In Silico Clinical Trials
Medical school teaches us there are in vitro and in vivo clinical trials. But the trend is going towards in silico - trials without human patients.
Artificial intelligence experienced a breakthrough in clinical trials in September 2020, when the new guidelines were introduced. But there’s another trend and a way of medical research that’s emerging and getting more and more accepted by regulators.
In silico is still very much in its early days, but there are two different types. The first one is a hybrid model called "Organ Chips" or (organs on chips) developed by The Wyss Institute for Biologically Inspired Engineering of Harvard Medical School. The base is made of polymer, but it contains living human cells inside it. This is how they describe it on their website:
Each Organ Chip is composed of a clear flexible polymer about the size of a computer memory stick that contains hollow microfluidic channels lined by living human organ-specific cells interfaced with a human endothelial cell-lined artificial vasculature, and mechanical forces can be applied to mimic the physical microenvironment of living organs, including breathing motions in lung and peristalsis-like deformations in the intestine. They are essentially living, three-dimensional cross-sections of major functional units of whole living organs. Because they are translucent, they provide a window into the inner workings of human cells in living tissues within an organ-relevant context.
In other words, we can say this is the replacement for animals in drug trials. In essence, it’s an upgraded in vitro model as it can be the host of many human body environments and also a precise simulation environment for different purposes. This for example includes smoking, according to people’s habits.
It gets even more holistic as many of these “organs” can be interconnected to simulate the human body in greater details and research the pharmacokinetic and pharmacodynamic responses to drugs.
The only problem could be that it’s still too reductionist. The following solution does an even better job, at the possible cost of validity - precise computational models of the human body, where scientists could run drug and clinical trials in the future.
Novadiscovery builds “mathematical models of disease and treatment based on real-world data and literature”. They try to improve three key steps in drug design that would make it easier for the developers to set the right target for the drug, optimise clinical development and have easier access to the market as it’s easier to demonstrate value to the payers.
Just as it says on their website, in silico models allow for a greater range of experimentation and no consequences whatsoever for the “participants”. And if you think about it, a very detailed in silico human model is better than animals, as there’s zero to no discrepancies in physiology.
Advanced models would also give precision medicine a boost since it would allow for two things. Firstly, drugs could be developed and applied to the individual. Secondly, it could give researchers answers to why specific drugs work for some people and not for others.
It’s almost not worthy of writing that it would also speed-up drug development and potentially make them cheaper across the board - i.e. more accessible.
There are many reservations, though. Firstly, in silico is challenging because it lacks validity, and it’s questionable when it will. As mentioned and written about, artificial intelligence now has standards in place for better and safer trial design. In silico not so much. The basis for this is ASME V&V 40-2018 or “Assessing Credibility of Computational Modeling through Verification and Validation: Application to Medical Devices”, but is primarily intended for medical devices.
Secondly, healthcare and patients would have to comply with the fact that some drugs simply would not be tested on humans, but used by them. This is almost absurd (though theoretically possible) by itself. The patients would be faced with a great deal of uncertainty, which is again tied back to the previous point about validity.
But it has to start somewhere.