Sex Differences in Risk of Lung Cancer Associated with Methylene-tetrahydrofolate Reductase Polymorphisms

Introduction

Lung cancer causes the greatest number of cancer-related deaths in the United States, accounting for an estimated 172,570 new cases and 163,510 deaths in 2005 (1). Although tobacco smoking is the primary risk factor for lung cancer (2), fewer than 20% of lifetime smokers develop lung cancer. This observation suggests that other factors, such as interindividual variations in genetic susceptibility (3) or dietary habits (4), may influence lung cancer risk associated with exposure to the chemical carcinogens in tobacco.

Epidemiologic studies have provided evidence that high consumption of vegetables and fruits is associated with a reduced risk of lung cancer (5-8). Folate is one of the constituents found in vegetables and fruits, and dietary folate may be one of the micronutrients that provide protection against lung carcinogenesis (9). We have previously shown that low dietary folate intake was associated with reduced DNA repair capacity as measured by the host-cell reactivation assay (10) and suboptimal DNA repair capacity was associated with an increased risk of lung cancer (11). Many other nutrients, also existing in vegetables and fruits, such as vitamins B₁₂ and B₆, have anticancer properties, and deficiencies in such nutrients could lead to DNA damage, such as single- and double-strand breaks and oxidative lesions also induced by tobacco carcinogens (12). Thus, there is a biologically plausible link between dietary nutrient intake and susceptibility to lung cancer. Therefore, identifying biomarkers that underlie genetic susceptibility could facilitate risk assessment of tobacco-induced lung cancer.

As an important enzyme involved in folate metabolism, 5,10-methylenetetrahydrofolate reductase (MTHFR) catalyzes the irreversible conversion of 5,10-methylenetetrahydrofolate to 5-methyl-tetrahydrofolate (13), which serves as a substrate for the remethylation of homocysteine to methionine required for the subsequent synthesis of S-adenosylmethionine. S-adenosylmethionine is the universal methyl donor for many intracellular methylation reactions, especially DNA methylation (14). In addition, the substrate of MTHFR, 5,10-methylenetetrahydrofolate, is also required for thymidine synthesis via thymidylate synthase and indirectly for purine synthesis.

The MTHFR gene is located at 1p36.3 (15), and several polymorphisms of the MTHFR gene have been identified (15, 16), among which are three nonsynonymous single nucleotide polymorphisms (i.e., C677T, A1298C, and G1793A) that have minor allele frequencies of >5%. The roles of these polymorphisms in the risk of various cancers, including lung cancer, have been studied (17-24), but the results have been inconsistent (17, 22, 23), suggesting phenotypic and genotypic heterogeneity. In addition, these published studies did not include all three nonsynonymous single nucleotide polymorphisms and all have had relatively small sample sizes without the statistical power to identify a susceptible subgroup or to detect gene-environment interactions. Therefore, we genotyped the three MTHFR C677T, A1298C, and G1793A polymorphisms in a large hospital-based lung cancer study of non-Hispanic Whites. We tested the hypotheses that these polymorphisms are associated with risk of lung cancer and that there are interactions between these polymorphisms and dietary nutrient intake or tobacco smoking in the etiology of lung cancer.

Materials and Methods

Study Subjects

Patients were recruited consecutively between September 1995 and December 2003, without restrictions on age, sex, or cancer stage and histology from an ongoing molecular epidemiologic study of lung cancer conducted in the Department of Epidemiology at The University of Texas M.D. Anderson Cancer Center in Houston, TX. The study has been described in detail previously (3, 9). All patients included in this analysis had newly diagnosed, histopathologically confirmed, previously untreated (by radiotherapy or chemotherapy) lung cancer, and all were non-Hispanic White. The controls were selected from a pool of cancer-free subjects recruited through the Kelsey Seybold Clinic, a large multispecialty physician practice that has multiple clinics throughout the Houston metropolitan area. The controls were frequency matched to the cases according to age (±5 years), sex, ethnicity, and smoking status (ever or never). Exclusion criteria included having received a blood transfusion within the past 6 months for all participants and having had previous treatment with radiotherapy or chemotherapy for cases, or having had previous cancer for all participants. The participation response rate was over 70% for the cases and controls. After subjects provided written informed consent, interviews were conducted with a structured questionnaire to collect information on demographic data and risk factors such as smoking status and pack-years smoked.

Dietary Analysis

We used a modified version of the National Cancer Institute's Health Habits and History Questionnaire to collect the dietary data (25, 26). The Health Habits and History Questionnaire includes a food frequency list, an open-ended food section, and other dietary behavior questions pertaining to use of supplements, restaurant dining, and food preparation methods. The food frequency instrument assessed diet in the cases the year before diagnosis and in the controls the year before enrollment in the study. Data entry was conducted using DietSys (version 4.01) and DietSYS+Plus (DietSYS+Plus Analysis Software, version 5.9 Block Dietary Data Systems, Berkeley CA, 1999) programs. Dietary analysis was conducted using DietSYS+Plus (version 5.9). Source of total folate values was Standard Release 14 (27). Recipe adjustments for moisture changes and nutrient losses due to cooking were also made. Dietary total folate intake was adjusted by daily calorie intake and expressed as dietary total folate in μg/1,000 kcal/d. The study was approved by the institutional review boards of M.D. Anderson and the Kelsey Seybold Clinic.

Genotyping

Leukocyte pellets were obtained from the 200-μL buffy coat after centrifugation of a 1-mL whole blood sample. Genomic DNA was extracted from the leukocytes by using a Qiagen DNA Blood Mini Kit (Valencia, CA). DNA purity and concentrations were determined by spectrophotometric measurement of absorbance at 260 and 280 nm. We used previously published primers (17, 28) to amplify the region upstream of the MTHFR gene containing the C677T, A1298C, and G1793A polymorphic sites. The PCR reactions were done in a total volume of 10 μL containing 1× PCR buffer (50 mmol/L KCl, 10 mmol/L Tris-HCl, and 0.1% Triton X-100), 1.5 mmol/L MgCl₂, 0.15 mmol/L deoxyribonucleotide triphosphates, 100 nmol/L each primer, 1.5 units of Taq polymerase (Sigma-Aldrich, St. Louis, MO), and ∼20 ng of genomic DNA. The cycling conditions consisted of one cycle of 95°C for 5 minutes and 35 cycles of the following: 95°C for 30 seconds, annealing temperature for 45 seconds, and 72°C for 45 seconds, and a final extension at 72°C for 10 minutes. After PCR amplification the product was digested for 12 hours by the HinfI enzyme for C677T, the MboII enzyme for A1298C, and the BsrbI enzyme for G1793A (New England BioLabs, Beverly, MA). The PCR products were resolved in separate but adjacent wells on a 1.5% agarose gel for C677T, 4% metaphor gel for A1298C, and 3% Nusieve 3:1 agarose gel (Cambes, East Rutherford, NJ) for G1793A and stained with ethidium bromide. We evaluated the results without knowing the subjects' case or control status and at least 10% of the samples were tested, and the concordance was 100%.

Statistical Analyses

Participants who had smoked fewer than 100 cigarettes during their lifetimes were categorized as “never smokers,” and the rest were categorized as “ever smokers.” Participants who had drunk alcoholic beverages at least once a week for ≥1 year were categorized as “ever drinkers,” and the rest were defined as “never drinkers.” Dietary nutrient assessment has been described in detail elsewhere (9). Mean values of body mass index; total energy intake; dietary intake of total folate, vitamin B₆, vitamin B₁₂, and methionine; and pack-years smoked were compared for cases and controls by the Student's t test. Differences in categorized demographic variables, smoking status, alcohol consumption, and allele and genotype frequencies between the cases and controls were tested by the χ² test. Linkage disequilibrium between the polymorphisms was measured with the Two-locus Linkage Disequilibrium Calculator (2LD) written by Zhao et al. (29, 30). Associations between genotypes and risk of lung cancer were estimated by computing the odds ratios (OR) and their 95% confidence intervals (95% CI) from both univariate and multivariate logistic regression analyses. The reference groups were subjects carrying homozygotes of the major alleles (i.e., the MTHFR 677CC, MTHFR 1298AA, or MTHFR 1793GG genotype). Stratification analysis was used to estimate risk for subgroups by age, sex, smoking status, alcohol consumption, total folate intake, vitamin B₆, vitamin B₁₂, and methionine. The lower quartile (25%) value of continuous variables was used as the cutoff point. Ps for interactions were determined by the likelihood ratio testing of models with and without a multiplicative interaction term. All tests were two sided, and all statistical analyses were done by using the Statistical Analysis System software (version 8e; SAS Institute, Cary, NC).

Results

A total of 1,055 cases and 1,146 controls were available for this analysis. DNA samples from four cases and five controls could not be amplified; thus, the final analysis included 1,051 cases and 1,141 controls (Table 1). No significant differences in age, sex, or smoking status were found between the cases and controls, suggesting that the controls had been adequately frequency matched. However, the mean number of pack-years smoked for the cases (43.92 ± 35.12 years) was significantly higher than that for the controls (35.34 ± 30.70 years; P < 0.0001). This difference was further adjusted for in the multivariate analysis.

Because some participants did not provide information on nutrient intake and alcohol consumption, only 956 cases and 1,071 controls were included in the nutrient intake analysis, and 962 cases and 1,073 controls were included in the alcohol status analysis. Mean body mass index was significantly lower in the cases (26.14 ± 5.00 kg/m²) than in the controls (27.36 ± 5.03 kg/m²; P < 0.0001). In general, the cases consumed significantly less total folate (208.06 ± 69.38 μg/1,000 kcal/d) and vitamin B₆ (1.70 ± 0.62 mg/d) than did the controls (total folate, 220.44 ± 86.54 μg/1,000 kcal/d; vitamin B₆, 1.79 ± 0.76 mg/d; Table 1; P = 0.0004 for total folate and 0.002 for vitamin B₆). Fewer cases (64.3%) reported ever drinking than did controls (69.6%; P = 0.012).

The MTHFR variant allele and genotype distributions among the cases and controls are shown in Table 2. The MTHFR 677T allele frequencies were 0.318 in the cases and 0.336 in the controls. The MTHFR 1298C frequencies were 0.324 in the cases and 0.297 in the controls, and the MTHFR 1793A frequencies were 0.048 in the cases and 0.038 in the controls. These variant allele frequencies did not differ between the cases and controls (P = 0.418 for 677T, P = 0.109 for 1298C, and P = 0.125 for 1793A). The distributions of genotypes for the three MTHFR polymorphisms in the controls were all in Hardy-Weinberg equilibrium (P = 0.516 for C677T, P = 0.168 for A1298C, and P = 0.612 for G1793A; Table 2).

In the multivariate analysis with adjustment for age, sex, and square root of pack-years smoked, neither the 677CT nor 677TT genotypes were associated with risk of lung cancer (OR, 0.93; 95% CI, 0.78-1.11 for 677CT; OR, 0.81; 95% CI, 0.61-1.09 for 677TT), compared with the 677CC genotype. On the other hand, the 1298CC genotype was associated with significantly increased risk of lung cancer (OR, 1.36; 95% CI, 1.01-1.85), although the 1298AC genotype was not (OR, 1.07; 95% CI, 0.90-1.28) compared with the 1298AA genotype. Because the 1793AA genotype was relatively rare, we combined it with the 1793GA genotype for further analysis, and the combined 1793AA and 1793GA genotype was also not associated with risk (OR, 1.23, 95% CI, 0.90-1.67) compared with the 1793GG genotype (Table 2).

The three polymorphisms were in strong linkage disequilibrium [the normalized disequlibrium coefficient was −0.993 for the C677T and A1298C genotypes (χ² = 281.37; P = 0.0000; −0.856) for the C677T and G1793A genotypes (χ² = 105.17; P = 0.0000); and 0.884 for the A1298C and G1793A genotypes (χ² = 165.06; P = 0.0000)]. Because the frequencies of the 1793A variant genotypes were very low (<10% for the GA genotype and <0.6% for the AA genotype for both cases and controls; Table 2), further stratification analysis was done only for the more common C677T and A1298C polymorphisms.

We then evaluated genotype frequencies in age and sex strata. The MTHFR C677T and A1298C polymorphisms were not associated with lung cancer risk in any age subgroups except for a borderline increased risk associated with the 1298CC genotype in those >55 years (OR, 1.39; 95% CI, 0.98-1.99; Table 3). However, women carrying the 677TT genotype had a significantly lower risk of lung cancer than did women with the 677CC genotype (OR, 0.60; 95% CI, 0.40-0.92). In addition, female but not male subjects with the 1298CC genotype had a 2-fold increased lung cancer risk (OR, 2.09; 95% CI, 1.32-3.29) compared with those with the 1298AA genotype. However, no such patterns were found in men (OR, 1.09; 95% CI, 0.72-1.65 for 677TT and OR, 0.95; 95% CI, 0.62-1.45 for 1298CC).

Because of this observed sex difference in risk of lung cancer associated with two of the MTHFR polymorphisms (i.e., C677T and A1298C), we did further stratification analyses by smoking status and consumption of alcohol, total folate, vitamin B₆, vitamin B₁₂, and methionine for women and men separately. The subjects were classified as having “low” or “high” consumption of nutrients, based on the 25% percentile values in controls as the cutoff. The results of the analysis suggest evidence of gene-nutrient interactions in women (Table 4) and gene-smoking interactions in men (Table 5).

In women, there was interaction between the C677T genotypes and reported dietary intake of nutrients (i.e., vitamin B₆, vitamin B₁₂, and methionine). For example, compared with those carrying the 677CC genotype, there were trends of decreasing risks of lung cancer among those reporting high intakes and trends of increasing risks among those reporting low intakes as the number of the 677T allele increased (i.e., the 677CT and TT genotypes; tests for interactions: P = 0.006 for vitamin B₆, P = 0.020 for vitamin B₁₂, and P = 0.014 for methionine). However, such interactions were not evident for smoking (P = 0.966), alcohol use (P = 0.264), and total folate intake (P = 0.971), although the lowest risks were observed for those carrying the 677TT genotype who were either ever smokers (adjusted OR, 0.20; 95% CI, 0.10-0.43) or never drinkers (adjusted OR, 0.39; 95% CI, 0.19-0.77; Table 4).

Specifically, among those women who reported high intakes, the 677TT genotype was associated with significantly decreased risks (adjusted OR, 0.57; 95% CI, 0.34-0.95 for folate intake; adjusted OR, 0.29; 95% CI, 0.15-0.56 for vitamin B₆ intake; adjusted OR, 0.34; 95% CI, 0.18-0.63 for vitamin B₁₂ intake; adjusted OR, 0.29; 95% CI, 0.15-0.57 for methionine intake) compared with the 677CC genotype. For the MTHFRA 1298C polymorphisms, the 1298CC genotype showed increased risk of lung cancer in those women who reported ever smoking (adjusted OR, 2.25; 95% CI, 1.19-4.23), ever drinking (adjusted OR, 1.99; 95% CI, 1.04-3.82), high vitamin B₆ intake (adjusted OR, 2.15; 95% CI, 1.14-4.05), high vitamin B₁₂ intake (adjusted OR, 2.42; 95% CI, 1.32-4.46), or high methionine intake (adjusted OR, 2.49; 95% CI, 1.32-4.70; Table 4). However, only interaction between the MTHFR A1298C polymorphism and alcohol consumption was statistically significant (P = 0.021).

The results from women were not evident in men, however. For men, the lung cancer risk was not associated with any combination of the variant genotypes with any subgroups of nutrient intake (Table 5). Nevertheless, the MTHFR C677T and A1298C polymorphisms seemed to show interactions with smoking (P = 0.033 and P = 0.014, respectively). Compared with never smokers carrying 677CC genotypes, never smokers with the 677TT genotype had a significantly higher risk of lung cancer (adjusted OR, 12.97; 95%CI, 1.16-145.35). Using subjects with the 1298AA genotype who reported never smoking as the reference group, lower risk of lung cancer was found for never smokers with the 1298AC genotypes (adjusted OR, 0.33; 95%CI, 0.14-0.81) and ever smokers with 1298AA (adjusted OR, 0.42; 95%CI, 0.22-0.83) or 1298CC genotypes (adjusted OR, 0.44; 95%CI, 0.20-0.95; Table 5).

Discussion

We found overall that subjects with the MTHFR 1298CC genotype were at higher risk of lung cancer than were those with the MTHFR 1298AA genotype, and that this association was more pronounced among women. In addition, the MTHFR 677TT genotype in women was associated with a decreased lung cancer risk compared with carriers of the MTHFR 677CC genotype, and this association was influenced by an interaction with dietary intake of vitamin B₆, vitamin B₁₂ and methionine. However, these findings were not evident in men, among whom there was evidence for gene-environmental interactions between smoking and the MTHFR C677T and MTHFR A1298C polymorphisms. These findings suggest that the MTHFR polymorphisms may contribute to susceptibility to lung cancer, particularly in women, and that the risk of lung cancer seems influenced by sex-specific interactions between the MTHFR polymorphisms and dietary habits in women and smoking status in men. Although findings from subgroup analyses must be treated with caution due to reduced numbers of observations in the subgroups, some published studies seem to support a role of the MTHFR polymorphisms and a sex-specific difference in lung cancer risk.

The C677T polymorphism, the most extensively investigated MTHFR polymorphism, causes an alanine-to-valine substitution in exon 4 at codon 222, which consequently reduces enzyme activity. It has been shown that subjects with the 677TT genotype had a two-thirds reduction in enzyme activity (31). This reduction would result in a relative lack of the product, methyltetrahydrofolate, and thus decreased conversion of homocysteine to methionine, subsequently reducing the availability of methyl groups for methylation reactions that require adenosylmethionine. Theoretically, a reduction in the MTHFR activity may increase cancer risk due to altered DNA methylation resulting from lower levels of 5-methyltetrahydrofolate. Consistent with this hypothesis, increased risks of breast cancer and esophageal cancer have been reported in those with the 677TT genotypes (20, 21, 32).

On the other hand, elevated levels of methylenetetrahydrofolate that result from deficient MTHFR (i.e., the 677TT genotype) may provide more one-carbon groups for thymidylate synthesis, thereby enhancing DNA synthesis and repair ability. We have previously reported that suboptimal DNA repair capacity could increase the risk of lung cancer (11). Therefore, it is also possible that 677TT might be associated with reduced lung cancer risk, but such a dual function of folate metabolisms seems to depend on the bioavailability of folate, which in turn depends on dietary intake of vitamins, the coenzymes involved in the folate metabolism pathway (33, 34). Our findings of a protective effect of MTHFR 677TT in women with higher intakes of vitamin B₆ and vitamin B₁₂ are consistent with this hypothesis. Furthermore, other studies have also provided evidence to support such a hypothesis. In an Italian study, a >3-fold decreased risk of adult acute lymphoblastic leukemia was found associated with the 677TT genotype (OR, 0.28; 95% CI, 0.12-0.72) compared with the 677CC genotype (24), suggesting a protective role of the 677TT genotype. However, inconsistent results regarding the C677T polymorphism in colorectal cancer studies indicate the need for simultaneously investigating environmental factors, such as nutrient intake, and the MTHFR polymorphisms (18, 19).

Few studies have investigated the association between the MTHFR C677T polymorphisms and risk of lung cancer, and no conclusion can be drawn from the limited information available. For example, in a case-control study of 146 cases and 44 controls in a Caucasian population, the 677TT genotype was found associated with a significantly higher risk of non–small cell lung cancer (22). However, in our previous analysis (17) of 550 lung cancer cases and 554 controls, a subset of the present study, we did not find any significant association between the C677T and A1298C polymorphisms and increased lung cancer risk due to limited study power. In a case-control study of 59 cases and 232 controls in a Chinese population, a 30% decrease in lung cancer risk was found associated with the 677TT genotype, although the result was not statistically significant (OR, 0.71; 95% CI, 0.39-1.30) because of the small sample size (23). These inconsistent results suggest that the effects of the MTHFR polymorphisms may be complex and that many other factors, such as environmental exposure, dietary habits, and other genes in the same pathway, may all contribute to the outcome. To investigate possible interactions between these multiple etiologic factors, studies with larger sample sizes are required.

Low folate intake is thought to impair the metabolic pathway of folate in which MTHFR is involved, and other nutrients, including vitamin B₆, vitamin B₁₂, and methionine, are also thought to interact metabolically with folate in this pathway (33, 34). In vitro experiments showed that MTHFR C677T results in impaired stability and reduced activity under low folate conditions (31, 35, 36). Studies of elderly women showed that when the folate status is low, the 677TT genotype is associated with an increase in homocysteine concentration and DNA hypomethylation (37, 38). Considering the role of the MTHFR enzyme in DNA synthesis and methylation, a metabolic explanation can be drawn for the interactions between the MTHFR C677T polymorphisms and nutrient status. Our results suggested that in women who consumed high levels of vitamin B₆, vitamin B₁₂, or methionine, the MTHFR 677TT genotype seemed protective against lung cancer, whereas in women who consumed less vitamin B₆, vitamin B₁₂, or methionine, the MTHFR 677TT genotype seemed to increase lung cancer risk. These interactions found between the MTHFR C677T polymorphism and vitamin B₆, vitamin B₁₂, or methionine in women are consistent with the results from studies of colorectal cancer (39, 40).

We did not find any association between C677T and lung cancer risk in men. When stratified by smoking status, the 677TT genotype was associated with a significantly higher risk of lung cancer in men who had never smoked, which contradicts the results found in women. Considering that only five cases and two controls had the 677TT genotype among nonsmoking men, this result might be biased as a result of the small sample size. Unlike women, interactions were found between the MTHFR C677T and A1298C genotypes and smoking status in men but not between the polymorphisms and nutrient intake. The different interaction patterns may indicate a difference in susceptibility between men and women to tobacco-induced lung cancer. Studies have reported that smoking-related DNA adduct levels in tumors and adjacent lung tissue, with adjustment for smoking dose, were higher in women than in men in both smokers (41, 42) and nonsmokers (43). Our previous work indicates that women tend to have lower DNA repair capacity than men (11). These findings suggest that the high intake of nutrients that stimulate DNA repair and synthesis ability may benefit women more than men. In the present study, smoking was more prevalent in men (90% ever smokers) than in women (76%), suggesting that smoking had a more predominant effect on lung cancer risk in men than in women, and this effect was not affected by either nutrient intake or the MTHFR polymorphisms.

The A1298C polymorphism leads to substitution of alanine for glutamic acid in exon 7 at codon 429, which also results in decreased enzyme activity (44). Studies investigating the MTHFR 1298C allele have found inverse associations with colorectal cancer (45), colon cancer (46), breast cancer (20), and acute lymphocytic leukemia (47). On the other hand, a positive association was also reported between the 1298C allele and esophageal cancer (32). For bladder cancer, the 1298C variant genotypes were associated with a 4-fold increased risk in heavy smokers compared with nonsmokers with the 1298AA genotype (48).

We found that the MTHFR 1298CC genotype was associated with a higher risk of lung cancer in our study population. This polymorphism influences the specific activity of the enzyme, homocysteine levels, and plasma folate concentrations but to a lesser extent than the C677T polymorphism does (16). This result is consistent with those reported in studies of bladder and esophageal cancer (32, 48) but is in contrast to those reported in a study of colorectal cancer (45). Additional functional studies on the MTHFR A1298C polymorphism are needed to develop a more in depth understanding of its association with cancer susceptibility.

The novel MTHFR G1793A variant is rare in Whites and the allele frequency in our study population was consistent with that reported by Rady et al. (28). We found no increase in risk associated with the variant genotypes, a result consistent with our previous study of squamous cell carcinoma of the head and neck (49). The newly identified G1793A located in exon 11 at codon 594, causes an arginine-to-glutamine change (28). This polymorphism occurs much less often (0.07) among Whites than the other two MTHFR polymorphisms, and its functional relevance and association with cancer risk have not been reported. Our data suggest that this polymorphism may not be important in the etiology of lung cancer.

Because we typed all three potentially functionally relevant variants in MTHFR, we were able to evaluate evidence of linkage disequilibrium between the MTHFR variants. Previous studies have shown that MTHFR C677T and A1298C are in strong linkage disequilibrium (50-52). Rosenberg et al. (53) investigated the linkage disequilibrium of the C677T, A1298C, and three other intronic dimorphisms among white Israelis, Japanese, and Ghanaian Africans and found that the 677T allele was associated with one haplotype, G-T-A-C, in White Israelis and Japanese homozygotes. Our results provide evidence of linkage disequilibrium between the C677T and A1298C polymorphisms, and the G1793A polymorphism was also in linkage disequilibrium with both C677T and A1298C polymorphisms.

Possible limitations in our hospital-based study design could have introduced some selection bias. However, the genotype distributions of our study were similar to those reported in other studies. For example, the frequencies of the MTHFR 677 CC, CT, and TT genotypes among our 1,141 non-Hispanic White controls from Texas were 43.6%, 45.5%, and 10.9%, respectively, compared with 42.6%, 46.0%, and 11.4%, respectively, for 260 population-based Caucasian controls from Canada (54) and 45.1%, 43.3%, and 11.5% for 1,964 hospital-based controls in California (55). In addition, the frequencies of MTHFR 1298AA, AC, and CC among our 1,141 controls were 48.5%, 43.5%, and 8.0%, respectively, compared with 49%, 43%, and 8%, respectively, for 257 Caucasian controls from Italy (24) and 47.2%, 41.9%, and 10.9% for 1,964 controls in California (55). The frequencies of the MTHFR 1793GG, GA, and AA genotypes in our controls were 92.5%, 7.4% and 0.1%, respectively, compared with 93.8%, 6.2% and 0%, respectively, for Caucasians in Galveston, TX (28). Because our MTHFR genotype frequency estimates from the hospital-based controls are similar to those of population-based controls, any selection bias related to the genotype distributions is unlikely to be substantial.

The dietary data analysis had some limitations. Like all case-control studies of diet and cancer risk, the diet information obtained was retrospective and was for the year preceding the lung cancer diagnosis. However, several studies have examined the reproducibility of food frequency questionnaires under a wide variety of conditions and have shown correlations generally ranging from 0.4 to 0.7 for nutrient intakes measured at periods of 1 to 10 years apart (56-58). The average reduction in the intake of nutrients was 0.07 over the 5-year period, suggesting that dietary changes over time were not substantial and thus repeated administration of the questionnaire was not necessary (59). In this study, >80% of the cases were interviewed within 15 days after their diagnosis, thus reducing potential measurement errors attributable to recall bias as well as recent dietary changes after diagnosis.

In conclusion, in a large, non-Hispanic White population, we observed a sex difference in lung cancer risk associated with the MTHFR polymorphisms. Specifically, the MTHFR C677T and A1298C polymorphisms were associated with risk of lung cancer in women with significant interactions between the MTHFR C677T polymorphism and nutrient intake, but these observations were not evident in men. In contrast, the MTHFR C677T and A1298C polymorphisms interacted with smoking status in men but not in women. Because ours was a hospital-based case-control study, it is likely that some other unknown confounders could have contributed to this sex difference. Therefore, large, prospective studies as well as studies of the mechanisms of gene-diet interactions of MTHFR and related genes involved in the folate metabolic pathway are warranted.