Eli Lilly and Company (LLY) Earnings Call Transcript & Summary

Ladies and gentlemen, thank you for standing by, and welcome to the Alzheimer's Update Conference Call. [Operator Instructions] As a reminder, your conference is being recorded. I would now like to turn the conference over to your host, Sara Smith. Please go ahead.

Good morning. Thank you for joining us for Eli Lilly and Company's donanemab update. I'm Sara Smith, Director, Investor Relations. Joining me on today's call are Dr. Dan Skovronsky, Chief Scientific Officer; Dr. Mark Mintun, Vice President, Alzheimer's Disease Development Unit; and Dr. John Sims, Senior Medical Director for Donanemab. During this conference call, we anticipate making projections and forward-looking statements based on our current expectations. Our actual results could differ materially due to a number of factors, including the extent and duration of the effects of the COVID-19 pandemic as well as other factors listed on Slide 3 and those outlined in our latest Forms 10-K, 10-Q and any 8-Ks filed with the Securities and Exchange Commission. The information we provide about our products and pipeline is for the benefit of the investment community. It is not intended to be promotional and is not sufficient for prescribing decisions. I'll now turn the call over to Dan to provide some introductory remarks.

Thanks, Sara, and thanks to everyone for joining us today. This is an exciting time for Alzheimer's patients and for Lilly. As a scientist and a physician who's been working in Alzheimer's disease for the past 2 decades, I'm delighted to share additional details about donanemab, which showed significant slowing of Alzheimer's disease progression on both clinical outcomes and disease pathology in the TRAILBLAZER-ALZ study. We'll highlight the contribution we believe these results make to the field and share with you expected next steps for donanemab. Before we discuss detailed results, I'd like to start by highlighting our journey in understanding Alzheimer's disease, which laid the foundation for the exciting donanemab data we are discussing today. Starting on Slide 4. Alzheimer's disease remains one of the most significant unmet medical needs in the society. It is now the largest cause of death without approved treatment to modify disease course and is projected to exceed $1 trillion in annual cost to society in the U.S. alone by 2050. Despite decades of efforts and investments, there has been little progress for patients. Why is this? Some of the challenges are highlighted here. First of all, today, Alzheimer's disease is still principally diagnosed and treated based on symptoms, like a psychiatric illness, rather than based on pathology we now know is present in the brain. Imagine taking a symptoms-based approach to treating cancer and how little progress we would have made had we focused solely on symptoms. For over a decade, Lilly has set out to change how we diagnose and treat Alzheimer's. And we've created a set of tools that lets us see and measure brain pathology directly. This has unlocked new ways to conduct trials, and we believe this is clearly the future for continued progress in AD. You'll hear more about that shortly. The second related issue we face in Alzheimer's disease is that noisy trials have led to a replication crisis. Even nearly identical trials conducted simultaneously, evaluating the same drug can show discordant results. This is caused by 2 main factors. First, as I said, most trials set enrollment criteria based on symptoms rather than pathology. Thus, they enroll a heterogeneous population with wide variation in levels of underlying pathology, or worse, they enroll patients with different underlying diseases. Accordingly, these patients progress at very different rates and intra-group variability, measured by standard deviation of the mean, for example, is quite large in most AD trials. Second, the population heterogeneity problem is compounded by intra-subject variability, driving noise in the outcome measures. Cognitive performance can vary widely from day to day. Measurement error can be large, and outcome measures often rely on subjective assessments. Partially as a consequence of these 2 factors, disease-modifying trials in AD prior to TRAILBLAZER-ALZ have been quite noisy, failed their primary objective. And sponsors have been left picking through secondary endpoints and subpopulations looking for signal upon which to build. The repeated failure to generate replicable results has created confusion in the field, with important hypotheses remaining neither fully tested nor fully refuted and therapeutic nihilism becoming increasingly pervasive. Finally, the third major challenge we face is highlighted on the right side of the slide. Despite the fact that Alzheimer's disease is insidiously progressive and uniformly fatal without any disease-modifying therapies, society as a whole, inclusive of industry and governments, isn't willing to lean into Alzheimer's disease drug development in the same way we have for oncology: aggressively attacking validated targets, using biomarkers to stratify patients and using surrogates and accelerated regulatory pathways to bring new medicines to patients with urgency. In one sense, in the face of the diagnostic challenges, the noisy trials and the replication crisis that we've highlighted, this failure to lean in is actually a rational approach. But given the advances in the field, we predict the situation can and will change dramatically. As you can see on Slide 5, our progress in AD has been hard won. Despite what have been disappointing failures up until now, each setback informed subsequent advances on biomarkers, clinical outcome measures and statistical methods. For example, solanezumab's EXPEDITION1 and 2 were the first trials to use large-scale amyloid imaging at baseline. As a result, we were able to follow progression of amyloid-positive versus amyloid-negative patients and learned that amyloid-negative patients don't show progression of impairment and thus should not be enrolled in Alzheimer's trials. That once-controversial observation is now standard practice in the field, and nearly every trial screens for amyloid positivity at baseline. In EXPEDITION3, naturally, we screen for amyloid at baseline, but now we advanced the field again by including tau imaging. As a result, we learned that baseline tau predicted subsequent cognitive decline, opening the door to trials that stage patients based on both amyloid and tau. In time, we expect this will also become standard in the field. Obviously, these advances in clinical trial design wouldn't have been possible had it not been for advances in developing diagnostics and biomarkers. Those efforts include Amyvid, an amyloid PET tracer approved in 2012; Tauvid, the tau tangle PET tracer approved in 2020; and phospho-tau217, a well-validated blood-based biomarker for tau created at Lilly with an in vitro diagnostic currently in development. Slide 6 shows how we've woven our insights and capabilities together to enable development of donanemab. We've coupled our ability to pathologically stage patients by amyloid and tau levels with donanemab, a rapid and deep plaque-clearing anti-N3pG A-beta antibody designed to fully engage the target of amyloid plaque. We used our insights to design the trial to reduce noise and give us more confidence in the results by focusing on a homogenous population using an improved endpoint in iADRS and adding a sensitive statistical method that weighs the full patient course over time rather than unduly focusing on a single time point to assess efficacy. We're delighted that these strategies resulted in a successful outcome. The Phase II results give us confidence that donanemab has the potential to become a very important medicine for Alzheimer's disease. We're also excited about the contribution these results make to the Alzheimer's field. We've advanced the field in several ways here. First, our results link robust plaque clearing to slower cognitive decline, hence helping to confirm the amyloid hypothesis. Second, we've showed that tau staging can be used as an important future standard for running AD trials and selecting patients for therapy. Third, we provided data to validate regional tau spread as an important surrogate for disease progression and drug effect, akin to objective response rates in oncology. We hope these advances will set the stage for even more progress in Alzheimer's disease in years to come. Now I'll turn it over to Mark to share more details and insights from the TRAILBLAZER-ALZ trial.

Thanks, Dan. Before we get into the donanemab data from TRAILBLAZER-ALZ, I'd like to spend a few minutes to discuss our observations around the importance of staging patients by tau levels and the utilization of tau as a biomarker for disease progression. The solanezumab EXPEDITION3 trial was the first therapeutic trial to incorporate tau imaging. When we reviewed those scans, we noticed that baseline tau was telling us something important about the patient's cognitive decline in the study. On Slide 7, you can see the analysis we were able to conduct using pooled data from the patients across the EXPEDITION3 trial with those from a flortaucipir development trial. We have previously shown that amyloid positivity was critical for predicting decline, so we were eager to see if tau staging provided any additional prognostic information. The figure comes from dividing the 241 patients into 4 equal subgroups by their baseline global tau levels and plotting the subsequent decline over 18 months for each group as measured by the Mini-Mental State Examination, or MMSE. We immediately saw that despite them all being amyloid positive, the lowest quartile of patients, shown on the left, had very little apparent tau on their scan and then declined at only a fraction of the rates seen in the other groups. In contrast, the highest tau quartile declined at a rate almost double the group average. Visually, we noticed that tau scans showed not only higher intensity of uptake but a wider distribution of tau pathology throughout the brain. Finally, the 2 middle quartiles declined similarly to each other and, not surprisingly, similar to the average for the entire group. As we collected more longitudinal data and as our measurement tools become more sophisticated, we continue to investigate this relationship between baseline tau and subsequent cognitive and functional decline. On Slide 8, we subsequently show that baseline tau levels in EXPEDITION3 were a highly significant predictor of clinical progression over 18 months as measured by the iADRS. This is shown for the global brain measure of tau. But in addition to the global tau, the frontal lobe tau stood out as a region having highly significant predictive value for progression. And then as shown in the table below the correlation plots, this relationship was replicated with the data from the AMARANTH lanabecestat trial. These data and our own work suggested that higher levels of tau represent more advanced pathological disease and predict a more rapid decline in cognition and function. In addition, a growing body of independent evidence in the field helped reinforce this conviction that tau was a biomarker for predicting disease progression. Moving to Slide 9. We show here how we are applying this knowledge in TRAILBLAZER-ALZ, which is the first study to screen and enroll patients based on their tau pathology. We now recognize that simply knowing a patient has abnormal amyloid is not enough. A key predictor of clinical progression is the presence of sufficient tau. And Tauvid, our PET imaging diagnostic, enabled us to measure tau burden and appropriately stage in this trial. The TRAILBLAZER-ALZ study excluded patients that were in the no/very low tau group as they were expected to have no significant decline within the study's 18-month window. It is important to note that essentially all amyloid-negative subjects will fall into this no/very low tau group. So this screening step removes both amyloid-negative patients and the amyloid-positive patients with minimal tau levels. When screening patients with symptoms of early Alzheimer's, the excluded no/very low tau group represents about 40% to 60% of the population. The study also excluded high tau patients, roughly about 10% to 15% of that same population. And we hypothesized at such advanced tau pathology, they would not benefit from an anti-amyloid therapy. The study thus included those amyloid-positive intermediate tau patients who represented about 40% of patients presenting with symptoms of early Alzheimer's disease. Lilly has been a leader in staging patients for amyloid, which has now become the standard approach in Alzheimer's disease study. In the same way, staging patients by tau pathology is our new standard, and we expect the field to follow as it did for amyloid. In addition to tau staging, we also expect longitudinal measures of tau spread to become increasingly important, but more about that later. Slide 10. We outlined several key attributes of the TRAILBLAZER-ALZ study design. After being randomized, either donanemab or placebo, participants were observed for 76 weeks. I would note that when the study was initiated, it originally included a third arm, combining donanemab plus a base inhibitor. This arm enrolled 15 patients before we terminated the base inhibitor program and, therefore, discontinued that arm. Participants were dosed every 4 weeks, and those receiving donanemab would start with 700-milligram dose for the first 3 doses and then receive 1,400 milligrams every 4 week thereafter. This rapid up-titration was designed to achieve high amyloid clearance early in the trial. Amyloid scans were performed at 24 weeks and 52 weeks. It is important to note that once the patient on the donanemab arm achieved a negative amyloid scan below 25 centiloids, their dose was titrated down to either 700 milligrams of donanemab or down all the way to placebo based on steps detailed in the footnotes. You can see on the chart the impact of this treat-to-target dosing algorithm. As shown at 24 weeks, over 1/4 of the dose patients were receiving placebo. And at 52 weeks, over half the dose patients were receiving placebo. Moving to Slide 11. You will see the baseline characteristics for TRAILBLAZER-ALZ. Cognition and function metrics were comparable and well balanced across the arms as was the proportion of apoE4 carriers. The starting centiloid level of amyloid plaque was 101 in the placebo arm and 108 in the donanemab arm. We note that by screening based on pathology rather than on clinical measures, we actually included a wider clinical range of patients. Indeed, the average MMSE values of about 23.6 indicates that even the average patient in our trial had an MMSE score below that is worse than the minimum level for inclusion in some other trials in early Alzheimer's disease. So there might be something unexpected here. Previous trials in AD that were trying to focus on early disease used strict rules on how clinically impaired the patients could be, such as they had to have an MMSE greater and equal to 24. And that would minimize the number of patients with more advanced disease in their trial. But as we've seen over the last decade, clinical measures are a poor proxy for predicting pathology. We learned this about predicting amyloid pathology, and we're learning this now about tau pathology. So in TRAILBLAZER-ALZ, we limited the range of Alzheimer's pathology, excluding patients with pathology likely to advance to respond to anti-amyloid therapy. But we significantly widened the range of clinical presentations, allowing those down to MMSE of 20. We felt that patients that might be considered too impaired for treatment in a conventional development program might be found to actually have early AD pathology and still respond to treatment. We believe our current results support that hypothesis, and we see the potential ability to offer hope of an efficacious treatment even to patients with more advanced clinical impairment as an unexpected benefit. Moving to Slide 12. In TRAILBLAZER-ALZ, the safety profile revealed no new safety signals. ARIA-E was present in nearly 27% of the patients, similar to other trials with plaque-clearing agents, while 6% of treated patients had symptomatic ARIA-E. Turning to the treatment discontinuations due to AEs. ARIA was, by far, the largest reason for donanemab treatment discontinuation. This was due to a prescribed and, in retrospect, highly conservative rule set for stopping treatment after ARIA-E or ARIA-H, including those with superficial siderosis. Discontinuation of treatment was not at the discretion of the investigator or patient. Specifically, the rules for ARIA-H drove most of the ARIA-related discontinuations. Across both types of ARIA, a large majority of discontinued patients were symptomatic. And there was no evidence that patients with ARIA had differential treatment benefit compared to those without ARIA. While some patients opted to leave the study once they were no longer eligible for treatment, most stayed in the trial. And even though they were no longer on therapy, their data was collected and used in all the analyses. With respect to patients who did discontinue the trial, missing data was not imputed. Elsewhere on the trial -- on the slide, you will note that approximately 8% of patients treated with donanemab had infusion-related reactions. Antidrug antibody levels were consistent with Phase I. Despite a majority of patients showing ADAs, donanemab was still able to engage the target and resulted in 68% of the patients being amyloid negative at week 76. Before I discuss the details of the efficacy data seen in TRAILBLAZER-ALZ, I want to show the placebo arm performance across several outcome measures in the context of other relevant Alzheimer's trials. As we have previously stated, it's important to evaluate how the placebo group performed in a clinical trial to rule out the possibility that the placebo arm declined unusually fast, leading to a false-positive study result. You can see on Slide 13 that in line with our expectations, the 18-month placebo performance in TRAILBLAZER-ALZ showed a similar decline, if anything, it was a little bit slower decline to other Alzheimer's trials across several outcome measures. And therefore, we do not believe unusual performance in the placebo arm is responsible for the positive study results. Moving to Slide 14. The primary endpoint in the study was the MMRM analysis of the integrated Alzheimer's Disease Rating Scale, or iADRS, a composite tool created by adding 2 well-accepted measures, the AD Assessment Scale-Cognitive Subscale, or ADAS-Cog, and the AD Cooperative Study-instrumental Activities of Daily Living, ADCS-iADL. We believe the iADRS is more consistent and sensitive measure to detect treatment differences than other Alzheimer's disease scale. It was gratifying to see the novel design elements of TRAILBLAZER-ALZ lead to a remarkably consistent meaningful effect on the primary endpoint. The 32% slowing of decline compared to placebo translated into a roughly 6 months improvement in disease progression, which we believe is a very meaningful result for a study of this length. Although this study only followed patients for 18 months, we are also excited to see the potential impact on decline beyond that time point based on our current understanding of tau pathology, which I'll discuss in a moment. Donanemab is the first potentially disease-modifying drug for Alzheimer's disease to meet its primary endpoint at the primary analysis in a late-stage trial. Our confidence is bolstered by the curve starting to separate at 24 weeks, reaching nominal statistical significance at 36 weeks and continuing through the end of the study. While MMRM analysis of the 18-month time point was the primary endpoint, one should be cautious about over-interpreting any single time point in any trial, especially in Alzheimer's disease, where the last time point is often the noisiest for reasons we don't really fully understand. Even in large AD trials, a trial that looks like it will surely fail based on failure to see drug effect and earlier data, relying on earlier time points can be rescued by a surprisingly large effect measured at the last time point. And vice versa, a large trial with small but growing effects over time that looks like it will succeed can fail because the last time point has a surprisingly small drug effect measured. So we're encouraged in the case of TRAILBLAZER-ALZ that the interpretation of drug effect is not reliant on merely the last time point but rather can also be concluded based on the effects seen throughout the treatment period. Towards that end, the disease progression model is a preferred alternative as it assumes a proportional treatment effect relative to placebo and fully weighs treatment effect during the entire trial period. I'll describe that momentarily. On Slide 15, an exploratory analysis was performed to examine the efficacy, measured by iADRS, of participants by baseline tau levels within the intermediate tau population we enrolled. For this analysis, those same intermediate tau patients were further divided into terciles by baseline tau PET. Although we're cautious about making definitive conclusions given the small sample size now in each group, the data suggests that donanemab had the largest effect in participants at the lower tau levels -- with lower tau levels at baseline. This effect is best seen on iADRS given the small sample sizes and enhanced sensitivity of this outcome measure, but it can also be seen on the CDR Sum of Boxes. Why are patients with higher baseline tau potentially showing less drug effect? As seen on the right, perhaps patients who are at the top end of this intermediate tau category are potentially in or are nearing an amyloid-independent stage of disease, where amyloid therapies are less likely to show slowing of disease progression. By extrapolation, those patients who were excluded from TRAILBLAZER-ALZ due to high tau would be even less likely to have slowing of progression. We are going to test this directly in all -- in TRAILBLAZER-ALZ 2. This is important because treating patients for whom anti-amyloid therapies do not work creates risk due to ARIA or other side effects without potential benefits. It's also important to identify patient groups that might be likely to receive the greatest disease slowing from any new potential drug as these would be the most urgent candidates for treatment. Based on our understanding of AD biology, we would expect the impact of baseline tau to be a constant across all amyloid-lowering drugs. It's, therefore, likely it will be important for drugs in Alzheimer's disease, like many drugs in other therapeutic areas, the pathologically stage patients and understand the stages in which specific drugs are or are not likely to work. While patient stratification based on biomarkers is sometimes initially resisted, it's becoming a proven method to maximize patient benefit. And in some cases, the precision medicine ends up overtaking the medicine that does not stratify patients. We look forward to the potential for further insights into tau pathology from the larger TRAILBLAZER-ALZ 2 study that's under way. On Slide 16, you can see that while this trial wasn't powered to hit on multiple endpoints, nevertheless, the key prespecified secondary endpoints all moved in a similar positive direction and with roughly similar magnitude, underscoring our conviction and the strong efficacy observed in patients treated with donanemab. The slowing of both cognition and functional decline was consistent across endpoints and time points with all secondary measures achieving nominal significance at least one time point, and we observed early separation between placebo and donanemab arms across measures. While CDR Sum of Boxes did not achieve statistical significance at 76 weeks, we were pleased to note that it achieved nominal significance to earlier time points and delivered a percent slowing of 23% at the final time point. While cross-trial comparisons have limitations, such as differences in study design or patient population, this is as large of a percent slowing in effect size that has been seen to date. And again, this result was not dependent on an anomalously good performance at the 18-month time point. The variability seen across time points and outcomes such as CDR emphasize the need for a statistical model that better takes all the data into account. Moving to Slide 17. We also prespecified exactly such a second analysis approach, the Bayesian disease progression model, which generates a single estimate of progression over the entire 18-month study. Applying the DPM analysis for each time -- excuse me, each endpoint, we were encouraged that all the outcome measures showed slowing in the range of 21% to 32%. Importantly, all of the 95% credible intervals for donanemab drug effect separated from placebo, providing further support for the primary study results. Slide 18 shows the significant plaque clearance in patients treated with donanemab. 40% of the patients in the treatment arm achieved amyloid-negative level by week 24. Further, by the end of the study, nearly 70% of patients treated with donanemab were considered amyloid negative with an average plaque reduction of 85 centiloid observed at 76 weeks. An 85 centiloid reduction translates into an SUVR reduction of negative 0.46. And while cross-trial comparisons have limitations, this reduction stands out compared to a negative 0.23 and a negative 0.28 reduction for aducanumab high-dose arms in the 301 and 302 studies at a similar time point. Not only is the depth of clearance with donanemab noteworthy but also the speed at which it was achieved. For example, while cross-trial comparisons can have limitations at 6 months, donanemab patients achieved nearly a 68 centiloid reduction or a negative 0.37 SUVR compared to a reduction for aducanumab high dose of approximately 0.08 in this 302 study. The remarkable depth and rapidity of the amyloid removal may account for the magnitude of the clinical benefit as well as the consistency across endpoints. Indeed, across various fields of medicine, it is usually true that the more complete treatment of disease pathology ultimately results in greater benefit for patients. I started my remarks showing our historical data indicates tau predicts subsequent cognitive decline and that we sought to confirm this in the TRAILBLAZER-ALZ study. As our staging criteria excluded the lower and upper portions of the tau spectrum, we were pleased to see once again a correlation between baseline tau and cognition despite a very narrow range of tau in the study. Slide 19 shows again the 3 graphs we discussed for Slide 15, while adding in the data in the green box showing the significant correlation in baseline frontal tau SUVR with the change in iADRS and CDR Sum of Boxes over 76 weeks. While there were correlations observed for many regions, frontal tau is the region that appeared most predictive of subsequent decline. This informs our interpretation of longitudinal tau data. All subjects completing TRAILBLAZER-ALZ trial had longitudinal tau scan at 76 weeks to measure the change in tau over the course of the study, as shown on Slide 20. Global tau, as measured by a novel algorithm created by Invicro, called TauIQ, showed a nonsignificant reduction of approximately 10% at 76 weeks for patients treated with donanemab. We also performed prespecified exploratory analyses on the major brain regions using more conventional SUVR analyses. These regional analyses revealed significant slowing of the tau increases across the frontal, parietal and lateral temporal areas, not in the occipital lobe, though. It is worth highlighting that the percent slowing observed in the frontal lobe was nearly 60%. This leads us to believe that donanemab does have an impact on reducing regional tau load in the brain. The large impact on the frontal lobe tau accumulation appears particularly meaningful, given how pathological involvement of this region is likely related to future decline. So while we turn to Slide 21, I want to reinforce the comprehensive data set that we have generated as we close this section on TRAILBLAZER-ALZ. Donanemab is the first therapy of its kind to meet its prespecified primary endpoint at the primary analysis, showing a 32% slowing of decline as measured by iADRS. The totality of the data shows a strong efficacy signal and supporting biomarker data with consistent improvements observed across all secondary endpoints across analysis methods and across time points with early separation from placebo on functional scale. These data give strong support to our ability to replicate these results in the ongoing TRAILBLAZER-ALZ 2 study. And finally, the study provided further affirmation of the need to stage Alzheimer's disease patients by tau pathology in the development of treatments and highlights its potential as a surrogate for disease outcome. Now turning to Slide 22. The TRAILBLAZER-ALZ 2 study has been screening and enrolling patients since last summer. Like TRAILBLAZER-ALZ, we are staging amyloid-positive patients by their tau pathology with patient groups stratified for intermediate or high tau burden. As we incorporate this insight from completion of the first TRAILBLAZER-ALZ study, we are making several key changes. These changes include: changing the trial designation to a Phase III study, reflecting our confidence in the results for donanemab; focusing the study on intermediate tau patients for the primary analysis; continuing to enroll high tau patients to inform future treatment guidelines; increasing study size to 1,000 patients having intermediate tau burden; and changing the primary endpoint to iADRS with CDR Sum of Boxes moving to a key secondary endpoint. Excluding the high tau patients from the primary analysis population makes sense, given the decrease in treatment effect with increasing baseline tau levels seen in the intermediate tau terciles of TRAILBLAZER-ALZ. Continuing to enroll the high tau patients in the study as well as prespecifying subgroups of the intermediate tau patients for analysis is useful because it will potentially inform future treatment guidelines. Now there are several reasons why we're increasing sample size. These include to maximize the chance of success, not just for primary analysis but also for key secondary outcome measures. It allows us greater power to confirm subgroup effects, including based on tau substages such as the lowest tau tercile group, and it allows us to generate a larger safety database. As I previously noted, we had as large of an effect size as ever demonstrated on CDR Sum of Boxes in TRAILBLAZER-ALZ. While we were pleased to see the results, we still believe CDR Sum of Boxes has not been a reliable measure. We list on Slide 23 the only 2 sets of Alzheimer's disease trials in the last decade that were enrolled with similar populations, intending to show the same effect: EXPEDITION1 and 2 for solanezumab and Studies 301 and 302 for aducanumab. While cross-trial comparisons have limitations, you can see that despite recruiting similar and large populations, both pairs of trials failed to replicate the effect size of CDR Sum of Boxes from one trial to the other. Replication is absolutely key to successful drug development, and our goal is to minimize our dependence on a measure with questionable reliability. On the other hand, if we look at Slide 24, we can see that iADRS was a much more consistent measure in EXPEDITION1 and 2 than CDR Sum of Boxes. Recent data reported by Biogen also shows more consistency across their 2 trials when using an iADRS-like composite than with CDR Sum of Boxes. The consistency of iADRS grows ever more compelling with the logical use as a key measure because it's -- both of its components are valid and meaningful. Now we acknowledge that iADRS involves a risk if the 2 components are divergent as even a positive signal would need to be interpreted with caution. We were, however, pleased to see that both ADAS-Cog13 and the ADCS-iADL showed positive trends in TRAILBLAZER-ALZ. Now I'll turn the call back over to Dan to wrap up.

Thanks, Mark. Before I close, I want to highlight the next steps for donanemab shown on Slide 25. As we move forward with donanemab, we believe replication of TRAILBLAZER-ALZ results is important because there continues to be great uncertainty in the field. Specifically, the amyloid hypothesis continues to be questioned. Given failures of replication, this is not an unexpected viewpoint. Launching a new medicine into this environment could lead to suboptimal decisions for patients, prescribers and payers. For this reason, Lilly committed to the path of replication by starting TRAILBLAZER-ALZ 2 even before we had the TRAILBLAZER-ALZ 1 results. Our conviction around the necessity of this approach is unchanged. Further, replication allows us to answer important questions for the field. Most significantly, confirmation of subgroups that may show no benefit from amyloid lowering, such as the high tau patients. This will be important to inform future treatment guidelines for this potentially important class of medicines. Even so, we clearly recognize the depth of unmet medical needs in Alzheimer's disease and the urgency to bring new therapies to patients. We're committed to working with the FDA and global regulators to find a path that addresses urgency on behalf of patients and allows for replication. That's the focus of our ongoing efforts with TRAILBLAZER-ALZ 2. Our expectation is that despite the additional 1,000 patients, we'll still complete enrollment for the study in the second half of this year, with primary efficacy data from the 18-month time point available in the first half of 2023. Our base case expectation remains that TRAILBLAZER 1 and TRAILBLAZER 2 together as 2 positive trials will be required for approval. However, we also recognize that regulatory expectations could shift, and we'll watch actions from the FDA and other regulators closely and adjust our plans accordingly. Further, given the competing needs for both replication and for bringing solutions to patients as quickly as possible, there could be new pathways for accelerated approval at various time points between the 2 bookends of submission with TRAILBLAZER-ALZ alone and waiting for the full results of the second study. We will explore submission possibilities more fully through our ongoing discussions with regulators, and you can be confident that Lilly will be prepared to move forward with urgency on behalf of patients. Before we move to Q&A, a frequent question from investors is how to think about sizing the opportunity for donanemab to help patients. On Slide 26, you can see a few key considerations for how we're thinking about bringing this potential medicine to market. Clearly, we see limited treatment duration as an important benefit for patients and for health care systems. Ideally, we can time limit costs and side effects while hopefully creating a lasting benefit. Similarly, we see our investments in capabilities in diagnostics for Alzheimer's disease as a sustainable advantage, which is critical for detecting and staging disease. By combining our expertise across blood-based tests, imaging agents and therapeutics, we hope to be uniquely positioned to bring integrated solutions to patients and health care systems. I started the call recognizing the significant challenges and unmet need in Alzheimer's disease. On the right side of the slide, you can see some numbers related to the population we hope to help with donanemab. The current population of patients with symptoms of early Alzheimer's disease, inclusive of mild AD and mild cognitive impairment, is roughly 4.5 million in the U.S., 5 million in Europe and 4 million in Japan. This is a large population that continues to grow. Of these, we expect 30% to 45% to be the amyloid-positive intermediate tau patients, similar to those we've studied in the TRAILBLAZER-ALZ program, which translates to roughly 1 million to 2 million addressable patients in the U.S. and 4 million to 5 million more around the world. It's also important to keep in mind this population will likely grow over time and that every patient with Alzheimer's disease will progress through this phase at some point during the course of their illness. In summary, Alzheimer's remains a devastating disease that significantly impacts millions of patients as well as families and caregivers. The industry has spent decades trying to find medicines that could slow disease progression, and Lilly has made a very meaningful step forward with the positive TRAILBLAZER-ALZ results. A critical step necessary to meaningfully improving outcomes for patients is the generation of compelling efficacy data that is reproducible and unequivocal. We're committed to providing that firm foundation with donanemab and across our AD portfolio. Beyond Alzheimer's, we shared exciting readouts over the last several months for tirzepatide, LOXO-305 and our COVID-19 neutralizing antibodies. This string of positive readouts on new molecules is not a coincidence. But rather, it is a consequence of our R&D strategy as we focus on following the science to attack well-understood biology with highly validated targets, making differentiated molecules with unique properties to fully address these targets and then testing them in novel clinical trials with novel designs that follow the science and the patient needs, setting new standards in the field rather than restricting ourselves to the usual approaches. We're emboldened by these exciting readouts, and we are energized to continue our efforts to help patients across our therapeutic areas. This concludes our prepared remarks. Now I'll turn the call back over to Sara to moderate the Q&A session.

Thanks, Dan. [Operator Instructions] Lois, please provide the instructions for the Q&A session, and then we're ready for the first caller.

[Operator Instructions] And the first question is from Terence Flynn from Goldman Sachs.

Great. I guess 2 for me. I was wondering if you have regulatory buy-in on iADRS at this point. And how are you powering the -- that primary endpoint given the change for TRAILBLAZER-2? And then, Dan, given all your comments on the importance of tau, just wondering how you're thinking about your tau drug-targeting strategy there and remind us of any updates on kind of when we might see the next set of data from you guys.

Thanks, Terence. We'll go to Dan for the first question and Mark for the second on tau?

Yes. Thanks, Terence. Your first question on regulatory buy-in for the iADRS is an important one, for sure. I think I won't get into detail on back and forth discussions with regulators, specifically. But I'll say this is a conversation/discussion we've been having for years. As we've gotten more and more data on this endpoint, we've become more and more convinced of its utility. The TRAILBLAZER-1 readout that we just shared, I think, further reinforces that confidence. As you can see, even just focusing on the placebo group, how well behaved this outcome measure is. So as Mark commented, there is always risk in using a novel endpoint. In this case, with iADRS, the risk is related to the 2 different components of it, which could diverge. So for example, if you had a drug that had a dramatic improvement on the ADAS-Cog and that side of iADRS, but on the other hand, even modestly worsened patients on the ADLs, the functional side of iADRS, you could still have an overall positive score, but that wouldn't be a drug that would make sense to take forward for approval. So you can think about that kind of risk in iADRS. And in our case, we were encouraged that we don't see that. The 2 sides of it move together. Your second question was around powering the study with the new increased sample size and iADRS as a primary endpoint is very highly powered to show a positive result here, and we're confident it will do so. As I mentioned, by increasing the sample size, we're also adding to the power that we show positive effects on secondary endpoints, which although noisy, are -- should still be possible. I'll -- yes, to Mark for the second question.

Yes. So about our tau portfolio, and clearly, there -- it's not an accident. We've obviously been very focused on the tau pathology and the development of the -- our tau tracer with recent approval of Tauvid is certainly a component of the overall approach to attack this pathology. We -- as you know, we have a couple already in the clinic with our zagotenemab, which is an anti-tau aggregate antibody, and that's in Phase II that we'll report out later this year; and then an OGA inhibitor, a small molecule, which is in Phase I. We're very excited about those 2 approaches. They're very different. You'll notice one is an antibody against the -- that spread of tau. We believe that's the best explanation for the mechanism we're seeing in the preclinical data. And the spread of that tau from neuron to neuron can be interrupted, we hope, with that antibody. While the OGA inhibitor, we hope, can actually slow tau aggregation in the neuron. Very different approaches. And I think as you point out, our ability to actually image the tau both for staging the patients going into it, and we could see different hypothesis for what type of patients who would want to enroll in those 2 very different agents as well as being -- now to follow that with the evidence that we've seen here with donanemab being able to slow tau progression and the potential use of that as a proxy biomarker for success, we see that is playing an even bigger role in the anti-tau therapy program. So I think it is indeed highly linked, and we'll be integrating those as we've done here with donanemab.

The next question is from Ronny Gal from Bernstein. The next caller is Tim Anderson from Wolfe Research.

A couple of questions, please. So the stock is down 9%. You could claim some of that's due to a sell-on-the-news phenomenon or perhaps there's disappointment from investors who expected this to be a totally perfect and pristine data set. In your opinion, what have you seen from analysts or investors that say there might be some misunderstanding or mischaracterization of the data that could also be leading shares to be lower today? And then second question is one of the things I found puzzling with the data was a lack of correlation between plaque reduction and improvement in cognition and function. There was no data shown on this. It was only a mention in The New England Journal. But the obvious premise, I think, with plaque-targeting mAbs is that removing plaque drives the clinical effect. If so, then why wasn't there a correlation?

Thanks, Tim. We'll go to Dan for the first question and Mark for the second.

Yes. Thanks, Tim. I appreciate that expectations may have been for a perfect and pristine data set. That might be unreasonable for an Alzheimer's study, particularly for an Alzheimer's Phase II study. Having said that, what's the misunderstanding here? I don't know. I'm not focused on short-term stock movements here. I'm focused on the long-term future for Alzheimer's patients. I think that's gotten very much more bright with our data. The reason is because of the purpose of Phase II trials is really to understand whether or not you have a drug here. I think when you look at this data set, even more so than any other Phase II data set or even a Phase III data set that's out there, you see a package of results that are highly compelling that this will indeed become a medicine. That, as we've outlined, is based on the consistency across time, and I think that's unique here. The consistency across endpoints, and I think that, again, is unique; and then consistency across statistical methods. We've shown 2 different statistical methods that were applied here. I think based on questions we've received, we even looked at a third statistical method here, which also showed great consistency, in fact, with many of the secondary endpoints reaching stat significance. So I think that overall package data is probably one thing. The second thing that I'd just point at is the tau biomarker. The fact here that we're reducing, slowing in regions of the brain affected by Alzheimer's disease, particularly in the frontal lobe, which predicts longitudinal progression of disease, is incredibly encouraging and was quite a pleasant surprise for us as we dug into the data. Mark, for amyloid correlation?

Yes. I mean it's an interesting question. And of course, by category, there's certainly a relationship. It's not like half the placebo patients went down. And so we -- the placebo patients didn't change at all in amyloid and the treated group did. And they separated as far as the iADRS and as we've talked about on the other measures. So there's certainly a relationship between decreasing amyloid and having slowing of the cognitive decline. And of course, when you look across the whole field and you look at the different attempts at different dose levels, some with more amyloid removal than others, there certainly seems to be a trend towards the higher amyloid removals or having the higher effects on cognitive slowing. The -- when you actually look at a specific trial, we do not have a lot of dynamic range here because of the huge number of patients that went all the way down to basically effectively negative. There wasn't a lot of dynamic range for pulling out a signal there. The one thing that I can note is that we actually -- because of what the strongest single -- the strongest signal we have from this is the slowing in the tau accumulation. And one of the things -- that's the one that hit with the highest p-value. It has the greatest connection to the pathology, obviously. And one of the things that is interesting about that is when you do parse the data set into the people who did achieve amyloid negative versus the people who didn't achieve amyloid negative, there was a clear increase in the amount of slowing of tau accumulation. And there was more increase -- there was more slowing in the tau accumulation for the amyloid-negative group than the other. So if you do this by category, I think the signal is there. But one thing that maybe is at the heart of some of these questions is if you propose that a drug, an antiplaque drug works in many different ways, maybe that makes a difference. The neat thing about donanemab is that it really only does one thing. It removes plaque. So if we're seeing a cognitive slowing, which we are, then it's almost certainly due to removing plaque. I don't think that ends up really remaining the questions.

The next caller is Ronny Gal from Bernstein.

Let's hope we've got better luck now.

Yes, we hear you.

All right. Excellent. So I have 2 questions. The first one is about what is going to be a clinically meaningful difference here. In the first paragraph of discussion of your New England Journal of Medicine paper, you kind of mentioned you achieved 3.2 point difference on the iADRS, but you attempted to get 6 points. And what is the minimal requirement is not clear yet. Can you just discuss a little bit how you're thinking about it? How are you going to prove it? And how should we think about what will be good enough in terms of clinical use? Second and somewhat related, you haven't mentioned per-protocol efficacy. Given the level of dropouts you had, can you talk a little bit about how that have changed the data and whatever you are willing to share with us on the per-protocol efficacy of this treatment?

We'll go to Dan for those.

Yes. I'll start, and maybe my colleagues will weigh in. So first is what's a clinically meaningful change and why did we set up the study to look for 50% slowing? Look, we're ambitious. We want to actually stop Alzheimer's disease. I think that's something for the future. These data create the possibility that perhaps by going earlier, perhaps by hitting with multiple mechanisms, we could someday fully stop Alzheimer's disease. For now, I think we've shown, even in the whole population here, the 32% decline didn't achieve the 50% that we set out to do, of course, which has never been shown. But the 32%, as Mark has said, is really as good as any result has ever been seen in this very difficult disease. And when you look at the tau terciles, there's a hint here that there may be segments of the population that could do even better. And so we have to reproduce that. But if you look at that lowest tau tercile, I think we're getting close to or maybe even exceeding slowing disease progression by 50%, which is just utterly remarkable. Clearly, I think 30% slowing is clinically meaningful. I think most patients and physicians and caregivers will agree with that and, in fact, would take even less as meaningful and hopeful given the long disease course. The second question there was on per-protocol efficacy. That's what we're showing here. That's -- these are all per-protocol analyses. There's no imputation of missing data. And this is a conservative statistical approach.

Oh, John?

Yes. Thanks for the question. I might just add a little bit to what Dan said. So when we think about the amount of slowing that we've seen in this 18-month trial, it equates to approximately 6 months of less slowing in at least an 18-month study. The other thing that adds to our confidence in the meaningfulness of this is the fact that we're already seeing less tau in that frontal lobe and other parts of the brain. And all our data to date has shown that the amount of tau matters. It also predicts future decline. And so even though we've only studied them for 18 months, I think we're encouraged by the less tau in the brain, and that would also project forward into less decline even beyond the time period that we have studied.

The next caller is Chris Schott from JPMorgan.

Great. Just 2 for me. I guess just a little bit more in terms of, will there be an ability to take, I guess, earlier looks at TRAILBLAZER-2, given some of the separation we saw on iADRS from this first study? And I guess related to that, when do you expect you'll have more clarity on alternative filing pathways with regulators? Is this something that's going to evolve over time? Or is it just a meeting that hasn't happened yet, and with either like 1Q earnings or 2Q earnings, we'll have more clarity on that point? My second question, which is maybe coming back to the 2 components of iADRS, do you believe you're going to need to show significance on both of those components or just clear trends? I'm just trying to see the cut point of what would define kind of like an imbalanced kind of, I guess, symptoms versus signs kind of outcome on the composite.

Thanks, Chris. We'll go to Dan for both of those.

Yes. Sure. Two very good questions here. So your first question is, is it possible to take an earlier look, maybe an interim analysis in TRAILBLAZER-2? Look, we haven't made decisions around that. We haven't communicated anything about that at this moment in time. But like you, looking at the data from TRAILBLAZER-1, we were impressed with the early separation here. And really at 52 weeks, the data looked quite significantly positive across the primary and a number of secondaries. So there could be opportunities there. I think we have to work through exactly what that means and how do you keep a study running without unblinding it if you take interim looks that could lead to filing. You wanted to know more about when we'll know more about that. There's not a moment in time when that happens. That's probably a long dialogue, and we have some time as we enroll patients and follow them in this trial. I'll start on the second question, but my colleagues here will weigh in as well. You're asking if you need both components of iADRS to be stat sig to go forward to approval. I don't expect that's the case. If it were, we would need -- that would be akin to setting them both up as co-primaries, which sort of takes us back in time to how Alzheimer's studies used to be conducted. So if that's the regulatory standard, that would be a pretty big step backwards. But I think consistency for the 2 is not an unreasonable expectation.

Yes. And I'll just add a little bit more color to that. As we've used the iADRS now across many studies, including in the primary for zagotenemab trial that reads out later this year, we've gotten much more accustomed to how it performs. And all of our analyses to date show that the correlation between these 2 is just about right. It's about half. So they're each providing unique information, and yet they both won't -- tend not to disagree with each other. And so all the trials that we run, we've seen them largely agree with the other and not giving us any opposite result. So that also encourages that as one moves, so shall the other.

The next question is from Seamus Fernandez from Guggenheim Securities.

Great. So just hoping to get a better understanding of exactly the Phase III trial design now and how the team is stratifying the different patient population. Is the primary goal to start with the sort of lower tertile of the lower baseline tau patient population and then sort of in a hierarchical way proceed to higher levels of tau for primary analysis? Just trying to get a better understanding of that in terms of how we should think about the Phase III trial results as well as the approachable patient population in the context of the 1 million to 2 million patients that you guys mentioned. And then separately, a big discussion at AD/PD this weekend was really the correlation to blood-based biomarkers and tau as an absolutely critical driver of patient identification going forward. Can you guys give us a sense of where blood-based tau biomarkers are at this point as a potential corollary to patient identification utilizing relative to your multiple tau biomarkers on PET?

Thanks. We'll go to Dan for the first question and Mark for the second.

Thanks, Seamus. These are important questions. On TRAILBLAZER-2, our plan here is that the primary analysis population is the full cohort of intermediate tau patients, so the same as TRAILBLAZER-1. Of course, you'll note that we've dramatically here increased. Sample size is 4x higher. If we have a similar-sized effect in TRAILBLAZER-2 versus TRAILBLAZER-1, obviously, the stat sig on each of those time points for iADRS will be much stronger. So with that done, I think then we can turn to other populations. And we haven't worked out our full hierarchical statistical testing model, but it will be interesting and important to test the tau strata as you've noted and perhaps see if we can again replicate that patients who have the lowest tau have an even stronger effect. That could lead to important information when this trial is complete. Of course, we'll also want to go back to the high tau patients, who are being enrolled in TRAILBLAZER-2 but are not part of the primary efficacy population and see what we see in those patients as an additional analysis.

Mark?

Yes. Regarding the blood-based biomarkers, particularly as they relate for Alzheimer's disease development, it is exciting to see the really, really rapid, almost explosive growth of -- and progress we've made in figuring out how we can determine blood assays that could relate to Alzheimer's pathology. One of the things that we point to is the P-tau217 assay that's developed at Lilly has figured permanently in a lot of this excitement. We see that as a potentially transformative way that patients can be identified in the future. In the beginning, maybe for clinical trials for Alzheimer's disease and then maybe we hope later as a way of doing this in clinical application in the clinic. So we are pushing our own internal program hard to put that into action, both -- again, both for our own clinical trials, enable others around the world to be able to have access to that clinical trials and then ultimately enable that type of assay to be available for patients. The interesting thing about that assay is that our own data and what's already out there and published is that it looks like it can be very sensitive for picking up amyloid-positive patients over a wide range of tau levels. So it is not dependent on having high tau, for instance, or even intermediate. It can be down at very low levels of tau, it seems to be able to turn positive. But might not be true, though -- although we'd love like a perfect ability to predict the tau PET scan from a blood-based biomarker, but it looks like it may not be quite as good at being able to actually say how much tau -- how far the pathology is. So it may be better for screening for patients who have perhaps started even at the very, very early stages of the tau pathology of Alzheimer's disease but might not be good for predicting exactly what level. So we believe that over the next year, that type of data is going to come out. And I think we'll probably be in the very middle of that as we develop and hopefully make available a P-tau217-anchored blood assay.

The next caller is Umer Raffat from Evercore.

This is Mike DiFiore in for Umer. A few for me. Just digging into this trial's data. Could you just tell us more about the timing of discontinuation in TRAILBLAZER-1 or when the dropouts occurred? I asked because week 36, in particular, seem to drive most of the CDR-SB delta. Number two is would you consider continuous dosing rather than just stopping dosing with donanemab, just giving the narrowing of effect that we saw over time and because there was no widening of benefit over time? And then finally, for TRAILBLAZER-2, and forgive me if I'm misunderstanding this, but I'm trying to better understand why you'd want to remove the patients with low tau who are likely the best responders or -- in the trial. That's it.

Sure, Mike. We'll go to John for the first question; Mark, the second; and Dan, the third.

Yes, thanks. So I think we have to change our mindset a little bit on how we're thinking about this drug, right? Our goal has not been to maintain patients on drug for a lifetime. And so it's an isolated course of treatment. And when we look at the dropouts, they are largely balanced between the 2 studies. And so if you look across the 2 studies, we have the same amount of completers. What we have is somewhat an earlier...

Two arms.

In the 2 arms, yes. And then we have basically 2 -- early on, we have some people dropping out from the ARIA, which we know, as typical for most of these studies that they generally occur early and with the dose changes. So those are generally dropping early. But the patients or subjects are staying in the trial. And so they've all -- we've done a good job of keeping them in there and following them over time. So they are all contributing to that data set.

Thanks, John. Mark?

Yes. One of the interesting things is -- I'll pick up where John stopped, which is the way that they contribute to the data set. One of the things that we're able to do is to, as they stayed in the trial and continued, when we add up and we look at the distribution of doses -- total doses received for patients across the whole trial, you can divide them in various ways. But the bottom line, and we were really pleased to see this, is that if you just simply look at the people who sort of stopped dosing earlier for a variety of reasons, maybe because they turned amyloid negative at 6 months because a fair percentage of people did that or because they perhaps stopped dosage because of ARIA and then stayed in the trial. When we look at that group that stopped early versus the people who needed to continue, in other words they continued the drug because perhaps it took a little longer drug to get their amyloid levels down, they were basically overlapping lines. In other words, their responses were identical. They -- there's really no difference in the cognitive slowing scene in either of those 2 groups. So we feel pretty comfortable that our algorithm is working. It's not -- we don't know if it's perfect. But what we're comfortable is that stopping slowing, particularly after demonstrating amyloid removal versus continuing -- stopping dosing when there's amyloid removal or continuing the dose when there's not as much amyloid removal, both of them are leading to the same cognitive benefits.

Thanks, Mark. And I'll take the last question. Maybe just to clarify, the discontinuations that we speak of here are treatment discontinuations, by and large, which means they stopped getting dosed, which is fine. But they stayed in the study, and we were able to collect cognitive measures over time. And those patients didn't do worse. In fact, they had a very similar or same drug benefit as patients who stayed on therapy. That's because they had already had a plaque reduction that was extremely significant by their early time points. So discontinuation of dosing here has no effect on treatment efficacy. The second question is on discontinuation from the trial, people who dropped out of the trial. That's a smaller number. But again, our analyses suggest that there is no evidence at all that these results were influenced in either direction by study dropouts. So finally, just to put that question to rest. The third question there was on why are we not including the low tau patients. We are including the low tau patients in TRAILBLAZER-2. You might have been thinking about the people with no tau at the very end of the spectrum. Those are excluded because we know from past studies, they don't progress. So in 18 months, how can you show a drug benefit if the placebo group isn't progressing? Those could be great candidates for future studies that could be longer in duration that would be considered secondary prevention studies. And that's a really important consideration.

The next caller is Greg Gilbert from Truist Securities.

Dan, if aducanumab is approved based on existing data, does that change anything about your strategy or time line expectations? And secondly, curious whether Lilly is still not pursuing trials with GLP-1s or other antidiabetic agents in Alzheimer's disease? There seemed to be some at least academic interest in that during one of the discussions on Saturday.

Thanks. We'll go to Dan for those.

Yes. So your question is sort of how do we change our plans if aducanumab is approved on the basis of a single trial and a second trial that's clearly negative. Look, I think we have to wait and see what happens. I'll tell you one thing that won't change. That's our commitment to replication. We're running TRAILBLAZER-2. That study will enroll and read out regardless of what happens with adu. I think that's important. I think the field will welcome a drug that has clear data. And that's the situation that we're anticipating and planning for with donanemab. And as I said before, that's important not just from a regulatory framework but also from a marketplace framework for patients and prescribers and payers to actually have confidence and use a drug. It should not be launched in the face of controversial data. The second question there is on GLP-1s. Of course, we watch the field closely. I guess we're the key company here that has both leading expertise in incretin biology and in amyloid biology. It's exciting to contemplate how those 2 might work together. I think there are obviously different pathways here. GLP-1 appears to have benefits on brain health through probably most likely effects on vascular biology as we've seen in the outcome studies with a decreased incidence of symptomatic stroke and probably, therefore, a small micro infarct, say, symptomatic strokes that then, over time, contribute to cognitive decline. But at the moment, we do not have any studies that are investigating GLP-1s in Alzheimer's disease.

The next caller is Geoff Meacham with Bank of America.

Just have a couple of quick ones. The first one is what are the clinical and development implications of the antidrug antibodies? And how much of a wrinkle does that play into redosing? And the second question is, in a way to -- how you guys aggressively reduced A beta with donanemab, is the speed or magnitude of effect on tau more important when you look to the second half of this year when you have the zagotenemab data set?

Thanks. We'll go to Mark for the first question and Dan for the second.

Oh, I think we're going to do it the other way around, actually. So I'll take the first question on ADAs, and then Mark will take tau level. So on antidrug antibodies, you make a great point, which is, clearly, we have high levels of ADAs. We knew this from essentially the first patient that we dosed with this drug several years ago. The reason we continued on in development despite high levels of ADAs was because we could see a clear and rapid pharmacodynamic response, i.e., the clearing of amyloid plaques. Over time, we've studied that PD response as well as the PK of the antibody and found that the presence of ADAs in the majority of patients, for sure, doesn't impact the ability to clear plaques. We've seen this dramatic plaque clearance even in the face -- in the majority of patients with ADAs. It also doesn't appear to impact redosing. One thing you could look at in drugs with ADAs are, are there any very severe types of anaphylaxis or immune complex disease. That is a theoretical possibility. We haven't seen anything to that extent. We watch infusion reactions, which are present at a relatively low level. We don't know whether or not that's related to ADAs, lots of drugs have infusion reactions. Still, all things equal, it's better to have a drug that doesn't have antidrug antibodies. We've worked to create that. That's the drug we call N3pG 4. It's in clinical trials. I expect there, we'll have a drug that doesn't have ADAs. We'll probably be able to dose it at lower levels because we don't have to dose above the ADAs. And hopefully, we'll be able to generate the same levels of plaque clearing with lower doses. That's not a -- we're not switching horses here. We're sticking with donanemab, but that could be a life cycle play.

Thanks, Dan. Mark?

Yes. But I do want to verify. Maybe I don't know if he's still on. The -- you started with the fact that we were able to dramatically lower amyloid rapidly that might have been a key to seeing some of the tau outcome, but I didn't see how that connected -- you asked a question then about the Phase II studies we're doing with our anti-tau drug?

That's right.

Maybe. Okay. Then I will -- okay. I will point out that, obviously, we were -- one of the zagotenemab Phase II was actually also enrolled on the basis of tau window. So we were able to exclude those patients with extensive tau pathology, the same way we did for donanemab. One of the advantages of that in this particular type of mechanism of action or what we hope the mechanism of action for zagotenemab is that if we're trying to slow the spread of the tau from neuron to neuron, having extensive tau on a tau PET scan would imply that the spreading is essentially already over and there's already an extensive amount of tau throughout the brain. Our hope is that by limiting it and excluding that high tau pathology, we'll be able to see a clinical benefit and a clinical -- and we'll see on the tau scans, we'll be able to hopefully see a slowing of the tau spread as well. Unfortunately, with an anti-tau antibody like zagotenemab, we don't have the [ crisp ] biomarker that we have for donanemab. And so we don't have the equivalent of showing whether we're dramatically, for instance, tying up the tau aggregates in the brain. These have proven incredibly hard to measure in CSF. And so from that point of view, we are a little bit -- we have to fly a little bit blind as far as our exact target engagement. But we're still very, very hopeful. We know that this has been a very difficult area for the field. And we think that our extensive use of tau, essentially in every single patient, our tau imaging, will give us some insights on it and exactly whether -- not only is whether the drug is working but how it might be working when we finally open the envelope.

The next caller is Andrew Baum from Citi.

A couple of questions on TRAILBLAZER-ALZ 2. I was just being careful to what you said about using a Bayesian disease progression model to capture the treatment effect along the treatment period. I'm interested whether you're intending or hoping to use the Bayesian analysis for the primary of this trial and how the FDA would react to not only a novel composite but also a Bayesian analysis. And then second, for the interim time point in TRAILBLAZER-ALZ, I'm curious as to what percentage of those were in person versus virtual assessment just in terms of how much weight or potential bias maybe capture that? And then finally, if you have any plan to talk about the brain volume impact. I'm thinking about some of the early vaccine data with 1792 and what you saw there, if there's anything to take [ home ] if time allows.

Okay. We'll go to Dan for those.

Okay. Thanks, Andrew, and I'm going to -- from here on out, I'm watching the clock. So I'm going to answer the questions briefly. So don't take it personally, Andrew. So on TB-2, you're asking about whether we'll use DPM or not. That's our intent. But I think we need to work through the different statistical methods here. Those discussions haven't really happened in earnest. There's a third alternative that's been tossed around here called the cubic spline method. I mentioned earlier, we did run that on this trial every -- the primary, of course, still hit. But many of the secondaries, including CDR Sum of the Boxes hit at the last time point on cubic spline. So that's also an option. The interim time points were in person, and the brain volume effects are routinely seen with amyloid-lowering drugs. And we think that's a normal class effect.

Next caller is Vamil Divan from Mizuho Securities.

So maybe just two, if I could. One, on this intermediate tau that you showed for the first trial, in which you're talking about for the second, it sounds like you're keeping that the same sort of definition of 1.1 to 1.46. Just curious, have you thought about maybe narrowing that a little bit and kind of going more to a little lower threshold just because the upper 1/3 of patients in the first trial really seem to get no benefit, the lines almost overlapping. So curious, I know -- if you powered up enough, you might get statistical significance, but wondering if it makes sense to maybe narrow the range there a little bit. And then second one, I guess, for Dan. On the last earnings call back in late January, you had said you did not see any reason to make a change to the primary endpoint when you're sort of asked directly. To be clear, I do agree with the change you're making. I'm just wondering what has changed since the end of January and presumably had all this data already available to today when you're making the announcement. Did you have further regulatory discussion or see something further that got you comfortable in making the change?

Great, Vamil. So I'm going to go fast again here. So no, we're not going to change the definition of intermediate tau. This is about replication. But I take your point, future studies could go even lower than this, but that will be a separate population. Second, on the earnings call, I said I didn't see anything that required us to change from CDR Sum of Boxes. I still feel about the same way. We don't -- we didn't have to change. I think I'd be willing to bet on CDR Sum of the Boxes if we had to. On the other hand, why do that? iADRS is just clearly better. And as you look at the graphs, if all I had to go by were the placebo arms, you can just see looking at that Slide 16 that shows all the endpoints, the placebo arm on iADRS is so much cleaner. And then the same is true on drug arm. It's just the more you look at the data, the more you realize it's a better outcome. To specifically answer your question, yes, of course, over that period of time, there's been deeper looks at the data and regulatory discussions.

The next caller is Steve Scala from Cowen.

Should we expect a ceiling on efficacy given that more than 2/3 of patients reached amyloid negativity? And then secondly, if we just step back, TRAILBLAZER-2 appears to have some advantages and challenges versus TRAILBLAZER-1. On the positive side, it's larger, maybe fewer dropouts, narrowing of the endpoint, and collectively, these all provide greater statistical power. On the negative side, it's more diverse in centers and patients, which is never good. First of all, did I miss any positives or negatives? And how -- when you think through these positives and negatives, how do you come out positively?

Okay. This is Dan. I'll take the first one for sure. So do we think there's a ceiling efficacy? Probably amyloid-lowering drugs at this stage of the disease, this is probably the ceiling. This drug fully clears plaques in a majority of patients. I don't think there will be better effects from stronger drugs. Of course, going early in the disease course could offer larger benefits. Combinations could offer larger benefits. Those are probably 2 things to think about. Your second question was on the pros and cons of TB-2. I think the most important thing here is that it's larger and, therefore, more highly powered. And we could be much more confident about the probability of success here.

And the next caller is David Risinger from Morgan Stanley.

Yes. So my first question is on tau. The second is on clinical relevance. So first, given donanemab's impact on tau, would you in hindsight have expected better results in terms of efficacy relative to high-dose aducanumab? And why does donanemab reduce tau? And second, some observers have questioned the clinical relevance of donanemab's key efficacy endpoint results. How does Lilly plan to validate donanemab's ability to benefit patients in a clinically relevant manner?

Thanks, Dave. We'll go to Mark for the first one and Dan for the second.

One of the things that -- the way we think about the tau and the interaction with amyloid is go back for a second, the whole concept is that amyloid probably is initiated first, amyloid deposits and very likely, given our understanding of the disease, accelerate the tau. We don't know exactly what aspect of the amyloid deposits that the acceleration of the tauopathy is due to. But this data really does show that when removing the amyloid plaques, we're changing something that slows that acceleration, we believe, in the tau spread. The tau is probably one of the major, major -- the tau deposits is probably one of the major sources of the neuropathy and the ultimate death of neurons and dysfunction in neurons. And so our sense is the series of events is that as we remove whatever toxic or accelerating aspects of the amyloid plaques, we can slow the tauopathy from spreading. It's not perfect, as Dan just pointed out. There's almost certainly a ceiling effect here. The more tauopathy you already have in the brain, the less it can be maybe impacted by this effect. Now of course, we believe that we are actually sort of -- the tau change is actually a leading indicator. So what we're looking at is something that the amount of tau we're seeing slowing in different areas may actually itself then continue to be an indication of slower spread of tau and clinical decline in the future. So we think those -- that's the relationship we think is going on, and we think the data continue to support it. Dan?

Thanks, Mark. We'll actually go to John for the second question.

Yes. So the clinical relevance of the iADRS -- thanks for the question. So we're continuing to work on this. And what we've seen for this outcome measure is that it works across almost the entire symptomatic spectrum. So when we talk about absolute points, it's not as helpful as to talk about relative change because the absolute points changed depending on what the patient population you enroll. We're fortunate here at Lilly to have large data sets. In fact, we have a large observational studies. And we'll continue to work on publishing those anchor points that a lot of people look forward to seeing. And when we have those results, we'll release those.

The next caller is Carter Gould from Barclays.

I guess maybe just -- there've been a couple of allusions to it, but I don't think you've explicitly said out right. I guess, looking at the tertile analysis, what can you say explicitly around the percent of patients who achieved plaque clearance across those different cuts and if there were any differences? And I guess coming back to the differences in TB-2 versus TB-1, just trying to understand if the prespecified tau strata will be defined by the absolute levels or you have to sort of do another tertile analysis. Just thinking about reproducibility here.

Okay. Thanks. We'll go to Dan.

Yes. So your first question on differences in plaque lowering across the different tertiles, I don't expect we would see that. I'm not sure we've done that analysis. But it's just that everyone had such good plaque lowering. It will be hard to see -- as Mark said earlier, you can see it the other way around. So in other words, if you look at the patients who completely cleared plaques, they did have more tau lowering than patients who didn't achieve that full clearance. So that's something. With respect to the tertiles in TB-1 versus TB-2, we haven't really specified the subgroups within there. In TB-1, it was done as the name implies, just by 1/3, 1/3, 1/3.

The next caller is Louise Chen from Cantor.

So first question I had for you is with the meaningful decline in amyloid, what would you have expected to see in terms of tau burden? Did the study meet your expectations here? And second question I had for you is, do you think the high level of apoE4 carriers impacted your results in any way in terms of clinical efficacy but also in terms of tolerability?

Thanks, Louise. We'll go to Dan for those.

Yes. So the first one is what did we expect for tau levels. This exceeds our expectations. So I think there was very little data on this question. To see these huge reductions in tau spread reaching significance in region after region is incredibly encouraging for the amyloid hypothesis and efficacy of this drug in the long term. The second question on apoE4s, I think this is a really -- given the study size here, I'm not sure it would be productive to look at all of these different subgroups of populations. One thing we do know, though, is that E4 carriers in these kinds of studies with amyloid-lowering drugs do account for more of the ARIA events than noncarriers. So that's consistent across the class.

The next caller is Navin Jacob from UBS.

Very quickly, I just want to clarify. So the intermediate group in TRAILBLAZER-2 will include that -- the high end of that intermediate group, which you showed doesn't have as much efficacy, obviously, or the delta in the curves looks very minimal, so you will still be including that? And then the comment around no difference in efficacy regardless of whether you dose down or not or regardless of whether there were adverse events-related discontinuations, I mean one way to interpret that is that there is no dose effect associated with donanemab. Just wondering how you think about that.

Okay. And this time, I'll take both of those. I mean so first, to your question on the intermediate group, yes, it will include the high tau tercile. It's exactly the same as the TB-1 group. Note that TB-1 hits that sig even with 1/3 of the patients not contributing to efficacy. TB-2 being 4x larger should be fine. But it will be interesting to look at those subgroups. If we reconfirm that, that could have impacts for labeling. Second question...

Dose effects.

Dose effects. I think it's hard to say here. There's obvious confounds because the people who got less drug burden, so the people who switched over to placebo because they had complete clearance of plaque, had more PD effect with less drug. So should they respond better or worse? I think that the 2 things probably cancel out. And in the end, it's about the same. So we don't see a difference.

Next caller is Kerry Holford from Berenberg.

Just one last point of clarification, please. So in your view, do you see the expanded ALZ-2 study of the Phase II data we have today as submission for filing? Or do you expect to need to run any additional clinical trial to supplement that filing?

We'll go to Dan.

Thanks, Kerry. Exactly as you said. We expect 2 positive studies on this iADRS endpoint in the same population to be sufficient for studying. So that's TRAILBLAZER-1 and TRAILBLAZER-2.

Okay. That exhausts the queue. Now we'll go to Dan to close.

Thanks, everybody. In closing, we're grateful to have been able to have this time to share with you today what we see is potentially transformative progress we're making on Alzheimer's disease as well as very specifically the next steps for donanemab. We're appreciative of all the efforts from the Lilly team and our investigators and all of our collaborators to conduct this trial. But most of all, we're deeply thankful to the patients and caregivers who've participated in the TRAILBLAZER-ALZ study as well as those who've partnered with us in this journey in trials past. Thanks again for your participation in today's investor call and your interest in Eli Lilly and Company. Please follow up with our IR team if you have any questions we've not addressed on the call. Enjoy the rest of your day.

Thank you, and that does conclude our conference for today. Thank you for your participation and for using AT&T Event Conferencing. You may now disconnect.