BLOG: Interview with VP of Data Science Research on COVID-19 Studies with Humanigen

BioSymetrics partnered with Humanigen in 2021 to investigate whether subsets of patients in the company’s Phase 3 study of lenzilumab for severe COVID-19 responded differently to the drug. Below is an interview with BioSymetrics’ VP of Data Science Research, Victoria Catterson, PhD, about the work and subsequent publications.

Question: Findings from Humanigen's Phase 3 study of lenzilumab in patients with severe COVID-19 were recently published in the Lancet Respiratory Medicine and more papers are pending in other peer-reviewed journals. BioSymetrics partnered with Humanigen in support of these studies. Can you summarize what BioSymetrics did as part of this collaboration?  

Victoria: The successful Phase 3 clinical trial showed higher likelihood of survival without mechanical ventilation (SWOV) for patients treated with lenzilumab. SWOV is an endpoint which captures not only survival, but also quality of life of the surviving patients, and is therefore a strong marker of effectiveness. We applied our Clinical Insights workflow, from our phenotypic-driven drug discovery platform, to investigate whether there were some subsets of patients in the trial cohort who responded particularly well to lenzilumab. This type of patient stratification can unlock precision medicine, where therapeutics are chosen for an individual patient based on attributes of their specific case, rather than drugs being used for a disease cohort.
 
Question: Some of the BioSymetrics analysis prompted Humanigen to do further study on the dataset. Can you describe what that was, and how it impacted further study design and outcomes? 

Victoria: Our analysis found that when controlling for the effect of multiple variables at the same time, the strongest predictor of SWOV was not patient severity, age, or even treatment, but the patient’s initial level of C-Reactive Protein (CRP). CRP is a biomarker indicating inflammation in the body, and normal, healthy levels are under 10 mg/L. The inflammation caused by COVID-19 can drive these levels very high: the median value in the trial population was 79 mg/L, and 25% of the cohort had values over 137 mg/L. While no one in this patient population had a “healthy” level of CRP, our findings showed that those with initially lower CRP were more likely to survive. 

This prompted follow-up investigation focusing specifically on the interaction of lenzilumab and CRP. When stratifying patients according to CRP, those in the lowest quartile (<47 mg/L) showed the highest benefit of lenzilumab, with over 8 times the chance of SWOV if given lenzilumab versus placebo. At the other end of the CRP scale, those in the highest quartile (>137 mg/L) showed the most modest benefit from lenzilumab, with only a 17% increased chance of SWOV over those on placebo. 

The main take-away is that lenzilumab demonstrated efficacy across the whole cohort, but we uncovered the particularly strong effect in those with lower CRP in this cohort. Future trials can be structured to confirm the effect seen here, pointing the way to the use of CRP as a biomarker to personalize the use of lenzilumab. 

Question: What would you hope other pharma companies might take away from the work conducted with Humanigen, and how that might be applied to more studies in COVID-19?  

Victoria: At BioSymetrics, our mission is to use phenomics-driven drug discovery to translate human disease biology and advance precision medicines to ultimately improve people’s lives. Many diseases that were previously thought to be monolithic conditions have been refined into subtypes, such as when different genes within the same pathway trigger similar symptoms. We believe that the future is precision medicine, where we take account of the patient’s specific history, symptoms, demographics, genomics, and any other relevant factors in order to determine a treatment plan tailored to them. 

If you have longitudinal electronic health records (EHRs), genomics, medical images, notes, or some combination of health data, we can use our platform to uncover the phenogroups, or clusters of patient attributes, within an overarching disease diagnosis. Further stages of the platform can then focus drug discovery on one or more of these phenogroups. We believe that phenomics-driven drug discovery leads to a more targeted therapeutic, with more specific patient criteria for trial recruitment, and ultimately a higher chance of success at clinical trial. 

Question: What other work is BioSymetrics doing currently in COVID-19?  

Victoria: We have a broad program of work in COVID-19, starting with patient stratification for Janssen in 2020 while they were determining criteria for their vaccine’s trial enrollment. Prior to the pandemic we were actively working on Acute Respiratory Distress Syndrome (ARDS) in partnership with Northwell Health (the largest hospital network in New York), which positioned us well to expand into COVID-ARDS as it became the highest cause of mortality in COVID-19 patients. We have recently performed phenogrouping on COVID-ARDS patients and compared them with non-COVID-ARDS, highlighting that some phenogroups are similar between the two diseases, but COVID-ARDS also presents differently in a large proportion of patients.  

Finally, we have explored the genomics of susceptibility to severe COVID through a partnership with Sunnybrook Hospital in Toronto, which led to a follow-on study linking the genomics and EHR phenogroups, to uncover biomarkers for precision use of COVID therapeutics. 

For more information on the Humanigen collaboration and results, you can read the Phase 3 study of lenzilumab in the Lancet Respiratory Medicine and an analysis of CRP levels from the study on medRxiv.