Relationship between menopausal hormone therapy and incidence risk of breast cancer: systematic review and meta-analysis

Abstract

Background

Menopausal hormone therapy (MHT) is widely prescribed for alleviating menopausal symptoms. Prior studies have mainly focused on individual hormone therapy formulations. This meta-analysis comprehensively synthesized evidence across various regimens to systematically evaluate the association between MHT and breast cancer risk.

Materials and methods

We systematically searched CNKI, Wanfang, PubMed, and Web of Science from inception through August 2024. Eligible studies examining breast cancer risk following MHT were independently screened and assessed by two reviewers. The review and meta-analysis followed PRISMA guidelines.

Results

Thirty-four studies were included. The random-effects model showed a significant but heterogeneous overall association (OR = 1.15, 95% CI: 1.09–1.22; I² = 92.4%). Subgroup analysis identified hormone type, use status, and region as key determinants (P for interaction < 0.001). Stratification by hormone type resolved much of the heterogeneity, revealing risk confined to estrogen-progestin therapy (EPT; OR = 1.44, 95% CI: 1.26–1.64), while estrogen-only therapy (ET) showed no overall association (OR = 1.00, 95% CI: 0.91–1.10). Study type, region, and sample size were significant effect modifiers. For ET, randomized controlled trials demonstrated a protective effect (OR = 0.78, 95% CI: 0.70–0.87), contrasting with neutral findings from observational studies.

Conclusions

MHT is associated with a modest but significant increase in breast cancer risk, primarily driven by EPT. This risk was not observed with ET in observational studies, though trials suggested a protective effect. Crucially, the association shows marked geographical heterogeneity, indicating risk is modified by regimen and regional factors.

KEY MESSAGES

1. The increased breast cancer risk associated with MHT is mainly attributable to EPT, current user in Europe, with no overall risk observed for ET.

2. The apparent effect of ET is critically influenced by study design, with randomized trials suggesting a protective effect, in contrast to neutral findings from observational studies.

3. The association between MHT and breast cancer risk is not uniform but exhibits marked geographical heterogeneity, indicating that regional factors substantially modify the risk profile.

Introduction

Breast cancer, a prevalent hormone-dependent malignancy, ranks as the fifth most common cancer in women globally in terms of both incidence and mortality [1,2]. Despite extensive research, the precise etiology and molecular mechanisms of breast cancer remain unclear, though evidence suggests associations with genetic, environmental, and hormonal factors [3,4]. Menopausal hormone therapy (MHT) is an FDA-approved treatment for menopausal symptoms, including vasomotor disturbances and sexual dysfunction [5]. MHT, involving estrogen-only therapy (ET) and estrogen-progestin therapy (EPT), alleviates menopausal symptoms by supplementing exogenous sex hormones. It is a common and effective option typically initiated around menopause and continued for several years [6]. The development of MHT began in the 1890s with desiccated ovarian estrogen. Key milestones included the isolation of oestrone (1929) [7], oestradiol (1936) [8], and progesterone (1934) [9], leading to the marketing of early products like Progynon and conjugated estrogens. Initially uncontroversial, safety concerns emerged in the late 1970s when estrogen alone was linked to endometrial cancer [10], and then, the first report of MHT as a potential risk factor of breast cancer was released [11].

The WHI trial established that long-term (> 5 years) EPT increases breast cancer risk (HR = 1.26) [12], while ET may reduce the risk (RR = 0.77) [13]. Afterward, many well-designed studies were conducted to observe potential associations between MHT and breast cancer risk. Some studies showed a lower risk of breast cancer during short-term MHT (OR = 1.1 to 1.3), but the risk increased significantly (OR = 1.3 to 1.6) during a long-term use (> 5 years). Subsequent research further shows that risk increases with treatment duration and is influenced by progestin type, with synthetic versions like medroxyprogesterone acetate posing a higher risk than natural alternatives [14]. Current evidence on the association between MHT and breast cancer risk is inconsistent and often limited by a lack of focus on specific therapeutic variables. To address this, our study will quantitatively synthesize global evidence to determine the overall risk, while examining variations due to study methodology, patient characteristics, treatment type, duration, and cancer subtype.

Materials and methods

Registration and protocol

This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15] with detailed definitions presented in Additional file (PRISMA_2020_checklist). The study protocol was preregistered on the International Prospective Register of Systematic Reviews (PROSPERO, CRD42024613824).

Information sources and search strategy

We comprehensively and systematically searched the CNKI, Wanfang, PubMed, and Web of Science databases according to the predefined search terms from inception to August 2024. Relevant articles were identified using keywords and Mesh terms relating to menopause, hormone therapy, and breast cancer. As an example, the search strategy for the database PubMed is: ((menopause[MeSH Terms]) AND (Hormone Therapy[MeSH Terms])) AND (breast cancers[MeSH Terms]).

Study selection criteria

Studies were included according to predefined criteria:

(1) Conducted in women who were either aged ≥ 40 years or explicitly described as perimenopausal/postmenopausal (based on self-report, menstrual history, or biochemical parameters), without pre-existing breast cancer; (2) Employed observational (cohort or case-control) or experimental (randomized controlled trial) designs; (3) The intervention group comprised MHT users (receiving ET, PT, EPT, or Tibolone wherein ET refers to systemic estrogen administration, vaginal estrogen formulations were excluded from this category). The control group was defined according to the study design: in observational studies (cohort or case-control), controls were ‘never-users’ of MHT; in experimental studies (randomized controlled trials), controls received a placebo or no treatment; (4) Reported breast cancer incidence; (5) Provided comparative outcome data (e.g. cancer rates in ‘never-users’ vs. ‘ever-users’ [current/past]) or summary estimates: Odds Ratio (OR), Relative Risk (RR) or Hazard Ratio (HR) with confidence intervals or exact P-values.

Exclusion criteria encompassed: reviews, conference abstracts, case reports, non-comparative trials, studies with unavailable full texts, and those lacking extractable quantitative data.

Data synthesis

Literature screening was performed using EndNote. Records retrieved from databases were imported, deduplicated, and then screened by title and abstract to exclude irrelevant publication types. The remaining full-text articles were assessed for eligibility based on the predefined criteria.

Using a standardized data extraction form, two independent reviewers meticulously extracted: (1) Study characteristics: first author, country/region, publication year; (2) Population baseline: sample size, age distribution, intervention/follow-up duration; (3) Outcome measures: breast cancer incidence, metastasis events, or effect estimates (OR, RR, HR) with 95% confidence intervals (CIs); (4) Study design: randomized controlled trial, cohort, or case-control. All extracted data were systematically organized in Excel.

Literature quality assessment

The Newcastle-Ottawa Scale (NOS) was utilised to evaluate the quality of literature pertaining to both case-control and cohort studies [16]. The NOS employs an ‘ * ‘ system to rate the quality of literature on a scale of 0–9, with ≤ 3 designated low quality, 4 ∼ 6 as moderate quality, and ≥ 7 classified as high quality [17]. Use the modified Cochrane Collaboration tool to assess risk of bias for randomized controlled trials. Bias is assessed as a judgment (high, low, or unclear) for individual elements from five domains (selection, performance, attrition, reporting, and other) [18].

Statistical method

Data were analyzed using Stata (version 18.0). Pooled odds ratios (ORs) with 95% confidence intervals (CIs) were calculated as the summary effect measure. For studies that reported only relative risks (RR) or hazard ratios (HR), study-specific odds ratios (ORs) were computed from raw counts structured in a 2 × 2 format. In prospective studies, cells were defined as: (a) MHT users who developed breast cancer, (b) MHT users who did not, (c) non-users who developed breast cancer, and (d) non-users who did not. In retrospective studies, the corresponding cells were: (a) cases exposed to MHT, (b) controls exposed to MHT, (c) cases unexposed, and (d) controls unexposed. The OR was derived as (a × d)/(b × c). We converted these measures to ORs to maintain a consistent effect metric for meta-analysis. This conversion is considered valid under the assumption of a low baseline event risk (< 10%), which holds true for the annual incidence of breast cancer in the general population of postmenopausal women. The methodological basis for this approximation follows the guidance provided in the Cochrane Handbook.

All meta-analyses were performed using Stata software (version 18.0; StataCorp) with the ‘metan’ command, which calculated pooled odds ratios using the inverse-variance weighting method under a random-effects model. Between-study heterogeneity was quantified using the I² statistic.

In the meta-analysis, multi-arm studies with different hormone therapy groups were pooled as ‘MHT users’ or ‘Non-MHT users’ for overall effect estimation. In subgroup analyses, however, these groups were analyzed separately as: ‘Estrogen only (ET)’, ‘Progestin only (PT)’, ‘Estrogen plus progestin (EPT)’, and ‘Tibolone’.

Heterogeneity was assessed using the I² statistic, in accordance with Cochrane guidelines. A random-effects model, which accounts for both within- and between-study variance, was used to pool the effect estimates. Heterogeneity was quantified as low (I² ≤ 50%), moderate (I², 51–75%), or high (I² > 75%) [19]. Subgroup analyses were conducted to explore sources of heterogeneity based on study type (case-control, cohort, randomized controlled trial), region (North America, Europe, Asia, other), sample size (< 100,000, ≥ 100,000), age (< 50, ≥ 50 years), median follow-up (< 10, ≥ 10 years), intervention duration (< 5, ≥ 5 years), type of hormone therapy (ET, PT, EPT, Tibolone), hormone use status (current, past), and cancer subtype (breast cancer, invasive breast cancer, carcinoma in situ/mixed). The statistical significance of differences across subgroups was evaluated using appropriate interaction tests.

Given that previous research has identified hormone type as important determinants of breast cancer risk among MHT users [12,20–22], we further prespecified a stratified analysis by hormone type to more precisely characterize these associations. First, summary odds ratios (ORs) were calculated separately for ET and EPT. Subsequently, within each hormone type, additional subgroup analyses were conducted based on study design, region, sample size (< 10,000 vs. ≥ 10,000), age (< 50 vs. ≥ 50 years), median follow-up (< 10 vs. ≥ 10 years), intervention duration (< 5 vs. ≥ 5 years), and cancer subtype (breast cancer, invasive breast cancer, carcinoma in situ/mixed). Between-subgroup heterogeneity was formally tested using interaction tests.

The combined effect estimates were displayed using a forest plot. To assess the robustness of the synthesized results, a sensitivity analysis was conducted by excluding individual studies sequentially; the resulting variations in effect sizes and confidence intervals were visualized in a plot. A funnel plot was generated to visually assess publication bias, which was further quantitatively evaluated using Begg’s and Egger’s tests [23].

Results

Search results

Database searches yielded 12,102 records (Figure 1). After deduplication (n = 3,788), 8,314 records underwent title/abstract screening, leading to 44 articles for full-text review. Eight studies were excluded (2 for unspecified endpoints, 3 for thematic ineligibility, 3 for unextractable data), resulting in 34 studies eligible for inclusion in the synthesis and meta-analysis.

Figure 1.

PRISMA flow diagram of the study selection process.

Characteristics of included studies

This meta-analysis included 34 studies (10 randomized controlled trials, 9 cohort, and 15 case–control studies) involving 4,533,838 participants to evaluate the association between MHT and breast cancer risk (Table 1). Most studies were conducted in the United States (52.9%), with additional contributions from Europe (26.5%) and Asia (8.8%, including China and South Korea). The study population primarily comprised women aged 50–79 years, with sample sizes ranging from 30 to 1.27 million participants (median: 3,947; IQR: 739–43,183).

Table 1.

Basic characteristics of the included Literature.

Study	Country	Case	Sample size	Age (years)	Study design	Hormone type	Intervention (years)	Median follow-up (years)	Diagnosis
Harris (1992) [24]	US	412	1124	≥40	CCSa	ETd	3	Unknown	BCh
Stanford (1995) [25]	US	526	1029	50–64	CCS	ET, PTe, EPTf	2.5	Unknown	BC
Newcomb (1995) [26]	US	4053	6828	65–74	CCS	ET, EPT	2	Unknown	IBCi
Magnusson (1999) [27]	Sweden	2401	5096	50–74	CCS	ET, EPT	Unknown	Unknown	IBC
Ross (2000) [28]	US	1897	3534	55–72	CCS	ET, EPT	Unknown	Unknown	BC
Porch (2002) [29]	US	411	17835	≥45	CSb	ET, EPT	5.9	5.9	IBC
Kirsh (2002) [30]	Canada	404	807	20–74	CCS	ET, EPT	Unknown	Unknown	BC
Chen (2002) [31]	US	658	1397	50–74	CCS	ET, EPT	5.5	Unknown	IBC
Weiss (2002) [32]	US	2774	3823	35–64	CCS	ET, EPT	5	Unknown	IBC
Stahlberg (2004) [33]	Danish	244	10874	≥45	CS	ET, EPT, Tiboloneg	6.34	6.34	IBC
Bots (2006) [34]	US	9	866	45–79	RCTc	EPT, Tibolone	3	Unknown	BC
Swanson (2006) [35]	US	1	396	≥40	RCT	Tibolone	0.23	Unknown	BC
Borgquist (2007) [36]	Sweden	331	12504	≥42	CS	ET, EPT	Unknown	4.5	IBC
Cummings (2008) [37]	US	25	4506	60–85	RCT	Tibolone	5	2.83	IBC
Opatrny (2008) [38]	UK	5900	37863	50–75	CCS	ET, PT, EPT, Tibolone	6	Unknown	IBC
Chlebowski (2009) [21]	US	349	15397	50–79	RCT	EPT	2.6	2.4	IBC
Vieira (2009) [39]	Brazil	3	30	30–65	RCT	Tibolone	Unknown	1	BC
Chlebowski (2010) [20]	US	678	16608	50–79	RCT	EPT	5.6	7.9	IBC
Beral (2011) [40]	UK	15759	1129025	50–59	CS	ET, EPT, Tibolone	4	Unknown	IBC
LaCroix (2011) [41]	US	350	7645	50–79	RCT	ET	5.9	10.7	IBC
Crandall (2012) [42]	US	360	16608	50–79	RCT	ET, EPT	6.8	Unknown	IBC
Emilie (2013) [43]	France	669	1555	≥35	CCS	ET, EPT, Tibolone	2	Unknown	CISoMj
Chlebowski (2015) [22]	US	1565	27347	50–79	RCT	ET, EPT	7.2	13	IBC
Jones (2016) [44]	UK	775	39183	≥58	CS	ET, EPT	5.4	6	CISoM
Liu (2016) [45]	China	217	22929	≥45	CS	ET, EPT	Unknown	Unknown	BC
Salagame (2016) [46]	Australia	1223	2098	50–75	CCS	ET, EPT, Tibolone	1.5	Unknown	IBC
Salagame (2018) [47]	Australia	410	723	>50	CCS	unknown	1.5	Unknown	IBC
DeBono (2018) [48]	US	1474	2813	>40	CCS	ET, PT, EPT	8	Unknown	IBC
Wang (2020) [49]	US	6384	118760	50–71	CS	ET, EPT	10	15.1	CISoM
Prentice (2021) [50]	US	598	27347	50–79	RCT	ET, EPT	7.2	18	IBC
Abenhaim (2022) [51]	UK	43183	475013	≥50	CCS	ET	20	Unknown	IBC
Støer (2024) [52]	Norway	33654	1275783	≥45	CS	ET, EPT, Tibolone	14	12.7	IBC
Yuk (2024) [53]	Korea	17908	1246064	>40	CS	ET, Tibolone	18	Unknown	BC
Shang (2024) [54]	China	354	428	46–73	CCS	ET, PT	2.75	Unknown	IBC

^aCCS, Case-control study; ^bCS, Cohort study; ^cRCT, Randomized controlled trials; ^dET, estrogen-only therapy; ^ePT, progestin-only therapy; ^fEPT, estrogen-progestin therapy; ^gTibolone, a synthetic steroid with estrogenic, progestogenic, and androgenic properties; ^hBC, Breast Cancer; ⁱIBC, Invasive breast cancer; ^jCISoM, Carcinoma In Situ or Mixed.

EPT was the most frequently evaluated regimen (58.8% of studies). Treatment duration varied widely, from 3 months to 20 years (median: 5.4 years). The primary endpoint was invasive breast cancer (67.6% of studies), followed by nonspecific breast cancer (23.5%) and carcinoma in situ or mixed subtypes (8.8%). The median follow-up duration was 6.3 years (maximum: 18 years); however, 32.4% of studies did not report specific follow-up periods, which represents a potential source of survival bias.

Literature quality assessment

Among the 24 observational studies (Table 2), 13 were high quality and 11 were moderate; none were low. For the 10 RCTs (Table 3), 8 had a low risk of bias and 2 had an unclear risk.

Table 2.

Quality assessment of observational studies using the Newcastle-Ottawa scale (NOS).

Study	Study Design	Selection	Comparability	Outcome /Exposure	Overall Score	Quality Rating
Porch (2002)	CS	**	**	**	7	High
Stahlberg (2004)	CS	***	**	***	8	High
Borgquist (2007)	CS	****	*	*	6	Moderate
Beral (2011)	CS	****	**	**	8	High
Jones (2016)	CS	****	**	***	9	High
Liu (2016)	CS	****	*	*	6	Moderate
Wang (2020)	CS	****	**	***	9	High
Støer (2024)	CS	****	**	**	8	High
Yuk (2024)	CS	****	*	*	6	Moderate
Harris (1992)	CCS	***	**	*	6	Moderate
Stanford (1995)	CCS	***	**	*	6	Moderate
Newcomb (1995)	CCS	***	**	**	7	High
Magnusson (1999)	CCS	***	**	**	7	High
Ross (2000)	CCS	***	**	*	6	Moderate
Kirsh (2002)	CCS	***	**	**	7	High
Weiss (2002)	CCS	***	**	**	7	High
Chen (2002)	CCS	***	**	**	7	High
Opatrny (2008)	CCS	****	**	**	8	High
Emilie (2013)	CCS	**	**	**	5	Moderate
Salagame (2016)	CCS	***	**	*	6	Moderate
Salagame (2018)	CCS	***	**	*	6	Moderate
DeBono (2018)	CCS	***	**	*	6	Moderate
Abenhaim (2022)	CCS	***	**	**	7	High
Shang(2024)	CCS	***	**	*	6	Moderate

According to the NOS scoring criteria (total score 0–9), a score of ≤ 3 is considered low quality, 4–6 as moderate quality, and ≥ 7 as high quality.

Table 3.

Risk of bias assessment for randomized controlled trials using the Cochrane risk of bias tool.

Study	Selection bias	Performance bias	Detection bias	Attrition bias	Reporting bias	Other bias	Overall
Bots (2006)	Low	Low	Low	Low	Low	Unclear	Low
Swanson (2006)	Unclear	Low	Unclear	Low	Low	Unclear	Unclear
Cummings (2008)	Unclear	Low	Low	Low	Low	Unclear	Low
Vieira (2009)	Low	Low	Low	Low	Low	Unclear	Low
Chlebowski (2009)	Low	Low	Unclear	Low	Low	Unclear	Low
Chlebowski (2010)	Low	Low	Low	Low	Low	Unclear	Low
LaCroix (2011)	Low	Low	Low	Low	Low	Unclear	Low
Crandall (2012)	Low	Low	Low	Low	Low	Unclear	Low
Chlebowski (2020)	Unclear	Low	Low	Low	Low	Unclear	Low
Prentice (2021)	Unclear	Unclear	Unclear	Low	Low	Unclear	Unclear

The Cochrane Risk of Bias Tool was used for the assessment. ‘Low’ indicates a low risk of bias; ‘Unclear’ indicates that information was insufficient to permit a clear judgment or the risk level is uncertain.

Both Egger’s test (t = −0.58, p = 0.566) and Begg’s test (Z = 0.18, p = 0.859) for publication bias indicated no significant asymmetry, suggesting that there was no evidence of publication bias.

Effect values combined results

A total of 34 studies were included in this meta-analysis, and the combined results are shown as a forest plot (Figure 2). Given that I² = 92.4%, a random-effects model was used for the combined results. The random-effects model demonstrated a statistically significant increase in breast cancer risk among MHT users (OR = 1.15, 95% CI: 1.09–1.22, p < 0.001).

Figure 2.

Forest plot of the association between MHT and breast cancer risk.

Subgroup analysis

Given the presence of heterogeneity, conducting subgroup analyses is necessary. The subgroup analyses revealed a nuanced relationship between MHT and breast cancer risk, moderated by several key factors. As shown in Table 4, the intergroup heterogeneity for Hormone type (I² = 91.2%, p < 0.001) and Hormone use status (I² = 97.1%, p < 0.001) were highly significant, and Regions (I² = 92.4%, p = 0.010) was statistically significant. Geographically, the risk was most pronounced in Europe (OR = 1.30, 95% CI: 1.19–1.43, p < 0.001), with no significant associations found in North America, Asia, or other regions. The type of hormone therapy used was a critical determinant (Figure 3), a substantial increase in risk was associated with EPT (OR = 1.44, 95% CI: 1.26–1.64, p < 0.001), in contrast to ET, which showed no significant association (OR = 1.00, 95% CI: 0.91–1.10). Furthermore, current users faced a significantly higher risk (OR = 1.62, 95% CI: 1.27–2.06, p < 0.001) compared to past users, for whom no significant risk was identified. Notably, the risk was more marked for invasive breast cancer (IBC) than for breast cancer overall (OR = 1.18, 95% CI: 1.10–1.25, p < 0.001).

Table 4.

Subgroup analysis of the relationship between MHT and breast cancer risk.

Subgroup	N	Odds Ratio (95% CI)	Heterogeneity (I², %)	p
Study type			92.4	0.071
Case-control study	15	1.11(1.02,1.20)	84.0	<0.001
Cohort study	9	1.28(1.14,1.44)	97.4	<0.001
Randomized controlled trials	10	1.07(0.93,1.23)	66.1	0.002
Regions			92.4	0.010*
North America	19	1.06(0.97,1.15)	80.1	<0.001
Europe	9	1.30(1.19,1.43)	96.0	<0.001
Asia	3	1.11(0.73,1.67)	88.1	<0.001
other	3	1.09(0.94,1.26)	0	0.907
Sample size			92.4	0.131
<100000	29	1.12(1.03,1.22)	83.2	<0.001
≥100000	5	1.25(1.12,1.39)	98.4	<0.001
Age (years)			92.4	0.313
<50	16	1.21(1.06,1.37)	95.2	<0.001
≥50	18	1.13(1.06,1.19)	84.9	<0.001
Median follow-up (years)			90.4	0.115
<10	8	1.31(1.04,1.66)	85.2	<0.001
≥10	5	1.06(0.94,1.20)	93.1	<0.001
Intervention (years)			92.6	0.743
<5	11	1.10(0.96,1.27)	84.7	<0.001
≥5	17	1.13(1.06,1.22)	94.4	<0.001
Hormone type			91.2	<0.001***
ETa	24	1.00(0.91,1.10)	88.7	<0.001
PTb	4	1.31(0.93,1.84)	81.8	<0.001
EPTc	22	1.44(1.26,1.64)	92.2	<0.001
Tiboloned	9	1.19(0.86,1.65)	67.0	0.002
Hormone use status			97.1	<0.001***
Current	5	1.62(1.27,2.06)	85.6	<0.001
Past	5	0.97(0.85,1.12)	64.5	0.024
Cancer subtype			92.4	0.305
BCe	9	0.98(0.79,1.22)	86.1	<0.001
IBCf	22	1.18(1.10,1.25)	92.0	<0.001
CISoMg	3	1.17(1.01,1.35)	72.8	0.025

^aET, estrogen-only therapy; ^bPT, progestin-only therapy; ^cEPT, estrogen-progestin therapy; ^dTibolone, a synthetic steroid with estrogenic, progestogenic, and androgenic properties; ^eBC, Breast Cancer; ^fIBC, Invasive breast cancer; ^gCISoM, Carcinoma In Situ or Mixed. ***p < 0.001, intergroup heterogeneity is highly significant; *p < 0.05, intergroup heterogeneity was statistically significant.

Figure 3.

Subgroup analysis of breast cancer risk associated with MHT, by hormone type.

Stratified analysis

Given the substantial statistical heterogeneity observed in the overall pooled estimate (I² = 92.4%), which limits its clinical interpretability. As indicated in the subgroup analysis (Table 4, Figure 3), hormone type was a critical determinant of risk, with significant between-group heterogeneity (p < 0.001). Therefore, we prioritized a stratified analysis based on the pre-specified key effect modifier: hormone type (ET vs. EPT).

The association between EPT and breast cancer risk was analyzed using data from 22 studies (Figure 4). The forest plot presented that the pooled odds ratio (OR) was 1.44 (95% CI: 1.26–1.64, p < 0.001) using a random-effects model, indicating a statistically significant increased risk associated with EPT. Considerable heterogeneity was observed among studies (I² = 92.2%, p < 0.001). The pre-specified subgroup analysis (Table 5) revealed that the association between EPT use and breast cancer risk was significantly modified by Study type (I² = 92.2%, p = 0.044), Regions (I² = 92.2%, p = 0.007) and Sample size (I² = 92.2%, p = 0.028). The strongest associations were observed in cohort study (OR = 2.05, 95% CI: 1.40–3.01, p < 0.001) and Europe (OR = 2.02, 95% CI: 1.32–3.09, p < 0.001). A larger sample size (≥ 10,000) was also associated with a higher pooled risk estimate (OR = 1.60, 95% CI: 1.34–1.91, p < 0.001). Moreover, while short-term EPT use (< 5 years) was not significantly associated with breast cancer risk (OR = 1.16, 95% CI: 0.93–1.44), long-term use (≥ 5 years) showed a significant positive association (OR = 1.48, 95% CI: 1.25–1.76, p < 0.001). Subgroup analysis by Cancer subtype suggested a potential difference in risk estimates, although the test for subgroup differences did not reach formal statistical significance (p = 0.051). The association appeared strongest and most precise for invasive breast cancer (IBC) (OR = 1.42, 95% CI: 1.25–1.62, p < 0.001).

Figure 4.

Forest plot of the association between EPT and breast cancer risk.

Table 5.

Subgroup analysis of the relationship between EPT and breast cancer risk.

Subgroup	N	Odds Ratio (95% CI)	Heterogeneity (I², %)	p
Study type			92.2	0.044*
Case-control study	10	1.23(1.08,1.39)	77.8	<0.001
Cohort study	6	2.05(1.40,3.01)	96.0	<0.001
Randomized controlled trials	6	1.27(1.17,1.37)	0	0.831
Regions			92.2	0.007**
North America	14	1.23(1.11,1.37)	79.0	<0.001
Europe	6	2.02(1.32,3.09)	97.3	<0.001
Asia	1	1.12(0.83,1.52)	0	<0.001
other	1	2.39(1.47,3.87)	0	0.907
Sample size			92.2	0.028*
<10000	10	1.22(1.04,1.44)	78.3	<0.001
≥10000	12	1.60(1.34,1.91)	94.5	<0.001
Age (years)			92.2	0.828
<50	10	1.46(1.14,1.88)	90.0	<0.001
≥50	12	1.41(1.20,1.67)	93.8	<0.001
Median follow-up (years)			95.1	0.130
<10	6	2.01(1.29,3.12)	96.1	<0.001
≥10	3	1.38(1.15,1.66)	84.9	0.001
Intervention (years)			92.5	0.082
<5	6	1.16(0.93,1.44)	67.9	0.008
≥5	17	1.48(1.25,1.76)	94.4	<0.001
Cancer subtype			92.2	0.051
BCa	5	1.13(0.97,1.33)	17.2	0.305
IBCb	14	1.42(1.25,1.62)	87.1	<0.001
CISoMc	3	1.88(0.98,3.60)	98.1	<0.001

^aBC, Breast Cancer; ^bIBC, Invasive breast cancer; ^cCISoM, Carcinoma In Situ or Mixed. **p < 0.01, intergroup heterogeneity is very significant; *p < 0.05, intergroup heterogeneity was statistically significant.

Although the combined results (Figure 5) of 24 studies about the relationship between ET and breast cancer risk showed no significant association (OR = 1.00, 95% CI: 0.91–1.10), the forest plot of the subgroup analysis (Figure 6) indicated a potential protective effect (OR = 0.78, 95% CI: 0.70–0.87; I² = 0, p = 0.919) in randomized controlled trials with significant effect modification by Study type (I² = 88.1%, p < 0.001). Additionally, the subgroup analysis (Table 6) also showed significant effect modification for Regions (between-subgroup p < 0.001). Studies from North America suggested a possible risk reduction (OR = 0.87, 95% CI: 0.80–0.94), whereas those from Europe indicated an increased risk (OR = 1.24, 95% CI: 1.07–1.44).

Figure 5.

Forest plot of the association between ET and breast cancer risk.

Figure 6.

Subgroup analysis of breast cancer risk associated with ET, by study type.

Table 6.

Subgroup analysis of the relationship between ET and breast cancer risk.

Subgroup	N	Odds Ration (95% CI)	Heterogeneity (I², %)	p
Study type			88.1	<0.001***
Case-control study	13	1.01(0.90,1.14)	88.7	<0.001
Cohort study	7	1.16(0.96,1.41)	83.6	<0.001
Randomized controlled trials	4	0.78(0.70,0.87)	0	0.919
Regions			88.1	<0.001***
North America	13	0.87(0.80,0.94)	59.4	0.003
Europe	7	1.24(1.07,1.44)	80.7	<0.001
Asia	3	1.04(0.69,1.57)	70.4	0.034
other	1	1.62(1.12,2.34)	0	<0.001
Sample size			88.1	0.900
<10000	13	1.00(0.87,1.15)	80.2	<0.001
≥10000	11	1.01(0.91,1.13)	88.1	<0.001
Age (years)			88.1	0.612
<50	12	1.04(0.85,1.26)	88.1	<0.001
≥50	12	0.98(0.87,1.09)	88.9	<0.001
Median follow-up (years)			79.1	0.129
<10	4	1.21(0.80,1.81)	81.6	<0.001
≥10	4	0.86(0.72,1.02)	73.2	0.011
Intervention (years)			89.9	0.774
<5	6	0.99(0.83,1.19)	61.8	0.023
≥5	13	0.96(0.86,1.08)	91.6	<0.001
Cancer subtype			88.1	0.310
BCa	6	1.08(0.90,1.29)	76.0	<0.001
IBCb	15	0.95(0.83,1.08)	90.5	<0.001
CISoMc	3	1.16(0.88,1.53)	79.9	0.007

^aBC, Breast Cancer; ^bIBC, Invasive breast cancer; ^cCISoM, Carcinoma In Situ or Mixed. ***p < 0.001, intergroup heterogeneity is highly significant.

Sensitivity analysis

To assess the robustness of the pooled results between MHT and breast cancer risk, we performed a leave-one-out sensitivity analysis (Figure 7 A). This analysis indicated that no single study disproportionately influenced the overall estimate; the recalculated pooled odds ratios ranged from 1.13 to 1.17 (Table 7), all within the 95% confidence interval of the primary result (OR = 1.15, 95% CI: 1.09–1.22). These findings suggest that the meta-analysis results are robust and not driven by any individual study.

Figure 7.

Leave-one-out sensitivity analysis. A, Leave-one-out sensitivity analysis of breast cancer risk associated with MHT; B, Leave-one-out sensitivity analysis of breast cancer risk associated with EPT; C, Leave-one-out sensitivity analysis of breast cancer risk associated with ET in randomized controlled trials. The plot displays the pooled odds ratio (central vertical line) and its 95% confidence interval (horizontal ‘I-beams’) recalculated after sequentially removing each study (labeled on the left). The consistency across all rows indicates the robustness of the main result.

Table 7.

Leave-one-out sensitivity analysis of the association between MHT and breast cancer risk.

Exclusionary research	Odds ratio (95% CI)	I2 (%)	p
Harris (1992)	1.15 (1.09, 1.22)	92.60	<0.001
Stanford (1995)	1.16 (1.10, 1.23)	92.50	<0.001
Newcomb (1995)	1.16 (1.10, 1.23)	92.10	<0.001
Magnusson (1999)	1.14 (1.07, 1.20)	92.10	<0.001
Ross (2000)	1.16 (1.09, 1.23)	92.60	<0.001
Chen (2002)	1.15 (1.08, 1.22)	92.60	<0.001
Kirsh (2002)	1.15 (1.09, 1.22)	92.60	<0.001
Porch (2002)	1.16 (1.09, 1.22)	92.60	<0.001
Weiss (2002)	1.16 (1.10, 1.23)	92.40	<0.001
Stahlberg (2004)	1.14 (1.07, 1.21)	92.40	<0.001
Bots (2006)	1.15 (1.09, 1.22)	92.60	<0.001
Swanson (2006)	1.15 (1.09, 1.22)	92.60	<0.001
Borgquist (2007)	1.13 (1.07, 1.20)	92.10	<0.001
Cummings (2008)	1.16 (1.09, 1.23)	92.50	<0.001
Vieira (2009)	1.15 (1.09, 1.22)	92.60	<0.001
Chlebowski (2009)	1.15 (1.08, 1.22)	92.60	<0.001
Chlebowski (2010)	1.15 (1.08, 1.22)	92.60	<0.001
LaCroix (2011)	1.17 (1.10, 1.24)	92.30	<0.001
Beral (2011)	1.14 (1.08, 1.22)	91.80	<0.001
Crandall (2012)	1.15 (1.08, 1.22)	92.60	<0.001
Emilie (2013)	1.16 (1.09, 1.23)	92.60	<0.001
Chlebowski (2015)	1.16 (1.09, 1.23)	92.60	<0.001
Jones (2016)	1.15 (1.09, 1.22)	92.60	<0.001
Liu (2016)	1.16 (1.10, 1.23)	92.40	<0.001
Salagame (2016)	1.16 (1.09, 1.23)	92.60	<0.001
Salagame (2018)	1.15 (1.09, 1.22)	92.40	<0.001
DeBono (2018)	1.16 (1.10, 1.23)	92.40	<0.001
Wang (2020)	1.15 (1.08, 1.22)	92.40	<0.001
Prentice (2021)	1.16 (1.09, 1.23)	92.60	<0.001
Abenhaim (2022)	1.15 (1.08, 1.23)	92.60	<0.001
Stoer (2024)	1.16 (1.09, 1.23)	89.10	<0.001
Yuk (2024)	1.14 (1.08, 1.21)	89.60	<0.001
Shang (2024)	1.15 (1.08, 1.22)	92.60	<0.001

Each row represents the pooled OR and 95% CI calculated using a random-effects model after excluding the specified study.

Sensitivity analyses were also conducted for the stratified analysis. The leave-one-out sensitivity analyses (Figure 7 B) indicated that the conclusion regarding ‘EPT increases the risk of breast cancer’ is highly robust. Upon sequentially excluding any individual study, the recalculated ORs ranged from 1.36 to 1.46, with all 95% CIs excluding 1 (all p < 0.001). No single study substantially altered the pooled estimate, indicating that the meta-analysis result is robust. Visual inspection of the funnel plot (Fig S1) identified a notable outlier (Bots, 2006). After excluding this study, the results (Fig S2) remained robust, with a consistent and significant pooled estimate (OR = 1.44, 95% CI: 1.27–1.65, p < 0.001). The leave-one-out sensitivity analysis (Fig S3) confirmed the robustness of the null association between ET and breast cancer risk. The recalculated pooled odds ratios remained near null (range: 0.97–1.02), with all 95% confidence intervals including 1. No individual study exerted a disproportionate influence on the overall result. Given the protective effect of ET against breast cancer observed in randomized controlled trials (OR = 0.78, 95% CI: 0.70–0.87, p < 0.001; Figure 6), we performed a leave-one-out sensitivity analysis (Figure 7 C) on this subgroup to assess the robustness of the finding. After sequentially excluding each of the four trials, the recalculated pooled odds ratios remained consistently below 1 (range: 0.76–0.79). All corresponding 95% confidence intervals also remained below 1, indicating that the significant protective association was not driven by any single study.

Discussion

MHT, is commonly used to alleviate menopausal symptoms, prevent osteoporosis, and improve quality of life. The 2009–2010 NHANES reported that approximately 5% of U.S. women aged 40 years or older used oral estrogen therapy (estrogen-only or estrogen-progestin) [55,56]. However, after the Women’s Health Initiative linked combined MHT to increased risks of cardiovascular disease and breast cancer, MHT use declined sharply in 2002, accompanied by a observed decrease in breast cancer incidence in the U.S [57–59]. Although some countries have reported elevated breast cancer risk with MHT, this finding is not consistent worldwide, and the association remains internationally contentious [60,61].

Our meta-analysis of 34 studies (Figure 2) demonstrated that MHT significantly increases the risk of breast cancer by 1.15-fold (OR = 1.15, 95% CI: 1.09–1.22, p < 0.001). The established role of estrogen and progesterone in both mammary gland development and breast cancer pathogenesis, as noted by James et al. [62], along with the identified concern regarding breast cancer risk as a major barrier to MHT use (Kingsberg et al. [63], is strongly corroborated by the findings of the our study. Collaborative Group on Hormonal Factors in Breast Cancer pooled data from 58 studies (n = 570,000 women) to assess hormonal risk factors for breast cancer [64]. They found that all systemic MHT types, except vaginal estrogen, significantly increase breast cancer risk. The strongest association was observed in current users of combined estrogen-progestogen therapy. Furthermore, the risk demonstrated a positive correlation with the duration of MHT use. This finding by the Collaborative Group is consistent with our subgroup analysis.

Accumulating evidence indicates that long-term use or specific formulations of MHT may increase breast cancer risk. Notably, EPT was significantly associated with breast cancer risk [21,51,52,65], whereas ET has been linked to a neutral or potentially protective effect [12,21]. The Women’s Health Initiative (WHI) randomized trial reported a 23% higher hazard ratio for breast cancer in the EPT group (HR = 1.23, 95%CI:1.03, 1.47, p = 0.020), corresponding to an absolute risk increase of 8 cases per 10,000 person-years. In contrast, the ET group showed no significant increase in risk (HR = 0.77, 95% CI: 0.59, 1.01), a finding that might suggest a potential risk reduction, though possibly influenced by selective bias [21,66]. Jordan (2020) [67] advanced a key mechanistic hypothesis for the protective effect of ET in the WHI, positing that the trial’s design—enrolling women years after menopause—inadvertently created a state of long-term estrogen deprivation (LTED) in participants. Breast cancer cells adapt to this LTED state. Subsequent administration of estrogen (ET) does not stimulate growth but instead triggers severe endoplasmic reticulum stress and activates the mitochondrial apoptosis pathway, leading to tumor cell death. This mechanism explains the sustained decrease in breast cancer incidence and mortality seen in the WHI estrogen-alone trial. Supporting this, the EPIC cohort study found a substantially elevated relative risk among current EPT users (RR = 1.69, 95% CI: 1.50, 1.91), compared to a more modest increase for ET users (RR = 1.20, 95%: CI 1.01, 1.43) [68]. These findings are consistent with the results of our meta-analysis (Figure 4, Figure 5, Figure 6). Importantly, the elevated risk appears to decline following the discontinuation of therapy [12,69]. Furthermore, the risk profile may vary by progestogen type, with some evidence suggesting that micronized progesterone is associated with a lower risk compared to synthetic analogs [70–72]. The specific MHT regimen is a critical determinant of cancer risk. Among progestogens, synthetic progestins such as medroxyprogesterone acetate and norethisterone—the latter widely used in Europe—are associated with a significantly elevated risk of breast cancer [73,74]. Consequently, an accurate assessment requires careful consideration of regimen characteristics—specifically, the use of estrogen alone or combined with a progestogen, and the progestogen type—alongside individual patient factors.

Our analysis reveals significant geographical heterogeneity in the association between MHT and breast cancer risk. The strongest association between EPT and breast cancer risk was observed in Europe (OR = 2.05, 95% CI: 1.32–3.09, Table 5), and an increased risk for ET was also found in European studies (OR = 1.24, 95% CI: 1.07–1.44, Table 6). In contrast, studies from North America suggested a possible risk reduction for ET (OR = 0.87, 95% CI: 0.80–0.94, Table 6). This divergence likely reflects key differences in MHT prescribing patterns and healthcare practices across regions. The higher breast cancer risk observed in European studies may be attributable to regional variations in MHT regimens, such as the more frequent use of certain progestin types (e.g. norethisterone) that have been associated with elevated breast cancer risk [75,76], as well as differences in treatment duration, adherence, and concomitant medications. Additionally, variations in healthcare systems and screening practices may influence breast cancer detection rates, potentially contributing to the observed heterogeneity. It is also noteworthy that the distribution of study designs differs between regions, with a higher proportion of randomized controlled trials in North America and more observational studies in Europe; this methodological difference may partially explain the divergent effect estimates. Future research should prioritize direct comparisons of prescription patterns, progestin types, and treatment durations across regions, ideally using harmonized individual participant data, to better elucidate the factors driving geographical heterogeneity in MHT-associated breast cancer risk.

Regarding the issue of heterogeneity (I²) in our study, it is widely recognized that meta-analysis involves the synthesis of multiple independent studies. Heterogeneity typically stems from variations in study design and methodological quality. However, the I² statistic should not be regarded as unequivocal: the extent of heterogeneity should not be determined exclusively based on conventional thresholds for I² values (e.g. 25% = low, 50% = moderate, 75% = high). A meta-analysis with an I² of 25% may demonstrate greater dispersion in effect sizes than one with an I² of 75% [77]. Our study employed a random-effects model. The meta-analysis exhibited substantial between-study heterogeneity (I²= 92.4%). Subgroup analysis for overall MHT use identified hormone type as a major source of this heterogeneity (P for interaction < 0.001), with current use status (p < 0.001) and region (p = 0.010) also contributing significantly. In the stratified analysis for EPT, significant between-subgroup differences were observed by study type (p = 0.044), region (p = 0.007), and sample size (p = 0.028). Similarly, for ET, the heterogeneity was primarily explained by study type (p < 0.001) and region (p < 0.001). In summary, our subgroup analysis successfully quantified the substantial contributions of clinical and methodological factors to the overall heterogeneity observed. The observed heterogeneity related to sample size may partly reflect underlying differences in study design. In our subgroup analyses, study design significantly modified the association (P for interaction = 0.044 for EPT and < 0.001 for ET; Table 5, Table 6), with randomized controlled trials often having smaller sample sizes and yielding different effect estimates compared to larger observational studies. This suggests that heterogeneity by sample size may be confounded by study design, underscoring the importance of considering methodological differences when interpreting meta-analytic findings.

Our inclusion criteria focused on women typically aged ≥ 40 years or in the menopausal transition/postmenopause. Therefore, our findings may not be generalizable to younger premenopausal women (e.g. < 40 years) who receive systemic hormone therapy for other indications, such as premature ovarian insufficiency or gender-affirming care. Additionally, while our subgroup analyses explored broad categories such as hormone type (ET, PT, EPT, Tibolone) and user status, the available literature lacked the granular data necessary to perform more refined stratifications. Specifically, we were unable to analyze the potential differential effects of specific estrogen formulations (e.g. estradiol vs. conjugated equine estrogens), specific progestins (e.g. medroxyprogesterone acetate vs. norethisterone), or routes of administration (oral vs. transdermal). Furthermore, data on the precise time since last hormone use were insufficient for a meaningful subgroup analysis. The inability to account for these nuances may obscure important variations in risk profiles associated with different MHT regimens.

Conclusion

A modest but significant overall increase in breast cancer risk associated with MHT, which is mainly driven by EPT, current user and in Europe. While the risk was not observed with ET in observational studies, a protective effect was suggested by randomized trials. Critically, the association exhibits marked geographical heterogeneity, underscoring that the risk profile of MHT is not uniform but is substantially modified by the specific hormone regimen used and regional factors.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethics statement

This article did not include studies with human participants performed by any of the authors.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.