Personal Use of Hair Dye and the Risk of Certain Subtypes of Non-Hodgkin Lymphoma

Yawei Zhang; Silvia De Sanjose; Paige M. Bracci; Lindsay M. Morton; Rong Wang; Paul Brennan; Patricia Hartge; Paolo Boffetta; Nikolaus Becker; Marc Maynadie; Lenka Foretova; Pierluigi Cocco; Anthony Staines; Theodore Holford; Elizabeth A. Holly; Alexandra Nieters; Yolanda Benavente; Leslie Bernstein; Shelia Hoar Zahm; Tongzhang Zheng


Am J Epidemiol. 2008;167(11):1321-1331. 

In This Article

Materials and Methods

A total of 4,461 patients with incident NHL (International Classification of Diseases for Oncology codes M-9590-9591, M-9595, M-9670-9673, M-9675-9676, M-9680-9688, M-9690-9691, M-9695-9698, M-9700, M-9702-9703, M-9705-9711, M-9713-9715, M-9823, and M-9827) and 5,799 controls from four case-control studies[14,17,19,20] that collected detailed information on personal hair-dye use were included in this pooled analysis. The characteristics of each study are presented in table 1 . All study protocols were approved by local institutional review boards, and written informed consent was obtained from all participants.

Within each study, NHL diagnoses and subtypes were confirmed by pathologists, and NHL cases were categorized by histologic type. For this pooled analysis, NHL cases were classified according to the World Health Organization classification system and as proposed by the InterLymph Pathology Working Group.[21] Five major subtypes of NHL were included for subtype analyses: diffuse large B-cell lymphoma, follicular lymphoma, chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), marginal-zone lymphoma, and T-cell lymphoma.

Information on hair-dye use and other potential risk factors was collected by trained interviewers using standardized, structured questionnaires. The Connecticut Women's NHL Study (from Yale University), the NCI/SEER (National Cancer Institute/Surveillance, Epidemiology, and End Results) Multi-Center Case-Control Study, and the International Case-Control Study of Lymphomas from Europe (EpiLymph) used very similar questionnaires pertaining to history of hair-dye use. For example, subjects in these studies were first asked whether they had ever used any hair-coloring products. If they had, these subjects were asked: 1) at what age they used each product for the first time; 2) how often they used each product (weekly, monthly, yearly, or other); 3) at what age they used each product for the last time; and 4) what type (permanent or nonpermanent, including semipermanent and temporary) and color (dark color, including black, brown, and red, or light color, including blond) of product they used. In the Epidemiology of NHL Study (from the University of California, San Francisco), participants were first asked whether they had ever used hair-coloring products on their own hair more than five times up to 1 year before the diagnosis or interview. If they had, they were asked about their ages at first and last use of each product and the frequency, duration (in years), type, and color of each product used. For analyses that included time period of use, the participants were categorized on the basis of whether they had started using hair dyes before 1980 or during 1980 or later (i.e., no use of any type prior to 1980).

From each study, we obtained original data for this pooled analysis in order to investigate the relation between personal hair-dye use and risk of NHL. The likelihood ratio test was used to test heterogeneity across studies by comparing the logistic regression models with and without the cross-product terms of hair-dye use (i.e., ever use of dark-colored dyes or ever use of permanent dyes) and study. A random-effects model was used to compute the pooled risk estimates and 95 percent confidence intervals, weighted by the inverse marginal variance (the sum of the study-specific variance and the variance of the exposure effect across studies or random study effects).[22] Risk estimates from the random-effects models were consistent with the results from dichotomous and polytomous unconditional logistic regression models. Therefore, we present odds ratios and 95 percent confidence intervals derived from dichotomous and polytomous unconditional logistic regression models. Continuous variables, including duration of hair-dye use, frequency of hair-dye use, and total number of applications of hair dye, were categorized into tertiles based on the distribution of any hair-dye use among controls. The final model was adjusted for age (continuous), sex (male/female), race (White, Black, or other), and study center. Adjustments for other variables, such as family history of hematopoietic cancer in first-degree relatives, tobacco smoking, alcohol consumption, and highest level of education, did not produce material changes in the risk estimates (<10 percent change in risk estimates) and thus were not included in the final model.

Formal testing between NHL and the linear trends for duration, frequency, and total number of applications of hair-dye was performed by assigning values of 0, 1, 2, or 3 for exposure groups. More specifically, restricted cubic splines were compared with the linear models and evaluated by likelihood ratio test and by visual inspection of the restricted cubic spline graphs.[23] Evidence from the cubic spline analyses did not support a departure from linear trend of a dose-response relation. Sensitivity analyses were performed by comparing the risk estimates obtained from models that systematically excluded each study population to determine whether any one population had a disproportionate influence on the summary estimated risk. All analyses were performed using SAS 9.1 (SAS Institute, Inc., Cary, North Carolina) and Stata 8.0 (Stata Corporation, College Station, Texas). All statistical tests were two-sided with a significance level of 0.05.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: