Dependence of Diagnostic Accuracy of Algorithms for Covid-19 Progression on Sars-Cov-2 Variant Shift

The COVID-19 pandemic posed a major challenge to global healthcare systems, demanding rapid development of diagnostic and treatment protocols. Amid overwhelming hospital admissions, timely identification of high-risk patients was critical. Researchers worldwide developed risk stratification algorithms incorporating clinical, laboratory, demographic, and comorbidity data, along with key immunological markers like cytokine levels. This study included 289 patients with confirmed COVID-19 (U07.2). Multiplex analysis assessed a broad panel of immune markers: interleukins (IL-1α, IL-6, IL-10, IL-17A), interferons (IFNα2, IFNγ), tumor necrosis factors (TNFα, TNFβ), chemokines (CCL2/MCP-1, CXCL10/IP-10), and growth factors (G-CSF, VEGF-A). Participants were infected with distinct SARS-CoV-2 variants: original Wuhan strain (n=56), Alpha (B.1.1.7, n=95), Delta (B.1.617.2, n=78), and Omicron (B.1.1.529, n=57). Two diagnostic algorithms were tested: disease severity classification (based on single-timepoint IL-6/IL-8 levels). Chi-square tests revealed significant accuracy variations across variants (p=0.0065), with predictive power declining as viral mutations accumulated; and outcome prediction (using combined IL-18/IL-6 levels). Again, accuracy differed significantly by variant (p<0.0001). Despite extensive data, cytokine-based prognostic models face limitations. Viral evolution drives unique immune responses, reducing the reliability of existing algorithms. Thus, continuous refinement is essential to address SARS-CoV-2’s variability.