A Mantel-Haenszel statistic for testing the association between a polychotomous exposure and a rare outcome

The Mantel-Haenszel statistic has been designed to test the association between two dichotomous variables using information from several 2 x 2 tables. This statistic is extended here for incidence data on a rare outcome arising at two or more levels of an exposure variable. When there are only two levels of exposure the statistic is reduced to the one suggested earlier by Miettinen. The relationship between degree of urbanization and female laryngeal cancer is studied as an example.