Although adjustment factors could have been applied to avoid undefined OR, for an extra layer of
conservatism, estimability was subject to a requirement of at least 1 observation per cell of the contingency table
(Argresti 1996). Evaluation of the quantitative trend between 1,4-dioxane and TCE was restricted to only those installations
where significant association could be definitively determined. Conditional association was tested by x2 using Proc Freq in
SAS Version 9.2 for Windows (SAS Institute 2008) and considered significant at p0.05