For distance-related exposure variables, any spatial bias in the recruitment of the study groups (such as a non-random selection of controls with a lower fraction from the distant part) would result in a biased risk estimate.

The SAS procedure SURVEYSELECT was applied for a random re-sampling of controls according to the spatial population distibution.

A bootstrap method was applied to estimate the odds ratios from 800 re-sampled groups.