Thimerosal and Autism: Reviewing the Evidence

Jay Liebermann, M.D., M.P.D.
July 12, 2007

There are three forms of mercury. One form is an organic mercury, of which the two primary forms and ethyl and methylmercury. Ethylmercury is the form used in some vaccines. Mercuric salts are found in batteries and elemental mercury, e.g., quicksilver. Mercury is ubiquitous in the global environment; 95 percent of environmental mercury resides in soil and is released into the environment by burning fossil fuels such as coal and petroleum. It can also be released from rock erosion and volcanoes. It has been estimated that increases in power-plant emissions and industrial uses over the past century have tripled the amount of environmentally available mercury. Atmospheric mercury can be deposited on the surface of bodies of water, where elemental mercury is converted to methylmercury, usually by bacteria, and ingested by fish. The principal source of human exposure for organic mercury is fish consumption, particularly fish higher in the food chain. For an example, a can of tuna contains about 28 micrograms of methylmercury. Mercury can also be found in breast milk if the mother has concentrations in her system and in some cosmetics used by certain ethnic groups, as well as some herbal remedies and dental amalgams.

Mercury is a known neurotoxin, and the fetal brain appears to be especially susceptible to exposure to organic mercury. There have been outbreaks described with high-dose exposures. One was in Japan in an area called Minimata, where industrial pollution heavily contaminated the ocean waters with mercury. When locals ate contaminated fish, it caused many cases of mercury poisoning, deaths and infants born with severe developmental disabilities, including cerebral palsy, mental retardation, and seizures. In Iraq, there was an incident of grain contamination in the early seventies, causing many cases of mercury poisoning and deaths, and infants were also born during that time with severe developmental disabilities. In this case, researchers were able to document to some degree a dose-related response, higher levels being associated with more severe disability.

In the 1980s, two large cohort studies tried to correlate prenatal methylmercury exposures with neurodevelopmental outcomes. One study in the Faroe Islands was able to show an association between mercury levels in the mother’s hair and cord blood and neurodevelopmental outcomes at seven years of age. This study is extremely important because the data were used to try to determine a level of mercury exposure to the fetus at which or below which one would not see effects on neurodevelopmental outcome. These were used to develop federal guidelines for safe levels of mercury exposure. A similar study was done in the Seychelles and no association was found, although the exposure to mercury was about the same. The reason for the difference is not clear. In the Seychelles, exposure is through fish, which they eat often. In the Faroe Islands, most of the mercury exposure comes from eating pilot whales, which are contaminated with other things such as PCBs. The role PCBs might be playing in the neurodevelopmental outcomes is not clear.

Thimerosal is about 50 percent ethylmercury, whereas methylmercury is the predominant form of organic mercury in the environment. Many childhood vaccines contain between 12.5 to 25 micrograms of mercury. Through the eighties, it was in flu vaccine and various diphtheria/tetanus/pertussis vaccines. In the late eighties and early nineties, as hepatitis B and Hib vaccines were added to the immunization schedule, there were now more vaccines in the childhood immunization schedule that contained thimerosal.

A goal of the 1997 FDA Modernization Act was to compile a list of drugs and foods that contained intentionally introduced mercury compounds and provide a quantitative analysis. The FDA concluded that infants who received thimerosal-containing vaccines at several visits could exceed the total mercury exposure recommended by Environmental Protection Agency guidelines. These guidelines for exposure were set to avoid toxicity to the fetus. They were based on studies of oral ingestion of methylmercury, usually through fish. It was assumed that, since digestion was going on daily over months, the half-life of methylmercury was about 50 days. The aim was to develop a no-effect level that would be safe and then put in a safety factor. For the EPA, that factor was about tenfold and they set the level at 0.1 micrograms per kilogram per day of methylmercury. FDA guidelines were a little higher because of how they analyzed the data and their safety factor, but these are not toxic levels, they are safe levels, set to minimize the chance of neurodevelopmental problems. Looking at different ages, specifically females in the first six months of life by different weights and different body percentiles, and calculating by the different standards for mercury (EPA, FDA, WHO, etc.), one can ask what an acceptable or safe level of mercury would be for that daily exposure.

These exposure limits had a number of assumptions. For vaccines, there’s an assumption that the toxicity and pharmacokinetics of ethylmercury in thimerosal and methylmercury are the same, but exactly how they relate is unknown. There’s some evidence that ethylmercury is excreted faster in the stools, and some evidence in monkeys that less goes to the brain. They also assume that the effects of low-dose oral exposure daily are the same as a bolus IM injection and that the susceptibility of the infant to toxicity is the same as that of the fetus.
There are also many assumptions when making the leap from what is happening to a developing fetus to giving an injection at two and six months of age. However the conclusion was that children receiving all possible thimerosal-containing vaccines could receive quite a bit of thimerosal compared to the EPA guidelines – 200 micrograms by six months of age. For girls, the recommended upper limit was between 65 and 106 micrograms depending on weight, and by two years, 275 micrograms.

In 1999, a joint statement by the AAP/FDA U.S. Public Health Service urged manufacturers to remove thimerosal from vaccines as soon as possible, as a precautionary measure to maintain the public’s trust in immunization. There was no evidence at the time of any harm caused by the low levels in vaccines. The big uproar came from pediatricians who saw their immunization schedule potentially disrupted.

Thimerosal as a preservative was removed from most childhood vaccines by 2001 and the last lots of thimerosal preservative-containing vaccines expired in January 2003. It is still being discussed only because of ongoing litigation in both state courts and the Vaccine Injury Compensation Program. Thimerosal is still contained in some of the flu vaccines routinely given to children since 2004. There are also vaccines that contain trace, often unmeasurable, levels of thimerosal.

During the nineties, the number of persons reported to be receiving services for autistic spectrum disorders increased substantially. Rates are tenfold higher than in the 1970s; in the MMWR, rates ranged from 4.5 to almost 10 per thousand eight-year-olds. Newspapers have reported that one in 150 children has autism. A study published in Nature Genetics from the Autism Genome Project analyzed genes from more than 1100 families with at least two children with autism. There is some very promising research identifying some loci of genes on chromosomes, particularly Chromosome 11.

While researchers are getting closer to understanding where autism comes from, especially regarding the strong genetic component, they still do not know if an environmental hit is also necessary. There’s some evidence that it occurs in a susceptible host in whom something else happens. Prenatal exposures, such as Thalidomide, have been linked with autism. Interestingly, those children who developed autism had very specific birth defects, which suggested a specific time of onset. Congenital rubella, as well, had been associated with autism. To date, there’s no good evidence that postnatal exposure to anything is linked with the onset of autism, but the concern about mercury in thimerosal in the vaccines remains. A paper published in Pediatrics several years ago looked at autism and some of the neurologic characteristics, comparing it to the known neurotoxicity of mercury. In comparing the two, they were dissimilar.

The Institute of Medicine was commissioned by the CDC and the NIH to review specific vaccine-safety topics, including vaccines and autism. In their 2001 report, they concluded that the evidence was inadequate to accept or reject a causal relationship between exposure to thimerosal from vaccines and autism, ADHD, speech and language delay. The reason was that there were no published epidemiologic studies examining the potential relationship between thimerosal exposure and neuro-developmental disorders. They did conclude, however, that because mercury is a known neurotoxin and because prenatal exposures to methylmercury have been documented to negatively affect early childhood development, a potential biological mechanism could at least be hypothesized and was worth investigating.

To test this hypothesis, ideally there would be a double-blind, randomized, placebo-controlled trial, randomizing children to get thimerosal-containing vaccines or not and following them for many years. However the sample size would be prohibitive and it couldn’t be done from an ethical standpoint either, since the goal was already to remove thimerosal from vaccines. This leaves retrospective cohort studies, looking at a well-defined cohort, their exposures to thimerosal and gradations of exposure, and neuro-developmental outcomes. In addition, ecological studies can track changes in incidence of autism and other neuro-developmental disorders as the use of thimerosal changes.

The first published study, from Denmark, was a population-based cohort study of all children born in that country between January 1990 and December 1996, a total of almost half a million. Their health care system was able to track and compare children vaccinated with thimerosal-containing vaccines with children vaccinated with thimerosal-free vaccines, in general, the diphtheria/tetanus/pertussis vaccines as they evolved. The use of thimerosal was discontinued in 1992. They were able to identify 440 cases of autism and 787 with other autistic spectrum disorders. They found that the risk of autism and autistic spectrum disorders did not differ significantly between groups. For example, for autism, the risk ratio was 0.85, showing no link between getting vaccines containing thimerosal and the development of autism and autistic spectrum disorders. In addition, they found no evidence for a dose-response association.

The next big study was done in the U.S., using the Vaccine Safety Datalink project, which was developed as a screening study and initiated in late 1999. The study was devised to examine the association between thimerosal exposure from childhood vaccines by one, three, and seven months of age and various diagnoses, including autism, speech and language delays, tics, and ADD. This study was in two phases; the first phase at two managed-care organizations (HMOs A and B) looking at all children born between January 1992 and December 1998. The second phase was done at a third managed-care organization (HMO C). The preliminary findings showed a statistically significant dose-response association between exposure to thimerosal at three months of age and any of various neuro-developmental disorders.

These findings were discussed at a meeting in Georgia, in early June 2000. Several consultants were invited to review the findings and advise the CDC. They were ultimately presented at the June ACIP and then at the Institute of Medicine. The meeting was the focus of the articles in Rolling Stone and Salon, which suggested that something nefarious was going on because the final published report findings were different from the preliminary findings. They differed because after peer review the authors corrected errors and refined their analytic methods, adjusting for utilization of healthcare services. The major influence on when children got the vaccine was how often they went to the doctor, which would also influence diagnoses and behaviors. They also extended the follow-up period, so they had more diagnoses. But the fact that the final results differed from the preliminary just fueled suspicions that something was being hidden.

In the published paper, in Phase I, the cumulative thimerosal exposure at three months was associated with tics. At HMO B, the cumulative exposure at three and seven months was associated with language delays, with relatively low relative risks. In Phase II at HMO C, there were no significant associations between cumulative thimerosal exposure at one, three, or seven months and speech or language delay, ADD, or tics. Only HMO B had a sufficient number of cases of autism to perform an analysis, and even in the preliminary analysis, autism was not significantly associated with thimerosal exposure. The conclusion was that there was no evidence of a clear association between thimerosal exposure in infant vaccines and specific neuro-developmental disorders. But results among the HMOs were inconsistent, and so further investigation was recommended.

The next study was a retrospective cohort study looking at over 100,000 children born in the U.K. between 1988 and 1997, which evaluated the relationship between exposure to thimerosal via DT or DTP vaccines and neuro-developmental outcomes. There was some evidence for higher risk of tics with increasing doses at four months, but no negative associations were found between thimerosal exposure and ADD, general developmental disorders, and unspecified developmental delay.

A problem with these cohort studies is that diagnostic accuracy of outcomes such as ADD or autism is not perfect. For example, in the study just presented, 90 percent of the tics were transient. The doses of thimerosal also differ among the studies. Cohort studies are best used to quantify a risk to exposure rather than to prove its absence.

Dr. Liebermann then talked about ecological studies. In 2000 the Institute of Medicine presented data showing the amount of thimerosal in vaccines and cases of autism diagnosed in California by year of birth. The data show increases in the amount of thimerosal given to children by birth cohort and increases in the rates of diagnosis of autism, and they track fairly well. The rate of autism rose in the mid-eighties, even before an increase in thimerosal. However, anything that increased in the nineties tracks with diagnoses of autism, including use of home personal computers and cell phones. This kind of data only shows that two things were increasing at the same time. It says nothing about a possible association.

An ecological study in Denmark analyzed data from almost 1,000 children diagnosed with autism over 30 years. Thimerosal was used in childhood vaccines from the early fifties until 1992, and there was no trend for an increase in autism up through 1990. From 1991 to 2000, the incidence of autism increased, after thimerosal was discontinued from vaccines, suggesting no relationship between thimerosal and autism in Denmark.

In Sweden, a similar ecological study analyzed inpatients diagnosed between two and ten years of age with autism over a 12-year period. It looked at the average cumulative dose of thimerosal using vaccine-coverage levels. Thimerosal was eliminated by 1993, after which the data show no decline in autism, in fact, it continued to rise. From 1980 to 1996 there was almost a doubling in cases. Looking at all these studies, the Institute of Medicine report in 2004 said the evidence favors rejection of a causal relationship between thimerosal-containing vaccines and autism.

There have been studies showing an association: two ecological studies and three studies using passive reporting data, all by the same authors, who are the only ones who ever found a relationship between MMR vaccine and autism. Quoting the Institute of Medicine, “These studies cited have serious methodological flaws. Their analytic methods were nontransparent, making the results uninterpretable and therefore noncontributory with respect to causality.” They used the VAERS database to calculate an incidence of neurodevelopmental disorders and heart disease following thimerosal-containing versus thimerosal-free DTaP vaccines. It might be possible with rare events, like intussusception or Guillain Barre, but not autism. They found an exponential distribution link between autism, speech disorders and heart arrest, and thimerosal dose. VAERS has many limitations. It’s passive and there’s underreporting, incomplete reporting, and, certainly, bias. It can’t be used to calculate incidence and the cases weren’t verified. These authors also did an ecological study with U.S. Department of Education data. They compared autism to thimerosal exposure by birth cohort and showed a linear relationship – as thimerosal in vaccines increased, autism increased. However, that does not prove causality and how they actually got their numbers is not clear.

A paper published by Sarah Parker and colleagues carefully analyzed the studies looking at autism and thimerosal exposure, emphasizing the quality of the studies. How good are the data? Are exclusion criteria defined and outcome measures precisely described? Is there a basis for the sample size? Is bias controlled? The paper concluded that while the studies that do not support a relationship are not perfect, one can understand the methodology. The studies that do support a relationship are worthless to evaluate any possible association between thimerosal and neurodevelopmental disorders.

Since the IOM report, there have been two new studies. In the U.K. a longitudinal study of more than 14,000 children determined the ages at which they got thimerosal-containing vaccines, calculated levels of mercury exposure, and compared them with development at between 6 and 91 months of age. Again, the results showed no evidence of any harmful effect of early exposure to thimerosal on neurological or psychosocial outcomes. Indeed, the unadjusted results suggested a beneficial effect of thimerosal exposure. For example, those who had thimerosal had less hyperactivity and better model development. When results were adjusted for possible confounders, eight of the nine significant associations actually showed a beneficial effect, not suggesting that thimerosal is protective, but no evidence of harm. The only negative association was poor pro-social behavior at 47 months.

The last study, published this past year from Montréal, looked at almost 30,000 children and evaluated the relationship between the development of pervasive developmental disorder and changes in vaccines, thimerosal exposure as well as MMR. The cumulative exposure to thimerosal increased and decreased over time. By 1996, it was out of the vaccines. There were a total of 182 children with pervasive developmental disorders. Looking at average thimerosal in the vaccines and diagnoses of autism by birth cohort, the same trend of increases in cases over time can be seen, even in those infants who were not exposed to thimerosal. This seems to be the most compelling evidence of no association. If thimerosal is responsible for autism, as it is removed from vaccines, cases of autism should decline, but that is not happening.

Data from California from 2002 to 2006, which has been submitted for publication, indicates that cases of autism in three- to five-year-olds and six- to nine-year-olds have continued to increase. There’s absolutely no evidence of any decrease in cases of autism since thimerosal has been taken out of vaccines. Instead, cases have continued to climb. The science shows that in well-designed epidemiologic studies, there’s no association between thimerosal exposure from vaccines and autism. The ecological studies show that autism does not go down when thimerosal is removed from childhood vaccines. Studies in tissue cultures and animals provide interesting information about toxicity, but they do not translate to what is going on in children.

Finally, Dr. Liebermann talked about unintended consequences. For example, the birth dose of hepatitis B vaccine is universally recommended to prevent against failures of screening, but hospitals discontinued the routine birth dose in 1999, after the joint statement until thimerosal-free vaccines became available. However, by 2006, vaccine-coverage rates for the birth dose were still below the 1999 levels. As a consequence, hundreds of children in the U.S.were born to hepatitis B-positive mothers and, for various reasons, not screened. The infants did not get their dose of hepatitis B vaccine and, therefore, were at risk of becoming chronic carriers of hepatitis B. Concerns about vaccine safety and the fear that the MMR vaccine caused autism caused MMR vaccination rates to fall in the U.K. with subsequent outbreaks of these vaccine-preventable diseases. Fear of thimerosal caused some high-risk children to avoid the recommended flu vaccine. So there are real consequences to every decision made.

California has passed legislation banning thimerosal-containing vaccines for children under three years and pregnant women in that state. A provision was put in for an exemption in case of emergency or vaccine shortage. In fall 2006, there was a delay in shipping some doses of flu vaccine for children three years of age and younger because one virus in the vaccine was slow growing. A group of medical organizations requested an exemption, which was granted. There are no data to show how this influenced flu vaccines in California, but flu vaccines are time-dependent and not having vaccine makes it very hard to vaccinate children. Anything that limits the availability of vaccines for children potentially puts them at risk.

In conclusion, the evidence does not support an association between thimerosal and autism. The consistency among these well-designed studies lends strength to their individual conclusions. Autism is increasing and research should be directed towards areas of more promise.

Dr. Harry Hull asked if there were any other possible reasons why the diagnosis of autism might be increasing. For example, there was something in the popular press recently saying that autism is going up but mental retardation is going down because it’s being reclassified, or that support programs for families with an autistic child might be reaching more children who are diagnosed with autism. Dr. Liebermann replied that part of the increase is related to changes in diagnostic criteria and accessibility to services. But among pediatricians, there’s also a sense of a true increase, which is why it’s important to investigate what is leading to that increase.
Dr. David Salisbury noted that the levels reported in the U.K. schedule were lower than those experienced in the U.S. However, the vaccines were given at two, three, and four months in the U.K. rather than two, four, six, in part because children are actively called in through registers. The fact that the U.K. studies showed no association is important because the ages map across quite reasonably and compliance with the recommended ages for vaccination is very tight.

This page was posted on July 12, 2007.