No Real-Life Placebo Required – Deriving Virtual Control Group Data by Machine Learning

Drug development is an arduous and time-consuming process requiring a huge investment of capital, manpower and equipment. The average research and development period lasts for 10 years, of which clinical trials take up about 7 years, let alone the fact that the amount of investment in prescription drugs can be as high as €2 billion ($2.21 billion). Furthermore, ethical issues that may be involved in clinical drug trials are gradually being taken seriously, and the rights of subjects in randomized controlled clinical trials should be adequately protected for the group receiving placebo.

Bayer AG from Germany has partnered with Finland’s Aalto University to make use of artificial intelligence in clinical trials to reduce or eliminate the need for a control group in some trials by creating a “virtual” control group from a medical database. This allows a larger proportion of patients to be treated with the study drug and a smaller proportion to be assigned to placebo or standard treatment.

Finland’s High Quality Health Data is Valued by Bayer

Bayer’s three-year Future Clinical Trials (FCT) project, launched in June 2021 in collaboration with Aalto University and Helsinki University Hospital. The project is focused on making drug trials more accessible to patients. In addition to remote monitoring of patients to reduce the number of visits, the application of artificial intelligence to improve the safety and efficiency of clinical drug studies is being evaluated. 

Professor Harri Lähdesmäki of Aalto University and the Finnish Center for Artificial Intelligence (FCAI) said that the formation of virtual control groups using AI-based medical databases would eliminate the need to recruit patients to control groups and also improve the cost efficiency of drug development.

Jussi Leinonen, a data scientist at Bayer, noted that the goal is to enable AI-based algorithms to identify rare messages related to the side effects of the drug, while improving sensitivity. The possibility of reducing the number of patients needed for clinical trials while ensuring the safety and reliability of the trials is still under investigation.

Independent Decision-Making Still Not Possible with AI

Clinical studies start with openness, in which all the collaborative approaches of all parties involved and process monitoring are open to review and validation. These features are equally important to AI researchers, with an emphasis on openness of data and algorithms. One of the current areas of research at FCAI is the development of artificial intelligence methods to search for suitable molecules for drug ingredients. Harri Lähdesmäki explained that researchers are trying to find alternative solutions to the previous need to experiment with millions of different drug molecules, replacing them with machine learning approaches. This will allow drug development to be done faster and with fewer resources by suggesting that certain molecules are worth exploring in greater depth in a laboratory setting.

Jussi Leinonen also pointed out that AI will not completely replace humans in clinical trials. In the near future, AI itself will not make independent decisions, but will identify situations that require a response or direct that response in a certain direction, which also applies to the AI technology used by physicians when treating patients.

Written and translated by Richard Chau

Author

GeneOnline