Seeking progress in disease modification in Parkinson disease

Abstract

Objective:

Disease modification in Parkinson disease (PD) has remained an elusive goal, in spite of large investments over several decades. Following a large meeting of experts, this review article discusses the state of the science, possible reasons for past PD trials’ failures to demonstrate disease-modifying benefit, and potential solutions.

Methods:

The National Institute of Neurological Disorders and Stroke (NINDS) convened a meeting including leaders in the field and representatives of key stakeholder groups to discuss drug therapy with the goal of disease modification in PD.

Results:

Important lessons can be learned from previous attempts, as well as from other fields. The selection process for therapeutic targets and agents differs among various organizations committed to therapeutic development. The areas identified as critical to target in future research include the development of relevant biomarkers, refinements of the targeted patient populations, considerations of novel trial designs, and improving collaborations between all stakeholders.

Conclusions:

We identify potential barriers to progress in disease modification for Parkinson’s and propose a set of research priorities that may improve the likelihood of success.

1. Introduction

Parkinson’s disease (PD) represents a large public health concern and a large financial burden [1]. The National Institutes of Health (NIH), pharma and biotech industries, and non-profit organizations have made large investments in PD research in general, and in potential therapeutics development. Despite this, a multi-decade history of attempts to achieve disease modification and slow the clinical decline in PD using drugs has not been successful.

The National Institute of Neurological Disorders and Stroke (NINDS) convened a group of experts from industry, academia, and the nonprofit sector, to discuss reasons for past failures in developing disease-modifying therapies for PD, and possible solutions. Using the key discussion topics as a starting point, we discuss here the history, major barriers to progress, and future directions.

2. History and framing the problem

There is a long history of attempts to achieve disease modification in PD. Agents studied to date in rigorous trials, without a successful outcome, include: tocopherol; deprenyl; TCH346; CEP-1347; Coenzyme Q10; levodopa; selegiline; minocycline; pioglitazone; rasagiline; pramipexole; creatine [2]. The most recent attempts involved: nilotinib [3, 4]; inosine; isradipine [5].

Many of the agents mentioned above were studied through the Neuroprotective Exploratory Trials in PD (NET-PD) program, which was established as a platform to accelerate therapeutic development [6]. Over 80 compounds were evaluated [7], and selected agents were moved into multiple pilot trials, running concurrently. One drug (creatine) went into a large pivotal trial, which was stopped for futility [8]. The program represented a notable effort to accelerate therapeutic development and develop collaborations, but it is possible that either the agents evaluated, the patient selection, or the underlying hypotheses were sub-optimal or incorrect.

The lessons from prior efforts include a need for optimized agent choice, better target selection, and better trial design. In addition, the choice of study population, given the heterogeneity of PD, is critical. A large number of the previous clinical trials have focused on motor function and symptoms. While this is adequate when testing symptomatic treatments, disease-modification trail designs need to use different endpoints, including long-term disability and major clinical milestones. Long-term follow-up is also critical to avoid errors, for example as recently demonstrated in extended follow-up of the ADAGIO study [9].

Awareness of these prior limitations resulted in clearer answers in the most recent NIH-funded large clinical trials, SURE-PD3 and STEADY-PDIII, but unfortunately these answers have been negative [5,10]. These studies had strong rationales and preliminary data, as well as supporting epidemiological data.

STEADY-PDIII was premised on neuroprotective effect of isradipine in animal models, and epidemiological data showing an apparent reduction in PD risk and slowing of PD progression with exposure to this drug class. The trial showed no effect of isradipine on the primary outcome. Potential reasons for failing to show a positive outcome include the lack of a human readout on target engagement, possible use of the wrong dose to provide sufficient effect, deploying the intervention too late in the disease, and the possibility that engaging this putative target by itself is insufficient to alter the disease progression.

SURE-PD3 was premised on the observation that serum urate, a major antioxidant, predicts a slower rate of decline. The study actually used a bioassay selection criterion (low urate) to recruit suitable participants. The trial was terminated when a prespecified interim analysis indicated that it would not show any benefit with inosine treatment, in spite of achieving the targeted urate concentrations. The reasons for this result are still being explored, along with additional analyses.

Much can be learned from these prior efforts, and a detailed discussion of the reasons the most recent trials were negative is ongoing. Among the important questions are whether the intended biological targets were appropriately engaged by the therapeutic agent, potential differences between drug formulations used in phase 2 and 3 trials, and whether the mechanisms of action were correctly understood and utilized to guide the design of the therapeutic and the clinical trial.

3. Lessons from other fields

Given this history, lessons from other disease areas, specifically those where some degree of disease modification has been achieved, can be valuable. Multiple sclerosis (MS) has a number of agents available for therapy, and early treatment changes the overall disease course and disability accumulation. Several elements can be identified as potential sources of success. Autoimmune encephalomyelitis is a useful (albeit imperfect) animal model of central nervous system (CNS) inflammation to answer questions about different aspects of the immune response, subsequent repair, and potential effect of therapeutic interventions [11]. New and active lesions on MRI have been established as a sensitive and specific biomarker of treatment response and now are the most commonly employed Phase II outcomes in relapsing MS [12,13]. The field continually integrated new science to improve measuring clinical outcomes, such as the Multiple Sclerosis Functional Composite to measure disability in a more sensitive manner [14]. Large-scale collaborations are exemplified by the Clinical Path Institute-led MS Outcome Assessment Consortium, which provided a pathway to obtain regulatory agreement on new clinical outcomes [15]. The International Progressive MS Alliance has linked together MS organizations, researchers, health professionals, pharmaceutical industry, trust, foundations, donors, and people affected by MS from over twenty countries into a global collaboration to address the unmet needs of people with progressive MS (www.ProgressiveMSAlliance.org).

Conversely, important lessons can also be taken from fields that have struggled to achieve disease modification goals, like several attempts targeting amyloid in Alzheimer disease [16]. Target engagement assessed by PET quantification of beta-amyloid (a modality as yet unavailable in PD) was demonstrated in nearly 70% of all anti-amyloid trials, yet without corresponding clinical benefits. Hence, an anti-protein approach may be focused on the “tail end” of many biological abnormalities but not on the pathogen triggering the protein aggregation [17]. Given the parallels between AD and PD, we must recognize that clarity about the precise pathologic role of misfolded proteins remains elusive, and can explain some of the lack of progress in clinic.

These findings are informative and can be utilized across the therapeutics development spectrum to build confidence in advancing therapeutics towards human use. Reliable biomarkers or surrogates are well worth aiming for; however, the field also needs to recognize that it is not always possible to achieve these goals in an organ that cannot be easily sampled and that many highly effective therapies lack a clear understanding of the specific biological target(s).

4. Therapeutic target selection

Some common pathological features of PD are well described, including mitochondrial dysfunction and the formation of Lewy bodies. It is not certain, however, that the observed tissue changes are pathogenic [18]. In terms of etiopathogenesis, a number of loci and genes are associated with PD, ranging from causal to low risk. The study of genetics has created the opportunity to study new therapeutic targets and develop disease-relevant models [19].

It is important to remember that PD is much more than a motor disease, and pathology extends far beyond motor control areas. The Braak staging suggests progressive involvement of the CNS, and also offers potential clues to disease onset [20]. The first sites affected in the brainstem can be consistent with a spread via the vagus nerve for example, and recent work has shown the potential for alpha-synuclein to spread to the brain after injection in the intestine [21]. Pathogenic proteins also appear to have the ability to propagate between cells, through direct penetration, endocytosis, or receptor-mediated internalization.

Extensive work investigating the cellular processes underlying neurodegeneration in PD has enabled the identification of putative therapeutic targets. These include alpha-synuclein [22], with strategies targeting protein clearance, pathologic aggregation, post-translational modification, and others [23]. Several agents are in trials [24–26]. Other targets include GLP-1 [27], LRRK2 [28], c-Abl [29], PARIS [30], PARP1 [31], and others [32]. Some of these targets can be applicable to a wide range of PD subjects, while others, particularly specific genetic therapies, are the domain of precision medicine [33]. The huge scale of this global effort to address unmet need in neuroprotection for patients with Parkinson’s disease has recently been documented in an overview of the pipeline [34]. Dialogue between stakeholders could allow broader understanding of the spectrum of therapeutic targets, and of the basic research and clinically defined targets, to harmonize research efforts.

Target validation has been proposed as a potential central element in the quest for improving therapeutic development. It is possible that this lack of validation is responsible for a large proportion of trial failures. Early investment in developing predictive tools and models and better understanding of the relevant pathogenic pathways would likely lead to success. Relevant biomarkers of target engagement can also significantly improve the probability of successful therapeutic development, and de-risk transition of therapeutics from preclinical to clinical stages.

Evidence of target engagement includes proof that the test agent interacts with the intended pathogenic target, with a direct or indirect biological consequence; and correlation with drug exposure, with ability to predict efficacy. Moreover, ability to corelate drug exposure with efficacy at the preclinical stage can potentially de-risk significant investment in clinical trials.

It is worth considering whether proving target engagement should be universally treated as a strict no-go criterion in phase 2 development, with reliable biomarkers of target engagement potentially allowing clear go/no-go decisions. It is important, however, to view target engagement in the disease context in a specific organ system, and understand that this can sometimes be impossible to achieve (for example, if brain tissue samples were the only option for measurements). A higher likelihood of success may be possible with specific genetic etiologies. For instance, LRRK2 activity can be measured in the CSF [35], blood [36], and urine [37]. Developments in detection technology can remove some barriers to achieving reliable readouts.

Animal models are a critical tool for identifying therapeutic targets. Toxic or acute injury models have reproduced the loss of dopaminergic neurons, but have limited relevance to the progressive disease process in PD. Genetic models, on the other hand, offer a chance for mechanistic relevance [38]. A particularly attractive option is alpha-synuclein overexpression [39]. Models based on this approach reproduce a much wider range of PD deficits than older injury models, with construct, face, and predictive validity, albeit with some limited phenotypic expression. This type of approach can potentially narrow the gap between animal work and human therapeutic relevance.

Epidemiologic data can inform hypotheses to be tested experimentally, with a goal of developing therapeutic strategies. In order to be useful, the data need to be strong, consistent, specific, coherent, and demonstrate exposure-response and plausible causal relationships – although by their nature these data cannot, typically, demonstrate causality. For better translation of epidemiologic findings into PD strategies, awareness of disease heterogeneity, improved biomarkers, and predictive preclinical models are needed. Epidemiological data can otherwise easily mislead, and some recent negative phase 3 trials were preceded by highly suggestive epidemiological signals [40].

Technologies such as induced pluripotent stem cells (iPSCs) can offer unique strength as research tools and therapeutic agents. A recently developed technology uses artificial intelligence (AI) for prediction of pluripotency [41], and restoring A9 dopaminergic function from allogeneic or autologous approaches is now possible [42].

5. Approaches to agent selection for phase II trials

Requisites for successful drug development include an understanding of the disease pathology, information on drug safety and target pharmacology, a characterized population with propensity to respond biologically and manifest measurable clinical response, clinical measures to assess relevant outcomes, and high-quality trial design and implementation.

Prioritizing compounds or approaches is an ongoing problem for research programs. Several approaches exist among organizations engaged in the funding and execution of research.

The Linked Clinical Trials (LCT) initiative (Cure Parkinson’s UK) is a program, initially focused on drug repurposing, specifically aimed at identifying drugs that can slow PD progression [43,44]. A global committee conducts yearly reviews of “dossiers” describing individual candidate compounds, which are triaged and ranked. The most promising compounds move forward for clinical testing. By 2020, from around 200 drugs evaluated over 9 years, >50 drugs have been prioritized by this committee, half of which have already entered (or completed) clinical trials, mostly in Phase 2 or Phase 3 neuroprotection studies.

The Michael J Fox Foundation for Parkinson’s Research (MJFF) is a global strategic funder of PD research and drug development. The organization’s active research reflects an increasing interest in targeting disease mechanisms, and efforts to better define and measure PD [45]. The process for selecting agents to move further in development entails assessing the rationale (genetic and pathological links, preclinical and clinical validation, epidemiology), the drug, the outcome measures, and the feasibility. Some of the perceived areas of needed improvement are better understanding of biological processes and targets, converting these into better measures of disease, better matching of drugs to target populations, and increased efficiency in extracting and sharing information from studies.

NINDS funds a large portfolio of neuroscience research. The key areas of focus to be considered are pre-clinical research, biomarkers of target engagement or downstream outcomes, and the clinical trial pipeline. The institute is supporting several clinical trial networks, as standing infrastructures that optimize the trial pipeline [46]. Novel trial designs are given strong consideration, including response-adaptive randomization (platform trials, described below). The availability of biomarkers would almost certainly increase the likelihood of success [47], and NINDS has significant investment in biomarker research and biorepositories.

The Parkinson Foundation (PF) is dedicated to understanding PD through research, ensuring better care for everyone with PD, and educating and empowering the PD community. While not directly funding phase 2 or above clinical trials, it focuses on meeting the requirements that can allow successful later trials.

The process for prioritizing targets and conducting research has specific challenges in industry. Prioritization and probability of success depend on technical, regulatory, and commercial considerations. In order to be considered, a therapeutic agent should ideally have proof of mechanism of action, of engaging the relevant disease biology, and of the ability to produce a measurable clinical effect. Included here is the need for assessment of appropriate dose range and therapeutic index. The use of any biomarkers is contingent upon their meeting requirements and clear roles – fit-for-purpose biomarkers.

6. Biomarkers (Fig. 1)

Fig. 1.

Biomarker types and staged applications.

Biomarkers have been defined as a characteristic that is objectively measured and evaluated as an indicator of normal biological process, pathogenic processes, or pharmacologic responses to therapeutic intervention. They can be applied for diagnosis, as measure of disease severity or response to therapy, as measures of target engagement as discussed above, and critically, for population selection, discussed below.

Biomarker development efforts have focused on various areas. This includes biofluids, especially CSF and blood [48], targeting mainly alpha-synuclein species; imaging, with techniques holding promise being network imaging - the Parkinson’s disease-related pattern (PDRP) has been replicated in multiple populations and used as an outcome measure in trials [49], and quantitative DAT scan as potential progression marker [50]; tissue, including skin and salivary glands, with some particularly promising results from skin synuclein seeding assays [51]. Combination or multi-modal biomarkers may prove superior to individual assays [52]. This can involve synergistic markers targeting various pathologic processes, or different approaches to the same putative pathologic process, like measuring alpha-synuclein in various tissues [53].

A different approach to outcome measures and biomarkers is required when considering prodromal disease [54]. In addition to the clinical markers, a number of candidate biomarkers are under investigation. If progress is achieved in the therapeutic pipeline, this will need to be paralleled by validation of prodromal biomarkers.

Large-scale efforts are dedicated to the development of PD biomarkers. This includes the Parkinson’s progression markers initiative (PPMI) [55]; the NINDS Parkinson’s Disease Biomarkers Program (PDBP) [56]; large collaborative efforts like the Accelerating Medicines Partnership – Parkinson’s Disease (AMP-PD) [57]; and other efforts on various scales.

7. Trial design

While a focus on early development is appropriate, rigorous design of late phase clinical trials cannot be neglected. The FDA Guidance for Industry and Review Staff regarding product profiles is an important strategic development process tool [58]. Differences in approach and tools available need to be considered when addressing the same question in different contexts, such as academic vs industry environments.

Trials of symptomatic therapies are comparatively more straight-forward to design than those focused on disease modification, given that success is possible without a biomarker of pathogenesis. In trials of progression in PD, a major issue has been defining “progression”, with little concordance across different outcomes. This endpoint choice is a critical issue to resolve. Ideally, an endpoint would reflect the primary objective of the trial, and could be measured in all study subjects, specified before the start of the trial, and as objective as possible. However, a large number of endpoints in a clinical trial can also be counter-productive.

Some of the most recent trials have been adequately powered, had appropriate follow-up duration, had well-chosen outcome points, and well-selected populations [59]. As a common limitation of all disease-modification PD research to date, evidence of target engagement and biomarkers were absent or limited. The primary limitation in phase 3 PD trials focusing on disease modification is likely the lack of a short-term, symptom-unrelated outcome measures.

Clinical outcome measures also bear revisiting, as the limitations of scales like the MDS-UPDRS are becoming apparent [60]. Composite outcomes extending beyond the MDS-UPDRS could improve precision of trial metrics. One recent example is the AI-derived composite measure PDCORE [61].

The recent progress in technologies dedicated to objective measurements of disease characteristics can potentially provide more reliable deep phenotyping in PD [62]. Current assessments tend to be categoric, episodic, and subjective. Technology-enabled deep phenotyping can improve the understanding of the disease, enable personalized care, and develop objective disease or outcome measures. Wearable and connected remote devices/instruments can provide objective, sensitive, frequent, and remote measures that cannot be otherwise obtained.

Historically, it appears that early phase trials had a relatively low bar for proceeding to pivotal studies. Phase 2 trials are typically the most difficult to design. Clear criteria for moving to phase 3 are needed, particularly no-go criteria that would stop further research if unlikely to succeed.

Master protocols, which constitute a collection of trials or substudies that share key design components and operational aspects, can potentially facilitate efficient trial design strategies to expedite development of drugs and biologics [63]. They can allow several innovative approaches:

• Basket trials, in which a master protocol is designed to test a single intervention in different populations (defined by disease stage, biomarkers, etc.)

• Umbrella trials, in which multiple investigational interventions (single or in combination) are tested in a single disease population

• Platform trials, in which multiple targeted interventions are studied in a single disease in a perpetual manner, with therapies entering and leaving the platform according to an algorithm.

The shared infrastructure can streamline trial logistics, improve data quality, and facilitate data collection and sharing. The use of common protocols enables a broader set of objectives to be met more efficiently. Shared placebo groups, which can apply to any of these setups, can result in significant sample size (and cost) savings.

Important limitations of this approach can include an inability to precisely match interventions to populations, requiring a way to build a personalized medicine approach into the platform. The dynamic nature of master protocols is well-suited for fast pace of precision medicine drug development, but the infrastructure requires more resources up front.

Examples of “defensible” trial design approaches in the context of current limitations include: prevention trials of low-risk interventions in pre-manifest individuals; large trials of treated PD patients asking simple, well-defined questions; pathway-targeted interventions in genetically defined PD patients or other situations with clear measures of target engagement. On the other hand, short-term trials using MDS-UPDRS as the main outcome have proven very limited.

Some proposed principles of good trial design include: the involvement of persons with Parkinson’s in protocol development and review of trial implementation; accounting for symptomatic effects, depending on endpoint; adequate powering to avoid type II error in phase 3 trials, and carefully selected approaches to gating progression in phases 1 and 2; early stopping rules for futility; tracking target engagement and imaging endpoints, as well as markers of proximal and downstream pharmacodynamics; selection of appropriate patient population; banking biosamples for future evaluation; assessment of both prespecified and post-hoc subgroups; and contributing anonymized data to public repositories.

8. Population selection

Population selection and matching the intervention to the right subjects is emerging as a potential central factor in the lack of success to date. PD is recognized as a heterogeneous disease (or even group of diseases), yet the typical clinical trial tends to include several subtypes of the condition. The therapy of interest needs to be matched to the population targeted.

Choosing the right patients at the right time for therapeutic trials has been critical in fields like oncology [64]. Patient stratification in clinical trials according to diagnostic, severity, and prognostic biomarkers offer increased chances of positive response. While many biomarker development efforts have favored convergent processes, like alpha-synuclein or dopamine transporter imaging, divergent biomarkers, present in some by virtue of identifying a pathogenic mechanism, but absent in others, are a critical component of precision medicine. Genetic predispositions can be leveraged for this purpose in PD [65]. Longitudinal progression of motor and cognitive features may be different in genetic forms of parkinsonism, and this information can guide outcome selection, and timing of patient enrollment in trials. The issue of timing is a critical component of matching populations to therapy, and even more so if an intervention is to be deployed in prodromal stages [66].

Most PD cases display mixed pathology. Some of the features considered hallmarks of the disease, like Lewy body pathology, have a limited correlation with the amount of cell loss [67]. Furthermore, this pathology is not required for a parkinsonian phenotype to occur. This problem extends to the use of biomarkers, as the candidate biomarkers studied to date have tended to show limited consistency across cohorts. This is because generating disease subtypes based on phenotype may not predict the pathobiology adequately.

A potential path forward involves reimagining the therapeutics development process. Biomarkers can be used as markers of underlying biology for subtyping [68]. Information from genetically-defined cohorts can be used to inform biomarkers and therapeutic development for those subtypes. This can help match the interventions to the right populations and drive higher chances of success.

The field of genetics research can focus trial design and identify more homogeneous and selective patient populations. Focused markers of drug activity can be obtained. And at-risk carriers can be identified for early intervention.

Ongoing efforts are directed at targeting LRRK2-associated PD [69] and GBA-associated PD [70]. As mentioned above, it is worth noting that even within genetically-defined conditions such as these, significant heterogeneity exists [71], and choices of disease stage, sub-phenotype, outcome measures, etc. are difficult, and need to be addressed if the opportunities presented by genetics are to be taken advantage of.

9. Increasing collaborations and stakeholder Synergy

Using lessons learned from prior efforts, including large scale collaborations like NET-PD and large industry trials, collaboration models between funding and advocacy agencies, academic centers, and industry/companies, can be outlined (Fig. 2). Organizations like the NIH have the advantage of non-dilutive flexible funding and can serve as bridges to other agencies, like the FDA and CMS. Academic centers can provide focused time and effort, and can excel in work on translational preclinical models, and trial design and execution. Industry partners can develop compelling drugs and have expertise and ability to execute longitudinal development, as well as to raise capital and attract investments. Foundations like the MJFF, PF, and collaboration networks can provide targeted program development and partnerships.

Fig. 2.

Collaboration model for PD drug development.

One way to achieve more efficient progress is to develop program-specific synergies among stakeholders and different research teams working towards similar goals. For example, a recent target of great interest has been c-Abl, based on extensive basic science identifying its role in pathogenesis [29]. Several parallel efforts have been undertaken to assess the therapeutic potential of c-Abl inhibitors. One agent, nilotinib, was studied in two simultaneous clinical trial efforts [3,72]. The results were similar and showed limited CSF penetration, discouraging further development of this agent. Future efforts would benefit from prospective efforts at harmonization and standardization, which would permit pooling of trial read-outs for joint analyses. These efforts should consider all potential stakeholders and be mindful of the limitations, such as protection of intellectual property and programmatic barriers.

In any such future efforts, including the voice of the patient community, with input into outcome measures, study design, and other aspects, will be critical.

10. Summary and recommendations

The following areas are identified as potential key elements for future research directions:

• Population Selection: A narrower focus in terms of population studied, for example defined by genotype and sub-biotype, is more likely to yield actionable results, which can then be modified and/or scaled to wider PD populations. Gene-targeted therapies may hold the most promise, at least in well-defined cases. Some of the workshop participants and the PD research community feel that this is likely the key limitation of prior efforts.

• Target Engagement: Markers of therapeutic target engagement can more precisely guide therapeutic efforts. Incorporating strong go/no-go criteria linked to effects on target, where appropriate, can increase the likelihood of success.

• Biomarkers: Biomarkers of population selection, pathogenic targeting, and treatment outcomes can be expected to accelerate therapeutic development.

In addition, the following areas can improve likelihood of success and optimize the research pipeline:

• Candidate selection: The selection of top candidate therapies would benefit from improved, scientifically rigorous approaches. Existing approaches could potentially be modified and scaled to fit this wider purpose.

• Trial design: Trial design should be updated and adapted to maximize chances of success. Master protocols and adaptive designs can, under the right circumstances, prove superior to traditional designs.

• Collaborative environment: Collaboration between stakeholders can reduce redundancies and generate beneficial synergies. Models like the one presented in Fig. 2 can serve as an initial basis for further development. A global view focused on programs rather than disparate projects may be the more efficient approach.

11. Conclusions

A multi-decade history of attempts to achieve Parkinson’s disease course modification using therapeutics has had limited success. Lessons learned from these efforts point to population selection, target selection, biomarkers and markers of target engagement, and clinical trial design as bottlenecks for therapeutics development.

These points can be incorporated into strategies to be pursued at the level of individual organizations as well as through collaborations among funding agencies, industry partners, academic institutions, and advocacy groups.

Additional strategic details need to be generated at the level of basic and translational science, and addressing non-drug therapies, to ultimately devise an actionable global PD disease modification strategy.