Early Immune Biomarkers Guide Severe COVID-19 Care

As COVID-19 management evolves, identifying which patients are most likely to benefit from advanced therapies beyond standard treatment remains a persistent challenge. In a post-hoc analysis of a randomized Phase 1/2 clinical trial, investigators explored whether early immunologic changes could be used to predict outcomes and guide treatment selection for patients with severe COVID-19 receiving either standard of care (SoC) alone or SoC combined with adoptive SARS-CoV-2–specific T-cell (CoV-2-ST) therapy.

The study analyzed data from 87 hospitalized adults with severe COVID-19 during the Delta variant era. Patients were randomized in a 2:1 ratio to receive CoV-2-ST therapy +  SoC (n=57) or SoC alone (n=30). Eligible patients had severe disease, were within the first 6 days of symptom onset, and had evidence of systemic inflammation. The original trial demonstrated improved outcomes with adoptive T-cell therapy, and the current analysis sought to determine whether early biomarker patterns could identify patients most likely to benefit from this approach.

CoV-2-ST therapy involves the infusion of virus-specific T lymphocytes derived from vaccinated or convalescent donors. These cells are designed to restore antiviral cellular immunity, a critical component of host defense that may be impaired in patients with severe COVID-19. To quantify these responses, investigators measured circulating CoV-2-STs using interferon-gamma (IFN-γ) enzyme-linked immunospot (ELISpot) assays, which detect antigen-specific T-cell activity after viral stimulation.

Immune Recovery Signals Emerge Within Days

While baseline biomarker profiles were largely similar between treatment groups, significant differences emerged by day 5. Patients receiving CoV-2-ST therapy demonstrated greater increases in CD3+ T cells (P=.04), CD8+ T cells (P=.01), natural killer (CD56+) cells (P<.001), and circulating CoV-2-STs (P=.02) compared with patients receiving SoC alone. These changes suggested a more robust restoration of both adaptive and innate immune function following adoptive cellular therapy.

Clinical outcomes also favored the CoV-2-ST group. By day 60, 64.9% of patients treated with CoV-2-STs + SoC were alive compared with 40.0% of patients receiving SoC alone, corresponding to a crude odds ratio for recovery of 2.8.

Machine Learning Identifies Predictive Biomarker Patterns

Among patients receiving CoV-2-ST therapy, shrinkage-LDA model performance was strong, with area-under-the-receiver-operating-characteristic-curve (AUROC) values ranging from approximately 0.86 to 0.88 and area-under-the-precision-recall-curve (AUPRC) values of 0.74–0.78. Sensitivity ranged from 0.89 to 0.91, while specificity ranged from 0.83 to 0.87. In the SoC-only cohort, predictive performance was more modest, with AUROC values between 0.72 and 0.76, although sensitivity remained high at approximately 0.95.

Several biomarkers emerged as key predictors of outcome. In patients receiving CoV-2-ST therapy, important variables included Karnofsky performance status, CD3+, CD4+, CD8+, and CD56+ counts, circulating CoV-2-ST levels, IL-6, C-reactive protein, ferritin, lactate dehydrogenase, and white blood cell counts. In contrast, only a smaller set of markers—circulating CoV-2-STs, CD3+, CD56+, and C-reactive protein—were strongly associated with outcomes in the SoC group.

Advancing Precision Treatment Strategies

The investigators also conducted Monte Carlo simulation analyses to model hypothetical treatment-switch scenarios. These simulations suggested that as many as 30% of patients initially treated with SoC alone might benefit from CoV-2-ST therapy, while misclassifying appropriate candidates and treating them with SoC alone could increase critical outcomes by up to approximately 22%. Although these analyses were exploratory and not designed to establish causality, they illustrate how early immune biomarkers could potentially support individualized treatment decisions.

Taken together, the findings highlight the potential value of integrating dynamic immune monitoring with interpretable machine-learning models to guide precision treatment strategies in severe COVID-19. While external validation in larger and more contemporary patient populations remains necessary, the study offers an intriguing framework for identifying high-risk patients early and matching them to therapies most likely to improve outcomes.

Reference

Savvopoulos S, Papadopoulou A, Karavalakis G, et al. Early Immunological Biomarkers for Personalized Treatment Selection in Severe COVID-19: Post Hoc Machine Learning Analysis of a Randomized Clinical Trial. JMIR Med Inform. 2026;14:e78471. doi:10.2196/78471