Identification of Sport Fish Consumption Patterns in Families of Recreational Anglers through Factor Analysis

Environmental Research Section A 89, 19}28 (2002) doi:10.1006/enrs.2002.4348, available online at http://www.idealibrary.com on

Identification of Sport Fish Consumption Patterns in Families of Recreational Anglers through Factor Analysis Gregory P. Beehler,* John M. Weiner,* Susan E. McCann,* John E. Vena,* and David E. Sandberg-?A *Social & Preventive Medicine, -Department of Psychiatry, and ?Department of Pediatrics, School of Medicine and Biomedical Science, University at Buffalo, The State University of New York, Buffalo, New York 14214-3000; and APediatric Psychiatry and Psychology, Children’s Hospital of Buffalo, 219 Bryant Street, Buffalo, New York 14222-2099 Received July 6, 2001

Key Words: sport Ash; children; consumption patterns; factor analysis; Great Lakes Region.

This paper reports on the identiAcation of sport Ash consumption patterns in angler families through the factor analysis technique. New York State recreational anglers and their spouses or partners were surveyed in 1991 about their consumption of Lake Ontario and Lake Erie sport Ash. Respondents were surveyed again in 1997 to report the number of sport Ash consumed by their children aged 5+10 years. Parental report revealed that 60% of children had consumed at least one sport Ash meal over their lifetime. Anglers’, partners’, and children’s variables were entered into three separate factor analysis models to assess consumption patterns. Factors for anglers and partners accounted for 65 and 59% of variance in consumption scores, respectively. Factors dealing with trout and salmon consumption accounted for the most variance. Children’s factors accounted for 82% of variance in consumption scores, showing separation in relation to type of Ash, body of water, and age at consumption. Children’s factors were then used as dependent measures of separate multiple regression runs in which parental factors were entered stepwise as predictors. SigniAcant associations between parental and children’s factors were noted, suggesting that sport Ash consumption patterns in parents are predictive of similar consumption patterns in children. Results suggest that sport Ash consumption advisories do not fully prevent consumption of contaminated sport Ash during childhood. Therefore, risk communicators may need to modify the current strategy of informing anglers and their families about sport Ash consumption recommendations.  2002 Elsevier

INTRODUCTION

Consumption of sport 7sh and game from the Great Lakes and tributaries has been a public health concern for several decades due to the contamination of these waters and 7sh with a number of different polychlorinated toxicants (Anderson et al., 1998; Cordle et al., 1982; Fiore et al., 1989; Kearney et al., 1999; Kosatsky et al., 1999; Kostyniak et al., 1999; Sonzogni et al., 1991). All of the states bordering the Great Lakes and Canada have issued health advisories designed to inform the population with regard to the dangers of consuming such wildlife and ways of preparing these foods to minimize risks. For example, the New York State advisories (NYSDOH, 2001) have recommended that children and women avoid consumption of these foods. However, awareness and compliance with sport 7sh consumption advisories have not necessarily been high (Burger et al., 1998, 1999a; P8ugh et al., 1999; Tilden et al., 1997). There are measurement problems associated with determining the nature of advisory compliance. One is the obvious recall of consumption patterns particularly when such consumption represents a lowfrequency, relatively rare event. A second problem is the time frame used to determine frequency of meals. Previous research on adult recreational anglers typically considers the total years of sport 7sh consumption (Falk et al., 1999), the number of meals in the past 12 months (Courval et al., 1996; Falk et al., 1999; Tilden et al., 1997), the average number of meals per month (Burger et al., 1999b), or the number of meals consumed 3 weeks prior to

Science (USA) 1 To whom correspondence should be addressed at Social & Preventive Medicine, School of Medicine and Biomedical Sciences, University at Buffalo, The State University of New York, Farber Hall Rm. 270, 3435 Main St., Bldg. 26, Buffalo, NY 14214-3000. Fax: (716) 829}2979. E-mail: [email protected].

19 0013-9351/02 $35.00  2002 Elsevier Science (USA) All rights reserved.

20

BEEHLER ET AL.

survey (Kosatsky et al., 1999). A third problem is the size of the meal consumed. Some investigators attempt to characterize this in terms of weight, while others employ volume estimates. This report focuses attention on two important issues in the assessment of consumption of 7sh from the Great Lakes. The 7rst is the amount consumed by children (ages 5}10 years). The second is the relation between the child’s pattern and that of his or her parents. These issues have not been discussed with supporting data in the environmental health sciences literature and are important in evaluating current public health interventions and planning new ones. Reporting consumption patterns is not straightforward if the intake of the foods of interest are low within any speci7c time period. With low levels of consumption, the typical descriptors of center (average or median) are reduced to numbers such as 0, 1, or 2 meals per time period. The range, however, associated with these estimates could be fairly broad, suggesting that a small percentage of the respondents may be consuming a larger number of meals. Hence, the description of center and spread would suggest that a better description of consumption might be a histogram, depicting the distribution of meals in the identi7ed time period. Such descriptions, however, are subject to dif7culty when multiple time periods are considered. Trying to depict the distributions across such time periods is dif7cult since the centers and spreads tend to be relatively stable. In other words, there is no progressive regression-like effect allowing the distributions to appear separately across time. Instead, the distributions appear to be superimposed. Individuals have attempted to correct the problem by considering lifetime consumption. This is calculated by summing the consumption in individual time periods. This is essentially a weighted sum with all weights equal to one. While that may be satisfactory as a 7rst approximation of an appropriate weighted sum, statistical models that allow more precise weightings have existed for more than 50 years. One such model is factor analysis. This statistical procedure is designed to create weighted sums of the original variates so that independent parts of the total variance are explained. That is, the 7rst factor would explain the largest percentage of the total variance associated with the multivariate system studied. The second factor would explain the next larger independent portion of the variance. This analytic process continues until all of the variance is explained. Factor analysis also offers other important advantages. The resulting factor scores (i.e., weighted

sums of the original variates) are independent of each other. Each factor approximates a normal distribution, thus better satisfying subsequent statistical exercises. Each factor can be interpreted in terms of contributions of the original variates to the score, thus allowing consideration of important determinants of wildlife consumption. In this situation, factors could be attributed to the bodies of water, the type of 7sh, and age of the children. These independent components of the total problem would be useful, then, in developing new hypotheses in situations where little previous data existed. This report illustrates the use of factor analysis in the development of measures of consumption in children and, separately, in their parents. These factors are used to identify potential sources of agreement between consumption patterns and can be valuable in determining the impact of the health advisories in assessing compliance. Rather than examining a multitude of issues that may be associated with consumption behaviors, such as demographic and risk perception variables, this paper focuses speci7cally on the identi7cation of 7sh consumption patterns and their associations between two generations. Brie8y, our 7ndings suggest that sport 7sh consumption patterns in parents are predictive of similar patterns in their children. MATERIALS AND METHODS

Study Participants Established in 1991, the New York Angler State Angler Cohort Study (NYSACS) is a prospective, epidemiologic investigation of the health-related effects of exposure to sport-caught 7sh in licensed anglers residing in 16 counties of upstate New York. Described in detail elsewhere (Vena et al., 1996), the cohort of reproductive-age anglers has been followed to the present and assessed on several different health outcomes under the broad categories of cancer, fertility, and child development. Approximately 30,000 licensed anglers aged 18}40 years were sent self-administered questionnaires requesting information on medical history, occupational and environmental risk perceptions, sport 7sh consumption history, 7sh consumption advisory awareness, and demographic information. A covering letter detailed the voluntary nature of the research and measures taken to ensure con7dentiality. As approved by the University Human Subjects Review Board, return of the questionnaire implied consent. Female partners of the male anglers were asked to complete questions with regard to medical and reproductive history, tobacco and alcohol use, and

SPORT FISH CONSUMPTION PATTERNS IN ANGLER FAMILIES

TABLE 1 Baseline Demographic Information for Anglers, NYSACS 1991}2001 Variable Age, in years

mean 31.6

SDa 5.5

n

%

Sex Male Female Total

10,518 913 11,431

92.0 8.0 100

Race/ethnic group White Nonwhite Missing Total

11,004 227 150 11,431

96.3 2.4 1.3 100

Highest grade of school completed
High school High school graduate Some college College graduate Missing Total

506 4,738 3,586 2,496 105 11,431

4.5 41.4 31.3 21.8 1.0 100

Total household income
15,000 15,000}24,999 25,000}39,999 40,000# Missing Total

956 2,104 3,859 3,972 540 11,431

8.4 18.4 33.8 34.7 4.7 100

a

Standard deviation.

sport 7sh consumption. Partners also reported name, birth date, sex, birth weight, and location of birth for any child born alive between 1986 and the time of the survey. Two follow-up contacts of nonresponders made by mail resulted in an overall response rate of 40%, comprising 10,518 male anglers, 913 female anglers, 6,651 female partners, and 4,226 offspring. Table 1 displays baseline demographic information for anglers. Anglers and partners who noted in the 1991 survey that they had one or more children born between June of 1986 and June of 1991 were contacted again between 1997 and 1998 to respond to a survey with regard to their children. Questions included 7sh consumption history, medical background, behavioral problems, and school performance. At time of survey, 1,692 households, representing 2,158 children between ages 5 and 11 years, were contacted. Of those contacted, 1,112 households (65.7%) representing 1,415 children returned completed surveys. Because multiple children per household were surveyed, only 7sh consumption data from 7rst-born children ages 5 to 10 years (n"564) were analyzed

21

in this report. Focus on the 7rst-born eliminates the statistical dependency among children sampled from the same household. In addition, the consumption patterns in 7rst-born may be different from those in siblings due to changes in family food patterns with time. Among 7rst-born children, 52% were males and 48% were females. Survey of Fish Consumption The survey methodology employed in the NYSACS included in-depth questions with regard to sport 7sh consumption of both male and female anglers, the female spouses or partners of male anglers, and their children using classical dietary assessment methods of standardized food-frequency questionnaires (Willett, 1998). These methods have well-documented reliability. Survey items were designed to characterize not just number of years of 7sh consumption over the lifetime, but also more detailed consumption information for the year preceding time of survey. To this end, anglers were asked to categorize the average number of 7sh meals eaten from New York State waters during the 1990 to 1991 season by quartiles, starting with June to August of 1990. Participants recorded their responses to the number of meals consumed as the following: none, one or less per month, two per month, three per month, one per week, two per week, three to four per week, or 7ve or more per week. Average number of meals of speci7c species of 7sh for 1990 to 1991 were recorded in the same increments for the following 7sh from Lake Ontario and tributaries: channel cat7sh, lake trout, Chinook salmon, coho salmon over 21 inches in length, coho salmon under 21 inches in length, rainbow trout over 25 inches in length, rainbow trout under 25 inches in length, brown trout over 20 inches in length, brown trout under 20 inches in length, carp, white perch, and yellow perch. (Advisory suggestions for certain 7sh, such as trout and salmon, vary depending on size. Therefore, participants are asked about these 7sh by length in separate items.) The same items with regard to type of 7sh from Lake Ontario were repeated for Lake Erie, but also included walleye or pickerel. Female partners of anglers were asked the questions in the same format as anglers, but these included questions only about Lake Ontario and tributaries. In 1997}1998, mothers (85%) and fathers or other members of the family (15%) acted as informant of the child’s 7sh consumption history. The parents of the index child reported the total number of sport 7sh consumed by the child during each year of life

22

BEEHLER ET AL.

for all sport 7sh, Lake Ontario sport 7sh, and Lake Ontario trout and salmon sport 7sh, speci7cally.

counting for 10% of variance or greater, were used as dependent variate-outcomes in each model. In this way, speci7c parental 7sh consumption patterns could be assessed as predictors of children’s consumption.

Patterns of Sport Fish Consumption Sport 7sh consumption patterns were identi7ed by principal components analysis (PCA) in SPSS for Windows, version 10. Standardized consumption variables were entered into three separate models as follows: 29 anglers’ variables as model 1 describing consumption patterns among anglers, 12 partners’ variables as model 2 describing consumption patterns among partners, and 30 childrens’ variables as model 3 describing consumption patterns in the 7rst-born offspring. Factors were rotated with an orthogonal (varimax) rotation to improve interpretability and minimize the correlation between the factors. The number of factors retained from each model was determined by the signi7cance of the eigenvalues, the amount of variance explained by each factor, and the factor interpretability. Labeling of the factors was primarily descriptive and based on interpretation of the pattern structures.

RESULTS

Initial inspection of the child’s 7sh consumption data showed that 60% of children had consumed at least one sport 7sh meal of any type from the ages of 1 to 10 years. Descriptive statistics in Table 2 illustrate that children were consumers of sport 7sh in as early as the 7rst year of life. We report these data by age at consumption for our sample of children (age 5}10 years at time of survey) to point to the early age at which children may become consumers of sport 7sh. However, because of the relatively low levels of consumption and the large percentage of children who never consumed (40%), it is most meaningful to give median and range values for those children who report ever consuming sport 7sh. Table 2 shows that by age 5 years, almost one half of children had consumed a least 1 meal of sport 7sh from any body of water. A median of 2 to 3 sport 7sh meals were reported to be consumed by children for each year of life, among those who reported consumption of any type of sport 7sh. For Lake Ontario sport 7sh consumption, the proportion of consumers drops to 17% by age 5 years, with a median of 1.5 to 2 meals per year. For Lake Ontario trout and meals speci7cally, the proportion of consumers by age

Relationship of Parental to Child Sport Fish Consumption To assess the relationship between parental and child sport 7sh consumption, pattern-speci7c scores derived from PCA were saved and subsequently included in stepwise multiple regression analyses. The 7rst four children’s factors, respectively, each ac-

TABLE 2 Sport Fish Meals Consumed by Anglers’ Children Age 5 to 10 years by Age at Consumption (n"564) Sport 7sh meals (any type)

Age at consumptiona

Number of participants eligible to consume during this yearb

1 2 3 4 5 6 7 8 9 10

564 564 564 564 564 554 432 332 233 124

a

Lake Ontario sport 7sh meals

Percentage of Median (range) Percentage of Median (range) eligible who number of eligible who number of meals consumed meals reported consumed reported by during this by consumers in during this consumers in this year this year of life year year of life 5.3 21.6 34.8 42.0 47.5 43.1 44.0 43.4 39.5 38.7

2.0 2.0 2.0 2.0 3.0 3.0 3.0 3.0 3.0 3.0

(49) (49) (49) (49) (49) (49) (49) (49) (49) (49)

Age at consumption, not age of child at time of survey. Children must have reached this age to report during this year.

b

1.4 7.4 12.1 15.8 17.2 15.5 15.3 17.5 14.6 11.3

1.5 2.0 2.0 2.0 2.0 2.0 2.0 2.0 1.5 1.5

(4) (11) (23) (23) (23) (24) (29) (29) (19) (24)

Lake Ontario trout and Salmon meals Percentage of eligible who consumed during this year 0.5 4.1 6.7 8.7 9.4 8.1 8.6 9.0 8.4 7.3

Median (range) number of meals reported by consumers in this year of life 1.0 2.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 2.0

(;) (7) (17) (5) (9) (7) (5) (3) (3) (3)

SPORT FISH CONSUMPTION PATTERNS IN ANGLER FAMILIES

5 years drops to 9%, with a median of 1 to 2 meals per year. Although we have presented the descriptive statistics for the children in this sample for illustrative purposes, we have purposefully avoided the summation of total sport meals over the lifetime of the child so that, like the parent’s data, the factor analysis technique may appropriately weight the contribution of each variate when considering the context of all variates. As shown in Table 3, analysis of children’s consumption variables by factor analysis yielded factors representing age at consumption, type of 7sh, and whether the 7sh was speci7cally from Lake Ontario. The 7rst factor dealt with sport 7sh meals over the lifetime of the child, age 1 to 10 years, from any waters. The other 7ve factors follow a similar patterning which characterizes childern’s sport 7sh consumption patterns. The factors accounted for about 82% of variance. Table 4 displays the analysis of anglers’ variables which yielded eight signi7cant factors accounting for about 65% of variance. Factors were patterned by 7sh type and by body of water. For example, factor one described a consumption pattern of Lake Ontario trout and salmon of all types, including Chinook salmon, coho salmon
21 inches, coho salmon 21 inches, brown trout
20 inches, brown trout 20 inches, rainbow trout
25 inches, rainbow trout 25 inches, and lake trout. Interpretations of a similar nature were possible for the remaining factors, each identifying a speci7c type of 7sh or water source. Partners’ analysis (Table 5) revealed patterning similar to that of the angler. Though female partners’ questions were limited to consumption history from Lake Ontario, analysis of 7sh type yielded only three factors, accounting for about 59% of variance. Factors dealt with speci7c type of 7sh from Lake Ontario. For example, component one represented trout consumption, regardless of size or type. With patterns of 7sh consumption identi7ed in both parents and children, associations between the two generations were examined by multiple regression analysis. Factor scores for parents’ components were entered stepwise as predictors into multiple linear regression models for each of the 7rst four children’s factors, which acted as dependent measures. The Coef7cient of Determination was used to assess signi7cance of the relation and to estimate the relative importance. Results are shown in Table 6. Children’s consumption of sport 7sh from age 1 to 10 years was the dependent variable for the 7rst model. This factor was best predicted by one

23

factor dealing with the anglers’ consumption of New York State sport 7sh from 1990 to 1991. As seen, the Coef7cient of Determination was 0.15, implying that 15% of the variance in the child’s 7rst factor of consumption was explained by the factor describing anglers’ consumption from 1990 to 1991. For the second regression model, the second children’s factor dealing with consumption of Lake Ontario sport 7sh from age 4 to 10 years acted as the dependent variable. Partners’ consumption of Lake Ontario perch and cat7sh, anglers’ consumption of any New York State sport 7sh, and angler’s consumption of Lake Ontario perch entered the model. (These factors may have been selected because of the popularity of perch for sport 7sh meals). Anglers’ sport 7sh consumption from Lake Erie showed negative standardized beta coef7cients, suggesting the lack of consumption of these 7sh being predictive of consumption of Lake Ontario sport 7sh. Since children of anglers would be consuming the same type of 7sh as their parent(s) who caught the 7sh, a children’s consumption pattern of eating Lake Ontario 7sh would be likely predicted by parents’ consumption of these same 7sh, rather than those from Lake Erie. Children’s factors three and four were representative of consumption of Lake Ontario trout and salmon from, respectively, age 2 to 7 years and from age 7 to 10 years. Though both models selected signi7cant predictors from parents’ factors, model three accounted for about 8% of variance in the child’s factor scores, while model four accounted for only 6% of variance in the child’s factor scores. Regardless of these relatively small correlations, the parents’ factors selected were those concerning Lake Ontario trout and salmon consumption, again pointing to the fact that parent’s consumption is predictive of children’s consumption, and therefore exposure, to contaminated sport 7sh. DISCUSSION

Understanding the exposure of children and families to environmental hazards has become an area of increasing attention. It is essential to examine with precision the nature of childern’s encounters with their environment. One such environmental exposure of concern, which has not been studied extensively in childern, is that of sport 7sh consumption. In the current study, the detailed questions posed in the surveys requesting speci7c information on type of 7sh and body of water lent themselves well to the factor analysis technique. Analysis revealed that type of 7sh and body of water were important

24

TABLE 3 Factor Loadings for Body of Water and Age at Consumption Associated with Fish Meal Consumption Patterns IdentiAed with Factor Analysis for Children Age 5}10 years, NYSACS 1991}2001

Factor 1: sport 7sh meals, age 1}10 years 0.895 0.889 0.884 0.880 0.866 0.854 0.804 0.799 0.745

Factor 3: LOa trout and salmon meals, age 2}7 years

Factor 4: LOa trout and salmon meals, age 7}10 years

Factor 5: LOa sport 7sh meals, age 2}5 years

LO sport 7sh, age 8 LO sport 7sh, age 10 LO sport 7sh, age 9 LO sport 7sh, age 7 LO sport 7sh, age 6 LO sport 7sh, age 5 LO sport 7sh, age 4 Any sport 7sh, age 10 Any sport 7sh, age 9

LO trout and salmon, age 3 LO trout and salmon, age 2 LO trout and salmon, age 6 LO trout and salmon, age 4 LO trout and salmon, age 5 LO sport 7sh, age 3 LO trout and salmon, age 7 LO sport 7sh, age 2 LO sport 7sh, age 5

LO trout and 0.905 salmon, age 9 LO trout and 0.890 salmon, age 8 LO trout and 0.836 salmon, age 10 LO trout and 0.647 salmon, age 7

LO sport age 2 LO sport age 3, LO sport age 4, LO sport age 5,

0.933 0.918 0.906 0.863 0.820 0.620 0.607 0.543 0.430

0.852 0.833 0.822 0.787

7sh, 0.759 7sh

0.720

7sh

0.620

7sh

0.483

Factor 6: LOa sport 7sh meals, age 1 year LO trout and salmon, age 1 LO sport 7sh, age 1

0.772 0.687

BEEHLER ET AL.

Any sport 7sh, age 4 Any sport 7sh, age 2 Any sport 7sh, age 6 Any sport 7sh, age 3 Any sport 7sh, age 5 Any sport 7sh, age 7 Any sport 7sh, age 8 Any sport 7sh, age 1 Any sport 7sh, age 9 Any sport 7sh, age 10

Factor 2: LOa sport 7sh meals, age 4}10 years

0.751 0.586 0.518 0.459 0.426

0.673

Variance 23.9 explained (%)

18.9

15.9

10.2

8.2

4.6

Note. Total variance explained"81.70%; patterns were identi7ed by entering standardized consumption variables into principal components analysis with orthogonal rotation. a Lake Ontario.

TABLE 4 Factor Loadings for Fish Type and Body of Water Associated with Fish Meal Consumption Patterns IdentiAed with Factor Analysis for Anglers, NYSACS 1991}2001

Factor 5: LEb perch, Factor 4: LEb salmon walleye, or pickerel

0.738

LE brown trout 20

0.776

NYS 7sh meals, 0.817 Sep}Nov 1990

LE coho salmon 21

0.851

LE yellow perch

0.724

LE rainbow trout
25

0.772

NYS 7sh meals, 0.798 Mar}May 1991

LE Chinook salmon

0.817

LE walleye 0.739 or pickerel

0.686

LE brown trout
20

0.766

NYS 7sh meals, 0.780 Jun}Aug 1990

LE coho 0.796 salmon
21

0.685

LE rainbow 0.717 trout
25 LE lake trout 0.638

NYS 7sh meals, 0.779 Dec}Feb 1991

Factor 1: LO trout and salmon LO Chinook salmon LO coho salmon, 21 inches LO coho salmon,
21 LO brown trout 20 LO rainbow trout 25 LO rainbow trout
25 LO brown trout
20 LO lake trout

0.680

LE white perch

0.823

Factor 6: LOa perchFactor 7: Carp

Factor 8: Channel cat7sh

LO 0.747 white perch LO 0.688 yellow perch

LE 0.769 channel cat7sh LO 0.761 channel cat7sh

LE carp

0.823

LO carp

0.805

0.642

0.674 0.607 0.603

Variance 13.54 explained (%)

11.47

9.86

7.94

6.55

5.84

5.29

4.38

SPORT FISH CONSUMPTION PATTERNS IN ANGLER FAMILIES

Factor 2: LEb trout

Factor 3: NYSc sport 7sh meals, 1990}1991

a

Note. Total variance explained"64.88%; patterns were identi7ed by entering standardized consumption variables into principal components analysis with orthogonal rotation. a Lake Ontario. b Lake Erie. c New York State.

25

26

BEEHLER ET AL.

TABLE 5 Factor Loadings for Fish Type and Body of Water Associated with Fish Meal Consumption Patterns IdentiAed with Factor Analysis for Anglers’ Female Partners, NYSACS 1991+2001 Factor 1: LOa trout

LO LO LO LO LO

brown trout
20 rainbow trout
25 brown trout 20 rainbow trout 25 lake trout

Factor 2: LOa salmon

0.821 0.778 0.762 0.701 0.665

Factor 3: LOa perch or cat7sh

LO coho salmon 21 LO coho salmon
21 LO chinook salmon

Variance explained (%) 25.69

0.849 0.781 0<694

LO white perch LO yellow perch LO channel cat7sh

17.92

0.815 0.742 0.564

15.29

Note. Total variance explained"58.90%; patterns were identi7ed by entering standardized consumption variables into principal components analysis with orthogonal rotation. a Lake Ontario.

aspects in appropriate weighting of variables to reveal underlying patterns of 7sh consumption. Anglers, anglers’ female partners, and children showed differentiation of factors based on type of 7sh in particular. For example, in anglers, the 7rst factor or pattern, characterized as Lake Ontario trout and salmon consumption, illustrates a particular weighting of variates which emphasizes different types and sizes of trout (rainbow and brown) and salmon (coho and Chinook) which are both conceptually distinct and statistically independent from angler’s patterns of Lake Erie trout (rainbow and brown) consumption and Lake Erie salmon (coho and Chinook) consumption. Similar patterning was repeated throughout the sample of anglers’ partners and childern. The use of highly detailed consumption questions and the

factor analysis technique created different combinations of the variates, re8ecting behaviors that would not be as evident using other statistical approaches. In addition to better representation of the individual’s behavior and group tendency, the question set may be made more detailed, enabling the respondent to more accurately report behaviors with regard to the foods consumed and when. Dietary patterns derived from the use of factor analysis have been found to be reproducible (Hu et al., 1999a). Reproducibility of identi7ed patterns was also a goal of the current study. To investigate this, the angler sample was initially divided into two random samples. Analysis revealed similar patterns based on type of 7sh and body of water in these two independent subsamples (results not shown).

TABLE 6 Multiple Regression Analysis of Parents’ Consumption Patterns on Children’s Consumption Patterns Regression model



Child’s factor

Parents’ factor(s)a

1

Total meals, age 1}10

Anglers NYSa meals in last year, 1990}1991

0.39

0.15

2

LOb meals, age 4}10

Partner’s LO perch/cat7sh Angler’s NYS meals in last year, 1990}1991 Angler’s LO perch Angler’s LEc perch, walleye, pickerel Angler’s LE trout

0.30 0.22 0.12 !0.10 !0.09

0.20

3

LO trout and salmon, age 2}7

Angler’s LO trout or salmon Partner’s LO salmon Partner’s LO trout

0.24 0.14 !0.13

0.08

4

LO trout and salmon, age 7}10

Partner’s LO salmon

0.25

0.06

Note. Parents’ factors were selected by a stepwise model. a New York State. b Lake Ontario. c Lake Erie.

R2

SPORT FISH CONSUMPTION PATTERNS IN ANGLER FAMILIES

Analysis was then repeated in the entire sample of anglers (as shown in Table 4). This process was repeated for the anglers’ female partner group, providing similar results and suggesting that 7sh consumption patterns were reproducible. How do these 7ndings impact environmental exposure assessment? The identi7cation of 7sh consumption patterns points to the fact that sport 7sh should not be treated as a single, uni7ed exposure. Sport 7sh consumption advisories give detailed information with regard to how much 7sh can safely be consumed because exposure varies by type of 7sh, size of 7sh, and body of water. The above analyses have detected that speci7c types of consumption patterns do exist. For example, in anglers, it was found that three of the four factors that account for the most variance among all consumption variables can be described as patterns of trout and salmon consumption. More speci7cally, a pattern of consumption of both trout and salmon from Lake Ontario accounted for the greatest single proportion of variance among the factors. This is an important 7nding given that the New York State 7shing advisory recommends particularly stringent consumption guidelines for trout and salmon from Lake Ontario (only one 7sh per month, or eat none) compared to Lake Erie (eat one 7sh per week). This 7nding points to the fact that Lake Ontario trout and salmon, which may be some of the larger and more contaminated species available in the area, are accounting for signi7cant patterns of consumption in anglers. Previous studies have shown that body burden of environmental toxicants can be related to sport 7sh consumption (Cordle, et al., 1982; Fiore et al., 1989; Kearney et al., 1999; Kosatsky et al., 1999; Kuwabara et al., 1997; Sjo din et al., 2000; Sonzogni et al., 1991), but have typically relied on a lifetime type of question posed with regard to consumption or on some form of diary. This study suggests that a compromise between the daily diary and the lifetime question is possible by focusing on short time periods, by focusing on speci7c types of 7sh in those time periods, and by focusing on speci7c waters. In other investigations, factor analysis-derived dietary patterns have been shown to be correlated with biological markers of exposure, such as serum vitamin levels (Fung et al., 2001; Hu et al., 1999b), providing further evidence that these patterns can provide meaningful exposure characterization. How the consumption patterns for the current study relate to biological markers of exposure, such as serum PCB level, warrants further investigation.

27

The multiple regression analysis conducted here illustrates the link between diet in children and parents with speci7c patterns of consumption in parents correlating with logically similar patterns in childern. The ability of parents’ diet to predict that of their children may be particularly true for consumption of sport 7sh because it is highly unlikely that this food would be available from alternate sources such as schools or daycare. Sport 7sh likely represents a unique food, meaning that estimates of consumption may be more accurate. This study seems to suggest that parents can provide detailed information about their children’s diet through this form of speci7c time and food questoning. Alternatively, parents might be more likely to report childern’s consumption simply as a function of their own. Because of this linkage of parent and child consumption, it is important to consider how risk communication messages are constructed. The children in this study are not consuming sport 7sh randomly; they are consuming in such a manner that logical patterns can be discerned. The questions posed to parents with regard to their children’s consumption were based largely on that of Lake Ontario 7sh. The most recent edition of the NYSDOH sport 7sh consumption advisories (NYSDOH, 2001) states that children under the age of 15 years should eat no 7sh from this body of water or its tributaries. Clearly, the message is not being received. In-depth studies of why sport 7sh consumption risk messages are not fully effective are generally lacking. However, a recent focus group investigation of an African American community of Buffalo, New York (Beehler et al., 2001) has suggested that risk messages posed to anglers as formal, written advisories distributed with 7shing licenses not only miss those anglers who are not licensed, but also present scienti7cally based information which anglers do not actively incorporate into their existing lay-notions of contamination and risk. If anglers are either unaware of the advisories, are not persuaded by them, or do not perceive sport 7sh consumption to be risky behavior, then they may not consider it to be a risk to their childern. This investigation has illustrated that childern and families of recreational anglers are consuming sport 7sh in patterns that con8ict with present New York State recommendations. Of particular concern are childern’s patterns of Lake Ontario sport 7sh consumption, the linkage of childern’s patterns to those of their parents, and the role of parents in providing these contaminated 7sh to their childern. It is important to consider these 7ndings when devising future risk communication messages and risk reduction strategies. Without revised advisories or

28

BEEHLER ET AL.

enhanced public health risk communications, consumption of contaminated sport 7sh remains a potentially adverse;yet fully preventable;childhood exposure. ACKNOWLEDGMENTS This project received approval from the University at Buffalo Human Subjects Review Board. All related research activities followed institutional and national guidelines for the protection of human subjects. The authors thank the families who partcipated in the study. This project was supported by grants from the Agency for Toxic Substances and Disease Registry (H75/ ATH298328) and the National Institute of Environmental Health Sciences, NIH (R03ES08500-01).

REFERENCES Anderson, H., Falk, C., Hanrahan, L., Olson, J., Burse, V., Needham, L., Paschal, D., Patterson, D., Jr. and Hill, R., Jr., and The Great Lakes Consortium. (1998) Pro7les of Great Lakes critical pollutants: A sentinel analysis of human blood and urine. Environ. Health Perspect 106, 279}289. Beehler, G. P., McGuinness, B. M., and Vena, J. E. (2001). Polluted 7sh, sources of knowledge, and the perception of risk: Contextualizing African American anglers’ sport-7shing practices. Human Organization 60, 288}297. Burger, J. (1998). Fishing and risk along the Savannah River: Possible intervention. J. Toxicol. Environ. Health Part A 55, 405}419. Burger, J., P8ugh K., Lurig, L., Von Hagen, A., and Von Hagen, S. (1999a). Fishing in urban New Jersey: Ethnicity affects information sources, perception, and compliance. Risk Anal. 19, 217}229. Burger, J., Stephens, W., Boring, C., Kuklinski M., Gibbons, J., and Gochfeld, M. (1996). Factors in exposure assessment: Ethnic and socioeconomic differences in 7shing and consumption of 7sh caught along the Savannah River. Risk Anal. 19, 427}438. Cordle, F., Locke, R., and Springer, J. (1982). Risk assessment in a federal regulatory agency: An assessment of risk associated with the human consumption of some species of 7sh contaminated with polychlorinated biphenyls (PCBs). Environ. Health Perspect. 43, 171}182. Courval, J., DeHoog, J., Holzman, C., Tay, E., Fischer, L., Humphrey, H., Paneth N., and Sweeney, A. (1996). Fish consumption and other characteristics of reproductive-aged Michigan anglers;A potential population for studying the effects of consumption of Great Lakes 7sh on reproductive health. Toxicol. Ind. Health. 12, 347}359. Falk, C., Hanrahan, L., Anderson, H., Kanarek, M., Draheim, L., Needham, L., Patterson, D., Jr. and the Great Lakes Consortium. (1999). Body burden levels of dioxin, furans, and PCBs among frequent consumers of Great Lakes sport 7sh. Environ. Res. 80, S19-S25. Fiore, B., Anderson, H., Hanrahan, L., Olson, L., and Sonzogni, W. (1989). Sport 7sh consumption and body burden levels

of chlorinated hydrocarbons: A study of Wisconsin anglers. Arch. Environ. Health. 44, 82}88. Fung, T. T., Rimm, E. B., Spiegelman, D., Rifai, N., To8er, G. H., Willett, C., and Hu., F. B. (2001). Association between dietary patterns and plasma biomarkers of obesity and cardiovascular disease risk. Am. J. Clin. Nutr. 73, 61}67. Hu, F. B., Rimm, E. B., Stampfer M. J., Ascherio, A., Spiegelman, D., and Willett W. C. (1999a). A prospective study of major dietary patterns and risk of coronary heart disease in men. Am. J. Epidemiol. 149, s49. Hu, F. B., Smith-Warner, S. A., Feskanich, D., Stampfer, M. J., Ascherio, A., Sampson, L., and Willett, W. (1996b). Reproducibility and validity of dietary patterns assessed with a foodfrequency questionnaire. Am. J. Clin. Nutr. 69, 243}249. Kearney, J., Cole, D., Ferron, L., and Weber, J. (1999). Blood PCB, p,p-DDE, and mirex levels in Great Lakes 7sh and waterfowl consumers in two Ontario communities. Environ. Res. 80, S138}S149. Kosatsky, T., Przybysz, R., Shantenstein, B., Weber, J. P., and Armstrong, B. (1999). Fish consumption and contaminant exposure among Montreal-area sport7shers: Pilot study. Environ. Res. 80, S150}S158. Kostyniak, P., Stinson, C., Greizerstein, H., Vena, J., Buck, G., and Mendola, P. (1999). Relation of Lake Ontario 7sh consumption, lifetime lactation, and parity to breast milk polychlorobiphenyl and pesticide concentrations. Environ. Res. 80, S166}S174. Kuwabara, K., Yakushiji, T., Watanabe, I., Yoshida, S., Yoyama, K., and Kunita, N. (1979). Increase in the human blood PCB levels promptly following ingestion of 7sh containing PCBs. Bull. Environ. Contam. Toxicol. 21, 273}278. New York State Department of Health, Division of Environmental Health Assessment. (2001). ‘‘Chemicals in Sport7sh and Game Health Advisories 2001}2002.’’ P8ugh, K., Lurig, L., Von Hagen, L., Von Hagen, S., and Burger, J. (1999). Urban anglers’ perception of risk from contaminated 7sh. Sci. Total Environ. 228, 203}218. SjoH Hdin, A., Hagmar, L., Klasson-Wehler, E., Bjork, J., and Bergman, A. (2000). In8uence of the consumption of fatty Baltic sea 7sh on plasma levels of halogenated environmental contaminants in Latvian and Swedish men. Environ. Health Perspect. 108, 1035}1041. Sonzogni, W., Maack, L., Gibson, T., Degenhardt, D., Anderson, H., and Fiore, B. (1991). Polychlorinated biphenyl congeners in blood of Wisconsin sport 7sh consumers. Arch. Environ. Contam. Toxicol. 20, 56}60. Tilden, J., Hanrahan, L., Anderson, H., Palit C., Olson, J., Mac Kenzie, W., and the Great Lakes Sport Fish Consortium. (1997). Health advisories for consumers of Great Lakes sport 7sh: Is the message being received? Environ. Health Perspect. 105, 1360}1365. Vena, J., Buck, G., Kostyniak, P., Mendola, P., Fitzgerald, E., Sever L., Freudenheim, J., Greizerstein, H., Zielenzy, M., McReynolds, J., and Olson, J. (1996). The New York Angler Cohort Study: Exposure characterization and reproductive and developmental health. Toxicol. Ind. Health. 12, 327}334. Willett, W. C. (1998). ‘‘Nutritional Epidemiology,’’ 2nd ed. Oxford Univ. Press, Oxford.