CAR-T Therapies and the Case for 'Data First, AI Later'

From Blunt Tools to Precision Weapons

Traditional treatments for autoimmune conditions and cancer relied on non-specific weapons—steroids, chemotherapy, and broad immunosuppressants. These were, in effect, blunt tools. Monoclonal antibodies (mAbs) marked a turning point. By using cloned immune cells to target specific proteins (like TNF-alpha), mAbs made it possible to design therapies with surgical precision.

The breakthrough in 1975 by Georges Köhler and César Milstein to produce monoclonal antibodies set the stage. CAR-T therapy took this one step further: instead of delivering an antibody, scientists now equip the patient's T-cells with a synthetic receptor—built using mAb fragments—that can seek and destroy specific cancer cells.

Health Data: The Hidden Enabler

Creating and delivering a CAR-T therapy requires more than lab skills. It demands data at every step:

• Genetic profiling to identify the target antigen

• Health records to determine eligibility

• Real-world data to monitor long-term effectiveness

• AI models to predict patient risk and guide infusion timing

But the data landscape is far from ideal. Fragmented records, biased datasets, and missing outcomes make it difficult to know if a €400,000 therapy is truly worth the cost, or who is most likely to benefit.

Enter HTA: The Value Guardians

Health Technology Assessment (HTA) bodies are tasked with deciding which therapies offer value for money. With CAR-T and similar high-cost interventions, HTA agencies face multiple challenges:

• Limited clinical trial data at the time of approval

• Uncertain long-term outcomes

• Difficulties quantifying non-economic value (hope, quality of life)

• Rising equity concerns when access is limited to high-income systems

To navigate this, HTA increasingly turns to AI-driven insights from electronic health records, registries, and patient-reported outcomes. Yet these tools are only as good as the data they rely on.

The "Data First, AI Later" Imperative

This is where the Health Data Forum's movement—Data First, AI Later—becomes essential. AI is not a silver bullet. If we rush to deploy algorithms trained on flawed or incomplete data, we risk amplifying existing inequities. If we let black-box AI decide who receives access to curative treatments like CAR-T, we risk losing public trust.

Instead, we must prioritize:

• Data integrity: Clean, complete, and interoperable data

• Data ethics: Consent, privacy, and transparency in how patient data is used

• Data diversity: Ensuring datasets represent the full spectrum of patients

• Collaborative governance: Involving patients, clinicians, and regulators in decision-making

CAR-T as a Teachable Moment

CAR-T is not just a miracle therapy. It is a case study in the future of healthcare: one that is high-tech, high-cost, and high-stakes. It shows us that precision medicine is only as precise as the data that fuels it.

To deliver on the promise of these therapies while protecting public health systems and ensuring equitable access, we must put data first.

Not because AI isn't transformative. But because trust is.

Published by Health Data Forum | July 2025

Image credits: Building blocks for institutional preparation of CTL019, Jul 2017 by Joseph Patrick McGuirk, Edmund K Waller, Muna Qayed, G. Douglas Myers.