While the HCR assay as described in this study assesses the overall DNA repair capacity, it preferentially measures repairs made by the nucleotide excision repair (NER) pathway. NER repairs bulky adducts caused by UV irradiation [38
] and many carcinogens such as the ones contributing to carcinogenesis presented by BC patients [5
]. Low levels of NER repair have been detected in peripheral blood lymphocytes and tumor samples of BC patients [5
]. In addition, exposure to ionizing radiation is a risk factor for BC [9
]. DNA lesions arising from ionizing radiation and many carcinogens causing bulky lesions are repaired by NER and deficiencies in the NER pathway can increase genomic instability.
The HCR assay, developed in the early 1990s, provided a way to obtain a quantitative, phenotypic measurement of DRC in lymphocytes [5
]. The subsequent development of a plasmid based on luciferase activity instead of radioactivity resulted in stable transfection efficiencies [38
]. Because of the high cost of this assay, its technical complexity, and its labor intensiveness, relatively few laboratories in the world currently utilize it for large-scale molecular epidemiology studies of cancer [9
Studies published to date of HCR assays of lymphocytes have frequently, but not always [34
], found a link between a low DRC and increased cancer risk. Our group [11
] was the first to report that a low DRC was a risk indicator for BC. That initial DRC study consisted of a small sample size (n = 36 patients) and did not have the benefit of ROC analysis or a well-defined percent DRC level (e.g. low, medium, high) as a BC risk estimator compared to the analyses presented in this study. In the current study, we furthered classified DRC into terciles; low, medium and high levels. This methodology has the potential to serve as a biomarker of risk of BC (United States Patent Trade Office US 8,163,473 B2). Our adjusted OR (2.3) for DRC as a continuous variable is within the range of similar ORs values for published studies for basal cell carcinoma skin cancer (Wei et al.
1993), lung cancer (Wei et al.
2000), head and neck cancer (Cheng et al.
1998), prostate cancer (Hu et al.
2004) and BC (Shi et al.
However, when DRC was analyzed either as a dichotomous or as a categorical variable (low, medium, high), the ORs are much higher. Such high ORs are unusual and have been reported only a few studies. For example, Latimer et al.
(2010) recently estimated that the sensitivity of detecting tumors based on reduced NER levels alone is 95%, the specificity is 74%, and the odds ratio is 53.8 (95% CI, 28.3–102.4). This group demonstrated the critical importance of the NER pathway in BC by measuring the NER capacity in breast tumors (n=19) and normal primary cultures expressed relative to the mean of these normal BC tissues. The mean NER capacity of the tumor samples was significantly lower than that of normal breast tissues, averaging only 44% of normal activity (P
< 0.001). Some other studies have also reported very high ORs when strong associations are found. This has occurred with the association of cervical cancer with HPV exposure. Some HPV studies have reported very powerful associations. For example, a case control study for HPV by Powell et al.
(2011) reported ORs as high as 2770; Chen et al.
(2011), as high as 75; Chuang et al.
(2012), ORs 18.1 and higher; and Almonte et al.
(2011), ORs of 16.1 [46
In this study, we observed an unusually strong increment on the OR after adjustment. Therefore, we repeated the analysis entering, first entering DRC alone in the model; then the covariates were entered one by one. In doing that, we found that the adjusted OR increased gradually as we added each covariate to the model.
No evidence of interaction terms were found in the multivariate logistic model (p>0.05), so DRC can be considered as a potential independent risk factor for DRC because it is biologically plausible, shows a strong association, and exhibits a dose–response relationship after adjusting for all confounders simultaneously.
Despite the usefulness of the HCR assay to estimate DRC and detect BC risk, future prospective studies are needed to further validate the use of this approach. Specifically, further studies are needed to ascertain 1) whether women without BC with a low DRC have a higher incidence of BC compared to those who have a high DRC and 2) whether women with BC and a low DRC are at an increased risk of recurrent BC compared to those with high DRC.
In this study, patients were evaluated for DRC after BC was diagnosed; thus, reverse causation could be present if BC lowers DRC, which would explain why we found a low DRC associated with BC. Nevertheless, reverse causation is an unlikely explanation of our findings because experimental evidence demonstrates that a low DRC or impairments in DNA repair pathways and/or genes is a mechanistic component involved in BC carcinogenesis [2
] or relapse [24
] for the transition to hormone independence [52
Given that newly diagnosed patients are not always truly incident patients, we compared DRC results to tumor size (data not shown) as a surrogate for the length of time period between disease onset and diagnosis to address the potential changes that DRC in lymphocytes might have had after a woman develops BC. With these analyses, we did not find any statistically significant DRC level differences by tumor size that could interfere with our comparisons.
DRC at its cutoff point selected based on its median value (4.3%) was found in this study to have 83.2% sensitivity and 77.6% specificity for detecting BC. ROC analysis showed a high accuracy detecting BC (88.4% of the area under the curve), which was statistically significant. Because any given specific value of DRC level is associated with a quantitative BC risk estimate, DRC has the potential to be a biomarker to rule out BC risk in low-prevalence or low-risk populations or confirm its presence in high-prevalence or high-risk populations. However, a higher specificity and sensitivity would increase the value of DRC as a detector of BC. In addition, these findings need to be validated in other populations and with future prospective studies.