Alzheimer’s Disease: Combination Therapies and Clinical Trials for Combination Therapy Development

Abstract

Alzheimer’s disease (AD) is a complex multifaceted disease. Recently approved anti-amyloid monoclonal antibodies slow disease progression by approximately 30%, and combination therapy appears necessary to prevent the onset of AD or produce greater slowing of cognitive and functional decline. Combination therapies may address core features, non-specific co-pathology commonly occurring in patients with AD (e.g., inflammation), or non-AD pathologies that may co-occur with AD (e.g., α-synuclein). Combination therapies may be advanced through co-development of more than one new molecular entity or through add-on strategies including an approved agent plus a new molecular entity. Addressing add-on combination therapy is currently urgent since patients on anti-amyloid monoclonal antibodies may be included in clinical trials for experimental agents. Phase 1 information must be generated for each agent in combination drug development. Phase 2 and Phase 3 of add-on therapies may contrast the new molecular entity, the approved agent as standard of care, and the combination. More complex development programs including standard or modified combinatorial designs are required for co-development of two or more new molecular entities. Biomarkers are markedly affected by anti-amyloid monoclonal antibodies, and these effects must be anticipated in add-on trials. Examining target engagement biomarkers and comparing the magnitude and sequence of biomarker changes in those receiving more than one therapy, compared with those on monotherapy, may be informative. Using network-based medicine approaches, computational strategies may identify rational combinations using disease and drug effect network mapping.

Key Points

Alzheimer’s disease is a complex disorder with many biological processes simultaneously affected.

Combination therapies are needed to address the complex pathology of Alzheimer’s disease.

Monoclonal antibodies directed at plaque amyloid slow the progression of Alzheimer’s disease, and add-on therapies may result in improved clinical benefit.

Computational strategies incorporating improved understanding of disease pathways and networks of drug effects promise to generate rational suites of combination therapy candidates.

Introduction

Alzheimer’s disease (AD) is a complex, progressive neurobiological illness that progresses through biomarker-defined stages and includes multiple pathological processes, including amyloid-beta protein (Aβ) aggregation in fibrillar plaques, tau protein aggregation into neurofibrillary tangles, neuroinflammation, neurodegeneration, oxidative cellular injury, mitochondrial dysfunction, abnormalities of autophagy, bioenergetic and metabolic changes, synaptic dysfunction, vascular disturbances, and epigenetic alterations [1–3]. Multiple proteinopathies including aggregation of Aβ, tau, transactive response DNA-binding protein-43 (TDP-43) and α-synuclein, are common findings at the time of postmortem examination of patients with AD [4, 5]. Genetic influences such as apolipoprotein E ε4 (APOE4), racial and ethnic variables, and sex differences further exaggerate the heterogeneity in AD biology [6–8].

The complex array of neurobiological processes occurring simultaneously and sequentially in AD demands a correspondingly complex multifaceted therapeutic intervention if disease progression is to be arrested or slowed. Combination interventions are required to influence multiple pathological processes. In this review, we consider the development of combination therapies including simultaneous development of more than one treatment to create a new combination therapy (combination products) and use of add-on therapies resulting in combinations of approved and experimental agents in trials. We note the potential development of single agents with multiple simultaneous effects and of combinations, such as traditional Chinese medicines (TCMs), that are combination therapies. We include pharmacodynamic (PD) and pharmacokinetic (PK) combinations in the review. We describe the potential utility of technologies such as computational strategies and biomarkers that may accelerate combination therapy development.

Biological Framework for Combination Therapy for Alzheimer’s Disease

Biological processes that have been identified in AD and related dementias (ADRD) comprise the Common Alzheimer’s Disease Research Ontology (CADRO) [9]. These processes provide the targets for monotherapy and combination therapies. Table 1 integrates the CADRO categories with the Alzheimer’s Association Revised Diagnostic Framework [10] to provide a conceptual architecture for combination therapies.

Table 1.

Categories of AD pathology that can be addressed in combination therapy approaches [based on the Alzheimer’s Association Diagnostic and Staging Framework and the Common Alzheimer’s Disease Research Ontology (CADRO)]

Classification	Targets for combination therapy
Core AD pathologies	Amyloid
	Tau
Non-specific biological reactions involved in AD pathophysiology	Neurodegeneration
	Inflammation (peripheral and central)
	Astrocytic reaction
	Bioenergetic and metabolic dysfunction
	Synaptic dysfunction
	Oxidative stress
	Apolipoprotein E ε4, lipid, and lipoprotein receptor effects
	Gut-brain axis
	Growth factors and hormones
	Neurotransmitter receptor changes
	Circadian rhythm disturbances
Non-AD co-pathology that may occur with AD	Vascular changes
	α-synuclein aggregation
	TDP-43 aggregation
	Proteostasis/proteinopathies

The diagnostic framework is based on biomarkers that can be used for diagnosis and staging of AD. The framework identifies two core pathologies (Aβ and tau); nonspecific tissue reactions involved in AD pathophysiology (neurodegeneration and inflammation); and non-AD co-pathology that may occur in individuals with AD (vascular injury and α-synuclein). Combination therapies for AD might address two core pathologies, two or more types of non-core tissue reactions, two or more types of non-AD pathologies, or combinations of core pathologies, reactive changes, and non-AD co-pathologies. Table 1 provides a conceptual structure for combination therapy based on the diagnostic framework and CADRO.

The neuropathology of AD evolves over the course of time [11]. Amyloid-β changes are the earliest detectable abnormality, with aggregation of tau into neurofibrillary tangles and neurodegeneration ensuing [12, 13]. A temporal sequence of combination therapies can be anticipated to optimize the biological effects of the intervention. Treatments might be timed to reduce existing abnormalities [e.g., administration of anti-Aβ monoclonal antibodies (MABs) to reduce Aβ plaque] or might be given in anticipation of changes predicted by plausible models of AD to delay the emergence of expected pathologies and the associated clinical deficits (e.g., administration of anti-tau therapies prior to the emergence of neurofibrillary tangles). Figure 1 provides a map for how such combination programs—including both treatment and prevention approaches—might be planned. Drugs to reduce inflammation and preserve synaptic plasticity are among the most common agents in the AD drug development pipeline and could be combined with proteinopathy-directed treatments in combinations therapies [14]. The continuum of combinations would extend from treatment of Aβ and prevention of tau-related neurofibrillary tangles in the earliest stages to symptomatic and palliative treatments in the final stages of the illness. Primary prevention trials targeting the initiating events of AD, such as Aβ aggregation or neurofibrillary tangle formation, with therapies to prevent one pathology, or using combination therapies to address both pathologies, are anticipated [15] .

Fig. 1.

Combinations of agents to be used to treat or prevent the pathological changes of AD as the disease evolves (©J Cummings; illustrator, M. de la Flor, Ph.D.). Abeta, amyloid-beta protein; BioM, biomarker negative; BioM+, biomarker positive; CU, cognitively unimpaired; MCI, mild cognitive impairment; NFT, neurofibrillary tangle

Types of Combination Therapies

Table 2 shows the possible types of currently recognized combination therapies. Pharmacodynamic combinations include two or more active therapies that are intended to effect a change in AD biology or behavior. The impact might be additive or synergistic. Pharmacodynamic combinations can include two or more approved products, a combination of one or more approved targets and one or more new molecular entities (NMEs), combinations of two or more NMEs developed separately but used simultaneously, sequential combinations of approved therapies or NMEs, combinations of two approved agents into a single pill or other formulation, multi-target drugs, and traditional medications such as TCMs usually comprised of multiple herbs or natural products. Pharmacodynamic combination therapies can also include one or more drugs with one or more other types of treatment modalities, such as devices or lifestyle interventions [16].

Table 2.

Types of combination therapies

Class	Combination composition	Existing or proposed example
Pharmacodynamic	2 or more approved products	Cholinesterase inhibitor and memantine
	1 or more approved product + 1 or more NMEs	AA/MAB + experimental agent
	2 or more co-developed products to be used together	AA/MAB + anti-tau MAB
	2 or more experimental products developed separately and used together	AA/MAB + anti-tau MAB
	Sequential combinations	AA/MAB followed by a gamma secretase modulator
	2 or more approved agents in a single pill	Namzaric®
	Multi-target drugs	Rasagline; ladostigil
	Traditional medications	Traditional Chinese medications
	1 or more active agents + a device	Cholinesterase inhibitor + deep brain stimulation
	1 or more active agents + a lifestyle intervention	Metformin + FINGER lifestyle plan
Pharmacokinetic	Active agent + a metabolic inhibitor	Dextromethorphan + bupropion
	Active agent + side-effect-blocking agent	Xanomeline + trospium
	Active agent + pharmacologic mechanism to cross the BBB	Gantenerumab + transferrin receptor “brain shuttle” (trontinemab)
	Active agent + device to assist crossing BBB	Aducanumab + focused ultrasound

AA/MAB, anti-amyloid monoclonal antibody; BBB, blood–brain barrier

Pharmacokinetic combinations include one or more active agents and one or more drugs intended to prevent or slow the metabolism of the active agent, antagonize peripheral side effects, or facilitate the access of the active agent to its central nervous system (CNS) target through impact on the blood–brain barrier (BBB). Blood–brain barrier strategies include use of transporters to escort drugs into the brain or use of devices such as ultrasound to temporarily interrupt barrier functions [17, 18]. Combination therapies of drugs that may reduce amyloid-related imaging abnormalities (ARIA) in conjunction with anti-Aβ MABs is a key area for drug development that would allow these agents to be administered more safely and used more widely. Growing evidence of a role for inflammation of Aβ-laden vessels as a predisposing factor for ARIA suggests that combination therapy with anti-inflammatory agents warrants exploration [19]. Table 2 includes examples of existing or proposed combination strategies in AD drug development.

In addition to the active pharmacologic ingredients involved in combination therapies, different formulations must also be considered. Oral agents (tablet, capsule film, and solution), intravenous drugs, intranasal agents, transdermal or “patch” drugs, intramuscularly injected agents, subcutaneously administered drugs, and intrathecally delivered drugs are formulations to be considered in a combination treatment regimen. The pharmacokinetic features of the agents, the physicochemical properties of the drugs, and patient convenience comprise aspects of the dialogue leading to the final decision of the combination of formulations to be advanced in the combination therapy. When combining an NME with an approved agent, the formulation of the approved agent is determined by past trials, and the formulation of the NME can be based on its planned use in combination with the approved therapy.

Combination Therapy Development

Overview

The US Food and Drug Administration (FDA) provides guidance for co-development of one or more NMEs known as “combination products.” Co-development programs should meet the following criteria (Table 3): the combination is intended to treat a serious disease or condition, there is a compelling biological rationale for use of the combination, evidence from a non-clinical model or short-term clinical study with an established biomarker suggests the combination has substantial activity and provides greater than additive activity or a more durable response then the individual agents administered by themselves, and there is a compelling reason for why the agents cannot be developed individually [20].

Table 3.

Food and Drug Administration criteria for the co-development of two or more investigational drugs in combination (FDA, 2010)

The combination is intended to treat a serious disease or condition

There is a compelling biological rationale for use of the combination (e.g., the agents inhibit distinct targets in the same molecular pathway, provide inhibition of both a primary and compensatory pathway, or inhibit the same target at different binding sites to allow use of lower doses to minimize toxicity)

A preclinical model in in vivo or in vitro or in short-term clinical study on an established biomarker suggests that the combination has substantial activity and provides greater than additive activity or a more durable response compared with the individual agents alone

There is a compelling reason for why the agents cannot be developed individually (e.g., one or both of the agents would be expected to have very limited activity when used as monotherapy)

Pharmacokinetic combinations meet all these criteria. Use of bupropion or quinidine to inhibit the cytochrome P450 2D6 enzyme allows administration of dextromethorphan in doses with acceptable peripheral side effects in trials of agitation and treatment of pseudobulbar affect, respectively [21, 22]. Trospium blocks the peripheral side effects of xanomeline in a combination being assessed for treatment of psychosis of AD [23]. Trontinemab capitalizes on a transferrin transporter mechanism to facilitate penetration of gantenerumab across the BBB in greater amounts than can occur without the transporter [24]. Focused ultrasound has been used in combination with aducanumab to produce local interruption of the BBB and enhance brain penetration of the anti-Aβ MAB [17]. These pharmacokinetic combinations meet the criterion of required co-development to achieve the prespecified activity of the test agent.

Pharmacodynamic combinations such as administration of an anti-Aβ agent plus an anti-inflammatory agent or an anti-Aβ agent plus a synapto-protective agent might not require co-development, could be developed and administered independently, would be assessed in add-on trial designs, and would not necessarily be approved and marketed for required co-administration. If two agents were shown to have a synergistic effect (e.g., two co-administered anti-inflammatory agents might result in more than additive effects), co-development of the two novel therapies might be warranted. Sequential pharmacodynamic combinations (e.g., novel anti-Aβ MAB followed by a gamma secretase modulator) might require co-development to optimize the timing and dosing of the sequence[25].

Co-development of Combination Therapies Comprised of Two or More New Molecular Entities

Development of two or more NMEs requires contemporary availability of two or more such agents. This may occur in a single company but could involve two or more collaborating enterprises. For the non-clinical aspect of the development program, the FDA recommends that each entity be assessed for toxicology prior to evaluating the combination. Non-clinical studies should provide evidence to support the biological rationale for the combination and should compare the activity of the combination with the activity of the individual components. Drug interaction studies may be required. These studies could be done in in vivo or in vitro models. An animal model relevant to the disease may provide valuable activity data as well as guidance for dosing the combination [26].

When two or more novel agents are co-developed, Phase 1 studies show the safety profile of each new investigational agent as would be done with a single drug, including determination of the maximum tolerated dose (MTD) or other approach to choosing the highest dose, the nature of dose limiting toxicity (DLT), and the agent’s PK parameters. Dose–response relationships on biomarker measures can add persuasive pharmacodynamic information. Assessment of the combination in Phase 1 is based on the safety observations of each individual agent to determine the starting doses, dosing intervals, and doses to be used in dose–response studies [20]. Phase 1 combination studies of the combination will provide insight into possible additive toxicity of the test agents.

Phase 2 proof-of-concept (POC) studies employed in co-development programs optimally use a four-arm trial design comparing each drug alone with the combination and with placebo or standard of care. An adaptive design allowing collapse of the single-agent arms is appropriate if it becomes evident that the single agents have much less activity than the combination [20].

Master protocols are defined as protocols designed with multiple substudies, which may have different objectives and involve coordinated efforts to evaluate one or more investigational drugs and one or more disease types within the overall trial structure [27]. Basket trials, umbrella trials, and platform trials are all examples of master protocol trial designs. Basket trials are designed to test a single investigational drug or drug combinations in different populations, in different disease stages, or identified by different genetic or other biomarkers or demographic characteristics [27, 28].

Phase 3 trial co-development designs build on the observations from Phase 2 studies. If the contribution of each agent is established, then a comparison of the combination to placebo or standard of care is appropriate. If the contribution of the individual agents is not determined, a four-arm factorial study or a factorial study using only active drug treatment arms without a placebo (individual agents are compared with the combination) may be proposed. For example, in a PK combination program, a three-arm study compared the effects of dextromethorphan, quinidine, and the combination for the treatment of pseudobulbar affect in patients with amyotrophic lateral sclerosis [29]. This trial established the superiority of the pharmacokinetic combination over the contribution made by each agent. A two-arm trial may be sufficient if Phase 2 suggests that one agent is accounting for most of the activity but that the combination appears to be more effective than either agent alone [20]. Discussion with regulatory authorities prior to initiating a combination development program will help anticipate issues that may arise later in development.

Combination Treatment Composed of an Approved Treatment with a New Molecular Entity

Add-on therapy with two cognitive enhancers—cholinesterase inhibitors plus memantine—is commonly utilized in the treatment of AD. The memantine development program followed the approval of cholinesterase inhibitors and included one monotherapy trial and one combination add-on trial in its Phase 3 studies [30, 31]. Memantine is approved for monotherapy and can be used in combination with cholinesterase inhibitors. This program provides a precedent for add-on treatment development.

The approval of anti-Aβ MABs has made understanding of more complex combination therapies an urgent requirement. Aducanumab (Aduhelm^®) and lecanemab (Leqembi^®) are approved for market use, and donanemab is currently under review by the FDA. These treatments produce a 25–40% decline in the rate of progression of AD depending on the specific trial and the specific outcome measure used [32–37]. The trial observations suggest that combining anti-Aβ MABs with therapies targeting non-plaque species of Aβ or non-Aβ aspects of AD is required to further amplify the treatment effect. In clinical trials, NMEs will be administered simultaneously with the MABs or sequentially following MAB treatment. Allowing patients on anti-Aβ MABs to enter trials, or to initiate treatment with a MAB during a trial, may be necessary to accommodate patients’ best interests. It is necessary to understand the add-on effects and how to interpret the trial observations for the NME when anti-Aβ MABs are being administered to some portion of or all the trial participants.

Simultaneous administration of anti-Aβ MABs with an NME might include combinations directed at any of the targets shown in Table 1. An anti-Aβ MAB might be combined with a tau agent addressing the biology leading to neurofibrillary tangles, an anti-inflammatory agent combating the inflammatory processes evident in the brain of the AD patient, or drugs with effects on growth factors or neurodegeneration (see below for a discussion of how computational strategies might help prioritize agents to be used in combinations) [38–40].

Sequential combinations might be considered in conjunction with anti-Aβ MABs. This approach would involve administration of an anti-Aβ MAB that reduces plaque burden, followed by an agent that prevents Aβ production or re-accumulation. In its Phase 2 and Phase 3 trials, donanemab was discontinued when Aβ plaque levels were reduced to below detectable limits [33, 34]. Amyloid-β then begins to reaccumulate, with a foreseeable time to detection estimated at 4 years [41]. Administration of a gamma secretase modulator might prevent re-accumulation and maintain low levels of toxic aggregated Aβ species if given following cessation of donanemab therapy [25]. Gamma secretase inhibitors and beta-site Aβ precursor protein cleaving enzyme (BACE) inhibitors could also be considered for a role in combination therapy, but these agents have toxic properties that would require further study of their biological effects and re-examining dosing strategies to ensure safe utilization [42, 43]. Anti-Aβ MABs, such as lecanemab, are directed at protofibrillar Aβ species, and are administered continuously after initiation until the patient exits the therapeutic window, possibly making this agent less of a candidate for sequential treatment [44]. Simultaneous combinations of lecanemab with agents directed at tau pathology, AD co-pathologies, or protein and vascular pathologies that co-occur in patients with AD are feasible.

Monoclonal antibodies produce marked reduction in plaque Aβ as visualized by Aβ positron emission tomography (PET) [32–37]. Plasma biomarkers show that the anti-Aβ MABs also reduce phospho-tau (p-tau) and glial fibrillary acidic protein (GFAP), suggesting that they are affecting multiple downstream biological pathways and may themselves represent combination therapies impacting multiple linked processes [45, 46]. The impact of anti-Aβ MABs on imaging, plasma, and cerebrospinal fluid (CSF) biomarkers complicates the assessment of experimental therapies in patients who have received or are receiving treatment with these agents (discussed below).

FDA guidance regarding development of combinations of previously marketed small molecule drugs or biologics in combination with NMEs indicates that non-clinical studies may be warranted if existing toxicologic studies suggest additive or synergistic adverse events, and drug–drug interaction studies maybe necessary if interactions between the members of the combination are possible [47]. Phase 1 will have been completed for the approved agent and will be done in a standard manner for the NME. Small clinical investigations may be required to assess the likelihood of drug–drug interactions and evaluate possible additive or synergistic adverse events. Once these issues are addressed, a Phase 2 POC trial comparing the NME to an add-on combination and standard of care with the approved agent may determine add-on efficacy and safety. A Phase 3 trial comparing combination therapy and the standard of care could be the basis for approval as an add-on treatment (Fig. 2 shows an example of a trial design). A Phase 3, three-arm trial comparing the approved therapy, the NME, and the combination therapy would provide additional information and could lead to approval of the NME as monotherapy. Discussions with the FDA about trial design and drug dosing is warranted for program planning [48].

Fig. 2.

Example of a program for development of a new molecular entity as add-on therapy to an approved agent (©J Cummings; illustrator, M. de la Flor, Ph.D.). POC, proof of concept

Comparisons of clinical outcomes in those on add-on therapy compared with NME monotherapy must be adequately powered, and the sample sizes required will be determined by the number of participants who are receiving or who have previously received treatment with anti-Aβ MABs.

Phase 4 trials and real-world evidence will provide vital information for both add-on combination therapies and co-developed combination products. Phase 4 studies may include patients with a wider range of comorbidities and concomitant medications, participants from underrepresented groups, patients with greater or lesser disease severity, biomarker-defined populations, or related diagnoses, such as another neurodegenerative disorder when the agent has a relevant mechanism of action. The participants in Phase 4 studies differ from those involved in controlled clinical trials where rigorous eligibility criteria create a more homogeneous population. Study of safety data derived from Phase 4 is particularly important in determining whether groups not included in Phase 3 trials experience different side effect profiles or different severities of side effects compared with those included in trials leading to regulatory approval. Label adjustments, or changes in risk management strategies for the combination, could ensue from these observations if sufficiently compelling [49].

Evidence of efficacy depends on a complex placebo comparison, not usually present in Phase 4 studies; effectiveness is determined by changes from baseline with symptomatic agents or perceived changes in trajectory by the clinician or patient and caregiver in the case of disease modifying therapies (DMTs). Historical controls can provide useful insight in Phase 4 studies [50]. Real-world evidence relevant to understanding combination therapies can be obtained from claims data (e.g., Centers for Medicare and Medicaid Services and insurance companies), electronic medical records, prescription data, diagnostics data (e.g., laboratory testing and imaging), registries, and safety and adverse event reporting [51, 52]. Informative data to be collected include persistence on therapy, dose adjustments after treatment initiation, observations from patient groups underrepresented in the clinical trial, safety and tolerability, treatment adjustments made by clinicians or patients/care partners, and safety and effectiveness comparisons to alternative therapeutic approaches [52].

Multi-functional and Traditional Medications

Two classes of medications that are relevant to combination therapy include NMEs with multiple possible biological effects and traditional medications often comprised of combinations of herbs and other natural products. These differ from combinations of approved medications such as Namzaric^®—a fixed combination of memantine and donepezil—where both active ingredients have been shown to be efficacious and well tolerated and are approved. Agents such as rasagiline and ladostigil have been tested in AD, and foundational research suggests that these agents have effects on multiple AD-relevant pathways [53–55]. No specific development pathway for multifunctional agents has been defined, and they would currently be treated as single NMEs. Claims regarding the effects of individual components of the multifunctional agents would depend on demonstrating the efficacy and safety of each of the components. Many drugs have effects on more than one process, and the development program focuses on the processes most likely to respond to intervention and to result in clinical benefit. Dose relationships require study to determine whether the multiple potential effects occur at similar doses. Biomarkers may be helpful in demonstrating effects on multiple targets or pathways.

Traditional medications such as TCMs are typically composed of multiple natural products, and individual elements may contribute to any observed therapeutic benefit. Development of such agents would be similar to those of multifunctional drugs. They would be regarded as single NMEs and claims about individual elements would depend on demonstrating the efficacy and safety of each element in a clinical trial.

Use of Biomarkers in Trials of Combination Therapies

Biomarkers play an increasingly important role in AD drug development. Accurate diagnosis, prognostic information, documentation of pharmacodynamic drug effects, prediction of which patients are likely to respond to therapy or develop side effects, and safety monitoring are all areas where biomarkers are making critical contributions to clinical trials [56, 57]. Anti-Aβ MABs have marked effect on biomarkers. In the donanemab Phase 3 trial, Aβ PET demonstrated that 80% of participants reached undetectable plaque levels by week 76 in the low/medium tau population [34]. Plasma p-tau 217 and GFAP were also significantly reduced in donanemab trials compared with the placebo group [45]. Similarly, lecanemab produced Aβ plaque clearance in 81% of participants at 18 months, and there was a significant increase in the plasma Aβ 42/40 ratio and decrease in plasma p-tau 181 [46].

Detection and characterization of biomarker changes induced by the NME in an add-on therapy trial will be challenging. Currently, trials are structured to compare an NME with placebo; both arms of the trial may include participants who are on anti-Aβ MABs or who have received treatment with anti-Aβ MABs in the past. Both the NME arm and the placebo arms of the trial will have individuals with complete or partial treatment-related Aβ plaque clearance. At trial termination, comparisons can be made between those on and not on anti-Aβ MABs and between those who have or have not received treatment with anti-Aβ MABs. Target engagement biomarkers specific to the NME will be of value in establishing the drug effect of the test agent. The magnitude or sequence of changes of biomarkers may be contrasted in those receiving add-on therapy compared with those on monotherapy with the NME (Table 4). The trial sample sizes required for drawing adequately powered conclusions from these comparisons must be determined and will depend on the number of participants who are on treatment or have previously received treatment with anti-Aβ MABs.

Table 4.

Analyses that may help elucidate drug effects of new molecular entities being studied as add-on therapy in patients being treated with anti-amyloid monoclonal antibodies

Subgroup analyses of patients on anti-amyloid monoclonal antibody therapy

Subgroup analyses of patients who have previously received anti-amyloid monoclonal antibody therapy

Comparison of target engagement biomarkers specific to the mechanism of action of the new molecular entity

Comparison of neurofilament light (NfL) in patients on the new molecular entity (keeping in mind that NfL has not changed in most trials of anti-amyloid monoclonal antibodies)

Comparison of the magnitude of change (clinical or biomarker) in patients with add-on therapy compared with those with new molecular entity monotherapy and standard of care monotherapy

Comparison of the sequence of biomarker change in those on add-on therapy compared with those on new molecular entity monotherapy and standard to care monotherapy

Computational Strategies for Development of Combination Therapies

To accurately model the effects of combination therapies on human diseases, it is necessary to integrate many sources of information including the drug’s chemical properties, their interactions with targets, and the molecular mechanisms of AD. Network medicine and machine learning can be used to predict potentially efficacious drug combinations using “omics,” real-world health data, drug databases with drug properties, and data from animal model observations.

Network medicine takes advantage of advances in understanding biological systems such as protein interactions and metabolic pathways as a basis for modeling human diseases [58–60]. Network medicine approaches represent proteins or metabolites as geometric “nodes” that have their interactions with one another expressed as lines (“edges”) connecting the nodes. Within these interaction networks, biological relationships are determined by how nodes cluster, the number of edges to which a node is connected (its “degree”), and the distance between nodes along the shortest path of edges (their “proximity”) [60]. Proteins associated with a specific disease such as AD cluster together into a “module” on the network [61, 62]. It is hypothesized that the closer a drug target module is to a disease module, the more likely it is that the drug will affect the disease. This approach has been successfully used to replicate and predict the target engagement of monotherapies by calculating the proximity of a disease module to a module of proteins targeted by the drug [63].

When network proximity mapping was applied to anti-cancer and anti-hypertensive drug combinations in a study by Cheng et al., it was found that the molecular characteristics, efficacy, and adverse effects of two drugs were reflected by their distance from one another. Drug target modules that were closer together on the network were more likely to share similar proteins, be involved in the same biological processes, and have similar chemical features [64]. These observations may provide the basis for combination therapy candidates for application to AD.

Network medicine has been used to support drug repurposing efforts for both mono- and combination therapies for AD. Patient health records were used in a survival analysis of 38 drugs by Fan et al. to find a protective effect of anti-psychotic and anti-depressant combination therapy against AD. This was validated through use of a network of AD and psychosis-specific protein interactions to identify combinations of aripiprazole with maprotiline or sertraline as candidate combination therapies [65].

Artificial intelligence and deep learning models have been designed to identify proteins relevant to disease networks. Graph neural network (GNN) deep learning is used to identify protein sub-clusters that have similar gene ontology (GO) functions on extensive human protein interaction networks [66]. This information was then used to prioritize 156 AD risk genes on the basis of their enrichment for gene regulatory features such as CpG islands and transcription factor binding sites. Calculating the network proximity of repurposed drugs to this deep-learning AD disease module led to the prioritization of ibuprofen, gemfibrozil, cholecalciferol, and ceftriaxone, which were further validated as possible AD therapeutics with the use of electronic health record data. Combinations of these agents can be explored to determine the extent to which co-administration might optimize efficacy. Network medicine studies might provide insights into which combinations should be prioritized from among available agents, and precision medicine strategies may link treatment combinations to specific pathologies indicated by biomarkers.

The use of network medicine and machine learning has helped to identify numerous drug candidates for repurposing in AD and can be used to identify promising combination therapies. The integration of side-effect modules into drug–disease module proximity will improve the prediction of adverse reactions of combinations where adverse events can be additive, synergistic, or subtractive. Network mapping approaches may provide information allowing construction of rational combinations of therapies appropriate for the heterogeneous pathobiology of AD and responsive to the evolving changes of the pathophysiology of AD as the disease progresses.

Conclusions

AD is a complex disease with multiple pathological processes that represent therapeutic targets of potential benefit for delaying disease onset, slowing progression, or improving symptoms. Combination therapies represent an opportunity to address multiple pathological processes simultaneously. Approval of the anti-Aβ MABs makes add-on combination therapies increasingly likely. This produces challenges in drug development. Phase 1 trials are required for each agent, and Phase 2 and Phase 3 trials are designed to provide insight into the efficacy of the combination therapy and the contribution made by the NME to the combination. Co-development of two or more NMEs (combination products) is complex, requires demonstration of efficacy of each of the ingredients, and requires justification for why combination therapy is superior to developing each agent individually. PK and PD combinations contribute to the repertoire of therapies being developed for treatment of AD. Biomarkers are central to successful drug development, and biomarker effects produced by anti-Aβ MABs require that subjects who are on or have received anti-Aβ MAB therapy be analyzed separately to understand the effects of each treatment on the participant's biomarker profile. New computational strategies promise to generate rational combinations of treatments on the basis of an enhanced understanding of disease pathways and network effects of combination interventions. As new drugs become available, combination therapy will be increasingly common as a strategy for managing or preventing the cognitive, behavioral, and functional manifestations of AD.

Declarations

Funding

J.L.C. is supported by National Institute of General Medical Sciences (NIGMS) grant P20GM109025, National Institute on Aging (NIA) grant R35AG71476, NIA R25 AG083721-01, the Alzheimer’s Disease Drug Discovery Foundation (ADDF), the Ted and Maria Quirk Endowment, and the Joy Chambers-Grundy Endowment. Funding from these sources was used to provide Open Access. A.L.-O., J.W.K., D.C., and J.C. declare no funding relevant to this article.

Conflict of interest

J.C. has provided consultation to Acadia, Acumen, ALZpath, Aprinoia, Artery, Biogen, Biohaven, BioXcel, Bristol-Myers Squib, Eisai, Fosun, GAP Foundation, Janssen, Karuna, Lighthouse, Lilly, Lundbeck, LSP/eqt, Merck, MoCA Cognition, New Amsterdam, Novo Nordisk, Optoceutics, Otsuka, Oxford Brain Diagnostics, Prothena, ReMYND, Roche, Scottish Brain Sciences, Signant Health, Simcere, sinaptica, TrueBinding, and Vaxxinity pharmaceutical, assessment, and investment companies. J.C. owns the copyright of the Neuropsychiatric Inventory. J.C. has stocks/options in Artery, Vaxxinity, Behrens, Alzheon, MedAvante-Prophase, and Acumen. D.C., J.P., A.L., and J. Chen have no disclosures.

Ethics approval

This is a narrative review, does not involve any patient-level information, and does not require review by an institutional review board.

Consent to participate

No patients were involved in this report and no informed consent was needed. All other participants were voluntarily engaged, and no documentation of their consent is required. All have approved this article.

Consent for publication

No patients were involved in this narrative review. No patient identification could be derived from this review, and no consent to publish is required.

Availability of data and material

Our narrative review is based on an extensive review of and knowledge of the relevant literature. All sources of information are documented and are provided in the reference section. Of particular interest is our complete documentation of the regulatory guidance documents from the Food and Drug Administration (FDA), which are key to aspects of this review.

Code availability

No statical analyses were conducted, and no availability of code comment is relevant.

Author contributions

All authors participated in the development of this narrative review. The idea was initiated by J.L.C. with important preliminary contributions from a A.O.-L. and J.W.K. A.L.O., J.W.K., D.C., and J.C. all were drafting authors of individual sections of this review. A.O.-L. and J.W.K. were responsible for the biomarker section. D.C. and J.C. were responsible for the computational strategy section. Each author performed the literature search relevant to their area of manuscript development. All of the authors were involved in revision of the work, all confirm that they have read and approve of the final submitted manuscript, and all agree to be accountable for the work.