CAR-T Cell Therapy Biomarkers: Can We Finally Predict Who Will Respond?

CAR-T Cell Therapy Biomarkers: Can We Finally Predict Who Will Respond?

You have a patient. Relapsed, refractory, eligible for CAR-T. The science is extraordinary, the product is approved, and the therapy genuinely cures some people. But you know that somewhere between 40 and 60% of patients will not respond. You infuse anyway because you have no validated biomarker to tell you otherwise.

That gap, between a therapy with curative potential and our inability to reliably predict who benefits, is one of the most important unsolved problems in immuno-oncology today.

This piece is not a comprehensive review of all things CAR-T. It is a focused look at where the biomarker science actually stands, what signals are earning clinical credibility, and where the field still has more confidence than evidence.

Key takeaways

• Pre-infusion T cell phenotype and inflammatory signatures are emerging as the most clinically actionable biomarkers for predicting CAR-T response.
• Tumor burden alone is no longer a sufficient predictor. Composite multi-parameter models show significantly stronger prognostic power.
• CAR-T resistance in solid tumors is driven by the tumor microenvironment as much as antigen biology.
• Biomarker validation strategy must be embedded in trial design from Phase 1, not retrofitted later.

Why prediction matters more than ever

CAR-T cell therapy costs somewhere between $400,000 and $500,000 per treatment course. Non-response is not just a clinical failure. It is an economic failure and an ethical one.

The case for predictive biomarkers is therefore not academic. It is urgent, financial, and patient-centered all at once.

If we can distinguish responders from non-responders before treatment, we improve cost-effectiveness, reduce unnecessary toxicity exposure, and create space for better alternatives for patients unlikely to benefit.

What we thought we knew: Tumor burden as the default signal

For years, tumor burden was the dominant pre-treatment variable. Lower is better. Higher correlates with worse outcomes.

The problem is that tumor burden captures disease load, not immune competence. Two patients with similar disease metrics can have radically different CAR-T outcomes because their immune systems are fundamentally different.

Tumor burden remains important. But it explains only part of the variance.

The pre-infusion window: your most underutilized data point

T cell phenotype at leukapheresis

Researchers identified a specific pre-treatment blood biomarker that predicts CAR-T response before manufacturing even begins.

Low frequencies of differentiated, exhausted T cells at leukapheresis are associated with stronger CAR-T outcomes.

Memory T cell subsets such as TCM and TSCM populations consistently associate with better CAR-T expansion, persistence, and clinical response.

Inflammatory signatures

Inflammatory baseline signatures are emerging as powerful predictors of CAR-T failure.

Composite inflammatory models integrating cytokines, CRP, ferritin, LDH, and end-organ markers are demonstrating strong prognostic power across lymphoma cohorts.

Moving beyond single markers: The case for composite models

Single biomarkers correlate with outcomes, but they do not fully capture the complexity of immune dynamics.

CAR-T and solid tumors: The harder problem

Predicting response in solid tumors is far more difficult than in blood cancers.

CAR-T cells face multiple barriers in solid tumors: trafficking challenges, extracellular matrix penetration, immunosuppressive microenvironments, and antigen heterogeneity.

The next-generation IO biomarker question is not simply whether CAR-T cells expand. It is whether the tumor microenvironment allows them to function at all.

• Multi-specific CAR constructs
• TME reprogramming strategies
• Synthetic circuit-based CAR systems

None of these approaches yet has a fully validated predictive biomarker.

Resistance mechanisms: What the tumor does next

Even in responders, durability is not guaranteed.

Antigen loss is the best-characterized resistance mechanism. Tumors can downregulate or completely lose the target antigen after CAR-T pressure.

T cell exhaustion, MHC class I downregulation, and epigenetic silencing of antigen-presenting machinery also contribute to relapse.

Circulating tumor DNA (ctDNA) is emerging as a powerful real-time monitoring tool for relapse detection and response tracking.

What the field needs to stop doing

The field continues to publish promising single-center biomarker signals without the infrastructure to validate them at scale.

Large multi-cohort validation studies, pre-competitive data sharing, and manufacturing standardization are essential for meaningful clinical deployment.

Companion diagnostic development also needs to begin in parallel with therapeutic development from the earliest clinical phases.

Where the data is pointing

The strongest predictive power is coming from composite models combining:

• Pre-infusion T cell phenotype
• Baseline inflammatory markers
• Early post-infusion kinetics

For solid tumors, future biomarker strategies will likely integrate:

• Spatial transcriptomics
• Tumor microenvironment profiling
• ctDNA tracking
• Multi-omics machine learning models

Bring your biomarker science to the table at Immunomark Summit

If you are working on CAR-T biomarker discovery, predictive modeling, TME profiling, or next-generation IO biomarkers, Immunomark Summit is designed for exactly these conversations.

The conference brings together biomarker scientists, translational researchers, and biotech leaders focused on advancing immuno-oncology science through rigorous, clinically meaningful biomarker research.

Register for Immunomark Summit