High hypoxia inducible factor-1α expression is associated with reduced survival in patients with breast cancer: A meta-analysis

Abstract

BACKGROUND

Hypoxia-inducible factor 1α (HIF-1α) plays a crucial role in the prognosis of breast cancer, but the current evidence remains inconclusive.

AIM

To provide comprehensive evidence about the correlation of altered HIF-1α expression with overall survival (OS) and disease-free survival (DFS) in breast cancer patients.

METHODS

A systematic search was conducted in PubMed, Embase, and Web of Science databases to collect relevant articles that were published before April 8, 2024. A meta-analysis was used to assess the impact of altered HIF-1α expression on the OS and DFS of breast cancer patients. Subgroup and sensitivity analyses were also performed in this meta-analysis.

RESULTS

This meta-analysis included 40 studies. The average percentage of breast cancer patients with high HIF-1α expression was 39.6%. The overall meta-analysis results demonstrated that high HIF-1α expression is strongly linked to poor outcomes in patients of breast cancer. Compared with low HIF-1α expression, the overall hazard ratio for OS in patients with high HIF-1α expression was 1.47 [95% confidence interval (CI): 1.29-1.69], and the overall hazard ratio for DFS was 1.82 (95%CI: 1.56-2.12). Furthermore, both OS [1.18 (95%CI: 1.01-1.38)] and DFS [1.79 (95%CI: 1.03-3.11)] were markedly shorter in triple-negative breast cancer cases with high HIF-1α expression. Subgroup analysis revealed that the antibody used to detect HIF-1α expression affected only the correlation linking HIF-1α expression to DFS in breast cancer patients (P = 0.0004). Furthermore, the sensitivity analysis demonstrates that the overall conclusions of the meta-analysis were unaffected by the removal of individual studies.

CONCLUSION

Compared to patients with low HIF-1α expression, those with high expression level had shorter OS and DFS. However, the prognostic significance of high HIF-1α expression varies across molecularly stratified breast cancer cohorts needs to be further elucidated.

Key Words: Breast cancer; Hypoxia-inducible factor-1α; Prognosis; Hypoxia; Systematic review; Meta-analysis

Core Tip: Hypoxia is a crucial characteristic of tumors that can influence the phenotype of cancer stem cells and contribute to cancer cell invasion and metastasis, thereby significantly impacting the prognosis of cancer patients. This meta-analysis offers a more exhaustive analysis of the relationship between high hypoxia-inducible factor 1α expression and the prognosis of breast cancer patients, providing clinicians with systematic evidence about the role of high hypoxia-inducible factor 1α expression in predicting breast cancer prognosis.

INTRODUCTION

According to 2022 Global Cancer Statistics, it is estimated that breast cancer accounted for 2.3 million incident cases, establishing it as the foremost oncological disease affecting women worldwide[1]. In China, approximately 357000 new breast cancer cases were reported among women in 2022, ranking second among all new cancer cases[1]. In the United States, it is estimated that there were 313000 new breast cancer cases in 2024, and breast cancer ranked first among all new cancer cases in American women[2]. Despite considerable progress in early diagnosis, comprehensive treatment strategies, and improvements in survival rates of breast cancer patients, more than 680000 women die each year from the recurrence and metastasis of breast cancer, which is the second highest contributor to cancer-related fatalities in females[1]. Therefore, an accurate understanding of the biology and behavior of breast cancer is crucial for making treatment decisions and prognostic predictions about breast cancer[3].

Hypoxia is a critical characteristic of tumors, and it influences the phenotype of tumor stem cells and participates in the invasion and metastasis of cancer cells[4]. Under hypoxic conditions, the expression of hypoxia-inducible factor 1α (HIF-1α) is increased in cancer cells. Subsequently, HIF-1α maintains the stemness of cancer cells through multiple mechanisms, affecting angiogenesis, epithelial-mesenchymal transition, extracellular matrix remodeling, and immune evasion, ultimately facilitating breast cancer invasion and metastasis[5-8]. Evidence from systematic reviews indicates a strong positive linking HIF-1α expression levels to lymph node metastasis and histological grade in breast cancer[9]. Furthermore, the risk of recurrence and death in breast cancer patients exhibiting high HIF-1α expression is approximately double that of patients with low HIF-1α expression. Specifically, breast cancer patients with high HIF-1α expression have shorter overall survival (OS), disease-free survival (DFS), distant metastasis-free survival, and recurrence-free survival (RFS)[9-11].

Although many clinical prognostic studies have provided evidence supporting the connection between high expression of HIF-1α and poor prognosis in individuals with breast cancer, definitive evidence about the role of HIF-1α expression in predicting the prognosis of breast cancer and guiding treatment decisions is still lacking. Although previous meta-analyses systematically evaluated and quantitatively analyzed published prognostic studies, they failed to examine all relevant studies through a more comprehensive search[12-16]. Moreover, in recent years, the publication of recent studies requires a re-evaluation through a systematic review to analyze the impact of HIF-1α expression on breast cancer prognosis[17-22]. Therefore, this study aimed to execute a systematic review together with meta-analysis, striving to evaluate more accurately how the expression of HIF-1α influences the prognostic significance regarding OS and DFS among individuals with breast cancer. This approach is intended to provide high-level clinical evidence for more precise prediction of prognosis and treatment-related decision-making for patients with breast cancer.

MATERIALS AND METHODS

The quantitative evidence synthesis was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, which provided the methodological framework to ensure reporting standards.

Literature search and inclusion criteria

Scholarly resource identification spanned three principal knowledge bases: PubMed, Embase, and Web of Science. A variety of combinations of the following keywords were used to search titles, abstracts, or full texts: “Breast cancer or Human Mammary Carcinoma”, “Cell Hypoxia or hypoxic or Hypoxia-Inducible Factor 1α”, and “Survival or outcome or prognosis or Recurrence”. The search covered publications up to April 8, 2024. Additionally, manual searches and the reference backtracking method were performed to identify further relevant literature.

The retrieved research studies satisfied the inclusion criteria detailed below: (1) A pathological diagnosis of human breast cancer; (2) The detection of HIF-1α expression in breast cancer tissues through immunohistochemistry or immunofluorescence; (3) Study endpoints of DFS/RFS and OS/breast cancer-specific survival (BCSS); (4) The provision of prognostic effect indicators hazard ratio (HR)/logHR and 95% confidence interval (CI)/standard error (SE) or raw data for HIF-1α expression and OS and DFS in breast cancer; and (5) Articles written in either Chinese or English.

Two researchers (Xue-Di Zheng and Huan-Yu Li) initially reviewed the titles and abstracts of the candidate studies independently, eliminating those that did not meet the inclusion criteria. Subsequently, the two researchers meticulously read the full texts of the remaining candidates to determine their eligibility for inclusion based on the established criteria. In cases of uncertainty about a study’s inclusion or a disagreement between the researchers, the study designer (Jiang-Bo Liu) was consulted for review and discussion until a consensus was reached. Additionally, if multiple studies were conducted on the same population after a comprehensive evaluation, the study selected for the meta-analysis was the one with either the largest sample size or the most recent publication.

Data extraction and quality assessment

Two researchers (Xue-Di Zheng and Huan-Yu Li) separately extracted the following data and information from every study that satisfied the inclusion criteria in accordance with a pre-prepared form: First author, publication year, study location, recruitment period, analytical method, study design, sample size, average age, tissue type, molecular subtype, antibodies used for HIF-1α detection, HIF-1α expression cutoff value, study endpoint, HR and 95%CI, and follow-up time.

The endpoints of this systematic review and meta-analysis focused on investigating the connections linking HIF-1α to the prognosis of breast cancer patients, were DFS/RFS and OS/BCSS. These endpoints were defined as follows: DFS/RFS indicated the period from the commencement of treatment (including surgery) to the recurrence or metastasis of the disease, and OS/BCSS represented the duration from the time of diagnosis and surgical intervention to the patient’s death due to any cause/breast cancer metastasis.

Statistical analysis

Review Manager 5.4 was utilized to conduct the meta-analysis. Data on the prognostic analysis of HIF-1α expression were extracted. If the included studies reported HRs and 95%CIs, the generic inverse variance method was used in Review Manager to calculate the logHR and SE for each study for meta-analysis. If studies did not directly report HRs and CIs, statistical data were obtained using methods such as survival curve data extraction and estimation of study endpoint events to calculate the logHR and SE for each study[23]. This meta-analysis primarily combined the multivariable Cox regression effect sizes (HR and CI) from each study; however, when studies reported only univariate HR and CI, they were included in the pooled analysis as univariate effect sizes. Finally, a pooled HR greater than 1 indicated a higher risk of recurrence, metastasis, or death and poorer prognosis for patients of breast cancer with high HIF-1α expression. If the 95%CI did not overlap with 1, it was considered significant according to statistical analysis (P value < 0.05).

The I² statistic was utilized to evaluate heterogeneity among the studies selected for the meta-analysis, which quantifies variability on a scale from 0% to 100%. Studies with an I² value below 50% or a P value exceeding 0.05 were considered to show no or low heterogeneity, justifying the use of a fixed-effects model for meta-analysis. In cases where these thresholds were not met, a random-effects model was employed. To further investigate potential sources of heterogeneity and examine the influence of specific variables, subgroup analyses were systematically performed.

Furthermore, only studies with large sample sizes (≥ 200) were included to investigate whether population sampling affected the sensitivity of this meta-analysis. Funnel plots and Egger’s test were used to investigate publication bias in the included studies. If funnel plot asymmetry was observed or the P value from Egger’s test was less than 0.05, publication bias was present[24]. To address possible publication bias, a Duval and Tweedie[25] trim-and-fill analysis was applied for both detection and adjustment in the meta-analysis. The analytical framework implemented two-tailed verification protocols, establishing 0.05 as the critical significance benchmark.

RESULTS

The initial literature search yielded 5864 articles. Through a rigorous screening process based on the established inclusion criteria, 40 eligible studies involving 7738 breast cancer patients were ultimately included in this systematic review and meta-analysis. Based on the Oxford Evidence levels, this provides level 2b evidence regarding prognosis. The process and results of the literature screening are presented in Figure 1.

Figure 1 Flow chart of literature screening for systematic review and meta-analysis.

Description of eligible studies

Detailed characteristics of the included studies are summarized Table 1. Among the 40 studies, 23[12-14,16,20,21,26-41] included European populations, 16[15,18,19,22,42-53] included Asian populations, and 1[17] included an African population. Twenty-seven studies[13,16,17,19,20,22,28-32,35-45,47,48,52-54] reported histological grading, including 2442 cases of G1-G2 disease and 1837 cases of G3 disease. Thirteen studies[15,18,20,22,26,38,41-43,45,48,50,51] reported clinical tumor-node-metastasis staging, including 1536 patients with stage I-II disease and 809 with stage III-IV. Seven studies[12,14,21,27,33,34,46] did not report breast cancer staging or grading. Among the 40 studies, 38[14-22,26-54] focused on female breast cancer patients (n = 7326), whereas 2[12,13] focused on male breast cancer patients (n = 412).

Table 1 Characteristics of included studies on prognostic impact of hypoxia-inducible factor 1α on survival in breast carcinoma patients.

Ref.	Recruitment period	Study design	Sample size	Age median mean year (range)	Histological grading, and TNM staging	Percentage of grading and staging	Molecular typing	N of HIFα positive (%), cut-off, and antibody	Survival endpoint (months)	Hazard ratio, 95%CI and P value	HR estimation	Follow-up months (range)
El-Guindy et al[17], 2023, Egypt	June 2019-June 2022	Prospective	60	47.5	G2: 24, G3: 36; NA	G1/G2: 40%; I/II: NA	TNBC	28 (46.6); ≥ 10%; EP1215Y	OS	(M) 7.56 (1.21-47.24), 0.03	Reported	24
Jögi et al[26], 2019, Sweden	1977-2007	NA	634 of 688 (BC2)	NA	NA; I: 273, II: 1 59, III: 83, Unknow: 119	G1/G2: NA; I/II: 82.6%	NA	111 (17.3); ≥ 1%; BD610959	OS	(U) 1.6 (1.2-2.2), 0.001	Reported	NA
										(M) 1.6 (1.0-2.5), 0.03
Laurinaviciu et al[31], 2015, Lithuania	1977-2007	NA	107	NA	G1: 25, G2: 51, G3: 31	G1/G2: 71%; I/II: NA	Luminal A: 60, luminal B: 29, HER2 +: 28	NA; score 14; EP1215Y	OS	(M) 0.23 (0.08-0.62), 0.002	Reported	84
Li et al[48], 2016, China	2005-2009	NA	156	50	G1: 32, G2: 78, G3: 46; I: 23, II: 60, III: 48, IV: 25	G1/G2: 70%; I/II: 53%	NA	83 (53.3); score 6; ab82832	OS	(U) 2.07 (1.58-2.71), < 0.001	(U) Survival curve (M) reported	60
										(M) 2.37 (1.09-5.15), 0.029
Zhuang et al[22], 2024, China	May 2014-August 2016	NA	197	NA	G1-G2: 130, G3: 57; I-II: 147, III- IV: 50	G1/G2: 69%; I/II: 75%	Luminal A: 37, luminal B: 133, HER2 +: 13, TNBC: 14	121 (61.4); score 3; ab51608	OS	(U) 5.36 (2.02-14.22), 0.001	Reported	NA
										(M) 4.94 (1.86-13.14), 0.001
Rajković-Molek et al[38], 2014, Croatia	2000-2004	Retrospective	208	65	G1: 46, G2: 113, G3: 49; I: 67, II: 68, III: 60, IV: 3, Unknown: 10	G1/G2: 76%; I/II: 68%	Luminal A: 111, luminal B: 46, HER2 +: 19, TNBC: 32	82 (39.4); ≥ 10 %; NB100-131	OS	(U) 1.63 (1.03-2.60), 0.0369	Reported	NA
Malfettone et al[39], 2012, Italy	NA	NA	187	50	G1: 31, G2: 87, G3: 69; NA	G1/G2: 63%; I/II: NA	NA	58 (31.0); ≥ 1%; H206	OS	(U) 1.52 (1.10-2.11), 0.011	Estimate	NA
Ni et al[42], 2013, China	January 2005-October 2006	NA	85 of 95	NA	G1: 16, G2: 35, G3: 24; I-II: 38, III: 37	G1/G2: 68%; I/II: 51%	NA	52 (61.2); score ≥ 1; NA	OS	(U) 1.56 (1.01-2.42), 0.045; (M) 2.25 [1.38-6.45], 0.037	U: Survival curve M: Reported	60
Peurala et al[54], 2012, Finland	2002-2005	NA	102	59	G1: 30, G2: 35, G3: 37; NA	G1/G2: 63%; I/II: NA	NA	27 (26.4); score ≥ 2; NA	BCSS	(U) 2.4 (0.75-7.5), 0.126	Reported	NA
Kornegoor et al[13], 2012, Netherlands	1986-2010	NA	MBC; 126	66	G1-G2: 81, G3: 44; NA	G1/G2: 50%; I/II: NA	NA	51 (40.5); ≥ 5%; NA	OS	(M) 2.50 (1.1-5.6), 0.029	Reported	NA
Deb et al[12], 2014, Australia	1990-2007	Retrospective	MBC; 286	70	NA	NA; NA	NA	68 (23.8); score 6; NA	OS	(U) 3.8 (1.5-9.8), 0.006	Reported	96
Dales et al[34], 2005, France	1986-1995	Retrospective	745	56.1	NA	NA; NA	NA	543 (72.9); > 10 %; H206	OS; DFS	OS: (U) 1.21 (1.03-1.43), 0.019	U: Survival curve	162
										(M) 1.20 (1.02-1.41), 0.03	M: Estimate
										DFS: (M) 1.12 (0.96-1.32), 0.158
Kronblad et al[37], 2006, Sweden	1984-1991	Prospective	377	NA	G1: 41, G2: 142, G3: 183, Unknow: 12; NA	NA; NA	NA	91 (24.1); > 2%; NB100-123H2	OS; RFS	OS: (U) 1.21 (0.95-1.53), 0.11; RFS: (U) 1.27 (1.00-1.61), 0.048	Survival curve	166.8
										LN-: 1.17 (0.66-2.09), 0.59
										LN+: 1.69 [1.11-2.57], 0.014
Bos et al[32], 2003, Netherlands	1985-1993	NA	150	60	G1: 35, G2: 49, G3: 66; NA	G1/G2: 56%; I/II: NA	NA	51 (34.0); > 5%; H1α67	OS; DFS	LN+: (M) OS: 6.37 (1.32-30.67), 0.021; (U) 1.21 (1.05-1.39), 0.008; DFS: (M) 4.19 (1.56-12.08), 0.008; (U) 1.38 (1.11-1.72), 0.004	M: Reported	106
										LN-: (M) OS: 2.16 (0.95-4.89), 0.067; DFS: 1.67 (0.88-3.17), 0.115	U: Survival curve
Gruber et al[29], 2004, Switzerland	August 1988-June 1998	NA	77	57	Moderate/G2: 43, poor/G3: 32; NA	G1/G2: 57%; I/II: NA	NA	43 (55.8); score 1; H1α67	OS; DFS	(U) OS: 1.34 (0.85-2.11), 0.21	M: Reported	36
										DFS: 1.61 (1.02-2.55), 0.04	U: Estimate
										(M) OS: 2.66 (0.83–8.51), 0.09; DFS: 1.68 (0.62–4.47), 0.30
Giatromanolaki et al[27], 2022, Greece	2003-2009	Prospective	175	61	NA	NA; NA	Luminal A: 88, luminal B: 47, HER2 +: 24, TNBC: 16	39 (22.3); ≥ 50%; ESEE122	OS, DFS	(U) OS: 2.26 (1.06-4.82), 0.035	Survival curve	130
										DFS: 1.96 (1.01-3.80), 0.004
Ong et al[19], 2022, Singapore	2003-2013	Retrospective	307	55	G1/2: 48, G3: 288; NA	G1/G2: 14%; I/II: NA	TNBC	141 (45.9); score ≥ 1; NA	OS, DFS	(U) OS: 1.06 (0.65-1.75), 0.807	Estimate	NA
										DFS: 1.40 (0.90-2.17), 0.137
Schindl et al[33], 2002, Austria	NA	Prospective	206	52.3	NA	NA; NA	NA	48 (23.3); score 3; H1α67	OS, DFS	OS: (U) 1.89 (1.01-3.53), 0.045; (M) 1.41 (1.12-1.77), 0.003	M: Reported	87
										DFS: (U) 2.83 (1.54-5.19), 0.0008; (M) 1.40 (1.14-1.72), 0.001	U: Survival curve
Yamamoto et al[43], 2008, Japan	1993-2001	NA	171	NA	G1: 56, G2: 83, G3: 32	G1/G2: 81%; I/II: 90%	NA	63 (36.8); > 5%; H1α67	OS, DFS	OS: (U) 1.39 (1.17-1.65), 0.0002; (M) 2.15 (1.15-5.76), 0.016	M: Reported	60
					I: 45, II: 109, III: 17					DFS: (U) 1.7 (1.67-1.70), < 0.0001	U: Survival curve
										1.59 (1.05-2.43), 0.017
Lu et al[18], 2022, China	2017-2019	NA	189	52.0	NA; I-II: 91, III- IV: 98	G1/G2: NA; I/II: 48%	Luminal A: 61, luminal B: 43, HER2 +: 56, TNBC: 29	122 (64.6); score ≥ 2; ab51608	OS; DFS	(M) OS: 1.024 (0.85-1.23), 0.788; DFS: 1.07 (0.88-1.29), 0.506	Reported	60
Choi et al[52], 2013, South Korea	January 2000-December 2001	NA	276	49.0	G1: 33, G2: 147, G3: 96	G1/G2: 65%; I/II: NA	Luminal A: 121, luminal B: 33, HER2 +: 25, TNBC: 97	13 (4.7); > 10%; EP1215Y	OS; DFS	OS: (U) 1.95 (1.49-2.54), < 0.001; (M) 4.54 (1.46-14.11), 0.009; DFS: (M) 5.21 (1.84-14.78), 0.002	M: Reported	67
					NA						U: Estimate
Dong et al[53], 2013, China	April 1999-October 2008	NA	378	48.5	G1: 15, G2: 127, G3: 56, Unknown: 178; NA	G1/G2: 72%; I/II: NA	NA	195 (50.1); score 4; MAB5382	OS; DFS	(U) OS: 1.05 (0.81-1.36), 0.714	Estimate	73.8
										DFS: 1.01 (0.81-1.25), 0.963
Huang et al, [46], 2014, Taiwan	1991-2001	NA	96	46.5	NA	NA; NA	NA	29 (30.2); score ≥ 3; NA	OS; DFS	(M): OS: 1.61 (0.63-4.13), 0.322	Reported	NA
										DFS: 2.18 (0.93-5.11), 0.073
Koo and Jung[49], 2010, South Korea	January 2000-December 2001	NA	182 of 224	48.8	NA; I: 32, II: 123, III: 69	G1/G2: NA; I/II: 69%	Luminal A: 115, luminal B: 20, HER2 +: 29, TNBC: 60	33 (18.1); ≥ 1%; EP1215Y	OS; RFS	(U): OS: 2.94 (1.98-4.37), < 0.0001	Survival curve	89.6
										RFS: 2.78 (1.42-5.46), 0.003
Ramírez-Tortosa et al [20], 2022, Spain	NA	Prospective	88 of 95	20	G1: 20, G2: 37, G3: 34, unknown: 4	G1/G2: 63%; I/II: 68%	Luminal A: 31, luminal B: 28, HER2 +: 20, TNBC: 13, Missing: 3	35 (39.8); ≥ 5%; NA	OS; DFS	OS: (M) 1.31 (0.81-2.13), 0.295; DFS: (U) 1.70 (1.12-2.57), 0.013; (M) 2.5 (1.0-6.2), 0.047	DFS: U: Survival curve OS: Estimate	88.8
					Stage: II: 63, III: 29, unknown: 3
Sato-Tadano et al[15], 2013, China	2004-2008	NA	118	57	NA; I: 68, II: 31, III: 19	G1/G2: NA; I/II: 84%	Luminal A: 23, luminal B: 15, HER2 +: 4, TNBC: 10	62 (52.5); score ≥ 3; NA	BCSS; DFS	(U) BCSS: 1.12 (0.78-1.61), 0.54; DFS: 1.07 (0.74-1.53), 0.72	Estimate	57
Trastour et al[30], 2007, France	1993	Retrospective	132	62	G1: 57, G2: 44, G3: 10	G1/G2: 91%; I/II: NA	NA	59 (44.7); > 1%	OS, DFS	(U) OS: 1.84 (1.31-2.5), 0.0005; DFS: 1.64 (1.28-2.1), 0.0001	U: Estimate	138
										(M) OS: 1.25 (0.89-1.76), 0.2	M: DFS reported
					NA			Antiserum 2087		DFS: 4.2 [2.1-8.5], < 0.001	M: OS estimate
Schoppmann et al[40], 2006, Austria	NA	Prospective	119	50.9	G1: 9, G2: 54, G3: 56	G1/G2: 53%; I/II: NA	NA	30 (25.2); score 3; H1α67	OS; DFS	OS: (U) 1.60 (1.05-2.41), 0.027; DFS: (U) 1.59 (1.05-2.40), 0.029; (M) 1.61 (1.06-2.43), 0.025	Estimate	110
					NA
Yan et al[36], 2009, Australia	NA	NA	125	NA	G1: 7, G2: 30, G3: 67, Unknow: 19; NA	G1/G2: 36%; I/II: NA	Luminal: 49, HER2 +: 6, TNBC: 37	55 (44.0); score 1; NA	DFS	(U) 3.25 (1.01-10.51), 0.049;	Survival curve	64
Nie et al[51], 2018, China	January 2013-December 2014	NA	220	46	NA; II: 54, III: 166	G1/G2: NA; I/II: 25%	Luminal A: 8, luminal B: 127, HER2 +: 50, TNBC: 35	150 (68.2); score ≥ 1; NA	DFS	(M) 4.17 (1.01–17.17), 0.048	Reported	220
Chen et al[47], 2007, Taiwan	1988-2002	Retrospective	104	NA	G1: 33, G2: 43, G3: 28; NA	G1/G2: 73%; I/II: NA	NA	47 (45.2); score 3; H1α67	DFS	(U): 3.82 (2.14-6.84), < 0.0001	Reported	> 120
Cui and Jiang[45], 2019, China	January 2012-December 2015	Retrospective	87 of 126	50.4	G1: 5, G2: 48, G3: 34	G1/G2: 70%; I/II: 84%	TNBC	36 (41.4); score ≥ 1; ab51608	DFS	(U): 2.03 (1.36–3.51), < 0.001; (M): 2.22 (1.47–3.77), < 0.001	Reported	24
					I: 19, II: 54, III: 14
Generali et al[35], 2006, Italy	January 1997-December 2001	Prospective	187	NA	G2: 95; G3: 135; missing: 3	G1/G2: 41%; I/II: NA	NA	138 (73.8); score ≥ 1; ESEE122	DFS	(U) 1.83 (1.10-3.04), 0.02	Survival curve	53
					NA					ER +: 1.39 (1.00-1.92), 0.05
Jin et al[44], 2016, South Korea	2003-2006	NA	270	NA	G2: 59, G3: 211; NA	G1/G2: 22%; I/II: NA	TNBC	39 (14.4); ≥ 1%; NA	DFS	(U) 1.26 (1.04-1.52), 0.017	U: Survival curve M: Reported	NA
										(M) 2.62 (1.33-5.15), 0.05
Kuijper et al[14], 2005, Netherlands	NA	NA	37	NA	NA	NA; NA	NA	15 (40.5); ≥ 1%; NA	DFS	(U) 4.39 (1.14-16.99), 0.032	Survival curve	NA
Vleugel et al[28], 2005, Netherlands	NA	NA	200	NA	G1: 61, G2: 78, G3: 61; NA	G1/G2: 70%; I/II: NA	NA	88 (44.0); ≥ 1%; NA	RFS	(M) 2.23 (1.18-4.21), 0.01	Reported	105
Shi et al[50], 2017, China	September 2004-September 2008	Retrospective	60	53	NA; I: 20, II: 28, III: 12	G1/G2: NA; I/II: 80%	NA	20 (33.3); ≥ 5%; ab85886	DFS	(U): 4.76 (2.17, 10.44), < 0.001	Survival curve	60
Tan et al[16], 2007, United Kingdom	NA	NA	295	57	G1: 37, G2: 66, G3: 50; NA	G1/G2: 67%; I/II: NA	NA	125 (42.4); score ≥ 2; ESEE122	DFS	(U) 1.60 (1.02-2.42), 0.04	Reported	105
Shamis et al[21], 2022, United Kingdom	1995-1998	NA	Cohort I: 289 of 373	NA	NA,	NA; NA	ER +	39 (13.5); NA; H1α67	DFS	(U) 1.52 (1.08-2.13), 0.015	Survival curve	> 12
Marton et al[38], 2012, Croatia	2001-2005	NA	31	61.7	G1: 7, G2: 19, G3: 5; NA	G1/G2: 84%; I/II: NA	NA	7 (22.6); score ≥ 1; NA	DFS	(U) 2.20 (0.95-5.11), 0.066	Estimate	144
										(M) 2.74 (1.18-6.36), 0.019

TNM: Tumor-node-metastasis; HIFα: Hypoxia-inducible factor 1α; NA: Not available; TNBC: Triple-negative breast cancer; OS: Overall survival; RFS: Recurrence-free survival; DMSF: Distant metastasis-free survival; DFS: Disease-free survival; DSS/SS: Disease-specific survival; HER2 +: HER2 enrich; MBC: Male breast cancer; BDC: Breast ductal cancer; RCS: Randomized controlled study; BCSS: Breast cancer-specific survival; U: Univariate; M: Multivariate; HR: Hazard ratio; CI: confidence interval.

Forty studies reported 57 analyses, including 28 analyses (n = 5833) of the connection linking HIF-1α expression with OS in breast cancer patients and 29 analyses (n = 5691) of the connections linking HIF-1α expression to DFS among individuals with breast cancer. The most commonly reported cutoff values for HIF-1α expression were nuclear staining exceeding 1% (n = 7) and 5% (n = 5), as well as HIF-1α expression scores (adopting a scoring approach that merges stain intensity with positive cell percentage) exceeding 1 (n = 8) and 3 (n = 6). There were differences in the antibodies employed across the various studies, with 12 different antibodies involved. The most common was HIFα67 (n = 6)[29,32,33,40,43,47], followed by EP1215Y (n = 4)[17,31,49,52].

Results of the meta-analysis

Effect of HIF-1α on OS of breast cancer patients: Figure 2 displays the forest plot of HIF-1α expression among individuals with breast cancer and its relation to OS. Compared to cases exhibiting diminished HIF-1α expression, those with high HIF-1α expression had a greater risk of death and shorter OS (HR = 1.47; 95%CI: 1.29-1.69); I² = 65%; P < 0.00001). According to the leave-one-out sensitivity analysis, leaving out the study that showed the greatest effect size[42] did not affect the influence of HIF-1α on patients’ OS (HR = 1.51; 95%CI: 1.31-1.74; I² = 62%; P < 0.00001). Sensitivity analyses were also carried out by including only studies with large sample sizes or high quality, and these analyses showed consistent results (Table 2). The funnel plot appeared asymmetrical, and the statistically significant result obtained from Egger’s asymmetry test (P = 0.02) (Figure 3), which is below the 0.05, suggesting potential publication bias among the included OS-related studies (Figure 3A). The trim-and-fill method analysis revealed that 9 missing studies were input to account for potential bias in the OS analysis (Figure 3B). Furthermore, observations showed that the trimmed funnel plot exhibited no significant asymmetry, and after adjusting for publication bias, the significant overall meta-analysis effect size suggests the presence of limited or non-significant publication bias in the OS studies.

Figure 2 Forest plot of hypoxia-inducible factor 1α-positive and negative phenotypes compared on overall survival in breast cancer patients. Risk ratios and associated 95% confidence intervals were calculated using a random-effects model. CI: Confidence interval.

Figure 3 Funnel plot illustrates the correlation between hypoxia-inducible factor 1α expression levels and breast cancer prognosis. A: Correlation of high hypoxia-inducible factor 1α expression in overall survival in breast cancer patients. An asymmetric funnel plot and Egger’s test P value (P = 0.02) less than 0.05 suggested potential publication bias in the included studies of overall meta-analysis; B: The trim-and-fill method analysis showed that there was no significant asymmetry in the trimmed funnel plot, and the relevant overall survival meta-analysis effect size remained significant after adjusting for publication bias; C: Correlation of high hypoxia-inducible factor 1α expression in disease-free survival in breast cancer patients. An asymmetric funnel plot and Egger’s test P value (P = 0.001) less than 0.05 suggested potential publication bias in the included studies of overall meta-analysis; D: The trim-and-fill method analysis showed that there was no significant asymmetry in the trimmed funnel plot, and the relevant disease-free survival meta-analysis effect size remained significant after adjusting for publication bias. OS: Overall survival; DFS: Disease-free survival; CI: Confidence interval.

Table 2 Meta-analysis results and sensitivity analysis.

Analysis	Number of studies	Pooled HR ratio (95%CI)	I2 statistic (%)	χ2 P value for heterogeneity	Analytical model	P value for overall effect
Primary analyses
OS	28	1.47 (1.29-1.69)	65	< 0.00001	REM	< 0.00001
DFS	29	1.82 (1.56-2.12)	71	< 0.00001	REM	< 0.00001
Sensitivity analyses
OS
Exclusion of study with the largest effect size[42]	27	1.51 (1.31-1.74)	62	< 0.0001	REM	< 0.00001
Sample size ≥ 200[12,19,26,33,34,37,38,52,53]	9	1.20 (1.06-1.36)	48	0.05	FEM	0.004
NOS scoring ≥ 7[12,13,19,20,32,37,53]	7	1.46 (1.08-1.97)	62	0.02	REM	0.02
DFS
Exclusion of study with the largest effect size[34]	28	1.88 (1.60-2.22)	69	< 0.00001	REM	< 0.00001
Sample size ≥ 200[16,19,21,28,33,34,37,44,51-53]	11	1.44 (1.21-1.71)	63	0.003	REM	< 0.0001
NOS scoring ≥ 7[16,19-21,32,36,37,45,53]	9	1.57 (1.24-1.99)	63	0.005	REM	0.0002

HR: Hazard ratio; CI: Confidence interval; NOS: Newcastle-ottawa scale; OS: Overall survival; DFS: Disease-free survival; FEM: Fixed-effects model; REM: Random-effects model.

Effects of HIF-1α on recurrence-free and metastatic survival of breast cancer patients: Figure 4 presents a forest plot describing the correlation linking HIF-1α expression to DFS in breast cancer patients. High HIF-1α expression was related to a greater risk of recurrence and metastasis, resulting in a shorter DFS (HR = 1.82; 95%CI: 1.56-2.12, I² = 71%; P < 0.00001). This finding substantiates the idea that high HIF-1α expression serves as a prognostic factor influencing recurrence and metastasis in breast cancer (Table 2).

Figure 4 Forest plot of hypoxia-inducible factor 1α positive and negative phenotypes compared on disease-free survival in breast cancer patients. Risk ratios and associated 95% confidence intervals were calculated using a random-effects model. CI: Confidence interval.

A leave-one-out sensitivity analysis was employed. Excluding the study that has the greatest effect size[34], and this did not change the influence of HIF-1α on DFS (HR = 1.88; 95%CI: 1.60-2.22; I² = 69%; P < 0.00001). Sensitivity analyses restricted to large sample sizes or high-quality studies alone also did not alter the overall HR (Table 2). The funnel plot exhibited asymmetry, with the calculated P value (P = 0.001) from Egger’s asymmetry test falling below 0.05, highlighting possible selection bias in the systematic review of DFS interval studies (Figure 3C). Furthermore, the trim-and-fill method analysis revealed that twelve missing studies were input to account for potential bias in the DFS analysis (Figure 3D). However, after adjusting for publication bias, the significant overall meta-analysis effect size suggests the presence of limited or non-significant publication bias in the DFS studies.

Subgroup analysis of the effect of HIF-1α on survival of breast cancer patients: Table 3 and Table 4 present the detailed outcomes of the subgroup analysis, suggesting that age, study design, publication year, HIF-1α expression cutoff value, tumor stage, and tumor grade do not influence the connection linking HIF-1α expression to OS or DFS in breast cancer patients. In other words, HIF-1α expression served as a prognostic factor in various subgroups (P > 0.05) (Tables 3 and 4).

Table 3 Subgroup analysis of hypoxia-inducible factor 1α and overall survival in breast cancer patients.

Stratified analysis	Number of studies	Pooled HR ratio (95%CI)	I2 statistic (%)	χ2 P value for heterogeneity	Analytical model	χ2 P value for subgroup differences
Age (median)
< 53[17,18,20,33,39,40,46,48,49,52,53]	11	1.35 (1.21-1.51)	76	< 0.00001	REM	-
≥ 53[12,13,15,19,27,29,30,32,34,38,54]	11	1.52 (1.33-1.74)	45	0.05	FEM	0.18
Location
Asia[15,18,19,22,42,43,46,48,49,52,53]	11	1.55 (1.39-1.73)	31	0.11	FEM	-
Europe[12,13,20,26,27,29-34,37-40,54]	16	1.58 (1.39-1.79)	29	0.14	FEM
Africa[17]	1	1.20 (1.02-1.41)	NA	-	-	0.002
Antibody
H1α67[29,32,33,40,43]	5	1.52 (1.26-1.83)	27	0.24	FEM	-
EP1215Y[17,31,49,52]	4	1.42 (1.22-1.64)	87	< 0.0001	REM	-
ab51608[18,22]	2	2.23 (1.37-3.64)	0	0.85	FEM	-
H206[34,39]	2	1.65 (1.21-2.25)	70	0.07	REM	-
NB100- 123H2[37]	1	1.21 (0.95-1.54)	NA	-	-	-
NB100- 131[38]	1	1.63 (1.03-2.58)	NA	-	-	-
BD610959[26]	1	1.61 (0.63-4.11)	NA	-	-	-
Ab82832[48]	1	0.23 (0.08-0.66)	NA	-	-	-
Antiserum 2087[30]	1	1.41 (1.12-1.77)	NA	-	-	-
MAB5382[53]	1	1.05 (0.81-1.36)	NA	-	-	-
ESEE122[27]	1	2.26 (1.06-4.82)	NA	-	-	0.16
Cut-off value
Percentage ≥ 1%[26,30,37,39,49]	5	1.49 (1.30-1.71)	72	0.006	REM	-
Percentage ≥ 5%[13,17,20,27,32,34,38,43,52]	9	1.36 (1.20-1.55)	59	0.01	REM	-
Scoring ≥ 1 [18,19,29,31,42,54]	6	1.21 (1.05-1.39)	69	0.007	REM	-
Scoring ≥ 3[12,15,22,33,40,46,48,53]	8	1.25 (1.06-1.47)	70	0.002	REM	0.71
Study design
Retrospective[12,19,30,34,38]	5	1.48 (1.23-1.77)	59	0.04	REM	-
Prospective[17,20,27,33,37,40]	6	1.27 (1.13-1.44)	26	0.24	FEM	0.18
Publish date
Before 2013[13,15,29,30,32-34,37,39,40,42,43,49,52-54]	16	1.32 (1.22-1.44)	68	< 0.0001	REM	-
After 2013[12,17-20,22,26,27,31,38,46,48]	12	1.35 (1.19-1.54)	64	0.001	REM	0.53
Grading
G1/G2 < 65%[13,17,19,20,29,32,37,39,40,54]	10	1.35 (1.21-1.50)	40	0.09	FEM	-
G1/G2 ≥ 65%[22,30,31,38,42,43,48,52,53]	9	1.19 (1.07-1.33)	71	0.0005	REM	0.54
Staging
I/II < 70%[18,20,38,42,48,49]	6	1.26 (1.09-1.45)	87	< 0.0001	REM	-
I/II ≥ 70%[15,22,26,43]	4	1.32 (1.03-1.69)	11	0.34	REM	0.69
Molecular typing
TNBC[17,19]	2	1.18 (1.01-1.38)	0	0.62	FEM	-
Gender
Male[12,13]	2	1.85 (1.25-2.72)	62	0.1	REM	-
Female[15,17-20,22,26,27,29-34,37-40,42,43,46,48,49,52-54]	26	1.29 (1.21-1.38)	65	< 0.00001	REM	0.08

HR: Hazard ratio; CI: Confidence interval; FEM: Fixed-effects model; REM: Random-effects model; TNBC: Triple negative breast cancer.

Table 4 Subgroup analysis of hypoxia-inducible factor 1α and disease-free survival in breast cancer patients.

Stratified analysis	Number of studies	Pooled HR ratio (95%CI)	I2 statistic (%)	χ2 P value for heterogeneity	Analytical model	χ2 P value for subgroup differences
Age median
≤ 53[18,20,33,40,45,46,49,51-53]	10	1.32 (1.18-1.47)	74	< 0.001	REM	-
>53[15,16,19,27,29,30,32,34,41,50]	10	1.35 (1.20-1.53)	76	< 0.001	REM	0.74
Location
Asia[15,18,19,43-47,49-53]	13	1.37 (1.23-1.53)	80	< 0.001	REM	-
Europe[14,16,20,21,27-30,32-37,40,41]	16	1.73 (1.45-2.07)	60	0.001	REM	0.51
Antibody
H1α67[29,32,33,40,43,47]	6	1.61 (1.38-1.87)	63	0.02	REM	-
EP1215Y[49,52]	2	3.34 (1.90-5.90)	0	0.32	FEM	-
ESEE122[16,27,35]	3	1.79 (1.36-2.34)	0	0.82	FEM	-
ab51608[18,45]	2	1.23 (1.03-1.47)	90	0.002	REM	-
H206[34]	1	1.12 (0.96-1.31)	NA	-	-	-
MAB5382[53]	1	1.01 (0.81-1.26)	NA	-	-	-
NB100- 123H2[37]	1	1.27 (1.00-1.61)	NA	-	-	-
ab85886[50]	1	4.76 (2.17-10.44)	NA	-	-	-
Antiserum 2087[30]	1	4.20 (2.10-8.40)	NA	-	-	0.0004
Cut-off value
Percentage ≥ 1%[14,28,30,37,44,49]	6	1.70 (1.40-2.05)	73	0.002	REM	-
Percentage ≥ 5%[20,27,32,34,43,50,52]	7	1.36 (1.19-1.55)	81	< 0.001	REM	-
Scoring ≥ 1[16,18,19,29,35,36,41,45,51]	9	1.40 (1.21-1.61)	60	0.01	REM	-
Scoring ≥ 3[15,33,40,46,47,53]	6	1.31 (1.15-1.48)	78	0.0005	REM	0.25
Study design
Retrospective[19,30,34,45,47,50]	6	1.42 (1.25-1.62)	88	< 0.01	REM	-
Prospective[20,27,33,35,37,40]	6	1.46 (1.28-1.66)	8	0.37	FEM	0.79
Recruitment period
Before 2013[14-16,21,28-30,32-37,40,41,43,47,49,52,53]	19	1.37 (1.26-1.48)	73	< 0.001	REM	-
After 2013[18-21,27,44-46,50,51]	10	1.29 (1.15-1.44)	71	0.0003	REM	0.63
Grading
G1/G2 < 65%[19,20,29,32,35-37,40,44,45]	10	1.63 (1.40-1.89)	37	0.12	FEM	-
G1/G2 ≥ 65%[16,28,30,41,43,47,52,53]	8	2.25 (1.47-3.45)	82	< 0.001	REM	0.35
Staging
I/II < 70%[18,20,49,51]	4	1.22 (1.01-1.47)	76	0.005	REM	-
I/II ≥ 70%[15,43,45,50]	4	1.65 (1.33-2.05)	79	0.003	REM	0.85
Molecular typing
TNBC[19,44,45]	3	1.79 (1.03-3.11)	72	0.03	REM	-

HR: Hazard ratio; CI: Confidence interval; TNBC: Triple negative breast cancer; FEM: Fixed-effects model; REM: Random-effects model.

The meta-analysis results from four studies[17,19,44,45] on triple-negative breast cancer verified that HIF-1α expression remains a prognostic factor for these patients. Analysis of two studies[17,19] focusing on OS revealed that the OS of patients with high HIF-1α expression is shorter than that of patients with low expression (HR = 1.18; 95%CI: 1.01-1.38, I² = 0%; P = 0.03). Three studies[19,44,45] focusing on DFS indicated that patients exhibiting high HIF-1α levels have significantly shorter DFS compared to those with low expression (HR = 1.79; 95%CI: 1.03-3.11, I² = 72%; P = 0.03) (Tables 3 and 4).

Furthermore, the antibody that was used to detect HIF-1α expression can influence the relationship between HIF-1α expression and breast cancer prognosis. The impact of HIF-1α expression detected with the EP1215Y antibody on DFS was greater than that of HIF-1α expression detected with other antibodies (P = 0.0004) (Table 4). However, the effect of the antibody that was used to detect HIF-1α expression on OS was not significant (P = 0.16) (Table 3). Additionally, the region where the study population was located had a significant effect on the correlation linking HIF-la expression with OS in patients with breast cancer (P = 0.002) (Table 3). It is worth noting that two studies[12,13] on male breast cancer revealed that HIF-1α expression remained a prognostic factor for patients, with a shorter OS observed in patients with high HIF-1α expression (HR = 1.85; 95%CI: 1.25-2.72, I² = 62%, P = 0.002) (Table 3).

DISCUSSION

Identifying the risk of disease recurrence and death among breast cancer patients is crucial for guiding the monitoring of recurrence and risk of death and for selecting adjuvant therapies. However, accurately identifying these risks for individual patients remains challenging[55]. HIF-1α serves as the master transcriptional driver enabling cellular adaptation under hypoxia, and it influences the initiation and progression of breast cancer through various mechanisms. HIF-1α can regulate breast cancer invasion, metastasis, and immune evasion. Numerous clinical prognostic studies have confirmed that HIF-1α expression can affect the prognosis of breast cancer patients[10]. This systematic review and meta-analysis comprehensively searched and systematically analyzed 40 published studies examining the correlation linking HIF-1α expression with the risk of recurrence and death in breast cancer patients. These findings indicate that compared to patients with low HIF-1α expression, those with high HIF-1α expression not only have an increased risk of recurrence but also a significantly elevated risk of death. Notably, this systematic review combined with meta-analysis additionally validated the notable prognostic significance of HIF-1α expression within triple-negative breast cancer, which has the worst prognosis. Thus, this study represents the most up-to-date and comprehensive evaluation of the prognostic role of HIF-1α expression in breast cancer patients, and it is believed that this study will provide valuable evidence for the prediction of breast cancer prognosis.

Hypoxia can regulate breast cancer invasion, metastasis, and immune evasion through multiple mechanisms, thereby influencing breast cancer progression and prognosis. First, under hypoxic conditions, cancer cells overexpress HIF-1α, upregulate the transcription of angiogenesis-related factors such as vascular endothelial growth factor (VEGF) and platelet-derived growth factor, and promote breast cancer progression by regulating angiogenesis[56,57]. Second, hypoxia can induce the transcription of drug resistance-related genes, such as MDR1, MRP1, and BRCP, leading to chemotherapy resistance and enabling breast cancer cells to evade the effects of cytotoxic drugs, resulting in metastasis[58,59]. Furthermore, HIF-1α can induce epithelial-mesenchymal transition and establish an immunosuppressive tumor microenvironment; during these processes, cancer cell intercellular contacts are disrupted, and these cells acquire motility, accelerating tumor metastasis[60]. Metastasis significantly increases the risk of mortality. The results of this systematic review and meta-analysis revealed that patients of high HIF-1α expression breast cancer are more prone to recurrence and metastasis than those with low expression, suggesting that upregulated HIF-1α expression may affect breast cancer recurrence and metastasis by regulating angiogenesis and drug resistance, thereby influencing the prognosis of breast cancer. Therefore, targeting HIF-1α or VEGF may help facilitate vascular network stabilization in neoplastic tissues, enhance chemotherapeutic agent penetration, mitigate oxygen-deprived tumor niches and immune-evasive milieus, counteract therapeutic resistance mechanisms, and potentially reverse the prognosis of breast cancer recurrence and metastasis[61]. Research has demonstrated that HIF-1α is involved in multiple signaling pathways, such as the HIF-1, Wnt, and VEGF signaling pathways, and regulates the proliferation and invasion of triple-negative breast cancer cells by recruiting coactivator associated arginine methyltransferase 1[62]. Notably, the findings of this systematic review indicate that triple-negative breast cancer patients with high HIF-1α expression have shorter OS and DFS, suggesting a role of HIF-1α in promoting tumor effects, specifically in triple-negative breast cancer recurrence and metastasis. However, subgroup analysis for other molecular subtypes was not possible due to the lack of studies analyzing the prognostic implications for HIF-1α expression in other molecular subtypes of breast cancer separately. Nevertheless, based on the prognostic significance of HIF-1α in the overall population of breast cancer patients, it can be inferred that HIF-1α also has a prognostic impact on other molecular subtypes of breast cancer.

The results of this meta-analysis suggest that breast cancer patients exhibiting high levels of HIF-1α have a higher mortality rate compared to those with lower expression levels. Among the different age subgroups, although the mortality rate is higher in high-HIF-1α-expression patients, the subgroup of patients over 53 years shows less heterogeneity. These findings indicate that the prognostic significance of HIF-1α expression is more consistent in older breast cancer patients, demonstrating that HIF-1α expression may be a more reliable prognostic predictor in this high-risk age group. In terms of disease stage, the subgroup with a greater proportion of Stage I/II disease (over 70%) demonstrated less heterogeneity, suggesting a more uniform prognostic outcome in earlier stages, including the prognostic significance of tumor markers such as HIF-1α. When considering histological grade, the subgroup with a higher proportion of G1/G2 tumors (above 65%) showed significant heterogeneity in the prognostic role of HIF-1α expression in breast cancer. These findings suggest that the impact of HIF-1α-regulated hypoxia-related transcription in more highly differentiated breast cancers on prognosis varies considerably. The relationship between HIF-1α expression and OS of breast cancer exhibits significant differences across various regions, yet further investigation is required to elucidate potential regulatory role of HIF-1α in breast cancer progression among different populations.

Furthermore, the choice of antibodies that are used to detect HIF-1α expression can affect the ability of HIF-1α expression to predict DFS in breast cancer patients. Therefore, it is recommended to refer to the evidence from this systematic review when selecting antibodies for the detection of HIF-1α expression in studies examining its prognostic role in breast cancer to ensure the selection of antibodies with more stable prognostic significance. Additionally, evidence from prospective cohort studies on the predictive significance of HIF-1α expression for breast cancer prognosis is more consistent, suggesting that rigorous and well-designed study protocols can ensure the consistency and accuracy of prognostic research results. This finding also supports the evidence from subgroup analyses of prospective cohort studies, which are more compelling. Sensitivity analyses performed using the leave-one-out method and by omitting small or low-quality studies did not reveal any significant influence of individual studies or specific types of studies on the overall meta-analysis results, confirming the stability of the findings of this meta-analysis.

Although we comprehensively evaluated the correlation linking HIF-1α expression in breast cancer with patient prognosis, this systematic review and meta-analysis have some limitations. First, there are differences in the cutoff values used in the included studies, so the determination of high HIF-1α expression may vary significantly across studies. This may affect the true correlation that is linking HIF-1α expression levels to breast cancer prognosis, resulting in substantial heterogeneity among studies. Second, due to the lack of detailed data for individual molecular subtypes of breast cancer in the included studies, accurately predicting the impact of different molecular subtypes on the established linkage HIF-1α overexpression and patient prognosis was not possible. Furthermore, restricting the analysis to studies published in Chinese or English may lead to language bias. Thus, considering these factors, we recommend that the results of this meta-analysis be interpreted with caution. Moreover, to improve the quality of prognostic tumor marker reports, we recommend following the Reporting Recommendations for Tumor Marker Prognostic Studies guidelines when new related research is conducted.

CONCLUSION

In summary, this meta-analysis confirms that HIF-1α expression can influence the prognosis of patients with breast cancer. Higher HIF-1α expression is associated with shorter DFS and OS. Hypoxia, which is a significant characteristic of cancer, plays a crucial role in the initiation and progression of breast cancer. Our study revealed that HIF-1α, which is a key indicator of hypoxia in the tumor microenvironment, serves as a prognostic marker for adverse outcomes in breast cancer patients. These findings offer valuable insights for evaluating the biological behavior of breast cancer and guiding decisions on adjuvant treatment. Nevertheless, although the overall meta-analysis suggests that HIF-1α expression affects the prognosis of all breast cancer patients, due to the limited molecular subtype information in the included studies, we can only precisely assess the risk of recurrence, metastasis, and death in a subgroup of triple-negative breast cancer patients exhibiting high HIF-1α expression. Hence, designing more rigorous studies to determine the prognostic significance of HIF-1α expression across diverse molecular subtypes in breast cancer may result in more accurate predictions and customized treatments based on these subtypes.

ACKNOWLEDGEMENTS

The authors wish to acknowledge Dr. Ping ZG, Professor of College of Public Health, Zhengzhou University, for his help in guiding and reviewing the statistical methods of this study.