My thoughts on Sarepta

Although I do not have any positions in Sarepta Therapeutics (SRPT), yesterday I was consistently berated by the entire Twitterverse for suggesting that their phase 2 for Eteplirsen in Duchenne Muscular Dystrophy (DMD) study would not in fact be sufficient to pass the FDA’s sniff test for accelerated approval.  Here I will spell out why:

1.       Regulatory Precedent

The FDA has laid out guidance for any sponsor developing a therapy for rare diseases which is nicely summarized here.   The presentation includes a nice description of the size of NDA submission programs for recently approved orphan therapies.

The good news is that orphan products are eligible for Fast-Track, Accelerated approval paths and Priority reviews, however, the same regulatory standards that apply to regular drugs still have to be met:

Regulatory Approval Standard

                        Orphan drugs are held to the same statutory requirements for demonstrating effectiveness and safety

                        Orphan drugs must: Demonstrate substantial evidence of effectiveness/clinical benefit (21CFR 314.50)

                        Substantial evidence of benefit requires: Adequate and well-controlled clinical study(ies) (§314.126)

 “Study has been designed well enough so as to be able “to distinguish the effect of a drug from other influences, such as spontaneous change…, placebo effect, or biased observation” (§314.126)”

So, does the P2 study in question have “Substantial Evidence of Effectiveness”, is it “adequate and well-controlled”?

There is absolutely no rational argument that n=4 patients receiving therapy is adequate to establish conclusive efficacy.  I have yet to see any baseline demographics on these study participants so who knows if it was well controlled between the two study arms, not to mention the selection bias for who was entered into this study to begin with.  Only the initial 24 weeks were blinded to treatment so the risk of observer bias cannot be excluded for the 48 week data presented yesterday; the investigators knew all study participants were on treatment at that data point.  Single center studies also introduce an element of bias which must be taken into consideration when reviewing these results.  Bias aside, all participants had to be on steroids to be enrolled in the study.  Steroids can effect ambulation performance; can anyone tell me what dosage of steroid these participants had throughout the course of the study and if there were any differences between the two groups?

According to the FDA Guidance:

What constitutes substantial evidence of effectiveness?

-Typically two adequate and well-controlled studies

Independent substantiation of experimental results – single clinical experiment not usually considered adequate scientific support for conclusion of effectiveness

Let’s just assume for one wildly optimistic second that that FDA was to consider Accelerated approval with a single study, what would be the regulatory requirements?

                        Accelerated Approval  is based on surrogate endpoint  where  the surrogate is reasonably likely to predict clinical benefit

                        and usually requires post-marketing studies to verify and describe clinical benefit

OK, that doesn’t sound too bad. Afterall we have the muscle biopsy data.  But wait a second, let’s take a closer look at that.  According to the company data which can be found here:

Treatment Arm			Mean Change from Baseline in % Dystrophin-Positive Fibers		p-value
Eterplirsen (both doses): 48 wks of Tx (n=8)			47.0		≤0.001
	Eteplirsen 50 mg/kg (n=4)		41.7		≤0.008
	Eteplirsen 30 mg/kg (n=4)		52.1		≤0.001
Placebo/Delayed Tx: 24 wks of Tx (n=4)			38.3		≤0.009
	Placebo/50 mg/kg Delayed-Tx (n=2)		42.9		ns
	Placebo/30 mg/kg Delayed-Tx (n=2)		34.2		ns

                         

The patients in the 30mg/kg arm actually generate more dystrophin positive fibers than the higher dose group, yet, this did not translate into a significant improvement on the 6-minute walking test.

Treatment Arm		Mean Change from Baseline in 6MWT (meters)		Estimated Treatment Effect (Eteplirsen minus Placebo/Delayed-Tx)		p-value
Placebo/Delayed-Tx (n=4)		-60.3
Eteplirsen 50 mg/kg (n=4)		+27.1		87.4 m		≤0.001
Eteplirsen Both Doses (n=6)		+7.3		67.3 m		≤0.001
Eteplirsen 30 mg/kg (n=2)		-31.5		28.8 m		ns

In other words, taken as is, this surrogate does not predict clinical benefit (as defined by distance on the 6MWT) at all.  Can anyone explain this?  I doubt it, I am certainly not convinced this is an accurate biomarker, nor is it particularly pleasant to be subjecting these participants to multiple muscle biopsies.  (For the sake of brevity, I’m going to skip what I have learned about DMD trial end points and biomarkers from a brief review of the literature).  Surely a blood based biomarker would be more practical?  It seems that is what the company had in mind.  According to the original study protocol here they were prospectively testing this as a secondary outcome.

Secondary Outcome Measures:

·         Secondary efficacy endpoints will be the change from baseline in: CD3, CD4, and CD8 lymphocyte counts in muscle biopsy tissue

The presence of these lymphocytes is correlated with muscle damage in DMD, so you would hope treatment would show a depletion in counts with treatment. So what happened to these data, did they mysteriously disappear?  I have feeling they were negative and the company just decided not to disclose them.

2.       The regulatory bar set for other companies developing DMD therapies

Obviously Sarepta is not the only company in the race to develop a therapy for DMD.  A quick search on clinicaltrials.gov reveals numerous open, industry-supported, phase 3 clinical programs in this condition -in other words, phase 3 programs are required despite the unmet medical need.

The most similar agent to Eteplirsen in development is GSK2402968. Of note, they have N=54 patients in their phase 2 study

and over N=180 in their phase 3 programs.

If you examine the P2 GSK protocol, you can see that they actually reduced variability in the results of their study by having subjects complete 2 separate 6MWT at baseline as part of the inclusion criteria. 

“Able to complete 6 Minute Walk Distance (6MWD) test with minimal distance of at least 75 meters, with reproducible results (within 20% of each other) at Screening Visit 1, Screening Visit 2 and at the baseline visit prior to randomization”.  This reduces the risk that any differences at important study visits were due to normal variability.

They also had tighter controls around changes to steroid doses that were allowed during the study. 

Sarepta did not assess the reproducibility of their results at baseline at all.

What I find particularly interesting is the statistical powering of the GSK phase 3 program:

Primary Outcome Measure: To assess the efficacy of subcutaneous 6 mg/kg GSK2402968 versus placebo; specifically to have 90% power to detect a difference in 6MWD between GSK2402968 and placebo of 30 meters, assuming a common standard deviation of 55meters.    In other words, 6MWT is so variable that a 55 meter difference between 2 individuals would not even be considered significant.

Clearly the clinical development folks at GSK have put a lot thought into their study protocol and have looked at baseline individual variability a bit harder than Sarepta.  What would the Sarepta data look like if they were adjusted for such a measure of standard deviation? 

To reiterate, I have no financial position in Sarepta.  I advise everyone to take a closer look at the data and consider the prospect that this drug may not be all it has been hyped up to be.  Based on these facts, the suggestion that it will be approved via an accelerated approval pathway based on the existing data is frankly ridiculous.

This article is not intended as trading advice.  Conduct your own due dilligence prior to buying or selling securities.