As an outside observer to the rare disease community, I find the recent acquisition of Prosensa (RNA) by Biomarin (BMRN) puzzling. To me, Prosensa’s drisapersen was just another case where promising phase II results were not corroborated in phase III. This happens frequently with oncology drugs (metmab, iniparib, tivantinib, palifosfamide etc.) and typically leads to termination of the program. Judging by Biomarin’s optimism, drisapersen’s fate may be different, which raises questions regarding approval of drugs despite negative P3 outcome. As background, Prosensa’s lead drug, drisapersen, is being developed for the treatment of DMD (Duchenne muscular dystrophy), a devastating and fatal muscle-wasting disease that affects young boys. The drug addresses 13% of DMD cases and works by a new mechanism (exon skipping). Sarepta (SRPT) is developing a competing drug (eteplirsen) that works by the same mechanism.
Promising phase 2 data…
Prosensa went public based on positive results from a 53-patient randomized phase II (DEMAND II). The trial compared 2 regimens of drisapersen (continuous and intermittent) to placebo, with 6-minute walk distance (6MWD) as the primary endpoint. 6MWD measures the distance a patient is able to walk over a period of 6 minutes and is regarded as a reliable way to assess disease progression. In DMD, the minimally clinical improvement is considered to be 30 meters.
The continuous drisapersen arm demonstrated a ~35 meters after 24 weeks, which was maintained at 48 weeks. The difference at 24 weeks was statistically significant whereas the benefit after 48 weeks was almost statistically significant (p= 0.051).
Based on these promising results, Prosensa advanced drisapersen to phase III, which failed to show any benefit after 24 or 48 weeks. The absolute difference at 48 weeks was 10.3 meters in favor of drisapersen, which was not statistically significant (p= 0.42) and is not considered clinically meaningful. In parallel, Prosensa and its partner at the time, GSK (GSK), released results from another small phase II (51 patients) which evaluated two doses of drisapersen vs. placebo. The higher dose led to a 27-meter benefit over placebo after 24 weeks but the difference was not statistically significant (although it was close: p=0.069).
This created a situation where the drug had positive readouts from two different randomized phase II trials but a completely negative readout from a large phase III study.
In order to reconcile this discrepancy, Prosensa and Biomarin argue that the phase II trials, coupled with other long term open label studies, represent the true activity of drisapersen. The phase III failure was blamed on trial design issues such as patient heterogeneity, a high proportion of more advanced DMD patients and insufficient follow-up.
A quick comment on data presentation
In an attempt to convince investors and regulators that the phase III had a different patient population, Prosensa created a slide (see below) that compared baseline characteristics across the different studies. To me, the slide is misleading because the graphical presentation makes the differences look more dramatic than what they are (X and Y axis do not cross at 0). For example, the “time since diagnosis” graph gives a sense of a 3-4 fold difference while in fact the actual difference was only 29%. When companies use these tricks it should be viewed as a major red flag.
Going back to Biomarin’s claims regarding trial design issues as the reason for the phase III failure, the arguments provided to date focus on disease severity and insufficient follow up.
Disease severity – Subset analysis of the phase III data based on age and disease severity did not reveal much in terms efficacy. Subset analyses are unreliable by nature and in this case, they didn’t even lead to a significant benefit. The best subset analyses improved absolute difference between the drisapersen arm and placebo to 21-25 meters but this did not come close to statistical significance, with a sample of 47 and 79 patients, respectively. This is notable given the fact that the two positive phase II studies were able to show a statistically significant (or on the verge of stat. significance) with 34-36 patients.
PTC Therapeutics (PTCT), which is developing ataluren for another subset of DMD patients, also used subset analysis from its randomized phase II and reached the opposite conclusion. PTC decided to recruit older patients with more advanced stage DMD for their ongoing phase III study based on the idea that it will be hard to show a clinical benefit in patients with early-stage disease because they experience very limited deterioration initially (up to 48 weeks). This can also explain the spectacular results with drisapersen in small single arm studies: Some patients do very well, regardless of treatment (see right graph below).
Longer follow up – The need for long- term follow up in order to see an effect contradicts the signals observed in the phase II studies, where a clear effect was seen even after 24 weeks. Therefore, this argument discredits the efficacy signals from the phase II trials.
At JPM, Biomarin’s CEO presented updated results from the phase III extension study, in which patients from the placebo arm crossed over to receive drisapersen (patients originally assigned to drisapersen vs. patients who crossed over from the placebo arm). After 2 years there was a 49.2 meter difference between the two arms. No p value was provided so one can assume the difference wasn’t statistically significant.
Regardless, this analysis is very hard to interpret as the drug is compared to itself from the 48 week time point and included a small number of patients. As not all patients continued to receive the drug, the two arms were probably not carefully balanced. Therefore, the difference between the two arms could be a result of natural variability between the groups. In addition, landmark analyses are unreliable if they are not defined prospectively before results are available. Without this, one can go over the graphs looking for a single time point in which the difference is maximal.
Drisapersen generated intriguing signals that may merit further evaluation but in my opinion, the totality of data fails to prove drisapersen provides any benefit in DMD. If Biomarin truly believes the drug is effective, the right thing to do is a new pivotal trial applying the new selection criteria and endpoints. Instead, Biomarin intends to file for approval with a predominantly negative package which is full of inconsistencies and inherent discrepancies.
The phase III failure and the lack of any positive subset analysis are an order of magnitude more reliable than any positive indication from smaller studies (either placebo controlled or single arm). Despite this, Biomarin believes drisapersen could receive accelerated approval, which is quite unusual for a drug that proved ineffective in phase III. There is no doubt that new DMD treatments are urgently needed but one has to wonder whether this can be an excuse for approving drugs based on unmet need while ignoring basic tenets of evidence-based medicine.
We are selling one of three positions in Foundation Medicine (FMI) following the 100% jump earlier this month (in order to limit exposure to 10% of total holdings).
Portfolio holdings – January 25TH, 2015