Medivation: Commentary

(from NI January 2009)

While our uncertainty regarding the generalizability of the 183 pt Russian dataset remains, Pfizer found it convincing enough to offer Medivation a deal too good to refuse--$225 million up front, $500 million in potential milestones. There is a catch, in that Medivation must fund 40% of the development costs, and with Pfizer announcing they need two more Phase III studies (not a surprise), covering mild-to-moderate AD and Dimebon as an adjunct to Aricept, delaying the NDA at least until 2011, that will be a nine figure expense for Medivation. But it was still a very impressive deal given the sparse POC data. A 525 pt Phase III trial is underway. It should also be noted that a Karolinska group concluded that Dimebon does have a mitochondrial protectant effect, which suggests the possibility of disease-modification. The Huntington's data was less impressive in our view than had been the across-the-board cognitive benefit from that Russian AD trial; it did not produce benefit on the UHDRS, the only neuropsych test component utilized that taps the frontal dysfunction often seen in HD. Medivation also plans to take its prostate cancer program into Phase III this year.

Getting A Handle on Dimebon (On-line comment, addendum 7/10/08)

Medivation's Dimebon is a slippery story, one whose substance continues to be perplexing and ambiguous for NI, even given as we are to avidly following and applauding any hints of therapeutic efficacy in the neurodegenerative disease area. The just-released "top-line" results for Dimebon in Huntington's Disease are another case-in-point. In this 90 patient, three month duration trial, Dimebon produced statistically significant results (p=.03) on the Mini-Mental State Examination, or MMSE. It did not produce any hints of effect on the cognitive component of the UHDRS, or on the ADAS-cog. There were a couple of other positive glimmers that were highlighted in the presentation, including a relatively lower rate of falls, 16% compared to 9% (not statistically significant), and a positive trend (undisclosed) on the behavioral component of the UHDRS. While the percentage change was stated on the incident of falls, when percentage changes were inquired about for other measures, Medivation consistently replied that they were only discussing "top-line" data, though they had already disclosed such a percentage on one which happened to look favorable for Dimebon.

Besides the selectivity of disclosure, which seems based on spin as much as anything, there are some other issues of note here:

First is the relative weighting of the MMSE in the assessment of Dimebon's potential effect on cognition. Medivation's statement that the MMSE is the most widely used instrument in the assessment of neurodegenerative disease is a bit disingenuous: It happens to be the quickest screening test (5-10 minutes), hence once that can be used by a wide range of clinicians as a screening device. As a measure of cognitive change, it is a blunt instrument, indeed it makes the oft-criticized ADAS-cog look compellingly detailed in comparison. The fact that no trend was found on the ADAS-cog (a fact that was disclosed only on direct inquiry) cannot be dismissed simply on the basis of the trial not being 'powered for efficacy.' Medivation previously made much of the consistent results across five different cognitive tests in their Russian Alzheimer's trial, if consistency mattered there in the decision to move into Phase III, why is consistency unimportant in understanding the import of these Huntington's data? Inconsistency can be due to differences amongst the tests used, but as is described below, looking at what these tests measure does not add support to the Dimebon in HD premise.

Second, we were struck by the incongruous "enthusiastic yes" from the CEO when asked about Dimebon's readiness for Phase III in Huntington's. Out of the three cognitive endpoints, Dimebon showed effect on one, the least sensitive, nothing on the other two. Leaping from a 90 patient, three-month duration trial into Phase III is not entirely surprising for Medivation, given their history in Alzheimer's, but it still makes no sense, other than presenting Dimebon as being Phase III ready in two indications, rather than none, as is the case. The MMSE data and selective behavioral findings are encouraging, and would certainly warrant a longer-duration Phase IIb study. The fact that preceding Huntington's trials hinting at possible cognitive or behavioral benefit from other drugs (rivastigmine and Namenda) were even smaller and more poorly controlled (if at all) than this one does not make this data yet worthy of pivotal trial status.

A parenthetical note: Several observers jumped on Medivation's "mistake" in having cited efficacy as the primary endpoint in filings over time, with safety and tolerability then stated as primary in the current discussion. It's true that this allowed for a better headline, but we will take them at their word that this was unintentional. It doesn't really matter: the data are what they are regardless of whether efficacy was tiered first or second, and the data are inconsistent and preliminary at best. Medivation would have been better served by a more sober and balanced appraisal of these results, so that their comments about their Alzheimer's program could be judged as coming from a more measured perspective. As it is, our spin filter continues to be calibrated at its maximum setting when it comes to Dimebon information. Now we will wait for the ICAD meetings, when Medivation states that more information about Dimebon will be presented.

Addendum from a Neuropsych Perspective (added 7/10/08)

Looking at the inconsistency of the data of the three tests used by Medivation, where only the MMSE showed benefit, one must consider the particular focus of the tests, which one(s) would be most credible as an outcome measure.

The MMSE, where Dimebon's effect was reported as statistically significant, does not have any component which measures executive function deterioration, the 'frontal' signs which are most emblematic of Huntington's-associated dementia. The ADAS-cog has very limited coverage of executive system functioning, but it does cover more cognitive ground than does the MMSE. The literature describes the MMSE as not being particularly useful in HD, because HD patients tend to score at or near the 'ceiling' of the test.

In contrast, the literature supports the use of the UHDRS cognitive section, which by design, is aimed at the deficits that crop up most saliently in Huntington's. The three tests which constitute the UHDRS-cog are:

The Stroop Test; Symbol Digit Modalities Test; and the Oral Letter Fluency test. All three of these are sensitive to changes in frontal/executive system functioning.

In other words, the presence of a perceived benefit on the MMSE is much less important than the lack of benefit on the UHDRS-cog when it comes to assessing the utility of Dimebon in Huntington's. This is the reverse of what would be the case for Alzheimer's, where the early losses in memory functioning are considered paradigmatic for that disease, and the ADAS-cog in particular is intended to tap these areas (albeit inadequately in our opinion), the MMSE does so in a crude snapshot fashion (delayed recall of three objects). The absence of effect on the ADAS-cog cannot be extrapolated as saying anything about Dimebon's benefit in Alzheimer's, since this system is relatively intact in HD, thus there is little deterioration to be prevented.

(from NI February 2008)

Medivation announced that they will initiate a 2 dose-level, 525pt Phase III trial for Dimebon in Alzheimer's in 2Q, and that the FDA has OK'd the use of the oft-cited 183pt Russian Phase II trial as the other pivotal trial. They tellingly emphasized that "We are now a Phase 3 company." Being a Phase III company may impress the Street, but the goal is to get an approved drug, isn't it? Here is our view of their current clinical trial plan for Alzheimer's. And calling a Phase II drug 'Phase III' doesn't make it so. There seems to be little middle ground on Medivation, observers either love their prospects, or hate them.

1) Being told by the FDA that they are OK with the use of the Russian PhII trial as one of the two pivotal trials does not mean that it will pass muster under the far harsher scrutiny of NDA review. Companies have not infrequently had the goalposts moved as they completed their NDA process, one of the reasons that the SPA system, wherein the FDA in theory agrees to be bound by its word (why should that require a special protocol?), was devised. Medivation states that they don't need a SPA because there is a well-defined path to approval for Alzheimer's drugs. That is true, but while they use the same endpoints as their predecessors, they are trying to do so with a comparatively minuscule patient sample: 92 (presuming half of the 183 patient total) Phase II patients from 11 sites in Russia. Plus 175 (one-third of the 525 pt total) coming from some undefined mixture of US, European, and South American sites. Which yields a grand total of 267 patients receiving the 20mg TID dose for which approval is being sought. That, in our view, is a critical departure from the well-trodden path.
2) Even if--and we consider this unlikely--this Phase III trial shows efficacy at the same extraordinary level as the Phase II, relying upon 267 patients has a fatal flaw: It does not adequately sample potential problems. The FDA has been under intense attack for permitting drugs to be sold which turn out to have a risk of low-incidence but serious side effects. 267 patients will not be enough to assuage any doubts about a medication which in theory could be taken by millions. Nor will the 1500 patient safety database that Medivation plans to cobble together from a smorgasbord of past, present, and future trials. Nor will any substandard safety data from its previous use as an antihistamine. Compared to the several thousand pivotal patient samples usually submitted for a drug tapping this size CNS market, 267 seems woefully inadequate.n re 3) There are some criticisms of the trial program that we think are specious. One analyst who is very bearish on Medivation cites the necessity of patients being off of other AD drugs as hamstringing the Dimebon patients with a high number of treatment failures. Medivation is correct is noting that patients stopping other drugs frequently do so because of side effects, often GI in nature, which would not make them treatment 'failures' in the usual sense. Furthermore, so long as they are still in the mild-moderate range when they enroll, having previously been on Aricept or Namenda should not be a significant problem for this trial. eve 4) Medivation hopes to complete this trial in 2010 and file later that year. They may be trying to streamline the pivotal trial process in order to save money, placate investors, and/or sidestep the black hole in AD trial enrollment that is the Elan/Wyeth 18 month duration Phase III for bapineuzemab, which is enrolling 4000 patients (albeit to show disease modification, which is a higher hurdle to surpass). wit 5) Our prediction: If the trial produces statistically significant results, the FDA will decide that a total of 267 patients receiving the target dose does not constitute an adequate trial for a drug that might be used by millions, and will issue an 'approvable' letter, requesting that another trial be done.indi 6) With all due respect to the Huntington's and prostate cancer programs, Medivation's $480 million market cap rests completely upon Dimebon in AD. If Medivation were serious about something other than boosting nearterm valuation, they would be designing a two trial Phase III program to be conducted in parallel. It would slow them down, but not by the several years they cite, and would take the Russian trial wild card out of the equation. Instead, they seem to be trying to figure out the bare minimum requirements, in order to not do anything more than that. We believe this is a serious strategic error.nssen (Wilkinson & Truyon) meta-analysis of trials of three ChEi's reported that Janssen's galantamine (surprise!) kept ADAS-cog levels at baseline after 12 months, with modest declines for drugs from Novartis and Pfizer. This is stabilization on the FDA's favorite endpoint over twelve months.

2) Medivation reported that 69% of patients in their trial were the same or improved over baseline global measures after 12 months. In Borkowska's 2005 report of a study of patients on Aricept or Reminyl, she reported that 83% had stable or improved ADAS-cog scores after 12 months. While this is a more narrow endpoint, it does indicate a marked degree of stabilization. Rachelle Doody, who has been out front with Medivation's results, had herself reported a few years ago that Aricept maintained ADAS-cog over baseline for at least one year.

3) The five positive endpoints, covering several types of cognitive and behavioral function, have also been highlighted. However, broad-brush improvements have been reported before. In 1998, the CoEnzyme Q analog idebenone was reported (in a 450pt trial) to have produced statistically significant improvement on the ADAS-cog and four other measures. While not identical to Medivation's menu, they did attempt to cover a similar range of behaviors and cognition.