Association of IL-1β, IL-6, IL-8, IL-12, TNF-α, and IFN-α With Severe Pediatric COVID-19 and Multisystem Inflammatory Syndrome Authors: Kateryna Kozak, Halyna Pavlyshyn Published: 2026 ABSTRACT AbstractBackgroundThe cytokine profile may provide clinically relevant information for assessing the inflammatory response in infectious diseases, including SARS-CoV-2. However, the threshold concentrations of cytokines associated with severe pediatric COVID-19 and cytokine patterns related to multisystem inflammatory syndrome in children (MIS-C), remain insufficiently characterized.ObjectivesThis study aimed to evaluate serum levels and exploratory threshold concentrations of IL-1β, IL-6, IL-8, IL-12, TNF-α, and IFN-α associated with severe pediatric COVID-19 and MIS-C, and to characterize cytokine patterns across different clinical forms of SARS-CoV-2 infection in children.DesignThe cohort study included 200 children with laboratory-confirmed SARS-CoV-2 infection, including mild COVID-19 (n=106), moderate COVID-19 (n=72), and severe COVID-19 (n= 22), 40 children diagnosed with MIS-C, and 45 children with negative PCR and ELISA results for SARS-CoV-2. Participants were aged 1 month to 17 years (6.47±5.53 years), and none had received a COVID-19 vaccine at the time of inclusion.MethodsSerum concentrations of IL-1β, IL-6, IL-8, IL-12, TNF-α, and IFN-α were measured within the first 24 hours of hospitalization, before treatment initiation, using immunoenzymatic assays.ResultsCytokines associated with severe COVID-19 included IL-1β (cut-off 6.96 pg/mL), IL-6 (cut-off 68.37 pg/mL), IL-12 (cut-off 34.53 pg/mL), TNF-α (cut-off 60.63 pg/mL), and IFN-α (cut-off 26.96 pg/mL). All measured cytokines demonstrated adequate discriminatory performance for identifying MIS-C: IL-1β (cut-off 7.28 pg/mL), IL-6 (cut-off 100.64 pg/mL), IL-8 (cut-off 16.93 pg/mL), IL-12 (cut-off 34.91 pg/mL), TNF-α (cut-off 80.19 pg/mL), and IFN-α (cut-off 30.56 pg/mL). Random Forest analysis identified IL-1β and IL-8 as variables with the highest model-derived importance for COVID-19 severity classification, whereas age and sex showed lower model-derived importance than cytokine biomarkers.ConclusionCytokine profiling may support a more nuanced evaluation of hyperinflammatory immune responses in pediatric SARS-CoV-2-associated disease. However, its additional clinical value beyond routine clinical and laboratory markers requires validation in larger independent pediatric cohorts.Keywords: IL-1β, IL-6, IL-8, IL-12, TNF-α, IFN-α, COVID-19, MIS-C 1. Introduction The clinical spectrum of SARS-CoV-2 infection in children ranges from asymptomatic cases (5–21%) to severe forms (1–8%) and the multisystem inflammatory syndrome in children (MIS-C) associated with COVID-19. 1 Notably, the mortality rate among pediatric patients remains exceptionally low (0–0.2%), whereas in adults it may reach 7%. 1 The relatively mild disease course in children is attributed to a combination of factors, including a more active innate and adaptive immune response, higher levels of IgM and IgG antibodies, lymphocytes, cross-reactive immunity, limited immunosenescence, the absence of comorbidities, higher melatonin levels, distinct gut microbiota, lower expression of angiotensin-converting enzyme 2 (ACE2) receptors, reduced binding affinity of these receptors to the viral spike protein, and reduced viral exposure and load.2,3 Age-related differences in immune regulation, particularly in interferon signaling and cytokine production, further shape disease outcomes.1,4 Adults typically exhibit higher levels of IL-1, IL-2, IL-6, IL-8, and TNF-α compared with children, which contributes to excessive inflammation, endothelial injury, and cytokine storm development.5,6 MIS-C, a distinct hyperinflammatory condition associated with COVID-19, predominantly occurs in children and is less frequently observed in young adults. 7 MIS-C is characterized by systemic inflammation involving the gastrointestinal tract, cardiovascular, nervous, and renal systems, accompanied by markedly elevated IL-6 levels.8,9 Although other cytokines have been implicated in MIS-C pathogenesis, they are not yet included in its formal diagnostic criteria. Cytokine profiling serves as a powerful tool for elucidating the immunopathogenesis of SARS-CoV-2 and for identifying potential biomarkers of disease severity. In adults, elevated levels of IL-1β, IL-1Ra, IL-2, IL-6, IL-8, IL-10, IL-17, IL-18, IL-33, TNF-α, IP-10, G-CSF, and YKL-40 have been associated with worse outcomes.10-14 An uncontrolled cytokine storm represents a central mechanism underlying severe and critical disease.12,13,15 Dysregulated type I interferon (IFN-α and IFN-β) responses also play a pivotal role: decreased IFN-α levels correlate with poor outcomes in adults, while elevated IFN-α concentrations have been reported in children with MIS-C compared to those with acute COVID-19.16-21 Recent pediatric studies have additionally demonstrated that elevated levels of IL-1β, IL-6, IL-8, and TNF-α are closely associated with hyperinflammatory responses and MIS-C development, supporting their role as potential biomarkers of disease severity and immune dysregulation in children with SARS-CoV-2 infection.22-24 Despite increasing evidence, the threshold levels of cytokines predictive of severe COVID-19 in children remain insufficiently characterized, as do the cytokine patterns distinguishing MIS-C from acute infection. Identifying the principal immunological determinants of disease severity in the pediatric population is therefore essential for refining diagnostic criteria and improving early risk assessment. Based on current evidence, we included IL-1β, IL-6, IL-8, IL-12, TNF-α and IFN-α in this study, as these cytokines represent central mediators of inflammation and immune regulation in SARS-CoV-2 infection. IL-1β is a potent proinflammatory cytokine that contributes to endothelial permeability and acute lung injury and has been linked to poor outcomes in adults; however, its systemic levels often vary due to its short half-life and predominant local pulmonary production.25-28 IL-6, one of the most extensively studied cytokines in COVID-19, correlates with disease severity, structural lung changes, and systemic inflammation, and remains a hallmark of MIS-C.12,29-31 TNF-α, a key driver of both acute and chronic inflammation, is associated with cytokine release syndrome and poor prognosis in adults and may indicate progression toward severe disease in children.13,32-34 IL-8 serves as a chemotactic factor for neutrophil recruitment and an indicator of early inflammatory activation; however, findings in pediatric populations remain inconsistent.8,35-37 IL-12, a crucial modulator of antiviral immunity, bridges innate and adaptive responses by stimulating IFN-γ and CD8+ T-cell activity, making it particularly relevant for understanding age-dependent immune mechanisms in SARS-CoV-2 infection.10,38 Collectively, these cytokines represent complementary aspects of the inflammatory response, providing a mechanistic framework for identifying potential biomarkers associated with severe COVID-19 and MIS-C. 2. Materials and Methods 2.1. Study Design and Participants The study was conducted in Ukraine and included 200 children with laboratory-confirmed SARS-CoV-2 infection, including 106 children with mild COVID-19, 72 with moderate COVID-19, and 22 with severe COVID-19, 40 children diagnosed with MIS-C, and 45 children with negative PCR and ELISA test results for SARS-CoV-2. The control group consisted of children who attended outpatient pediatric departments for routine health examinations and had negative PCR and ELISA results for SARS-CoV-2. These children had no clinical signs of acute infectious or inflammatory disease at the time of examination and were not hospitalized. They were therefore considered clinically healthy outpatient controls. Infants and toddlers were not represented in the control group because informed consent for inclusion in the study could not be obtained from the parents or legal guardians of healthy children aged from 1 month to 3 years. None of the children had been previously vaccinated against COVID-19 at the time of inclusion in the study. All patients with confirmed SARS-CoV-2 infection or a diagnosis of MIS-C received inpatient treatment at medical facilities in Ternopil, Ukraine. Children were included in the COVID-19 group if they had laboratory-confirmed SARS-CoV-2 infection and clinical manifestations of acute COVID-19. Children were included in the MIS-C group if they met the WHO diagnostic criteria for multisystem inflammatory syndrome in children associated with SARS-CoV-2 infection. The control group included clinically healthy outpatient children without signs of acute infectious or inflammatory disease at the time of examination and with negative PCR and ELISA results for SARS-CoV-2. Exclusion criteria were absence of informed consent, previous COVID-19 vaccination, incomplete clinical or laboratory data required for analysis, visible hemolysis or excessive lipemia of serum samples, known chronic inflammatory or autoimmune disease, primary or secondary immunodeficiency, malignant disease, ongoing immunosuppressive therapy, and administration of immunomodulatory treatment before blood sampling. In patients with MIS-C, intravenous immunoglobulin and systemic corticosteroids had not been administered before sample collection. Detailed data on possible coinfections were not consistently available for all participants; therefore, coinfection status could not be included as a covariate in the adjusted models and was considered a study limitation. Disease duration before hospitalization was recorded for all hospitalized patients with acute COVID-19 and MIS-C and was defined as the interval between the onset of clinical symptoms and hospital admission. The severity of COVID-19 was stratified according to the World Health Organization (WHO) criteria. 39 Mild infection was defined as the presence of clinical symptoms of COVID-19 without signs of pneumonia or hypoxia. Moderate severity was characterized by the presence of pneumonia with typical clinical manifestations (fever, cough, dyspnea, tachypnea, retractions), while SpO2 remained ≥ 90% when breathing atmospheric air. Severe COVID-19 was defined as the presence of severe pneumonia with the development of respiratory distress syndrome, a decrease in SpO2 < 90% in room air, or the presence of threatening clinical manifestations such as impaired consciousness, convulsions, refusal to eat or drink. The diagnosis of multisystem inflammatory syndrome in children (MIS-C) was established according to WHO criteria. 9 These criteria require the patient to be between the ages of 0 and 19 years, have a fever lasting for at least 3 days, and exhibit at least two of the following clinical signs: skin rash, bilateral non-purulent conjunctivitis, or signs of inflammatory lesions on the skin or mucous membranes; arterial hypotension or shock; signs of heart damage (such as myocarditis, pericarditis, valve damage, or coronary artery abnormalities), elevated levels of troponin or NT-proBNP; coagulopathy (elevated D-dimer, prothrombin time, or APTT); acute gastrointestinal symptoms (such as diarrhea, vomiting, or abdominal pain); signs of systemic inflammation (elevated ESR, C-reactive protein, or procalcitonin); exclusion of other possible causes of inflammation, including bacterial sepsis, staphylococcal or streptococcal toxic shock syndrome; and confirmation of SARS-CoV-2 infection (positive PCR, antigen test, or serological detection of antibodies) or established close contact with a COVID-19 patient within the last 4 weeks. 9 The study included children from 1 month to 17 years of age, with an average age of 6.47 ± 5.53 years. The observation group consisted of 155 boys (54.39%) and 130 girls (45.61%). Table 1 presents a detailed breakdown of patients by age and sex, categorized by the severity of their disease. Table 1.Age and Sex Characteristics of Children Included in the StudyParameterCOVID-19MIS-CControlMildModerateSevere Sex Male50 (47.17)45 (62.50)8 (36.36)24 (60.00)28 (62.22)Female56 (52.83)27 (37.50)14 (63.64)16 (40.00)17 (37.78) Age groups Infancy34 (32.08)19 (26.39)3 (13.64)−−Toddler13 (12.26)7 (9.72)1 (4.55)5 (12.50)−Early childhood28 (26.42)16 (22.22)3 (13.64)12 (30.00)8 (17.78)Middle childhood17 (16.04)12 (16.67)5 (22.73)17 (42.50)19 (42.22)Early adolescence14 (13.21)18 (25.00)10 (45.45)6 (15.00)18 (40.00)Note 1. MIS-C – multisystem inflammatory syndrome in children.Note 2. Pearson’s Chi-square test for sex vs. severity groups: χ2=8.64, p=0.071.Note 3. Pearson’s Chi-square test for age groups vs. severity groups: χ2=68.05, p<0.001. Age and Sex Characteristics of Children Included in the Study Note 1. MIS-C – multisystem inflammatory syndrome in children. Note 2. Pearson’s Chi-square test for sex vs. severity groups: χ2=8.64, p=0.071. Note 3. Pearson’s Chi-square test for age groups vs. severity groups: χ2=68.05, p<0.001. To analyze the age characteristics of the patients, we utilized the Pediatric Terminology classification proposed by the National Institute of Child Health and Human Development. 40 This classification divides participants into the following age groups: infants (0–12 months), toddlers (1–2 years), early childhood (2–5 years), middle childhood (6–11 years), and early adolescence (12–17 years). Comorbid conditions were documented in 80 of 285 children (28.1%). The frequency of comorbidity increased with COVID-19 severity: comorbidities were present in 22 of 106 children with mild disease (20.8%), 23 of 72 children with moderate disease (31.9%), and 11 of 22 children with severe disease (50.0%). In the MIS-C group, comorbid conditions were observed in 18 of 40 children (45.0%), whereas in the control group they were recorded in 6 of 45 children (13.3%). The most frequent comorbidity categories in the overall cohort were gastroenterological/hepatobiliary disorders (9.8%), endocrine conditions (7.4%), nephro-urological disorders (4.2%), allergic conditions (3.9%), and cardiovascular disorders (3.5%). Less frequent categories included hematological disorders (2.1%), ENT disorders (2.1%), neurological disorders (1.1%), ophthalmological disorders (1.1%) and congenital malformations (1.1%). The most common individual comorbidities were biliary tract dysfunction and overweight/obesity. Other documented conditions included allergic rhinitis, iron-deficiency anemia, atopic dermatitis, congenital heart disease, gastroesophageal reflux disease, drug allergy, adenoids, bronchial asthma, tonsillar hypertrophy, myopia, underweight, renal pyelectasis, and chronic tonsillitis. The number of comorbidity categories exceeded the number of children with comorbidity because some patients had more than one documented condition. The study was conducted in accordance with the principles of the Helsinki Declaration (2024 version) and was approved by the Bioethics Committee of I. Horbachevsky Ternopil National Medical University of the Ministry of Health of Ukraine (Protocol No. 71, dated 25 October 2022). 2.2. Cytokine Analysis Venous blood samples were collected from all participants during the first 24 hours of hospitalization, before the initiation of hospital-based therapy. Before inclusion in the study, patients received only symptomatic treatment, which was not expected to substantially affect the serum levels of the investigated cytokines. In patients with MIS-C, immunomodulatory therapy, including intravenous immunoglobulin and systemic corticosteroids, had not been administered before blood sampling. Trained medical personnel used standard aseptic techniques. Blood was collected into sterile, disposable, non-endotoxin tubes. Samples with visible hemolysis or excessive lipemia were not analyzed. Blood was centrifuged at 2000 revolutions per minute for 10 minutes to obtain serum, which was aliquoted and stored at –80°C until testing. Repeated freeze–thaw cycles were avoided. Serum cytokine levels were measured using commercial ELISA kits (Elabscience) according to the manufacturer’s instructions: Human IL-1β (Interleukin 1 Beta) ELISA Kit (Cat. No.: E-EL-H0149); Human IL-6 (Interleukin 6) ELISA Kit (Cat. No.: E-EL-H6156); Human IL-8 (Interleukin 8) ELISA Kit (Cat. No.: E-EL-H6008); Human IL-12 (Interleukin 12) ELISA Kit (Cat. No.: E-EL-H0150); Human TNF-α (Tumor Necrosis Factor Alpha) ELISA Kit (Cat. No.: E-EL-H0109); Human IFN-α (Interferon Alpha) ELISA Kit (Cat. No.: E-EL-H6125). For IL-12, the kit documentation identifies the assay target as human IL-12 and does not specify IL-12p70. The analytical characteristics of the ELISA kits were as follows: IL-1β, sensitivity 4.69 pg/mL, detection range 7.81–500 pg/mL; IL-6, sensitivity 0.94 pg/mL, detection range 1.56–100 pg/mL; IL-8, sensitivity 4.69 pg/mL, detection range 7.81–500 pg/mL; IL-12, sensitivity 9.38 pg/mL, detection range 15.63–1000 pg/mL; TNF-α, sensitivity 4.69 pg/mL, detection range 7.81–500 pg/mL; and IFN-α, sensitivity 4.69 pg/mL, detection range 7.81–500 pg/mL. According to the manufacturer’s specifications, the repeatability coefficient of variation for each kit was <10%. Standards were run according to the manufacturer’s protocol to generate the calibration curve for each assay. Serum samples were analyzed according to the manufacturer’s instructions. Due to limited sample volume and reagent availability, duplicate measurements could not be performed for all samples. Samples were analyzed without preliminary dilution unless the optical density exceeded the upper limit of the standard curve. In such cases, samples were re-tested after appropriate dilution, and the final concentration was calculated by multiplying the measured value by the dilution factor. Values below the lower limit of detection were recorded as below the detection limit. 2.3. Statistical Analysis Statistical analysis was performed using IBM SPSS Statistics 21, GraphPad Prism 8.4.3, and JASP 0.19.3 software. Descriptive statistics of quantitative variables were presented according to data distribution. Normally distributed variables were expressed as mean ± standard deviation (Mean ± SD), whereas non-normally distributed variables were presented as median with interquartile range (Median [Lq; Uq]). Categorical variables were summarized as absolute numbers and percentages. Frequency tables were analyzed using Pearson’s chi-square test. Differences were considered statistically significant at p < 0.05. For comparisons of quantitative variables among more than two independent groups, the Kruskal–Wallis H-test was used. If the Kruskal–Wallis test showed a statistically significant difference (p < 0.05), post-hoc pairwise comparisons were performed using adjusted significance values for multiple comparisons implemented in SPSS. Correlation analysis was performed using Spearman’s rank correlation coefficient. Linear regression analysis was used to evaluate associations between clinical characteristics (age group, sex, comorbidity status, and disease duration before hospitalization) and cytokine concentrations. Results were presented as standardized regression coefficients (β), standard errors (SE), t-values, and p-values. Logistic regression analysis was performed to evaluate the independent association of cytokine concentrations with severe COVID-19 and MIS-C after adjustment for age group, sex, comorbidity status, and disease duration before hospitalization. Results were presented as adjusted odds ratios (ORs) with corresponding 95% confidence intervals (95% CI) and p-values. ROC curve analysis was used to assess the discriminatory performance of cytokine concentrations for identifying severe acute COVID-19 and for differentiating MIS-C from acute COVID-19. Severe COVID-19 cases were compared with non-severe acute COVID-19 cases, whereas MIS-C cases were compared with acute COVID-19 cases. For each ROC analysis, the area under the curve (AUC), standard error, 95% confidence interval for AUC, p-value, optimal cut-off value, sensitivity, specificity, and positive likelihood ratio were calculated. Because multiple ROC analyses were performed for the cytokine panel, Benjamini–Hochberg false discovery rate correction was applied to ROC-derived p-values separately for severe COVID-19 and MIS-C analyses. FDR-adjusted p-values were reported as q-values, with q < 0.05 considered statistically significant after correction for multiple comparisons. Missing data were handled using an analysis-specific complete-case approach. Participants with missing cytokine or clinical values were excluded only from analyses requiring those variables. No imputation of missing data was performed. The Random Forest model included only participants with complete data for all model variables. To identify cytokines associated with severe COVID-19, an exploratory Random Forest classification model was developed. The outcome variable was severe COVID-19, with patients classified as severe COVID-19 versus non-severe COVID-19, the latter including mild and moderate cases. Cytokine concentrations (IL-1β, IL-6, IL-8, IL-12, TNF-α, and IFN-α), age, and sex were included as input variables. Only patients with complete datasets for all analyzed variables were included in the machine-learning analysis. The Random Forest model was constructed using 300 decision trees, with two variables randomly selected at each split. Model performance was internally assessed using an 80:20 train/test split. Classification performance was evaluated using test accuracy, sensitivity, specificity, positive predictive value, negative predictive value, balanced accuracy, and AUC. A confusion matrix was additionally used to describe classification results in the testing subset. Feature importance was assessed using mean decrease in accuracy, total increase in node purity, and mean dropout loss. 3. Results 3.1. Age Distribution of Study Participants Analysis of the age distribution of children included in the study revealed significant differences between groups with varying severity of COVID-19 and MIS-C (χ2 = 68.05, p < 0.001). Infants were the most common age group in the mild SARS-CoV-2 infection group (32.08%), while this age group was not represented in the MIS-C group. Between the ages of 1 and 5, there was no significant difference in the frequency of SARS-CoV-2 infection and MIS-C. However, from age 6 onward, there was an increase in the frequency of severe clinical forms: MIS-C was more commonly diagnosed in school-age children (42.50%). In comparison, severe COVID-19 was more prevalent in early adolescence (45.45%). A comparative analysis within the COVID-19 patient group showed that mild disease predominated in children under 5 years old: 60.71% of infants (34/56), 61.90% of children under 2 years old (13/21), and 59.57% of patients aged 2 to 5 years (28/47) had mild forms. Conversely, the highest proportions of severe COVID-19 were observed in early adolescence (23.81%, 10/42) and middle childhood (14.71%, 5/34). 3.2. Serum Cytokine Levels and Association With Disease Severity Levels of IL-1β and IL-12 did not differ significantly between the control group and children with acute SARS-CoV-2 infection (p > 0.05). However, patients diagnosed with MIS-C showed significantly elevated levels of IL-1β and IL-12 compared to both the control group and children with COVID-19. Analysis of IL-6, IL-8, and TNF-α levels revealed increases during SARS-CoV-2 infection and at the onset of MIS-C relative to controls. Notably, MIS-C was characterized by significantly higher IL-6, IL-8, and TNF-α levels compared to COVID-19. The level of IFN-α in healthy children differed substantially from those in the COVID-19 and MIS-C groups; IFN-α levels decreased 3.75-fold in COVID-19 and 6.33-fold in MIS-C compared to controls (Figure 1). Post-hoc pairwise comparisons were interpreted using adjusted significance values for multiple comparisons. Serum cytokine levels in children with COVID-19 (n=200), MIS-C (n=40), and healthy control (n=45). Notes. Bar graphs represent median values with interquartile ranges for each group: Control (blue columns), COVID-19 (purple columns), MIS-C (orange columns). Red brackets]* indicate statistically significant differences between groups based on post-hoc pairwise comparisons with adjusted significance values for multiple comparisons (adjusted p < 0.05) A comparative analysis of cytokine levels according to the severity of COVID-19 revealed that the concentrations of most cytokines, except for IL-8, did not differ significantly between groups with mild and moderate disease courses. Meanwhile, children with severe COVID-19 and MIS-C exhibited substantially higher levels of IL-1β, IL-6, IL-8, IL-12, and TNF-α compared to those with mild infections. It is important to emphasize that when comparing patients with moderate COVID-19 and MIS-C, the levels of IL-1β, IL-6, IL-12, and TNF-α were significantly higher in the MIS-C group. In addition, TNF-α levels increased significantly with worsening disease severity, from moderate to severe COVID-19. Regarding IFN-α levels, children with severe COVID-19 and MIS-C had significantly lower levels compared to those recorded in patients with mild disease (Table 2). Table 2.Cytokine Level in Children With Different COVID-19 Courses and MIS-CParameterMild COVID-19Moderate COVID-19Severe COVID-19MIS-C1234IL-1βpg/mL2.11 3, 4 (0.71; 3.73)4.88 4 (1.47; 6.89)9.56 (4.01; 12.87)10.20 (7.90; 14.31)H, pH=31.95; p < 0.001*IL-6pg/mL51.45 3, 4 (24.14; 181.55)65.28 4 (32.75; 209.45)125.35 (90.14; 388.41)295.40 (120.20; 608.88)H, pH=27.96; p < 0.001*IL-8pg/mL12.23 2, 3, 4 (8.44; 17.39)21.52 (13.48; 26.46)26.19 (12.60; 43.75)28.63 (17.05; 53.53)H, pH=17.81; p < 0.001*IL-12pg/mL14.27 3, 4 (8.39; 25.89)16.70 4 (11.96; 41.43)40.53 (30.29; 70.75)79.17 (35.01; 178.30)H, pH=16.87; p < 0.001*TNF-αpg/mL21.29 3, 4 (12.37; 37.43)39.94 3, 4 (19.17; 70.48)99.32 (82.69; 205.00)112.34 (57.75; 466.30)H, pH=42.38; p < 0.001*IFN-αpg/mL66.44 3, 4 (27.33; 104.10)38.12 (32.02; 52.96)20.94 (17.39; 44.09)22.21 (16.20; 29.09)H, pH=15.52; p = 0.001*Note 1. MIS-C – multisystem inflammatory syndrome in children.Note 2. 2, 3, 4 – statistically significant differences compared with the corresponding group based on post-hoc pairwise comparisons with adjusted significance values for multiple comparisons (adjusted p < 0.05).Note 3. H − Kruskal-Wallis test and p – level of its significance.Note 4. * −a statistically significant result of the Kruskal–Wallis test (p < 0.05). Cytokine Level in Children With Different COVID-19 Courses and MIS-C Note 1. MIS-C – multisystem inflammatory syndrome in children. Note 2. 2, 3, 4 – statistically significant differences compared with the corresponding group based on post-hoc pairwise comparisons with adjusted significance values for multiple comparisons (adjusted p < 0.05). Note 3. H − Kruskal-Wallis test and p – level of its significance. Note 4. * −a statistically significant result of the Kruskal–Wallis test (p < 0.05). 3.3. Age- and Sex-Stratified Cytokine Patterns Considering the age heterogeneity of the sample, a stratified analysis of cytokine levels was conducted across different age groups within the three studied cohorts: children with COVID-19, patients with MIS-C, and the control group. First, it is essential to note that in the control group (children without confirmed SARS-CoV-2 infection), no significant age-related differences were found in serum levels of IL-1β, IL-6, IL-8, IL-12, TNF-α, and IFN-α (p > 0.05). In patients with acute SARS-CoV-2 infection and MIS-C, no statistically significant age-related differences were observed in the levels of IL-1β, IL-8, TNF-α, and IFN-α (p > 0.05). The level of IL-1β in patients with MIS-C was significantly higher in each age category compared to patients with COVID-19 and healthy controls (p < 0.05). IL-6 showed age dependence in patients with COVID-19: its level in early adolescence (125.35 pg/mL) was significantly higher than in patients aged 2 to 5 years (45.07 pg/mL) (p < 0.05). This age-related pattern was not observed in the MIS-C group. For all age groups, an increase in IL-6 levels was characteristic during COVID-19 and MIS-C compared to healthy peers, with IL-6 being significantly higher in the MIS-C group compared to the COVID-19 group between ages 2 and 11 years (p < 0.05). In the 6–11-year age range, IL-8 levels did not differ significantly between the observation groups (COVID-19, MIS-C, and control) (p > 0.05). However, in early childhood and early adolescence, IL-8 levels were significantly higher in the COVID-19 and MIS-C groups compared to healthy controls (p < 0.05). IL-12 levels in children with COVID-19 also demonstrated age differences − the level in adolescence (37.13 pg/mL) was significantly higher than in early childhood (13.01 pg/mL) (p < 0.05). Meanwhile, even in children aged 2–5 years with MIS-C, IL-12 did not exceed control values (p > 0.05). TNF-α levels were significantly elevated across all age groups in patients with MIS-C and COVID-19 compared to controls (p < 0.05). Moreover, TNF-α values in MIS-C patients were considerably higher than those in children with acute COVID-19 infection (p < 0.05). It should be emphasized that IFN-α levels in children infected with SARS-CoV-2 and in those with MIS-C were significantly reduced compared to healthy children across all age groups (Figure 2). Age-stratified heatmaps of serum cytokine levels in pediatric patients with COVID-19 (n=200), MIS-C (n=40), and healthy control (n=45). Notes. Red arrows indicate statistically significant differences between groups after adjustment for multiple comparisons (adjusted p < 0.05). “X” marks denote the absence of participants in the corresponding age category. Color intensity reflects cytokine concentration, with darker orange representing higher levels Analysis of sex differences in the levels of the studied cytokines found no significant differences between boys and girls within each observation group (p > 0.05) (Figure 3). Sex-based differences in serum cytokine levels among pediatric patients with COVID-19 (n=200), MIS-C (n=40), and healthy control (n=45). Notes. Bar graphs represent median values with interquartile ranges for each group: Boys (blue columns) and Girls (pink columns). No statistically significant differences were observed between groups (p > 0.05) 3.4. Clinical Covariates and Adjusted Associations of Cytokines With Severe COVID-19 and MIS-C Given the presence of comorbid conditions in the examined cohorts, univariable linear regression analyses were performed separately in children with COVID-19 and MIS-C to evaluate the association between comorbidity status and cytokine concentrations (Table 3). In the COVID-19 group, comorbidity status was significantly associated only with higher IL-12 concentrations (β = 0.249, p = 0.047), while a borderline association was observed for TNF-α (β = 0.160, p = 0.066). No significant associations were found for IL-1β, IL-6, IL-8, or IFN-α. In the MIS-C group, comorbidity status was significantly associated with lower IL-6 concentrations (β = -0.484, p = 0.019), whereas no significant associations were identified for IL-1β, IL-8, IL-12, TNF-α, or IFN-α. Table 3.Univariable Regression Analysis of the Association Between Comorbidity Status and Cytokine Concentrations in SARS-CoV-2 Infected ChildrenCytokineCOVID-19MIS-CβSEtpβSEtpIL-1β0.1170.1610.7240.473-0.1990.253-0.7870.444IL-6-0.0460.087-0.5310.596-0.4840.191-2.5340.019*IL-80.0840.1270.6680.507-0.2740.278-0.9860.344IL-120.2490.1232.0230.047*0.3950.2651.4910.162TNF-α0.1600.0861.8530.066-0.1670.215-0.7770.446IFN-α-0.1090.135-0.8090.422-0.3770.239-1.5770.136Abbreviations: β – standardized regression coefficient, SE − Standard Error.Note. * − a statistically significant result (p < 0.05). Univariable Regression Analysis of the Association Between Comorbidity Status and Cytokine Concentrations in SARS-CoV-2 Infected Children Abbreviations: β – standardized regression coefficient, SE − Standard Error. Note. * − a statistically significant result (p < 0.05). To evaluate the potential influence of symptom duration before hospitalization on cytokine concentrations, Spearman correlation analyses were performed separately in the COVID-19 and MIS-C cohorts (Table 4). Patients with MIS-C had a longer disease duration before hospitalization compared with children with acute COVID-19 (4.45 ± 2.29 vs. 3.35 ± 2.34 days, respectively; p = 0.006). In children with COVID-19, a weak positive correlation was observed between TNF-α concentrations and disease duration (r=0.20, p=0.025). No significant correlations were identified for IL-1β, IL-6, IL-8, IL-12, or IFN-α. Similarly, in the MIS-C cohort, none of the investigated cytokines demonstrated significant correlations with disease duration before hospitalization (all p>0.05). Table 4.Correlations Between Cytokine Levels and Disease Duration Before HospitalizationParameterCOVID-19MIS-CrprpIL-1β0.040.773-0.270.316IL-60.160.0720.150.566IL-80.180.1510.100.755IL-120.190.126-0.220.480TNF-α0.200.025*-0.100.704IFN-α-0.150.2830.030.925Abbreviations: r – Spearman coefficient of correlation.Note. * − a statistically significant result (p < 0.05). Correlations Between Cytokine Levels and Disease Duration Before Hospitalization Abbreviations: r – Spearman coefficient of correlation. Note. * − a statistically significant result (p < 0.05). To determine the independent contribution of demographic and clinical factors to cytokine variability in children with SARS-CoV-2-associated disease, separate multivariable linear regression models were constructed for each cytokine (Table 5). Age group and sex were not independently associated with any of the investigated cytokines. Comorbidity status remained independently associated with higher IL-12 concentrations (β = 0.328, p = 0.005). Disease duration before hospitalization demonstrated a positive association with IL-6 concentrations (β = 0.183, p = 0.029) and a negative association with IFN-α concentrations (β = -0.252, p = 0.042). No independent determinants were identified for IL-1β, IL-8, or TNF-α concentrations in the multivariable linear regression models. Table 5.Multivariable Linear Regression Analysis of Clinical Determinants of Cytokine Concentrations in Children With SARS-CoV-2-Associated DiseaseCytokinePredictorβStandard errortp-valueIL-1βAge group0.1050.1150.9100.365Sex-0.1210.118-1.0280.307Comorbidity0.1050.1170.8930.374Disease duration0.0800.1170.6880.493IL-6Age group0.0060.0820.0750.940Sex-0.0660.080-0.8260.411Comorbidity-0.0880.081-1.0870.280Disease duration0.1830.0832.2140.029*IL-8Age group0.1310.1161.1300.262Sex-0.2100.118-1.7910.079Comorbidity0.0660.1180.5590.578Disease duration0.0890.1170.7540.453IL-12Age group0.1850.1121.6600.101Sex-0.1180.115-1.0250.308Comorbidity0.3280.1142.8860.005*Disease duration0.0950.1130.8390.404TNF-αAge group-0.0350.083-0.4280.669Sex-0.0210.081-0.2640.792Comorbidity0.1220.0821.4930.139Disease duration0.0960.0831.1520.253IFN-αAge group-0.0300.117-0.2590.796Sex-0.1140.122-0.9310.354Comorbidity-0.1000.123-0.8150.417Disease duration-0.2520.121-2.0690.042*Abbreviations: β – standardized regression coefficient.Note. * − a statistically significant result (p < 0.05). Multivariable Linear Regression Analysis of Clinical Determinants of Cytokine Concentrations in Children With SARS-CoV-2-Associated Disease Abbreviations: β – standardized regression coefficient. Note. * − a statistically significant result (p < 0.05). Logistic regression analysis was performed to evaluate the independent association of cytokine concentrations with severe COVID-19 and MIS-C after adjustment for age group, sex, comorbidity status, and disease duration before hospitalization. After adjustment for multiple testing using the Benjamini–Hochberg false discovery rate correction, IL-1β, IL-6, and TNF-α retained statistically significant associations with severe COVID-19. In the MIS-C analysis, IL-1β, IL-6, IL-12, and TNF-α remained significant after FDR correction (Table 6). Table 6.Adjusted Associations of Cytokine Concentrations With Severe COVID-19 and MIS-CParameterSevere COVID-19MIS-CAdjusted OR95% CIp-valueq valueAdjusted OR95% CIp-valueq valueIL-1β1.1791.023–1.3590.0200.040*1.1791.058–1.3120.0020.003*IL-65.9521.818–19.4900.0030.009*10.5213.031–36.522<0.0010.003*IL-81.0130.986–1.0410.3380.4061.0190.999–1.0390.0540.055IL-121.0020.986–1.0180.7770.7771.0211.008–1.0350.0020.003*TNF-α24.2904.985–118.360<0.0010.006*4.7031.879–11.795<0.0010.003*IFN-α0.9590.919–1.0010.0510.0770.9700.938–1.0010.0550.055Abbreviations: OR − odds ratio; CI − confidence interval; q-value − FDR-adjusted p-value calculated using the Benjamini–Hochberg false discovery rate correction.Note 1. Each model was adjusted for age group, sex, comorbidity status, and disease duration before hospitalization.Note 2. * − a statistically significant result after FDR correction (q < 0.05). Adjusted Associations of Cytokine Concentrations With Severe COVID-19 and MIS-C Abbreviations: OR − odds ratio; CI − confidence interval; q-value − FDR-adjusted p-value calculated using the Benjamini–Hochberg false discovery rate correction. Note 1. Each model was adjusted for age group, sex, comorbidity status, and disease duration before hospitalization. Note 2. * − a statistically significant result after FDR correction (q < 0.05). 3.5. Cytokine Associations With Severe COVID-19 and MIS-C ROC analyses were performed to evaluate two clinically distinct comparisons: identification of severe acute COVID-19 among children with acute SARS-CoV-2 infection and differentiation of MIS-C from acute COVID-19. The cytokines demonstrating the strongest discriminatory performance for severe COVID-19 were: IL-1β (AUC = 0.700; 95% CI: 0.561–0.838), IL-6 (AUC = 0.706; 95% CI: 0.617–0.796), IL-12 (AUC = 0.701; 95% CI: 0.566–0.836), TNF-α (AUC = 0.796; 95% CI: 0.708–0.886), and IFN-α (AUC = 0.723; 95% CI: 0.581–0.865). In the MIS-C analysis, all investigated cytokines demonstrated adequate discriminatory performance for differentiating MIS-C from acute COVID-19. After Benjamini–Hochberg false discovery rate correction, IL-1β, IL-6, IL-12, TNF-α, and IFN-α remained significantly associated with severe COVID-19 in ROC analysis, whereas IL-8 did not reach statistical significance. In the MIS-C ROC analysis, all investigated cytokines remained statistically significant after FDR correction. Among the analyzed biomarkers, TNF-α showed the strongest discriminatory performance for severe COVID-19 (AUC = 0.796), while IL-1β demonstrated the highest discriminatory performance for MIS-C (AUC = 0.861), indicating their potential role as key inflammatory biomarkers associated with severe disease manifestations. Notably, IL-1β and IL-6 in MIS-C had likelihood ratios (LR+) over 3, suggesting acceptable discriminatory performance within the analyzed cohort. For severe COVID-19, the strongest discriminatory value was found for TNF-α (LR+ = 2.92), IL-12 (LR+ = 2.59), and decreased IFN-α levels (LR+ = 2.56). ROC analysis allowed determination of threshold concentrations for each cytokine associated with severe COVID-19 and MIS-C (Tables 7 and 8; Figures 4 and 5). Confidence intervals for sensitivity, specificity, and AUC estimates are presented in Tables 7 and 8 and likely reflect the relatively limited number of patients in the severe COVID-19 subgroup. Nevertheless, the identified cytokine threshold concentrations consistently demonstrated associations with severe inflammatory manifestations of pediatric SARS-CoV-2-associated disease within the study cohort. Table 7.Cytokine Levels Associated With Severe COVID-19 in ChildrenParameterIL-1β, pg/mLIL-6, pg/mLIL-8, pg/mLIL-12, pg/mLTNF-α, pg/mLIFN-α, pg/mLAUC0.7000.7060.6430.7010.7960.723Std. Error0.0710.0460.0830.0690.0450.07395% CI AUC0.561−0.8380.617−0.7960.481−0.8050.566−0.8360.708−0.8860.581−0.865Р value0.0180.0020.1130.026< 0.0010.011q value0.027*0.006*0.1130.031*0.006*0.022*Cut-off value6.9668.3717.1834.5360.6326.96Sensitivity (95 % CI)64.29 (38.76−83.66)95.24 (77.33−99.76)66.67 (39.06−86.19)75.00 (46.77−91.11)90.48 (71.09−98.31)61.54 (35.52−82.29)Specificity (95% CI)68.92 (57.66−78.31)54.19 (46.34−61.84)60.53 (49.29; 70.75)71.05 (60.04−80.04)69.03 (61.37−75.78)76.00 (65.22−84.25)Likelihood ratio2.072.081.692.592.922.56Note 1. AUC − area under the curve; 95% CI AUC − 95% confidence interval for the area under the curve; q-value − FDR-adjusted p-value calculated using the Benjamini–Hochberg false discovery rate correction.Note 2. * − a statistically significant result after FDR correction (q < 0.05). Cytokine Levels Associated With Severe COVID-19 in Children Note 1. AUC − area under the curve; 95% CI AUC − 95% confidence interval for the area under the curve; q-value − FDR-adjusted p-value calculated using the Benjamini–Hochberg false discovery rate correction. Note 2. * − a statistically significant result after FDR correction (q < 0.05). Table 8.Cytokine Levels Associated With MIS-C in ChildrenParameterIL-1β, pg/mLIL-6, pg/mLIL-8, pg/mLIL-12, pg/mLTNF-α, pg/mLIFN-α, pg/mLAUC0.8610.7990.7860.7830.7870.763Std. Error0.0420.0430.0620.0760.0600.06195% CI AUC0.778−0.9440.715−0.8820.664−0.9080.634−0.9310.671−0.9040.643−0.883Р value< 0.001< 0.0010.001< 0.001< 0.0010.001q value0.0002*0.0002*0.0011*0.0011*0.0002*0.0011*Cut-off value7.28100.6416.9334.9180.1930.56Sensitivity (95 % CI)86.67 (62.12−97.63)86.96 (67.87−95.46)84.62 (57.77−97.27)78.57 (52.41−92.43)73.91 (53.53−87.45)75.00 (50.50−89.82)Specificity (95% CI)76.71 (65.83−84.92)63.40 (55.23−70.62)62.67 (51.35; 72.74)74.32 (63.35−82.90)74.51 (67.06−80.76)70.83 (59.49−80.06)Likelihood ratio3.722.382.273.062.902.57Note 1. AUC − area under the curve; 95% CI AUC − 95% confidence interval for the area under the curve; q-value − FDR-adjusted p-value calculated using the Benjamini–Hochberg false discovery rate correction.Note 2. * − a statistically significant result after FDR correction (q < 0.05). Cytokine Levels Associated With MIS-C in Children Note 1. AUC − area under the curve; 95% CI AUC − 95% confidence interval for the area under the curve; q-value − FDR-adjusted p-value calculated using the Benjamini–Hochberg false discovery rate correction. Note 2. * − a statistically significant result after FDR correction (q < 0.05). Receiver operating characteristic (ROC) analysis of cytokines for identifying severe COVID-19 in children. Notes. The red diagonal lines represent random classification; red dots indicate optimal cutoff points Receiver operating characteristic (ROC) analysis of cytokines for differentiating MIS-C from acute COVID-19. Notes. The red diagonal lines represent random classification; red dots indicate optimal cutoff points In particular, the cutoff value of IL-1β for MIS-C was 7.28 pg/mL, which exceeded the corresponding value for severe COVID-19 (6.96 pg/mL). Sensitivity (86.67%) and specificity (76.71%) of the IL-1β cutoff for MIS-C were also higher compared to severe COVID-19 (64.29% and 68.92%, respectively). An IL-6 level exceeding 68.37 pg/mL demonstrated a positive likelihood ratio of 2.08 for severe COVID-19 within the analyzed cohort, with a sensitivity of 95.24% and specificity of 54.19%. For differentiating MIS-C from acute COVID-19, the IL-6 cutoff was >100.64 pg/mL (sensitivity, 86.96%; specificity, 63.40%). Practically identical cutoff values of IL-8 were determined for severe COVID-19 (17.18 pg/mL) and MIS-C (16.93 pg/mL). A similar trend was observed for IL-12: cutoff for severe COVID-19 was 34.53 pg/mL (sensitivity 75%, specificity 71.05%), while for MIS-C it was 34.91 pg/mL (sensitivity 78.57%, specificity 74.32%). Particular attention should be paid to TNF-α, the diagnostic cutoff for which in MIS-C (80.19 pg/mL) exceeded that for severe COVID-19 (60.63 pg/mL) by 32.26%. Within the interferon response, cutoff values for IFN-α were lower in severe COVID-19 (26.96 pg/mL; sensitivity 61.54%, specificity 76%) compared to MIS-C (30.56 pg/mL; sensitivity 75%, specificity 70.83%) (Tables 7 and 8). 3.6. Machine Learning-Based Severity Prediction for COVID-19 Unlike MIS-C, for which clearly defined clinical criteria exist, the clinical differentiation between mild, moderate, and severe COVID-19 in children remains less distinct. This necessitates the identification of potential immunological markers associated with disease severity, particularly concentrations of proinflammatory cytokines that may play a key role in the pathogenesis of severe disease forms. In this context, a classification model based on the Random Forest machine learning algorithm was developed to stratify the severity of COVID-19 in children. The model was developed as an exploratory supplementary approach to distinguish patients with severe disease (class 1) from those with mild or moderate disease (class 0) based on the concentrations of six cytokines (IL-1β, IL-6, IL-8, IL-12, TNF-α, IFN-α), as well as age and sex. The model was built using 300 decision trees and internally assessed using an 80:20 train/test split. Given the relatively limited number of severe COVID-19 cases, the Random Forest model was considered an exploratory supplementary approach for evaluating cytokine associations with severe disease. The model achieved an overall classification accuracy of 75.0% on the test dataset, indicating acceptable discriminatory performance within the analyzed cohort. ROC curve analysis demonstrated high discriminative ability, with an average AUC value of 0.906 (AUC for class 0 was 0.938, and for class 1 − 0.875) (Figure 6). Additional classification performance metrics included a sensitivity of 75.0%, specificity of 75.0%, balanced accuracy of 75.0%, positive predictive value of 83.3%, and negative predictive value of 83.3%. Confusion matrix analysis demonstrated accurate classification of 75.0% of non-severe COVID-19 cases within the testing subset, whereas 25.0% of severe cases were classified as non-severe, likely reflecting the exploratory nature of the analysis and the limited size of the severe COVID-19 subgroup. Receiver Operating Characteristic (ROC) curve for the Random Forest model classifying COVID-19 severity in children. Notes: The colored lines represent the ROC curves illustrating the model’s ability to classify each severity class. The pink line corresponds to Class 0 (mild and moderate COVID-19), and the teal line corresponds to Class 1 (severe COVID-19), showing the trade-off between the True Positive Rate (TPR) and False Positive Rate (FPR) across various classification thresholds. The black dashed diagonal line represents the performance of a random classifier. The gray circle labeled «Perfect separation» near the coordinate (0.0, 1.0) indicates the ideal performance point, where TPR is 1.0 (perfect sensitivity) and FPR is 0.0 (perfect specificity) The importance of predictors in the Random Forest model was assessed using three metrics: Mean Decrease in Accuracy, Total Increase in Node Purity, and Mean Dropout Loss. The biomarkers with the highest model-derived importance for classification of severe COVID-19 in children were IL-1β and IL-8. Conversely, IL-6 demonstrated the lowest model-derived importance. Sociodemographic variables, including age and sex, showed low model-derived importance compared with cytokine biomarkers in this exploratory Random Forest model (Table 9). Table 9.Feature Importance Evaluation in the Random Forest ModelParameterMean decrease in accuracyTotal increase in node purityMean dropout lossIL-1β0.0200.1070.334IL-80.0050.0440.220TNF-α0.0020.0420.140IL-12-0.0020.0170.028IFN-α0.0040.0150.192IL-6-0.0020.0090.078Sex0.001-0.0040.000Age-8,922x10-4-0.0060.000 Feature Importance Evaluation in the Random Forest Model Given that MIS-C is defined and clinically classified according to established international criteria, including WHO and CDC criteria, separate machine-learning-based modeling of MIS-C severity was beyond the scope of the present study.9,41 MIS-C was therefore analyzed as a distinct SARS-CoV-2-associated hyperinflammatory condition, whereas the Random Forest model was applied as an exploratory supplementary approach to classify severe versus non-severe acute COVID-19. Future studies involving larger MIS-C cohorts with predefined severity categories are warranted to determine whether machine-learning approaches may provide incremental value for MIS-C classification or severity stratification. 4. Discussion The clinical course of SARS-CoV-2 infection in children is predominantly mild or even asymptomatic, consistent with observations reported worldwide.1-3 In our cohort, an age-related trend in disease severity was observed: mild forms of COVID-19 were most common among children under 5 years of age (60.48%), whereas severe cases were more frequently identified in school-aged children (14.71%) and adolescents (23.81%). Epidemiological data indicate that infants under 3 months are the most vulnerable to severe COVID-19, with hospitalization rates being higher among infants under 1 year and adolescents over 15 years old.16,42 These age-dependent differences likely reflect developmental variations in immune maturation and cytokine regulation, which modulate both antiviral defense and susceptibility to hyperinflammatory responses. Characterizing the age-related cytokine response in pediatric COVID-19 and MIS-C is crucial for identifying children at risk of clinical deterioration. Although no age-dependent differences in cytokine levels were observed among healthy children in our cohort, prior studies indicate that cytokine profiles in healthy children are age-dependent. Data from the Canadian Laboratory Initiative on Pediatric Reference Intervals (CALIPER) show that TNF-α levels decline after 8 years of age, IL-1β peaks between 4 and 14 years, and IL-6 exhibits a general tendency to decrease with age.43,44 Other studies report an approximate 7% annual decline in TNF-α concentrations in healthy children. In contrast, antigenic stimulation induces an age-dependent increase in pro-inflammatory cytokine production, with TNF-α and IL-6 rising by 7–8% and 6–19% per year, respectively. 45 These findings suggest progressive maturation of the immune system, with increasing capacity for antigen-specific pro-inflammatory responses, which is limited in infancy. Our results demonstrate age-related variability in cytokine responses in pediatric COVID-19. TNF-α levels were elevated in MIS-C and COVID-19 across all age groups compared to healthy children, indicating excessive pro-inflammatory activation exceeding physiological age-related fluctuations. Significant age-related differences were observed for IL-6 and IL-12. Adolescents exhibited markedly higher IL-6 levels compared with children aged 2–5 years, suggesting an age-related intensification of the immune response to SARS-CoV-2. Similarly, IL-12 concentrations were significantly higher in adolescents compared to younger children with COVID-19; however, children aged 2–5 years with MIS-C did not exhibit a significant increase in IL-12 (p > 0.05). These findings align with prior evidence that monocytes in early childhood have a reduced capacity to express IL-12 after lipopolysaccharide stimulation, reflecting immaturity of the monocytic compartment and insufficient Th1 response activation. 46 Age-related enhancement of pro-inflammatory cytokine production via Toll-like receptor (TLR) signaling, particularly through TLR3, TLR4, and TLR7/8, has also been reported, promoting IL-12, IL-1β, and IL-6 secretion with increasing age.47-49 IL-8, a potent neutrophil chemoattractant, also displays age-dependent regulation. In children under 5 years, TLR-1/2, TLR-4, TLR-5, and TLR-9 stimulation induces higher IL-8 secretion, facilitating rapid neutrophil recruitment and effective early innate defense.48,50 This mechanism declines with age, likely due to increased activity of FoxP3+ regulatory T cells, which suppress excessive innate activation and contribute to the formation of a more controlled immune response. 50 Consistent with this, IL-8 levels in the 6–11-year age group did not differ significantly between COVID-19, MIS-C, and healthy controls, suggesting that a balanced immune response is attained during this developmental period and limiting excessive IL-8 production in response to infectious challenges. Another hallmark of immune immaturity in early childhood is reduced production of IFN-α, a key antiviral cytokine. Although TLR9 stimulation induces IFN-α secretion, levels remain lower than in adults until 12–18 months of age.49,51 Responses to TLR7/8 and TLR9 may reach adult levels by 1 year, whereas TLR3 responses decrease until 2 years and subsequently increase. 52 IFN-α production by peripheral blood mononuclear cells in children aged 12–59 months in response to respiratory viral infections is significantly lower than in adults, reflecting the immaturity of the antiviral response in early childhood. 53 In line with this, IFN-α levels in children with COVID-19 and MIS-C were significantly reduced across all age groups compared to healthy controls, which may reflect suppression of the natural antiviral response in these pathological conditions, regardless of patient age. In addition to age-related differences, sex is an essential determinant of immune responses, influencing the clinical course of infectious diseases in boys and girls. Our observations did not reveal such sex differences, which is likely due to the small sample size, particularly in the MIS-C and healthy control groups. Nevertheless, previous studies have reported that boys with MIS-C have significantly higher levels of pro-inflammatory cytokines (IL-1β, IL-6, IL-12, and TNF-α) compared to girls, suggesting a more pronounced pro-inflammatory response in males. 54 Moreover, healthy girls have been shown to exhibit lower TNF-α levels than boys. 45 Males generally exhibit lower type I interferon levels, which may contribute to an increased susceptibility to viral infections. 28 However, our study did not detect sex differences in IFN-α levels during childhood. Hormonal and genetic factors primarily influence sex differences in the immune response. Females possess two X chromosomes, enabling higher expression of immunoregulatory genes located on the X chromosome, including Toll-like receptors (TLR7, TLR8), transcription factors (FOXP3), chemokine receptors (CXCR3), and CD40L, which is central to immune activation via the CD40L–CD40 pathway. 55 The X chromosome also contains genes encoding interleukins and other biologically active molecules that can affect the intensity of the immune response. 56 A key mechanism underlying heightened female immune activity is the “escape” from X-chromosome inactivation (XCI). Approximately 15% of X-linked genes in females remain transcriptionally active on both alleles, resulting in increased expression of immunoregulatory genes. 55 This is associated with enhanced type I interferon production, a critical antiviral defense component.57,58 XCI of the CD40L gene further contributes to macrophage activation, T-cell stimulation, B-cell-mediated antibody production, and elevated levels of TNF-α, IL-21, and IL-27.55,59 FOXP3 expression, regulated via the IL-2/JAK/STAT pathway, is essential for Treg-mediated immunosuppression and concurrently suppresses NF-κB signaling to prevent excessive T-cell activation. 55 Conversely, males may exhibit reduced immune responses due, in part, to loss of the Y chromosome (LOY) in immune cells, including NK cells, T and B lymphocytes, and monocytes. LOY leads to altered expression of over 500 genes, significantly impairing immune cell function and overall immune competence in males.55,60 Numerous studies have established the pivotal role of interleukins in the pathogenesis of COVID-19, particularly in the development of cytokine imbalance, which underlies severe disease. Although findings vary due to differences in study populations, sample sizes, and methodologies, several immunological markers have consistently shown a correlation with disease severity. These include members of the IL-1 family (IL-1β, IL-1RA, IL-18, IL-33), IL-6, Th1-associated cytokines (IL-2, IL-2R, IL-12, IL-15, IL-27, TNFα, sTNFR1), Th2-associated cytokines (IL-4, IL-5, IL-10, IL-13), Th17-associated cytokines (IL-17A, IL-22), as well as chemokines (IL-8, MCP-1, MCP-3, MIP-1α, MIP-1β, MIP-3β, CXCL9), interferon response proteins (IFN-γ, IP-10), and growth factors (IL-7, TGF-β, GM-CSF, G-CSF).14,61-63 Particular attention has been given to the synergistic effects of IL-6, IL-8, IL-1β, and TNF-α, which act as central mediators of hyperinflammation and may determine the clinical course of SARS-CoV-2 infection. 14 In the pediatric population, analysis of 45 cytokines and chemokines revealed elevated levels of multiple interleukins in both severe COVID-19 and MIS-C, with IL-6, MIG, and IP-10 identified as core components of the cytokine storm. 16 Italian studies similarly reported increases in IL-1β, CD25, IFN-γ, IL-6, and TNF-α in children with COVID-19, although logistic analysis did not identify any of these cytokines as independent predictors of disease severity. 64 Notably, elevations of sIL-2R, IL-1RA, IL-10, IL-18, IL-8, IFN-γ, IL-27, IL-17F, and TNF-α were observed exclusively in MIS-C. 16 Recent studies have further expanded the understanding of hyperinflammatory responses in pediatric SARS-CoV-2-associated disease. Lawrence et al reported significantly elevated levels of type-2 and proinflammatory cytokines, including IL-4, IL-5, IL-6, IL-8, IL-10, IL-13, and IL-33, in children with MIS-C. 65 Similarly, Pidkova et al identified significant associations between soluble inflammatory biomarkers and MIS-C severity, while Brimacombe et al described distinct multivariate immune-response signatures associated with MIS-C.22,23 Damodar et al additionally reported elevated levels of IL-1β, IL-6, IL-8, and TNF-α in children with severe COVID-19 and MIS-C, further supporting the role of these cytokines in hyperinflammatory immune responses. 24 Consistent with these observations, our study demonstrated a statistically significant elevation of IL-1β, IL-6, IL-8, IL-12, and TNF-α in children with SARS-CoV-2 infection compared with healthy controls, further supporting the central role of proinflammatory cytokine activation in the pathogenesis of pediatric SARS-CoV-2-associated disease. Importantly, our analysis identified specific interleukins and exploratory threshold concentrations associated with severe COVID-19 and MIS-C. In addition, our study evaluated cytokine alterations across the full spectrum of pediatric SARS-CoV-2-associated disease severity and included patients from 1 month to 17 years of age, thereby extending current evidence on age-related inflammatory responses in pediatric COVID-19 and MIS-C. Furthermore, unlike studies based on high-dimensional proteomic or multiplex immune profiling platforms, our investigation relied on ELISA-based cytokine assessment, which may facilitate broader clinical applicability of cytokine-based inflammatory profiling, particularly in resource-limited settings. Although several cytokines demonstrated acceptable discriminatory performance in distinguishing severe COVID-19 and MIS-C within the analyzed cohort, the identified threshold concentrations should be interpreted in the context of the observational design of the study and the limited number of patients with severe COVID-19. The corresponding confidence intervals for sensitivity, specificity, and AUC estimates likely reflect the small size of the severe COVID-19 subgroup. Therefore, the reported ROC-derived threshold concentrations should be considered exploratory findings associated with hyperinflammatory manifestations of pediatric SARS-CoV-2-associated disease and require validation in larger independent pediatric cohorts. However, the evaluated cytokines are not specific exclusively to SARS-CoV-2-associated inflammation, as elevated levels of IL-1β, IL-6, IL-8, and TNF-α may also occur in other pediatric inflammatory conditions, including Kawasaki disease and community-acquired pneumonia.66-70 Therefore, these biomarkers should be interpreted as indicators of hyperinflammatory immune activation rather than disease-specific inflammatory markers exclusive to SARS-CoV-2 infection. Notably, the ROC analysis for MIS-C was designed to differentiate MIS-C from acute COVID-19 rather than from healthy controls. This distinction is important for interpreting the identified cytokine thresholds, as the MIS-C analysis reflects differentiation between two SARS-CoV-2-associated clinical conditions rather than simple case-control discrimination. In the exploratory Random Forest analysis, IL-1β, IL-8, TNF-α, and IFN-α showed the highest model-derived importance for classification of severe COVID-19. These findings suggest that markers related to proinflammatory and interferon-mediated immune responses may contribute to discrimination between severe and non-severe disease within the analyzed cohort. However, feature importance measures derived from Random Forest models do not imply statistical significance or causal relevance and may be influenced by correlations among predictors and characteristics of the analyzed dataset. Moreover, the interpretation of machine-learning findings in clinical research should extend beyond aggregate accuracy and feature importance. As emphasized by Tan and Benos, AI-driven diagnostic tools require evaluation of explainability, fairness across subgroups, privacy and security, and clinical accountability, in addition to diagnostic performance. 71 In the present study, the Random Forest model was exploratory and was not intended as a deployable diagnostic tool. Therefore, the low model-derived importance of age and sex should not be interpreted as evidence of absence of subgroup effects. Future validation studies should evaluate model performance across age, sex, and other clinically relevant subgroups before any potential clinical application. These methodological considerations do not diminish the biological relevance of the cytokines identified in the present analysis, but they underscore the need to interpret machine-learning-derived findings together with established immunopathophysiological mechanisms. Therefore, the observed importance of IL-1β, IL-8, TNF-α, and IFN-α should be discussed in the context of their known roles in innate immune activation, neutrophil recruitment, systemic inflammation, and antiviral response during SARS-CoV-2-associated disease. IL-1β is primarily secreted by monocytes, macrophages, and neutrophils and is retained intracellularly until triggered by pro-inflammatory stimuli. In the context of tissue damage characteristic of SARS-CoV-2 infection, increasing hypoxia, reactive oxygen species (ROS), and danger-associated molecular patterns (DAMPs) activate the nuclear factor κB (NF-κB) pathway. This induces transcription and translation of pro-inflammatory genes, including those encoding IL-1α, IL-1β, IL-18, and components of the NLRP3 inflammasome (NACHT, LRR, and PYD domains-containing protein 3).27,72-74 Previous studies have demonstrated a pathogenic link between IL-1β and interleukin 6 (IL-6), as well as TNF-α.25,28 Our findings of elevated IL-1β levels in patients with acute SARS-CoV-2 infection compared to the control group (healthy individuals) are consistent with adult studies, where IL-1β is recognized as a key inflammatory biomarker in COVID-19.75-77 Notably, the threshold values associated with severe disease for IL-1β identified in our pediatric cohort exceed those reported in adults. This discrepancy may reflect the absence of disease severity stratification in previous studies and age-related differences in IL-1β-mediated responses. Adults more frequently present with pulmonary involvement, potentially resulting in predominantly local IL-1β synthesis in respiratory tissues without a corresponding systemic increase, unlike in children. IL-8 also demonstrated a strong association with disease severity in pediatric COVID-19, with concentrations rising significantly from mild to severe disease. In severe cases, IL-8 levels were nearly identical to those observed in MIS-C, indicating pronounced systemic proinflammatory activation. These findings align with previous reports identifying IL-8 as a key marker of both SARS-CoV-2 infection severity and MIS-C.24,78,79 IL-8 is produced by macrophages, as well as epithelial and endothelial cells, and functions as a chemoattractant for neutrophils via binding to G-protein-coupled receptors CXCR1 (CD181) and CXCR2 (CD182). This interaction activates ERK1/2, PI3K, and PLC-γ signaling pathways, leading to neutrophil activation, chemotaxis, lysosomal enzyme release, upregulation of adhesion molecules, increased intracellular calcium, and oxidative stress.12,80,81 Although IL-6 did not emerge as the primary determinant of disease severity among all studied cytokines, its levels were significantly higher in severe COVID-19 cases compared to those with mild and moderate forms. This observation is consistent with current evidence identifying IL-6 hyperproduction as a marker associated with severe COVID-19 in both adults and children, including as an independent predictor of COVID-19-associated pneumonia.5,76,79,82,83 IL-6 is predominantly produced by monocytes and macrophages, with contributions from T and B lymphocytes, as well as fibroblasts and endothelial cells. Its synthesis is triggered via Toll-like receptor activation and further enhanced by IL-1 and TNF-α. 15 Previous studies have shown that IL-6 levels in healthy individuals typically range from 0 to 43.5 pg/mL. 84 Elevated IL-6 disrupts immune balance by causing an imbalance in Th17/Treg and Th1/Th2 responses, promoting Th17 differentiation while inhibiting TGF-β-mediated regulatory T cell (Treg) development, which possess anti-inflammatory properties. 15 Concurrently, IL-6 stimulates the differentiation of naïve CD4+ T cells into Th2 cells, which may exacerbate cytokine production, contribute to the development of a cytokine storm, and suppress Th1-mediated antiviral responses, ultimately reducing interferon activity and impairing cellular clearance of SARS-CoV-2. 15 According to the results of our study, IL-12 levels increased significantly with the severity of COVID-19 and were highest in children with MIS-C. Polish studies report elevated IL-12 (p40) levels in pediatric COVID-19, though no significant difference was observed for IL-12 (p70) between patients and controls. 85 IL-12 is a potent proinflammatory and immunostimulatory cytokine secreted by macrophages, dendritic cells, B lymphocytes, and monocytes.38,76 Viral infections induce the expression of the IL-12 gene, resulting in sustained high levels of IL-12 production not only during the acute phase of disease but also after the cessation of viral replication. 76 This may explain the elevated IL-12 levels observed in MIS-C. Elevated TNF-α levels have been associated with COVID-19 progression, particularly with severe disease and the requirement for intensive care unit admission. 76 Consistently, in our study, higher TNF-α concentrations correlated with disease severity. Activated macrophages primarily produce TNF-α, but lymphocytes, monocytes, neutrophils, and alveolar epithelial cells also contribute to its synthesis. 15 Notably, Curatola et al did not observe a correlation between TNF-α or IL-6 levels and disease severity in pediatric COVID-19, underscoring the need for further investigation into the role of these cytokines in children. 64 Our findings additionally indicate a progressive decrease in interferon responses with increasing severity of SARS-CoV-2 infection, with the most pronounced suppression observed in patients with MIS-C. SARS-CoV-2 is recognized as a weak inducer of type I interferon responses, resulting in low serum interferon levels and poorer prognosis in COVID-19 patients. 19 Contributing factors to reduced IFN-α levels include inborn errors of type I interferon immunity, gene polymorphisms, and the presence of neutralizing autoantibodies against type I interferons.86-89 Research into the role of interferons in COVID-19 pathogenesis is ongoing, particularly as IFN-α is being actively explored as a therapeutic agent in COVID-19 management. 19 However, some studies have shown conflicting results. A meta-analysis evaluating the role of IFN-α in COVID-19 pathogenesis, considering variables such as age, sex, timing of blood collection, and analytical methods, did not find statistically significant differences in IFN-α concentrations between patients with mild and severe COVID-19. 90 These observations underscore the need for further research to clarify the role of interferons in the pathogenesis of SARS-CoV-2 infection. 5. Conclusion Our study, which evaluated associations and exploratory threshold levels of cytokines related to severe COVID-19 and MIS-C, may provide a basis for further refinement of inflammatory profiling in pediatric SARS-CoV-2-associated disease. Expanding laboratory assessment to include cytokine profiling may contribute to a more detailed characterization of severe inflammatory manifestations and support further investigation of risk stratification approaches in children with SARS-CoV-2 infection. However, the additional clinical value of cytokine profiling beyond routine clinical and laboratory markers requires validation in larger independent pediatric cohorts. 6. Limitations This study has several limitations that should be considered when interpreting the findings. The assessment was restricted to a panel of six cytokines. Although these markers were selected based on their relevance to SARS-CoV-2-associated inflammation, evaluation of a broader spectrum of cytokines, chemokines, and other immune mediators could provide a more comprehensive characterization of immune dysregulation in pediatric COVID-19 and MIS-C. The relatively small number of patients with severe COVID-19 may have affected the stability and generalizability of subgroup analyses, ROC-derived threshold estimates, and machine-learning results. The ROC-derived threshold concentrations were generated and evaluated within the same study cohort and should therefore be considered exploratory findings requiring validation in independent pediatric populations before potential clinical implementation. Some subgroup and covariate analyses were limited by the structure of the study cohort. Age- and sex-stratified analyses within the MIS-C group were constrained by the relatively limited number of MIS-C patients. In addition, infants and toddlers were not represented among controls because informed consent could not be obtained from parents or legal guardians of healthy children aged from 1 month to 3 years. Therefore, the control group was not fully age-matched to the COVID-19 and MIS-C cohorts, which should be considered when interpreting comparisons between patients and controls. Detailed data on possible coinfections were also not consistently available for all participants; therefore, coinfection status could not be included as a covariate in the adjusted regression models. The study did not include comparator cohorts with non-SARS-CoV-2 inflammatory diseases, such as Kawasaki disease or community-acquired pneumonia of non-SARS-CoV-2 etiology. Consequently, the specificity and differential diagnostic utility of the evaluated cytokines could not be assessed. The observed cytokine alterations should therefore be interpreted as markers of hyperinflammatory immune activation associated with SARS-CoV-2 infection rather than disease-specific diagnostic biomarkers. Finally, the Random Forest model should be interpreted with caution. Because of the limited number of severe COVID-19 cases, model performance estimates may be unstable and potentially optimistic. Although internal train/test splitting was used, no external validation cohort was available, and stratified cross-validation, bootstrapping, and calibration analyses were not performed. Feature importance estimates derived from the Random Forest model should not be interpreted as measures of statistical significance or causal relevance. In addition, the model was not evaluated for explainability, fairness across clinically relevant subgroups, privacy-related implementation issues, or clinical accountability. Therefore, the machine-learning findings should be considered exploratory and require validation in larger independent pediatric cohorts before any potential clinical translation. Footnotes Author Contributions: KK: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Visualization, Writing – original draft; Writing – review & editing; HP: Conceptualization, Methodology, Investigation, Resources, Supervision, Project administration, Writing – review & editing. Funding: The authors received no financial support for the research, authorship, and/or publication of this article. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ORCID iD Kateryna Kozak https://orcid.org/0000-0002-5328-4647 Ethical Considerations The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Bioethics Committee of I. Horbachevsky Ternopil National Medical University of the Ministry of Health of Ukraine (Protocol No. 71, dated 25 October 2022). Consent to Participate Written informed consent was obtained from the parents or legal guardians of all participating children. Data Availability Statement The datasets generated and analyzed during the current study are not publicly available due to ethical considerations related to pediatric participant confidentiality. De-identified data and additional analytical details related to the machine-learning analysis may be made available from the corresponding author upon reasonable request and subject to appropriate ethical approval.* References Associated Data Data Availability Statement The datasets generated and analyzed during the current study are not publicly available due to ethical considerations related to pediatric participant confidentiality. De-identified data and additional analytical details related to the machine-learning analysis may be made available from the corresponding author upon reasonable request and subject to appropriate ethical approval.*