The effect of migration on comparison of disease rates in geographic studies in the United States.

Cancer rates are often compared between counties or other geographic units as a method of testing for risk from environmental exposures. Migration between geographic areas greatly reduces the sensitivity of this method. Under simplifying assumptions the quantitative effect of migration on risk estimates is shown using migration and cancer incidence data for the United States. For example, 40--50% of the relative excess risk, defined as the relative risk minus one, is not reflected in the estimated risk for most cancers, when rates are compared between exposed and unexposed counties and migration has taken place during a 30-year latent period. More extreme losses of sensitivity also occur. Under the simplifying assumptions, the quantitative effect of migration on risk estimates is shown as a function of cancer site, latent period, and the type of geographic units for which rates are calculated--states, counties, or places. Also discussed are some implications of these findings for geographically-based studies and additional data needs for assessing the effect of migration.