Research reading group · call recap
The kidney paper and what a composite endpoint hides
19 October 202513 attendedHeld by Sunil
FLOW — Perkovic V, et al. New England Journal of Medicine 2024;391(2):109 to 121. We had covered it jointly with the type 2 circle in the spring; this was the proper technical read, requested by four members who wanted to go at the endpoint itself. It was one of the better arguments this group has had and I changed my mind about one thing in it.
What the paper is
Semaglutide against placebo in people with type 2 diabetes and chronic kidney disease, stopped early for efficacy on a composite primary outcome. The composite bundles kidney failure events, a substantial sustained fall in kidney function, and death from kidney or cardiovascular causes.
Stopping early for efficacy is itself worth understanding, and we spent time on it. It is not a scandal and it is not free: an early stop tends to produce a larger apparent effect than a trial run to completion, and it truncates what you can say about the longer term.
Where the argument was
Whether a composite endpoint of that kind is informative. My position going in was that composites are frequently used to manufacture significance out of soft components. Nils took it apart, fairly: in kidney disease the components are hard, clinically serious and conventionally bundled, and the individual components moved in the same direction rather than one carrying the whole result. That is a much stronger composite than the ones I had in mind.
I said on the call that I had been importing a scepticism from cardiology trials where it does not fit. Two members said afterwards that a facilitator changing position out loud was the most useful thing in the hour, which is a slightly humbling piece of feedback.
What the room concluded
That the result is meaningful in the population studied, that early stopping means the size of the effect should be held loosely, and that the popular version — good for kidneys — is doing a lot of work that the paper does not do. Three members with kidney involvement left with a question about whether the trial population resembles them, which is the only question that matters at an individual level.