Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 250
Toxicological Effects of Methylmercury 6 COMPARISON OF STUDIES FOR USE IN RISK ASSESSMENT UNTIL recently, the data base available for risk assessments of MeHg has been limited to high-dose poisoning episodes in Japan and Iraq. More recently, however, epidemiological studies have been conducted on the health effects of exposure to low doses of MeHg (for details of health effects, see Chapter 5). The low-dose MeHg exposure studies are more relevant to levels of exposures in the United States and, therefore, more appropriate for use in risk assessments. The two largest and most comprehensive studies to address the health effects of MeHg — the Seychelles Child Development Study (SCDS) and the Faroe Islands studies — reached different conclusions. A range of adverse neuropsychological and neurophysiological outcomes were found to be associated with prenatal Hg exposure in the Faroe Islands study, whereas adverse effects were not found in the main Seychelles study. This chapter compares those two studies, as well as data from the pilot phase of the SCDS, and a smaller study carried out on a cohort in New Zealand. MeHg exposure in the SCDS and Faroe Islands studies were similar; the arithmetic mean maternal hair Hg concentration in the Seychelles cohort (6.8 µg/g) was slightly higher than the geometric mean reported in the Faroe cohort (4.3 µg/g). Several differences in research design and cohort characteristics have been identified that might account for the discrepant findings. Some of those explanations seem less persuasive, however, when the data from the New Zealand study are consid-
OCR for page 251
Toxicological Effects of Methylmercury ered. That study found associations with MeHg exposure in a population whose sources of MeHg exposure were similar to those in the Seychelles and used end points similar to those examined in the Seychelles. Although the New Zealand data have been available for some time, they have not been used extensively for risk assessment, possibly because until recently, they had not been subjected to standard peer-review procedures. A re-analysis of the New Zealand data by Crump et al. (1998), which underwent peer review, reported associations of prenatal MeHg exposure with several end points (when one extreme outlier was excluded), including four that were not found to be related to MeHg in the Seychelles study. The New Zealand study has been criticized for errors in matching exposed children to controls and for testing exposed children and controls at different ages (Myers et al. 1998). Those errors occurred in the 4-year follow-up but were corrected in the 6-year follow-up, which is the data set reviewed in this section. In addition, there is no information that would suggest the presence of differential measurement error across the studies. Any error of that type is likely to be nondifferential (i.e., unbiased), and it would reduce the likelihood of detecting associations between MeHg exposure and neurobehavioral test scores. Data from the peer-reviewed pilot SCDS of 217 children assessed at 5.5 years (Myers et al. 1995) are also considered in this chapter. (Note that the nonstandard treatment of the data from the Revised Denver Developmental Screening Test (DDST-R) discussed in Chapter 5 was not an issue in the 5.5-year follow-up since the DDST-R was not given at that age.) Two of the four outcomes that were tested in both the pilot and the main Seychelles studies at 5.5 years of age were found to be associated with prenatal Hg exposure in the pilot study. The Seychelles investigators were cautious about drawing inferences from their pilot data, because the effects were substantially weaker when four outliers were excluded from the analyses and because socioenvironmental influences were not adequately assessed and controlled statistically. It is not clear, however, that it is appropriate statistically to exclude influential data points; many statisticians would instead recommend the use of data transformation to reduce their influence. Exclusion is appropriate only where a value appears biologically implausible (see discussion of the New Zealand outlier in the Benchmark Analysis section in Chapter 7). With regard to socioenvironmental influences, T.W. Clarkson
OCR for page 252
Toxicological Effects of Methylmercury (principal investigator in the SCDS, personal commun., January 20, 2000) indicated to the committee that the most heavily contaminated fish consumed in the Seychelles islands — swordfish, shark, and tuna — tend to be among the most expensive fish, so that, if anything, exposure levels might be higher among mothers with higher socioeconomic status. There is, therefore, no reason to expect a confounding of exposure with lower socioeconomic status, and low socioeconomic status is not likely to explain the association of Hg exposure with adverse development outcomes in the Seychelles pilot study. ASSESSMENT OF PRENATAL Hg EXPOSURE: CORD BLOOD VERSUS MATERNAL HAIR AND TIMING OF EXPOSURE The principal measure of prenatal exposure in the Faroe study was Hg in cord blood; in the Seychelles, it was Hg in maternal hair. The Faroe investigators also analyzed maternal-hair samples, but no cord-blood specimens were obtained in the Seychelles. In a recently published analysis, the Faroe investigators compared the relation of the cord-blood and maternal-hair Hg measures with their 7-year end points (Grandjean et al. 1999). As shown in Table 6-1, cord-blood Hg concentration was significantly associated with a slightly larger number of end points than maternal-hair Hg concentration, and in most cases the associations were slightly stronger. For various pharmacokinetic and neurodevelopmental reasons, cord-blood measurements might be more sensitive indicators of the neurodevelopmental effects of MeHg. However, given that hair Hg concentrations in the Faroe Islands study were only a slightly weaker predictor of Hg effects than cord blood, it would be reasonable to expect that, if children were affected in the main Seychelles study, some indication of an association between child performance and maternal-hair Hg concentration would be apparent in that study. With the possible exception of the Bender Gestalt scores for boys, there is no indication of even a trend in the predicted direction in the data published to date from the main SCDS (e.g., see Figures 5-7 and 5-8). It should be noted that the maternal-hair samples obtained in the Faroe and Seychelles studies did not necessarily reflect exactly the same period of pregnancy. In part, this is because the Seychelles study ob-
OCR for page 253
Toxicological Effects of Methylmercury TABLE 6-1 Change in Neuropsychological Test Performance Associated with a Doubling of the Hg Concentration in the Faroese Cohorta Cord Blood Hg Maternal Hair Hg Test Changeb pc Changeb pc NESd Finger Tapping Preferred hand −5.37 0.049 −5.99 0.039 Other hand −1.97 0.460 −4.40 0.120 Both hands −4.11 0.136 −6.64 0.024 NES Hand-Eye Coordination Error score 3.70 0.187 5.40 0.070 NES Continuous Performance Test Missed responses 10.08 0.024 5.14 0.241 Reaction Time 15.93 <0.001 8.99 0.035 Wechsler Intelligence Scale for Children-Revised Digit Spans −5.62 0.049 −4.39 0.147 Similarities −0.37 0.902 −2.07 0.525 Block Designs −4.36 0.109 −2.86 0.322 Bender Visual Motor Gestalt Test Error on copying 3.83 0.154 3.60 0.208 Delayed recall −4.64 0.104 −1.26 0.679 Boston Naming Test No cues −9.75 <0.001 −6.98 0.016 With cues −10.47 <0.001 −7.47 0.009 California Verbal Learning Test (Children) Learning −4.33 0.123 −3.96 0.184 Immediate recall −6.64 0.019 −5.93 0.049 Delayed recall −5.69 0.047 −5.15 0.092 Recognition −4.24 0.151 −3.15 0.318 Source: Adapted from Grandjean et al., 1999, Table 2. aAll Hg values were log transformed for the analyses presented here. bExpressed as % of standard deviation. cStatistical significance of the effect associated with Hg exposure in a mutiple regression including potential confounders. dNeurobehavioral Evaluation System.
OCR for page 254
Toxicological Effects of Methylmercury tained Hg concentrations from a 9-cm length of hair reflecting average MeHg exposure during pregnancy and the Faroe study obtained concentrations from hair samples of variable length, some of 3-cm (reflecting late second and third trimester) and some of 9-cm in length. Additionally, the Faroe maternal-hair samples, which were obtained at delivery, did not include the last 3 weeks of gestation, because it takes approximately 20 days after ingestion for the newly formed portion of the hair strand to emerge above the scalp. If the third trimester is particularly important for the development of the neural substrate for cognitive and neuromotor function, it is perhaps not surprising that the maternal-hair sample obtained in the Faroe Islands might be somewhat less sensitive than the cord-blood sample, which primarily reflects third-trimester exposure (see Chapter 3). A maternal-hair sample that reflected the last 20 days of pregnancy might have been more sensitive. In the SCDS, the maternal-hair samples were obtained at delivery and at 6 months postpartum. The portion of the hair strands corresponding to the pregnancy period was analyzed, assuming 1 cm of hair growth per month. If third-trimester exposure is critical for the neurodevelopmental end points, the SCDS measure of Hg exposure averaged across the entire pregnancy might be less sensitive in detecting them, compared with cord blood, which primarily reflects third-trimester exposure. It might be informative for the SCDS group to re-analyze their data using the concentration of Hg in the portions of hair corresponding only to the third trimester as the exposure measure. It is also of interest that the neurophysiological end points in the Faroe study (e.g., brain-stem auditory-evoked potentials) were associated only with the maternal-hair Hg measure, not with the cord-blood Hg. Because the hair measure presumably reflects an earlier period of gestation, the data suggest an earlier sensitive or critical period for the neurophysiological effects. In summary, it does not appear that the failure of the SCDS to collect cord-blood Hg samples can account for the discrepancies between their findings and those in the Faroe study because, in the latter study, associations were found between neurobehavioral test scores and both cord-blood Hg and maternal-hair Hg concentrations (Table 6-1). Moreover, the findings reported in New Zealand and the pilot SCDS were based solely on maternal-hair-sample data averaged across the entire pregnancy.
OCR for page 255
Toxicological Effects of Methylmercury DIFFERENCES IN THE NEUROBEHAVIORAL END POINTS ASSESSED AND THE CHILDREN'S AGES AT ASSESSMENT The Faroe and Seychelles studies used very different neurobehavioral test batteries. For the most part, the tests selected for the SCDS are considered apical or omnibus tests (e.g., the McCarthy Scales of Children's Abilities), which yield global scores that integrate performance over many separate neuropsychological domains. In contrast, because the Faroe investigators hypothesized multifocal domain-specific neuropsychological effects, their test battery largely consisted of highly focused tests selected from those commonly used in clinical neuropsychology (e.g., California Verbal Learning Test — Children and Boston Naming Test). The Faroe test battery does not include an apical test of global function. The subscales from the McCarthy test (verbal, perceptual-performance, quantitative, memory, and motor) that assess specific domains of function might be expected to be more directly comparable to the tests administered in the Faroe Islands. For instance, given the finding in the Faroe study that memory, as assessed by the California Verbal Learning Test, was significantly associated with prenatal Hg exposure, it would be expected that children's scores on the McCarthy memory scale in the SCDS would be associated with Hg exposure. However, they were not. In fact, prenatal Hg exposure was not associated with scores on any of the McCarthy subscales. It is important to examine in detail the extent to which the individual McCarthy subscales are comparable to the domain-specific tests selected for the Faroe study. Psychometrically, they are different. The California Verbal Learning Test, for example, involves five learning trials of a 12-word list, with free- and cued-recall trials following short and long delays, and a recognition trial. None of the 18 tests that contribute to scores on the McCarthy scales examine rate of learning, and the memory scale combines scores on four tests that involve recall of differing types of information: pictorial (six common objects arrayed on a page), auditory sequence (xylophone notes), word list (ranging from 3 words in a specified sequence to a 13-word sentence with 9 key words that are scored), connected discourse (recall of individual story elements), and numbers (forward and backward recall of strings of numbers up to seven digits long.) Clearly, a child's score on
OCR for page 256
Toxicological Effects of Methylmercury the McCarthy memory scale integrates performance on a much wider variety of memory skills than does either the short- or long-delay free-recall trials of the California Verbal Learning Test. Scores on some of the 18 specific subscales of the McCarthy test might offer greater comparability with the key end points of the California Verbal Learning Test assessed in the Faroe study. Each of the 18 subscales is quite brief, however, and thus less psychometrically sound than the richer California Verbal Learning Test, which assesses only one domain of function but does that in considerable depth. Similarly, although the Boston Naming Test, which was included in the Faroe Islands test battery, and the preschool language scale and the verbal scale of the McCarthy verbal scale which were included in the SCDS 66-month test battery of the SCDS can be considered tests of language skills, the specific skills they assess are quite different. The Boston Naming Test specifically assesses confrontational naming skills, consisting of line drawings of common objects that a child has to name under time pressure (20 seconds). If the child cannot retrieve the correct name spontaneously, semantic and then phonemic cues are provided. In contrast, the total score on the preschool language scale (PLS) integrates a child's performance on the auditory comprehension and expressive communication subscales, both of which assess a broad range of language skills (eg., comprehension and production of vocabulary; concepts of quantity, quality, space, and time; morphology; syntax; and inference drawing). Like the total score on the PLS, the total score on the McCarthy verbal scale integrates a child' s performance across many language-relevant domains in the following tests: pictorial memory (same as test described for memory scale), word knowledge (pointing to the picture of an object named by the examiner, providing the name for four pictured objects, and providing word definitions), verbal memory (same as test described for memory scale ), verbal fluency (generating words in 20-sec trials to fit specific semantic constraints, such as things to eat or animals), and opposite analogies (providing antonyms). Thus, although the four items of the word-knowledge test that assess naming could be isolated and considered an index of confrontational naming, similar to the Boston Naming Test, the four items are unlikely to possess the same sensitivity insofar as the latter test consists of 60 items. Thus, although the Faroe Islands and SCDS test batteries include tests of language and memory, it is not appropriate to view the end points
OCR for page 257
Toxicological Effects of Methylmercury used in the studies to assess each domain to be equivalent either in terms of the specific skills assessed or the test sensitivity. Although the Bender-Gestalt Test was administered in the Faroe study and SCDS as a measure of visual-spatial abilities, different scoring systems were used (the Gottingen system in the Faroe Islands and the Koppitz system in the Seychelles). The finding of a significant association with Hg in the former but not the latter study is similar to the finding reported by Trillingsgaard et al. (1985) that scores derived using the more-detailed Gottingen system were significantly associated with low-dose lead exposure, and scores on the Koppitz system were not. Thus, the Gottingen system used in the Faroe Islands might be more sensitive. Although the Seychelles data could be rescored using the Gottingen system, the committee was told that the data might still not be comparable, because the more sensitive memory for design conditions was not administered in the Seychelles study. To help determine the degree to which the discrepant results from the Faroe study and SCDS are attributable to differences in the neurobehavioral tests used, the Seychelles group administered several of the more domain-specific tests from the Faroe battery in their 8-year followup. The results of those assessments, however, are not yet available. A second important difference in the assessment batteries used in the Faroe study and SCDS relates to the age of assessment — 7 years in the Faroe Islands and 5.5 years of age in the SCDS. The final assessment in the New Zealand cohort was at 6 years of age. Generally speaking, developmental assessments are likely to be less sensitive in detecting subtle neurotoxic effects when they are administered during a period of rapid developmental change. The period covering ages 60-72 months, when the SCDS and New Zealand cohorts were evaluated, is one such period during which marked individual differences in the rate of cognitive maturation are likely to eclipse subtle differences in function attributable to a teratogenic exposure (Jacobson and Jacobson 1991). The assessments performed in the SCDS during infancy, particularly the 19-and 29-month Bayley scales, were also not administered at optimal age points. Studies of prenatal exposure to alcohol and other substances that have administered the Bayley scales at multiple ages have repeatedly failed to detect effects at 18 months, probably because it too is a period of rapid cognitive maturation, involving the emergence of spoken language. Twenty-nine months is likely to be an insensitive testing
OCR for page 258
Toxicological Effects of Methylmercury point for the Bayley scales because it is at the end of the age range for which the version of this test used in the Seychelles was standardized, leading to a substantial risk of a “ceiling effect” (i.e., too many children receiving the highest possible scores on numerous items). The next round of testing in the Seychelles will be at 8 years of age, a point in development that should be more optimal for detecting neurodevelopmental effects. Although differences in end points assessed and age of assessment might explain the failure of the SCDS to detect the associations found in the Faroe Islands study, findings from the New Zealand study and the Seychelles pilot study suggest that the discrepancies between the Faroe Islands and the main Seychelles studies are probably not due to differences in the assessments. The New Zealand study found associations between MeHg exposure and scores on the McCarthy Scales of Children's Abilities (the primary outcome measure used in the SCDS) at about the same age of assessment as in the Seychelles study, in a study with full control for potential confounding influences. Associations with prenatal Hg exposure were even seen on the McCarthy scales and the PLS in the 217-member Seychelles pilot study at 5.5 years of age, albeit with only limited control for socioenvironmental influences. STABLE VERSUS EPISODIC PATTERN OF EXPOSURE The predominant source of Hg exposure in the Seychelles is daily fish consumption. Maternal fish consumption averages 12 meals per week. Hg exposure in the Faroe Islands, by contrast, is often more episodic. In the Faroe Islands, pilot-whale meals are relatively infrequent (less than once per month on the average), but whale meat has concentrations of MeHg between 10 and 20 times greater than those in many fish consumed in the Faroe Islands (Grandjean et al. 1992); thus, the whale meals might represent toxicologically more significant peak or bolus doses. Laboratory animal experiments on prenatal alcohol exposure have demonstrated that maternal ingestion of a given dose of alcohol over a short time causes greater neuronal (Bonthius and West 1990) and behavioral impairment (Goodlett et al. 1987) than that caused by gradual ingestion of the same total dose over several days. Thus, it is possible that the more episodic exposure pattern in the Faroe Islands, with
OCR for page 259
Toxicological Effects of Methylmercury heavier doses per occasion, has a more adverse impact on neuronal development than the more gradual exposure in the Seychelles. However, it is difficult to compare the 12 fish meals per week reported in the Seychelles with the three fish “dinners” per week in the Faroe Islands, because the types of fish eaten and their Hg concentrations are different. Moreover, the exposure-associated differences in neurobehavior found in the New Zealand cohort and the Seychelles pilot study where no whale meat was eaten suggest that bolus doses are not necessary to generate cognitive deficits at those levels of exposure. The importance of high episodic (“spiking”) exposures is unclear. However, as discussed in Chapter 4, the degree of spiking in the Faroe study is likely to be in the low-to-moderate range (i.e., less than a doubling in hair Hg concentrations, assuming an individual at the Faroe Islands median exposure level consumes three consecutive 4-ounce whale meals). Spiking might also occur in the Seychelles given the availability of fish species with characteristically moderate-to-high concentrations of Hg (e.g. tuna), although the absence of dietary data does not allow this issue to be examined further. STUDY DIFFERENCES IN CONTROL FOR CONFOUNDERS A potential confounder is a variable related to both the exposure of interest (e.g., MeHg) and to the outcome of interest (e.g., neurobehavior). If the relation between exposure and outcome is no longer significant after controlling statistically for a potential confounder, it is inferred that the relation between exposure and outcome is spurious and due to confounding by the control variable being examined. Because random assignment to predetermined exposure levels cannot be used to control for confounding in human exposure studies, it is important to assess whether a broad range of control variables confound any associations observed between exposure and outcome. Table 6-2 lists the control variables examined in the Faroe and Seychelles studies. Both studies evaluated most of the variables that are known to be at least moderately related to childhood cognitive outcome, including maternal cognitive competence (e.g. Ravens test), child age, gender, maternal alcohol consumption and smoking during pregnancy, and parental income. A few variables that are sometimes modestly related to those outcomes
OCR for page 260
Toxicological Effects of Methylmercury TABLE 6-2 Control Variables Assessed in the Faroe Islands and the Seychelles Studies Covariates Faroe Islands Study Seychelles Study Birth weight X Maternal cognitive competence (Ravens) X X Child's age X --a Child's sex X X Gestational age X Smoking during pregnancy X X Alcohol during pregnancy X X Duration breast feeding X X PCBs X --b Education (mother and father) X Employment (family income) X X Obstetric care X Daycare X X Computer acquaintance X Examiner X Birth order X Maternal age X Child's medical history X Language at home N/A X Maternal medical history X Maternal hair lead Xc Xd aTest scores adjusted for age, based on U.S. norms. bNo PCBs were detected in any of the 49 serum samples obtained at 66 months postpartum. These data support the assumption that there is virtually no PCB exposure in this population. cWhole pregnancy. dAt parturition, i.e., pregnancy minus 45 days. were assessed in one study but not in the other (e.g., maternal age and birth order in the Seychelles study; obstetrical care in the Faroe study). However, the influences of those variables on cognitive outcome are
OCR for page 261
Toxicological Effects of Methylmercury probably too weak to account for any major inconsistencies between the two studies. Parental education was not assessed in the Seychelles study, but it is likely to have added little information over and above family income and maternal cognitive competence. Source: Adapted from NIEHS 1998. At a workshop sponsored by the White House Office of Science and Technology Policy in November 1998, the Faroe investigators noted that, apparently due to social-class differences, the maternal Ravens scores and the child verbal-test scores were generally higher among families residing in one of the three Faroe towns than among those living in the countryside. Because more fish and whale meat are consumed by rural residents, the associations of Hg exposure with child verbal-test scores could be spurious, reflecting those social-class differences. (Although the Ravens scores were controlled statistically in the analyses, that single test might not have fully controlled for social-class confounding.) Data presented at the workshop showed, however, that these associations remained significant, even after controlling for a dichotomous town-country control variable (Table 6-3). Although that analysis is reassuring, it would not be appropriate to control for town routinely in all analyses. Because fish and whale consumption constitute a large proportion of the rural diet, the disappearance of associations after controlling for residence could be due to the fact that residing in a rural area leads to increased Hg exposure which, in turn, causes an adverse outcome. It would not necessarily indicate that the lower social class associated with rural residence is the true cause of the Hg-associated deficit. The disappearance of an association between Hg and neurobehavior under those circumstances would be very difficult to interpret, because the interpretation would depend upon what condition is considered the reason for the association between living in a rural area and poor outcome (i.e., lower social class or greater Hg exposure). Because the rural residents had to travel relatively long distances to the testing site in Torshavn, there has been concern that fatigue and the strangeness of the urban setting might have caused the rural children to perform more poorly. As noted above, however, the data in Table 6-3 make clear that the regression coefficients for prenatal Hg exposure remain significant even after controlling for child's residence. The neuropsychological test examiner is one potentially important factor that was routinely controlled for in the Faroe Islands study (see NIEHS 1998, section 3.5), but was not controlled for in the SCDS. The
OCR for page 262
Toxicological Effects of Methylmercury TABLE 6-3 Effects of Residence (Town vs. Country) and Prenatal Hg Exposure on Developmental Outcomes in the Faroe Islands Study Residencea Hg Without Controlling for Residenceb Hg Controlling for Residencec Test bd p b p b p NES Finger Tapping Preferred Hand 0.03 0.95 −1.10 0.05 −1.14 0.04 Other Hand −0.47 0.26 −0.39 0.46 −0.55 0.31 Both Hands −1.41 0.11 −1.67 0.14 −2.04 0.07 NES Hand-Eye Coordination Error Score 0.001 0.94 0.03 0.19 0.04 0.18 NES Continuous Performance Missed Responses −0.12 0.16 0.27 0.02 0.24 0.04 Reaction Time −14.55 0.06 40.30 <0.001 35.34 0.002 WISC-R Digit Spans 0.06 0.58 −0.27 0.05 −0.26 0.07 Similarities 0.04 0.91 −0.05 0.90 −0.08 0.84 Block Designs 0.19 0.02 −0.17 0.11 −0.12 0.26 Bender Visual Motor Gestalt Test Error on Copying −1.03 0.005 0.67 0.15 0.45 0.35 Delayed Recall 0.35 0.004 −0.25 0.10 −0.17 0.28 Boston Naming Test No cues 1.10 0.005 −1.77 <0.001 −1.51 0.003 With cues 1.24 0.001 −1.91 <0.001 −1.60 0.001 CVLT Learning 0.89 0.16 −1.25 0.12 −1.10 0.18 Immediate recall 0.37 0.06 −0.57 0.02 −0.49 0.05 Delayed recall 0.02 0.92 −0.55 0.05 −0.56 0.05 Recognition 0.15 0.92 −0.29 0.15 −0.30 0.14 aControlling for all independent variables except Hg; 0 = country, 1 = town. bControlling for all independent variables except residence. CControlling for all independent variables. dRaw (unstandardized) regression coefficient. Abbreviations: NES, Neurobehavioral Evaluation System; CVLT, California Verbal Learning Test. Source: Adapted from Appendix III-B, Table 3, NIEHS (1998).
OCR for page 263
Toxicological Effects of Methylmercury omission of that control variable might not seem important in light of the lack of observed effects in that study. However, if one examiner who is less adept at eliciting optimal performance from the subjects tested a large proportion of less exposed children, the results could be affected. If those children performed more poorly than they otherwise would have on the test, an association between Hg concentration and test scores might be obscured by failure to control for the examiner. That result could also occur if an adept tester tested a large proportion of the more heavily exposed children, leading them to achieve higher scores than they would have if tested by other examiners. The SCDS controlled for age by converting the raw test scores to age-corrected standard scores with conversion tables based on U.S. norms (NIEHS 1998). In contrast, the Faroe study analyzed the raw scores by adjusting statistically for the child's age (measured in days since birth). The latter approach is preferable for three reasons. First, the applicability of U.S. norms to these study populations is uncertain. Second, the use of age-corrected standard scores can reduce the sensitivity of the test, because several adjacent raw scores are treated as equivalent in converting to standard scores. Moreover, because age-corrected standard scores use 3-month intervals, for the purposes of conversion of raw to standard scores, a child whose age is 4 years, 3 months, and 31 days is considered to be the same age as a child who is 4 years, 0 months, and 1 day. However, that same child is considered to be different in age from a child who is only 1 day older (i.e., 4 years, 4 months, and 1 day). Finally, the Faroe approach of controlling statistically for age by multiple regression seems appropriate, because the effect of age is likely to be linear across the relatively short age period (3 months in both studies). Although it seems unlikely that the difference in approach to controlling for age could account for the discrepancies in the findings of those two studies, it would be of interest to see a re-analysis of the SCDS data using the approach that was used in the Faroe study. There appears to have been no need to control for PCB exposure in the Seychelles, because PCB body burdens in that population are exceptionally low. In contrast to North America and Europe, where these contaminants are routinely detected in serum samples, 28 samples obtained from Seychelles study children showed no detectable concentrations of any PCB congeners. In the Faroe study, prenatal PCB exposure was measured in umbilical cord tissue rather than cord blood or maternal blood or milk, as in most previous studies, and specimens were
OCR for page 264
Toxicological Effects of Methylmercury obtained for only half the newborns. Cord-tissue PCB concentration has never been validated in relation to blood or milk concentration, and because cord tissue is lean, it might provide a less reliable indication of total PCB body burden. Although PCBs are measured most accurately on a lipid-adjusted basis in most tissue, the wet-weight measure used in the Faroes was probably more valid for cord tissues, because the gravimetric approach used to measure fat content is not sufficiently reliable in a lean medium. With respect to confounding by PCBs, prenatal PCB exposure was associated with four of the eight outcomes whose relation to cord Hg concentration was statistically significant. Those outcomes related primarily to verbal and memory performance, domains found in previous studies to be associated with prenatal PCB exposure (Jacobson and Jacobson 1996; Patandin et al. 1999). When PCBs and Hg were included together in the model, one outcome — continuous performance test (CPT) reaction time — was independently related to Hg exposure (Grandjean et al. 1997, Table 5). For the other three outcomes, however, the associations with both PCB and Hg fell short of conventional levels of statistical significance. One likely explanation is that both of those contaminants adversely affect those outcomes, but their relative contributions cannot be determined given their co-occurrence in the Faroe population (r = 0.41). It is unfortunate that cord specimens were not obtained from a greater proportion of the children. In a second set of analyses (Budtz-Jørgensen et al. 1999), potential confounding by prenatal PCB exposure was reduced by dividing the sample into tertiles in terms of the infants' cord PCB concentrations. Regressions assessing the associations between Hg exposure and the five principal 7-year outcomes were then run separately for each of the groups. The regression coefficients for Hg in the lowest PCB tertile were no weaker than those among the infants exposed to moderate or heavy PCB doses, lending additional support to the conclusion that the associations between Hg and these outcomes are not attributable to confounding by prenatal PCB exposure. POPULATION DIFFERENCES IN VULNERABILITY Vulnerability to prenatal Hg exposure might be enhanced or attenu-
OCR for page 265
Toxicological Effects of Methylmercury ated by differences in genetic susceptibility, diet, or exposure to other contaminants. The SCDS cohort is predominantly African in descent; the Faroe cohort is Caucasian. Moreover, the Faroe population is thought to be descendant from a small number of “founders,” which could increase genetic vulnerability to toxic insult. Although racial differences in vulnerability are possible, it should be noted that such differences have not been seen for environmental exposure to lead, which has been studied in racially diverse samples. Moreover, evidence of MeHg neurotoxicity was found in the genetically heterogeneous and racially diverse sample assessed in New Zealand, a sample that was predominantly non-Caucasian. In principle, poor nutrition might also make a population particularly vulnerable to teratogenic insult. However, the data on birth weight and gestation length in the Faroe and Seychelles studies suggest that energy and macronutrient (protein and carbohydrate) deficiency is unlikely. Nevertheless, micronutrient deficiency in association with low intakes of fortified or unrefined grains or fruits and vegetables is possible. It is also possible that children in one or both samples might have been weaned to breast-milk substitutes or milks of other species that provide inadequate amounts of iron, other minerals, or vitamins. Conversely, certain nutrients found in fish eaten by the Seychelles residents (e.g., omega-3 fatty acids and selenium) could attenuate adverse effects of Hg exposure. The general health status of a population might also enhance or attenuate vulnerability to teratogenic exposure. Because the Faroe and Seychelles populations apparently both receive excellent health care, however, health status seems unlikely to explain the differences in the study findings. As stated above, one substantial difference between the Faroe and the Seychelles populations relates to their PCB exposure. Whereas PCB concentrations in the Seychelles population are among the lowest observed anywhere in the world, the portion of the Faroe population that eats whale blubber accumulates unusually high PCB body burdens. Although it is conceivable that PCB exposure in the Faroe Islands might enhance fetal vulnerability to Hg, that hypothesis is speculative at present; experimental animal studies would be needed to test its plausibility. The possibility of effect modification by PCB exposure was examined in regression analyses that, in addition to confounders, also included the Hg and PCB exposure variables and their product (Hg ×
OCR for page 266
Toxicological Effects of Methylmercury PCBs; Budtz-Jørgensen et al. 1999). All five p-values for the Hg × PCBs interaction terms exceeded 0.20, suggesting an absence of potentiation of the effects of one of the contaminants by the other. Thus, it seems unlikely that differences in vulnerability due to PCB exposure can explain the differences between the Faroe Islands and the Seychelles findings. The sample in the main Seychelles study appears to have been developmentally robust. There was an exceptionally low number of abnormal scores on the Denver Developmental Screening Test, an unusually high mean Psychomotor Development Index score, and a very low rate of referral for mental retardation. On the other hand, the means and standard deviations of the cognitive measures administered at later ages were similar to U.S. norms. It is unclear to what extent the developmental robustness of that particular sample might have buffered it from any adverse effects of prenatal Hg exposure. RANDOM VARIATION IN THE DETECTABILITY OF EFFECTS AT LOW EXPOSURES The magnitude of the associations found in the MeHg studies resembles that reported with respect to other environmental contaminants, such as low-dose lead and PCBs. When the magnitude of an association is subtle, it is not surprising that it is not detected in every cohort studied. With respect to lead exposure, a strong scientific consensus has developed that blood lead concentrations in excess of 10 µg/dL place a child at increased risk of poor developmental outcomes. However, not all lead studies have found an association, and substantial variability exists in the magnitudes of the reported effects (Bellinger 1995). If two studies from the lead literature were chosen randomly, it is likely that the results of the two would not be entirely concordant. The uncertainties inherent in such studies (e.g., the assessment of exposure histories, the measurement of critical population characteristics, the idiosyncratic pattern of potential confounding factors, and the measurement of neurodevelopmental outcomes) render it likely that evidence of neurotoxicity will not be detected in some of the study cohorts assessed. With respect to the SCDS, the evidence consistent with such effects found in the pilot phase, coupled with the suggestion of unusual devel-
OCR for page 267
Toxicological Effects of Methylmercury opmental robustness in the main study, suggest that the failure to detect apparent adverse effects in the main study could be due to the substantial sample-to-sample variation expected when trying to identify weak associations in an inherently “noisy” system of complex, multi-determined neurobehavioral end points. Given the large sample size in the main Seychelles cohort, it might seem surprising that that study could lack the power needed to detect an association between increased MeHg exposure and neurobehavioral impairments. However, power analyses that are based on total sample size can be misleading if adverse effects occur primarily among the most heavily exposed children, who typically comprise a very small proportion of the sample. Although the sample size of 700 children in the SCDS would seem to be more than adequate, only about 35 children were exposed at 15 µg/g or higher. Because multiple regression analysis examines associations that are averaged across the entire distribution of exposure, associations that hold only for the most highly exposed children can be difficult to detect. Thus, if adverse effects of prenatal MeHg exposure occur primarily in the upper range, the power to detect them will be limited, and it would not be surprising if associations found in one Seychelles cohort (the pilot study) were not detected in the next cohort (the main study) (see Chapter 7 for further discussion on the issue of statistical power). CONCLUSIONS Three well-designed, prospective, longitudinal studies have examined the relation of prenatal MeHg exposure to neuropsychological function in childhood. MeHg was associated with poorer performance in the Faroe Islands study but not in the SCDS. Little attention has been paid to the New Zealand study because, until recently, it had not been subjected to peer review. Differences in the primary biomarker of Hg exposure (cord blood versus maternal hair), type of neuropsychological tests administered (domain specific versus global), age of testing (7 versus 5.5 years), and sources of exposure (whale meat versus fish) between the Faroe study and the SCDS have been suggested to account for the differences in the findings of the two studies. When the New Zealand
OCR for page 268
Toxicological Effects of Methylmercury data are considered, however, those differences no longer seem determinative, because the New Zealand study, in which the exposure and research design were very similar to the SCDS, also found associations between higher MeHg levels and worse neurobehavioral test scores, as did the pilot SCDS. There is no empirical evidence or hypothesized mechanism to support the suggestion that PCB exposure might enhance vulnerability to MeHg. The lack of any evidence of statistical interaction between Hg and PCB exposure in the Faroes data also makes it unlikely that a difference in PCB exposure can explain the differences between the Faroe Islands and the Seychelles findings. The lack of statistical control for examiner in the SCDS, population differences in susceptibility among the study populations, and dietary factors might explain some of the differences among the study findings. It is possible that the differences are primarily due to between-sample variability in the expression of neurotoxicity at low doses. Even large sample studies can lack adequate power to detect adverse associations if a relatively small number of children are exposed in the upper ranges of the exposure distributions, where the adverse effects are most likely to be found. Although none of the between-study differences noted above appears to be determinative, the combined influence of two or more of these factors is difficult to predict. For example, it is possible that slightly reduced vulnerability in the Seychelles population combined with the use in that study of a biomarker of exposure that averages across pregnancy could make it difficult to detect neurocognitive effects that might be specific to third trimester exposure. When the two studies reporting associations between MeHg and neurobehavior are compared, the strengths of the New Zealand study include an ethnically heterogeneous sample and the use of developmental end points with greater predictive validity. The advantages of the Faroe study include a larger sample size, the use of two different biomarkers of exposure, and extensive scrutiny in the epidemiological literature. The Faroe data have also undergone extensive re-analyses in response to questions raised by panelists in the NIEHS (1998) workshop and by this committee in the course of its deliberations.
OCR for page 269
Toxicological Effects of Methylmercury RECOMMENDATIONS It would be helpful to obtain more comprehensive nutritional data from all three populations as well as single-strand hair analyses to address more effectively the issue of spiking or bolus dose. A reanalysis of the 5.5-year SCDS data controlling statistically for examiner might also be useful. Most of the MeHg exposure standards currently in effect are based on extrapolations from the Iraqi MeHg poisoning episode, in which exposure was due to the consumption of highly contaminated grain and resulted in body burdens that greatly exceeded those found in the general population of fish consumers. Given the availability of data from three well-designed epidemiological studies in which prenatal MeHg exposures were in the range of general-population exposures, exposure standards should be based on data from these newer studies. REFERENCES Bellinger, D. 1995. Interpreting the literature on lead and child development: The neglected role of the “experimental system” . Neurotoxicol. Teratol. 17(3):201-212. Bonthius, D.J., and J.R. West. 1990. Alcohol-induced neuronal loss in developing rats: Increased brain damage with binge exposure. Alcohol Clin. Exp. Res. 14(1):107-118. Budtz-Jørgensen, E., N. Keiding, P. Grandjean, and R. White. 1999. Methylmercury neurotoxicity independent of PCB exposure. [Letter]. Environ. Health Perspect. 107(5):A236-A237. Crump, K.S., T. Kjellström, A.M. Shipp, A. Silvers, and A. Stewart . 1998. Influence of prenatal mercury exposure upon scholastic and psychological test performance: benchmark analysis of a New Zealand cohort. Risk Anal. 18(6):701-713. Goodlett, C.R., S.J. Kelly, and J.R. West. 1987. Early postnatal alcohol exposure that produces high blood alcohol levels impairs development of spatial navigation learning. Psychobiology 15(1):64-74. Grandjean, P., P. Weihe, P.J. Jørgensen, T. Clarkson, E. Cernichiari, and T. Videro. 1992. Impact of maternal seafood diet on fetal exposure to mercury, selenium, and lead. Arch. Environ. Health 47:185-195 Grandjean, P., E. Budtz-Jørgensen, R.F. White, P.J. Jørgensen, P. Weihe, F. Debes, and N. Keiding . 1999. Methylmercury exposure biomarkers as
OCR for page 270
Toxicological Effects of Methylmercury indicators of neurotoxicity in children aged 7 years. Am. J. Epidemiol. 150(3):301-305. Grandjean, P., P. Weihe, R.F. White, F. Debes, S. Araki, K. Yokoyama, K. Murata, N. Sørensen, R. Dahl, and P.J. Jørgensen. 1997. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol. Teratol. 19(6):417-428. Jacobson, J.L., and S.W. Jacobson. 1991. Assessment of teratogenic effects on cognitive and behavioral development in infancy and childhood. Pp. 248-261 in Methodological Issues in Controlled Studies on Effects of Prenatal Exposure to Drugs of Abuse, Research Monograph 114, M.M. Kilbey, and K. Asghar, eds. Rockville, MD: National Institute on Drug Abuse. Jacobson, J.L., and S.W. Jacobson. 1996. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. N. Engl. J. Med. 335(11):783-789. Myers, G.J., P.W. Davidson, and C.F. Shamlaye. 1998. A review of methylmercury and child development. Neurotoxicology 19(2):313-28. Myers, G.J., P.W. Davidson, C. Cox, C.F. Shamlaye, M.A. Tanner, O. Choisy, J. Sloane-Reeves, D.O. Marsh, E. Cernichiari, A. Choi, M. Berlin, and T.W. Clarkson. 1995. Neurodevelopmental outcomes of Seychellois children sixty-six months after in utero exposure to methylmercury from a maternal fish diet: Pilot study. Neurotoxicology 16(4):639-652. NIEHS (National Institute of Environmental Health Sciences). 1998. Scientific Issues Relevant to Assessment of Health Effects from Exposure to Methylmercury. Workshop organized by Committee on Environmental and Natural Resources(CENR) Office of Science and Technology Policy (OSTP) The White House, November 18-20, 1998, Raleigh, NC. Patandin, S., C.I. Lanting, P.G. Mulder, E.R. Boersma, P.J. Sauer, and N. Weisglas-Kuperus. 1999. Effects of environmental exposure to polychlorinated iphenyls and dioxins on cognitive abilities in Dutch children at 42 months of age. J. Pediatr. 134(1):33-41. Trillingsgaard, A., O.N. Hansen, and I. Beese. 1985. The Bender-Gestalt Test as a neurobehavioral measure of preclinical visual-motor integration deficits in children with low-level lead exposure. Pp. 189-193 in WHO Environmental Health, Document 3. Neurobehavioral Methods in Occupational and Environmental Health, Second International Symposium, Copenhagen, Denmark, Aug.6-9, 1985. Copenhagen, Denmark: World Health Organization.
Representative terms from entire chapter: