FDA Briefing Document On Amylyx Pharmaceuticals’ AMX0035 Says ‘it does not appear that this data can be considered independent confirmatory evidence as it uses the same data as the primary analysis.’

FDA Position:
This post hoc analysis is highly correlated with the primary analysis. Both change from
baseline slope and pre-study slope were used in the primary analysis; thus, this is not
independent data. Therefore, it does not appear that this data can be considered
independent confirmatory evidence as it uses the same data as the primary analysis.
Also, notwithstanding the Applicant’s explanation, it is unclear why the Applicant has
chosen to compare the treatment effect at Week 18, rather than Week 24 (the primary
analysis endpoint). We note that the effect size on the primary endpoint was larger at
Week 18 than Week 24.
Additionally, the Applicant claims that using participants as their own control may
provide independent evidence of effectiveness; however, this analysis doesn’t
completely use participants as their own controls. The analysis is still a comparison
between groups, comparing the response rate of patients receiving treatment to the
response rate of patients receiving placebo. A true within-patient analysis that uses
patients as their own control should calculate a treatment difference for each patient
enrolled in the study; this is not possible for patients assigned placebo in this study
design, and thus, is not an independent analysis from the primary analysis. Additional
concerns with this analysis include:
– The pre-study slope was not directly measured and was calculated based on
retrospectively collected data.
o The variability of the change in slope may not be constant, because
“delFS” depends on time from symptom onset, which varies widely across
patients and the mean time from symptom onset of 59 weeks is
significantly longer than the double-blind follow-up of 18 weeks considered
in this analysis.
o Pre-randomization slope is based on a baseline measurement/presumed
score of 48 (normal) at disease onset. It is unlikely that patients would
have a maximum score at the time of diagnosis because they would have
signs and symptoms of ALS that prompted the work-up and subsequent
diagnosis.
o Linearity seems questionable for the pre-randomization slope because this
slope is calculated over a period of up to 608 days, which is much longer
than the 24 weeks of the primary efficacy study, for which the prior review
showed linearity may not hold.
o There is no way to check linearity of the pre-randomization slope.
– There is post-baseline starting of ALS medications, more in the drug arm, which
could confound this analysis.
Therefore, these data appear limited in their ability to provide independent
substantiation for the observed effect on [the primary endpoint.