Month: July 2018

Analysis by Buzzword

There was a poignant visualization of what science is really like rendered by Stanisław Lem (science fiction writer).  In a story, a strange machine created science; groups of people in white lab coats appearing to read and write papers, quarrel about them, burn them, shout and carrying them above their heads in excitement, all against their detractors etc, etc. (the human element of something abstract)

Science when it percolates to the public, relies on buzzwords, and every new one is better than the one before it.  You have to be deep in the weeds to read papers and understand all the connections and implications which would drive the next set of  papers to be written, so I relied on buzzwords, or rather I ask myself where are their limitations.

On many forums there is a cornucopia of posts citing ANAVEX 2-73, and Precision Medicine together, with admixture of  the resulting euphoria .  I have lived through many “revolutions” in science and industry, which ended with those buzzword being forlorn and missing from public memory after few years.  Do you remember superconductivity in room temperature ( 1980’s ) – Oh! the world was supposed to be so different from the one we have today!

On July 25th, AVXL revealed that some common variations of two genes were correlated with response to ANAVEX 2-73.  The variants of gene of related to morphology of SIGMAR1 receptor is no brainer since ANAVEX 2-73 binds to it and any changes can either attenuate the effect or strengthen it.  I do not think that getting to the bottom of interaction between the ligand and drug is on agenda of Dr Missling at this moment – way too costly.

Nobody knows why ANAVEX 2-73 has the effect on SIGMAR1 and that might be a saga of its own.  I do not know whether anybody mapped (3D structure) of the ligand or any drug with affinity to it. The market is yawing because many tried to affect the ligand and got nowhere.  As a consequence,  ANAVEX 2-73 will come as a surprise to the pundits and investing public alike.

Once this happens number of ventures will start looking into interaction of the ligand and drug to copy the interaction, and not the drug, to circumvent the patent.

The second gene called out is COMT, its  variations are connected to tradeoffs between ability to perform under stress and cognition (ah! heterogeneity of human race and evolution!).  The researches at smaller companies spread in all directions, in search of paths less taken, so some pursuit now the connection between mental stress and Alzheimer via inflammation.

In the light of involving in the mix inflammation, (a splendid buzzword by itself) one can only mention that two main approaches to tackling Alzheimer besides current benjamins of Big Pharma; Amyloid β and Tau, which are inflammation and mitochondrial health.

So what is in perfect world Precision Medicine?  A patient’s genomic and phenotype profile is used to make drugs to seamlessly interact with both to restore health. With ~60,000 genes and phenotype multiplying this way high because of epigenetic filter on activation of genotype, I think we are way over our heads, with what current status of medicine can handle.

The effort by ANAVEX to use the genetic analysis is just a foray into complex and vast field Precision Medicine, but somebody has to make first step in thousand step journey.

Please, curb your enthusiasm and be ready to reignite it again!

 

Brain Model for Dummies

I like to look at the brain from bottom up that is from neuronal health through synaptic health then I move my attention to networks (specific parts of brain), and network of local networks (the whole nine yards – your cognition).  Disruption can come at any level and upset the workings.  This being evident at the top, in scores of test like MMSE measuring cognition or quality of daily life.  

Of course this is a poor model of an amateur but I can’t afford a better one without going hours upon hours into literature on this subject.  

Yet the model is general and logical enough to at least hold some limited yet relevant degree of fidelity to the real thing.

 

 

The motley rectangles starting with Mitochondria (at the bottom) refer to levels of increasing complexity of the brain.  The arrow at right points to directions of spreading “disturbance”.  I think it does not require any explanation when it moves from cell to cognition, yet to some degree there is tendency to identify way in which mental stress can influence incidence of cellular damage especially through chronic inflammation. 

Trends and Their Buzzwords

Following big trends are and their buzzword are in bold type:

1. Cellular energy metabolism dysfunction in Mitochondria and Cell Glucose Metabolism.
2. Structural damage by Free Radicals and Omega 3 Fatty Acid Deficiency.
3. Amyloid Plaque and Tau Tangles due cellular metabolism dysfunction  ( due to inflammation?).
4. Chronic Inflammation triggered by infection or cellular damage (many threads in this story).
5. Neural Growth Factors (from stem cell research) affecting neurogenesis and neurotransmitters production.
6. Nutrition and Stress Reduction.

 

Since AVXL investors keep asking themself whether is there a better mouse trap?  Let’s examine the trap makers in each of the therapeutic trends at next insatments

!

it is indeed after July 25th, 2018 and that is when AVXL made claims that ANAVEX 2-73 should perform on average over period of a 1 years with following parameters Δ=+2 MMSE (+4.9 ADCS-ADL) starting from baselineof  >= 20 MMSE.  I do not exactly know how they got to that point but I take it as article of faith (You’ve gotta believe someone).  Of course, that is after dropouts and the unresponsive variants of SIGMAR1 and COMT genes are removed from the ITT population.  This is comparable to the average performance of the STRONG from CTAD, November 2017 presentation. This make me confident enough to be tempted to peek into the future and attempt to calculate the impact of ANAVEX 2-73 on GDP of Unites States.

We know that after 2 years into Phase 2a the STRONG response bifurcated into two groups: 3 patients group whose MMSE scores kept on steadily going up on gentle incline and those who (2 patients) after initial response succumbed to further decline, 1/3 the decline of undruged patients on natural course of the disease. This illustration says almost all.

Y Axis: MMSE (Mini Mental State Examination) Scores

X Axis: Weeks into trial.

It is rather obvious that the SET A improves over time and will “never” (in terms of the study) need to be cared for.  SET B must be evaluated vs. some group that is  those who comprise the rest of Alzheimer patients (natural progress).  The natural decline is slope -3.8MMSE/year (that is excluding those who droptout from study).  After calculating this from SET B slope on the plot this metric is in term of MMSE points over period of year is -1.04MMSE/year.  About 15 MMSE points is the target at which point into consideration comes institutionalizing a patient or 24 hour care. If we start from 24 MMSE score than with normal decline it takes about 2.1 years to reach that point of dysfunction but for those in SET B it is 7.7 years. Since the incidence of Alzheimer disease is exponentially growing with age let’s assume that only 2/3 of that population would be overtaken by morbidity before facing hospitalization or require serious care.

Hence from the AAIC, July 2018 poster we know that 80% of ITT population should respond in this way and that by 2 years only 61% population stays in the trial as the rest have declined so swiftely that they dropout (>15 MMSE level(?)) we are faced with about 50% of overall population sick with Alzheimer being responsive to ANAVEX 2-73 as the model predicts.

From 5.7 million suffers, half is potential responders = 2,850,000

The ones “cured” are 60% of that population = 1,710,000

Saving accrued = 1,710,000 * $30,000/year=$51,300,000,000/year

second branch is (the 40% responding like in SET B) = 1,140,000 * .66 * $30,000/year

= $22,572,000,000 (2/3 only saving in population due to morbidity before hospitalization)

That is about ~$74 billion savings/year

This about $74/$20,000=.37% GDP (  $20 trillion GDP )

I assumed cost of $30,000 which is combination of cost of hospitalization ($87,000) and part time care.  This should give basic insight into the cost, you are encourage to plug in your own numbers.

I already got a pitcher of beer out of this…..Do you really want to keep me sober?

Every time there is a news release the data provided is formated in different way and there are things omitted  so if somebody tries to get to the bottom of it he has to make assumptions.  I have the deep seated impression that he company strategically refrains from spilling the beans on ANAVEX 2-73.  It gingerly threads between making us believe “we have the cure!” and “it is just another Alzheimer trial”.  I have the impression that even Phd readers who have seen too many of these to fail are kept in this camp too.

I have to confess that I tried to get to the bottom of the format of the data presented by AVXL but ultimately gave up hence my remark about Phd part of humanity.  I can infer certain things but only at the peril of making assumptions and even then there are holes in the logic through which one can drive a decent size car, but not a truck.

The only to two parts of this poster I can relate to square with the data from other releases, that is the beginning and the conclusions of the poster.  The rest is like: oh! look green bar is higher than the orange so it must be good! 

 

From the red square on the left, I learned that by second year the number of patients still in trial, since at the data dump after 57 weeks that number was 26, is now 21 patients with one patient an odd man out, which I would identify as patient number 1009, who probably had been was wrongly diagnosed.  The other piece of info is that varying doses are kept into the extension and that there can be patients with doses below 14mg/Kg, that previously had been determined to be minimum therapeutic dose. Trials like this have natural rate of dropouts I used reference given in post (link):  ANAVEX 2-73 Phase 2a vs. Alzheimer’s Progress

In this post I made assumptions to format the data consistent across the reference on Natural Progression of Disease and Phase 2a data.  This bit of information has confirmed correctness of my assumption.  Let us plot this for both trials:

 

Green line gives the number of participants as in the reference study, red line the same for Phase 2a and the purple is projection if the number stay the same till now.

The population ITT (Intent To Treat) is 62% of overall Alzheimer population multiplied by .80 of those still in the trial.  So we are looking at 50% of Alzheimer total population as ITT population for ANAVEX 2-73.

Performance of this population is said to be Δ=+2 MMSE (Avg) @57 weeks(?) by comparison for the six patients (Strong) whose data was released up to 109 weeks in CTAD November 2017 presentation this metric was Δ=+2.5 MMSE (Avg) @57 weeks.  Also The average MMSE score for the predisposed group at start of trail was >= 20 so compared with six patient (Strong) MMSE=22 implies lesser performance since the starting score is a bit lower. See previous post for remarks on these and other aspects of the expected performance of ANAVEX 2-73.

My scratch pad is posted here so you cna check my work.

 

I have plotted the data from CTAD 2017 presentation together with some linear extrapolation from the 57 week data and computed certain averages.  Once you do this you get the instantaneous insight into the data, it is amazing.  Here is the plot:

From top are:
1. High Concentration Ascending (4 patients)
2. High Concentration Initially Better (2 patients)
3. Average for High Concentration Cohort (10 patients)
4. Progress of Alzheimer Disease as reference from previous post
4. High Concentration Descending (4? patients) (?=dropouts)

Lines at 30, 24, 16 and 6 delineate MMSE scores ranges named Mild, Moderate and Severe (from top). 30 is top score called Healthy.

There is plenty of insight here, but I will limit myself to the essential ones.  Is there an indication of efficacy?  When you get the average score for all groups (High Average) in the cohort, and then compare it with referenced Natural Course labeled line you notice that the difference is amounting to ~2.5 MMSE point per year. Over 4 years that is 10 points, this might make the difference for many not to have requiring expensive care. This result includes all groups and encompasses those who are going really fast  below 16 towards 6 points.

The line with label (High Descending) shows the group of patients for whom there is no therapeutic outcome.  The line is extrapolated from one year data and shows quick descent into Severe range, probably dropouts from trial accelerate afterwards.  This phenomenon validates normalizing data in  previous post.

As I had pointed out in previous post there was some selective action on those healthier than the average and it showed in the average score for two other groups. The higher average the stronger effect. If this would be the only criteria, since we don’t know and have no way of knowing because have only partial raw data, there would be no need for genetic and phenotype inquiry and establishing correlations with the effect.  This inquiry will either/and validate the data or discover some needed criteria for ITT (Intent To Treat) population.

If ~50% patients of 5 million will never go to institutions or will show no need of specialized care at home then we look at (3 years stay, $30k/year) about 200 billion savings to taxpayer and families (1.1% of GDP). This reminds me of Louis Pasteur’s invention of rabies vaccine which created so much wealth in France that it defrayed all the cost of war reparations to Prussia after 1870.

In the High Ascending data set I included patient 101009 who improved greatly though did not show high concentration in blood. This could be explained by heterogeneity with propensity of clearing Anavex 2-73 at higher rate than normal which does not preclude therapeutic effect.

One more thing, at one year of  observation the values for decline starting at 21 MMSE score and Standard Deviation for Natural Course of Progress were 17.2 +/-4.3,  and the same for High Concentration Average were 19.6 +/-5.9.  That was 138% of SD for Natural, either we are having just so, as it happens, larger dispertion in data and it should be dismissed, or this discrepancy is due to bifurcation (selective therapeutic action of the drug) i.e. the area on which 68% of data resides is increased.

 

This analysis is just as bad as the rest of the stuff on this blog.  So let’s start with percentages going to different types costs in producing a car as that of the MSRP price of ICE car (sticker price). Here it is…

If we sum up the costs roughly attributed to generally manufacturing like “Raw Mat’l”, “Procured Parts”, “Internally Produced Parts” and Inbound Logistics” as the cost of the parts from which a car assembled then we get 48.7% of sticker price, then add “Manufacturing”, probably meaning assembly.  At that point we look at 61.5%. The German Estimate was $18,000 for parts and $10,000 production cost. This is comparable to ICE sticker price that would be $45,500.  “Product Development and Overhead” in this model accounts for 8.5% if you enter this into the cost structure then you get even 70% to be the cost for “Factory Gated” Tesla Model 3.  Indeed, this is the 30% Gross Margin. Wow, right on the money!  What an accounting feat, such a precision! The other costs include the OEM profit margin (7%), being a Dealer (10.3%) and Marketing (10%).  We are after the Profit Margin so let’s leave it alone for now.  Let’s set being a Dealer makes 10% of the sticker for TESLA, Direct marketing has been outsourced to Mr Hype, Ms. Save The Planet and Mr. Elon Musk twitter account (0%).

At this point profit per Model 3, @ $45,500 sticker price, is $8,950. But, but there is the fly in the ointment …..

If we compare efficiency of $TM Toyota and $TSLA as number of cars per employee then we get this:

$TM  cars made worldwide ~8,900,000 / 364,000 employee = 24.5 cars/employee

$TSLA cars made ~180,000 (assumption, full year) / 30,000 employees = 6 cars/employee

The model above takes into consideration decently run company like Toyota.  People who work are required to eat, sleep and rest but this costs money so $TSLA pays them at least 4 time the money per car Toyota does.  If we identify labor at 8% contribution to all costs and 1 percent of sticker price is at $455.25 (the quarter killed $TSLA) ergo additional cost to TSLA is $10,920 (didn’t I tell you about that quarter?).

In this half-assed (technical term) assessment even before financials are taken into considerations there ain’t no money in it for $TSLA (-1,970), are there any for you?

No wonder, that after exhausting all options to jack the price up Mr. Elon Musk might come up with better and improved Model 3:  Model XXX.  We are waiting with baited breath!

1. ANAVEX 2-73 Phase 2a trial has been discounted by some on the grounds that the sample size produced statistically insignificant results and that there was no placebo arm.
2. Source of data for Natural Course of the Alzheimer Disease can be found in scientific literature and rudimentary comparative analysis can be made of vs. Phase 2a.
3. Using data released for the first year and partial one after the first year, analysis revealed that:

1.  High Concentration Cohort data could signal strong therapeutic effect.

2.  Therapeutic effect might not  be explained away by a fluke distribution claim.

3.  Bifurcation of outcomes in High Concentration Cohort might explain the need for biomarkers to identify ITT population.

The management of AVXL has been criticized and even ridiculed by some over the choice of 32 patient for phase 2a trial of its Alzheimer drug ANAVEX 2-73. The tradeoffs here, were between risky and expensive trial, possibly, bringing the company down in the case of failure (in context of history of 98% trials ending in failure), versus trial aiming to seek putative signals of efficacy, without being statistically significant (standard of the industry) and having the advantage cost to be within the finances of yearly cash burn.  On a lighter note, I always find hilarity in PRs dishing words of some not closely defined “efficacy” together with dash of “statistically significant” to make you swallow the bait for further financing. No, this is not the case, here!

The analysis is done by replotting publicly available data and making assumptions explained in the text. The author has not been trained in statistics so any criticism and pieces of advise are welcome.  I cannot help myself but to quip that if you lower the threshold for things improbable even dragons are possible and stone might strike up a conversation (attributed to Stanislaw Lem).  To draw up on previous sentence, reality is not locked into statistics but shouldn’t be totally out of it, either. 

The crucial information here is data contained in paper Variability in Annual Mini-Mental State Examination Score in Patients With Probable Alzheimer Disease  Arch Neurol. 1999;56(7):857-862. doi:10.1001/archneur.56.7.857.  Link is given here:  https://jamanetwork.com/journals/jamaneurology/fullarticle/775209 

Natural Course of the Alzheimer Disease presented in the study can be best visualized as a bar graph with 3 classes of patients: Those who did not decline up to this moment (No Decline): Those who declined less than 3 points on MMSE scale in any year followed; classified as insignificant decline (Decline<3): Those who keep declining more than 3 points in a given period (Decline>3).  Since the data is formatted in that way, almost identical classes has been established from raw data of ANAVEX 2-73 study.  The raw data can be found on pages 23 and 29 of Corporate Presentation first released at CTAD 2017. 

To visualize Natural Course of the Alzheimer Disease see the first bar graph below. The trial starts with 372 patients, after first year there is 6% dropout, after 2 years the dropout is 39%. The dropout is mostly due of to rapid and profound deterioration. The class of patients No Decline quickly diminishes in numbers, and after 4 years they are nonexistent. Since human beings vary greatly in genotype and phenotype the variation of MMSE scores between individuals follows. The average MMSE score of the patients starting and staying on the trial can not be precisely verified since there is no raw date on that, but can be approximately taken to be around 20 MMSE. For the ANAVEX 2-73 study this was 21 MMSE, so these are comparable studies.

Illustration of Natural Progression Of Alzheimer Disease

 

Illustration of Progress of Disease in High Concentration Cohort of ANVEX 2-73

Since the data is limited to two years for the only cohort which displayed therapeutic effect (High Concentration Cohort) the comparison can be conducted so far.  Also, the number of people dropping out in the cohort were adjusted to mimic those in the Natural Course of the Alzheimer Disease.

These assumptions allowed to have the cohort to start initially with10 people, reaching at year one 9 (10% dropout), and again at the end of second year coming down to 6 (40% dropout rate at  year two).

Clearly, the graphs allow for direct visual comparison as they are almost identical in format.  Over first two years of the ANAVEX 2-73 trial there is an increase in probability of a patient, for those still staying in the trial, to be counted in the two classes (colors; yellow and green) defining the therapeutic effect versus Natural Course of the Alzheimer Disease.

When there is no putative therapeutic effect for given patient he is counted in the color blue Decline>3 class.  The presentation of results defines the probability of a patient at the start as 100% (certainty) to find himself/herself in the No Decline class and then splits the 100% into probabilities for the three classes. The goal is to show larger swats of yellow and green vs. blue.

Probability for patient to find himself/herself in one of three classes as long as he/she stays in trial – Natural Course of the Alzheimer Disease.

Probability for patient to find himself/herself in one of three classes as long as he/she stays in trial – High Concentration Cohort ANVEX 2-73 Phase 2a 

As we are not informed on the dropout rate at 2 year end of Phase2a ANVEX 2-73 study we took liberty to assume that that is the same as for the Natural Course of the Alzheimer Disease, with this assumption the difference in probabilities for both studies are easier to be visually recognized.  The putative therapeutic effect can be measured in the width of No Decline, and in limited capacity in Decline<3.  This suggests expectations of a strong signal of efficacy in standard industry trials. 

The efficacy of ANAVEX 2-73 visualized here, might be dismissed by claiming that we are falling victims of skewed distribution of outcomes due to small sample size producing what is called statistically insignificant results, questioning their degree of validity.  Logically, had this been the case for the sample size one would have expected skewed distribution on the other branches of the study for ANAVEX 2-73 which did not shown therapeutic effects.  If this is the case then the argument gains on validity.  Refuting the argument bring us closer to raising the degree of signal validity even with the limited sample size.

Average Declines for Populations in AD trials over one year (see legend above)

The lines for Natural Course of the Alzheimer Disease and Low+ Medium Concentration are almost identical so are their distributions so by extension Low+ Medium Concentration exhibits natural for AD distribution disturbed by any other effects.  This observation would bring increase confidence in existence of signal of efficacy in High Concentration Cohort.  

The High Concentration Declining patients decline at the rate of 165% of Natural Course of the Alzheimer Disease. There might be suspicion that ANAVEX 2-73 besides making some better can also make some worse, that might be called negative synergy between ANAVEX 2-73 and AD hence the efforts by the company to identify ITT (Intent To Treat) population to eliminate those who could potentially be harmed in Phase 3.  On the other hand the cohort seem to consists of two subsets; one of people who are healthier than average with higher initial scores who might have benefited, and on other side those below average that can not be rescued due to selective action of the drug and are naturally declining at higher rate.  This divide might be responsible for the bifurcation.  The clue can be seen in the fact that those who were helped by ANAVEX 2-73 (High Concentration Advancing) are having their initial average scores at 22.5 MMSE vs. for All Cohorts that is 21 MMSE.  

The delay with FDA go ahead might be caused by the need for some data to assuage regulator’s fears of realization of negative synergy between ANAVEX 2-73 and AD in some patients treatment. 

The hypothesis that the bifurcation is rather a natural phenomenon due to selective action of the drug, either before point of no return for patients with Alzheimer disease, or synergy with genotype/phenotype combination to trigger the drug action, could be vaguely supported or refuted by comparing average score for only High Concentration Advancing necessary to bring the High Concentration (whole cohort average) (High Concentration Declining’s score is known) to decline with the rate of Natural Course of the Alzheimer Disease. This turns out to be +1.33 MMSE per annum (vs. +2.53), close but not enough to be within first Standard Deviation (-3.8 avg. +/-4.3) of Natural Course of the Alzheimer Disease.  It might be that there is more at play than just higher MMSE score connected to point of no return.  In general, the author lacks analytical skills and data to move beyond this set of concepts, both should be found in presentation on results of genetic and phenotype patient signatures collections and analysis helped by Artificial Intelligence (automated due to overwhelming number of connections) and scheduled to be released at AAIC 2018 Chicago on July 25th, 2018.

The delay with FDA to go ahead with Phase 3 might be caused by the need for some data to assuage regulator’s fears of realization of the negative synergy between ANAVEX 2-73 and AD in some patients treated with ANAVEX 2-73, but this is only a wild conjecture. 

I am convinced that such population has been found, hints could be found in mentions by the company of some genes whose specific expression is connected to prevalence of sigma 1 receptors on the cell walls.  I should wrapped up with short conjecture on prevalence of ITT population.  Out of 9 patients 5 responded or rather tentatively can be counted as such.  If the High Concentration cohort had followed normal distribution it would suggest that combined factors of low MMSE initial scores and other factors might make it quite common to place patients in ITT population.  Ultimately, this can drive the Phase2b/3 to resounding success in the future and make ANAVEX 2-73 shoe in for next Standard Of Care status.

Conclusions 

1. There are reasons to conclude that Phase 2a points to strong efficacy in industry standard Phase 3 trials.
2. Expected existence of genomic and/or phenotype signature for responders validates claims for efficacy even in light of small sample size.