Genetic Variations in Xenobiotic Metabolic Pathway Genes, Personal Hair Dye Use, and Risk of Non-Hodgkin Lymphoma

Abstract

From 1996 to 2000, the authors conducted a population-based case-control study among Connecticut women to test the hypothesis that genetic variation in xenobiotic metabolic pathway genes modifies the relation between hair dye use and risk of non-Hodgkin lymphoma. No effect modifications were found for women who started using hair dyes in 1980 or afterward. For women who started using hair dye before 1980 as compared with never users, a statistically significantly increased risk of non-Hodgkin lymphoma was found for carriers of CYP2C9 Ex3-52C>T TT/CT genotypes (odds ratio (OR) = 2.9, 95% confidence interval (CI): 1.4, 6.1), CYP2E1 -332T>A AT/AA genotypes (OR = 2.0, 95% CI: 1.2, 3.4), a homozygous or heterozygous 3-base-pair deletion in intron 6 of GSTM3 (OR = 2.3, 95% CI: 1.3, 4.1), GSTP1 Ex5-24A>G AA genotypes (OR = 1.8, 95% CI: 1.1, 2.9), or NAT2 genotypes conferring intermediate/rapid acetylator status (OR = 1.6, 95% CI: 1.0, 2.7). The observed associations were mainly seen for follicular lymphoma. In contrast, no significantly increased risk was observed for starting hair dye use before 1980 (relative to never use) among women who were homozygous wild-type for the CYP2C9, CYP2E1, or GSTM3 polymorphisms, women carrying 1 or 2 copies of the variant GSTP1 allele, or women who were slow NAT2 acetylators. A possible role of genetic variation in xenobiotic metabolism in the carcinogenicity of hair dye use needs to be confirmed in larger studies.

Personal use of hair dye has been suggested to be associated with the risk of non-Hodgkin lymphoma (NHL), particularly for women who started using hair dyes before 1980 (1–4). This possible relation is supported by the fact that hair dye products manufactured before 1980 contained a variety of arylamine carcinogens, such as 2,4-diaminoanisole, 2,4-diaminotoluene, 4-amino-2-nitrophenol, etc., which were demonstrated to produce tumors in experimental animal studies (5). A possible relation between hair dye use and risk of follicular lymphoma, a major subtype of NHL, among women who started using hair dyes in 1980 or afterward was suggested by a recent pooled analysis, in which Zhang et al. (4) postulated a potential carcinogenic effect of certain chemicals used in current hair dye formulations or possible carcinogenic contaminants of hair dye products. Given the high prevalence of personal hair dye use in the general population, especially among women (approximately 75% of women in the United States have used hair dye products (3)), there is a great deal of public health concern regarding the role of personal hair dye use in NHL tumorigenesis.

Xenobiotic metabolic enzymes constitute one of the first lines of defense against environmental carcinogens and include the cytochrome P-450s (CYPs), the glutathione S-transferases (GSTs), and the N-acetyltransferases (NATs). Environmental carcinogens can be activated into more toxic forms by CYPs (phase I metabolism) and then conjugated by GSTs or NATs (phase II metabolism) to increase the solubility of the products and subsequently facilitate their excretion (6). Many of these enzymes are genetically polymorphic and may be responsible for different patterns of individual susceptibility to the effect of environmental carcinogens. Recent evidence shows that genetic variation in NAT1 and NAT2 may modify the relation between hair dye use and risk of NHL (2). However, to our knowledge, no study has investigated the role of genetic variation of other enzymes involved in arylamine activation and/or detoxification (i.e., CYPs and GSTs) in the relation between hair dye use and risk of NHL. Using data from a population-based case-control study of NHL in Connecticut women, we examined the influence of genetic mutations in xenobiotic metabolic pathway genes (CYP1A1, CYP1A2, CYP1B1, CYP2C9, CYP2E1, GSTP1, GSTM3, NAT1, and NAT2) on the relation between hair dye use and risk of NHL.

MATERIALS AND METHODS

Study population

A detailed description of the study population has been published previously (3, 7). Briefly, among female Connecticut residents, 601 cases with histologically confirmed incident NHL diagnosed between 1996 and 2000 and 717 population-based controls aged 21–84 years with no prior history of cancer, aside from nonmelanoma skin cancer, were enrolled and completed in-person interviews. Cases were identified through the Yale Cancer Center's Rapid Case Ascertainment Shared Resource, histologically confirmed by 2 independent study pathologists, and classified into NHL subtypes according to the World Health Organization classification system (8). Control women were identified via random-digit dialing (for those aged <65 years) or random selection from Centers for Medicare and Medicaid Services records (for those aged ≥65 years) and were frequency-matched to the cases on age.

The case participation rate was 72%. The participation rate for controls was 69% among persons identified via random-digit dialing and 47% among persons identified from the Centers for Medicare and Medicaid Services records. Approximately 75% of the study subjects (76.7% of the cases and 74.6% of the controls) provided blood samples, and approximately 10% of the subjects (11.0% of the cases and 10.4% of the controls) provided buccal cell samples for genotyping. These analyses were conducted among the 461 cases and 535 controls who provided blood samples out of the approximately 835 cases and 1,264 controls invited to participate, yielding effective response rates for cases and controls of approximately 55% and 42%, respectively.

Genotyping

DNA was extracted from blood or buccal cell samples using phenol-chloroform extraction (9). A total of 19 single nucleotide polymorphisms in 9 xenobiotic genes, including CYP1A1 (rs1048943), CYP1A2 (rs762551), CYP1B1 (rs1056836), CYP2C9 (rs1799853), CYP2E1 (rs2070673 and rs2031920), GSTM3 (rs1799735), GSTP1 (rs1695 and rs1138272), NAT1 (rs4987076, rs13249533, rs1057126, and rs15561), and NAT2 (rs1041983, rs1801280, rs1799929, rs1799930, rs1208, and rs1799931), were selected and genotyped in blood-based DNA samples because of the insufficiency of DNA samples from buccal cells. Genotyping was conducted using real-time polymerase chain reaction on an ABI 7900HT sequence detection system (Applied Biosystems, Inc., Foster City, California) as described on the Web site of the Cancer Genome Anatomy Project (http://snp500cancer.nci.nih.gov) at the National Cancer Institute Core Genotyping Facility (10). Duplicate samples from 100 study subjects and 40 replicate samples from each of 2 blood donors were interspersed throughout the plates used for genotype analysis. The concordance rates for quality control samples were 100% for all assays. All genotyping frequencies among control populations were in Hardy-Weinberg equilibrium (P > 0.05). The single nucleotide polymorphism data were used to assign the most likely NAT1 and NAT2 alleles and NAT2 acetylation phenotypes by Dr. David W. Hein at the University of Louisville (Louisville, Kentucky) (11).

Interview

Information on use of hair dyes was obtained through in-person interviews. The study subjects were first asked whether they had ever used any hair-coloring products. Those who had were asked to provide information on age and calendar years at first and last use, as well as frequency and duration of use and type and color of each product they had used. For time period of use, the subjects were categorized on the basis of whether they had started using hair dyes before 1980 or in 1980 or later (i.e., no use of any type prior to 1980). Information on potentially confounding variables, including cigarette smoking, alcohol consumption, farming history, dietary intake, and demographic factors, was also collected during the interview.

Statistical analysis

Unconditional logistic regression models were used to estimate the associations of hair dye use and genetic variation in metabolic pathway genes with risk of NHL, overall and by the following subtypes: diffuse large B-cell lymphoma, follicular lymphoma, chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), marginal zone B-cell lymphoma, and T-cell lymphoma. To increase statistical power, we combined heterozygous and homozygous variant genotypes for CYP1A1 (rs1048943), CYP1A2 (rs762551), CYP1B1 (rs1056836), CYP2C9 (rs1799853), CYP2E1 (rs2070673 and rs2031920), GSTM3 (rs1799735), and GSTP1 (rs1695 and rs1138272). According to previous research (12, 13), the NAT1*10 allele was treated as the at-risk allele for NAT1 analyses, while NAT2 slow-acetylator alleles were treated as the referent phenotype and intermediate/rapid acetylator status was treated as the at-risk group for NAT2 analyses. The false discovery rate method with the false discovery rate set at 0.2 was used to control for multiple comparisons in the different sets of models. Potentially confounding variables included in the final models were age (years; continuous) and race (white vs. nonwhite). Adjustment for other variables, such as cigarette smoking, alcohol consumption, and farming history, did not result in material changes in the observed associations; thus, these variables were not included in the final models. Analyses were also repeated in the subgroup of white, non-Hispanic participants (444 cases and 504 controls). Odds ratios and 95% confidence intervals were calculated using the SAS statistical software package (version 9.1; SAS Institute Inc., Cary, North Carolina).

RESULTS

The associations between hair dye use and NHL risk in the population with blood samples only (results not shown) were similar to the results observed in the overall population, which were reported previously (3). A 40% increased risk of NHL was observed among women who started using hair dyes before 1980 (versus a 30% increase among these women in the overall study population), while no increased risk was found among women who started using hair dyes in 1980 or afterward, which echoed the finding for the overall study population. For women who started using hair dyes before 1980, the statistically significantly increased risk was mainly driven by associations with follicular lymphoma (odds ratio (OR) = 1.9, 95% confidence interval (CI): 1.1, 3.3) and CLL/SLL (OR = 1.9, 95% CI: 1.0, 3.9) (results not shown).

Table 1 presents the main effects for NHL overall and the 3 most common NHL subtypes—diffuse large B-cell lymphoma, follicular lymphoma, and CLL/SLL—for all 11 metabolic variants considered in these analyses. None of these variants were associated with a significant change in NHL risk, overall or for any of the subtypes considered.

Table 1.

Associations Between Variations in Metabolic Genes and Risk of Non-Hodgkin Lymphoma (Overall and by Subtype) Among Connecticut Women, 1996–2000^a

Single Nucleotide Polymorphism	All Non-Hodgkin Lymphoma				Diffuse Large B-Cell Lymphoma				Follicular Lymphoma				CLL/SLL
	No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI
CYP1A1 (rs1048943)
AA	418	487	1.0	Referent	129	487	1.0	Referent	98	487	1.0	Referent	49	487	1.0	Referent
AG + GG	33	39	1.0	0.6, 1.6	15	39	1.5	0.8, 2.7	6	39	0.8	0.3, 1.9	4	39	1.0	0.4, 3.0
CYP1A2 (rs762551)
AA	211	267	1.0	Referent	72	267	1.0	Referent	50	267	1.0	Referent	25	267	1.0	Referent
AC + CC	245	267	1.2	0.9, 1.5	74	267	1.1	0.7, 1.5	55	267	1.1	0.8, 1.7	28	267	1.1	0.6, 1.9
CYP1B1 (rs1056836)
CC	159	185	1.0	Referent	57	185	1.0	Referent	33	185	1.0	Referent	23	185	1.0	Referent
CG + GG	297	350	1.0	0.8, 1.3	90	350	0.9	0.6, 1.3	72	350	1.2	0.8, 1.9	30	350	0.7	0.4, 1.2
CYP2C9 (rs1799853)
CC	358	411	1.0	Referent	115	411	1.0	Referent	75	411	1.0	Referent	45	411	1.0	Referent
CT + TT	97	122	0.9	0.6, 1.2	31	122	0.8	0.5, 1.3	29	122	1.2	0.8, 2.0	9	122	0.7	0.3, 1.5
CYP2E1 (rs2070673)
AA	295	353	1.0	Referent	100	353	1.0	Referent	68	353	1.0	Referent	36	353	1.0	Referent
AT + TT	162	181	1.1	0.9, 1.5	47	181	1.0	0.7, 1.4	37	181	1.2	0.7, 1.8	17	181	0.9	0.5, 1.7
CYP2E1 (rs2031920)
CC	420	490	1.0	Referent	137	490	1.0	Referent	101	490	1.0	Referent	49	490	1.0	Referent
CT + TT	23	33	0.8	0.5, 1.4	6	33	0.6	0.3, 1.5	3	33	0.4	0.1, 1.5	2	33	0.6	0.1, 2.7
GSTM3 (rs1799735)
+ / +	321	372	1.0	Referent	105	372	1.0	Referent	79	372	1.0	Referent	34	372	1.0	Referent
+ /− and −/−	138	159	1.0	0.8, 1.4	42	159	0.9	0.6, 1.4	27	159	0.8	0.5, 1.3	19	159	1.4	0.7, 2.5
GSTP1 (rs1695)
AA	187	230	1.0	Referent	51	230	1.0	Referent	43	230	1.0	Referent	24	230	1.0	Referent
AG + GG	271	302	1.1	0.9, 1.4	96	302	1.4	1.0, 2.1	62	302	1.1	0.7, 1.7	30	302	1.0	0.5, 1.7
GSTP1 (rs1138272)
CC	384	440	1.0	Referent	122	440	1.0	Referent	87	440	1.0	Referent	44	440	1.0	Referent
CT + TT	75	95	0.9	0.6, 1.2	25	95	0.9	0.6, 1.5	18	95	0.9	0.5, 1.6	10	95	1.1	0.5, 2.2
NAT1 genotype
NAT1*any (except*10)/NAT1*any (except*10)	289	328	1.0	Referent	104	328	1.0	Referent	69	328	1.0	Referent	32	328	1.0	Referent
NAT1*10/*any (except*10) or NAT1*10/*10	164	194	1.0	0.8, 1.3	40	194	0.7	0.4, 1.0	37	194	0.9	0.6, 1.5	22	194	1.2	0.7, 2.1
NAT2 phenotype
Slow acetylator	257	315	1.0	Referent	88	315	1.0	Referent	58	315	1.0	Referent	30	315	1.0	Referent
Fast/intermediate acetylator	199	218	1.1	0.9, 1.5	58	218	0.9	0.6, 1.4	47	218	1.2	0.8, 1.8	24	218	1.2	0.7, 2.1

Abbreviations: CI, confidence interval; CLL/SLL, chronic lymphocytic leukemia/small lymphocytic leukemia; CYP, cytochrome P-450; GST, glutathione-S-transferase; NAT, N-acetyltransferase; OR, odds ratio.

^aNumbers of cases and controls for each of the genes for a given non-Hodgkin lymphoma outcome may not sum to the total because of incomplete data for some of the single nucleotide polymorphisms studied.

^bAdjusted for age (continuous) and race (white vs. nonwhite).

Table 2 presents the effects of hair dye use on NHL risk according to the homozygous wild-type and heterozygous/homozygous variant genotypes of CYP2C9 Ex3-52C>T and CYP2E1 -332T>A. Among women who carried the variant CYP2C9 allele (TT or CT genotypes), an approximately 3-fold increased risk was found for those who started using hair dyes before 1980 (OR = 2.9, 95% CI: 1.4, 6.1), and the risk reached 6-fold for follicular lymphoma (OR = 6.3, 95% CI: 1.6, 24.7). No statistically significant risk difference was observed for women who carried the CYP2C9 CC genotypes. The P value for interaction between starting hair dye use before 1980 and the CYP2C9 genotypes with respect to NHL overall was statistically significant (P = 0.033), and that for the interaction with respect to follicular lymphoma was borderline significant (P = 0.055). Women who carried CYP2E1 AT or AA genotypes, but not those who carried TT genotypes, experienced a 2-fold increased risk (OR = 2.0, 95% CI: 1.2, 3.4) if they had started using hair dyes before 1980, and the risk was almost 4-fold for follicular lymphoma (OR = 3.9, 95% CI: 1.4, 11.1). The P values for the interaction terms did not achieve statistical significance (for NHL overall and for follicular lymphoma, P = 0.245 and P = 0.161, respectively). No discernible difference in risk of NHL associated with hair dye use was observed for the CYP1A1 Ex7 + 131A>G rs1048943, CYP1A2 IVS1-154C>A rs762551, or CYP1B1 Ex3 + 251G>C rs1056836 genotype (data not shown).

Table 2.

Associations Between Use of Hair Dye and Risk of Non-Hodgkin Lymphoma According to Cytochrome P-450 Genotype Among Connecticut Women, 1996–2000

	CYP2C9 (rs1799853)								P Valuea	CYP2E1 (rs2070673)								P Valuea
	CC				CT + TT					TT				AT + AA
	No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI		No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI
Total
Never use	100	124	1.0	Referent	16	33	1.0	Referent		75	98	1.0	Referent	41	59	1.0	Referent
Ever use	258	287	1.1	0.8, 1.5	81	89	1.9	1.0, 3.8	0.154	220	255	1.1	0.8, 1.6	121	122	1.4	0.9, 2.3	0.434
Year of starting hair dye use (vs. never use)
Before 1980	166	170	1.2	0.9, 1.7	58	43	2.9	1.4, 6.1	0.033	143	148	1.3	0.9, 1.8	81	65	2.0	1.2, 3.4	0.245
1980 or later	92	117	0.9	0.6, 1.4	23	46	1.0	0.4, 2.3	0.889	77	107	1.0	0.6, 1.5	40	57	0.9	0.5, 1.7	0.808
Non-Hodgkin lymphoma subtype among women who started using hair dye before 1980 (vs. never use)
Diffuse large B-cell lymphoma	46	170	1.0	0.6, 1.6	15	43	2.0	0.7, 5.9	0.212	43	148	1.1	0.6, 1.9	18	65	1.2	0.5, 2.7	0.918
Follicular lymphoma	35	170	1.4	0.7, 2.6	20	43	6.3	1.6, 24.7	0.055	34	148	1.4	0.7, 2.8	21	65	3.9	1.4, 11.1	0.161
CLL/SLL	25	170	1.7	0.8, 3.5	6	43	4.6	0.5, 40.1	0.386	20	148	1.5	0.6, 3.4	11	65	3.2	0.9, 12.4	0.302

Abbreviations: CI, confidence interval; CLL/SLL, chronic lymphocytic leukemia/small lymphocytic leukemia; CYP, cytochrome P-450; OR, odds ratio.

^aP value for the interaction between a given hair dye exposure variable and genotype variations in a logistic regression model, with adjustment for age (continuous) and race (white vs. nonwhite).

^bLogistic regression model with adjustment for age (continuous) and race (white vs. nonwhite).

Table 3 shows the results for hair dye use and NHL according to GSTM3 IVS6 + 22AGG>- and GSTP1 Ex5-24A>G genotypes. An increased risk associated with starting hair dye use before 1980 was observed for NHL overall (OR = 2.3, 95% CI: 1.3, 4.1), follicular lymphoma (OR = 3.7, 95% CI: 1.3, 10.6), and marginal zone B-cell lymphoma (OR = 8.7, 95% CI: 1.0, 73.8; data not shown) in women who carried a homozygous or heterozygous 3-base-pair deletion in intron 6 of the GSTM3 allele. No significant risk increase was found in women who carried 2 GSTM3 wild-type alleles. The P value for interaction between starting hair dye use before 1980 and the GSTM3 genotypes was statistically significant with respect to NHL overall (P = 0.047) but not with respect to follicular lymphoma (P = 0.104) or marginal zone B-cell lymphoma (P = 0.173; data not shown). An increased risk was also observed for women with GSTP1 Ex5-24A>G AA genotypes (for NHL overall, OR = 1.8, 95% CI: 1.1, 2.9; for follicular lymphoma, OR = 2.2, 95% CI: 0.9, 5.3) but not for those with GSTP1 AG or GG genotypes. However, the interaction between starting hair dye use before 1980 and the GSTP1 genotypes did not achieve statistical significance (for NHL overall and for follicular lymphoma, P = 0.247 and P = 0.486, respectively).

Table 3.

Associations Between Use of Hair Dye and Risk of Non-Hodgkin Lymphoma According to Glutathione S-Transferase Genotype Among Connecticut Women, 1996–2000

	GSTM3 (rs1799735)								P Valuea	GSTP1 (rs1695)								P Valuea
	+/+				+/− and −/−					AA				AG + GG
	No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI		No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI
Total
Never use	84	100	1.0	Referent	35	58	1.0	Referent		42	65	1.0	Referent	75	93	1.0	Referent
Ever use	237	272	1.0	0.7, 1.4	103	101	1.7	1.0, 2.9	0.114	145	165	1.3	0.8, 2.1	196	209	1.1	0.8, 1.6	0.632
Year of starting hair dye use (vs. never use)
Before 1980	158	165	1.1	0.8, 1.7	64	46	2.3	1.3, 4.1	0.047	103	90	1.8	1.1, 2.9	120	120	1.3	0.8, 1.9	0.247
1980 or later	79	107	0.8	0.5, 1.2	39	55	1.3	0.7, 2.4	0.404	42	75	0.9	0.5, 1.5	76	89	1.0	0.7, 1.6	0.521
Non-Hodgkin lymphoma subtype among women who started using hair dye before 1980 (vs. never use)
Diffuse large B-cell lymphoma	45	165	0.9	0.5, 1.6	16	46	1.9	0.8, 4.6	0.190	26	90	1.4	0.7, 3.0	35	120	1.0	0.6, 1.8	0.461
Follicular lymphoma	39	165	1.4	0.7, 2.6	16	46	3.7	1.3, 10.6	0.104	24	90	2.2	0.9, 5.3	30	120	1.7	0.8, 3.4	0.486
CLL/SLL	21	165	2.0	0.8, 5.2	9	46	1.9	0.6, 5.8	0.863	15	90	2.2	0.7, 6.5	16	120	1.8	0.7, 4.5	0.791

Abbreviations: CI, confidence interval; CLL/SLL, chronic lymphocytic leukemia/small lymphocytic leukemia; GST, glutathione-S-transferase; OR, odds ratio.

^aP value for the interaction between a given hair dye exposure variable and genotype variations in a logistic regression model, with adjustment for age (continuous) and race (white vs. nonwhite).

^bLogistic regression model with adjustment for age (continuous) and race (white vs. nonwhite).

The associations between hair dye use and risk of NHL, overall and by subtype, among women who carried 1 or 2 NAT1*10 alleles did not differ significantly from those for women who did not carry any NAT1*10 allele (Table 4). None of the P values for interaction were statistically significant at the 0.05 level. Among women who were intermediate/rapid NAT2 acetylators, those who had used hair dye before 1980 (relative to never users) had slightly higher risks for NHL overall (OR = 1.6, 95% CI: 1.0, 2.7), follicular lymphoma (OR = 2.8, 95% CI: 1.1, 7.2), and CLL/SLL (OR = 3.2, 95% CI: 1.0, 10.2), whereas these increases were not observed among women who were slow acetylators. None of the P values for interaction with the various outcomes were statistically significant.

Table 4.

Associations Between Use of Hair Dye and Risk of Non-Hodgkin Lymphoma According to N-Acetyltransferase NAT1*10 Genotype and NAT2 Phenotype Among Connecticut Women, 1996–2000

	NAT1*10								P Valuea	NAT2								P Valuea
	NAT1*any (except*10)/*any (except*10)				NAT1*10/*any (except*10) or NAT1*10/*10					Slow Acetylator				Rapid/Intermediate Acetylator
	No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI		No. of Cases	No. of Controls	ORb	95% CI	No. of Cases	No. of Controls	ORb	95% CI
Total
Never use	66	91	1.0	Referent	49	66	1.0	Referent		69	93	1.0	Referent	47	65	1.0	Referent
Ever use	215	235	1.2	0.9, 1.8	123	136	1.2	0.8, 1.9	0.729	188	222	1.1	0.8, 1.6	152	153	1.3	0.9, 2.1	0.549
Year of starting hair dye use (vs. never use)
Before 1980	142	132	1.5	1.0, 2.2	80	77	1.4	0.9, 2.3	0.678	129	132	1.4	0.9, 2.0	94	79	1.6	1.0, 2.7	0.485
1980 or later	73	103	0.9	0.6, 1.5	43	59	1.0	0.6, 1.7	0.963	59	90	0.9	0.5, 1.4	58	74	1.1	0.6, 1.8	0.590
Non-Hodgkin lymphoma subtype among women who started using hair dye before 1980 (vs. never use)
Diffuse large B-cell lymphoma	44	132	1.1	0.7, 2.0	15	77	1.0	0.4, 2.1	0.747	36	132	1.1	0.6, 2.1	24	79	1.1	0.5, 2.2	0.919
Follicular lymphoma	34	132	1.7	0.9, 3.4	21	77	2.2	0.9, 5.3	0.745	31	132	1.6	0.8, 3.3	24	79	2.8	1.1, 7.2	0.416
CLL/SLL	18	132	2.4	0.9, 6.7	13	77	1.7	0.6, 4.5	0.345	15	132	1.4	0.6, 3.6	16	79	3.2	1.0, 10.2	0.216

Abbreviations: CI, confidence interval; CLL/SLL, chronic lymphocytic leukemia/small lymphocytic leukemia; NAT, N-acetyltransferase; OR, odds ratio.

^aP value for the interaction between a given hair dye exposure variable and genotype variations in a logistic regression model, with adjustment for age (continuous) and race (white vs. nonwhite).

^bLogistic regression model with adjustment for age (continuous) and race (white vs. nonwhite).

Restricting the analyses to white non-Hispanics (data not shown) did not substantially change the results. However, after we used the false discovery rate approach to adjust for multiple comparisons in the interaction assessments, none of the gene-hair dye interactions or stratum-specific hair dye effects for the genotypic variants highlighted in Tables 2–4 remained statistically significant.

We also separately examined the magnitude of the gene-hair dye interaction odds ratios in relation to NHL overall for women who started using hair dyes before and after 1980 (data not shown). We observed that the interaction effects tended to cluster around 1.0 for women who started using hair dyes during the later time period, versus a tendency towards interaction effects of higher magnitude for those who started using hair dyes before 1980.

DISCUSSION

The results of our population-based case-control study are consistent with the hypothesis that both phase I and phase II metabolic enzymes may modify the relation between hair dye use and risk of NHL, although none of our findings achieved statistical significance after adjustment for multiple comparisons. Specifically, the effect modification by genetic variations in CYP2C9, CYP2E1, GSTM3, and GSTP1 genes in our study was mainly observed among women who started using hair dyes before 1980. However, the lack of findings among women who started using hair dyes in 1980 or later could have been due to lack of statistical power, as there were fewer subjects who started using hair dyes in 1980 or later (n = 281) than before 1980 (n = 435).

Thus far, very few investigators have studied the relation between genetic polymorphisms in xenobiotic metabolic pathway genes and risk of NHL (14–19). Results from the limited studies available have been inconsistent. One possible explanation for the contradictory results is that few investigators have considered environmental carcinogenic exposures when examining the gene-disease relation. Populations involved in different studies may have different levels of exposure to various substrates.

To date, only Morton et al. (2) have examined genetic variation in NAT1 and NAT2 in relation to NHL risk and hair dye use. They found an increased risk of NHL associated with hair dye use among women who carried the NAT2 rapid/intermediate acetylator phenotype and started using hair dyes before 1980, but not among women who carried the slow NAT2 acetylator phenotype. In our study, we found a similarly increased risk of NHL associated with starting use before 1980 among women with rapid/intermediate NAT2 phenotypes versus slow NAT2 phenotypes, although a statistically significant increase was observed only for women with rapid/intermediate NAT2 phenotypes. NATs are involved in metabolism of a variety of aromatic and heterocyclic amines through N-acetylation (detoxification) and/or O-acetylation (activation) (20). NAT2 rapid acetylators have greater O-acetylation activity, resulting in increased bioactivation of aromatic and heterocyclic amines and formation of DNA adducts that have been shown to lead to cancer in rodents, including lymphoma (11, 21, 22). As such, the observed higher risk associated with hair dye use for NAT2 rapid/intermediate acetylators is biologically plausible. However, the effect modification by NAT*10 genotypes observed in the previous study (2) was not replicated in our study.

In addition to effect modification by NAT2 acetylator status, we observed effect modification by CYP and GST genes on the risk of NHL associated with hair dye use. CYP2C9 is involved in bioactivation of several carcinogens, such as polycyclic aromatic hydrocarbons and hetereocyclic aromatic amines, which are present in hair dyes (23). The CYP2C9 Ex3-52C>T rs1799853 polymorphism has been shown to alter CYP2C9 activity (24, 25). As such, the finding of increased risk of NHL associated with hair dye use among women who carried the CYP2C9 variant allele, but not among those who carried 2 wild-type alleles, is biologically plausible. CYP2E1 is also a key enzyme that is involved in metabolic activation of a variety of carcinogenic and toxic compounds, including benzene and N-nitrosamines (26). Although the functional significance is currently unclear for the variant allele of CYP2E1 -332T>A rs2070673, the finding from our study that women who carried the variant allele appeared to have higher risk of NHL associated with hair dye use, despite nonsignificant results for the gene-hair dye interaction term in the models, requires further investigation. Future functional studies on this single nucleotide polymorphism are necessary to explain the meaning of the observed modification effect of CYP2E1 -332T>A.

GSTM3 plays a pivotal role in conjugation and detoxification of environmental carcinogens, such as arylamines (27). GSTM3 intron 6 rs1799735, a 3-base-pair deletion polymorphism, produces a binding site for the transcription factor YY1 [Yin Yang 1], which influences the expression of GSTM3 (28). The GSTM3 3-base-pair deletion allele has been shown to alter expression in cytosol, which may result in varying efficiency of carcinogen detoxification and therefore may predispose a person to cancer (29). This polymorphism has been reported to modify susceptibility to cigarette and tobacco carcinogens (30–33). Our study revealed a modulation of risk associated with hair dye use by this polymorphism.

Our study also found that the GSTP1 Ex5-24A>G rs1695 polymorphism modified the relation between hair dye use and risk of NHL. This polymorphism results in an Ile¹⁰⁵Val alteration in the encoded amino acid sequence. It has been shown that the Val¹⁰⁵ variant allozyme accommodates less bulky substrates than the Ile¹⁰⁵ allozyme and subsequently displays different substrate specificities (34). In addition, the Val¹⁰⁵ allozyme exhibits different thermal stability than the Ile¹⁰⁵ allozyme (35). These characteristics may be responsible for the apparent difference in the risk of NHL associated with hair dye use with respect to GSTP1 allozymes, although the P value for interaction was not statistically significant.

In this study, we did not observe significant modulation of risk associated with hair dye use with respect to CYP1A1, CYP1A2, CYP1B1, GSTT1, or NAT1 genotypes. It is possible that the single nucleotide polymorphisms we studied for these genes are not functional single nucleotide polymorphisms. As such, a possible role of those genes cannot be ruled out.

Several strengths of this study lend confidence to the observed associations. All NHL cases were histologically confirmed by 2 study pathologists who are experienced in diagnosis of lymphoma, which minimized disease misclassification, particularly for NHL subtypes. Using a rapid case ascertainment system to identify newly diagnosed NHL cases reduced potential survival bias. Detailed information on lifetime hair dye use and potentially confounding variables allowed us to examine the effect of hair dye use by time period of use, with control for several potential confounders.

Although we had a relatively large sample, the numbers of subjects became small for stratified analyses by both hair dye use and genetic variation. Given the large number of tests conducted in the study, the possibility that some or all of the observed associations were due to chance cannot be ruled out. In this study, we included women only; therefore, the results may not be generalizable to men. Differential recall bias regarding hair dye use among study subjects could have occurred if NHL patients believed that hair dye use or use of a specific type (or color) of hair dye increased a person's risk of NHL. The lack of association between overall use of hair dye or hair-coloring products after 1980 and major NHL subtypes (e.g., diffuse large B-cell lymphoma) argues against recall bias’ having played a major role in the observed association. Additionally, although we were able to examine allelic/phenotypic classifications for NAT1 and NAT2, respectively, such detailed information is not known for the other metabolic genes studied here. Thus, we were limited to the examination of individual single nucleotide polymorphisms for the other genes, which may not have captured meaningful information on gene function.

In summary, the results from this study suggest effect modification of the relation between hair dye use (particularly use that started before 1980) and NHL by variation in several genes involved in xenobiotic metabolism: CYP2C9 Ex3-52C>T, CYP2E1 -332 T>A, the GSTM3 deletion in intron 6, GSTP1 Ex5-24A>G, and NAT2 acetylator status. Although there is biologic plausibility for the majority of these findings, the results should be replicated in other studies with larger samples before stronger conclusions can be made about the roles of these genes in the carcinogenicity of hair dye use.

Glossary

Abbreviations

CI

confidence interval

CLL/SLL

chronic lymphocytic leukemia/small lymphocytic lymphoma

CYP

cytochrome P-450

GST

glutathione S-transferase

NAT

N-acetyltransferase

NHL

non-Hodgkin lymphoma

OR

odds ratio