Sage Journals: Discover world-class research

Abstract

Background:

Hepatocellular carcinoma (HCC) is a highly recurrent tumor after resection and has been closely related to hypoxia. Hypoxia-inducible factors 1α and 2α (HIF-1α and HIF-2α) have been shown to contribute to tumor progression and therapy resistance in HCC. We evaluated the prognostic and clinicopathological significance of HIF-1α and HIF-2α in HCC patients.

Methods:

We systematically searched Embase, Cochrane, PubMed, Scopus and Web of Science (WOS) from inception to 1 June 2020 for studies evaluating HIF-1α and/or HIF-2α expression in HCC. Selected articles evaluate at least one factor by immunohistochemistry (IHC) in HCC patients who underwent surgical resection, and its relationship with prognosis and/or clinicopathological features. Study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO; CDR42020191977). We meta-analyzed the data extracted or estimated according to the Parmar method employing STATA software. We evaluated the overall effect size for the hazard ratio (HR) and odds ratio (OR) with 95% confidence interval (CI), as well as heterogeneity across studies with the I² statistic and chi-square-based Q test. Moreover, we conducted subgroup analysis when heterogeneity was substantial. Publication bias was assessed by funnel plot asymmetry and Egger’s test.

Results:

HIF-1α overexpression was correlated with overall survival (OS), disease-free survival (DFS)/recurrence-free survival (RFS) and clinicopathological features including Barcelona Clinic Liver Cancer (BCLC), capsule infiltration, intrahepatic metastasis, lymph node metastasis, tumor–node–metastasis (TNM), tumor differentiation, tumor number, tumor size (3 cm), vascular invasion and vasculogenic mimicry. We also detected a possible correlation of HIF-1α with alpha-fetoprotein (AFP), cirrhosis, histological grade, tumor size (5 cm) and albumin after subgroup analysis. Initially, only DFS/RFS appeared to be associated with HIF-2α overexpression. Subgroup analysis denoted that HIF-2α overexpression was related to OS and capsule infiltration.

Conclusions:

HIF-1α and HIF-2α overexpression is related to poor OS, DFS/RFS and some clinicopathological features of HCC patients, suggesting that both factors could be useful HCC biomarkers.

Keywords

clinicopathological features hepatocellular carcinoma hypoxia-inducible factor 1α hypoxia-inducible factor 2α prognosis

Introduction

Hepatocarcinoma or hepatocellular carcinoma (HCC) is the main type of primary liver tumor, which currently represents the sixth most common cancer and the fourth cause of cancer-related death worldwide.^1–3 HCC is frequently diagnosed in advanced stages, resulting in a high mortality rate. Moreover, despite early diagnosis allowing curative resection, the recurrence rate in these patients usually reaches up to 60%.⁴ Typical biomarkers such as alpha-fetoprotein (AFP) have been shown not to have enough value to predict HCC prognosis and metastatic recurrence.⁴ Recent research has focused on discovering further useful biomarkers which include serum metabolites or enzymes.^5,6 Therefore, determining new effective biomarkers is necessary to predict the clinical prognosis and treatment response of HCC individuals accurately.

Hypoxia is a shared phenomenon among solid tumors, such as HCC, that plays a critical role in tumor development and progression, and is also associated with resistance to both radiation and sorafenib treatment in HCC.^7,8 The cellular response to low oxygen tension is mainly mediated by the hypoxia-inducible factors (HIFs), heterodimeric transcription factors comprising a constitutively expressed subunit (HIF-β) and an oxygen-regulated subunit (HIF-1α, HIF-2α and HIF-3α). Although both factors are frequently overexpressed in HCC, HIF-1α mediates acute hypoxia whereas HIF-2α likely drives the chronic hypoxia response.^8,9

HIF-1α has been reported to be overexpressed in several tumors,^10–17 denoting a correlation between HIF-1α high expression and tumorigenesis, cancer progression and worse prognosis. Even though various research supports the role of HIF-1α overexpression in prompting invasion^18,19 and HCC patients’ survival shortening,¹⁹ the relationship of HIF-1α with every clinicopathological feature and prognosis in HCC still remains inconclusive. HIF-2α upregulation has also been linked to poor prognosis in diverse malignancies^{11,14,20–24} while, in HCC, HIF-2α has been shown to promote invasion and metastasis.^25,26 However, there is a lack of documentation on the association between HIF-2α overexpression and a poorer outcome in HCC, given that existing results are controversial and inconsistent.^25,26

In the present study, we conducted a systematic review with meta-analysis of the available evidence on the relationship between HIF-1α or HIF-2α expression and prognosis and clinicopathological features in HCC. Our aim was to assess the strength of this association to understand better the development and progression of HCC as well as to make better clinical decisions and to improve HCC patients’ survival further.

Methods

This analysis was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplemental Table 1).²⁷ The study protocol was previously registered in the International Prospective Register of Systematic Reviews (PROSPERO, registration number CRD42020191977).

Study objectives

We firstly aimed at evaluating the prognostic value of HIF-1α or HIF-2α expression in HCC patients who underwent surgical resection, concerning overall survival (OS) and disease-free survival (DFS)/recurrence-free survival (RFS). The second purpose was to explore the association of HIF-1α or HIF-2α expression with tumor and patient characteristics.

Search strategy

A comprehensive literature search of Embase, Cochrane, PubMed, Scopus and Web of Science (WOS) databases was performed prior to 1 June 2020. Studies eligible for this analysis were identified using the following search strategy: (“HCC” OR “hepatocarcinoma” OR “hepatocellular carcinoma”) AND (“HIF” OR “hypoxia-inducible factor”) (Supplemental Table 2).

Criteria for inclusion and exclusion

Studies meeting the following criteria were selected: (1) patients with distinctive HCC diagnosis by pathology; (2) HIF-1α or HIF-2α protein expression determined using immunohistochemistry (IHC); (3) samples obtained via surgical resection; (4) relationship between the HIF-1α or HIF-2α expression in HCC and clinicopathological features or survival information was examined; (5) appropriate statistical methodology was used; (6) articles in English.

We excluded studies complying with the following: (1) studies conducted only on cell lines or animals; (2) reviews, case reports, letters, book chapters or meeting communications; (3) tumor samples without intratumoral tissues, or just involving the paracarcinoma tissues; (4) the detection method was not IHC; (5) studies in which the required data were not provided or could not be calculated/estimated; (6) articles without English full text.

Data extraction and quality assessment

Three authors, CMB, PFP and FF independently screened the full text of selected studies to confirm eligibility, assess quality, and extract data. Discrepancies were resolved by discussion and consensus.

The baseline characteristics of each included study were extracted and are shown in Table 1.^29–62 The Newcastle–Ottawa scale (NOS) score was used for assessing the quality of selected articles, which ranged from 0 to 9.²⁸ Studies with scores ⩾5 were regarded as high-quality studies; while low-quality studies were not included in the quantitative synthesis. Furthermore, we collected in Supplemental Table 3 the IHC antibodies and the staining procedure employed in the included articles.

Table 1.

Baseline characteristics of included articles.

Study (reference)	Year	Sample size	Sample size (M/F)	Intervention	Age range	Mean/ median age	Study quality	HIF-1α measurement	Survival analysis	Hazard ratios	High HIF-1α definition	Number of patients with ‘high’ HIF-1α
Huang et al.²⁹	2005	36	32/4	Resection	19–77	45.90	6/9	IHC	OS	Estimated	Positive staining^a	24
Wada et al.³⁰	2006	60	45/15	Resection	44–79	63	6/9	IHC	DFS	Reported	>1% nuclear staining and/or strong cytoplasmic staining	7
Xie et al.³¹	2008	72	59/13	Surgical resection	23–71	50.57	7/9	IHC	OS/DFS	Reported	III and IV^b	37
Dai et al.³²	2009	110	95/15	Hepatectomy	28–75	52.40	7/9	IHC/RT-qPCR	OS/DFS	Reported	III and IV^b	39
Liu et al.³³	2010	200	169/31	Radical resection	NR	NR	7/9	IHC	OS/TTR	Reported	>50% with nuclear staining	126
Xiang et al.³⁴	2011	309	262/47	Curative resection	NR	NR	7/9	IHC	NR	NR	⩾10% nuclear staining	85
Li et al.³⁵	2012	35	30/5	Curative hepatectomy	34–68	50 ± 9.19	5/9	IHC	NR	NR	Positive staining^c	28
Xia et al.³⁶	2012	406	331/75	Curative resection	NR	51.10	7/9	IHC	OS/TTR	Estimated	4–5 (+) or 6–7 (++)^d	212
Xiang et al.³⁷	2012	69	61/8	Curative resection	NR	NR	7/9	IHC	OS/RFS	Reported	⩾10% nuclear staining	30
Cui et al.³⁸	2013	55	34/21	Surgery	20–73	41	7/9	IHC	OS	NEP	3 (moderate staining) and 4–6 (strong staining)^e	30
Ma et al.³⁹	2013	207	156/51	Resection	23–80	57	6/9	IHC	NR	NR	⩾4^f	147
Wang et al.⁴⁰	2014	45	34/11	Hepatectomy	36–78	NR	6/9	IHC/RT-qPCR	OS	NEP	⩾3^g	32
Yang et al.⁴¹	2014	126	110/16	Surgical resection	19–66	48.80	7/9	IHC/WB/RT-qPCR	CS/DFS	Estimated	III and IV^b	72
Huang et al.⁴²	2015	47	35/12	Surgery	33–74	53	6/9	IHC	NR	NR	⩾10% cyplasmic staining	19
Li et al.⁴³	2015	102	87/15	Hepatectomy	NR	NR	6/9	IHC	OS/DFS	Estimated	2–4^h	64
Srivastava et al.⁴⁴	2015	179	142/37	Curative hepatectomy	NR	57.50	5/9	IHC	OS/RFS	Reported	III and IV^b	108
Zhao et al.⁴⁵	2015	97	NR	Surgical resection	NR	NR	5/9	IHC	CS	Estimated	3–6ⁱ	63
Tang et al.⁴⁶	2016	143	130/13	Curative resection	21–70	49.47	6/9	IHC	CS/TTR	Estimated	Scores determined by software based on the percentage of positively stained cells and the staining intensity	72
Wang et al.⁴⁷	2017	201	169/32	Hepatectomy	NR	NR	6/9	IHC	CS	Estimated	⩾4^j	94
Dai et al.⁴⁸	2018	90	84/6	Curative hepatectomy	13–81	54	6/9	IHC	OS/DFS	Reported	2 and 3^b	39
Tian et al.⁴⁹	2018	65	38/27	Surgery	25–77	46.50 ± 2.80	6/9	IHC/WB	NR	NR	>3^k	30
Wang et al.⁵⁰	2018	419	313/106	Surgery	NR	NR	7/9	IHC	OS	Reported	⩾30%	223 (6 missing)
Zou et al.⁵¹	2018	138	116/22	Surgery	NR	NR	6/9	IHC/RT-qPCR/WB	CS/TTR	Estimated	Optimal cut-off point of the relative integrated optical densities based on patients’ outcome	73
Gong et al.⁵²	2019	137	115/22	Primary surgical resection	NR	NR	7/9	IHC	OS/TTR	Estimated	⩾6^l	68
Wu et al.⁵³	2019	119	NR	Curative resection	NR	NR	5/9	IHC	OS	Estimated	⩾6^l	67
Zhou et al.⁵⁴	2020	90	74/16	Surgery	NR	NR	2/9	IHC	OS	NEP	NR	NR
Qian et al.⁵⁵	2020	111	NR	Surgery	NR	NR	5/9	IHC	OS	Estimated	>2^f	57 (2 missing)
Study (Reference)	Year	Sample size	Sample size (M/F)	Intervention	Age range	Mean/ median age	Study quality	HIF-2α measurement	Survival analysis	Hazard ratios	High HIF-2α definition	Number of patients with “high” HIF-2α
Bangoura et al.⁵⁶	2004	97	76/21	Resection	34–78	61.40 ± 8.90	6/9	IHC	NR	NR	++ (dark brown)	31
Bangoura et al.⁵⁷	2007	315	260/55	Curative surgical resection	46–79	60.80	7/9	IHC	CS	Estimated	Positive staining^m	219
Sun et al.⁵⁸	2013	246	198/48	Curative resection	NR	NR	6/9	IHC/WB/RT-qPCR	OS	Reported	>50%	118
Yang et al.⁴¹	2014	126	110/16	Surgical resection	19–66	48.80	7/9	IHC/WB/RT-qPCR	CS/DFS	Estimated	III and IV^b	17
Yang et al.⁵⁹	2016	206	177/29	Radical resection	31–84	57.20	7/9	IHC/WB	OS/RFS	Reported	>50%	67
Jiang et al.⁶⁰	2018	84	70/14	Curative surgery	NR	NR	7/9	IHC/WB/RT-qPCR	NR	NR	⩾4ⁿ	34
Chen et al.⁶¹	2019	139	116/23	Hepatectomy	NR	NR	7/9	IHC/WB/RT-qPCR	OS/DFS	Estimated	Median value of final scores (product of the percentage of stained cells by the staining strength) as a cut-off	67
Cao et al.⁶²	2020	328	NR	Surgery	NR	NR	5/9	IHC	OS	NEP	3–9^j	NR

CS, cumulative survival; DFS, disease-free survival; F, female; HIF, hypoxia-inducible factor; ICH, immunohistochemistry; M, male; NEP, no estimation possible; NR, not reported; OS, overall survival; RFS, recurrence-free survival; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; TTR, time to recurrence; WB, western blot.

The sum of sections with weak and strong staining.

Nuclear staining in 10–50% of cells and/or distinct/moderate cytoplasmic staining (III or 2), and nuclear staining in >50% of cells and/or strong cytoplasmic staining (IV or 3).

The sum of cases with weak (10–25%), moderate (26–50%) and strong staining (>51%).

Final scores were assessed by the sum of the intensity (0, negative; 1, weak; and 2, strong), and the staining extent based on the percentage of positive tumor cells (0, negative; 1, 1–25%; 2, 26–50%; 3, 51–75%; and 4, 76–100%).

Final scores were assessed by the sum of the intensity (0, negative; 1, weak; 2, intermediate; 3, strong), and the extent of immunoreaction (0, 0%; 1, <5%; 2, 5–50%; and 3, >50%).

Final scores were assessed by multiplying the result of the percentage of positive cells (⩽5%, 0; –25%, 1; –50%, 2; –75%, 3; >75%, 4) by the staining intensity (colorless, 0; pale yellow, 1; deep yellow and brownish red, 2; sepia, 3).

Final scores were assessed by the sum of the cytoplasmic staining degree (0, no or negligible staining; 1, pale yellow staining; 2, brown-yellow staining; 3, brown staining) and the punctuation obtained based on the percentage of positively stained cells (0, <5%; 1, 5–25%; 2, >25–50%; and 3, >50%).

Final scores were assessed by determining the percentage of immunoreactive tumor cells (0,0%; 1, 1–25%; 2, 26–50%; 3, 51–75%; 4, 76–100%).

Final scores were assessed by the sum of the positivity extent (0, <10%; 1, <25%; 2, <50%; 3, >50%) and the staining intensity (0, no appreciable; 1, barely detectable; 2, readily visible; 3, dark brown staining).

Final scores were assessed by multiplying the positivity extent (0, <10%; 1, <25%; 2, <50%; 3, >50%) by the staining intensity (0, no appreciable; 1, barely detectable; 2, readily visible; 3, dark brown staining).

Final scores were assessed by multiplying the degree of staining (negative control, 0; light yellow, 1; tan, 2; sepia, 3) by the scoring of the positive cells proportion (⩽10%, 1; 11–50%, 2; 51–75%, 3; >75%, 4).

Final scores were assessed by multiplying the stained area (1, 1–25%; 2, 26–50%; 3, 51–75%; 4, 76–100%) by the staining intensity (0, 1, 2, or 3)

More than 65% of cells were stained intensely (+++) or moderately (++) or weakly (+).

Final scores were assessed by multiplying the results of the percentage (0, <10%; 1, 10–30%; 2, 31–60%; 3, >61%) and the intensity (0, lack of any immunoreactivity; 1, light yellow; 2, yellow-brown; and 3, brown) of immune-staining cells.

Statistical analysis

STATA software version 16 (Stata Corporation, College Station, TX, USA) was employed to assess the correlation between HIF-1α or HIF-2α expression and prognosis and clinicopathological features in HCC.

We measured the effect of HIF expression on HCC in two steps. Firstly, we pooled the OS, DFS, RFS and time to recurrence (TTR) by hazard ratio (HR) and 95% confidence interval (CI) to calculate the effective value to assess the correlation between each HIF and HCC prognosis. OS was measured from the intervention date until either the day of death or the last follow-up visit. DFS, RFS and TTR were defined as the period from the intervention date to the date of last follow-up or recurrence. HR and the corresponding 95% CI were combined across studies. The Parmar method⁶³ was used to extract data when no direct information could be obtained from the primary study. Secondly, the strength of association between HIF-1α or HIF-2α overexpression and tumor clinicopathological features was evaluated by estimating the odds ratio (OR) with 95% CI. Combined HR >1 and OR >1 suggested a higher risk of poor survival and a higher incidence of the analyzed feature, respectively, related to HIF overexpression. These relationships were significant when p < 0.05.

Heterogeneity was tested using the chi-square-based Q test, showing significant levels when the p-value was < 0.1. The I² statistic, a quantitative measure of inconsistency across studies, was also calculated. The I² varies from 0% (no observed heterogeneity) to 100% (maximal heterogeneity). I² ⩾ 50% was considered to represent substantial heterogeneity. The restricted maximum likelihood (REML) method as the random-effect model was employed when heterogeneity was confirmed by at least one statistical method. Otherwise, the fixed-effects model with inverse variance (IV) method was used. To explore the heterogeneity sources, we conducted subgroup analyses based on sample size, NOS score, follow-up and median age.

We assessed the possibility of publication bias by evaluating funnel plot asymmetry and Egger’s test. When Egger’s p-value was < 0.05 and the funnel plot was asymmetric, significant publication bias existed. In this case, the trim-and-fill method was used to estimate a corrected effect size after adjustment, which helped to determine whether the publication bias substantially affected the robustness of the pooled results.

Results

Study characteristics

A total of 3888 applicable studies were identified after the database search, 2172 studies were duplicates, and after scanning titles and abstracts another 1386 articles were omitted for the following reasons: animal or cells studies, non-HCC or HIF articles, reviews or similar. The full text of 330 articles was assessed for eligibility, finding 24 with full text in Chinese, 264 without HIF IHC or analysis about survival or clinicopathological features, seven did not employ surgical resection and one was about HIF-3α. Thus, these 296 papers were also excluded from our study. After screening, 34 studies^29–62 were assessed for quality and data extraction. Cao et al.⁶² and Zhou et al.⁵⁴ did not provide enough data to calculate HR and its 95% CI. Moreover, Zhou et al.⁵⁴ did not reach the quality threshold (Table 1). Eventually, 32 publications were eligible for quantitative meta-analysis: 25 on HIF-1α, six on HIF-2α and one about both factors (Figure 1).

Figure 1.

PRISMA flowchart of study selection.

The baseline of included articles and results of quality assessment are summarized in Table 1. These studies were published from 2004 to 2020 and a total of 3578 (eight were ‘missing’ in HIF-1α expression analysis) and 1213 HCC patients consisted of HIF-1α and HIF-2α, respectively. All included patients came from Asia, mostly from China. For HIF-1α, patients’ number across studies ranged from 35 to 419, and 1846 (51.7%) had HIF-1α overexpression. From 84 to 315 patients by study were enrolled to HIF-2α analysis, 553 (45.6%) showed HIF-2α overexpression. Among the 25 HIF-1α articles, 18 provided data on OS, eight on DFS/RFS and 23 on clinicopathological features; while for HIF-2α, five of seven articles supplied OS, three DFS/RFS and all of them clinicopathological features. Five HIF-1α studies evaluated TTR; nonetheless, only one article reported the HR and estimation was not possible according to the Parmar method. Hence, TTR analysis was not included.

Patients in all studies underwent surgical resection. In Wada et al.³⁰ and Dai et al.³² some of the enrolled patients received preoperative antitumor therapy. Moreover, patients undergoing external beam radiotherapy had postoperative adjuvant treatment in Xiang et al.³⁷ The rest of the included studies had no intervention prior to surgery. Regarding etiology, within the 32 papers included 21 evaluated patients with hepatitis B,^{29–31,33–39,41,43,44,46,50–52,57–59,61} nine assessed patients with hepatitis C,^{30,34.36,41,43,44,57–59} and only one alcoholic patient was assessed;⁴⁴ it should be noted that all the studies were performed in the Asiatic population, where hepatitis B is the key etiology factor. Likewise, 20 articles evaluated HCC patients derived from cirrhosis.^{29–34,36,39,41,43,44,46,48,50–52,56–59} In summary, we extracted these data and among the patients included in the present meta-analysis, 78.2% had hepatitis B, 6.6% had hepatitis C and 66.1% had cirrhosis.

Association of HIF protein expression with prognosis

Based on the meta-analysis, we evaluated the prognostic value of HIF-1α and found that high expression correlated with OS (HR 1.73; 95% CI 1.54–1.94; p = 0.00) and with DFS/RFS (HR 1.64; 95% CI 1.36–1.99; p = 0.00), not finding significant heterogeneity (Figure 2A).

Figure 2.

Meta-analysis of the prognostic value of HIF-1α and HIF-2α in HCC patients. Forest plot of OS and DFS/RFS for (A) HIF-1α and (B) HIF-2α.

We also assessed the correlation between HIF-2α protein levels and prognosis. The results suggest that there is no significant association between HIF-2α high expression and OS in HCC patients (HR 1.25; 95% CI 0.68–2.32; p = 0.48), assuming heterogeneity among studies. However, HIF-2α overexpression appears to be associated with DFS/RFS (HR 1.37; 95% CI 1.05–1.79; p = 0.02), and no heterogeneity was shown (Figure 2B).

Association of HIF protein expression with clinicopathological features

Otherwise, we evaluated the possible correlation between HIF expression and different clinicopathological features of HCC patients.

High HIF-1α protein levels were positively associated with Barcelona Clinic Liver Cancer (BCLC) staging (OR 2.49; 95% CI 1.56-3.98; p = 0.00), capsule infiltration (OR 2.48; 95% CI 1.29–4.77; p = 0.01), intrahepatic metastasis (OR 2.90; 95% CI 1.62–5.20; p = 0.00), lymph node metastasis (OR 3.74; 95% CI 1.73–8.07; p = 0.00), tumor–node–metastasis (TNM) classification (I, II–III) (OR 1.59; 95% CI 1.21–2.09; p = 0.00), TNM (I-II, III) (OR 2.62; 95% CI 1.69–4.08; p = 0.00), TNM (I–II, III–IV) (OR 2.23; 95% CI 1.37–3.64; p = 0.00), tumor differentiation (OR 1.78; 95% CI 1.07–2.96; p = 0.03), tumor number (OR 1.50; 95% CI 1.15–1.96; p = 0.00), tumor size (3 cm) (OR 3.70; 95% CI 1.29–10.63; p = 0.02), vascular invasion (OR 2.61; 95% CI 1.82–3.75; p = 0.00) and vasculogenic mimicry (OR 2.61; 95% CI 1.67–4.09; p = 0.00) (Figure 3). We also found that there is no statistical significance with other tumor features, including AFP levels (20 ng/ml) (OR 1.39; 95% CI 0.92–2.09; p = 0.11), AFP (400 ng/ml) (OR 1.49; 95% CI 0.67–3.33; p = 0.33), age (50 years) (OR 0.86; 95% CI 0.57–1.31; p = 0.49), age (60 years) (OR 1.03; 95% CI 0.68–1.55; p = 0.90), albumin (35 U/L) (OR 0.60; 95% CI 0.26–1.38; p = 0.23), alanine aminotransferase (ALT) (40 U/L) (OR 0.86; 95% CI 0.60–1.24; p = 0.42), ALT (80 U/L) (OR 1.04; 95% CI 0.67–1.62; p = 0.86), bilirubin (1 μmol/L) (OR 1.65; 95% CI 1.00–2.73; p = 0.0501), capsule formation (OR 0.89; 95% CI 0.69–1.13; p = 0.33), Child–Pugh score (OR 1.52; 95% CI 0.91–2.53; p = 0.11), cirrhosis (OR 1.32; 95% CI 0.97–1.80; p = 0.08), distant metastasis (OR 6.14; 95% CI 0.83–45.48; p = 0.08), Edmondson grading (OR 1.35; 95% CI 0.83–2.20; p = 0.22), gender (OR 0.93; 95% CI 0.77–1.14; p = 0.51), hepatitis B (OR 0.99; 95% CI 0.82–1.21; p = 0.96), hepatitis C (OR 1.25; 95% CI 0.72–2.16; p = 0.42), histological grade (OR 1.54; 95% CI 0.70–3.40; p = 0.28) and tumor size (5 cm) (OR 1.40; 95% CI 0.88–2.22; p = 0.16) (Figure 4).

Figure 3.

Forest plot of the clinicopathological features significantly associated with HIF-1α overexpression in HCC patients.

Figure 4.

Forest plot of the clinicopathological features not significantly associated with HIF-1α overexpression in HCC patients.

All HIF-2α studies were included in the analysis of clinical and pathological features. Positive HIF-2α expression was not significantly associated with any feature analyzed: AFP levels (400 ng/ml) (OR 0.88; 95% CI 0.60–1.30; p = 0.52), age (50 years) (OR 1.17; 95% CI 0.79–1.73; p = 0.44), capsule formation (OR 1.31; 95% CI 0.93–1.83; p = 0.12), capsule infiltration (OR 1.82; 95% CI 0.54–6.13; p = 0.33), cirrhosis (OR 1.22; 95% CI 0.91–1.64; p = 0.19), Edmondson grading (OR 11.05; 95% CI 0.02–6167.72; p = 0.46), gender (OR 0.95; 95% CI 0.68–1.35; p = 0.79), hepatitis B (OR 1.03; 95% CI 0.76–1.39; p = 0.86), histological grade (OR 0.93; 95% CI 0.43–1.99; p = 0.85), necrosis (OR 1.32; 95% CI 0.25–6.98; p = 0.74), TNM (I–II, III–IV) (OR 1.12; 95% CI 0.40–3.10; p = 0.83) tumor number (OR 1.44; 95% CI 0.92–2.27; p = 0.11), tumor size (5 cm) (OR 1.20; 95% CI 0.36–3.99; p = 0.77) and vascular invasion (OR 1.16; 95% CI 0.67–2.00; p = 0.60) (Figure 5).

Figure 5.

Forest plot of the association between HIF-2α overexpression and clinicopathological features in HCC patients.

Subgroup analysis

To explore the potential heterogeneity sources, subgroup analysis for heterogeneity parameters was performed.

When subgroups for HIF-1α were based on sample size, HIF-1α expression was related to AFP levels (20 ng/ml) (n ⩾ 100: OR 1.59; 95% CI 1.03–2.46; p = 0.04), cirrhosis (n < 300: OR 1.48; 95% CI 1.06–2.07; p = 0.02), tumor size (5 cm) (n ⩾ 200: OR 1.46; 95% CI 1.04–2.06; p = 0.03) and vascular invasion (n ⩾ 100: OR 2.39; 95% CI 1.58–3.61; p = 0.00) (n < 100: OR 3.22; 95% CI 1.87–5.55; p = 0.00) (n ⩾ 200: OR 1.32; 95% CI 1.04–1.68; p = 0.02) (n < 200: OR 3.54; 95% CI 2.69–4.66; p = 0.00) (n ⩾ 300: OR 1.39; 95% CI 1.08–1.79; p = 0.01) (n < 300: OR 3.22; 95% CI 2.13–4.88; p = 0.00) (n < 400: OR 3.04; 95% CI 2.08–4.45; p = 0.00); nonetheless, heterogeneity continued to be substantial in some cases. Sample size subgroups provided assumable heterogeneity for AFP levels (20 ng/ml) (n < 100, n ⩾ 200/300), age (50 years) (n ⩾ 100, n < 200), albumin (n ⩾ 100), cirrhosis (n < 100), tumor size (5 cm) (n < 100, n⩾300) and vascular invasion (n < 100, n ⩾ 200, n < 200, n ⩾ 300, n ⩾ 400). NOS score subgroups displayed an association between HIF-1α overexpression and albumin (n < 7: OR 0.45; 95% CI 0.21–0.93; p = 0.03), tumor differentiation (n ⩾ 6: OR 1.76; 95% CI 1.04–2.97; p = 0.04), tumor size (5 cm) (n = 6: OR 2.27; 95% CI 1.10–4.70; p = 0.03) (n < 7: OR 2.45; 95% CI 1.20–4.99; p = 0.01) and vascular invasion (n ⩾ 7: OR 1.99; 95% CI 1.36–2.90; p = 0.00) (n < 7: OR 4.00; 95% CI 2.14–7.46; p = 0.00), where only albumin subgroup presented low heterogeneity (n < 7). In addition, NOS classification resolved heterogeneity for AFP (20 ng/ml) (n ⩾ 7), AFP (400 ng/ml) (n ⩾ 7), age (50 years) (n ⩾ 7), cirrhosis (n ⩾ 7) and histological grade (n ⩾ 7), but not showing association with protein expression. Curiously, the single elimination of Zou et al.⁵¹ in AFP (20 ng/ml), and Xia et al.³⁶ and Tian et al.⁴⁹ in tumor differentiation led to an assumable heterogeneity, although there was no correlation with HIF-1α overexpression. However, the deletion of Xiang et al.³⁴ and Ma et al.³⁹ for cirrhosis, and Wang et al.⁴⁰ for histological grade, achieved low heterogeneity and a significant association with HIF-1α (Table 2).

Table 2.

Subgroup analysis of prognostic and clinicopathological features of HIF-1α and HIF-2α.

HIF-1α
Subgroups	Number of studies (n)	Number of cases (n)	HIF-1α⁺ (n)	HIF-1α⁺ (%)	Pooled data			Test for heterogeneity		Model used
					OR	95% CI	p-value	p-value	I² (%)	Model used
AFP (20 ng/ml)
Sample size (n)
⩾100	5	1106	481	43.49	1.59	1.03–2.46	0.04	0.06	55.89	REM
<100	2	116	49	42.24	0.74	0.35–1.59	0.45	0.40	0.00	FEM
⩾200/300	2	715	297	41.54	1.31	0.91–1.89	0.15	0.32	0.00	FEM
<200/300	5	507	233	45.96	1.39	0.73–2.64	0.31	0.02	64.06	REM
⩾400	1	406	212	52.22	1.12	0.69–1.81	0.65	−	−	REM
<400	6	816	318	38.97	1.45	0.88–2.40	0.14	0.04	57.84	REM
NOS score
⩾7	4	894	366	40.94	1.27	0.93–1.75	0.13	0.20	35.04	FEM
<7	3	328	164	50.00	1.61	0.67–3.88	0.29	0.03	68.87	REM
Without Zou et al.⁵¹	6	1084	457	42.16	1.21	0.91–1.60	0.19	0.40	3.17	FEM
AFP (400 ng/ml)
Sample size (n)
⩾100	2	263	140	53.23	1.16	0.45–2.98	0.75	0.06	72.69	REM
<100	2	162	76	46.91	2.03	0.38–10.90	0.41	0.02	82.78	REM
NOS score
⩾7	3	335	177	52.84	1.11	0.72–1.71	0.65	0.14	49.57	FEM
<7	1	90	39	43.33	4.86	1.73–13.62	0.00	−	−	REM
Age (50 years)
Sample size (n)
⩾100	4	566	298	52.65	1.00	0.70–1.43	0.99	0.32	13.66	FEM
<100	3	231	106	45.89	0.70	0.26–1.94	0.50	0.03	71.05	REM
⩾200	1	201	94	46.77	1.50	0.81–2.79	0.20	−	−	REM
<200	6	596	310	52.01	0.77	0.55–1.09	0.14	0.15	38.69	FEM
NOS score
⩾7	4	404	207	51.24	0.95	0.63–1.43	0.82	0.20	35.07	FEM
<7	3	393	197	50.13	0.74	0.33–1.66	0.46	0.03	68.99	REM
Albumin
Sample size (n)
⩾100	2	275	141	51.27	0.82	0.45–1.51	0.53	0.31	2.23	FEM
<100	1	90	39	43.33	0.21	0.05–0.84	0.03	−	−	REM
NOS score
⩾7	1	137	68	49.64	1.12	0.48–2.59	0.80	−	−	REM
<7	2	228	112	49.12	0.45	0.21–0.93	0.03	0.21	36.71	FEM
Cirrhosis
Sample size (n)
⩾100	10	1878	958	51.01	1.40	0.95–2.07	0.09	0.01	59.07	REM
<100	4	258	107	41.47	1.18	0.67–2.06	0.56	0.42	0.00	FEM
⩾200	4	1122	570	71.36	1.35	0.73–2.48	0.34	0.03	68.62	REM
<200	10	1014	495	48.82	1.32	0.90–1.95	0.16	0.08	36.74	REM
⩾300	2	715	297	41.54	0.89	0.48–1.66	0.72	0.14	54.49	REM
<300	12	1421	768	54.05	1.48	1.06–2.07	0.02	0.07	35.53	REM
⩾400	1	406	212	52.22	1.16	0.73–1.86	0.53	−	−	REM
<400	13	1730	853	49.31	1.35	0.95–1.92	0.10	0.02	48.04	REM
NOS score
⩾7	7	1360	639	46.99	1.20	0.91–1.58	0.19	0.16	35.56	FEM
<7	7	776	426	54.90	1.37	0.76–2.45	0.30	0.03	57.23	REM
Without Xiang et al.³⁴ and Ma et al.³⁹	12	1620	833	51.42	1.33	1.05–1.69	0.02	0.12	33.52	FEM
Histological grade
Sample size (n)
⩾100	1	126	72	57.14	1.15	0.43–3.04	0.78	−	−	REM
<100	4	267	115	43.07	1.70	0.58–4.95	0.33	0.06	60.50	REM
NOS score
⩾7	2	198	109	55.05	1.67	0.79–3.54	0.18	0.24	28.40	FEM
<7	3	195	78	40.00	1.40	0.32–6.02	0.65	0.04	68.67	REM
Without Wang et al. 2014⁴¹	4	348	155	44.54	2.04	1.12-3.69	0.02	0.53	0.00	FEM
Tumor differentiation
Sample size (n)
≥100	5	1168	534	45.72	1.53	0.86-2.74	0.15	0.00	76.26	REM
<100	4	216	107	49.54	4.82	0.59–39.55	0.14	0.01	81.86	REM
⩾200	3	915	423	46.23	1.59	0.70–3.60	0.27	0.00	83.89	REM
<200	6	469	218	46.48	2.36	0.85–6.51	0.10	0.01	75.68	REM
⩾300	2	715	297	41.54	1.93	0.58–6.40	0.28	0.00	90.39	REM
<300	7	669	344	51.42	1.68	0.92–3.07	0.09	0.01	53.66	REM
⩾400	1	406	212	52.22	3.54	2.15–5.83	0.00	−	−	REM
<400	8	978	429	43.87	1.42	0.93–2.17	0.10	0.01	37.68	REM
NOS score
5	1	35	28	80.00	3.04	0.15–62.85	0.47	−	−	REM
6	3	255	121	47.45	7.25	0.47–111.92	0.16	0.01	90.53	REM
7	5	1094	492	44.97	1.38	0.79–2.43	0.26	0.00	72.22	REM
NOS (threshold 6)
⩾6	8	1349	613	45.44	1.76	1.04–2.97	0.04	0.00	70.54	REM
<6	1	35	28	80.00	3.04	0.15–62.85	0.47	−	−	REM
NOS (threshold 7)
⩾7	5	1094	492	44.97	1.38	0.79–2.43	0.26	0.00	72.22	REM
<7	4	290	149	51.38	5.47	0.79–37.76	0.08	0.03	80.72	REM
Without Tian et al.⁴⁹	7	913	399	43.70	1.25	0.92–1.69	0.16	0.40	3.09	FEM
Tumor size (5 cm)
Sample size (n)
⩾100	11	2079	1052	50.60	1.75	0.90–3.39	0.10	0.00	91.66	REM
<100	8	467	250	53.53	0.88	0.59–1.31	0.53	0.27	20.05	FEM
⩾200	5	1323	664	50.19	1.46	1.04–2.06	0.03	0.08	51.84	REM
<200	14	1223	638	52.17	1.37	0.71–2.65	0.35	0.00	84.81	REM
⩾300	2	715	297	41.54	1.15	0.84–1.57	0.38	0.37	0.00	FEM
<300	17	1831	1005	54.89	1.45	0.85–2.46	0.17	0.00	84.36	REM
⩾400	1	406	212	52.22	1.28	0.87–1.91	0.21	−	−	REM
<400	18	2140	1090	50.93	1.41	0.86–2.31	0.18	0.00	84.50	REM
NOS score
5	1	35	28	80	15.00	0.78-287.68	0.07	−	−	REM
6	9	1027	575	55.99	2.27	1.10–4.70	0.03	0.00	84.53	REM
7	9	1484	699	47.10	0.84	0.57–1.24	0.38	0.01	65.64	REM
NOS (threshold 6)
⩾6	18	2511	1274	50.74	1.34	0.84–2.13	0.22	0.00	85.49	REM
<6	1	35	28	80	15.00	0.78–287.68	0.07	−	−	REM
NOS (threshold 7)
⩾7	9	1484	699	47.10	0.84	0.57–1.24	0.38	0.01	65.64	REM
<7	10	1062	603	56.78	2.45	1.20–4.99	0.01	0.00	83.02	REM
Vascular invasion
Sample size (n)
⩾100	11	2233	1106	49.53	2.39	1.58–3.61	0.00	0.00	75.38	REM
<100	5	327	137	41.90	3.22	1.87–5.55	0.00	0.13	44.29	FEM
⩾200	4	1334	646	48.43	1.32	1.04–1.68	0.02	0.38	2.54	FEM
<200	12	1226	597	48.69	3.54	2.69–4.66	0.00	0.11	35.11	FEM
⩾300	3	1134	520	45.86	1.39	1.08–1.79	0.01	0.41	0.00	FEM
<300	13	1426	723	50.70	3.22	2.13–4.88	0.00	0.00	57.60	REM
⩾400	2	825	435	52.73	1.28	0.97–1.70	0.08	0.87	0.00	FEM
<400	14	1735	808	46.57	3.04	2.08–4.45	0.00	0.00	57.00	REM
NOS score
⩾7	9	1848	892	48.27	1.99	1.36–2.90	0.00	0.01	63.60	REM
<7	7	712	351	49.30	4.00	2.14–7.46	0.00	0.02	58.51	REM
HIF-2α
Subgroups	Number of studies (n)	Number of cases (n)	HIF-2α⁺ (n)	HIF-2α⁺ (%)	Pooled data			Test for heterogeneity		Model used
					HR	95% CI	p-value	p-value	I² (%)	Model used
Overall Survival
Sample size (n)
⩾200	3	767	404	53.67	1.00	0.39–2.60	1.00	0.00	96.06	REM
<200	2	265	84	31.70	1.83	1.18–2.84	0.01	0.93	0.00	FEM
≥300	1	315	219	69.52	2.58	2.08-3.20	0.00	−	−	REM
<300	4	717	269	37.52	1.01	0.55-1.87	0.97	0.00	84.09	REM
NOS score
⩾7	3	660	353	53.48	1.44	0.61–3.42	0.41	0.00	93.51	REM
<7	2	372	135	36.29	1.00	0.35–2.84	1.00	0.01	85.67	REM
Follow-up (months)
>72	2	454	286	63.00	2.47	2.02–3.03	0.00	0.29	11.74	FEM
⩽72	3	578	202	34.95	0.82	0.44–1.54	0.54	0.02	81.14	REM
Subgroups	Number of studies (n)	Number of cases (n)	HIF-2α⁺ (n)	HIF-2α⁺ (%)	Pooled data			Test for heterogeneity		Model used
					OR	95% CI	p-value	p-value	I² (%)	Model used
Capsule infiltration
Sample size (n)
⩾100	2	441	236	53.51	1.15	0.25–5.27	0.85	0.04	77.39	REM
<100	1	97	31	31.96	4.76	1.54–14.70	0.01	−	−	REM
⩾200	1	315	219	69.52	2.26	1.19–4.28	0.01	−	−	REM
<200	2	223	48	21.52	1.54	0.16–14.79	0.71	0.01	85.41	REM
NOS score
⩾7	1	315	219	69.52	2.26	1.19–4.28	0.01	−	−	REM
<7	2	223	48	21.52	1.54	0.16–14.79	0.71	0.01	85.41	REM
Median age (years)
⩾50	2	412	250	60.68	2.71	1.55–4.73	0.00	0.26	21.15	FEM
<50	1	126	17	13.49	0.47	0.13–1.75	0.26	−	−	REM
Tumor size (5 cm)
Sample size (n)
⩾100	5	1032	488	47.29	1.34	0.27–6.71	0.72	0.00	96.16	REM
<100	2	181	65	35.91	0.89	0.13–6.10	0.90	0.01	87.06	REM
⩾200	3	767	404	52.67	1.17	0.07–18.75	0.91	0.00	98.22	REM
<200	4	446	149	33.41	1.25	0.43–3.60	0.69	0.00	81.42	REM
⩾300	1	315	219	69.52	19.74	9.40–41.45	0.00	−	−	REM
<300	6	898	334	37.19	0.74	0.30–1.85	0.52	0.00	88.12	REM
NOS score
⩾7	5	870	404	46.44	1.32	0.26-6.80	0.74	0.00	95.19	REM
<7	2	343	149	43.44	0.91	0.15–5.39	0.92	0.00	90.05	REM
Vascular invasion
Sample size (n)
⩾100	4	893	421	47.14	1.13	0.53–2.42	0.75	0.01	75.80	REM
<100	2	181	65	35.91	1.27	0.62–2.60	0.51	0.27	16.26	FEM
⩾200	3	767	404	52.67	1.30	0.55–3.07	0.54	0.00	82.57	REM
<200	3	307	82	26.71	1.08	0.56–2.06	0.82	0.31	15.06	FEM
⩾300	1	315	219	69.52	3.05	1.53–6.10	0.00	−	−	REM
<300	5	759	267	35.18	0.90	0.64–1.27	0.55	0.38	3.84	FEM
NOS score
⩾7	4	731	337	46.10	1.24	0.57–2.70	0.59	0.04	63.40	REM
<7	2	343	149	43.44	1.00	0.42–2.35	0.99	0.09	66.09	REM

AFP, alpha-fetoprotein; CI, confidence interval; FEM, fixed-effects model; HIF, hypoxia-inducible factor; HR, hazard ratio; NOS, Newcastle–Ottawa scale; OR, odds ratio; REM, random-effects model.

For HIF-2α, when subgroups were based on sample size, HIF-2α expression was linked to OS (n < 200: HR 1.83; 95% CI 1.18–2.84; p = 0.01), where heterogeneity was solved. Moreover, heterogeneity was eliminated from vascular invasion (n < 100, n < 200, n < 300), but no association was found. The follow-up subgroup showed a relationship between OS and HIF-2α overexpression (follow-up ⩾72: HR 2.47; 95% CI 2.02–3.03; p = 0.00) with reduced heterogeneity. Capsule infiltration was analyzed according to median age, displaying correlation to HIF-2α (years ⩾ 50: OR 2.71; 95% CI 1.55–4.73; p = 0.00) and low heterogeneity. Subgroup classification by NOS did not exhibit any change. Thus, tumor size heterogeneity could not be resolved (Table 2).

Subgroups with only one study were not considered. The results of subgroup analysis revealed that sample size, NOS score, median age and follow-up time likely triggered heterogeneity.

Publication bias

There was pronounced asymmetry denoting publication bias on the OS parameter for HIF-1α that was confirmed by Egger’s test (p = 0.0027). Hence, the trim-and-fill method was used, in which seven studies were imputed and the global effect size was corrected (HR 1.559; 95% CI 1.405–1.731). Instead, there was no asymmetry detected for DFS/RFS (p = 0.0631) (Table 3; Figure 6A). Concerning the clinicopathological features of HIF-1α, asymmetry was only identified in gender (p = 0.0255), tumor differentiation (p = 0.0428) and vascular invasion (p = 0.0016). In gender, trim-and-fill estimated the new global effect (OR 0.827; 95% CI 0.687–0.997) and imputed five studies. Likewise, six ‘missing’ studies were included in the vascular invasion funnel plot, adjusting effect size (OR 1.749; 95% CI 1.121–2.729). Conversely, trim-and-fill analysis did not report any ‘missing’ study for tumor differentiation. All imputed studies for HIF-1α parameters were on the left side of the funnel plot (Table 3; Figure 7).

Table 3.

Assessment of publication bias on prognostic and clinicopathological features of HIF-1α and HIF-2α.

HIF-1α
Survival	Number of studies	Egger’s test (p-value)	Model used	Trim-and-fill HR (95% CI)	Imputed studies
OS	18	0.00*	FEM	1.56 (1.41–1.73)	7
DFS/RFS	8	0.06	FEM	−	−
Clinicopathological feature	Number of studies	Egger’s test (p-value)	Model used	Trim-and-fill OR (95% CI)	Imputed studies
AFP (20 ng/ml)	7	0.59	REM	−	−
AFP (400 ng/ml)	4	0.41	REM	−	−
Age (50 years)	7	0.18	REM	−	−
Age (60 years)	4	0.24	FEM	−	−
Albumin	3	0.07	REM	−	−
ALT (40 U/L)	3	0.33	FEM	−	−
ALT (80 U/L)	3	0.59	FEM	−	−
BCLC	3	0.58	FEM	−	−
Bilirubin	2	0.51	FEM	−	−
Capsule formation	7	0.93	FEM	−	−
Capsule infiltration	2	0.38	FEM	−	−
Child–Pugh score	3	0.85	FEM	−	−
Cirrhosis	14	0.97	REM	−	−
Distant metastasis	2	†	REM	−	−
Edmondson grading	3	0.08	FEM	−	−
Gender	19	0.03*	FEM	0.83 (0.69–1.00)	5
Hepatitis B	16	0.91	FEM	−	−
Hepatitis C	4	0.94	FEM	−	−
Histological grade	5	0.82	REM	−	−
Intrahepatic metastasis	3	0.70	FEM	−	−
Lymph node metastasis	3	0.39	FEM	−	−
TNM (I, II–III)	4	0.35	FEM	−	−
TNM (I–II, III)	2	0.99	FEM	−	−
TNM (I–II, III–IV)	3	0.50	FEM	−	−
Tumor differentiation	9	0.04*	REM	1.78 (1.07–2.96)	0
Tumor number	8	0.32	FEM	−	−
Tumor size (3 cm)	2	0.47	FEM	−	−
Tumor size (5 cm)	19	0.47	REM	−	−
Vascular invasion	16	0.00*	REM	1.75 (1.12–2.73)	6
Vasculogenic mimicry	3	0.33	FEM	−	−
HIF-2α
Survival	Number of studies	Egger’s test (p-value)	Model used	Trim-and-fill HR (95% CI)	Imputed studies
OS	5	0.93	REM	−	−
DFS/RFS	3	0.55	FEM	−	−
Clinicopathological feature	Number of studies	Egger’s test (p-value)	Model used	Trim-and-fill OR (95% CI)	Imputed studies
AFP (400 ng/ml)	4	0.02*	FEM	1.00 (0.69-1.44)	1
Age (50 years)	4	0.27	FEM	−	−
Capsule formation	3	0.17	FEM	−	−
Capsule infiltration	3	0.63	REM	−	−
Cirrhosis	5	0.46	FEM	−	−
Edmondson grading	2	^†	REM	−	−
Gender	6	0.67	FEM	−	−
Hepatitis B	4	0.54	FEM	−	−
Histological grade	2	0.19	FEM	−	−
Necrosis	2	^†	REM	−	−
TNM (I–II, III–IV)	2	^†	REM	−	−
Tumor number	3	0.08	FEM	−	−
Tumor size (5 cm)	7	0.89	REM	−	−
Vascular invasion	6	0.46	REM	−	−

AFP, alpha-fetoprotein; ALT, alanine aminotransferase; BCLC, Barcelona Clinic Liver Cancer; CI, confidence interval; DFS, disease-free survival; FEM, fixed-effects model; HIF, hypoxia-inducible factor; HR, hazard ratio; OR, odds ratio; OS, overall survival; REM, random-effects model; RFS, recurrence-free survival; TNM, tumor–node–metastasis.

p-value <0.05.

†

Convergence not achieved during tau2 estimation.

Figure 6.

Publication bias analysis of the prognostic value of HIF-1α and HIF-2α. Funnel plot of OS and DFS/RFS for (A) HIF-1α with trim-and-fill funnel plot for OS, and (B) HIF-2α.

Figure 7.

Publication bias analysis of the association between HIF-1α overexpression and clinicopathological features by funnel plot asymmetry. For gender, tumor differentiation and vascular invasion the trim-and-fill funnel plot is also represented.

Conversely, there was no asymmetry evidence on OS (p = 0.9273) and DFS/RFS (p = 0.5480) for HIF-2α (Table 3; Figure 6B). Concerning the clinicopathological features, only AFP levels (p = 0.0213) presented asymmetry, where the trim-and-fill method imputed one study on the right side of the funnel plot and corrected the global effect (OR 1.001; 95% CI 0.694–1.444) (Table 3; Figure 8).

Figure 8.

Publication bias analysis of the association between HIF-2α overexpression and clinicopathological features by funnel plot asymmetry. For AFP (400 ng/ml) the trim-and-fill funnel plot is also represented.

Discussion

Hypoxia is a common microenvironment characteristic of solid tumors, such as HCC, which arises as consequence of defective vascularization and intense metabolic activity.⁸ Despite HCC being characterized by being one of the most hyper-vascularized tumors, hypoxic regions are frequently present in HCC due to rapid proliferation of tumor cells and the formation of aberrant blood vessels.^19,26 Although a reduction in oxygen supply is initially harmful for cell survival, some tumor cells adapt to the hypoxic microenvironment by decreasing energy consumption and enhancing anaerobic metabolism.²⁵ This adaptive response is mainly accomplished by HIFs, which entails a set of pro-survival changes implicated in aggressive tumor progression, therapy resistance, selection of more invasive clones and poor clinical outcomes.⁸

Due to difficult detection and the high recurrence rate of early HCC,⁴ discerning the risk of recurrence and mortality in HCC patients is key to guide surveillance and determine possible adjuvant therapies. Thus, in this study we evaluated the main hypoxia response mediators, HIF-1α and HIF-2α, as potential clinical biomarkers for predicting HCC prognosis. To investigate the relationship between HIF protein expression and HCC, the present meta-analysis aimed to examine the association between HIF-1α or HIF-2α overexpression and the prognosis and clinicopathological features of HCC patients.

A total of 26 high-quality studies (3570 patients) were included for HIF-1α analysis. Pooled results showed that overexpression of HIF-1α leads to poor OS and DFS/RFS in HCC. Previous studies conducted by Zheng et al.¹⁹ and Cao et al.,¹⁸ which enrolled seven and eight articles, respectively, evaluated HIF-1α expression in HCC. According to our results, both studies showed association between high HIF-1α levels and DFS, and Zheng et al.¹⁹ also reported correlation with OS.

Various research performed in lung cancer,¹⁰ renal cell carcinoma,¹¹ pancreatic cancer,⁶⁴ esophageal cancer,¹² gastric tumors,¹³ colorectal cancer,¹⁴ head and neck cancer,¹⁵ oral squamous cell carcinoma,¹⁶ bone tumors,¹⁷ breast cancer,⁶⁵ endometrial cancer,⁶⁶ and epithelial ovarian cancer,⁶⁷ also revealed a significant correlation between HIF-1α overexpression and poor prognosis. Moreover, HIF-1α expression has been associated with worse prognosis in advanced cancer patients treated with chemotherapy and/or radiotherapy; thus, in addition to a useful biomarker, targeting HIF-1α could be an interesting therapeutic approach to improve survival in advanced cancer patients.⁶⁸

Here, significant results were observed in some clinicopathological features, including BCLC staging, capsule infiltration, intrahepatic metastasis, lymph node metastasis, TNM classification, tumor differentiation, tumor number, tumor size (contrasting greater or less than 3 cm), vascular invasion and vasculogenic mimicry. After subgroup analysis according to sample size and NOS score, we also observed a possible HIF-1α correlation with AFP levels, cirrhosis, tumor size (5 cm) and albumin, while histological grade showed association after removing one study.

The two previous meta-analyses^18,19 found a correlation with vascular invasion, and Cao et al.¹⁸ reported no significant association with capsule formation, cirrhosis, tumor differentiation or tumor size. Nevertheless, differences could be explained by the lower number of studies included in such meta-analyses and the absence of subgroup analyses.^18,19

Meanwhile, seven high-quality articles (1213 patients) were employed to analyze HIF-2α. Initially, only DFS/RFS appeared to be associated with HIF-2α expression; nonetheless, subgroup analysis denoted that HIF-2α overexpression is also markedly related to OS when grouped by sample size and follow-up time.

A meta-analysis by Yao et al.²⁶ and another by Luo et al.²⁵ evaluated HIF-2α in HCC and multiple types of cancer, respectively. Luo et al.²⁵ found a relation between OS and HIF-2α with multivariate but not with univariate analysis. Likewise, this meta-analysis reported that HIF-2α overexpression results in poor OS in additional tumors including lung or colorectal cancers, among others.²⁵ In contrast, Yao et al.²⁶ did not observe significant correlation among OS and this transcription factor, which can be explained based on the fact that our work included most recent articles and excluded those with full text in Chinese.²⁶ No previous study has evaluated the impact on DFS or RFS.

Other research has observed the association between HIF-2α overexpression and a worse prognosis in further tumors such as lung cancer,²¹ renal cell carcinoma,¹¹ gastric cancer,²³ colorectal cancer,^14,24 oral squamous cell carcinoma²² or head and neck cancer.²⁰

In addition, concerning the clinicopathological parameters, only capsule infiltration was related to HIF-2α high levels when the median age is ⩾50 years. This result agrees with Yao et al.,²⁶ but they also found an association with vein invasion and histological grade. Besides, both that meta-analysis and ours observed no correlation of HIF-2α with cirrhosis, necrosis and tumor size.²⁶

Hypoxia is the principal physiological stimulus inducing angiogenesis in HCC through the upregulation of angiogenic factors. Vascular endothelial growth factor (VEGF) is crucial for blood vessel formation by promoting the growth and migration of endothelial cells, and it is transcriptionally regulated by both HIFs.⁶⁹ Some studies included in this meta-analysis showed that high VEGF levels were associated with angiogenesis, microvessel density, vasculogenic mimicry and poor prognosis; having a positive correlation between VEGF and HIF-1α.^{29,30,34,44,45} Other angiogenic factors have also been linked to HIF regulation and have been described to be involved in angiogenesis in HCC, such as angiopoietin-2,³⁰ HIF-1α targets bone morphogenetic protein 4 (BMP4), while HIF-2α targets stem cell factor (SCF) and plasminogen activator inhibitor-1 (PAI-1), and both HIFs target erythropoietin and platelet-derived growth factor (PDGF) expression.⁶⁹ Consequently, hypoxia, and more specifically HIFs, contribute to angiogenesis in HCC and could be related to the results obtained in this meta-analysis associating high HIF-1α expression with increased vascular invasion and vasculogenic mimicry.

Furthermore, epithelial to mesenchymal transition and metastasis can be induced under hypoxia in HCC cells.⁶⁹ Several studies have described that the increase in invasion and worse prognosis in HCC patients could be related to the expression of invasion-related proteins such as metalloproteinases, interleukin-8 (IL-8) or E-cadherin.^{32,34,40,43,44} In preclinical studies similar results were also found, associating high invasion and metastasis with the HIF-1α targets SNAIL-1, granulocyte chemotactic protein-2 (CXCL6), IL-8 and Rab11-family interacting protein 4 (Rab11-FIP4); and the HIF-2α targets SERPINB3, CUB domain-containing protein 1 (CDCP1) and SCF.^69–71 These findings can explain the relationship between both HIFs and the clinicopathological factors related to metastasis, such as intrahepatic or lymph node metastasis and capsule infiltration.

This meta-analysis is a highly comprehensive study performing a detailed quantitative analysis of clinical evidence on HIF-1α and HIF-2α correlation with prognostic variables such as OS and DFS/RFS and with clinicopathological features in HCC patients who underwent surgical resection. Moreover, this work includes an assessment of heterogeneity, subgroup analysis and publication bias. Hence, this is the first meta-analysis analyzing both main hypoxia-inducible factors. Previous meta-analyses evaluating the relationship between HIF-1α or HIF-2α expression and tumor outcome, which were included in the discussion of our results, were performed prior to 2015 and, thus, comprised a lower number of studies and evaluated fewer parameters. Moreover, some studies did not assess publication bias and/or subgroup analysis.

Even though in the present meta-analysis we exhaustively evaluated the association between both HIFs and tumor outcome, some limitations should be acknowledged. Despite full-text articles being obtained, those written in Chinese were excluded and, therefore, likely relevant data. Data extraction was not always possible because of the absence of required variables for the estimation, such as patient follow-up or patient numbers in each group. Furthermore, studies included employed diverse or unspecified antibodies for HIF detection, with miscellaneous or insufficient staining procedure description and different cut-off values for IHC scores of HIFs, which could lead to higher heterogeneity. All the articles included were performed with an Asiatic population, mainly from China, where hepatitis B is the key etiology factor.¹ In accordance with this, most of the articles evaluated the number of patients with hepatitis B or C; however, there are no available studies enrolling populations from other origins, such as western countries where etiological factors such as obesity or non-alcoholic fatty liver disease (NAFLD) prevail.¹ The research volume in HIF-2α analysis was low due to the low number of studies found in the literature; then, more high-quality articles would be needed. Besides, some variables were not collected uniformly between studies, hindering the assessment of established aims and the potential sources of heterogeneity. Finally, publication bias was denoted in some parameters.

In summary, HIF overexpression is linked to a more aggressive behavior of HCC. In this meta-analysis, 51.7% and 45.6% of patients displayed HIF-1α and HIF-2α overexpression, respectively. HIF-1α and HIF-2α seem to act as negative prognosis markers, being linked to poor OS, DFS, RFS and some clinicopathological features of HCC patients. This evidence suggests that both HIFs are useful biomarkers for predicting HCC prognosis that may improve clinical decisions, especially when combined with other prognostic-related markers. These results have been represented graphically in Figure 9.

Figure 9.

Graphical abstract.

Supplemental Material

sj-pdf-1-tam-10.1177_1758835920987071 – Supplemental material for Prognostic and clinicopathological significance of hypoxia-inducible factors 1α and 2α in hepatocellular carcinoma: a systematic review with meta-analysis

Supplemental material, sj-pdf-1-tam-10.1177_1758835920987071 for Prognostic and clinicopathological significance of hypoxia-inducible factors 1α and 2α in hepatocellular carcinoma: a systematic review with meta-analysis by Carolina Méndez-Blanco, Paula Fernández-Palanca, Flavia Fondevila, Javier González-Gallego and José L. Mauriz in Therapeutic Advances in Medical Oncology

Footnotes

Acknowledgements

CMB is supported by the Asociación Española Contra el Cáncer (AECC)-Junta Provincial de León. PFP and FF are supported by the Ministry of Education of Spain (FPU17/01995 and FPU16/05277 grants). CIBERehd is funded by Instituto de Salud Carlos III, Spain.

Author contributions

All authors were responsible for study conception and design, interpretation of the data and drafting of the manuscript. Systematic literature review, data extraction and data analysis were performed by CMB, PFP and FF. Moreover, JGG and JLM carried out the study supervision. The final version of the manuscript was approved by all authors.

Conflict of interest statement

The authors declare that there is no conflict of interest.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Javier González-Gallego

José L. Mauriz

Supplemental material

Supplemental material for this article is available online.

References

Bray

Ferlay

Soerjomataram

, et al Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018; 68: 394–424.

Fernández-Palanca

Fondevila

Méndez-Blanco

, et al Antitumor effects of quercetin in hepatocarcinoma in vitro and in vivo models: a systematic review. Nutrients 2019; 11: 2875.

Fondevila

Méndez-Blanco

Fernández-Palanca

, et al Anti-tumoral activity of single and combined regorafenib treatments in preclinical models of liver and gastrointestinal cancers. Exp Mol Med 2019; 51: 109.

Hsu

Hsieh

Chen

, et al Axl and autophagy LC3 expression in tumors is strongly associated with clinical prognosis of hepatocellular carcinoma patients after curative resection. Cancer Med 2019; 8: 3453–3463.

Pascut

Sukowati

CHC

Antoniali

, et al Serum AP-endonuclease 1 (sAPE1) as novel biomarker for hepatocellular carcinoma. Oncotarget 2019; 10: 383–394.

Banales

Iñarrairaegui

Arbelaiz

, et al Serum metabolites as diagnostic biomarkers for cholangiocarcinoma, hepatocellular carcinoma, and primary sclerosing cholangitis. Hepatology 2019; 70: 547–562.

Méndez-Blanco

Fondevila

Fernández-Palanca

, et al Stabilization of hypoxia-inducible factors and BNIP3 promoter methylation contribute to acquired sorafenib resistance in human hepatocarcinoma cells. Cancers (Basel) 2019; 11: 1984.

Méndez-Blanco

Fondevila

García-Palomo

, et al Sorafenib resistance in hepatocarcinoma: role of hypoxia-inducible factors. Exp Mol Med 2018; 50: 134.

Prieto-Domínguez

Méndez-Blanco

Carbajo-Pescador

, et al Melatonin enhances sorafenib actions in human hepatocarcinoma cells by inhibiting mTORC1/p70S6K/HIF-1α and hypoxia-mediated mitophagy. Oncotarget 2017; 8: 91402–91414.

10.

Wang

Rui

, et al Prognosis value of HIF-1α expression in patients with non-small cell lung cancer. Gene 2014; 541: 69–74.

11.

Fan

, et al Prognostic significance of hypoxia-inducible factor expression in renal cell carcinoma: a PRISMA-compliant systematic review and meta-analysis. Med (United States) 2015; 94: e1646.

12.

Jing

Wang

Expression of hypoxia inducible factor 1 alpha and its clinical significance in esophageal carcinoma: a meta-analysis. Tumor Biol 2017; 39: 1–12.

13.

Chen

Liu

, et al Clinical and prognostic significance of HIF-1α, PTEN, CD44v6, and survivin for gastric cancer: a meta-analysis. PLoS One 2014; 9: e91842.

14.

Chen

Xia

, et al Prognostic value and clinicopathological differences of HIFs in colorectal cancer: evidence from meta-analysis. PLoS One 2013; 8: e80337.

15.

Swartz

Pothen

Stegeman

, et al Clinical implications of hypoxia biomarker expression in head and neck squamous cell carcinoma: a systematic review. Cancer Med 2015; 4: 1101–1116.

16.

Zhou

Huang

Wang

, et al Clinical and prognostic significance of HIF-1α overexpression in oral squamous cell carcinoma: a meta-analysis. World J Surg Oncol 2017; 15: 104.

17.

Luo

Ren

Zhang

, et al Clinicopathological and prognostic value of hypoxia-inducible factor-1α in patients with bone tumor: a systematic review and meta-analysis. J Orthop Surg Res 2019; 14: 56.

18.

Cao

Yang

, et al Protein expression of hypoxia-inducible factor-1 alpha and hepatocellular carcinoma: a systematic review with meta-analysis. Clin Res Hepatol Gastroenterol 2014; 38: 598–603.

19.

Zheng

Chen

Yin

, et al Prognostic significance of HIF-1α expression in hepatocellular carcinoma: a meta-analysis. PLoS One 2013; 8: e65753.

20.

Gong

Zhang

Zhou

, et al Prognostic value of HIFs expression in head and neck cancer: a systematic review. PLoS One 2013; 8: e75094.

21.

, et al Prognostic role of hypoxic inducible factor expression in non-small cell lung cancer: a meta-analysis. Asian Pacific J Cancer Prev 2013; 14: 3607–3612.

22.

Qian

Wenguang

Zhiyong

, et al Hypoxia inducible factor: a potential prognostic biomarker in oral squamous cell carcinoma. Tumor Biol 2016; 37: 10815–10820.

23.

Zheng

Zhao

Clinicopathological differences and prognostic value of hypoxia-inducible factor-2α expression for gastric cancer evidence from meta-analysis. Med (United States) 2016; 95: e2871.

24.

Han

Huang

, et al Association between Hypoxia-Inducible Factor-2α (HIF-2α) expression and colorectal cancer and its prognostic role: a systematic analysis. Cell Physiol Biochem 2018; 48: 516–527.

25.

Luo

Liu

Lin

, et al The clinicopathologic and prognostic value of hypoxia-inducible factor-2α in cancer patients: a systematic review and meta-analysis. Cancer Epidemiol Biomarkers Prev 2019; 28: 857–866.

26.

Yao

Pan

, et al Prognostic significance and clinicopathological features of hypoxic inducible factor-2alpha expression in hepatocellular carcinoma. Saudi Med J 2015; 36: 170–175.

27.

Moher

Liberati

Tetzlaff

, et al Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009; 6: e1000097.

28.

Wells

Shea

O’Connell

, et al The Newcastle–Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. The Ottawa Hospital Research Institute, http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp9 (2019; accessed 5 June 2020).

29.

Huang

Yang

WQ.

Expression of hypoxia-inducible factor 1α and vascular endothelial growth factor in hepatocellular carcinoma: impact on neovascularization and survival. World J Gastroenterol 2005; 11: 1705–1708.

30.

Wada

Nagano

Yamamoto

, et al Expression pattern of angiogenic factors and prognosis after hepatic resection in hepatocellular carcinoma: importance of angiopoietin-2 and hypoxia-induced factor-1α. Liver Int 2006; 26: 414–423.

31.

Xie

Song

Liu

, et al The expression of hypoxia-inducible factor-1α in hepatitis B virus-related hepatocellular carcinoma: correlation with patients’ prognosis and hepatitis B virus X protein. Dig Dis Sci 2008; 53: 3225–3233.

32.

Dai

Gao

Qiu

, et al Hypoxia-inducible factor-1 alpha, in association with inflammation, angiogenesis and MYC, is a critical prognostic factor in patients with HCC after surgery. BMC Cancer 2009; 9: 418.

33.

Liu

Zhu

Wang

, et al Activation of β-catenin by hypoxia in hepatocellular carcinoma contributes to enhanced metastatic potential and poor prognosis. Clin Cancer Res 2010; 16: 2740–2750.

34.

Xiang

Zeng

Fan

, et al Gene expression profiling of fixed tissues identified hypoxia-inducible factor-1α, VEGF, and matrix metalloproteinase-2 as biomarkers of lymph node metastasis in hepatocellular carcinoma. Clin Cancer Res 2011; 17: 5463–5472.

35.

Yao

Dong

, et al Abnormal expression of hypoxia inducible factor-1α and clinical values of molecular-targeted interference in hepatocellular carcinoma. Chinese–German J Clin Oncol 2012; 11: 125–129.

36.

Xia

Huang

, et al The TNF-α/ROS/HIF-1-induced upregulation of foxMI expression promotes HCC proliferation and resistance to apoptosis. Carcinogenesis 2012; 33: 2250–2259.

37.

Xiang

Zeng

Fan

, et al The expression of HIF-1α in primary hepatocellular carcinoma and its correlation with radiotherapy response and clinical outcome. Mol Biol Rep 2012; 39: 2021–2029.

38.

Cui

Huang

Chen

, et al The role of Aurora A in hypoxia-inducible factor 1α-promoting malignant phenotypes of hepatocelluar carcinoma. Cell Cycle 2013; 12: 2849–2866.

39.

Tan

Expression of HIF-1α in hepatocellular carcinoma and its relationship with vasculogenic mimicry and clinical pathology. Chinese–German J Clin Oncol 2013; 12: 528–531.

40.

Wang

Ding

Fan

, et al Expression and significance of MMP2 and HIF-1α in hepatocellular carcinoma. Oncol Lett 2014; 8: 539–546.

41.

Yang

Liu

Jiang

, et al The correlation of expression levels of HIF-1α and HIF-2α in hepatocellular carcinoma with capsular invasion, portal vein tumor thrombi and patients’ clinical outcome. Jpn J Clin Oncol 2014; 44: 159–167.

42.

Huang

Meng

, et al Correlations between ADC values and molecular markers of Ki-67 and HIF-1α in hepatocellular carcinoma. Eur J Radiol 2015; 84: 2464–2469.

43.

Yang

Biskup

, et al Co-expression of CXCL8 and HIF-1α is associated with metastasis and poor prognosis in hepatocellular carcinoma. Oncotarget 2015; 6: 22880–22889.

44.

Srivastava

Thakkar

Yeoh

, et al Expression of proteins associated with hypoxia and Wnt pathway activation is of prognostic significance in hepatocellular carcinoma. Virchows Arch 2015; 466: 541–548.

45.

Zhao

Sun

Zhao

, et al Role of Bcl-2 and its associated miRNAs in vasculogenic mimicry of hepatocellular carcinoma. Int J Clin Exp Pathol 2015; 8: 15759–15768.

46.

Tang

Liu

, et al 14-3-3ζ promotes hepatocellular carcinoma venous metastasis by modulating hypoxia-inducible factor-1α. Oncotarget 2016; 7: 15854–15867.

47.

Wang

Zhao

Zhu

, et al HIF-1α promoted vasculogenic mimicry formation in hepatocellular carcinoma through LOXL2 up-regulation in hypoxic tumor microenvironment. J Exp Clin Cancer Res 2017; 36: 60.

48.

Dai

Yang

S li

, et al Association of PD-L1 and HIF-1α coexpression with poor prognosis in hepatocellular carcinoma. Transl Oncol 2018; 11: 559–566.

49.

Tian

Liu

, et al Expressions and correlation analysis of HIF-1α, survivin and VEGF in patients with hepatocarcinoma. Eur Rev Med Pharmacol Sci 2018; 22: 3378–3385.

50.

Wang

Zhang

, et al Hypoxia inducible factor 1α in hepatocellular carcinoma with cirrhosis: association with prognosis. Pathol Res Pract 2018; 214: 1987–1992.

51.

Zou

Liu

Zhang

, et al The expression of FAP in hepatocellular carcinoma cells is induced by hypoxia and correlates with poor clinical outcomes. J Cancer 2018; 9: 3278–3286.

52.

Gong

Zou

Peng

, et al Nuclear GAPDH is vital for hypoxia-induced hepatic stellate cell apoptosis and is indicative of aggressive hepatocellular carcinoma behavior. Cancer Manag Res 2019; 11: 4947–4956.

53.

Zhou

Yin

, et al Blocking triggering receptor expressed on myeloid cells-1-positive tumor-associated macrophages induced by hypoxia reverses immunosuppression and anti-programmed cell death ligand 1 resistance in liver cancer. Hepatology 2019; 70: 198–214.

54.

Zhou

Dong

Xiu

, et al Meloxicam, a selective COX-2 ihibitor, mediates hypoxia-inducible factor- (HIF-) 1α signaling in hepatocellular carcinoma. Oxid Med Cell Longev 2020; 2020: 7079308.

55.

Qian

Zheng

, et al High methylation levels of histone H3 lysine 9 associated with activation of hypoxia-inducible factor 1α (HIF-1α) predict patients’ worse prognosis in human hepatocellular carcinomas. Cancer Genet 2020; 245: 17–26.

56.

Bangoura

Yang

Huang

, et al Expression of HIF-2α/EPAS1 in hepatocellular carcinoma. World J Gastroenterol 2004; 10: 525–530.

57.

Bangoura

Liu

Qian

, et al Prognostic significance of HIF-2α/EPAS1 expression in hepatocellular carcinoma. World J Gastroenterol 2007; 13: 3176–3182.

58.

Sun

Yang

, et al HIF-2α inhibits hepatocellular carcinoma growth through the TFDP3/E2F1-dependent apoptotic pathway. Hepatology 2013; 57: 1088–1097.

59.

Yang

Liu

Niu

, et al Downregulation and pro-apoptotic effect of hypoxia-inducible factor 2 alpha in hepatocellular carcinoma. Oncotarget 2016; 7: 34571–34581.

60.

Jiang

Liu

Liang

, et al Association of PHD3 and HIF2α gene expression with clinicopathological characteristics in human hepatocellular carcinoma. Oncol Lett 2018; 15: 545–551.

61.

Chen

Huang

, et al HIF-2α upregulation mediated by hypoxia promotes NAFLD-HCC progression by activating lipid synthesis via the PI3K-AKT-mTOR pathway. Aging (Albany NY) 2019; 11: 10839–10860.

62.

Cao

You

Cui

, et al Cross talk between oxidative stress and hypoxia via thioredoxin and HIF-2α drives metastasis of hepatocellular carcinoma. FASEB J 2020; 34: 5892–5905.

63.

Parmar

MKB

Torri

Stewart

Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998; 17: 2815–2834.

64.

Zhang

Bai

, et al Hypoxia-inducible factor 1α expression and its clinical significance in pancreatic cancer: a meta-analysis. Pancreatology 2014; 14: 391–397.

65.

Sun

Wang

Correlation between HIF-1α expression and breast cancer risk: a meta-analysis. Breast J 2014; 20: 213–215.

66.

Sadlecki

Bodnar

Grabiec

, et al The role of hypoxia-inducible factor-1α, glucose transporter-1, (GLUT-1) and carbon anhydrase IX in endometrial cancer patients. Biomed Res Int 2014; 2014: 616850.

67.

Jin

Wang

Liang

, et al Pathological and prognostic significance of hypoxia-inducible factor 1α expression in epithelial ovarian cancer: a meta-analysis. Tumor Biol 2014; 35: 8149–8159.

68.

Han

Huang

Hou

, et al The prognostic value of hypoxia-inducible factor-1α in advanced cancer survivors: a meta-analysis with trial sequential analysis. Ther Adv Med Oncol 2019; 11: 1–21.

69.

Chen

Lou

Hypoxia inducible factors in hepatocellular carcinoma. Oncotarget 2017; 8: 46691–46703.

70.

Cannito

Turato

Paternostro

, et al Hypoxia up-regulates SERPINB3 through HIF-2α in human liver cancer cells. Oncotarget 2014; 6: 2206–2221.

71.

Cao

Gao

Zhou

, et al HIF-2α regulates CDCP1 to promote PKCδ-mediated migration in hepatocellular carcinoma. Tumor Biol 2016; 37: 1651–1662.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.20 MB