Designing Trials to Consider the Full Spectrum of Patient Outcomes

ACT: How is clinical trial design and analysis changing in response to more comprehensive evaluation methods (methods that consider the full spectrum of patient outcomes)?

Buyse: I'm a statistician, so this is the type of thing that I do for a living: designing clinical trials. And it's true that, in fact, this methodology which combines benefits and risks, or which combines several efficacy outcomes in a single analysis, actually creates a lot of problems. It's much more complicated to do the design of a trial if you have a multivariate outcome than if you have a single outcome, so it does complicate our life. No question that statisticians will have a lot more difficulties designing trials simply because of the complexity of the outcomes that we look at. Instead of having a single one, we now have a range of outcomes. Very often we use what we call prioritized outcomes, so we have, for example, three or four outcomes that matter to patients. We may have duration of life as the first, we may have the time to progressive disease as the second, we may have the achievement of tumor response as the third. And we may have, for example, severe grade-three or four toxicities as the fourth outcome. So that means when we design the trial, we have to figure out the effect of the new treatment on each and every one of these four outcomes, and we have to account for the correlation of these outcomes; there is an association statistically between these outcomes. They are not independent of each other. For example, to come back to what I said before, a patient who has toxicity, very often is a patient who has a response to treatment as well. So the two are kind of correlated, and so all of this has to be taken into account when we design trials. So yes, it will make the life of statisticians more challenging, but also more interesting, because the end-product, the design of the trial itself, will probably be better suited to address the question of interest, both from a clinical point of view, but also from a patient-centric point of view. And so I think it's additional work that is well worth embarking on. And accepting the fact that the paradigm changes and makes things more complex but also much more interesting.

Coppe: If I may just add a comment, which is really important to connect the dots between the statistical analysis and what clinicians and patients will take out of the clinical trial. How can you interpret those kinds of statistical analyses Marc was mentioning? If you just do what you call statistical analysis, you just assess one outcome at a time, and at the end of the day, it's always hard to balance two outcomes and see which one is the most important for a clinician and for a patient. You may do it correctly from a statistical standpoint, but it's not always easy to understand and to share it later on. It's also really important to be able to offer a robust statistical analysis, but a statistical analysis that offers to listen to patients. Which of those kinds of patient and relevant outcomes do you want to consider in this, in this single statistical analysis, so that you can cooperate those outcomes together. The second very important aspect is being able to interpret correctly, the results of the statistical method and being able to tell the patients and clinicians what it means. It means that globally, the next treatment benefit is more positive based on the different outcomes you defined prior for this treatment of this one. So, that's really making this statistical methodology impactful, robust and reliable, but also really easy, and understandable for patients and clinicians to buy in.