Response to Grant, WB: “Ultraviolet exposure and non-Hodgkin’s lymphoma: beneficial and adverse effects?”

Abstract

As we discussed in our recent manuscript, our findings of modest but significant positive associations between ambient ultraviolet (UV) radiation and non-Hodgkin lymphoma (NHL) risk in the Nurses’ Health Study (NHS) were unexpected and inconsistent with most of the prior literature on this topic [1]. Dr. Grant offers alternative explanations for our results. In response to his comments, we have conducted additional data analyses. First, Dr. Grant questions whether a majority of NHS participants reside in regions of relatively low total and UV-B radiation. We agree that results from the NHS may not be generalizable to other populations, particularly those in the areas of higher UV radiation. Because nurses originally were recruited from 11 US states (i.e., California, Connecticut, Florida, Maryland, Massachusetts, Michigan, New Jersey, New York, Ohio, Pennsylvania, and Texas) [2], a majority of participants (59%) resided in the Northeast at baseline in 1976. A similar proportion of women reported Northeast residence at birth, age 15, and age 30. Over the course of 30 years of follow-up, 55% of the person-time included residence in the northeast region of the USA. We used the distribution of average annual UV-B flux values in our study population, which ranged from 93 to 196 (in R-B counts 910) as nurses changed residence over the 30 years of follow-up, to guide our choice of cutpoints for analysis. We reported that NHL risk was increased among women who resided in regions with UV-B flux[113 compared to those with lower exposure (\113) [1]. In response to Dr. Grant’s comment, we dichotomized the top category of UV-flux at its median value (164) and re-analyzed the data using this 4-category variable. Our findings were similar to the results presented in the manuscript, with increased risks most apparent in the highest category of ambient UV-B exposure. For example, the multivariable-adjusted rate ratio (RR) for NHL associated with UV-B flux [164 at age 15, which included residence in Texas, Florida, Arizona, New Mexico, and Louisiana, compared to UV-B flux\113 was 1.72 (95% confidence interval (CI): 1.30, 2.29). Further, it is important to note that we observed statistically significant linear trends for the association of average annual UV-B flux, assessed at various time points and considered as a continuous variable, and risk of NHL (p-trend\0.05). Moreover, we found no evidence of departure from linearity when restricted cubic splines were examined. Second, Dr. Grant suggests that the association between UV-B radiation and NHL risk may differ by latitude because of differences in the relative amount of UV-B and UV-A radiation. We speculated in our paper that increased risks of NHL could plausibly be mediated through immunosuppressive effects of UV exposure. As previously discussed, modest increased risks were most apparent for individuals living in areas of higher UV-B radiation (i.e., outside of the northeast). Unfortunately, data on ambient UV-A levels were not available for analysis, and therefore, we could not evaluate the relative effects of UV-A vs. UV-B exposure nor could we directly evaluate the immunosuppression K. A. Bertrand (&) F. Laden Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA e-mail: kbertran@hsph.harvard.edu