OK, then, I went to go back to some of my recent posts and for some revisioning.
I could just edit the posts, though it will be much more transparent if I actually
just repost my edits here.
My first egregious error is from my July 08, 8:55 post where I wrote of the probable
decline of the treatment group while they were off treatment after the futility analysis.
My suggestion that these patients could achieve absolute stasis in cognitive scores
was clearly untrue. Even when the patients were at full dose they still did not achieve
stasis. One should expect that once the treatment is stopped then continued decline
would likely resume. However, one interesting note is that the half life of the treatment
is quite long. ~21 days. It is thus possible that patients could actually drift for some time
without immediate cognitive deterioration, thought this would probably not last for
a year (?). Nevertheless, some widening (though probably a smallish amount) in the difference
from treatment to placebo might have occurred during the break. The big item to account
for is the big hidden placebo understatement that would need to unwind at some point.
[Edit: A big question that I have now relates to the nature of the driving force of AD pathology.
In particular, the high dose patients after ~78 weeks of treatment in the EMERGE trial were
had such minimal amyloid levels that they were approaching the limits of detection. If amyloid
has been so throughly removed from the brain, then would tau continue to accumulate? Isn't
amyloid a necessary, though insufficient condition for AD progression? The current aducan
extension should help resolve this question.]
The confusion about the labeling of the time points in the PRIME and the two phase 3s is worth
reiterating. The phase 3s really only treated for 52 weeks. AD trials have almost always typically
been 18 month treatment trials. PRIME showed that at the 24 week mark there was almost
no difference between placebo and treatment. It is something of a mystery as to why at the
52 week of active treatment (week 78 in the phase 3s) there did not appear to be more benefit.
With thousands of patients one might think that such evidence should be highly visible. The problem
seems to have been a lack of placebo decline. At 78 weeks, a 4.5 point decline in MMSE is expected,
EMERGE reported 3.3. 1.2 more MMSE points to the treatment effect would give 1.8; close to the 2.2
figure reported in PRIME.
Instead of a 3 1/3 point decline in placebo over a 3 year period (=10), a more conservative 3 gives us 9.
For the extra weeks to bring us up to week 164, add 0.5. So the placebo decline expected to week
164, is ~9.5 MMSE points. Now instead of a 6 point treatment gain in MMSE, we have a gain of 5.5.
That is still very large. That would be of clearly apparent and highly relevant to those caring
for someone with dementia.
From my first post, I was confused about how large the treatment difference seemed to be between
PRIME and the phase 3s, apparently this related to slow placebo decline.
For my second post I suggested that 2.7/3.3 was a ~19% decline. In fact this is a 22.9% decline almost exactly the
23% figure reported. I only included 2 decimal places from the online app so I might have hit 23% exactly. This
shows how accurate the online tool is for extracting coordinates.
The comment about non-AD patients in the phase 3s is perhaps misplaced. All the patients underwent amyloid measurement and were required to be positive, thus the patients should be ~ true AD.
Using the patient registry idea would have made so much sense.
Being able to characterize patients before they arrived on site would have been a big plus in the phase 3s.
Post Three -- Hmm, I can also take the correlation between the SUVR and CDR-sb for ENGAGE. I was quite impressed with the 98.5% correlation with EMERGE. Yet, I was nagged by the idea that ENGAGE correlation would not be as strong. If so, is ENGAGE even accurate?
Go back to page 15 of the third link of the second post of this thread; right hand side ENGAGE.
Correlation between ENGAGE CDR-sb change and SUVR: -0.97 (p=0.15?). There were only 3 points involved!
This is an apples to apples comparison (i.e., those who received 10 or more injections in EMERGE and ENGAGE). There is nearly 100% correlation between amyloid reduction and change in cognitive score during the course of the trials.
These calculations were based upon large subgroups including 340-550 for ENGAGE placebo, 300-550 for EMERGE placebo
and 100-120 ENGAGE with sufficient high dose treatment and 130 150 EMERGE with sufficient high dose treatment. The claim that ENGAGE did not show efficacy is entirely unsupported by ... the evidence. If it is reasonable to group all the data points into an average and then find the correlation, then knowing the level of amyloid removal nearly perfectly predicts the decline in CDR-sb.
In fact, one could use the amyloid decline from either of the phase 3 trials and with near perfect foresight predict
the cognitive scores in either trial. The predictor is invariant to treatment sample: a near universal regression law
of amyloid neuropathology for AD. It probably can also be derived from the PRIME trial. Such a strong predictor
definitively proves that the trials were a success. The top line numbers for ENGAGE were largely only noise.
The logic is not entirely closed, though it is close. There could be 100% correlation and yet no significant treatment effect. Yet, we know that there was a significant treatment effect for EMERGE. That should do it.
A final note: strangely, even the placebo decline for the CDR-sb apparently is off. From my perusal of the evidence, the typical placebo decline for the CDR-sb at 78 weeks is 2.7! On page 15 on the right look up -- way up. 2.7 is off the chart! That is where the placebo is expected to have been. Apparently the CDR-sb placebo was just as far off as the MMSE placebo. This true placebo decline should probably be included with these figures. The "true" p-values must be massive for both trials. I want to double check this, though it does seem that the expected placebo result at week 78 does not even fit onto the chart.
Last edited by J11
on Sun Jul 12, 2020 8:59 am, edited 3 times in total.