The petition against Cassava Sciences was supplemented on Sep 8. We shall examine some questions arising from this supplement and provide our comments in this post.
Concern #1: Authenticity and Availability of Phase 2a Data—Western Blots
The petition claims the original western blot images should be available at Cassava’s premises
Can someone point us to a regulation that requires this?
The petition claims corresponding authors in a paper must store, maintain, and validate data.
The role of the corresponding author according to International Committee of Medical Journal Editors (ICMJE) is shown below
Concern #2: Phase 2b Biomarker Redo Was Not Done by an “Outside Lab”
Concern #3: Which Outside Lab Did the First Phase 2b Biomarker Study?
These concerns were already presented in the original petition and are just repeated in the supplement. Here is the relevant paragraph from Cassava Sciences’ Form 10-K where they mention that blinded conditions were maintained.
We also tried to find other study protocol designs where names of the specific lab(s) were listed and mostly came away empty-handed. Most big and small pharmas do not list the names of the labs they use; many do not even upload study protocols or results onto clinicaltrials.gov. This list includes drugs that were granted SPA. When we did find a trial granted SPA with a study protocol, the shipping address for the samples in that study was blacked out. But by carefully studying the protocol we found that the shipping recipient was the sponsoring pharma. Our point here is simple – the independence of labs is only truly guaranteed by blinding requirements and not affiliation. If this were not the case, it would be impossible to trust trial sites and labs used in developing countries, as many big pharmas do.
On this front we plan to watch out for the study design of Cassava Sciences’ phase 3 studies for Simufilam where we may learn how they plan to leverage their new CRO.
Concern #4 Likely Lab Errors or Manipulation in the Phase 2b Redo Study
The petition claims that CSF/plasma albumin ratio baseline does not match clinically observed ratios
In the methods section of the phase 2b paper, it is mentioned that optical densitometry is used in determining CSF albumin levels. It is therefore fairly clear that the CSF/Plasma albumin baseline ratio also used the same measure. Our team member highlighted this fact on twitter first.
Albumin constitutes a high portion of plasma protein (~60%) and can easily hit the upper end of the optical density range due to saturation. This usually causes an under measurement of plasma albumin and can bump up the CSF/plasma albumin ratio significantly as seen in the baseline provided by Cassava. The clinical ratio Qalb, which is usually measured by a more precise ELISA assay, cannot therefore be directly compared to this technique.
In Cassavas’ defense, the change in CSF albumin (not the ratio) is the biomarker of interest and the comparisons there between placebo and dosage arms show a treatment effect and a positive reduction of CSF albumin.
We also want to point out the malice in the chart included in the supplement as seen below.
The petition referred to multiple past studies where the CSF/Albumin ratio was reported in the range 4 -9. This is, in reality, 0.004 – 0.009 (without the 1000x factor to make QAlb read more easily in clinical charts). They placed Cassava’s value of ~0.25 alongside this in an attempt to magnify the difference. This is extremely misleading and can easily be shared on social media to propagate misinformation. The supplement conveniently omits mentioning the use of optical density in the paper. As a result, the above misleading chart is completely lacking context.
The petition claims that CSF IL-6 baseline does not match clinically observed ratios
The petitioner states that “In other AD studies, CSF IL-6 levels are 1-5 pg/mg.” That’s actually a typo and should be “pg/ml.” But let’s assume they meant 1-5 pg/ml. They further state that the baseline IL-6 levels for Cassavas’ studies are 33-34 pg/ml and are far outside the norm (they volunteered to provide references on request). Why are the values so different? It seems as if all the “Exemplary published studies” referenced in the petitioner’s supplement used ECLIA (Electro-chemiluminescence Immunoassay) in the measurement of CSF IL-6 levels whereas Wang et al used ELISA (Enzyme Linked Immunosorbent Assay). It was even stated in Dr. Wang’s publication, “Although these assays are not directly comparable, values from each are reported in pg/mL.” In essence, they are correct in pointing out that the values are drastically different. But what they fail to point out, however, is that they were measured using different methods. It is a continuation of the same theme of misleading comparisons made with seemingly incriminating charts.
Gruber et al., showed elevated CSF IL-6 levels for a variety of conditions. These levels were measured using ELISA. 
· Neuro-lupus (71.40±5.89pg/mL)
· CNS infection (374.24±92.61 pg/mL)
· Other inflammatory, neurological conditions (33.92±29.36 pg/mL)
· Controls without inflammatory CNS disease (4.35±3.00 pg/mL)
Note that the CSF IL-6 baseline values measured by Dr. Wang et al fit well within the inflammatory, neurological conditions range above. Again – both methods being ELISA.
We acknowledge that ECLIA may be a superior method over ELISA, but “Currently, enzyme-linked immunoassay (ELISA) is the most widely used approach for the detection of AD core biomarkers in CSF. However, these ELISA methods often show considerable inter and intra-lab variability that prevents the use of standard cut-off values and precludes the wide use of CSF biomarkers in clinical practice.” (reference). Regardless of the CSF IL-6 baseline values, or how they were obtained, it is crucial to note that it is the change from baseline that is important. And that is what was measured in the Phase 2b study.
The petition claims that CSF sTREM2 baseline does not match clinically observed levels
Similarly, the petitioner claims that the sTREM2 levels obtained by Cassava Sciences are far lower than other published studies. Another simple literature search can find studies with measured CSF sTREM2 levels even lower than Cassava’s, in the range of 172.5 to 305.4 pg/ml. (Heslegrave et al. Molecular Neurodegeneration (2016) 11:3)
Concern #5 Apparent False and/or Misstatements in Cassava’s Phase 3 SPA Request
We did not find any deliberate misstatements. There were some errors, but minor errors do not constitute fraud and do not in any way call into question the safety of Simufilam or its potential efficacy. In fact, publishing all this data and facing scrutiny is what has set up Cassava for a more robust rollout of its phase 3 trials.
Remi Barbier’s recent fireside chat clarified a few things. Phase 3 trials are on track, and he confirmed that cognition and biomarker improvements are not tainted and show real treatment effects. He stated that while the results are encouraging, clear links have not been established between cognition and biomarkers. We appreciate him for volunteering to provide transparency on a chain of transfer of data relating to the upcoming 12-month results of the Open Label study. This shows confidence and will serve to reassure investors. We eagerly await those 12-month results. In a recently released corporate presentation, it was specified that there is a < 10% drop rate for patients transitioning to the CMS study – this is a very low rate that shows patients are willing to risk being put on placebo.
In summary, we find the fraud allegations in the citizen petition are filled with malice as they continue to attempt to discredit Simufilam and Cassava. The allegations do not explain how phase 2 clinical results show improvements in cognition scores measured across multiple clinical trial sites and are obfuscating the blinding processes put in place.
We are confident that the FDA will reject the allegations in the petition.
Note: In the event of any verifiable errors, we will take steps to update the post. Please add comments on the blog to help us ensure the accuracy of our findings. Thank you!