Dose of Medication

Finding the right dose of medication by trial and error can be a painful experience for patients who are already suffering because of their disease. Artificial intelligence can greatly accelerate this process thanks to a new algorithm created by Adrien Coulet; in collaboration with researchers from Stanford University. By analyzing patients’ digital data; this innovative tool can predict in advance whether patients; will need a lower dose of medication thus reducing the suffering caused by side effects.

Using artificial intelligence 

Using electronic patients’ data from Stanford University Hospital; the team of researchers was able to identify those who required a lower dose than the standard one for certain drugs; even before treatment started. They used an automatic learning method called “Random Forest Classifier”; this algorithm combines the results of decision trees based on slightly different subsets of data.
So the innovative side of this project lies; in the preparation and selection of the data used. Scientists compiled information from very different sources; on groups of patients who needed to have their dosage of drugs changed or not.
Patients’ data from Stanford University Hospitals been computerized for years so the algorithm had multiple data on the highly variable responses of patients to their treatments. This computerization of the health system allowed the algorithm to learn and then predict whether patients who had never taken the drugs in question at the time would need a lower dosage.

Refined predictions about complex drugs

Adrien Coulet and his collaborators particularly interested in drugs that interact with the family of the P450 enzymes. Malfunctions of this enzyme family are linked to adverse reactions to several types of medications; including some anti-cancer drugs. These adverse reactions can be limited when the dosage of the anti-cancer drugs prescribed as part of chemotherapy are adjusted to each individual patient according to his or her needs.

Traditionally these dosages are very difficult to adjust because the activity of the P450 family of enzymes is highly dependent on the individual characteristics of the patients. Thus the effects felt by patients taking these drugs, whether positive or negative; can vary greatly depending on their genes, their personal history or their way of life.

The precision and completeness of the algorithm have made it possible; to take this multitude of possibilities into account to apply it to new patients and make accurate predictions. Researchers now wish to adopt this technology to the data and the scientific; social and legal environment of French hospitals. The computerization of the health system can provide valuable support to health personnel and avoid unnecessary inconvenience and suffering for patients.