Jump to main content.

Report to Congress

Two fisherman fishing in Lake Superior

A cooperative project by:

Agency for Toxic Substances and Disease Registry (ASTDR) logo

Other Related Topics

exit EPA (About PDF)

back to index

The Effects of Great Lakes Contaminants on Human Health

photo: an eagle
An Eagle

I. Introduction

Historical Overview

The Great Lakes hold one-fifth of the fresh water on the earth's surface, and it is a valuable resource for both the United States and Canada. Approximately 25 percent of the Canadian population and 10 percent of the United States population live within the Great Lakes watershed. For more than 200 years, the Great Lakes basin has been used as a resource for industry, agriculture, shipping and recreation. By the early 1960's, the environmental quality of the Great Lakes had deteriorated significantly as a result of eutrophication, overfishing, and the widespread presence of toxic substances. During the 1970's, the use of the Great Lakes as a disposal site for agricultural, industrial and domestic wastes became an increasingly widespread concern due to detrimental effects on fish and wildlife, and the potentially adverse effects on human health.

Some pollutants of the Great Lakes have been found in significant concentrations in fish and wildlife (Keith 1966; Aulerich and Ringer 1977; Nisbet 1989). Wildlife investigators have demonstrated that these pollutants are implicated in interfering with reproduction in birds and lake trout, contribute to congenital abnormalities in birds, and have implicated these pollutants in the development of neoplastic tumors in bottom-dwelling fish (Gilbertson et al. ,1976; Maccubbin et al., 1985; Price and Weseloh 1986). Concentrations of PCBs in some Great Lakes fish have been significant enough at least since the mid-1970's to warrant advisories to restrict consumption of these fish, particularly by children and women of childbearing age.

Except for some migratory waterfowl, the fish and wildlife in the Great Lakes basin generally live their entire lives within the basin, being continually exposed to toxic contaminants in the water and/or food. A combination of factors contribute to the Great Lakes acting as a reservoir for these relatively nondegradable lipophilic chemicals. Factors such as the following: 1) long retention times, the time it takes for the lake water to be completely exchanged ranges from about 3 years for Lake Erie to 173 years for Lake Superior; 2) internal recycling of contaminants; pollutants tend to adsorb onto particles and eventually settle to the bottom of the lakes. However, the surficial sediments are subject to resuspension by storm actions and other natural processes, and the contaminants are dispersed throughout the water column again.

The concentrations of contaminants in the water column of the Great Lakes declined measurably after significant decreases in the loadings of the pollutants. Sediment-water interactions, however, greatly extend the overall response time of the system to loading reductions. For example, data from mass balance studies and inventories in the Great Lakes system indicate that there is a reservoir of PCBs in the sediments that will continue to release PCBs into the lakes at significant rates for decades to come (Andre et al., 1993; Beltran, 1992; Richardson, 1993; U.S. EPA 1989).

Fish tissue concentrations of some contaminants, including DDT and PCBs, have declined significantly in the Great Lakes basin since the mid-1970s. At that time, PCB concentrations in lake trout from Lake Michigan were about 22 mg/kg. By the early 1980's they had declined to about 6 mg/kg, a level not greatly more than found now. Recently, the rate of decline of PCB concentrations in these fish has slowed, and smaller incremental reductions in the fish tissue levels will continue to occur.

As stated previously, fish tissue contaminant levels are lower than in the 1970's, which is when some of the early human health studies were done. However more recent human studies that have taken advantage of newer robust tools (e.g., improvements in analytical capabilities and clinical testing sophistication) continue to demonstrate the potential for subtle effects in exposed at risk human populations.

In 1972, the United States and Canada signed the Great Lakes Water Quality Agreement to initiate remedial actions to improve the quality of Great Lakes waters, focusing on the issue of excessive loadings of nutrients. In 1978, the two countries signed a revised Agreement to "restore and maintain the chemical, physical and biological integrity of the waters of the Great Lakes ecosystems" (International Joint Commission 1978), with an emphasis on reducing toxic contaminants. The Agreement was amended again in 1987 expanding many of the programs and recognizing that an ecosystem approach was required to truly restore and protect the Great Lakes ecosystems.

Persistent Toxic Substances

The presence of toxic substances in the Great Lakes continues to be a significant concern in the 1990s. In the U.S., some 70,000 commercial and industrial compounds are now in use. More than 30,000 chemicals are produced or used in the Great Lakes region. There have been 362 contaminants identified in the Great Lakes system; of these 362 approximately one-third have been evaluated for their potential toxic effects on aquatic life, wildlife, and human health. (An Inventory of Chemical Substances Identified in the Great Lakes Ecosystem ,Vol. 1-6. IJC, Windsor, Ontario 1983 and Federal Register March 23, 1995 Page 15367) in 1985, eleven of the most persistent and widespread toxic substances were identified as "critical Great Lakes pollutants" by the International Joint Commission (IJC). The critical pollutants identified by the IJC are:

polychlorinated biphenyls (PCBs)
dichlorodiphenyl trichloroethane
(DDT) and metabolites
(a member of a class of
substances known as PAHs)
hexachlorobenzene mirex
alkylated leadfurans

All eleven of the persistent toxic substances tend to bioaccumulate in organisms, biomagnify in food webs, and persist at high levels in some areas of the ecosystem of the Great Lakes. In the Great Lakes basin, 43 geographical locations in the United States and Canada have been identified as "Areas of Concern" because of high concentrations of these toxic pollutants. Of these 43 locations, 26 are located within the boundaries of the United States and 5 are shared with Canada, (i.e., connecting channels.

Eight of these pollutants are organochlorine compounds and are potentially harmful because of their persistence, ubiquity, and associated toxicity. Both alkylated lead and methylmercury are heavy metals that are potentially harmful because of their chemical characteristics and demonstrated toxicity. Eight of the eleven IJC critical pollutants are lipophilic and biomagnify within the aquatic food chain, thereby making them available to higher forms of life, including humans. PAHs, methylmercury, and alkylated lead also bioaccumulate in human tissues.

Pathways of Exposure

Potential environmental pathways of human exposure to Great Lakes pollutants include inhalation of air, ingestion of water, foodstuffs or contaminated soil, and dermal contact with water or airborne particulates. Multimedia analyses indicate that the majority (80-90%) of human exposure to chlorinated organic compounds comes from the food pathway, a lesser amount (5-10%) from air, and minute amounts (less than 1%) from water (Birmingham et al. 1989; Newhook 1988).

Most of the data available on human exposure to toxic substances in the Great Lakes come from the analyses of contaminant levels in drinking water and sport fish. The consumption of contaminated sport fish and wildlife can significantly increase human exposure to the Great Lakes critical pollutants. A spectrum of these major contaminants has been identified in cooked Great Lakes fish. Investigators have demonstrated that blood serum levels of these contaminants are significantly increased in consumers of contaminated Great Lakes sport fish as compared to nonfisheaters (Humphrey 1983a,b; Jacobson et al. 1989).

Even though residents of the Great Lakes basin are exposed to toxic substances from many sources originating within and outside the region, the main routes of human exposure to contaminants from the waters of the Great Lakes are ingestion of fish and to a lesser extent ingestion of drinking water (National Health and Welfare Canada 1991). Also, several investigators have shown that exposure from fish far outweighs atmospheric, terrestrial, or water column sources (Swain 1983; Humphrey 1983b). These patterns may vary for populations living in the vicinity of industrialized areas such as refineries or smelting plants.

Several epidemiologic investigations have been conducted to investigate the association between water pollutants in the Great Lakes and the health of people in the Great Lakes states. These studies have demonstrated increased tissue levels of toxic substances (body burdens) in these populations that may be associated with or potentially result in reproductive, developmental, behavioral, neurologic, endocrinologic, and immunologic effects. The following review of human health literature in the Great Lakes basin summarizes and compares the findings concerning exposure to Great Lakes pollutants.

II. A Review of the Great Lakes Human Health Literature

Several human health studies have been conducted in the Great Lakes basin. The following review summarizes some of these studies and other studies concerning exposure to Great Lakes pollutants.

Michigan Sport Fishermen Cohort Study

photo: walleye fish
walleye fish

The Michigan Sports Fishermen Cohort study was initiated in 1974 to assess the potential exposure of humans to PCBs from consumption of sport fish from Lake Michigan (Humphrey 1976). The population consisted of 336 adults from 18 western counties of Michigan that border Lake Michigan. These adults made up two groups: those who consumed large quantities of Lake Michigan sport fish, especially trout and salmon, and those who did not eat fish from Lake Michigan. Blood samples were taken from each group and the serum was analyzed for PCB compounds. The resulting data were compared to information received from each individual through interviews. Fish consumption variables were compared with PCB levels. The results from this initial study demonstrated a statistically significant correlation between fish consumption patterns and PCB levels in this population.

In the late 1970s and early 1980s, 123 members of this original cohort were re-examined using the same protocol. On average, sport fish-eaters consumed 32 pounds of fish a year, and some ate as much as 262 pounds a year. This average was approximately five times the national per capita fish consumption rate commonly used in risk estimates at that time. Individuals who regularly ate 24 pounds a year or more of Great Lakes fish, especially trout and salmon, had higher (P < 0.001) serum concentration levels of PCBs than individuals who seldom or never ate such fish.

A third cohort was included in these studies to test the above conclusion. In 1982, 572 individuals from the Lake Michigan shore who ate 24 or more pounds of fish a year were compared with 419 randomly selected residents from the same communities along the Lake Michigan shoreline. Again, the data indicated that individuals in the upper range of fish consumption had serum PCBs concentrations approximately four times greater than those of unexposed individuals. These studies support a positive correlation between human uptake of toxic pollutants in the Great Lakes and Great Lakes fish consumption (Humphrey 1976, 1983a, 1988a, 1988b, 1989).

Wisconsin Sports Fish-Consumers Study

photo: round whitefish
round whitefish

Another study of sport anglers conducted in Wisconsin (Wisconsin Sports Fish-Consumers Cohort Study) was similar in design to the Michigan Sports Fishermen Cohort Study. Investigators in Wisconsin surveyed 801 sport anglers for their fishing and consumption habits, and for comprehension of and compliance with the Wisconsin fish consumption health advisory (Fiore et al. 1989; Sonzogni et al. 1991). In this cross-sectional study of Wisconsin anglers, the investigators also measured the angler's body burden levels of PCBs and DDE to examine possible correlations between sport fish consumption and the body burden of these substances. Of the 801 persons surveyed, 198 individuals provided blood samples for analysis of PCBs and DDE.

The survey results indicated that the mean number of sport fish meals eaten annually was 18, 72 percent of the anglers were aware of the fish advisory, and 57 percent changed their fishing or fishing habits as a result of the advisory. Of the 198 blood samples drawn for PCBs and DDE analysis, 6 were determined to be inadequate for analysis. The mean PCB serum level for the 192 anglers was 2.2 ug/l; of the 13 PCB congeners for which standards were available, only 2 were not detected in sera. The investigators also determined that the congeners most frequently identified in human sera were also the most abundant congeners in the tissues of a variety of Wisconsin fish (Maack and Sonzogni 1988). Additionally, there was a statistically significant Spearman correlation (R = .21, p = .003) between the PCB congener sum and the total number of sport fish meals. The mean DDE level was 6.3 ug/l. Of the 109 samples tested, 65 of the subjects had a serum level between 5.0 and 10.0 ug/l, and 44 had levels from 10.0 to 40.0 ug/l. There was also a statistically significant positive correlation (R =0.14, p =0.04) between sera levels and the total number of sport fish meals.

The Wisconsin Sports Fish-Consumers Cohort is the first study of anglers for whom PCB-specific congeners were determined in human sera. Using capillary column gas chromatography as opposed to the older, packed column technique, these researchers were able to detect PCB congeners 153, 138, 180, and 118 (order based on frequency of occurrence). The authors also suggest that the capillary method produces PCB results that are approximately 30 percent more sensitive than those obtained by packed column chromatography, allowing for the summation of individual congeners to estimate total PCB burden. The differences in laboratory techniques between the Wisconsin and Michigan studies (used packed column chromatography) makes the comparison of results very difficult.

Minnesota Ecologic Epidemiologic Study

In Minnesota, a researcher conducted an ecologic epidemiologic study to examine fetal and neonatal death rates, as well as congenital malformations rates of residents in counties that were in close proximity to the lakes (Schuman et al. 1982). An ecologic study is a descriptive study in which measures that represent characteristics of entire populations are used to describe disease in relation to some factors of interest. In ecologic studies the unit of observation is a group of people rather than individuals and the unit of exposure pertains to geographical areas rather than to individuals. The assumption in this study was that residents in lake- bordering counties were more likely to have environmental and occupational exposure to pollutants discharged into the air and water. Information on the above human end points was collected for 1950-1975 and 1977 for six of the eight Great Lakes States (excluding New York and Pennsylvania) and analyzed. The investigator found no significant differences among fetal death rates, neonatal death rates, and congenital malformations in populations bordering the lakes; however, results are considered inconclusive because the rates were not adjusted for confounding variables, such maternal age, race, or sex.

Michigan Maternal and Infant Study

In Michigan, the Maternal/Infant Cohort Study was designed to assess the impacts of consumption of contaminated fish on pregnant women and their newborn infants. The Michigan Maternal/Infant Cohort Study consisted of 313 infants and their mothers from four hospitals in the Michigan area (Fein et al. 1983). The 313 infants included 242 infants born to women who consumed moderate to high amounts of Lake Michigan fish for at least 6 years before their pregnancy (average consumption was 6.7 kilograms/year, which is equivalent to about 2.5 pounds of salmon or lake trout meals a month), and 71 infants whose mothers did not eat Lake Michigan fish.

The initial study examined two parameters: (1) the relationship between maternal fish consumption and exposure to PCBs and (2) the potential for in utero PCB exposure based on maternal fish consumption. The investigators analyzed for PCBs in maternal serum and in umbilical cord blood. The authors reported that no relationship was found between PCB levels in umbilical cord serum and maternal fish consumption. However, a statistical correlation was found between the maximum yearly ingestion of Lake Michigan fish and PCB levels in maternal serum. The concentration of PCBs in serum was predictable from the quantity of fish eaten by the mothers. These studies also indicated a highly significant (P< 0.0001) correlation of umbilical cord serum to maternal serum, which strongly suggests transplacental transfer of PCB molecules (Jacobson et al. 1983, 1984a; Schwartz et al. 1983).

Several adverse developmental health outcomes were observed in newborn infants of mothers who consumed more than 11.8 kilograms of contaminated fish (Fein et al. 1984). These health outcomes were birth weight, head circumference, and gestational age, as estimated by the mothers' reports of their last menstrual periods and on the Ballard Examination for Fetal Maturity. The exposed infants showed statistically significant decreases in infant birth weights (160 to 190 grams lighter than controls) and gestational age (average 4.9 days less than controls), and head circumference (average 0.6 centimeters smaller than controls). The authors attributed these effects to intrauterine exposure to PCBs via the mothers' diet of contaminated Lake Michigan fish. Maternal serum PCB concentrations for the samples as a whole averaged 5.5 +/- 3.7 ng/ml. According to the authors, this value is comparable to other midwestern American samples. Reported maternal fish consumption did predict maternal serum PCB levels (r= 0.37, P < 0.001), which in turn predicted umbilical cord PCB levels in their infants (r= 0.41, P < 0.001). The infants' umbilical cord serum averaged 2.5 +/- 1.9 ng/ml.

A follow up of these infants at 5 months postpartum indicated that the observed difference in birth weights still persisted (Jacobson and Jacobson 1988). Neonates who were exposed to 3 ppb or greater of PCBs, as indicated in their umbilical cord serum, were approximately 260 grams lighter than those infants exposed to less than 3 ppb PCBs in their cord serum. Further studies of these children at 4 years of age indicated a continuous deficit in weight gain when compared to the controls. Children with cord serum PCB levels of 5.0 ng/ml or more weighed 1.8 kg less on average than the lowest exposed children after adjustment for covariates and potential confounders (Jacobson et al. 1990a,b).

Neurological effects were also observed in the infants of this Michigan study. Prenatal exposure to low levels of PCBs from maternal consumption of contaminated Great Lakes fish was shown to have result in subtle behavioral effects (Jacobson et al. 1984a,b,c).

The Brazelton Neonatal Behavioral Assessment Scale (NBAS; Brazelton 1973) was used to assess behavioral outcomes, and the Ballard Examination for Fetal Maturity (Ballard et al. 1979) was administered to assess neuromuscular and physical maturity in 242 infants from the Michigan Maternal/Infant Cohort, 48 to 72 hours following birth (Jacobson et al. 1984b,c). Neonatal behavioral deficits associated with maternal consumption of contaminated fish included greater inclination to startle; poorer motor reflex and neuromuscular functioning; and depressed responsiveness, as evidenced by a greater number of hypoactive reflexes and limited adaptability of states of mind when compared to control infants. The degree of maternal fish consumption predicted a linear combination of alterations in neonatal health, including autonomic maturity (p >.025), number of abnormal reflexes (p >.05), and range of state (p >.005). Such subtle effects are categorized in the "worrisome" range by Als et al. (1979).

In the same infants from the Michigan Maternal/Infant Cohort, umbilical cord serum samples were analyzed for the presence of PCBs (Aroclors 1016 and 1260) as a direct measure of infant exposure. However, the authors found no correlation between cord serum values and behavioral effects in infants (Jacobson et al. 1984a). Difficulty in obtaining cord serum levels for all of the infants, especially those with the highest fish consumption exposure levels may have biased these results. Also, the behavioral effects associated with fish consumption in the Michigan Maternal/Infant Cohort may be due to unmeasured contaminants, other than PCBs, present in the fish (Jacobson and Jacobson 1988; Stone 1992).

When the infants from the Michigan Maternal/Infant Cohort reached 7 months of age, Jacobson et al. (1985) showed that infant visual recognition memory was predicted by both cord serum PCB levels and maternal report of Lake Michigan fish consumption, but not postpartum PCB exposure from breast-feeding (Jacobson and Jacobson 1988). Visual recognition memory performance was assessed using Fagan's Visual Recognition Memory Test (Fagan and McGrath, 1981) in infants at 7 months. Although the infants used in this study appeared clinically normal at birth, poorer performance on Fagan's Test was associated with intrauterine PCB exposure (measured by umbilical cord serum PCB levels and maternal fish consumption) indicating that some behavioral deficits resulting from secondary exposure may not be evident at birth and that follow up examinations are necessary. Jacobson and Jacobson (1988) concluded that visual recognition memory testing is capable of detecting subtle performance deficits in otherwise clinically normal infants.

Reduced performance on recognition memory testing in children exposed to contaminants through maternal consumption of Great Lakes fish may result from processing of information at a slower speed, poor visual discrimination, poorer attention, or difficulties in encoding and retrieving of information (Jacobson and Jacobson 1988). In addition, infant recognition memory has been shown to be a valid predictor of later intelligence, as measured by verbal intelligence quotient (IQ) scores among children (Fagan and McGrath 1981). Therefore, it is plausible that poorer performance in memory testing at an early age may be indicative of long-term cognitive deficits. Animal data subjected to the risk assessment model proposed by the U.S. Environmental Protection Agency (EPA) for developmental toxicity (IRIS 1994) resulted in derivation of a "reference dose" for PCBs of 7 X 10-5 mg/kg/day, based on reduced birth weights. The ATSDR has derived a chronic oral Minimal Risk Level (MRL) for PCBs of 0.00002 mg/kg/day, based on immunological effects (ATSDR 1993).

Additional follow up studies evaluated 4-year-old children from the Michigan Maternal/Infant Cohort previously examined for PCB-related deficits as infants (Jacobson et al. 1989, 1990a,b). At age 4, study participants were administered the McCarthy Scales of Children's Abilities (McCarthy 1972) to assess cognitive function. Exposure due to maternal consumption of contaminated fish from Lake Michigan was measured by PCB levels in cord serum (average = 2.5 +/- 2.0 ng/ml), maternal serum (average = 5.9 +/- 3.6 ng/ml), maternal milk (average = 835.9 +/- 388.4 ng/ml), and body burden at 4 years of age (breast-fed for at least 6 months, 5.1 +/- 3.9 ng/ml; breast-fed for less than 6 months, 1.2 +/- 1.6 ng/ml; and 0.3 +/-0.7 ng/ml for those not breast-fed). High cord serum PCB levels predicted poorer performance on McCarthy Verbal and Memory Scales in a dose-dependent manner. These results are consistent with earlier findings of an association between poorer memory and prenatal exposure in infants (Jacobson et al. 1985).

Postnatal exposure from nursing based on maternal milk PCB levels and weeks of nursing was not related to memory deficit in these children (Jacobson et al. 1989). In addition, the children's serum PCB level at 4 years (related primarily to postnatal nursing) also failed to predict cognitive outcomes in this study. Although a larger quantity of PCBs is transferred postnatally through lactation, prenatal exposure was found to be associated with decreased cognitive function in young children. In summary, findings of the Fein et al. (1983, 1984) and Jacobson studies are significant in that both initially identified associations between maternal consumption of contaminated fish and adverse human health outcomes in their children.

Smith's Wisconsin Maternal and Infant Study

The Wisconsin Maternal and Infant Cohort study also investigated the association of maternal exposure to toxic substances and adverse health effects in infants (Smith 1984). Results from this study indicated an association between maternal serum PCB levels and the number and type of infectious diseases suffered by their infants. The categories of infectious diseases included colds, earache, and flu symptoms. The author concluded that in utero exposure to PCBs resulted in an increase in the number of illnesses in infants during the first 4 months of life. However, the design of this study prevented access to either umbilical cord blood or infant serum acquired immediately after birth to analyze for the presence of PCBs. Therefore, the issue of exposure was presumed and not actually addressed by the Wisconsin study (Swain 1991).

Dar's Wisconsin Maternal and Infant Study

A more recent study in Wisconsin examined Great Lakes fish consumption and reproductive outcomes in Green Bay (Dar et al. 1992). In this prospective study 1,112 women who were seen at the time of a positive pregnancy test were asked to complete a questionnaire on fish consumption, health and reproductive history, and other relevant issues, and to provide blood samples for PCB analysis. After the pregnancy, the following reproductive outcomes were abstracted from hospital labor reports and measured: fetal wastage and stillbirths, birth weight, birth length, head circumference, ponderal index, and birth weight percentiles for live births. The typical negative associations between birth size measures and confounding variables, (i.e., consumption of caffeine and alcohol, and smoking) were found. Birth size was positively associated with gestational age, birth order, weight gain during pregnancy, male babies, and rural residence. However, contrary to expectations, a positive association was found (p<0.044) between birth size and PCB exposure (via sport fish consumption) for most mothers (the exception being those mothers who gained more than 34 pounds during pregnancy).

Serum PCB levels were performed on 106 women in this group, and a positive correlation (Pearson p = 0.666) was found between the amount of Lake Michigan fish mothers reported to consume and their PCB serum levels (Dar et al. 1992). The PCBs concentration was based on the sum of 13 individual congeners; only 5 of 13 PCB congeners were identified in maternal serum. Also, only 23 percent of the study group had PCBs above the detection limit of 0.6 ng/ml for each congener. Additionally, the consumption of highly contaminated sport fish was very low in this cohort. The authors concluded that the Green Bay mothers were exposed to very low levels of PCBs, compared with exposures in other studies, (e.g. Fein et al. 1983,1984). This low estimated level of exposure does not appear to have had a negative effect on birth size. The authors also suggest that if adverse reproductive effects occur in infants because of maternal fish consumption, perhaps there is a threshold effect below which there are no negative effects.

New York Ecologic Epidemiologic Study

An ecologic epidemiologic study was conducted within the Great Lakes drainage basin of upstate New York (Kagey and Stark 1992). The hypothesis for this study was "maternal residence within the Great Lakes drainage basin from 1968-1987 adversely impacted birth weight, gestational age, and congenital anomaly rates of their offspring. In addition, within the basin area, those infants whose mothers resided within the counties adjacent to the lakes were at greater risk of lower birth weight, shorter gestational age, and increased congenital anomaly rates associated with their close proximity to the lakes." In defining the exposure area as the Great Lakes drainage basin, it was assumed that potential exposure to contaminants was through one or more of the following routes: inhalation of specific contaminants that evaporated from the lakes, consumption of sport fish, home garden produce, and drinking water from the lakes or from point sources within the basin.

The authors determined that the mean birth weight of infants born within the Great Lakes drainage basin was comparable to the mean birth weight of infants born outside the basin: 3,313 versus 3,319 grams, respectively. Mothers residing in counties adjacent to the Great Lakes gave birth to infants who weighed 28 grams less than infants born to mothers who resided in nonadjacent areas. Approximately 6 percent of all births in this study population were low birth weight (less than 2,500 grams). This percentage was slightly higher in infants born to mothers who resided near the lakes, however.

The mean gestational age for the entire study population was 279 days; no differences were found in mean gestational age by study area or density quintile. Differences in congenital anomaly prevalence rates were also difficult to detect because of changes in coding and reporting practices over time. The authors concluded that the two major limitations of ecologic epidemiologic studies were the lack of individual measurements on exposure or outcome and the inability to control for confounding variables.

III. Human Health Studies Outside of the Great Lakes - Exposure to Similar Persistent Toxic Substance

Other epidemiologic studies have been conducted on mothers exposed to toxic substances, similar to those identified in Great Lakes fish, whose children have exhibited either reproductive and developmental or neurobehavioral effects.

The North Carolina Breast Milk and Formula Project

This study investigated the effects of prenatal exposure to PCBs (Rogan et al. 1986). Mothers were exposed to environmental or background levels of PCBs. A total of 856 breast-fed infants were identified at birth and periodically examined for up to 60 months. No association was found between maternal PCB levels and birth weight, head circumference, or gestational age. However, neurological effects were observed in these children. Children with higher in utero PCB exposure (assessed as breast milk fat PCBs concentration) exhibited neurological deficiencies, such as hyporeflexia and hypotonicity similar to those observed in the Michigan Maternal and Infant Cohort Study.

At 6 and 12-month follow up examinations, using the Bayley's Scales of Infant Development, the psychomotor index indicated continued decreases in behavioral effects (Gladen et al. 1988). At subsequent examinations these behavioral effects were not observed (Gladen and Rogan 1991).

Japan and Taiwan PCBs Studies

Two separate but similar studies were conducted on accidental exposure to PCBs involving human populations in Japan and Taiwan (Hsu et al. 1985). In Japan in 1968 and in Taiwan in 1979, accidental mass poisoning incidents occurred from cooking oil contaminated by thermally degraded polychlorinated biphenyls. In 1985, 117 children born to women in Taiwan who ingested contaminated rice oil and 108 unexposed controls were examined and evaluated.

The authors indicated that the exposed children were shorter and weighed less than the controls. Additionally, the exposed children showed delayed developmental milestones, deficits on formal developmental testing, and abnormalities on behavioral assessment. Intrauterine exposure of Japanese infants to PCBs was associated with prematurely and decreased size for gestational age (Higuchi 1976; Wong and Huang 1981). In follow up studies of these exposed children, growth deficiencies were evident. Autonomic disturbances were observed; sluggish, clumsy, and jerky movements were exhibited; and the average intelligence quotient in these children was in the low 70s. The children in this study were exposed to PCBs but also to the heat degradation products of PCBs including highly toxic polychlorinated dibenzofurans.

Occupationally Exposed Female Capacitor Workers

This study investigated the relationship of PCB exposure to birth weight and gestational age in 356 infants of mothers occupationally exposed to PCBs during the manufacture of capacitors in upstate New York (Taylor et al. 1989). The results from this study indicated that infants exposed in utero to PCBs were also smaller at birth, on average 153 grams less than unexposed infants, and that their mean gestational age was 6.6 days shorter than infants from mothers employed in the same facility but not directly exposed to PCBs. The authors concluded that there is a significant association (p<0.02) between increasing serum PCB level and decreased birth weight and gestational age, and that the decrease in birth weight is at least partially related to shortened gestational age. The authors also stated that the magnitude of these effects was quite small and the clinical significance of these results was questionable. However, the authors suggest that unknown or unmeasured confounding variables that influence birth weight and gestational age will of course alter the results of their study.

Epidemiologic studies of exposed human populations usually provide the most convincing evidence of human health effects. However, epidemiologic studies may vary with design (i.e., prospective studies, retrospective studies, ecologic studies) or investigators may use different analytical techniques; therefore, it becomes very difficult to evaluate or compare studies. The following section on study limitations will discuss some of these differences.

IV. Limitations of Human Health Studies

The Michigan Sport Fishermen Cohort Studies was the first epidemiologic study to assess exposure to PCBs in humans from consumption of sport fish from the Great Lakes. This study supports a positive correlation between human uptake of toxic pollutants in the Great Lakes and Great Lakes fish consumption. Generally, this study was well designed and executed but limited by the analytical methodology used, which was state-of-the-art at the time. For example, to determine PCB levels in serum, the samples from the cohort were pooled to determine "total PCBs." This methodology did not allow study of individual PCB levels within a cohort. Additionally, the investigators used packed column chromatography to determine total PCBs, which resulted in a lack of sensitivity to detect individual PCB congeners. But, again the methodology used in these late 1970s studies was the best available at the time.

The differences in laboratory techniques makes evaluating or comparing studies very difficult. For example, the two sport angler studies (Michigan and Wisconsin) were both conducted to determine the possible correlation between sport fish consumption and body burden of toxic substances. The analytical methods used to determine total PCBs in human sera were packed column in the Michigan study and capillary column gas chromatography in the Wisconsin study. The interlaboratory differences strongly contributed to variation in levels of total PCB levels in sera in the two studies. The highest PCB congener sum in the Wisconsin study was 27.1 ug/l; the highest value in the Michigan study was 366 ug/l. Using the older packed column technique, the investigators could only determine total PCB. Using capillary column gas chromatography the investigators could determine PCB by summation of individual congeners to estimate total PCB body burden. The Michigan and Wisconsin values may be attributed to design differences, (e.g., different methods of selecting survey participants such as age and sex), different methods to obtaining fish consumption histories, different times of sample collection [1973 and 1980 for Michigan versus 1986 for Wisconsin], or different consumption and fish preparation patterns). Because of the differences in analytical methodology and design, the results of these studies are therefore limited in providing convincing evidence of the status of human health from exposure to Great Lakes contaminants.

The Michigan Maternal/Infant cohort study is widely referenced because it was the first epidemiologic investigation that demonstrated the association between maternal consumption of Great Lakes fish and adverse human health outcomes in their children. The establishment of this cohort was an excellent idea in an attempt to associate maternal fish consumption, PCB serum levels, and adverse health outcomes. However, several shortcomings and unanswered questions regarding this study remain. The following discussion briefly evaluates the major limitations of this study.

Of the 8,482 women delivering infants in four Michigan hospitals, only 4 percent (343) were considered eligible to participate based on reported high fish consumption (Fein et al. 1983, 1984). The control population comprised only 71 women who had reported eating no Great Lakes fish. The statistical power of the study is limited by the size of the control group, which is only one-fifth the size of the exposed group.

A nonrandom sampling technique was used to define the study population. Investigators purposely selected individuals who reported consuming large amounts of Lake Michigan fish; therefore, the final study group consisted of a small homogeneous population. Random sampling and heterogeneity of the investigative cohort are important because the ideal would be to have the unbiased ascertainment of eligible exposed and control individuals.

A number of differences existed in individual characteristics within the exposed group, as well as between the exposed mothers and the control group in these studies. Thirty-seven potentially confounding factors that may have affected the observed health outcomes in these studies have been identified (Fein et al. 1984). These factors include health history, maternal health, obstetrical care, pregnancy and delivery, and fetal exposure to drugs, such as alcohol, caffeine, nicotine, and cold medications. These differences raise the possibility that the exposed and control groups may differ in their overall health condition and that these differences may affect their pregnancy outcome or infant development or both.

Estimation of maternal fish consumption in this study was based on the ability of the mother to recall past patterns of fish consumption for at least 6 years. The investigators defined the total fish consumption using a "weighted sum" (amount of fish meals consumed a year) calculation. The accuracy of this method is questionable, especially when trying to determine maternal exposure, and precluded the investigators from making dose comparisons.

In addition, maternal estimates of cumulative fish consumption is poorly correlated with maternal serum PCB levels in some studies. In other studies, there are correlations between fish consumption and serum measurements. The reason for these discrepancies is unknown. Also, there was no correlation between maternal fish consumption and umbilical cord levels of PCB which are considered to be an accurate measure of transplacental exposure. Given these limitations, it appears that either the estimation of maternal fish consumption was inaccurate or the measurement of total PCB levels was not a good indictor of total fish consumption. Also, the standards used (Aroclor 1016 and 1260) as references to quantify total PCBs accounted for only a small portion of the PCB congeners detected (Swain 1991). Therefore, the analytical methods used to measure PCB levels may not have been appropriate.

In summary, the limitations of this study were the sampling technique used to define the population, uncertainties associated with estimation of exposure, lack of correlation between fish consumption and PCB levels in serum in some studies, and a number of confounding variables not taken into account. In addition, the health effects observed in this study may also be attributed to exposure to contaminants other than PCBs present in fish. Researchers have demonstrated that the fish consumed probably contained a mixture of other chlorinated toxic chemicals, such as DDE, DDT, aldrin, and dieldrin, as well as heavy metals, such as mercury or lead. These substances have been found in contaminated fish from the Great Lakes region (National Health and Welfare Canada 1991). Currently, follow up studies using the Michigan Maternal/Infant Cohort are under way. New information from these efforts may provide answers for some of the remaining questions.

In Dar's Wisconsin Maternal and Infant study which examined fish consumption and reproductive outcomes in women from Green Bay, Wisconsin, the investigators identified a positive correlation between PCB exposure and birth weight (Dar et al. 1992). In this study, the estimates of PCB intake may have been inaccurate because they were based on questionnaire responses subject to recall. Even if no recall bias existed, uncertainties still exist because the amount of fish eaten in a fish meal varies from individual to individual and the amount of PCBs and other toxic substances varies within the same species of fish because of age, size, length, and fat content.

In addition, serum samples were taken early in pregnancy (women with positive pregnancy tests) within this population, and it has been demonstrated that the serum half-lives of three different congeners detected in this study ranged from 124 to 338 days (Buhler et al. 1988). Of the 1,112 women in this study, only 106 maternal serum samples were analyzed for PCBs. Because a small number of maternal serum samples were analyzed the values obtained may not be a true representation of the entire study population. No additional maternal blood during pregnancy or fetal cord blood after birth was analyzed for PCB levels to determine if maternal levels of PCBs varied during pregnancy or to determine exposure levels to the fetus.

Also during the time of this study, a lot of publicity in the state focused on the potential for adverse health effects in infants from mothers who consumed sport-caught fish and on prenatal counseling, including information on toxins in sport fish. The authors concluded that this publicity probably made the population more sensitive to fish contamination and therefore, fish consumption during pregnancy may have been less than usual before and during pregnancy.

In comparing Dar's Wisconsin Maternal and Infant study to the Michigan Maternal and Infant study, differences exist in study design. The most critical difference in methodologies was in the selection of exposed populations. In the Michigan study, the investigators selectively chose individuals who consumed large amounts of Lake Michigan fish, (i.e., populations that would have very high PCB exposure). In comparison the participants in the Wisconsin cohort consumed much lower amounts of contaminated fish. When the exposure scores of the Michigan and Wisconsin studies were compared, only 49 persons would have qualified for Michigan's exposed group. Of the 49 persons who would have qualified, only 4 participants were estimated to be exposed at the high Michigan exposure level. Because of the differences in study design, therefore, the status of human health from exposure to Great Lakes contaminants remains uncertain.

The Minnesota Ecologic Epidemiologic Study and the New York Ecologic Epidemiologic Study both hypothesized that populations that live near the lakes would be potentially exposed to pollutants in air and water, thereby resulting in adverse reproductive health outcomes (Shuman et al. 1982; and Kagey and Stark 1992). Neither study was able to definitely prove its hypothesis because of the study limitations.

Ecologic studies have two major limitations: the lack of individual measurements on exposure or outcome and the inability to control for confounding variables. If individual measurements for exposure or outcome are lacking in a geographical area as large as the Great Lakes, misclassification is possible. If exposure was classified incorrectly, it was assumed to be random. This would tend to underestimate the relationship between residence and outcome. The second limitation, the inability to control for confounding variables, may have greatly altered the data outcome. Factors known to affect birth weight, such as smoking history and alcohol consumption are not recorded on birth certificates, from which the data for the study were obtained. When data on potential confounding variables were not recorded, the investigators assumed that they were distributed equally among exposure areas in the Great Lakes drainage basin. However, this assumption may not have been correct.

Epidemiologic studies of exposed human populations provide the most convincing evidence of human health effects. Of the three major routes of human exposure in the Great Lakes -- air, drinking water, and fish consumption -- fish consumption is generally thought to present the greatest risk of exposure. The most direct evidence for adverse human health effects from environmental pollution is found in a series of studies linking PCB exposure to consumption of contaminated fish (Fein et al. 1983, 1984; Jacobson et al. 1984a; Jacobson and Jacobson 1988). Replicating improving upon, and continuing these types of epidemiologic studies should provide the most relevant and convincing evidence of the status of human health from exposure to Great Lakes pollutants.

V. Program Areas of Research

Given the implications of the association between contaminants in the Great Lakes and adverse human health outcomes, the following research is being conducted to clarify the relationships among exposure, contaminant levels in human biological tissues and fluids, and human health effects. Below are specific research data gaps that are in the process of being filled, which will provide necessary information for assessing human risk from exposure to Great Lakes contaminants.

Characterization of exposure and determination of the profiles and levels of Great Lakes contaminants in human biologic tissues and fluids

Many studies have identified the presence of numerous toxicants found in a variety of species in the Great Lakes food chain, including humans (Phillips and Birchard 1991a,b; Humphrey 1987; Kreiss 1985). Exposure to contaminants has been shown to be associated with health effects not only in individuals consuming fish from the Great Lakes basin but also in their children.

Exposure to contaminants via consumption of contaminated fish over an extended period of time allows for continuous exposure that may increase the potential for adverse human health effects. The persistence and latency of some of the Great Lakes contaminants have been demonstrated to be important factors in their ability to cause adverse health effects, such as tumor promotion, hepatic effects, and reproductive/developmental effects (Humphrey 1988b). Studies are being conducted to:

Identification of sensitive and specific human reproductive end points

Human reproductive effects have been initially assessed in females following exposure to Great Lakes contaminants. (Fein et al. 1984; Jacobson and Jacobson 1988). It is known that Great Lakes contaminants can also accumulate in male reproductive fluids and organs such as seminal fluid and testes (Bush et al. 1986; Dougherty et al. 1980). Studies are ongoing to:

Determination of the short- and long-term risk(s) of adverse health effects in the children of exposed parent

Growth retardation, decreased birth weights, and neurological effects have been observed in the offspring of mothers who consumed contaminated fish from the Great Lakes region. There are no reports about adverse health effects in children from paternal exposure to Great Lakes contaminants. The following research is ongoing to provide some answers regarding parental exposure and potential health outcomes in their children:

Feasibility of Establishing registries and/or surveillance cohorts in the Great Lakes

Currently, there are no registries of individuals exposed to multiple contaminants in the Great Lakes region via fish consumption. Cohorts of populations considered at risk because of current or past dietary patterns of consumption of contaminated fish from the Great Lakes exist, feasibility analyses are ongoing to:

This ongoing research will provide information required for ATSDR and state agencies to perform public health assessments of people at risk from exposure to hazardous substances. In addition, this research will provide valuable information for the assessment of human risk from simultaneous exposure to chemical mixtures in the Great Lakes region. The results of these research projects will extend current knowledge of the effects of Great Lakes contaminants on human reproduction, especially male reproductive toxicity, and resulting adverse health effects in the children of exposed individuals. This research will also establish new cohorts for research efforts in the Great Lakes basin that can be used for future investigation of other human toxicological end points. Finally, this research will also eliminate a number of uncertainties in the risk assessment process, thereby providing direction for future research.


Local Navigation

Jump to main content.