Research Time Trends in Sport-Caught Great Lakes Fish Consumption and Serum Polychlorinated Biphenyl Levels among Michigan Anglers, 1973-1993 J I A N - P I N G H E , * ,† A R Y E H D . S T E I N , †,‡ HAROLD E. B. HUMPHREY,§ NIGEL PANETH,† AND J E A N N E M . C O U R V A L †,| Department of Epidemiology and Institute for Environmental Toxicology, Michigan State University, East Lansing, Michigan 48824, and Department of International Health and Department of Epidemiology, Rollins School of Public Health of Emory University, Atlanta Georgia 30322
Polychlorinated biphenyls (PCBs) remain public health concerns because of their persistence in the environment and their potential health impact. We linked data from three mixed cross-sectional/longitudinal surveys of Michigan anglers conducted by the Michigan Department of Public Health in 1973-1974, 1979-1982, and 1989-1993 to examine the association between sport-caught fish consumption and serum PCBs. The reported weight of fish consumed declined over the three surveys (median 40, 38, and 31 lb/yr). Serum Aroclor 1260 levels were 2-3 times higher in fish-eaters than in nonfish-eaters in all three surveys in both men and women. In nonfish-eaters, serum PCB levels rose between 1973-1974 and 19791982 [adjusted change ) 0.30 log(ppb), p ) 0.01] and then declined between 1979-1982 and 1989-1993 [adjusted change ) -0.16 log(ppb), p ) 0.002]. Among fish-eaters, serum PCB levels also rose between 1973-1974 and 19791982 [adjusted change ) 0.45 log(ppb), p < 0.001] but were unchanged between 1979-1982 and 19891993 [adjusted change ) -0.09 log(ppb), p ) 0.14]. Predictors of serum PCB levels included annual fish consumption, gender, and age. We conclude that background human serum levels of Aroclor 1260 had declined by 19891993 from earlier peak levels. Among consumers of sportcaught Great Lake fish, serum PCB levels did not significantly decrease, probably due to continued exposure and the long half-life of PCB.
Background Polychlorinated biphenyls (PCBs) are a mixture of chlorinated biphenyl congeners with the empirical formula C12H10-nCln. * Corresponding author phone: (517)353-8623; fax: (517)432-1130; e-mail:
[email protected]. † Department of Epidemiology, Michigan State University. ‡ Department of International Health, Rollins School of Public Health of Emory University. § Institute for Environmental Toxicology, Michigan State University. | Department of Epidemiology, Rollins School of Public Health of Emory University. 10.1021/es001067p CCC: $20.00 Published on Web 12/15/2000
2001 American Chemical Society
PCB exposure has been linked to reproductive dysfunction (1-3), immune suppression (4, 5), tumor formation (6, 7), and behavioral and learning problems (8, 9). As PCBs bioaccumulate in Great Lakes fish, consumption of sportcaught Great Lakes fish is an important human route of exposure (10-19). Most studies to date have been cross-sectional (11, 16, 17, 20); longitudinal associations have not been extensively examined. In this paper, we describe 20-yr trends in sport fish consumption and serum PCB levels in Michigan anglers.
Methods Population and Settings. The first survey of human exposure to PCBs from consumption of sport-caught fish in Lake Michigan anglers was performed in four Michigan lakeshore counties in 1973-1974 (12). Anglers were recruited from boat launch sites, piers, bait shops, sporting goods shops, and fishing clubs. To recruit avid sport anglers, a “snowball approach” was used. Interviewees were asked to give leads to other persons who might also be interested in participating in the study, including family members. Participants were selected for study if preliminary interviews disclosed that their recreational sport-caught Great Lakes fish consumption fell into one of two categories: “high” consumption (24 lb or more of sport fish consumed in the past 12 months; hereafter designated as fish-eaters) or “low” consumption (less than 6 lb in the past 12 months, including no consumption; hereafter designated as nonfish-eaters). A household survey was also conducted in each community to identify nonfish-eater controls. The strategy of selecting extremes of fish consumption was designed to amplify the association between fish consumption and human serum PCB burden, should an association exist. A total of 156 individuals (including 73 fish-eaters) were recruited. In 1979-1982, a second large multisite PCB survey was conducted (12, 13). Eleven target communities in Western Michigan, along the shoreline of Lake Michigan, were grouped into three regions: northern (Petoskey, Traverse City, Manistee, and Ludington); middle (Whitehall, Grand Haven, White Lake, and Muskegon); and southern (Holland, South Haven, and St. Joseph). All participants in the first survey were invited to participate. The second survey used the same approach to recruiting additional fish-eaters as had been adopted for the first survey. A total of 610 fish-eaters participated in the survey of whom 55 had also participated in the 1973-1974 survey. Random digit dialing was used in each region to recruit a “nonfish-eating” comparison group for the survey: 419 agreed to participate. The second survey included 115 individuals from the first survey and added 1140 new participants, for a total sample size of 1255. Participants from the second survey were recontacted in 1989-1993 and invited to participate in a followup survey. Nearly 57% (717/1255) agreed to participate (12). A few (n ) 11) additional individuals were recruited at this time. At each survey, participants were encouraged to provide a nonfasting blood sample for analysis of serum levels of PCB and other contaminants. The numbers of participants with information on both fish consumption and serum PCB levels in the three surveys were 152, 1091, and 638, respectively. Ascertainment of Fish Consumption. The questionnaires used in the surveys consisted of a fish consumption module and a general information module. In the first survey, the fish consumption module contained questions about sportVOL. 35, NO. 3, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
435
caught Great Lakes fish consumption behaviors, including the number of years the respondent ate sport-caught Great Lakes fish, number of fish meals eaten in the past year, and usual portion serving size, based on a photograph of an 8-oz fish dinner. About two-thirds of participants (103/156) also provided detailed information on the most frequently consumed species of fish (e.g., trout, salmon, perch, catfish, whitefish) and the most frequent cooking (e.g., pan fried, deep fried, broiled, baked, etc.), preparation (trimming away dorsal or belly fat, removing skin prior to cooking), and preservation (freezing, canning) methods. In the second survey, separate questionnaires were developed for fisheaters and nonfish-eaters. For fish-eaters, the questionnaire included both demographic and fish consumption modules, whereas for nonfish-eaters, the questionnaire included only the demographic module. In the third survey, all participants were administered the complete interview. Measurement of Serum PCB Levels. All serum samples were analyzed at the Health Risk Assessment Laboratory of the Michigan Department of Public Health in Lansing, MI. Serum PCB levels were determined using a modification of the Association of Official Analytical Chemists approved Webb-McCall packed-column gas chromatography technique (21). The laboratory analytical methodology used was reported earlier (19) and has been described in detail (22, 23). The procedure used a methanol ether/hexane extraction, microflorisil column cleanup, and silica gel column separation technique prior to injection into a programmed electron capture gas chromatograph. Congener-specific PCB analysis was not available during the study period. In the 1972-1973 and the 1979-1982 surveys, quantitation was based on Aroclor 1254 and 1260 standards. In the 1989-1993 survey, Aroclor 1016 and 1260 standards were used. As Aroclor 1260 was the only standard that was measured consistently across the three surveys, serum levels of Aroclor 1260 were considered the outcome (dependent) variable in this analysis. Evaluations based on the Aroclor 1260 standard are not statistically different from those that use the less chlorinated 1254 or 1016 standards for in-vivo blood evaluations (24). This was recently confirmed by congener-specific quantitation of blood from a subgroup of this cohort where the predominant congeners detected were those with chlorine content (penta-octa) that is captured by the 1260 standard in packed-column work (25). Laboratory values reported as less than the detectable limit for Aroclor 1260 (3 ppb) were coded as 1.5 ppb. Approximately 1.9%, 5.1%, and 5.3% of samples were below this threshold in the three surveys, respectively. Estimation of Exposure to Sport-Caught Great Lakes Fish. Individuals were categorized as fish-eaters or nonfisheaters at baseline, based on the reported total weight of sportcaught Great Lakes fish consumed during the 12 months preceding their first enrollment into the study. Consumption of sport-caught Great Lakes fish at each survey was considered the primary exposure variable. Current exposure was expressed by two continuous variables: annual sport-caught Great Lake fish meals and total weight of fish consumed in previous 12 months. We computed annual sport-caught Great Lakes fish meals by summing the reported number of fish meals consumed in each of the preceding 12 months. We then estimated the total weight of fish consumed by multiplying the number of meals by the usual portion serving size. This approach assumes that the portion serving size was consistent across the preceding 12 months. The total weight of fish consumption was coded as zero in participants with missing fish consumption; most of these were nonfisheaters (based on the screening instrument administered at recruitment) for whom the fish consumption module was not administered in 1979-1982. 436
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 3, 2001
TABLE 1. General Characteristics of 1255 Michigan Anglers Interviewed in 1979-1982, by Fish-Eating Statusa fish-eaters no. of participants 608 gender: (%) male 66.3 female 33.7 age in years (mean, SD) 46.0, 13.6 age: (%) 15-24 3.6 25-34 20.0 35-44 26.3 45-54 21.2 55-64 17.9 65+ 11.0 education: (%) elementary school 4.6 some high school 12.4 high school graduate 42.0 some college 23.3 college graduate 17.7 region of residence: (%) south 39.5 middle 31.6 north 28.9
nonfish-eaters 645 42.9 57.1 43.1, 15.0
p
