1/ Our study of genetic prion disease age of onset & its implications for clinical trials, is now on @biorxivpreprint!
Blog post forthcoming next week. For now, a quick tweetstorm.
Blog post forthcoming next week. For now, a quick tweetstorm.
2/ From day 1 of predictive genetic testing, Sonia & I wanted to know WHEN will she develop the disease? We asked collaborators to help us aggregate the world's largest (and only public) dataset on age of onset in genetic prion disease (N=1,094) to try to answer this question.
3/ Result: age of onset is incredibly variable (e.g. age 12 to 89 for Sonia's mutation) and as far as we can tell today, nothing predicts it. It seems to be totally random.
4/ Our goal is to develop a drug to delay onset of Sonia's disease. So next research question: how can you show that a drug delays onset, if onset is so variable?
5/ Turns out it's very hard to follow people to a clinical endpoint in a trial. Because onset is so variable, you can't identify a decent-sized group of people likely to develop disease in the next five years.
6/ To illustrate this, we made this plot, inspired by the legendary Charles Joseph Minard plot of Napoleon's invasion of Russia. Recruit young people - OK numbers but their annual risk is very low. Recruit older people - higher annual risk but very few people.
7/ Power calculation: if a drug reduced annual risk of onset by half, then to have 80% power at P=.05 to show that it delayed onset, you would need to randomize 813 people age 40+ for 5 years.
8/ That's a v. long and v. expensive trial. Moreover, it's more people than we even *have* in this rare disease. Mutation carrier rate may be ~1 in 100K, but due to underdiagnosis & low genetic testing rates, we think only ~60 ppl in U.S. meet criteria for trial modeled above.
9/ Conclusion: if our goal is to develop a drug to delay onset of prion disease, clinical evaluation of that drug will require a biomarker. This motivates our work on biomarkers, e.g. Sonia's recent @biorxivpreprint on CSF PrP levels: biorxiv.org/content/early/…
10/ Final question: if a biomarker could one day get us Accelerated Approval of a preventive drug, how do we ultimately convince ourselves and regulators that the drug really does delay onset?
11/ Answer: of course many possible designs for post-marketing confirmatory studies. We raise one option that merits evaluation: surveil drug-treated people long term and compare their survival to historical controls.
12/ You get increased power from longer trial duration, large control N, and maybe continuous enrollment, and maybe can recruit larger N for an approved drug than a trial. There are of course issues with use of historical controls, but worth exploring and trying to make it work.
13/ Conclusion: the mandate for us as prion scientists: 1) Develop biomarkers for genetic prion disease, 2) Develop infrastructure for post-marketing surveillance. This give us a route forward in case a drug is effective only at prevention and not in symptomatic patients.
14/ Final note: public dataset & source code for this study are available on GitHub for your reproducing pleasure: github.com/ericminikel/pr… *END*
