Diagnosing, prognosing, and predicting psychiatric illnesses more precisely is critical for early intervention and better disease management. Traditional diagnostic and treatment pathways primarily rely on clinical interviews, somatic symptoms, and clinicians' subjective opinions. Precision psychiatry aims to combine diverse markers, such as clinical presentation, biological, neuroimaging, genetic, metabolomic, and digital biomarkers, to create a more objective and holistic phenotype for patients.1
The success of precision medicine in cardiology and oncology offers a roadmap for psychiatric drug development—but it’s complicated. In cancer, the explosion of molecular targets fragmented once-large diagnostic categories, such as non-small cell lung cancer, into ever smaller subgroups defined by the presence or absence of actionable biomarkers.
Major depressive disorder (MDD) is equally heterogeneous, complex, and variable. However, single biomarkers have thus far failed to identify phenotypic subtypes of MDD accurately.2 Cortisol, serotonin, and brain-derived neurotrophic factor (BDNF) are biomarkers for depression, but social, environmental, and genetic predictors also play a significant role in whether some patients get depressed, while others don’t.3
Rather than starting with an illness definition, precision psychiatry seeks to link symptoms to specific neurobiological pathways that novel treatments can target.
At Parexel, we increasingly collaborate with sponsors to explore how to incorporate precision medicine techniques in developing psychiatric treatments. It’s a nascent effort with a long road ahead, but we do see progress in critical areas, including:
Diagnostic biomarkers can identify patient subgroups
Many sponsors are working to characterize and identify smaller subpopulations of psychiatric patients, such as those with treatment-resistant conditions or specific pharmacogenomic biomarkers.
Recently, we worked with a sponsor to operationalize a novel diagnostic subgroup of patients with treatment-resistant bipolar depression. While there is medical literature describing this subgroup, it is not well defined by regulatory precedent. The company needed to develop a clear definition and diagnostic pathway that was acceptable to the FDA. We helped them seek regulatory advice and present their criteria for diagnosing patients to enroll in a late-stage efficacy trial.
In May 2024, Denovo Biopharma announced results from a Phase 2b clinical trial of treatment-resistant depression designed to validate a novel pharmacogenomic biomarker—the study marks one of the first times a genetic biomarker has been used to select patients for treatment in psychiatric disease. 4
While researchers are discovering more blood biomarkers, they have yet to be widely utilized in psychiatry, and few standards exist. However, many protocols collect biomarkers as exploratory endpoints. At Parexel, we routinely support studies that collect novel blood and digital biomarker data as well as pharmacogenomic and epigenetic samples. Sponsors often collect these data to evaluate exploratory endpoints that could inform Phase 3 trial designs and outcome selection.
Digital biomarkers, based on features from voice samples and patterns of physical activity (captured via actigraph) and engagement, are increasingly used as secondary endpoints in conditions such as MDD. Sponsors incorporate the data in multivariate models to measure their prevalence in the target patient population and test whether they can predict treatment outcomes for defined subgroups.
Imaging biomarkers are less common in psychiatric trials, in contrast to their increasing use and utility for diagnosing and staging Alzheimer’s disease and other dementias. Progress has been bumpy. One recent study analyzed whether machine learning could identify a multivariate neuroimaging biomarker for MDD that could accurately classify MDD patients. The study collected structural and functional magnetic resonance imaging and polygenic risk scores for depression but did not identify an informative individual-level MDD diagnostic biomarker.5
Symptom clusters that reflect biological pathways can stratify patients
Sponsors are grouping and stratifying patient populations with greater precision. Clinical trials now routinely divide heterogeneous indications such as depression and schizophrenia into discrete symptom clusters that better represent biological pathways and provide clear drug targets.
For example, anhedonia—an inability to experience joy or pleasure—afflicts up to 75% of MDD patients and is common in other mental health conditions as well. In the not-too-distant past, most trials of MDD patients failed to break out anhedonia-positive and anhedonia-negative subgroups. At Parexel, we increasingly work on trials that test new agents in MDD patients with moderate-to-severe anhedonia rather than all comers. Likewise, sponsors can enroll MDD patients with insomnia, excluding those without.
The heterogeneity of psychiatric illnesses contributes to variable treatment responses. The goal is to target specific biological mechanisms in subgroups to help stratify these heterogeneous populations with focused treatments. Recently, we ran three Phase 3 studies in patients with cognitive impairment in schizophrenia (CIAS). Researchers are increasingly interested in CIAS patients because approved antipsychotics for schizophrenia modulate hallucinations and delusions rather than cognitive impairment, a core symptom that is associated with poor clinical outcomes.6 Cognitive markers served as the primary endpoints for the trials rather than conventional symptom scoring, such as the Positive and Negative Syndrome Scale (PANSS).
We also see sponsors refining the target patient population based on metabolic profiles. Some seek to stratify patients by measuring the level of CYP2D6, an enzyme produced by the CYP2D6 gene that affects how drugs are metabolized. Based on their CYP2D6 activity, patients can vary between low, intermediate, and high metabolizers, and quantifying helps physicians choose the correct dose or exclude certain patients. Of course, this is not unique to neuroscience.
Rather than starting with an illness definition, precision psychiatry seeks to link symptoms to specific neurobiological pathways that novel treatments can target. Recently, we worked with a sponsor to study the effects of an already-marketed therapeutic on impulsivity in adult ADHD patients. Impulsivity is a symptom characteristic of a broad range of mental illnesses. In clinical trials, it can be measured by questionnaires and behavioral laboratory tasks. Using these endpoints as proxies, studies can estimate the prevalence of impulsivity in specific patient populations. The sponsor is exploring the best metrics to measure impulsivity in an initial trial to gather pharmacokinetics data, define population variability and effect size after treatment, and inform endpoint selection.
Developing a symptom-targeted drug (in this case, impulsivity) across a variety of mental illnesses is new territory in neuroscience. In oncology, the tumor-agnostic approach, which treats patients according to their tumors' genetic mutation(s) rather than their type or location in the body, is now well-established.
Sponsors are grouping and stratifying patient populations with greater precision. Clinical trials now routinely divide heterogeneous indications such as depression and schizophrenia into discrete symptom clusters that better represent biological pathways and provide clear drug targets.
RWE can generate adherence data and optimize protocols
In psychiatry, real-world evidence (RWE) has the potential to be more valuable to patients and payers than data from randomized controlled trials (RCTs). RWE can capture the impact of culture, socioeconomic status, and education on patient adherence and outcomes. Designs can be more patient-centric and decentralized, lowering the cost of long-term data collection.
We have worked with a vendor on several studies that use a smartphone app to track adherence by recording patients each time they self-administer a treatment. If doses are missed or other issues arise, the app communicates with the site or monitoring team in real-time to allow early intervention. However, external technical support staff must troubleshoot adequately when studies incorporate smartphone apps.
We have also participated in real-world studies of digital therapeutics. For example, some sponsors have asked whether regular cognitive assessments could be performed on patients’ cell phones, which can theoretically help detect changes earlier than through conventional visits alone.
Measuring smartphone usage can allow deeper digital phenotyping by capturing activity, sleep, and online support tools via a companion app that can be used at times convenient for the patient. It also helps manage the visit burden. However, capturing continuous digital biomarker data in a large clinical trial produces a lot of data to analyze and interpret. At Parexel, we have developed protocols for managing the analytics of large datasets, which is essential to execute digital therapeutics studies successfully.
Real-world data can help refine and support study designs and protocols, particularly quantifying patient characteristics such as concomitant medications, medical histories, and geographic locations. It can also identify patient populations that may otherwise be missed by conventional studies. We use multiple secondary data sources, including census, pharmacy, and insurance claims data, to inform trial designs. The data also help sponsors achieve FDA-mandated levels of diversity in late-stage trials.
Correlating biomarkers with outcomes will accelerate precision psychiatry
Serious mental illnesses decrease life expectancy on average by a decade and impose lifetime societal costs of $1.85 million per patient.7 Despite the burdens these disorders impose globally, investment and innovation in psychiatry lag behind other therapeutic areas, notably cancer.8 Gaps in basic research, diagnosis, treatment, and outcomes persist.
Precision medicine in psychiatry is more challenging because we don’t have clear molecular targets or etiologies. However, as more researchers and sponsors collect and validate novel, multivariate forms of biomarker data and correlate them with outcomes, we draw closer to a diagnosis and treatment framework that can offer patients more personalized and effective therapies.
Contributing Expert
