Health advisories for consumers of Great Lakes sport fish: is the message being received?

Nationwide, 45 states issue health advisories for sport fish consumers. Chemical contaminants in some Great Lakes (GL) sport fish include compounds suspected of causing adverse reproductive and developmental effects. Although advisories to reduce consumption of contaminated fish, especially by women, have been issued by GL states (i.e., Illinois, Indiana, Michigan, Minnesota, New York, Ohio, Pennsylvania, and Wisconsin) since the mid-1970s, little is known about advisory awareness and GL sport fish consumption in the general population. To estimate the prevalence of GL sport fish consumption and health advisory awareness, we conducted a population-based telephone survey of 8,306 adult residents of the eight GL states. We gathered information concerning respondents' demographic characteristics, fish consumption during the preceding year, and sport fish consumption advisory awareness. The survey response rate was 69%. GL sport fish were eaten during the preceding year by 8.4% -95% confidence interval (CI), 7.6-9.2- of adults in the GL states, approximately 4.7 million persons. Women accounted for 43.9% (CI,39. 4-48.4) of consumers. Although 49.9% of GL sport fish consumers were aware of a health advisory, awareness varied significantly by sex: 58.2% (CI, 51.7-64.7) of males and 39.1% (CI, 32.6-45.6) of females were aware. Using logistic regression, we found awareness associated with male sex -odds ratio (OR) = 2.3; CI, 1.5-3.5), white race (OR = 4.2; CI, 1.9-9.1), college degree (OR = 3.1; CI, 1.3-7.6), and consuming >=24 GL sport fish meals/year (OR = 2.4; CI, 1.4-4.3). Only half of GL sport fish consumers reported awareness of a health advisory concerning eating GL sport fish. Awareness was especially low among women, suggesting the need of targeted risk communication programs for female consumers.

Nationwide, 45 states issue health advisories for sport fish consumers. Ch cal coninnts in some Great Iakes (GL) sport fish include compounds suspeted of causing advese reproductve and developmental effects. Although advisories to reduce consumption of contaminated fish, espeially by women, have been issued by GL states (.e., Illinois, Indiana, Mich n, Minnesota, New York, Ohio, Penyani, and Wison) ce the mid-1970., l i about adsory awarness and GL sport fish cons pionin the general population. To estimate the prevalence of GL sport fish ptio an health advisory awarenes, we otd a pulationbased teephone survey of  Great Lakes (GL) sport fish consumption advisories were first issued in the 1970s after extensive testing detected chemical contaminants in fish tissue. Of particular concern were environmentally persistent lipophilic pollutants such as polychlorinated biphenyls (PCBs) and DDT. These compounds are often found in the highest concentrations in the older and larger predatory fish, which are sought and preferred by many sport anglers. In 1991, an estimated 2.55 million licensed anglers fished on one or more of the Great Lakes (1). Thus, the popularity of this sport presents a significant potential for human exposure to chemical residues in fish tissue. Although two decades of environmental regulation have substantially reduced chemical residues (2-3), some sport fish still contain levels thought to be potentially harmful to human health (Anderson et al., unpublished data).
Studies of long-term consumption of GL sport fish confirmed an association between contaminated fish consumption and increased PCB or DDT\DDE body burdens (47). The potential adverse health effects of these contaminants have been studied extensively (8)(9)(10). The EPA has classified PCBs and DDT\DDE as probable human carcinogens. Consistent with low exposure effects observed in nonhuman primates, some human epidemiologic studies have found associations between maternal/fetal PCB levels, Lake Michigan sport fish consumption, and adverse reproductive and developmental effects (11)(12)(13). These studies are not conclusive, however, and scientists disagree regarding interpretation of their findings (14)(15)(16)(17). This level of uncertainty complicates the process of communicating risk information to sport fish consumers.
Current GL sport fish consumption advisories seek to 1) inform the public about the chemical contaminants contained in some sport fish, 2) educate consumers as to how they can minimize their exposure to contaminants, 3) remind consumers of the health benefits of fish consumption (18,19), and 4) present advisory information in a manner conducive to maximal voluntary compliance (Anderson et al., unpublished data). Because of potential adverse reproductive and developmental effects, all current advisories make specific consumption frequency recommendations for women of childbearing age. GL advisories seek to help individual consumers make informed decisions regarding sport fish consumption.
In the GL states, the level of advisory awareness among licensed anglers has been used to assess the adequacy of advisory communication programs (20). Licensed anglers are a relatively easy-to-identify group of GL sport fish consumers who can be surveyed by mail (5,(21)(22). One drawback of this approach is that the results of surveys conducted solely on anglers with fishing licenses might not be generalizable to all persons who eat GL sport fish. To overcome this limitation, we conducted a random-digit-dial telephone survey of the general population. We surveyed adult residents of the eight GL states to characterize the types of fish eaten by the general population, the demographic characteristics of persons who ate GL sport fish, and the level of advisory.awareness among these GL sport fish consumers.

Materials and Methods
Survey design. From April 1993 through February 1994, trained interviewers from the University of Wisconsin Survey Research Laboratory conducted a population-based random-digit-dialed telephone survey (23)(24)(25) of adults (i.e., persons >18 years of age) residing in Indiana, Illinois, Michigan, Minnesota, New York, Ohio, Pennsylvania, and Wisconsin). One randomly selected adult from each participating household was interviewed. Interview dates were scheduled so that approximately equal numbers of interviews were completed during each of the four seasons. Informed consent was obtained before the start of the interview, and the data acquisition procedures maintained complete respondent anonymity.
Interviewers used a standard questionnaire to collect information regarding demographic characteristics, the types of fish eaten Articles * Awareness of sport fish health advisories during the preceding 12 months, and an estimate of the average number of fish meals eaten during a time interval selected by respondents (e.g., during a week, month, or year). Respondents were characterized as persons who either 1) ate no fish, 2) ate only commercially purchased fish, 3) ate sport fish not caught in a Great Lake, or 4) ate some GL sport fish. Respondents who reported eating GL sport fish provided additional information, including estimates of the number of GL sport fish meals they consumed during the preceding 12 months, whether they had heard of their state's health advisory for sport fish consumers, and if they had followed specific advisory recommendations (e.g., cleaning and cooking practices). They provided consumption estimates for six GL sport fish groups: lake trout, carp/catfish, brown trout, rainbow trout/chinook salmon/coho salmon, perch/smelt/walleye, and all other GL sport fish. These fish species were grouped according to the level of chemical contaminants reported by federal and state monitoring programs. These groups are listed in approximate rank-order, from highest to lowest, by level of chemical contaminants detected. We defined advisory awareness as a self-report of having heard of the health advisory. Compliance was defined as a selfreport of always or usually following an advisory recommendation. None of the selfreported information was independently verified.
Analytic methods. We calculated overall and state-specific response rates using standard procedures (26) (see Appendix 1). To obtain population-based estimates, we weighted each respondent's information (27,28). Appendix 2 describes the case weighting methodology.
The low number of respondents with less than a high school degree was insufficient to weight separately; thus, only two categories of educational attainment were used to calculate weights. These categories were persons with a high school degree or less and persons with at least some post-high school education. In all analyses, respondent age was coded as a threelevel categorical variable (i.e., 18-34 years, 35-44 years, and 45 or more years of age) based on the likelihood of childbearing. We initially explored the data through univariate and stratified analysis. Using SUDAAN statistical software (Research Triangle Institute, Research Triangle Park, NC) (29), we calculated overall prevalence estimates for four different fish consumption patterns, advisory awareness among GL sport fish consumers, and self-reported compliance among GL sport fish consumers who reported advisory awareness. We conducted a stratified analysis to determine how advisory awareness varied by sociodemographic group and we conducted a multivariable logistic regression analysis (30). Possible interactive effects between sex and other variables were assessed. Models of significant variables were derived by a backward model-selection method (31) and by comparison of hierarchical models using likelihood ratio chi-square statistics. Age group and state of residence variables were kept in all models. To adjust for the study's sampling design, the final logistic regression analysis used SUDAAN software.

Results
A total of 8,306 persons were interviewed, including over 1,000 adult residents per state. The overall survey response rate was 69% (see Appendix 1). State-specific response rates ranged from 57% in New York to 78% in Wisconsin. Persons who had less than a high school degree were underrepresented, comprising 14% of our weighted sample compared with the 23% indicated by the 1990 census for these states (28).
Fish consumption. The estimated percentages of persons in the eight GL states who ate fish obtained from different sources are shown in Table 1. The median number of total fish meals eaten (from all sources) was not significantly different between groups. An estimated 8.4% [95% confidence interval (CI), 7.6-9.2] of adult residents in the GL states had eaten GL sport fish during the preceding year (approximately 4.7 million persons). Ninety-two percent of GL sport fish consumers in this survey were white, the median age was 39 years, 44% were women, and 52% lived in either Michigan or Ohio. Results of the 1990 census indicates that, among adult residents of the GL states, 84% were white, 53% were women, and 26% lived in either Michigan or Ohio.
The estimated distribution of GL sport fish consumers by fish consumption level and sex is shown in Figure 1. A median of 6.5 fish meals were eaten per year (range, 1-292 fish meals per year). We estimated that 830,000 persons in the eight GL states had eaten .24 GL sport fish meals per year. Men reported having eaten GL sport fish more frequently than did women. The median consumption levels for men and women were 8.2 (CI, 6.7-9.6)] and 5.8 (CI, 5.0-7.6) meals, respectively.
Sport fish consumption advisory awareness. The prevalence of advisory awareness among adult GL sport fish consumers by various sociodemographic characteristics is shown in Table 2. Approximately half of all GL sport fish eaters reported that they had heard of their state's health advisory. Advisory awareness varied significantly by race, sex, educational attainment, GL sportcaught fish consumption level, and state of residence. Advisory awareness was not significandy different by age category. Our sample contained small numbers of GL sport fish eaters from some states. This resulted in unstable state-specific estimates of advisory awareness as indicated by wide 95% confidence intervals. In Minnesota and New York, the level of awareness indicated by the weighted sample was approximately 10% less than that indicated by the unweighted data. In these states, persons with relatively high weighting (e.g., males 18-34 years of age who did not have a high school degree) tended to report that they had not heard of the advisory.
The final logistic regression model for variables associated with advisory awareness among GL sport fish eaters is shown in Table 3. The risk estimates displayed in Table 3 are odds ratios (OR). In the analysis, Wisconsin was used as the referent group in the logistic regression model because the level of GL sport fish consumption and advisory awareness among Wisconsin anglers had been documented previously. Statistical significance (p<0.05) was assessed by 95% confidence intervals. Parameters with confidence intervals that did not include an OR of 1.0 were therefore the ones that were significant. After adjusting for other variables in the model, advisory awareness was significantly associated with male sex, white race, having a college degree, and eating .24 GL sport fish meals per year. GL sport fish consumers in Ohio were significantly less likely than GL sport fish eaters in Wisconsin (the referent state) to be aware of an advisory. None of the interaction terms investigated were statistically significant. Of those GL sport fish consumers who were aware of an advisory, compliance with advisory recommendations differed significantly between men and women ( Table 4). Use of recommended cleaning and cooking methods was the most frequently reported risk reduction practice, reported by 69% of men and 55% of women. Compliance was significantly lower for advisory recommendations that required changes in fishing Abbreviations: Cl, 95% confidence interval; GL, Great Lakes. aModel includes 666 out of the 671 total persons for whom complete data was known for all variables in the model; three persons had missing race information and two persons had missing GLSCF consumption level information (see also

Discussion
The popularity of sport fishing on the Great Lakes is indicated by our estimate that approximately 4.7 million persons had eaten GL sport fish during the year preceding the survey. Our results indicate that the majority of these persons infrequently ate GL sport fish and that they tended to eat fish species that contain relatively low contaminant levels (i.e., perch/smelt/walleye). The GL sport fish consumption levels reported by most respondents were highly unlikely to result in body burden levels previously associated with adverse human health effects. However, the frequency of GL sport fish consumption varied widely, suggesting substantially different potentials for chemical contaminant exposure among individual consumers. Because environmentally persistent chemical contaminants are present in some GL sport fish, consumers will probably continue to seek information regarding the risks and benefits of eating GL sport fish.
In the United States, state governments are primarily responsible for managing potential risks associated with contaminants in locally caught sport fish (32). As of September 1993, 45 of the 50 states had issued one or more fish or shellfish consumption advisories or bans (33). Each of the eight GL states has issued sport fish consumption advisories since the mid-1970s. The overall effectiveness of sport fish advisory programs depends on the effectiveness of both the recommendations made and communication methods used (21,34,35). Unfortunately, inconsistency between state-issued advisories has been recognized as a nationwide problem (32). Inconsistencies in GL sport fish advisories have at times confused consumers and limited the effectiveness of advisories in the region. Representatives of the eight GL states have drafted a uniform advisory for the Great Lakes (Anderson et al., unpublished data), but it has not yet been adopted. In this survey, although ability to make interstate comparisons was limited, state of residence predicted the level of advisory awareness, suggesting that significant differences in advisory programs continue to exist.
In general, GL sport fish advisories encourage sport fish consumers to eat fewer of the fish species and sizes known to contain elevated levels of chemical contaminants and recommend the use of cleaning and cooking methods that can substantially reduce the levels of PCBs and other fat-soluble contaminants contained in a fish meal [Anderson et al., unpublished data; (36)(37)(38)]. Since GL advisories were first issued, several studies have documented a decline in GL sport fish consumption (22,39,40). In our survey, the most widely accepted advisory recommendation was deaning and cooking methods. The results of a survey of New York anglers indicated that the use of these cleaning and cooking methods was significantly higher among persons who were aware of that state's health advisory (21). These findings support the belief that sport fish consumption advisories can decrease chemical contaminant exposures among sport fish consumers if effective communication programs are used.
Communication programs in the GL states have traditionally targeted licensed anglers, who are predominantly white men. Written advisory information (e.g., fishing regulation booklets and advisory brochures) has usually been distributed through the recreational fishing industry and governmental offices. GL sport fish consumers who do not purchase licenses might not have access to information distributed in this way. Because we did not ask respondents if they had purchased fishing licenses, we could not directly assess advisory awareness among licensed anglers and nonlicensed GL sport fish consumers. However, our estimated number of GL sport fish consumers was roughly twice that of the most recent estimate of licensed GL anglers, suggesting that a substantial proportion of GL sport fish consumers were not licensed anglers. Only half of the adults in our population-based survey who had eaten GL sport fish had heard of their state's health advisory. When contrasted with the overall high levels of advisory awareness found by previous mail surveys of licensed GL anglers (7,21,22), it appears that substantial numbers of nonanglers are not receiving advisory messages.
Our survey results indicate that existing advisory communication programs are less effectively reaching women, nonwhites, and persons with lower levels of educational attainment. The results of previous surveys of licensed GL anglers also support this conclusion (21,225. Well-accepted principles of risk communication indicate that messages designed for specific audiences are often needed to reach diverse subgroups within a population (41,42). The weight of evidence from the risk communication literature, surveys of licensed GL anglers, and this population-based survey of GL sport fish consumers suggests that expanded use of targeted advisory communication methods are needed.
To accomplish this, the EPA has developed a detailed and comprehensive guidance document (42) to assist health professionals with the task of fish consumption advisory risk communication. The manual provides specific advice to health professionals on all aspects of health advisory risk communication, including risk communication as a process of sharing information, problem analysis and program objectives, audience identification and needs assessment, communication strategy design and implementation, program evaluation, and methods of responding to public inquiries. By following the principles outlined in the EPA guidance, state jurisdictions can develop successful risk communication programs for those who have not been receiving this very important message.
In an era of shrinking governmental budgets, individual state agencies must balance the costs of managing potential health risks against the magnitude of the potential health risk. It is unlikely that sport fish advisory programs will have sufficient resources to mount extensive communication campaigns. However, targeted communication programs designed to reach specific groups are possible. Our results suggest that additional communication efforts are particularly needed to reach persons most sensitive to the effects of the chemical contaminants (i.e., women of childbearing age). Greater collaboration between government agencies and risk communication specialists would probably reduce the costs of developing and evaluating new communication programs, and this collaboration would reap the additional benefit of increased consistency among state-issued advisories. .The final weight is a product oft deign w.eig9ht and ithe poststratification weight, and the constant.25. The :const-anintcluded so that all four p.li waves could be ud estima annual preva-lence>.The design!wei.ght is constructed front thesotat specific responsp:.. rate, the hou.shold unit..selectio.npr obiliy,eInumber of resiea p.hones inthe household u eheume of ts residing in the Polteiiain atI.v r hF u & c constr tJrec lco mbinatio of. eighsats three age ou i, w sexi rdtwoeducaion groups. Deinwegt were su nndwrtehbsnainitth statificatio *... cet serve as1* the tenor in stern. The lSSc ur..i population estimate was obtaRined for each ataifie io tLThs th. l_cistatiiato weight was the ratio of the census populatin eimtes to the sum of the desin weights in each statification cell.