As previously described, PBDEs are widespread in household products and have now moved into the food chain , although little research has addressed human health effects. The lack of association between children's circulating levels of PBDEs and autism case status does not preclude a role for PBDEs in autism etiology. A weakness in this pilot study was the examination of current levels of PBDEs as a proxy for exposures that preceded the neuropathologic changes leading to autism. In retrospective research such as the CHARGE Study and other case-control investigations, obtaining etiologically relevant exposure measurements is challenging. Half-lives of PBDEs vary. Higher brominated compounds appear to have the shortest half-lives, e.g., 15 days for #209, 39 days for #207, and 94 days for #183, based on a study involving occupational exposures  and therefore, correlations of internal measurements over a period of years would not be expected to be high. Half-lives of lower brominated congeners were estimated based on daily intake and total body burden as 1.8, 2.9, 1.6, and 6.5 years for BDE-47, -99, -100, and -153, respectively . Nevertheless, even for those compounds that are retained, correlations over time can vary widely and will be a function primarily of the introduction of new sources, elimination of old sources, and changes in behaviors that influence extent of contact with those sources. The correlation between current and past exposure will be greatest when the sources are constant over time for an individual, or when external sources have been eliminated and only past exposures are present, with gradual ongoing excretion. Exposures were rising in the U.S. between the births and the time of data collection for these subjects, as a result of continued production of household furnishings and electronics with PBDEs, and increasing levels in the food supply, and therefore, it is unlikely participants' exposures remained steady . We also note that the concentrations in this sample are, to our knowledge, higher than previously reported in any other population, even from the 2003-2004 NHANES sample , and are seven to ten times higher than those reported to predict lower cognitive scores in a sample of children from New York City .
During the prenatal period, the mother's exposures will be critical. Her exposures will be partially based on diet, and partially based on pathways related to the home environment, including inhalation of both chemicals in the gas phase and resuspended particulate matter, as well as dermal exposure and non-dietary ingestion of dust. During early infancy, exposures, particularly for lower brominated congeners  appear to be primarily through breast milk , again reflective of the mother's exposure, or infant formula, as well as some contribution from inhalation of air and dust, possibly from mattresses, upholstered furniture, and cushions. With the introduction of solid food, the diet becomes more varied, with new sources of PBDE exposures, even though some toddlers choose to eat a limited number of food items. As children start to interact more independently with their environment toward the second half of the first year, intake becomes increasingly influenced by what is present in the home environment, in large part due to increases in frequency of contact with surfaces, hand to mouth activity, and object to mouth activity [37, 38]. Higher concentrations have been found in toddlers than in school-age children or adults, supporting a critical role for non-food ingestion during the early years . All of these factors underscore the differences between exposures measured at 24 to 60 months of age versus those that occurred during a critical time period, which may very well be the prenatal period. The issue is further complicated when comparing children with developmental disorders to those with typical development: the former may have different timing or frequency in their hand to mouth activity, delays in explorative play, or a more limited diet due to feeding difficulties [39–41]. Similar to what others have reported, in our study sample, both food and non-food sources appeared to contribute to the children's body burdens .
Given the uncertainties outlined above regarding exposure and intake, particularly for infants and toddlers, and the dramatic changes between birth and age two-five years in diet and play behaviors, current PBDE concentrations (even when half-lives are long, e.g., 4 years) may not adequately estimate exposures in the prenatal or early postnatal period. House dust or banked specimens from adults (whose diet and behaviors change less than their children's) might provide better estimates of PBDE exposures during the critical time windows. Breakdown of the deca-BDE (209) may occur in house dust, but for other congeners, stability of levels in house dust may be an indication that it serves as a more valid repository of long-term exposures than blood plasma, at least in young children .
Although few risk factors have been established for autism, we evaluated more than ten potential confounders that had no appreciable effect on the coefficients relating PBDEs to diagnosis. Similar to the findings at an older age , during the first two years of life, children with autism were less likely to eat ocean fish (primarily tuna), but in this small sample, the differences were neither large nor statistically precise. As Hg was not associated with case status and showed low correlations with PBDEs, it could not have confounded the analyses of PBDEs and AU/ASD or developmental delay. Confounding by known risk factors or other sociodemographic characteristics seems unlikely to have been responsible for the null results of this study.
Another consideration in the interpretation of these null findings is that the levels of PBDEs may be sufficiently high that all children who are predisposed to develop autism have had exposures sufficient to surpass their individual thresholds. It has previously been noted  that when exposure is high and ubiquitous, no study will be able to detect the effect of that exposure; within such populations, diseases will instead appear to have only genetic influences. Despite the speculative nature of this proposition in the case of autism and PBDEs, the problem of widespread environmental chemical pollution poses a serious obstacle to identifying health effects. Even their interaction with genes will be elusive and research could appear to support a purely genetic etiology either when the exposure variability is low, or when levels are so high as to exceed the threshold in all susceptible individuals.
The lower concentration of BDE #197 in children with developmental delay is most likely a chance finding, given the small sample size in this pilot study and the large number of comparisons made. Children with developmental delay often have more, rather than less, hand-to-mouth activity, thus it is unclear whether behaviors could explain this particular association. For the lower brominated congeners, the more extreme values in some children from the DD group is interesting, and might reflect different behavioral patterns in children with cognitive and adaptive impairments.
Thus, although our current measurements of PBDEs do not predict risk of autism, this pilot project should not be construed as the last word on PBDEs and autism. On the one hand, our null findings could have been the consequence of substantial misclassification relative to exposure in the etiologically critical time windows: measurements in plasma samples collected at least a year and potentially as much as five years after the key events in neurodevelopment may have been poor approximations of the relevant prenatal or early postnatal exposures. On the other, the biological plausibility of a causal association between prenatal or early postnatal PBDE exposures and the development of autism is supported by numerous documented toxicologic mechanisms relevant to CNS development.
Further research can be directed towards several issues: a better understanding of sources and pathways of exposure for pregnant women and young children; design of methods to obtain more accurate measurements of exposure, especially to hydroxylated metabolites, during periods critical for autism etiology; analysis of larger sample sizes; and expansion of animal and in vitro models of neurodevelopmental toxicity to define the molecular mechanisms. Of potential relevance are Ca2+ signaling pathways, endocrine disruption, or other cellular, intracellular and transport processes, and their regulation by genes that may have been associated with autism.