Advera Clinical Evidence - Statistical Measures

Summary

Our current statistical measures use the Cochran-Mantel-Haenszel1,2 methods (CMH) in order to provide estimates of fixed effect for dichotomous outcomes (AE vs. Not-AE) between Control & treatment groups.

The formula for the CMH Odds Ratio is given below:
​

Suitability of the CMH Measures for Pooled Analyses

From the Cochrane Handbook for Systematic Reviews3 - considered to be the authoritative guide on comprehensive meta-analysis reviews of medical interventions:

"The Mantel-Haenszel methods (Mantel 1959, Greenland 1985) are the default fixed-effect methods of meta-analysis...When data are sparse, either in terms of event rates being low or study size being small, the estimates of the standard errors of the effect estimates that are used in the inverse variance methods may be poor. Mantel-Haenszel methods use a different weighting scheme that depends upon which effect measure (e.g. risk ratio, odds ratio, risk difference) is being used. They have been shown to have better statistical properties when there are few events. As this is a common situation in Cochrane reviews, the Mantel-Haenszel method is generally preferable to the inverse variance (I^2) method. In other situations, the two methods give similar estimates."

Caveats & Notes

Our calculations for the CMH OR use "continuity correction" - that is for any cell values in the 2x2 table for a trial which are 0 - then these are automatically changed to 0.5. This is a standard practice and is used in order to prevent division by 0 scenarios.

In cases where cell counts are very low (0 or 1), or where there is great variance in the directionality of the measure between studies in the pool- results can be less reliable and misleading. This is where a different model (Random effects) might be better.

References