Opportunities for use of neuroimaging in de-risking drug development and improving clinical outcomes in psychiatry: an industry perspective

Abstract

Neuroimaging, across positron emission tomography (PET), electroencephalography (EEG), and magnetic resonance imaging (MRI), has been a mainstay of clinical neuroscience research for decades, yet has penetrated little into psychiatric drug development beyond often underpowered phase 1 studies, or into clinical care. Simultaneously, there is a pressing need to improve the probability of success in drug development, increase mechanistic diversity, and enhance clinical efficacy. These goals can be achieved by leveraging neuroimaging in a precision psychiatry framework, wherein effects of drugs on the brain are measured early in clinical development to understand dosing and indication, and then in later-stage trials to identify likely drug responders and enrich clinical trials, ultimately improving clinical outcomes. Here we examine the key variables important for success in using neuroimaging for precision psychiatry from the lens of biotechnology and pharmaceutical companies developing and deploying new drugs in psychiatry. We argue that there are clear paths for incorporating different neuroimaging modalities to de-risk subsequent development phases in the near to intermediate term, culminating in use of select neuroimaging modalities in clinical care for prescription of new precision drugs. Better outcomes through neuroimaging biomarkers, however, require a wholesale commitment to a precision psychiatry approach and will necessitate a cultural shift to align biopharma and clinical care in psychiatry to a precision orientation already routine in other areas of medicine.

Fig. 1. Key considerations regarding uses of neuroimaging biomarkers in Phase 1–3 of drug development as well as ultimate clinical care once a drug is approved.

Elements of these biomarker uses in each phase likewise feed forward into use in the next phase, which together act to progressively de-risk development. This begins with pharmacodynamic biomarkers in Phase 1 and culminates in scalable patient selection biomarkers with drug approval and clinical deployment. Most useful modalities for each phase are noted, in line with the uses and considerations involved.

Fig. 2. Feed-forward and feed-back flow of biomarker-related information on pharmacodynamic and clinical effects by systematically collecting biomarkers across drugs, populations and development phases.

Through iterative cycling through these processes over time and development programs, the predictive utility of early biomarker outcomes on later clinical development risk and clinical efficacy can be tightened, approximating progress already being made in other areas of biomedicine.

Fig. 3. Conceptual overview of the development and validation of a biomarker-based patient selection approach in clinical trials.

Central to this approach is the requirement for prospective validation of a biomarker’s ability to stratify populations on clinical outcome in an independent group of patients. Upon successful replication, a patient selection approach can be taken, in line with the FDA’s enrichment guidelines.

Fig. 4. Different Phase 2 trial designs for the development and validation of a biomarker for patient selection.

Three example approaches of progressively greater cost and time requirements are shown, along with how discovery and test data are divided for analysis and associated predictions. The first approach (a) is best suited for situations where external placebo datasets exists in which relevant biomarkers were collected. The use of a single drug arm design is the most streamlined way to identify and replicate a patient selection biomarker that will also translate best to how the drug would be used in clinical practice. The second and third approaches (b, c) are best suited to indications in which little data are available and biomarker-related knowledge must be generated entirely through the data being collected (RCT = randomized clinical trial). If there is substantial concern about whether biomarker discovery depends on drug exposure context (single arm vs. RCT) then the third approach (c) may be best. However, in this situation, there may be concern about how well a biomarker will generalize to clinical care where treatment more closely resembles the single arm context.