Exemplary QOL analyses that avoid change-from-baseline blunders?

So our Citizen Petition has been submitted to the FDA on need and rationale for
supplementary comparison of Quality of Life-related patient reported outcome (QoL-PROs)* for surrogate endpoints based on tumor assessments.

full text: http://www.lymphomation.org/Citizen%20Petition-QoL-feb2022-fin.pdf

We waited for Congress to approve a commissioner and avoided advancing methods (thanks to guidance provide here) for capturing, comparing, or reporting QoL-PROs.

Opening statement:

To amend the Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics Guidance for the Industry to require or strongly urge supplementary comparison of Quality of Life-related patient reported outcome (QoL-PROs) for the following surrogate endpoints in randomized controlled clinical trials:*

Endpoints Based on Tumor Assessments

Disease-Free Survival (and Event-Free Survival) | Objective Response Rate
Complete Response | Time to Progression and Progression-Free Survival (PFS) | Time to Treatment Failure

What I hope to see is a universal set of QoL-PRO in key domains applied across all studies and reported in a way that promotes physician and patient understanding.
We anticipate this change in industry registration trials would provide a good many benefits (see text of petition) and would be an aid to regulatory decision making in close calls:

3 scenarios:
Major improvement in time to relapse with modest impairment in QoL.
(Approval could still be justified)

Modest improvement in time to relapse with impairment of QoL
(Longer followup justified)

Modest improvement in tumor response with improvement in QoL
(Approval could still be justified)

Suggestions on models, instruments, analysis, and reporting in plain language most welcome!

1 Like

I’m glad you are pushing this. A key analytical challenge is that you can’t really separate QOL from mortality, since mortality blocks observation of QOL. So I don’t recommend separation of clinical and QOL outcomes but rather a single comprehensive ordinal analysis. Ordinal longitudinal analysis as motivated here provides clear estimands. For example one can compute as a function of time and treatment the probabiity that a patient has QOL level x or worse, where “or worse” includes both a worse QOL scale, recurrence, or death. From the same model one can estimate P(recurrence or death by time t), P(death by time t), P(recurrence but alive at time t). Extensive material related to this may be found at https:/hbiostat.org/proj/covid19.

4 Likes

Thank you, Dr. Harrell!. (Pardon my typo!) If the FDA acts on this (no expectations) I assume that they will convene experts and advocates to select the appropriate instrument(s) and analytics. My part is done!

Having said that I am surprised that mortality would be a problematic issue in trials that use surrogate endpoints as the primary endpoint - my impression being that surrogates are used because differences in survival are not anticipated during the course of the study and the defined followup.

1 Like