Researchers Can Now Better Predict COVID-19 Severity
New discovery may help those living with SARS-CoV-2 who often fare the worst, including people with untreated HIV
By Jeffrey Laurence, M.D.
A remarkable feature of infection with the SARS-CoV-2 virus is its highly variable outcome, ranging from mild symptoms with no persistent effects to severe disease requiring hospitalization and, in some cases, intubation, dialysis, and development of prolonged symptoms known as long COVID. Non-vaccinated and immune-suppressed individuals, including those with untreated HIV, may be particularly prone to advanced COVID-19. However, most of the biologic causes of these divergent outcomes in humans remain unknown.
Some of the pathways influencing outcomes of SARS-CoV-2 infection have been identified, but the focus has tended to be on a relatively small number of immune-related pathways. This study sought to assess thousands of potential biomarkers associated with COVID-19 severity by measuring blood proteins. Machine learning methods were then used to develop a model predictive of COVID-19 outcome. 4,701 unique proteins were examined in 986 individuals, with blood samples collected within two weeks of onset of COVID-19 symptoms.
Participants were recruited from two cities, Montreal and New York City. This permitted examination of an initial dataset to “train” the machine to predict outcomes, and then confirmation of any findings from those patients in a “test” group. 92 proteins out of the 4,701 studied yielded a predictive value for COVID-19 severity of 89%, and over half of these proteins were unrelated to the immune system. This level of prediction was not improved by adding well-known clinical features linked to COVID-19 outcome early on in the pandemic, such as age and sex.
The authors conclude that proteins measured in blood at early stages of COVID-19 are “reasonably accurate” predictors of ultimate disease severity. They may also be relevant to discovery of new drugs to treat the disorder. The authors also emphasize that these proteins must be assessed in additional groups of patients. This should include those with HIV and the immune-suppressed.
amfAR was a funder of this research. Authors of this paper include amfAR grantee Daniel Kaufmann, M.D.
Dr. Laurence is amfAR’s senior scientific consultant.