ORIGINAL ARTICLE Cancer incidence and soil arsenic exposure in a historical gold mining area in Victoria, Australia: A geospatial analysis Dora Claire Pearce 1,2 , Kim Dowling 1 and Malcolm Ross Sim 3 Soil and mine waste around historical gold mining sites may have elevated arsenic concentrations. Recent evidence suggests some systemic arsenic absorption by residents in the goldfields region of Victoria, Australia. Victorian Cancer Registry and geochemical data were accessed for an ecological geographical correlation study, 1984--2003. Spatial empirical Bayes smoothing was applied when estimating standardised incidence ratios (SIRs) for cancers in 61 statistical local areas. The derived soil arsenic exposure metric ranged from 1.4 to 1857 mg/kg. Spatial autoregressive modelling detected increases in smoothed SIRs for all cancers of 0.05 (95% confidence interval (CI), 0.02--0.08) and 0.04 (0.01--0.07) per 2.7-fold increase in the natural log-transformed exposure metric for males and females, respectively, in more socioeconomically disadvantaged areas; for melanoma in males (0.05 (0.01--0.08) adjusted for disadvantage) and females (0.05 (0.02--0.09) in disadvantaged areas). Excess risks were estimated for all cancers (relative risk 1.21 (95% CI, 1.15--1.27) and 1.08 (1.03--1.14)), and melanoma (1.52 (1.25--1.85) and 1.29 (1.08--1.55)), for males and females, respectively, in disadvantaged areas in the highest quintile of the exposure metric relative to the lowest. Our findings suggest small but significant increases in past cancer risk associated with increasing soil arsenic in socioeconomically disadvantaged areas and demonstrate the robustness of this geospatial approach. Journal of Exposure Science and Environmental Epidemiology (2012) 22, 248--257; doi:10.1038/jes.2012.15; published online 21 March 2012 Keywords: cancer; soil; arsenic; exposure; geospatial INTRODUCTION Arsenic exposure through contaminated drinking water has been linked to cancers including bladder, colon, kidney, liver, lung and skin. 1,2 The mode of action of arsenic as a multiorgan human carcinogen has not been fully elucidated, 3 although inhibition of DNA repair mechanisms may enable it to act as a cocarcinogen. 4 The soil environment has also been identified as a contributor to arsenic exposure. 5-7 Typically associated with gold mineralisation, residual arsenic contamination from historical gold mining activity persists in the goldfields region of Victoria, Australia, where elevated arsenic concentrations have been observed in mine waste and some residential soils, surface and ground waters. 7 - 10 Socioeconomic status may bias ecological studies not only because of poorer health outcomes but also because lower socioeconomic status increases the chance of living closer to sources of environmental pollution. 11,12 Further, exposure mis- classification due to crude exposure metrics may undermine the likelihood of identifying a true relationship between an environ- mental hazard and disease outcomes. 13,14 In a previous aggregation of data from 22 geographical regions irregularly distributed across Victoria, each with evidence of elevated environmental arsenic concentrations or visible mine tailings, small but significantly increased risk of all cancers combined, prostate and breast cancers, melanoma and chronic myeloid leukaemia were identified. 9 However, potential limitations of this study may have arisen from a lack of adjustment for socioeconomic status, low statistical power due to sparse cancer data in some rural areas over the 10-year period 1982--1991 and a reliance on categorisation of environmental arsenic contamination to determine exposure classification in these mainly disconnected regions. Further investigation was therefore warranted to confirm associations observed with melanoma and leukaemia, not previously linked to arsenic exposure, prostate cancer, typically associated with arsenic-contaminated drinking water, 15 and lack of expected associations with lung and bladder cancers. We aimed to investigate associations between arsenic in soil and mine waste and a priori selected cancers in the goldfields region of Victoria at a refined spatial resolution using an improved exposure metric, taking into account socioeconomic disadvantage (SED). Further objectives were to: (1) assess the validity of this approach by comparing associations detected with those previously observed in Victoria; and (2) evaluate the robustness of associations detected with derived exposure metrics using spatial autoregressive modelling. MATERIALS AND METHODS Study Design This small-area ecological study investigated the incidence of all cancers combined and a priori selected cancers, and their association with two derived soil arsenic exposure metrics. Underpinning this analysis was the availability of 20 years of Victorian Cancer Registry (VCR) data, 1984 - 2003, georeferenced to statistical local areas (SLAs), and geochemical data from the University of Ballarat and GeoScience Victoria comprising soil arsenic concentrations with geospatial coordinates of sampling sites, facilitating derivation of exposure metrics. We used the Geocentric Datum of Australia (GDA94) as the spatial coordinate system, map projection VICGRID94, and Received 28 July 2011; accepted 19 December 2011; published online 21 March 2012 1 University of Ballarat, Mt Helen, Victoria, Australia; 2 Melbourne School of Population Health, The University of Melbourne, Melbourne, Victoria, Australia; 3 Monash University, Melbourne, Victoria, Australia. Correspondence to: Dr. Dora C. Pearce, Melbourne School of Population Health, The University of Melbourne, Level 3, 207 Bouverie Street, Melbourne, VIC 3010, Australia. Tel.: þ 61 3 9035 3343. Fax: þ 61 3 9349 5815. E-mail: dpearce@unimelb.edu.au Journal of Exposure Science and Environmental Epidemiology (2012) 22, 248 - 257 & 2012 Macmillan Publishers Limited All rights reserved 1559-0631/12 www.nature.com/jes