Background

Cancer is the second leading cause of morbidity and mortality worldwide [1]. One major feature of cancer is uncontrolled cell proliferation, which can then invade adjacent parts of the body and spread to other organs, the latter process is referred as metastases, which are the major cause of death from cancer [2]. The most common causes of cancer deaths are due to cancers of the: lung (1.59 million deaths), liver (745,000 deaths), stomach (723,000 deaths), colorectal (694,000 deaths), breast (521,000 deaths) and esophageal (400,000 deaths) [1, 2]. Alongside, metabolic alterations and tumor hypoxia have consistently been identified as classical features with aggressive malignancy [3, 4]. Hypoxia regulates tumor cell phenotype mainly by altering genes that are sensitive to oxygen pressure [5]. However, the exact mechanism of carcinogenesis is yet to be elucidated. In recent years, an increasing number of studies have focused on understanding the relationship between genetic factors and cancer risk [3, 4]. Through the years, it has become well accepted that single nucleotide polymorphisms (SNPs) are the most common and effective type of genetic variations studied in association with disease susceptibility and are the markers of many complex diseases [6].

Hypoxia inducible factor 1α (HIF1A), is a transcription factor that has major impacts in the process of development and progression of cancers [7]. HIF1A regulates the expression of over 100 genes that control the major cellular functions including apoptosis, cell proliferation, glucose metabolism, erythropoiesis, iron metabolism and angiogenesis. It is a master regulator of oxygen homeostasis [7]. In the scientific community, HIF1A has been a research focus and a number of SNPs within HIF1A gene have been identified in association with cancers, with the most widely studied polymorphisms are C1772T (rs11549465) and G1790A (rs11549467) polymorphisms [8–38]. These two SNPs are located within the same domain (ODD/ pVHL) in exon 12 of the HIF1A gene [8, 9]. Recently a meta-analysis has revealed that C1772T is not in substantial linkage disequilibrium (LD) with G1790A [38]. A number of studies have suggested that these two nonsynonymous mutations might alter the transcriptional activity of HIF1A gene by causing structural changes with varied stability, which in turn, might influence the downstream target genes expression and regulation [8, 9, 38]. In the recent years, a good number of studies have investigated the impact of HIF1A polymorphisms on cancer risk in different populations; however reported results varied across studies and remain inconclusive [10–38]. In this study, the effect of rs11549465 (1772 C/T) and rs11549467 (1790 G/A) polymorphisms of HIF1A gene and its association with cancers were investigated systematically through meta-analysis.

Methods

Search study and study selection

The PubMed, PubMed Central and Google Scholar databases were searched systematically to retrieve compatible and pertinent peer reviewed publications of empirical studies. Published articles of last 15 years (ended on December 2014), in English language were only considered for this study. The search terms included were (1) HIF1A, (2) GWAS, (3) SNPs, (4) polymorphisms, (5) C1772T/ P582S, (6) A1790G/ A588T, (7) case-control study, and (8) cancer.

Eligibility criteria

Two authors independently investigated titles and abstracts of all the articles. Irrelevant and incompatible studies were excluded primarily. For final review, criteria’s for further study elimination were: if (1) the study population was not defined completely; (2) it is not a case-control study; (3) not a genome wide association study; (4) incomplete information of allele frequency; and (5) the year of study conducted was not specified. Also, reviews, editorials, meta-analysis and non-human researches were excluded. Only case-control studies, genome wide association study (GWAS) and human researches were considered for the final review. Further, the references of the selected studies were screened carefully for incorporation of additional relevant studies. Only English language articles were considered for this study. Discrepancies and difficulties were discussed with corresponding authors where necessary. Following information were extracted from each study: (1) authors name, (2) year of study, (3) ethnicity of the study subjects, (4) cancer type and (5) allelic frequency (Fig. 1).

Fig. 1
figure 1

Flow diagram of study selection for HIF1A 1772 C/T and 1790 G/A polymorphisms; where “n” in the boxes is the number of corresponding studies

Full size image

Meta-analysis

For HIF1A 1772 C/T polymorphism 22 case-control studies were included of sample size 19024 and for 1790 G/A polymorphism 19 case-control studies were included with sample size 10654. The meta-analysis was prepared in accordance with PRISMA statement [39].

Statistical analysis

Meta-analysis of genome wide association studies (GWAS) of HIF1A were conducted for two polymorphisms, 1772 C/T and 1790 G/A using odds ratios (ORs). A slightly amended estimator of OR was used to avoid the computation of reciprocal of zeros among observed values in the calculation of the original OR [40]. Pooled ORs with 95 % CIs were calculated using random effects model (REM) incorporating the inverse variance weighted method [41]. Heterogeneity among studies was assessed using the Q statistic [42] and quantified using I^2 index [43]. Subgroup analyses were carried out by cancer type and ethnicity. The Hardy Weinberg Equilibrium (HWE) test was performed for the controls of each study. The studies with control not in HWE were supervised for sensitivity analysis. Publication bias was assessed visually by conventionally constructed funnel plot where the inverse of the standard error (1/se) of the effect estimates were plotted against the logarithm transformation of Odds Ratios [log(OR)] [44]. Furthermore, Egger’s test was performed to provide quantitative evidence of publication bias [45]. “Gap: Genetic analysis package” was used to perform the Hardy Weinberg Equilibrium (HWE) test [46, 47]. All analyses were conducted using “meta” package in R environment [46].

Summary measures

Odds Ratios (OR) with a 95 % confidence interval (CI) were calculated to evaluate the genotype contrasts. The genotype contrasts for the HIF1A 1772 C/T polymorphisms were: homozygote comparison [TT versus CC], heterozygote comparison [CT versus CC], and dominant model [TT + CT versus CC], recessive model [TT versus CT + CC] and T allele versus C allele. For HIF1A 1772 C/T polymorphism, three studies were found with genotype information of CC and CT + TT. These three studies were included only to evaluate genotype contrast of dominant model [TT + CT vs. CC]. The genotype contrasts for the HIF1A 1790 G/A polymorphism were: homozygote comparison [AA versus GG], heterozygote comparison [GA versus GG] and dominant model [AA + GA versus GG], recessive model [AA versus GA + GG] and [G versus A allele].

Results and discussion

Study characteristics

In the meta-analysis of the HIF1A 1772 C/T polymorphism, ten different types of cancers consisted of 22 studies with 8149 cancer cases and 10,875 controls were included. The types of cancer included in these studies were prostate cancer, colorectal cancer, renal cell carcinoma, breast cancer, lung cancer, oral squamous cell carcinoma (OSCC), head-neck cancer, cervical cancer, bladder carcinoma and pancreatic cancer. For the following cancer types: head-neck, cervical, bladder and pancreatic only one study of each were found for the final review. So, these cancer types with single studies were incorporated in subgroup analysis as Other Cancers (Table 1).

Table 1 Characteristic of eligible studies included in meta-analysis of HIF1A 1772 C/T polymorphism
Full size table

For the meta-analysis of HIF1A 1790 G/A polymorphism, 19 studies with eleven different cancer types consisted of 4681 cancer cases and 5973 controls were included. The cancer types associated with this polymorphism were: renal cancer, prostate cancer, breast cancer, lung cancer, oral squamous cell carcinoma (OSCC), head-neck cancer, gastric cancer, hepatocellular carcinoma, lymph node metastasis, pancreatic cancer and colorectal cancer. For final review, only one study of each of the following cancer types was found: head-neck cancer, gastric cancer, hepatocellular carcinoma, lymph node metastasis, pancreatic cancer and colorectal cancer. These cancer types with single studies were incorporated in subgroup analysis as Other Cancers (Table 2).

Table 2 Characteristic of eligible studies included in meta-analysis of HIF1A 1790G/A polymorphism
Full size table

Association of the HIF1A 1772 C/T polymorphism with cancer risk

The pooled ORs for overall cancer suggested that the HIF1A 1772 C/T polymorphism was significantly associated with increasing cancer risk for the dominant model [TT + CT vs. CC: OR 1.30, 95 % CI (1.06-1.59), p-value: 0.0115] and [T vs. C allele: OR 1.32, 95 % CI (1.07-1.63), p-value: 0.0098] (Fig. 2).

Fig. 2
figure 2

Forest plot of HIF1A polymorphism 1772 C/T for overall cancer

Full size image

Subgroup analyses performed by cancer type

The subgroup analyses of prostate cancer, colorectal cancer, breast cancer and oral squamous-cell carcinoma suggested no significant association of the HIF1A 1772 C/T polymorphism. However, the subgroup analyses of renal cell carcinoma suggested that the HIF1A 1772 C/T polymorphism is significantly associated with lowering renal cell carcinoma risk in homozygote comparison [TT vs. CC: OR 0.27, 95 % CI (0.08-0.90), p-value:0.0335]. Interestingly, the results of subgroup analyses of lung cancer suggested that the HIF1A 1772 C/T polymorphism is highly associated with increasing lung cancer risk in homozygote comparison [TT vs. CC: OR 4.88, 95 % CI (2.42-9.84), p-value: <0.0001], recessive model [TT vs. CT + CC: OR 4.04, 95 % CI (2.02-8.08), p-value:<0.0001]. The subgroup analyses of Other Cancers suggested that the HIF1A 1772 C/T polymorphism is highly associated with increasing Other Cancer risk in homozygote comparison [TT vs. CC: OR 27.20, 95 % CI (5.04-146.78), p-value: 0.0001], heterozygote comparison [CT vs. CC: OR 2.16, 95 % CI (1.46-3.18), p-value: 0.0056], dominant model [TT + CT vs. CC: OR 1.92, 95 % CI (1.17-3.14), p-value: 0.0093], recessive model [TT vs. CT + CC: OR 17.5, 95 % CI (3.49-87.70), p-value: 0.0005] and [T vs. C allele: OR 2.42, 95 % CI (1.55-3.77), p-value: <0.0001] (Table 3).

Table 3 Meta-analysis of the HIF1A 1772 C/T polymorphism association with cancer
Full size table

Subgroup analyses by ethnicity group

The analyses data for the HIF1A 1772 C/T polymorphism suggested that there was no significant effect on the Caucasian population. However, the subgroup analyses of the Asian population suggested that the HIF1A 1772 C/T polymorphism was significantly associated with increasing cancer risk in homozygote comparison [TT vs. CC: OR 4.98, 95 % CI (2.66-9.31), p-value: <0.0001], heterozygote comparison [CT vs. CC: OR 1.30, 95 % CI (1.01-1.69), p-value: 0.0455], dominant model [TT + CT vs. CC: OR 1.41, 95 % CI (1.08-1.84), p-value: 0.0109], recessive model [TT vs. CT + CC: OR 4.28, 95 % CI (2.31-7.95), p-value:<0.0001] and [T vs. C allele: OR 1.43, 95 % CI (1.07-1.90), p-value: 0.0156] (Table 3). The subgroup analyses of mixed ethnic groups suggested that there were no significant association between HIF1A 1772 C/T polymorphism and cancer risk (Table 3).

Sources of heterogeneity

There were significant heterogeneity observed in the analyses of HIF1A 1772 C/T polymorphism for overall cancer heterozygote comparison [CT vs. CC: Q = 69.67, d.f = 18, p-value 0.0001, I^2 = 74.2 % (59.5 %-83.5 %)], dominant model [TT + CT vs. CC: Q = 90.25, d.f = 21, p <0.0001, I^2 = 76.7 % (65.1 %-84.5 %)], and [T vs. C allele: Q = 96.87, d.f = 18, p <0.0001, I^2 = 81.4 % (71.9 %-87.7 %). To detect the sources of heterogeneity subgroup analyses by cancer type and ethnicity group were performed. In the subgroup analyses by cancer type heterogeneity was significantly reduced. The results suggested that the studies in prostate cancer, renal cell carcinoma, lung cancer, Caucasian ethnicity and Asian ethnicity were the main sources of heterogeneity (Additional file 1).

Association of the HIF1A 1790 G/A polymorphism with cancer risk

The pooled ORs for overall cancer suggested that the HIF1A 1790 G/A polymorphism was significantly associated with increasing cancer risk for homozygote comparison [AA vs. GG: OR 5.10, 95 % CI (3.12-8.33), p-value: <0.0001, heterozygote comparison [GA vs. GG: OR 1.74, 95 % CI (1.20-2.52), p-value: 0.0033, dominant model [AA + GA vs. GG: OR 1.82, 95 % CI (1.26-2.62), p-value: 0.0014], recessive model [AA vs. GA + GG: OR 3.79, 95 % CI (2.34-6.15), p-value: <0.0001] and [A vs. G allele: OR 1.82, 95 % CI (1.31-2.52), p-value: 0.0003] (Fig. 3).

Fig. 3
figure 3

Forest plot of the HIF1A polymorphism 1790 G/A for overall cancer

Full size image

Subgroup analyses by cancer type

The analyzed data of prostate cancer suggested no significant association with the HIF1A 1790 G/A polymorphism. The subgroup analyses of renal cancer suggested that the HIF1A 1790 G/A polymorphism was significantly associated with increasing cancer risk for homozygote comparison [AA vs. GG: OR 5.11, 95 % CI (2.24-11.66), p-value: 0.0001], recessive model [AA vs. GA + GG: OR 3.05, 95 % CI (1.36-6.84), p-value: 0.0068] whereas the subgroup analyses of breast cancer showed that the HIF1A 1790 G/A polymorphism was significantly associated with decreasing cancer risk for [A vs. G allele: OR 0.30, 95 % CI (0.09-1.00), p-value: 0.0495]. The subgroup analyses of lung cancer suggested that the HIF1A 1790 G/A polymorphism was significantly associated with increasing cancer risk for homozygote comparison [AA vs. GG: OR 5.41, 95 % CI (2.74-10.69), p-value: <0.0001], heterozygote comparison [GA vs. GG: OR 1.76, 95 % CI (1.25-2.49), p-value: 0.0013], dominant model [AA + GA vs. GG: OR 2.20, 95 % CI (1.60-3.03), p-value:<0.0001], recessive model [AA vs. GA + GG: OR 4.51, 95 % CI (2.31-8.81), p-value:<0.0001] and [A vs. G allele: OR 2.31, 95 % CI (1.77-3.02), p-value: <0.0001]. Also, the subgroup analyses of oral squamous cell carcinoma (OSCC) suggested that the HIF1A 1790 G/A polymorphism was significantly associated with increasing cancer risk for homozygote comparison [AA vs. GG: OR 12.68, 95 % CI (1.43-112.64), p-value: 0.0227], heterozygote comparison [GA vs. GG: OR 4.69, 95 % CI (1.96-11.21), p-value: 0.0005], dominant model [AA + GA vs. GG: OR 5.17, 95 % CI (1.99-13.43), p-value: 0.0008], recessive model [AA vs. GA + GG: OR 10.12, 95 % CI (1.14-89.72), p-value: 0.0376] and [A vs. G allele: OR 5.00, 95 % CI (2.10-11.97), p-value: 0.0003] (Table 4). The subgroup analyses of Other Cancers suggested that the HIF1A 1790 G/A polymorphism is highly associated with increasing Other Cancer risk heterozygote comparison [GA vs. GG: OR 1.96, 95 % CI (1.05-3.65), p-value: 0.0336], dominant model [AA + GA vs. GG: OR 1.96, 95 % CI (1.05-3.67), p-value: 0.0341], and [A vs. G allele: OR 1.91, 95 % CI (1.06-3.44), p-value: 0.0306] (Table 4).

Table 4 Meta-analysis of the HIF1A 1790 G/A polymorphism association with cancer
Full size table

Subgroup analyses by ethnicity group

For Caucasian population, the analyzed data suggested that the HIF1A 1790 G/A polymorphism was highly associated with increasing cancer risk for homozygote comparison [AA vs. GG: OR 5.68, 95 % CI (2.57-12.58), p-value: <0.0001], recessive model [AA vs. GA + GG: OR 3.42, 95 % CI (1.57-7.45), p-value: 0.002]. For the Asian population, the subgroup analyses of ethnicity group suggested that the HIF1A 1790 G/A polymorphism was highly associated with increasing cancer risk for homozygote comparison [AA vs. GG: OR 4.76, 95 % CI (2.55-8.91), p-value: <0.0001], heterozygote comparison [GA vs. GG: OR 1.94, 95 % CI (1.38-2.72), p-value: 0.0001], dominant model [AA + GA vs. GG: OR 2.04, 95 % CI (1.44-2.87), p-value: <0.0001], recessive model [AA vs. GA + GG: OR 4.05, 95 % CI (2.18-7.51), p-value: <0.0001] and [A vs. G allele: OR 2.03, 95 % CI (1.46-2.81), p-value: <0.0001] (Table 4).

Sources of heterogeneity

There were significant heterogeneity observed in the analyses of HIF1A 1790G/A polymorphism for overall cancer heterozygote comparison [GA vs. GG: Q = 77.05, d.f = 18, p-value: <0.0001, I^2 = 76.6 % (63.8 %-84.9 %), dominant model [AA + GA vs. GG: Q = 79.66, d.f = 18, p-value: <0.0001, I^2 = 77.4 % (65.1 %-85.4 %)], and [A vs. G allele: Q = 71.09, d.f = 18, p-value: <0.0001, I^2 = 74.7 % (60.4 %-83.8 %)]. To detect the sources of heterogeneity subgroup analyses by cancer type and ethnicity group were performed. The results suggested that the studies in renal cell carcinoma, oral squamous cell carcinoma (OSCC), Caucasian ethnicity and Asian ethnicity were the main sources of heterogeneity (Additional file 2).

Publication bias

To investigate the evidence of publication bias of the HIF1A 1772 C/T polymorphism for T versus C allele and HIF1A 1790 G/A polymorphism for G versus A allele funnel plot were used. The conventionally constructed funnel plot (log odds ratio [log(OR] vs 1/standard error, 1/se) of HIF1A polymorphism 1772 C/T for T vs. C allele suggested that there was evidence of publication bias (Fig. 4). Also the funnel plot of HIF1A polymorphism 1790 G/A for A vs. G allele suggested that there was evidence of publication bias (Fig. 4). However, the Egger’s linear regression analyses suggested no evidence of significant publication bias in [T vs C allele: t = 1.83, d.f = 17, p-value 0.0847] for HIF1A 1772 C/T polymorphism. Also, for HIF1A 1790 G/A polymorphism results showed no significant evidence of publication bias in [A vs G allele: t = -1.87, d.f = 17, p-value 0.0787] (Additional file 3).

Fig. 4
figure 4

Funnel plot of HIF1A polymorphism (a) 1772 C/T for T allele vs. C allele and (b) 1790 G/A for A allele vs. G allele; showing visual evidence of publication bias

Full size image

Sensitivity analysis

Studies which were not in HWE were excluded to evaluate the stability of the acquired results. The statistical significance of the results was not shifted after omitting the studies which were not in HWE which confirmed the obtained results of the meta-analysis were stable and robust.

Conclusion

Results generated from this meta-analysis indicated that both 1772 C/T and 1790 G/A polymorphisms are significantly associated with increasing overall cancer risk. The subgroup analyses by cancer type showed that both 1772 C/T and 1790 G/A polymorphisms have significant association with lung cancer, whereas these two polymorphisms showed no significant association with prostate cancer. In oral squamous cell carcinoma (OSCC) subgroup analyses data showed that only 1790 G/A polymorphism has significant association whereas the HIF1A 1772 C/T polymorphism showed no significant association. However, the 1772 C/T polymorphism has indicated significantly decreased risk in renal cell carcinoma. Also, 1790 G/A polymorphism has increased the cancer risk significantly in both Caucasian and Asian ethnicity. Taken together all analyzed data, HIF1A could be a prognostic marker useful for early detection and diagnosis for cancers. In future, further experimental validations would be necessary to confirm the results.