Mercury and Autism: What Do We Know? What Don't We Know?

Background: The role of mercury and specifically, thimerosal-containing vaccines, in autism has received considerable attention. A few large studies have addressed this concern, using databases from health practices, health maintenance organizations, and the CDC's Vaccine Safety Datalink.

Objectives: To evaluate the strength of the evidence regarding a role for thimerosal-containing vaccines in the risk for autism. To determine what hypotheses have been addressed in those studies.

Methods: Searches were conducted using PubMed, Medline, and citations listed in previously identified original and review articles. Reports with scientifically valid data, a minimum of 25 autism cases and a primary analysis focused on thimerosal-containing vaccines were included in the review. A total of six studies met these criteria. These studies were then evaluated for design characteristics (ecological vs. individual), presence of dose information, and the nature of the hypothesis that was tested. Potential biases in selection and uncontrolled confounding were also assessed.

Results: Two were of an ecological design (Madsen et al 2005, Schecter & Grether 2007) and four used individual-level data on vaccines administered. These four had number of vaccines administered to each child and conducted an analysis by dose or a surrogate for dose.. Doses were lower in studies conducted outside the US. In studies from Denmark (Madsen et al 2003, Hviid et al 2003), the sources of cases in the post-thimerosal period included outpatients whereas only inpatients were included in the thimerosal-exposed period, suggesting lack of comparability in case ascertainment for the two exposure periods. Several UK-based studies (Andrews et al 2004, Heron et al 2004) addressed the role of early exposures, e.g., prior to 3, 4 or 6 months of age, comparing those who received thimerosal-containing vaccines by a given age vs. those who did not and hence may have received them at a later age. In these studies, most adverse developmental outcomes were less prevalent in those receiving the early vaccines, and several were significantly less prevalent, raising the possibility of a ‘healthy vaccinatee effect' (similar to the well-characterized ‘healthy worker effect' found in occupational studies). In the CDC study (Verstraeten et al 2003), a relatively large percentage of births (~20%), including many who would be considered at increased risk for autism based on current knowledge (low birth weight, maternal perinatal conditions, malformations of genetic or chromosomal origin), were excluded. In a multi-factorial developmental disorder, those with predisposing conditions may be most vulnerable to a later insult. Thus, the exclusion criteria may have lowered statistical power and introduced selection bias.

Conclusions: Of published studies meeting pre-established minimal criteria, none supports an association between thimerosal-containing vaccines and autism. However, in each of the null studies, methodologic weaknesses limit the conclusions that can be drawn. The strength of evidence in support of the hypothesis that thimerosal-containing vaccines play no role in autism may have been overestimated in previous reviews and editorials. Bias and confounding, including that due to indications for early vs. later vaccine administration, should be addressed in future studies.