Upping the Ante for Predicting the Success of Alzheimer Disease Treatments

Image credit: BillionPhotos.com | stock.adobe.com

Over the past 20 years, less than a handful of completed Phase III Alzheimer disease (AD) clinical trials have been successful. Despite large investments of capital and research, more than 200 investigational programs have failed or have been abandoned in the past decade alone.¹ This lack of progress has significantly hampered advancing the care of those with AD, as clinical trials are the only means of generating safety and efficacy data that can lead to a drug’s approval and widespread availability.

AD is already a pressing public health challenge, and as the population of the United States ages, that challenge deepens alarmingly. An estimated 6.7 million Americans aged 65 and older are living with AD dementia today, a number that could grow to 13.8 million by 2060 barring the development of medical breakthroughs to prevent, slow, or cure AD.²

What can we do to increase the chances of success in AD clinical trials? Although clinical trials are inherently difficult, the complexity of AD biology makes the drug development process even more so. However, the use of modeling and biosimulation can help predict potential outcomes and improve confidence in therapeutic candidates. This predictive power will ultimately help drug developers to avoid common pitfalls by learning from the many failed attempts to bring effective therapies to people living with AD and their families.

Over the past 25 years, the cumulative expense of conducting clinical trials for new AD therapies was estimated at $42.5 billion, with the greatest costs incurred in late-stage drug development. Better means of reducing and distributing costs, sharing risks, and improving development success are needed.³

First described in 1992, the amyloid hypothesis postulates that aggregation of the protein amyloid-beta (Aβ) is the first driver of neuropathology in AD.⁴ It is proposed that aggregation of Aβ fragments into neurotoxic protofibrils and amyloid plaques is a key event leading to abnormal neuron function and brain cell death.

Aβ fragments or malfunctioning neurons facilitate abnormal phosphorylation of tau protein promoting the formation of neurofibrillary tangles.⁴ These events induce inflammation in the brain (neuroinflammation) associated with the cognitive decline caused by AD.⁵ This means the longer the wait to start treatment, the more damage there may already be.

A case in point is when Eisai and Biogen announced positive top-line results for their anti-amyloid antibody lecanemab. This trial became the first successful, completed Phase III AD drug study in the Western world in more than 20 years.⁶ Certara, in partnership with Eisai, developed the first in-silico approach in support of a successful disease-modifying AD therapy based on the unique pharmacological properties of lecanemab.

This combined Physiologically-based Pharmacokinetics/Quantitative Systems Pharmacology (PBPK/QSP) model correctly predicted efficacy biomarker outcomes and generated a new hypothesis to mitigate adverse effects of this treatment.⁷ The model was developed based on the synthesis and degradation of Aβ fragments Aβ40 and Aβ42, and their aggregation into plaques and clearance by the brain’s “trash collectors” or microglial cells. The QSP platform is calibrated with pharmacological and clinical information on longitudinal observational studies and interventions with six amyloid-targeting antibodies.

Using both fluid and imaging amyloid biomarker changes, the model determined mechanistic differences between lecanemab and other treatment modalities targeting amyloid peptides illuminating the biology driving the different clinical outcomes. The model was able to describe target exposure of monoclonal antibodies and simulate dynamics of cerebrospinal fluid (CSF)—the fluid that is around the brain volume—and plasma biomarkers, including CSF Aβ42 and the ratio of plasma Aβ42/Aβ40 levels, respectively.

Changes over time in patients with AD were simulated using an age-dependent decrease in Aβ clearance and recapitulated the time-dependent relationship between fluid biomarkers and brain amyloid load in naturalistic studies. The model also allowed for the prediction of Aβ PET imaging load as measured by Standardized Uptake Value Ratio (SUVR). Importantly, the model confirms plasma Aβ42/Aβ40 levels and phospho-tau may be used in lieu of expensive and invasive PET scans to document the central amyloid changes after treatment.

In addition to known biomarkers for AD, the model also predicted the effect on other biomarkers that can't be measured but may help predict a compound’s efficacy and safety. Examples include the dynamics of insoluble forms such as protofibrils, that are hypothesized to drive clinical AD cognitive effects rather than the changes in plaques. These can be derived from changes in SUVR that are recognized as a surrogate marker for efficacy by the FDA.

By incorporating antibody-bound plaque mediated immune cell activation, the model also simulates the incidence of amyloid-related imaging abnormalities with edema (ARIA-E). By implementing a time-to-event approach, the model was able to capture the time-dependent increase of ARIA-E adverse effects (AEs), which was more pronounced in the first weeks of treatment, as observed in the clinical trials.

The model suggests that the observed edema is correlated with the amount of antibody bound to plaques in the area surrounding brain arteries, rather than the amount of antibody bound to protofibrils. This generated a new hypothesis that may be useful in the optimization of the pharmacology and titration strategy for new anti-amyloid antibodies. This model has demonstrated its predictive ability and the important role biosimulation may play in the R&D and regulatory approaches for new AD drugs, such as informing treatment strategies for novel formulations of amyloid therapies, including combination trials.

In future studies, biosimulation and modeling may also be used to optimize clinical trial design for novel antibody modalities, use specific changes in peripheral biomarkers to determine the time at which brain amyloid negativity is achieved, and support development of maintenance-dose strategies. Importantly, the model can support the development of non-antibody based amyloid interventions, such as small molecules modulating aggregation or clearance processes.

Modeling can help maximize amyloid SUVR reduction and minimize ARIA-E AE liability for any new drug profile. This amyloid model is currently being integrated with tau pathology⁸ and neuroinflammation QSP models to potentially support combination trials. In general, this approach may be applicable to other neurodegenerative diseases in need of disease-modifying therapies, including Parkinson disease.

It has become increasingly evident that there are several AD subtypes caused by different pathologies, including amyloid composition, tau protein distribution, other toxic proteins, neuroinflammation, and cerebrovascular pathologies that all lead to different clinical trajectories.⁹ Rather than a single pathological process, AD therapy is now moving from an amyloid-centered treatment to a multi-targeted approach.

New technologies, such as integrated informatics and text analytics software using natural language processing, allow the analysis of big data and can help identify other important factors in the development of AD, such as cholesterol metabolism, inflammation, immune factors, and others. The identification of new biomarkers can help detect people at risk sooner, allowing them to start treatment earlier and potentially improving outcomes.

Modeling and biosimulation are powerful tools to describe the interactions between these different processes and biomarkers in a quantitative and actionable way that can support new treatment paradigms. Overall, QSP modeling could support the clinical trial design of different amyloid-modulating interventions, define optimal titration and maintenance schedules, and provide a first step to understanding the variability of biomarker response in clinical practice.

As of January 2023, there were 187 trials assessing 141 unique treatments for AD, of which 79% were disease-modifying therapies.¹⁰ The model suggests that the relative affinity of the antibody for the protofibrils and plaques is a major determinant for ARIA-E to occur during treatment.

With a disease such as AD, time is of the utmost importance, as the sooner treatment begins, the higher the potential for better outcomes. By saving valuable time, predictive modeling and biosimulation may be used to bring new effective therapies to people and families in need before it’s too late.

About the Authors

Hugo Geerts, PhD, Pharma MBA, is Head of Quantitative Systems Pharmacology (QSP) Neurosciences at Certara with more than two decades of experience in mechanism-based QSP modeling in neurology and psychiatry. Piet van der Graaf, PharmD, PhD, is Senior Vice President, Quantitative Systems Pharmacology and contributes to the strategic development of Certara based on his more than 20 years of experience in the pharmaceutical industry.

References

1. Yiannopoulou KG, Anastasiou AI, Zachariou V, Pelidou SH. Reasons for Failed Trials of Disease-Modifying Treatments for Alzheimer Disease and Their Contribution in Recent Research. Biomedicines. 2019;7(4):97. Published 2019 Dec 9. doi:10.3390/biomedicines7040097
2. 2023 Alzheimer's disease facts and figures. Alzheimers Dement. 2023;19(4):1598-1695. doi:10.1002/alz.13016
3. Cummings JL, Goldman DP, Simmons-Stern NR, Ponton E. The costs of developing treatments for Alzheimer's disease: A retrospective exploration. Alzheimers Dement. 2022;18(3):469-477. doi:10.1002/alz.12450
4. Hardy JA, Higgins GA. Alzheimer's disease: the amyloid cascade hypothesis. Science. 1992;256(5054):184-185. doi:10.1126/science.1566067
5. Lecca D, Jung YJ, Scerba MT, et al. Role of chronic neuroinflammation in neuroplasticity and cognitive function: A hypothesis. Alzheimers Dement. 2022;18(11):2327-2340. doi:10.1002/alz.12610
6. Rogers MB. How will alzheimer’s trials, treatment change in 2023 and beyond? ALZFORUM. October 7, 2022. Accessed August 14, 2023. https://www.alzforum.org/news/research-news/how-will-alzheimers-trials-treatment-change-2023-and-beyond
7. Geerts H, Walker M, Rose R, et al. A combined physiologically-based pharmacokinetic and quantitative systems pharmacology model for modeling amyloid aggregation in Alzheimer's disease. CPT Pharmacometrics Syst Pharmacol. 2023;12(4):444-461. doi:10.1002/psp4.12912
8. Geerts H, Bergeler S, Walker M, van der Graaf PH, Courade JP. Analysis of clinical failure of anti-tau and anti-synuclein antibodies in neurodegeneration using a quantitative systems pharmacology model. Sci Rep. 2023;13(1):14342. Published 2023 Sep 1. doi:10.1038/s41598-023-41382-0
9. Ferrari C, Sorbi S. The complexity of Alzheimer's disease: an evolving puzzle. Physiol Rev. 2021;101(3):1047-1081. doi:10.1152/physrev.00015.2020
10. Cummings J, Zhou Y, Lee G, Zhong K, Fonseca J, Cheng F. Alzheimer's disease drug development pipeline: 2023 [published correction appears in Alzheimers Dement (N Y). 2023 Jun 28;9(2):e12407]. Alzheimers Dement (N Y). 2023;9(2):e12385. Published 2023 May 25. doi:10.1002/trc2.12385