COVID-19 Cytokines Linked to Severity, Not 3-Month Symptoms

Key Takeaways

• Among U.S. Veterans in the EPIC3 cohort, serial cytokine patterns were associated with 30-day COVID-19 severity, with higher early HGF, IL-1RA, IL-18, IP-10, and VEGF-A tracking with more severe illness.
• Higher post-acute MIP-1α and VEGF-A were also associated with greater 30-day COVID-19 severity.
• Adjusted models did not identify single-timepoint cytokine associations with later dyspnea, fatigue, or cognitive impairment.
• Over time, mild disease and absence of later symptoms were associated with a more resolving cytokine pattern, including higher EGF, MIP-1β, and RANTES and lower MIP-1α or Eotaxin.

The transition from acute severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection to later dyspnea, fatigue, and cognitive symptoms remains difficult to map in a U.S. Veteran population with substantial comorbidity burden. Blood sampling in the first week after the first positive test and again 14 to 42 days later, followed by symptom questionnaires completed 60 to 135 days after enrollment, allowed the investigators to relate cytokine measurements to 30-day severity and later symptoms across early and post-acute study windows.

A prospective longitudinal analysis of U.S. Veterans in the EPIC3 veteran cohort cytokine study in Frontiers in Immunology recruited across 16 Veterans Affairs medical centers from July 2020 to September 2022 included 503 SARS-CoV-2-positive participants with cytokine data for severity analyses, including 387 early samples, 191 post-acute samples, 72 paired longitudinal samples, and 260 participants in the long-term symptom subset. Median sampling times were 2 days (IQR 0-3) and 32 days (IQR 29-35) after the first positive test; investigators measured a 45-analyte peripheral blood cytokine panel on the Luminex platform, defined severity by the highest Veterans Affairs Severity Index for COVID-19 (VASIC) category within 30 days with severe disease and death combined, and assessed long-term outcomes with the modified Medical Research Council (mMRC) dyspnea scale, the Patient-Reported Outcomes Measurement Information System (PROMIS) Fatigue 6a and PROMIS Cognitive Function 4a using the earliest questionnaire completed 60 to 135 days after enrollment, with adjusted models including age, sex, and Charlson Comorbidity Index (CCI).

In serial cytokine associations with COVID-19 severity and long-term symptoms, higher early hepatocyte growth factor (HGF; q < 0.001), interleukin-1 receptor antagonist (IL-1RA; q = 0.014), interleukin-18 (IL-18; q = 0.009), interferon gamma-induced protein-10 (IP-10; q < 0.001), and vascular endothelial growth factor-A (VEGF-A; q = 0.011) tracked with increasing severity, while higher post-acute macrophage inflammatory protein-1 alpha (MIP-1α) and VEGF-A also tracked with greater severity at q < 0.05. In an early pairwise comparison, higher monocyte chemoattractant protein-1 (MCP-1) was associated with severe versus moderate disease, with an odds ratio of 1.37, 95% CI 1.11-1.70, and q = 0.036. Other early pairwise signals for HGF, IP-10, IL-18, and VEGF-A generally moved in the same direction.

Adjusted regression models did not identify associations between single-timepoint cytokine concentrations and later dyspnea, fatigue, or cognitive impairment in either the early or post-acute sampling period. In the paired-sample subset, mild disease was associated with increases in epithelial growth factor (EGF; q = 0.020), macrophage inflammatory protein-1 beta (MIP-1β; q = 0.026), and Regulated upon Activation, Normal T Cell Expressed and Presumably Secreted (RANTES; q = 0.020) and with a decrease in MIP-1α (q = 0.026) over time. Absence of long-term symptoms was associated with increases in MIP-1β (q = 0.029) and RANTES (q = 0.007) over time; additional RANTES associations were reported for unimpaired cognition and absence of fatigue or dyspnea, while Eotaxin decreased over time in participants without fatigue or cognitive impairment but not in those with those symptoms.

Early and post-acute samples overlapped incompletely, and participants contributing post-acute samples were younger and more often outpatients, so between-period contrasts cannot be read as purely temporal. The positivity definition could also capture earlier infections or false positives, so the windows reflect study timing around testing and enrollment rather than precise biologic stages. Cytokines with more than 20% missingness or out-of-range values were excluded from the main analysis, including interleukin-6 (IL-6), interleukin-1 beta (IL-1β), and tumor necrosis factor-alpha (TNF-α), and long-term symptom analyses were limited to the approximately 3-month window; sensitivity analyses were broadly concordant where sample sizes were adequate. Within this dataset, adjusted single-timepoint models were null, while unadjusted analyses in the smaller paired subset identified several longitudinal cytokine changes associated with later symptom status; the observational design does not establish causality.

Serial cytokine patterns were associated with 30-day COVID-19 severity in this veteran cohort, and at about 3 months, long-term symptom status was associated with longitudinal cytokine changes, whereas adjusted single-timepoint cytokine models were null.

Clinician Questions

Which cytokines were associated with worse 30-day COVID-19 severity in U.S. Veterans during early versus post-acute infection?

In the EPIC3 veteran cohort, higher early HGF, IL-1RA, IL-18, IP-10, and VEGF-A and higher post-acute MIP-1α and VEGF-A were associated with increasing VASIC severity within 30 days, with severe disease and death combined for analysis.

Were single-timepoint cytokine levels linked to dyspnea, fatigue, or cognitive impairment about 3 months after SARS-CoV-2 infection?

Adjusted regression models in the EPIC3 veteran cohort did not identify associations between cytokine concentrations measured during either the early or post-acute sampling period and later dyspnea, fatigue, or cognitive impairment assessed 60 to 135 days after enrollment.

What longitudinal cytokine changes were associated with mild COVID-19 in the paired-sample EPIC3 subset?

Among 72 participants with paired early and post-acute samples, mild COVID-19 was associated with increases in EGF, MIP-1β, and RANTES and a decrease in MIP-1α over time, as reported by the investigators.

Which cytokine trajectories were associated with absence of long-term symptoms after COVID-19 in the EPIC3 veteran cohort?

In the EPIC3 veteran cohort, absence of long-term symptoms was associated with increases in MIP-1β and RANTES over time, and Eotaxin decreased over time in participants without fatigue or cognitive impairment, whereas those changes were not seen in participants with those symptoms.