BREEZE-3 Hits on 3 of 4 Endpoints, Misses 24 week Secondary
On October 13, 2011, Depomed (DEPO) announced that top-line results from the BREEZE-3 trial. Efficacy data from the trial were positive and statistically significant for three of the four pre-specified primary endpoints of frequency and severity at four and 12 weeks. Data for the key secondary endpoints of frequency and severity at 24 weeks did not achieve statistical significance.
Safety results showed that Serada was well-tolerated. The most common adverse events were dizziness and somnolence. The incidence of dizziness in the active arm was 12.7% compared to 3.4% for the placebo arm. Somnolence was 6.0% in the active arm compared to 2.7% in the placebo arm. Withdrawals due to adverse events in the active arm were 17%, compared to 12% in the placebo arm.
… Management Overthinks Design …
Unfortunately for Depomed, the results fall short of FDA guidelines, at least as analyzed by the company’s non-parametric analysis. A non-parametric analysis is used when the distribution is expected to be not a normal bell curve. Therefore, the analysis is done around the median, instead of the mean. Depomed chose this type of analysis because the distribution in the first BREEZE-1 and -2 programs was not a normal bell curve. It skewed high, with some patients reporting as many as 99 hot flashes per day. The non-parametric analysis was used to reduce the effect of these outliers on the overall assessment.
However, Depomed also dramatically altered the trial design for BREEZE-3. The company employed a two-week run in period to prior to randomization, rather than one week to better standardize the baseline. Enrollment criteria also called for a cap on the maximum number of hot flash events participants can have per day to reduce outlier. The protocol also called for limited patient interactions to minimize caregiver impact. What we believe happened is that these protocol adjustments forced a normal (bell curve) distribution in BREEZE-3. Therefore, the ideal analysis for a normal distribution is a parametric analysis of covariance, or ANCOVA, that focused on the mean.
In fact, ANCOVA analysis of BREEZE-3 shows the trial hit statistical significance at all four primary endpoints and the frequency and severity secondary endpoints at 24 weeks. Conversely, the non-parametric analysis would have shown statistical significance for the BREEZE-1 and -2 trials. There is a responder analysis currently ongoing. We believe it will show statistical significance for all endpoints.
… Where Does Depomed Go From Here …
We think Serada works. We think management used the wrong statistical analysis to analyze the data – again. They had a skewed distribution for BREEZE-1 and -2, and used a parametric analysis. They had a normal distribution for BREEZE-3 and used a non-parametric analysis. We believe the FDA is clearly interested in persistence of efficacy at the 24 week timeframe. We will be waiting for management to release additional data from BREEZE-3 in the next few months so we can see “just how close” they came on the 24 week endpoints.
The next step for management is to meet with the U.S. FDA to discuss the data from all three BREEZE trials. In total, there were over 1600 patients in the program. As analyzed, using a p-value of 0.05 to show statistical significance, management was 10-for-12 on primary endpoints. If analyzed with the correct statistical analysis plan (SAP) for each trial, they would have been 12-for-12. We expect this meeting with the U.S. FDA will take place in the first quarter 2012.
It remains to be seen what the FDA will say. As noted above, we believe the drug works to reduce both the severity and frequency of menopausal hot flashes. We are unfamiliar with an instance, however, where the FDA allows varying statistical analysis plans for a new drug application (NDA). It is quite possible that the FDA will want to see yet another trial in BREEZE-4. We estimate this could take another 12-18 months and $10+ million. We do not know if management at Depomed will be willing to undertake such a venture. The Serada use patent expires in 2020. Serada needs to be on the market within the next 2-3 years to be a viable product in our view.
… Previous Data …
Below are the data from BREEZE-1 and BREEZE-2. We are only showing the data from the 1800mg dose (dosed as 1 600mg pill in the AM and 2 600mg pills in the PM). We are also showing statistical significance at p=0.05. The original designs of the trials called for a 2-1 randomization which included a 1200mg dose. Therefore, the statistical significance as run was with a p-value of 0.025.
What To Do Now
We are maintaining our Outperform rating and lowering our price target from $10 to $8 per share. We have completely removed Serada from our financial model. Our DCF valuation assumes the following “sum-of-parts” to arrive at our $8 per share target: Gralise = $3.50, Glumetza = $2.00, Cash = $2.00, Pipeline/Technology = $0.50.
Our advice to investors is to buy Depomed stock at a level where Serada expectations are zero - which we believe is today. Expectations around Gralise are varied. We happen to believe the drug has $400 million potential at peak in the U.S. We understand this may be aggressive. Therefore, even if our forecasts for Gralise are too high, we still see a floor for the stock at $4 per share based on the cash (currently over $150 million) and royalties from Glumetza. We note the recent deal with Santarus (SNTS) dramatically improves the cash flow from Glumetza. The rest of the pipeline, including deals around the AccuForm technology with Merck (MRK), J&J (JNJ), Boehringer Ingelheim, and Covidien (COV) are worth at least another $0.50. We do not know how the FDA will respond to the BREEZE data on Serada. Again, Serada is a drug we believe works and has utility for the reduction of frequency and severity of menopausal hot flashes. There may still be a path to market. Then again, there may not. We like buying the stock when expectations are low, or in this case, non-existent. Good news offers quick upside early next-year. And even if Serada is dead, the stock has long-term upside potential driven by Gralise, Glumetza, and the pipeline.