Statistics Glossary → Breslow-Day test (homogeneity of odds ratios)
Breslow-Day test (homogeneity of odds ratios)concept
A test, for 2 x 2 tables split by the strata of a confounding variable, of whether the odds ratio is the same in every stratum.
The Cochran-Mantel-Haenszel method combines 2 x 2 tables split into strata (age groups, centres) into a common odds ratio adjusted for the confounder. That common ratio only makes sense if the odds ratio is roughly the same in each stratum. The Breslow-Day test compares the observed counts in each stratum with those expected under the common Mantel-Haenszel odds ratio; the null hypothesis is 'the odds ratios are homogeneous across strata', and the statistic approximately follows a chi-square distribution with the number of strata minus 1 degrees of freedom. Tarone's correction refines the test slightly.
If the test is significant, the strength of the association varies by stratum (effect modification or interaction), and stratum-specific odds ratios should be reported instead of a single common one. With many strata or small counts within strata the test has little power, so a non-significant result should not be taken as proof of homogeneity.
An everyday example
A study examines the link between smoking and chronic cough separately in three age groups. If the odds ratios are 2.1 in the young, 2.3 in the middle-aged and 2.0 in the elderly, the Breslow-Day test does not reject homogeneity and a common ratio of about 2.1 is meaningful. If they were 1.0, 2.5 and 6.0, the test would be significant and the conclusion that the effect of smoking depends on age would come to the fore.
In MerQur
In the Cochran-Mantel-Haenszel analysis, the Breslow-Day homogeneity test and the Tarone test are reported alongside the Mantel-Haenszel common odds ratio; uninformative (zero-margin) strata are removed for these two tests.
Analyses where this concept appears
Cochran-Mantel-Haenszel (CMH) Test
Related concepts
Mantel-Haenszel common odds ratioStratified analysis (stratum variable)