Variant genotypes and haplotypes of the epidermal growth factor gene promoter are associated with a decreased risk of gastric cancer in a high‐risk Chinese population

Abstract

Epidermal growth factor (EGF), a ligand of the EGF receptor, plays a critical role in the development of gastric cancer. Genetic variants in its promoter region may influence transcription activity and contribute to gastric cancer predisposition. To test this hypothesis, we genotyped three EGF promoter polymorphisms (G61A, G‐1380A, and A‐1744G) in a case–control study of 675 gastric cancer cases and 704 cancer‐free controls. We found that the variant genotypes of EGF 61GA/AA were associated with a significantly decreased risk of gastric cancer (OR = 0.77, 95% CI = 0.61–0.95), when compared with wild‐type homozygote 61GG. In the combined analysis with all three loci of EGF, subjects carrying one or more variant loci had a significantly decreased risk of gastric cancer in a dose–response manner (adjusted OR = 0.58, 95% CI = 0.42–0.80 for subjects carrying one variant locus and OR = 0.46, 95% CI = 0.32–0.66 for those carrying two to three variant loci, respectively; trend test: χ² = 16.14, P < 0.001). Compared with the most common haplotype GGA, haplotypes AGA, GGG and GAA (each containing one variant allele) were associated with 33%, 29% and 34% significantly decreased risk of gastric cancer (adjusted OR = 0.67, 95% CI = 0.55–0.82 for AGA; OR = 0.71, 95% CI = 0.57–0.88 for GGG and OR = 0.66, 95% CI = 0.52–0.84 for GAA, respectively). Our findings indicate that variant genotypes and haplotypes of EGF promoter might play a role in gastric carcinogenesis. (Cancer Sci 2007; 98: 864–868)

Abbreviations: CHO

Chinese hamster ovary

CI

confidence interval

EGF

epidermal growth factor

EGFR

epidermal growth factor receptor

ELISA

enzyme‐linked immunosorbent assay

GBM

glioblastoma multiforme

HNF1

hepatocyte nuclear factor 1

LD

linkage disequilibrium

OR

odds ratio

PCR

polymerase chain reaction

PIRA

primer‐introduced restriction analysis

RFLP

restriction fragment length polymorphism

RTK

receptor of tyrosine kinase

SNP

single nucleotide polymorphism.

Gastric cancer is the second leading cause of cancer‐related mortality worldwide, accounting for ~700 000 deaths annually.^{( 1 )} Almost 40% of the gastric cancer cases occur in China with a remarkable geographic variation.^{( 2 )} Epidemiological studies suggest that some environmental exposures (e.g. salty diet, tobacco smoking and Helicobacter pylori infection) are important for the development of gastric cancer.^{( 3} , ^{4 )} However, accumulating evidence indicates that host factors and genetic alterations may also play an important role in gastric carcinogenesis through gene–environment interactions.^{( 5 )}

The EGF gene encodes a ligand for the EGFR, a receptor of tyrosine kinase. When binding with EGFR, EGF can activate multiple signaling pathways, regulating cell proliferation and differentiation.^{( 6} , ⁷ , ⁸ , ^{9 )} Studies showed that EGF and EGFR were highly expressed in gastric cancer,^{( 10} , ^{11 )} cooperating with H. pylori and inflammatory cytokines in gastric carcinogenesis.^{( 12} , ^{13 )}

Shahbazi et al. analyzed the EGF gene region from position –1350 to 164 and identified a G to A substitution at position 61 in the 5′ untranslated region, where the presence of the variant 61 A allele leads to a decreased in vitro EGF production in peripheral blood mononuclear cells.^{( 14 )} Therefore, it was hypothesized that this promoter variant might be associated with risk of gastric cancer. In a hospital‐based case‐control study in Japan (200 cases and 230 controls), Hanai et al. reported that EGF G61A (rs4444903) was involved not only in the occurrence but also in the progression of gastric cancer.^{( 15 )} However, this result was not supported by a later study in Japan (202 cases and 454 controls), although the main effect of EGF G61A was in the same direction.^{( 16 )} Because the single locus may not represent the functional region of the gene promoter, it is biologically possible that other promoter variants may be also involved in gastric cancer susceptibility through a haplotype effect. Therefore, we used the public SNP database (http://www.ncbi.nlm.nih.gov/) to select SNPs in the promoter region of EGF. Apart from G61A, we chose two SNPs, G‐1380A (rs11568835) and A‐1744G (rs3756261), with a minor allele frequency >0.05 in a Chinese population.

To evaluate the effects of these three EGF promoter SNPs and haplotypes in gastric cancer susceptibility, we carried out genotyping analyses for SNPs of G61A, G‐1380A and A‐1744G in 675 gastric cancer cases and 704 cancer‐free controls frequency‐matched to the cases on age and sex.

Materials and Methods

Study subjects.  A total of 675 incident gastric cancer patients were consecutively recruited from the Yang‐Zhong and Yi‐Xing counties, two areas with high gastric cancer mortality, in Jiangsu Province, China, between January 2003 and July 2005. All of the cases were of local ethnic Han Chinese residents. Residents with histopathologically diagnosed adenocarcinoma in a stable medical condition as determined by their physician and who were willing to participate in the study and provide blood samples were included in the study. Subjects were not restricted by age or sex. The response rate of the cases was 89.4% (675/755). The eligible controls were cancer‐free individuals who had lived in the study areas for at least 5 years without a self‐reported history of any kind of cancer and were selected from a name list of the residents in each selected village, according to the frequencies of age (±5) and gender groups of the cases (frequency‐matching). A total of 704 controls completed interviews and donated blood samples and the response rate was 84.8% (704/830).

After informed consent was obtained, each subject was personally interviewed face‐to‐face by trained interviewers using a standard questionnaire to obtain information on demographic data (e.g. age and gender) and related factors, including tobacco and alcohol use. After the interview, an approximately 5‐mL venous blood sample was collected from each subject. Individuals that smoked one cigarette per day for over one year were defined as smokers, and those that consumed three or more alcohol drinks a week for over 6 months were considered alcohol drinkers. IgG antibodies to H. pylori infection was detected by an ELISA according to the manufacturer's instructions (Anti‐H. pylori enzyme immunoassay, Bell Biotech Inc. Beijing, China). The study was approved by the Institutional Review Board of Nanjing Medical University.

Genotyping.  Genomic DNA was extracted from a leukocyte pellet by proteinase K digestion and was followed by phenol‐chloroform extraction and ethanol precipitation. The EGF G61A was genotyped by the PCR‐RFLP assay as described previously.^{( 14 )} Briefly, we used a pair of primers of 5′‐TGTCACTAAAGGAAAGGAGGT‐3′ (sense) and 5′‐TTCACAGAGTTTAACAGCCC‐3′ (antisense) to generate a 242 bp PCR product, which was digested by restriction enzymes of AluI (New England BioLabs, Beverly, MA, USA) and separated on a 3% agarose gel. The polymorphic (61A) allele produces four fragments of 102‐, 91‐, 34‐, and 15‐bp and the wild‐type (61G) results in 193‐, 34‐ and 15‐bp fragments. The SNPs of G‐1380A and A‐1744G of EGF were genotyped using a PIRA‐PCR assay,^{( 17 )} as we previously reported.^{( 18 )} For the EGF G‐1380A polymorphism, a mismatched C was introduced into the sense‐prime to replace T at –2 bp from the polymorphic site and a restriction site of HpaII was created (sense‐5′‐CCTTCCATTGCTGTCATCCG, antisense‐5′‐CATTGCTTTCTGGACTGAGTCAGA). The 148‐bp PCR products were then digested by HpaII (New England BioLabs) and separated on a 3% agarose gel. The wild‐type (–1380G) produced two fragments of 130‐ and 18‐bp and the polymorphic (–1380A) allele resulted in a single 148‐bp fragment. Similarly, for EGF A‐1744G, a mismatched C was introduced into the sense‐primer to replace T at –2 bp from the polymorphic sites and a PstI restriction site was created. The primers were sense‐5′‐AGAGCTACCCAACTGGGAAGGATCT and antisense‐5′‐GGCCTCGATGCGCTTCCGCTTCA. The 121‐bp PCR product was digested by PstI (New England BioLabs) and separated on a 3% agarose gel. The variant allele EGF–1744G produced two fragments of 97‐ and 24‐bp and the wild‐type allele –1744A produced only one fragment of 121‐bp.

Genotyping was carried out without knowing the subjects’ case or control status and approximately equal numbers of the cases’ and the controls’ samples were assayed in each 96‐well PCR plate with a positive control of a sample with a known heterozygous genotype. To further confirm the genotyping results, PCR products of the three loci with different genotypes were selected for direct sequencing using an automated sequencer (ABI model 377 genetic analysis; Perkin‐Elmer Applied Biosystems).

Statistical analyses.  Differences in the distributions of demographic characteristics, selected variables, and genotypes of the EGF variants between the cases and controls were evaluated using the χ² test. The associations between EGF genotypes and risk of gastric cancer were estimated by computing the ORs and their 95% CIs using logistic regression analyses for crude ORs and adjusted ORs when adjusting for age, sex, smoking and drinking status. We used the PHASE 2.0 program,^{( 19 )} to infer haplotype frequencies based on the observed EGF genotypes. The Hardy–Weinberg equilibrium was tested by a goodness‐of‐fit χ² test to compare the observed genotype frequencies with the expected ones among the control subjects. All statistical analyses were carried out with SAS 9.1.3 (SAS Institute, Cary, NC, USA).

Results

Among the 675 cases and 704 controls with DNA samples, the genotyping was successful for all three polymorphisms in 617 gastric cancer cases and 660 controls, resulting in an overall success rate of 92.6% (G61A, 94.8%; G‐1380A, 98.0%; A‐1744G, 97.0%). Therefore, a total of 617 cases and 660 controls with complete genotype information for the above three SNPs were included in the final analyses. The mean age was 60.5 years (±9.4 years, ranging 30–82 years) for the case group and 59.6 years (±10.4 years, ranging 24–84 years) for the control group (P =0.131). As shown in Table 1, there were no significant differences in terms of distributions on age (<60 and =60 years old) and gender between the cases and the controls (P = 0.275 and 0.433, respectively), suggesting that our frequency matching of the demographic characteristics was satisfactory. In addition, there were no significant differences between the cases and the controls in smoking and drinking status (OR = 1.01, 95% CI = 0.81–1.27, P = 0.921 for smoking and OR = 0.80, 95% CI = 0.63–1.01, P = 0.051 for alcohol use, respectively). However, the positive rate of IgG antibodies against H. pylori was higher in control subjects (61.0%) than in gastric cancer patients (53.1%) (P = 0.020) (Table 1).

Table 1.

Distribution of selected demographic variables and risk factors in gastric cancer cases and controls

Variable	Cases (n = 617)		Controls (n = 660)		P‐value*
	n	%	n	%
Age (years)					0.275
<60	271	43.9	310	47.0
≥60	346	56.1	350	53.0
Sex					0.433
Male	424	68.7	440	66.7
Female	193	31.3	220	33.3
Tobacco use					0.921
Never	376	60.9	404	61.2
Tobacco user	241	39.1	256	38.8
Alcohol use					0.051
Never	411	66.6	405	61.4
Alcohol user	206	33.4	255	38.6
H. pylori †					0.020
Negative	150	46.9	250	39.0
Positive	170	53.1	391	61.0

^* Two‐sided χ² test.

^{† †} 320 gastric cancer cases and 641 controls were detected.

The allele and genotype distributions of EGF G61A, G‐1380A and A‐1744G polymorphisms in the cases and controls are shown in Table 2. The observed genotype frequencies for these three SNPs were all in Hardy–Weinberg Equilibrium in the controls (P = 0.407, 0.088 and 0.119 for EGF G61A, G‐1380A and A‐1744G, respectively). In the single locus analyses, none of the three polymorphisms achieved significant differences in the genotype distributions between the cases and the controls (P = 0.062, 0.445, and 0.054 for EGF G61A, G‐1380A and A‐1744G, respectively). The logistic regression analyses revealed that the 61GA heterozygote was associated with a significantly reduced risk of gastric cancer (adjusted OR = 0.78, 95% CI = 0.62–0.99) and the 61AA homozygote was associated with a non‐significantly decreased risk (adjusted OR = 0.68, 95% CI = 0.44–1.05), compared with the 61GG wild‐type homozygote. When we combined the variant genotypes (61GA/AA) assuming a codominant genotype effect, the combined 61GA/AA variant genotypes were associated with a significantly decreased risk of gastric cancer (adjusted OR = 0.77, 95% CI = 0.61–0.95). For EGF G‐1380A and A‐1744G SNPs, the combined variant genotypes of both –1380GA/AA and –1744AG/GG were associated with a non‐significantly decreased risk (adjusted OR = 0.85, 95% CI = 0.66–1.08 for –1380GA/AA and OR = 0.81, 95% CI = 0.65–1.02 for –1744AG/GG, respectively), compared with their wild‐type homozygotes, respectively. Specifically, instead of the EGF–1744GG variant homozygote, the –1744AG heterozygote was associated with a significantly decreased risk of gastric cancer with adjusted OR and 95% CI of 0.77 and 0.61–0.98, compared with the –1744AA wild‐type genotype (Table 2).

Table 2.

Logistic regression analyses of associations between epidermal growth factor promoter polymorphisms and risk of gastric cancer

Genotype	Cases (n = 617)		Controls (n = 660)		Crude OR (95% CI)	Adjusted OR † (95% CI)
	n	%	n	%
EGF G61A
GG	333	54.0	314	47.6	1.00	1.00
GA	242	39.2	289	43.8	0.79 (0.63–0.99)	0.78 (0.62–0.99)
AA	42	6.8	57	8.6	0.70 (0.45–1.07)	0.68 (0.44–1.05)
GA/AA	284	46.0	346	52.4	0.77 (0.62–0.96)	0.77 (0.61–0.95)
A allele	326	26.4	403	30.5
EGF G‐1380A
GG	451	73.1	462	70.0	1.00	1.00
GA	143	23.2	173	26.2	0.85 (0.66–1.10)	0.83 (0.64–1.08)
AA	23	3.7	25	3.8	0.94 (0.53–1.69)	0.95 (0.53–1.70)
GA/AA	166	26.9	198	30.0	0.86 (0.67–1.10)	0.85 (0.66–1.08)
A allele	189	15.3	223	16.9
EGF A‐1744G
AA	408	66.1	403	61.1	1.00	1.00
AG	181	29.3	234	35.5	0.76 (0.60–0.97)	0.77 (0.61–0.98)
GG	28	4.5	23	3.5	1.20 (0.68–2.12)	1.25 (0.71–2.22)
AG/GG	209	33.9	257	38.9	0.80 (0.64–1.01)	0.81 (0.65–1.02)
G allele	237	19.2	280	21.2
Combined analysis ‡
No variant locus	117	19.0	74	11.2	1.00	1.00
One variant locus	343	55.6	373	56.5	0.58 (0.42–0.81)	0.58 (0.42–0.80)
Two to three variant loci	157	25.5	213	32.3	0.47 (0.33–0.67)	0.46 (0.32–0.66)

^† Adjusted for age, sex, smoking status and drinking status.

^‡ Assuming dominant genetic model in each locus, that is, heterozygote and variant homozygote versus wild type homozygote. CI, confidence interval; EGF, epidermal growth factor gene; OR, odds ratio.

To evaluate the combined effects of the three EGF promoter SNPs on gastric cancer risk, we defined the number of heterozygotes or variant homozygotes of the three loci as the number of variant locus. As shown in Table 2, compared with the subjects carrying the wild‐type genotypes for all three loci, those carrying one or more variant loci had a significantly decreased risk of gastric cancer in a dose–response manner (adjusted OR = 0.58, 95% CI = 0.42–0.80 for subjects carrying one variant locus; and adjusted OR = 0.46, 95% CI = 0.32–0.66 for those with two to three variant loci; trend test: χ² = 16.14, P < 0.001) (Table 2).

In the LD analyses, we found that all of the three loci were in LD with each other (χ² = 170.4, P < 0.001 and D′ = 0.93 for G61A and G‐1380A; χ² = 187.5, P < 0.001 and D′ = 0.85 for G61A and A‐1744G; χ² = 88.8, P < 0.001 and D′ = 0.84 for G‐1380A and A‐1744G), suggesting that there might be haplotype effects among these three SNPs. When we combined these three loci and carried out the haplotype inference using the PHASE 2.0 program, totally seven possible haplotypes were derived from their known genotypes (Table 3). Among them, four common (>10%) haplotypes (GGA, AGA, GGG, GAA) represented 97.6% of the chromosomes for cases and 98.4% of that for controls, and the haplotype distribution between the cases and the controls was statistically different (χ² = 31.72, d.f. = 6, P < 0.001). Specifically, haplotype AGA, GGG and GAA, each containing one of the three variant alleles, were less common in the cases (AGA: 0.247, GGG: 0.177 and GAA: 0.137) than in the controls (0.293, 0.198 and 0.164, respectively) (P < 0.001 for AGA, P = 0.002 for GGG and P < 0.001 for GAA, respectively). Compared with the most common haplotype GGA (consisting of the common allele from each polymorphic site), the AGA, GGG and GAA haplotypes were associated with 33%, 29% and 34% significantly decreased risk of gastric cancer (adjusted OR = 0.67, 95% CI = 0.55–0.82 for AGA; OR = 0.71, 95% CI = 0.57–0.88 for GGG and OR = 0.66, 95% CI = 0.52–0.84 for GAA, respectively).

Table 3.

Frequencies of inferred haplotypes based on observed genotypes in gastric cancer cases and cancer‐free controls

EGF haplotypes			Allele frequencies*				OR (95% CI)
G61A alleles	G‐1380A alleles	A‐1744G alleles	Cases (n = 1234)		Controls (n = 1320)
			N%	%	N	%
G	G	A	512	41.5	435	32.9	1.00
A	G	A	305	24.7	386	29.3	0.67 (0.55–0.82)
G	G	G	218	17.7	262	19.8	0.71 (0.57–0.88)
G	A	A	169	13.7	217	16.4	0.66 (0.52–0.84)
Others †			30	2.4	20	1.6	1.27 (0.71–2.28)

^* P < 0.001 for haplotypes distribution among cases and controls.

^† Including AGG, GAG and AAA haplotyope (haplotype frequencies < 0.05). CI, confidence interval; EGF, epidermal growth factor gene; OR, odds ratio.

The dichotomized genotypes (one or more variant loci versus no variant locus) were further examined for subgroups by selected variables according to the number of variant loci. As shown in Table 4, the decreased risk associated with the combined genotypes with one or more variant loci was more pronounced in subjects who were non‐drinkers (OR = 0.49, 95% CI = 0.33–0.73) and carriers of H. pylori (OR = 0.51, 95% CI = 0.31–0.85). However, there were no significant differences in the magnitude of the associations between the combined genotypes and gastric cancer risk for subjects with different age, gender and smoking status (Table 4). In addition, we also did not find any significant gene–environment interactions in relation to risk of gastric cancer (data not shown).

Table 4.

Stratified analyses between the combined genotypes of epidermal growth factor promoter polymorphisms and gastric cancer risk

Variant loci	Cases (n = 617)		Controls (n = 660)		Adjusted OR* (95% CI)		P‐value*
	0	≥1	0	≥1	0	≥1
	n (%)	n (%)	n (%)	n (%)
All subjects	117 (19.0)	500 (81.0)	74 (11.2)	586 (88.8)	1.00	0.53 (0.39–0.73)	<0.001
Age (years)
<60	51 (18.8)	220 (81.2)	40 (12.9)	270 (87.1)	1.00	0.63 (0.40–0.99)	0.050
≥60	66 (19.1)	280 (80.9)	34 (9.7)	316 (90.3)	1.00	0.45 (0.29–0.70)	<0.001
Sex
Male	76 (17.9)	348 (82.1)	53 (12.1)	387 (88.0)	1.00	0.61 (0.42–0.90)	0.012
Female	41 (21.2)	152 (78.8)	21 (9.6)	199 (90.5)	1.00	0.40 (0.22–0.70)	0.002
Tobacco use
Never	71 (18.9)	305 (81.1)	43 (10.6)	361 (89.4)	1.00	0.50 (0.33–0.75)	0.001
Tobacco user	46 (19.1)	195 (80.9)	31 (12.1)	225 (87.9)	1.00	0.58 (0.35–0.96)	0.032
Alcohol use
Never	78 (19.0)	333 (81.0)	42 (10.4)	363 (89.6)	1.00	0.49 (0.33–0.73)	0.001
Alcohol user	39 (18.9)	167 (81.1)	32 (12.6)	223 (87.5)	1.00	0.60 (0.35–1.01)	0.053
H. pylori †
Negative	26 (17.3)	124 (82.7)	31 (12.4)	219 (87.6)	1.00	0.69 (0.39–1.24)	0.212 †
Positive	32 (18.8)	138 (81.2)	43 (11.0)	348 (89.0)	1.00	0.51 (0.31–0.85)	0.010 †

^* Adjusted for age, sex, smoking and drinking status.

^† 320 gastric cancer cases and 641 controls were detected for Helicobacter pylori infection. CI, confidence interval; OR, odds ratio.

Discussion

In this population‐based case‐control study, we investigated the associations of three SNPs in the promoter region of EGF with risk of gastric cancer in a high‐risk Chinese population. We found that the variant genotypes of EGF 61GA/AA, –1380GA/AA and –1744AG/GG were associated with a decreased risk of gastric cancer, compared with their wild‐type homozygotes. In the combined analyses with these three loci of EGF, the more variant loci of three promoter SNPs of EGF there were, the lower risks of gastric cancer were observed. Compared with the most common haplotype GGA, haplotypes AGA, GGG and GAA (each containing one variant allele) were associated with a significantly decreased risk of gastric cancer. To the best of our knowledge, this is the first study that investigated the association of these three variant genotypes and haplotypes in the promoter region of EGF and the risk of gastric cancer.

EGF regulates cell proliferation, differentiation and survival of normal cells by binding and activating EGFR.^{( 20 )} The EGFR and the EGF‐family of peptide growth factor play an important role in the development and progression of diverse carcinoma types, including gastric cancer.^{( 11} , ¹³ , ^{21 )} Dysregualtion of EGFR pathway may occur through mutations in EGFR and EGF, resulting in constitutive activation.^{( 20} , ^{21 )} Therefore, genetic variants in the EGF gene were hypothesized to play a critical role in carcinogenesis. Shahbazi et al. firstly identified a G to A substitution at position 61 in the 5′ untranslated region of the EGF gene, and they found that the presence of the 61G allele leads to increased EGF production in vitro in peripheral blood mononuclear cells.^{( 14 )} Furthermore, they showed that the 61G allele conferred a 2.7‐fold increased risk of melanoma,^{( 14 )} although this was not supported by later studies.^{( 22} , ²³ , ²⁴ , ^{25 )} In addition, Bhowmick et al. found that EGF expression was significantly lower in tumors with 61AA genotype when compared with 61GA (P < 0.011) or 61GG (P < 0.004) and the frequency of G allele in GBM patients was significantly greater than that in normal controls (P < 0.001).^{( 26 )} More recently, Vauleon et al. showed that the EGF 61G allele could result in a 40% higher luciferase activity in CHO cells than 61A allele in an in vitro study, because the G allele might enhance the affinity of the HNF1 transcription factor at position 61.^{( 27} , ^{28 )} Taken together, this evidence suggests that the EGF G61A polymorphism might be a functional variant in modifying the risks for different kinds of cancers.

In a small case‐control study of gastric cancer with 200 cases and 230 controls in a Japanese population, Hamai et al. first investigated the association of EGF G61A with gastric cancer risk and found that the 61GA/AA genotypes were associated with a significantly 40% decreased risk of gastric cancer (OR = 0.60, 95% CI = 0.41–0.88).^{( 15 )} In another case‐control study in a Japanese population (202 cases and 454 controls), Goto et al. also showed a protective but not significant effect on gastric cancer for the variant 61A allele.^{( 16 )} In the present study with relatively a large sample size, we found that variant 61A allele of EGF promoter was associated with a 23% significantly decreased risk of gastric cancer in this high‐risk Chinese population, which was consistent with the results reported by Hamai.^{( 15 )} In addition, we found that the variant alleles of the EGF promoter, –1380A and –1744G, were also associated with a decreased risk of gastric cancer with borderline significance, indicating that they might play a similar role in gastric carcinogenesis through linkage disequilibriums between these loci. It is also possible, however, that these three polymorphisms might be in linkage disequilibrium with other putative etiological genetic variants.

Interestingly, we found for the first time that EGF haplotypes AGA, GGG and GAA were associated with a significantly decreased risk of gastric cancer when compared with the most common haplotype GGA (containing three wild‐type alleles). These three haplotypes, each containing one variant allele, could represent the corresponding variant alleles of the polymorphic sites in this region, suggesting that these three SNPs might contribute to gastric cancer susceptibility in an allele‐specific manner. However, these three haplotypes incorporating all three loci showed a stable and additive effect on a decreased risk of gastric cancer as compared with any single locus (with corresponding variant allele), which might result from the fact that the effect of haplotypes was greater than a single locus. Recent studies supported that haplotype analyses might be of advantage in investigating the disease associations, compared with single locus analyses.^{( 29} , ³⁰ , ^{31 )}

In the two Japanese case‐control studies, the allele frequencies of EGF 61A in controls were 0.304 and 0.313, respectively,^{( 15} , ^{16 )} which were comparable with that in this Chinese population (0.305). However, 61A allele frequencies were significantly higher in Caucasian populations (above 0.50),^{( 14} , ²² , ²³ , ²⁴ , ²⁵ , ^{26 )} and therefore, population stratification should be paid attention to in further studies. Because this was the first molecular epidemiological study that investigated the association between G‐1380A and A‐1744G polymorphisms and cancer susceptibility, no comparison between published studies could be made at this time. According to the NCBI dbSNP database (http://www.ncbi.nlm.nih.gov/), the allele frequencies in 44 unrelated Han Chinese in Beijing and 24 Han Chinese in Los Angeles were 0.068 and 0.188 for –1380A allele and 0.239 and 0.125 for –1744G allele, respectively, which was not different from our Chinese population: –1380A allele (0.169) and –1744G allele (0.212).

One limitation of the current study is that detailed information on clinical staging, metastasis and survival of gastric cancer were not available, which restricted our further analysis on the role of EGF in cancer progression and prognosis. Another limitation is the lack of the related phenotypic and functional evaluations on the EGF SNPs, which limited our inquiry into the functional consequence of these variants. Furthermore, the plasma positive rate of IgG antibodies to H. pylori by ELISA may not represent the real H. pylori infection before the disease occurred and therefore we obtained negative association between H. pylori infection and gastric cancer risk. A possible explanation was that the loss of H. pylori from the stomach occurred during gastric carcinogenesis or after the long‐term antibiotic treatment. It was also possible that the immune response to H. pylori infection was reduced with the development of gastric cancer.^{( 4} , ^{32 )} Finally, occupational exposure and certain dietary components might act as potential confounders in the analysis. Unfortunately, information on these factors in our case‐control study was not available.

In summary, our results indicate that genetic variants in the promoter region of EGF may play a role in gastric cancer susceptibility. To explore the exact biological mechanism of EGF genotypes and haplotypes and their interaction with environmental factors, further functional studies and larger well‐designed prospective studies are warranted.