A novel Alzheimer’s drug is being tested in clinical trials in USA. The US company Cassava Sciences is ready to earn billions of dollars, convinced to have the game-changer, verified in the lab and in soon in clinic. And then concerns are publicly raised with FDA about falsified preclinical research data with that same drug, funded by that same company, done in the New York lab of their long-term board member, Hoau-Yan Wang. Cassava retaliated with a full–blown attack on its critics via a press release. Which provoked Elisabeth Bik to have a look as well.
On 23 August 2021, the US medical regulatory authority FDA posted this document, a kind of “Citizen’s Petition” by a law firm addressed to the FDA:
Statement of Concern Regarding the Accuracy and Integrity of Clinical and Preclinical Data Supporting the Ongoing Clinical Evaluation of Compound PTI-125, Also Known As Simufilam re Citizen Petition from Labaton Sucharow
This happens when concerned investors employ a law firm to investigate the evidence of research misconduct and get their money back from the accused company. For example, something like this is currently happening to Athira Pharma:
Here, it is about the company Cassava Sciences, founded by its CEO Remi Barbier. The law firm Labaton Sucharow acted apparently on some internal whistleblower information, because there was no PubPeer evidence at that time, and engaged a scientific image integrity expert to verify and summarise the concerns. Part A of the lawyers’ report is an Executive Summary:
“For over 15 years, Cassava Sciences (previously Pain Therapeutics, Inc, PTI) has funded the lab of Dr. Hoau-Yan Wang at City University of New York (CUNY). Together with Dr. Lindsay Burns at Cassava, Dr. Wang has published nearly a dozen papers connecting Filamin A protein with pain and Alzheimer’s disease (AD).
Cassava Sciences created a drug candidate called simufilam (previously PTI-125) that they claim binds Filamin A and has beneficial effects in biochemical and animal models of AD. The studies from Drs. Wang and Burns discussed in this dossier were used by Cassava Sciences to garner NIH grants and to open an investigational new drug (IND) application to study simufilam in AD patients. They form the basic science foundation for two completed clinical trials (phase IIa and IIb) which exposed over 70 patients to simufilam. Cassava Sciences is
currently recruiting 200 additional patients for a follow-up open-label trial.
This report raises concerns about the quality and integrity of the laboratory-based studies surrounding this drug candidate.
To preface the analysis that follows, no other labs have confirmed this research connecting Filamin A to pain or AD. No other labs have confirmed that simufilam binds or modifies Filamin A or has effects in AD models.”
In Part B, the lawyers explain that their sources are public (i.e., published research papers) and that
“This letter details a long-standing pattern of seemingly intentional data manipulation and misrepresentation in scientific papers and corporate disclosures authored primarily by Drs. Hoau-Yan Wang, Associate Medical Professor, City University of New York, and Lindsay A Burns, Sr. Vice President of Neuroscience at Cassava Sciences.“
Hoau-Yan Wang is member of Cassava’s advisory board, according to the lawyers for over 15 years already.
What follows soon after, is Part C, where evidence of research fraud is presented. Mostly, it’s about falsified western blots. This was the first criticised paper:
H.-Y. Wang, E. Friedman, M.C. Olmstead, L.H. Burns Ultra-low-dose naloxone suppresses opioid tolerance, dependence and associated changes in mu opioid receptor-G protein coupling and Gbetagamma signaling Neuroscience (2005) doi: 10.1016/j.neuroscience.2005.06.003
Cassava immediately issued a rebuttal to the Labaton Sucharow accusations via a press release, point-by-point retorting to “Fiction” with “Fact”. To me it would kind of make sense actually, if one swaps the Fiction and Fact labels. For example, regarding the western blots like the above, Cassava declared:
“Fiction: Western blots data are identical, more evidence of image manipulation.
Fact: The Western blots bands shown in the allegation are control bands. Control bands are supposed to be highly similar (since they show equal amounts of protein between lanes). Bands show clear differences when expanded. In addition, image manipulation of control bands makes no sense since these would not change the end data.“
Well, that is an old and tired “it’s just controls excuse”. Without controls, a scientific experiment (or a clinical trial) becomes largely meaningless. Imagine Cassava would sell a weight loss supplement instead of an Alzheimer drug. Their advertisement shows a Before and After photos, on the right a lean muscular man, on the left the same picture of same man just photoshopped to make him look fat. What? You aren’t interested in the supplement anymore? It’s just controls! The two men are supposed to be highly similar! Look, the fat guy shows clear differences when expanded. And anyway, manipulation of control male models makes no sense since these would not change the end data, right?
Elisabeth Bik confirmed the concerns reported to the FDA, and challenged by the Cassava statement, searched herself and then found more, much more. She later published a blog post detailing some of the new evidence.
Blue boxes mark very similar panels which differ by brightness only while representing different experiments. On top of that, Bik spotted that the the Gai panel in Figure 2 looks fishy. It does indeed appear that bands were digitally pasted or selectively manipulated otherwise in Photoshop.
Then there were even more duplicated gel bands. Where the Labaton Sucharow letter reported “Telltale signs that the Gα bands in Figure 5a likely come from different gels“, Bik followed up on the “Additional Suspicious Western Blots” in Section E and found:
- “Blue and green boxes: It was noted in the report mentioned above that some of the “double decker” bands look similar to each other. The resolution is low, but maybe the authors could take away these concerns by showing the uncropped blots, if they still are available?
- Pink boxes: An additional concern not noted in the report is that two of the G-alpha bands, representing seemingly different experiments (Morphine vs NLX), look quite similar, while their surrounding bands are very different.
- Orange arrows: some artifacts around bands suggestive of splicing.“
That paper doesn’t have a conflict of interest section, apparently back in 2005 this concept was not invented by Elsevier yet.
As mentioned above, that Wang et al 2005 paper contained data later reused in Wang et al 2010:
Hoau-Yan Wang, Kalindi Bakshi, Changpeng Shen , Maya Frankfurt, Caryn Trocmé-Thibierge, Philippe Morain S 24795 limits beta-amyloid-alpha7 nicotinic receptor interaction and reduces Alzheimer’s disease-like pathologies Biological Psychiatry (2010) doi: 10.1016/j.biopsych.2009.09.031
That study with a proprietary drug was funded by another pharma company, Servier, which also paid Wang as consultant. Bik found that paper was also falsified:
“Figure 1A appears to show a lighter background around some, but not all bands in the top a7nAChR blot. A similar lighter area is visible around all bands in the lower blot.“
Such things happen when a gel image gets assembled or otherwise modified in Photoshop, with the purpose of data forgery.
Moving on to the next paper:
Hoau-Yan Wang, Maya Frankfurt , Lindsay H. Burns High-Affinity Naloxone Binding to Filamin A Prevents Mu Opioid Receptor–Gs Coupling Underlying Opioid Tolerance and Dependence PLoS ONE (2008) doi: 10.1371/journal.pone.0001554
The paper mentions as conflict of interests that “This study was supported by Pain Therapeutics, Inc. and LHB is an employee of this company“, but Wang forgot to mention that he was paid by Pain Therapeutics (now Cassava) as advisory board member.
The lawyer’s report criticised gel splicing and added “Figure 7a of that paper appears to show two IDENTICAL panels (red arrows) for what are reported as different experiments.”
Cassava rebutted such evidence with this statement:
“Fiction: Cassava Sciences’ Western blots data appear overexposed and highly processed, evidence of image manipulation.
Fact: High quality bands are supposed to look sharp6. Smudged bands can be evidence of inexperience, depending on levels of protein in the band.“
I love this. If you gels lack sharp edges around the bands, it’s because you are an inexperienced loser. This pigheaded arrogance is breathtaking, do they really think everyone else is as scientifically illiterate as they are?
The next discussed paper about curing Alzheimer’s with Filamin A was this:
Hoau-Yan Wang , Kalindi Bakshi , Maya Frankfurt , Andres Stucky , Marissa Goberdhan , Sanket M Shah , Lindsay H Burns Reducing amyloid-related Alzheimer’s disease pathogenesis by a small molecule targeting filamin A The Journal of neuroscience (2012) doi: 10.1523/jneurosci.0354-12.2012
At least here, conflicts of interests were declared: “L.H.B. is an employee of and H.-Y.W. is a consultant for Pain Therapeutics, Inc.“. But not the massive fraud in the presented data!
The Labaton Sucharow letter explains:
“In Figure 1a, the four Filamin A bands in the top set are more similar to each than can be expected by chance and appear to be duplicates. The images at right are magnified, showing that the pixels containing the bands are essentially identical. Additionally, the blots are not aligned and the spacing is irregular. Because FLNA is a large protein (~290kDa), it does not migrate in the gel very far; therefore, this degree of misalignment is suspicious. Moreover, the thin white halos surrounding each band are concerning. There are optical reasons why a halo (or ringing artifact) could occur, but this artifact is most common when components from multiple images are combined using photo editing software. This halo artifact is more prominent in the questionable blots, and extends in some cases into the frame around the blot which is hard to explain as an
This was Cassava’s retort:
“Fiction: “Halo” effects in certain bands indicate fraud.
Fact: A “Halo” effects in certain bands is a direct result of very dense dark loading control bands.7
Fiction: Unusual looking bands on Western blots were pieced together from multiple sources.
Fact: Proteins can and do stick to the side of a lane and migrate that way, resulting in ‘candy-wrapper’ appearance or other fictional images.“
It wasn’t the “loading control bands” which have a halo effect, even outside the gel. And the “‘candy-wrapper’ appearance” derives from digital gel band splicing, where some kack-handed fraudster accidentally cuts into neighbouring bands. But yes, I agree, the images presented in those papers are “fictional“.
Bik found more, like the completely fake western blots of IRbeta in Figures 6A and 6B, which were most obviously assembled in Photoshop, sometimes from copy-pasted bands:
Figures 9A and 11 A proved to be fraudulent disasters also:
Those above are fake western blots, made with one single band copy-pasted several times, as the arrows highlight. To add some more colours to that misery, Bik found a recycled immunohistochemistry image, to stand in for different antibody stainings of different experiments.
On to the last paper criticised in the lawyer’s letter to FDA!
Hoau-Yan Wang, Kuo-Chieh Lee , Zhe Pei , Amber Khan , Kalindi Bakshi, Lindsay H. Burns PTI-125 binds and reverses an altered conformation of filamin A to reduce Alzheimer’s disease pathogenesis Neurobiology of Aging (2017) doi: 10.1016/j.neurobiolaging.2017.03.016
The Figure 12 of that Wang et al 2017 paper contains a skilfully falsified western blot. Impressive is that Cassava sees no problem with same gel having either 12 lanes or 13, depending which protein is shown:
In Figure 3, Bik reported more:
“One of the bands representing a 10-month sample, in the right blot, appears to be surrounded by a rectangle of a different background than the rest of the blot.“
That is because that gel band was digitally inserted in Photoshop, and/or selectively modified inside the rectangle area.
Bik then found similar issues in Figure 10B, gel bands digitally inserted, likely to once again to replace undesired real results:
Also this study’s mouse behavioural data in the so-called Y-Maze experiments was questioned, because “if the result were legitimate, the drug treatment changing the mice’s behavior to closer to 50% spontaneous alternation (i.e., closer to random) would be more accurately interpreted as evidence of worse memory performance.“
Cassava reacted with:
“Fiction: Changes in the Y-maze test for transgenic mice could be interpreted as a decline in cognition.
Fact: A panel of independent, peer-reviewers believe these changes represent an improvement, along with significant improvements in two other behavior tests.“
Which basically means, haha, it doesn’t matter if our results make sense, they passed peer review so you are legally not allowed to criticise them anymore. I guess same applies to the Photoshop fabrications?
Indeed, that paper, as well as the above Wang J Neuroscience 2021 plus Wang et al J Neuroscience 2009, all passed peer review, despite being biologically improbable, as the Labaton Sucharow letter explains:
“In one key line of experiments, the authors report that this entire mechanism can be observed in post-mortem human brain tissue from subjects with
Alzheimer’s disease and neurological controls. […] In these experiments, post-mortem human brain tissue is warmed from -80°C to -20°C and chopped into 200micron x 200micron x 3mm blocks with a McIlwain chopper (as a side
note, a McIlwain chopper doesn’t effectively cut frozen tissue). The resulting chopped tissue is treated with β-amyloid and the experimental drug for 1 hour. They then report a massive increase in tau phosphorylation (modification of the tau protein by enzymatic addition of a phosphate group to the protein; up to 10 fold) from β-amyloid treatment in untreated samples; and that tau phosphorylation was blocked by addition of PTI-125. It is unlikely that the enzyme
responsible for phosphorylation would survive the initial -80°C freezing step. Moreover, the phosphorylation experiments are reported to have been performed at 4°C, but it is unlikely that the enzyme responsible for phosphorylation would be active at 4°C (enzymes generally work best at body temperature—37°C)“
Indeed, valid points, which suggest either that the peer reviewers were either not reading the paper they approved, or were incompetent and scientifically illiterate themselves. But the situation was even worse:
“The methodology for the post-mortem human brain experiments among the three studies are virtually word-for-word identical. The age and post-mortem interval for the groups of subjects are the same (down to the decimal points) in each of the three papers. It is therefore reasonable to assume the same human brain specimens were used across the studies from 2008-2017, so the results are premised on the enzymes in the human brain extracts remaining active up to 13 hours post-mortem before freezing, remaining active after nearly 10 years in frozen archival without any advanced cryopreservative techniques, and being active at 4°C.”
The company retorted with:
“Cassava Sciences is not aware of an industry-wide ‘expiration date’ on human post-mortem brain tissue that is properly collected, processed and stored.“
I wish they would instead explain how they made the native brain enzymes work under such conditions though.
Those were not the only problematic papers by Wang. Here one with the company’s founder and CEO Remi Barbier and the Chief Medical Officer Nadav Friedmann as co-authors:
H.-Y. Wang, Z. Pei , K.-C. Lee , E. Lopez-Brignoni , B. Nikolov , C.A. Crowley , M.R. Marsman , R. Barbier , N. Friedmann, L.H. Burns PTI-125 REDUCES BIOMARKERS OF ALZHEIMER’S DISEASE IN PATIENTS The Journal of Prevention of Alzheimer s Disease (2020) doi: 10.14283/jpad.2020.6
Bik criticised that the paper was peer reviewed in just 6 days. And she found a problematic gel in Figure 3A:
- “Slanted blue arrows: Some bands appear to float over the background, surrounded by a white halo, and with a sharper appearance than other bands
- Pink arrows: The right-most band in the pT231Tau lane appears to be surrounded by a rectangular box with a different background appearance as the other lanes.”
Jay J. Paquette , Hoau-Yan Wang, Kalindi Bakshi, Mary C. Olmstead Cannabinoid-induced tolerance is associated with a CB1 receptor G protein coupling switch that is prevented by ultra-low dose rimonabant Behavioural Pharmacology (2007) doi: 10.1097/fbp.0b013e3282f15890
Bik found problems with Figure 4A.
- “Orange boxes: Two of the double decker G alpha bands appear to look similar, if one of them is flipped horizontally. Note a slanted stripe in the middle of the top band.
- Green boxes: Two other double decker bands in the bottom left blot look similar
- Teal boxes: Two single bands in different G alpha blots look similar“
Presumably, if your western blots lack duplicated bands, it is the sign of your “inexperience”.
Kalindi Bakshi, Raminder Parihar, Satindra K. Goswami , Melissa Walsh , Eitan Friedman, Hoau-Yan Wang Prenatal cocaine exposure uncouples mGluR1 from Homer1 and Gq Proteins PLoS ONE (2014) doi: 10.1371/journal.pone.0091671
Bik found two figures very skilfully falsified. For example, Figure 2A.
“In the Homer 1 / Frontal cortex blot, unexpected sharp horizontal and vertical background transitions appear to be visible. Of particular note, the top band in the right-most lane appears surrounded by a rectangular box of a different background. Shown with blue arrows.“
Finally, one very new Wang paper (The authors declare no competing financial interests) was flagged by an anonymous PubPeer commenter:
Gavin M. Meade , Lily S. Charron , Lantz W. Kilburn , Zhe Pei , Hoau-Yan Wang , Siobhan Robinson A model of negative emotional contagion between male-female rat dyads: Effects of voluntary exercise on stress-induced behavior and BDNF-TrkB signaling Physiology & Behavior (2021) doi: 10.1016/j.physbeh.2020.113286
Section E of the lawyer’s letter to FDA is about “Six Additional Areas of Concern“, discussing “suspicious” scientific claims and flagging “Additional Suspicious Western Blots“, the latter largely confirmed by Bik.
Part D asks FDA to suspend Cassava’s scheduled phase 3 clinical trial NCT04994483 with 750 participants, and to audit all past clincial trials this company performed. NIH and the New York University are asked to investigate Wang’s research, the journals are invited to retract the falsified papers.
The Labaton Sucharow lawyers also discuss irregularities in the trial data, like this:
“it seems that the primary biomarker data set we have with simufilam in Alzheimer’s disease that was entirely produced and finalized by an external lab found that the drug had no effect on clinical biomarkers. Cassava replaced this with a reanalysis that was finalized by an academic lab (presumably Dr.
Wang) and showed that simufilam showed remarkable benefit. Second, plasma biomarker data from these same patients, which were just presented by Cassava Sciences, contains evidence of manipulation. If there’s no biomarker signal, and there is apparent misrepresentation of clinical data the continuation of the ongoing Cassava trials may put patients at risk without the claimed evidence of biomarker benefit.”
Cassava’s reply to it sounds like what Chinese cheaters post on PubPeer blaming an “external company”:
“Fiction: Biomarker data is generated by Cassava Sciences or its science collaborators and therefore are falsified.
Fact: Cassava Sciences’ plasma p-tau data from Alzheimer’s patients was generated by Quanterix Corp., an independent company, and presented at the recent Alzheimer’s Association International Conference1.“
Problem is, Quanterix was not prepared to play the scapegoat and was quoted:
“In Friday’s statement, Quanterix said it is “widely recognized for its commitment to business integrity and to upholding the highest standards of quality.” Quanterix’s stock rose 4.5% in premarket trading, while Cassava shares tumbled 21.2%”
Apparently unconvinced by Cassava’s explanations, The City College of New York opened a research misconduct investigation of Wang’s papers. Tony M. Liss, Provost and Senior Vice President for Academic Affairs, confirmed it to me:
“The College takes accusations of research misconduct very seriously. When such an accusation is made the College Research Integrity Officer follows the CUNY Policy Regarding the Disposition of Allegations of Research Misconduct . Until that process is completed, we are unable to comment or respond to inquiries.“
Maybe this is why Cassava started to distance themselves from Wang in their next statement from 27.08.2021:
“No, the preclinical and clinical foundations linking Filamin A to Alzheimer’s disease do not derive only from the publications of Drs. Wang and Burns.“
As if that proves anything. The Seeking Alpha article (allegedly written by a “Joe Springer”, if you believe that) lists several research papers on Filamin A (who gave those to Joe?), and then warns: “Trials are only halted for safety, and simufilam is shown safer than placebo, there is no chance the FDA will act on the petition’s requests.“
Apparently they think research fraud is not a reason to suspend a clinical trial.
The above mentioned Athira Pharma placed their own founding CEO on leave after Leen Kawas’ papers were reported for data manipulation. But Cassava Sciences keeps lashing out at critics instead of dismissing Wang and Burns. This pharma startup seems to be at best utterly ignorant how science is done properly, and at worst, well, you saw the gel pictures and Cassava’s press releases. Not really reassuring to the investors, no wonder that the stocks are in freefall. What will happen to the clinical trials now?
On 29 July 2021, Cassava proudly announced interim results of an open-label clinical trial with Simufilam (PTI-125) with 150 participants. The trial is described to have started in March 2020, and was not only unblinded, but it also didn’t have a control arm (which makes that whole trial rather pointless).
The description of the NCT04388254 trial, which somehow managed to be Open-Label (unblinded) and quadruply blinded, as well as both with and without control arm, is confusing at best (hence my correction of the previous update):
“Approximately two hundred (200) patients will be enrolled into the study. All participants will receive open-label simufilam 100 mg b.i.d. for a year. At Month12, participants will be randomized (1:1) to continue taking simufilam 100 mg b.i.d. or to be switched to placebo for 6 months. At Month 18, all participants will enter a final 6-month treatment period of open-label simufilam 100 mg b.i.d.“
The switch from “A 12-Month, Open-Label Safety Study of PTI-125” to “Simufilam Followed by a 6-Month Randomized Withdrawal and 6 Additional Months Open-Label” happened on 30 July 2020. Also, the number of patients to be recruited jumped from 100 to 200, but now the press release changed it to 150: “The open-label study has completed its target enrollment of 150 subjects“. It all looks like an insane trick to somehow turn something extremely biased and unreliable into a gold standard of a blinded randomised controlled clinical trial.
The now announced open-label trial was founded by NIH in 2020 (grant #1R44AG065152-01) as a continuation of the previous open-label trial by Cassava NCT03748706, which took place March-May 2019 and treated 13 patients. Its results were published by our trustworthy friends Wang and Burns (the latter is listed as principal investigator of both the old and the new open-label trial):
Wang HY, Pei Z, Lee KC, Lopez-Brignoni E, Nikolov B, Crowley CA, Marsman MR, Barbier R, Friedmann N, Burns LH. PTI-125 Reduces Biomarkers of Alzheimer’s Disease in Patients. J Prev Alzheimers Dis. (2020). doi: 10.14283/jpad.2020.6
I apologise for the confusion in the update before, I never saw a quadruple-blinded but open-label while one-arm but placebo-controlled trial with 150 or maybe 200 participants before 😉
Elisabeth Bik wrote a follow up, and she found several inconsistencies in the Cassava clinical data from the contested phase 2 trial NCT04079803 and its results presented as poster Wang et al at a conference in July 2021. Bik notes:
“The poster states that 64 patients were included in the study and randomly assigned to three groups. There are no details given about how many patients were assigned to each group, but we would expect roughly 20 people per group. Yet, Figure 4 only shows 20, 15, and 17 data points for the placebo, 50 mg, and 100 mg Simufilam groups, respectively. That is a total of 52 patients. What happened to the 12 other patients?“
Well, maybe these patients were all misbehaving outliers, you know… Look, two of them left the placebo group eventually, one disappeared, the other felt so well that participant joined the treatment group:
“While Figure 4 shows 20/15/17 (sum=52) data points for the three groups, Figure 5 appears to show 18/15/18 data points (sum=51). The numbers of the placebo and 100 mg treatment groups therefore do not match.“
Actually, the patients walked in both directions. Those whose Alzheimer’s decease worsened, got up and joined the placebo group:
“Of particular note, in Figure 5, the P-tau181 value from a patient in the 100 mg treatment group went up from ~2.1 to ~5.2, which is an increase of 150%. In other words, one of the patients in the 100 mg treatment group showed a large increase of the Alzheimer’s disease biomarker . Yet, this particular data point is missing in the 100 mg treatment group in Figure 4. Instead, the 150% increase data point is included in the placebo group.“
Watching all that pathetic data fudgery is PAINful, pun intended.
A reader pointed me to this preprint with same trial results:
Hoau-Yan Wang, Zhe Pei, Kuo-Chieh Lee, Yaneicy Gonzalez Rojas, Tamara Doehner, John Puente, Patrick Sciara, Brian Beck, Evelyn Lopez-Brignoni, Boris Nikolov, Carrie Crowley, George Ben Thornton, Remi Barbier, Nadav Friedmann, Jeffrey L. Cummings, Lindsay Burns Effects of simufilam on cerebrospinal fluid biomarkers in Alzheimer’s disease: A randomized clinical trial ResearchSquare (2021) DOI: 10.21203/rs.3.rs-249858/v1
According to this source (apparently based on this paywalled STAT News article), Remi Barbier and Lindsay Burns are married. So now we know why Barbier keeps defending the obvious Photoshop forgeries his wife published with Wang:
“I am aware that the allegations have caused a storm of opinions and counter-opinions on the internet around the fine visual features of photograph of certain Western blots. We all know that once a photograph is on the internet, the pixels that make up that photograph can easily be Photoshopped, cropped or otherwise distorted to mean anything you want it to mean.
Furthermore, internet photos are resolution dependent. This means an internet photo can quickly lose quality and look blurry or pixelated, or whatever. I don’t trust the authenticity of photos on the internet, and neither should you.“
It must be a global conspiracy against Cassava which involves everyone in scholarly publishing and academia, going back many years already?
Cassava Sciences never replied to my emails when asked to comment about the growing PubPeer evidence or the clinical trial registration issues.