based on person years of exposure. So if there is differential exposure between treatment
groups, the computation of this estimate adjusts for that difference. With low event rates
and small differences in exposure, this estimate will not differ from a risk ratio computed
as the ratio of incidences. For Saxagliptin, we thought this estimate was important for the
short-term data where dropouts were generally significantly greater in the placebo group
than the Saxagliptin group, and not as important for the short-term plus long-term period,
although we will show estimates from both of the time periods.
computed stratifying on study. I will explain what each of these offers to the
interpretation of the CV results. For the risk difference, studies with no events are
included in the analysis. This contrasts with methods for computing odds ratios where
studies with no events are only included if a continuity correction is used. With the risk
difference, we are able to assess the influence of the no event trials on the interpretation
of risk when we look at this estimate against an odds ratio.
set by the guidance was the odds ratio based on an exact test. Our summaries all contain
these estimates. The confidence interval for this estimate tends to be conservative
compared to intervals computed by other methods and therefore I think an attractive
approach for assessing CV risk.
The incidence rate ratio, that is the second one listed here, is computed
Now the estimate we use for assessing risk in the context of the boundary
Scribes, LLC Toll Free 1-800-675-8846 www.scribesllc.com
Lastly, we thought it was important to analyze the results with more than one statistical
method. From our experience with Avandia and other drugs, we knew that the method
used could impact the results particularly with rare events and with multiple studies with
Now here are the estimates we considered. All these estimates are