Credit: By Blair O'Neill :
Much has been said among analysts ( here and here ), and Seeking Alpha authors ( here and here ), about Celladon's ( CLDN ) MYDICAR's monumental upside if the upcoming Phase 2 trial results are positive. However, very little has been said about the chance for successful results. Therefore, the focus of this report is to look at the following:
- The preclinical and clinical data supporting Celladon's case
- Odds for CUPID-2's success
- Potential outcomes for the stock price
MYDICAR
That enzyme is called SERCA2a and is critical for the contraction of the cardiac muscle cell. If the enzyme becomes depleted, the heart's contraction cycle becomes unbalanced, which can lead to end-stage heart failure.
If it were widely believed that increasing SERCA2a levels would lead to a meaningful clinical benefit, then maybe MYDICAR could have been approved based upon a surrogate endpoint. However, the FDA is wary when it comes to the use of unproven surrogate endpoints, and so Celladon must instead show that MYDICAR produces a meaningful clinical benefit.
Preclinical data in pigs
So far, Celladon has some preclinical data to show that MYDICAR can increase SERCA2a expression in the cardiac tissue. As for improving heart function, the data is a bit of a mixed bag.
Celladon measured the following mean values at baseline, during the injection period 2 months later, and 2 months after that:
- Peak left-ventricle (LV) pressure: Peak blood pressure during a contraction
- End-distolic pressure: Blood pressure in ventricle immediately before contraction
- End-systolic pressure: Blood pressure in ventricle at the end of contraction
- LV internal distolic diameter: Width of LV immediately before contraction
- LV internal systolic diameter: Width of LV immediately after contraction
- Left-ventricle ejection fraction (LVEF): Percentage of blood pumped out of the heart with each contraction
- A couple of endpoints relating to how quickly the peak LV pressure is reached
We only know of the LV internal systolic diameter, LVEF, and peak LV pressure rate endpoints showing a statistically significant benefit.
Upon further examination, it appears that the endpoints relating to how quickly peak LV pressure was reached were marred by baseline imbalances, and so we're left with the other 2 remaining endpoints. My concern is that the baseline imbalances seen with the peak LV pressure rate could be a sign that the control arm had weaker cardiac muscles. However, such imbalances were not evident in the reported mean baselines for LVEF.
Interestingly, after the heart failure was induced in the pigs, by poking a hole in the mitral valve, the experimental arm was measured as leaking 9% more blood than the control arm. Another 2 months later, the control arm would be leaking 30% more blood than the experimental arm. There is no clear explanation for why the leaking rate increased so dramatically in the control arm, when it had remained relatively unchanged in the experimental arm.
Regardless, what makes the efficacy data such a mixed bag is the fact that no pigs in the control arm died after the injection of the placebo, while 3 would die after injection in the experimental arm (1 from general deterioration, and another from worsening HF). The data from the deceased pigs were excluded from all efficacy analysis. It seems likely that if no pigs had died, then the performance of the experimental arm would have been significantly impacted.
In conclusion, the preclinical data seems to support MYDICAR's ability to increase SERCA2a expression. As for MYDICAR's ability to improve or maintain LVEF, that endpoint was marred by dead experimental pigs, worsening valve leaks among control pigs, and/or potential imbalances in cardiac muscle strength. So it remains unanswered whether or not the use of MYDICAR led to a meaningful enough increase in SERCA2a prevalence in order to cause the observed cardiac benefits.
In later presentations , the company would use median, instead of mean, to assess the preclinical endpoints. Considering how few animals were studied, I feel that the use of the median would produce even more unreliable results.
CUPID-1's efficacy data
This study began as a Phase 1 open-label dose-escalation study. It would compare 4 different dosages of MYDICAR, before moving onto the 2nd phase of the study. The Phase 2 portion would be a placebo-controlled randomized dose-ranging study.
A key pre-specified efficacy endpoint , which happens to be the primary endpoint in the on-going CUPID-2 trial, was the number of patients that experience at least 1 HF-related hospitalization event, or worsening heart failure, after 12 months post-injection. Otherwise worded as:
Here were the results for that endpoint in CUPID-1:
As we can see, there was no dose response between the low and mid doses, and neither of those dosages showed a benefit that was statistically significant with respect to placebo at 12 months.
Another important endpoint would be the change in LVEF at 12 months. Those results were:
Here we see an escalating efficacy response among the different doses. However, the placebo performed better than the low and mid doses, and so we really can't say there was actually a dose response. In addition, the LVEF data was confounded by dying patients.
Alarmingly, only 1 of the remaining 16 efficacy endpoints showed some kind of dose response while also exhibiting a consistent benefit over the placebo arm. That 1 efficacy endpoint was the mortality rate, but there were so few deaths within the 12-month period that it is probably foolish to trust it.
Even if we just ignored the lower dosage arms, and just focused on the high dose arm, Celladon still wouldn't have a good case for MYDICAR's efficacy. Why? Extensive patient baseline imbalances (courtesy of @AndyBiotech ).
In almost all of the measurements, the experimental arms had significantly healthier baseline levels than the placebo arm. Even among the experimental arms, the higher dosed patients typically had healthier baseline levels than seen among the lower dosed patients.
Here were the prevalence of pre-existing conditions and prior use rates of other therapies (courtesy of @chasingthealpha ):
Once again, we see that the baseline characteristics favor the high-dose experimental arm by a considerable margin. You can look at a more comprehensive review of the CUPID-1 trial and results under the Publications section of this page.
Alternative explanations for the positive CUPID-1 data
It is quite possible that the benefits seen in CUPID-1 was driven not by an increase in SERCA2a prevalence, but, instead, by healthier baseline characteristics among the higher dosed patients. Perhaps the lack of a prior diagnosis for ischemic heart disease, and not an increased SERCA2a prevalence, was a more reliable indicator for clinical success.
We have to consider that MYDICAR isn't improving the prevalence of SERCA2a in the human heart as much as Celladon hoped, which could possibly be fixed with a higher dosage. That might be why Celladon, as of January 2015, has designed a Phase 2 study to assess the efficacy of a dose 2.5 times higher than the high-dose in CUPID-1 and CUPID-2.
In addition, if a meaningful clinical benefit can, in fact, be derived from dosing patients with a high-dose of MYDICAR, the effect size may not be large enough to produce a statistically significant result in CUPID-2.
Ultimately, I consider the efficacy results of CUPID-1 to be inconclusive. It seems more likely that baseline patient characteristics provided more relevant factors for success, than did improved SERCA2a levels. Therefore, in the larger CUPID-2 trial where patient baselines are more balanced, it seems unlikely that MYDICAR will achieve the needed 45% risk reduction on the primary endpoint in order to show statistical significance. I think Celladon needs to either run a larger trial to show a more modest clinical benefit with the current primary endpoint, use a more aggressive dosing regimen, and/or use a different primary endpoint.
CUPID-2's possible outcome scenarios
- If the trial fails to produce any statistically significant results in the intent-to-treat population, or among patients who did not experience a cardiac event prior to dosing, nor does the trial produce any sufficiently statistically significant results in a clinically meaningful subgroup, then the company is back to square 1. They either have an ineffective therapy, or still need to discover what dosing and/or clinical endpoints they should be using. Price target: Cash value, or
- If the trial produces statistically significant results in the ITT or mITT (dosed patients), then the stock will multiply several times over. Price target: >$50. This estimate assumes peak sales of $3 billion during 2023 at a 20% royalty rate. 5% probability (or random chance).
- If the trial produces a very statistically significant result in a subgroup, then the price target will be an equally proportional percentage of the $50. Price target: ~$20, assuming the subgroup makes up a third of the wider market. 15% probability, but could still see a sell-off in the near term.
Currently, the market has valued the company as if the odds of success is 50%. Looking at the totality of the data, I think a most optimistic bull should keep that probability at no more than 30%, and so the company should not be valued much more than $11. (Edit: The 50% and 30% figures maintain my 3:1 ratio for the 2 different positive outcomes, as illustrated by my previously described 5%+15% scenarios.)
Note: All price targets use the current fully diluted share count of ~26.4M.
I am not short the stock, since I trade from a registered account. I only decided to write this report because I know how easy it will be for the losers in ATHX to want to jump into another catalyst in order to quickly recover losses.
See also Syneron Medical's ( ELOS ) CEO Amit Meridor on Q1 2015 Results - Earnings Call Transcript on seekingalpha.com
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
