Introduction
The aromatic and heterocyclic amines are among the constituents of tobacco smoke that have been identified as carcinogens [
1,
2]. These carcinogens require host-mediated metabolic activation to electrophiles, which readily bind nucleophilic DNA, to induce a mutation and, ultimately, cancer [
3]. Two pathways metabolize aromatic amines [
4]. First, aromatic amines can be
N-acetylated in the liver by
N-acetyltransferase-2 (NAT2) or
N-acetyltransferase-1 (NAT1) [
5]. This is a detoxifying pathway. Second, aromatic amines and heterocyclic amines can be
N-oxidized by P450 enzymes in the liver or in extrahepatic tissues [
4]. This oxidation competes with the hepatic
N-acetylation for aromatic amines but not for heterocyclic amines. The product of the oxidation is then either
O-acetylated by NAT2 or NAT1, a reaction that yields the activated electrophile, or detoxified by competing enzymatic pathways [
4,
6]. The NAT2 enzyme therefore has a dual role: it detoxifies aromatic amines hepatically but may also play a role in activation of aromatic amines and heterocyclic amines in extrahepatic tissues such as the breast.
NAT2 is a polymorphic enzyme in humans [
5,
7‐
10]. Those who possess homozygous wild-type alleles are classified as fast acetylators – because they have a higher rate of metabolic activation of aryl aromatic amines – and those who possess certain polymorphisms are called slow acetylators because they have a lower rate of metabolic activation of these amines [
9,
11]. Depending on which metabolic pathway predominates at critical junctures of exposure and tissue susceptibility, fast acetylators may be at higher or lower risk of smoking-induced breast carcinogenesis than slow acetylators. Postmenopausal women who smoke and have a reduced ability to detoxify by-products of tobacco smoke, as measured by their NAT2 genotype (slow acetylators), have an excess risk of breast cancer [
12‐
14]. In one of these studies, this excess risk was found to be limited to women who had smoked for 20 years or more [
14]. In another study, the postmenopausal women who were rapid acetylators were found to be at highest risk [
15]. In a fifth study, the association between smoking and breast cancer showed little dependence on acetylation rate [
16].
Most epidemiologic studies that have examined the relation between active cigarette smoking and breast cancer have found weak or null associations [
17,
18]. A meta-analysis of the studies that excluded from the analysis those women who had been passively exposed reported that the risk of breast cancer for active smokers was more than twice as much as that for women never actively or passively exposed to tobacco smoke [
19]. Studies comparing women who were passive smokers with women who had never been either active or passive smokers have also shown consistent elevations in breast cancer risk associated with smoking [
20‐
23]. Recently, two case–control studies [
24,
25] have reported effect modification by acetylation status for both active and passive smokers. Both studies found stronger associations between breast cancer risk and passive exposure to smoke among rapid acetylators. Though both studies also found an association between active smoking and breast cancer risk, the magnitude of the risk was greater among slow acetylators in the study by Chang-Claude and colleagues [
24] and among fast acetylators in the study by Morabia and colleagues [
25]. Inconsistent findings have prevented any meaningful conclusions from being drawn about the interaction of acetylation status and exposure to tobacco smoke in the etiology of breast cancer.
We collected genetic and behavioral information from incident primary breast cancer cases arising in five different sites across the United States. We used a case-only design to examine the potential interaction between acetylation status – as assigned by NAT2 genotype – and self-reported active or passive smoking status. We hypothesized that slow acetylators would be at increased risk of breast cancer associated with both active and passive smoking, and that these risks would be more pronounced among women whose exposure began before their first pregnancy or at an early age. The case-only design is optimal for assessing multiplicative interaction when the genotype and environmental exposure are independent of one another. This investigation is the largest case-only study to examine the interaction between NAT2 acetylation status and history of tobacco exposure as it relates to the risk of breast cancer.
Results
Among the Cape Cod population, 330 of 483 eligible women agreed to receive a sample collection kit; the remainder refused or could not be contacted. Of the 330 who received a kit, 272 returned a sample and 269 samples yielded DNA that could be genotyped. Among the second study population (from California, Rhode Island, Minnesota, and North Carolina), 372 of 410 eligible women agreed to receive a sample collection kit and the remainder refused or were unable to be contacted. Of the 372 who received a kit, 321 returned a sample and 233 had samples that yielded DNA that could be genotyped and had the requisite interview data. In both studies, 56% of eligible participants were genotyped and included in the analysis. The proportion of smokers among nonparticipants was not significantly different from that among participants in either study population. The mean age was greater among nonparticipants than among participants (mean ages 66 years versus 61, respectively, in the Cape Cod population,
P = 0.0001; and 74 versus 73 in the second study population,
P = 0.03), reflecting greater losses to follow-up among older women. Age was not associated with genotype among the participants. The proportion of participants who were slow acetylators, active smokers, and passive smokers did not vary significantly with their site of enrollment. Among the genotyped controls in the Cape Cod study, the OR for association of acetylation status (fast versus slow) with exposure to tobacco smoke (women who had ever smoked actively versus all the others studied) was 1.06 (
P = 0.90). This finding is consistent with those of earlier studies [
12,
14‐
17,
24,
25], in which acetylation status and active smoking were not significantly associated among controls.
Table
2 provides demographic and risk factor characteristics for the 502 breast cancer patients in the analytic sample according to acetylation status. The distribution of age, family history of breast cancer, and history of benign breast disease was nearly identical for fast and slow acetylators. There were small differences in alcohol consumption and BMI between fast and slow acetylators and a noticeable difference in the proportion of women who had had one to three live births (46% versus 61%, respectively). The great majority of the participants (97%) were white (data not shown).
Table 2
Distribution of breast cancer risk factors according to arylamine N-acetyltransferase activity (acetylation status)
Age (years) | | | | |
<50a
| 22 (12) | 38 (12) | 1.0 | - |
50 to 59 | 15 (8) | 28 (9) | 1.1 | 0.5 to 2.4 |
60 to 69 | 59 (32) | 103 (32) | 1.0 | 0.5 to 1.9 |
70+ | 86 (47) | 151 (47) | 1.0 | 0.6 to 1.8 |
Alcohol use | | | | |
Nondrinkera
| 45 (14) | 32 (18) | 1.0 | - |
≤ 1 drink/month | 75 (23) | 43 (24) | 1.2 | 0.7 to 2.2 |
Few drinks/month | 77 (24) | 39 (21) | 1.4 | 0.8 to 2.5 |
Few drinks/week | 76 (24) | 39 (21) | 1.4 | 0.8 to 2.5 |
Almost every day | 40 (12) | 24 (13) | 1.2 | 0.6 to 2.3 |
Unknown | 7 (2) | 5 (3) | 1.0 | 0.3 to 3.4 |
Body mass index (BMI) (kg/m2) | | | | |
<20.0a
| 30 (17) | 59 (19) | 1.0 | - |
20.0 to 24.9 | 114 (65) | 196 (62) | 0.9 | 0.5 to 1.4 |
25.0 to 29.9 | 24 (14) | 50 (16) | 1.1 | 0.5 to 2.0 |
30.0+ | 8 (4) | 11 (3) | 0.7 | 0.3 to 1.9 |
First-degree family history of breast cancerb
|
Noa
| 140 (78) | 240 (77) | 1.0 | - |
Yes | 39 (22) | 72 (23) | 1.1 | 0.7 to 1.7 |
Parity | | | | |
Nulliparousa
| 72 (23) | 27 (15) | 1.0 | - |
1 live birth | 22 (7) | 22 (12) | 0.4 | 0.2 to 0.8 |
2 live births | 56 (17) | 43 (24) | 0.5 | 0.3 to 0.9 |
3 live births | 70 (22) | 48 (25) | 0.5 | 0.3 to 1.0 |
4 live births | 46 (14) | 21 (12) | 0.8 | 0.4 to 1.6 |
5 or more live births | 54 (17) | 21 (12) | 1.0 | 0.5 to 1.9 |
Geographical location | | | | |
Cape Cod, Massachusettsa
| 91 (50) | 178 (56) | 1.0 | - |
Los Angeles, California | 27 (15) | 33 (10) | 0.6 | 0.3 to 1.1 |
Rhode Island | 17 (9) | 38 (12) | 1.1 | 0.6 to 2.1 |
Minnesota | 28 (15) | 34 (11) | 0.6 | 0.3 to 1.1 |
North Carolina | 19 (10) | 37 (12) | 1.0 | 0.5 to 1.8 |
History of benign breast disease |
Noa
| 116 (65) | 206 (65) | 1.0 | - |
Yes | 63 (35) | 110 (35) | 1.0 | 0.7 to 1.4 |
We observed no substantial departure from multiplicativity between acetylation status and history of ever having smoked actively (adjusted OR estimate of departure from multiplicativity (AOR) = 0.9) or of ever having experienced passive residential exposure to tobacco smoke (AOR = 0.7) (Table
3, which also shows confidence intervals). The ratios of the upper limits of the intervals to their lower limits were about 3 and 3.7 for the crude and adjusted estimates of effect, respectively (Table
3). These ratios measure the precision of the estimates of effect and indicate adequate precision about these estimates.
Table 3
Departure from multiplicativity between acetylation status and smoking status among patients with breast cancer
Active smoker | 162 | 85 | 1.04 | 0.59 to 1.82 | 0.88 | 0.45 to 1.70 |
Passive smoker | 114 | 73 | 0.85 | 0.48 to 1.52 | 0.75 | 0.39 to 1.45 |
Nonsmoker | 44 | 24 | 1.0 | - | 1.0 | - |
Estimates for the departure from multiplicativity between acetylation status and the various measures of intensity, duration, and timing of active and passive tobacco exposure are presented in Table
4. For active smokers, we found estimates lacking consistent directionality. The AOR estimates for women in the categories with the highest intensity (packs/day) and greatest length (in years) of smoking were in opposite directions. For example, the departure from multiplicativity was above the null for women who had smoked two or more packs per day (AOR = 1.8) but below the null for women who had smoked for 40 or more years (AOR = 0.7). For the variables describing the age at which a woman began smoking, when she began smoking in relation to her first live birth, and the time elapsed since she quit smoking, we observed estimates of departure from multiplicativity both above and below the null.
Table 4
Departure from multiplicativity between acetylation status and smoking status among patients with breast cancer
Nonsmokers
| 44 | 24 | 1.0 | 1.0 | - |
Active smokers
|
Packs (per day) | | | | | |
<1 | 96 | 47 | 1.11 | 0.97 | 0.48 to 1.95 |
1 to <2 | 53 | 33 | 0.88 | 0.74 | 0.35 to 1.60 |
≥ 2 | 12 | 2 | 3.27 | 1.80 | 0.33 to 9.81 |
Data missing | 1 | 3 | | | |
Duration (years) | | | | | |
<20 | 39 | 29 | 0.73 | 0.59 | 0.26 to 1.35 |
20 to <40 | 79 | 25 | 1.72 | 1.32 | 0.62 to 2.81 |
40+ | 43 | 30 | 0.78 | 0.74 | 0.33 to 1.63 |
Data missing | 1 | 1 | | | |
Duration (years) in relation to first birth |
Nulliparous | 41 | 11 | 2.03 | 1.05 | 0.32 to 3.41 |
All before first | 11 | 8 | 0.75 | 0.60 | 0.17 to 2.10 |
Before and after first | 90 | 48 | 1.02 | 1.03 | 0.48 to 2.20 |
All after first | 19 | 17 | 0.61 | 0.52 | 0.19 to 1.37 |
Data missing | 1 | 1 | | | |
Age started (years) | | | | | |
≤ 13 | 4 | 3 | 0.73 | 0.63 | 0.11 to 3.48 |
14 to 15 | 22 | 7 | 1.71 | 1.94 | 0.61 to 6.19 |
16 to 17 | 35 | 18 | 1.06 | 0.79 | 0.32 to 1.94 |
18 to 21 | 67 | 35 | 1.04 | 0.94 | 0.45 to 1.99 |
22 to 29 | 19 | 11 | 0.94 | 0.84 | 0.31 to 2.30 |
≥ 30 | 14 | 10 | 0.76 | 0.60 | 0.21 to 1.71 |
Data missing | 1 | 1 | | | |
Quit before diagnosis date (years) |
Current or <5 | 32 | 12 | 1.45 | 1.25 | 0.50 to 3.19 |
5 to 15 | 37 | 19 | 1.06 | 0.82 | 0.35 to 1.95 |
>15 | 71 | 42 | 0.92 | 0.86 | 0.41 to 1.80 |
Data missing | 22 | 12 | | | |
Passive smokers
|
Duration (years)b
| | | | | |
<20 | 34 | 18 | 1.03 | 0.99 | 0.43 to 2.33 |
20 to <40 | 48 | 36 | 0.73 | 0.61 | 0.29 to 1.28 |
40+ | 29 | 18 | 0.88 | 0.81 | 0.33 to 2.01 |
Data missing | 3 | 1 | | | |
Duration (years) in relation to first birth |
Nulliparous | 20 | 14 | 0.78 | 0.52 | 0.16 to 1.69 |
All before first | 20 | 8 | 1.36 | 1.85 | 0.59 to 5.83 |
Before and after first | 56 | 34 | 0.90 | 0.91 | 0.39 to 2.10 |
All after first | 6 | 5 | 0.65 | 0.39 | 0.09 to 1.73 |
Data missing | 12 | 12 | | | |
First lived with smoker (age in years) |
<12 | 51 | 35 | 0.84 | 0.83 | 0.29 to 2.34 |
12 to 20 | 18 | 5 | 3.11 | 2.43 | 0.41 to 14.3 |
>20 | 35 | 28 | 0.89 | 0.48 | 0.17 to 1.42 |
Data missing | 10 | 5 | | | |
The estimates for departure from multiplicativity between acetylation status and the measures of duration and timing of passive exposure to tobacco also lacked consistency. For the variable describing the duration of passive exposure to tobacco in the residence, we observed null and less than null associations with slow acetylation status: AOR = 1.0 for <20 years, 0.6 for 20 to <40 years, and 0.8 for 40+ years. We observed a departure from multiplicativity between slow acetylation status and passive exposure occurring exclusively before a woman's first live birth (AOR = 1.9), and a positive departure for women whose first passive exposure to tobacco smoke occurred between the ages of 12 and 20 (AOR = 2.4).
Discussion
In this, the largest case-only study to evaluate the interaction between acetylation status and exposure to tobacco smoke, we found little evidence to support a departure from multiplicativity between acetylation status and a history of active smoking for women with breast cancer. There is some suggestion that women who were slow acetylators were at higher risk from passive exposure to tobacco smoke before their first live birth than women who had never been either passive or active smokers. A similar positive departure was observed for women who were first passively exposed between the ages of 12 and 20. The effect estimates observed in this study for measures of intensity, duration, and timing of exposure showed no consistent pattern and in some instances were statistically unstable.
Our study is one of only a few to assess the interaction between exposure to tobacco smoke and acetylation status in relation to breast cancer risk. Hunter and colleagues [
16], in addition to examining the association between slow acetylation status and the risk of breast cancer, for which they reported a null association, found no evidence of an interaction between recent smoking status and NAT2 acetylation status among 706 postmenopausal women (cases and controls). Recently, two studies that removed passive smokers from the analysis of the unexposed group found suggestions of an interaction between tobacco exposure and acetylation status. Both reported a greater breast cancer risk among passive smokers who were fast acetylators [
24,
25]. The findings among active smokers were not consistent, however. Morabia and colleagues [
25] found that active smokers who were fast acetylators were at greater risk, whereas Chang-Claude and colleagues [
24] found the greater risk from active smoking among slow acetylators. By parsing their cases into contingency tables (genotype by smoking group) and applying a case-only analysis, we obtained estimates of departure from multiplicativity for both studies very similar to ours, but with wider CIs.
In the only other case-only analysis, Ambrosone and colleagues [
12] found a strong positive departure from multiplicativity between acetylation status and smoking at an early age (<18) and for smoking 20 or more cigarettes 20 years previously. Overall, we found departures from multiplicativity between acetylation status in relation to both active and passive exposure to tobacco smoke below the null. In subanalyses, we did find a positive departure from multiplicativity between acetylation status and smoking initiation between 14 and 15 years of age, and, separately, for first passive exposure to tobacco smoke between the ages of 12 and 20. Both findings are consistent with the hypothesis that environmental insults to developing breast tissue may increase the tissue's susceptibility to carcinogenesis, and thus may increase a woman's risk of breast cancer [
38,
39]. However, the lack of a consistent directionality to our estimates for the other age-at-initiation categories (≤ 13 and 16 to 17 years) suggests that these may be chance findings.
To date, numerous polymorphisms on the NAT2 gene have been identified (Table
1), which has furthered our understanding of NAT2 phenotypes and improved our ability to assign acetylation status to the breast cancer cases in this study. The genotyping procedures employed in this analysis are more accurate than the PCR-RFLP (PCR–restriction fragment length polymorphism) techniques used in previous studies [
12,
16,
25]. Consequently, the rates of misclassification of acetylation status in this study should be less than in those studies.
Misclassification of either the genetic or environmental variables involved in an assessment of interaction by case–control design can give rise to the appearance of interaction when, in fact, there is none [
40]. Our analysis of interaction using case-only data provides greater control over the impact of potential misclassification errors, because there are only two variables that are susceptible to misclassification – acetylation status and smoking status. If the misclassification rates are nondifferential, as one would expect, then the estimates of departure from multiplicativity will be biased towards the null [
41]. As discussed above, in the previous case–control analyses, the impact of misclassification is less predictable. It is therefore possible that findings from previous studies evaluating the interaction of acetylation status and exposure to tobacco smoke in relation to breast cancer risk may have generated spurious estimates of interaction, even if the misclassification was nondifferential. As discussed above, case-only estimates derived from these studies were similar to ours. The attenuation of the interaction after reanalysis using the case-only design further suggests that the published case–control results may have been more susceptible to misclassification. By genotyping more SNPs with a more accurate method and by implementing a case-only design, our analysis provides a more valid assessment of the multiplicative interaction between NAT2 genotype and exposure to tobacco smoke in relation to breast cancer.
Weighing against this advantage of the case-only design is the limitation that only departure from multiplicativity can be assessed. Many epidemiologists weigh departure from additive interaction more heavily, arguing that the additive scale corresponds better to the biologic meaning of synergistic effects [
42]. A further limitation of the case-only design is its reliance on the assumption that the genetic polymorphisms and environmental exposure are independent of one another [
43]. Violations of this assumption can substantially distort the estimates of interaction. However, NAT2 polymorphisms and smoking history were not associated among the genotyped controls in the Cape Cod study or among the controls in earlier studies [
12,
14‐
17,
24,
25]. The absence of association supports the assumption of independence required to validly estimate departure from multiplicativity with the case-only design.
These results must be interpreted with the following additional limitations in mind. First, only 56% of eligible cases were available for analysis. Participation was not related to smoking status and although participation was related to age, age was not related to genotype. We expect that the selection of participants introduced no substantial bias, although we acknowledge that our study of breast cancer survivors may have influenced the estimates of effect in ways that we are unable to anticipate. Second, haplotypes were inferred from genotyping data by assigning the haplotype with the maximum probability to each case. Forty-one percent of haplotype assignments had probabilities of 100% and 91% had probabilities of 80% or better. Less than 5% had probabilities of less than 50%. We expect that the procedure used to infer haplotypes introduced little error.
Acknowledgements
This project was primarily supported by grant K07 CA87724 from the National Cancer Institute, National Institutes of Health. Interview data for the Cape Cod population was collected with support by grant number 2P42 ES07381 from the National Institute of Environmental Health Sciences, National Institutes of Health, with funds from the Environmental Protection Agency. Interview data for the second population were collected with support from grants R01 CA/AG70818 from the National Cancer Institute and National Institute on Aging, National Institutes of Health. The publication's contents are solely the responsibility of the authors and do not necessarily reflect the official views of the NIEHS, the NIH, or the EPA.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TL conceived of the study, collected genotyping samples, participated in data analysis, and drafted the manuscript. BB conducted data analysis and drafted the manuscript. JW inferred haplotypes from the genotyping data using SNPHAP and drafted the manuscript. AA collected interview data from the Cape Cod population and drafted the manuscript. All authors read and approved the final manuscript.