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 »

103 MMRM 1.JPG
103 MMRM ex 1.JPG
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 found the careful scrutiny of the lecanemab CDR-sb vs SUVR curve by doses to be highly insightful. Considering each trial one by one (as was done in the FDA pharma review) for this relationship allows us to see what assumptions are being made and where some of the noise might be entering into the analysis (and how perhaps it can be removed (e.g. by averaging it out by combining into combination arms). Assembling many of these results into a comprehensive meta-figure to a large extent obscures the strong correlations that are present within each study because no allowance is made for the complexity of the datasets. There is a certain art to understanding the numbers that is lost by simply tossing everything together. When this nuance is allowed to be included, many of these trials demonstrate, surprisingly, very strong correlations.

The above figures are the CDR-sb vs SUVR curves for trial 103. What is of interest here is that there were two sets of analyses done for 103. The recent FDA pharma review presents the MMRM approach. However, the 1 mg/kg dose in 103 using MMRM has always seemed "off". The figure for 1 mg/kg in the review posted a few posts above also shows the MMRM with the 1 mg/kg seemingly too close to the x-axis. I wasn't quite sure what was wrong. Importantly, the 1 mg/kg has much better alignment using the ANCOVA in the figures gives 3 and 4 figures above. The problem might be as below with titration there could be even a single outlier patient in this small n treatment arm that greatly distorts the result using ANCOVA. Removing the 1 mg/kg point from ANCOVA helps to normalize the regression. {The 1 mg/kg arm might have been affected by the change in the placeo SUVR level. In 103 there is reported to be a -0.064 difference from placebo for 1 mg/kg, yet the placebo itself seems to have increased by ~ 0.01 to 0.02. 1 mg/kg might have been more affected by this placebo change; the 1 mg/kg might really only have an amyloid reduction of -0.04 to -0.054.}

Also as noted below the titration dose included in the 103 study was not comparable to other aducan trials as it was 44 weeks instead of 26 in 301/302. This is why it was removed in one of the above figures in the ANCOVA analysis. The titration arm overperformed because it was given more aducan exposure. When the titration is removed the regression result becomes stronger. Alternatively, one of the titration dose patients dropped out early after 26 weeks of treatment with a 6.5 point worsening; the MMRM included this patient while the ANCOVA did not -- this might explain the substantial titration deviation from the regression line with ANCOVA.

The ANCOVA with the titration arm removed gives an impressive result. y = 4.4 x+ 0.053 R^2 = 0.974 . As noted the titration is quite a bit off in this regression possibly due to the exclusion of the rapid progressor. With MMRM, the best results appears to be by excluding the 1 mg/kg arm. Here, y= 3.7x + 0.03. R^2 = 0.957 . I realize that it is not appropriate to simply take whatever result is more favorable, though for some reason ANCOVA reported a value for 1 mg/kg of ~-0.20 and MMRM a value of ~-0.06. The MMRM value does not seem plausible. There might be another issue where even a single patient can make a large change in the result. Probably good to simply present these different outcomes and acknowledge that there are a range of possible results. What however is again quite notable is how strong the linear correlation is for either of the above scenarios. Instead of concentrating on the random noise of the trials, it can clearly be seen that when more dose is applied --> more amyloid is removed --> larger cognitive benefit.

This is seen time and time again in the results. This is what the totality of the datasets from the anti-amyloid trials demonstrate. The implication which should be made explicit is that any trial that does not follow this strongly established pattern is the outlier; the strongly established pattern is not the outlier. Yet, this is the argument that is made in the FDA Briefing Document. 302 high dose is portrayed as some sort of chance event which is of equal likelihood as the 301 high dose strong outlier. From all the figures that have been presented here for the different anti-amyloid trials this is obviously untrue. 301 high dose should not posteriorly have been weighted the same as 302 because it is so far off from the totality of the evidence. It is surprising that such a highly aberrant result could ever have been presented as anything other than the strong outlier that it is.



"The results presented by the Applicant are an accurate representation of the prespecified
analyses for Study 103. An MMRM model (the same primary analysis method that was used
in Studies 301 and 302), which more appropriately handles dropouts, with the same covariates
as the ANCOVA model specified in the SAP was used as a supplementary analysis and
provided similar estimates of the treatment effect for the 10 mg/kg treatment arm for both
CDR-SB and MMSE.

Of note, the titration arm of Study 103 has a nominal p-value of 0.0432 using ANCOVA.
However, the MMRM analysis of the titration group is not significant (p=0.2044) likely due
to the inclusion of a titration group dropout with a 6.5 point worsening at Week 26 which is
excluded from the Applicant’s by-Visit ANCOVA analysis of Week 54.

The titration arm in Study 103 is not the relevant dose to compare to the high dose arm in
Study 302 because the titration period was much longer in Study 103 (44 weeks) compared to
Study 302 (24 weeks). Subjects in the titration arm in Study 103 received 3 doses of 10 mg/kg
compared to 14 doses of 10 mg/kg in the high dose arm in Study 302. In fact, the reduction of
brain β-amyloid for the titration arm in Study 103 at Week 54 was more similar to the lower
dose in Study 302."

Page 90 FDA Briefing Document

Page 214 for the MMRM


That is the study by study regression result for the lecanemab phase 2 and the aducanumab 103. Both of them have very highly correlated results in the regression.

This makes me wonder why the trials do not have sub-grouping within the trial arms. Instead of having one BIG point that they report for each arm with very good balancing within each arm, why not have balancing (if possible) for every 20, 30 or 50 patients within each treatment arm? This might allow for a more granular look at the results, perhaps with the ability to reach out to farther SUVR values. As it was with 301/302 the ends of the SUVR became almost uninterpretable as the patients were no longer balanced for covariates. All we were shown then were individual level correlations which lacked statistical validity. Sub-groups would give a better view. There could also be some pre-specified program that could combine them into larger groups.
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 »

All Amyloids 1 Edit 1.jpg
Leca Phase 2 All Amyloid 1.JPG
Leca Phase 2 All Amyloid 2 with Placebo.JPG
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 »

Leca Phase 2 ex Placebo 1.JPG
Leca Phase 2 ex Placebo 2 Bayesian.JPG
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 took the figure (first figure two posts up) and extracted the dots for the lecanemab phase 2 trial as presented in the FDA pharmacology review and then applied linear regression to the result. This is the second figure two posts up. What is remarkable is how far off this regression is! The regression origin is at 0.53 with a slope of ~4.2! That is way way off. Even looking at the dots you can see that the line will be far off from the 0,0 origin where we know by definition the placebo must be. Even when we force the origin as in the third figure; it is still far off. A 4.2 slope makes no sense! That would mean that another 0.1 SUVR would give 0.42 more CDR-sb; that is more than the 0.39 from the high dose 302. Creating a strong narrative with consistent patterns allows us to see how wrong the high dose 301 result was; yet with the with regressions given this becomes hidden.

The figures directly above this post are from my own regressions as I made various adjustments (e.g., Bayesian, weighted, "reducing" etc.). The regression (bottom figure of above post) is much closer to the origin and is much more consistent with what is expected with a slope ~1.6 and with high correlation (R^2 = 0.995 , even without placebo!). It would be best in all of these figures if regression equations, correlations, n values are all presented to be transparent and to allow reproducible science to be possible. The previously posted best Bayesian, reduced, and weighted with a placebo point added was even closer to prior expectation.
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 »

All Amyloids 1 Edit 2.jpg
103 Steep.JPG
302 Regression Line.JPG
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 »

Everyone thank you very much for supporting the thread! Perhaps given my quantitative bent I have been tracking the thread visits. Lately, there has been a noticeable uptick in thread traffic. This is somewhat surprising to me as I have continued my near endless posting on what might seem somewhat repetitive posts about the correlations of aducan. J11 will dare to be dull but informative.

I think it is important to keep on posting until I really feel comfortable with how aducan behaves; we are already ~700 posts on this thread-- could be a few hundred more, as long as it takes. Thank you all for showing an interest in the actual patterns of the dataset. I have looked around online and there is a near complete absence in interest in thinking quantitatively about aducan, yet that is where the argument should be won or lost. The opposition simply is not making any meaningful effort to contribute substance to the conversation.

The above post is more of the same. The first figure emphasizes once again how the recent FDA pharmacology review somewhat obscures the nature of the regression line for the lecanemab phase 2. The pink line is a drawn by eye regression for the lecanemab gold triangles. What is quite apparent is that this line is so far off target. The y-intercept is ~0.5 yet we know that 0,0 should be the true intercept. The regressions provided for the lecanemab phase 2 in my previous posts actually shows that lines which are highly consistent with our previous findings can be drawn even right out of the box with no data massage. After making some reasonable data "manipulations" the data falls almost exactly on the line.

One problem with the first figure above is the scaling that is used. The figure has a highly compressed y-axis so everything seems to bunch up near the top left. This scaling underplays the substantial benefits that the treatments can offer those with AD. The middle figure shows what happens when the y-axis scale is elongated. This is the same data as presented in earlier figures for 103 though this time with a steeper line. It clearly looks like aducan is helping much much more, even though it is the same dataset. Here this is the ANCOVA result instead of the MMRM that was used in the FDA pharmacology review (In the ANCOVA the 1 mg/kg point is much more aligned to the regression and the 10 mg/kg point move down to ~1.2). Nevertheless, the middle figure certainly does highlight the effectiveness of aducan in a way that is almost entirely lost in the other figures. Such use of different scaling can be highly manipulative, though in this instance it should be recognized that the bottom of the scale is set to 1.65 which is the expected placebo decline. The elongated figure does not so much manipulate the data as fairly present it. Basically, if the patients could receive 1.65 CDR-sb of benefit from aducan that's that- they would experience no decline-- they would achieve stasis -- a cure? The middle figure highlights how close to this goal they actually were. The upper figure almost completely obscures this finding. The upper figure just has a blob of dots without any actual indication of where the goal posts are and a highly distorting scaling with an extended x-axis and a compressed y-axis.

Final figure is the 302 high, low and placebo doses. I had thought it would be best to put the 301 and 302 results all into one figure. Yet, when I put the 302 results into their own figure I noticed something interesting: The high and low doses fall almost exactly onto a straight line. The 302 low dose slope is almost exactly the same as the 302 high dose slope from the placebo origin. I had thought
(when looking at the other anti-amyloid results) that perhaps the 302 high dose was actually somewhat of an overachiever-- that its "true" results was less than what were reported. Now I am not as sure. The 302 low dose is almost exactly in line with the 302 high dose. We had been seeing regression slopes of ~1.3 for most of the anti-amyloids with aducan we now see a slope of ~1.4 with a fairly large n value.
User avatar
floramaria
Support Team
Support Team
Posts: 1423
Joined: Tue Jul 04, 2017 11:22 am
Location: Northern New Mexico

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

Post by floramaria »

J11 wrote:Everyone thank you very much for supporting the thread! Perhaps given my quantitative bent I have been tracking the thread visits. Lately, there has been a noticeable uptick in thread traffic. This is somewhat surprising to me as I have continued my near endless posting on what might seem somewhat repetitive posts about the correlations of aducan. J11 will dare to be dull but informative.
Thank you, J11, for all of your efforts. I think a lot of us have been following your thread to get deeper insights than we can get from news articles. I really appreciate the time and energy you have put into this and will continue to follow it.
I still question the premise underlying the FDA approval, that removing amyloid will result in cognitive benefits.
Did you look at the thread, " A Replacement therapy May Be Able to Rescue the Brain from Alzheimer's"? the article it links to is interesting and presents an opposing view.
Here is the article from the Cincinnati Public Radio station WVXU:
https://www.wvxu.org/post/replacement-t ... s#stream/0

Here is a quote from the article:
Professor of Neurology Alberto Espay is frustrated the FDA recently approved a drug he says is based on a faulty theory. He says the problem isn't clumps of plaque but missing liquid protein. He says this protein is needed in its original, soluble form to keep the brain healthy.

Sometimes the protein clumps into amyloid plaques but Espay says that's not the problem.

"No matter how high the amyloid gets, if individuals are able to keep a level of the soluble fraction of the product, which is the normal protein, they can do very well. And that is the opposite theory to what's prevailing, what's been with us all along," says Espay.

. What do you think about this approach? Can you see a way that this would fit into any of the models you have been analyzing?
Functional Medicine Certified Health Coach
IFM/ Bredesen Training in Reversing Cognitive Decline (March 2017)
ReCODE 2.0 Health Coach with Apollo Health
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 »

Flora, thank you for responding.

The highly consistent and multiply reproduced finding of a strong linear relationship across the anti-amyloids is very comforting for me. This linear relationship is biologically plausible and creates a strong base to support the validity of the 302 high dose result. At the same time this evidence also helps to establish how aberrant the 301 high dose result was. The 301 high has no connection to the 301 low dose; it is a complete anomaly. In the FDA Briefing Document and elsewhere the question is repeatedly raised: What should you do with one positive trial and one negative trial? You can't just toss the negative trial because it could be the correct answer, while the positive trial might be the incorrect answer. Yes, you can! The 301 trial has no relationship with any of the other evidence; it is an outlier of all the other research. 301 should be the one downweighted because it clearly stands apart from all the other findings.

The size of the treatment effect in different patient populations also stands out for me. The 103 study found quite large treatment effects of aducan of >1 CDR-sb. Similarly large effects were seen in the lecanemab phase 2 in the APOE 4+ and mild AD subgroups.

Large differences were also noted in the aducan extension trial when conditioning on those who had achieved amyloid negativity by week 78 (The figure for this was posted not too long ago on thread). None of these patients demonstrated runaway cognitive decline in the same way that occurred for those who were amyloid positive at week 78. Maintaining high levels of amyloid through time in those with MCI/Mild AD apparently creates a substantial risk for rapid cognitive decline to occur. Arguing that aducan can only have small cognitive benefits is not consistent with the currently reported research.
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 »

With respect to the soluble amyloid idea, yes I read the Alzheon research (and the research that you cited in your post) and it did create some uncertainty for me. The Alzheon perspective is that removing amyloid plaque is largely unnecessary and possibly counter-productive. After going to all this effort and experiencing all of this ARIA-- and then it might not have even been needed? There are some who have amyloid plaque who never develop dementing illness; there is something of a logic gap to ignore this paradox. While amyloid plaque might not be at the direct center of toxicity, removing the plaque on this interlinked amyloid pathway appears to offer cognitive benefit.
Post Reply