Medical products require human trials.

Most of us are familiar with and support the idea that before a drug or medical device can be prescribed or sold, it needs approval from a national watchdog and regulatory entity. In the US that’s the FDA. And to get FDA approval for products that are classified as sufficiently risky, that means the manufacturer needs to provide proof the product is safe and effective. In a large majority of cases that proof is based on data derived from clinical trials. Sometimes first with animals, but always to include carefully constructed and observed trials with humans.

For sufficiently risky products those trials are expensive (think millions of dollars) and take a long time (think 3 to 12 months). And there is never just one such trial. There will be a whole series of human trials. Collecting performance data to get convinced that the product is worth producing and seek FDA approval means upfront investments. Often a Research and Development organization will need to be financed for years without any immediate returns.

Modelling the human biology, with the ever expanding AI technologies, should lead to being able to predict the outcome of interventions, drug delivery, control systems and more. Over time this modeling can replace most of what animal and human trials are needed for.

High costs of drugs and medical devices

Not surprisingly, companies want to recoup this investment through the price they charge when (or if) the product hits the shelves. The high prices of drugs and medical products can thus in part be attributed to this desire for a high ROI and in part to the cost of the animal and human trials and the associated years-long R&D efforts.

AlphaFold

One of the breakthroughs in the rapid evolving landscape of AI is demonstrated by AlphaFold. AlphaFold is an AI system developed by Google DeepMind that predicts a protein’s 3D structure from its amino acid sequence. It regularly achieves accuracy competitive with conventional experiments. AlphaFold has immediate potential to help us understand the structure of proteins and advance biological research. It is also substantial faster than purely human driven research. As it turns out, its focus on protein folding is very relevant to drug discovery.

Modelling the human biology

In 2023 the FDA published ‘Technical Considerations of Medical Devices with Physiologic Closed-Loop Control Technology guidance’. It’s an example of the growing notion that risk management, in this case risks associated with closed loop drug delivery systems, can in part be expressed through a classification of feedback (reactive) and feedforward (predictive) biomarkers, and control systems theory. This underlying theory can be developed into a systematic outline for rigorous device development, testing, and implementation. Part of that can replace (early) animal or human trials.

Modelling the human biology, with the ever expanding AI technologies, should lead to being able to predict the outcome of interventions, drug delivery, control systems and more, to such an extent that they can replace most human trials.

Also in 2023 (and not a coincidence), Lane Desborough (previously chief engineer at Medtronic Diabetes and Bigfoot Biomedical) gave a presentation during which he argued:

• Model and simulate your closed-loop control system: Gain valuable insights into system behavior before implementation, enabling early identification and mitigation of potential risks.
• Generate production-ready code: Automatically generate code from your model, ensuring consistency and reducing the potential for errors.
• Perform comprehensive verification and validation: Utilize powerful testing tools to ensure your system meets all necessary safety and performance requirements.
• Streamline documentation and regulatory compliance: Generate comprehensive documentation automatically, facilitating regulatory submissions and audits.

By now these activities could be substantially taken on by AI driven agents.

Faster innovation cycles and lower costs

Let’s combine the example of AlphaFold and the progress made by researchers and the FDA in recognizing modelling as a reliable methodology, for safety and effectiveness assessment.

There ought to be a focus on modeling, simulation, and the control of physiological systems for medical product development through reliable and accurate AI models and agents. They will produce results faster and be cheaper than traditional trials. Over time this modeling can replace most of what animal and human trials are needed for. It should result in shorter innovation cycles and lower costs.