Plasma Organochlorine Levels and Risk of Non–Hodgkin Lymphoma in the Nurses' Health Study

Polychlorinated biphenyls (PCB) and organochlorine pesticides such as dichlorodiphenyl trichloroethane have been the focus of several recent investigations into the etiology of non–Hodgkin lymphoma (NHL; refs. 1-8). It has been hypothesized that the organochlorine-NHL association may be mediated through immunotoxic mechanisms (9). Although manufacturing and new uses of PCBs and dichlorodiphenyl trichloroethane were banned in the United States in the 1970s, these compounds persist in the environment and store in adipose tissue and the lipid components of blood and breast milk. Because they are resistant to metabolism and have long half-lives, measurements of these compounds in biological media represent cumulative exposures over time (10). We examined the association of blood levels of PCBs and p,p′-dichlorodiphenyl dichloroethane (p,p′-DDE), the primary metabolite of dichlorodiphenyl trichloroethane, with risk of NHL among women in a case-control study nested in the Nurses' Health Study.

NHL cases (n = 145) and two controls per case, matched on age, race, month of blood draw, and fasting status, (n = 290), were identified from participants in the Nurses' Health Study blood cohort (11). NHL diagnoses, including chronic lymphocytic leukemia/small lymphocytic lymphoma, were identified by annual follow-up questionnaires and all cases were confirmed by a review of medical records and pathology reports. Women with a diagnosis of NHL before or within 6 months of blood collection and those with a prior diagnosis of cancer (other than nonmelanoma skin cancer) were excluded. Histologic subtype was determined according to the WHO classification of lymphomas (12, 13).

Organochlorine analyses for 51 individual PCB congeners and p,p′-DDE were done at the Harvard School of Public Health Organic Chemistry Analytical Laboratory. The laboratory methods have been described in detail elsewhere (13). Plasma PCB and p-p′-DDE concentrations were adjusted for total serum lipids calculated using the formula by Phillips et al. (14) and are reported in units of nanograms of organochlorine per gram of lipid (ng/g).

Our primary interest was in p-p′-DDE and a priori groupings of PCB congeners based on suspected immunotoxicity (that is, International Union of Pure and Applied Chemistry 66, 74, 105, 118, 156, and 167; ref. 15). We also evaluated other groupings including the sum of PCBs (ΣPCB) and the sum of the four most prevalent congeners (that is, 118, 138, 153, and 180) as well as these individual PCB congeners. The difference between the means of the cases and those of their matched controls was computed using generalized estimating equations, adjusting for the matching factors. We categorized organochlorine concentrations into quartiles based on the distribution among controls. In separate models for each organochlorine or group of organochlorines, we used conditional logistic regression, stratifying on the matched case-control triplets to estimate odds ratios (OR) and 95% confidence intervals (CI) for risk of NHL associated with each quartile of exposure relative to the lowest quartile. Tests for trend were done using the natural log-transformed lipid-adjusted organochlorine concentrations as continuous variables.

Multivariable conditional logistic regression models including height as a continuous variable and indicator variables for region of residence (Northeast, Midwest, West, and South), smoking history (never, past, and current), body mass index (BMI; tertiles based on the control distribution), and alcohol intake (≥1 time/d, 1-6 times/wk, 1-3 times/mo, rarely/never) were used to adjust simultaneously for potential confounding by these factors. Additional analyses of individual PCB congeners and congener groups were done including p-p′-DDE in the model. We also examined whether the associations between organochlorines and NHL were modified by known or suspected risk factors for NHL as well as region of residence, in unconditional logistic regression models. Tertiles of organochlorines rather than quartiles were used in the stratified analyses because of the decreased sample size in each stratum. Additionally, we performed polytomous logistic regression to test for heterogeneity in effect estimates for the most common NHL subtypes (that is, diffuse large B-cell lymphoma, follicular lymphoma, and chronic lymphocytic leukemia/small lymphocytic lymphoma).

Cases and controls, ranged in age from 44 to 69 years at blood draw, were 96% white and were similar with respect to height, and BMI. Cases were more likely to reside in the South and less likely to reside in the Northeast, Midwest, or West compared with controls. Additionally, cases were slightly more likely to be current smokers than controls. The median time to diagnosis among cases was 5.8 years. The distributions (in ng/g lipid) of PCBs and DDE were not statistically significantly different between cases and controls (ΣPCB cases: median, 621.0; interquartile range (IQR), 252.0; max, 1,957.5; min, 279.5; ΣPCB controls: median, 625.0; IQR, 322.2; max, 3,012.2; min, 221.6; P = 0.35; DDE cases: median, 996.2; IQR, 1,293.7; max, 6,079.4; min, 7.5; DDE controls: median, 1,002.3; IQR, 1,152.4; max, 7,042.1; min, 54.6; P = 0.26). Results from conditional logistic regression analyses of total NHL and the different organochlorine metrics are presented in Table 1. Similar analyses of NHL subtypes are presented in Table 2. There was no evidence of confounding by DDE or effect modification by lactation, current smoking status, region, or follow-up period. We observed a suggestive positive linear association of ΣPCB with NHL in obese women (P_trend = 0.09). However, this is based on only 20 cases.

Previous reports (2, 4, 6, 7, 16), including pilot analyses in this cohort using controls selected for a study of breast cancer (3), have found significant evidence of an association between plasma concentrations of PCBs and risk of NHL. In contrast, we observed no association of NHL or NHL subtypes with PCBs or DDE. The levels of organochlorines measured in this general population sample were low; however, they are consistent with or even higher than (7) other studies that have observed positive associations. Different laboratories and laboratory methods were used to measure PCBs in the pilot and this study; however, we observed a significant positive association between PCBs and NHL in men using the same laboratory as this study (13). In the pilot analyses, the median time to diagnosis for cases was only 1 year versus 5.8 years here. It is possible that a biased case or control sample was selected by chance in one or both studies. In conclusion, there was no consistent evidence of an association of NHL with prospectively measured blood levels of PCBs or DDE in this population based study of U.S. women.

No potential conflicts of interest were disclosed.

We thank David Hunter and Jaime Hart for their scientific input.

Grant Support: Grant CA098122 from the National Cancer Institute, grant support in part by T32 ES007155 (K.A. Bertrand) and by T32 MH073122 (S.K. Sagiv).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.