Day-5 IL-9 levels may reflect an evolving risk in severe COVID-19, but the association with in-hospital mortality was sensitive to multiple-testing correction and model instability, making findings exploratory and requiring confirmation in larger cohorts.
Key Findings
Results
Pulmonary embolism was confirmed in 40.4% of severe COVID-19 patients evaluated for PE, and in-hospital all-cause mortality was 27.7%.
47 hospitalized adults with severe COVID-19 and suspected PE were enrolled in this prospective observational study.
PE was confirmed using computed tomography pulmonary angiography in 19 of 47 patients (40.4%).
13 of 47 patients (27.7%) died during hospitalization.
Serum cytokines, coagulation parameters, and cardiac injury biomarkers were assessed at enrollment and on hospital day 5.
Results
Higher standardized day-5 IL-9 and IL-15 levels were associated with in-hospital mortality in univariable analyses.
A panel of 15 cytokines was measured on hospital day 5.
Both IL-9 and IL-15 showed associations with mortality in univariable analyses.
Neither IL-9 nor IL-15 remained statistically significant after Benjamini-Hochberg false discovery rate correction (IL-9, q=0.180; IL-15, q=0.210).
The loss of significance after correction indicates these findings are susceptible to false discovery given the multiple comparisons across the 15-cytokine panel.
Results
In an age-adjusted multivariable model, day-5 IL-9 remained independently associated with in-hospital mortality.
The adjusted odds ratio per 1-SD increase in day-5 IL-9 was 3.331 (95% CI, 1.289–8.608; P=0.013).
This association persisted after adjusting for age in the multivariable binary logistic regression model.
Apparent model discrimination was moderate (AUC 0.774; 95% CI, 0.563–0.984; P=0.011).
Bootstrap internal validation revealed marked coefficient instability, undermining confidence in the model's reliability.
Results
Bootstrap resampling validation demonstrated marked coefficient instability in the age-adjusted IL-9 mortality prediction model.
Internal validation was performed using bootstrap resampling methodology.
Despite moderate apparent discrimination (AUC 0.774), bootstrap validation showed the model coefficients were unstable.
The wide confidence interval for the AUC (0.563–0.984) further reflects uncertainty in model performance.
The authors interpret this instability as indicating the findings are exploratory rather than confirmatory.
Conclusions
The study authors concluded that the association between day-5 IL-9 and mortality in severe COVID-19 requires confirmation in larger, externally validated cohorts.
The small sample size of 47 patients limits statistical power and generalizability.
The association between IL-9 and mortality did not survive false discovery rate correction across the full cytokine panel.
The authors explicitly describe their findings as 'exploratory.'
External validation in larger cohorts was identified as a necessary next step before clinical application.
What This Means
This research suggests that in patients hospitalized with severe COVID-19 who were being evaluated for blood clots in the lungs (pulmonary embolism), certain immune signaling proteins measured on the fifth day of hospitalization may be linked to the risk of dying in the hospital. Specifically, a protein called interleukin-9 (IL-9), measured on day 5, was associated with higher odds of in-hospital death even after accounting for patients' ages. About 40% of the 47 patients studied were found to have pulmonary embolism, and about 28% died during their hospital stay, reflecting the severity of illness in this group.
However, the study has important limitations that mean these findings should be interpreted with caution. When the researchers applied a statistical correction to account for the fact that they tested 15 different immune proteins at once (which increases the chance of a false positive result), the association between IL-9 and mortality was no longer statistically significant. Additionally, when they used a technique called bootstrap resampling to internally validate their prediction model, they found the results were unstable, meaning the model's predictions varied considerably depending on which patients were included in the analysis.
This research suggests that tracking how immune activity changes over the first few days of severe COVID-19 hospitalization — particularly proteins like IL-9 — could eventually help identify patients at higher risk of dying. However, because this was a small, single-center study with only 47 patients and showed signs of statistical instability, these findings are considered preliminary and exploratory. Larger studies with external validation would be needed before such measurements could be used to guide clinical decision-making.
Repečka T, Ališauskas A, Naudžiūnas A, Barauskienė G, Grabauskytė I, Vitkauskienė A, et al.. (2026). Dynamic Cytokine and Coagulation Profiling in Patients With Severe COVID‑19 Evaluated for Pulmonary Embolism: A Prospective Cohort Study.. Medical science monitor : international medical journal of experimental and clinical research. https://doi.org/10.12659/MSM.952644