STS - Sure, but it is a bit complicated. When doing a statistical analysis there are two types of effects you can look at: Main Effects and Interactions. Main effects are straightforward effects of X variable on Y outcome. In R-It, the effect of drug condition (V versus placebo) on MACE outcomes is a main effects analysis. Interaction effects are a bit more complicated, and refer to situations where the effect of X on Y depends on the status of variable Z. In R-It, the interaction that everyone is discussing relates to whether the beneficial effects of V on MACE outcomes depend on primary/secondary prevention status. If an interaction is significant, you then have to see where the interaction is coming from, so you examine the effect of X on Y outcome separately within each category of Z variable. In R-It, this would mean that if the Drug Condition by Primary/Secondary status interaction was significant, you would then have to examine the effects of V on MACE outcomes separately in the primary versus secondary prevention subsamples. These so-called "simple effects analyses" are then interpreted based on the pattern of significant findings within each subgroup. So, hypothetically, if the interaction was significant, and the effect of V on MACE was significant in the secondary prevention subsample but NOT in the primary subsample, you would conclude the the interaction resulted from a significant effect of V on MACE outcomes in the secondary prevention subsample that was absent in the primary prevention subsample. This is how all of the negative commentators are interpreting the R-It data.
In the actual R-It results, the interaction between Drug Condition and primary/secondary prevention was not significant for the most clinically important hard MACE outcome, so there was no obligation to go look at whether effects of V on MACE were significant in the individual primary and secondary prevention subgroups. The lack of significant interaction they reported means the effects of V on MACE were not statistically different between the primary and secondary prevention subsamples, and THAT is the end of the story. V worked for everyone studied in terms of hard MACE outcomes. I think the one stats reviewer at the adcom made her positive comments about approval in both the primary and secondary prevention populations based on this absence of interaction.
Now, with all of that said, I will admit that technically, there was what the investigators considered a barely significant interaction for the soft MACE outcome, with V working better in the secondary prevention than in the primary prevention subgroup. So technically, for the soft 5 point MACE outcome, the commentators are correct in saying that results support V being effective in the secondary but not the primary prevention sample. It is important to note that the sole reason they got a significant interaction for soft MACE was due to their a priori choice to have a VERY lenient criterion for determining significance of this interaction (p<.15). This is atypical in my experience, but I am sure they must have had some reason to be so lenient. I honestly do not know what that reason was though. If they had used the traditional p<.05 level of significance for testing this interaction, it would not have been significant, and again we would conclude that the benefits of V on soft MACE were statistically similar regardless of primary vs. secondary subgroup status. I think what I have described above is the source of some of the confusion on how to interpret the results.
For the soft MACE situation I described, the lack of significance of V effects on soft MACE in the primary prevention group was very much influenced by the relatively smaller sample size for the primary subgroup. They did not have the statistical power to show that the effect they got within this subgroup, which is likely clinically meaningful, was statistically significant. If they had added another 1,000 patients to this subgroup (estimate), the exact same effect they actually reported would have been significant and we would not be having the conversations that we have been having.
My take on all of this is to pay more attention to the most clinically important hard MACE outcome, which found that primary vs. secondary prevention did not make any difference in the benefits of V. We will see if the FDA agrees, but my hunch is that they will. I think we get the R-It entry criteria as the most likely label scenario, with the best case scenario being a label that includes patients with "existing CV disease or with multiple risk factors for CV disease." I think statin use and having TGs >150 will also be on any label we get.
News 
Market Data 
Discover 
AI
