BOULDER, Colo.—Diagnostic solutions company Biodesix Inc. reported the publication of a paper—“Predicting Prognosis in COVID-19 Patients using Machine Learning and Readily Available Clinical Data”—in medRxiv regarding a new artificial intelligence (AI)-based COVID-19 algorithm and data from a related study. The algorithm was developed using Biodesix’s Diagnostic Cortex AI platform in close collaboration with a leading academic center, and as noted in the paper, it is capable of rapidly and accurately predicting the risk of severe outcomes for patients with COVID-19 infections by using patient data collected upon admission to the hospital. This AI-based algorithm can determine which patients are likely to require intervention (such as admission to intensive care units or being put on ventilators) or develop acute respiratory distress syndrome (ARDS), and which are not.
“We are extremely excited about this algorithm based on our proprietary AI platform and its ability to support treatment of patients with COVID-19,” said Scott Hutton, CEO of Biodesix. “Implementing this during the initial patient assessment can offer physicians the necessary insight to quickly and confidently choose the appropriate treatment for each patient. This ability is especially valuable now, during the pandemic, when timely treatment decisions are crucial and critical resources such as ICU beds and ventilators can be scarce.”
The algorithm was evaluated in a study of 559 COVID-19 patients, 229 of whom were hospitalized with COVID-19 and were used in the development of the algorithm, and 330 of whom were a blinded, independent validation. By applying 26 variables—such as patient characteristics, vital signs, and test values from readily available laboratory tests—the study demonstrated that Biodesix’s algorithm accurately determined which patients would likely need intervention and which would not.
“The ability to rapidly determine a patient’s risk of severe COVID disease—early, easily, and accurately—can help physicians to identify those who could benefit from specialized treatment and early interventions and get them the care they require in the optimal treatment window,” stated Dr. Robert Georgantas III, senior vice president of Research and Translational Science at Biodesix. “At the same time, the AI-based algorithm may help to avoid unnecessary treatment for patients deemed to be at lower risk, for whom heroic interventions may cause harm.”
In moving forward with this algorithm, Biodesix is working with a U.S.-based COVID-19 consortia and another international COVID-19 consortia to conduct additional blinded, independent validation testing, as well as evaluating options for making the algorithm widely available for hospitals.
In other recent news for the company, Biodesix inked an agreement in January with biotechnology company HiberCell to further develop an enzyme-linked immunosorbent assays (ELISA) as a companion diagnostic for registrational trials in breast cancer for Imprime PGG programs. The ELISA test will be designed to test for Anti-β Glucan IgG Antibody (IgG ABA) expression in breast and melanoma serum samples to assess cancer patients' eligibility for Imprime PGG therapy.
As described on HiberCell's website, Imprime PGG is “a systematically administered innate immune activator,” one that the company says “has the potential to significantly increase the number of patients who benefit from combinatorial approaches with immuno-oncology therapies such as checkpoint inhibitors. Imprime PGG primes the innate immune system and reprograms the immunosuppressive tumor microenvironment to drive enhanced antigen presentation, triggering T cell activation and ultimately enhancing the immune response against tumors.”
"Biodesix is excited to collaborate with HiberCell across their promising portfolio," Hutton remarked. "Our leadership in clinical proteomics along with our comprehensive approach to diagnostic discovery, development and commercialization will help streamline and accelerate HiberCell's ability to bring critical therapies to patients. Working together on a companion diagnostic for the Imprime PGG program is only touching the surface of what we can do together. This approach to identifying patients for this drug is very novel and holds tremendous promise."