OBJECTIVE:To examine the analytic role of arsenic exposure on cancer mortality among the low-dose (well water arsenic level <150 μg/L) villages in the Blackfoot-disease (BFD) endemic area of southwest Taiwan and with respect to the southwest regional data. METHOD:Poisson analyses of the bladder and lung cancer deaths with respect to arsenic exposure (μg/kg/day) for the low-dose (<150 μg/L) villages with exposure defined by the village median, mean, or maximum and with or without regional data. RESULTS:Use of the village median well water arsenic level as the exposure metric introduced misclassification bias by including villages with levels >500 μg/L, but use of the village mean or the maximum did not. Poisson analyses using mean or maximum arsenic levels showed significant negative cancer slope factors for models of bladder cancers and of bladder and lung cancers combined. Inclusion of the southwest Taiwan regional data did not change the findings when the model contained an explanatory variable for non-arsenic differences. A positive slope could only be generated by including the comparison population as a separate data point with the assumption of zero arsenic exposure from drinking water and eliminating the variable for non-arsenic risk factors. CONCLUSION:The cancer rates are higher among the low-dose (<150 μg/L) villages in the BFD area than in the southwest Taiwan region. However, among the low-dose villages in the BFD area, cancer risks suggest a negative association with well water arsenic levels. Positive differences from regional data seem attributable to non-arsenic ecological factors.