Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
3 Epidemiologic Studies of Occupational and Environmental Exposures to Perchlorate THIS chapter considers observational epidemiologic studies of perchlorate exposures and measures of thyroid function and thyroid diseases in selected populations. The committee considered published and unpublished epidemi- ologic data in forming its conclusions regarding the possible effects on thyroid function of exposure to perchlorate and in assessing the relevant exposure limits. More weight was given to published reports, and unpub- lished data were considered only when sufficient information was made available to the committee to evaluate the methods used. Most of the unpublished data considered were presented to the committee in open meetings. The committee considered one unpublished masterâs thesis (Schwartz 2001). In weighing the evidence on perchlorate exposure, the committee emphasized studies with the soundest scientific methods and studies that included biologically sensitive groups, such as pregnant women, fetuses, and neonates. Of the epidemiologic studies, one is an unpublished mortality study of workers at a chemical plant who had simultaneous exposures to many agents, including perchlorate (Rockette and Arena 1983); three are cross-sectional studies of occupational cohorts that had respiratory exposures to perchlorate and for which assessments of thyroid function were obtained (Gibbs et al. 1998; Lamm et al. 1999; Braverman et al. 2004); and the remainder are primarily ecologic investigations that compared measures of thyroid function or disease and neurodevelopmental outcomes in neonates, children, and adults in geographically defined groups with and without detectable perchlorate in community water supplies (Lamm and Doemland 1999; Brechner et al. 2000; Crump et al. 2000; F.X. 75
76 Health Implications of Perchlorate Ingestion Li et al. 2000; Z. Li et al. 2000; Li et al. 2001; Schwartz 2001; Chang et al. 2003; Kelsh et al. 2003; Lamm 2003; Buffler et al. 2004; Gibbs 2004a,b). The ecologic study of Morgan and Cassady (2002) assessed cancer inci- dence in a community exposed to both perchlorate and trichloroethylene in drinking water. No epidemiologic studies, either published or unpublished, have measured both thyroid outcomes and perchlorate exposure from drinking water in the same people. Many of the epidemiologic data related to the effects of perchlorate are derived from ecologic studies. The smallest units on which exposure or outcome data are available are geographically defined units, most com- monly counties, states, or countries; exposure data, outcome data, or both are available only at that level, not on individual subjects. Because ecologic studies do not include information about exposure and outcome in individu- als, they are considered to be the weakest type of observational studies. In ecologic studies, comparisons are made between exposures and outcomes among large units. An example of an ecologic study would be one that compares the mean serum concentration of thyroid-stimulating hormone (TSH) in infants born in a county in which the drinking-water supply contains perchlorate with the mean serum TSH concentration in infants born in a county in which the drinking water does not contain perchlorate. That design is subject to what is referred to as the ecologic fallacy: associations observed at the ecologic level may not apply at the individual level. For example, an observation that the average serum TSH concentra- tion is higher in newborns in a city in which the water supply contains perchlorate than in newborns in cities in which the drinking water does not contain perchlorate would be compatible with a perchlorate effect on thyroid function. In that example, being compatible with an association assumes that no other important variations between the cities could account for the difference in serum TSH concentrations. However, it is not known from such a study whether infants who have high TSH concentrations were themselves exposed to perchlorate during gestation. How well ecologic studies are able to characterize individual exposure depends, in part, on how much variability of exposure there is in the geographic unit. For example, ecologic studies of the correlation of dietary fat intake and breast-cancer mortality by country would undoubtedly suffer from considerable variation in the average dietary fat intake among the persons in a given country. The use of average community exposures for ecologic studies of perchlorate in municipal drinking water may be less subject to error, depending on the population coverage of the water supply. In many of the exposed communi- ties included in the literature, an entire communityâs water supply has
Epidemiological Studies of Occupational and Environmental Exposures 77 measurable perchlorate. Thus, it is likely that everyone has some exposure to perchlorate, depending on how much of their water intake comes from the communityâs water supply. Individual variation in water exposure undoubtedly still occurs, however, because of the use of wellwater or bottled water or because of nonuniform distributions of contaminants in a geographic area. Some degree of individual variability undoubtedly exists, but it may not be as great as for other types of exposures that are not part of a communitywide, common-source exposure. Another limitation of ecologic studies is that their design cannot control for many confounding factors, because such data are not usually available at the population level. Results of ecologic studies can be useful in provid- ing supporting data on a possible causal relation, but they cannot them- selves provide direct evidence of causation. The available pertinent occupational and epidemiologic studies are summarized in Table 3-1 and discussed in the following sections. STUDIES IN OCCUPATIONAL COHORTS AND ADULTS An early study of an occupational cohort by Rockette and Arena (1983) reported mortality patterns for 59 selected causes of death among workers at the Niagara Plant of Hooker Chemical. It included people with at least 1 year of employment from January 1, 1949, to December 31, 1978. The cohort consisted of 3,963 workers (3,715 men and 248 women) at the plant in Niagara Falls, New York. Only 13 deaths were recorded among women, and detailed analyses of the female data were not conducted because the sample was so small. The following results are related to men. Results showed statistically significant excess mortality from stomach cancer (standardized mortality ratio [SMR], 178.9; p < 0.05) and respiratory cancers (SMR, 145.9; p < 0.01). In the respiratory-cancer category, the SMR was significantly increased for cancer of the lung (SMR, 142.4; p < 0.01). A review of work areas and production indicated that exposure to magnesium perchlorate occurred in 1970-1976 with simultaneous expo- sure to dozens of other chemicals at the plant. Of the work areas in the plant, 12 departments were identified as having exposure to âgroups of chemicals.â The department referred to as âarea 4â included exposure to magnesium perchlorate and 23 other chemicals. Job titles in the department were not provided, so specific jobs were not assessed with regard to expo- sure to magnesium perchlorate. A comparison of department-specific
78 TABLE 3-1 Summary of Epidemiologic Studiesa Study Adjustment Reference Design Population Exposure Factors Outcomes Major Findings Comments Studies of Occupational Groups and Adult Cohorts Rockette SMR study, 3,963 chemical Magnesium Age, Cause-specific mor- For males 1970-1978, all Very few deaths recorded and Arena 1970-1978 workers, 94% perchlorate, calendar tality cancer SMR = 143.7, p < in females, so only data on 1983 male unknown amount, time 0.01; respiratory cancer males analyzed in detail; in combination with SMR = 169.6, p < 0.01; multiple chemical 23 other chemicals stomach cancer SMR = exposures; no adjustment 236.7, p < 0.05 for cigarette-smoking or potential confounders other than age Gibbs et al. Repeated 18 workers Respiratory Duration of Compared preshift No significant relation of Respiratory exposure; small 1998 cross- exposed to exposure at 0.2-436 shift and postshift serum perchlorate dose to change number of exposed sectional ammonium :g/kg, depending free T4 index, total in thyroid hormone and TSH workers; subject to (preshift perchlorate and on work area T4, and TSH measures across shift potential bias of affected and 83 nonexposed workers who left postshift) workers employment before study Gibbs et al. Historical 53 high dose, 44 Lifetime None Compared medical Neither estimated Respiratory exposure; 1998 cohort low dose, 192 cumulative records on thyroid cumulative lifetime exposure subject to potential bias of nonexposed exposure to per- hormone and TSH nor exposure category affected workers who left chlorate; mean in measures and significantly associated with employment before study; âlow-doseâ group, disease; medical abnormalities of thyroid no statistical adjustment for 3,500 :g/kg; mean surveillance data hormone measures and TSH BMI or activity level âhigh-dose,â from cohort 38,000 :g/kg
Lamm et Cross- 31 ammonium Absorbed None Serum TSH, T4, and No significant differences in Respiratory exposure; al. 1999 sectional perchlorate work- perchlorate T3; free T4 index; any outcome measure subject to potential bias of ers and 21 azide categorized as thyroid hormone between perchlorate- affected workers who left workers at the none, low, medium, binding ratio; exposed and azide-exposed employment before study; same site; ages and high (mean, 1, thyroid peroxidase workers; no clinical small number in study 20-56 yr 4, 11, 34 mg/shift) antibodies; urinary evidence of thyroid groups reduces ability to iodide and abnormalities in perchlorate- detect differences creatinine; clinical exposed workers examination Li et al. Ecologic Medicaid en- Residence in Clark None Medicaid payment Prevalence of thyroid Most thyroid diseases were 2001 rollees in Nevada County (drinking- diagnosis of one of diseases or thyroid cancer uncommon, so numbers by county of water mean = 8.9 nine thyroid not significantly higher in were small in county residence, 1997- ppb), Washoe diseases or thyroid Clark County than in either comparisons; no adjustment 1998 County, or cancer Washoe or all other nonex- for differences among three remaining counties posed counties combined county groups in age, sex, combined race or other potential (nonexposed) confounders Morgan Ecologic Residents of 13 Drinking-water Age, sex, O/E numbers of SIR for thyroid cancer = 1.0 Timing and duration of and contiguous exposure to TCE race, or cancer cases; total (95% CI = 0.63-1.47); exposure to perchlorate Cassady census tracts in (0.09-97 ppb) ethnicity and site-specific and cancers of colon or rectum unclear; exposure to TCE 2002 Redlands, CA measured since and separately for and lung or bronchus had also present; no adjustment (San Bernardino 1980 and calendar children < 15 yr old significantly lower SIRs and for other potentially County), 1988- ammonium time melanoma and uterine confounding variables; 1998 perchlorate (5-98 cancer had significantly expected numbers were ppb) since 1997 higher SIRs; no cancers were based on surrounding four observed more often than counties and included study expected in children county data 79
80 TABLE 3-1 (Continued) Study Adjustment Reference Design Population Exposure Factors Outcomes Major Findings Comments Braverman Repeated 29 workers at Respiratory None Serum TSH, T4, T3, Decrease in mean RAIU Pre- and postshift et al. 2004 cross- ammonium exposure to free T4 index, 14-hrpostshift compared to evaluation allows for sectional perchlorate ammonium radioactive iodide preshift and increase in comparison within the same (pre- and production perchlorate; mean uptake (RAIU), iodide excretion postshift; individuals; data are postshift) facility and 12 absorbed dose per urinary iodide serum perchlorate preliminary; subject to nonworker shift 0.3 mg/kg excretion, serum undetectable after 3 days off potential bias of affected volunteers perchlorate, work but elevated postshift; workers leaving thiocyanate and TSH unchanged, T4, and T3, employment prior to study; nitrate; thyroid size free T4 index showed small small number of but significant increases nonexposed volunteers postshift Studies of Neonatal and Pediatric Populations Lamm and Ecologic Clark County Drinking water; Hispanic Congenital O/E cases = 1.0 (0.90-1.16) Exposure only sporadic in Doemland SMR NV; six southern NV, 4-16 ppb ethnicity hypothyroidism CA; questionable 1999 CA counties; continuous; CA, 5- comparison rates used; not 700,000 8 ppb sporadic adjusted for birthweight newborns screened 1996- 1997 Brechner Ecologic Newborns in ~ 6 ppb in Yuma Race or Mean log TSH and Adjusted mean log TSH Not adjusted for et al. 2000 Yuma, AZ drinking water; ethnicity T4 in serum significantly higher in Yuma birthweight or gestational (1,099) and none in Flagstaff and age at than in Flagstaff; no age; 1999 perchlorate used newborns in water sample significant difference in T4 to characterize the study Flagstaff, AZ collection period, 1994-1997 (443) 1994-97
Crump et Ecologic 6- to 8-yr olds in Drinking water; Age, sex, Serum TSH, T4, freeFor life-long residents, no Drinking-water perchlorate al. 2000 three cities in Antofagasta, no urinary T4, T3, liver and significant differences in measured in sample from Chile, 1999; detectable; iodide kidney function, TSH; adjusted free T4 signi- faucets used by children newborns in the Chanaral, 5-7 ppb; urinary iodide and ficantly higher in Chanaral studied; newborn TSH not three cities, 1996- Taltal, 100-120 ppb creatinine; clinical and Taltal than in Antofa- adjusted for birthweight or 1999 examination for gasta; no significant differ- gestational age goiter; family ences in prevalence of goiter history of goiter by in children (~20% in each questionnaire city); family history goiter ~5 times as high in Taltal; log TSH significantly lower in newborns from Taltal Z. Li et al. Ecologic Newborns in Las Las Vegas drinking Sex, age at Mean T4 in newborn No significant difference in Excluded birthweights 2000 Vegas, NV water (9-15 ppb for sample blood adjusted mean T4 <2,500 g and >4,500 g (17,308), 7 mo and undetect- collection, concentrations between newborns in able for 8 mo; birthweight newborns in Las Vegas and Reno, NV estimated cumula- newborns in Reno; no (5,882), 1998- tive exposure in difference in prevalence of 1999 pregnancy, 0.9-4.2 low T4 (#10th percentile of mg); Reno birth-cohort distribution) (nonexposed) F.X. Li et Ecologic Newborns in Las Las Vegas drinking Age at Mean log TSH in No significant difference in Excluded birthweights al. 2000 Vegas, NV (407) water (9-15 ppb for sample newborn blood adjusted mean log TSH <2,500 g and >4,500 g; and Reno, NV 7 mo and undetect- collection, concentrations between adjustment for age at (133), 1998-1999 able for 8 mo; sex newborns in two cities specimen collection estimated cumula- dichotomous tive exposure in pregnancy, 0.9-4.2 mg); Reno (nonexposed) 81
82 TABLE 3-1 (Continued) Study Adjustment Reference Design Population Exposure Factors Outcomes Major Findings Comments Schwartz Ecologic 507,982 CA Drinking water; Sex, Newborn blood Adjusted mean T4 Potential misclassification 2001 newborns, 1996 estimates based on ethnicity, screening T4 and significantly lower and mean of perchlorate exposure due averages obtained multiple TSH, presumptive log TSH significantly higher to incomplete from 1997 testing birth, and confirmed with increasing perchlorate measurements; failure to of samples from set birthweight, congenital exposure; odds of account for multiple of stateâs water age at hypothyroidism presumptive congenital laboratories doing newborn systems; categor- specimen hypothyroidism screening tests ized as none, low collection significantly increased in (1-2 ppb), medium low and medium exposure (3-12 ppb), high categories, but not in high; (>12 ppb) confirmed congenital hypothyroidism not related to perchlorate exposure Chang et Ecologic NV Medicaid County of None Proportion of No significant differences Analysis does not account al. 2003 recipients, 1996- residence; Clark Medicaid recipients between exposed and for changes in residence; 2000, <18 yr old County (Las Vegas) with diagnosis of nonexposed counties in the validity of diagnoses not with diagnosis of (exposed; median ADHD or autism proportion of Medicaid established; exposure time ADHD or autism 23.8 ppb), Washoe and national recipients <18 yr old with may not have been County (Reno) percentile rankings diagnosis of ADHD or biologically relevant; (nonexposed), all for fourth-grade autism; no significant diagnoses made by multiple remaining counties proficiency test differences in fourth-grade health-care providers (nonexposed) scores performance scores without use of standard criteria; no statistical adjustment for potential confounding variables
Kelsh et al. Ecologic Newborns in two Drinking water; Age at Primary congenital Residence in exposed Analyses based on only two 2003 CA communities, Redlands and specimen hypothyroidism and community not associated cases of congenital 1983-1997; Mentone (exposed, collection, high serum TSH with higher prevalence of hypothyroidism and six of Redlands and up to 9 ppb; mean 1 sex, race, or (mostly >25 congenital hypothyroidism high TSH in exposed Mentone ppb), San ethnicity, :U/mL, sometimes (SPR = 0.45, 95% CI = 0.06- community; exposure data (15,330), San Bernardino and birthweight, >16 :U/mL) 1.64) or high serum TSH in from single year used to Bernardino and Riverside multiple newborns (SPR = 0.72, 95% characterize entire interval Riverside (nonexposed) births, cal- CI = 0.28-1.54) (696,530) endar year Lamm Ecologic Newborns in Drinking water; None Median serum TSH Median serum TSH Not adjusted for differences 2003 Yuma, San Luis, Yuma exposed; San concentration higher in in birthweight, gestational and Somerton Luis, Somerton, Yuma (20.8 mU/L) and San age, ethnicity (Yuma County), Flagstaff Luis and Somerton (21 and Flagstaff, AZ nonexposed mU/L) than in Flagstaff September 1994- (14.5 mU/L); newborns in June 1998 Yuma have exposure to perchlorate, those in San Luis and Somerton do not Buffler et Ecologic Newborns in CA, Drinking water, Age at Primary congenital No statistically significant Study was extension of al. 2004 1998 (342,257) 1997-1998; specimen hypothyroidism and relation between residence in Kelsh et al. (2003) exposed collection, high serum TSH exposed community and addressing lack of exposure communities had sex, (generally >25 prevalence or odds of assessment for complete mean perchlorate ethnicity, :U/mL) primary congenital study period and small >5 ppb; birthweight hypothyroidism or high TSH number of cases, although nonexposed still lacking exposure communities had assessment of individuals; perchlorate <5 ppb in Buffler et al. (2004), exposure assessment corresponded to newborn screening data period; 15 cases of primary congenital 83
84 TABLE 3-1 (Continued) Study Adjustment Reference Design Population Exposure Factors Outcomes Major Findings Comments hypothyroidism and 147 cases of high serum TSH observed a Includes published papers and final study reports. Abbreviations: ADHD, attention-deficit-hyperactivity disorder; AZ, Arizona; BMI, body-mass index; CA, California; CI, confidence interval; g, gram; mg, milligram; :g/kg, microgram per kilogram; :U/mL, microunits per milliliter; mU/L, milliunits per liter; mo, month; NV, Nevada; O/E, observed over expected; ppb, parts per billion; SIR, standardized incidence ratio; SMR, standardized mortality ratio; SPR, standardized prevalence ratio; T3, triiodothyronine; T4, thyroxine; TCE, trichloroethylene; TSH, thyroid-stimulating hormone; yr, year.
Epidemiological Studies of Occupational and Environmental Exposures 85 mortality of workers in the plant with U.S. and local Niagara County mortality by calendar year found no statistically significant excess mortality for 1970-1978; magnesium perchlorate exposure occurred in âarea 4â of the Niagara Falls plant in 1970-1976. However, analyses of the entire cohort by calendar period showed a statistically significant excess of deaths from malignant neoplasms as a group (SMR, 143.7; p < 0.01) and for respiratory cancers (SMR, 169.6; p < 0.01) and stomach cancers (SMR, 236.7; p < 0.05) for 1970-1978. Because of the multiple exposures of those workers and the inability to adjust for important confounding variables, such as cigarette smoking, it was not possible to determine whether perchlorate exposure itself was related to any increase in SMRs. In another study of occupational exposure among 254 employees at an ammonium perchlorate production facility in Nevada, those who agreed to participate in the study were investigated for possible acute or chronic effects of perchlorate exposure on thyroid function (Gibbs et al. 1998). Two analyses were conducted: a single-shift study and a working-lifetime study. For the single-shift study, differences in the triiodothyronine (T3) resin uptake assay, the free thyroxine (T4) index, total T4, and TSH in serum obtained before and after a work shift were compared in exposed workers (n = 18) and control workers (n = 83), defined as persons who had not worked in the production areas for the preceding 30 days. Exposures to ammonium perchlorate in this production facility were through the respira- tory tract and were estimated by personal breathing-zone sampling and air sampling of work areas in the plant. Full-shift, personal breathing-zone filters in closed-face cassettes were used to characterize the average air- borne exposure to ammonium perchlorate in each work area, which was estimated at 0.2-436 :g/kg (mean, 36 :g/kg) per day, depending on work area. Exposure estimates were obtained for the exact shift in which a worker had volunteered for thyroid hormone and TSH measurements to be taken and were adjusted downward by 65% for areas in which workers wore respiratory protection. Workers taking thyroid medications (all of whom were potential controls) were excluded from analyses. Cumulative working-lifetime exposure estimates were also calculated (Gibbs et al. 1998). Estimates of lifetime exposure of individual workers were based on the product of mean exposures for their work areas, the number of years they worked in those areas, and 2,000 (the estimated number of hours worked per year). Work histories used in the calculation of cumulative lifetime exposures were obtained by reviewing personnel records and interviewing workers. In addition to the individual exposure estimates, three exposure categories were arbitrarily created for the
86 Health Implications of Perchlorate Ingestion cumulative-exposure analysis; controls, âlowâ exposure (cumulative life- time exposure, 500-7,000 :g/kg; mean, 3,500 :g/kg), and âhighâ exposure (cumulative lifetime exposure, 8,000-88,000 :g/kg; mean, 38,000 :g/kg). Duration of employment of the âlowâ and âhighâ exposure groups com- bined ranged from 1 to 27 years (mean, 8.3 years). Absolute cumulative- exposure estimates by individual and by exposure group were examined separately in relation to thyroid hormone and TSH test results. The thyroid hormone and TSH measures assessed in the cumulative-exposure analysis were the same as those examined in the single-shift analysis. Rather than being contemporaneously obtained, however, thyroid hormone and TSH measures were obtained from reports of routine medical surveillance in the workersâ medical records. Those routine medical-surveillance profiles included blood tests of liver, kidney, and bone marrow function, which were also examined in the analysis of cumulative exposure data. Multiple regression analyses were used to compare thyroid hormone and TSH measures with exposure dose estimates, controlling for race, sex, and age for both single-shift and cumulative assessments and, for the single-shift assessments, controlling also for hours awake before the preshift test, number of hours slept in the period before testing, time of day of test, and shift length. Results were not controlled for BMI or activity level. Results from the single-shift analysis indicated no significant associa- tions between estimated exposure and alterations in any of the thyroid hormone and TSH measures (Gibbs et al. 1998). Only duration of shift was related to serum TSH: the postshift mean serum TSH concentration was higher than the preshift concentration and exceeded it by more after a 12-hr shift than after an 8-hr shift, a difference expected because there is a circa- dian increase in serum TSH concentrations. In the analysis of cumulative lifetime exposure to ammonium perchlorate, neither the dose estimate nor the categorical exposure variable was significantly related to measures of thyroid hormone and TSH production. They were unrelated to all but one (white blood cell count) of the measures of bone marrow, liver, or kidney function. The average number of years of exposure to ammonium perchlo- rate in the working-lifetime study was 8.3 (range, 1-27 years); the results of these analyses appear to reflect the effects of chronic exposure in these workers. The working-lifetime data are based, however, only on employees still working at the plant when the study was done; they do not account for those who may have left employment because of thyroid disease or other disease or toxicity. One of the primary limitations of both the single-shift and cumulative exposure analyses is that the route of exposure to perchlo- rate was inhalation and thus may not reflect health effects of ingestion of
Epidemiological Studies of Occupational and Environmental Exposures 87 drinking water. The authors argue, however, that because ammonium perchlorate is very water-soluble, it is likely that a high percentage of the inhaled dose is absorbed through the respiratory tract. Urinary measures of perchlorate obtained from workers in a later study (see below) indicated significant absorption of perchlorate by the respiratory tract. Other limita- tions include the small sample size, particularly of the exposed group, which minimizes the statistical power to detect a meaningful difference; and the low participation rate, which raises the possibility that selection bias will be increased by the limitation of studying only current workers. Lamm et al. (1999) compared the thyroid health status of 37 workers at an ammonium perchlorate production plant in Iron County, Utah, with that of 21 workers involved in the production of sodium azide at the same industrial site. The workers were 20-56 years old. Exposure of workers to perchlorate at both sites was estimated by measuring air concentrations of total and respirable perchlorate particles (exposures to airborne perchlorate ranged from 0.004 to 167 mg/day), and job assignments were categorized as âlow,â âmedium,â and âhighâ exposure on the basis of visible dust generated in the production area. The mean absorbed perchlorate was about 1, 4, 11, and 34 mg/shift for the azide-exposed and low, medium, and high perchlorate-exposure groups, respectively. The validity of that categoriza- tion for capturing variable exposures was demonstrated by comparisons of creatinine-adjusted median concentrations of absorbed perchlorate mea- sured in urine samples collected before and after the work shift. Question- naires were administered that included information on alcohol and tobacco use, medications, and family history of diseases (diabetes, hypertension, rheumatoid arthritis, thyroid disease, and cancer). Thyroid status was assessed by measuring serum TSH, T4, T3, free T4 index, thyroid hormone binding ratio, and thyroid peroxidase antibodies and by conducting clinical examinations. Postshift blood specimens were obtained for a complete blood count and chemistry panel, and urinary iodide and creatinine concen- trations were measured to determine whether workers had adequate dietary iodide intake. No statistically significant differences were observed in any of the measures of thyroid status between the azide workers and any of the three perchlorate-exposure groups. No clinical evidence of thyroid abnor- malities or evidence of hematotoxicity was observed in any perchlorate- exposure group. The sample sizes in each group were small, however, so it was difficult to detect meaningful differences between exposure groups. This study and that of Gibbs et al. (1998) were cross-sectional investiga- tions of worker populations. Thus, they are not able to account for any effects of exposure that might have occurred in workers who have left
88 Health Implications of Perchlorate Ingestion employment for any reason. If perchlorate-exposed workers who developed thyroid disease retired or changed jobs because of their illness, any associa- tion of exposure with adverse outcomes would be underestimated or missed altogether. Preliminary analyses of data from another study of workers in the Utah ammonium perchlorate production facility were presented at a public meeting of the committee in May 2004 and later in a draft manuscript (Braverman et al. 2004). Twenty-nine of 40 workers employed for at least 2 years in ammonium perchlorate production were assessed at two times: after 3 days off work (preshift) and during the last of three 12-hr shifts in the plant (postshift). Exposure to perchlorate was respiratory. The authors also examined 12 nonrandomly selected community volunteers who were not working in the plant. Measures included serum T4, free T4 index, total T3, thyroxine-binding globulin, and TSH; 14-hr radioactive iodide uptake (RAIU); urinary iodide excretion; and serum perchlorate, thiocyanate, and nitrate. The perchlorate doses absorbed over a shift averaged 0.3 mg/kg. Mean postshift RAIU (13.5%) was lower than preshift RAIU (21.5%) but preshift RAIU values were higher than those in the nonexposed controls, suggesting upregulation of iodide transport activity. An increase in urinary iodide excretion, 230 :g/g of creatinine (postshift) vs 148 :g/g of creatinine (preshift), was concomitant with postshift RAIU reduction. Serum per- chlorate was not detectable after 3 days off work, but was increased in the postshift sample (850 ppb [850 :g/L]). Serum TSH and thyroxine-binding globulin concentrations were within the normal range and were similar in all three comparisons (preshift and postshift workers and nonexposed volunteers). Serum T4 (8.3 vs 7.7 :g/dL), free T4 index (2.4 vs 2.2), and total T3 (147 vs 134 ng/dL) concentrations were significantly higher in postshift than in preshift samples. Thyroid volumes and patterns assessed by ultrasonography were not significantly different between workers and volunteers. The data suggest small acute changes in some measures of thyroid function that are in an opposite direction to an antithyroid effect of perchlorate. No apparent long-term effects on thyroid size or function were observed. As previously mentioned, the analyses are subject to potential bias because of loss from the workforce of workers whose thyroid function has been adversely affected; thus, the study group may have been âselectedâ for the absence of effects. Because of the preliminary nature of the data, the committee was not able to consider the results of the study in arriving at its conclusions. Li et al. (2001) analyzed Nevada Medicaid data for the period January 1, 1997, through December 31, 1998, to compare the prevalence of specific
Epidemiological Studies of Occupational and Environmental Exposures 89 thyroid diseases among Medicaid recipients by county of residence as characterized by the presence or absence of perchlorate in public drinking water. Clark County, which includes Las Vegas, is the only Nevada county in which perchlorate was detected in public drinking water during the time of the study; all other counties were considered to be ânonexposed.â Data on the size of the Medicaid-eligible population in each county were pro- vided by the state. The prevalence proportions for each of the specific thyroid diseases were calculated for eight county categories (seven individ- ual counties and all other counties combined). Prevalence proportions were then compared among three county groupings: Clark County, with perchlo- rate in drinking water at 4.1-14 ppb (mean, 8.9 ppb [8.9 :g/L]); Washoe County, which includes Reno, with no detectable perchlorate; and all other counties in the state combined, also with no detectable perchlorate. The thyroid diseases studied were simple and nonspecified goiter, nontoxic nodular goiter, thyrotoxicosis with or without goiter, congenital hypothy- roidism, acquired hypothyroidism, thyroiditis, other disorders of the thy- roid, and malignant neoplasms of the thyroid (ICD-9 Codes 193 and 240- 246). Residence as noted in the Medicaid database was used to determine exposure status (county of residence). Most of the thyroid diseases studied were uncommon, with a 2-year prevalence in Nevada ranging from 1 in 10,000 for congenital hypothyroid- ism to 121 in 10,000 for acquired hypothyroidism (Li et al. 2001). There were no statistically significant differences in the prevalence of any thyroid disease between Clark County (the âexposedâ county) and Washoe County (ânonexposedâ and urban) or between Clark County and all remaining counties combined (ânonexposedâ and rural). The prevalence ratio for thyroid cancer in Clark County (28 cases in 122,519, or 2 per 10,000 Medicaid-eligible people) versus Washoe County (nine cases in 29,622, or 3 per 10,000 Medicaid-eligible people) was 0.75 (95% confidence interval [CI], 0.35-1.59), indicating no significant increase in thyroid cancer associ- ated with exposure to perchlorate at the concentrations reported for those areas. No adjustment of prevalence proportions for differences in sex or age distributions among the three geographic areas could be made, because the pertinent data were not available in this ecologic study. In the absence of the ability to adjust for potential confounding variables, comparisons between Clark and Washoe Counties are probably the most informative in that both counties include large metropolitan areas. Clark County had about 4 times more Medicaid enrollees than did Washoe County at the time of the study, so prevalence estimates from Clark County are more stable. The infrequency of some of the thyroid diseases resulted in small numbers of
90 Health Implications of Perchlorate Ingestion cases, which may have accounted for the failure to achieve statistical significance in several of the comparisons. In another ecologic study, Morgan and Cassady (2002) compared the observed and expected numbers of incident cancer cases among residents of 13 contiguous census tracts in Redlands, California, in San Bernardino County. That area is served by the Desert Sierra Cancer Surveillance Program, a regional cancer registry that includes four counties in Southern California, one of which is San Bernardino County. All cancers in the region have been reportable to the California cancer registry by law since 1988. Testing of the drinking water from about 20 wells serving Redlands for trichloroethylene (TCE) was initiated in 1980 and for ammonium perchlorate in 1997. Perchlorate in the wells in 2001 was reported at 5-98 ppb (5-98 :g/L), with drinking-water concentrations not exceeding 18 ppb (18 :g/L). TCE in the wells initially ranged from 0.09 to 97 ppb (0.09-97 :g/L) but after water treatment or removal of highly contaminated wells from service has not exceeded 5 ppb (5 :g/L) in drinking water since 1991. Thus, residents of the 13 census tracts are known to have been exposed to various concentrations of TCE since 1980 and of ammonium perchlorate since 1997. It is assumed that perchlorate contamination in the wells was present as early as 1980. The numbers of observed cancers by site and age for the period January 1, 1988, through December 31, 1998, were compared with expected numbers for the same period. Residence at time of diagnosis was used to identify eligible cases. A total of 3,098 cancers occurred among residents during the period. Expected numbers were calculated by applying the average annual age, sex, and race-ethnicity-specific incidence rates within the four-county region for 1988-1992 to the population size of the 13 census tracts reported in the 1990 census and extrapolating popula- tion growth through 1998 without accounting for changes in age, sex, or racial or ethnic distributions. Analyses were done for all cancers combined, by specific sites, and separately for children younger than 15 years old. Standardized incidence ratios (SIRs) were not significantly different from 1.0 (when a conservative 99% confidence interval was used to account for multiple statistical tests) for all cancers combined (SIR, 0.97; 99% CI, 0.93-1.02) or for any specific cancer site, except for colon and rectum (SIR, 0.86; 99% CI, 0.74-0.99) and lung and bronchus (SIR, 0.71; 99% CI, 0.61- 0.81), which were lower than expected, and melanoma of the skin (SIR, 1.42; 99% CI, 1.13-1.77) and cancer of the uterine corpus (SIR, 1.35; 99% CI, 1.06-1.70), which were higher than expected. There was neither an excess nor a deficit of thyroid cancer (SIR, 1.00; 99% CI, 0.63-1.47); this estimate was based on 40 observed cases. The types of thyroid cancer were
Epidemiological Studies of Occupational and Environmental Exposures 91 not specified. No cancers were significantly increased in the analysis confined to children, in which all cancers combined, brain tumors, the leukemias, and thyroid cancer were individually examined. The authors speculated that the lower numbers of cancers of some sites were due to lower cigarette-smoking and greater use of screening programs in the community, which is relatively affluent. The higher incidence of uterine cancer was thought to reflect a higher prevalence of postmenopausal estro- gen therapy, and it was postulated that the higher incidence of melanoma was a result of failure to adjust completely for the risk associated with fair skin and of the increased use of health care, which led to earlier diagnosis of low-grade melanomas. No excess of thyroid cancer was observed in the residents, whose drinking water contained measurable TCE and ammonium perchlorate (Morgan and Cassady 2002). The duration of perchlorate contamination of some wells in Redlands before the onset of cancer may have been as short as 8 years or as long as 18 years if contamination began as early as 1980. The duration is uncertain, however, because perchlorate was not measured in well water until 1997. Thus, exposure duration may have been too short to detect outcomes that have long latency, such as thyroid cancer. Expected numbers were derived from the four-county region as a whole, which includes the exposed community, not from a ânonexposedâ area. That could result in an underestimate of the SIR. The proportion of the total registry derived from Redlands is not reported. Finally, the authors were unable to adjust for factors that might have confounded the analysis of drinking-water contaminants and cancer in the community. STUDIES IN NEONATES, CHILDREN, AND PREGNANT WOMEN Ecologic Studies Based on Newborn Screening Data Congenital Hypothyroidism A primary concern regarding exposure to perchlorate in drinking water is its potential effect on the developing fetus and the possibility of inducing or contributing to hypothyroidism in the newborn period. One of the first ecologic studies of perchlorate exposure and thyroid disease in newborns was done by Lamm and Doemland in 1999. The number of observed cases of congenital hypothyroidism in six counties in California and one county
92 Health Implications of Perchlorate Ingestion in Nevada (Clark County) was compared with that expected on the basis of statewide rates. All cases for 1996-1997 were identified through statewide mandatory neonatal blood screening programs. The criteria used to define congenital hypothyroidism were not stated. Nearly all the water supply for Clark County, which includes Las Vegas, comes from Lake Mead, which is known to be contaminated with perchlorate. Perchlorate in the countyâs water supply was measured at 4-16 ppb (4-16 :g/L). In California, per- chlorate in countiesâ water supplies is variable, and exposure is intermittent. Nevertheless, the six counties in California were assumed to be âexposedâ because they receive water from the Colorado River, in which perchlorate had been detected at 5-8 ppb (5-8 :g/L). Among the nearly 700,000 births in the 2-year period, 249 cases of congenital hypothyroidism were identi- fied, compared with 243 expected, adjusted for Hispanic ethnicity (SIR, 1.0; 95% CI, 0.90-1.16). There were no significant differences between the number of observed cases of congenital hypothyroidism and the number expected for any individual county or for all counties combined. In this study, the expected number was not based on the rates in nonexposed counties but rather on the rate in the entire state, which includes the exposed counties and potentially results in an underestimate of the SIR. Births in the exposed counties amounted to about 60% of all births in the two states. It is implied although not explicit that state-specific rates were used in the individual analyses of Nevada and California data. The expected concentra- tions were adjusted for county differences in the percentage of newborns of Hispanic ethnicity, but other potential confounding variables, such as birthweight, were not considered. Thyroid Hormone and TSH Production A series of ecologic studies examined differences in thyroid hormone and TSH production in newborns in Las Vegas (mean monthly concentra- tions of perchlorate in drinking water, undetectable to 15 ppb [15 :g/L]) and Reno, Nevada (no detectable perchlorate). The first study compared mean T4 concentrationsâobtained as part of the Nevada newborn blood screening programâin infants born in Las Vegas (n = 17,308) with concen- trations in those born in Reno (n = 5,882) over a 15-month period, from April 1998-June 1999 (Z. Li et al. 2000). The infants included had birth- weights of 2,500-4,500 g, had blood samples taken within 4 days, and had not been admitted to a neonatal intensive care unit. During that time, Las Vegas drinking water had perchlorate at 9-15 ppb (9-15 :g/L) for 7 months
Epidemiological Studies of Occupational and Environmental Exposures 93 (referred to as period A, April-June 1998 and March-June 1999) and no detectable perchlorate for 8 months (period B, July 1998-February 1999). T4 concentrations were measured in screening blood specimens taken within the first 4 days of life and in samples taken at infantsâ first pediatric visits within 60 days of birth. All T4 analyses were made for the state of Nevada by the Oregon State Public Health Laboratory. The estimated cumulative exposure to perchlorate in drinking water during pregnancy was 0.9-4.2 mg. When stratified by study period (that is, A or B), there was no significant difference in mean T4 concentrations between infants born in Las Vegas and those born in Reno. In both cities, T4 values were significantly higher in period B. After adjustment for infantsâ sex, age at sample collection, and birthweight, there was no significant difference between the two cities in mean newborn T4 concentrations. There were also no temporal differences between the cities in the relation of mean T4 concentration and the presence or absence of detectable perchlorate in Las Vegas water. There were no differences between cities in the prevalence of low neonatal T4 values (at or below the 10th percentile). Postneonatal T4 concentrations obtained within 60 days of birth also did not differ. Restriction of the study to newborns of normal birthweight (that is, excluding babies under 2,500 g and those over 4,500 g) may have reduced the studyâs ability to detect differences in mean T4 concentrations between Las Vegas and Reno newborns. Infants with congenital hypothyroidism have a higher average birthweight, so some cases may have been missed by excluding newborns with birthweights over 4,500 g. Low-birthweight and preterm infants, who were also excluded, are likely to be most vulnerable to the effects of iodide deficiency. At the time of those studies, Nevada used a two-stage screening proce- dure for thyroid function in which newborns who had T4 concentrations below the 10th percentile had a follow-up TSH blood test. Data on TSH concentrations, excluding those obtained in the first day of life, were compared in newborns in Las Vegas (n = 407) and Reno (n = 133) with birthweights of 2,500-4,500 g in the period December 1998-October 1999 (F.X. Li et al. 2000). Crude mean concentrations of TSH did not differ significantly between Las Vegas newborns (11.5 :U/mL) and Reno new- borns (12.5 :U/mL). Mean log TSH values also did not vary significantly by city after adjustment for age at specimen collection (2-7 days vs 8-30 days of life) and sex. TSH concentrations were significantly higher in the first age interval than in the second and higher in male than in female infants. Graphic presentation of the trends in mean monthly TSH values for those 2-7 days old by city and mean monthly perchlorate concentrations in Las Vegas water indicated no important differences between newborn TSH
94 Health Implications of Perchlorate Ingestion values in Las Vegas and Reno and no consistent temporal variations in perchlorate and TSH concentrations in Las Vegas infants. Exclusion of infants with low or high birthweights reduced the potential for confounding of the city comparisons by birthweight. Because a two-stage process was used, only infants with low T4 values were tested for TSH, so the distribu- tions of TSH concentrations in the two cities might be expected to be similar inasmuch as only one part of the distribution was examined. Brechner et al. (2000) compared median TSH concentrations, obtained as part of the Arizona newborn blood screening program from October 1994 to December 1997, in newborns in Yuma, which receives all its drinking water from a perchlorate-contaminated source, the Colorado River, and Flagstaff, which receives none of its drinking water from the Colorado River. In a two-stage screening program, infants who had the lowest 10% of T4 values were retested for TSH. TSH concentrations were determined in about 17% of the 7,599 newborns in Yuma, and in about 15% of the 3,539 newborns in Flagstaff. All assays were done in the Arizona Depart- ment of Health Services Laboratory. Perchlorate concentrations in commu- nity water were not available for the period of the study. However, 1999 measurements indicated perchlorate at 6 ppb (6 :g/L) in raw and finished water in Yuma and nondetectable concentrations in Flagstaff, and it was assumed that those findings applied to the earlier study period. Median TSH concentrations were statistically significantly higher in Yuma new- borns than in Flagstaff newborns (19.9 mU/L vs 13.4 mU/L). A higher percentage of samples were taken before the third day of life from Yuma newborns (69% for non-Hispanic and 71% for Hispanic infants) than from Flagstaff newborns (35% for non-Hispanic and 43% for Hispanic infants). After adjustment for age at sample collection and race or ethnicity, the mean log-transformed TSH values were still statistically significantly higher in Yuma than in Flagstaff (p = 0.017). The actual adjusted concentrations for each city were not reported. Neonatal T4 values did not differ signifi- cantly between Yuma and Flagstaff after adjustment for race or ethnicity. However, follow-up testing of TSH is done only in infants with the lowest 10% of T4 concentrations, so the absence of differences in T4 concentrations between the cities is not especially informative inasmuch as only the lowest part of the entire distribution was compared. In the Brechner et al. (2000) ecologic study, perchlorate exposures in infantsâ mothers were not directly measured; in fact, drinking-water concentrations of perchlorate were not derived from the same period as the newborn screening results. It is not known whether water perchlorate concentrations changed between 1994- 1997 and 1999, when they were measured for this study, although some
Epidemiological Studies of Occupational and Environmental Exposures 95 data suggest variations in measured exposures of 4-6 ppb (4-6 :g/L) and nondetectable for Yuma during the relevant period (Lamm 2003). Informa- tion on birthweight or gestational age was not available, so those potential confounders could not be addressed in the analysis. It has been noted by others (Lamm 2003) that there are several potentially relevant differences between Yuma and Flagstaff other than water perchlorate content. Flagstaff is about 7,000 ft above sea level, and Yuma is near sea level, at 192 ft (U.S. Geological Survey 2004). Infants born at higher elevations often require supplemental oxygen at birth and are generally of lower birthweight than infants born closer to sea level (Nahum and Stanislaw 2004). Low-birth- weight infants are more likely to have lower T4 concentrations without the magnitude of the TSH surge observed in term infants (Mercado et al. 1988). Thus, it is possible that more of the infants with low T4 who had a follow-up blood test for TSH in Flagstaff were low-birthweight infants who would not have had as great a TSH surge, keeping the mean and median concentra- tions in Flagstaff lower than in Yuma. In a commentary and follow-up analysis of the Brechner et al. (2000) study, Lamm (2003) conducted an alternative analysis in an attempt to account for geographic, ethnic, and medical-care differences between Yuma and Flagstaff that might have influenced the results of the earlier analysis. Newborn blood TSH concentrations in Yuma, Flagstaff, and San Luis and Somerton in September 1994-June 1998 were compared. San Luis and Somerton are in Yuma County, near the city of Yuma, but they do not get their municipal water from the Colorado River and thus are assumed to have no exposure to perchlorate. Their combined population is about half that of the city of Yuma. Data on TSH were obtained from the Arizona State Department of Health. The median TSH concentrations were 20.8 mU/L in Yuma, 21.0 mU/L in San Luis and Somerton, and 14.5 mU/L in Flag- staff. Yuma County, as a whole, had the highest median newborn TSH concentrations in the state. Those results suggest that Yuma County as a whole has relatively high TSH screening concentrations, regardless of whether a community has detectable perchlorate in its municipal water supply. However, socioeconomic, racial, ethnic, and low-birthweight differences between Yuma and San Luis and Somerton were not evaluated or controlled, so their influence on these values cannot be assessed. An unpublished ecologic study by Schwartz (2001) examined the relation of perchlorate exposure to prevalence of abnormal newborn blood screening concentrations of T4 and TSH and to presumptive and confirmed cases of congenital hypothyroidism. The study included 507,982 California infants born in 1996 for whom newborn screening results were available.
96 Health Implications of Perchlorate Ingestion Newborn screening samples were processed by eight laboratories in the state. Exposure to perchlorate was estimated on the basis of tests, con- ducted in 1997 or later, of samples from drinking-water sources that con- tributed to 380 public water systems in the state. Testing for perchlorate in California was initiated in systems that were suspected of being contami- nated. The 380 systems tested first served about half the stateâs population. As of 2001, 43% of the water sources of the 380 systems had been tested for perchlorate. The remaining, nontested water systems in the state were assumed to be negative for perchlorate exposure for the purpose of this study. Testing of 820 previously untested systems after the study found that 786 (96%) were negative for perchlorate (J. Schwartz, Impact Assessment, Inc., personal communication, April 9, 2004). Estimates of perchlorate exposure attempted to account for the mixing of water sources, and it was assumed that perchlorate concentrations measured after 1996 reflected those in 1996. Concentrations were averaged across samples from each water source, and average exposures were assigned to each ZIP code. Infantsâ ZIP codes were based on the motherâs residence at time of delivery. Be- cause of the lack of precision of measurements, perchlorate exposure was categorized as none (n = 251,026), low (1-2 ppb [1-2 :g/L]; n = 125,373), medium (3-12 ppb [3-12 :g/L]; n = 129,661), or high (at least 13 ppb [13 :g/L]; n = 1,922). The unadjusted mean concentrations of T4 (:g/dL) by perchlorate exposure category were as follows: none, 17.09; low, 16.21; medium, 16.06, and high, 15.05. The crude mean concentrations of TSH (:U/mL) in the same groups were 7.6, 7.6, 7.7, and 7.9. After statistical adjustment for infant sex, ethnicity, multiple birth, birthweight, and age at time of specimen collection, perchlorate category was significantly associated with progressive decrements in mean T4 concentrations (!0.97, !1.12, and !1.82 :g/dL in the low, medium, and high exposure categories, respectively) and with small but statistically significant increases in ln TSH values (0.029, 0.030, and 0.13 :U/mL). Absolute concentrations for multivariate-adjusted TSH by perchlorate category were not stated. The reductions in T4 concen- trations associated with low birthweight and with early age at specimen collection (7-18 hr after birth) were about 4-7 times those in the highest perchlorate category. Perchlorate category was significantly associated with the presumptive diagnosis of congenital hypothyroidism but not in the predicted direction. Compared with infants living in ZIP codes that had no measurable perchlorate in the water, the odds of presumptive congenital hypothyroidism by category of exposure were 1.15 (95% CI, 1.12-1.17), 1.07 (95% CI, 1.05-1.10) and 1.05 (95% CI, 0.91-1.21) in the low, medium,
Epidemiological Studies of Occupational and Environmental Exposures 97 and high exposure categories, respectively. Because of the small number of confirmed cases of congenital hypothyroidism, perchlorate exposure was dichotomized as ânoneâ vs âany.â On the basis of that categorization, there was no significant association between diagnosed congenital hypothy- roidism and living in a ZIP code that had measurable perchlorate in the drinking water. The study of Schwartz was based on a large number of newborns and thus was able to detect small differences in blood screening concentrations of T4 and TSH among exposure groups if they existed. The clinical impor- tance of the decreases in T4 with increasing perchlorate is difficult to determine, because the decreases were accompanied by very small changes in TSH. No significant association was observed between perchlorate and confirmed congenital hypothyroidism when exposure was dichotomized. Although important confounders were considered in the analyses, the potential for misclassification of perchlorate exposure remains. No mea- surement of perchlorate had been done in a substantial percentage of water sources that were used to characterize each water system; it was assumed that later measurement reflected 1996 measurements and that nontested systems were negative, and, as in other ecologic studies, there was no measurement of perchlorate exposure in the water to which individual infants were exposed. Nondifferential misclassification of exposure tends to bias associations toward the null value, but only for dichotomous expo- sure variables. When exposures are categorized at more than two levels, nondifferential misclassification can introduce biases both toward and away from the null in the same dataset (Rothman and Greenland 1998). In addition, the investigator was unable to control for laboratory variation in thyroid hormone and TSH measurements. Because eight laboratories in California are responsible for conducting newborn screening tests, small variations in results among geographic areas served by different laboratories may have contributed to the observed differences. Kelsh et al. (2003) examined the frequency of congenital hypothyroid- ism and of high blood TSH concentrations among newborns in 1983-1997 whose mothers resided in one of two California communities and who were screened as part of the California Newborn Screening Program. Measure- ments of perchlorate in drinking water were made by the California Depart- ment of Health Services Drinking Water Program, which began testing in 1997. The âexposedâ community included 13 census tracts in Redlands and Mentone serviced by the Redlands Municipal Water District in which perchlorate had been detected in the water system at up to 9 ppb (9 :g/L), with a calculated mean concentration below 1 ppb (1 :g/L). The compari-
98 Health Implications of Perchlorate Ingestion son communities were San Bernardino and Riverside counties, which were adjacent to Redlands and used the same newborn screening laboratories but in which perchlorate had not been detected in the water supply. Because assessment of thyroid function was ascertained in the immediate newborn period, the relevant exposure was assumed to be fetal. Exposure data obtained in 1997 were assumed to apply to the entire study period. Cases were infants in whom congenital hypothyroidism was diagnosed or whose screening concentration of TSH was âelevated,â usually defined as over 25 :U/mL but sometimes over 16 :U/mL. During the study period, the California Newborn Screening Program measured TSH only among infants who first screened in the lower 5% of T4 concentrations. Thus, TSH con- centrations are available only for infants who also had low T4 concentra- tions. Potentially confounding variables included in the analyses were age at specimen collection (in hours since birth, categorized as less than 6, 6 to less than 12, 12 to less than 24, and 24 and higher), sex, race or ethnicity, birthweight, multiple birth status (yes or no), and calendar year of birth. About two-thirds of the newborns in both âexposedâ and nonexposed communities had their TSH screening blood specimens collected at 18 hr or more after birth (Kelsh et al. 2003). A total of two cases of congenital hypothyroidism were found among the 15,348 Redlands births, compared with the expected 4.66 cases based on data from San Bernardino and Riverside births. The standardized prevalence ratio (SPR) for congenital hypothyroidism in Redlands was 0.45 (95% CI, 0.06-1.64), adjusted for infantâs sex, ethnicity, birthweight, and birth year. Multivariate analysis found no statistically significant relation between residence in Redlands and the odds of âelevatedâ TSH among newborns whose specimens were collected 18 hr or more after birth (odds ratio [OR], 0.72; 95% CI, 0.28- 1.54). The analysis was based on six cases of âelevatedâ TSH in the Redlands newborns. Additional analyses, specifically excluding from the control communities infants who were born in areas with any measurable perchlorate or infants in communities that receive water from the Colorado River, did not alter the results in any meaningful way. As an ecologic study, the Kelsh et al. (2003) study is limited in that exposure of individual mothers is unknown. In addition, exposure data from a single year were used to characterize exposures over the entire 15 years of the study. Whether residence in areas with measurable perchlorate was associated with adverse thyroid outcomes in infants of low birthweight was not specifically addressed. Birthweight was analyzed as a continuous variable in the multivariate analysis, in spite of the reported observation of a U-shaped relation between birthweight and TSH concentrations.
Epidemiological Studies of Occupational and Environmental Exposures 99 Buffler et al. (2004) extended the Kelsh et al. (2003) study by analyzing data from the California Newborn Screening program for 1998. They included all births in the state to mothers residing in communities in which the drinking water had been tested for perchlorate in 1997-1998. âEx- posedâ communities were those in which the mean perchlorate concentra- tion in drinking water was over 5 ppb (5 :g/L), and ânonexposedâ commu- nities were those in which perchlorate was not detectable or the average concentration was 5 ppb (5 :g/L) or lower. Data on congenital hypothy- roidism and high blood TSH (generally defined as a screening value over 25.0 :U/dL) were compared. In 1998, the two-tiered TSH sampling strat- egy was discontinued, and all samples collected at the age of 24 hours or later were included. SPRs for congenital hypothyroidism were calculated as the ratio of the observed number of cases in the âexposedâ group to the number expected on the basis of rates in the ânonexposedâ group, standard- ized for ethnicity, sex, and birthweight. The odds of congenital hypothy- roidism and of high TSH values (categorized as âhighâ or ânormalâ) in âexposedâ compared with ânonexposedâ newborns was estimated with logistic regression methods, with adjustment for infantsâ sex, ethnic status, birthweight (continuous for TSH analysis, categoric for primary hypothy- roidism analysis), year of birth, and age at blood collection (for TSH). The authors reported no statistically significant relation between exposure to perchlorate over 5 ppb (5 :g/L) and the prevalence of or odds of congenital hypothyroidism or increased TSH values (Buffler et al. 2004). Their analysis addressed some of the limitations noted in the earlier study of Kelsh et al. (2003). Specifically, the timing of the drinking-water assess- ments was concordant with the birth cohort, and because the data covered the entire state, the observed numbers of cases of congenital hypothy- roidism (15) and of increased TSH level (147) in the âexposedâ areas were considerably higher than in the prior study. However, exposure assessment on an individual level was still lacking. Thyroid Hormone and TSH Production in Children Crump et al. (2000) compared thyroid hormone and TSH production and the frequency of thyroid disease in newborns and in school-age children in three Chilean cities with different perchlorate concentrations in their municipal water supplies. Perchlorate was measured in drinking-water samples in each city drawn from potable-water faucets in schools, homes of the children, and public buildings near the schools. This is the only study
100 Health Implications of Perchlorate Ingestion in children that was based on outcome measures in individual children and on perchlorate measurements in water from taps accessible to the children for drinking water. The analysis of data from newborn screening is a purely ecologic design. The three cities were Antofagasta (no detectable perchlor- ate), Chanaral (5-7 ppb [5-7 :g/L]), and Taltal (100-120 ppb [100-120 :g/L]). The cities are west of the Andes Mountains in the Atacama Desert of northern Chile, one of the most arid regions on Earth. The region re- ceives measurable rainfall as infrequently as once in 5-20 years, so there is no farming. Water is supplied to Antofagasta from a pipeline on the west- ern margins of the Andes, an area that contains large amounts of naturally occurring perchlorate. Potable water for all other cities in the area comes from ground water in wells in alluvial basins. All water samples were analyzed in the same laboratory, and technicians did not know the city from which the samples came. For the study of school-age children, first- and second-graders 6-8 years old (50-60 children per city) were recruited from one or two public schools in each city in 1999. They were of similar ethnicity and socioeconomic status. More than 90% of their parents agreed to their participation; this involved completion of a questionnaire by parents and collection of blood and urine specimens from the children. The questionnaire included items related to family history of thyroid disease and residential history. Blood samples were drawn from the 162 participating children between 9 a.m. and 3 p.m. Studies included serum TSH, T4, free T4, and T3 concentrations and tests of liver and kidney function. A first-void, spot urine sample was obtained for measurement of urinary iodide and creatinine. Children were also examined for evidence of goiter by one of three endocrinologists who did not know the perchlorate concentrations in the study cities. Group analyses were performed for the 127 children who had lived their entire lives, from conception to testing, in the same city. Neonatal thyroid screen- ing has been mandatory in Chile since 1992, and all samples for the entire country are processed in a single facility. For the present study, newborn samples were obtained from the 9,784 children born from February 1996 to January 1999 in the three cities. Among children who were life-long residents in their city, there were no statistically significant differences in serum TSH levels among the three cities after adjustment for age, sex, and urinary iodide concentration (Crump et al. 2000). Serum free T4 levels were significantly higher in children in Chanaral and Taltal than Antofagasta after adjustment for age, sex, and urinary iodide levels; the difference was opposite the direction predicted on the basis of competitive perchlorate inhibition of iodide uptake. There were
Epidemiological Studies of Occupational and Environmental Exposures 101 no significant differences among the cities in the prevalence of goiter among the schoolchildren (22%, 20%, and 26% in Antofagasta, Chanaral, and Taltal, respectively). Compared with Antofagasta, however, the multivariate-adjusted odds of having a relative with a history of goiter, hypothyroidism, or subtotal thyroidectomy was nearly 5 times as high in Taltal (OR, 4.97; 95% CI, 1.29-19.17) and was not significantly different in Chanaral (OR, 1.04; 95% CI, 0.21-5.09). Family history of goiter was reported on the study questionnaire and included parents, siblings, grand- parents, great-grandparents, aunts, uncles, and cousins. Crump et al. (2000) suggest that the high prevalence of a family history of goiter, which was seen only in Taltal, may have resulted from a combination of high perchlor- ate and low iodide intake before the introduction of iodized salt into the region in 1982. The country has since been identified by the Iodine Defi- ciency Disease Prevalence and Control Program as an area of iodide excess, with a population median urinary iodide excretion of 54 :g/dL (Interna- tional Council for the Control of Iodine Deficiency Disease 2001, 2002). Multivariate-adjusted comparisons for serum T3 concentrations were not presented, but univariate analyses indicated no differences in mean values among the three cities for all children (Crump et al. 2000). Mean urinary iodide levels in the 6- to 8-year-olds did not differ significantly: 75.6 Â± 40.4 :g/dL in Antofagasta, 61.4 Â± 35.7 :g/dL in Chanaral, and 76.6 Â± 47.4 :g/dL in Taltal. No statistically significant differences were ob- served among children from the three cities in measures of bone marrow, liver, or kidney function. After adjusting for sex and age of testing (in days), newborns in Taltal had statistically significantly lower log serum TSH concentrations; mean log TSH levels did not vary significantly between newborns in the other two cities (Crump et al. 2000). The lower mean in Taltal is, again, opposite of what would be expected in association with increased exposure to per- chlorate. During the period of observation, seven presumptive cases of congenital hypothyroidism (TSH, at least 25 :U/mL) were identified, all in infants born in Antofagasta. The committee thinks that the study of Crump et al. (2000) had impor- tant strengths. Although individual exposures were not assessed, it is one of the few studies that measured perchlorate in drinking water in samples taken directly from the environment of the children studied, such as homes and schools. In addition, it was possible to compare assessments of thyroid function, other end points, and potential risk factors obtained in a systematic manner from all participants and adjusted for a number of important co- variates, and participation was high. All laboratory assessments were done
102 Health Implications of Perchlorate Ingestion at the same facility, and assessments of thyroid status and other measures were done by observers unaware of the perchlorate exposure of the chil- dren. All newborn screening tests in Chile are done at a single laboratory. Numerous critiques of the Crump et al. (2000) study have been done either directly by the U.S. Environmental Protection Agency (EPA) in its risk-assessment document or by others at the request of EPA (Park 2001; Marcus 2003). Reanalysis by EPA was also done in 2002 with only the published data. The authors have responded to each critique. With respect to the analyses done among school-age children, the study was criticized for failing to adjust for differences among the three cities in ethnicity and socioeconomic status. Ethnicity has been shown in some studies to be related to risk of congenital hypothyroidism or to thyroid hormone concen- trations, at least in newborns (Lamm and Doemland 1999; Brechner et al. 2000; Schwartz 2001; Kelsh et al. 2003). The authors provided data to support their contention that there are no important differences among the three cities in ethnicity, socioeconomic status, or access to medical care (Gibbs 2003a). Many of the data supplied in the Gibbs (2003a) letter come from a study of pregnant women in the three cities, which is described below. The analyses of newborn screening data have been criticized on the basis of failure to adjust for differences among the cities in iodide intake, ethnicity, and birthweight. No other studies have included adjustment for maternal iodide intake, so this concern is not peculiar to the study in Chile. In the analysis of data from school-age children, there were no differences in urinary iodide excretion between the lowest- and highest-exposure cities; urinary iodide in Chanaral (the middle-exposure city) was about 15 :g/dL lower than those in Antofagasta and Taltal. Similar findings are reported in unpublished preliminary data from the study of pregnant women. The committee does not think that failure to adjust for maternal iodide intake can explain the significantly lower TSH concentrations in Taltal newborns. There are no differences in ethnicity, so it is not necessary to adjust for this variable. Finally, comparisons of newborn thyroid hormone screening concentrations were criticized because of failure to adjust for birthweight. Information was not provided on the distributions of birthweights or gesta- tional ages in the three cities, so it is not possible to assess whether the newborn screening results might have been influenced by differences in them. The study was also criticized because of differences in the median postnatal day on which newborn samples were tested for serum TSH. The committee considers this issue adequately addressed by Crump et al. (2000),
Epidemiological Studies of Occupational and Environmental Exposures 103 in that they compared TSH concentrations by single days of age (Table 8, p. 609), and age at sample collection was included in the multivariate analysis of TSH (Table 9, p. 610). The possibility was raised that differences among the three cities in ambient indoor and outdoor temperatures may have affected measures of thyroid hormone in newborns. The authors provide data to show that the climate is very similar in the three areas and that, in any case, no differences would be expected in the delivery suites in the three locations (Gibbs 2003a). The committee considers this criticism adequately addressed by the authors. The Marcus (2003) critique also points to high coefficients of variation (CVs) in the endocrine measures of the study by Crump et al. (2000), citing CVs (%) of 45-54% for serum TSH and 50-64% for urinary iodide. Marcus (2003) suggests that this variability is evidence of poor laboratory measure- ment techniques and an absence of quality-control procedures. In response, the authors provide data comparing the CVs for measurements in their study with those of other published work (Gibbs and Crump 2003). Those data represent interassay variation rather than interindividual variation and indicate no differences in precision of measures between Crump et al. (2000) and reports from other laboratories. Many of the issues raised in regard to the study by Crump et al. (2000) apply equally to other ecologic studies and are not peculiar to the study done in Chile. They include the possibility of âuncontrolled confounders,â which can be present in any epidemiologic study, and questions regarding the characterization of exposure. The major criticisms of the study of Crump et al. (2000) are related to the high dietary iodide intake in the populations and to the high prevalence of goiter in the children themselves (as assessed by physical examination) and their family members (based on mothersâ reports). EPA considered the high prevalence of goiter to be an indication of thyroid abnormalities in the population and further evidence of the unsuitability of the data for inference to the U.S. experience. The mean urinary iodide excretion in children in the three Chilean cities (61-77 :g/dL) is about 3 times as high as that in U.S. children 6-9 years old in 1988-1994 (geometric means, 25.2 and 20.3 :g/dL in boy and girls, respectively) (Crump et al. 2000; Soldin et al. 2003). The prevalence of goiter at which iodide deficiency is considered to be a public- health problem was defined by the World Health Organization (WHO) in 1994 as 5% (WHO 1994). The proportions of lifelong resident, 6- to 8- year-old children who had clinically diagnosed goiter were 22%, 20%, and 26% in Antofagasta, Chanaral, and Taltal, respectively, exceeding the WHO
104 Health Implications of Perchlorate Ingestion value of 5% in spite of the apparently high iodide intake. Goiter in the children was assessed by palpation, and the presence of goiter was based on the 1960 WHO criteria, which, although more valid than the revised, 1994 WHO criteria, has a specificity of only about 76% in a high-prevalence area (Peterson et al. 2000). Basing the diagnosis of goiter solely on the results of palpation has been shown to have good reproducibility within observers but not between observers and to have poor specificity compared with ultrasonography, resulting in a high number of false positives (Peterson et al. 2000). Therefore, the committee did not consider the higher than ex- pected prevalence of goiter to be grounds for dismissing the results of the Chilean study. It is suggested in EPAâs critiques that the high dietary iodide intake makes this geographic area unsuitable for studies of the thyroid effects of perchlorate exposure. The contention is that the iodide intake is so high that substantial competition with perchlorate cannot be detected. Because EPA did not supply data to support that contention, the committee performed calculations on the effect of basal serum iodide concentrations in humans on the inhibition of iodide uptake by the thyroid sodium (Na+)/iodide (IG) symporter (NIS) in the presence of perchlorate. The Michaelis-Menten competitive-inhibition equation was used for the calculations; the assump- tions used in each model are included in Appendix D. A second set of calculations produced a series of curves of perchlorateâs inhibition of iodide uptake across a 105-fold range of basal serum iodide concentrations (Appen- dix D). On the basis of those models, serum iodide concentrations in humans over a 10-fold range (1.5-15 :g/dL) are not likely to have a pro- found influence on the ability to detect a perchlorate effect on the thyroid NIS. In addition, a basal serum iodide concentration over 100 :g/dL would be required to shift the dose-response curve for perchlorateâs inhibition of iodide uptake; this suggests that serum iodide concentrations within 0-100 :g/dL would be equally sensitive to perchlorateâs effects. On the basis of the calculations, the committee concludes that iodide intake in the popula- tion in question does not interfere with the studyâs ability to detect effects of perchlorate exposure at up to 120 ppb (120 :g/L) on iodide uptake, if effects are present. It should also be noted that in spite of the contention that dietary iodide is too high in the population to allow the detection of a perchlorate effect, an effect on serum free T4 concentrations was observed (albeit inverse). The mean serum T4 value was highest in Taltal, the city with the highest exposure to perchlorate. In addition to those calculations, the committee compared serum con- centrations of TSH and thyroid hormones reported by Crump et al. (2000)
Epidemiological Studies of Occupational and Environmental Exposures 105 in children 6-8 years old with those children of similar ages in the United States during a comparable period (Zurakowski et al. 1999) as another approach to assessing whether data from the Chilean study might be useful for evaluating the U.S. experience. With the exception of serum T3, which is known to vary more depending on the assay used, the mean serum thy- roid hormone concentrations were not markedly different between U.S. and Chilean children. The mean TSH concentrations were somewhat higher in Chile but were not accompanied by decreases in serum T4 concentrations. Those results suggest that mean serum thyroid hormone concentrations in Chilean children were similar during the period to those in U.S. children of the same age. On the basis of the iodide-inhibition analyses, the additional comparisons, and a review of information on urinary iodide excretion, the committee concluded that the data from Chile could be considered in the evaluation of the U.S. experience with perchlorate in drinking water. Thyroid Hormone and TSH Production in Pregnant Women and Their Newborns A second study among pregnant women and their newborns in each of the three cities in Chile examined the relation between perchlorate exposure during pregnancy, indexes of maternal thyroid function, and thyroid hor- mone concentrations in the newborns (Gibbs 2003b). Serum T3, free T4, TSH, thyroid peroxidase antibodies, antithyroglobulin antibodies, and thyroglobulin and urinary iodide were measured in mothers during the first and third trimesters and after birth. Maternal blood specimens from the first trimester, maternal urine specimens from the first and third trimesters and after birth, umbilical cord blood, and samples of breast milk were tested for perchlorate. Water samples from the pregnant womenâs homes were ana- lyzed for perchlorate to provide individual home-exposure estimates. Preliminary unpublished results on maternal urinary iodide, serum T3, free T4, thyroglobulin, and TSH concentrations at the first and second prenatal visits, cord blood thyroid hormones, and perchlorate in breast milk were presented at a public meeting in May 2004 (Gibbs 2004a). Analyses of accrued data were provided to the committee in a written report in August 2004 (Gibbs 2004b) and are presented here. Mean urinary iodide in mothers at the first prenatal visit was significantly different among the three cities, with mean levels being lowest in Taltal (323 :g/g creatinine in 62 samples compared with 407 :g/g creatinine in 61 samples in Anto- fagasta, and 363 :g/g creatinine in 39 samples in Chanaral). Maternal mean
106 Health Implications of Perchlorate Ingestion serum free T4 and TSH concentrations from the first prenatal visit were similar among the three cities, although mothers in Chanaral had the highest mean TSH values (2.81 :U/mL vs 2.63 :U/mL in Antofagasta and 2.61 :U/mL in Taltal). Mean serum T3 was significantly different in first-tri- mester samples. Mean T3 was similar in mothers in Antofagasta and Taltal (183 ng/dL and 187 ng/dL, respectively), but it was higher in Chanaral (207 ng/dL). Mothers identified at the first prenatal visit as hypothyroid (serum TSH, at least 4.5 :U/dL) were treated. Five such mothers were identified in each of the three cities. At the second prenatal visit, serum T3 values were significantly different, with the lowest mean in Taltal (173 ng/dL), but they were not adjusted for week of gestation, which was longest at the time of sampling in Taltal (33.2 weeks). Mean serum free T4 was also signifi- cantly different among the cities. Serum free T4 values were similar in Chanaral and Taltal (0.82 ng/dL and 0.83 ng/dL, respectively), but higher in Antofagasta (0.86 ng/dL), again, no adjustment was made for differences in length of gestation. Median iodide concentrations in samples of breast milk from mothers were 36.9 :g/dL in Antofagasta (16 samples), 29.5 :g/dL in Chanaral (16 samples), and 38.4 :g/dL in Taltal (25 samples). Median perchlorate concentrations in breast milk were less than 0.5, 19.3, and 103.8 ppb (:g/L) respectively (Gibbs 2004b). Only mean T3 was significantly different in cord blood samples; the value was lowest in Chanaral at 73 ng/dL compared with 79 ng/dL in Antofagasta and 82 ng/dL in Taltal. In data presented in May but not updated in August, perchlorate was detectable in cord blood specimens from infants born in Taltal, but not in Chanaral. It should be noted that no analyses have been adjusted by gestational age or birthweight, so the results presented for newborns are preliminary. The study had not been published at the time of the committeeâs deliberations; therefore, it did not use the results in formulating its conclusions. Ecologic Studies of Neurodevelopment in Children There are sparse data concerning the relation between perchlorate exposure and neurodevelopmental disorders in children. One such study is that of attention-deficit-hyperactivity disorder (ADHD) and autism by Chang et al. (2003), which examined the association between residence in Nevada communities with and without detectable perchlorate in the public water supply and diagnoses of either ADHD or autism in children less than 18 years old who were recipients of Medicaid. Scores on tests of 4th grade
Epidemiological Studies of Occupational and Environmental Exposures 107 performance were also compared with national values for 1998-1999 and 2001. The three comparison groups were in Clark County, which includes Las Vegas and in which the public water supply has perchlorate ranging from undetectable to 23.8 ppb (23.8 :g/L) (median, 10.5 ppb [10.5 :g/L]) as measured in 1997-2001; Washoe County, which includes Reno and has no detectable perchlorate in the public water supply; and the remainder of Nevada, was considered a âruralâ control for Las Vegas. The rural areas have no detectable perchlorate in public water supplies. Data from the service records of the Nevada Medicaid Program for calendar years 1996- 2000 were used to ascertain the total number of Medicaid recipients less than 18 years old for each year and the number of cases with a first, under- lying diagnosis of ADHD or autism. The midpoint Medicaid population size for that period and the average number of cases per year in each of the three geographically defined groups were used to calculate the average annual proportion of Medicaid recipients under 18 years old with ADHD or autism. For ADHD, the annual numbers of cases per 1,000 Medicaid recipients under 18 years old were 17, 28, and 29 for Clark, Washoe, and the rest of the state, respectively; for autism, the annual numbers of cases per 1,000 were 1.4, 2.8, and 1.2. Thus, Clark County, the only area exposed to perchlorate in drinking water, did not have a higher proportion of ADHD or autism than non- exposed communities. Similarly, 4th grade proficiency scores, expressed as national percentile ranks, did not differ significantly among the three areas. A comparison of 2001 4th grade achievement scores allowed inclu- sion only of children who had resided in the same community for at least 3 years. The results also showed no important differences with respect to 4th grade achievement among the three exposure groups. Other than the single analysis of the 2001 4th grade achievement scores, the data cannot account for any changes in childrenâs residence. They also do not reflect exposures to perchlorate at a time in neurologic development that is considered biolog- ically relevant, at least for autism, namely early gestation. As an ecologic study, the study measured neither exposure nor outcome in individuals. In addition, the specific criteria used to make individual diagnoses of ADHD or autism are unknown; diagnoses were made by multiple health-care providers and without the use of uniform diagnostic criteria. Thus, the validity of the diagnoses cannot be assessed, and geographic differences could reflect diagnostic practices in local communities rather than true differences in risk. Furthermore, no adjustment was made for potentially confounding factors, such as age, sex, race, and ethnicity. Finally, the end points studied are not the most sensitive indicators of the kinds of neuro-
108 Health Implications of Perchlorate Ingestion developmental outcomes that might be predicted on the basis of prior studies of the effects of hypothyroidism on the developing nervous system. The indicators include subtle impairments in cognitive and motor function, such as those observed in children who have untreated or inadequately treated congenital hypothyroidism. In addition, neither autism nor autistic- spectrum disorder has been observed previously in association with thyroid hormone deficiencies. SUMMARY Limitations of Existing Data Epidemiologic studies have examined possible associations between environmental exposure to perchlorate in drinking water at about 4-120 ppb (4-120 :g/L) and abnormalities of thyroid hormone and TSH production in newborns, as well as thyroid diseases (such as congenital hypothyroidism and goiter) and cancer in children and adults (Lamm and Doemland 1999; Brechner et al. 2000; Crump et al. 2000; F.X. Li et al. 2000; Z. Li et al. 2000; Schwartz 2001; Morgan and Cassady 2002; Kelsh et al. 2003; Lamm 2003; Buffler et al. 2004). Occupational studies of respiratory exposures up to 0.5 mg/kg perchlorate per day and abnormalities of thyroid hormone and TSH production in adult workers have been conducted (Gibbs et al. 1998; Lamm et al. 1999; Braverman et al. 2004). Only one study has examined a possible relation between perchlorate exposure and adverse neurodevelopmental outcomes (ADHD and autism) in children (Chang et al. 2003). A number of the studies have samples that are too small to detect differences in frequency of outcomes between exposure groups, and adjust- ment for potentially confounding factors was limited. Nearly all the studies were ecologic, including those in newborns and children, the groups poten- tially most vulnerable to the effects of perchlorate exposure. Ecologic studies can provide supporting evidence of a possible association but cannot themselves provide definitive evidence regarding cause. Perchlorate expo- sure of individuals is difficult to measure and was not assessed directly in any of the studies conducted outside the occupational setting. The only study with measures of perchlorate made directly from drinking-water samples taken from faucets potentially used by people who were studied was that done in Chile by Crump et al. (2000). No studies have examined the relation of perchlorate exposure and adverse outcomes, either in thyroid function or in neurodevelopment,
Epidemiological Studies of Occupational and Environmental Exposures 109 among especially vulnerable groups, such as low-birthweight or preterm infants. In addition, the available studies do not assess the possibility of adverse outcomes associated with perchlorate exposure in infants born to mothers who had inadequate dietary iodide intake. Thus, no direct human data are available regarding a possible interaction between maternal iodide intake and perchlorate exposure. Although the ecologic design is inherently limited with respect to establishing causality, results of such studies can be informative when combined with other data on the biology of the thyroid gland, experimental studies of the effects of acute exposure to perchlorate, and studies of occu- pational perchlorate exposure. Committee Conclusions Drawn from Epidemiologic Data on Specific Health Outcomes Congenital Hypothyroidism Ecologic data alone are not sufficient to demonstrate whether or not an association is causal, but they do provide evidence that can be used to evaluate possible associations. Acknowledging that ecologic data alone are inherently limited in drawing causal inferences, the committee concludes that the available epidemiologic evidence is not consistent with a causal association between perchlorate exposure and congenital hypothyroidism as defined by the authors of the studies reviewed here. All studies of this association were negative. Perturbation of Thyroid Hormone and TSH Production in Newborns Again given the limitations of ecologic data in inferring causation, the available epidemiologic evidence is not consistent with a causal association between perturbations of thyroid hormone and TSH production in normal newborns (that is, not low-birthweight or preterm) and exposure during gestation to perchlorate in drinking water at up to 120 ppb (120 :g/L). Most studies do not show either significantly lower T4 concentrations or significantly higher TSH concentrations among infants born in geographic areas in which the water supply has measurable perchlorate.
110 Health Implications of Perchlorate Ingestion However, no epidemiologic studies are available on the association between perchlorate exposure and thyroid dysfunction among low-birth- weight or preterm newborns, offspring of mothers who had iodide defi- ciency during gestation, or offspring of hypothyroid mothers. Those are the groups of greatest concern with respect to potential effects of perchlorate exposure. Neurodevelopmental Outcomes The epidemiologic evidence is inadequate to determine whether or not there is an association between perchlorate exposure and adverse neurodevelopmental outcomes in children. The only relevant study used an ecologic design and examined autism and ADHD as end points. Subtler neurodevelopmental outcomes have not been assessed in human popula- tions. Although inclusion of ADHD was considered plausible, the commit- tee questioned the appropriateness of autism as an end point. Autism has not been observed among the spectrum of clinical outcomes in children who had congenital hypothyroidism and were evaluated prospectively (Rovet 1999, 2002, 2003). Thyroid Diseases and Hypothyroidism in Adults On the basis of data from studies of chronic occupational exposures to ammonium perchlorate and ecologic investigations in adults, the committee concludes that the epidemiologic evidence is not consistent with a causal association between exposure to perchlorate at the concentrations studied and thyroid diseases in adults. The thyroid diseases and thyroid measures investigated included simple and nonspecified goiter, nontoxic nodular goiter, thyrotoxicosis with or without goiter, acquired hypothyroidism, thyroiditis, and other disorders of the thyroid, and measures of serum TSH and the two thyroid hormones. In occupational studies, perchlorate expo- sures as high as 0.5 mg/kg per day have not been associated with adverse effects on thyroid function in workers, but small samples in some studies may have reduced the ability to identify important differences, and studies were limited to workers who remained in the workforce.
Epidemiological Studies of Occupational and Environmental Exposures 111 Thyroid Cancer in Adults The epidemiologic evidence is insufficient to determine whether there is a causal association between perchlorate exposure and thyroid cancer. Only two studies related to this issue have been done, and both were eco- logic. In one study, the number of thyroid-cancer cases was too small to have a reasonable chance of detecting an association if one existed (Li et al. 2001). In the second, larger study (Morgan and Cassady 2002), mixed exposures were present (to perchlorate and TCE). In neither study was it possible to adjust for potential confounding variables. The committee notes, however, that on the basis of its understanding of the biology of human and rodent thyroid tumors, it is unlikely that perchlorate poses a risk of thyroid cancer in humans. REFERENCES Braverman, L.E., X. He, S. Pino, M. Cross, B. Magnani, S.H. Lamm, K. Crump, and J. Gibbs. 2004. The effect of perchlorate, thiocyanate, and nitrate on thyroid function in long-term workers exposed to perchlorate. Presentation at the Fourth Meeting of the Committee to Assess the Health Implications of Perchlorate Ingestion, May 24, 2004, Woods Hole, MA. Brechner, R.J., G.D. Parkhurst, W.O. Humble, M.B. Brown, and W.H. Herman. 2000. Ammonium perchlorate contamination of Colorado River drinking water is associated with abnormal thyroid function in newborns in Arizona. J. Occup. Environ. Med. 42(8):777-782. Buffler, P.A., M.A. Kelsh, E.C. Lau, C.H. Edinboro, and J.C. Barnard. 2004. Epidemiologic Studies of Primary Congenital Hypothyroidism and Newborn Thyroid Function Among California Residents, Final Report. April 2004, Berkeley, CA. Chang, S., C. Crothers, S. Lai, and S. Lamm. 2003. Pediatric neurobehavioral diseases in Nevada counties with respect to perchlorate in drinking water: An ecological inquiry. Birth Defects Res. Part A Clin. Mol. Teratol. 67(10):886- 892. Crump, C., P. Michaud, R. Tellez, C. Reyes, G. Gonzalez, E.L. Montgomery, K.S. Crump, G. Lobo, C. Becerra, and J.P. Gibbs. 2000. Does perchlorate in drinking water affect thyroid function in newborns or school-age children? J. Occup. Environ. Med. 42(6):603-612. Gibbs, J.P., R. Ahmad, K.S. Crump, D.P. Houck, T.S. Leveille, J.E. Findley, and M. Francis. 1998. Evaluation of a population with occupational exposure to airborne ammonium perchlorate for possible acute or chronic effects on thyroid function. J. Occup. Environ. Med. 40(12):1072-1082.
112 Health Implications of Perchlorate Ingestion Gibbs, J.P. 2003a. Comments and Some Information About Specific Issues, Rafael Tellez. Letter to R. Johnston, Committee to Assess the Health Implications of Perchlorate Ingestion, from J.P. Gibbs, Kerr-McGee Corp., Oklahoma City, OK. December 16, 2003. Gibbs, J.P. 2003b. Comments Regarding Previous and Ongoing Studies in North- ern Chile. Protocol and Preliminary Data from an Ongoing Study Among Pregnant Women and Newborns in the Same Three Cities as Studied by Crump et al (2000). Letter to E.K. Mantus, National Academy of Sciences, from J.P. Gibbs, Kerr-McGee Corp., Oklahoma City, OK. October 27, 2003. Gibbs, J.P. 2004a. Preliminary: Chronic Environmental Exposure to Perchlorate in Drinking Water and Thyroid Function During Pregnancy and the Neonatal Period. Presentation at the Fourth Meeting of the Committee to Assess the Health Implications of Perchlorate Ingestion, May 24, 2004, Woods Hole, MA. Gibbs, J.P. 2004b. Chronic Environmental Exposure to Perchlorate in Drinking Water and Thyroid Function During Pregnancy and the Neonatal Period. August 8, 2004 Update. Letter to R. Johnston, Committee to Assess the Health Implications of Perchlorate Ingestion, from J.P. Gibbs, Kerr-McGee Corp., Oklahoma City, OK. August 7, 2004. Gibbs, J., and K. Crump. 2003. Analysis of Review of Original Statistical Meth- ods and Alternate Analyses for Ecological Epidemiological Study of Crump et al. (2000) by A.H. Marcus, National Center for Environmental Assessment, in memorandum to A.M. Jarabek, National Center for Environmental Assess- ment, Washington, DC, September 23, 2003. December 11, 2003. International Council for the Control of Iodine Deficiency Disease. 2001. The Western Hemisphere Nears Iodine Sufficiency. IDD Newsletter 17(1):2. [Online]. Available: http://www.people.virginia.edu/%7Ejtd/iccidd/newsletter/ feb2001.htm [accessed July 8, 2004]. International Council for the Control of Iodine Deficiency Disease. 2002. Iodine Deficiency Disease (IDD) Prevalence and Control Program Data: Chile. [Online]. Available: http://www.people.virginia.edu/~jtd/iccidd/mi/idd_034. htm [accessed July 8, 2004]. Kelsh, M.A., P.A. Buffler, J.J. Daaboul, G.W. Rutherford, E.C. Lau, J.C. Cahill, A.K. Exuzides, A.K. Madl, L.G. Palmer, and F.W. Lorey. 2003. Primary congenital hypothyroidism, newborn thyroid function, and environmental perchlorate exposure among residents of a southern California community. J. Occup. Environ. Med. 45(10):1116-1127. Lamm, S. H. 2003. Perchlorate exposure does not explain differences in neonatal thyroid function between Yuma and Flagstaff. [Letter]. J. Occup. Environ. Med. 45(11):1131-1132. Lamm, S.H., and M. Doemland. 1999. Has perchlorate in drinking water increased the rate of congenital hypothyroidism? J. Occup. Environ. Med. 41(5):409- 411. Lamm, S.H., L.E. Braverman, F.X. Li, K. Richman, S. Pino, and G. Howearth. 1999. Thyroid health status of ammonium perchlorate workers: A cross- sectional occupational health study. J. Occup. Environ. Med. 41(4):248-260.
Epidemiological Studies of Occupational and Environmental Exposures 113 Li, F.X., D.M. Byrd , G.M. Deyhle, D.E. Sesser, M.R.Skeels, S.R.Katkowsky, and S.H. Lamm. 2000. Neonatal thyroid-stimulating hormone level and perchlo- rate in drinking water. Teratology 62(6):429-431. Li, Z., F.X. Li, D. Byrd, G.M. Deyhle, D.E. Sesser, M.R. Skeels, and S.H. Lamm. 2000. Neonatal thyroxine level and perchlorate in drinking water. J. Occup. Environ. Med. 42(2):200-205. Li, F.X., L. Squartsoff, and S.H. Lamm. 2001. Prevalence of thyroid diseases in Nevada counties with respect to perchlorate in drinking water. J. Occup. Environ. Med. 43(7):630-634. Marcus, A.H. 2003. Review of Original Statistical Methods and Alternate Analy- ses for Ecological Epidemiology Study of Crump et al. (2000). Memorandum to A.M. Jarabek, National Center for Environmental Assessment, from A.H. Marcus, National Center for Environmental Assessment, Washington, DC. Sept. 23, 2003. [Online]. Available: http://www.epa.gov/ncea/perchlorate/ references/documents/crump_memo.pdf [accessed September 30, 2004]. Mercado, M., V.Y. Yu, I. Francis, W. Szymonowicz, and H. Gold. 1988. Thyroid function in very preterm infants. Early Hum. Dev. 16(2-3):131-141. Morgan, J.W., and R.E. Cassady. 2002. Community cancer assessment in re- sponse to long-time exposure to perchlorate and trichloroethylene in drinking water. J. Occup. Environ. Med. 44(7):616-621. Nahum, G.G., and H. Stanislaw. 2004. Hemoglobin, altitude and birth weight: Does maternal anemia during pregnancy influence fetal growth? J. Reprod. Med. 49(4):297-305. Park, R.M. 2001. Perchlorate Health Effects in the Epidemiological Literature. Letter to A. Jarabek, National Center for Environmental Assessment, Office of Research and Development, Washington, DC, from R.M. Park, National Institute for Occupational Safety and Health, Cincinnati, OH. [Online]. Available: http://www.epa.gov/ncea/perchlorate/references2/documents/100842.pdf [accessed September 30, 2004]. Peterson, S., A. Sanga, H. Eklof, B. Bunga, A. Taube, M. Gebre-Medhin, and H. Rosling. 2000. Classification of thyroid size by palpation and ultrasonography in field surveys. Lancet 355(9198):106-110. Rockette, H.E., and V.C. Arena. 1983. Mortality Pattern of Workers in the Niag- ara Plant. Department of Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA. June 1983. Rothman, K.J., and S. Greenland. 1998. Precision and validity in epidemiologic studies. Pp. 115-134 in Modern Epidemiology, 2nd Ed., K.J. Rothman and S. Greenland, eds. Philadelphia: Lippincott-Raven Publishers. Rovet, J.F. 1999. Long-term neuropsychological sequelae of early-treated congeni- tal hypothyroidism: Effects in adolescence. Acta Paediatr. 88(432):88-95. Rovet, J.F. 2002. Congenital hypothyroidism: An analysis of persisting deficits and associated factors. Neuropsychol. Dev. Cogn. Sect. C Child Neuropsy- chol. 8(3):150-162.
114 Health Implications of Perchlorate Ingestion Rovet, J.F. 2003. Long-term follow-up of children born with sporadic congenital hypothyroidism. Ann. Endocrinol. 64(1):58-61. Schwartz, J. 2001. Gestational Exposure to Perchlorate is Associated With Mea- sures of Decreased Thyroid Function in a Population of California Neonates. M.S. Thesis, University of California, Berkeley. Soldin, O.P., S.J. Soldin, and J.C. Pezzullo. 2003. Urinary iodine percentile ranges in the United States. Clin. Chim. Acta 328(1-2):185-190. U.S. Geological Survey. 2004. The National Map Viewer. [Online]. Available at: http://nmviewogc.cr.usgs.gov/viewer.htm [accessed July 8, 2004]. WHO (World Health Organization). 1994. Iodine and Health: Eliminating Iodine Deficiency Disorders Safely Through Salt Iodization: A statement by the World Health Organization. WHO/NUT/94.4. Geneva: World Health Organi- zation. Zurakowski, D., J. DiCanzio, and J.A. Majzoub. 1999. Pediatric reference inter- vals for serum thyroxine, triiodothyronine, thyrotropin, and free thyroxine. Clin. Chem. 45(7):1087-1091.