Adsorption, oxidation, and bioaccessibility of As(III) in soils.

At As-contaminated sites, where the ingestion of soil by children is typically the critical human-health exposure pathway, information on the bioavailability of soil-bound As is often limited. The influence of various soil physical and chemical properties (iron and manganese oxides, pH, cation exchange capacity, total inorganic and organic carbon, and particle size) on As(III) adsorption, sequestration, bioaccessibility (as a surrogate for oral bioavailability), and oxidation was investigated in 36 well-characterized soils by use of a physiologically based extraction test (PBET). These results were compared to an earlier published study with As(V) on the same set of soils. The properties of the soils were able to explain >80% of the variability in the adsorption and sequestration (as measured by the reduction in bioaccessibility over time) of As(III) in these soils. The initial bioaccessibility of As(III) was significantly higher than the initial bioaccessibility of As(V) on the same set of soils. However, over a 6-month period of aerobic aging, a significant portion of the solid-phase As(III) on these soils was oxidized to As(V), decreasing its bioaccessibility markedly. A multivariable linear regression model previously developed to predict the steady-state bioaccessibility of As(V) in soils was able to predict the bioaccessibility in As(III)-spiked soils within a root-mean-square error (RMSE) of 16.8%. Generally, soils having a higher iron oxide content and lower soil pH exhibited lower bioaccessibility. This model was also able to predict the in vivo bioavailability of As in contaminated soils previously used in an independent juvenile swine dosing trial within an RMSE of 15.5%, providing a greatly improved yet conservative estimate of bioavailability relative to the typical default assumption of 100%. However, the model was not able to accurately predict the bioavailability of As in a different set of contaminated soils previously used in an independent Cebus monkey dosing trial, consistently overpredicting the bioavailability, resulting in an RMSE of 42.7%. This model can be used to provide an initial estimate of As bioavailability in soil to aid in screening sites and justifying expensive site-specific animal feeding studies. Further, as the model is based on major soil properties, the resulting estimates are valid as long as the major soil properties do not change, thus providing some confidence in the long-term applicability of the estimates.