Cancer Incidence among Refinery and Petrochemical Employees in Louisiana, 1983–1999 SHAN P. TSAI, PHD, VIVIEN W. CHEN, PHD, ERIN E. FOX, MS, JUDY K. WENDT, MPH, XIAO CHENG WU, MD, MPH, DELIA E. FOSTER, MS, AND ALISTAIR E. FRASER, MB, CHB
PURPOSE: The purpose of this study is to determine the incidence of cancer among employees at two petrochemical facilities in south Louisiana, and to compare their cancer rates to those of the general population of south Louisiana. METHODS: Records on 4639 active and former employees and retirees from the two plants were linked to the Louisiana Tumor Registry (LTR) database by LTR staff to ascertain incident cases of cancer. Standardized incidence ratios (SIRs) were then calculated using the south Louisiana population as the comparison and adjusted for age, race, and time period. RESULTS: There was a significant 16% deficit of overall cancer cases for males in this cohort (SIR ⫽ 0.84; 95% CI, 0.74–0.95). The only significantly elevated SIR in males was for cancer of the bone and joint (SIR ⫽ 6.89; 95% CI, 1.42–20.1). This result was based on three non-fatal cases of bone cancer with different histologies, occurring in different parts of the body. These cases worked in different units of one plant. Significant deficits were seen for lung cancer, non-Hodgkin’s lymphoma, and cancer of the oral cavity and pharynx. Cancer incidence among 719 female employees was nonsignificantly increased (SIR ⫽ 1.24; 95% CI, 0.81–1.82). Breast cancer accounts for the excess (SIR ⫽ 1.46; 95% CI, 0.73–2.61). Seventy percent of the breast cancer cases worked in an office setting. CONCLUSIONS: This study found little evidence of any association between cancer incidence and employment at these two petrochemical facilities. The increased incidence of bone cancer is unlikely to be due to occupational exposures. The non-significant excess of breast cancer may be due to early detection or other important unmeasured confounders, such as certain reproductive factors. Ann Epidemiol 2004;14:722–730. 쑕 2004 Elsevier Inc. All rights reserved. KEY WORDS:
Cancer, Incidence, Refinery, Petrochemical, Occupational Health.
INTRODUCTION As part of a comprehensive occupational health surveillance program, Shell Oil Company (Shell) routinely collects and evaluates mortality data on its employees and pensioners. The most recent mortality analyses have not shown significant increases for any cause of death, including site-specific cancers (1–4). However, mortality studies may be subject to survival bias, especially in occupational studies in which employees may have access to better health care than the general population. Cancer incidence data provide a useful
From the Shell Oil Company, Shell Health Services, Houston, TX (S.P.T., E.E.F., J.K.W., D.E.F., A.E.F.); and Louisiana Tumor Registry, School of Public Health, Louisiana State University Health Sciences Center, New Orleans, LA (V.W.C., X.C.W.). Address correspondence to: Dr. Tsai, Shell Oil Company, Shell Health Services, One Shell Plaza, P.O. Box 2463, Houston, Texas 77252-2463. Tel.: (713) 241-6078; Fax: (713) 241-5875. E-mail:
[email protected] Work by Louisiana Tumor Registry staff was supported fully by the Louisiana State University Health Sciences Center, the Centers for Disease Control and Prevention Cooperative Agreement # U75/CCU618724 and U55/CCU621886, and the National Cancer Institute’s SEER Contract # NO2-PC-15106. Received October 15, 2003; accepted January 22, 2004. 쑕 2004 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
supplement to cancer mortality information previously described for this work force (1–4). Incidence and its associated effect measures, such as standardized incidence ratios (SIRs), are direct reflections of cancer occurrence regardless of survival. To determine the existence of any unrecognized associations between cancer and employment, Shell, in collaboration with the Louisiana Tumor Registry (LTR), performed a cancer incidence study in Louisiana, where two Shell manufacturing plants (at Norco and Geismar) are located, and the general population is served by LTR. Ascertaining incident cancer cases from registry data makes it possible to identify cancer occurrence in a more accurate, detailed, and complete manner than if using mortality data or relying on employee self-report. If significantly elevated cancer rates exist among Shell employees in Louisiana, a cancer incidence study will be able to detect that association with greater statistical power than a mortality study. The feasibility and importance of linking occupational cohorts with population-based cancer registries has been previously shown in the literature (5–12). This study furthers the methodology of obtaining detailed residential histories over more than 15 years. The purpose of this study was to determine cancer incidence in employees at two Shell facilities in 1047-2797/04/$–see front matter doi:10.1016/j.annepidem.2004.01.005
AEP Vol. 14, No. 9 October 2004: 722–730
Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
Selected Abbreviations and Acronyms CI ⫽ confidence interval Exp ⫽ expected number of cases LTR ⫽ Louisiana Tumor Registry Obs ⫽ observed number of cases SEER ⫽ Surveillance, Epidemiology, and End Results SIR ⫽ standardized incidence ratio SMR ⫽ standardized mortality ratio
south Louisiana and to compare their cancer rates to those of the south Louisiana general population.
METHODS Study Population Two Shell facilities located in south Louisiana were included in this study. The Norco Manufacturing Complex is an oil refinery and chemical plant that is located on the Mississippi River and has been in operation since the 1920s. It is one of the largest petrochemical manufacturing facilities in the United States, producing ethylene, propylene, automotive gasoline, jet fuel, heating oil, petroleum coke, calcium chloride, butadiene, and catalysts. Other potential exposures at the facility include hydrogen sulfide, polynuclear aromatics, asbestos, and various solvents. The second facility is a chemical manufacturing operation in Geismar that opened in 1968. The plant operates processes that produce long chain olefins, converting these olefins to surfactants or hydrophobes for detergents. Other chemical products such as ethylene oxide, ethylene glycol, and glycol ethers are also produced at this facility. Currently, there are approximately 1850 employees working at these two facilities. However, the total number of active employees, former employees, and pensioners eligible to be included in this study was 4639. These study subjects were identified from two existing cohorts that contained employees from Norco and Geismar (1, 13). The study population consists of active/former and retired employees. The active/former employee group includes all full-time employees at Norco and Geismar who: 1) were actively employed on or after January 1, 1976; 2) worked more than 6 months before December 31, 1999 (end of study); 3) lived in south Louisiana any time from 1983 to 1987 and/or in Louisiana from 1988 to 1999 (also referred to as the LTR catchment area); and 4) were alive as of January 1, 1983. The retired population includes all fulltime employees who: 1) were pensioned prior to 1976; 2) were alive as of January 1, 1983; and 3) lived in the LTR catchment area. Active employees, pensioners, and former employees represent 44.1%, 29.6%, and 26.3% of the entire cohort, respectively. Due to a lack of complete data, contractors were not included. The study covers the period from January 1, 1983 to December 31, 1999.
723
The comparison population is south Louisiana, which includes 35 of the 64 Louisiana parishes (counties) and about two-thirds of the state population. South Louisiana is considered to be the most appropriate comparison because: 1) both study facilities are located in south Louisiana; 2) LTR began population-based cancer registration for south Louisiana in 1983; 3) statewide cancer registration was not in place until 1988; and 4) lifestyle and behavioral characteristics of the cohort members are likely to be similar to this population (14). Employee Demographic and Work History Data Demographic and work history data were obtained from Shell’s personnel and Health Surveillance Systems. Vital status was ascertained by a National Death Index search, which was completed for all members of this cohort through 1999 (1, 13). Work history records, which have been computerized since 1977, contain detailed employee work assignments, including dates of changes in job type or location. Gaps in employment due to leaves of absence, unemployment, or unknown assignment were also coded. This study has no information on exposures to specific substances or combinations of substances. Subgroup analysis was therefore limited to surrogates of general exposure that could be determined from work history records. The potentially exposed subgroup includes hourly production and maintenance employees and salaried employees with routine field or laboratory assignments. Employees with job titles such as accountant, draftsman, administrative assistant, and medical and security personnel were not included in the exposed group. Time of hire and duration of employment analyses were restricted to the potentially exposed subgroup. Residential History Data Residential history since 1977 for eligible employees was retrieved from company personnel and pension record systems, where more than 95% of active employees and pensioners have up-to-date addresses. However, since no address data exist for employees post-termination, data from Louisiana voter registration records and a private address database (LexisNexis) were used to supplement company address histories. We found at least one address for more than 99% of the entire cohort using the three address sources. None of the 2046 current employees, 3 of 1373 (0.2%) retirees, and 32 of 1220 (2.6%) terminated employees were lost to follow-up. Every attempt was made to ascertain each employee’s complete residential history; nevertheless, some employee addresses were missing for individual years during the study period. The percentage of the cohort that left the LTR catchment area was 14.7% over the 17-year period from 1983 to 1999.
724 Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
Cancer Incidence Data LTR staff linked Shell data to LTR cancer incidence data for the years 1983 to 1999. LTR has been collecting incident cancer cases in south Louisiana since 1983 and statewide since 1988. Invasive cancer cases plus in situ bladder cancer were counted as incident cases if the date of diagnosis occurred between January 1, 1983 and December 31, 1999. Only primary cancers were enumerated. More than 90% of LTR cancer cases have diagnostic confirmation by positive histology, cytology, or positive hematological findings (15). LTR uses the International Classification of Diseases for Oncology coding systems for primary site and morphology (16–17). Stage of disease and other variables were coded according to the rules of the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program (18) and the North American Association of Central Cancer Registries (19). LTR staff performed the data linkage using AutoMatch (currently owned by Ascential Software, Westboro, MA), a probabilistic linkage software program. Records were matched on last name, first name, social security number, date of birth, sex, and race. LTR and Shell staff manually reviewed all possible matches, and any inconsistencies were resolved by consensus. Upon resolution of the matched cases, all names and social security numbers were removed and replaced with a non-identifiable study number. This file, containing study ID numbers and demographic data, underlying causes of death (decedents only), LTR cancer diagnosis data, and Shell work history and residential history data, was used for analysis and cannot be linked back to either Shell or LTR databases. In addition, all dates were limited to month and year to further protect confidentiality.
AEP Vol. 14, No. 9 October 2004: 722–730
most conservative and most liberal estimates (see Appendix). The analysis presented in this article used the most conservative method (exit date based on last known address within the LTR catchment area), although the most liberal method (exit date based on the earliest known address outside the LTR catchment area) was also assessed. Standardized incidence ratios were calculated as the ratio of observed to expected number of cancer cases. Cancer incidence rates from south Louisiana were used as the comparison. The number of expected cases was adjusted for age, race, and time period of follow-up (1983–1989, 1990–1994, and 1995–1999). 95% confidence intervals for each SIR were calculated assuming a Poisson distribution for the observed cases and a two-sided test of significance (20). The software packages used for data analysis were Occupational Cohort Mortality Analysis Program (OCMAP) (21) and SAS (22). OCMAP was modified to accommodate cancer incidence data, recoded from ICD-O to ICD-9. SIRs based on fewer than three observed or expected cases were not calculated. Additionally, any site-specific cancers with one or fewer observed or expected cases are not presented. SIRs for potentially exposed subgroups, defined by calendar year of hire (1950 or earlier, 1950⫹) and duration of work (less than 10 years, 10⫹ years) were calculated. In addition, an analysis excluding employees who retired prior to 1973 was conducted to examine potential survival bias due to inclusion of these healthier long-term survivors. Job title and type information cannot be used to form subgroups, because detailed work histories are not complete for all subjects before 1977. The small number of females in the study cohort limited the analysis to all cancer combined and breast cancer. A separate analysis of Norco employees was also performed.
Data Analysis Employees entered follow-up on January 1, 1983 if they met the eligibility criteria and lived in south Louisiana at that time. Otherwise, they entered the study 6 months after the start of employment at either Shell facility. The residential criterion included all of Louisiana starting in 1988 to reflect the expansion of the LTR catchment area. If an employee at one of these facilities moved to the catchment area after the start of employment, he entered the study on his moving date. All analyses were based on persontime accumulation ending on the earliest of: 1) the date the employee moved out of the LTR catchment area, 2) the date of death, or 3) December 31, 1999. A sensitivity analysis was also performed using the date of diagnosis as the endpoint of person-time accumulation. Because some residential history data points were missing, LTR and Shell staff developed standardized rules to determine two sets of entry and exit dates, reflecting the
RESULTS Norco employees comprised approximately 80% of the 4639 study cohort members. The study cohort was mostly white and male, and the average age at entry into the cohort was 39.4 years for males and 30.1 years for females (Table 1). Few female employees were hired before the 1970s. More than half of female employees were employed for less than 10 years, while more than half of the male employees were employed for 20 or more years. Males contributed 86% of the 56,265 total person-years accumulated in this study. There were 297 cancer cases among the 4639 employees in the study cohort, 91% occurring in males. Twenty-six employees (24 males) were diagnosed with more than one primary cancer. The most common multiple primary combination was bladder and prostate (n ⫽ 5). Overall, there was a significant 16% deficit of cancer cases among males with 271 observed and 322.3 expected
AEP Vol. 14, No. 9 October 2004: 722–730
Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
TABLE 1. Distribution of cohort demographic characteristics by sex* Demographic variables Race (%) White Non-white Average age at entry (years) Year of hire (%) Before 1950 1950–59 1960–69 1970–79 1980–89 1990–99 Duration of employment (%) ⬍ 10 years 10–19 years 20–29 years 30⫹ years
Male (n ⫽ 3920)
Female (n ⫽ 719)
88.1 11.9 39.4
78.7 21.3 30.1
8.5 11.9 11.8 35.5 16.9 15.4
0.6 2.4 4.3 37.6 26.2 29.1
24.9 18.6 34.6 21.9
56.2 22.7 17.4 3.8
*Total percent in each group may not add up to exactly 100% due to rounding.
cancer cases (SIR ⫽ 0.84; 95% CI, 0.74–0.95), compared with south Louisiana males (Table 2). However, a significant excess was observed in males for cancer of the bone and joint, based on 3 observed cases (SIR ⫽ 6.89; 95% CI, 1.42–20.1). The three cases had different histologies, and occurred in different parts of the body. These cases had jobs with potential exposure to various chemicals at Norco, although in different units of the plant. In addition, all were white, had a history of smoking, and were alive at the end of the study. Non-significant excesses were seen in male employees, including multiple myeloma (SIR ⫽ 1.23; 95% CI, 0.34–3.15) and cancers of the soft tissue (SIR ⫽ 1.59; 95% CI, 0.33–4.64), bladder (SIR ⫽ 1.30; 95% CI, 0.85– 1.90), and endocrine system (SIR ⫽ 1.71; 95% CI, 0.47– 4.38). Significant deficits in male employees were seen for lung cancer (SIR ⫽ 0.64; 95% CI, 0.46–0.85), non-Hodgkin’s lymphoma (SIR ⫽ 0.41; 95% CI, 0.13–0.95), and cancer of the oral cavity and pharynx (SIR ⫽ 0.32; 95% CI, 0.09– 0.82). Non-significant deficits were noted for leukemia (SIR ⫽ 0.43; 95% CI, 0.09–1.25), colorectal cancer (SIR ⫽ 0.77; 95% CI, 0.52–1.10), and cancer of the larynx (SIR ⫽ 0.66; 95% CI, 0.21–1.54). There were no cases of mesothelioma, and this was verified by examining death certificate information among the deceased cohort members. The observed number of cancers in men was approximately the same as expected (non-significant and less than a 20% increase or decrease) for melanoma and cancers of the pancreas, kidney, prostate, stomach, and brain. In 719 female employees contributing a total of 7959 person-years, a non-significantly increased SIR of 1.24 (95%
725
CI, 0.81–1.82) was found, based on all-cancer incidence rates in south Louisiana females (26 observed vs. 21 expected). Breast cancer accounts for this excess, with 11 observed and 7.5 expected cases (SIR ⫽ 1.46; 95% CI, 0.73– 2.61). No other sub-category or site-specific cancer contained more than two observed or expected cases. Of the 11 observed cases of primary female breast cancer, one employee accounted for two cases. Therefore, there were 10 female employees diagnosed with breast cancer. Of these, 80% were white and non-Hispanic and 70% were under age 50. Three cases occurred before age 40. Seven of the ten cases worked in an office setting as a secretary, office assistant, or accountant, and therefore were unlikely to be exposed to chemicals in either plant. Job titles for the remaining three were process technician (duration of employment of less than 2 years), laboratory technician (3.5 years of employment), and chemist (more than 20 years of employment). Nine out of the eleven breast cancer cases (82%) were diagnosed with stage 1 localized disease. In the analysis of potentially exposed male workers and a subgroup of those who had 10 or more years of employment, results were similar to those of the entire male cohort (Table 3). Elevated SIRs were seen for the same sites as in the allmale analysis and there were no differences in the strengths of association. There were two bone cancer cases in the 10 or more years of employment group compared with 0.4 expected. Separate analyses performed for males actively employed after 1973 and males hired after 1950 also showed the same of pattern of SIRs as the male cohort as a whole (data not shown). Similar SIRs for all cancer sites were observed for exposed Norco employees, who experienced 234 total cancer cases with 282 expected (SIR ⫽ 0.83; 95% CI, 0.73– 0.94). A sensitivity analysis was performed to examine the effect of basing the end of person-time accumulation on last known address in the LTR catchment area. No differences in risk patterns were seen when the data were analyzed using earliest known address outside the catchment area as the exit date (data not shown). Stage of disease at diagnosis was examined in both the Shell and comparison populations for female breast, prostate, and colorectal cancers (Table 4). For each cancer site, the study population had a higher proportion of localized (early stage) tumors and a generally lower proportion of distant (late stage) or unknown stage tumors than south Louisiana. Eighty-two percent of breast cancers were diagnosed at a localized stage in the study group compared with 55% in the south Louisiana general population. Localized tumors accounted for 74% of prostate cancer cases in the study population compared with 64% in south Louisiana. Additionally, 37% of colorectal cancer cases were diagnosed with localized disease in the study population vs. 29% in the comparison group.
726 Tsai et al.
AEP Vol. 14, No. 9 October 2004: 722–730
CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
TABLE 2. Standardized incidence ratios (SIRs) of major cancer groupings for all male employees compared to south Louisiana males, 1983–1999 (Employees at risk ⫽ 3920, total person-years ⫽ 48,306) Cancer categories (ICD-9 codes)
Obs
Exp
SIR*
95% CI
All cancer combined (140–208) Oral cavity & pharynx (140–148) Digestive system (150–159) Esophagus (150) Stomach (151) Colon & rectum (153–154) Colon (153) Rectum (154) Liver & other biliary tract (155, 156.8–156.9) Pancreas (157) Respiratory system (160–165) Larynx (161) Lung & bronchus (162) Bone & joints (170) Soft tissues (171) Skin (172–173) Melanoma of the skin (172) Male reproductive system (185–187) Prostate (185) Urinary system (188–189) Urinary bladder (188) Kidney & renal pelvis (189) Eye & orbit (190) Brain & nervous system (191–192) Brain (191) Endocrine system (193–194) Thyroid (193) Lymphomas (201–202) Non-Hodgkin’s lymphomas (202) Multiple myeloma (203) Leukemias (204–208) Other, ill-defined, & unknown sites (195–199)
271 4 48 2 6 29 21 8 2 9 54 5 45 3 3 10 8 83 81 37 26 11 2 5 5 4 2 6 5 4 3 5
322.26 12.55 64.05 4.51 6.59 37.73 26.36 11.37 4.57 8.16 80.81 7.60 70.76 0.44 1.89 9.77 6.82 79.27 75.41 31.19 20.04 10.27 0.50 4.41 4.24 2.34 1.99 14.10 12.28 3.25 7.01 9.93
0.84 0.32 0.75 0.91 0.77 0.80 0.70 1.10 0.67 0.66 0.64 6.89 1.59 1.02 1.17 1.05 1.07 1.19 1.30 1.07 1.13 1.18 1.71 0.43 0.41 1.23 0.43 0.50
0.74–0.95 0.09–0.82 0.55–0.99 0.33–1.98 0.52–1.10 0.49–1.22 0.30–1.39 0.51–2.10 0.50–0.87 0.21–1.54 0.46–0.85 1.42–20.1 0.33–4.64 0.49–1.88 0.51–2.31 0.83–1.30 0.85–1.34 0.84–1.64 0.85–1.90 0.54–1.92 0.37–2.65 0.38–2.75 0.47–4.38 0.16–0.93 0.13–0.95 0.34–3.15 0.09–1.25 0.16–1.17
*All SIRs are adjusted for age, race, and calendar time.
DISCUSSION Overall, male employees had significantly fewer incident cancer cases (SIR ⫽ 0.84) than the general south Louisiana population, while female employees had slightly more observed cases of cancer than expected (SIR ⫽ 1.24). The SIR for all-cancers combined among males is similar to those reported in previous studies (5–6, 8, 10–12). In these studies, SIRs for males ranged from 0.78 to 1.04. An exception is the non-significant SIR of 1.7 found in a relatively small cohort of male employees with potential occupational exposure to pesticides in Iowa (9). Few studies have presented all-cancer SIRs for women, mostly because of the limited number of women in petrochemical cohorts. The Finnish refinery study identified 69 cancer cases in female employees, representing a 7% excess (10). However, the study in pesticide-exposed workers in Iowa identified only three female cancer cases, representing no excess (9). Neither SIR was statistically significant.
The significant increase in bone cancer incidence among males was unexpected. No increase in bone cancer deaths was seen in previous mortality studies of this cohort. However, a couple of studies have found significant increases in bone cancer mortality in refinery or petrochemical workers. Wen and colleagues reported a statistically significant increase in bone cancer mortality, but questioned the validity of death certificate coding and classification because bone is a common metastatic site (23). In an update of that study, Satin et al. concluded that primary bone cancer was not related to employment at the facility (24). Exposure to ionizing radiation is the only agent consistently shown to be associated with certain types of bone cancer (25); however, the current study has no information on previous treatment with radiotherapy. There are no known associations of bone cancer with any occupational chemical (25). It is not possible to draw conclusions regarding the etiology of these cases, because of the small number of cases with different histologies, although it seems unlikely that they were caused by occupational exposure.
AEP Vol. 14, No. 9 October 2004: 722–730
Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
727
TABLE 3. Standardized incidence ratios (SIRs) for exposed male employees and exposed male employees who worked ten or more years compared to south Louisiana males, 1983–1999 Ever exposed and worked ⭓ 10 years2
Ever exposed employees1 Cancer categories (ICD-9 codes)
Obs
SIR*
95% CI
Obs
SIR*
95% CI
All cancer combined (140–208) Oral cavity & pharynx (140–148) Digestive system (150–159) Esophagus (150) Stomach (151) Colon & rectum (153–154) Colon (153) Rectum (154) Pancreas (157) Respiratory system (160–165) Larynx (161) Lung & bronchus (162) Bone & joints (170) Soft tissues (171) Skin (172–173) Melanoma of the skin (172) Male reproductive system (185–187) Prostate (185) Urinary system (188–189) Urinary bladder (188) Kidney & renal pelvis (189) Eye & orbit (190) Brain & nervous system (191–192) Brain (191) Endocrine system (193–194) Thyroid (193) Lymphomas (201–202) Non-Hodgkin’s lymphomas (202) Multiple myeloma (203) Leukemias (204–208) Other, ill-defined & unknown sites (195–199)
262 4 45 2 5 28 20 8 8 53 5 44 3 3 10 8 81 79 34 23 11 2 5 5 4 2 6 5 4 3 5
0.84 0.33 0.73 – 0.78 0.77 0.78 0.73 1.01 0.68 0.68 0.65 7.10 1.64 1.06 1.21 1.06 1.08 1.13 1.19 1.11 – 1.17 1.22 1.77 – 0.44 0.42 1.27 0.44 0.52
0.74–0.95 0.09–0.85 0.53–0.97 – 0.25–1.83 0.51–1.11 0.48–1.21 0.32–1.44 0.44–2.00 0.51–0.89 0.22–1.59 0.47–0.87 1.47–20.8 0.34–4.79 0.51–1.94 0.52–2.39 0.84–1.31 0.86–1.35 0.78–1.58 0.75–1.78 0.55–1.99 – 0.38–2.74 0.40–2.85 0.48–4.52 – 0.16–0.96 0.14–0.98 0.35–3.25 0.09–1.29 0.17–1.21
248 4 43 2 5 26 18 8 8 48 4 40 2 3 9 8 77 75 33 23 10 2 5 5 4 2 6 5 4 3 5
0.83 0.35 0.72 – 0.81 0.73 0.72 0.75 1.04 0.63 0.57 0.60 – 1.80 1.06 1.31 1.03 1.04 1.13 1.21 1.06 – 1.27 1.32 1.96 – 0.48 0.45 1.32 0.46 0.54
0.73–0.93 0.10–0.89 0.52–0.97 – 0.26–1.89 0.48–1.07 0.43–1.14 0.33–1.49 0.45–2.05 0.47–0.84 0.15–1.45 0.43–0.82 – 0.37–5.27 0.48–2.01 0.57–2.59 0.81–1.29 0.82–1.31 0.78–1.58 0.77–1.82 0.51–1.94 – 0.41–2.97 0.43–3.07 0.54–5.02 – 0.18–1.04 0.15–1.05 0.36–3.37 0.10–1.35 0.18–1.26
Employees at risk ⫽ 3807, total person-years ⫽ 46,956. Employees at risk ⫽ 2867, total person-years ⫽ 39,722. *All SIRs are adjusted for age, race, and calendar time. 1 2
A non-significant increase in female breast cancer was also observed in the study population. The 46% increase is similar to the 50% increase seen in the Finnish refinery study (10), but higher than that seen in Illinois petrochemical research workers (8). Unfortunately, information was not available on established risk factors such as parity, age at menarche, breast-feeding, age at menopause, alcohol consumption, and body mass index (26), making it difficult to speculate on the etiology of these cases. Seventy percent of
breast cancer cases in the study population worked in administrative support. Some studies have found slight breast cancer increases among female administrative staff (27–29). However, these studies were conducted without occupational exposure information and usually did not control for reproductive or socioeconomic confounders. Although female employees had an increased incidence of breast cancer, they were also more likely to be diagnosed with localized tumors compared with the south Louisiana
TABLE 4. Stage distribution (in percent) for selected cancer sites diagnosed in south Louisiana and the study cohort, 1983–1999* Breast cancer SEER summary stage Localized Regional Distant Unknown
Prostate cancer
Colorectal cancer
South Louisiana
Study cohort
South Louisiana
Study cohort
South Louisiana
Study cohort
55.4 34.0 6.5 4.2
81.8 – 9.1 9.1
64.4 13.9 11.1 10.7
74.1 17.3 3.7 4.9
29.0 38.2 24.3 8.6
36.7 46.7 16.7 –
*Total percent in each group may not total 100% because of rounding.
728 Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
population (82% vs. 55%). This finding suggests that female employees were diagnosed earlier than their south Louisiana counterparts, perhaps because of employee insurance that pays for an annual screening mammogram. Male employees had a slightly higher incidence of prostate cancer compared with south Louisiana males (SIR ⫽ 1.07). However, a larger proportion of male employees were diagnosed with localized disease than the comparison population (74% vs. 64%). The slightly higher incidence of prostate cancer may be due to increased use of the prostate-specific antigen (PSA) screening test in the study population. Incidence of colorectal cancer was lower in the study cohort compared with the south Louisiana population (SIR ⫽ 0.77), and a higher proportion of the study population was diagnosed with localized tumors (37% vs. 29%). Certain methods of colorectal cancer screening (colonoscopy and sigmoidoscopy) allow for immediate removal of pre-cancerous lesions (e.g., adenomatous polyps) and are covered by employer-based health insurance. It is possible that more employees underwent colonoscopy and sigmoidoscopy tests compared with south Louisiana residents, resulting in a deficit of colorectal cancer in the study population. Another finding of interest was the significant deficit in lung cancer incidence for male employees (SIR ⫽ 0.64). This was not unexpected, as a similar finding (SMR ⫽ 0.70; 95% CI, 0.51–0.93) was seen in a previous mortality study at Norco (1). Lung cancer incidence rates among white males in Louisiana and south Louisiana have been significantly higher than national rates, but rates in the Industrial Corridor along the lower Mississippi River have been lower than either south Louisiana or Louisiana as a whole (30, 31). Completeness of case ascertainment is important in cohort studies examining cancer incidence. The North American Association of Central Cancer Registries, a professional organization that evaluates and certifies state cancer registries in North America, estimated LTR case ascertainment at 95% or higher for the period between 1995 and 2000 (32). This estimate indicates that some cases may be missed by LTR, especially Louisiana residents who are diagnosed and treated exclusively in a physician’s office or at out-of-state facilities. To improve the capture of these cases, LTR routinely screens data from pathology laboratories and death certificates. At the end of every year, Louisiana death certificate files listing cancer as a cause of death are matched against the LTR master file. All unmatched cancer deaths are followed back to determine the date and residency at diagnosis. Eligible cases are then included in the LTR data file. To address the issue of LTR case ascertainment, we compared cancer deaths from the National Death Index with
AEP Vol. 14, No. 9 October 2004: 722–730
cancer cases from LTR and identified 13 cancer deaths in the cohort that were not in the registry. A closer examination showed that five cases died in 1983, but were diagnosed prior to 1983, before the start of the study; three cases died shortly after 1983, but were probably diagnosed prior to 1983 since they were cancers with generally good prognosis (bladder, prostate, lymphoma); two cases were missed (in regional databases but not in the central database); and three cases may have been missed (one melanoma case who died in 1996, one lung cancer case who died in 1985, and one case at an unspecified site who died in 1998). These three undetermined cases all died in Louisiana, were not in the LTR database, and no other information could be found on them. The five cases, had they been ascertained by LRT, would not have changed this study’s conclusions. This study has several strengths. Utilizing incidence rather than mortality data minimized survival bias. In addition, linking to LTR resulted in the highest possible ascertainment of cancer cases while minimizing diagnostic misclassification. This study, along with others, demonstrates the feasibility of conducting a collaborative study between industry and a state cancer registry. Migration out of Louisiana is not an issue for this study. 14.7 percent of the study cohort left Louisiana over the 17-year period from 1983 to 1999, while 9.2% of working-age (20–64 years old) residents out-migrated from Louisiana between 1995 and 2000, according to the 2000 US census (33). Assuming that Louisiana residents migrated out of their state at similar rates over the entire period from 1983 to 1999 and that there were no differences by region of the state, our cohort experienced less out-migration than the comparison population, supporting our assumption that cohort members did not leave the LTR catchment area at higher rates for healthrelated reasons. And although residential history data was less than 100% complete, this study’s methods are an improvement over using current residence to determine eligibility with less than 1% loss to follow-up. Despite this study’s numerous strengths, there are some limitations to consider. No detailed information on occupational exposure or lifestyle was available. Without detailed exposure information, it was not possible to link specific chemical agents to any increased risk in cancer. Data on potential confounders, such as socioeconomic status and reproductive history, were also not available and therefore were not adjusted for. In conclusion, this study found little evidence of any association between employee cancer risk and working at these two refining and petrochemical facilities. Possible causes for the increase in bone cancer are not clear, but it is unlikely that occupational exposures played a large role. Early detection of breast cancer among female employees may have contributed to the non-significant increase of the
AEP Vol. 14, No. 9 October 2004: 722–730
Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
disease although other important, but unmeasured confounders, such as socioeconomic status or certain reproductive factors may also play an important role. The authors thank Dr. Elizabeth Delzell and Dr. Sharon Cooper, who served as the external advisors for this study.
APPENDIX PROCEDURES USED TO PROCESS RESIDENTIAL HISTORY DATA 1. If the earliest address and the latest address were located in south Louisiana for the time period 1983 to 1987 and in Louisiana for the time period 1988 to 1999, it was assumed that the subject remained in the LTR catchment area throughout the study time period. 2. If the only one address was found and indicated residence in the LTR catchment area, it was assumed that the subject lived in the LTR catchment area throughout the study period. 3. If a gap in residential history is present and residence inside the LTR catchment area changes during this gap, the midpoint of the range of dates will be used to count person-years. 4. If records from multiple sources did not agree for a given year, the record that confirmed a previous address from a different source was kept and the nonconfirming record was discarded. 5. Address records with effective dates after death were discarded. 6. Address data extracted from work history was supplemented when possible. 7. If the all addresses found during the study period from all residence history and/or work history sources indicated residence outside the LTR catchment area, the subject was excluded.
729
5. Huebner WW, Chen VW, Friedlander BR, Wu XC, Jorgensen G, Bhojani FA, et al. Incidence of lymphohaematopoietic malignancies in a petrochemical industry cohort: 1983–94 follow up. Occup Environ Med. 2000; 57(9):605–614. 6. Bond GG, Austin DF, Gondek MR, Chiang M, Cook RR. Use of a population-based tumor registry to estimate cancer incidence among a cohort of chemical workers. J Occup Med. 1988;30(5):443–448. 7. Olsen GW, Gondek MR, Flannery J, Cartmill JB, Bodner KM. A thirtyeight year comparison of cancer incidence and mortality among employees at a Connecticut chemical plant. Conn Med. 1997;61(2):83–89. 8. Sathiakumar N, Delzell E, Rodu B, Beall C, Myers S. Cancer incidence among employees at a petrochemical research facility. J Occup Environ Med. 2001;43(2):166–174. 9. Leet T, Acquavella J, Lynch C, Anne M, Weiss NS, Vaughan T, et al. Cancer incidence among alachlor manufacturing workers. Am J Ind Med. 1996;30(3):300–306. 10. Pukkala E. Cancer incidence among Finnish oil refinery workers, 1971– 1994. J Occup Environ Med. 1998;40(8):675–679. 11. Christie D, Robinson K, Gordon I, Bisby J. A prospective study in the Australian petroleum industry. II. Incidence of Cancer. Br J Ind Med. 1991;48(8):511–514. 12. Gun R, Pilotto L, Ryan P, Griffith E, Pratt N, Ewing S, et al. Health Watch Eleventh Report. South Australia: The University of Adelaide Department of Public Health; 2000. 13. Tsai SP, Wendt JK, Cardarelli K, Fraser AE. A mortality and morbidity study of Geismar chemical plant employees. Unpublished internal report; April 8, 2002. 14. Correa P, Fontham ETH, Chen V, Craig JF, Falk R, Pickle LW. Diet, nutrition, and cancer. J Louisiana State Med Soc. 1988;140:43–49. 15. Chen VW, Correa CN, Wu XC, Andrews PA, Fontham ETH, Blount GA, et al. Cancer Incidence in Louisiana, 1988–1992. Cancer in Louisiana. v. 8(1). New Orleans: Louisiana Tumor Registry; 1996. 16. World Health Organization. International Classification of Diseases for Oncology (ICD-O-1). 1st ed. Geneva, Switzerland: World Health Organization; 1976. 17. Percy C, Van Holton V, Muir C. International Classification of Diseases for Oncology (ICD-O-2). 2nd ed. Geneva, Switzerland: World Health Organization; 1990. 18. Fritz A, Ries L, eds. The SEER Program Code Manual. 3rd ed. Bethesda, MD: US National Cancer Institute; 1998. 19. Johnson CH, ed. Standards for Cancer Registries. Volume II: Data Standards and Data Dictionary Version 8. 4th ed. Sacramento, CA: North American Association of Central Cancer Registries; 1999. 20. Bailar JC, Ederer F. Significance factors for the ratio of a Poisson variable to its expectation. Biometrics. 1964;20:304–312. 21. Marsh G, Preininger ME. A user-oriented occupational cohort mortality analysis program. Am Stat. 1980;27:220–222. 22. SAS Institute Inc. SAS Procedure Guide; Release 8.12. Cary, NC: SAS Institute, Inc; 1999.
REFERENCES 1. Tsai SP, Wendt JK, Cardarelli KM, Fraser AE. A mortality and morbidity study of refinery and petrochemical employees in Louisiana. Occup Environ Med. 2003;60:627–633.
23. Wen CP, Tsai SP, McClellan WA, Gibson RL. Long-term mortality study of oil refinery workers. I. Mortality of hourly and salaried workers. Am J Epidemiol. 1983;118(4):526–542.
2. Tsai SP, Wendt JK. Health findings from a mortality and morbidity surveillance of refinery employees. Ann Epidemiol. 2001;11:466–476.
24. Satin KP, Wong O, Yuan LA, Bailey WJ, Newton KL, Wen CP, et al. A 50-year mortality follow-up of a large cohort of oil refinery workers in Texas. J Occup Environ Med. 1996;38(5):492–506.
3. Tsai SP, Gilstrap EL, Colangelo TA, Menard AK, Ross CE. A mortality study of oil refinery and petrochemical employees. J Occup Environ Med. 1997;39(5):448–454.
25. Miller RW, Boice JD, Curtis RE. Bone cancer. In: Schottenfeld D and Fraumeni JF, eds. Cancer Epidemiology and Prevention. 2nd ed. New York: Oxford University Press; 1996:971–983.
4. Honda Y, Delzell E, Cole P. An updated study of mortality among workers at a petroleum manufacturing plant. J Occup Environ Med. 1995;37(2): 194–200.
26. Hankinson S, Hunter D. Breast cancer. In: Adami H-O, Hunter D, Trichopoulos D, eds. Textbook of Cancer Epidemiology. Oxford: Oxford University Press; 2002:301–339.
730 Tsai et al. CANCER INCIDENCE AMONG REFINERY AND PETROCHEMICAL WORKERS
27. Calle EE, Murphy TK, Rodriguez C, Thun MJ, Heath CW Jr. Occupation and breast cancer mortality in a prospective cohort of US women. Am J Epidemiol. 1998;148(2):191–197. 28. Coogan PF, Clapp RW, Newcomb PA, Mittendorf R, Bogdan G, Baron JA, et al. Variation in female breast cancer risk by occupation. Am J Ind Med. 1996;30(4):430–437. 29. Morton WE. Major differences in breast cancer risks among occupations. J Occup Environ Med. 1995;37(3):328–335. 30. Groves FD, Andrews PA, Chen VW, Fontham ETH, Correa P. Is there a ‘cancer corridor’ in Louisiana? J Louisiana State Med Soc. 1996;148: 155–165.
AEP Vol. 14, No. 9 October 2004: 722–730
31. Wu XC, Correa CN, Andrews PA, Hsieh MC, Schmidt BA, Ahmed MN, et al. Cancer Incidence and Mortality in Louisiana, 1993–1997. New Orleans: Louisiana Tumor Registry; 2001. 32. North American Association of Central Cancer Registries. Registry Certification. Available at: http://www.naaccr.org/index.asp?Col_SectionKey= 12&Col_ContentID⫽54. Accessed September 30, 2003. 33. US Census Bureau. Gross and Net Migration by Age for the Population 5 Years and Over for the United States, Regions, and States: 2000. Available at: http://www.census.gov/population/cen2000/phc-t23/tab03.xls. Accessed December 23, 2003.
