FDA taps Cerner AI for Sentinel drug safety initiative

FDA taps Cerner AI for Sentinel drug safety initiative

Oracle’s Cerner Enviza unit has been recruited by the FDA to develop artificial intelligence tools that will be used to glean information from patients’ electronic health records (EHRs), to help the regulator monitor the effects of medicines at the population scale.

The project – which will also call on the expertise of AI and natural language processing (NLP) company John Snow Labs – will be part of the FDA’s Sentinel initiative, a national electronic database first launched in 2008 that is used to monitor the safety of drugs, vaccines, biologics, and medical devices.

Cerner Enviza is the R&D data management unit of Cerner – one of the top two sellers of EHR software in the US, alongside Epic Systems – which was acquired by tech giant Oracle in a $28 billion deal that completed last year.

A new two-year project between the regulator and the two companies will focus on Merck & Co’s asthma drug Singulair (montelukast) and generics, and probe a possible link between the drug and mental health side effects.

Use of the decades-old drug has been associated with changes in behaviour, as well as mood related changes, such as aggression, agitation, irritability, confusion, anxiety and depression, obsessive-compulsive symptoms, and suicidal thinking. Since 2020, it has carried a boxed warning on its label in the US about neuropsychiatric side effects.

Dubbed MOSAIC-NLP, the project will use NLP technologies to analyse unstructured EHR data, in the hope of showing that it is feasible to draw usable insights that can be used to support pharmacoepidemiology studies and direct public health initiatives.

Cerner Enviza is taking the lead on the project, with participation from researchers at Mass General Brigham and Harvard Pilgrim Health Care Institute, which run the FDA’s Sentinel Innovation Centre.

The partners note that traditional manual methods for analysing clinician notes often can be a bottleneck for fully understanding the symptoms and outcomes that patients experience at the population level, which could be overcome through the use of AI and NLP.

The study will use “de-identified EHR data to help transform unstructured clinical notes into validated and usable data for physicians and researchers,” commented Mike Kelly, global head of Cerner Enviza.

“Connected technologies and unified data can accelerate innovation and, in turn, help providers realise better recommendations and outcomes for their patients,” he added.

The use of AI and NLP technologies to spot adverse events is one of the top trends for the life sciences industry this year, highlighted in a recent IQVIA report.