Abstract
Concerns continue to be raised regarding the potential role of mercury exposure in the etiology of autism. In a recent article published in this journal (“Reduced Levels of Mercury in First Baby Haircuts of Autistic Children”), Holmes, Blaxill, and Haley (2003) report intriguing findings of H1: Low hair mercury concentration is a reliable and valid measure of impaired excretion capability in young children. H2: Impaired ability to excrete mercury in some children may, in the presence of mercury exposure, contribute to the development of autism. H3: The greater the impairment in excretion, the more severe the autism. H4: Autistic children have normal mercury excretion but have a lowered uptake in first baby hair. H5: Study bias or chance account for the findings in this study.
To our knowledge, the relationship of mercury concentration in hair to overall excretion capability in infants and young children has not been established. High hair levels are considered markers of high methylmercury exposure, at least in adults (Grandjean, Weihe, and Nielsen 1994). High maternal hair levels of methylmercury at delivery have been associated with neuropsychological deficits in their young children and high levels in hair of 7-year-old children were associated with subtle deficits in performance (Grandjean, Weihe, and White 1997), (Grandjean et al. 1999; Weihe et al. 2002). While it is possible that excretion in infants and very young children follows a different biologic trajectory than in older children and adults, the interpretation put forward by Holmes et al. of their unexpected findings requires confirmation in a study specifically designed to test this hypothesis.
As acknowledged by Holmes and colleagues, their study had several methodological limitations that might explain the surprising findings. Both the case and control groups represent “convenience samples” that undoubtedly bias the findings to an unknown extent. The children with autism were referrals to the lead author who is an advocate for chelation therapy for children with autism. Although the overall gender pattern among the children with autism in the study is consistent with the literature, the high percentage of females in the mildly affected case group is contrary to available data from numerous sources that consistently show increasing imbalance in the sex ratio with decreasing severity of impairment. Similarly, controls were recruited through autism parent groups and autism newsletters and are unlikely to be representative of the population of typically developing children.
As reported in the article, the data collection methods may also have introduced further bias. Interviews with parents were conducted by the lead author, ASH, who was not blinded to the study hypothesis nor to the diagnostic status of the children. It is not stated if study respondents were kept blinded to the hypothesis, but this seems unlikely since parents of affected children had sought the lead author’s help for evaluation of mercury toxicity in their children. This raises the possibility that respondents knew what the study was about and may have responded accordingly. Of further concern, Holmes et al. found that 46% of mothers in their case group reported being treated with Rho D immunoglobulin during pregnancy, a strikingly increased proportion compared to the control group and to population estimates. Rho D immunoglobulin contained the ethyl mercury preservative, thimerosal, until late in the 1990s and the authors interpret their case-control difference as indicating that Rho D immunoglobulin treatment is a risk factor for autism. Alternatively, the findings could indicate substantial recall bias in maternal reports (which were not validated through review of medical records) and/or a highly selected and unrepresentative group of cases. The authors also report that maternal dietary consumption of fish was not significantly associated with autism in their study. This is surprising given that maternal fish consumption during pregnancy is typically the source of greatest prenatal mercury exposure.
The study also raises concerns related to specimen gathering, processing, and laboratory analysis. Under the best of circumstances, analysis of hair samples is a highly imperfect methodology for measuring past exposures to metals. Unfortunately, sometimes it is the only available way to obtain such data. An expert panel on hair analysis convened by the federal Agency for Toxic Substances and Disease Registry (ATSDR) in 2001 reported that the lack of standard procedures for collecting, washing, and analyzing hair samples presents notable limitations in using hair samples in etiologic studies (ASTDR 2003). Holmes et al. reported that the laboratory testing was conducted in a blinded fashion but it is not clear that the specimens were submitted to the laboratory during the same time period and thus randomly analyzed. Without further information to evaluate this possibility, we are concerned that the observed differences between mercury levels in specimens from case and control subjects could be due to variability over time in laboratory analysis.
Whether or not the degree of bias introduced by these limitations is sufficient to invalidate the provocative findings of this study can only be determined by further study. An adequate test of the hypothesis that impaired capability to excrete mercury contributes to autism would require population-based case and control groups and data collection methods that minimize recall and reporting biases. For evaluation of first baby hair, the concern about lack of standardization in collection procedures can probably only be solved in a prospective study that enrolls baby sibs of ASD-affected and typically developing children. Other concerns about laboratory analyses can be alleviated by use of research rather than commercial laboratories and by assuring that case and control specimens are randomly and blindly analyzed. Research on etiology of autism is, unfortunately, still in very early stages and we don’t yet have sufficient answers to guide prevention and medical treatment. We urge caution in acceptance of the conclusions of this interesting study and join with the authors in calling for additional, rigorously designed, research to replicate the findings.