Jonzobot:

Agreed that I oversimplified the "all benefit - no benefit" separation between  the eGFR-based CKD and nonCKD subgroups.  The nonCDK subgroup did derive some benefit from apabetalone but it was small, a 6% improvement over control (the HR for the group was 0.94 according to slide 19).  But the high LDL group derived zero mean benefit from apabetalone (the HR was greater than 1). So here is how I did the arithmetic from the two analyses:

In the eGFR analysis the total MACE benefit averaged 50% in 10% of the trial population and 10% in 90% of the trial population:(0.5 x 0.1) + (0.06 x 0.9) = 0.104

In the LDL analysis the total MACE benefit averaged 40% in 50% of the trial population and 0% in 50% of the trial population:(0.4 x 0.5) + (0 x 0.5) = 0.20

These are the numbers I could not reconcile. But now I see the error of my ways.  The analysis doesn't depend on the proportion of patients in each group, so much as it depends on the proportion of MACE in each group. From slide 19 the total number of MACE is 274.  Since the small CKD group of patients accounted for more than 10% of the MACE (48 or 17.5% in apabetalone and control combined), and the larger low LDL group accounted for fewer than 50% (126 or 46% in apabetalone and control combined) of the MACE events in the study, the above equations are wrong.  Too tired to correct them tonight.  But I see that the two analyses of the MACE data are likely not discrepant as I initially considered them to be. 

Jupe