Warning message

You need to be logged in to add this content to your library.

Scientists Re-Analyze Aduhelm Data, Try to Parse Who Benefits

Series - AD/PD 2022: Advances in Science and Therapy (Hybrid Virtual and In-Person):

Last June’s accelerated FDA approval of Aduhelm based on amyloid plaque removal left unresolved whether the antibody truly shores up cognitive function. At the 16th International Conference on Alzheimer’s and Parkinson’s Diseases, held March 15-20 online and in Barcelona, Spain, speakers picked at this question with new analyses of the old Phase 3 data, and some argued that it does. Biogen presented an updated meta-analysis of anti-amyloid immunotherapies, reporting that plaque removal does slow cognitive decline across multiple trials and molecules. Other speakers sussed out patient characteristics that could help predict who will benefit from immunotherapy, showing analyses from the aducanumab and donanemab programs that hinted at possible effects of age and baseline p-tau181. Meanwhile, Philip Scheltens of VU University, Amsterdam, offered a sobering assessment of how heterogeneity in the AD patient population muddies clinical endpoints and may mask drug effects.

  • New analysis of Aduhelm’s Phase 3 data correlates plaque removal with CDR-SB score.
  • Meta-analysis of recent amyloid immunotherapy trials concurs.
  • Biomarkers could pick out those most likely to benefit from treatment.

Despite their longstanding difficulties in demonstrating a meaningful benefit, pharma scientists remain focused on anti-amyloid therapy. They note its strong effects on multiple biomarkers of amyloid, tau, and neuroinflammation. “Biomarkers show changes in the underlying pathological state,” said Mark Mintun of Eli Lilly, which makes donanemab, during a panel discussion. “At Lilly, we have a lot of confidence that this is the right track.”

Aduhelm Controversy Continues
Nonetheless, the first antibody to gain approval in the U.S. is still treading a rocky road. After Biogen’s mixed Phase 3 results got a thumbs-down from the FDA's advisory committee, the agency granted accelerated approval based on plaque removal being “reasonably likely” to translate into a clinical benefit (Nov 2020 news; Jun 2021 news). The Phase 3 efficacy data on which this decision rests remained unpublished for three years after these trials were stopped, and nearly a year after aducanumab was approved.

Now the data are published, in the March 16 Journal of Prevention of Alzheimer’s Disease, an 8-year-old specialty journal. According to press reports, in 2021 Biogen withdrew the paper from consideration by JAMA (Axios news). Aducanumab's Phase 1 data had appeared in Nature (Sevigny et al., 2016).

JPAD ran three accompanying editorials. Ron Petersen, Mayo Clinic Rochester, Minnesota, cautioned that the results of the EMERGE trial must be interpreted in the context of the negative results of ENGAGE, meaning the impact of plaque removal at this stage of disease may be minimal. Zaven Khachaturian of the Campaign to Prevent Alzheimer’s Disease offered no specific opinion on the aducanumab data, but drew a historic analogy with the flawed, first-approved acetylcholinesterase inhibitor, Tacrine. He wrote, “Even if in the long run [aducanumab] fails the promise of becoming a treatment of choice, the FDA’s conditional approval has already created an encouraging environment for further investments in R&D of drugs in this class.” Lon Schneider, University of Southern California, Los Angeles, reminded the journal's readers of the flaws in the two trial's data analysis and presentation after the futility analysis ended the trials early, calling the JPAD paper's claim of a clinically meaningful impact "so wrong on many levels." 

Like Schneider, other commentators were unimpressed. Madhav Thambisetty at the National Institute on Aging, Bethesda, Maryland, noted that the paper skirted key questions such as whether the faster placebo decline in the positive EMERGE study could have driven the apparent treatment effect. Rob Howard at University College London was quoted as saying, “The peer review has apparently been so gentle and unrevealing that they might as well not have bothered” (Endpoints news).

At AD/PD, some presentations attempted to calm the roiling waters with fresh looks at the Phase 3 data. Biogen's Kumar Kandadi Muralidharan presented an exposure-response analysis of the ENGAGE and EMERGE trials that linked reduction in amyloid PET SUVR to slowed decline on the CDR-SB. Only about a third of participants had PET scans. Biogen used data on how their drug exposure related to their plaque reduction to impute SUVR values for the rest, i.e., the majority of the cohort. With these imputed numbers, Biogen created an exposure-response model that rendered a statistically significant effect of amyloid removal on CDR-SB score. The effect predicted by this model was the same for both ENGAGE and EMERGE participants, as well as for people whose disease progressed at different rates. Overall, the model posits that 18 months of aducanumab titrated to 10 mg/kg would slash a person's plaque load by 75 percent and slow his or her slide on the CDR-SB by 0.38 points on average.

Sam Dickson of Pentara Corporation, which sells biostatistics services to drug sponsors, took a different tack. He combined the data from all endpoints into a single univariate scale and, based on that, claimed that even in the negative ENGAGE trial, this new endpoint favored aducanumab in both dose groups.

Ultimately, new efficacy data will be required to convince the field at large. On March 30, Biogen announced that it has submitted the final protocol for Envision, its required confirmatory trial, to the FDA for approval and expects to begin screening patients in May and to complete the trial in four years. Biogen will enroll around 1,500 participants, and pledged at least 18 percent of those from the U.S. will be black or Latino.

Also at AD/PD, Changyu Shen of Biogen argued that data from other anti-amyloid immunotherapy trials strengthen the case for cognitive benefits with this treatment approach. He revisited an oft-cited meta-analysis by Maria Glymour and colleagues at the University of California, San Francisco, which reported no significant correlation between amyloid lowering and MMSE scores. This meta-analysis combined 14 trials of eight purported amyloid-lowering drugs, including early immunotherapies, γ-secretase and BACE inhibitors, and bexarotene (Ackley et al., 2021).

Shen repeated this analysis, using a publicly available web interface Glymour and colleagues had provided to allow their findings to be updated with new data. Shen added two trials not previously included, the aducanumab Phase 1 and donanemab Phase 2 studies. He also corrected what he said were data inconsistencies in the original meta-analysis. Shen's updated meta-analysis found a statistically significant effect of amyloid removal on the three endpoints examined: CDR-SB, MMSE, and ADAS-Cog. Shen reported larger effect sizes than did Ackley et al. For example, the effect of a 0.1 SUVR reduction in amyloid rose from 0.06 points on the CDR-SB to 0.09. Restricting the meta-analysis to only immunotherapy trials further strengthened the effect size and statistical significance, Shen noted.

These findings broadly align with a recent review of immunotherapy trials by Eric Karran at AbbVie in Cambridge, Massachusetts, and Bart De Strooper at the UK Dementia Research Institute, London. They concluded that plaque needs to be lowered to 20 centiloids or less to produce a noticeable cognitive benefit, with a lag time of several months between amyloid removal and clinical effect. Therefore, trials such as the Phase 3 aducanumab program, where most people became amyloid-negative only at the end of the study, might not demonstrate a convincing clinical benefit (Karran and De Strooper, 2022). 

Who Gets the Most Out of Adu?
Even if removing plaque eventually restores cognition, researchers agree that the effects will vary significantly because people’s rates of progression and even the confluence of genetic, environmental, and lifestyle factors leading to AD differ so much from one person to the next. At AD/PD, Scheltens provided a glimpse at how noisy Alzheimer's progression data can be. VU scientists analyzed cognitive change over 18 months in amyloid-positive, cognitively impaired participants in the ADNI observational cohort. They saw a broad range of outcomes. For example, on the CDR-SB, 95 percent of people had anything from a drop of 0.35 to a gain of the same amount over this time period. Most recent immunotherapy trials posted drug effect sizes within this range, meaning what signal there is could easily be lost in this noise, Scheltens noted.

Would selecting patients based on risk factors like an APOE4 allele or tau pathology tighten this variability? Alarmingly, it would not, Scheltens reported. The VU study found even more clinical variability in E4 carriers than noncarriers, and in people with higher concentrations of CSF total tau than in people with less CSF total tau. Variability was higher in younger than older people, and higher in women than men. Running longer studies to try to see an effect might not work either, as clinical variability increased over time (Jutten et al., 2021). 

Instead, Scheltens suggested enrolling more people, using more sensitive outcome measures such as ADCOMS, and, importantly, comparing individual patient trajectories rather than group differences. For such trajectories, acquiring some run-in data before people start on drug can further help changes pop out, Scheltens noted. Some trials, such as the DIAN-TU platform, are already doing this.

Other talks at AD/PD offered tantalizing if inconclusive hints of factors that could influence whether a given patient responds. Oliver Peters of Charité University Hospital, Berlin, examined data from the 15 people who participated in the ENGAGE trial at his site, took aducanumab, and had CSF data. He noticed that the dose a person was on correlated with his or her change in CSF Aβ42/40 but not p-tau181. To parse this, Peters separated the eight participants who had a large drop in p-tau181 on aducanumab from the seven who did not. The main difference between these two groups was a higher baseline p-tau181 in the responders, though both groups started above the threshold for abnormal p-tau181.

Peters believes that within this abnormal range, low baseline p-tau181 may mark patients who have atypical AD and are therefore less likely to respond to immunotherapy. Although both groups had similar cognitive scores at baseline, and both declined while on aducanumab, those whose CSF p-tau181 dropped on drug declined less than the p-tau181 “non-responders” on the MMSE, ADAS-Cog, CDR-SB, and Activities of Daily Living scales.

“Aducanumab showed best results in typical AD patients who tested amyloid- and tangle-positive,” Peters wrote to Alzforum. He thinks this lends support to the use of tau PET as a trial inclusion criteria, as Lilly is doing with its donanemab studies.

Meanwhile, Oskar Hansson of Lund University, Sweden, spotted a relationship between p-tau181 change and an unexpected factor linked to treatment response. He previously reported that a dip in this marker in the Phase 3 aducanumab trials correlated with plaque clearance and a slowing of cognitive decline at the group level (Nov 2021 conference news). At AD/PD, Hansson added subgroup analyses, finding that a person's sex, APOE genotype, disease stage, or standard AD medications had no bearing here. Age did, however. In people older than 75, p-tau181 dropped more dramatically with plaque clearance than it did in younger people. Hansson was at a loss to explain this. “I would love to get ideas from anyone here on why this might be the case,” he told the audience.

Analysis of a different therapeutic antibody program, Lilly’s donanemab, implicated drug exposure as the main factor determining treatment response. Lilly’s Ivelina Gueorguieva presented a pharmacokinetic analysis based on serum samples from 304 participants in donanemab Phase 1 or 2 trials; 177 took drug, 127 placebo. Participants had amyloid PET scans at baseline, six, 12, and 18 months. Only one factor affected their amyloid removal: People who maintained an average serum concentration of at least 4.4 μg/mL donanemab during the trial cleared plaque rapidly, while those with less donanemab cleared it more slowly. About 80 percent of the cohort were in the former category. What determined this threshold effect? Gueorguieva does not know. People with a higher body weight, or more anti-donanemab antibodies in their blood, cleared donanemab from their bodies fastest, but neither factor had much impact on plaque removal, she said.

While scientists are slicing and dicing old trial data to squeeze out some additional learning, all eyes are on new trial results. Gene Kinney of Prothena noted that upcoming trial readouts for gantenerumab, lecanemab, and donanemab will give better information on patient selection, biomarker response, and the sensitivity of clinical endpoints. “I think we’ll have a rich dataset on which to learn more about every one of those components,” Kinney said in Barcelona.

For its part, Prothena announced on March 28 that it has begun a Phase 1 trial of a new anti-amyloid antibody, PRX012. It binds Aβ oligomers, plaques, and pyroglutamated forms with high affinity and can be injected under the skin.—Madolyn Bowman Rogers

Comments

  1. In the pre-dementia stage of AD, removing the negative impact of pathological amyloid might help to improve some aspects of cognition, according to this new paper in Brain (Düzel et al., 2022). However, the protein TMEM106B can cause the formation of amyloid filaments in human brain in a way that is age-dependent, and which might not be related to disease (Schweighauser et al., 2022). 

    References:

    Düzel E, Ziegler G, Berron D, Maass A, Schütze H, Cardenas-Blanco A, Glanz W, Metzger C, Dobisch L, Reuter M, Spottke A, Brosseron F, Fliessbach K, Heneka MT, Laske C, Peters O, Priller J, Spruth EJ, Ramirez A, Speck O, Schneider A, Teipel S, Kilimann I, Jens W, Schott BH, Preis L, Gref D, Maier F, Munk MH, Roy N, Ballarini T, Yakupov R, Haynes JD, Dechent P, Scheffler K, Wagner M, Jessen F. Amyloid pathology but not APOE ε4 status is permissive for tau-related hippocampal dysfunction. Brain. 2022 May 24;145(4):1473-1485. PubMed.

    Schweighauser M, Arseni D, Bacioglu M, Huang M, Lövestam S, Shi Y, Yang Y, Zhang W, Kotecha A, Garringer HJ, Vidal R, Hallinan GI, Newell KL, Tarutani A, Murayama S, Miyazaki M, Saito Y, Yoshida M, Hasegawa K, Lashley T, Revesz T, Kovacs GG, van Swieten J, Takao M, Hasegawa M, Ghetti B, Spillantini MG, Ryskeldi-Falcon B, Murzin AG, Goedert M, Scheres SH. Age-dependent formation of TMEM106B amyloid filaments in human brains. Nature. 2022 May;605(7909):310-314. Epub 2022 Mar 28 PubMed.

    View all comments by Charles Stromeyer
    • User Profile Image Maria Glymour
      University of California, San Francisco
    • Posted:

    I am pleased that Dr. Shen and Biogen have taken up our invitation to add internal and new data to our prior meta-analysis (Ackley et al., 2021), and I look forward to seeing their detailed methods. Our analyses showed the potential power of pooled analyses but was handicapped by lack of data access. When we built and posted a publicly available interface to repeat our analyses as data updates became available, it was in the sincere hope that our tools would accelerate progress in this domain.

    Alzheimer's research on amyloid-reducing therapies could make much faster progress if the companies involved all released their data for harmonized, pooled analyses of amyloid-targeting therapies. Such pooled analyses would be better powered than any single company's evidence, and would allow for precise estimation of the important threshold and subgroup effects hypothesized above.

    References:

    Ackley SF, Elahi F, Glymour MM. Instrumental variable meta-analysis of aggregated randomized drug trial data for evaluating proposed target mechanisms. BMJ. 2021 Feb 25;372:n346. PubMed.

    View all comments by Maria Glymour

Make a Comment

To make a comment you must login or register.

References

Therapeutics Citations

  1. Aduhelm
  2. Donanemab
  3. Bexarotene
  4. Gantenerumab
  5. Leqembi

News Citations

  1. FDA Advisory Committee Throws Cold Water on Aducanumab Filing
  2. Aducanumab Approved to Treat Alzheimer’s Disease
  3. Aduhelm Lowers Tau; Registry to Track Real-World Performance

Paper Citations

  1. Sevigny J, Chiao P, Bussière T, Weinreb PH, Williams L, Maier M, Dunstan R, Salloway S, Chen T, Ling Y, O'Gorman J, Qian F, Arastu M, Li M, Chollate S, Brennan MS, Quintero-Monzon O, Scannevin RH, Arnold HM, Engber T, Rhodes K, Ferrero J, Hang Y, Mikulskis A, Grimm J, Hock C, Nitsch RM, Sandrock A. The antibody aducanumab reduces Aβ plaques in Alzheimer's disease. Nature. 2016 Aug 31;537(7618):50-6. PubMed.
  2. Ackley SF, Zimmerman SC, Brenowitz WD, Tchetgen Tchetgen EJ, Gold AL, Manly JJ, Mayeda ER, Filshtein TJ, Power MC, Elahi FM, Brickman AM, Glymour MM. Effect of reductions in amyloid levels on cognitive change in randomized trials: instrumental variable meta-analysis. BMJ. 2021 Feb 25;372:n156. PubMed.
  3. Karran E, De Strooper B. The amyloid hypothesis in Alzheimer disease: new insights from new therapeutics. Nat Rev Drug Discov. 2022 Apr;21(4):306-318. Epub 2022 Feb 17 PubMed.
  4. Jutten RJ, Sikkes SA, Van der Flier WM, Scheltens P, Visser PJ, Tijms BM, Alzheimer's Disease Neuroimaging Initiative. Finding Treatment Effects in Alzheimer Trials in the Face of Disease Progression Heterogeneity. Neurology. 2021 Jun 1;96(22):e2673-e2684. PubMed.

External Citations

  1. Axios news
  2. Ron Petersen
  3. Zaven Khachaturian
  4. Lon Schneider
  5. Endpoints news
  6. announced
  7. announced 

Further Reading

Annotate

To make an annotation you must Login or Register.