Even if such factors are assessed, and the most precise summary measures are employed, studies using the approaches examined here would need to be powered and interpreted in light of the likely effect of an important degree of non-differential measurement error (i.e. noticeably attenuated ORs). Furthermore, water quality monitoring databases may not be suitable for estimating absolute levels of exposure to tap water contaminants. As evidenced by some samples being conservatively reported (i.e. as upper bounds), it may be that these types of records would tend to overestimate absolute contaminant levels. In general this would not be a major limitation for association studies. Precision and "average measurement error" are independent [16]. However, care must be taken in combining these estimates with other sources of exposure (e.g. diet), and in interpreting "cut-points" for categories (e.g. quartiles) as being meaningful.