Conclusions
Historical water quality databases may be useful in epidemiologic studies that categorize participants by relative levels of tap water nitrate in order to assess the association of this exposure with health outcomes. Such records-based approaches must be applied carefully to avoid introducing bias in ORs or other measures of relative risk. In addition, results must be interpreted with care so that studies that fail to observe an association are not overstated. The use of such methods for arsenic may be limited at present.