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

[J11]

NF52, thank you for your comment.
Yes, this figure has been out there for quite a while.
I would like to see some more granular amyloid scans at different ages and for different genotypes.

One of the bigger lessons that I can relate from hard experience is:
Solve problems as soon as they arise.
Pretending that problems do not exist and kicking them down the road is one of the worst strategies possible.
Costs escalate exponentially when problems are not solved.

What I see in that figure (given the solve problems right away perspective) is amyloid should be addressed
as soon as it begins to accumulate. This might even be in the 20s for those with 44. Waiting for the symptomatic
onset is a hopelessly bad strategy. The frustrating problem here is that this research might never be done given
the regulatory and patent environment.

How much amyloid do 30 year old 44s have?
At what age might 44s be motivated to microdose?
From what I vaguely recall, an article spoke of amyloid being present for "decades" before symptomatic onset.
One clearly does not need clairvoyance to see massive demand from the worried well for an anti-amyloid.

[J11]

I will not be able to go line by line on the minutes of the adcomm transcript, though there are some general comments that are of interest.

The second question at the adcomm concerned the data set that was chosen as the primary analysis.
https://www.fda.gov/media/145691/download (page 58). This point had been somewhat confusing with the FDA Briefing Document. There were three different data sets presented: ITT, ITT uncensored and opportunity to complete. It is emphasized that ITT was chosen as the primary data set for consideration: everything was double blinded and all pre-specifications were adhered to. I would have liked a more complete description of why ITT uncensored was not thought worthy of more prominence. Wasn't the blind maintained even after the trial had been "terminated"? The ITT uncensored provided the most comprehensive insight into the results of the trial. Of importance, the p-value fell by a factor of ~4 from ITT to ITT uncensored. It would become especially difficult to call into a question the top line 302 high dose p-value of 0.25% from the ITT uncensored data set.

The follow-up questions clarified that the ITT was pre-specified in the statistical analysis plan, so ITT would seem to be the required dataset. The rebuttal that could be imagined is that the futile futility analysis wasn't part of the plan. The analysis was going off-road --
ITT uncensored (though unspecified) might then not be an entirely second choice. Perhaps the only reason that ITT uncensored was not mentioned in the plan was because no one had thought that there would even be ITT uncensored.

The pressing on the p-value calculation related to the conditioning on the 302 result seems out of place to me. Why not simply have this added into the agenda of the meeting before the meeting? There is minimal value in pressing a gotcha moment when it could all be handled smoothly as a preset meeting item.


Reading through all the back and forth about what was and what was not prespecified certainly makes me wonder why the statistical analysis plan had not been reworked after all of the oddness of the trials. Keeping the same plan even with changing the dosing in the high dose group for the APOE+, not having the same treatment dosing for 301 and 302 etc. wouldn't seem to reflect the true nature of the actual datasets. If it had been me, at some point, I might have simply said we'll go to statistical analysis plan B. Perhaps the data unlock could happen in stages. One obvious fall back would have been to simply ask for the results for 301 and 302 in the patients that had substantial numbers of high dosing. I am still unsure how the data set for high dose aducan makes sense when so many on high dose never actually received any high doses. The entire argument that 301 high dose was ineffective disappears once those on actual high dose are considered. How is receiving treatment even an assumption? The obvious workaround that could have been considered is that patients would remain on trial until they actually received some minimum number of high doses (e.g., >10).

Later in the transcript they move on to the rapid progressors. I am surprised that the rapid progressors were ever allowed to gain such importance. The problem was that they were using the "average". These rapid progressors had the weights of upwards of 23+ typically progressing patients. Yet, what people really want to know is how treatment will benefit these typical patients not the highly distorted "average" created by "rapid progressors". Less consideration should have been focused on rapid progression and more should have been focused on median patients. What is of further interest as was noted on many occasions on this thread was that strangely enough the rapid progressors who were on treatment typically did not actually receive many high doses. I am not sure if all the data was presented, but some of these high dosers did not even receive any high doses. Nonetheless, the potential that aducan treatment was driving this extreme progression actually occupied quite a bit of thought. BUT THEY DIDN'T RECEIVE ANY HIGH DOSES! Shouldn't the question have been reversed? Why were these rapid responders who did not receive high doses (though they were nominally designated as high dose) progress so rapidly?

As a parenthetical aside, one of the earlier figures from this thread illustrated how the different cognitive tests had varying discriminatory power across the range of cognitive impairment studied. So on the item response theta scale, MMSE and CDR-sb
are pretty much hopeless in discriminating from ~-4 to -1 theta. ADAS-cog does much better in this range. CDR-sb does much better
from ~-1 to |+2 theta. This certainly has me wondering whether some sort of a patchwork scale might be best. Use the psychometric instrument with the highest discriminatory ability for different impairment ranges.

Another highlight of the transcript for me is the reporting of the US result from the phase 3. The number mentioned was 0.55 CDR-sb benefit (302?). That is quite a bit more than the average top-line. The non-US number must have been quite a bit less. This problem has cropped up on other global AD trials. It seems that some nations really do not have a great grasp of what AD actually is.
However, in nations such as the US, public awareness of the features of AD has increased. Other places in the world might have a vaguer idea that AD is merely cognitive impairment; perhaps that it is a form of alcohol abuse neurodegeneration. While embracing a more international perspective is admirable, such an approach might hinder clinical progress as not all nations are at the same stage in understanding the evolving AD crisis.

Page 70, specifically considers the question of the mix of MCI and mild AD patients. The commentary noted that removal of amyloid earlier might be more beneficial. Yet, this really was not what the trials found. Mild AD patients had much larger CDR-sb benefit than did the MCI patients. This was also true in the lecanemab phase 2. There is so much heterogeneity in the results by subgroups that it really would be for the best if subgrouping by demographics were done. Now with tau staging significant results could be achieved rapidly in some of the subgroups.

Another general comment is that in the transcript whenever the discussion begins to become especially interesting there is a call to move on as time is scarce. Yet, then the transcript have 50 pages of boilerplate. The whole concept of the flipped classroom has not
fully taken root. The idea here is that you do not emphasize what everyone understands, but what everyone does not understand.
The value of a meeting is exactly in discussing the difficult parts and minimizing all of the ceremonial and pro forma aspects of the agenda. The transcript for the adcomm repeatedly curtails the most vitally important components of the discussion only to replace them with the least insightful relevant components.

For example, on page 81 a discussion of principal components analysis is gaining speed but then it is thought best not to discuss this further. This was a particularly interesting topic for me. The position that AD is primarily driven by 1 or 2 factors does seem to have substantial intuitive appeal. AD does certainly coalesce as a unitary disease, though differing features are also present. One approach to measuring this might be to note which patients were especially affected by the different domains (e.g., were below 1 SD on function, neuropsych, or cognition). This would give a more multidimensional view of progression. Without such an approach you could imagine that some AD patients might score highly on a purely cognitive test such as the MMSE, yet do much less well on the NPI (had hallucinations). Looking out for the more severe impairments on the different dimensions would more accurately represent the nature of the illness.

After the lunch break, members from the public were encouraged to make statements. I found it notable how some of the patients reported substantial improvement while on aducan and then the loss of this improvement when off treatment and regaining the benefit when back on therapy. This seemed to be a big missed opportunity to actually turn anecdote into statistical evidence. Statistics is the language of these types of meetings and the actual patient experiences can then somehow be thought of being less
valid than the numbers. Why then did none of them call for this evidence to be compiled? There is a large accumulation of further results for aducan. Why has this not been scrutinized? The fact that patients could be compared against their own cognitive trajectories while on aducan and while off it, would seem to offer a uniquely powerful way to make a true comparison. The trials only compared the decline of those on treatment versus those on placebo. This introduces a great many variables: you aren't actually comparing the same person against their own pattern of cognitive decline. The speakers who did comment on this issue were able to offer powerful personal narratives of how much of a difference treatment made for them. The statistical power of thousands of other patients could offer very powerful evidence of efficacy.


Later on in the transcript they move to the actual voting. What was of particular interest was how the adcomm understood the question about 302 success. This starts around page 200. I had not understood how there could be any confusion: 302 high dose hit its top-line cleanly. The contention from the adcomm was that they did not see the 302 as being an independent separate result from the 301 study. The adcomm largely dismissed the framing of the question. The problem was that the question was leading to an answer and the adcomm did not accept this tilt. The vote that was almost unanimously against really should not be understood as a negative vote because in their comments they made clear that they rejected the framing of the question itself.

A later question had a similar problem. The FDA's stated position was that any difference from placebo should be understood as clinically meaningful. That was there position. It is not totally clear whether that really is entirely reasonable, though that was there position. The adcomm also rejected this tilting of the question. They voted down the question not so much on the basis was non-superior to placebo, but because it did not have enough evidence of having a clinically meaningful difference to placebo. Technically the adcomm is correct on this point. aducan posted a 22% benefit over placebo on high dose in the 302. 25% was thought needed for a clinically meaningful difference. Using different measures (e.g. ITT uncensored, the higher dosed patients, probably APOE+) such clinically meaningful differences would have been achieved. The recently published leca phase 2 did report strong evidence supporting efficacy. For example, the APOE+ on the highest dose had a 99.2% chance of being superior to placebo (with only 45 included). The result would have been much much stronger if the decision to stop dosing APOE+ at the highest dose had not been made. The current Phase 3 Clarity trial is dosing everyone in the treatment arm at 10 mg/kg bi-monthly. The 218 APOE+ who were dosed in the phase 2 leca trial has a 95.6% probability of being superior to placebo. The discussion that anti-amyloids are not better
placebo is no longer even felt to be worth arguing about. The goal posts have moved yet for whatever reason people are continuing to argue the point. The new discussion is whether there is a clinically meaningful difference with treatment. This argument also appears to be largely over. With only 45 patients, the leca phase 2 highest dose reported a 93.6% chance of having a clinically meaningful difference against placebo. These results posted after the adcomm make much of that discussion mute. The dona phase 2 results also means that selecting responder patients will amplify the reported treatment effects. aducan now has a massive amount of more support than it did at that time of the meeting. The science has advanced a great deal since the meeting and this science strongly favors aducan.

One of the recurring problems that was noted at the adcomm and elsewhere was about the lack of correlation between plaque removal and cognitive scores. It would be very interesting to add in the tau imaging to see whether this brings everything into focus.
We saw in the dona study that was tremendous clarity with tau and amyloid together. It is extremely exciting to watch the clinical understanding advance with additional research.

[J11]

I want to go through a last minute aducanumab cram session. Final Examination is Monday, an all-nighter or two might be needed to get everyone exam ready. I continue to read online commentary that does not seem to align with the facts-- I will want to help clarify these facts. Also I want to be clear and concise about all my previous hundreds of posts to this thread. After building up all this background about aducan, I can now draw upon these previous posts to make direct replies.

[J11]

Top line question: Is aducanumab effective?

[J11]

Yes.

[J11]

1. Aducanumab was reversed engineered from healthy cognitive super-agers. These super-agers were highly selected exactly because they did not develop cognitive impairment as they aged. It is difficult to rebut this feature of aducan's development. Sample sizes of these super-agers with naturally occurring aducan could be increased as large as liked, though they will show the same result -- nanodosed aducan throughout the lifespan reduces amyloid burden and it prevents the emergence of later life cognitive impairment.

[J11]

2. Numerous plots on this thread have shown that a range of anti-amyloids (including aducanumab, lecanemab, donanemab, and gantenerumab (perhaps others for example gamma entrainment) have a strong correlation between amyloid removal and cognitive benefit typically measured by CDR-sb. This correlation was remarkably strong; even in the least selected of these plots the correlations were strong. The only truly anomalous result was for 301 high dose aducan. Yet, even in that study large subgroups actually were near the expected regression line.

The fixation on comparison to placebo seemed misguided (the p-values). The question that I became more fixated on was the dose-response relationship: not so much was the treatment effective but was the treatment effect as expected by the regression? For aducan and the other anti-amyloids the treatment effect was often nearly exactly what would be expected. Ironically, with the low doses of aducan, if they had shown a significant treatment effect, from the perspective of this strong dose relationship, it would have raised questions for me whether the dose relationship was in fact valid. The FDA Briefing Document, however, interpreted the lack of low dose significance as somehow diagnostic of lack of treatment effect.

Aducanumab was the breakthrough anti-amyloid that extended out SUVR amyloid change to ~0.27. It is no great mystery to see now why earlier anti-amyloids were ineffective; they did not remove much amyloid. It was surprising how little some of these anti-amyloids did actually remove. Apparently, some of them did not remove any oligomer amyloid. We have the benefit of hindsight, yet it is very unclear to me how these early anti-amyloids could have possibly been helpful to patients. Those who mention the long history of unsuccessful anti-amyloid treatment are largely referring to this early era. Yet, the current era demonstrates benefit for all of the aducan class drugs. It has merely become a question of dosing up. Perhaps newer dosing schedules can be devised that are more effective at reducing side effect risk. Possibly even more careful titrating up at the beginning when high amyloid is present could be useful.

The point about the lack of correlation between amyloid and cognitive benefit was repeatedly noted in the FDA Briefing Document. For some reason I was somewhat confused by this point, notwithstanding the endless correlation plots on this thread that showed ----- a very strong correlation between amyloid removal and cognitive benefit (admittedly at the group level).

[J11]

3. The other aducan class anti-amyloids have recently reported very strong results-- almost unequivocally strong. For example, the phase 2 lecanemab epsilon 4s on highest dose (n=45) had a 99.2% probability of being superior to placebo. The second highest dose for the epsilon 4s was 95.6% (n=218). The goal posts have moved; it no longer seems even relevant to argue anymore about straight efficacy against placebo -- science moves forward. The discussion has shifted now to clinically significant change. However, it should be noted that the FDA Briefing Document clearly stated that the definition of success for aducan was superiority to placebo (not clinical significance). aducan was near the line in terms of clinical significance, though in the higher dosing and other subgroups, it was quite a bit past this line. The donanemab phase 2 was also reasonably strong especially with tau selection. Also the gamma therapy unpublished results.

[J11]

4. The subgroups have also shown consistent differences in response across several of these trials. In some of these sub-groups the treatment effect appears to be quite large. For example, in the leca phase 2 the CDR-sb benefit in Mild AD was ~1 point. That is big. In the other trials large differences also were reported. Such differences were obscured largely by the demographics of the different studies. aducan phase 3s had 80% MCI which also reported much smaller benefit than did the 20% Mild AD. I had understood why the patient population was selected in this way, though I now suspect that this would allow a larger benefit to be reported in the longer term follow-up. Treating earlier should offer larger benefit, however I am somewhat less certain about this perspective after it was noted that with aging there is an opening of the BBB. This has me questioning whether treating younger patients with anti-amyloid anti-bodies might not have the same benefit as their BBB might be less leaky.

[J11]

5. The effectiveness of aducan is most dramatically demonstrated in the longer term extension. This was seen particularly in the 103 LTE. The benefit of aducan extended out for years and years. They have reported out to ~4 years and the benefit apparently persisted.
Claiming that a 20-40% slowing of progression is not meaningful is only plausible over the short time frame of an 18 month clinical trial.
When you look out to 4 years, this treatment effect would become visibly self-apparent to those without any training in AD. From what I projected, the patients in the 103 on placebo could now be approaching total disability, while the high dose patients would likely still be capable of somewhat independent activity (~moderate AD).

Questioning the effectiveness of aducan simply does not seem to be a particularly strong line of argument. This is all the more true because Biogen has thousands of patient years of extended dosing; some of these patients have been on treatment for ~5 years. It has already been suggested that these patients have (from the phase 3s) replicated the study 103 long term benefits. I am very unclear why so many commentators have completely overlooked this dataset. Basically, Biogen can back up the truck any time they want and unload. I do not even think this is necessary, I believe them this dataset must be overwhelmingly in favor of efficacy. Yet, for whatever
reason the commentators have no insight into how weak their hand is and how persuasive existing LTE evidence could be in support of aducan.