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

Alzheimer's, cardiovascular, and other chronic diseases; biomarkers, lifestyle, supplements, drugs, and health care.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

CDR-sb 2.gif
You do not have the required permissions to view the files attached to this post.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

This is an exciting one!
http://adni.loni.usc.edu/adni-publicati ... opsych.pdf
I had to look around a bit to find it, though it was worth the effort.

What we see is a correspondence table for ADAS-cog, MMSE and CDR-sb.
In the figure I boxed the MMSE and CDR-sb scores that had a single set of MMSE scores for each CDR-sb.
So, 2 MMSE scores of 28 had a functional mapping to a CDR-sb score of 1.
And the CDR-sb scores in the red boxes functionally mapped back to the MMSE scores.
One exception was the 0 CDR-sb score which also mapped to 29 MMSE.

How exactly is this exciting?
What I find interesting is that ADAS-cog has no less than 6 different scores for MMSE 30!
It has 7 scores for a CDR-sb score of 0!!

Exciting, n'est pas?

The idea here is that MMSE and CDR-sb are unable to make distinctions between
all of the finer gradations that ADAS-cog is making. ADAS-cog clearly has the potential
to be inherently more sensitive than the other measures. One expects that there should
be ways of clinically validating this. Basically repeatedly test an Alzheimer population and
observe how the different tests are able to match the decline. Alzheimer's involves essentially
continuous cognitive decline so frequent online testing could yield deep insights into
dementia measurement. Presumably one could leverage the fact that decline is continuous
and then find the cognitive items that were able to detect various intermediate steps
in the process of cognitive decline.

You could measure CDR-sb and receive scores of 0,0,0,0,0,0,0 yet at the same time
there could be scores of 0,1,2,3,4,5,6 for ADAS-cog. The CDR-sb would have no idea when
a score of 0.5 might happen, though ADAS-cog would have extraplatory power to do this.

This would seem to be a fairly important observation because in the aducan trial there
were MCI stage patients who would have had scores in the MMSE range of 27, 28, and 30;
as well as CDR-sb scores that corresponded to those MMSE, yet ADAS-cog would be able
provide much finer distinctions within each of these ranges.

In fact over the MMSE range of 23-30 (8 different MMSE scores) there are 24 ADAS-cog scores possible
and 10 CDR-sb scores. I can certainly see the argument for the superiority of ADAS-cog. Very strange that
MMSE was used as the gatekeeper for the secondary measures in the aducan trials. Also somewhat odd that
CDR-sb was used as the primary top-line.

These issues concerning measures to use and stages of patients have been argued about for a decade or more
and it still is not fully resolved. When you look at the aducan results, ADAS-cog was only at 27% vs. 22% for CDR-sb
which is fairly close. It is with the ADCS-ADL that there is a big difference at 40% slowing. IT might be worth a look
into the ADCS-ADL.
Last edited by J11 on Wed Mar 31, 2021 5:36 pm, edited 6 times in total.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

Correspondence.GIF
You do not have the required permissions to view the files attached to this post.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

This is another look at the correspondence. They are using item response theory (IRT) with x-axis = theta.

Look at the black line at the top as it moves form -4 to -1 x-axis (i.e., worsening cognitive dysfunction) corresponding to an MMSE of 30.
(black curve is MMSE). Notice that the CDR-sb (red line) is also flat from -4 to -1. Surprisingly, both MMSE and CDR-sb are entirely unable to detect any cognitive difference over this wide range of theta. That entire cognitive ability range is placed into a single bin (i.e., MMSE 30 or CDR-sb 0). Yet over this range ADAS-cog increases from 0 to 5. It is able to detect differences in this range. One problem with this figure is that they squished the ADAS-cog scale to fit with the other scales. This makes it appear as though CDR-sb has higher discriminatory power to detect score changes than ADAS-cog (ie., higher slope, for example at theta = 1). That isn't true ADAS-sog has finer gradations that either MMSE and CDR-sb.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

dona ADAS-cog.GIF
iADRS tau cor.GIF
Brain tau.GIF
You do not have the required permissions to view the files attached to this post.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

Top figure in the above series appears to show the ADAS superiority. The ADAS result gradually increases the benefit of treatment which would seem to be what should be expected. However, with the ADCS shows a ballooning of the benefit and then a deflation of the balloon that is not what one might expect; one is expecting more of a linear trend from both arms. Notice that the measurement error would built for both the placebo and treatment arms. Due to the less granular ability of the non-ADAS-cog there would be a lumpier read-out: that is CDR-sb would read out 0,0,0,0,0,0,0 and then suddenly it would be 0.5. Yet, with ADAS-cog there would be more of a constant evolution. This is what the figures seem to show: one smooth, one lumpy.

Good to see the tau and measured correlation. Later this report notes a ~25% correlation for tau and measured result (cognitive/functional) for the latest dona result. Moving to individual level not group aggregated dot correlation introduces the exciting potential of precision medicine. Possibly the new tau imaging with selection by tau level will help to further increase the correlation.

Last one is also pretty exciting. There is a huge tau SUVR change in the frontal lobe of 59%. Outsiders would not realize that this is an emotional red flag. At 59% reduction, we are start marching!

If this translates into protection of the frontal motor cortex, then that would have a profound functional impact on dementia management. When dementia patients can no longer walk the caregiving burden increases by at least 2 orders of magnitude. As soon as you can't walk, you have lost your functional independence and your ability to exert your agency as a person. If anti-amyloids (anti-taus) treatments can substantially change the trajectory of ambulatory independence, then this would be a line in the sand for the AD community. This is a firm functional endpoint that has profound significance to caregivers. It is a categorical yes/no that does not require fancy testing.

This bottom figure also introduces the exciting prospect of a more refined type of imaging prediction of cognitive and/or functional ability. Perhaps an imaging scan could reveal that a patient had more tau removal in the occipital lobe and this would suggest less hallucinatory experiences and somewhat less in the frontal motor cortex and thus less reduced motor decline. This information could be of practical importance in managing care. Perhaps different anti-amyloids have different abilities to manipulate amyloid/tau in different brain regions in different patients?
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

Corr.GIF
You do not have the required permissions to view the files attached to this post.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

I am feeling a wave of optimism for aducan.
The results are starting to make sense to me and it feels as though everything is aligning.

I have previously posted a great many of the correlation plots similar to the one above (as readers are no doubt all too aware).
The correlation across 3 of the anti-amyloids is surprisingly strong: close to 100%.
Probably would have been best to keep some extra noise in there to increase the level of
confidence in the result. It's best not to find a result that seems too good to be true.
Trying less hard in data massaging would have been productive.
Perhaps this almost implausibly good result is a consequence of choosing trials that have fairly
substantial n's in the different arms.

Nevertheless, the correlations are so strong that the result is close to being unequivocal.

If it were me I think I would emphasize this type of thinking more the endless reporting of p-values.
All of the p-values (versus placebo) for the points in the correlation above could have been insignificant
statistically and a strong correlation would still override such results. With the correlation approach
you are using the totality of all the data. With aducan, they were only ever using ~300 patients in each
treatment arm against a similar number in placebo. In my correlation plots there are a lot of patients
involved. There are over 1,000 patients in the aducan treatment arms alone. The placebo also has more than
1,000 patients. The N that is present in the correlation plots simply overrides all pairwise p-value calculations.

Also if it were me, I would probably even move away from placebo. If you think in correlation terms what is
the benefit of placebo? Placebo is a fairly empty readout. We know that AD placebo have close to 0 SUVR change.
No information there. We also have a fairly good idea about 78 week cognitive change ~1.5- 1.7 on CDR-sb at 78 weeks.
What would seem to make more sense would be to dose the placebo with active treatment and receive more actual
correlation data. Why receive such a minimal information yield from these patients? What information are the placebo
patients actually providing? The linear association between amyloid and cognition is much more important to clarify.
What keeps everyone honest is that now it is about correlation, it is no longer enough to say we hit a strong top-line
result. What is important is to achieve a strong linear correlation. Thumbing the scale would no longer be advantageous.
Without the placebo, it is no longer obvious how one could go about thumbing the scale; what needs to happen is the points
need to be on the line. Merely adding extra scores to patients even on treatment would not be that helpful.

Another suggestion is that the correlation plots could be given more as an area mass than as a single dot. This would give a much clearer impression of the range of results especially if the patients were heterogeneous.

One additional insight that I have had is that once the correlation of anti-amyloid and cognition is beyond argument
there is an important implication. (For those still not convinced about the correlation
I suppose additional points could be retrieved though I am unclear how many more of
these dots would be welcomed.)

The important implication is that once the regression line is accepted, then one could
reasonably extrapolate this line. There is no especially compelling reason to believe that
aducan fully maxed out its treatment window. In fact the zero point for SUVR apparently is
0.372. According to the regression equation above that gives us 0.466 "on the line" for aducan
benefit (i.e., another 0.08 CDR-sb benefit). Notably, the phase 3 302 high dose aducan slightly
overperformed its expected regression potential. Other secondary measures could also yield
super-low results if they were extrapolated as well.

Another 19.5% benefit relative to the reported 22% for CDR-sb.
Giving 26.3% over placebo on the top line.
The p-value of that would be really small.

There are not that many unlikely hand wavy assumptions being made:
- The regression line is highly confirmed in multiple clinical trials, in multiple treatments at multiple dosing levels.
- It is highly reasonable to expect that aducan did not max out the benefit; higher total doses in aducan trials confirm this.


There is also quite a bit left on the table regarding patient selection.
When they started with AD trials, they did not even check to see if the patients had AD (i.e., amyloid).
aducan checked for amyloid.

AD also is defined by tau.
The aducan trials did not universally screen for tau.
dona did.
This is a large result that has emerged post-adcomm.
dona used substantial selection on tau (~3:1).
Could even use ~9:1 or higher to truly capture AD responders.

This result has redefined the AD clinical landscape.
I am not sure what you are supposed to do when you write the Briefing Document and then
something new lands on the desk. Perhaps this is why the FDA called for more time on the clock.

I mean the aducan results now are clearly understating the treatment
benefit to the respondent class. It is hard to imagine how aducan was even able to do as well
as it did when it was so loaded up with known non-responders. dona excluded those too early
to benefit (not enough tau), they excluded those too late to benefit (too much tau). They included
the intermediate tau. Yet, even in the intermediate stage two thirds had too much tau and also did not benefit.
Basically, you can use 9 fold or greater selection in early AD to find the actual respondent class.
aducan took everyone and still swept the topline for 302.

Given that all of these anti-amyloids are doing the same thing, amyloid and tau are now the universal
biomarkers to screen for responders. It is a little murky to try and see through the murk but
there could be large reportable AD benefits in future trials.Only 1 in 9 patients are rowing the boat
the other 8 in 9 are along for the ride and resting their arms for the buffet table. The dona figures
did not show as large a result as I had expected, though if only 1 in 9 are the true responders, then it
is inevitable that the result would be substantial.


Not only were the treatment arms stacked against aducan, so too were the placebo arms.
There were a lot of patients that we now know were too early to progress (not enough tau).
These were the substantial number of placebo patients who did not get worse, some even got better!
How can you show that an AD treatment is effective when the two normal curves overlap so much?
AD patients in AD clinical trials need to have amyloid, the right amount of tau, and they need to
be in the active phase of cognitive decline (hopefully, we can move more towards the super-earlier stage
where there is no cognitive decline and amyloid is understood to be a neurotoxin).


Amyloid science has advanced and clearly the as reported aducan results are underestimates
of the true treatment benefit for responders. This of course does raise the question what other
exclusions could be made for potentially other non-amyloid type proteinopathies? TDP-43 etc.?
Possibly mixed Parkinson's? Genotyping might be one way to narrow down such questions.

Might also suggest concentrating more on APOE 44s who might have a purer and more homogeneous
form of AD.
Last edited by J11 on Wed Mar 31, 2021 5:44 pm, edited 1 time in total.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

Dona Phase 3.GIF
tau terciles.GIF
Dona Phase 3 iADRS.GIF
You do not have the required permissions to view the files attached to this post.
J11
Contributor
Contributor
Posts: 3351
Joined: Sat May 17, 2014 4:04 pm

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

Post by J11 »

Correlation
Dona intermediate low tau p-value 54 sample.GIF
Dona intermediate low tau p-value 100 sample.GIF
Dona intermediate tau p-value 100 sample.GIF
You do not have the required permissions to view the files attached to this post.
Post Reply