Sequence variations in the DNA repair gene XPD and risk of lung cancer in a Chinese population


  • Gang Liang,

    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for more papers by this author
  • Deyin Xing,

    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for more papers by this author
  • Xiaoping Miao,

    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for more papers by this author
  • Wen Tan,

    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for more papers by this author
  • Chunyuan Yu,

    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for more papers by this author
  • Wenfu Lu,

    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for more papers by this author
  • Dongxin Lin

    Corresponding author
    1. Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    • Department of Etiology and Carcinogenesis, Cancer Institute, Chinese Academy of Medical Sciences, Beijing 100021, China
    Search for more papers by this author
    • Fax: +86-10-67722460


Variation in DNA repair capacity, which is believed to be largely determined by genetic traits, is linked to risk of certain cancers. The Asp312Asn and Lys751Gln polymorphisms in the xeroderma pigmentosum complementary group D (XPD) gene may alter DNA repair capacity. We thus examined the hypothesis that these 2 XPD polymorphisms are associated with risk of lung cancer via a large hospital-based, case-control study among Chinese. The study subjects consisted of 1,006 patients with primary lung cancer and 1,020 age- and sex-matched population controls. XPD genotypes were determined using PCR-RFLP techniques, and the associations between genotypes and risk of lung cancer were estimated by odds ratios (ORs) and their 95% confidence intervals (CIs) calculated by unconditional logistic regression. Subjects homozygous for the 312Asn/Asn genotype had an increased risk of lung cancer (adjusted OR = 10.33, 95% CI = 1.29–82.50) compared with subjects homozygous for the 312Asp/Asp genotype. The 751Gln/Gln genotype was also associated with increased risk for the cancer compared with the 751Lys/Lys genotype (adjusted OR = 2.71, 95% CI = 1.01–7.24). Stratification analysis revealed that the increased risk was mainly confined to lung squamous cell carcinoma, with the ORs being 20.50 (95% CI = 2.25–179.05) for the 312Asn/Asn genotype and 4.24 (95% CI = 1.34–13.38) for the 751Gln/Gln genotype, respectively. Haplotype analysis with the 2 polymorphisms suggested these polymorphisms might be in linkage disequilibrium with a different causative locus or act together with other functional variants in or close to the XPD locus. © 2003 Wiley-Liss, Inc.

The DNA repair system plays an important role in protecting against mutagenesis and carcinogenesis. It has been documented that the defect in DNA repair causes several hereditary cancer syndromes1 and the development of some common sporadic cancers may also associated with reduced DNA repair capacity.2, 3, 4 Accumulating evidence indicates that variation in DNA repair capacity is likely to be largely determined by genetic traits. The defect of DNA repair often results from gene mutations. However, single nucleotide polymorphisms (SNPs), when located within the coding and/or regulating regions of the gene, can also spoil DNA repair capacity due to the amino acid substitution or diminished protein expression. XPD (xeroderma pigmentosum complementary group D), an important DNA repair protein, encodes an evolutionarily conserved ATP-dependent helicase that participates in both nucleotide excision repair and basal transcription as part of the transcription factor TFIIH.5 Mutations at different sites in XPD that destroy XPD protein function cause 3 severe syndromes: Cockayne's syndrome, trichotiodystrophy and xeroderma pigmentosum, which exhibits a >1,000-fold incidence of sun-induced skin cancer and elevated risk of internal cancers.6, 7 Several SNPs have also been identified in the XPD locus. Among them, a G-to-A transition in codon 312 of exon 10 results in an Asp→Asn substitution in an evolutionarily conserved region, and another transversion, A-to-C in codon 751 of exon 23, produces a Lys→Gln substitution.8 The 2 sites are reported to be in linkage disequilibrium and appear to have phenotypic significance, although contradictory results exist regarding which allele is associated with impaired DNA repair capacity.4, 9, 10, 11, 12, 13, 14

Since XDP is one of the important components in the nucleotide excision repair (NER) and NER is the most flexible pathway that has the ability to remove a broad range of DNA damage such as BPDE-DNA adducts induced by benzo(a)pyrene,15, 16 a major constituent of tobacco smoking, the impact of these 2 genetic variations in XPD on risk of cancer in addition to skin cancer has been attracting research interest. Some case-control studies have been conducted in different ethnic populations to investigate the associations between the aforementioned polymorphisms in XPD and risk of cancers at various sites, including lung4, 11, 17, 18, 19, 20, 21, 22 head and neck,23 skin24, 25, 26 and bladder.27 However, the results from these molecular epidemiological studies, as in the phenotype analysis studies, are confusing rather than conclusive. For instance, of the 8 studies on lung cancer, 4 showed increased risk related to the rare alleles, i.e. 312Asn and 751Gln,4, 11, 20, 22 but the rest either suggested the common alleles (312Asp or 751Lys) as risk alleles17, 19 or showed null association.18, 21

Although the reasons for the discrepant results remain to be discovered, most earlier studies have suffered from a small sample size, racially diverse study populations or inappropriate selection of control subjects in terms of matching the cases. This has raised a requirement for more careful studies with a large sample size. In this study, we recruited 1,006 incident lung cancer cases and 1,020 healthy population controls, all of homogenous Han Chinese origin, and genotyped the aforementioned 2 polymorphisms in XPD to further test the hypothesis that these genetic polymorphisms are associated with risk of lung cancer.


Study subjects

Lung cancer patients were selected consecutively from an ongoing molecular epidemiological study of lung cancer conducted in the Department of Etiology and Carcinogenesis at the Cancer Institute/Hospital, Chinese Academy of Medical Sciences (Beijing). The 1,006 cases were patients with primary lung cancer recruited from January 1997 to June 2002 at our institution, who were newly diagnosed, histopathologically confirmed and previously untreated (by radiotherapy or chemotherapy). There were no age, stage nor histology restrictions and all the subjects were unrelated ethnic Han Chinese. The exclusion criteria included previous cancer, metastasized cancer, previous radiotherapy or chemotherapy. A portion of the cases were participants in a pilot study on XPD and lung cancer as described previously.20 Population controls included 1,020 subjects who were randomly selected from a pool of cancer-free subjects recruited from a nutritional survey conducted in the same region.28 The controls were frequency-matched to the cases on age (±5 years), sex and ethnicity. At recruitment, informed consent was obtained from each subject and each participant was then interviewed to collect detailed information on demographic characteristics and lifetime history of tobacco use. The participation response rate of this study was 91% among patients and 87% among controls. The study was approved by the Institutional Review Board of the Chinese Academy of Medical Sciences Cancer Institute.

XPD genotyping

Genomic DNA from control and most of case subjects was extracted from the leukocyte pellet obtained from each blood sample's buffy coat by centrifugation of 2 ml of whole blood. Approximate 30% DNA samples from cases were isolated from surgically resected normal tissues adjacent to the tumor of lung cancer patients. The XPD genotypes at the Asp312Asn and Lys751Gln were analyzed by PCR-restriction fragment length polymorphism (RFLP) assays as described previously.20 The restriction enzyme StyI (New England Biolabs, Beverly, MA) was used to type the Asp312Asn polymorphism. The wild-type homozygotes (Asp/Asp) had 2 DNA bands (507 and 244 bp), the homozygous variant genotype (Asn/Asn) resulted in 3 bands (474, 244 and 33 bp) and heterozygote (Asp/Asn) displayed all 4 bands (507, 474, 244 and 33 bp). The restriction enzyme PstI (New England Biolabs) was used to type the Lys751Gln polymorphism. PstI digestion resulted in 2 fragments of 290 and 146 bp for the wild-type homozygotes (Lys/Lys); 3 fragments of 227, 146 and 63 bp for the variant homozygotes (Gln/Gln) and 4 fragments at 290, 227, 146 and 63 bp for the heterozygotes (Lys/Gln). Since the Lys751Gln polymorphism can also be typed by the restriction enzyme MboII (New England Biolabs),12 a 10% masked, random samples of cases and controls was tested again by this enzyme to compare with the results obtained by PstI digestion. MboII digestion produced 2 fragments of 219 and 217 bp for the Lys/Lys genotype; 3 fragments of 436, 219 and 217 bp for the Lys/Gln heterozygotes; and 1 fragment of 436 bp for the Gln/Gln variant genotype. The restricted products were analyzed by electrophoresis in 3% agarose gels containing ethidium bromide. Furthermore, the genotypes of Lys751Gln polymorphism identified by PstI or MboII digestion were further confirmed by DNA sequencing. Three different allelic PCR products were directly analyzed with an ABI PRISM 377 automatic sequencer using a dye terminator sequencing kit and sequences were compared with a published XDP sequences.29

Ten percent of random samples were tested twice by different persons and the results were 100% concordant. The 312Asp and 751Lys alleles are in linkage disequilibrium, that is, carriers with the XPD312Asp allele tend to have the 751Lys allele.4, 14, 20, 24 Therefore, after we had obtained genotype data from all subjects, we reanalyzed the samples with the 312Asp allele but not 751Lys and with 312Asn but not 751Gln to ensure the genotyping. All results were identical to the first test.

Statistical analysis

A chi-square test was used to evaluate the differences in select demographic variables, smoking status, pack-years smoked and the distribution of the XPD genotypes and haplotypes between case patients and control subjects. Those who had smoked less than 100 cigarettes in their lifetimes were defined as nonsmokers as traditionally used in epidemiological studies; otherwise they were considered as smokers. Information was collected on the number of cigarettes smoked per day, the age at which the subjects started smoking and the age at which ex-smokers stopped smoking. Hardy-Weinberg equilibrium was tested by a goodness-of-fit chi-square test to compare the observed genotype frequencies to the expected genotype frequencies among the control subjects. Haplotype frequencies and linkage disequilibrium coefficient were estimated using EH (EH-plus) software.30 The associations between XPD genotypes and risk of lung cancer were estimated by computing the odds ratios (ORs) and their 95% confidence intervals (CIs) from both univariate and multivariate logistic regression analyses. The ORs were also adjusted for age, gender and smoking status because of the use of frequency matching. These statistical analyses were performed with Statistical Analysis System software (Version 6.12, SAS Institute, Cary, NC).


The distributions of demographic characteristics of the 1,006 lung cancer cases and 1,020 control subjects were summarized in Table I. The cases and controls appeared to be matched adequately on sex and age. The mean age was 58.2 (± 6.6) years for the cases, which was not significantly different from 57.6 (± 8.9) years for the controls. However, smokers were overrepresented among lung cancer patients compared with control subjects (65.71% vs. 48.92%; χ2 = 57.6, p < 0.0001). Furthermore, 66.57% of smokers among the cases were heavy smokers (pack-years > 25); this percentage was significantly higher than that (49.70%) among the controls (χ2 = 24.0, p < 0.0001). Of the 1,006 lung cancer cases, 461 (45.8%) were classified as squamous cell carcinoma (SCC), 300 (29.8%) as adenocarcinomas and 245 (24.4%) as others carcinomas, including undifferentiated cancers, bronchioalveolar carcinomas and mixed cell carcinomas.

Table I. Distribution of Select Characteristics in Lung Cancer Cases and Controls
VariableCase patients (n = 1006)Control subjects (n = 1020)p-value1
  • 1

    Two-sided χ2 test.

  • 2

    SCC, squamous cell carcinoma; other includes undifferentiated cancers (n = 97), bronchioalveolar carcinomas (n = 88) and mixed cell carcinomas (n = 60).

Sex    0.583
Age (years)    0.153
 > 70110(10.9)85(8.3) 
Smoking status    0.000
Pack-years smoked    0.000
 ≤ 25221(33.4)251(50.3) 
 > 25440(66.6)248(49.7) 
Histological type2     

The XPD genotypes identified by PCR-RFLP analysis were readily discerned. Figure 1 shows representative gel pictures for genotyping of the Lys751Gln polymorphism. We utilized 2 restriction enzymes, PstI (Fig. 1a) or MboII (Fig. 1b), to identify the genotypes. The results obtained with the 2 methods were concordant for all repeated samples (n = 202). In addition, DNA sequencing (Fig. 2) confirmed the A→C transversion at this polymorphic site. These results clearly demonstrate that the genotyping methods we used are completely reliable.

Figure 1.

A representative gel picture showing PCR-RFLP analysis of XPD Lys751Gln polymorphism in human genomic DNAs with the restriction enzyme PstI (A) or MboII (B). Lanes 1,3, and 6–9, Lys/Lys genotype; Lanes 2,10,and 11, Lys/Gln genotype and Lanes 4 and 5, Gln/Gln genotype. The results obtained with the 2 methods were completely concordant.

Figure 2.

Partial DNA sequences of 3 different allelic PCR products analyzed directly with an ABI PRISM 377 automatic sequencer showing an A-to-C transition at codon 751 of the XPD gene. (a) Lys/Lys genotype, (b) Lys/Gln genotype and (c) Gln/Gln genotype.

Genotyping data (Table II) show that the allele frequencies for XPD 312Asn and 751Gln were 0.065 and 0.087 among the control population compared with 0.073 and 0.090 among lung cancer cases. The observed genotype frequencies of both Asp312Asn and Lys751Gln sites among controls were not significantly different from the expected frequencies, indicating that they were in Hardy-Weinberg equilibrium. However, the observed genotype frequencies among cases did not fit Hardy-Weinberg equilibrium. The distributions of these XPD genotypes were then compared, respectively, among cases and controls. It was found that the frequencies of 312 Asp/Asp and Asp/Asn genotypes among cancer cases were almost the same as those among controls (86.48 and 12.43% vs. 87.16 and 12.75%, P = 0.948). However, the frequencies of Asn/Asn homozygotes, even though were extremely rare, differed significantly among cases and controls (1.09% vs. 0.09%, P = 0.009). Similarly, although the homozygous wild-type and heterozygous genotypes (Lys/Lys and Lys/Gln) at the Lys751Gln site were not significantly different among cases and controls, the rare homozygous variant genotype (Gln/Gln) appeared to be more prevalent among cases than controls (1.39% vs. 0.59%, p = 0.067). By using logistic regression analysis, we evaluated the association between the 321Asn/Asn or 751Gln/Gln genotype and risk of lung cancer. Compared with 321Asp/Asp homozygotes, 321Asn/Asn homozygotes had a more than tenfold increased risk for overall lung cancer (adjusted OR, 10.33; 95% CI, 1.29– 82.50). The homozygous variants of the Lys751Gln site also had an increased risk for the disease although the association was less strong (adjusted OR, 2.71; 95% CI, 1.01–7.24). However, there was no evidence showing any association between the heterozygous genotypes of both Asp321Asn and Lys751Gln sites at the XPD gene and risk of the cancer.

Table II. XPD Genotypes of Case Patients and Control Subjects and Their Association With Risk of Overall Lung Cancer
XPD genotypeCase patients (n = 1006)Control subjects (n = 1020)Crude OR (95% CI)Adjusted OR (95% CI)1
  • 1

    ORs were adjusted for sex, age, smoking status and pack-years smoked in a logistic regression model.

 Asp/Asn125(12.43)130(12.75)0.98 (0.76–1.28)1.03 (0.82–1.35)
 Asn/Asn11(1.09)1(0.09)11.24 (1.45–87.24)10.33 (1.29–82.50)
 Lys/Gln153(15.21)166(16.27)0.93 (0.73–1.18)0.95 (0.74–1.22)
 Gln/Gln14(1.39)6(0.59)2.36 (0.90–6.17)2.71 (1.01–7.24)

Since lung cancers are derived from different cell types and carcinogenesis of different subtypes of lung cancer may initiated by diverse DNA damage, risk related to the XPD polymorphisms was further evaluated among SCCs, adenocarcinomas and other carcinomas of the lung (Table III). It was found that increased risk associated with the polymorphisms was only evident among lung SCCs, with the adjusted OR being 20.05 (95% CI, 2.25–179.05) for the 321Asn/Asn genotype and 4.24 (95% CI, 1.34–13.38) for the 751Gln/Gln genotype, respectively. No patient with lung adenocarcinma was 321Asn/Asn homozygote, although 1 of these patients carried the 751Gln/Gln genotype. The potential interactions between these 2 genetic polymorphisms and smoking on risk of lung cancer were also analyzed in a logistic regression model. However, no effect was observed (data not shown), apparently due to the small numbers involved.

Table III. Risk of Subtypes of Lung Cancer Related to the XPD Genotypes
 XPD Asp315Asn genotypeXPD Lys751Gln genotype
Asp/Asp or Asp/AsnAsn/AsnCrude OR (95% CI)Adjusted OR1 (95% CI)Lys/Lys or Lys/GlnGln/GlnCrude OR (95% CI)Adjusted OR1 (95% CI)
  • 1

    ORs were adjusted for sex, age, smoking status and pack-years smoked in a logistic regression model.

  • 2

    Other includes undifferentiated cancers (n = 97), bronchioalveolar carcinomas (n = 88) and mixed cell carcinomas (n = 60).

  • 3

    Not calculated because of cell with 0 subjects.

Controls1,0191  1,0146  
Overall cases9951111.2710.34992142.39 (0.91–6.23)2.73 (1.02–7.30)
SCC453818.0020.0545382.99 (1.03–8.65)4.24 (1.34–13.38)
AC3000Not calculated3Not calculated329910.57 (0.07–4.71)0.62 (0.07–5.27)
Other2242312.6313.4724053.52 (1.07–11.63)3.84 (1.15–12.85)

The haplotype frequencies of XPD among cases and controls were estimated and the results are presented in Table IV. No significant difference was observed between cases and controls in terms of haplotype frequencies of the XPD polymorphisms. We also performed linkage disequilibrium analysis to examine the linkage between the polymorphisms at the 2 loci. The χ2 test of statistical significance for a 2-locus disequilibrium gave a test statistic value of 929.3 (D′ = 0.76) for the cases, 753.6 (D′ = 0.71) for the controls, and 1496.8 (D′ = 0.71) for all subjects. The 2-locus disequilibrium was statistically significant (p = 0.0000), indicating that the 2 polymorphisms are tightly linked in our study population.

Table IV. Global Two-Marker Haplotype Analysis
HaplotypeCases (%)1Controls (%)1
  • 1

    Haplotype frequencies were estimated using EH software with the assumption of independence.

Significancep > 0.05, d.f. = 3
Marker order5′-(codon 312)-(codon 751)-3′


In agreement with the majority of the previous studies on XPD polymorphisms and lung cancer risk, the results obtained from this study demonstrate that polymorphisms at the XPD locus affect risk of lung cancer in Chinese population. On the basis of the analysis of 1,006 patients and 1,020 frequency-matched controls, we found that the variant XPD 312Asn/Asn and 751Gln/Gln genotypes were associated with increased risk of lung cancer. Our findings are consistent with the previous case-control studies by Spitz et al.,4 Hou et al.,11 Xing et al.,20 and Zhou et al.,22 who reported that increased risk of lung cancer is associated with the variant alleles (i.e. 312Asn or 751Gln) of XPD. In addition, our findings are comparable to those of the studies on other cancers such as SCC of the head and neck23 and basal cell carcinoma.25 Supporting data also come from the phenotype analysis studies. Hemminki et al.9 found that subjects with both 312Asn/Asn and 751Gln/Gln genotypes had an approximately 50% lower rate of repair of ultraviolet-specific cyclobutane pyrimidine dimers. Recently, Qiao et al.10 have shown that DNA repair capacity among subjects homozygous for the variant XPD genotypes was significantly lower than that among those homozygous for the wild-type genotypes. These data together with our findings in the present study demonstrate that both 312Asn and 751Gln alleles are risk alleles, most likely due to reduced DNA repair capacity. Another significant finding in the current study is that the positive association was confined to lung SCC but not adenocarcinoma and other subtypes of lung cancer. As have been fully discussed in our previous report,20 this result is parallel to the observation that lung SCC is linked to exposure to tobacco carcinogens polycyclic aromatic hydrocarbons such as benzo(a)pyrene,31, 32, 33, 34 which can preferentially bind to p53 gene to form DNA adducts35, 36 that are mainly repaired by the nucleotide excision repair pathway.16 Therefore, one may expect that the genetic variants of XPD, which have reduced nucleotide excision repair capacity,4, 9, 10 are a risk modifier of smoking-related lung SCC.

However, these data compared with some of other studies reported previously are in direct contrast. In the study of XPD and risk of lung cancer, Butkiewicz et al.17 found increased risk of lung cancer related to the 312Asp/Asp genotype. Lunn et al.12 and Dybdahl et al.26 reported that subjects with the 751Lys/Lys genotype had a higher number of radiation-induced chromatid aberrations and a higher risk of skin cancer than those with 751Gln/Gln or Lys/Gln genotype. On the other hand, David-Beabes et al.18 showed a null association between the Lys751Gln polymorphism and risk of lung cancer. Several factors may account for the conflicting results of these molecular epidemiological studies. Inadequate study design such as ethnically mixed study populations,18 small sample size12, 17, 26 and perhaps inappropriate subjects evaluated are anticipated to be potential sources of bias in the previous studies. Moreover, the effect of a low-penetrance susceptibility gene on disease risk is likely to be influenced by modifying genes and environment factors. The different genetic background and different carcinogen exposure in different populations, and different types of DNA damage in the initiation of different cancers may to some extent explain the different risk estimates associated with the variant alleles.

Alternatively, the association observed in the present study between the XPD polymorphisms and lung cancer risk in Chinese population may be secondary to linkage disequilibrium with a yet unidentified, but tightly linked, lung cancer locus. To further examine this hypothesis, we compared haplotype frequencies among cases and controls. If the 312Asp→Asn was etiological and the Asn allele a dominant disease-causing allele, all haplotypes containing the Asn allele would be expected to be over-represented in cases; however this does not appear to be true. These findings may suggest that the XPD polymorphism is not a determinant of lung cancer susceptibility but instead is in linkage disequilibrium with a putative etiological variant. Another explanation for these results is that the effect is polyallelic, with several tightly linked polymorphisms influencing lung cancer risk. It could be that the contribution of the XPD polymorphisms to lung cancer risk is influenced positively or negatively by one or more additional polymorphisms within or close to the XPD gene. Depending on the combination of polymorphic variants, the effect of the 312Asn allele may be masked due to the presence of other unidentified risk alleles associated with lung cancer. This may account for the discrepancy in the role of the XPD polymorphisms among different ethnic populations.

Prevalence of the variant XPD alleles and genotypes varies markedly with ethnicity. Among Caucasians, the 312Asn and 751Gln alleles were varied from 0.29 to 0.44,4, 8, 10, 11, 18, 22 whereas among African-Americans, the 751Gln allele was shown to be 0.31.18 However, in the present study with 1,020 healthy controls, we found that the 312Asn and 751Gln alleles were 0.065 and 0.087, respectively, among Chinese, which are much less than those among Caucasians and African-Americans. These data are in accordance with another set of XPD genotyping results reported by our group in the study of esophageal cancer.37 Park et al.21 have recently shown that the frequency of the XPD 751Gln allele among Koreans was 0.055 in controls (n = 163) and 0.062 in cases with lung cancer (n = 250). Their results are very similar to ours, except that they did not observe a significant association between this variant allele of XPD and the disease, most likely due to the fact that this study was designed with too few subjects to analyze the very low homozygous genotype. The rarity of the variant allele of XPD in Asians has also been shown among Japanese. Hamajima et al.38 genotyped 240 cancer-free Japanese and reported a frequency of 0.052 for the XPD 751Gln allele. However, it is surprising to note that a much higher frequency of the 751Gln allele in a Chinese population was reported recently.19 In this study with small sample size, the 751Gln allele and 751Gln/Gln genotype was shown to be 0.40 and 18.3%, respectively, in the controls and increased risk of lung cancer was inversely associated with this allele. The contradictory results between their study and ours made us consider the possibility that there were errors in genotyping or classification, since it seems unlikely to have this great variation within the same ethnic population. In the present study, we utilized 2 different methods to distinguish Lys751Gln genotypes and direct DNA sequencing to detect the mutation. In addition, since the 312Asn and 751Gln alleles are in tight linkage disequilibrium, we reanalyzed the samples with the 312Asp allele but not 751Lys and with 312Asn but not 751Gln to ensure the genotyping and the results were the same as the first test. These efforts clearly demonstrate that our genotyping data are reliable.

However, it should be noted that even if our study consisted of a large sample size, the number of subjects in the subgroup of analysis was relatively low due to rarity of the homozygous variant genotypes (i.e., 312Asn/Asn and 751Gln/Gln genotypes) in our study population. The limitation of the results is reflected in relatively wide range of confidence intervals of the odds ratios. The findings should therefore be interpreted with caution before being confirmed in future studies.