Re: Hair dye use, genetic variation in N-acetyltransferase 1 (NAT1) and 2 (NAT2), and risk of non-Hodgkin lymphoma, author response

Dear Sir,

We thank Drs Bailey and Skare for their letter and clarification regarding the term ‘depth’ of color for the concept we labeled ‘intensity of tone’ (1). It will improve future reports on hair coloring to use the industry terminology. We are also grateful for the opportunity to clarify the content and origins of personal communication data regarding color depth. On 29 April 2003, Dr William Dressler of Proctor and Gamble/Clairol presented a graph displaying concentrations of total permanent dyes and p-phenylenediamine (PPD) by color shade and depth at a meeting of the Clairol Epidemiology Peer Review Group organized by Johns Hopkins University under contract with Clairol. By the time of our publication (1), Dr Dressler had retired and, in response to our request, Dr Skare gave us permission to use the information and stated that she should be cited as the source. She did not, however, see or approve of how the graph's information was used in the manuscript prior to publication. Descriptions and interpretation of the graph were our own and not Dr Skare's.

The graph displayed the total dye and PPD concentrations for two to seven individual products per color/depth combination. As expected, as a group, black, dark brown and dark red hair dyes have greater concentrations of permanent dyes than blonde dyes. The surprising feature of the graph was that some individual dark blonde products appeared to have higher PPD concentrations than some of the individual brown and red dye products. This contrasted with the general assumption that blonde dyes universally have much lower concentrations than dark color dyes. There was more overlap in the range of PPD concentrations for dark blonde products with those for the dark color dyes than we expected. We pursued this by developing a new analysis combining dark blonde products with other products of greater color depth, in addition to the more traditional approach of combining dark colors. Because of the concerns expressed by Drs Bailey and Skare, we have reanalyzed our data excluding dark blonde users from the group of users of products of greater color depth (Table I). The previously published results (1) based on black, dark brown and dark blonde users differ very little from results excluding the dark blonde users.

Table I.

Number of non-Hodgkin lymphoma cases and controls and odds ratios according to permanent hair coloring product use (women only)

	Black, dark brown or dark blonde permanent hair dye usea			Black or dark brown permanent hair dye use (excludes use of permanent, dark blonde dyes)
	Controls	Cases	OR (95% CI)b	Controls	Cases	OR (95% CI)b
No hair coloring product use	98	101	1.0 (referent)	98	101	1.0 (referent)
First use before 1980	36	61	1.6 (0.9, 2.7)	31	50	1.5 (0.9, 2.6)
First use in 1980 or after	42	33	0.6 (0.4, 1.1)	25	20	0.7 (0.3, 1.5)
Frequency of use before 1980 (applications/year)
1–4	9	20	2.2 (0.9, 5.3)	9	16	1.8 (0.8, 4.4)
5–7	10	22	2.1 (0.9, 4.9)	8	20	2.3 (0.9, 5.7)
≥8	17	19	1.0 (0.5, 2.1)	14	14	0.9 (0.4, 2.0)
Duration of use before 1980 (years)
1–4	20	30	1.5 (0.8, 2.9)	17	25	1.5 (0.7, 3.0)
5–14	12	14	1.0 (0.4, 2.4)	11	10	0.7 (0.3, 1.9)
≥15	4	17	3.9 (1.2, 12.5)	3	15	4.6 (1.2, 17.0)
Cumulative applications before 1980
1–24	17	27	1.6 (0.8, 3.2)	14	23	1.6 (0.8, 3.4)
25–99	9	18	1.7 (0.7, 4.2)	9	14	1.3 (0.5, 3.4)
≥100	10	16	1.4 (0.6, 3.4)	8	13	1.5 (0.6, 3.8)
NAT2 phenotype
Slow
No hair coloring product use	38	37	1.0 (referent)	38	37	1.0 (referent)
First use before 1980	14	23	1.5 (0.6, 3.6)	11	22	1.8 (0.7, 4.5)
Rapid/intermediate
No hair coloring product use	41	32	1.0 (referent)	41	32	1.0 (referent)
First use before 1980	11	25	3.3 (1.3, 8.6)	11	20	2.6 (1.0, 7.0)
NAT1 genotype
NAT1 any
No hair coloring product use	42	45	1.0 (referent)	42	45	1.0 (referent)
First use before 1980	16	26	1.5 (0.7, 3.3)	15	22	1.3 (0.6, 3.1)
NAT1*10/any + NAT1*10/*10
No hair coloring product use	30	23	1.0 (referent)	30	23	1.0 (referent)
First use before 1980	10	22	2.5 (0.9, 7.6)	8	16	2.5 (0.8, 8.1)
NHL subtype
Follicular lymphoma
No hair coloring product use	98	24	1.0 (referent)	98	24	1.0 (referent)
First use before 1980	36	17	2.0 (0.9, 4.4)	31	15	2.1 (0.9, 4.7)
DLBCL
No hair coloring product use	98	34	1.0 (referent)	98	34	1.0 (referent)
First use before 1980	36	16	1.1 (0.5, 2.4)	31	11	0.8 (0.4, 1.9)

OR, odds ratio; CI, confidence interval; NAT, N-acetyltransferase; NHL, non-Hodgkin lymphoma; DLBCL, diffuse large B-cell lymphoma.

^aAs originally published.

^bOR (95% CI), adjusted for age, race and Surveillance, Epidemiology, and End Results area.

Drs Bailey and Skare asked for clarification of the colors included in ‘dark colors’ and ‘intense tones’. As shown in Table II of our original paper, dark colors included black, brown and red dyes (regardless of depth) and intense tones included black, dark brown and dark blonde (1). Our questionnaire did not ascertain information on depth for red hair dyes, so they could not be considered in the ‘intense tone’ analysis.

It is interesting to note that Turesky et al. (2) observed 4-aminobiphenyl in black, red and blonde hair dyes but not in brown hair dyes (2). 4-Aminobiphenyl has not been linked to lymphoma, as stated in our paper, but the report illustrates that hair dye products are complex mixtures with sometimes unexpected compounds and that blonde dyes should not automatically be assumed to have lower concentrations of dye-related compounds than the dark colors.

Proctor and Gamble is to be commended for funding work to develop and test questionnaires to collect data on hair dye use. Because hair dye use has many characteristics, changes over time and the details may be difficult to recall, exposure misclassification is a challenge. The methods work funded by Proctor and Gamble has generated valuable data and tools that should improve exposure assessment. We also appreciate the color groupings suggested in Dr Bailey's and Dr Skare's letter, although we cannot apply them because we lack depth information for red dyes. It is a shame, however, that Drs Bailey and Skare did not include their quantitative data on dye concentrations by color shade and depth, which presumably provide the foundation for the color groupings they suggest. If such data were made publicly available, researchers and reviewers could better judge exposure metrics used in studies of hair dye use and cancer, and great advances could be made in improving the analysis and interpretation of results from these studies.