Celebration Thread! Biogen is going to the FDA with Aducan.

[J11]

The bottom figure discusses how ARIA would be managed. Stopping dosing for those with symptoms (e.g., headache would seem reasonable), though from what I understand this was not the protocol.

[J11]

Table 25.GIF

Table 26.GIF

[J11]

Yeah, that is all of the tables!
I have now remixed the FDA position statements, the Figures and the Tables.
This now covers most of the important topics in the first segment of the Document.

I can make some concluding comments and then move to the Appendices.

This main section of the Briefing Document did clearly present persuasive evidence for the efficacy of aducan. Almost all of the outstanding issues were addressed and there were not that many loose threads left unexplained. This portion of the Document concentrated on the topline results which reflect the centroid of the trials. This totality of the findings was strongly positive. When the high dose patients are specifically considered the results become even more impressive. This is all the more positive given the nonnormality of the results. To overcome the gamma distribution problem shows how effective aducan truly was. AD patients would not be expected to improve that much on (an absolute basis) and they did not in these trials, this left heavy lifting in the scrum which aducan was able to win. The statistics that included a nonnormality perspective found that aducan did even better than what was found on the top line primary (normality assumption).

Also of note is that the actual calculations involved in the aducan trial did not appear to anticipate the results to be as strong as those which were reported. Over 3,000 patients were not necessary to prove efficacy: there is excess evidence that accumulated from the dataset. Consider that BAN2401's phase 3 determined that 1,500 patients and a 25% reduction in decline would be sufficient to prove effectiveness. There is a great deal of extra evidence now for aducan.

[J11]

I am happy that I have been able to achieve a good level of comprehension of the main section of the FDA Briefing Document. Yet, there are a few ideas that are still somewhat unclear to me (I would like to have better insight into how logit linear, ANCOVA, MMRM, and imputation were operationalized in the studies.) It is probably for the best that technical issues did not muddle the strong results that were reported. The main portion of the Briefing Document effectively communicated the substantial clinical effects of aducan from the perspective of the primary analysis. Other points of view follow in the document and they might also be convincing, nevertheless, staying with the primary analysis gives confidence in the conclusions advanced by the FDA and Biogen.

[J11]

After all my recent posting I'll be on vacation for about a week.

First though, Figure 19 page 5, is quite a surprise. 1623 patients went on the long term extension after the phase 3s and received at least one 10 mg/kg dose? This information should have been prominently featured during the presentation. 792 of these patients were from the 301 or 392 high dose arms? This is important.

From Figure 39 we know that marginal dosing can greatly change the effect versus placebo. Extending out the 160 mg/kg cumulative dosing level could produce quite a strong response; the already reported dose is 160 /mg/kg expected change of 30 per cent on CDRsb. 2-3 more doses might move up the treatment response to close to 45%. The document does not seem to mention this long term extension elsewhere nor is there information about the distribution of dosing, Would it not be considerably ironic if the answer were related to the information in Table 19 page 5 which perhaps many chose to steer clear of.

What perhaps happened was that as the patients finished on trial treatment, they began to enroll in the extension trials. However, this was overlooked after the phase 3 trials themselves were shut down to futility.

Encouraging patients to get back to the clinic and receive aducan is one of our best strategies after COVID is over. 2400 patients on 10/mg/kg would rapidly demonstrate effects.

[J11]

Over the last few days I was able to make a great deal of progress in writing some R code that would calculate polygenic scores. It is not much (possibly a 20 line program), though this could be of substantial benefit to others. There are a great many polygenic scores available from the research (e.g,. Alzheimer's, schizophrenia, IQ etc. ). If everything works to plan the program could accept a file with rs numbers and betas and then provide a PGS. This could be a very powerful way of exploring a genome. I will post the code on the $100 genome sequence thread.

[J11]

I am searching online for the p-value for the original futility analysis, though I have yet to find it. This could provide us a sense of the change in p-value as the number of patient in each trial increased. Knowing this would inform our consideration of these p-values.
Are the p-values consistent as more patients enroll or is there variation?

Some of the terminology from the Document is now better understood by me. For example, conditional power. I was quite unsure of what this term meant.

"conditional power for CDR-SB,which is the probability calculated on the data at the interim analysis that the final analysis would show statistical significance in favor of aducanumab"

Also investigated principal component analysis especially in comparison to factor analysis. They are clearly different. Factor analysis is a latent variable technique that finds linear combinations of factors (latent variables ) that relate to a dependent variable. Principal components fits weights to observations creating an index variable. The combination found is the optimal such combination.

I have found a fun way to graphically display alpha and beta. I will discuss this when appropriate later.

The first section of the FDA Briefing Document continues to become more clear to me. There are now no longer many issues of uncertainty remaining. From the presentation given so far (accepting the way in which the dataset has been organized) I would agree that aducan is a highly impressive AD drug.

There is no doubt in my mind that I would want this drug if we were dealing with a family member with any level of Alzheimer's disease. Patients have waited so long for a new drug. One of the patients on the trial had been diagnosed over 20 years ago. More than anything else this long delay is what is causing the side effects for the patients. After waiting for 20 years, high dose is then given for 18 months and this is the most dangerous way of administering it. If there were years of lead time aggressive dosing would never be needed and the risk of ARIA would be much lower.

[J11]

Vacation has been cancelled!

Very exciting!
Here's the idea.
Use the dot plotter to find the dot locations of Figure 7 from Appendix 2 of the Briefing Document. Check
Go into R and take these points and plot them as a distribution. Check.

Here is the R code to plot the blue points (302 placebo).

library(ggplot2)
library(MASS)

x<-Blue$V2 +5

den <- density(x)
den

dat <- data.frame(x = den$x, y = den$y)

# Plot density as points

ggplot(data = dat, aes(x = x, y = y)) +
geom_point(size = 3) +
theme_classic()

Blue Distribution.GIF

[J11]

Ok, so what is the above?
I read in the data points using https://apps.automeris.io/wpd/ ,
First are the Blue points (placebo) for 302 as seen in Figure 7 from Appendix 2.
The points were in x,y format. For the distribution I am only interested in the CDRsb scores which was the y-vslues.
The y-values were in the file Blue$V2.
The program would not fit a gamma distribution with negative values so I added 5 to everything.
The above distribution is the starting point.

I am sure everyone knows where I am going with this now.

[J11]

Blue Distribution 2.GIF

This is the R code to fit a gamma distribution to the data.

# Fit parameters (to avoid errors, set lower bounds to zero)

fit.params <- fitdistr(x, "gamma", lower = c(0, 0))
fit.params

# Plot using density points

ggplot(data = dat, aes(x = x,y = y)) +
geom_point(size = 3) +
geom_line(aes(x=x, y=dgamma(x,fit.params$estimate["shape"], fit.params$estimate["rate"])), color="red", size = 1) +
theme_classic()



The shape and fit parameters used for the gamma were:

shape rate
11.265466 1.700323
( 1.962672) ( 0.302923)