Sunday, December 3, 2023

FDA, Multi-Cancer Screening, and Mortality? Consider Three Additional Points

 As we move towards an era of blood-based multi-cancer screening, there have been arguments that adoption must follow overall mortality benefits used as an endpoint.   This arose in the FDA's advisory panel on the topic November 29.   For that, see the FDA's meeting web page here, and see a pre-meeting essay by Mullen & Gibbs here.   

The FDA panel was  big deal in the MCED industry, but didn't get much other press.  I added some materials like an auto-transcript and auto-summary to my blog about the meeting (here, look for keyword, Update November 30).

With regard to the FDA advisory board, much discussion of the role of mortality or overall survival as an outcome variable.   To my mind, a lot of the discussion was confused, perhaps caused by the FDA's scanty briefing materials (noted by Mullen, link above).

I think there are three critical ways of thinking about overall survival.

First, overall survival is extremely difficult to show.  People are being screened, they have no symptoms, and their overall survival  is many decades!  And any one, or even several, causes of death by cancer is a tiny widget of all  causes of death, and, for any cancer, at best, MCED will only prevent some of the deaths (like any cancer screening method).  The US has about 3.5M deaths per year.  Of those, deaths from pancreatic cancer are about 50,000, deaths from ovarian cancer, 13,000.  So even if you (incredibly) halved the death rate from pancreatic cancer and from ovarian cancer, it would be a truly minute  percentage of all deaths. 

Added to that, there is no perfect 50,000 patient, 10 year trial.  Such trials have high dropout and lost-to-follow-up rates, and numerous puzzling subgroup differences will accrue.   Let alone the fact, that 12 or 15 years from now, any particular medical technology will be several generations advanced from that used 15 years ago when the trial commenced.  In the real world, that definitive 50,000 patient 10 year trial is forever a mirage on the horizon.

So while you can, and would, show that you prevented 5000 ovarian cancer deaths, showing that you changed the mortality of millions of people is practically nearly impossible.

Second, would MCED just provide a stage-shift, picking up patients earlier without much impact on morbidity or survival?   Under this concern, a cancer with a five-year survival is picked up (per patient) two years earlier, and the result is merely seven-year survival from the same cancer.   This is a worthwhile topic of study, and people are well aware of that, but it is different from overall survival.  

Third, what about the risk the MCED causes increased mortality?  To me, this seems like a fairly distant hypothetical, but on principal, discussing it, or assuring that this does not occur, is logical.

In short, questions of overall survival and mortality can be important, but shouldn't be lumped into one bucket.  Each of the three scenarios listed above has its special ins-and-outs.  When the topic of overall survival is launched into the air, it's rare to see these components laid out concurrently.

On the topic of RCT trial designs and how they calibrate unexpected harms, see a new review article Junqueira et al., 2023, J Clin Epidem, here.