We have conducted an experiment with 2 groups and a dichotomous outcome on 3 occasions and for each occasion we have a 2x2 contingency table.
Rather than pooling the 3 contingency table we want to determine if we have a consistent difference in proportions across the repeats.

When we analyze the data in SPSS we get a MH statistic of 13.19, df 1, p=0.00 and further to that we get an estimate of the common odds ratio with confidence intervals (0.23). The Breslow-Day test returns a p value of 0.298. Based on this I think we can conclude that the difference in proportion we see in the experiments is not confounded by the different occasions we have conducted the experiment. Am I correct?

However, for another study entirely with a similar design (resulting in three tables (2x2)) we have also used Mantel-Haenzsel statistics. On this occasion we have MH statistic of 22.8, df 1, p=0.00 and a common odds ratio of 0.143. However, the Breslow-Day test returns a significant p value (0.046). How does this alter the interpretation?

I'd really appreciate some help with the interpretation
Thanks in advance