What do you think of the state of surrogate endpoint validation in chronic kidney disease?

I’ve seen a number of papers now including this recent Nature Medicine paper purporting to validate candidate surrogate endpoints for chronic kidney disease (CKD).

All things considered, it seems like a sound meta-analysis of the correlation between putative surrogate and endpoint. However, we in this community all know that “a correlate does not a surrogate make” and all that. I don’t see anyone in CKD paying attention to everything we know about best practices to evaluate and validate surrogate endpoints.
[Disclaimer: There is much more work out there beyond what is covered in the review. I just offer this as a pointer to this large body of work.]

Would be curious to understand what the community here thinks, particularly clinicians like @Pavlos_Msaouel who understand the challenges of surrogate endpoint validation, or those with expertise in kidney disease.

3 Likes

Agree, the sooner we evolve beyond simple correlation the better. Cannot recommend enough the Elliott review you linked. Truly an excellent resource.

I have heard horror stories from statisticians trying to highlight this 10-20 years ago and finding zero traction in oncology journals. But last year during the annual meeting of the Kidney Cancer Association we included a think tank that essentially discussed this thread. Was very pleasantly surprised by how well it was received by clinicians and patient advocates with lots of engagement and great input. This is being drafted into a consensus statement manuscript to be submitted for peer reviewed in the coming weeks. Will link to it once it is out.

There are a lot of challenges to overcome and we will learn tons from practical usage in the coming years. But this appears to be a case where ideas took time but now are coming into the spotlight. Hopefully the momentum will not fizzle away.

3 Likes

Many thanks for pointing to this review paper :+1: @mnarayan and to the thread @Pavlos_Msaouel

Adding here on the specific clinical topic that the European Medicines Agency (EMA) recently issued an Opinion that “[…] qualifies GFR slope as a validated surrogate endpoint for CKD progression in randomised controlled clinical trials to support marketing authorisation and extension of indication approvals when the use of clinical endpoints is not feasible due to rarity of the disease or need for a very long study period.” (2023-12-21).

Details under the link below, including parts which I trust the community here finds interesting:

  • the Qualification opinion (see e.g. p 5 on treatment effects and trial-level surrogacy and p 7 for the conclusion cited above)
  • comments received during the public consultation on a draft opinion in 2023
  • responses to questions / issues identified during EMA’s scientific assessment of the proposal
  • the briefing book for the proposal (see e.g. p 36 and following for statistical methods, section “7.2 Trial-level analysis relating treatment effects on the clinical endpoint to treatment effects on GFR slope”)

The qualification was proposed the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) and National Kidney Foundation (NKF).

https://www.ema.europa.eu/en/human-regulatory-overview/research-development/scientific-advice-protocol-assistance/opinions-letters-support-qualification-novel-methodologies-medicine-development#gfr-slope-as-a-validated-surrogate-endpoint-for-rct-in-ckd-32672

(Disclaimer: Am working at the EMA to support, amongst others, pre-competitive R&D for regulatory science tools. Separately, am maintaining a collection, https://paediatricdata.eu/regscitools)

2 Likes

Hard to believe that a measure assuming linearity in the time-response profile is accepted as an outcome.

1 Like

This issue was raised and is discussed in several places in the documents (incl. briefing document, list of issues); excerpt from the conclusion: “Qualification of GFR slope (total or chronic) as a validated surrogate endpoint for CKD progression is for population-level analysis. Individual predictions of kidney function are not included in the Context of Use” and “Importantly, the study design should take into consideration the trial duration and frequency of GFR assessment for reliable assessment of the linearity of the slope.”

For such new endpoints, trial sponsors typically interact with regulators on planning the design of a specific study.

2 Likes

Thanks for these pointers @Ralf_Herold. These are some developments that I didn’t know about. Also great to know about your work on regulatory science tools!

2 Likes

That is nice to know that norms are changing and pockets of biomedical communities are open to such input.

2 Likes