Can innovative trial designs work in neuroscience?

Innovations that accelerate clinical trials are now commonplace in oncology and other therapeutic areas. Adaptively designed and biomarker-driven studies test efficacy in defined patient populations, and trials may morph seamlessly from safety and dose-finding to efficacy. For example, the speed and success of COVID-19 vaccine development were partly due to the adoption of seamless global Phase 1/2/3 trials with parallel activities and decision-making. 

Neuroscience trials remain traditional because of the heterogeneity of neurologic, neurodegenerative, and psychiatric diseases and the relative absence of actionable biomarkers. Yet adaptive approaches and biomarkers could optimize trial designs and make development more efficient.

Recent advances in the field, including the emergence of new blood and imaging biomarkers in Alzheimer’s disease (AD), multiple sclerosis (MS), and major depressive disorders (MDD), suggest the time has come for invention and cross-fertilization. At Parexel, we advise sponsors on using innovative trial designs, novel biomarkers, and external data to accelerate neuroscience drug development.

Innovative trial designs

An innovative trial design can answer multiple questions about one or more compounds in one or more conditions or patient subgroups. Pre-specified, “adaptive” modifications of the trial protocol can allow changes during the study based on interim data analyses. For example, researchers can increase the sample size of a trial or drop a dose arm that is ineffective or has undesirable side effects. Trial designs with adaptive elements require close collaboration between biostatisticians, clinical operations, data management, medical experts, and project leadership. 

Neuroscience sponsors have yet to widely adopt adaptive trial designs because they require a detailed understanding of disease etiology, staging, and progression. Many cancer trials use biomarker data to trigger adaptations, but in many neurological diseases—especially neurodegenerative and psychiatric diseases—we still don’t have validated, actionable biomarkers. 

At Parexel, we have worked with neuroscience sponsors to use adaptive designs to measure a drug’s impact on shorter- and longer-term outcomes. For example,  we recently helped a sponsor design an adaptive Phase 2 epilepsy trial that assessed two drug doses versus a placebo. The first portion of the trial measured efficacy by the standard endpoint of seizure reduction and included an interim analysis at 12 weeks. If either dose did not show a reduction, it would be dropped, followed by an extended evaluation of the drug’s impact on cognition and mood endpoints. This adaptive design enables the sponsor to evaluate a drug’s effects on critical clinical endpoints and comorbid conditions not captured in traditional epilepsy trials and efficiently informs future development. Although our adaptation for this protocol was based on the drug’s specific mechanism of action, the design could be applicable more widely.

We recently helped a sponsor design an adaptive Phase 2 epilepsy trial that enables them to evaluate a drug's effects on key clinical endpoints and critical comorbid conditions not captured in traditional epilepsy trials.

Novel biomarkers

Early-stage drug development has typically been a linear progression from preclinical work to safety to proof-of-concept to pharmacokinetics and pharmacodynamics (PK/PD), ending with the optimal effective dose for confirmatory trials. However, biomarker data can inform smarter and earlier product advancement and termination decisions.

Although the FDA is increasingly open to novel biomarkers, sponsors must be mindful of their purpose and limitations. When selecting, testing, and establishing biomarkers for use in traditional or innovative early-stage trials—such as integrated Phase 1-2 or adaptive trials—sponsors can engage with regulators early to test their rationales and data.

At Parexel, we advise sponsors to assess whether biomarkers can be converted into surrogate endpoints for accelerated approval (AA) in the United States or conditional marketing approval (CMA) in the EU. Where an experimental agent addresses a patient population with high unmet needs, in a disease where the pathogenesis is well-understood and the biomarker(s) make sense scientifically, regulators may show greater flexibility. However, AA and CMA regulatory pathways require careful planning. If a study was designed with endpoints for traditional approval, but the outcome was unexpected, and the sponsor wishes to pivot to AA or CMA, regulators are unlikely to allow it.

We advise sponsors to assess whether biomarkers can be converted into surrogate endpoints for accelerated approval in the United States or conditional marketing approval in the EU.

In neuroscience, novel biomarkers are emerging rapidly, offering ways to stratify patients and construct surrogate endpoints for clinical trials that could advance patient care. For example,  investigators recently used precision functional mapping tools, including serial multi-echo functional magnetic resonance imaging (fMRI) scans, to map the brain networks of people with depression. They found that a brain circuitry called the “salience network” was twice as large in the cortex of depressed individuals as it was in healthy ones and may predict depressive symptoms. Another team of researchers reported that a simple blood test analyzed in a lab, the amyloid probability score 2 (APS2), detected AD pathology accurately 91% of the time; by contrast, AD pathology is only detected 73% of the time by dementia specialists and 61% by primary care physicians.

Reliable biomarkers hold tremendous potential for optimizing trial design and patient care.

Improving development efficiency

According to a recent report, 127 compounds are being tested in 164 Phase 1 to 3 clinical trials for Alzheimer’s: 32 drugs in 48 Phase 3 trials, 81 in 90 Phase 2 trials, and 25 in 26 Phase 1 trials. These active AD trials require 51,398 participants. And AD is just one of many nervous system and brain disorders that impose heavy financial and societal burdens. Recruiting patients is becoming a bottleneck for drug development.

One solution is combining separate trials into a single platform trial that compares multiple treatments simultaneously against a single control. For instance, the Healey ALS Platform Trial is a perpetual adaptive trial that tests experimental agents and regimens to treat amyotrophic lateral sclerosis (ALS) utilizing a shared placebo group and central infrastructure. A team of ALS experts selects investigational treatments to add to the Master Protocol on a rolling basis.

Another approach to speeding development is combining anonymized patient-level data from the treatment and control arms of completed clinical trials. The Pooled Resource Open-Access ALS Clinical Trials Database (PRO-ACT) contains more than 11,600 records from 29 Phase 2/3 clinical trials. By providing information on endpoints, PRO-ACT can guide future study design and make it more efficient.

In neuroscience, real-world evidence (RWE) and artificial intelligence (AI) could also accelerate development by providing external control data that could reduce the number of patients required for trials. 

For example, researchers recently used a probabilistic deep learning model of AD progression to predict outcomes using the baseline data of 184 AD patients from the placebo arm of a completed Phase 2 study in a “digital twin” virtual trial arm.  Another study suggested that virtual brain twins could improve the diagnosis, treatment, and prognosis of patients with neuroscience disorders and test potential pathological mechanisms.

Sponsors can accelerate drug development by working with regulators to explore innovative trial designs, novel biomarkers, and RWE and AI-aided approaches to generating valid external control data.

Recruiting patients is becoming a bottleneck for drug development. One solution is combining separate trials into a single platform trial that compares multiple treatments simultaneously against a single control.

Contributing Experts