Specialists of the Local Ethics Committee at the N.N. Petrov Oncology Research Center have introduced software based on the Yandex neural network (YandexGPT) for processing clinical trial documents, TASS reported. The solution provides for automatic analysis of hundreds of pages, checking compliance with the requirements, finding errors and suggesting ways for their correction. The initial verification of the data used to take several weeks, but now it has been reduced to several minutes.
The task of the Local Ethics Committee, which processes up to 1,000 submissions per year, is to verify the ethics of each clinical trial, that is, the risk/benefit ratio and respect for the rights of the trial participants.
“One clinical trial application contains hundreds of pages with graphs, tables, procedure descriptions, scientific justifications, and the results of previous trials. All these must strictly comply with both Russian and international requirements for safety, ethics, and study design. It is not an easy task to sort through such an array of data manually,” explained Artem Poltoratsky, the head of the Department of Preclinical and Clinical Trials at the N.N. Petrov Oncology Research Center.
According to him, the full approval cycle has been reduced from several months to 5-10 business days. The software enabled doctors to not only accelerate the process but also enhance its precision, which is vital for maintaining ethical standards and ensuring patient safety. Now it is planned to implement the technology in cancer centers of the BRICS countries and other medical institutions involved in clinical trials. This will accelerate the development of new drugs and treatment methods.
The software was developed by Yandex Cloud in collaboration with the Petrov National Research Medical Center of Oncology and Raft company. The solution is a web service on the Yandex Cloud platform, where documents are uploaded. The YandexGPT 5 Pro neural network analyzes hundreds of pages of protocols, instructions, and scientific justifications, classifies them by type, verifies compliance with the requirements, and identifies errors. In 95% of cases, the initial verification of documents is performed without human participation.
Earlier it was reported that a proposal had been made to simplify patient participation in clinical trials of drugs. A bill was submitted to the State Duma for consideration, which suggests removing the existing key restrictions.