidamide with exemestane significantly prolonged median PFS,¹³ but OS improvements were not observed.¹⁴ Entinostat, a novel HDAC inhibitor, selectively targets Class I and Class IV HDAC enzymes. Preclinical studies have shown that entinostat can upregulate estrogen receptor and aromatase expression, thereby restoring tumor sensitivity to endocrine therapy and overcoming primary endocrine resistance.¹⁵ Additionally, entinostat has demonstrated the ability to reverse acquired resistance, supporting its potential as a novel approach to improve outcomes in patients with HR + /HER2- breast cancer.¹⁶

Previous studies on the efficacy of entinostat in ABC have yielded mixed results. The phase II ENCORE301 trial demonstrated significant PFS and OS benefits with entinostat in combination with exemestane for patients with HR + ABC who had progressed following endocrine therapy.¹⁷ The subsequent phase III E2112 trial, conducted predominantly in North American patients, failed to show similar efficacy for entinostat.¹⁸ In contrast, the randomized, double-blind, placebo-controlled phase III EOC103A3101 trial in a Chinese population demonstrated a significant improvement in PFS with entinostat compared to placebo (6.32 vs. 3.72 months), with an acceptable safety profile.¹⁹ Entinostat in combination with aromatase inhibitors has been approved by the National Medical Products Administration (NMPA) for the treatment of HR + /HER2- locally advanced or metastatic breast cancer that has recurred or progressed following endocrine therapy. The most recent Chinese guidelines recommend the use of HDAC inhibitors as a treatment option for HR + /HER2- ABC after progression on CDK4/6 inhibitors.²⁰ Entinostat has shown a favorable safety profile, with a lower incidence of grade ≥3 AEs compared to chidamide, particularly for hematological toxicities such as decreased neutrophil count, decreased platelet count, and decreased white blood cell count. Furthermore, the population in these trials of entinostat is not restricted to a specific molecular subtype of HR + ABC, which indicates its potentially broader clinical applicability. Despite promising PFS data, the long-term OS benefit of entinostat remains unclear and warrants further investigation.

Consistent with the published PFS results from the EOC103A3101 trial, the previous analysis on the secondary endpoint of OS showed a trend favorable to entinostat treatment.¹⁹ In this report, we present the long-term OS outcomes from the EOC103A3101 trial. Additionally, we report updated long-term safety data to provide a comprehensive assessment of entinostat in HR + /HER2- ABC.

Results

Patients

A total of 354 patients were enrolled and randomized to receive either entinostat (n = 235) or placebo (n = 119). All participants received at least one dose of the study drug and were included in the full analysis set (FAS) and safety set (SS). The per protocol set (PPS) comprised 331 patients (93.5%), including 218 in the entinostat-exemestane group and 113 in the placebo-exemestane group. Twenty-three patients (6.5%) were excluded from the PPS due to the absence of post-treatment tumor assessment data (16/23, 69.6%), poor adherence to the treatment protocol (4/23, 17.4%), use of prohibited medications (3/23, 13.0%), and meeting exclusion criteria (1/23, 4.3%). Detailed baseline demographic and clinical characteristics for the study population have been previously reported and are summarized in Supplementary Table 1.

Overall survival

As of April 8, 2024, the median follow-up time for the overall population was 31.8 months (range, 0.4–58.5), during which 138 (58.7%) deaths occurred in the entinostat-exemestane group and 76 (63.9%) in the placebo-exemestane group, leading to 60% OS data maturity. The median OS was 38.6 months (95% confidence interval [CI]: 29.6–44.0) in the entinostat-exemestane group compared with 29.2 months (95% CI: 22.5–41.6) in the placebo-exemestane group, corresponding to a 17% relative reduction in the risk of death with entinostat (hazard ratio [HR] = 0.83, 95% CI: 0.62–1.10, P = 0.184) (Fig. 1a). The 3-year OS rates for the entinostat-exemestane and placebo-exemestane groups were 51% (95% CI: 45–58%) and 47% (95% CI: 37–56%), respectively, and the 4-year OS rates were 41% (95% CI: 34–47%) and 36% (95% CI: 27–45%), respectively. In the PPS, the median OS was 39.0 months for the entinostat-exemestane group and 34.0 months for the placebo-exemestane group, representing a 22% relative reduction in the risk of death (HR = 0.78, 95% CI: 0.59–1.05, P = 0.101).

Fig. 1.

Overall survival in the a overall study population and b prespecified subgroups. CI confidence interval, ECOG Eastern Cooperative Oncology Group, PR progesterone receptor, CDK4/6 cyclin-dependent kinase 4/6

Subgroup analyses demonstrated consistent results with those in the overall population, supporting the OS benefits from entinostat (Fig. 1b). Notably, despite the limited sample size among patients with prior salvage chemotherapy (n = 137), the median OS was 25.3 months in the entinostat-exemestane group and 13.8 months in the placebo-exemestane group, yielding an HR of 0.62 (95% CI: 0.40–0.94).

Overall survival after adjustment for subsequent antitumor treatments

As of the data cutoff, 184 patients (78.3%) in the entinostat-exemestane group and 101 patients (84.9%) in the placebo-exemestane group received subsequent antitumor treatments following the discontinuation of study treatment. As detailed in Table 1, the most common subsequent treatments included chemotherapy (56.6% vs. 66.4%), endocrine therapy (41.3% vs. 44.5%), CDK4/6 inhibitors (16.2% vs. 26.1%), and herbal extracts (9.8% vs. 3.4%, with details provided in Supplementary Table 2). Given the potential heterogeneity in post-discontinuation treatments between the two groups, post-hoc Cox regression analyses were performed to adjust for these factors to further estimate the effect of entinostat on OS. As shown in Table 2, within the FAS, adjusting for CDK4/6 inhibitor, endocrine therapy, chemotherapy, and herbal treatments yielded HRs ranging from 0.75 to 0.78, consistently indicating a relative reduction in the risk of death with entinostat. In the PPS, OS benefits were statistically significant (P < 0.05) after adjustments for CDK4/6 inhibitor alone, or CDK4/6 inhibitor plus endocrine therapy and chemotherapy, as well as after incorporating herbal treatments into the model.

Table 1.

Subsequent antitumor treatments in the entinostat and placebo-exemestane groups after discontinuation of study treatment

Treatment pattern	Entinostat-exemestane group (n = 235)	Placebo-exemestane group (n = 119)
Any subsequent antitumor treatment, No. (%)	184 (78.3)	101 (84.9)
Chemotherapy	133 (56.6)	79 (66.4)
Paclitaxel	76 (32.3)	38 (31.9)
Capecitabine	70 (29.8)	36 (30.3)
Vinorelbine	39 (16.6)	20 (16.8)
Gemcitabine	20 (8.5)	18 (15.1)
Docetaxel	18 (7.7)	12 (10.1)
Carboplatin	18 (7.7)	8 (6.7)
Cisplatin	12 (5.1)	13 (10.9)
Endocrine therapy	97 (41.3)	53 (44.5)
Fulvestrant	58 (24.7)	33 (27.7)
Letrozole	19 (8.1)	11 (9.2)
Exemestane	17 (7.2)	13 (10.9)
Estrogen receptor degraders	4 (1.7)	6 (5.0)
CDK4/6 inhibitors	38 (16.2)	31 (26.1)
Palbociclib	12 (5.1)	12 (10.1)
Abemaciclib	12 (5.1)	6 (5.0)
Herbal and plant extracts	23 (9.8)	4 (3.4)
Others	43 (18.3)	22 (18.5)
Chidamide‌	7 (3.0)	6 (5.0)

Treatments listed include those with a frequency >5% in at least one treatment group

Table 2.

Multivariate Cox model analysis for overall survival after adjustment for subsequent anticancer treatments

Adjusted factors	FAS		PPS
	HR (95%CI)	P value	HR (95%CI)	P value
Unadjusted	0.83 (0.62, 1.10)	0.184	0.78 (0.59, 1.05)	0.101
Adjusted for endocrine therapy	0.80 (0.61, 1.07)	0.129	0.78 (0.58, 1.04)	0.088
Adjusted for chemotherapy	0.82 (0.62, 1.09)	0.179	0.82 (0.62, 1.10)	0.193
Adjusted for CDK4/6 inhibitors	0.77 (0.58, 1.03)	0.076	0.74 (0.55, 0.99)	0.041
Adjusted for endocrine therapy, CDK4/6 inhibitors and chemotherapy	0.78 (0.58, 1.03)	0.080	0.74 (0.56, 1.00)	0.048
Adjusted for endocrine therapy, CDK4/6 inhibitors, chemotherapy and herbal treatments	0.75 (0.57, 1.00)	0.052	0.72 (0.54, 0.96)	0.028

HR hazard ratio, CI confidence interval, FAS full analysis set, PPS per protocol set, CDK4/6 cyclin-dependent kinase 4/6

Given the limited number of patients who received herbal extract treatment, a sensitivity analysis excluding these patients was conducted. As shown in Supplementary Table 3, the adjusted analyses in the FAS remained supportive of OS benefits from entinostat, with HRs ranging from 0.73 to 0.78. Statistical significance was achieved in the FAS after adjusting for CDK4/6 inhibitors, or the combination of endocrine therapy, CDK4/6 inhibitors, and chemotherapy. The sensitivity analysis based on PPS demonstrated even more pronounced OS benefits, with the original model reaching statistical significance (HR = 0.73, 95% CI: 0.54–0.99, P = 0.046).

Time to post-discontinuation chemotherapy and chemotherapy-free survival

As shown in Fig. 2a, the median time to post-discontinuation chemotherapy (TTC) for the entinostat-exemestane group was 11.7 months, approximately 5 months longer than the 6.8 months observed in the placebo-exemestane group (HR = 0.70, 95% CI: 0.53–0.92, P = 0.012). Similarly, chemotherapy-free survival (CFS) was also extended in the entinostat-exemestane group. As shown in Fig. 2b, the median CFS was 9.3 months in the entinostat-exemestane group, representing an extension of 3.6 months compared to 5.6 months in the placebo-exemestane group (HR = 0.75, 95% CI: 0.59–0.96, P = 0.023).

Fig. 2.

Kaplan‒Meier curves for a time to post-discontinuation chemotherapy (TTC) and b chemotherapy-free survival (CFS). CI confidence interval

Safety

The long-term follow-up safety data were consistent with previously reported findings,¹⁹ with no new safety signals identified. Treatment-emergent adverse events (TEAEs) of any grade were reported in 97.9% of patients in the entinostat-exemestane group and 79.8% in the placebo-exemestane group, with 95.7% and 69.7% being related to treatment, respectively (Supplementary Table 4). Grade ≥3 TEAEs occurred in 65.5% and 20.2% of patients, respectively. The most common grade ≥3 TEAEs in the entinostat-exemestane group included decreased neutrophil count (43.8%), decreased platelet count (8.5%), decreased white blood cell count (6.8%), increased γ-glutamyltransferase (6.4%) and anemia (4.7%) (Table 3).

Table 3.

Long-term safety profiles

Events, n (%)	Entinostat-exemestane group (n = 235)		Placebo-exemestane group (n = 119)
	Any grade	Grade ≥3	Any grade	Grade ≥3
At least one TEAE	230 (97.9)	154 (65.5)	95 (79.8)	24 (20.2)
TEAEs occurred in ≥10% of patients in either group
Neutrophil count decreased	176 (74.9)	103 (43.8)	6 (5.0)	1 (0.8)
Platelet count decreased	159 (67.7)	20 (8.5)	15 (12.6)	1 (0.8)
White blood cell decreased	152 (64.7)	16 (6.8)	8 (6.7)	0
Aspartate aminotransferase increased	67 (28.5)	7 (3.0)	30 (25.2)	3 (2.5)
Anemia	58 (24.7)	11 (4.7)	12 (10.1)	4 (3.4)
Hyperglycemia	52 (22.1)	3 (1.3)	12 (10.1)	3 (2.5)
Alkaline phosphatase increased	50 (21.3)	3 (1.3)	9 (7.6)	2 (1.7)
Alanine aminotransferase increased	49 (20.9)	4 (1.7)	27 (22.7)	2 (1.7)
Hypocalcemia	49 (20.9)	1 (0.4)	5 (4.2)	0
Fatigue	48 (20.4)	1 (0.4)	11 (9.2)	0
Urinary tract infection	47 (20.0)	1 (0.4)	20 (16.8)	0
GGT increased	46 (19.6)	15 (6.4)	13 (10.9)	6 (5.0)
Hypokalemia	44 (18.7)	10 (4.3)	8 (6.7)	1 (0.8)
Hypoalbuminemia	38 (16.2)	0	8 (6.7)	0
Nausea	34 (14.5)	0	4 (3.4)	0
Hypophosphatemia	32 (13.6)	9 (3.8)	3 (2.5)	0
Anorexia	28 (11.9)	1 (0.4)	4 (3.4)	0
Hypertriglyceridemia	27 (11.5)	8 (3.4)	4 (3.4)	0
Fever	17 (7.2)	0	13 (10.9)	0
Arthralgia	12 (5.1)	0	12 (10.1)	0

TEAE treatment-emergent adverse event, GGT γ-glutamyltransferase

Biomarker analysis

Peripheral blood mononuclear cell samples were analyzed using mass spectrometry to detect histone acetylation sites. Univariate analysis identified pre-treatment histone acetylation markers associated with OS (Supplementary Table 5, Fig. 3a). Among these, risk-associated histone acetylation markers were functionally enriched in pathways related to DNA damage, telomere stress, and HDACs that deacetylate histones (Fig. 3b). Specifically, the acetylation site H2BC1_K59 exhibited a notable protective prognostic effect, with higher acetylation levels associated with longer survival (Supplementary Table 5, Fig. 3c). For post-treatment histone acetylation markers, risk-related factors were also enriched in pathways such as HDACs that deacetylate histones (Supplementary Table 5, Fig. 3d–f). H2BC1_K59 remained a significant protective factor, along with UGT2A3_K332, which also demonstrated a protective association with prognosis (Supplementary Table 5, Fig. 3g). Regarding changes in histone acetylation (post-treatment versus pre-treatment), increased acetylation levels of H4C1_K32 and H2BC21_K58 were identified as protective factors (Supplementary Table 5, Fig. 3h–j).

Fig. 3.

Prognostic biomarkers of histone acetylation related to overall survival (OS). a Volcano plot of univariate Cox analysis for pre-treatment histone acetylation markers associated with OS. The hazard ratio (HR) was calculated using the low-acetylation level group as the comparator (coded as 1) and the high-acetylation level group as the reference (coded as 0). Thus, an HR > 1 indicates that a higher acetylation level is associated with a lower risk of death, representing a protective factor. b Functional enrichment bubble plot for risk-associated pre-treatment markers. c Determination of the optimal cutoff for pre-treatment H2BC1_K59 level and its corresponding Kaplan‒Meier survival curves. d Volcano plot of univariate Cox analysis for post-treatment histone acetylation markers associated with OS. The HR was calculated using the low-acetylation level group as the comparator (coded as 1) and the high-acetylation level group as the reference (coded as 0). Thus, an HR > 1 indicates that a higher acetylation level is associated with a lower risk of death, representing a protective factor. e Functional enrichment bubble plots for protective post-treatment markers. f Functional enrichment bubble plots for risk-associated post-treatment markers. g Optimal cutoffs for post-treatment H2BC1_K59 and UGT2A3_K332 levels and Kaplan‒Meier survival analysis. h Volcano plot of univariate Cox analysis for histone acetylation markers based on the change in acetylation levels (post- versus pre-treatment) associated with OS. The HR was calculated using the low-acetylation level group as the comparator (coded as 1) and the high-acetylation level group as the reference (coded as 0). Thus, an HR > 1 indicates that a higher acetylation level is associated with a lower risk of death, representing a protective factor. i Functional enrichment bubble plot for risk-associated markers based on the change in acetylation levels. j Optimal cutoffs for changes in H4C1_K32 and H2BC21_K58 levels and their Kaplan‒Meier survival curves

To further explore the association between dynamic changes in acetylation biomarkers and treatment response, we performed a machine learning-based analysis using changes in acetylation levels (post-treatment versus pre-treatment). Patients were categorized as having disease control (partial response or stable disease) or progressive disease. Feature selection identified a panel of candidate biomarkers associated with treatment response, including TMEM192_K201, CALM2_K116, and COPS7B_K6, among others (Supplementary Table 6). A logistic regression model constructed using these markers demonstrated discriminatory ability in predicting treatment response, as shown by receiver operating characteristic (ROC) analysis (Supplementary Fig. 1). The area under the receiver operating characteristic curve (AUC) was 0.75 (95% CI, 0.68–0.83).

Discussion

In this study, the combination of entinostat and exemestane was associated with a numerically longer OS in patients with HR + /HER2- ABC after relapse or progression on endocrine therapy, with a median OS difference of 9.4 months. Nevertheless, this difference did not reach statistical significance in the primary analysis, which was mainly because this study was not powered for the test of OS. This survival benefit aligns with the previously reported PFS advantage,¹⁹ reinforcing the clinical efficacy of this treatment approach. Notably, the OS benefit was more pronounced in the subgroup of patients who had received prior salvage chemotherapy, with an OS extension of approximately 11.5 months. Entinostat also significantly delayed the initiation of subsequent chemotherapy and improved CFS. Importantly, after adjusting for the types of post-discontinuation treatments received, the OS benefit was even greater, further supporting the robustness of these findings. Long-term follow-up data reaffirmed the manageable safety profile of entinostat, with no new safety signals observed.

The findings of our study are consistent with the results of the phase II randomized ENCORE301 trial,¹⁷ both demonstrating significant improvements in PFS and OS with entinostat in combination with exemestane compared with exemestane plus placebo in patients with HR + /HER2- ABC who had progressed following endocrine therapy. However, these findings contrast with those of the international phase III E2112 trial, which did not show a significant difference in PFS with exemestane plus entinostat or placebo plus exemestane.¹⁸ The E2112 trial reported a median OS of 23.4 months compared to 21.7 months.¹⁸ There are notable differences between the trials in the study population and baseline characteristics. The proportion of patients in the experimental and control cohorts who had received prior fulvestrant treatment was comparable (26.4% vs 25.2% in our study and 29.5% vs 31.7% in the E2112 trial), indicating that baseline endocrine resistance was similar across studies. However, the E2112 trial had a higher proportion of patients who had received CDK4/6 inhibitors (approximately 35%), which may have contributed to poorer overall outcomes. Additionally, the E2112 trial was an international multicenter investigator-initiated phase III randomized controlled trial, with notable differences in study design. The E2112 trial also faced notable challenges in study execution. Despite plans to recruit from 658 centers (NCT02115282), only 111 centers actively enrolled patients, and 14% of patients were reported to have violated the inclusion/exclusion criteria.¹⁸ The median treatment duration in the entinostat-exemestane group of the E2112 trial was only 79.5 days, substantially shorter than the 106 days observed in the ENCORE301 trial¹⁷ and 140 days in our study, which may reflect reduced drug exposure and a potential impact on efficacy. Furthermore, the different results between trials could also be due to variations in epigenetic profiles and gene mutations across regions and ethnic groups,^21–23 which may influence how patients respond to HDAC inhibitors. Given that the therapeutic efficacy depends on sustained drug exposure to achieve epigenetic modulation, the shorter treatment duration may have directly impacted its therapeutic effect. Recent evidence from a systematic review and meta-analysis of randomized trials evaluating entinostat plus exemestane demonstrated a significant improvement in PFS, supporting the efficacy of this combination.²⁴ However, no statistically significant improvement in OS was observed, although this finding should be interpreted with caution given the immaturity of survival data in some included studies. Other endpoints, including objective response rate and clinical benefit rate, numerically favored the entinostat combination without reaching statistical significance.

Notably, a more pronounced trend toward improved OS was observed in patients with prior chemotherapy exposure. Although this finding should be interpreted with caution, several hypotheses may be considered. Prior chemotherapy may induce epigenetic alterations, including changes in HDAC-related transcriptional regulation, potentially increasing tumor susceptibility to HDAC inhibition.²⁵ In addition, HDAC-mediated mechanisms have been implicated in resistance to multiple chemotherapeutic agents, and HDAC inhibitors may partially reverse such resistance. Furthermore, the observed prolongation of time to chemotherapy and CFS with entinostat may delay the need for further chemotherapy and contribute to improved clinical outcomes. These hypotheses remain exploratory and warrant further investigation.

The results of the PALOMA-3 and MONARCH2 trials showed that adding CDK4/6 inhibitors to fulvestrant in the second-line setting was associated with a median OS prolongation of approximately 6.9 to 9.4 months (P = 0.09 and P = 0.01, respectively), providing a clinical context for interpreting the magnitude of OS differences observed in this setting.^26,27 The phase III ACE trial evaluated chidamide, another HDAC inhibitor, in combination with exemestane in patients with HR + /HER2- ABC who had progressed after endocrine therapy.¹³ That study demonstrated an improvement in PFS but did not show a statistically significant OS benefit.¹⁴ Compared with the ACE trial, our study included a population with a relatively higher disease burden and broader clinical characteristics, including both premenopausal and postmenopausal patients. While cross-trial comparisons should be interpreted with caution, the findings from our study suggest that entinostat in combination with exemestane was associated with a numerically longer OS, alongside a manageable safety profile, without new safety signals observed during long-term follow-up.

CDK4/6 inhibitors have an established role in the treatment of HR + /HER2- ABC following endocrine therapy failure. In the PALOMA-3 trial, the combination of palbociclib with fulvestrant significantly improved PFS, although the observed OS extension did not reach statistical significance.^28,29 In the present study, entinostat combined with exemestane was associated with a numerically longer OS. Notably, the median OS in the placebo-exemestane groups across studies was comparable, providing context for interpretation, although cross-trial comparisons should be approached with caution. In addition, the entinostat-exemestane group demonstrated prolonged time to chemotherapy and CFS, which may have clinical relevance given the impact of chemotherapy on quality of life. However, only a small proportion of patients in this study had prior exposure to CDK4/6 inhibitors, reflecting differences in treatment availability during the study period. Therefore, the role of entinostat in the post-CDK4/6 inhibitor setting warrants further evaluation in future studies. Emerging targeted therapies, such as capivasertib in combination with endocrine therapy, have demonstrated efficacy in biomarker-selected populations, although benefits may be limited in unselected patients.^30–32 Future research should focus on identifying biomarker-defined subgroups that may derive greater benefit from HDAC inhibitors, as well as exploring rational combination strategies with other targeted agents or immunotherapies.

Currently, the exploration of biomarkers for HDACis remains relatively limited. Prior studies have primarily focused on global changes in lysine acetylation rather than site-specific acetylation events or their associated biological pathways. For example, exploratory pharmacodynamic analyses from the E2112 study demonstrated increased lysine acetylation levels in peripheral blood mononuclear cells following entinostat treatment, while no substantial changes were observed in the control group.¹⁸ Similarly, the ENCORE301 study using multiplex flow cytometry reported increased protein lysine acetylation in peripheral blood cells after HDACi exposure.¹⁷ However, these studies largely provided a global assessment of acetylation without detailed characterization of specific acetylation sites or their functional relevance, highlighting an important gap in the current understanding of HDACi-related biomarkers. In this context, our study applied acetylation mass spectrometry to enable a more granular evaluation of site-specific acetylation changes and their potential association with clinical outcomes. We identified several histone acetylation markers associated with survival, including H2BC1_K59 and H4C1-related sites, which were also supported by the machine learning-based analysis, underscoring the potential relevance of histone acetylation in treatment response. The biomarker model demonstrated moderate discriminatory ability (AUC 0.75), and TMEM192_K201 contributed substantially to the model, suggesting a potential role in mediating the response to HDAC inhibition. TMEM192 is a lysosomal membrane protein implicated in the regulation of the lysosome-autophagy-apoptosis axis,^33,34 processes that are closely linked to the mechanism of action of HDAC inhibitors. Previous studies have shown that HDAC inhibitors can modulate autophagy through pathways such as mTOR inhibition, LC3 and Beclin-1 upregulation, and activation of the PIK3C3-BECN1 complex, ultimately influencing cell fate through lysosome-dependent autophagic flux.^35–37 These observations provide a biologically plausible framework linking acetylation changes to therapeutic response, although further mechanistic validation is needed. The prognostic significance of the identified biomarkers, including H2BC1_K59, H4C1_K32, and H2BC21_K58, and their potential role in predicting sensitivity to HDAC inhibitor therapy warrant further investigation. In addition, post-treatment UGT2A3_K332 was identified as a potential protective prognostic factor. UGT2A3 has been associated with cancer prognosis in prior studies,^38–40 although its mechanistic role remains incompletely understood. Besides, UGT2A3 is primarily expressed in organs pivotal to drug metabolism such as the liver, small intestine, and kidneys, and its acetylation status might indirectly influence treatment efficacy by modulating drug pharmacokinetics. Notably, emerging evidence suggests a close relationship between HDAC activity and hepatic metabolism,^41,42 raising the possibility that metabolic pathways may influence the efficacy of HDAC inhibition. Given that entinostat targets HDACs, these findings suggest that acetylation-related metabolic processes may contribute to treatment response and merit further investigation. Overall, these results provide preliminary insights into the potential biological mechanisms underlying HDAC inhibitor activity and may inform future efforts to optimize patient selection.

Functional enrichment analysis of prognosis-associated acetylation sites revealed that a substantial proportion of pathways were enriched in histone acetylation and deacetylation processes, such as “HDACs deacetylate histones” and “HATs acetylate histones.” This finding underscores the critical role of inhibiting histone deacetylase pathway activity by entinostat in influencing prognostic risk. Additionally, several other prognosis-related acetylation-enriched pathways, including DNA damage, DNA repair, and cell cycle, warrant attention, as they are closely linked to the mechanisms underlying entinostat-induced tumor cell apoptosis. Furthermore, the enrichment of the pathway “cellular response to hormone stimulus” is also notable, given that entinostat has been shown to functionally upregulate estrogen receptor and aromatase, thereby enhancing sensitivity to aromatase inhibitors.¹⁵ This may represent a key mechanism by which entinostat contributes to long-term prognosis in breast cancer. Together, these enrichment pathway analyses provide preliminary insights for further mechanistic investigation and therapeutic optimization.

This study has several limitations. First, the selection of subsequent treatments following discontinuation of entinostat was not randomized, resulting in differences in the distribution of post-discontinuation therapies between the two groups, which may have introduced potential bias in the OS analysis. Although a Cox regression model was employed to adjust for the most commonly used subsequent treatment modalities, the residual confounding effects of unadjusted treatments cannot be entirely excluded. Nevertheless, the potential influence of this bias is likely minimal and does not compromise the validity of the study’s conclusions. Second, due to historical factors at the time of enrollment, the proportion of patients with prior exposure to CDK4/6 inhibitors in this study was low, largely reflecting the timing of their clinical adoption in China. Although a similar trend toward improved OS was observed in the small subgroup of patients with prior CDK4/6 inhibitor exposure, the sample size was limited, and no definitive conclusions can be drawn. Therefore, the efficacy of entinostat in patients with HR + /HER2- ABC who have progressed following CDK4/6 inhibitor therapy warrants further investigation in future studies. In addition, biomarker analyses were conducted only in the entinostat-treated group, which limits the ability to distinguish predictive effects. Future studies incorporating both treatment and control groups are warranted to validate these findings and better define the clinical utility of these biomarkers.

In conclusion, this study shows that the combination of entinostat and exemestane significantly improves PFS in patients with HR + /HER2- ABC who have progressed following endocrine therapy and was associated with a numerically longer OS. The OS advantage is particularly pronounced in patients who had received prior salvage chemotherapy. Long-term follow-up data further support the favorable safety profile of entinostat, with no new safety signals observed. These findings suggest entinostat as a promising therapeutic option for patients with HR + /HER2- ABC, offering both efficacy and safety in a setting where treatment options remain limited.

Materials and methods

Trial design and patients

This multicenter, randomized, double-blind, placebo-controlled phase 3 clinical trial was conducted across 35 clinical centers in China to evaluate the efficacy and safety of entinostat combined with exemestane in patients with HR + /HER2- ABC who had experienced recurrence or progression following prior endocrine therapy.

Eligible patients were women aged 18–75 years with HR + /HER2-ABC, whose disease had recurred or progressed after endocrine therapy, with Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, and an anticipated life expectancy of at least 12 weeks. The use of prior chemotherapy for the metastatic setting should be no more than one line. Patients with current or prior central nervous system (CNS) metastases were excluded. Participants were randomized in a 2:1 ratio to receive oral exemestane (25 mg/day) in combination with either entinostat (5 mg/week) or placebo (5 mg/week) in 28-day treatment cycles. Randomization was implemented using a minimization method to balance the following covariates: previous treatments with CDK4/6 inhibitors, fulvestrant, and salvage chemotherapy, as well as the presence of visceral metastases. Blinding of treatment allocation was maintained for patients, investigators, research staff, imaging pathologists, and the sponsor. The trial methodology has been previously reported,¹⁹ and the detailed protocol is available in the Supplementary Materials.

The trial protocol was reviewed and approved by the ethics committees of all participating centers and adhered to the principles of Good Clinical Practice, the Declaration of Helsinki, and relevant Chinese regulatory requirements (Ethics Committee Approval Number: 17-190/1446 for National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College; Supplementary Table 7). Written informed consent was obtained from all participants prior to study enrollment. The trial was registered with ClinicalTrials.gov (NCT03538171).

Endpoints

The primary endpoint of this study was PFS as assessed by an independent radiographic committee (IRC) and has been previously reported.¹⁹ The prespecified secondary endpoint in this study was OS, defined as the time from randomization to death from any cause. This analysis also included post hoc endpoints: TTC, defined as the interval from randomization to the initiation of subsequent chemotherapy following study treatment discontinuation; and CFS, defined as the duration from randomization to the initiation of subsequent chemotherapy following study treatment discontinuation, or death, whichever occurred first. In addition to efficacy endpoints, this report also provides an update on long-term safety outcomes. An exploratory biomarker endpoint was also included to evaluate the association between histone acetylation and the treatment efficacy of entinostat-exemestane.

Biomarker exploration

Peripheral blood mononuclear cell samples were collected from patients in the entinostat-exemestane group before treatment and on the first day of the second treatment cycle. Histone acetylation levels were measured in these samples. Three biomarker-efficacy association analyses were performed: (1) association between pre-treatment histone acetylation levels and treatment efficacy; (2) association between post-treatment histone acetylation levels and efficacy; and (3) association between the change in histone acetylation levels (post-treatment versus pre-treatment) and efficacy. Detailed methodologies for biomarker analysis are described in the Supplementary Methods.

Statistical analysis

The sample size calculation for this study primarily focused on PFS, as previously described.¹⁹ Following intent-to-treat (ITT) principle, the FAS included all randomized patients, while the per protocol set PPS comprised FAS patients who had no major protocol deviations that might have impact on the efficacy evaluation. The SS included all patients who received at least one dose of the study drug. The cutoff date for this prespecified OS analysis was defined at approximately 60% OS data maturity. Median OS and survival curves were estimated using the Kaplan‒Meier method, and HRs with 95% CIs were derived using Cox proportional hazards regression models. OS was also evaluated in prespecified subgroups. Considering that a proportion of patients received new antitumor treatments after discontinuation, post hoc multivariate Cox regression analyses for OS were performed to adjust subsequent antitumor treatments. TTC and CFS were analyzed similarly using Kaplan‒Meier estimations. The above statistical analyses were performed using SAS software version 9.4. The statistical analytic methods involved in biomarker analysis such as data processing, bioinformatic analysis, and association analyses, are detailed in the Supplementary Methods. All the P values reported in this paper were nominal, with a two-sided P value of <0.05 considered statistically significant.

Author contributions

Binghe Xu, Jiani Wang and Qingyuan Zhang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Binghe Xu and Jiani Wang were involved in the study concepts and design. All authors (Binghe Xu, Jiani Wang, Qingyuan Zhang, Xichun Hu, Tao Sun, Wei Li, Quchang Ouyang, Jingfen Wang, Zhongsheng Tong, Min Yan, Huiping Li, Xiaohua Zeng, Changping Shan, Xian Wang, Jian Zhang, Yue Zhang, Liang Zhang, Hongyan Shang) were involved in the acquisition, analysis and interpretation of data. Jiani Wang supervised the analysis. Binghe Xu was involved in the draft of the manuscript. All the authors have read, critically revised and approved the manuscript.

Data availability

The clinical datasets generated and/or analyzed during the current study are not publicly available due to ethical and regulatory restrictions related to patient privacy and informed consent. De-identified individual participant data and relevant supporting clinical trial documents may be made available from the corresponding author upon reasonable request, subject to review and approval of a research proposal and a data access agreement in accordance with institutional and regulatory requirements. The mass spectrometry proteomics data generated in this study have been deposited to the ProteomeXchange Consortium (https://proteomecentral.proteomexchange.org) via the iProX partner repository with the dataset identifier PXD078835.^43,44

Competing interests

Hongyan Shang is an employee of Taizhou EOC Pharma Co., Ltd, Taizhou, China. All other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41392-026-02899-8.