Biotech Values

[biomaven0]

Here is the perfenidone NEJM disclosure of the statistical model FGEN also used:

The slope analysis is a mixed model repeated measurement analysis with linear time effects for patients. Specifically, the mixed model of change in FVC volume uses time-by-treatment, age-by-sex, height-by sex terms as fixed effects, and patient, time-by-patient as random effects, where time is defined as number of days starting from randomization.

So I assume the model basically assumes FVC changes linearly over time and fits a slope for each patient - but there are other ways to do this so I may not be completely correct here.


Here's the description from Nintedanib at the FDA:

The primary efficacy variable was the annual rate of decline in FVC from baseline to week 52 and was analyzed by random coefficient regression model, assuming linear decline over time.

https://www.accessdata.fda.gov/drugsatfda_docs/nda/2014/205832Orig1s000SumR.pdf


More on mixed models with repeated measurements here:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2971698/

http://www.stat.cmu.edu/~hseltman/309/Book/chapter15.pdf