By Deborah Borfitz 

February 23, 2022 | Researchers from Massachusetts General Hospital (MGH) and the University of Cyprus (Cyprus) have come up with a mathematical model that can predict treatment response for different subpopulations of patients with COVID-19 complicated by acute respiratory distress syndrome (ARDS) and determine biologic mechanisms of patient outcome. The treatment efficacy scoring system indicates “how each treatment is moving the patient closer to baseline, or initial state,” according to Lance Munn, Ph.D., deputy director of the E.L. Steele Laboratories for Tumor Biology at MGH and an associate professor of radiation oncology at Harvard Medical School (HMS).

Unlike machine learning approaches that look for patterns in data, the model was built from “the bottom up” based on simulations of the biological mechanisms of the SARS-CoV-2 virus, including the damage it does to different tissues and how it causes the blood to coagulate, Munn says. Disease course was modeled in six distinct patient types—young (healthy), diabetic (but not hyperinflamed), older, hyperinflamed, hypertensive, and obese—based on known risk factors for severe COVID-19.

For each of those patient populations, researchers then predicted how they would fare relative to one another if given either the broad-spectrum immunosuppressor dexamethasone (a synthetic corticosteroid) or targeted treatment with anti-IL6 or anti-IL6R anti-inflammatory therapies, all of which are currently in widespread use. The scoring system comprises 130 potential outcomes, including oxygen saturation, number of CD8 T cells that mediate viral clearance, and amount of pro-inflammatory cytokines in the bloodstream, says Munn.

The analysis indicates that the three treatments are not as useful for patients with diabetes or obesity but are among older and hyperinflamed patients. For these patients, the therapy type and timing of initiation determine treatment efficacy, Munn says. Conversely, patients in the hypertensive group benefit from all the treatments regardless of when they are started.

Response scores are determined by changes from baseline measures three weeks after treatment. “If it takes them all the way back to baseline on all 130 parameters, then they are completely cured, and they don’t have any disease left,” Munn says, “so this gives us a quantitative way to look at how well the treatments are doing.”

A closer look at the modeled outcomes revealed that the number of neutrophils is a key driver of treatment response in older patients, Munn continues. Among hyperinflamed patients, the amount of angiotensin in the bloodstream is an important marker, while anti-inflammatory cytokines (IL-6, IL1β) and micro thrombus in liver and cardiac vessels are the more relevant feature for patients who are obese, diabetic, and hypertensive.

The findings, which published recently in eBioMedicine (DOI: 10.1016/j.ebiom.2021.103809), were derived from an analysis of data on 15,000 patients tested for COVID-19 at MGH. The study included a clinical analysis of two of the proposed biomarkers, IL-6 and D-dimer (a coagulation measure), among roughly 5,300 patients that were predictive of outcomes for each of four phenotypes (older, obese, hypertensive, and diabetic). 

Interest in ARDS in the scientific and medical community has been heightened because of COVID-19, but the life-threatening lung injury is a longstanding topic of investigation, notes Rakesh K. Jain, Ph.D., director of the Steele Labs and professor of radiation oncology at HMS. It is associated with a variety of diseases—including other coronaviruses such as severe acute respiratory syndrome and Middle East respiratory syndrome—and is a common cause of mortality and morbidity worldwide. 

“To predict disease progression and personalize treatment [for ARDS], it is necessary to determine the associations among clinical features, biomarkers and underlying biology,” says Jain. “Although this can be achieved over the course of numerous clinical trials, the process is time-consuming and extremely expensive.” 

Precision Medicine 

The six patient types were chosen based on emerging reports about which comorbidities were linked to more severe COVID-19 disease. But one of the initial goals of the mathematical model, pre-pandemic, was to look at the renin-angiotensin system that controls hypertension, says Munn. 

Motivation for the project came from decades of work advancing precision medicine for cancer, according to Jain. One anti-hypertensive drug that appears to be beneficial in treating cancer regulates the same pathway used by the SARS-CoV-2 virus.

Jain joined hands with his colleagues (Triantafyllos Stylianopoulos and Chrysovalantis Voutouri) at the University of Cyprus and approached folks at MGH for access to the human data. As shared by Charles Corey Hardin, M.D., Ph.D., who takes care of COVID-19 patients at MGH, the clinical challenge in treating patients and conducting clinical trials is knowing which drug to give to which patients.

“If you do a trial by mixing apples and oranges and bananas you are not going to get the right answer,” Jain says. So, the newly assembled research team went about grouping patients based on their biological similarities. 

Endpoints of patient treatment benefit have variably included oxygen saturation level, D-dimer level, elevated concentrations of IL-6, and neutrophil count per the observed association of those variables with disease severity. Although clinicians as a matter of practicality must limit the number of tests that they can order to assess patients’ status, oxygen saturation is the major progress marker because it is easy to measure on a continuous basis, says Munn.

Other measures require blood tests involving more cost and inconvenience. “The power of the model is that it can suggest things that clinicians might do to improve their personalized treatment,” Munn says, such as monitor neutrophil elastase myeloperoxidase (made by neutrophils).

Whether ARDS caused by COVID-19 is different from ARDS associated with other conditions has been a matter of controversy over the past two years, says Munn. “But the consensus is that COVID ARDS is similar to non-COVID ARDS, but with more of a coagulopathy component… and our model predicts that is the case. The virus is affecting not just the lung but other tissues and that causes damage systemically [leading to thrombosis].” 

Next Up: Variants, Antiviral Drugs 

The big revelation of the study for Jain was the importance of treatment timing to patient benefit. “That’s what came out of the model and, fortunately, has been validated by [clinical] trials.” 

In another study published early last year in PNAS (DOI: 10.1073/pnas.2021642118), for example, the MGH and University of Cyprus research team discovered that dexamethasone should be initiated on day seven of a patient’s admission to the intensive care unit for optimal results. But other drugs—including heparin and anti-IL6 and anti-IL6R—are better administered on day three.

For Munn, the “pleasantly surprising” finding was that the team succeeded in creating a mathematical model that boils down many biological variables to the critical few that are backed by clinical trials. The outcomes of studies match what they had predicted. 

The model is now being extended to COVID-19 vaccines, specifically who needs a second booster shot, says Jain. Israel, Denmark, and Chile are already asking their citizens—especially those 65 and older—to get a fourth shot. 

At the same time, researchers are using the model to help discern the next variants and kinds of mutations that will be most worrisome, adds Munn. “Is it replication? Is it tighter binding to the receptor? Is it evasion of the antibodies? We can use the model to answer all of those questions.” 

Data on the Omicron variant and associated mutations are in limited supply, Munn says, but “the model lets us play with that at will... and then wait for people to [hopefully] show we were right with clinical studies.” More immediately, adds Jain, the model is being applied to antiviral drugs and monoclonal antibodies that have been recently approved (formally or conditionally) by the U.S. Food and Drug Administration to help guide clinical decision-making regarding who gets which drug and at what time and dose.