Human Exposure to Poultry and Poultry Products and the Risk of Death from Hematopoietic & Lymphatic Cancers

Purpose: The purpose of this study was to obtain preliminary information on occupational exposures responsible for the excess of hematopoietic & lymphatic (H&L) cancers previously observed in cohort mortality studies of workers in poultry slaughtering and processing plants. Methods: A pilot case-cohort study was conducted nested within a combined cohort of 30,411 Short Communication poultry workers and 16,408 controls, identified from several United Food & Commercial Workers (UFCW) unions across the United States. Interviewed cases were 48 deaths out of a total of 130 deaths (37%) from H & L tumors that occurred in the cohort between 1990-2003. Controls were N=152 subjects that were successfully interviewed out of 1516 subjects (10%) randomly selected from all cohort members alive as of January 1, 1990. Telephone interviews were obtained directly from live control subjects or next-of-kin for deceased cases and controls. Mortality risk was assessed using logistic regression odds ratios and hazard ratios. Results: Poultry farming [OR=10.8 (95% CI: 3.0-39.1)] and spreading chicken wastes as manure [OR=5.6 (95% CI: 1.5-20.4)] were significantly associated with lymphoma; handling raw eggs in supermarkets [OR=4.3 (95% CI: 1.0-18.0)] was significantly associated with leukemia. Non-poultry exposures significantly associated with these tumors included coal by-products, selling seafood, and killing of pigs. Conclusion: This preliminary study identifies possible occupational exposures that may be associated with excess deaths from H & L tumors in poultry workers. Case-control studies of sufficient statistical power are now needed to confirm these findings and discover new ones.


INTRODUCTION
Workers in poultry slaughtering and processing plants perform a variety of tasks, ranging from unloading live birds from trucks when they arrive at the plant, to shackling, killing, scalding, eviscerating, chilling, further processing, packing, storing in the cold room, and loading processed products into trucks for delivery to retail establishments [1]. In the process, the workers come into contact with the blood, internal organs, secretions, and skin of thousands of chickens daily. In some plants, up to 175,000 chickens are killed each day. This brings them into intimate contact with poultry blood, organs, and secretions. Frequent cuts and injuries from sharp knives and bone splinters provide a portal of entry for microorganisms to enter the body through the skin [2]. Additional hazards to workers include airborne transmission of microbial agents that also occurs in poultry plants [3].
Oncogenic viral exposures in poultry plants include viruses such as avian leukosis/sarcoma viruses (ALSV), reticuloendotheliosis virus (REV), and Marek's disease virus (MDV), all of which have been shown to cause hematopoietic and lymphatic (H & L) cancers in poultry [4]. Also, poultry workers are known to have a high prevalence of warts caused by human papilloma viruses (HPV) [5,6].
Humans are widely exposed to these viruses. What remains unknown is whether ALSV, REV and MDV can also cause cancer in humans. The literature examining the health risks due to occupational exposures among workers in poultry slaughtering and processing plants has shown excess mortality risks for H & L cancers [7][8][9][10][11][12]. In vitro experiments have revealed that these viruses can infect and cause malignant transformation in human cells [13]. In addition, some studies have reported the presence of antibodies to all of these viruses in the sera of poultry workers and in general population subjects [14][15][16], while others have not [17]. The workers are also potentially exposed to chemical carcinogens at work. This paper attempts to provide preliminary data for the first time, on which specific occupational exposures in poultry slaughtering and processing plants may be candidates to explain the reported increased mortality due to these cancers among poultry workers.

MATERIALS AND METHODS
The source population consisted of the 30 1,1990, and all subjects lost to follow-up were assumed to be alive at the end of the study, the group of subjects alive as of January 1, 1990 was essentially a closed cohort.
Cases were defined as deaths from all types of H & L cancers (ICD, Ninth Revision, codes 200, 202-208; or ICD Tenth Revision, codes C82-C96) that occurred in the base population between January 1, 1990, and December 31, 2003 (N=130). The comparison group was a subcohort that consisted of 1,516 live subjects randomly sampled from the base population, some of whom later died during the study period.
Because this was a pilot study, exhaustive attempts were not made to trace study subjects or their next-of-kin. Thus, we report here on the first 48 deaths (37%) from H & L cancers (cases) whose next-of-kin were traced, and completed a telephone interview that provided information on the deceased cases within the limited time available to do the study. Similarly, we report on the first 152 members of the subcohort (10%) for whom an interview was obtained either directly if the subject was alive, or from the next-of-kin if the subject was deceased.
The questionnaire was administered to the nextof-kin of deceased study subjects, or to the subject himself or herself if alive, using a computer-assisted telephone software (Questionnaire Development System, NOVA Research Company, Bethesda, Maryland). It contained over 600 questions on occupational poultry specific exposures, history of exposures at work (non-poultry related), medical history, medications, lifestyle, immunizations, and diet, that were intentionally broadly framed to facilitate recall. The questionnaire took an average of 40-60 minutes to complete. Risk associated with each job exposure was calculated for "ever/never" responses. Dates and duration of working at each task are not presented because of sparse data.
To obtain information on the reliability of responses from proxies, the questionnaire was administered to a small subset of seven pairs of live control study subjects and their next-of-kin.
Baseline data from cases and controls were compared using chi-square tests. Odds ratios (ORs) and 95% confidence intervals (95% CI) were estimated using unconditional logistic regression (SAS 9.1, SAS Institute, Cary, NC). Since time-to-event information was available, a Cox regression analysis was used to estimate hazard ratios (HR) and 95% CI. Failure time for formation of a risk set was the difference between age at study entry (January 1,1990) and age at time of a case-related death. At failure time, a risk set was formed consisting of the case and all available controls at risk at that time. All failures were included, regardless of whether they occurred in the sub-cohort or not. A case outside of the subcohort was not at risk until just before its failure, and therefore was not included in the earlier risk sets. All variables were tested to determine whether they satisfied the proportional hazards assumptions. Risk estimates were adjusted for both categorical age and union status (Chicago, Baltimore, and Missouri). Analyses of poultry and non-poultry occupational exposures were performed across all H & L histologic subtypes in aggregate. Exploratory analyses were further conducted when possible across specific H & L types (leukemia, lymphoma, and multiple myeloma).

RESULTS
Of the first 61 cases who died from cancer of the H & L systems whose next-of-kin were traced, interviews were obtained for 48 (79%) within the limited time available to do the study. Similarly, of the first 214 subjects in the subcohort (n = 1,516) or their next-of-kin that were traced during the time available, 152 (71%) completed phone interviews with the same questionnaire either directly (if alive), or through their next-of-kin if deceased. None of the interviewed cases were members of the subcohort control group.
The limited comparison that was performed on exposure information provided by seven pairs of live control study subjects with their respective next of kin indicated that of the 245 direct responses obtained for dichotomous questions, there was an agreement of 80% to 100% between the pairs for nearly 75% of the responses, with less than 60% agreement for only 8% of the responses.
Baseline demographic information on interviewed cases and controls are given in Table 1.
The results obtained by logistic regression and by Cox regression analyses were similar (Table 2). Histologic subtype analyses results for leukemia and lymphoma for the four poultry exposures with elevated ORs for all H & L tumors combined are given in Table 3.
In Table 4, risk estimates are given for potentially carcinogenic exposures in poultry plants other than oncogenic viruses.
The risk estimates for non-poultry occupational exposures are provided in Tables 5 and 6.

DISCUSSION
In Table 1, interviewed cases were similar to all cases, and interviewed controls were similar to the subcohort for demographic variables. The differences in ages between cases and controls are expected since controls were a random sample of the cohort. Accordingly, we adjusted for age, for which there was a significant difference. We also adjusted for union status because geographic variability in unknown factors may affect the results, especially as the Missouri cohort was all poultry and did not have subjects outside the poultry industry. Because of sparsity of the data, we could not adjust for race.

Carcinogenic Exposures in Poultry Slaughtering and Processing Plants
There are six potentially carcinogenic exposures that occur in poultry plants: 1) poultry oncogenic viruses, such as ALSV, REV, MDV, and papilloma viruses [4,5,13], which may be found in live animals, raw poultry products, raw eggs, or in the air; 2) aflatoxin that is present in the air inside plants, that is produced by the fungus aspergillus [18]; 3) polycyclic aromatic hydrocarbons (PAHs) emitted during the smoking of poultry [19][20][21]; 4) benzene, phthalate, and PAHs present in fumes emitted from the wrapping machine during the wrapping of poultry [22,23]; 5) heterocyclic amines emitted during the cooking or frying of poultry meat [24,25]; and 6) exposure to nitrosamines during the curing of poultry meat [25][26][27].

Carcinogenic exposure to poultry oncogenic viruses
Working as a poultry farmer, spreading of poultry wastes, handling raw eggs, working in the kill room, and contact with poultry blood involve tasks that would have brought workers in contact with high exposure to oncogenic viruses. These jobs were clearly associated with high risks of lymphoma or leukemia, except for killing and contact with blood, for which the increase was modest. One contributing factor to the lower risk estimates in this study was the relatively higher proportion of controls that were poultry workers (27%) compared to 1% in Metayer et al.'s study [8]. Controls in the current study were a random sample of the entire cohort, which also included poultry workers, while in the Metayer et al. [8].
study, the comparison group consisted of workers exclusively employed outside the meat industry. Our findings are consistent with those of the casecontrol study in which the reported risk for H & L tumors was twice as high in workers exposed to high levels of poultry and red meat blood combined [28]. The results are also consistent with the reported association of poultry farm exposure and non-Hodgkin's lymphoma [29], and also between growing up in a poultry farm and hematologic cancers [30].
The association of leukemia with handling eggs in supermarkets is interesting, bearing in mind that one of our studies showed that 1 in every 12 eggs displayed for sale in supermarkets in the New Orleans metropolitan area in the United States carried ALSV [31].
All of the poultry workers would have been potentially exposed to aflatoxin that would be present in the air inside the plants [18], or to papilloma viruses, because of the high prevalence of warts among them [5,6]. The role of these two exposures in the occurrence of these tumors cannot be ruled out, although it will be difficult to isolate them as independent risk factors in an epidemiologic study like this.

Carcinogenic exposure during smoking of poultry meat
Smoking of poultry meat that involves exposure to PAHs appeared to be associated with both leukemia and lymphoma. Although the ORs were not significant, the findings are consistent with the established causal association between PAHs and leukemia and lymphoma in humans [32].

Carcinogenic exposures during cooking or curing of poultry meat
The odds ratios for cooking or curing poultry meat were not elevated, suggesting that these exposures may not be involved in the excess occurrence of H & L tumors in poultry workers.

Carcinogenic exposure during wrapping of poultry meat
Surprisingly, the previously postulated possible association of these tumors with wrapping meat in supermarkets [7,8,33] was not evident here. A plausible explanation for the absence of an association in this report may be because all the responses for cases (who were all deceased) for exposure to fumes from the wrapping machine were from proxies who may not have known much about this specific exposure, while information for controls was mainly provided by the live study subjects themselves. In fact, the question on exposure to fumes from the wrapping machine was the least reliably reported exposure in the questionnaire, and, in the small validation study, the agreement between study subject and next-ofkin on this question was only 40%. Another reason is that this exposure was high and prevalent before 1975, and, after that date, it was drastically reduced to insignificant levels. The data in this study, however, are too sparse to investigate this exposure by calendar year of exposure. Thus, this study is unable to evaluate the effects of exposure to fumes from the wrapping machine.

Exposures Outside the Poultry Industry
Working in a stockyard handling cattle, pigs, sheep, or other animals appears to be associated with risk of tumors of the H& L systems. This is a consistent finding for other cancers in previous reports involving these same cohorts [8,[34][35][36]. Although this activity does not involve exposure to poultry, it does involve possible exposure to the oncogenic viruses of cattle, pigs, sheep and goats such as the bovine leukemia virus, Jaagsiekte sheep retrovirus, and papilloma viruses of these animals. It also is consistent with the observed increased risk of leukemia and lymphoma associated with occupational exposure to beef cattle and working in the meat industry [28,[37][38][39][40].
Selling seafood, killing pigs, and exposure to coal tar, turpentine, naphthalene, natural gas, paraffin, and smoke were all significantly associated with H & L tumors. The association between H & L tumors and coal tar and carbon products is established [32], and that with killing pigs is consistent with other reports [40]. However, the finding for selling seafood, especially in relation to multiple myeloma, is novel and unexpected, and would need to be confirmed in other studies.
We considered other explanations for the findings in this study. Though possible, selection bias is not likely to be a serious limitation in this study. All the cases that occurred in the cohort were included for study, and the controls were a random sample of the baseline cohort. Also, interviewed cases and controls were similar to their underlying source populations on demographic variables. Furthermore, 79% of the next-of-kin of cases that were contacted granted an interview, and for controls the response rate was 71%. Bias in the recall of exposures could potentially occur since responders for all cases and deceased controls were next-of-kin proxies, while responders for live controls were themselves the study subjects. As shown above, in the limited assessment of this issue, the agreement between the responses from proxies and those from the study subject themselves was quite good for the vast majority of responses. We lacked sufficient information regarding the duration of time involved in a specific occupational task. Hence, we were unable to determine whether there was a doseresponse pattern. Also, the occupational task exposures were not mutually exclusive. Some employees had multiple jobs and therefore multiple exposures within this industry. In this case, effect estimates may have been incorrectly estimated because of overlapping exposures. The diagnosis of H & L cases was based on underlying cause of death as coded in the death certificates. Previous studies found the detection and confirmation rates for hematopoietic cancers as the underlying cause of death were very high, with over 96% for leukemias and over 80% for lymphomas [41]. Thus, misclassification of disease should be less of a concern in our study.

CONCLUSION
In summary, this is a report of a pilot study, and limitations associated with pilot studies can be expected to be operative, especially the lack of statistical power, and therefore the findings should be regarded as preliminary. The study suggests an association between jobs in poultry slaughtering & processing plants involving exposure to oncogenic viruses and mortality from leukemia and lymphoma. This finding may be important considering that the general population is widely exposed to these viruses from contact with live chickens or ingestion of their raw or undercooked products including eggs, or through vaccination with vaccines manufactured by growing vaccine virus in contaminated eggs [14,15,16,31,42]. It does not appear that the excess occurrence of these tumors in poultry plants is associated with cooking or curing of poultry meat. The study could not adequately investigate carcinogenic exposures that occur during the smoking or wrapping of poultry meat, and these two exposures need to be investigated in studies with adequate statistical power.
The observed association with selling seafood, killing of pigs and working in a stockyard with these tumors are preliminary, and in the case of selling seafood new, and could be chance findings. The study though small was able to replicate the established association between exposure to byproducts of coal tar and natural gas that may contain PAH and benzene that cause leukemia and lymphoma. Also, leukemia and lymphoma have been linked with exposure to gasoline in gasoline stations and gasoline storage facilities and pesticides [32]. The finding that these established relationships were also replicated to a degree in this small study gives some credence to the new associations reported here. The uniqueness of the study lies in its singular focus on poultry workers, a group that carries the distinction of having one of the highest known human exposures to the oncogenic viruses of poultry. Ours is the only study of its kind to provide any detail on potential associations between occupational exposures in poultry plants and the occurrence of hematopoietic and lymphatic malignancies. The findings should be regarded as preliminary, but importantly, they point to possible risk factors (both poultry and non-poultry related) for these tumors that may need further investigation and replication in other studies with sufficient statistical power, especially as they also have implications for the general population. Finally, this pilot study indicates full blown larger studies are feasible, and point to issues that may need to be addressed in these studies. These findings call for a possible role of agencies such as the International Agency for Research on Cancer (IARC) and the National Institute for Occupational Safety & Health (NIOSH) to promote such studies, and for an agency such as the US Occupational Safety & Health Administration (OSHA) to begin to consider biological monitoring of exposures in the workplace for this occupational group.

CONSENT
Access to study subjects' union records was granted by the United Food & Commercial Workers International Union in Washington DC. Live study subjects and next-of-kin were read an informed consent document over the phone prior to interview, and granting of the interview was considered consent given. This measure was approved by the IRB.

ETHICAL APPROVAL
Approval was provided by the Human Subjects Committee (Institutional Review Board) of the University of North Texas Health Science Center -Approval Number IRB # 25-32.

FUNDING
This study was funded by grant # 1 RO1 OH008071 from the National Institute for Occupational Safety & Health.