Autism

Autism is a complex and severe developmental disorder characterized by impairments of social interaction, impairments in verbal and nonverbal communication, and restricted or repetitive and stereotyped patterns of behaviors and interests (APA, 1994; Filipek et al., 1999). Over time, research has identified subtle differences in the onset and progression of autistic symptoms. The term “autistic spectrum disorders” (ASD), synonymous with “pervasive developmental disorders” (PDD), refers to a continuum of related cognitive and neurobehavioral disorders that reflects the heterogeneity of these symptoms. ASD encompasses the more narrowly defined “autistic disorder” as well as childhood disintegrative disorder, Asperger's syndrome, Rett's syndrome, and pervasive developmental disorder not otherwise specified (PDD-NOS, or atypical autism). Box 1 lists diagnostic criteria for autistic disorder.

BOX 1

DSM-IV Criteria for Autistic Disorder (299.0). A. A total of at least six items from (1), (2) and (3), with at least two from (1), and one each from (2) and (3): 1. Qualitative impairment in social interaction, as manifested by at least two of the following: (more...)

Research has established a strong genetic component in the etiology of autism, but other factors, including infectious, neurological, metabolic, immunological, and environmental insults, may play important roles. Although the consensus of most scientific experts is that most cases of autism are caused by early prenatal exposures such as valproic acid (Moore et al., 2000) or thalidomide (Stromland et al., 1994), or are linked to early developmental genes (Ingram et al., 2000; Persico et al., 2001; Wassink et al., 2001), significant gaps still remain in our understanding of the risk factors and etiological mechanisms of ASD.

There is considerable uncertainty about the prevalence of autistic disorder and other ASDs. Two large reviews of epidemiological studies conducted outside the United States conclude that the best conservative estimate of prevalence of autistic disorder is approximately 10 per 10,000 (Fombonne, 2001b; Gillberg and Wing, 1999). This estimate does not include other categories of ASD. A recent study conducted in Britain suggests that the prevalence of autistic disorder may be higher than previously thought; this study estimated the prevalence of autistic disorder to be 16.8 per 10,000, and the prevalence of ASD to be 62.6 per 10,000 (Chakrabarti and Fombonne, 2001). Information about the prevalence of autism in the United States is limited, reflecting a lack of epidemiological research on autism in this country.

Attention Deficit/Hyperactivity Disorder (ADHD)

Attention deficit/hyperactivity disorder (ADHD) is a behavioral disorder characterized by persistent patterns of inattention and/or hyperactivity. (In this report, the terms ADHD and Attention Deficit Disorder (ADD) are used interchangeably. ADHD is the term used in the Diagnostic and Statistical Manual (DSM-IV), while ADD refers to a specific ICD-9 code (314.0) that can include hyperactivity symptoms. Such patterns must be present before a child is seven years old, and must cause significant interference and impairment in at least two areas of life—such as home and school, or home and work (APA, 1994). Standard diagnostic criteria for ADHD are listed in Box 2. There is currently no clinical test for diagnosing ADHD. Rather, diagnosis depends on careful observation and evaluation of the child in multiple settings. Psychometric tests, such as the Continuous Performance Test (CPT), a standard measure of attention, are often used, however, there is no relation between CPT scores alone and a diagnosis of ADHD.

BOX 2

DSM-IV Criteria for Attention-Deficit/Hyperactivity Disorder (ADHD) (314.00, 314.01). A. Either (1) or (2) 1. Six (or more) of the following symptoms of inattention have persisted for at least 6 months to a degree that is maladaptive and inconsistent (more...)

The most commonly cited prevalence rate suggests that 3–5% of school-age children are affected by ADHD, making it the most commonly diagnosed behavioral childhood disorder (NIH, 1998). Boys are diagnosed with ADHD more often than girls, although the exact ratio is unclear. Three- to nine-fold differences between the sexes are reported (Eme and Kavanaugh, 1995), and sex differences are also seen in the prevalence of subtypes of ADHD (Wolraich et al., 1996b). However, prevalence of the disorder is still debated. Wolraich and colleagues (1996a) reported an increase in prevalence to 12.4% as a result of a change in diagnostic criteria in 1995. A recent review also found that prevalence varied, depending on which DSM criteria were used in evaluation (Brown et al., 2001).

Speech or Language Delay

Medical diagnostic codes (ICD-9) for speech or language delay correspond with three communication disorders described in DSM-IV, namely expressive language disorder, mixed receptive-expressive language disorder, and phonological disorder (see Box 3). Expressive language disorder is characterized by a significant impairment in vocabulary, difficulty forming sentences of developmentally appropriate length, and misuse of verb tense (APA, 1994). Mixed receptive-expressive language disorder has similar features, with additional difficulties comprehending words and sentences. Phonological disorder is characterized by difficulties in forming age-appropriate speech sounds.

BOX 3

DSM-IV Criteria for Speech or Language Delay (315.31 and 315.39). 315.31 Expressive Language Disorder The scores obtained from standardized individually administered measures of expressive language development are substantially below those obtained from (more...)

Close clinical observation and standardized tests of language and intelligence are used to diagnose speech or language delay. For expressive language disorders, a child's scores are lower on tests of expressive language than they are for receptive language and intelligence. For mixed receptive-expressive language disorder, a child will score lower on tests of receptive and expressive language than on tests of nonverbal intellectual capacity (APA, 1994). These communication disorders can also be diagnosed when a child's speech development and language skills are significantly behind those of other children of the same age (Leung and Kao, 1999).

Estimates of the prevalence of speech or language delay vary, largely because of differences in terminology and diagnostic criteria used by practitioners, the subjective nature of parent and practitioner observations (Leung and Kao, 1999), and the age at which the child was evaluated. For example, 9–17% of two-year olds were reported to have a language delay, but prevalence was 1–3% by age 5 (Whitehurst and Fischel, 1994). A recent review (Leung and Kao, 1999) reported speech delay for 3–10% of all children, which is consistent with additional rates reported in the literature (Boyle et al., 1994; Shriberg et al., 1999; Silva et al., 1983). The researchers also noted that speech delay affects boys three to four times more often than girls (Leung and Kao, 1999). Other researchers reported a median prevalence of speech or language delay of 6% for children aged 0–7 years, and that this prevalence had remained constant over the past 30 years (Law et al., 2000).

Mercury Toxicokinetics

Mercury occurs in inorganic and organic forms. Inorganic mercury includes elemental, or metallic mercury (Hg⁰), and the mercurous (Hg⁺) and mercuric (Hg²⁺) salts. Organic mercury, defined as mercury compounds containing carbon bonds, includes ethylmercury (CH₃CH₂Hg⁺) and methylmercury (CH₃Hg⁺), as well as other compounds. Mercury exposure can occur through inhalation of metallic mercury vapor, ingestion (with almost complete absorption of methylmercury), and absorption through the skin, as well as through intravenous, intramuscular, and subcutaneous injections (ATSDR, 1999; NRC, 2000). The toxicokinetic profile of mercury exposure can differ depending on the form of mercury, the route of exposure, the size of the dose, and age at exposure, but the toxicity of all mercury compounds is due to the mercury itself.

Thimerosal is a thiosalicylate salt of ethylmercury. Upon administration, ethylmercury quickly dissociates from thiosalicylic acid and binds to blood or other tissue. The toxicological profile of ethylmercury from thimerosal is thought to be similar to that of ethylmercury from other sources (Magos, 2001b; Suzuki et al., 1963, 1973).

Mercury is clearly neurotoxic and nephrotoxic. Organic mercury forms of interest, ethyl- and methylmercury, are metabolized to mercuric mercury. The effects of mercuric mercury are greatest in the kidneys, whereas the effects of organic mercury are greatest in the central and peripheral nervous systems (Aschner and Aschner, 1990). Generally, mercury is thought to induce cytotoxicity through inhibition of protein synthesis and cellular enzyme-mediated reactions, resulting in structural changes of cells, interference with cellular metabolism, and inhibition of cell migration (Clarkson, 1972). More specifically, mercuric mercury exposure can cause acute renal failure and toxicity, characterized by proteinuria, oliguria, and hematuria. Case studies have also reported congested medulla, pale and swollen cortex, extensive necrosis, and degeneration of tubular epithelium (ATSDR, 1999). Similar measures of toxicity such as necrosis of proximal tubules, proteinuria, and general renal nephropathy have been demonstrated in animal studies (ATSDR, 1999).

Comparisons of Ethylmercury and Methylmercury

The limited data on the toxicology and pharmacokinetics of thimerosal and ethylmercury are summarized here, along with information about methylmercury, which has been studied extensively (see ATSDR, 1999; EPA, 1997; NRC, 2000; for detailed reviews). Many features of the toxicity of ethylmercury are thought likely to be qualitatively similar to those of methylmercury (Ball et al., 2001).

Methylmercury has a whole-body half-life in the range of 70 to 80 days. The half-life of ethylmercury is not known, but may be shorter given its more rapid conversion to inorganic mercury. Methylmercury in blood appears to have a half-life of 50 days (WHO, 1990). Methylmercury is excreted primarily in feces and bile, mostly in the form of inorganic mercury (NRC, 2000). Organic and inorganic mercury complexed with glutathione are excreted in bile (Ballatori and Clarkson, 1984). Methylmercury secreted in bile undergoes subsequent re-absorption in the intestinal tract. Urinary excretion plays a minor role in the elimination of methylmercury from the body. Urinary excretion may be higher after ethylmercury due to its more rapid conversion to inorganic mercury. In general, based on studies of methylmercury in rodents (Ballatori and Clarkson, 1984), suckling infants do not excrete methylmercury. At weaning, the process of biliary excretion is suddenly activated.

After ethylmercury administration in mice, more total mercury has been found in the blood and kidney—compared with methylmercury—and less in the brain (Suzuki et al., 1963). Methylmercury can pass through the blood-brain barrier and into nerve cells as part of a methylmercury-cysteine complex (Kerper et al., 1992). It is unclear whether ethylmercury passes readily through the blood-brain barrier at the concentrations that result from exposure through vaccines. This may result from lack of an amino acid transport system like that available to methylmercury. Glutathione, present in high concentrations inside cells, readily binds to methylmercury and may play a protective role. Methylmercury is transported out of cells probably as a complex with glutathione, as will be discussed later. Little is known about the transport mechanisms for ethylmercury out of cells, but it probably follows those of methylmercury. Ethylmercury also is converted more rapidly than methylmercury to mercuric mercury, which does not cross the blood-brain barrier as readily as the organic compounds (Magos, 2001b). It has also been postulated that mercury has a direct toxic effect on the blood-brain barrier at high doses, making it more permeable (Aschner and Aschner, 1990).

Once present in brain tissue, organic mercury is converted into inorganic mercury. Studies have shown that ethylmercury has a significant and fast conversion to inorganic mercury and that the rate of conversion is slower for methylmercury. Suzuki and colleagues (1973) reviewed several studies and compared inorganic mercury levels in the brain after methylmercury or ethylmercury exposure. One study found that up to 75% of the total mercury in the brain was inorganic mercury three days after injection of ethylmercury or thimerosal. Other studies suggest that the percentage of inorganic mercury resulting from ethylmercury exposure is lower (11–46%). In comparison, the total mercury in the brain at one to ten days after injection of methylmercury was 2.8% (reviewed in Suzuki et al., 1973). Furthermore, the amounts of inorganic mercury in the brain increased over time following exposure to ethylmercury (Suzuki et al., 1973). The characteristics of exposure also influence the conversion rate. Studies have found a higher fraction of inorganic mercury after chronic low dosing of methylmercury than after acute exposure (Aschner and Aschner, 1990).

Upon conversion from organic mercury, mercuric mercury does not as readily cross the blood-brain barrier to move into the blood stream for elimination. Thus it is retained in the neural tissue for a longer period of time (ATSDR, 1999). However, the relative contributions of inorganic and organic mercury to neural cytotoxicity are not known at this time. It is known that methylmercury can inhibit neuronal protein synthesis and several enzymatic processes that control cell metabolism and respiration (Chang et al., 1980). Unpublished data from in vitro studies presented to the committee suggest that ethylmercury from thimerosal in vaccines binds to various neuronal cellular proteins (Haley, 2001). However, because many metals bind to cellular proteins, the significance of this finding with respect to biological plausibility is unclear. Following organic mercury injection in the mature brain, distribution of mercury (reviewed in Clarkson, 1972 and Chang et al., 1980) indicates greatest retention in the calcarine cortex, frontal, temporal, and occipital cortex, cerebellum, and the spinal dorsal root ganglions. In the developing brain, methylmercury exposure leads to widespread damage, affecting virtually all areas of the brain (Clarkson, 1997).

Mercuric mercury accumulated in the kidney is bound to the protein metallothionein, which may play a protective role in renal toxicity. In the blood, ethyl- and methylmercury are bound to cysteine residues of hemoglobin and to plasma proteins (Clarkson, 1972; NRC, 2000; Takeda et al., 1968). Methylmercury exposure results in a greater ratio of mercury in the red blood cells to plasma compared with ethylmercury (reviewed in Suzuki et al., 1973). The blood-distribution ratio of organic mercury is species-dependent.

Methylmercury is avidly accumulated from the blood stream into scalp hair (for review, see WHO, 1990). The average hair to blood concentration ratio is 250:1. Once incorporated into the formed elements of the hair strand, the concentration remains unchanged. Thus, longitudinal analysis along the length of the hair strand will recapitulate previous blood levels (Amin-Zaki et al., 1976). Hair grows at an approximate rate of 1 cm per month. Thus depending on the length of the hair strand, recapitulation can take place over many months or even years. Hair is especially useful as a biological monitor for prenatal exposure as a hair sample that covers the entire period of pregnancy can be collected from the mother at or soon after delivery. Thus hair has been used as the primary biological monitor in all epidemiological studies of prenatal exposure or in one case to complement the use of cord blood (NRC, 2000). Especially important, it has recently been shown that levels in the mother's hair at delivery predict levels in autopsy brain samples from newborn infants that died of various causes (Cernichiari et al., 1995).

Organic mercury is also known to cross the placenta easily, allowing for prenatal mercury exposure. In addition, methylmercury is excreted in breast milk (Sundberg and Oskarsson, 1992; Yoshida et al., 1992).

Health Effects of High-Dose Exposures to Thimerosal, Ethylmercury, and Methylmercury

High-doses of thimerosal, ethylmercury, and methylmercury have been found to produce severe health problems. High-dose thimerosal exposures have been iatrogenic or, in one case, resulting from a deliberate use in a suicide attempt. High-dose ethylmercury poisonings have resulted from environmental and occupational exposures, with poisoning from ethylmercury-containing fungicides being frequently fatal. Methylmercury poisoning has resulted from consumption of contaminated fish and grain. Neurological effects were prominent in all of these poisonings, and prenatal exposures to methylmercury resulted in birth defects and serious neurodevelopmental deficits. Although these exposures are at levels far higher than those resulting from vaccination, the evidence of neurotoxicity offers some support for, but does not establish, the biologic plausibility of neurotoxic effects at lower doses.

Thimerosal Exposure. Several cases of high-dose exposure to thimerosal have been reported in the literature. Six patients (age 6 weeks to 39 years), who were given intramuscular injections of improperly formulated chloramphenicol, each received 71–330 mg/kg of thimerosal—1,000 times the correct level (Axton, 1972). Symptoms were mainly neurological, including restlessness, slurred speech, hemiparesis, confusion, unsteady gait, hallucinations, and coma; five of the six patients died. An 18-month-old girl, who ingested 127 mg/kg of thimerosal (in a merthiolate solution) over a one-month period, developed ataxia, hand tremors, vomiting, staring spells, and renal and hepatic failure before dying (Rohyans et al., 1984). (Chelation therapy did not appear to be associated with clinical improvement in neurological function.) Ten of 13 infants treated with topical applications of a 0.1% tincture of thimerosal for omphaloceles (herniations of the bowel through the umbilicus) died (Fagan et al., 1977). Autopsy studies of three infants found high levels of mercury in the liver (11.8–26.6 µg/g), the kidney (2.36–4.6 µg/g), and the brain (0.65–5.1 µg/g). A neurological follow-up at 10 years on a surviving patient found no abnormal focal neurological findings but included third-person reports of restlessness and easy distractibility in school.

Ingestion of 5g of thimerosal in a suicide attempt resulted in delirium, sensorimotor polyneuropathy culminating in mechanical ventilation, and coma, along with acute renal failure (which resolved after 40 days), fever, oral exanthema, and gingivitis (Pfab et al., 1996). The patient recovered completely at 148 days post-ingestion, except for sensory defects in two toes.

Toxicity was also reported in a 44-year-old male who had received approximately 20 mg of thimerosal from HBV immunoglobulin, administered intramuscularly or intravenously, after a liver transplant for end-stage liver failure due to hepatitis B, C, and D infection (Lowell et al., 1996). Beginning on the third postoperative day, the patient began developing symptoms that included paranoid thoughts, slurred speech, slowed movements, decreased muscle strength, inability to ambulate, and tremor in both hands. The patient's symptoms completely resolved in five weeks following chelation therapy with DMSA (dimercaptsuccinic acid). The diagnosis of mercury toxicity in this case has been questioned because of the brief exposure period, the absence of visual field constriction, and a blood-mercury level (104 µg/L on day nine) considered too low for the development of the symptoms described when compared with other cases (Magos, 2001b).

In contrast to these reports of high-dose thimerosal toxicity, early studies in animals and humans failed to show neurotoxic effects (Jamieson and Powell, 1931). Rabbits tolerated intravenous thimerosal doses of 25 mg/kg of body weight; at higher doses, fatalities resulted from kidney and intestinal disease. In rats, the tolerated dose was 45 mg/kg body weight. A report on 22 human subjects who received total doses of up 180 ml of a 1% solution of merthiolate (10 mg/kg) noted only local skin irritation (Powell and Jamieson, 1931). However, the study was not designed to test toxicity (Ball et al., 2001).

Other Ethylmercury Exposures. Mercury poisonings occurred in Iraq from consumption of bread made from grain treated with a fungicide containing 7.7% ethylmercury p-toluene sulphonanilide (Damluji, 1962; Jalili and Abbasi, 1961). Symptoms generally occurred one to two months after consumption of the treated wheat, and patients differed in the mix and severity of their symptoms. Among the neurological symptoms seen were ataxia, difficulty in walking, speech disturbances, constriction of visual fields, and blindness. Effects were also seen on the renal system, the skin, the cardiac system, and the gastrointestinal system.

In China, similar effects were seen from consumption of rice treated with ethylmercury chloride (2–2.5%) (Zhang, 1984). Symptoms began one to two weeks following consumption of the rice and continued for several months. Common neurological symptoms included weakness, dizziness, numbness of extremities, paresthesia, ataxia, and unsteady gait. Fewer people experienced symptoms of impaired vision, coma, speech disturbance, or hand tremor. The mildest cases resulted from doses of 0.5–1 mg/kg body weight, moderate cases from 1–2 mg/kg, severe cases from 2–3 mg/kg, and lethal cases above 4.0 mg/kg body weight.

Four cases of mercury poisoning occurred in Romania from consumption of pork from animals fed seed treated with ethylmercury chloride fungicides (Cinca et al., 1980). Symptoms were similar to those seen in Iraq and China, including ataxia, dysarthria, dysphagia, increased tendon reflexes, inability to stand or walk, coma, and constricted visual fields. Autopsy results from two boys aged 10 and 15 showed the greatest nerve-cell loss and proliferation of neuroglia in the cerebral cortex, especially in the calcarine cortex, the midbrain, and the bulbar reticular formation. Demyelination was apparent in the ninth and tenth cranial-nerve roots, and moderate cell loss and other lesions were found in the granular and Purkinje cells of the cerebellum. Motorneuron loss was evident in the ventral horns of the spinal cord.

Fatal ethylmercury poisoning has also resulted from a seven-week occupational exposure to ethylmercury in an insecticide factory (Hay et al., 1963). The presenting symptoms included dysarthria, ataxia, leg weakness, bilateral nerve deafness, increased reflexes, and poor coordination. At autopsy, the brain showed loss of neurons from areas of the cerebral cortex (especially calcarine cortex) and some degeneration of Purkinje and granule cells of the cerebellum. In contrast to typical findings, a greater amount of mercury was found in the brain, especially in the corpus callosum, than in the liver.

Methylmercury Exposures. Methylmercury at high doses is a well-documented neurotoxicant (NRC, 2000). The most severe neurological effects reported in humans occurred after accidental methylmercury-poisoning episodes in Japan and Iraq. Major epidemics of methylmercury poisoning occurred in Minamata and Niigata, Japan, during the 1950s and 1960s following consumption of contaminated fish and seafood, and first raised awareness of the neurological sequelae of high-dose exposures to methylmercury (Harada, 1995; Tsubaki and Irukayama, 1977). Poisoning episodes in Iraq in the 1960s and 1970s resulted from consumption of homemade bread made from grain treated with a methylmercury-containing fungicide (Bakir et al., 1973).

In adults, these high-dose exposures to methylmercury resulted in symptoms that included paresthesia, ataxia, and impairments of speech, hearing, and vision. In children exposed during fetal development, there were severe neurological dysfunctions and developmental abnormalities, including mental retardation, cerebral palsy, deafness, blindness, and dysarthria (Harada, 1995; Marsh et al., 1987; NRC, 2000). In both Japan and Iraq, the neurological effects observed in children exposed to methylmercury in utero were more serious than those observed in adults, and sometimes occurred at lower doses than in adults, indicating the increased susceptibility of the fetus to methylmercury exposure (NRC, 2000). The exposures that produced these effects were very high, and precise dose-response relationships at low doses were not established (NRC, 2000).

Health Effects of Low-Dose Exposures to Thimerosal and Methylmercury

Studies of low-dose exposures to thimerosal and methylmercury would be potentially more informative regarding the biological plausibility of neurodevelopmental effects from thimerosal exposures from vaccination. The data on thimerosal are limited, however, and the studies of the effect of ingested methylmercury are not directly applicable.

Hypersensitivity Reactions to Thimerosal. Immune-mediated reactions to mercury-containing compounds are well documented in humans and in experimental animals (for reviews see Enestron and Hultman, 1995; Griem and Gleichman, 1995; Pollard and Hultman, 1997). The most common manifestation is contact allergy (i.e., delayed-type hypersensitivity), although immune-complex-mediated disorders, such as glomerulonephritis, and immediate-type hypersensitivity (i.e. allergy) reactions have also been reported. Thimerosal can induce contact hypersensitivity responses in humans. Positive patch tests have been seen in 1–16% of individuals tested (reviewed in Cox and Forsyth, 1988). However, the frequency of clinically important contact hypersensitivity is much lower, and its significance is a matter of controversy (Ball et al., 2001).

Ethylmercury in vaccines and other pharmaceutical products may rarely cause immediate-type hypersensitivity reactions, including anaphylaxis, although conclusive proof that mercury in thimerosal is the causative component is lacking. In one case in which anaphylaxis occurred in association with hepatitis B vaccine administration, a repeat challenge suggested that thimerosal was not responsible (Ball et al., 2001). Similarly, one report of laryngeal obstruction after a patient used a throat spray containing thimerosal (Maibach, 1975) might reflect the corrosive effect of mercury rather than a hypersensitivity reaction.

Acrodynia has been reported in a patient receiving gammaglobulin injections containing .01% thimerosal (Matheson et al., 1980). Acrodynia, or pink disease, occurs in a small fraction of exposed individuals, primarily children, and may represent a hypersensitivity response to mercury. It presents with a pink, pruritic skin rash, especially on the soles and palms. Other symptoms can include photophobia, irritability, and lethargy. Acrodynia is most closely associated with infants exposed to inorganic mercury in products like teething powders, which are no longer marketed. The 20-year-old patient receiving gammaglobulin had congenital agammaglobulinemia and a history of allergic reactions and sensitivity to sulphonamide drugs. He had been receiving gammaglobulin shots over the course of 15 years. It was estimated that the patient was exposed to 40–50 mg of mercury during the year in which he presented with symptoms.

Methylmercury Exposure. The lack of data on the effects of low-dose ethylmercury exposure has led those concerned about thimerosal-containing vaccines to examine studies of chronic low-dose exposures to methylmercury, occurring primarily through consumption of fish and marine mammals. The neurodevelopmental effects in children resulting from moderate to low-level prenatal and postnatal exposures to methylmercury are less clear than the serious effects seen in high-dose poisonings (see NRC, 2000 for a review of epidemiological studies).

Two large prospective cohort studies are currently under way to evaluate the more subtle endpoints of neurotoxicity resulting from low-dose prenatal exposure to methylmercury. The first study is being conducted in the Faroe Islands, in the North Sea, and the second in the Republic of the Seychelles, a nation of islands in the Indian Ocean off the coast of East Africa. In the Faroe Islands, the predominant source of mercury exposure is from consumption of pilot whale meat. In the Seychelles, mercury exposure comes from consumption of ocean fish. To date, the two studies have reached different conclusions.

The Faroe Islands study is following a cohort of approximately 1,000 children born in 1986–1987. Developmental outcomes were assessed for 917 of the children at age 7 using a battery of domain-specific neuropsychological and neurophysiological tests. Prenatal methylmercury exposure, as measured by umbilical cord blood levels, was associated with subtle neuropsychological deficits most notable in the areas of attention, memory, and language, and to a lesser extent in visuospatial and motor functions. Postnatal exposures, measured by the child's hair mercury concentration at ages 1 year and 7 years, were less useful risk predictors (Grandjean et al., 1997).

The Seychelles study is following two cohorts of children, a Pilot Study cohort (N=789) and a Main Study cohort (N=779), enrolled in 1987 and 1989, respectively. The cohorts have been followed longitudinally and outcomes have been assessed at multiple ages. Prenatal and postnatal mercury exposures were measured in maternal hair and child hair cut at each testing, respectively. Evaluations conducted through 5.5 years of age, using global assessments, found no adverse associations between prenatal or postnatal exposure to methylmercury and childhood developmental outcomes (Davidson et al., 1995, 1998, 2000). Preliminary analyses from evaluations of pilot cohort members at age nine using domain-specific tests also show no adverse associations (Davidson et al., 2000; Myers, 2001).

The discrepant findings in these two studies may reflect differences in study design or in the characteristics of the populations examined. There were differences in exposure measures (cord blood versus maternal hair), types of neurological and psychological tests administered (domain-specific versus global developmental outcomes), the age of testing (7 years versus 5.5 years of age), possible confounders (PCB exposure or genetic differences in the populations studied), and biostatistical issues related to data analysis (NRC, 2000). Additional analyses of the Seychelle Islands data are under way and will address some of these differences (Myers, 2001).

The committee carefully considered the significance of the evidence of subtle neurological deficits from the Faroe Islands study for its assessment of the biological plausibility of a causal association between exposure to thimerosal-containing vaccines and the neurodevelopmental disorders of autism, ADHD, and speech or language delay. The conclusion was that those findings add indirect support to the biological plausibility of such an association but do not demonstrate a direct biological mechanism or provide evidence of causality. As was stated in the recent FDA risk assessment of thimerosal in vaccines (Ball et al., 2001), extrapolating the toxicity of methylmercury exposure to ethylmercury exposure in thimerosal-containing vaccines is problematic. Data on the comparative toxicology of ethyl- and methylmercury are limited. In addition, data on the metabolism and excretion of ethylmercury compared with methylmercury have not been well established. The comparability of the effects of chronic low-dose exposure to methylmercury via ingestion versus those of intermittent exposure to ethylmercury via intramuscular injection is unknown. The relative susceptibility of the fetus compared with the infant is also unknown. Furthermore, the neuropsychological deficits detected in the Faroe Island studies are not reliable predictors for the specific neurodevelopmental diagnoses of autism, ADHD, and speech or language delay.

Investigations Related to Mercury and Heavy Metals in Children with Autism

Similarities between autism and mercury intoxication have been offered as evidence that thimerosal in childhood vaccines could cause autism (Bernard et al., 2001; El-Dahr, 2001). The identified similarities include psychiatric disturbances, speech or language deficits, sensory abnormalities, motor disorders, cognitive impairments, unusual behaviors (such as head banging and sleep difficulties), physical disturbances (such as rashes, diarrhea, abnormal feeding behaviors), and biological indices (such as metal metabolism, immune system effects, CNS structure, neurochemistry, and neurophysiology). However, the analogy is weakened by differences in mechanistic understanding of some of the symptoms of mercury intoxication and of autism. For example, impaired visual fixation associated with mercury toxicity is due to problems with motor control of eye muscles. This is unrelated mechanistically to problems in joint attention associated with autism, which are most likely a problem of social reciprocity, not motor control (Tanguay, 2000). In addition, there are manifestations from ethylmercury toxicity not seen in autism, for example, polyuria, ataxia, and tremor. Analogies like these are important for hypothesis generation, but they have limited value in causality assessments.

Others have argued that thimerosal may cause autism based on the observation that some autistic children have abnormal blood-metal profiles. One method being used to indicate levels of mercury in autistic children is chelation therapy. Three unpublished reports of chelation-based estimates of mercury in autistic children were presented to the committee (Bradstreet, 2001). In one study, higher urinary mercury levels were found in autistic children than in neurologically normal children after provocation with the chelator DMSA. In a case report that also used DMSA provocation, mercury levels in an autistic boy were much more elevated than those of his mother, father, or brother. Finally, in a study of 27 autistic children in Indonesia, 70% had detectable mercury levels, with the levels in 30% of the children exceeding adult mercury levels. Some have reported improvements in functioning of autistic children after chelation therapy (Cave, 2000; Sykes, 2001).

Chelation therapy has not been established to improve renal or nervous system symptoms of chronic mercury toxicity (Sandborgh Englund et al., 1994) and has had no effect on cognitive function when used for excretion of another heavy metal, lead (Rogan et al., 2001). Because it is unlikely to remove mercury from the brain, it is useful only immediately after exposure, before damage has occurred (Evans, 1998). Moreover, chelation therapy is not without risks; for example, some chelation therapies might cause the release of mercury from soft-tissue stores, thus leading to increased exposure of the nervous system to mercury (Wentz, 2000). In addition, chelation therapies are not specific to one metal.

The presence of abnormal metal profiles in autistic children does not mean that the metal burden is the cause of autism. It could indicate a comorbidity between autism and an inability to handle heavy metals. Further, a favorable response to chelation therapy is not proof that the mercury levels caused the neurological dysfunction. Chelation therapy is non-specific, and the observed effects could be caused by other metals or other factors.

Application of Methylmercury Exposure Guidelines to Thimerosal Exposure from Vaccines

Concern over the use of thimerosal in childhood vaccines was originally triggered by calculations showing that vaccines on the recommended childhood immunization schedule might result in cumulative ethylmercury exposures that exceed estimated limits of safe exposure based on some federal guidelines for methylmercury intake. Three U.S. federal agencies—EPA, ATSDR, and FDA—have developed guidelines for assessing exposure to methylmercury. The EPA is responsible for monitoring mercury concentrations in the environment and regulating industrial releases of mercury to air and surface water. The FDA is responsible for regulating commercially sold fish and seafood based on mercury concentrations. Although ATSDR does not have regulatory authority, it monitors the potential for human exposure to methylmercury and investigates reported health effects (NRC, 2000). Currently, each of the agencies has a different guideline for assessing safe exposure to mercury, and different limits of methylmercury intake, ranging from 0.1 µg/kg/day for EPA (EPA, 2001), to 0.3 µg/kg/day for ATSDR (ATSDR, 1999), to approximately 0.4 µg/kg/day for FDA (FDA, 1979).³ The differences in the agencies' recommended limits can be attributed to the use of different risk-assessment methods and uncertainty factors, variation in primary data sources, and the different mandates of each agency (NRC, 2000). Although a consensus for safe mercury exposure does not exist, all of the recommended limits are within the same order of magnitude.

The methylmercury exposure limits calculated by these agencies are not limits above which injury is sure to occur. Rather, they should be interpreted as general levels of exposure below which there is confidence that adverse effects will be absent, although the margin of safety reflected in those limits is uncertain (EPA, 1997). Definitions of the exposure measures used in the three federal guidelines are provided in Box 4.

BOX 4

Definition of Measures Used in Federal Exposure Guidelines. Environmental Protection Agency Reference Dose (RfD): An estimate (with uncertainty spanning perhaps an order of magnitude) of daily exposure to the human population (including sensitive subgroups) (more...)

It is important to understand the specific data and assumptions on which these guidelines are based, however, when considering the possible implications of thimerosal exposures that exceed the guidelines on methylmercury intakes. As a result, the computation of the federal guidelines is briefly reviewed here, although an in-depth discussion of the development of the guidelines is beyond the scope of this report. (The interested reader is referred to a recent review in NRC, 2000.) The estimates of the maximum mercury exposures associated with recommended childhood immunization are then compared against the recommended exposure limits based on the federal guidelines. Also discussed are sources of additional uncertainty that arise from the application of these methylmercury-based guidelines to ethylmercury exposures resulting from the use of thimerosal-containing vaccines.

Computation of Federal Guidelines for Methylmercury Exposure. Although the details differ, the basic approaches used to compute EPA's Reference Dose (RfD), ATSDR's Minimal Risk Level (MRL), and FDA's Acceptable Daily Intake (ADI) are similar. Each calculation begins with a “point of departure” exposure that can be loosely interpreted as the lowest dose that has been empirically associated with a specific type of illness or injury in typical members of the population. The point of departure dose may then be divided by one or more factors that account for uncertainty or variability in the risk estimate. These factors, which are often referred to as uncertainty, adjustment, safety, or modifying factors, typically range in value from one to ten. These factors serve the same two general purposes. First, they are used to protect against the possibility that the true minimum exposure at which injury or disease occurs in typical members of the population is lower than the identified level (i.e., the point of departure dose). Second, they are used to protect the health of “sensitive” members of the population, who could for either pharmacokinetic reasons (more of the active agent is delivered to the target tissue per unit exposure) or pharmacodynamic reasons (the tissue reacts to a greater extent per unit of the active agent delivered), be adversely affected at a lower level than the level at which typical members of the population are adversely affected. Dividing by factors that exceed one makes the resulting standard more stringent or protective. Thus, for example, if it is suspected that another uninvestigated health effect might occur at exposure levels ten times lower than the effect used to compute the point of departure, the point of departure can be divided by a factor of ten. Similarly, if it is suspected that sensitive members of the population are adversely affected at an exposure level one-third as great as the level of exposure at which typical members of the population are adversely affected, the point of departure can be divided by a factor of three.

To set its RfD of 0.1 µg/kg/day for methylmercury intake, EPA originally used findings from a study of 81 children in Iraq exposed in utero to high levels of methylmercury from maternal consumption of seed grain (Marsh et al., 1987). Developmental effects, quantified in terms of a composite score reflecting “late walking, late talking, mental symptoms, seizures” (NRC, 2000) or adversely affected neurological function, were observed at lower levels of exposure than any other health effects, making them the most sensitive indicator of health risk. EPA estimated that with 95% confidence, the maternal mercury exposure that increased developmental risk for an unborn child by 10% corresponded to hair mercury concentrations of no less than 11 ppm (parts per million), estimated as equivalent to daily consumption of 1.1 µg of mercury per kg of bodyweight (1.1 µg/kg/day). The EPA applied uncertainty factors—with a total value of 10—to produce the exposure limit value of 0.1 µg/kg/day (EPA, 2001).

A National Research Council (2000) review of EPA's methylmercury RfD recommended maintaining the exposure limit of 0.1 µg/kg/day but suggested basing the value on data from the Faroe Islands study of the effects of chronic dietary exposure to mercury rather than on the acute high-exposure data from Iraq. As described above, the Faroe Islands study used a variety of performance tests to measure neurological and cognitive function among children with prenatal exposures to methylmercury.

Several sources of uncertainty and variability were included in the EPA's composite uncertainty factor of 10. First, a factor of 3 accounted for pharmacokinetic differences among members of the population. Another factor, also 3, addressed pharmacodynamic variability and uncertainty. EPA rounded the product of the uncertainty factors—in this case, 9—to one significant digit, thus yielding 10 (EPA, 2001).

In contrast, the ATSDR exposure limit of 0.3 µg/kg/day, established in 1999 (ATSDR, 1999), is based on data from the Seychelle Islands study (described above) (Davidson et al., 1998), which did not find an association between prenatal methylmercury exposure and deficits in neurological function. ATSDR used these data to identify the maximum level of mercury exposure not associated with adverse health effects. The highest exposure group in that study had hair mercury concentrations of 15.3 ppm, which ATSDR estimated to be associated with consumption of 1.3 µg/kg/day of methylmercury. To obtain the MRL, the ATSDR applied a composite factor of 4.5, which reflects (1) an uncertainty factor of 1.5 for pharmacokinetic differences in the population, (2) plus an uncertainty factor of 1.5 for inter-individual variability (pharmacodynamic) differences, (3) multiplied by an additional modifying factor of 1.5 to address the possibility that domain-specific tests, as employed in the Faroe Islands study, might be able to detect subtle neurological deficits not tested for in the Seychelles cohort (ATSDR, 1999).

The FDA set its action level of 1 ppm of methylmercury in the edible portion of fish in 1979, based on an ADI of 30 µg/day (equivalent to approximately 0.4 µg/kg/day assuming a 70 kg average body weight). The ADI estimates were based primarily on Swedish studies of methylmercury poisonings from consumption of highly contaminated fish in Niigata, Japan (FDA, 1979). Unlike the EPA and ATSDR exposure limits, which are based on effects of prenatal exposure to methylmercury, the FDA limit is based on exposure effects in adults. In deriving its ADI, FDA estimated the lowest daily intake at which adverse effects could appear in adults (300 µg/day). The FDA then applied a ten-fold margin of safety that it uses when human data are available, to obtain the 30 µg/day ADI (FDA, 1979).

In summary, exceeding federal exposure guidelines for mercury does not mean that adverse health effects should necessarily be expected to result. All the guidelines discussed here have additional factors built in to protect members of the population who might be more sensitive for either pharmacokinetic or pharmacodynamic reasons and address the possibility that (perhaps different) effects will occur at lower levels than those that have been empirically observed.

Comparing Childhood Vaccine Exposures with Federal Guidelines. As part of the FDA risk assessment of thimerosal, Ball and colleagues (2001) sought to compare the maximum mercury exposures resulting from the recommended childhood immunization schedule with the estimated cumulative limits for mercury exposure based on the three federal guidelines, as well as those of the World Health Organization (WHO).⁴ Table 3 shows the calculated exposure limits based on application of these guidelines to a female infant at the 5^th, 50^th, and 95^th percentile body weight between birth and 26 weeks, the time when most vaccines are given. The maximum vaccine-related cumulative mercury dose at age 6 months (based on the mercury content of the recommended vaccines in 1999; see Table 2) was 187.5 µg (and 200 µg for children who also received the influenza vaccine), which exceeded the EPA limits calculated for each bodyweight category and the ATSDR limits for the lowest-weight infants who also received the influenza vaccine (see shaded portions of the table).

There is, however, no scientific or clinical basis for knowing that this is the appropriate way to compare vaccine-related mercury exposure with the federal guidelines. Ball and colleagues essentially averaged exposures over the first six months of life, but because mercury exposures associated with vaccines are episodic, one can argue for an even shorter averaging period. Changing the averaging period can have a substantial impact on comparisons with federal guidelines. Halsey (1999a) made alternative comparisons in which exposures were averaged over periods of one month, one week, and one day. For example, using an averaging period of one month, a single dose of hepatitis B vaccine with 12.5 µg mercury given to a child with an average weight of 4 kg yields an average daily dose of 0.1 µg/kg/day, which is equal to EPA's RfD. Alternatively, if the exposure is averaged over a single day, it amounts to approximately 3.1 µg/kg/day, which exceeds EPA's standard by more than a factor of 30. Figure 2 shows Halsey's estimates of maximum single-day mercury exposure from recommended vaccines given at birth and ages 2, 4, and 6 months, for a range of body weights at each age. He assumed mercury doses of 12.5 µg at birth, 62.5 µg at age 2 months, 50 µg at age 4 months, and 62.5 µg at age 6 months.

FIGURE 2

Mercury (µg/kg) administered by age and weight if thimerosal-containing vaccines are given for Hepatitis B, Hib, and DTaP. Amount of Hg received (in micrograms)=12.5 at birth, 62.5 at 2 and 6 months, 50 at 4 months.

An alternative approach produces a substantially lower estimate of exposure. Because the federal guidelines are defined in terms of exposures over a “long period of time” or even over a lifetime, the daily doses assumed by Halsey could conceivably be averaged over a typical lifetime duration of approximately 25,000 days. Doing so yields doses of at least a factor of 400 below the EPA's RfD.

In order to resolve the apparently arbitrary selection of an averaging duration, the committee compared the maximum blood mercury concentration associated with vaccine exposures to the blood mercury concentrations that would be associated with long-term exposures equal to EPA's RfD, the most stringent of the mercury intake standards reviewed. NRC (2000) estimated the cord blood concentration associated with its point of departure dose to be 58 µg/L. Based on that estimate, it follows that EPA's RfD corresponds to a blood mercury concentration of 5 µg/L.⁵ This assumes that the blood mercury concentration is an appropriate indicator of the biologically relevant dose of mercury.⁶

To estimate blood mercury levels resulting from environmental exposures, the committee used data from the National Health and Nutrition Examination Survey (NHANES) (CDC, 2001d). The NHANES measurements in children one to five years old (N=248) show median blood mercury concentrations of 0.2 µg/L (95% CI 0.2–0.3 µg/L) and 90^th percentile values of 1.4 µg/L (95% CI 0.7–4.8 µg/L).

Estimates of blood mercury levels resulting from children's vaccine exposures were taken from unpublished theoretical estimates and both published and unpublished observational data reported to the committee at its July 2001 meeting.

Brown (2001) presented unpublished results to the commmitte based on a biokinetic model for methylmercury exposures from fish (Ginsberg and Toal, 2000) to produce a range of estimates based on assumptions about body weight and patterns of mercury elimination. Assuming an average body weight during the first 18 months of life and adult patterns of mercury excretion, vaccine doses produced peak blood mercury levels of 14 µg/L; peak levels corresponded with recommended vaccine doses at ages two, four, and six months. With assumptions of low body weight and no elimination of mercury, blood mercury levels rose over time with each recommended vaccine dose and reached an outer bound estimate of 40 µg/L at age six months. This model is useful only as a starting point for estimating maximum possible blood mercury levels following vaccine exposures. Furthermore, the model is based on methylmercury and must be more fully developed to be useful in risk assessments of thimerosal-containing vaccines.

In a recent published observational study (Stajich et al., 2000), total blood mercury levels were measured before and after the administration of the birth dose of hepatitis B (containing 12.5 µg of ethylmercury) in a group of 15 preterm infants and a control group of 5 term infants. Measurements of mercury levels were taken within 48 to 72 hours after vaccination. Comparisons of pre-and post-vaccination mercury levels demonstrated elevated blood mercury levels after a single dose of hepatitis B vaccine in both preterm and term infants. Preterm infants (mean=.54 µg/L; +/–.79 SD) had a tenfold higher mean blood mercury level at baseline compared with term infants (mean=.04 µg/L; +/–.09 SD), although the difference was not significant. Post-vaccination mean blood mercury levels in preterm infants (mean=7.36 µg/L•;+/– 4.99 SD) were 3 times higher than those in term infants (mean=2.24 µg/L;+/–.58 SD), and differences were significant.

Unpublished observational data funded by NIAID (Pichichero et al., 2001; Sager, 2001) show lower blood mercury levels than Brown's model-based estimates or the observational study by Stajich et al. (2000). Among two-month-olds (N=16) receiving an average of 45.6 µg of mercury (range 37.5–62.5 µg), blood mercury concentrations at 3 to 20 days after vaccination (average 11.25 days) averaged 1.5 µg/L (range < 0.75–4.11 µg/L). Among six-month-olds (N= 20) receiving an average cumulative dose of 111.3 µg mercury (range 87.5–175 µg), blood mercury concentrations 4 to 27 days post vaccination (average 13.3 days) averaged 0.98 µg/L (range < 0.05–1.5µg/L).

Two differences among the three studies should be noted. The delay between vaccination and blood mercury sampling in the NIAID data means that these measurements are not directly equivalent to the peak levels estimated by Brown or the measures by Stajich et al. (2000) taken within 48 to 72 hours. The delay was not substantial, given mercury's relatively long half-life, but the half-life of ethylmercury is probably shorter than the half-life of methylmercury (Brown, 2001), implying that earlier observation would have shown higher blood mercury levels. The findings of the NIAID study are consistent with excretion of ethylmercury. The children in the NIAID study also received lower cumulative mercury doses from their vaccines than the theoretical maximums modeled by Brown. Once again, however, this difference in dose (roughly a factor of 2) does not explain the differences between the modeled and measured blood mercury levels reported by Brown and NIAID, respectively. However, Brown's model is preliminary and is based on assumptions about methylmercury, which may explain why the results differ substantially from the empirical findings by Stajich et al. (2000) and NIAID. Furthermore, given the small size of the study populations in the two observational studies, the significance of the findings cannot be assessed.

Sources of Uncertainty. As the previous discussion shows, comparing mercury exposures from vaccines with federal mercury-exposure guidelines does not provide proof of either excess risk or adequate safety. In addition to considerations like the choice of an appropriate averaging duration, there are several other key factors that make comparisons between vaccine doses of ethylmercury and federal exposure guidelines uncertain. Many of these points are also raised elsewhere in this report.

The EPA and ATSDR exposure guidelines were set based on adverse effects of prenatal exposures to methylmercury resulting from maternal consumption of contaminated food. This means that vaccine exposures differ in terms of the age of exposure, the route of exposure, and the type of mercury. The developing fetal brain is known to be highly susceptible to toxic exposures, but data on the effects of early postnatal exposures are limited, for both methylmercury and ethylmercury. There is also little basis for determining whether the effects of exposures through injection are comparable to those from ingestion.

Furthermore, because the pharmacokinetics of ethylmercury are not well documented, it is difficult to know whether the biomarkers used in establishing the methylmercury guidelines can be applied to ethylmercury exposures. Magos (2001b) points out that blood mercury levels following exposure to ethylmercury appear to correspond to lower concentrations of mercury in the brain, and therefore perhaps less risk of neurotoxicity, than are found with similar blood levels resulting from methylmercury exposure. A key uncertainty is the rate of excretion of ethylmercury in infants. Although data from animal tests and data on methylmercury excretion in infants suggest little excretion in infants, the preliminary data from the NIAID study suggest significant mercury elimination in children given thimerosal-containing vaccines.

However, abnormalities or genetic variations in mercury metabolism might lead to an underestimate of mercury levels from thimerosal-containing vaccines. The preliminary report that autistic children have a higher risk than non-autistic children for metallothionein metabolic disturbances (Walsh and Usman, 2001), together with case reports from some pediatric practices that some autistic children have high mercury and other heavy-metal profiles, might contribute to the concern of some that a subgroup of children with NDDs might be at risk for mercury toxicity at levels of exposure that are safe for other children. This does not, however, imply that mercury exposures have caused the NDDs.

Due to differences in drug distribution or clearance, concentrations of mercury reached in specific tissues may vary between individuals receiving the same amount of mercury (or whose intake of mercury is similar). Other individuals may have genetic differences that make them more susceptible to mercury-induced injury at a given concentration of mercury in a particular tissue. It might be prudent to reduce mercury exposure as much as possible in these children. The role of the mercury metabolism (or other metabolic defects) in the genesis of concomitant NDDs is unclear, however, and cause and effect should not be inferred.

Case Reports and Case Series

VAERS. A search of the Vaccine Adverse Event Reporting System (VAERS), for the period from its inception in November 1990 through May 2001, identified 176 unique reports based on the following search strategy. During this 11-year period, about 120,000 reports were submitted to VAERS, including approximately 5,000 foreign reports (CDC, 2001b). Reports of interest were identified by searching all text fields using the terms “thimerosal,” “thiomersal,” “mercury,” or “merthiolate,” or a portion thereof (e.g., for “mercury,” the text string “merc” was used).

The reports referred to the following thimerosal-containing vaccines: hepatitis B (79), influenza (48), diphtheria-tetanus (DT) (3), diphtheria-tetanus-pertussis (DTP) (4), Haemophilus influenzae type b (Hib) (1), tetanus-diphtheria (Td)(9), DTP-Hib (5), pneumococcal (3), tetanus (4), rabies (1) and concurrent but separate administration of influenza and pneumococcal (3), influenza and TD (1), DTwP and hepatitis B (1), DTP and Hib (5), and DTaP, Hib, and hepatitis B (1). Eight reports listed an adverse event in response to the cumulative childhood vaccine schedule. Reports involving only non-thimerosal-containing vaccines (e.g., MMR or OPV) were excluded from the reporting.

Thirty-three reports were for children ages 15 years and younger and were distributed as follows: 6 months and younger (6 reports), 7 months to 2 years (11 reports), 3 to 5 years (9 reports), and 6 to 15 years (7 reports).

The reported adverse reactions were categorized into three groups: the neurodevelopmental outcomes addressed in this report (autism, ADHD, speech or language problems); other neurological outcomes; and non-neurological outcomes. Table 4 details the type and number of outcomes reported in these three categories.

TABLE 4

Summary of Outcomes Attributed on VAERS Reports to Thimerosal.

In the first category, there were nine reports of autism and nine reports of speech or language problems. There were no reports of ADHD, although there was one report of attention problems (not otherwise specified) for a child with autism.

In the second category, the types of neurological outcomes reported include: headache (15), paresthesia (11), dizziness (9), asthenia (6), non-febrile seizure (4), developmental delay (3), cognitive abnormality (2), and hypertonia (2). Five reports describe mercury poisoning in the patient, but not did not identify a specific outcome. See Table 4 for less frequently reported neurological outcomes.

In the third category, 155 people reported non-neurological outcomes. The most commonly reported event was delayed hypersensitivity reactions. Because the focus of this report is on neurodevelopmental outcomes, non-neurological outcomes are not reported in detail here.

There were 12 rechallenge cases identified involving 11 non-neurological outcomes, and one neurological outcome (headache). The one neurological-related rechallenge case reported a headache after receiving the first and second doses of the hepatitis B vaccine.

The committee concluded that these reports were uninformative with respect to causality. VAERS and other case reports submitted to the committee are useful for hypothesis generation, but they are generally inadequate to establish causality. The analytical value of data from passive surveillance systems is limited by such problems as underreporting, lack of detail, inconsistent diagnostic criteria, and inadequate denominator data (Ellenberg and Chen, 1997; Singleton et al., 1999).

Epidemiological Studies

Published Studies. None.

Unpublished Studies. Controlled Observational Studies. One unpublished controlled epidemiological study, funded by CDC, has tested whether or not certain neurodevelopmental and renal disorders are related to exposure to thimerosal-containing vaccines. The study was based on data from the Vaccine Safety Datalink (VSD), a large-linked database that includes vaccination, clinic, hospital discharge, and demographic data. The VSD, formed as a partnership between CDC and seven health maintenance organizations (HMOs), was initiated in 1991 and covers approximately 2.5 percent of the U.S. population (Verstraeten, 2001).

The study was conducted in two phases. Phase I was designed to screen data for potential associations between exposures to mercury from thimerosal-containing vaccines and selected neurodevelopmental and renal outcomes. Phase II was designed to test the hypotheses generated in the first phase. Both phases were designed as retrospective cohort studies.

In Phase I, the original study population included children born between 1992 and 1997 who were enrolled continuously during their first year of life in one of two large West Coast HMOs and who received at least two polio vaccinations by one year of age. Excluded from the study were infants with diagnoses of congenital or severe perinatal disorders or who were receiving hepatitis B immunoglobulins, and premature infants with gestational ages less than 38 completed weeks. Of the approximately 213,000 infants born into these HMOs between 1992 and 1997, approximately 110,000 met the eligibility criteria for inclusion in the analyses. (The eligible children from both HMOs were combined into a single cohort in the original Phase I analysis.) Separate analyses were conducted for the premature infants (Stehr-Green, 2000).

For each child, cumulative vaccine-related mercury exposure was calculated at the end of the first, second, third, and sixth months of life from individual automated vaccination records. Each vaccine a child received was assumed to contain the mean amount of thimerosal reported by manufacturers to the FDA, meaning that the ethylmercury content per dose of each childhood vaccine was assumed to be as follows: diphtheria-tetanus-pertussis (whole-cell or acellular), 25.0 µg; hepatitis B, 12.5 µg; Haemophilus influenzae type b, 25.0 µg; measles-mumps-rubella, polio, pneumococcal, and varicella, 0.0 µg. The level of mercury exposure was categorized in increments of 12.5 µg. The outcomes studied included a range of plausible neurological and renal disorders identified in the mercury toxicity literature, and defined by specific ICD-9 diagnostic codes.

Cox proportional hazard models were used to compare the risk of adverse developmental outcomes. The endpoint of the observation period for each child was defined as the date of the first of the following events: first diagnosis, first disenrollment, or the close of the study period (December 31, 1998). The analysis was stratified by HMO, year, and month of birth; adjustments were made for sex. To obtain sufficient power to detect any association, the study examined only outcomes with more than 50 identified cases.

Preliminary results of the Phase I analysis produced statistically significant but weak associations (relative risk ratios < 2.00 per 12.5 µg increment of mercury) between various cumulative exposures to thimerosal and the following neurodevelopmental diagnoses: unspecified developmental delays; tics; attention deficit disorder (ADD); language and speech delay; and general neurodevelopmental delays. No association was found for renal disorders or other neurological disorders with more than 50 cases, including autism (Stehr-Green, 2000, 2001). The researchers interpreted the analysis as showing a possible association between certain neurological developmental disorders and exposure to mercury from thimerosal-containing vaccines prior to six months of age (Stehr-Green, 2000).

An outside review panel (known as the Simpsonwood Panel), convened in June 2000, concluded that the Phase I VSD screening analyses did not provide adequate evidence to support or refute a causal relationship between exposure to thimerosal-containing vaccines and specific neurodevelopmental disorders, but recommended that these issues be vigorously investigated. The Simpsonwood panel also noted several concerns about factors that could affect the interpretability of the Phase I study results (Stehr-Green, 2000, 2001):

• An ascertainment or health-care-seeking bias could exist. Children whose parents make greater use of health care services may be more likely to have received all recommended vaccinations—and therefore the highest doses of thimerosal-containing vaccines—and to have had a greater opportunity to receive a diagnosis for a neurodevelopmental disorder.
• The inexactness in the diagnosis of neurodevelopmental disorders, especially for young children, and inconsistencies in diagnoses across clinicians, clinics, and HMOs could result in misclassifications or biases that affect the clinical significance of the findings.
• The lack of data on known familial and genetic predispositions to the neurobehavioral outcomes could mask factors that confound the relationship between thimerosal exposure and the outcomes being studied.
• The data and analyses are limited in their ability to distinguish any effects of thimerosal exposure from excess risks attributable to other vaccine components or other vaccine-related associations.

The Simpsonwood panel also expressed reservations about the applicability of findings from the studies of methylmercury exposures to possible effects of thimerosal exposures from vaccines (Stehr-Green, 2000, 2001).

Re-analyses of the Phase I data, reflecting modifications to address some of the concerns of the Simpsonwood panel and others, were reported at the IOM committee's scientific meeting in July 2001 (Verstraeten, 2001). Among the modifications made in the re-analysis were: conducting separate analyses for each HMO (referred to as HMO A and HMO B); adjusting for additional variables in the models; making additional controls and adjustments to avoid a health-care-utilization bias; and checking for outcome misclassification through chart review of select diagnoses.

Of the 22,647 children born from 1992 through 1997 into HMO A, 15,309 children were eligible and retained in the final cohort for analysis. Of the 184,723 children born from 1992 through 1998 into HMO B, 114,966 children were eligible and were retained in the final cohort. Table 5 shows the neurological and renal outcomes with more than 50 identified cases that were examined in the analysis. Two control diagnoses, flat feet/toe deformities and injury at unspecified site, were also included in the analysis to check for a potential health-care-seeking bias. These diagnoses were thought to be associated with the worry of parents and increased health-seeking behavior, but a priori were not expected to be associated with thimerosal exposure. Because the final cohort at HMO B is about eight times as large as the cohort at HMO A, there are relatively more children included in the outcome categories at HMO B. Not all of the outcomes that were examined at HMO B could be examined at HMO A due to insufficient numbers of cases in some outcome categories at HMO A.

TABLE 5

Neurodevelopmental, renal outcomes, and control diagnoses > 50 children.

As in the original Phase I analysis, the re-analysis identified positive but weak associations (relative risk ratios < 2.00) with several neurodevelopmental diagnoses (see Table 6). Results from unadjusted and adjusted models were presented to the committee. The adjusted models, which are used in epidemiology to control for potential confounders, included the following additional variables: birthweight, gestational age, mother's age at delivery, race and ethnicity, and Apgar score at 5 minutes (the last at HMO A only). The adjusted models included only approximately 80 percent of the eligible children because variables were missing for some children. For HMO A, positive but weak associations were found between exposures to thimerosal and stammering, sleep disorders, and emotional disturbances. The association with sleep disorders did not persist in the adjusted models, and the association with emotional disturbances appeared only in the adjusted models. For HMO B, positive but weak associations were found for: any neurodevelopmental disorder, stammering, language delay, speech delay, attention deficit disorder, and tics. The association with attention deficit disorder and tics did not persist in the adjusted models. In both rounds of analysis of Phase I data, the results failed to show a consistent dose-response pattern.

TABLE 6

Relative risks by increase of 12.5 µg ethylmercury—Summary of positive associations *: p < 0.05 **: p < 0.01.

The inconsistent findings between HMO A and B in the re-analysis could be due to several factors. At HMO B, a positive association was found in both unadjusted and adjusted models between thimerosal exposure and the control diagnosis of flat feet/toe deformities, indicating a potential health-care-seeking bias at HMO B. In addition, the sample size was considerably larger at HMO B, providing more power to detect small differences in risk, if they exist. It should also be noted that the chances of finding an association between thimerosal exposure and these outcomes, if it exists, depends on the range of thimerosal exposures in the population; an HMO that is particularly successful in vaccinating its children would have less range of exposure and ability to detect positive findings.

The Phase II study was conducted in mid-2000 to examine more closely the significant associations identified in the original Phase I analysis by replicating the Phase I study design with data from an East Coast HMO. However, the study population had only approximately 17,500 eligible children, providing a sufficient number of cases for analysis of only two of the outcomes, ADD and speech delays. The Phase II analysis identified no significant differences in risk with the receipt of thimerosal-containing vaccines and these two outcomes; however, the small sample size limited the power of the study to detect a small effect, if it exists (Verstraeten, 2001).

As a result of the inconsistent findings from the first two phases, a protocol for a Phase III follow-up study has been developed, although funding for the study is not confirmed. As described to the committee at its scientific meeting in July 2001 (Stehr-Green, 2001), the study will use a retrospective cohort design and will focus on primary diagnoses from the Phase I analysis, including ADD, language and speech deficits, and tics. Autism is not included because a substantially larger study population (and therefore a substantially more expensive study) would be needed to identify a sufficient number of cases for a retrospective cohort study. A separate case-control study of thimerosal exposure and autism has been proposed, but no study protocol has been developed. The Phase III protocol is discussed further in the research recommendations section of this report.

The committee notes several limitations of the Phases I and II of the VSD study. First, the consistency and accuracy of the diagnoses have not been established. Second, the variation between HMOs in the results of the updated Phase I analysis indicates an ascertainment or health-care-seeking bias could exist at HMO B among parents of study subjects who received the highest doses of thimerosal-containing vaccines. The exclusion of children who disenrolled from the HMOs is another potential source of bias. In addition, the effect sizes are small and thus difficult to interpret. Furthermore, with an observation period of only one year for some children, diagnoses usually made at older ages may be under-represented. With its restriction to diagnoses with at least 50 cases, the study is uninformative about other important but less common diagnoses. In addition, there is not a consistent dose-response relationship. Also, information on prenatal exposure to mercury is sparse. Finally, the findings from Phase I and Phase II of the study are inconsistent. Given these caveats, the committee believes that the Phase I and II VSD analyses are inconclusive with respect to causality.

Uncontrolled Observational Studies. An unpublished ecological analysis comparing aggregate trends in autism rates in California with the trends in mercury exposure from thimerosal-containing vaccines was presented to the committee (Blaxill, 2001). To estimate the average level of mercury exposure for children 19–35 months of age in a given year, the mercury content of all recommended vaccine doses (assumed to be 37.5 µg for 3 doses of hepatitis B, 75 µg for 3 doses of Hib, and 75–100 µg for 3 or 4 doses of DTP) was weighted by estimates from the National Health Interview Survey of coverage rates for the specific vaccines for the specified year. Estimates of the number of children with autism were obtained from annual caseload data from the California Department of Developmental Services database. These data were retabulated by year of birth, and birth-cohort prevalence rates were calculated using census data on births by year in California. The presenter concluded that the increasing trends seen in these data, in both the prevalence of autism and the levels of mercury exposure from thimerosal-containing vaccines, are consistent with the hypothesis that mercury exposure had a direct role in an increasing incidence of autism in the 1990s.

As noted in previous IOM reports (IOM, 1994 a,b, 2001), a positive ecological correlation constitutes only weak evidence of causality, and additional research would be needed to establish a causal association. In addition, the appropriateness of the transformation of the cross-sectional autism-caseload data into birth-cohort prevalence data, as was done in this analysis, is questionable. The data used in this analysis are cross-sectional and reflect the number of children with a diagnosis of autism who are registered in the California Developmental Services system. Transforming the cross-sectional data by dividing by birth cohort, creates the artificial impression of a cohort effect when the data are cross-sectional in nature (Fombonne, 2001a). Furthermore, the authors of a report on the California autism caseload data stress that their study was not designed to measure trends in autism and that the data should therefore be interpreted with caution (California Department of Developmental Services, 1999). The analytical value of the data is limited by the inability to account for any changes over time in diagnostic concepts, case definitions, or age of diagnosis (Fombonne, 2001a). As a result, the committee cannot assess trends in the prevalence of autism from these data. The committee concludes that this unpublished ecological analysis is uninformative with respect to causality.

Epidemiological Studies

• The committee recommends case-control studies examining the potential link between neurodevelopmental disorders and thimerosal-containing vaccines. The studies should use clearly defined outcomes, specifically diagnoses (e.g., identified in terms of ICD-9 coding) that correspond to current practice. Gradients of performance on neurodevelopmental tests should be used as outcome measures only if they can be linked to specific diagnoses. Furthermore, the committee recommends examining multiple cognitive outcomes, including autism. In addition, because thimerosal poisonings were associated with adverse renal effects, renal outcomes should also be included in the epidemiological, clinical, and basic science studies of thimerosal exposure recommended in this report. Although there are many challenges that will arise in planning and conducting these studies (e.g., appropriate control group selection, conducting a retrospective assessment of mercury exposure from other sources), the committee believes that multiple case-control studies are an efficient approach for seeking answers to the causality questions.
• The committee is aware of several cohorts of children outside the United States who did not receive thimerosal-containing doses as part of a clinical trial of DTaP vaccine. The committee recommends further analysis of neurodevelopmental outcomes in these populations. Although the exposure levels to thimerosal in these children are lower than exposure levels in the United States, these cohorts have the powerful analytic benefit of randomized assignment to thimerosal-free or thimerosal-containing vaccines. Studies using these already-defined cohorts could probably be completed more quickly than a study that would have to define and recruit some other population in the United States or elsewhere.
• The removal of thimerosal from vaccines on the recommended childhood immunization schedule in the United States presents a unique opportunity to study whether this change affected the rates of neurodevelopmental disorders in children. The committee recommends conducting epidemiological studies that compare the incidence and prevalence of neurodevelopmental disorders before and after the removal of thimerosal from vaccines. These studies should focus on multiple renal and cognitive outcomes, including autism. Collaborations with existing research studies, such as NIMH's Epidemiological Catchment Area studies or the Collaborating Program of Excellence in Autism, might be efficient and should be explored. Ecological studies are not well-suited to making individual-level causal inferences and comparisons before and after the removal of the thimerosal are problematic when the criteria for diagnosis, or the manner in which the criteria are applied, change over time, as they have with autism. However, the committee believes that well-designed ecological studies can provide support to other epidemiological studies.
• The committee recommends an increased effort to identify the primary sources and levels of prenatal and postnatal background exposure to thimerosal (e.g., Rho (D) Immune Globulin) and other forms of mercury (e.g., maternal consumption of fish) in infants, children, and pregnant women. Data on background exposures to mercury are sparse; additional studies may identify populations or quantify the number of children at higher risk for mercury toxicity. A feasibility assessment should be done to see if a meaningful study could be conducted regarding the risk of neurodevelopmental disorders due to thimerosal exposures through Rh immunoglobulin during pregnancy.

Where possible, studies of the rates of specific neurodevelopmental diagnoses in cohorts of children exposed to varying levels of thimerosal would be of interest. Such studies have to be large enough to have sufficient variation in exposure and in incidence rates of the diagnoses of interest to detect an association. They also should include uniform data on thimerosal exposure, clear definitions of neurodevelopmental disorders using current diagnostic rubrics, the use of standard clinical protocols for diagnosing neurodevelopmental disorders, and assessment of neurodevelopmental diagnoses by professionals who do not know the patients' exposure status.

The committee is aware of the planning under way for a two-stage follow-up study (Phase III) to the findings from the Phase I and II VSD studies. The plans for Phase III have been presented to several groups, including the committee at its July 2001 meeting (Stehr-Green, 2001). The Phase III study is conceived as a two-part retrospective cohort study in which participants, seven- to nine-year-old children, would be enrolled into one of three groups, based on exposure to thimerosal-containing vaccines at specified points in the past. The three groups would be defined according to two exposure factors: level of mercury exposure at birth, based on whether the birth dose of hepatitis B vaccine was received; and cumulative level of exposure to all thimerosal-containing vaccines during the first three months of life. The decision to group participants on the basis of levels of exposure during the earliest ages of life reflects the advice of expert consultants and a review of the methylmercury literature. The study will collect data on all vaccine exposures so as to permit analyses based on cumulative exposure at any age.

The study would focus on specific primary diagnostic outcomes identified from the Phase I screening analysis, including ADD, language and speech deficits, and tics. Autism is not included as an outcome in this study because of the high cost of the large sample size that would be needed to identify a sufficient number of cases of autism for a retrospective cohort design. A separate case-control study of autism has been proposed, although no study protocol has been developed.

The proposed follow-up study uses a two-part approach to assessing outcomes. In the first stage, a standardized set of neuropsychological tests would be administered to all study participants to identify children whose performance suggests the presence of selected neurodevelopmental disorders (NDD). In the second stage, confirmatory neuropsychological tests would be administered for ADD and speech or language delay specifically. The study would also develop measures of potential confounders (e.g., exposures to mercury, lead, PCBs, alcohol and other drugs, genetic predisposition, family medical history) and would attempt to evaluate the independent contribution of other vaccine antigens and components.

The committee recognizes that this study is still in the planning phase. Three issues loom large regarding the proposed Phase III study: feasibility, resources, and several technical design issues. The committee has reservations about such an ambitious and therefore resource-intensive study. It will be a few years until results and meaningful analyses are available. In addition, the power of the study to detect small relative risks is limited. The proposed Phase III study could thus be contributory, but would best be undertaken as part of an overall package of research and only if it accurately identifies neurodevelopmental conditions of concern.

Clinical Studies

• Very little is known about the pharmacokinetics of ethylmercury exposure in humans. Better understanding of these mechanisms would have greatly facilitated the risk assessment of thimerosal in vaccines. The committee recommends research on how children, including those diagnosed with neurodevelopmental disorders, metabolize and excrete metals—particularly mercury. Studies of proteins known to be associated with metal metabolism, such as glutathione and metallothionein, could be undertaken. Some of this research could be done expeditiously through autism research centers, which have great expertise in identifying and working with autistic children.
• The committee recommends continued research on theoretical modeling of ethylmercury exposures, including the incremental burden of thimerosal with background mercury exposure from other sources. It would be particularly helpful to have such theoretical modeling done in conjunction with empiricists, so that the models could be better constructed.
• The committee is aware of several specialized pediatric practices that use chelation therapy to treat autistic children. These practitioners report unusual metal profiles in their patients as well as clinical improvement following chelation. However, chelation therapy is currently indicated only for acute, high-dose mercury poisonings. Given that chelation therapy is not a benign treatment, the committee recommends careful, rigorous, and scientific investigations of chelation when used in children with neurodevelopmental disorders, especially autism. Although studies of chelation would not be able to link excreted metal specifically with vaccine exposure, and therefore would not contribute to causality assessments, it is important to pursue these uncontrolled clinical observations in order to establish an evidence base for appropriate therapeutic uses of chelation.

Basic Science Studies

• Many countries continue to use thimerosal-containing vaccines given the proven benefits of thimerosal as a vaccine preservative for many years and the benefits of continued immunization. Complete risk assessments have not been done for other countries that would indicate the need to switch to thimerosal-free vaccines. However, the committee recommends research to identify a safe, effective, and inexpensive alternative to thimerosal for countries that decide they need to switch.
• Comparative animal studies of the toxicity of ethylmercury and methylmercury are limited, though the teratological effects of methylmercury in rodents and primates is well established. Several important questions regarding the potential risk of thimerosal exposure could have been informed by a wider and deeper literature from studies in laboratory animals. The committee recommends research in appropriate animal models on neurodevelopmental effects of ethylmercury. These would help elucidate the comparability and validity for ethylmercury of the risk assessments based on methylmercury.