Although some significant results were observed in several subgroups, the present study does not support the hypothesis that genotypes or haplotype/diplotype of XPC
Ala499Val and Lys939Gln have an independent etiological role in breast cancer nor do they interact with cigarette smoking or PHA-DNA adducts. This is consistent with a recent epidemiological study indicating no relationship between XPC
polymorphisms and breast cancer risk in Caucasians, African Americans or mixed population,12;14
but is in conflict with reports in Finnish11
populations. Of course, the different ethnic groups included in various studies may partly account for the conflicting results on whether or not XPC
polymorphisms have independent effects and interaction with cigarette smoking for breast cancer risk, but other explanations should also be considered.
Methodological issues of inadequate study design should be considered, such as nonrandom sampling, limited sample size, lack of statistical power and precision in subgroups.19
Most previous studies with positive findings may have been underpowered, which may contribute to spurious findings. According to one estimate, approximately 68% of stratum-specific analyses had less than 80% power to detect an OR of 2.0 between genotypes and breast cancer risk modified by cigarette smoking status.19
Thus, earlier findings of a positive association between XPC
polymorphisms and cancer risk in select subgroups with small sample sizes are likely false positives due to chance. In the present study, we estimated the power of detecting an OR of 1.5 at significance level of 0.05 based on our sample size and the minor allele frequency. The power for XPC
codon 499 and 939 variant alleles were, respectively, 97% and 99%. The lack of an association between XPC
polymorphisms and breast cancer risk in the current study is thus not due to insufficient statistical power.
Another explanation for the inconsistent results across studies may be related to the opposing effects of smoking on breast cancer risk.18;38
Oestrogen is a known risk factor for breast cancer. The carcinogenic effects of DNA-damaging agents such as B[a]P and other aromatic hydrocarbons in cigarette smoke may have an antioestrogenic effect by acting as non-steroidal oestrogens39
or activate the oestrogen receptor α.40
A recent study41
that found a significant inverse trend for cumulative pack-years of cigarette smoking and breast cancer risk among parous women supports our observations. In this recent report, cigarette smoking was found to have a harmful effect among women who began smoking before the birth of their first child, but among women who began smoking after the first live birth, the risk for breast cancer was reduced. These results suggest that the risk of breast cancer from cigarette smoking may vary with the timing of the exposure over the lifecourse.41
Although these observations have not been consistently reported across investigations,16;42–44
it is possible that the lack of consideration of a possible interaction with oestrogen exposure, or for the effect of cigarette smoking during different periods, may mask or dilute real associations.
A third possible explanation is that most previous studies did not consider the complex gene-environment interactions. As part of the parent LIBCSP study,21;22
participants were characterized in terms of PAH-DNA adducts, enhancing the precision of defining individual PAH body burden. Most previous studies lack an exposure biomarker, possibly biasing (overestimating) the independent or interaction effects of XPC
polymorphisms and cigarette smoking for breast cancer risk due to the inaccurate categorization of cigarette smoke exposure.
Finally, only two polymorphisms in XPC
gene involved in recognition of DNA damage were explored in present study. There were more than thirty other genes involved in the NER pathway were not considered. Crew and colleagues observed that a combined effect of polymorphisms in NER genes revealed a modest, statistically significant positive association between number of putative high-risk alleles and breast cancer risk in LIBCSP.45
A previous study conducted in the same population observed that a polymorphism in XPD
codon 751, also functioning in the NER pathway, statistically significantly increased breast cancer risk from PAH-DNA adducts and cigarette smoking exposure.46
has a much broader role in affecting other DNA repair pathways.47
But we did not observe any gene-gene interactions between the two XPC
polymorphisms and XPD
codon 751 in the present study (data not shown). These results suggest that the two XPC
polymorphisms may not independently play a critical role in breast tumourigenesis, or at least, have a less important role than XPD
polymorphisms. Even if the XPC
polymorphisms affect breast cancer risk, it may not be through interaction with exposure to cigarette smoking or PAH-DNA adducts. Although increased breast cancer risk was observed in subgroups carrying the CC-CC diplotype and exposed to moderate/heavy smoking or detectable PAH-DNA adducts, the lack of a priori
functional data as well as the limitation of haplotype/diplotype imputation limits the interpretation of the data.48
The relatively large sample size, population-based study design, and both genotyping and exposure biomarker (PAH-DNA adducts) available, are the strengths of this study. Recall bias is one limitation that is common to case-control studies; but because neither cases nor controls were aware of their genotype and the exposure biomarker at the time of the interview, the accuracy of the response is not likely to be related to genotype or exposure status. Although the exposure information on cigarette smoking may be differentially recalled by cases and controls due to the cases’ cancer diagnosis, this misclassification would not differ by genotype and PAH-DNA adducts status. Although our results were obtained from a mixed population, most of them (>93%) were Caucasians. The potential influence of ethnic heterogeneity might not bias the conclusion. Extensive information collected in the study for evaluating confounding factors and effect modifiers allows us to control for confounding in data analysis. The OR for XPC polymorphism and breast cancer risk changed less than 10% when considering potential confounding factors individually or in multivariate models indicating that confounding did not dramatically affect the results. Age at reference included in logistic models, minimized any possible residual confounding effect of age.
In conclusion, we did not find any independent main effects of XPC
polymorphisms for breast cancer risk or interaction with cigarette smoking/PAH-DNA adducts, although several significant differences were observed for subgroups with XPC
variant alleles or diplotypes and specific carcinogenic exposures. There is evidence suggesting that the two XPC
polymorphisms are linked with a functional SNP in 3′ of intron 11 splice acceptor site and may decrease DNA repair activity of single strand breaks.49;50
However, the biological effects of the two SNPs under conditions of carcinogenic exposure remain to be determined.