key: cord-0748773-ihqzpdar
authors: Narain, Sonali; Stefanov, Dimitre G.; Chau, Alice S.; Weber, Andrew G.; Marder, Galina; Kaplan, Blanka; Malhotra, Prashant; Bloom, Ona; Liu, Audrey; Lesser, Martin L.; Hajizadeh, Negin; Cohen, Stuart L.; Cookingham, Jennifer; Hirschwerk, David A.; Maria, Naomi I.; Satapathy, Sanjaya K.; Sison, Cristina; Taylor, Matthew; Qiu, Michael
title: Comparative Survival Analysis of Immunomodulatory Therapy for COVID-19 'Cytokine Storm'
date: 2020-10-17
journal: Chest
DOI: 10.1016/j.chest.2020.09.275
sha: 7df246d42d24c95521af63557ea2753308292ff8
doc_id: 748773
cord_uid: ihqzpdar

Background Cytokine storm is a marker of COVID-19 illness severity and increased mortality. Immunomodulatory treatments have been repurposed to improve mortality outcomes. Research Question To identify if immunomodulatory therapies improve survival in patients with COVID-19 cytokine storm. Study Design and Methods We conducted a retrospective analysis of electronic health records across the Northwell Health system. COVID-19 patients hospitalized between March 1, 2020 and April 24, 2020 were included. Cytokine storm was defined by inflammatory markers: ferritin >700ng/mL, C-reactive protein >30mg/dL or lactate dehydrogenase >300U/L. Patients were subdivided into six groups—no immunomodulatory treatment (standard of care) and five groups that received either corticosteroids, anti-interleukin 6 antibody (tocilizumab) or anti-interleukin-1 therapy (anakinra) alone or in combination with corticosteroids. The primary outcome was hospital mortality. Results 5,776 patients met the inclusion criteria. The most common comorbidities were hypertension (44-59%), diabetes (32-46%) and cardiovascular disease (5-14%). Patients most frequently met criteria with high lactate dehydrogenase (76.2%) alone or in combination, followed by ferritin (63.2%) and C-reactive protein (8.4%). More than 80% of patients had an elevated D-dimer. Patients treated with corticosteroids and tocilizumab combination had lower mortality compared to standard of care (Hazard Ratio (HR):0.44, 95% confidence interval (CI): 0.35-0.55; p<0.0001) and when compared to corticosteroids alone (HR:0.66, 95%CI: 0.53-0.83; p-value=0.004), or in combination with anakinra (HR:0.64, 95%CI:0.50-0.81; p-value=0.003) . Corticosteroids when administered alone (HR:0.66, 95%CI:0.57-0.76; p<0.0001) or in combination with tocilizumab (HR:0.43, 95%CI:0.35-0.55; p<0.0001) or anakinra (HR:0.68, 95%CI:0.57-0.81; p<0.0001) improved hospital survival compared to standard of care. Interpretation The combination of corticosteroids with tocilizumab had superior survival outcome when compared to standard of care and corticosteroids alone or in combination with anakinra. Furthermore, corticosteroid use either alone or in combination with tocilizumab or anakinra was associated with reduced hospital mortality for patients with COVID-19 cytokine storm compared to standard of care.

In March 2020, New York City and its metropolitan area became the epicenter for coronavirus disease 2019 in the United States, with over 250,000 cases and greater than 17,000 deaths by early May 2020. 2 Throughout this outbreak, physicians and scientists have struggled to understand the pathogenesis and clinical course of this infection. Early retrospective data from China and Italy showed increased mortality in those with elevated inflammatory markers, such as ferritin, C-reactive protein (CRP), lactate dehydrogenase (LDH), interleukin 6 (IL-6) and D-dimer. 3 Uncontrolled and unabated cytokine release and a hyperinflammatory response termed as COVID-19 "cytokine storm" (CCS), was described as a major determinant of poor survival. 4 Limited data existed to guide clinical decision-making in the absence of FDA-approved COVID-19 specific therapies. Faced with rapidly increasing rates of infection and hospitalizations, physicians repurposed immunomodulatory treatments in an attempt to curtail morbidity and mortality. Although initial reports discouraged the use of corticosteroids, later publications suggested survival benefits. 3, 5, 6 Small retrospective studies reported improved outcomes in CCS by using anti-IL-6 (i.e., tocilizumab [Roche, Basel, Switzerland]) and anti-IL-1 therapies (i.e., anakinra [Sobi, Stockholm, Sweden]) 7-9 that are commonly used for inflammatory conditions such as cytokine release syndrome and macrophage-activation syndrome. Further evidence supporting the use of anti-IL-1 was based on previous reports of improved survival in a subgroup of patients with sepsis and hyperferritinemia. 10 Within Northwell Health, the largest private nonprofit health system in New York state, a multidisciplinary committee consisting of pulmonology, infectious disease, immunology and rheumatology specialists was formed to create COVID-19 treatment protocols. This included the identification of CCS, which we defined as ferritin>700ng/mL 11 or CRP>30mg/dL 3, 12 or LDH>300U/L 3 . Treatment protocols with corticosteroids, tocilizumab and anakinra as potential immunomodulatory therapies were based on the available literature at the time. 3, 11, 12 Due to the rapidly evolving data and surge of patients in a short timeframe, there was wide variation in the use of these drugs across the health system. In this retrospective study, we leveraged this natural experiment to compare mortality in patients meeting criteria for CCS who received different combinations of these immunomodulatory drugs.

Study Population: We retrospectively analyzed electronic health record data of patients admitted to the 12 hospitals and emergency departments within the Northwell Health system between March 1, 2020 and April 24, 2020. The Institutional Review Board for the Feinstein Institutes of Medical Research at Northwell Health approved this study as minimal-risk research and waived the requirement for informed consent. Inclusion criteria were: COVID-19 positivity as determined by polymerase chain reaction testing of nasopharyngeal swabs; age > 18; meeting CCS criteria: ferritin>700ng/mL 11 or CRP>30mg/dL 3, 12 or LDH>300U/L 3 (e- Figure 1 ). "T 0 " was identified as the time at which a patient first met this definition. Patients who received any of the pre-specified immunomodulatory drugs before T 0 were excluded from this study.

Group definition: There were six groups identified based on whether they received any of the pre-defined immunomodulatory drugs. One group consisted of those who received none of the medications, labeled as standard of care (SoC). Five treatment groups received varying combinations of the three immunomodulatory drugs: corticosteroids only (S), corticosteroids and tocilizumab (ST), corticosteroids and anakinra (SA), tocilizumab only (T) and anakinra only (A). In the timeframe of this analysis, hydroxychloroquine, azithromycin, colchicine and vitamin C, either alone or in combination, were administered to COVID-19 patients as part of institutional protocols (Supplemental e- Table 1 ).

Primary objective: To compare in-hospital mortality among COVID-19 patients with CCS who received combinations of immunomodulatory treatments versus SoC.

Covariates: Potentially confounding variables (covariates) were included in the multivariable model based on clinical experience and the COVID-19 literature at the time. These included demographic data such as age, gender, race/ethnicity, smoking history, insurance status and whether patients were treated in a tertiary versus community medical center. Comorbidities examined included chronic lung disease (i.e., asthma, chronic obstructive pulmonary disease), cardiovascular disease, hypertension, diabetes, renal disease, hemodialysis, liver disease, cancer, autoimmune disease, the Charlson comorbidity index (CCI) and body mass index.

Laboratory data included CRP, ferritin, D-dimer, LDH, hemoglobin, platelet count, serum sodium, serum transaminases and neutrophil to lymphocyte ratio (NLR). We also included disease severity surrogates, such as use of invasive mechanical ventilation (IMV; at any time prior to T 0 ) and vasopressor use (within 24 hours of T 0 ).

Treatment groups were compared using demographic variables, comorbidities and baseline lab values using the χ 2 , Fisher's exact or Kruskal-Wallis tests, as appropriate. Categorical variables were summarized using percentages. Continuous variables were summarized using medians with 25-75 th percentiles (PCT). Labs considered clinically important were included in the analysis. Baseline laboratory values in this study were defined as the value closest to T 0 within the 96 hours prior to T 0 . Exceptions were for CRP, ferritin and D-dimer, which were defined as within 96 hours prior to T 0 and up to 12 hours after T 0 , due to laboratory ordering practices.

Patient survival was calculated from T 0 to the time of in-hospital death. Data from patients discharged from the hospital or remaining in the hospital on April 24, 2020 were considered censored.

Patient survival was compared between treatment groups using the Cox regression model, adjusting for all covariates outlined above. The proportional hazards assumption was assessed and deemed acceptable. Missing lab values were handled using multiple imputation, using 50 imputed datasets. We used the fully conditional method with a discriminant function for the imputation of the lab categories (e.g. 'low', 'normal' or 'high' as specified in Table 2 ). PROC MI (SAS 9.4), with all variables from Table 1 , was used for the multiple imputation. Holm's stepdown procedure for multiple comparisons was used to account for the 15 pairwise tests resulting from the six groups. The final model included all clinically important covariates regardless of their statistical significance (the "full model"). Table   1 . There were twice as many males as females. A significant difference in the racial distribution across treatment groups was noted, with more blacks in the 'A' group and whites in the 'T' group. A higher proportion of patients identifying as other/multiracial race were noted to be in the 'A' group. Majority of the cohort (>65%) had never smoked. The most common comorbidities across groups were: hypertension (44-59%), diabetes (32-46%), cardiovascular disease (5-14%), chronic kidney disease (5-12%), cancer (5-11%) and asthma (3-12%). Less than 2% of patients were on hemodialysis prior to T 0 . Approximately, 40% of the patients in the cohort had low predicted 10-year survival rate based on CCI (>5). There were more patients with moderate to high CCI (>3) in the 'T' group as compared to other treatment groups. More patients in the 'S', 'ST' and 'T' treatment groups were on IMV and vasopressors at T 0 .

More than 80% of the patients who met criteria for CCS had an elevated D-dimer, of which ~20% had levels greater than five times the upper limit of normal. The most common criterion met for CCS definition was high LDH, which was found in 76.2% of patients, either alone or in combination with other criteria, followed by ferritin (63.2%) and CRP (8.4%). The definition of CCS was met by only one criterion in 56.0% of patients, by two criteria in 40.2% and by three criteria in 3.8% of patients. The distribution of CRP, ferritin and LDH is provided in Supplemental e- Figure 1 . There was a statistically significant difference between treatment arms with respect to CRP, ferritin and LDH (p<0.0001) with 'SoC' group having lower median CRP, ferritin and LDH compared to the 'S', 'ST' and 'SA' groups.

Kaplan-Meier (unadjusted) survival estimates for treatment groups are presented in Figure 2 . A Cox proportional hazards regression model was used to compare treatment groups, adjusting for clinically important variables. In this model, demographic covariates that were statistically significant association with increased mortality were older age, unknown smoking status, Medicaid and self-pay insurance ( Table 2) . Higher mortality was associated with presence of asthma, interstitial lung disease (ILD) and the need for IMV at T 0 . Higher mortality was also associated with elevated D-dimer, thrombocytopenia, low GFR, transaminitis, hypernatremia and abnormal NLR. Lower mortality was associated with hypertension and black race. Table 2 ). The average number of days of steroids use was approximately 4.5 days except for methylprednisolone in the 'ST' and 'SA' groups where the average duration was 6.5 days. The average length of steroid use was 12 -15 mg for dexamethasone, 85 -89 mg for methylprednisolone and 29 -33 mg for prednisone.

Rates of culture positive bloodstream infections in the treatment groups are reported in Supplemental e- Table 3 . Approximately 5% of patients in the 'S' group had bacteremia compared to 10% in the 'SA' and 'ST' group. Similarly, 2-3% of patients in the steroid groups 'S', 'ST' and 'SA' were noted to have fungemia. In comparison the rate of bacteremia in the 'SoC' group was 1.6% and rate of fungemia was 0.4%. No bacteremia or fungemia was reported in the 'T' or 'A' groups.

This large retrospective observational study leverages natural heterogeneity in practice patterns for CCS patients. We describe hospital survival outcomes in patients receiving different combinations of immunomodulatory therapy with careful consideration of potential confounders available in the electronic health records. Our findings suggest that corticosteroids used alone or in combination with tocilizumab or anakinra were associated with lower mortality as compared to SoC. This association remained after controlling for covariates that influence mortality in COVID-19.

Age was associated with increased mortality regardless of treatment group, consistent with other COVID-19 survival analyses. Contradictory to previous reports, black race was associated with better overall survival compared to whites. There may have been inherent differences in clinically important covariates in this population that may have contributed to better survival which could not be further analyzed. Medicaid and self-pay insurance were associated with increased mortality. We speculate that this may be due to factors such as hospital admission later in disease course or socioeconomic disadvantages. For surrogates of illness severity, the need for IMV prior to T 0 was associated with increased mortality, while the need for vasopressors was not.

Prior diagnosis of ILD was associated with increased mortality, consistent with existing literature. 13 Surprisingly, those with comorbid hypertension had lower mortality which is contradictory to other reports. 1, 14 One study suggested that use of angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers via renin-angiotensin pathway modulation may confer a protective effect in the setting of CCS. 15 Our analysis did not include consideration of home medications. Alternatively, adjustments for covariates in our model may have uncovered an association between hypertension and COVID-19 outcomes which could be further investigated. Interestingly, increased mortality was associated with asthma, but not with COPD. Early in the pandemic, chronic lung disease, including asthma, was reported as one of the comorbidities associated with hospital admissions. 16 Later studies failed to demonstrate increased mortality in patients with asthma and COPD, 17 although preexistent asthma was reported to be associated with prolonged intubation time. Atopic asthma and treatment with inhaled corticosteroids were reported to correlate with lower sputum cell expression of ACE2, [18] [19] [20] implying decreased susceptibility and morbidity in these patients.

High D-dimer was significantly associated with in-hospital mortality. This is consistent with known evidence that elevated D-dimer is associated with worse outcomes 21 and predicts a higher chance of requiring intensive care unit (ICU) admission and increased 28-day mortality. 5, 22, 23 Thrombocytopenia is associated with severity of SARS-CoV-2 infection. 24 We also found thrombocytopenia to be associated with higher mortality. Both thrombocytopenia and elevated D-dimer reflect the known coagulopathy in COVID-19. 25 IL-6 is an important mediator of inflammation that plays an essential role in host response to viral infection (Chau et al., in press). Higher IL-6 levels were observed in patients with severe COVID-19 compared to mild disease. 3, 26 Therefore, tocilizumab was proposed early in the pandemic as a potential treatment for those with CCS. 27 Anti-IL-1 therapy has been an attractive choice in the treatment of COVID-19 due to its short half-life, safety and tolerability profile. IL-1β has been implicated in lung inflammation, fibrosis 36 and indirectly, with activation of the inflammatory cascade. [37] [38] [39] [40] A study examining cytokine kinetics during COVID-19 showed an IL-1 peak prior to the apex of respiratory distress and the surge of other inflammatory cytokines. 41 Anakinra has also been shown to improve survival in a subset of sepsis patients with hyperferritinemia and hepatobiliary dysfunction 10 when compared to placebo. 42 Small studies report improvement in clinical outcomes with use of anakinra in COVID-19. 8 In our study, although patients treated with anakinra in combination with corticosteroids had improved survival compared to 'SoC', anakinra alone did not. The dose of anakinra suggested in our health system protocol (100mg subcutaneous four times per day for three days, followed by a suggested taper) was modest in comparison to some of the above studies. The lack of benefit with anakinra may have been due to lower doses, delayed time to treatment and subcutaneous administration, leading to decreased drug availability, especially in the critically ill.

Biological effects of anakinra and tocilizumab are slower when compared to steroids. Also with anakinra, we observed a delay in drug initiation when combined with corticosteroids. This leads us to question whether the timing to drug administration and the time to onset of action influenced the outcome amongst our treatment groups. Statistical analysis of variation of drug administration across treatment groups was not feasible in this study. Further analysis of our data is needed to evaluate the effects of immunomodulatory treatments on disease progression, including rates of thrombosis.

Given the small sample sizes in the groups receiving tocilizumab or anakinra only, we should be cautious in interpreting the relative lack of survival advantage in these groups. To test the robustness of the model, we performed sensitivity analyses by removing groups with small sample sizes (those who received either 'A' or 'T'). The results remained consistent with those of the full model.

There were increased rates of bacteremia and fungemia in the steroid groups compared to standard of care (Supplemental e- Table 5 ). However, despite this increase in the infection rate, improved survival remained in these cohorts.

Although we were rigorous in our approach to the study design and data analysis, there are intrinsic limitations that preclude definitive conclusions in retrospective studies. While the effect of variability in systematic practices across the individual hospitals in the health system could not be evaluated, we did look at differences between tertiary versus community hospitals. Despite similar use of immunomodulatory therapies in tertiary and community centers, tertiary facilities had a higher survival. Potential explanations for this could include a greater number of ICU beds, sub-specialist availability or differences in patient demographics between hospitals.

Our study is the largest retrospective analysis to date reporting on outcomes comparing the use of immunomodulatory therapies such as corticosteroids, tocilizumab and anakinra in the treatment of COVID-19 CCS. Our findings suggest that patients receiving steroids and tocilizumab had the lowest mortality of all treatment groups. Corticosteroid use, either alone or in combination with tocilizumab or anakinra, was associated with lower hospital mortality compared to standard of care. A randomized clinical trial with head-to-head comparison of tocilizumab plus corticosteroids versus corticosteroids alone is warranted. Further investigation into the effect of dosing and timing of these drugs also needs to be elucidated. Results of the multivariable model of in-hospital mortality for COVID-19 patients meeting inclusion criteria with CCS. Hazard Ratios are presented with 95% Confidence Intervals and p-values in comparison to reference within each variable. Hazard ratios for treatment groups represent adjustment for covariates in the model, comparing to SoC treatment as reference. Treatment group hazard ratios are not adjusted for multiple comparisons between treatment groups. Refer to Figure 3 (and e- Table 6 ) for treatment differences using Tukey's adjustment for multiple comparisons between treatment groups.

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the

An interactive web-based dashboard to track COVID-19 in real time. The Lancet Infectious Diseases

Risk Factors Associated With Acute Respiratory Distress Syndrome and Death in Patients With Coronavirus Disease

COVID-19: consider cytokine storm syndromes and immunosuppression

Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan, China

Low-dose corticosteroid therapy does not delay viral clearance in patients with COVID-1

Tocilizumab for the treatment of severe COVID-19 pneumonia with hyperinflammatory syndrome and acute respiratory failure: A single center study of 100 patients in

Interleukin-1 blockade with high-dose anakinra in patients with COVID-19, acute respiratory distress syndrome, and hyperinflammation: a retrospective cohort study. The Lancet Rheumatology

Targeting the inflammatory cascade with anakinra in moderate to severe COVID-19 pneumonia: case series

Interleukin-1 Receptor Blockade Is Associated With Reduced Mortality in Sepsis Patients With Features of Macrophage Activation Syndrome: Reanalysis of a Prior Phase III Trial

Similar but not the same: Differential diagnosis of HLH and sepsis

Clinical and immunological features of severe and moderate coronavirus disease 2019

Acute kidney injury in patients hospitalized with COVID-19

Baseline Characteristics and Outcomes of 1591 Patients Infected With SARS-CoV-2 Admitted to ICUs of the Lombardy Region

Renin-angiotensin system inhibitors improve the clinical outcomes of COVID-19 patients with hypertension

Hospitalization rates and characteristics of patients hospitalized with laboratoryconfirmed coronavirus disease 2019-COVID-NET, 14 States

Association of cardiovascular disease and 10 other pre-existing comorbidities with COVID-19 mortality: A systematic review and metaanalysis

Expression of SARS-CoV-2 Receptor ACE2 and Coincident Host Response Signature Varies by Asthma Inflammatory Phenotype

COVID-19 related genes in sputum cells in asthma: relationship to demographic features and corticosteroids

Association of respiratory allergy, asthma, and expression of the SARS-CoV-2 receptor ACE2

D-dimer is Associated with Severity of Coronavirus Disease 2019: A Pooled Analysis

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Abnormal coagulation parameters are associated with poor prognosis in patients with novel coronavirus pneumonia

Thrombocytopenia is associated with severe coronavirus disease 2019 (COVID-19) infections: A meta-analysis

Mechanism of thrombocytopenia in COVID-19 patients

Interleukin-6 in Covid-19: A systematic review and meta-analysis

Why tocilizumab could be an effective treatment for severe COVID-19?

Rational Use of Tocilizumab in the Treatment of Novel Coronavirus Pneumonia

Tocilizumab for Treatment of Severe

Patients: Preliminary Results from SMAtteo COvid19 REgistry (SMACORE)

Tocilizumab among patients with COVID-19 in the intensive care unit: a multicentre observational study. The Lancet Rheumatology

Tocilizumab in patients with severe COVID-19: a retrospective cohort study

Prolonged glucocorticoid treatment is associated with improved ARDS outcomes: analysis of individual patients' data from four randomized trials and trial-level meta-analysis of the updated literature

Dexamethasone in Hospitalized Patients with Covid-19 -Preliminary Report

Association Between Administration of Systemic Corticosteroids and Mortality Among Critically Ill Patients With COVID-19: A Meta-analysis

Transient expression of IL-1beta induces acute lung injury and chronic repair leading to pulmonary fibrosis

Inflammation. Neutrophil extracellular traps license macrophages for cytokine production in atherosclerosis

Novel Role of IL (Interleukin)-1beta in Neutrophil Extracellular Trap Formation and Abdominal Aortic Aneurysms

Extracellular DNA, Neutrophil Extracellular Traps, and Inflammasome Activation in Severe Asthma

The authors thank the Northwell COVID-19 Research Consortium for facilitating the study.