Evaluating the risk of elevated IL-6 levels in pediatric patients: a stratified approach and evidence of interaction effects

Abstract

Background

Distinguishing between an appropriate and a pathologically dysregulated elevation of interleukin-6 (IL-6) is challenging. This study aimed to explore how physiological function abnormalities could help identify subpopulations in whom IL-6 levels were associated with disease severity.

Methods

This retrospective study included pediatric patients with serum IL-6 test results. After selecting laboratory parameters that effectively identify a subpopulation in which IL-6 levels can predict adverse clinical outcomes, the chosen parameters were categorized into function groups. Using IL-6 levels and the number of impaired functions as stratifying parameters, the risk of adverse outcomes for each stratum was calculated. To assess the interaction effect between functional impairment and IL-6 elevation on adverse outcomes, the relative excess risk due to interaction (RERI) was calculated.

Results

A total of 56,170 pediatric patients were included. Fourteen laboratory parameters in pediatric patients could be used as stratifying parameters, where abnormal values suggest that IL-6 levels could be used as a prognostic marker [area under the receiver operating characteristic curves (AUCs) >0.7]. These parameters were categorized into five functional groups: immunological [white blood cell (WBC), neutrophils, monocytes], hepatic [aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bile acids], renal (creatinine, uric acid), coagulation [platelets, activated partial thromboplastin time (APTT), thrombin time], and myocardial enzymes [myoglobin, creatine kinase (CK), CK-MB]. For the rate of adverse outcomes, the rate difference (RD) between the IL-6 >85^th and ≤85^th percentile groups ranged from 0.7% [95% confidence interval (CI): 0.2%, 1.2%] in the stratum with 1 impaired function to 16.9% (95% CI: 12.5%, 21.3%) in the stratum with ≥4 impaired functions. Similar trends were observed for the rate of pediatric intensive care unit (PICU) admission and the length of hospital stays. For adverse outcomes, the RERI between the increased number of impaired functions and IL-6 levels increased from 1.06 (1 impaired function + IL-6 >85^th percentile; 95% CI: 0.73, 1.38) to 42.79 (4 impaired functions + IL-6 >85^th percentile; 95% CI: 34.49, 51.09). Consistent results were observed in the validation cohort and in disease-specific subpopulations.

Conclusions

The risk associated with elevated IL-6 levels increases as the number of impaired functions grows, with an intensified interaction effect between IL-6 elevation and functional impairment on clinical outcomes, offering a simple method for clinical risk stratification.

Highlight box.

Key findings

• Interleukin-6 (IL-6) >50 pg/mL (85th percentile) correlates with higher adverse outcomes, pediatric intensive care unit admission, and longer hospital stay in pediatric patients.

• Fourteen lab markers across five physiological systems (immune, hepatic, renal, coagulation, myocardial) identify high-risk subgroups where IL-6 predicts poor outcomes.

• Adverse event risk rises with the number of impaired functions; elevated IL-6 confers minimal risk with intact functions but marked risk with ≥4 impaired functions.

• Significant positive interaction exists between IL-6 elevation and functional impairment, intensifying with more impaired functions.

What is known and what is new?

• IL-6 is associated with disease severity, but the implications of elevated IL-6 levels remain difficult to determine.

• Assessment of physiological function impairment may improve risk stratification based on IL-6. Combined evaluation of IL-6 and functional impairment outperforms single biomarker assessment.

What is the implication, and what should change now?

• Use routine lab parameters to count impaired functions for IL-6 risk stratification.

• Prioritize intensive monitoring and intervention for children with elevated IL-6 plus multiple organ dysfunction.

Introduction

Interleukin-6 (IL-6) is a pleiotropic cytokine that regulates immune response, hematopoiesis, the acute phase response, and inflammation (1,2). It is produced by various types of cells, such as T-cells, B-cells, monocytes, fibroblasts, keratinocytes, endothelial cells, mesangial cells, adipocytes, and some tumor cells. Consistent with its multifunctional nature, IL-6 has been found to play pathological roles in various diseases, including inflammatory, autoimmune, and malignant conditions (3-5). The serum concentrations of IL-6 in healthy individuals are low, typically around 1–5 pg/mL (6). Elevated IL-6 levels have been observed in many diseases and are linked with disease severity and adverse outcomes, such as in sepsis (7-9), coronavirus disease 2019 (COVID-19) (10), injury (11), and breast cancer (12).

However, the interpretation of IL-6 concentrations in the blood is not straightforward (13,14), as distinguishing between appropriate and pathologically dysregulated elevations of IL-6 is challenging (9,14). The elevation of any cytokine could be a sign of disease deterioration, but it could also be an indicator of a vigorous inflammatory response, not necessarily linked to pathogenesis (13,15). Therefore, it has been suggested that evaluating cytokine balance is a superior predictor of outcomes than measuring IL-6 or other cytokines alone (13,14).

In this study, we propose a novel approach to evaluate the risk associated with IL-6 elevation based on the physiological functions of individuals. We hypothesize that the status of functional impairment can be used to identify a subpopulation in which IL-6 levels are associated with disease severity. The hypothesis consists of three parts: (I) Among those with impaired physiological functions, an elevation of IL-6 is associated with a higher risk of adverse outcomes; (II) the degree of functional impairment is positively associated with the strength of the relationship between IL-6 levels and disease severity; and (III) there is an interaction effect between functional impairment and IL-6 levels on disease severity, meaning the combined effect of these two factors is stronger than the sum of their individual effects. These hypotheses are based on the following reasoning: Among those with relatively intact physiological functions, which indicate a relatively equilibrium status, an elevation of IL-6 is likely to under a balanced feedback regulation, therefore could be interpreted as a tolerable inflammatory response. Among individuals with impaired functions, the cause of the function impairment could also be the cause of IL-6 elevation. Functional impairment indicates a state of disrupted equilibrium, where IL-6 levels may be associated with disease severity. Due to the cytokine cascade, elevated IL-6 could further induce or amplify cytokine storms, leading to tissue and organ damage (14). The organ damage and elevated IL-6 levels indicate a systemic imbalance; therefore, the associated risk is greater than the effect of each factor alone.

To identify a subpopulation in which IL-6 levels could be used as an indicator of disease severity and to verify our hypotheses, this study: (I) investigated the distribution of adverse outcomes across different IL-6 levels and identified a threshold for IL-6 stratification; (II) explored associations between IL-6 levels and adverse outcomes in subpopulations stratified by common laboratory parameters, and selected those laboratory parameters that could effectively identify a subpopulation in which IL-6 levels can predict adverse clinical outcomes; (III) categorized the selected laboratory parameters into different physiological function groups and calculated the number of impaired functions; (IV) conducted a stratified analysis according to IL-6 levels and the number of impaired functions, and calculated the risk of adverse outcomes for each stratum; and (V) estimated the interaction effect between functional impairment and IL-6 elevation on adverse outcomes. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0470/rc).

Methods

Study population

This study included pediatric patients hospitalized at Hunan Children’s Hospital from January 1, 2018, to August 8, 2023. Patients admitted between January 2018 and December 2022 were included in the primary analysis set, while those admitted between January 2023 and August 2023 were included in the validation set. The inclusion criteria were as follows: at least one serum IL-6 test performed and age ≤18 years. The exclusion criteria were unclear hospital discharge status or missing essential laboratory test results. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of the Hunan Children’s Hospital (No. HCHLL-2024-325). Informed consent was waived by the Medical Ethics Committee because of the retrospective design. The authors had no access to information that could identify individual participants during and after data collection.

Variables

The first IL-6 test results and the nearest test results for other potential stratifying laboratory parameters were extracted for analysis. This approach was designed to reflect real-world clinical practice, in which IL-6 is interpreted together with contemporaneous indicators of organ function to support integrated risk assessment. Potential stratifying laboratory parameters were selected from those commonly tested in complete blood cell counts, liver and renal function tests, coagulation function tests, and myocardial enzyme tests. A total of 36 parameters were examined, including: white blood cell (WBC) count, red blood cell count, platelet count, neutrophil count, monocyte count, lymphocyte count, lymphocyte percentage, eosinophil count, basophil count, hemoglobin, hematocrit, plateletcrit, platelet distribution width, aspartate aminotransferase (AST), alanine aminotransferase (ALT), globulin, total bilirubin, direct bilirubin, indirect bilirubin, total bile acids, creatinine, uric acid, thrombin time (TT), prothrombin time, international normalized ratio, activated partial thromboplastin time (APTT), fibrinogen, fibrinogen degradation products, D-dimer, antithrombin III, lactate dehydrogenase, creatine kinase (CK), CK-MB, myoglobin, procalcitonin, and C-reactive protein.

The primary outcome was a composite endpoint of adverse outcome, defined as in-hospital mortality or discharge status of no improvement. In the hospital information system, discharge outcomes were routinely recorded as four categories: recovery, improvement, no improvement, and death. For the purpose of this study, recovery and improvement were classified as favorable outcomes, whereas no improvement and death were classified as adverse outcomes. This grouping strategy was used to capture clinically relevant poor outcomes in hospitalized children, where both failures to improve and in-hospital death represent unfavorable disease trajectories. The secondary outcome variables were pediatric intensive care unit (PICU) admission during hospitalization and length of hospital stay. Demographic, clinical, and laboratory data were extracted from electronic medical records.

General statistical analysis

Categorical variables were presented as absolute values and percentages (%). Continuous variables were presented as medians with quartile 1 (Q1) and quartile 3 (Q3). The Wilcoxon rank-sum test or Chi-squared test was used for between-group comparisons. Trend analysis was conducted using the Cochran-Armitage test or Spearman’s rank correlation, as appropriate. All tests were two-tailed, with a type 1 error rate set at 5%. Missing data were not imputed. Statistical analyses were performed using SAS (version 9.4, SAS Institute). Figures were generated using GraphPad Prism 8 (GraphPad Software, San Diego, CA, USA).

Screening parameters for stratification

To explore the association between IL-6 levels and adverse outcomes, IL-6 was categorized into percentile-based groups using 5^th-percentile intervals. Given the lack of a universally accepted clinical cutoff for IL-6 in pediatric populations and substantial inter-assay variability, no predefined threshold was applied. This approach was used to explore potential non-linear associations between IL-6 levels and outcomes. Adverse outcome rates were calculated across percentile strata, and thresholds for further stratified analysis were determined based on observed changes in the trend of outcome risk across adjacent percentile categories.

To examine whether the ability of IL-6 to discriminate adverse outcomes improves in specific subpopulations stratified by common laboratory parameters, nine stratified analyses were conducted for each parameter using thresholds at every 10^th percentile from the 10^th to the 90^th percentile. Patients were stratified into lower and higher percentile groups, and the area under the receiver operating characteristic curve (AUC) and 95% confidence interval (CI) for IL-6 in discriminating adverse outcomes in each subpopulation was calculated. An AUC greater than 0.7 demonstrates a clinically valuable discriminative ability and was selected for subsequent analysis. To determine whether the normal reference range of selected laboratory parameters could be used as thresholds to stratify patients where IL-6 had good discriminative ability, a further stratified analysis was conducted, and the AUCs in the lower and higher percentile subpopulations were compared.

Stratified analysis

The stratifying laboratory parameters were then grouped according to physiological functions, and for each patient, the number of impaired functions was calculated by summing the function types with abnormal values. The study population was stratified by the number of impaired functions and IL-6 levels. The incidences of adverse clinical outcomes and PICU admission in each stratum were calculated. Within each stratum defined by the number of impaired functions, outcomes between the lower and higher IL-6 subgroups were compared, and the rate difference (RD) and 95% CI were calculated. The length of hospital stay was compared between the lower and higher IL-6 groups within each function impairment stratum.

Interaction effect

Using the stratum without impaired functions and lower IL-6 levels as the reference group, the odds ratio (OR) and 95% CI for adverse outcomes in each stratum were calculated using a logistic regression model, adjusting for age and sex (16). The relative excess risk due to interaction (RERI) and 95% CI were calculated to estimate the additive interaction effect between impaired physiological functions and elevated IL-6 levels on adverse outcomes, adjusting for age and sex (17). An RERI >0 and a lower 95% confidence limit >0 demonstrate a significant and positive interaction effect.

Sensitivity analysis

To assess whether the developed method of risk stratification is valid among patients with certain diseases, we conducted a sensitivity analysis among derived subpopulations diagnosed with pneumonia, sepsis, cytopenias or lymphopenia, autoimmune diseases, leukemia, and hemophagocytic lymphohistiocytosis, which are common conditions where IL-6 measurements are relevant. Both the universal threshold identified in the main analysis for IL-6 stratification and disease-specific thresholds were used for stratification.

Results

Study population and IL-6 levels

The primary analysis set included 52,095 pediatric patients with IL-6 test results, and the validation set included 4,075 patients. The study flowchart is shown in Figure 1. The rates of adverse outcomes and PICU admissions were calculated for each 5^th percentile of IL-6 levels (Figure 2A). Both rates increased gradually from the lower percentile groups and rose sharply after an inflection point. For the rate of adverse outcomes, the inflection point was around the 90^th percentile. From the 0^th to the 90^th percentile, the rate of adverse outcomes increased from 0.8% to 1.7%, and after the 90^th percentile, the rate rose sharply to 23.8%. Similarly, for PICU admissions, the inflection point was at the 85^th percentile. Before the 85^th percentile, the rate fluctuated below 10%, but in the 85^th to 100^th percentile groups, the rate increased sharply from 11.9% to 58.2%.

Figure 1.

Study flow chart. IL-6, interleukin-6.

Figure 2.

Association between serum IL-6 levels and adverse outcomes. (A) The rate of adverse outcomes and PICU admission by IL-6 percentiles. (B) ROCs for IL-6 to discriminate adverse outcomes and PICU admission. (C-P) AUCs and 95% CI for IL-6 to discriminate adverse outcomes among subpopulations stratified by laboratory parameters, using percentiles as thresholds. The reference line at an AUC of 0.7 indicates clinically valuable discriminative ability. Adverse outcomes refer to a hospital discharge status of death or no improvement. ALT, alanine aminotransferase; APTT, activated partial thromboplastin time; AST, aspartate aminotransferase; AUC, area under the receiver operating characteristic curve; CI, confidence interval; CK, creatine kinase; IL-6, interleukin-6; PICU, pediatric intensive care unit; ROC, receiver operating characteristic; WBC, white blood cell.

Based on the inflection points for adverse outcomes and PICU admissions, the 85^th percentile value (50 pg/mL) was selected as the threshold for IL-6. The demographic characteristics and clinical outcomes of the two groups are shown in Table 1. Significant but small differences were observed between the two groups in age (median: 2 years for the ≤85^th group and 2.25 years for the >85^th group, P=0.004) and sex (male: 62.2% for the ≤85^th group and 60.4% for the >85^th group, P=0.002). Compared with the ≤85^th group, the >85^th percentile group showed higher rates of adverse outcomes (3.2% vs. 1.1%, P<0.001), PICU admissions (17.2% vs. 7.0%, P<0.001), and longer hospital stays (median: 8 vs. 7 days, P<0.001).

Table 1. Characteristics and clinical outcomes of the study population according to IL-6 level.

Characteristics/outcomes	All (n=52,095)	IL-6 level
		≤85th percentile† (n=44,332)	>85th percentile† (n=7,763)	P
Age (years)	2.1 [0.7, 5.0]	2 [0.7, 5.0]	2.25 [0.8, 4.9]	0.004
Sex
Female	19,816 (38.0)	16,740 (37.8)	3,076 (39.6)	0.002
Male	32,274 (62.0)	27,587 (62.2)	4,687 (60.4)
Adverse outcomes‡
Yes	718 (1.4)	471 (1.1)	247 (3.2)	<0.001
No	51,377 (98.6)	43,861 (98.9)	7,516 (96.8)
PICU admission
Yes	4,425 (8.5)	3,090 (7.0)	1,335 (17.2)	<0.001
No	47,670 (91.5)	41,242 (93.0)	6,428 (82.8)
Length of hospital stay (days)	7 [5, 10]	7 [5, 10]	8 [6, 14]	<0.001

Data are presented as median [Q1, Q3] or n (%). ^†, the 85^th percentile value was 50 pg/mL. ^‡, adverse outcomes refer to a hospital discharge status of death or no improvement. IL-6, interleukin-6; PICU, pediatric intensive care unit; Q1, quartile 1; Q3, quartile 3.

Screening for stratifying laboratory parameters

The AUC for IL-6 in discriminating adverse outcomes and PICU admissions was 0.637 and 0.644, respectively (Figure 2B). The AUCs for IL-6 in discriminating adverse outcomes in subpopulations stratified by percentile thresholds of common laboratory parameters are shown in table available at https://cdn.amegroups.cn/static/public/tp-2026-0470-1.docx. Fourteen parameters were able to stratify subpopulations for which IL-6 had AUCs >0.7, including WBC, neutrophils, monocytes, AST, ALT, total bile acids, creatinine, uric acid, platelets, APTT, thrombin time, myoglobin, CK, and CK-MB (Figure 2C-2P). For WBC, neutrophils, monocytes, and platelets, IL-6 showed greater ability to identify adverse outcomes in the lower percentile subpopulations. For other parameters, the AUCs were higher in the higher percentile subpopulations compared to the lower percentile subpopulations, and increased with higher percentile thresholds. The AUCs in subpopulations stratified by local laboratory reference cut-offs as thresholds are shown in Figure 3. For most of the selected parameters, IL-6 demonstrated better ability to discriminate high-risk patients in subpopulations with abnormal values compared to those with normal values; AUCs in the abnormal subpopulation were greater than 0.7. Using the upper limit of the normal reference range for AST (40 IU/L), ALT (40 IU/L), total bile acids (9.67 µmol/L) or creatinine (33 µmol/L) as stratifying thresholds showed AUCs <0.7. Instead, the 90^th percentile thresholds (66 IU/L for AST, 56 IU/L for ALT, 20 µmol/L for total bile acids, and 42 µmol/L for creatinine) were used, successfully stratifying subpopulations where IL-6 showed AUCs >0.7 in discriminating patients with adverse outcomes (Figure 3).

Figure 3.

ROCs for IL-6 to discriminate adverse outcomes among subpopulations stratified by common laboratory parameters. The lower or upper limits of the normal reference range were used as thresholds, except for AST, ALT, total bile acids, and creatinine, where the 90th percentile thresholds were used. Adverse outcomes refer to a hospital discharge status of death or no improvement. ALT, alanine aminotransferase; APTT, activated partial thromboplastin time; AST, aspartate aminotransferase; AUC, area under the receiver operating characteristic curve; CK, creatine kinase; IL-6, interleukin-6; ROC, receiver operating characteristic; WBC, white blood cell.

Stratified analysis

According to physiological function, the stratifying parameters were categorized into five function groups: immunological (WBC, neutrophils, monocytes), hepatic (AST, ALT, total bile acids), renal (creatinine, uric acid), coagulation (platelets, APTT, thrombin time), and myocardial enzymes (myoglobin, CK, CK-MB). For each patient, the number of impaired functions was calculated. The study population was stratified by the number of impaired functions (0, 1, 2, 3, ≥4) and IL-6 levels (≤85^th percentile and >85^th percentile). Due to the small number of patients with 4 or 5 impaired functions, they were combined into one category, resulting in a total of eight subgroups. The incidences of adverse outcomes in each subgroup are shown in Table 2. In both IL-6 strata, the incidence of adverse outcomes increases as the number of impaired functions grows (all P values for trend <0.001). Within each function impairment stratum, the IL-6 >85^th percentile group showed a higher rate of adverse outcomes (all P values <0.001), except for the rate of adverse outcomes in the stratum without function impairment (P=0.68). For the rate of adverse outcomes, the RD between the higher and lower IL-6 groups increased from 0.7% (95% CI: 0.2%, 1.2%) in the stratum with 1 impaired function to 16.9% (95% CI: 12.5%, 21.3%) in the stratum with ≥4 impaired functions. Similar trends were observed for the rate of PICU admission, with the RD between the higher and lower IL-6 groups ranging from 2.8% (95% CI: 1.7%, 3.9%) in the stratum without impaired functions to 32.2% (95% CI: 26.5%, 37.9%) in the stratum with ≥4 impaired functions.

Table 2. Stratified analysis of hospital discharge status and PICU admission by the number of impaired functions and IL-6 levels.

Outcomes	No. of impaired functions†	IL-6 level					P	RD (95% CI)
		≤85th percentile			>85th percentile
		N	Mortality, n (%)		N	Mortality, n (%)
Adverse outcomes‡	0	14,811	84 (0.6)		2,354	15 (0.6)	0.68	0 (−0.3, 0.3)
	1	16,809	147 (0.9)		2,706	42 (1.6)	0.001	0.7 (0.2, 1.2)
	2	8,854	115 (1.3)		1,522	41 (2.7)	<0.001	1.4 (0.6, 2.2)
	3	3,009	73 (2.4)		785	58 (7.4)	<0.001	5 (3.1, 6.9)
	4	849	52 (6.1)		396	91 (23.0)	<0.001	16.9 (12.5, 21.3)
PICU admission	0	14,811	574 (3.9)		2,354	157 (6.7)	<0.001	2.8 (1.7, 3.9)
	1	16,809	970 (5.8)		2,706	319 (11.8)	<0.001	6 (4.7, 7.3)
	2	8,854	834 (9.4)		1,522	332 (21.8)	<0.001	12.4 (10.2, 14.6)
	3	3,009	471 (15.7)		785	287 (36.6)	<0.001	20.9 (17.3, 24.5)
	4	849	241 (28.4)		396	240 (60.6)	<0.001	32.2 (26.5, 37.9)

^†, the number of impaired functions was calculated based on five function groups: immunological (WBC, neutrophils, monocytes), hepatic (AST, ALT, total bile acids), kidney (creatinine, uric acid), coagulation (platelets, APTT, thrombin time), and myocardial enzymes (myoglobin, CK, CK-MB). ^‡, adverse outcomes refer to a hospital discharge status of death or no improvement. ALT, alanine aminotransferase; APTT, activated partial thromboplastin time; AST, aspartate aminotransferase; CI, confidence interval; CK, creatine kinase; IL-6, interleukin-6; No., number; PICU, pediatric intensive care unit; RD, rate difference; WBC, white blood cell.

The results of the stratified analysis on the length of hospital stays are shown in Table S1. Among those with IL-6 levels ≤85^th percentile, the median length of hospital stay was 7 to 8 days across the 0 to ≥4 function impairment strata (Spearman correlation coefficient, r_s=0.06, P<0.001). In contrast, among those with IL-6 levels >85^th percentile, the median length of hospital stays gradually increased from 7 days (Q1–Q3, 6–10 days) in the stratum without functional impairment to 17 days (Q1–Q3, 10–29 days) in the stratum with ≥4 impaired functions (r_s=0.27, P<0.001). Comparisons between the IL-6 ≤85^th and >85^th groups across all function impairment strata showed significantly longer hospital stays in the >85^th group (all P values <0.001).

Interaction effect

Table 3 shows the results of logistic regression and RERI estimation. For the risk of adverse outcomes, the adjusted RERI between the increased number of impaired functions and IL-6 levels increased from 1.06 (RERI_{1 impaired function + IL-6 >85th}, 95% CI: 0.73, 1.38) to 42.79 (RERI_{4 impaired functions + IL-6 >85th}, 95% CI: 34.49, 51.09). For the outcome of PICU admission, the adjusted RERI between the number of impaired functions and IL-6 levels increased from 1.01 (RERI_{1 impaired function + IL-6 >85th}, 95% CI: 0.87, 1.15) to 42.79 (RERI_{4 impaired functions + IL-6 >85th}, 95% CI: 27.4, 32.09).

Table 3. Interaction effect of the number of impaired functions and IL-6 levels on clinical outcomes.

Outcomes	No. of impaired functions†	IL-6 level					RERI (95% CI)‡
		≤85th percentile			>85th percentile
		OR (95% CI)‡	P		OR (95% CI)‡	P
Adverse outcomes§	0	1 (reference)			1.14 (0.66, 1.98)	0.65
	1	1.57 (1.2, 2.05)	0.001		2.8 (1.93, 4.07)	<0.001	1.06 (0.73, 1.38)
	2	2.34 (1.76, 3.1)	<0.001		4.89 (3.35, 7.14)	<0.001	2.38 (1.87, 2.89)
	3	4.41 (3.21, 6.06)	<0.001		14.1 (10, 19.9)	<0.001	9.82 (8.02, 11.62)
	4	11.55 (8.1, 16.48)	<0.001		52.64 (38.25, 72.45)	<0.001	42.79 (34.49, 51.09)
PICU admission	0	1 (reference)			1.76 (1.47, 2.12)	<0.001
	1	1.53 (1.37, 1.7)	<0.001		3.32 (2.88, 3.83)	<0.001	1.01 (0.87, 1.15)
	2	2.6 (2.33, 2.9)	<0.001		7.12 (6.14, 8.25)	<0.001	3.68 (3.4, 3.96)
	3	4.72 (4.14, 5.37)	<0.001		14.88 (12.57, 17.6)	<0.001	8.98 (8.29, 9.67)
	4	10.31 (8.68, 12.24)	<0.001		40.15 (32.25, 49.99)	<0.001	29.74 (27.4, 32.09)

^†, the number of impaired functions was calculated based on five function groups: immunological (WBC, neutrophils, monocytes), hepatic (AST, ALT, total bile acids), kidney (creatinine, uric acid), coagulation (platelets, APTT, thrombin time), and myocardial enzymes (myoglobin, CK, CK-MB). ^‡, the ORs and RERIs were adjusted for age and sex. ^§, adverse outcomes refer to a hospital discharge status of death or no improvement. ALT, alanine aminotransferase; APTT, activated partial thromboplastin time; AST, aspartate aminotransferase; CI, confidence interval; CK, creatine kinase; IL-6, interleukin-6; No., number; OR, odds ratio; PICU, pediatric intensive care unit; RERI, relative excess risk due to interaction; WBC, white blood cell.

Validation

The results of the validation for the stratified analysis on PICU admission and length of hospital stay are displayed in Tables S2-S4. Due to the relatively small sample size in the validation set, the number of patients with adverse outcomes was low, leading to a lack of statistical power for stratified analysis; therefore, this outcome was not analyzed. The validation analysis showed similar trends for PICU admission and hospital stays across strata, consistent with the results of the primary analysis (Tables S2-S4).

Sensitivity analysis

Among the study population, the 85^th percentile value of IL-6 was 50 pg/mL, and the 85^th percentile value varied across subpopulations with different diseases: pneumonia (45.58 pg/mL), sepsis (318.13 pg/mL), cytopenias or lymphopenia (42.93 pg/mL), autoimmune diseases (198.9 pg/mL), leukemia (128.5 pg/mL), and hemophagocytic lymphohistiocytosis (129.13 pg/mL). The results of the stratified analysis for the risk of PICU admission showed that, using either the threshold of 50 pg/mL or the disease-specific thresholds, the risk associated with elevated IL-6 levels increased as the number of impaired functions grew, although the absolute risk varied depending on the threshold used (Table S5).

Discussion

This study identified fourteen laboratory parameters in pediatric patients where abnormal values suggest that IL-6 levels could be used to evaluate disease severity. These parameters were categorized into five functional groups: immunological (WBC, neutrophils, monocytes), hepatic (AST, ALT, total bile acids), renal (creatinine, uric acid), coagulation (platelets, APTT, thrombin time), and myocardial enzymes (myoglobin, CK, CK-MB). Using IL-6 levels and the number of impaired functions as stratifying parameters, the stratified analysis showed that the incidence of adverse outcomes increases as the number of impaired functions grows; within each function impairment stratum, the subpopulation with IL-6 levels above the 85^th percentile had a higher rate of adverse outcomes, increased PICU admissions, and longer hospital stays. Interaction effects between elevated IL-6 levels and functional impairment on clinical outcomes were also detected, and this effect intensified with the number of impaired functions.

IL-6 was found to be correlate with disease severity in many disease conditions, like infection (18), inflammation (19), trauma (20,21), autoimmune disease (22,23), and cancer (12,24). For example, in COVID-19, IL-6 serum levels were found to be able to predict severity (10,25). Meta-analysis showed that the mean serum IL-6 in the severe and non-severe COVID-19 group was 56.8 (range, 41.4 to 72.3 pg/mL) and 17.3 pg/mL (range, 13.5 to 21.1 pg/mL), respectively (18). As one of the earliest and most important mediator of the physiologic short-term phase reaction to injury, early IL-6 response was correlated with injury severity, Multiple Organ Dysfunction Score (MODS) score, MODS development, ventilator days, intensive care unit length of stay, hospital length of stay, and mortality (11). In sepsis, IL-6 has been proposed as both diagnostic and prognostic marker (7,9,26,27). Although the correlation between IL-6 elevation and disease severity has long been recognized, how to interpret the IL-6 value in a given patient is not that straightforward. Because an elevation in IL-6 could be a vigorous response to injury, inflammation, or infection, not necessarily linked to disease deterioration, the clinical implication of IL-6 needs to identify circumstances in which the elevation is harmful. Studies have proposed methods for using a combination of IL-6 and other indicators to better evaluate disease severity. For example, the combination of IL-6 and IL-10 has been used to predict the prognosis of invasive breast cancer and COVID-19 patients (28,29), and a combined score including IL-6, IL-8, and IL-10 has been used to predict mortality in severe sepsis (30). Although promising results have been published, most of these methods require additional tests. We wondered whether readily available information, such as the results of commonly tested laboratory parameters, could help in interpreting the risk associated with elevated IL-6 levels.

In spite of the different causes of IL-6 elevation, a systemic approach based on the equilibrium between IL-6 levels and organ damage provides opportunities for cross-disease interpretation. We first hypothesized that stratifying a subpopulation based on physiological function impairment status could identify individuals for whom IL-6 levels are strongly associated with the risk of adverse outcomes. To test this hypothesis, we explored associations between IL-6 levels and adverse outcomes in subpopulations defined by the percentiles of candidate stratifying laboratory parameters. These parameters were selected from commonly used tests for organ system functions. Our analysis demonstrated that the ability of IL-6 to discriminate high-risk patients changes with the status of organ system functions. We identified fourteen parameters where, among patients with abnormal values, an elevation in IL-6 could be interpreted as a sign of increased risk. To develop a simple method for risk stratification, we grouped these parameters according to physiological functions and explored whether the number of impaired functions could be used for risk stratification. Our results showed that the risk associated with elevated IL-6 levels gradually increased as the number of impaired functions grew. For adverse outcomes, the RDs between the higher and lower IL-6 subpopulations within the 0 to ≥4 impaired functions strata were 0%, 0.7%, 1.4%, 5%, and 16.9% (comparisons between IL-6 groups, all P values <0.05). For PICU admission rates, the RDs between the two IL-6 strata across the five function impairment strata were 2.8%, 6%, 12.4%, 20.9%, and 32.2% (all P values <0.05). The difference in the length of hospital stays between the IL-6 strata also increased with the number of impaired functions. These results demonstrate that evaluating common laboratory parameters can assist in the interpretation of IL-6 levels.

The interaction effect, that refers to the joint effect of two or more exposures on a disease or outcome (31), has been widely observed in biological systems and in disease studies (32-34). Detecting interaction effects can provide valuable information for risk management, and further exploration of the mechanisms behind the observed interaction effect may lead to potential interventions. Damage to tissue or organ function could disrupt the balanced feedback regulation of IL-6 (3,15). The elevated IL-6, along with other cytokines, due to self-reinforcing feedback and excessive activation of immune cells, results in hyperinflammation and further aggravates organ damage (35). Therefore, we speculated that the superimposed IL-6 elevation and functional impairment could amplify the effects of each other. Based on this reasoning, we hypothesized that the combined effect of functional impairment and elevated IL-6 levels is stronger than the sum of their individual effect. Our results demonstrated positive interaction effects between IL-6 levels and functional impairment, and these interaction effects increased as the number of impaired functions grew. For adverse outcomes, the interaction effects, estimated by RERI, increased from 1.06 (1 impaired function, 95% CI: 0.73, 1.38) to 42.79 (≥4 impaired functions, 95% CI: 34.49, 51.09). Similar trends in interaction effects were also observed for the risk of PICU admission. The detection of interaction effects not only provides valuable information for risk evaluation but also suggests that for patients with multiple functional abnormalities, it is important to consider the likelihood of elevated IL-6 levels when evaluating risk. For those with elevated IL-6, recognizing the significantly increased risk is crucial for timely intervention.

In this study, we developed a simple method to interpret the risk associated with IL-6 elevation by evaluating physiological function abnormalities using common laboratory parameters. The effectiveness of risk stratification was validated in an independent cohort and across several disease subpopulations. This study had several limitations. First, due to the cost of IL-6 tests, most patients underwent only one IL-6 test, usually within 3 days of hospital admission. Therefore, our analysis only included the results of the first IL-6 test. Longitudinal analyses are needed to elucidate the sequence and causal relationship between IL-6 levels and function impairment, which could help guide monitoring and intervention strategies. Second, we only analyzed laboratory parameters that were structured in the electronic records of the study site; functional test results for other systems, such as the respiratory, circulatory, and central nervous systems, were not available in the dataset and therefore were not included in this study. Third, the definition of adverse outcomes was based on the local electronic medical record discharge classification system, which may limit direct comparability with studies using differently defined or more granular clinical endpoints. Fourth, IL-6 testing was performed based on clinician discretion rather than predefined indications, which may have introduced selection bias, as patients with more severe illness were more likely to be tested. Although subgroup analyses were conducted to explore robustness across major disease categories, residual selection bias cannot be excluded in this retrospective study. Fifth, the testing method and sample preparation procedure could affect the IL-6 test results, so the generalizability of our threshold for IL-6 needs validation at other sites. Additionally, the distribution of IL-6 varies among different diseases and likely to be age-dependent (36,37). Although sensitivity analysis showed that both the universal threshold of 50 pg/mL and the disease-specific threshold using the 85^th percentile could effectively stratify risk, for more precise risk evaluation, the selection of disease-specific thresholds needs further refinement.

Conclusions

The presence of physiological function abnormalities could help interpret the risk associated with elevated IL-6 levels. The risk associated with elevated IL-6 levels increased as the number of impaired functions grew. There was an interaction effect between elevated IL-6 levels and functional impairment on clinical outcomes, which intensified as the number of impaired functions increased.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of the Hunan Children’s Hospital (No. HCHLL-2024-325). Informed consent was waived by the Medical Ethics Committee because of the retrospective design.

Data Sharing Statement

Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0470/dss

tp-15-08-323-dss.pdf ^{(70.4KB, pdf)}

DOI: 10.21037/tp-2026-0470