Kerrisdale Capital – Prothena Corp. PLC short thesis
Prothena Corp. PLC (NASDAQ:PRTA) is a $2.4bn development-stage biotechnology company whose stock has climbed over 700% since its spin-off from Elan in December 2012. We are certain that its lead asset, NEOD001, will fail its ongoing Phase 2b and Phase 3 trials.
NEOD001 is a monoclonal antibody designed to treat amyloid light chain amyloidosis (AL amyloidosis). AL amyloidosis is a condition caused by light chain proteins misfolding, which then leads to the light chains sticking together and forming amyloid structures. These amyloids are then deposited within tissues and organs of the body where they interfere with organ function, and can potentially lead to organ failure/death. NEOD001 is intended to bind to and induce the removal of amyloid deposits.
Charlie Munger: Invert And Use “Disconfirming Evidence”
While this sounds good in theory, all clinical and scientific evidence points to disappointment. In the Phase 1/2 trial, NEOD001 failed to achieve meaningful clinical response, durable response, dose response, control over catastrophic events, or any apparent benefit at all.
The Phase 2b and Phase 3 trials will be no different. The imminent failure of NEOD001 comes as no surprise to amyloidosis researchers; in the words of one amyloid antibody co-inventor, the probability of NEOD001 succeeding in its Phase 3 trial is “almost zero.” Another co-inventor believes that NEOD001 “is just not going to work.” Antibodies targeting AL amyloidosis have been around since at least 2000 but have performed poorly in practice. The principal issue is that AL amyloid deposits are far too heterogeneous for a single antibody to work consistently among patients. On top of the heterogeneity, researchers believe that organ-specific occlusion and amyloid proteolyzation could further contribute to the inability of AL amyloid antibodies to bind to their target epitopes and achieve meaningful responses. Radioimaging studies for other AL amyloid antibodies have shown that even in situations where investigators have achieved binding between antibodies and amyloid deposits in patients, there is virtually no effect on amyloid deposits in the heart or kidneys — the key targets in Prothena’s trials. Prothena conspicuously opted against publishing such a radioimaging study for NEOD001.
In its Phase 1/2 trial, Prothena declares any patient who achieves a decline in NT-proBNP of ?30% from baseline at any single point during its trial to be a responder. But NT-proBNP has been shown to have a week-to-week clinical variance of 35% — so seeing 53% of patients temporarily achieve a ?30% decline in NT-proBNP at one point over a median of 12+ monthly measurements tells investors nothing. Using best response also masks patients who see their condition worsening because it only reflects the single best change in NT-proBNP. For instance, one patient saw his NT-proBNP levels skyrocket by 300% at which point he quit the trial, and another patient died during the trial. Both patients were recorded as “responders” under Prothena’s hollow best response endpoint — but this will still not translate to statistically significant results in the placebo-controlled Phase 2b and Phase 3 trials because precedent published data shows that a control group can be expected to achieve a comparable NT-proBNP response rate. Prothena’s purported cure is nothing but a deception to prop up its stock.
I. Investment Highlights
Kerrisdale Capital: What would you guess the probability is that their [Prothena’s] Phase 3 trial will be successful? That they will get statistically significant results?
Amyloid antibody co-inventor #1: Almost zero.
Amyloid antibody co-inventor #2: This disease [AL amyloidosis] is so heterogeneous that you cannot treat them like a population, you have to treat them like individuals… The way it [AL amyloidosis] presents and the way it works, it [NEOD001] is just not going to work. You cannot do it.
The experts agree — NEOD001 does not work. We spoke with numerous amyloidosis experts who were familiar with NEOD001 and its reported results, including researchers specializing in AL amyloidosis and amyloid antibodies and principal investigators involved with Prothena’s Phase 1/2, Phase 2b, and Phase 3 trials. Most, including ones involved with Prothena’s trials, were skeptical that NEOD001 actually works.
The data agrees — NEOD001 does not work. Prothena’s “positive” data presented to investors and analysts is distorted and misleading for the following reasons:
- Prothena’s initial Phase 1/2 data lacks all elements of a successful trial. There is no dose response, there is no duration benefit, and the drug does not appear to have any influence on catastrophic cardiac events. Prothena subsequently stopped publishing a patient-stratified NT-proBNP time series and switched to exclusively providing best response data — indicating that the initial poor results did not improve later.
- Prothena’s “best response” is an uninformative measure that substitutes variance for efficacy, and Prothena provides this in lieu of meaningful data because NEOD001 does not work. Prothena makes misleading apples-to-oranges comparisons with other publications, comparing their best response rates with responses over fixed durations to try to show that NEOD001 is efficacious. The average patient was in the NEOD001 Phase 1/2 trial for 12.8 treatment cycles as of the final results presentation. Prothena is cherry-picking the best data point out of those ~12 data points, and then comparing it to studies using a single data point at the end of a pre-specified length of time. Prothena’s outcome names (“responders” and “stable”) imply much better outcomes than were actually observed because they are purely based off of the single best data point for each patient.
- One patient saw NT-proBNP skyrocket by more than 300% from baseline while on NEOD001 and subsequently stopped taking the drug — yet still falls under Prothena’s definition of “best responder” since he saw NT-proBNP decline by more than 30% before it shot up.
- One “responder” saw an NT-proBNP best response of -58% — but later died during the Phase 1/2 trial, illustrating the fact that best response is primarily driven by variance and is useless as a clinical endpoint. This patient was counted as a successful responder anyway.
- Prothena repeatedly makes a misleading apples-to-oranges comparison between its 53% best response rate and the 26.5% response rate in Comenzo et al 2012 to convince investors that NEOD001 will succeed in Phase 3 — but Comenzo 2012 measures cardiac response rate based on the final change in NT-proBNP six months after baseline, not the cherry-picked best monthly measurement over a 12 month period. We can however triangulate the fixed duration response rates (response rates at the end of a specific number of months as opposed to best response rate) from Prothena’s Phase 1/2 trial based on responder data provided by the company and make an appropriate comparison with Comenzo 2012. As we show in this report, we see that Prothena’s month-to-month cardiac response rates are in the 20-30% range and in-line with what should be expected from a control group in the upcoming Phase 2b trial.
- The 53% cardiac best response rate achieved in Prothena’s Phase 1/2 trial is also in-line with what investors should expect from the control group in Prothena’s current Phase 2b and 3 trials. In the study Jaccard et al 2014, the authors carved out a subgroup of patients who had survived for at least three months after starting therapy. This subgroup is a good comparator to Prothena’s results and of the patients in Jaccard 2014’s >3 month survival cohort, cardiac best responses were documented in 45% (19/42).
- The change in best response rates over the course of the NEOD001 trial indicates that NEOD001 does not work. Initial data published in 2014 showed a best response rate of 56%, with the majority of responders being treated with ?1 mg/kg of NEOD001. Prothena’s final results presentation showed a 53% response rate when all patients were titrated up to the maximum dose (24 mg/kg) and average duration increased significantly. Increased doses and administrations (and therefore potential best-response data points) should be linked to higher best response rates — instead, Prothena saw the rates decline. This would be a highly improbable event if NEOD001 actually worked.
- Even Prothena’s best-response rates are an artifact of the efficacy of prior treatments rather than the efficacy of NEOD001. The median time since last PCD treatment in Prothena’s trial was approximately 6 months.1 However, it is well-known that chemotherapy causes a transient increase in NT-proBNP which subsequently reverses.
Gibbs et al 2009 showed that there is a transient increase in NT-proBNP levels following chemotherapy in AL amyloidosis patients, with a median increase of 83% six months following diagnosis which naturally reverses afterwards in nearly all patients. The study showed that 92% of patients saw subsequent declines in NT-proBNP levels at twelve months, with the median decline being nearly 50% -- bringing NT-proBNP to approximately where it was prior to chemotherapy.
Article by Kerrisdale Capital
See the full PDF below.