Anavex Life Sciences Corp (AVXL)

[Investor2014]

CGI-I Anchored RSBQ AUC without the maths

Well Missling is explaining it all again as I write and I believe he has got it right, but here is my attempt.

RSBQ starts out with a baseline measure (Caregiver/parent questionnaire consisting of 45 items), which is completed as the patient is enrolled. Then the same questionnaire is completed at the end of the trial (EOT) for each patient. In the case of Trofinitide with 12 weeks in between the just two samples.

This happens for both the placebo and drug arm patients. Once the data is unblinded the scores between the two questionnaires for each patient are compared to see if they improved or declined.

Finally, those delta scores are compared between the placebo and drug arm to test if the improvements are statistically significant with a 95% Confidence Level, meaning that there is only 5% chance that a positive result is a random occurrence.

The RSBQ and CGI-I topline results with this method is based on an average score across all the 45 items.

CGI-I (Physician questionnaire) is in parallel also the measured at baseline and EOT for each patient. CGI-I uses a 7-point scale from 1=very much improved, 2=much improved, 3=minimally improved, 4=no change, 5=minimally worse, 6=much worse, 7=very much worse.

In the Trofinetide trial RSBQ was the primary outcome measure and CGI-I the secondary.

CGI-I is assumed more reliable given that the professional physician scoring vs. the presumably more emotionally involved biased caregiver/parent RSBQ scores. The two questionnaires covers similar facets of Rett, but are not identical in questions or scoring scale.

Given that Rett patients can have good and bad days, and that observeresations are subjective, both RSBQ and CGI-I have limited accuracy/reliability. Comparing the trend of the two observation scoring methods should help detect a trend/correlation. However, given the variability having only two sample points one at baseline and the other at the end of trial, the end result is at the mercy of both the drugs effectiveness and whether a sample on one day or the other for each patient happens to be good or bad day.

In Rett there are no established objective biomarkers to compare the questionnaire scores to. So what can one do to more reliably gauge actual outcomes with the tools at hand?

CGI-I Anchored RSBQ AUC magic?

AUC, means Area Under the Curve. This is in short a Pharmacokinetic/Pharmacodynamic (PK/PD) measure for how much drug the patient is exposed to throughout the body and how it is metabolises and expelled over a time horizon. In the Anavex Rett trial therefore there would be several sample points e.g. once per week I believe, so baseline + week 1, 2, 3, 4, 5, 6, 7, eight in all.

Having both several drug PK/PD sample points (several blood samples) together with several coinciding RSBQ and CGI-I questionnaires, a picture emerges that relates drug exposure to response. So that's the AUC part.

Anchoring RSBQ to CGI-I, now we want to enhance the reliability of the RSBQ scores by aligning them with the professionally (and presumably less biased) physician CGI-I scores. To do that, the somewhat different questionnaire items are grouped and anchored together making the scores comparable. In CGI-I the first 3 scores on the scale are improvement and only those are Clinically Meaningful, which is much harder to achieve and even more important that statistical significance, although the two much go together for approval (unless of course if the drug happens to be sponsored by Biogen in which case neither are required).

With CGI-I Anchored RSBQ AUC we have several time point samples combined with drug exposure cancelling out pot luck good/bad day effects associated with the just a baseline and EOT score AND we have a clear expression of Clinical Meaningfulness that RSBQ and CGI-I do not measure well individually. So ACAD/Neuren with Trofinetide could in principle just have been scraping by on pot luck, imo.

Zero doubt in my mind that CGI-I Anchored RSBQ AUC is the superior measure with the tools at hand. As Missling points out there is further correlations with objective biomarkers, regardless of whether they are yet not on their own considered reliable outcome measures.