Baseline Interleukin-6 Level Predicts Risk of Severe COVID-19: A Two-Center, Retrospective Study

Background:The severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) pandemic has affected almost every country. Interleukin-6 (IL-6), a cytokine secreted by CD4+ T cell, has been shown to be a reliable marker of disease severity and a useful parameter for monitoring progression of coronavirus disease-2019 (COVID-19). However its value as a predictor of severe disease has not been assessed. Methods:A total of 160 laboratory-conrmed COVID-19 patients admitted to two hospitals were enrolledand separated into two groups according to whether or not they progressed to develop severe illness. Demographic and clinical characteristics at admission were compared between the groups. Results: Patients who developed severe COVID-19 had signi�cantly higher baseline IL-6 levels than patients who had mild disease course in hospital (P< 0.001). Patients were further grouped according to quartiles of IL-6 level. The cumulative incidence of severe illnesswas signi�cantly higher in the third and fourth quartiles groups than in the �rst quartile group (55% vs. 15% and 80% vs. 15%, respectively;bothP< 0.001). In multivariate logistic regression analysis, the risk for developing severe disease was markedly higher in the highest IL-6 quartile than in the lowest quartile (odds ratio: 14.95; 95% con�dence interval: 3.65–61.30; P< 0.001). Receiver operating characteristic curve analysis of potential predictive variables showed the area under the curve to be largest for baseline IL-6, with the value of 5.20 pg/mL having the best balance of sensitivity and speci�city for predicting risk of severe COVID-19. Conclusion: Serum baseline IL-6 appears to be a reliable predictor of risk of severe COVID-19. Early intervention may be advisable in patients with serum IL-6 levels >5.20 pg/mL, even if initial symptoms are mild.


Background
In the eight months since the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) was identi ed as the pathogen of anovel pneumonia in Wuhan, China, the organism has spread rapidly across the world [1,2].As of 9 August 2020, a total of 19,462,112 COVID-19 cases have beencon rmed worldwide, and 722,285 patients havedied.Disease severity varies widely, with mortality being much higher in patients with severe COVID-19 than in those with mild illness [3].Amethod for rapid triage of patients is therefore urgently needed.Interleukin-6 (IL-6), a cytokine secreted by CD4 + T cell,has been shown to be a reliable marker ofdisease severity and a useful parameter for monitoring COVID-19 progression [4][5][6], but its value for predicting risk of severe illness has not been examined.The objective of this retrospective study was to determinewhether serum IL-6 level at admission can serve as a predictor of risk for severe COVID-19 in hospitalized patients.

Association between baseline IL-6 and risk of severe COVID-19
After the patients were separatedinto IL-6 quartile subgroups, the incidenceof severe illness was found to increase with increase in IL-6 level.The cumulative incidence was signi cantly higher in the third and fourth quartile subgroups than in the rst quartile subgroup (55% vs. 15% and 80% vs. 15%, respectively; bothP < 0.001; Fig. 2A).Duration of hospitalization was also signi cantly longer in the higher quartile patient groups (all P < 0.01; Fig. 2B).The comorbidity-adjusted odds for severe illness increased signi cantly with increase inIL-6 level (Table 2; Model 1The trend did not change even after adjustment for age, sex, and comorbidityin multivariate logistic regression (Model 2) or after further adjustment for leukocyte subgroups (Model 3).

Discussion
The COVID-19 pandemic poses a great challenge to human society.Although case mortality for COVID-19 is relatively low, it has so far caused far more deaths than severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS) combined [7,8].Becausemortality is very high inpatients with severe COVID-19 [3], it is essential to nd reliable early predictors ofrisk for severe illness.
Previous studies in patients with severe COVID-19 have shownchanges in immune-in ammation parameters, e.g., low counts of lymphocytes and its subgroups, high neutrophil counts, high NLR, and elevated serum IL-6 level [4,6,[9][10][11]; these changes were attributed to the severe and aberrant host immune response [12].In our study, we found similar changes in patients who developed severe disease versus those who had mild disease:lymphocytes and all its subsets were signi cantly lower in the SC group, while neutrophil, NLR, N8R, and IL-6 were signi cantly higher (all P < 0.05).Reduction in T lymphocyte counts may due to the direct invasion by SARS-CoV-2, similar to the mechanism reported in MERS-CoV infection [13].In addition, autoimmune antibodies triggered by the viruscould decrease production and differentiation of T lymphocytes and other lymphocytes viainhibition of growthand suppression of hematopoiesis [11,14].Elevation of neutrophil counts usually indicates an in ammatory response, while decrease in lymphocyte counts suggests damage to the immune system; therefore, as the results of our study suggest, high NLR and N8R can serve as diagnostic markers for severe COVID-19 [9] and also be useful predictors of severe illness.
Huang et al.were the rst to suggest that a cytokine storm could be associated with severe disease [1]; this was con rmed by subsequent studies [15,16].IL-6 is a multifunctional cytokine involved in cell signaling and regulation of immune cells, and usually shows more signi cant serum levels changes than other cytokinesin in ammatory disease [17,18].IL-6 re ects the intensity of in ammatory response in the body.Liu et al. and Zhe et al. have demonstrated that IL-6 plays a pivotal role in the severity of COVID-19 and thereforehas value for monitoring the progress of severely ill patients [4,5].In our study, we found the cumulative incidence of severe illnessto be related to the baseline IL-6 level.Patients in the higher IL-6 quartile groups were signi cantly more likely to develop severe illness (Fig. 2A) and to need longer hospitalization (Fig. 2B).Since the severity of the disease corresponded to the intensity of in ammation, former phenomena could be explained.
Comorbidities such as diabetes and cancer can affect baseline IL-6 level [19,20]; we therefore used logistic regression models to adjust for the effect of potential confounders.However, the signi cant association between baseline IL-6 level andrisk of severe illnesspersisted even after adjustment for all potential confounders (Models1-3; Table 2).
Neutrophil count, NLR, N8R, and IL-6 level have all been previously shown to be useful prognostic indicatorsin severe COVID-19 [4][5][6]9].We performed ROC analysis to identify the variable with the best predictive value.The AUC of IL-6 was signi cantly larger than those of neutrophil count and N8R.The AUC of IL-6 was comparable to that of NLR, which is currently widely accepted as a predictor of outcome in patients with COVID-19.However, the sensitivity of IL-6 was much higher than that of NLR (91.67% vs.66.67%).For any predictive or screening index, sensitivity is more important than speci city;the higher the sensitivity, the fewer the false-negative results.Thus,in addition to being a useful parameter for monitoring progression of disease in patients with COVID-19, baseline IL-6 could be a powerful predictor of risk of severe disease.
This study had some limitations.First, this was a retrospective analysis of a small sample and some bias is inevitable.Second, because of inconsistencies in data recording between the two centers, some clinical parameters could not be included in the analysis.

Conclusion
Baseline IL-6 levelappears to be areliable predictor of the risk for developing severe COVID-19.In patients with mild COVID-19 features on admission but serum IL-6 level > 5.20 pg/mL, early interventionmight helpprevent progression to severe disease.

Data source and collection
This retrospective observational study was jointly performed by the Fourth A liated Hospital, School of Medicine, Zhejiang University, and the General Hospital of the Central Theater Command, and was approved by the Ethics Committees of both institutions (No.K20200027 and No. [2020]017 − 1, respectively).
A total of 160 patients hospitalized for treatment of COVID-19 between January 10, 2020,and March 9, 2020, were included in this retrospective study.COVID-19 was diagnosed according to the Chinese National Health Committee guidelines (version 7) [21].The patients were separated into two groups according to the course of the disease during hospital stay: the mild steady (MS) group (n = 88), comprising patients who had mild symptoms on admission and continued to have mild disease until recovery, and the severe change (SC) group, comprising patients who developed severe illness during hospitalization (n = 72).Severe illness was de ned as COVID-19 with any of the following: 1) shortness of breath, with respiratory rate ≥ 30/min; 2) oxygen saturation by pulse oximeter ≤ 93% in resting state; 3) ratio of partial pressure of arterial oxygen (PaO 2 ) to fraction of inspired oxygen (FiO 2 ) ≤ 300 mm Hg (1 mm Hg = 0.133 kPa); or 4) increase in lesion size by ≥ 50% over 24-48hourson lung imaging.Data on demographic characteristics, comorbidities, radiographic ndings, and laboratory tests on admission, and duration of hospitalization, were extracted from the electronic medical records and compared between the groups.All data were cross-checked for consistency and accuracy before analysis.For a closer exploration of the association between baseline IL-6 and risk of severe illness, we also grouped patients by IL-6 quartiles.

Statistical analysis
Continuous variables were expressed as means(± standard deviation) or medians (and interquartile ranges), and compared using the Student's t-test orthe Mann-Whitney U test.Categorical variables were expressed as counts and percentages and compared using thechi-square test or Fisher exact test.
Correlations between IL-6 and other variables were analyzed by Spearman correlation analysis.One-way analysis of variance (ANOVA) with Bonferroni correction was used to compare the duration of hospitalization between IL-6 quartilepatient subgroups.Multivariate logistic regression analysis was performed to assess the association between IL-6 level and risk of severe COVID-19.Odds ratios (OR) and 95% con dence intervals (95% CI) were calculated.To compare the predictive values of different laboratory parameters, receiver operating characteristic (ROC) curves were plotted, and the areas under the curves (AUC) were compared.The Youden index was used to nd the optimal cutoff values of each parameter forpredicting future severe illness.Statistical analysis was performed using SPSS 20.0 (IBM Corp., Armonk, NY, USA).Statistical signi cance was at P ≤ 0.05 for all tests.

Figure 2 The 2
Figure 2

Table 2
Risk for severe COVID-19 in patients grouped according to quartiles of baseline IL-6 OR, odds ratio; CI, con dence interval; IL-6, interleukin-6.Model 1: Adjustment made for comorbidities; Model 2: Adjustment made age, sex, and comorbidities; Model 3: Adjustment made for age, sex, comorbidities, neutrophil count, lymphocyte count, CD4 + T cell count, CD8 + T cell count, B cell count, and NK cell count.

Table 3
Optimal cutoff values of potential predictors of severe COVID-19 and their sensitivity and speci city # P < 0.05 vs. IL-6