Genetics and MRI offer a compelling way to think about biological trajectories across the schizophrenia–bipolar spectrum. Framing this as a hypothesis-generating stratification model, rather than a clinical decision-making tool, is not a limitation. It is a deliberate choice, and a consequential one.

Why does the framing matter so much?

Unlike in academic research, where imaging can serve as a relatively flexible exploratory tool, clinical trials for drug development add extra layers of regulatory, operational, and evidentiary responsibility. Even when academic imaging protocols are approved by ethics committees and are methodologically rigorous, their role is usually different from their role in a drug-development pathway.

One critical dimension is the role assigned to the neuroimaging biomarker.

In regulatory terms, this is a context-of-use problem: the question is not only what the biomarker measures, but what decision the biomarker is allowed to support.

These considerations apply particularly from Phase 2 onward. In Phase 1, PET imaging is routinely used to characterize target engagement, receptor occupancy, and pharmacodynamic signals. Regulatory expectations for this use are well-established, and the risks described here do not yet apply.

If neuroimaging is used in a trial, when does it remain a learning tool? And when does it become part of the evidence that must be reproduced, justified, and potentially carried into later development stages, and possibly into the approved label?

The regulatory risk does not come from acquiring MRI data. It comes from making MRI necessary for interpreting, reproducing, or applying the treatment effect. At lower levels, imaging can help explore mechanisms, characterize patients, or generate subgroup hypotheses. At higher levels, imaging can begin to define who enters the trial, who is considered likely to benefit, or which population the evidence for efficacy supports.

This is why stratification and enrichment should not be treated as interchangeable. Stratification can support learning, balance, or subgroup analysis. Enrichment starts to shape the population in which the treatment effect is demonstrated.

This distinction is especially important in psychiatry. Neuroimaging is not part of routine treatment selection in most psychiatric clinical workflows. Therefore, if an imaging biomarker were used to define the population for a pivotal trial and later contributed to defining the on-label population, the imaging measure would need to be scalable, standardized, reproducible, interpretable, and accessible to the target psychiatric population.

The same reasoning applies to genetic or omics-based profiles. A biomarker can be scientifically attractive, but if it becomes part of the treatment-eligibility logic, it must also be operationally and clinically deployable.

So the question is not simply: can neuroimaging stratify patients?

The more difficult question is: how far can we let neuroimaging advance through the clinical-development pathway before it becomes a regulatory commitment?

How far should neuroimaging go in clinical trials?

Regulatory dependence scale · imaging role vs label risk

Increasing regulatory dependence
8
Companion / eligibility test
Label-defining
7
Pivotal enrichment biomarker
Very high risk
6
Required inclusion criterion
High risk
5
Pre-specified subgroup analysis
Moderate risk
▲  Decision-defining layer
· commitment threshold ·
▼  Evidence-generating layer
4
Secondary endpoint
Moderate risk
3
Stratification factor for analysis
Low to moderate
2
Exploratory biomarker
Low risk
1
Mechanistic readout
Low risk

The more the trial depends on imaging to prove efficacy in a selected population,
the more likely imaging becomes part of the approved treatment population.

This scale is not meant as a formal regulatory classification. It is a practical way to visualize how regulatory dependence increases as imaging moves from supporting biological interpretation to defining eligibility, enrichment, or treatment-use logic.

In other words, the regulatory issue is not the scientific value of imaging. It is the dependency created by the trial design. The more the trial relies on imaging to select patients, interpret efficacy, or define the population that benefits, the more the sponsor must be prepared to defend the biomarker's biological rationale, technical validity, operational feasibility, and clinical deployability.

This is where the regulatory gap begins.

For psychiatric trials, this gap is particularly sensitive. MRI can be powerful for research, mechanism exploration, and biological subgroup discovery. But if it is pushed too early into an eligibility-defining role, it may create a development pathway that is difficult to reproduce in real-world psychiatric care.

A more cautious strategy would be to keep neuroimaging first in the evidence-generating layer: to support mechanistic understanding, characterize heterogeneity, and test whether biologically meaningful subgroups exist. Only after reproducibility, clinical relevance, incremental value, and feasibility are established should imaging move toward the decision-defining layer.

The same principle applies to genetics, omics, and other high-dimensional biomarkers. A stratification signal is not automatically a trial-ready enrichment tool. It must earn that role.