With the details from the DEMAND-III study, the full extent of the failure of Prosensa’s exon-skipping drug drisapersen to arrest progression of Duchenne Muscular Dystrophy (DMD) has become clear.
There are few grounds for optimism among the plethora of secondary end-points, no trends, not even a statistically significant post hoc sub-group analysis. Only pages and pages of convincingly negative data.
While its obviously a relief that concerns about liver toxicities proved unfounded – the treatment was very safe – the safety profile is unlikely to be much consolation for such a convincing lack of efficacy.
But clear-cut as the data is, it leaves one very important question entirely unanswered: was the failure due to insufficient dystrophin production, or – more worryingly – because successfully elevating dystrophin is much less effective than has been assumed?
There are also lessons to be learned beyond DMD: the infinite subtlety of biology means that no phase 3 study is a “slam dunk” – failure lurks round every corner, even in the most unexpected of places. Public market investors, in particular, can pump up asset valuations on an over-optimistic assessment of early clinical data (as the recent Intercept episode illustrates). The chastening experience at Prosensa should serve as a warning that translating promise and potential into regulatory approval and sales is often more difficult than it first appears.
Somewhat surprisingly, Prosensa got as far as conducting a phase 3 study in almost 200 boys with DMD without complying with Sabry’s Rule. This ‘rule’, attributed to James Sabry, the visionary Senior Vice President of Genentech Partnering, mandates the existence of a “target engagement biomarker” before considering any project for clinical development. Quite simply, you have to know that your therapeutic agent has done what it was intended to do (at the molecular level) in the patients you intend to treat, before you determine if that molecular intervention has successfully treated the clinical symptoms of the disease.
Prosensa obviously intended to do this in two ways: with a qualitative PCR assay for the presence of exon skipping, and a semi-quantitative assessment of dystrophin levels in biopsy samples. But both approaches were completely equivocal. Of the 12 placebo patients examined for exon-skipping, 10 were positive (presumably due to the low level of natural skipping that yields detectable full length dystrophin in most DMD patients). Not surprisingly, therefore, the percentage of patients with exon skipping in the drisapersen treated group was no higher. In a similar vein, the extreme variation in dystrophin levels measured in biopsies even among the placebo group precluded any robust assessment of whether dystrophin levels had been increased by drispersen treatment for 48 weeks.
Always remember how bumpy a race-course drug development really is
These analytical findings are not inconsistent with earlier, smaller studies. Even with 180 patients on study, it was unlikely the analytical methods employed would yield a clear affirmation of Sabry’s Rule. But, in 2013, more sophisticated analytical approaches are certainly available, and had compliance with Sabry’s Rule been given the appropriate priority, such an answer could surely have been gleaned earlier in the development pathway.
Had such studies been done, and they had clearly demonstrated that dystrophin levels were not effectively elevated by drisapersen treatment presumably the DEMAND-III trial would never have been initiated, and the developers could instead have focused on improving the agent itself. It is interesting to speculate how much of the strategy was driven by the head-to-head competition between Prosensa and Sarepta, who are developing an alternative exon-skipping therapy almost in parallel. Without the fear that the opposition would gain a lead, might Prosensa have taken the time to properly evaluate their agent in accordance with Sabry’s Rule?
The situation, then, is even more perilous for Sarepta, as we await the pivotal trial data for eteplirsen, an agent with very similar mechanism of action to drisapersen. Their data, showing an increase in dystrophin levels with the agent, is subject to the same issues as the Prosensa data. It cannot be said with much certainty that either agent induces a material increase in dystrophin levels in boys with DMD.
But, as the FDA noted in the recent letter to Sarepta, even if dystrophin is elevated by eteplirsen, the DEMAND-III data (together with the prior failure of PTC124, another drug that apparently increases dystrophin levels by a different mechanism) “raises considerable doubt about the biomarker, and consequentially, its ability to reasonably likely predict clinical benefit”.
That, presumably, comes as something of a shock for the investors in Prosensa and Sarepta. The assumption from day one has always been that raising dystrophin levels sufficiently, were that only technically possible, would yield a cure.
But that was never so certain.
DMD shares much in common with other inherited myopathies, including cardiomyopathies. In each case, muscle contraction strength is weakened due to mutations in structural proteins. But the catastrophic degeneration of function is caused by changes in muscle architecture that results from the initial weakness. If that were not the case, there would be no reason for muscle weakness to increase as the patient ages (since the direct consequence of the genetic mutation remains the same throughout life).
The culprit is the load-coupling mechanism that induces muscle hypertrophy in response to increased functional demands (such as exercise). In healthy muscle, its almost impossible push this too far – but in myopathy, when the muscle strength is weak, this loop drives not only cellular hypertrophy but critically also extracellular matrix production (through increased production of TGF-beta superfamily cytokines). Eventually, the amount of extracellular matrix begins to disorder the fiber structure of the muscle, leading to further declines in strength (because the fibers are no longer neatly aligned along the long axis of the muscle). This decline in function is no longer a direct consequence of the original mutation, and – beyond a certain threshold – is self-sustaining.
If that threshold has already been passed in boys as young as 6 years old (the population treated in DEMAND-III), as seems possible because they are already suffering continuous decline in muscle function (as opposed to a consistent stable weakness that would be the direct result of the genetic mutation), then even restoration of normal levels of functional dystrophin would be ineffective at halting the decline in muscle function.
DrugBaron raised this spectre as long ago as 2000, in an editorial for the Quarterly Journal of Medicine. In response to the surge of enthusiasm for gene therapy that accompanied the completion of the human genome sequence, DrugBaron asked (rhetorically) in how many genetic diseases would complete correction of the causative genetic defect lead to a clinical cure? And using DMD as an example argued that in many cases it likely would not (for precisely the reasons outlined here).
The FDA are right to doubt the utility of dystrophin levels as a biomarker for clinical improvement. Investors in Prosensa and Sarepta maybe should have been more cautious. Certainly, they should be cautious now given the double-whammy of uncertainty they are facing: the DEMAND-III data is far from convincing that dystrophin levels are materially increased; but more frightening still is the very real possibility that DMD, even in children as young as six, cannot be cured by restoring near-normal levels of dystrophin.
Does DEMAND-III signal the beginning of the end for exon-skipping in DMD?
With such convincingly negative data it would be easy to draw that conclusion. But DEMAND-III teaches something important about DMD itself: the inherent variability between patients. The same variability that rendered attempts to measure dystrophin levels inconclusive surely also contributed to the inability to see an effect of treatment on 6-minute walking time. One solution is to consider pivotal open label trials, using each patient as his own control. Placebo effect is not a major issue in these trials, and insisting upon a randomized, placebo-controlled study design (rightly the gold standard in other indications) may deny DMD patients an effective therapy.
Secondly, DEMAND-III robustly demonstrates the safety of the exon-skipping approach, at least with drisapersen. That opens the door to initiating therapy even earlier. Apparently, post hoc analysis suggests a possible effect of drisapersen on the youngest patients in the DEMAND-III cohort, providing further encouragement for a trial in the very young.
Lastly, its important to remember that 48 weeks sounds like a long trial (and it is), but the muscle weakening in these boys had been developing since birth (an average of between seven and eight years). If the consequences of the dystrophin loss were expressed that slowly, its possible that the benefits of restoring dystrophin may also take years to be maximal. Much longer trials may not be feasible, but at least using an open label design may increase the power to see a small effect sufficient for approval that may grow with continued use of the therapy.
This Prosensa case study amply demonstrates both the importance of Sabry’s Rule and the need for flexibility in clinical trial design, to match the characteristics of the disease under study. But it also teaches us a third lesson: there are no certainties, even in late-stage drug development. Those of us engaged in drug development every day can no longer be surprised by failure, marveling over and over again at the complexity of biology. The worry is that investors, and public market investors in particular, over-value assets backed by a sound hypothesis and promising early data – under-estimating the risk that remains in late-stage development.
Already in 2014, we have seen Intercept achieve a $5billion market value on the basis of phase 2 efficacy data for its NASH drug obeticholic acid. These same investors reacted similarly to the phase 2 data from Prosensa and Sarepta. Such huge valuations of still-risky assets is dangerous for the biotech sector – huge losses from the “surprise” failure of DEMAND-III damage confidence, and repeated too often can lead generalist investors to shun smaller healthcare companies for years or even a decade or more. Where Prosensa has gone, others could easily follow. So next time you see an odds-on favorite, remember just how bumpy a race-course drug development really is – and that the last furlong is no smoother than the rest of the track.