Biogen Inc. (BIIB) Earnings Call Transcript & Summary

Allow me to welcome, everyone, to this very important presentation by our colleagues from Biogen. This is John Dwyer, President of the Global Alzheimer's Platform Foundation. This was initially intended as a discussion of the science and clinical aspects of Biogen's drug aducanumab, and we broadened it to the public, which I think is terrific. So let me just say a word about who GAP is for those of you on this call that don't know us. The Global Alzheimer's Platform Foundation's mission is to reduce the duration and cost and improve the quality of Alzheimer's disease clinical therapeutic studies. Not a trivial mission, to say the least. Over the last 6 years, we have combined with our network of over 80 truly leading clinical trial sites in North America to try and accelerate and improve clinical trials for a variety of different Alzheimer's therapies. We're proud to say one of those was Biogen's. And it is with that background that we have in our audience, all of our sites, I hope, their PIs and staff who have worked so hard to bring aducanumab to the place where it is and other very important therapies to meet our singular goal of trying to accelerate a therapy for this scourge known as Alzheimer's. With that in mind, let me just mention the foundation's conflicts, such as they are. We have been a proud recipient of grants over the last 6 years from Biogen as well as having worked with them on a clinical trial in a fee-for-service capacity. And the only other conflict worth mentioning is that both my Chairman, George Vradenburg, and myself, have testified before the FDA in support of the Biogen application, recognizing that there is no disease-modifying therapy for either mild cognitive impairment or mild AD. Just a word about this particular endeavor. We are so keen to have Biogen particularly as represented by Samantha, to take on the opportunity to really tell us what Biogen sees as their data, as their clinical benefit to patients and say the things in a clear, sort of, unbroken way that might not have been available to us during the ADCOM. And the passage of time has also allowed them to learn a bit more, I suspect. There will be questions. We asked our PIs and colleagues to provide questions. Those questions have been provided to Samantha in advance, that is to reflect the unique regulatory circumstances we operate in, and it was the only prudent thing to do. I will be asking a set of questions that came from our PIs, and she will answer them. So with that, and without further ado, let me introduce Samantha Budd Haeberlein, a woman who has been living with this particular application for quite some time, the Head of the Neurological Disease Group at Biogen. And Samantha, thank you very much for coming. I look forward to your remarks.

Thank you very much, John, for your very kind introduction. And thank you to John and Cindy and the GAP platform for inviting us to present our data today. As you mentioned, there are a number of questions. There were many questions submitted from the GAP organization. Thank you, everyone. It really showed a great deal of interest in this work. And I look forward to answering as many of those as possible. We've got a lot lined up to answer at the end of the presentation. Let me share my screen and see if I succeed in that. Can you now see the presentation?

Yes.

Wonderful. And you can still see it?

Yes.

Okay. Very good. Thank you. So without further ado, I'll jump in. This slide is a necessary slide just to ensure it's clear that I may be making forward-looking statements through this presentation. And as a reminder -- let me just minimize that -- as a reminder, aducanumab is an investigational drug, and it is currently under regulatory review here in the U.S. but also in the EU and in Japan. And it's not approved for use in any country at this time. Just wanted to start by outlining the main data that I'll be speaking to today comes from the Phase III trials that recruited 3,285 patients at 348 sites in 20 countries. And as John just mentioned, a number of those sites in the United States are part of the Global Alzheimer's Platform. And really, it is because of the patients, their families, the patient advocacy, such as GAP and the clinical trial sites, because of their efforts that we're able to actually report this data. So I'd like to start by thanking everybody for their commitment and for advancing the research in Alzheimer's disease. As John mentioned, there was an Advisory Committee. So as part of the FDA review of aducanumab last year, there was an FDA Advisory Committee and this is held or hosted by the FDA. I just want to really make sure it's clear that we do stand behind the analysis and the results of aducanumab that were presented at that Advisory Committee, and I will be presenting many of those here. It was an unusual Advisory Committee. There was a core joint briefing book that was the first ever joint briefing book issued by the neurology division from the FDA. And the way to look at the data that we have submitted for regulatory approval is that it's the totality of data that provides substantial evidence of clinical effectiveness of aducanumab. And this is demonstrated by Study 302, study also called EMERGE, and is supported by Study 103, which is also referred to as PRIME. Biogen and the FDA did conclude jointly that the partially discordant results from Study 301 or ENGAGE, do not meaningfully detract from the persuasiveness of Study 302. We have applied the highest scientific rigor and integrity in the analysis that we submitted. And many of the analysis to understand the differences between the studies was conducted in collaboration with the FDA. We do, however, recognize the complexity of this data set and the challenges associated with the first positive Phase III study in Alzheimer's disease. This audience really almost doesn't need this slide. But for me, it's always important to realize and recognize what we're up against. Alzheimer's disease is a significant unmet medical need, and not only to the patients but to their families and society as a whole. As of 2018, we know the numbers. There are about 50 million people living with dementia, with Alzheimer's as the predominant diagnosis among those individuals. Alzheimer's is a progressive neurological disorder. Results in memory loss, but also behavioral symptoms and a loss of ability to perform daily activities. In advanced stages, that actually means that these patients become completely dependent on care from families or from society. And Alzheimer's disease is ultimately fatal. And there's no treatment today that alters the course of the disease. And that's where we have been putting our efforts for quite some time. In order to bring forward new treatments that address the underlying pathology of disease, you have to understand the disease. And what we know about Alzheimer's is that there are 2 main pathological hallmarks. One is amyloid or beta amyloid, which is a peptide formed inside of neurons, which is then released from neurons. And in Alzheimer's disease, there's an imbalance in the production and clearance of this peptide a beta, such that it is then able to aggregate from the small monomers into oligomers, fibrils and then into amyloid beta plaques. These are the clumps of beta amyloid that we can visualize in the brains of Alzheimer's patients using amyloid PET imaging. Amyloid beta is responsible for initiating the dysfunction of the second pathology, the hyperphosphorylation and aggregation of tau protein, which is found inside of neurons. Together, the tau fibrils, tangles and amyloid beta plaques lead to synaptic dysfunction, inflammation and neuronal death. And there is a great deal of extracellular activity with microglia, dying neurons, which is also occurring at the same time. Aducanumab then, as a mechanism of action, targets directly the pathology by binding to oligomers, fibrils and the amyloid plaques that are deposited in the brains of individuals with Alzheimer's disease. And so thereby reduces the progression of Alzheimer's disease, and is different from treating the symptoms of Alzheimer's disease and is aimed to stop or slow the progression of the disease, so-called disease modification. Aducanumab is not the only, not even the first, such approach to remove and reduce the degree of amyloid in the brain in Alzheimer's disease, but it is different from those that have gone before. It benefited from a great deal of learning from the first generation of anti-amyloid approaches. For example, at the molecular level, aducanumab specifically binds to the neurotoxic aggregated forms of a beta and retains effector function, which actually enables the microglia-mediated clearance of the aggregated a beta. In the clinical trials, there were a number of elements that were implemented and that you're well familiar with that are now considered important to enable the ability to see the efficacy of an approach such as aducanumab. For example, ensuring patients did have biomarker-confirmed Alzheimer's disease, enrolling patients much earlier in their disease progression. Aducanumab was the first of the current generation of antibodies to demonstrate proof-of-concept before initiating the Phase III trials, and I'll recap on what that is from the PRIME study. And so accordingly it's, therefore, having completed the Phase IIIs is the only program at this time to have read out with a positive result in the Phase III trials. We know there are other programs that are now also showing a reduction in amyloid and association with reduced clinical decline. So aducanumab is now among others that have shown this effect. Here's an overview of the clinical development program. And for this audience, you can very much appreciate the long timeframes of a clinical development program. And a big deal of this is how we are able to recruit and enroll patients into these clinical trials. Starting in 2011 and then with the Phase III trials, having read out in 2019, having read out early, in fact. There were 8 clinical trials. And the first efficacy trial was Study 103 or the PRIME study. And then the bulk of this presentation will be on Study 301 and 302 which were initiated in 2015. Happening today, is Study 304, and this is a redosing study. So all patients who were previously in an aducanumab clinical trial, those who are eligible are able to come back to Study 304, which is an open-label extension study where all patients are -- it is possible to receive aducanumab. There's no placebo arm in that study. Okay. Just to recap then on the Phase Ib study or Study 103. At the time this data read out, which was 2014, it was really quite remarkable data. On the left here, I'm showing the brain imaging, the amyloid PET imaging results from that study, where patients who were treated with aducanumab, and you can see here the doses on the right 3, 6 or 10-milligram per kilogram, you saw a dose-dependent and statistically significant reduction in amyloid plaque at the 1-year time point. And this was quite a robust reduction in amyloid compared to programs that had previously attempted this approach. But what was also remarkable was that even in this small study, there was also an effect on the 2 clinical endpoints. I'm showing here the CDR sum of boxes. And here, you can also see that dose dependency of the reduction in clinical progression on CDR sum of boxes at 1 year. And you can see in the high dose, 10-milligram per kilogram and titration to 10-milligram per kilogram, this was statistically significant even in this small study. The clinical endpoints were exploratory. And I would say somewhat unexpected, given that the study was designed to be able to test the effect on the reduction of amyloid plaque in the Phase Ib. And so with these strong but surprising results, we were able to design the Phase IIIs, which I'll spend the rest of this presentation on. Okay. So moving into the Phase III studies, then. I'll build this slide out. The Phase IIIs were 2 identically designed studies. They were 18 months in duration. They were randomized, double-blind, a placebo-controlled. And as I mentioned, they recruited 3,285 patients. The patients were early Alzheimer's disease, which was the same population or similar to that in the Phase Ib, including patients with MCI due to Alzheimer's disease and mild Alzheimer's disease, and with an MMSE range of between 24 to 30. Two doses were tested in this study. And the primary endpoint was CDR sum of boxes changed from baseline at 18 months. And there were other clinical endpoints, including the secondaries MMSE, ADAS-Cog 13 and ADCS-ADL-MCI. And we had sub-studies for a number of key biomarkers, and I'll show the results from those as well. The Phase III study was 18 months in duration, which was longer than the Phase I. The Phase I was 12 months in duration. And one of the reasons for that longer duration was because we had -- we made changes to the dosing regimens in Phase III to mitigate for the incidence of amyloid-related imaging abnormalities, or ARIA. ARIA was detected in the Phase I study. And as I mentioned, it is an imaging abnormality, so it is found on MRI. And it's believed to be due to an increased permeability in the cerebrovasculature due to antibodies binding to amyloid and removing it. It's found with aducanumab and with some other anti a beta antibodies. And so we made 2 changes to the dosing regimen to mitigate for ARIA. One was a 24-week titration to reduce the incidence of ARIA. And you can see on the schema the steps through doses 1, 3 and 6 before reaching the target dose of 10-milligram per kilogram. And then in the study, the target dose was 14 doses of 10-milligram per kilogram. But we also had a second impact from ARIA in that we had differential dosing by APOE4 carrier status. And this is because ARIA is found to be both dose dependent, but also has a greater incidence in APOE4 carriers. So on this next slide then, these are the 2 dosing regimens that we had in Phase III. The top 2 lines represent the low dose group, which was 3-milligram per kilogram for APOE4 carriers and 6-milligram per kilogram for APOE4 noncarriers. And in the high dosing regimen, it's the 2 joint lines at the bottom here, at the beginning of the study, APOE4 noncarriers were titrated to 10-milligram per kilogram, but APOE4 carriers were titrated only to 6-milligram per kilogram. Partway through the recruitment for these studies, we received data from the 103 Study that showed that it was safe to titrate patients, APOE4 carriers also to the 10-milligram per kilogram dose. And so partway through, we amended the protocols, such that for the high-dose regimen, both APOE4 carriers and noncarriers could be titrated to 10-milligram per kilogram. APOE4 carriers are, in fact, the majority of the population. 2/3 of patients with Alzheimer's disease are APOE4 carriers. And so this had an appreciable effect on the dose in the high dose arm. At the beginning of the study, so before Protocol Version 4, as you see the acronym here pre-PV4, the average dose was 116-milligram per kilogram. Whereas in patients after -- recruited after Protocol Version 4, the median cumulative dose was 153-milligram per kilogram. So you can see that had a significant impact on dose. This actually turns out to be quite important for differences that we saw in the results between the 2 studies. Now I mentioned that the study started in 2015. It, in fact, took 35 months to recruit the full patients in both studies. And in 2019, so well into the study and about a year before the end of the studies, we terminated the 2 trials following a futility analysis. However, that futility analysis was later determined to be inaccurate, and let me just describe that a little bit. The futility analysis is a prediction, and it used a methodology called conditional power. And that's the probability that the primary efficacy endpoint analysis, what would occur in the future, would be statistically significant at final analysis. And so using this calculation on approximately half of the study data, the futility analysis predicted that the studies would not be positive at the end of the trials. However, when we brought in the full data after the futility analysis, we learned that 2 key assumptions in the prediction did not hold. And so the futility analysis did not accurately predict those future results. Those assumptions were: one, that identically designed studies would lead to similar results, therefore the methodology that was used, pooled conditional power is appropriate. That turned out to be incorrect because we had different results for the 2 studies. The second assumption that the treatment effect would remain consistent over time. While with the protocol changes and the dosing changes, that also turned out not to hold. At the time we conducted the futility analysis in March of 2019, the results did already indicate that Study 302 was trending positive, whereas Study 301 was not. However, the upshot of this is given the futility prediction, we terminated the trials in March of 2019. As the studies were almost complete, we then brought in the remainder of the data, and we conducted the analysis according to the prespecified statistical analysis plan. And it's with that large data set that we determined that: one, the futility analysis was incorrect; and two, that we actually had efficacy in Study 302. And the remainder of this presentation is based on the larger data set. First, just to reflect on the clinical endpoints that are used in these trials. It's very simple and straightforward to say that Alzheimer's is characterized by a loss of memory. But in fact, there are a number of important and distinct symptoms that characterize Alzheimer's disease. And so we, as do others, use multiple clinical rating scales in the Phase III trials. The scales that we used are listed on the right here, and these are validated and widely used in the early Alzheimer's disease population that we recruited. They do cover the full scope of symptoms experienced by patients with Alzheimer's disease. And in addition to the different symptoms, they include a range of paradigms, including the clinical judgment based assessment of patients and caregiver input. Also, patient and caregiver direct reporting and cognitive performance tests, the one thing that people think of most frequently in regards to Alzheimer's disease. So these capture a range of paradigms and a range of symptoms. The scales are also quite distinct in the dimensions that they measure. And there's quite minimal overlap between them. It's between only 5% to 25% overlap in questions or items in one scale that is measuring something similar in another scale. So let's start with Study 302 then. And here is the primary objective of CDR sum of boxes at week 78. In the green bar, the high dose shows a 22% reduction compared with placebo, and this is statistically significant. In the low dose, we see a 15% reduction. This is not statistically significant, but is numerically in the direction of aducanumab. CDR sum of boxes, as an endpoint, actually has 6 domains, 3 each are covering aspects of cognition and function, and those domains are listed on the left here: orientation, community affairs, home and hobbies, judgment and problem solving, memory and personal care. The gray bars that I'm showing here are the individuals in the placebo arm of Study 302. And you can see that on each of the domains in the CDO sum of boxes, there is progression from baseline to week 78. For patients in the high dose arm, the green bars here, you can see that there is a reduction in progression at each and every one of the domains measured by the CDR sum of boxes. And so therefore, supporting that the overall result is not driven by any one domain in this particular scale. As I mentioned, other endpoints were measured. Here are the secondaries in the rank order of MMSE, ADAS-Cog and the ADCS-ADL, where the high dose is statistically significant and has reductions between 18% and 40% compared to placebo. And in the low dose, we also see more clearly, for ADAS-Cog and ADCS, an intermediate effect also for aducanumab, although these are not statistically significant. Among the tertiary endpoints, we had one additional clinical efficacy endpoint was the neuropsychiatric index version 10, which assesses behaviors such as aggression, agitation, anxiety. And in this endpoint, we also see in Study 302, a statistically significant effect, an 87% difference versus placebo in the high dose and again, an effect but not statistically significant in the low dose in this particular endpoint. The [ NPI ] also captures caregiver input on their distress. So there's a separate score that is conducted in regards to how the caregiver feels about these types of symptoms. And the caregivers of patients who received high dose reported an 84% less burden compared to the caregivers of patients who received placebo, which is a very important data point as we do understand that these symptoms, in particular, not just important to patients, but they are some of the most distressing for the families and caregivers. As in Study 103, we did also assess the reduction in amyloid plaque pathology. And here, instead of showing pictures of the brain, this is a quantification looking at SUVR. And you can see that there is a dose and time-dependent reduction, which is statistically significant. And the absolute value of 0.278 in the high dose at week 78 is very similar to what was seen with 14 doses of 10-milligram per kilogram in the Phase I study. So a robust reduction in amyloid pathology. In addition in this study, we had sub-studies assessing disease-specific to tau pathophysiology and neurodegeneration biomarkers. On the left is CSF phospho-tau. And while this is a small subgroup, we had not very large uptake of CSF in the subgroups. We nonetheless see a dose-dependent and statistically significant reduction compared with placebo in phospho-tau. And over on the right, we see a similar pattern with the biomarker of total tau. We did also implement towards the end of the study, across both studies, a small tau PET imaging substudy to assess using a novel tau PET ligand to MK6240, the effect of aducanumab on the reduction of tau pathophysiology directly using imaging. And here, I'm showing 3 different composite regions of the brain in which one would expect tau deposition to occur in this patient population and indeed to be increasing over the period of this clinical trial. And the gray boxes are the placebo arm, so the individuals in the placebo arm. And you can see by the increase from baseline in each of these composite regions that, that's indeed occurring. In the patients who do not have aducanumab, there is an increase in tau neurofibrillary tangles. Whereas in the low and high dose, we see a reduction compared to placebo in each of these regions. And so with both CSF and imaging biomarkers, which are independent methodologies, we can see that aducanumab is not only affecting amyloid pathology, but it's also having an impact on the downstream pathophysiology. We were interested to understand at the individual patient level whether there is a correlation between the reduction in amyloid and the consequent impact to the clinical outcomes. And that's shown in this schematic here, the orange representing the reduction in amyloid. And on the 4 primary and secondary endpoints that we assessed, we can see a statistically significant correlation with the clinical outcomes at the individual level. There's also a strong correlation between the reduction in amyloid and the reduction in tau -- phospho-tau as measured in the CSF, that's what the 0.52 stands for here. And there is also a relationship between the reduction in phospho-tau and the clinical outcomes on the 4 clinical outcomes measured. And you can see that 2 of those even achieved statistical significance, as shown on the right here. Moving to Study 301. We have a different picture with partially discordant results. Here, and this slide, similar to what I showed before, is the primary endpoint and the secondary endpoints for Study 301. The green bar, again, is the high dose. And directly you see on CDR sum of boxes with the 2% in the wrong direction. Study 301 is a negative study. We have some reduction in ADAS-Cog and ADCS-ADL, but none of these changes reach the level of a p-value of less than 0.05. Interestingly, the low dose, the blue bars, has an intermediate effect and this intermediate effect, which is all in the direction of aducanumab is very similar to the magnitude that we also saw with the low dose in Study 302, and which is the reason that we refer to these results as partially discordant. This picture is also repeated in the biomarkers. So for example, on the left here is the amyloid plaque reduction in Study 301. It looks very similar. You see the time and dose dependency. You even see statistical significance. But in fact, the high dose is, at 0.235, 16.5% smaller than in the high dose of Study 302, whereas the low dose is very similar to that in Study 302. What we learned later is that, and here in this PET subgroup, the cumulative dose in this group is 10.4% smaller than that in Study 302. We see a similar effect in the CSF biomarkers, and I'm just showing one of them here, in that we do have a reduction compared with placebo. None of these reached statistical significance or a p-value of less than 0.5. However, the low dose is, again, numerically very similar to the low dose in Study 302. And the high dose is not. It's, in fact, very similar to the low dose. So we have this blunted effect occurring in the high dose in the biomarkers. And here in the CSF subgroup, the treatment effect is only 51% in Study 301 high dose. One way to try and put the 3 studies together and try and understand the relationship between the reduction in amyloid and the clinical effect is to plot as we've done here, the amyloid PET SUVR on the x-axis, and then the treatment effect on CDR sum of boxes on the y axis. The pink dots, these are the dosing arms from the PRIME study 103, 1, 3, 6 and 10. And then you also have the titration group, which has an average dose of 5.3. And you can see how those -- they're small dots. They relate to the size of the population, but nonetheless have quite clearly falling along a line of association between amyloid plaque reduction and reduction in clinical decline. Adding in the doses, low and high from both of the Phase IIIs, you see that the 301 and 302 low doses fall into a very similar place in terms of degree of reduction of amyloid plaque. And they fall between the 3 and 6 milligrams from the Phase Ib, as one might expect. And then the Study 302 high dose arm falls in a region quite similar to that from Study 103, 10-milligram per kilogram arm. And it's the 301 high dose arm, which is falling outside of this correlation analysis. And so we refer to the results as partially discordant. The low dose are similar in both clinical and biomarker endpoints. The high doses are not similar in biomarker or clinical endpoints. But the exposure response relationship, when we assess patients on the basis of their exposure, is concordant. Here's another way of looking at that degree of consistency and here using the data from the 3 studies, looking at the common endpoints. The dark green color here is filling in the squares where each of the results has a p-value of less than 0.05 favoring aducanumab. And we can see that all of the measurements in Study 103 are in dark green, as are the high dose arm in Study 302. And then the light green, these are the cells where the results are numerically supporting aducanumab but do not achieve that p-value. And you can see that in the high dose of Study 301, here is where we have the 2 cells that are no longer supporting directionally and is inconsistent with the rest of the data set. So why are the 2 results different? Well, we spent an extensive period of time trying to understand the differences between the study, given that there were so many consistencies. And we conducted this work together with the FDA. And we concluded from that work that the demographics, the disease characteristics as well as the frequency, severity and management of ARIA were, in fact, all similar between the studies. The underlying pharmacology of aducanumab is similar in both studies. And this is quite critical, i.e., if a patient has received sufficient dosing and exposure of aducanumab, even in Study 301 high dose, the response has the same relationship as in Study 302. And what we determined the differences were largely driven by was, in Study 301, overall, a lower exposure to 10-milligram per kilogram dosing and an imbalance in a small number and distribution of rapidly progressing Alzheimer's disease patients. Actually a very small number. But what we did conclude from that work was that even in Study 301, patients who were randomized to have the full opportunity for 10-milligram per kilogram did have results similar to Study 302. And so this is one way of looking at that type of data. You have Study 301 at the top here and Study 302 at the bottom. And if we include the subgroups, who had the opportunity for 10-milligram per kilogram, i.e., not including those APOE4 carriers at the beginning of the study, who were only titrated to 6-milligram per kilogram, then in both studies, the weighted means of these groups on CDR sum of boxes gives a 23% reduction versus placebo. And so that is consistent for both studies. Moving then to look at the top line safety data from the Phase III trials. ARIA, as I mentioned earlier, was a safety finding of interest. It does refer specifically to radiographic abnormalities observed with aducanumab and other anti a-beta antibodies. And there are 2 types of ARIA that are defined and reported. One is ARIA-E, or ARIA-edema, and this is vasogenic edema or sulcal effusion, fluid movements in the brain. And the other is ARIA-Hemorrhage, which are either brain microhemorrhages or localized superficial siderosis. And as I mentioned, we believe that these may result from increased permeability as a consequence of antibody binding. Here are the most common adverse events associated with aducanumab in the placebo and the high dose 10-milligram per kilogram. And we can see that ARIA-E is the most common, followed by headache. And then ARIA-H, falls and then ARIA-H superficial siderosis. Serious hypersensitivity reactions were, in fact, rare with an incidence of less than 0.1%. And compared to placebo, aducanumab was not associated with other abnormalities in vital signs. Taking a bit of a closer look at the clinical and MRI characteristics of ARIA-E, given that these are identified by MRI imaging, it's important to understand that actually the majority of these, 74% of them in the high-dose arm, are not associated with symptoms, patients are unaware that they have this radiographic finding. 26% did report symptoms and the most common symptoms were headache, confusion, dizziness and nausea, and most symptoms reported were mild or moderate in severity. Further, the MRI findings of ARIA-E were typically mild or moderate in severity and transient. We had 98% of them resolving. So the characteristics of ARIA-E in these trials is very similar to what we and others have reported previously. I think I'll skip over this scale. So in closing then, we believe that these 3 trials establish the safety and efficacy of aducanumab. Study 302 is a positive study with robust and internally consistent results across a number of clinical and biomarker endpoints. Study 103 is an independent and second study that provides supportive evidence. While Study 301 is a failed study formally, we believe that we understand the reasons for the differences between the studies, and we find in post hoc subgroups support for Study 302 and 103 in this study. In summary, we believe that consistent exposure to 10-milligram per kilogram aducanumab is effective at reducing the clinical decline in patients with early symptomatic Alzheimer's disease and has a favorable benefit/risk profile. And I'll end there. Thank you very much for listening.

Thank you, Samantha. That was great. I would -- in the interest of time and our audience jump right into the questions. The first group is about the specifics around the aducanumab studies 301 and 302. Question one. The FDA has a guideline on criteria for substantial evidence of efficacy. Does aducanumab satisfy those criteria?

Thank you, John. The answer is, yes, we believe that the data we have on aducanumab is -- does achieve the threshold for substantial evidence of effectiveness. Now that is for the FDA to decide, and that's obviously deliberations that are ongoing. But if we look back at the guidelines for substantial evidence of effectiveness -- I don't know if you guys are hearing my little dog barking at the door. So just to refer to the guidances, there's a draft guidance, at the moment, for substantial evidence of effectiveness so 2019. But fundamentally, the threshold for evidence of effectiveness has not changed since 1998 at the FDA. And in that guidance, it does support that the FDA does have regulatory flexibility. There are -- there is the ability to approve data on the basis of one robust single trial with corroborating data, and there are different ways that data can corroborate. And we do believe that 302 is a large, robust clinical trial with internally consistent results and supported not just by clinic -- multiple clinical endpoints, but also by biomarkers. Study 103 could fulfill the criteria for a supportive study in the context of that guidance. So I hope that's helpful.

All right, question 2. The placebo decline in Study 302 was substantially larger than the placebo decline in Study 301. Couldn't this difference account for the apparent drug placebo difference in the 302 study?

Thanks for the question. I know there's a lot of interest around this one. So right at week 78, CDR sum of boxes in Study 301 is slightly lower than Study 302. But we don't believe that the difference in placebo between the 2 studies is driving the effect. And let me outline, there's a number of points that we don't believe that this is the case. First, we actually planned for a placebo decline of 2.0 units throughout the 18 months of these trials. So in fact, both trials have a lower placebo decline than was programmed, making it more difficult for either of the trials to have shown efficacy. So the degree of placebo decline was lower, but it's nonetheless quite consistent with contemporary clinical trials in this patient population. The results that we have through the 18 months, if you look at the placebo at every 200 patients, is also -- actually, there's no systemic difference between the 2 studies. And then another thing to point out is, while there's a focus on CDR sum of boxes, if people look closely at the data, the MMSE actually is in the other direction. So in fact, the placebo decline in MMSE on 302 is less than on Study 301. But you do -- you still see an effective aducanumab in the high dose in Study 302. So that goes against the idea that it's placebo that's driving it. And last but not least, both studies have a fairly consistent and numerical difference on the low dose. So irrespective of the placebo decline, you have a consistency of effective aducanumab in the low dose. Had the placebo been driving the effect, it would have also squashed that low dose in 301, and that's not what we see happening. So on closer inspection, placebo is really important. You have to have decline to be able to see a difference, but we don't believe that placebo decline is driving that difference between the 2 studies.

Another interest of some breadth -- a question of interest is, are there lessons from EMERGE and ENGAGE regarding the use of futility analysis in trials that you might share with us?

Another good question. I think there was a bit of a knee-jerk reaction that occurred after our futility analysis. And some even had the position of, "Hey, you shouldn't do futility analysis." That's not a good idea. Futility analyses play an important role. If we can assess earlier in a clinical trial that a treatment truly is not effective, then we want to guard against patients staying in a clinical trial where we've determined that there is no efficacy or, worse, if there is a safety issue happening. And so conduct of futility analyses, they are important to do for the benefit of patients. However, as we have learned, despite incredibly intelligent and diligent work to predefine a futility analysis, I would say that the results educated us that the unexpected can still happen. And so one needs to certainly not believe that -- believe that one knows everything going into clinical trials. I think we need to have some humility and we need to take some time to understand results that come out from these types of assessments. So I would suggest that we try our very best to design these futility analyses, but we need to ensure that we take time to understand what we are seeing when we conduct them.

Thank you, Samantha. Now I think you've addressed this in your presentation, but it's probably worthy of a bit of repetition. Why should we discount the 301 study and value the 302 study? We hear this question all the time. Both studies had the same protocol and roughly the same completion rate, but 301 showed no drug-related improvement while 302 showed a small but statistically significant improvement for the high dose.

Yes. Thanks, John. I think it's a really good question. And the first thing is -- well I hope it doesn't come across that we are discounting Study 301. In fact, we're definitely not doing that. I think the formal language of statistics at least traditional statistics, one trial positive, one trial negative, gives that impression that we are viewing them very much in those ways. And formally, per prespecified statistical analysis, those are the outcomes of those studies. But you know what? There are thousands of patients in those studies, and we spent a great deal of time trying to understand both studies. We started from the standpoint of which of these trials is the right outcome? Where we have concluded is that 302 is positive. It's very, very unusual to see such consistency of clinical endpoints showing an effect. So that we have a positive primary endpoint, that's one thing. But we're also seeing the same effect on other types of tests. So cognitive performance tests, caregiver-input based tests. And so using all of these different paradigms that are not highly correlated at a point in time. So we determined that the result in the high dose in 302 is very unlikely to be a false positive. And so we also then need to understand so why wasn't 301 also positive? Again, we don't discount 301. It was, in fact, the low dose that opened our eyes, both the clinical and biomarker data that suggested something had occurred in the high dose. And so we really did focus our investigations to try and understand what the differences were between the 2 studies and what those differences may have caused. So the other discipline to statistical analysis is more of an exposure response, a quantitative pharmacological approach. And this -- today, we have the computing power to look at each patient and their exposure and their response on these clinical endpoints. And when we look at the data from those perspectives, it's much more of a continuum, patients having low versus high exposure. And those with the higher exposures having the greater effect on the clinical endpoints. When we look at the data from that perspective, that's how we learn that 301 has the same pharmacological response, but overall, had less dosing and had other interruptions as well. So we're definitely not discounting Study 301. We've learned a great deal by truly trying to understand Study 301.

So in light of everything you've seen living with this data, what might you, Biogen, do to move away from the old school, CDR sum of boxes into more objective approaches for measuring drug efficacy, the holy grail?

That's a tricky question. Well, there's a tension for trying to measure something in a patient's progression, and a change from that something in progression. And in these very early patients, they are not progressing very far in the timeframes that are reasonable to conduct clinical trials. So I really -- I share the angst that we're trying to measure something with tools that are, frankly, quite blunt in that period of time. But the tension is that for a treatment to obtain approval, we need to be demonstrating that what we're measuring is meaningful to the patients, right? So it needs to have meaning for the patients, and it needs to be a meaningful difference. So I think that in earlier clinical trials, trying to understand, is there a signal at all is where there is space for looking at more sensitive cognitive instruments. But in the domain where we're trying to prove efficacy, we were kind of trapped by that tension to show that clinical meaningfulness at the same time. One of the ways I think we are capturing clinical meaningfulness is not just focusing on the primary endpoint, but actually having a number of clinical measures covering, as I mentioned, different domains. So activities of daily living, independence of patients, behavioral symptoms of patients as well as what we refer to as cognition and function. So I think it's actually the combination of the endpoints that is more meaningful. Of course, that -- in trials that are aiming for regulatory approval, we have this issue with statistical analysis and hierarchies. So we have to be very thoughtful about the selection of endpoints. But I think for now, we may need to continue to use CDR sum of boxes for this stage of Alzheimer's patients. Although I think we and others have been happy to see the performance of ADAS-Cog-13 and the modified ADL scales in this patient population.

Great. So thank you, Samantha. Take a couple of deep breaths. And I'm going to move on to an area that is very much of interest to the broader community, which hypothesizes around if aducanumab is approved, what are the ripples effect associated with that approval? Specifically, first question, do patients need to have a positive PET result to be eligible to receive treatment if aducanumab is approved?

Okay. Well, we can't speculate what the label might be if we are approved. That's really the domain of the FDA. Although we do have a position in that we do believe that patients -- Alzheimer's disease is characterized by the presence of amyloid. And so we do think that it is important to ascertain the presence of amyloid pathology before initiating treatment, should we be approved with aducanumab. We used PET imaging exclusively in our trials for aducanumab. But we know that others use a combination of either PET or CSF. And there's a great deal of work being done to be able to have other methodologies, CSF predominantly, but maybe also in the future, blood-based tests for amyloid. So we do believe that, that is an important step. And we are very invested to try to help innovation in that space as well.

So the opposite question actually, in a way. Several researchers and treating physicians wanted to know when might -- what would your guidance be in terms of how to determine when to stop treatment?

Initiating or stopping treatment is really a patient-physician conversation. The data we have at the moment suggests or informs us that patients who have had sufficient doses for a period of time, are those individuals who are -- who may benefit from treatment with aducanumab. We do not, today, know the long-term basis of that. So how long does one need to continue to treat for? What are the signs, symptoms in regards to no longer treating with aducanumab? So the window of benefit is unknown. It's something we need to continue to learn, hopefully, with the community in conducting studies with longer duration. But also it's a very individual conversation, John, for a given family, for a given patient and in conversation with the physician in regards to -- aducanumab slows down the progression of disease, it doesn't stop it. And so there may be a point, and we don't know at what point that is, that a patient decides that it's no longer the right thing for them to be doing. So these are important questions and questions that we need to learn.

So this is not meant to be a yes, no question, but it just might be. Will the health care system be ready for the approval of aducanumab based on your PDUFA date?

Unfortunately, it is a yes, no question, but I want to give it more response than that, John. We don't have this type of paradigm of treatment available today. And as we were just discussing, diagnosis with a biomarker is likely to be an important step. That one piece right there is we know definitely an infrastructure that is not ready today. We're talking about patients at earlier stages of disease. We know how difficult it is to achieve those numbers of patients in the clinical trial setting. We know that patients are going undiagnosed at that stage of disease today. We know that the primary care setting isn't yet set up to funnel this particular disease through to specialists or to treatment. So those are a couple of things. There was a report by RAND Corporation a couple of years ago that spoke to, actually the number of specialists, the number of neurologists available would be another bottleneck. And then treatment-specific factors for anti-Abeta antibodies as a class would be the MRI monitoring, which some may be required to ensure the safe initiation of treatment, as well as these antibodies are given IV. And so infusion capacity is another place. So it's sad to give you a multi-part response of there are many areas that would not be ready at the point and time of the PDUFA date, God willing that we are approved. But I do know that there are efforts to work towards this. So we know that a lot of work will need to be done.