U.K. to use AI tool to process 'high volume' of expected adverse reactions from COVID vaccine

The British government is funding the development of an artificial intelligence tool to track and log what it anticipates will be a "high volume" of adverse reactions to the upcoming COVID-19 vaccine once it becomes widely distributed. 

A "contract award notice" posted to the European Union public procurement tracker Tenders Electronic Daily states that the U.K.'s Medicines and Healthcare products Regulatory Agency plans to deploy "an Artificial Intelligence (AI) software tool" to "process the expected high volume of Covid-19 vaccine Adverse Drug Reaction (ADRs) and ensure that no details from the ADRs' reaction text are missed."

"It is not possible to retrofit the MHRA's legacy systems to handle the volume of ADRs that will be generated by a Covid-19 vaccine," the contract notice continues. "Therefore, if the MHRA does not implement the AI tool, it will be unable to process these ADRs effectively.

"This will hinder [the MHRA's] ability to rapidly identify any potential safety issues with the Covid-19 vaccine and represents a direct threat to patient life and public health."

The contract, which is worth $2 million, was awarded in September to Genpact (UK) Ltd. The posted announcement states that "reasons of extreme urgency" related to the pandemic have "accelerated the sourcing and implementation of a vaccine specific AI tool."

COVID vaccine safety expected to be 'similar to other types of vaccines'

A spokeswoman for the MHRA confirmed the agency's intended use of artificial intelligence to document the negative health effects associated with a COVID vaccine. 

"We have a range of resources and technology to support the safety monitoring of any COVID-19 vaccination programme," an agency representative said via email. "The use of AI will be one element of that. We take every report of a suspected side effect seriously, and we combine the review of these individual reports with statistical analysis of clinical records."

"The purpose of the AI tool we are introducing as part of our Yellow Card system is to help us rapidly evaluate such reports after approval, and not as part of the approval process," she continued. "It will help us by reducing the amount of manual coding for each report, thereby saving resource in processing cases and ensuring they are rapidly available for scientific analysis."

The agency did not directly answer when asked what kinds of adverse reactions health authorities are expecting to accompany the eventual COVID vaccine.

"Our past experience with other new immunisation campaigns," the spokeswoman said, "is that we tend to receive around 1 Yellow Card report per 1,000 doses administered and we are preparing our surveillance systems on that basis. Most of these are mild and short-term, and not everyone gets them."

"Based on the available published reports from the Phase One and Two clinical trials," she said, "we don’t currently anticipate any specific safety concerns with COVID-19 vaccines. We expect the general safety profile to be similar to other types of vaccines."

Multiple scientists and researchers working on COVID-19 vaccines in the United Kingdom did not respond to queries seeking more information about the frequency and severity of adverse reactions to the drugs they are developing. 

But Deborah Fuller, a microbiology professor at the University of Washington School of Medicine and a division chief of infectious diseases and translational medicine at Washington National Primate Research Center, suggested that the MHRA in utilizing the AI tool may be seeking to "address the ... unknowable."

Fuller drew a distinction between "adverse reactions" — what she described as "severe life-threatening situations requiring hospitalization" — and the less dire instances of "vaccine reactogenicity," which she described as comparatively mild symptoms such as low-grade fevers and soreness at the vaccine injection site.

Noting the MHRA's explicit use of "adverse drug reactions," Fuller said: "That suggests to me they are, indeed, referring to adverse (life threatening) situations." 

"We won't know what the rate of adverse reactions will be with these vaccines until we start mass vaccinations," Fuller noted. "Will it be 1 in a million or more like 1 in 100,000? We just don't know yet. That's because phase 3 clinical trials are structured to enroll 30-40,000 subjects. As such, they will pick up adverse reactions if they occur at a rate of about 1:10,000."

Artificial intelligence to survey population-scale reactions to vaccines "would be very valuable," Fuller said. 

"As the world gears up to vaccinate billions of people in a relatively short time," she said, "how are we going to monitor such a high volume of vaccinated people for rare adverse events? To me, that's what it sounds like they're really after — not necessarily to detect a high volume of adverse events but rather, providing the capability to screen a high volume of vaccinated people for rare adverse events."

In ongoing COVID-19 vaccine trials, severe events do seem to be rare. In July, the American drug company Moderna reported that "more than half" of participants in a COVID-19 vaccine trial reported adverse events, but most of those reactions were for symptoms such as "fatigue, chills, headache, myalgia, and pain at the injection site." Some participants experienced severe chills, headache and muscle pain.

AstraZeneca's COVID-19 vaccine trial in the U.K., on the other hand, was briefly halted in September after a patient who received the drug reportedly developed transverse myelitis, a potentially serious inflammation of the spine that can result in paralysis. The trial was eventually allowed to resume.