Intra-individual variability in toenail arsenic concentrations in a Michigan population, USA

Arsenic concentration in toenail clippings is used as a biomarker of exposure in epidemiological studies, often under the assumption that a single measurement represents long-term exposure. For this assumption to hold, the measured arsenic concentrations must be stable over time, yet temporal variability has not been adequately assessed. This study aims to evaluate temporal variability in multiple toenail samples collected from a population exposed to drinking water arsenic levels <100 microg/l. Our objectives are to investigate factors responsible for biomarker variability and to assess the suitability of single versus multiple measurements for determining exposure in epidemiological studies. Multiple toenail and drinking water samples were collected from 254 participants enrolled in a case-control study of arsenic exposure and bladder cancer in Michigan, USA; participants also answered questions on water consumption. Toenail samples collected an average of 14 months apart were positively correlated, although a substantial amount of variability was detected (r=0.43, P<0.0001, n=236). Arsenic concentration in drinking water was stable and small changes in drinking water arsenic concentration did not explain variability in toenail arsenic concentration. Change in drinking water consumption, however, was significant in predicting differences in toenail arsenic concentration. Stronger correlations between drinking water arsenic concentration and intake and toenail arsenic concentration were observed when two toenail samples were averaged, suggesting that multiple measurements may more accurately reflect exposure. When exposure was categorized into tertiles and other pre-determined categories, 25-40% of exposures were differentially classified. Only a small percentage (<4%), however, were classified as having low exposure using a single measurement and high exposure when an average of two measurements was used. These results suggest that the use of multiple measurements is unlikely to affect exposure classification of individuals into high- or low-exposure groups; however, collection of multiple samples may be advantageous for more refined exposure classification.