Organochlorines and risk of non-Hodgkin lymphoma
Organochlorine chemicals and polychlorinated biphenyls (PCBs) have been suspected as possible risk factors for non-Hodgkin lymphoma (NHL). We investigated PCBs and organochlorine pesticides and risk of NHL in a population-based case–control study in British Columbia, Canada. Congeners of PCBs (including dioxinlike congeners) and pesticides or pesticide metabolites were measured in plasma of 422 pretreatment cases and 460 control subjects. This is so far the largest study to examine organochlorines in plasma to date. Several dioxin-like PCB congeners were associated with increased risk of NHL, including dioxin-like PCB nos. 118 and 156 with odds ratios (OR) for the highest versus lowest quartile between 1.6 and 1.8. Several non-dioxin-like congeners also showed significant associations. The PCB congener with the strongest association was no. 180 with an OR for the highest versus the lowest quartile of 1.83 (95% confidence interval = 1.18–2.84). Six pesticide analytes also showed a significant association with NHL; β-hexachlorocyclohexane, p,p′-DDE, hexachlorobenzene, mirex, oxychlordane and trans-nonachlor. The strongest association was found for oxychlordane, a metabolite of the pesticide chlordane (highest vs. lowest quartile OR = 2.68, 95% confidence interval = 1.69–4.24). Our results provide further evidence that organochlorines contribute to NHL risk. © 2007 Wiley-Liss, Inc.
The incidence of non-Hodgkin lymphoma (NHL) has been rising steadily for the past 30 years worldwide, but has leveled off in the recent years.1, 2 It has been suggested that the increase and recent leveling off may be at least partially due to changing pattern of exposure to polychlorinated biphenyls (PCB) and organochlorine pesticides, compounds which have long half-lives and accumulate in human adipose tissue.
The epidemiologic evidence for associations between organochlorine insecticides and PCBs with NHL has been recently reviewed.3, 4 Some studies have suggested an increased risk, but many others have failed to demonstrate an association. Reasons for the inconsistent results include lack of power and imprecise measurement of exposure.
Recent studies in the United States and Sweden have evaluated biological organochlorine levels and risk of NHL.5, 6, 7, 8, 9, 10, 11 These studies found associations between NHL and high levels of specific PCB congeners and organochlorine pesticides, although the specific congeners and pesticides identified were not consistent between studies. To further investigate the association between organochlorine compounds and NHL, we conducted a population-based case–control study in British Columbia, Canada.
Material and methods
The study population consisted of a subset of participants from a case–control study of NHL conducted in British Columbia, Canada, hereafter referred to as the “parent study.” Between March 2000 and February 2004, subjects from the Greater Vancouver Regional District (GVRD) and the Capital Regional District (CRD), which includes the city of Victoria, were enrolled from the British Columbia Cancer Registry. Together, these 2 large districts represent about 59% of the population of the province. Cases included subjects with newly diagnosed NHL aged 20–79 without evidence of HIV infection. Study material was made available in the 4 most commonly spoken languages in the catchment area, English, Chinese, Punjabi and Tagalog. Subjects were asked to complete a questionnaire and provide a blood or saliva sample. Questionnaires were completed by self-report and a computer-assisted telephone interview (CATI). Questionnaires contained information on demographic characteristics, sunlight exposure, medical history and other factors.
Written informed consent was obtained from each participant. The study was approved by the BC Cancer Agency-University of British Columbia Research Ethics Board. All cases were reviewed by one of 2 pathologists (RDG, BRB) and classified using the World Health Organization classification.12
The total number of eligible cases available in the parent study was 1,263. Of these eligible cases, 133 (10.5%) died before contacted with a median time between diagnosis and death of 32 days (interquartile range 6–78 days), 62 (4.9%) were unable to be located and 1,068 (84.6%) were contacted. Of those contacted, 73 (6.8%) could not participate due to poor health, 8 (0.7%) could not participate due to language, 147 (13.8%) refused and 840 (78.7%) consented. After pathological review, 12 cases with insufficient material to classify or with prior transplantation were eliminated leaving 828 cases for analysis. All 828 cases completed at least part of the questionnaire, 769 (92.9%) provided a blood sample, 28 (3.4%) provided only a saliva sample and 31 (3.7%) did not provide a blood or saliva sample.
This study of risk associated with organochlorines included cases that provided blood prior to the onset of chemotherapy because chemotherapy treatment has been found to significantly change plasma levels of organochlorines.13 This resulted in a total of 455 cases with organochlorine levels measured. It has been shown that rapid weight loss can increase plasma organochlorine levels.14 Weight loss prior to diagnosis is a prognostic indicator and is routinely recorded in the patient charts. The occurrence of weight loss prior to diagnosis in cases was checked through chart reviews and physician interviews. Information on weight loss was obtained for 447 (98%) of the cases. Twenty-four cases were found to have over 10% weight loss prior to diagnosis and were dropped from all analyses. The remaining 8 cases without information on weight loss were also excluded. One additional case was excluded from all analyses because no lipid-adjusted organochlorine measurements could be obtained, as the case did not have a sufficient quantity of blood in the sample to measure lipid values. Thus, a total of 422 cases were available for organochlorine analysis.
Population controls were frequency matched to cases by sex, age (within 5-year age group) and residential location (GVRD or CRD) in an ∼1:1 ratio. Controls were randomly selected from the Client Registry of the BC Ministry of Health. There were 2,373 controls selected in the parent NHL study. Of these selected, 13 (0.5%) were deceased, 504 (21.2%) could not be contacted and 1,856 (78.2%) were contacted. Of these contacted, 103 (5.5%) could not participate due to poor health, 49 (2.6%) could not participate due to language, 856 (46.1%) refused and 848 (45.7%) consented. All consenting controls completed at least part of the questionnaire, 679 (80.1%) gave a blood sample, 113 (13.3%) gave only a saliva sample and 56 (6.6%) did not provide either sample.
A total of 463 frequency matched controls who gave a blood sample were randomly chosen to have organochlorine levels measured. Three controls did not have enough blood to perform lipid measurements and were therefore excluded from all analyses. A total of 460 controls were used in the organochlorine analysis.
The blood sample of 40 ml was drawn into 4 ACD and 2 EDTA tubes. All blood was processed at the BC Cancer Agency in the same laboratory. For organochlorine testing, 2 ml of plasma was separated from the blood by centrifugation in the EDTA tubes, transferred immediately to vials with a Teflon stopper, which prevents leaching of organic compounds, and frozen at −80°C. The samples were shipped in batches consisting of mixed cases and controls to the Toxicology Centre at the Québec National Public Health Institute for measurement. Samples were identified only by a unique sample code and thus were blinded for case–control status. The measured organochlorines included 14 PCB congeners (IUPAC nos. 28, 52, 99, 101, 105, 118, 128, 138, 153, 156, 170, 180, 183 and 187) and 11 pesticides or pesticide metabolites. The pesticides examined were aldrin, p,p′-DDT, hexachlorobenzene (HCB), β-hexachlorocyclohexane (β-HCCH), mirex and 4 chlordane compounds (α-chlordane, γ-chlordane, cis-nonachlor and trans-nonachlor). Two other analytes, p,p′-DDE, a metabolite of DDT, and oxychlordane, a metabolite of chlordane, were also examined.
Extraction of organochlorine residues from plasma was performed with hexane. The extract was then cleaned by passing through 2 Florisil columns and analyzed by gas chromatography using the HP-5890 series II gas chromatograph with dual column split/splitless injector and 2 electron capture detectors. A computer program developed at the Québec Toxicology Centre was used to identify the peaks corresponding to each compound. The detection limits for the various organochlorines using this protocol are 0.025 μg/DL for PCB no. 52, β-HCCH, DDT and DDE, and 0.015 μg/l for all other PCB congeners and pesticide analytes.
Free cholesterol (FC), total cholesterol (TC), triglycerides (TG) and phospholipids (PL) were also measured in each plasma sample using enzymatic methods. Total lipid concentration was calculated for each sample, using the Patterson formula.15 Lipid-adjusted organochlorine concentrations for each sample were calculated by dividing the whole-weight measurements by the total lipid concentration. Median detection limits of lipid-adjusted values were calculated by dividing the detection limit for each compound by the median total lipid concentration in all samples. Only lipid-adjusted measurements are reported.
Measurement reliability was assessed through the measurement of 69 randomly chosen quality control (QC) pairs that were replicates of single samples. The intraclass correlation coefficient (ICC) was calculated for each QC pair.
Variable coding and statistical analyses
Relationships between the different lipid-adjusted PCB congeners and pesticide analytes were assessed by the Spearman rank correlation; for this analysis, values below the limit of detection were assigned a value of the detection limit divided by √2.16 Only measurements with more than 80% values above the detection limit were included in this analysis.
Lipid-adjusted concentrations for organochlorine analytes were categorized according to their distribution in control samples. Values below the detection limit were always included in the lowest category. For analytes for which there were fewer than 25% of samples below the detection limit, the concentrations were categorized into quartile variables. PCB no. 183 had 38.4% of samples nondetectable, and so the concentrations were categorized into 3 groups with the lowest group being nondetectable. For PCB no. 28, PCB no. 105, cis-nonachlor, p,p′-DDT and mirex, more than half of the samples were below the detection limit; so, concentrations for these analytes were categorized as either nondetectable or detectable. The lowest quartile or lowest measurement group was used as the reference category.
The total sum of all PCB congeners measured, dioxin-like PCB congeners (PCB nos. 105, 118, 156) and non-dioxin-like PCB congeners17 were computed by summing the individual whole-weight measurements of the individual PCB congeners. Values below the detection limit for individual PCB congeners were assigned a value of the detection limit divided by √2 prior to computing the summary variables. After lipid adjustment, the summary measures were categorized into quartiles based on the distribution in controls.
Odds ratios (OR) and 95 percent confidence intervals (95% CI) for the risk of NHL were estimated using unconditional logistic regression for each of the organochlorine metrics.
Tests for trend across quartiles were conducted by creating a continuous variable with assigned values equal to the median level among controls within each category. To determine the median levels, values below the detection limit for individual analytes were assigned a value of the detection limit divided by √2. The change-in-estimate criterion was used to select confounders with more than a 5% change in the estimate considered important. The possible confounders included age, sex, region, ethnicity, education level, family history of NHL, body mass index (BMI) and farming.
Two age-group variables were tested; 12 categories of 5-year ranges and 4 categories (20–49, 50–59, 60–69 and 70+), which are approximate quartiles of age. Both age variables gave similar regression estimates when entered in the logistic regression model. The age variable with fewer categories was used for further regression modeling to improve efficiency.
Interactions between the PCB congeners and organochlorine pesticide analytes with sex, age (<60 vs. 60+), region, ethnicity (Caucasian vs. others), education (<high school, high school graduate, postsecondary graduate) and BMI (<25.3 vs. 25.3+) were examined by entering the interaction terms into the logistic regression model. Significance was based on the likelihood ratio p value for the interaction terms.
The forward stepwise selection was used to determine if combinations of organochlorines in the logistic regression model give better risk estimates than individual organochlorines. Trend variables of organochlorine levels were entered into the models for stepwise selection. Age, sex, region, education, BMI, ethnicity, farming and family history of NHL were included as confounders in the stepwise models.
Subgroup analyses of the associations between organochlorines, and NHL were also performed for histological subtypes. Follicular lymphoma and diffuse large cell lymphoma are the largest subtypes of NHL and are analyzed separately. All T-cell and other B-cell tumors (non-diffuse large cell, non-follicular) were also examined. Tests for heterogeneity were performed by analysis of the case group only using polychotomous logistic regression.18, 19
Aldrin, α-chlordane, γ-chlordane and PCB congeners nos. 52, 101 and 128 had less than 5% of samples with detectable levels and were excluded from further analyses, leaving 19 pesticide analytes and PCB congeners for analysis. Information on the number of samples measured above the detection limit for each analyte is shown in Table I. The median plasma total lipids concentration was 6.76 g/L; therefore, the median detection limits for the lipid-adjusted values were 3.70 ng/g-lipid for PCB no. 52, β-HCCH, DDT and DDE and 2.22 ng/g-lipid for all other PCB congeners and pesticide analytes. All analytes showed high reliability, with a median intraclass correlation of 0.998 and all ICCs greater than 0.93.
Table I. Results of Laboratory Measurements for PCB Congeners and Pesticide Analytes
| 28||158 (17.9)||17||0.968|
| 52||11 (1.2)||1|| |
| 99||721 (81.7)||53||0.967|
| 101||30 (3.4)||3||0.949|
| 105||275 (31.2)||22||0.933|
| 118||825 (93.5)||64||0.996|
| 128||2 (0.2)||0|| |
| 138||874 (99.1)||69||0.998|
| 153||880 (99.8)||69||0.999|
| 156||787 (89.2)||64||0.981|
| 170||846 (95.9)||68||1.000|
| 180||878 (99.5)||68||1.000|
| 183||543 (61.6)||41||1.000|
| 187||841 (95.4)||67||1.000|
| Aldrin||0||0|| |
| β-HCCH||779 (88.3)||58||0.998|
| α-Chlordane||0||0|| |
| γ-Chlordane||1 (0.1)||0|| |
| cis-Nonachlor||250 (28.3)||16||0.992|
| p,p′-DDE||882 (100.0)||69||1.000|
| p,p′-DDT||288 (32.7)||22||1.000|
| HCB||881 (99.9)||69||0.998|
| Mirex||328 (37.2)||19||0.974|
| Oxychlordane||848 (96.1)||68||0.990|
| trans-Nonachlor||863 (97.8)||68||0.997|
The study population characteristics for the organochlorine study, and the parent study is shown in Table II. Organochlorine study cases, who donated a blood sample prior to treatment or who were untreated, differed from the parent NHL study cases in that they were less likely to be of diffuse large cell histology and correspondingly more likely to be of follicular histology. All other characteristics were representative of the parent study. Organochlorine study cases were similar to their controls with regard to age, sex, region, education, ethnicity and family history of NHL. Organochlorine study cases had significantly lower measured plasma total lipids than controls (6.66 vs. 7.14.0 g/L, p < 0.001).
Table II. Characteristics of Cases and Controls in Parent Non-Hodgkin Lymphoma Study and Study of Plasma Organochlorines [Frequency (Percentage)]
| Follicular||141 (33.4)|| ||222 (26.8)|| |
| Diffuse large cell||67 (15.9)|| ||220 (26.6)|| |
| T cell||47 (11.1)|| ||80 (9.7)|| |
| Other B cell||167 (39.6)|| ||306 (37.0)|| |
| 20–49||83 (19.7)||90 (19.6)||162 (19.6)||241 (28.4)|
| 50–59||107 (25.4)||113 (24.6)||197 (23.7)||174 (20.5)|
| 60–69||99 (23.5)||109 (23.7)||219 (26.6)||217 (25.6)|
| 70+||133 (31.5)||148 (32.2)||250 (30.2)||216 (25.5)|
| Male||232 (55.0)||258 (56.1)||482 (58.2)||451 (53.2)|
| Female||190 (45.0)||202 (43.9)||346 (41.8)||397 (46.8)|
| GVRD (Greater Vancouver)||361 (85.5)||398 (86.5)||688 (83.1)||659 (77.7)|
| CRD (Greater Victoria)||61 (14.5)||62 (13.5)||140 (16.9)||189 (22.3)|
| Less than high school||77 (18.2)||76 (16.5)||158 (19.1)||120 (14.2)|
| High school graduate||225 (53.3)||249 (54.1)||425 (51.3)||469 (55.3)|
| Postsecondary graduate||111 (26.3)||130 (28.3)||226 (27.3)||249 (29.4)|
| Unknown||9 (2.1)||5 (1.1)||19 (2.3)||10 (1.2)|
| Caucasian||339 (80.3)||369 (80.2)||649 (78.4)||651 (76.8)|
| Asian||38 (9.0)||41 (8.9)||83 (10.0)||97 (11.4)|
| South Asian||13 (3.1)||21 (4.6)||29 (3.5)||43 (5.1)|
| Other/mixed||17 (4.0)||16 (3.5)||38 (4.6)||37 (4.4)|
| Unknown||15 (3.5)||13 (2.8)||29 (3.5)||20 (2.3)|
| <25||176 (41.7)||199 (43.3)||337 (46.8)||397 (46.8)|
| 25–27.5||114 (27.0)||122 (26.5)||225 (24.1)||204 (24.1)|
| 27.5–30||45 (10.7)||63 (13.7)||98 (11.8)||100 (11.8)|
| 30+||73 (17.3)||65 (14.1)||145 (14.5)||123 (17.5)|
| Unknown||14 (3.3)||11 (2.4)||23 (2.8)||24 (2.8)|
| Yes||17 (4.0)||11 (2.4)||27 (3.3)||16 (1.9)|
| No||386 (91.5)||447 (97.2)||772 (93.2)||828 (97.6)|
| Unknown||19 (4.5)||2 (0.4)||29 (3.5)||4 (0.5)|
|Family history of NHL|
| Yes||21 (5.0)||14 (3.0)||34 (4.1)||21 (2.5)|
| No||382 (90.5)||445 (96.7)||763 (92.1)||814 (96.0)|
| Unknown||19 (4.5)||1 (0.2)||31 (3.7)||13 (1.5)|
Thirty-two cases were excluded from analyses due to weight loss or unavailable weight loss information. Organochlorine levels were generally higher in these cases compared to cases without weight loss. For example, the median total PCB was 246.9 ng/g in the excluded cases versus 172.3 ng/g in the cases without weight loss. A sensitivity analysis showed that OR increased by roughly 10% when these cases were included.
Very few subjects in the study were ever involved in farming (n = 28). Living or working on a farm was only slightly more frequent among cases than controls (4.0% vs. 2.4%, p = 0.17).
Strong correlations were observed between plasma levels of PCB congeners and pesticide analytes (Table III). For PCB congeners nos. 170, 180 and 187, very strong correlations (r > 0.90) were observed for each pair. PCB congeners nos. 138, 153 and 156 were slightly less correlated (r > 0.80) with each other and with the former group. Strong correlations were also observed between PCB congeners nos. 99 and 138 (r = 0.85) and between oxychlordane and trans-nonachlor (r = 0.87).
Table III. Spearman Rank Correlations Between Individual PCB Congeners and Organochlorine Pesticide Analytes (with <20% of Samples Below the Detection Limit)
|PCB 118|| ||–||0.762||0.704||0.604||0.509||0.483||0.578||0.463||0.518||0.659||0.636||0.656|
|PCB 138|| || ||–||0.924||0.805||0.738||0.692||0.785||0.410||0.573||0.639||0.651||0.708|
|PCB 153|| || || ||–||0.867||0.879||0.857||0.891||0.376||0.523||0.620||0.654||0.708|
|PCB 156|| || || || ||–||0.888||0.848||0.815||0.293||0.287||0.541||0.638||0.662|
|PCB 170|| || || || || ||–||0.982||0.928||0.237||0.276||0.495||0.577||0.602|
|PCB 180|| || || || || || ||–||0.929||0.194||0.263||0.462||0.558||0.580|
|PCB 187|| || || || || || || ||–||0.3||0.413||0.540||0.600||0.669|
|β-HCCH|| || || || || || || || ||–||0.598||0.555||0.392||0.430|
|p,p′-DDE|| || || || || || || || || ||–||0.482||0.440||0.453|
|HCB|| || || || || || || || || || ||–||0.664||0.584|
|Oxychlordane|| || || || || || || || || || || ||–||0.870|
Overall, there was a significant association between total summed PCB level and NHL, with those in the highest quartile having an OR of 2.14 (95% CI = 1.38–3.30) compared to those in the lowest quartile (Table IV). Table IV also shows the associations with dioxin-like PCB level. A significant association was observed for the summed total dioxin-like PCB level (highest vs. lowest quartile OR = 2.40, 95% CI = 1.53–3.77). Looking at individual PCB congeners, 2 of the 3 dioxin-like PCB congeners analyzed showed significant associations with NHL with ORs for the highest versus lowest quartile of 1.77 (95% CI = 1.15–2.72) for no. 118 and 1.77 (95% CI = 1.14–2.74) for no. 156. PCB congener no. 105, with over 60% of the values below the detection limit, was not associated with NHL.
Table IV. Total PCB and Summed and Individual Dioxin-Like PCB Congeners and Association with NHL
|Total PCB summed1||≤100.9||81||115||1.00|| || ||0.001|
|Dioxin-like summed2||≤10.12||82||115||1.00|| || ||<0.001|
|PCB no. 1053||Not detected||281||316||1.00|| || ||0.675|
|PCB no. 1184||≤4.57||82||109||1.00|| || ||0.004|
|PCB no. 1562||≤3.65||85||114||1.00|| || ||0.004|
Table V shows the associations with non-dioxin-like PCB level. A significant association was observed for the summed total non-dioxin-like PCB level (highest vs. lowest quartile OR = 2.18, 95% CI = 1.41–3.38). Five of the 8 non-dioxin-like PCB congeners also showed significant associations with NHL, no. 138, 153, 170, 180 and 187, with ORs for the highest versus lowest quartiles ranging from 1.46 (95% CI = 0.98–2.18) for no. 138 to 1.91 (95% CI = 1.19–3.07) for no. 180.
Table V. Non-Dioxin-Like PCB Congeners and Association with NHL
|Non-dioxin-like summed1||≤88.57||85||115||1.00|| || ||<0.001|
|PCB no. 282||Not detected||348||376||1.00|| || ||0.779|
|PCB no. 993||≤3.06||106||113||1.00|| || ||0.045|
|PCB no. 1384||≤11.61||100||115||1.00|| || ||0.02|
|PCB no. 1534||≤25.29||90||115||1.00|| || ||0.002|
|PCB no. 1704||≤7.16||88||115||1.00|| || ||0.005|
|PCB no. 1805||≤21.93||85||111||1.00|| || ||0.005|
|PCB no. 1834||Not detected||162||177||1.00|| || ||0.113|
|PCB no. 1876||≤5.93||88||114||1.00|| || ||0.003|
Six of the 8 pesticide analytes were significantly associated with NHL (Table VI). These were β-HCCH, p,p′-DDE, HCB, mirex, oxychlordane and trans-nonachlor, with ORs for the highest versus the lowest quartile ranging from 1.42 (95% CI = 0.92–2.19) for p,p′-DDE to 2.68 (95% CI = 1.69–4.24) for oxychlordane. Subjects with mirex above the DL showed a significantly increased risk compared to those with level below the DL of 1.44 (95% CI = 1.08–1.92).
Table VI. Pesticide Analytes and Association with Non-Hodgkin's Lymphoma
|β-HCCH1||≤9.14||87||111||1.00|| || ||0.043|
|cis-Nonachlor2||Not detected||296||336||1.00|| || ||0.339|
|p,p′-DDE3||≤134.41||103||112||1.00|| || ||0.027|
|p′,p′-DDT2||Not detected||289||305||1.00|| || ||0.491|
|HCB4||≤11.45||83||108||1.00|| || ||0.001|
|Mirex5||≤1.43||248||306||1.00|| || ||0.013|
|Oxychlordane6||≤6.07||79||112||1.00|| || ||<0.001|
|trans-Nonachlor7||≤8.97||91||117||1.00|| || ||0.009|
Oxychlordane remained statistically significant after adjustment for all other organochlorine compounds. Table VII shows the OR for oxychlordane and the 3 summed PCB variables when adjusted for each other. After adjustment for oxychlordane, only summed total non-dioxin-like PCB level was significantly associated with NHL (p trend = 0.045, highest versus the lowest quartile OR = 1.61 (95% CI = 0.94–2.73). There was no interaction observed between oxychlordane and non-dioxin-like PCB congeners (p = 0.19). Although after adjustment for oxychlordane, the trend was not significant (p = 0.060), the risk of NHL for an individual with the highest quartile of dioxin-like PCB congeners remained significantly increased (OR = 1.80, 95% CI = 1.02–2.17). Nearly significant associations (p < 0.10) were observed for PCB 180, PCB 187 and mirex (not shown). Because of the moderate correlations between oxychlordane and the other organochlorine compounds (0.40 < r < 0.70), the risks for both oxychlordane and the other analytes were attenuated, with greater attenuation occurring for those compounds with less significant associations.
Table VII. Multivariate Models with Risk Estimates for Oxychlordane and Summed PCB Variables
| Oxychlordane||≤6.07||1.00|| || ||0.002|
| Non-dioxin-like||≤88.57||1.00|| || ||0.045|
| PCB summed||>88.57–136.19||1.09||0.68||1.76|| |
| Model 2|
| Oxychlordane||≤6.07||1.00|| || ||0.006|
| Dioxin-like||≤10.12||1.00|| || ||0.060|
| PCB summed||>10.12–15.35||1.28||0.79||2.08|| |
| Model 3|
| Oxychlordane||≤6.07||1.00|| || ||0.002|
| Total PCB summed||≤100.9||1.00|| || ||0.102|
No significant interactions were observed between PCB congeners or pesticide analytes and sex, age, region, education or BMI. There were significant interactions observed between ethnicity and both p,p′-DDT (p interaction = 0.003; above vs. below the DL, European, OR = 1.15 (95% CI = 0.82–1.62), p trend = 0.40; non-European, OR = 0.39 (95% CI = 0.21–0.73) p trend = 0.013) and PCB no. 105 (p interaction = 0.035; above vs. below the DL, European, OR = 0.90 (95% CI = 0.66–1.25), p trend = 0.54; non-European, OR = 2.04 (95% CI = 1.02–4.11) p trend = 0.045).
Table VIII shows the associations by histologic subtype of NHL. Except for PCB no. 28, the associations between the PCB congeners and pesticide analytes were consistent across the 4 NHL subtypes examined (all p values for heterogeneity > 0.25). For PCB 28, the heterogeneity approached significance (p = 0.052), primarily due to an increased risk for diffuse large cell lymphoma and a decreased risk for follicular lymphoma. Associations were generally strongest for the other B-cell category. Only PCB no. 118, HCB, oxychlordane and trans-nonachlor showed significant associations with diffuse large cell lymphoma, and only mirex, oxychlordane and PCB no. 180 showed significant associations with T-cell lymphoma. Oxychlordane had statistically significant highest versus lowest quartile OR of greater than 2.4 for all 4 subtypes and was strongest for follicular lymphoma (highest vs. lowest quartile OR = 3.26, 95% CI = 1.67–6.37).
Table VIII. Odds Ratios for Largest Versus the Smallest Quartile (95% CI) for Organochlorines by Histologic Subtype of NHL
|No. of Cases||422||141||67||47||167|
|Total PCB summed||2.1 (1.4–3.3)1||2.0 (1.1–3.7)2||1.8 (0.8–4.1)||1.7 (0.7–4.3)||2.8 (1.5–5.1)1|
|Dioxin-like summed||2.4 (1.5–3.8)1||2.5 (1.3–4.7)1||2.1 (0.9–4.9)||1.5 (0.6–3.8)||3.1 (1.6–5.8)1|
|PCB no. 1053||1.1 (0.8–1.4)||0.9 (0.6–1.4)||0.8 (0.5–1.5)||0.7 (0.4–1.5)||1.2 (0.8–1.7)|
|PCB no. 118||1.8 (1.2–2.7)1||2.0 (1.1–3.7)2||2.0 (0.9–4.7)2||1.2 (0.4–3.0)||1.8 (1.0–3.2)2|
|PCB no. 156||1.8 (1.1–2.7)1||2.4 (1.2–4.5)1||1.3 (0.6–3.0)||1.4 (0.6–3.4)||1.9 (1.0–3.4)|
|Non-dioxin-like summed||2.2 (1.4–3.4)1||2.1 (1.1–3.9)2||1.8 (0.8–4.1)||1.7 (0.7–4.1)||2.8 (1.5–5.2)1|
|PCB no. 283||1.0 (0.7–1.3)||0.7 (0.4–1.3)||1.3 (0.7–2.4)||1.1 (0.5–2.3)||0.9 (0.5–1.4)|
|PCB no. 99||1.3 (0.9–1.9)2||1.3 (0.8–2.3)||1.0 (0.5–2.0)||0.7 (0.3–1.7)||1.6 (1.0–2.7)2|
|PCB no. 138||1.5 (1.0–2.2)2||1.5 (0.9–2.7)||1.2 (0.6–2.6)||1.2 (0.5–2.8)||1.7 (1.0–2.9)2|
|PCB no. 153||1.8 (1.2–2.7)1||2.0 (1.1–3.7)2||1.3 (0.6–2.7)||1.6 (0.7–4.0)||2.0 (1.1–3.7)2|
|PCB no. 170||1.8 (1.2–2.8)1||1.5 (0.8–2.8)||1.6 (0.7–3.6)||1.9 (0.7–4.6)||2.5 (1.3–4.6)2|
|PCB no. 180||1.9 (1.2–3.1)1||1.6 (0.8–3.1)||1.2 (0.5–2.9)||3.5 (1.2–9.7)1||2.7 (1.4–5.2)1|
|PCB no. 1834||1.2 (0.9–1.7)||1.6 (1.0–2.7)2||0.8 (0.4–1.6)||1.1 (0.5–2.4)||1.2 (0.8–1.8)|
|PCB no. 187||1.9 (1.2–3.0)1||1.8 (1.0–3.3)||1.7 (0.7–4.0)||1.7 (0.7–4.3)||2.3 (1.2–4.1)1|
|β-HCCH||1.6 (1.0–2.5)2||1.2 (0.6–2.3)||1.5 (0.6–3.5)||1.4 (0.4–4.8)||2.1 (1.2–3.8)1|
|cis-Nonachlor3||1.2 (0.9–1.6)||1.1 (0.7–1.6)||1.1 (0.6–1.9)||1.1 (0.6–2.2)||1.3 (0.9–1.9)|
|p,p′-DDE||1.4 (0.9–2.2)2||1.8 (0.9–3.3)2||0.6 (0.2–1.5)||1.0 (0.4–2.6)||1.8 (1.0–3.2)2|
|p,p′-DDT||0.9 (0.7–1.2)||0.7 (0.5–1.1)||1.0 (0.6–1.7)||1.1 (0.6–2.1)||1.0 (0.7–1.5)|
|HCB||1.9 (1.3–3.0)1||2.4 (1.2–4.6)2||2.5 (1.0–6.0)2||1.4 (0.5–4.0)||1.6 (0.9–2.9)2|
|Mirex3||1.4 (1.1–1.9)2||1.3 (0.9–2.0)||1.3 (0.8–2.3)||2.2 (1.2–4.3)2||1.5 (1.0–2.1)|
|Oxychlordane||2.7 (1.7–4.2)1||3.3 (1.7–6.4)1||2.9 (1.2–7.2)1||2.4 (0.9–6.5)2||2.5 (1.3–4.7)1|
|Trans-Nonachlor||1.7 (1.1–2.6)1||1.7 (0.9–3.1)||1.9 (0.9–4.3)1||1.1 (0.5–2.9)||2.0 (1.1–3.6)2|
For the congeners and analytes that were not significantly associated with all NHL, only PCB congener no. 183 was significantly associated with a lymphoma subtype-follicular lymphoma (highest vs. lowest quartile OR = 1.64, 95% CI = 1.00–2.70).
The major strength of this population-based case–control study is the large number of subjects with measured organochlorine levels. Our study had over 4 times the number of cases as the next largest study.6 Because of the small sample sizes in the previous studies of organochlorines, negative findings may simply be reflective of lack of power. A further strength of our study is the elimination of cases with weight loss after diagnosis. Inclusion of these subjects in previous case–control studies may have led to observed associations that were biased upward.
The study has several weaknesses. Despite the relatively large sample size, the study still has limited power to detect associations for specific NHL subtypes or to detect interactions. Although previous studies have suggested that the associations of PCB congeners and organochlorine pesticides with NHL to be stronger in the presence of the Epstein-Barr virus (EBV),7, 8 we did not measure EBV markers in this study so could not confirm this interaction.
Although we did not assess organochlorine levels in adipose tissue, a strong correlation between organochlorines detected in plasma and adipose tissue has been documented.20 Weight loss information for controls and for cases after diagnosis was not available; however, controls with weight loss would likely have higher organochlorine levels, thus the bias from not removing controls with weight loss would be to reduce the observed association between organochlorines and NHL. Also, despite the importance of acquiring pre-chemotherapy blood samples and eliminating cases with excessive weight loss, other biologic changes around diagnosis of NHL may alter the organochlorine levels.
Residual confounding may also have biased risk estimates in this study. There were other factors that are recognized as potential confounders, but were not included in our analyses simply because of lack of data. Other important predictors of organochlorine uptake, elimination and overall body burden that may also be potential confounders but were not recorded include current BMI (we only collected weight one year prior to interview), body fat index, diet and lactation in women.21, 22
We did not measure dioxins, furans or coplanar PCB congeners, which along with the dioxin-like PCB nos. 105, 118 and 156 measured in this study, are inducers of the aryl hydrocarbon (Ah) receptor pathway. A long-term study of the Seveso residents exposed to high levels of dioxin, particularly the most toxic congener TCDD, revealed increased mortality from specific cancers including NHL.23 Two other studies have examined levels of dioxins, furans and coplanar PCB congeners in relation to NHL.6, 9 In a small Swedish study of organochlorine levels in adipose tissue of 28 cases and 17 controls, Hardell et al. showed no significant increased risks related to dioxins, furans or coplanar PCB congeners.9 In an analysis of plasma samples (100 cases and 100 controls) from a U.S. case–control study, De Roos et al. also found no associations with dioxins, however, found significant associations with several furan compounds and coplanar PCB no. 169.6
Another limitation of the study was that the high rate of nonresponse could have introduced bias if cases and controls differentially participated based on their organochlorine levels. Subjects knew of the objectives of the study, but were unlikely to know of their serum organochlorine levels prior to choosing whether to participate. Also, when the estimates of risk for organochlorines changed with adjusting for education or history of farming, these factors were included in the model. Finally, the results of our study are consistent with previous cohort and case–control studies. These considerations argue against a major response bias.
We found a significant association between NHL and 6 pesticide analytes; p,p′-DDE, HCB, β-HCCH, mirex, oxychlordane and trans-nonachlor. Several previous studies have found associations between biological measurements of organochlorine pesticide analytes (adipose tissue, plasma or dust) and NHL, although the results have not been consistent. Our study supports the previously reported associations between NHL and chlordane,7, 11 DDT,11, 24 HCB7, 11 and β-HCCH.11
In a nested case–control study of 74 cases and 147 matched controls from a U.S. cohort study, Rothman et al. found a weak nonsignificant association with serum concentration of DDT that disappeared after adjustment for total PCB.8 In a later publication utilizing the same subjects, Cantor et al. found no association with any organochlorine pesticide examined including chlordane, DDT, β-HCCH and HCB.5 De Roos et al. also reported no associations with chlordane, DDT, β-HCCH and HCB.6 Neither study had a sufficient number of cases with mirex above detectable levels for analysis.
In a Swedish case–control study of 82 cases and 83 controls utilizing a combination of blood and adipose tissue, Hardell et al. found an increased NHL risk with higher levels of cis-nonachlor, but not oxychlordane or trans-nonachlor.7 In a study of adipose tissue from cadavers (175 NHL cases, 481 controls who died from other causes), Quintana et al. also showed a near-significant association with HCB, but found that the associations with DDT and HCB weakened when adjusting for heptachlor epoxide, a metabolite of heptachlor.11 Heptachlor and chlordane are structurally related compounds and commercial preparations of each contain 10–20% of the other compound.25 They also observed strong positive associations with β-HCCH and dieldrin, a metabolite of aldrin, which was measured in our study, but with no samples measured above the detectable limit. It should be noted that this study had several methodological limitations.4 Finally, in a U.S. study of organochlorines measured in carpet dust from 682 cases and 513 controls, Colt et al. found increased risk associated with higher levels of DDE, but not chlordane.24
All 5 organochlorine pesticides we found to be associated with NHL are classified as possibly carcinogenic (2B) by the IARC monograph classification, with sufficient evidence in animals and inadequate evidence in humans.25, 26, 27 An exact mechanism is unknown but it may be due to direct mutagenicity, tumor promotion or other methods. All 4 pesticides have been shown to cause alterations to the immune system, including immune suppression,28, 29, 30, 31, 32, 33 however, the mechanisms by which organochlorine pesticides may increase the risk of NHL remain to be determined.
Our study confirms the associations with biological PCB congeners and NHL risk observed previously.6, 8, 9, 34 However, the individual PCB congeners measured and the actual PCB congeners associated with NHL differed between the studies. We found dioxin-like PCB congeners nos. 118, 156, non-dioxin-like PCB congeners nos. 138, 153, 170, 180 and 187 and summed dioxin-like, non-dioxin-like and total PCB to be associated with NHL. In particular, PCB congener no. 180 showed the strongest association among the PCB congeners. De Roos et al. found significant associations with total PCB and PCB congeners nos. 156, 180 and 194 (not measured in our study), however, found no association for PCB congeners nos. 118, 138, 153, 170 or 187.6 In the Swedish case–control study, Hardell et al. found a significant association with the sum of immunotoxic PCB congeners,35 but no association with total PCB.34 Associations with individual PCB congeners were not reported. In an earlier publication using a subset of the subjects with adipose tissue (28 cases and 17 controls), Hardell et al. found significant associations with PCB congeners nos. 156, 170 and 187 and borderline nonsignificant but positive associations with PCB congeners nos. 153 and 180, but observed no association with nos. 118 or 138.9 Rothman et al. found a significant association with total PCB, but did not report results on individual PCB congeners.8 In a recent combined analysis of this and 2 other cohorts, significant associations with PCB congeners nos. 118, 138 and 153 were found across the 3 cohorts.4 In the cadaver study, Quintana et al. failed to find an association with total PCB as estimated by Aroclor 1254 or Aroclor 1260.11 In the study of carpet dust, Colt et al. found increased risk associated with total PCB and PCB congeners nos. 153, 170 and 180, but not no. 138.36
PCBs are classified as probable carcinogens (2A) by the IARC monograph classification, with sufficient evidence in animals and limited evidence in humans.27 Although studied to a greater extent than the organochlorine pesticides, the mechanisms by which PCBs increase lymphoma risk are not clear. PCBs may increase the risk of NHL through immunotoxicity, immunosuppression or tumor promotion.35, 37, 38, 39, 40, 41, 42, 43 PCBs activate and are metabolized by the cytochrome P450 (CYP) enzymes; however, the CYP induction potential and specificity of CYP induction differ by specific PCB congener.40 Agonists of the Ah receptor pathway, such as coplanar PCB congeners, activate the CYP1A1 enzyme.44, 45 PCBs can also activate CYP2B enzymes independently of the Ah receptor pathway, and these PCB congeners are sometimes referred to as phenobarbital (PB)-like inducers.46, 47, 48 Although not as potent as coplanar PCB congeners, which are rare in the environment, other dioxin-like PCB congeners (nos. 105, 118 and 156) can bind the Ah receptor and activate CYP1A1, have PB-like induction potency and high potential toxicity.40 In our study, 2 of the 3 dioxinlike PCB congeners and the sum of the 3 dioxin-like PCB congeners showed significant associations with NHL. Other PCB congeners with mixed CYP induction potencies (nos. 138 and 170), pure PB-like inducers (nos. 153 and 180) or with no known activity (no. 187) as well as the sum of the non-dioxin-like PCB congeners were also found to increase the risk of NHL.
Our results support the hypothesis that certain PCB congeners and organochlorine pesticides increase the risk of NHL. We found that the increased risk was evident for both dioxin and non-dioxinlike PCB congeners and several pesticides analytes. The strongest association observed was with oxychlordane, a metabolite of the insecticide chlordane. Associations of the PCB congeners and organochlorine pesticides were fairly consistent across the 4 NHL subtypes examined.
The analysis of the joint effects of the organochlorine compounds provides evidence that both oxychlordane and non-dioxinlike PCB congeners independently increase the risk of NHL, however, which PCB congeners and organochlorine pesticides are the true risk factors is still unclear. It is possible that random variation or differences in measurement error between analytes led to the observed results. It is also possible that other unmeasured organochlorines, related to NHL and correlated with the measured analytes, are confounding the observed associations. Research examining the possible interactions between organochlorine measurements and exposure to other agents (e.g., viruses) and genetic susceptibility in needed to further clarify the biological mechanisms.
We thank the study staff, Ms. Kuldip Bagga, Ms. Agnes Bauzon, Ms. Betty Hall, Ms. Lina Hsu, Ms. Pat Ostrow, Ms. Lynne Tse and Mr. Anthony Tung, the computer support of Dr. Tim Lee and Ms. Zenaida Abanto, the blood processing provided by Ms. Rozmin Janoo-Gilani, Ms. Pat Lee and Mr. Stephen Leach and the support of the members of the BC Cancer Agency Lymphoma Tumour Group. Finally, we thank all the participants of the study for making this research possible. Ms. Carmen Ng was supported by the Michael Smith Foundation for Health Research.