key: cord-0986918-8hv8rx6h
authors: Dhar, S. K.; K, V.; Damodar, S.; Gujar, S.; Das, M.
title: IL-6 and IL-10 as predictors of disease severity in COVID 19 patients: Results from Meta-analysis and Regression
date: 2020-08-17
journal: nan
DOI: 10.1101/2020.08.15.20175844
sha: 72a76caef732c6d054e4d4595597d0dfe4484ff7
doc_id: 986918
cord_uid: 8hv8rx6h

Objectives: SARS-CoV-2, an infectious agent behind the ongoing COVID-19 pandemic, induces high levels of cytokines in patients contributing to the disease patho-physiology. Nonetheless, exact association and contribution of particular cytokines towards COVID-19 pathology remains poorly understood. This study performed a comprehensive meta-analysis to establish association between induced cytokines and COVID-19 disease severity to help in prognosis and clinical management. Methods: Scientific literature was searched to identify 18 clinical studies. Standardized mean difference (SMD) for cytokines IL-2, IL-4, IL-6, IL-10, TNF- and IFN-{gamma} between severe and non-severe COVID-19 patient groups were summarized using random effects model. A classifier was built using logistic regression model with cytokines having significant SMD as covariates. Results: Out of the 6 cytokines, IL-6 and IL-10 showed statistically significant SMD across the studies synthesized. Classifier with mean values of both IL-6 and IL-10 as covariates performed well with accuracy of ~ 92% that was significantly higher than accuracy reported in literature with IL-6 and IL-10 as individual covariates. Conclusions: Simple panel proposed by us with only two cytokine markers can be used as predictors for fast diagnosis of patients with higher risk of COVID-19 disease deterioration and thus can be managed well for a favourable prognosis.

COVID-19 pandemic caused by SARS-CoV-2 has emerged as a major threat to mankind affecting 18+ million people resulting in over 700,000 deaths 1 worldwide. Exuberant inflammation manifested as hypercytokinemia with elevated levels of cytokines such as IL-2, IL-6, IL-7, IL-10, IFN-γ, and TNF- is a hallmark of the SARS-CoV-2 infection often leading to critical conditions like ARDS (acute respiratory distress syndrome) and multiorgan failure 2 .

Innate immune response is the first step of defence mechanism against viral infection. Sustained elevated levels of the cytokines due to viral infections often lead to hypercyotokinemia, commonly referred as "cytokine storm" (CS) in which the homeostasis of immune response is disturbed leading to acute inflammation at systemic level. Acute lung damage is a common clinical manifestation of CS often leading to fatality 6 . Such heighted immune response is a signature of coronavirus family as reported earlier for MERS-CoV and SARS-CoV infections 7, 8 . In COVID-19, elevated levels of both proinflammatory (IL-1, IL-6, TNF-) and anti-inflammatory (IL-10, IL-1RA) cytokines are observed 9, 10 . IL-6 is a major actor in CS, with increased serum levels leading to induction of acute phase response proteins including C-reactive protein (CRP) and serum amyloid A (SAA). Recently published meta-analysis studies on the levels of cytokines in COVID-19 patients have focussed mostly on IL-6 due to potential repurposing of tocilizumab to suppress the inflammatory condition. Coomes is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint -52.9) 13 . All these meta-analysis summary results were associated with high level of heterogeneity (I 2 ~ 98 -100%). In a limited size meta-analysis over three studies, Rangel et al found the ratio IL-6/IFN- significantly elevated in severe patients 14 with SMD 0.74 (95% CI: 0.13 -1.38) with substantial heterogeneity (I 2 = 79%). IL-10, on the other hand is an anti-inflammatory cytokine which inhibits its pro-inflammatory counterparts.

Multiple clinical studies have reported elevated levels of IL-10 in COVID-19 patients 2,15,16 .

Levels of IL-6, IL-10 and TNF- were also found to be indicators of T-cell exhaustion in COVID-19 patients 17 .

Though the levels of cytokines are increased in severe COVID-19 patients, it's implication in a therapeutic perspective remains unclear. Corticosteroids that can potentially suppress cytokines by inhibiting NF-B transcription factor have been used on COVID-19 patients. However, a meta-analysis across studies involving infection of SARS-CoV, SARS-CoV-2 and MERS-CoV showed increased mortality risk ratio (RR 2.11, 95% CI: 1.13 -3.94) for patients treated with corticosteroids 18 . There are also a couple of case reports on anti-TNF- agents etanercept 19 and infliximab 20 therapy used successfully for COVID-19 patients with pre-existing auto-immune conditions. However, results of wider application of such therapy is still awaited. On the other hand, there have been several studies targeting IL-6 levels using siltuximab, an IL-6 inhibitor or tocilizumab, an IL-6 receptor inhibitor in severe COVID-19 patients. A review by Jamilloux et al 21 on few such studies did not reveal any conclusive benefits due to lack of appropriate control groups in design or inadequate statistical significance. Further, the timing for anti-IL-6 agent administration remains critical since an early administration may negatively affect the innate response, while the late administration may not yield desired benefits. Some of the ongoing clinical results using anti-IL-6 therapy will shed more light in near future.

Nonetheless, the use of the cytokines as a prognostic marker for severity of COVID-19 symptoms looked promising. One of the earliest clinical studies showed significant difference in levels of IL-6 and IL-10 along with CD4 + and CD8 + T-cell population between mild and severe groups 22 but did not carry out further analysis on prognostic value of these markers. Results of the meta-analysis studies described before [11] [12] [13] [14] concluded elevated levels of cytokines in severe COVID-19 patients but did not proceed to establish significance of the markers as a classifier of patient groups. A more recent study 23 presented univariate classifiers for severe and non-severe patient groups using levels of is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint IL-6, IL-10 and IFN- individually as covariates and failed to capture the possible interdependence of the markers which perhaps was the cause for lesser-than-optimal classifier performance. In this study, we present a comprehensive meta-analysis to establish difference in levels of commonly used cytokine markers between severe and non-severe patient groups and build a classifier using markers that present significant differences. This generalized and comprehensive approach can lead to a prognostic system to aid clinical management of COVID-19.

We performed the meta-analysis following PRISMA guidelines. Literature search was performed in the LitCovid portal of Pubmed, in Google Scholar as well as in preprint archives such as medRxiv, bioRxiv and SSRN library for articles in English published in

year 2020 with COVID-19-specific terms along with terms such as "cytokine level" and combinations of common cytokine names and gene symbols. Search strategy was reviewed by all authors and it was decided to retain non-peer reviewed articles in selection in view of the present emerging situation. Identified articles were screened and shortlisted to clinical studies that contained data for multiple cytokines including IL-6 for the 'severe' and 'non-severe' COVID-19 patient groups. Exclusion criteria for shortlisting included review articles, opinions and commentaries and studies that include other pathological conditions or complications associated with COVID-19.

Selected articles were reviewed independently by two authors (VK and SKD). Data was extracted from articles using standardized forms and was cross-validated. Cytokine levels reported as median (IQR) was converted to mean and standard deviation using standard methods 24 . To assess the effect of each marker, we used the standardized mean difference (SMD) of measured cytokine level (in pg/mL) between severe and non-severe groups including Hedges' correction for positive bias. All calculations were carried out using the metafor library 25 on R statistical software platform. Meta-regression of SMD of a marker was carried out using mixed-effects model with differences in age and sex (measured as percentage of male patients) and also with differences in neutrophil-tolymphocyte ratio and CD4 + /CD8 + T-cell count ratio across groups as modulators. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08. 15.20175844 doi: medRxiv preprint Publication bias in the studies was analyzed using Funnel Plot and Egger's regression test.

Sensitivity of meta-analysis results was assessed by repeating the analysis with leaving out one study at a time and observing significant changes in heterogeneity and summary effect.

We built a classifier using logistic regression model for categorization of patient groups in to severe and non-severe categories based on mean values of cytokines in the groups. 

The search yielded a total of 99 "hits" which were screened to select 18 articles containing cytokine levels of severe and non-severe patients reported from clinical studies (Fig. 1) .

All 18 selected studies were conducted in China and included 1,242 non-severe and 915 severe COVID-19 patients ( Table 1 ). Out of the 18 studies, 7 were non-peer reviewed studies [26] [27] [28] [29] [30] [31] [32] published online in preprint archives. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint NHC guideline: (a) Respiratory distress, RR≥30/min, (b) Oxygen saturation level (SpO2)  93% at rest and (c) Arterial blood oxygen partial pressure (PaO2) /oxygen concentration (FiO2) ≤300 mm Hg. Pulmonary lesion: >50% lesion progression within 24-48h in pulmonary imaging WHO guideline: Adolescent or adult with clinical signs of pneumonia (fever, cough, dyspnoea, fast breathing) plus one of the following: respiratory rate > 30 breaths/min; severe respiratory distress; or SpO2 < 90% on room air Age in mean (SD)

In most studies the severe patient group was designated based on National Health Commission of China guidelines 41 that specifies (a) Respiratory distress, RR≥30/min, (b) Oxygen saturation level (SpO2)  93% at rest and (c) Arterial blood oxygen partial pressure (PaO2) /oxygen concentration (FiO2) ≤300 mm Hg. For some studies, pulmonary lesion progression > 50% within 24-48 hours on imaging was used as additional criteria.

One study used WHO guideline 42 that categorizes severe patients as adolescent or adult with clinical signs of pneumonia (fever, cough, dyspnoea, fast breathing) plus one of the is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint following: respiratory rate > 30 breaths/min; severe respiratory distress; or SpO2 < 90% on room air. 

Meta-analysis for SMD value of each marker using a random-effects model showed moderate and statistically significant elevation in severe patients for only two cytokines viz. IL-6 (SMD 0.53, 95% CI: 0.26 -0.80, p < 0.001) (Fig. 2) and IL-10 (SMD 0.64, 95% CI: 0.38 -0.91, p < 0.0001) (Fig. 3) . Summary effect size of other markers IL-2 (p = 0.281), IL-4 (p = 0.305) and TNF- (p = 0.258) were not significant. Level of IFN-, the type II interferon, showed a weak elevation in severe group (SMD 0.11, 95% CI: -0,01 -0.23, p = 0.078) with moderate significance ( Table 2) . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint

Summary effects of both IL-6 and IL-10 showed substantial heterogeneity (I 2 ~ 72-73%) source of which was explored by carrying out meta-regression of IL-6 and IL-10 using (i) SMD of patient age and (ii) difference in male percentage between severe and non-severe groups as moderators in a mixed-effects model. Results revealed IL-10 levels having a dependence on the age difference (coefficient 0.63, p = 0.012), whereas the dependence of IL-6 on age was not significant (p > 0.05, Fig. 4) . No dependence of IL-6 and IL-10 levels were found on gender expressed as difference of male percentage between the two groups ( Supplementary Fig. S1 ). Above results ascribe some of the observed heterogeneity in IL-6 and IL-10 SMD to difference in average age between severe and non-severe group.

Neutrophil to Lymphocyte ratio (NLR) has been explored as a prognostic marker for COVID-19, with high level of NLR associated with severity of the disease 43, 44 or with higher mortality rates 45, 46 . More specific subsets of the Lymphocytes, especially the CD4 + and CD8 + T-cell counts are found to be depleted in severe COVID-19 patients 17, 47 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint However, meta-analysis across multiple studies did not indicate any significant change in the CD4+/CD8+ T-cell count ratio in severe COVID-19 patients 48 . Our synthesis shows similar results with SMD of NLR as 0.71 (95% CI: 0.14 -1.28, p = 0.014) ( Supplementary   Fig. S2 ) indicating significant increase in patients with severe disease, whereas the SMD of CD4+/CD8+ T-cell ratio was not significantly different across the groups ( Supplementary Fig. S3 ).

We performed a meta-regression analysis of the SMD values of IL-6 and IL-10 on NLR and CD4 + /CD8 + T-cell count ratio to assess if the observed heterogeneity could be ascribed to the variation of lymphocytes. Results showed IL-10 having a moderate dependence on NLR (p < 0.1) whereas IL-6 SMD was correlated with CD4+/CD8+ T-cell count ratio 

Our results indicate that out of the 6 cytokines analyzed, levels of only the proinflammatory IL-6 and the immuno-suppressive IL-10 are significantly elevated in the severe group of patients, an observation reported in other studies as well 23, 49 .

This prompted us to assess the prognostic potential of these markers by developing a logistic regression model using IL-6 and IL-10 mean values as covariates to classify the patient groups as severe and non-severe across all 18 studies. Model using both IL-6 and IL-10 as input parameters showed an overall accuracy of 91.7% and area under the corresponding ROC curve as 0.957, which has higher than the corresponding values using only IL-6 (accuracy 77.8%, AUC 0.821) and only-IL-10 (accuracy 80.6%, AUC 0.878) as input parameters (Table 3 , Fig. 5 ). Model with both IL-6 and IL-10 showed nearly 100% specificity and 83.3% sensitivity for classification in severe and non-severe categories. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint Primary limitation of this meta-analysis is that the studies are restricted only to Chinese ethnicity, which was not by design but due to the fact that most initial studies were reported from this geographic region. Also 7 of the 18 studies were non-peer reviewed publications selected as per search strategy. There was some heterogeneity in clinical criteria for severe disease as shown in Table 1 .

Publication bias towards higher SMD of IL-10 levels reported in smaller-size studies manifested as asymmetry in Funnel plot and Egger's regression test (p < 0.001), whereas the corresponding bias was not significant (p = 0.151) for IL-6 ( Fig. 6 ). Sensitivity analysis of the studies showed a large reduction of heterogeneity in IL-6 (I 2 value from 72% to 25%) and a moderate reduction for IL-10 (73% to 55%) by leaving out one particular study from the analysis (Supplementary Fig. S6 ). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint

As the spread of COVID-19 becomes prevalent in most societies, a suitable prognostic test that can predict possible progression of patient to a severe state of the disease with reasonable accuracy is the need of the hour. Though NLR is seen as a critical parameter that is elevated in the severe COVID-19 patients, it's use as a possible prognostic factor is not yet established. On the other hand, a test based on serum marker can have a wider deployment in varied clinical settings.

A recent proteomics study 50 predicted similar classification performance for severe and non-severe COVID19 patients using 22 serum proteins and 7 metabolites as covariates.

However, a proteomics infrastructure is not so common in a clinic, limiting clinical utility of this test.

Another study 23 presented univariate classifier results between mild and severe patients using each cytokine as a diagnostic factor individually and found good performance with IL-6, IL-10 and IFN- as individual covariates. This model used individual patient data and . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint showed maximal classifier performance using IL-6 as covariate (sensitivity 82.4% and specificity 78.5%). Our analysis, though using mean values of patient groups, shows much improved performance when both IL-6 and IL-10 are used as covariates.

Our findings indicate a possible dysregulation in immune response against COVID-19, characterized by two counter-acting cytokines IL-6 and IL-10, shifts the balance between non-severe to severe category of patients, and hence measurement of both markers is necessary to demarcate the boundary. Availability of individual patient-level IL-6 and IL-10 data of past studies or initiation of future clinical studies will help to validate this model and establish it for routine clinical use.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. 

Enclosed . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted August 17, 2020. . https://doi.org/10.1101/2020.08.15.20175844 doi: medRxiv preprint

COVID-19) Situation Report -199. World Health Organization

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Innate immune recognition of viral infection

Cytokines and chemokines: At the crossroads of cell signalling and inflammatory disease

IL-10: A multifunctional cytokine in viral infections

Into the Eye of the Cytokine Storm

Delayed induction of proinflammatory cytokines and suppression of innate antiviral response by the novel Middle East respiratory syndrome coronavirus: Implications for pathogenesis and treatment

Pathogenic human coronavirus infections: causes and consequences of cytokine storm and immunopathology

The COVID-19 Cytokine Storm; What We Know So Far

SARS-CoV-2 infection: The role of cytokines in COVID-19 disease

Interleukin-6 in COVID-19: A Systematic Review and Meta-Analysis. medRxiv

Association of inflammatory markers with the severity of COVID-19: A meta-analysis

Elevated Interleukin-6 and Severe COVID-19: A Meta-Analysis

High IL-6/IFN-γ ratio could be associated with severe disease in COVID-19 patients

Clinical and immunologic features in severe and moderate forms of Coronavirus Disease

Dysregulation of immune response in patients with COVID-19 in Wuhan, China

Reduction and Functional Exhaustion of T Cells in Patients With Coronavirus Disease 2019 (COVID-19)

The effect of corticosteroid treatment on patients with coronavirus infection: a systematic review and meta-analysis

Recovery from COVID-19 in a patient with spondyloarthritis treated with TNF-alpha inhibitor etanercept

Infliximab for severe ulcerative colitis and subsequent SARS-CoV-2 pneumonia: a stone for two birds

Should we stimulate or suppress immune responses in COVID-19? Cytokine and anti-cytokine interventions

Relationships among lymphocyte subsets, cytokines, and the pulmonary inflammation index in coronavirus (COVID-19) infected patients

Profiling serum cytokines in COVID-19 patients reveals IL-6 and IL-10 are disease severity predictors

Estimating the sample mean and standard deviation from the sample size , median , range and / or interquartile range

Conducting meta-analyses in R with the metafor

Clinical Characteristics and Short-Term Outcomes of Severe Patients with COVID-19 in Wuhan, China. medRxiv

Metabolic disturbances and inflammatory dysfunction predict severity of coronavirus disease 2019 (COVID-19): a retrospective study. medRxiv

Clinical Characteristics and Prognostic Factors of 148 COVID-19 Cases in a Secondary Epidemic Area

COVID-19 early warning score: a multi-parameter screening tool to identify highly suspected patients. medRxiv

Immune phenotyping based on neutrophil-tolymphocyte ratio and IgG predicts disease severity and outcome for patients with COVID-19. medRxiv

Potential Factors for Prediction of Disease Severity of COVID-19 Patients. medRxiv

Longitudinal Profiling of Cytokines and Chemokines in COVID-19 Reveals Inhibitory Mediators IL-1Ra and IL-10 Are Associated with Disease Severity While Elevated RANTES Is an Early Predictor of Mild Disease

The clinical course and its correlated immune status in COVID-19 pneumonia

Clinical characteristics and co-infections of 354 hospitalized patients with COVID-19 in Wuhan, China: a retrospective cohort study. Microbes Infect

Clinical Features of 69 Cases with Coronavirus Disease

Elevations of serum cancer biomarkers correlate with severity of COVID-19

Suppressed T cell-mediated immunity in patients with COVID-19: A clinical retrospective study in Wuhan

Predictive factors for disease progression in hospitalized patients with coronavirus disease 2019 in Wuhan

Viral load dynamics and disease severity in patients infected with SARS-CoV-2 in Zhejiang province

Clinical value of immune-inflammatory parameters to assess the severity of coronavirus disease 2019

National Health Commission & State Administration of Traditional Chinese Medicine. Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia

Clinical Management of COVID-19: Interim Guidance

Neutrophil-to-lymphocyte ratio and clinical outcome in COVID-19: a report from the Italian front line

Higher level of Neutrophil-to-Lymphocyte is associated with severe COVID-19

Neutrophil-to-lymphocyte ratio as an independent risk factor for mortality in hospitalized patients with COVID-19

Neutrophil-to-Lymphocyte Ratio and Outcomes in Louisiana Covid-19 Patients

Lymphocyte subset (CD4+, CD8+) counts reflect the severity of infection and predict the clinical outcomes in patients with COVID-19

CD4+T, CD8+T counts and severe COVID-19: A meta-analysis

The laboratory tests and host immunity of COVID-19 patients with different severity of illness

Article Proteomic and Metabolomic Characterization of COVID-19 Patient Sera ll Article Proteomic and Metabolomic Characterization of COVID-19 Patient Sera