Some new thoughts about extrapolation from JELIS
I’ve posted several times before about what I believed JELIS had to say about the prospects for RI. Recapping, it can be argued that the JELIS 19% RRR obtained with a 1.8g dose suggests RRR around 40% for RI with a 4g dose. Further adjustment for other differences between the two trials (LDL and background EPA levels, statin dose, placebo rate etc) seemed to suggest RI RRR even higher than that. (JL pioneered using JELIS to predict highly successful RI on the message board, several years ago.) So, while very pleased with the wonderful reported RI results, I have been puzzled why they weren’t even better. The CV death RRR substantially lower than for other types of CV events also seemed puzzling.
I now have a theory about what happened. The curves for the primary endpoint didn’t really start separating until after about a year, and the active arm didn’t really bend down until almost two years. Unlike JELIS, RI enrollment was spread out over a period of years, so a lot of the subjects had a relatively short time in the trial. If the first 18 months or so are just priming the pump, and the real benefits don’t start until 18 or 20 months in, then the actual RRR that would be attained in the long run would be substantially higher than what was reported for RI. For a quick ballpark estimate, with RI median follow-up about 20 quarters and with a startup period of 6 quarters, that suggests long-term RRR of 25% x (20)/(20-6) = 36%.
It could be that V is slowing the accumulation of plaque. The startup period could be because time is needed for the corresponding plaque in the placebo arm to block an artery. The amount of time could depend on the diameter of the artery, with larger arteries taking longer. Blockage of larger arteries seem more likely to be fatal, so for CV death the startup period could be longer than for non-fatal events. If so, then better CV death RRR might be observed by considering only post-startup patient years, allowing a substantial startup period for CV death. I think the later patient years will more accurately represent long-term results, and will show higher RRR. The trial is overpowered enough, omitting the startup period should still allow stat sig.
I am hopeful of even stronger results in future publications based on this kind of analysis. On top of that, subjects with full compliance and high plasma EPA might do even better. And, there may be convincing evidence that adding a gram or two to the dose could provide additional benefit for some subjects (especially older Western Males, as mentioned in an earlier post).
Lots of reasons for optimism IMO.
I'll be traveling with limited ability to discuss today, but welcome comments.
