key: cord-1024982-jz7zxgx8
authors: Cao, Yang; Wei, Jia; Zou, Liang; Jiang, Tiebin; Wang, Gaoxiang; Chen, Liting; Huang, Liang; Meng, Fankai; Huang, Lifang; Wang, Na; Zhou, Xiaoxi; Luo, Hui; Mao, Zekai; Chen, Xing; Xie, Jungang; Liu, Jing; Cheng, Hui; Zhao, Jianping; Huang, Gang; Wang, Wei; Zhou, Jianfeng
title: Ruxolitinib in treatment of severe coronavirus disease 2019 (COVID-19): A multicenter, single-blind, randomized controlled trial
date: 2020-05-26
journal: J Allergy Clin Immunol
DOI: 10.1016/j.jaci.2020.05.019
sha: 4d004ae7cae3787cf758a366e6fac77206cc8ef8
doc_id: 1024982
cord_uid: jz7zxgx8

Abstract Background Accumulating evidence proposed JAK inhibitors as therapeutic targets warranting rapid investigation. Objective This study evaluated the efficacy and safety of ruxolitinib, a Janus-associated kinase (JAK1/2) inhibitor, for COVID-19. Methods We conducted a prospective, multicenter, single-blind, randomized controlled phase II trial involving patients with severe COVID-19. Results Forty-three patients were randomly assigned (1:1) to receive ruxolitinib plus SoC treatment (22 patients) or placebo based on SoC treatment (21 patients). After exclusion of 2 patients (1 ineligible, 1 consent withdrawn) from the ruxolitinib group, 20 patients in intervention group and 21 patients in control group were included in the study. Treatment with ruxolitinib plus SoC was not associated with significantly accelerated clinical improvement in severe patients with COVID-19, although ruxolitinib recipients had a numerically faster clinical improvement. Eighteen (90%) patients from the ruxolitinib group showed CT improvement at D14 compared with 13 (61.9%) patients from the control group (P = 0.0495). Three patients in the control group died of respiratory failure, with 14.3% overall mortality at D28; no patients died in the ruxolitinib group. Ruxolitinib was well tolerated with low toxicities and no new safety signals. Levels of 7 cytokines were significantly decreased in the ruxolitinib group in comparison to the control group. Conclusions Although no statistical difference was observed, ruxolitinib recipients had a numerically faster clinical improvement. Significant chest CT improvement, a faster recovery from lymphopenia and favorable side-effect profile in ruxolitinib group were encouraging and informative to future trials to test efficacy of ruxolitinib in a larger population. This trial is registered at www.chictr.org.cn as ChiCTR-OPN-2000029580.

The end of 2019 witnessed an outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-94

2) infection and its associated coronavirus disease 2019 in Wuhan, China 1, 2 . The rapid 95 global spread has been classified as a pandemic by the World Health Organization 3 and has now 96 represented the most serious issue to public health globally. COVID-19 is a heterogeneous disease 97 population. Most patients are asymptomatic or exhibit mild to moderate symptoms 4-6 , but around 15% 98 progress to severe pneumonia and about 5% were eventually admitted to the intensive care unit (ICU) due 99 to acute respiratory distress syndrome (ARDS), septic shock and/or multiple organ failure 7-9 . Among 138 100 hospitalized patients with 26 .1% of patients were transferred to the ICU due to deteriorated 101 complications that mainly (61.1%) included acute respiratory distress syndrome (ARDS) 7 . Patients who 102 develop ARDS respond poorly to therapy and have an extremely dismal prognosis 8, 10 . 103

So far, only remdesivir showed to accelerate recovery from advanced COVID-19 based on a 104 preliminary data analysis 11 , while supportive therapies still have a fundamental role in the treatment of 105 COVID-19. Among the unmet medical needs related to COVID-19, one of the most urgent issues is to 106 assess existing conventional drugs in the treatment of severe/critical COVID-19 to improve the 107 unsatisfactory clinical outcomes. COVID-19, like SARS-CoV and MERS, are characterized by an 108 exuberant cytokine storm [12] [13] [14] [15] [16] . Upon virus infection, the body produces inflammatory cytokines to restrict 109 the spread/replication of the virus and eliminate the virus. However, highly pathogenic coronaviruses 110 often induce uncontrolled cytokine/chemokine response, known as a cytokine storm, which results in high 111 morbidity and mortality due to immunopathology 17 . While virus-induced direct pathogenic effects have 112 an essential role in disease severity, viral load is not correlated with the worsening of symptoms in SARS 113 [18] [19] [20] . Previous studies on SARS suggested that a dysregulated immune response results in an exuberant 114 inflammation and lethal disease 21 . In a recent study, which enrolled 41 cases with confirmed 115 one-third of patients were admitted to intensive care units, 10% patients required mechanical ventilation, 116 and six eventually died (14.6%); among these patients, cytokine storm was found to be associated with 117 disease severity 18 . There is accumulating evidence on the key pathophysiological role of cytokines during 118 the severe stage of COVID-19. In the context of lack of vaccine and specific antiviral agents, testing of 119 immunomodulatory agents to reduce excessive or uncontrolled inflammation before it results in 120 irreversible multi-organ dysfunction infection has received increasing research attention. 121

Ruxolitinib is a Janus-associated kinase (JAK1/2) inhibitor approved by the U.S. Food and Drug 122 Administration (FDA) and the European Medicines Agency for the treatment of polycythemia vera and 123 myelofibrosis 22 . It is also a promising option in the treatment of steroid-refractory acute graft-versus-host 124 disease after allogeneic hematopoietic stem cell transplantation 23 or secondary hemophagocytic 125 lymphohistiocytosis 24 by targeting the deleterious effects of aberrant host inflammatory response. Anemia 126 was the most common adverse event in patients receiving ruxolitinib though most anemia events were 127 mild to moderate in severity 25 . Non-hematological adverse events were generally low with long-term 128 ruxolitinib treatment 25 . Accordingly, we hypothesized that ruxolitinib might be effective against the 129 consequences of the elevated levels of cytokines in COVID-19 patients. Besides, the potential negative 130 impact of ruxolitinib on virus clearance and SARS-CoV-2 antibody production also need to be elucidated. 131

To evaluate the efficacy and safety of ruxolitinib for COVID-19, we conducted a randomized, 132 multicenter, placebo-controlled, single-blind phase II trial in patients hospitalized with severe COVID-19. 133

This prospective, single-blind, randomized controlled phase II trial included participants with COVID-19 137 enrolled for screening between February 9, and February 28, 2020 across three hospitals including Tongji 138 hospital, No.1 hospital and the Third Xiangya hospital in China. The original protocol included secondary 139 randomization in the treatment group for infusion of mesenchymal stem cells if the patient's clinical 140 response deteriorated at D 7 after the first randomization. Since no patients from the treatment group experienced deterioration at D 7 , secondary randomization was unnecessary after the first randomization, 142 and the protocol was updated correspondingly. This study followed the principles of the Declaration of 143

Helsinki and approved by the Medical ethics committees of Tongji hospital, Wuhan No.1 hospital and 144

Third Xiangya hospital. 145

Participants who met the following inclusion criteria were included in the study: (1) met the diagnostic 147 criteria for COVID-19; (2) 18 years or older and younger than 75 years; (3) severe cases. The diagnosis 148 and the illness severity of COVID-19 were defined according to the Chinese management guideline for 149 COVID-19 (version 5.0) 26 and the full translated edition of diagnostic criteria is available in 150

Supplementary Methods section in the Online Repository at www.jacionline.org. Exclusion criteria were 151 following: (1) patients with concomitant malignant tumors; (2) patients with severe cardiovascular and 152 metabolic disease that is not medically controlled; (3) patients with a mental or severe psychiatric 153 disorder; (4) patients in need of invasive mechanic ventilation at recruitment; (5) patients who could not 154 guarantee to complete all the scheduled treatment plans and follow-ups; (6) women of child-bearing age 155 with positive pregnancy tests or those in the lactating period; (7) patients whose condition was further 156 complicated with other active infections. Written informed consent was obtained from all participants. 157

The enrolled patients were randomly allocated into two groups (1:1 allocation ratio) by an independent 159 statistician using permuted blocks of 4 for all sites. The whole process of randomization was masked to 160 all treating physicians. Patient unique identification number and treatment allocation codes were provided 161 by a clinical research associate in sequentially numbered opaque envelopes. Treating physicians were 162 aware of group allocations for safety concern while the enrolled participants, the staff at trial sites, CT 163 radiologists, and laboratory personnel were masked to the trial group assignment.

The first day of randomization was designated as D 0 . The second and the fourth day after randomization 166 were designated D 1 and D 3 , respectively. D end was the day before discharge. The enrolled patients were 167 randomly separated into two groups: the treatment group (group B), which received oral intake of 168 ruxolitinib 5mg twice a day plus standard-of-care (SoC); the control group (group A), which was treated 169 with placebo (100mg vitamin C) twice a day with SoC. The dose of 5mg twice a day is a ruxolitinib dose 170 frequently used for treatment of autoimmune/inflammatory conditions and has shown effective inhibition 171 of inflammation proteins in previous trials 27 . All costs of ruxolitinib and vitamin C tablets were covered 172 by the funding from principle investigators. The SoC included antiviral therapy, supplemental oxygen, 173 noninvasive and invasive ventilation, corticosteroid, antibiotic agents, vasopressor support, renal-174 replacement therapy, and extracorporeal membrane oxygenation (ECMO). The safety profile was 175 monitored daily by two senior physicians from the safety monitoring committee of trial center. Adverse 176 events were classified according to the National Cancer Institute Common Terminology Criteria for 177 Adverse Events, version 5.0 28 . Non-contrast enhanced chest CT examinations were performed on D 0 and 178 should be followed-up at D 14 . Additional chest CT might be performed if the condition deteriorated. All 179 CT images were reviewed using Picture archiving and communication system (PACS). The CT features, 180 which were blindly evaluated by two independent senior radiologists, were supposed to meet at least one 181 of the following criteria to be considered as improved: the decreased presence of ground-glass opacities 182 (GGO), decreased lung opacification, reduced density of consolidation or decreased pleural effusion with 183 existence of fibrous stripes 29 . Disease progression is defined by comparing the scope, quantity, and 184 density of lesions detected in two chest CT scans. Obviously fusion of lesions, new lesions, and (or) 185 increased lung density was considered progression 30 . If no significant difference in the lesions between 186 the two CT independent senior radiologists was observed, the patient's condition was considered stable 30 . 187

Peripheral blood was taken from patients on D 1 , D 3 and D end for the determination of viral load by RT-188 ddPCR, SARS-CoV-2-specific IgM, and IgG and 48 cytokines (see the Supplementary Methods section 189 in the Online Repository at www.jacionline.org). Data from each participant were filled in one case 190 record form (CRF). All the CRF tables were input and saved by researchers into an electronic data capture system (EDCS) and validated by trial staff, including demographics, medical history, daily clinical 192 findings, oximetric measurements and laboratory data involving complete blood count, serum 193 biochemical parameters and high sensitivity C-reactive protein (hsCRP), etc. 194

The primary efficacy end point was the time to clinical improvement, defined as the time from clearance and its specific IgM and/or IgG-antibody formation and/or lymphocyte recovery were also 207 included into safety profile. Lymphopenia was defined as peripheral absolute lymphocytes were less than 208 1.0×10 9 /L. Lymphocyte recovery time was defined as the first day at which lymphocytes returned to the 209 normal levels within the observation period. The virus clearance time was defined as the time from 210 randomization to the first day of at least two consecutive negative RT-PCR assays separated by 24 h apart. 211

The secondary endpoint is the overall mortality at D 28. The investigational outcomes included the 212 dynamic changes of the virus copies, cytokine profile, SARS-CoV-2-speicific antibody and its correlation 213 with clinical treatment response. 214

The trial was initiated in rapid response to COVID-19 public health emergency. Since limited information 216 about clinical outcomes in hospitalized patients with COVID-19 was available at that time, we estimated 217 the sample size in two different ways. We assumed that the median clinical improvement for treatment 218 group is 7 days while control group is about 15 days and the estimated sample size was set at 70 to 219 provide the trial with 80% power (α=0.05). We also hypothesized that there was 40% difference gap in 220 term of CT improvement at D 14 between the two groups with approximately 90% patients had CT 221 improvement in treatment group. Accordingly, the estimated sample size was set at 40 to provide the trial 222 with 80% power (α=0.05). We decided to set the final sample size to 70 cases to reduce the possibility of 223 underpower of the trial. The number of cases in the experimental group and control group is 1:1. The 224 planned enrollment of 40 patients in the trial was accomplished quickly. Due to the slowing down of 225 pandemic in China and the fact that few newly diagnosed patients are available, recruitment to the trial 226 was stopped earlier than expected. 227

Continuous variables were expressed as median (IQR) and compared with the unpaired 2-sided 228 student's t test; categorical variables were expressed as number (%) and compared by χ 2 test or Fisher's 229 exact test. A modified intention-to-treat analysis that excluded two patients (1 ineligible, 1 consent 230 withdrawn) who did not take ruxolitinib in ruxolitinib group were performed. For primary endpoint, the 231 time to clinical improvement was portrayed by Kaplan-Meier plot and compared using a log-rank test. 232

Hazard ratios with 95% confidence intervals were calculated using Mantel-Haenszel approach. The 233 improvement rates of CT scan at D 14 were compared using Wilcoxon rank sum test. The clinical 234 improvement at D 7 , D 14 and D 21 , time to clinical deterioration, clinical deterioration at D 7 and D 14 and 235 mortality rate at D 28 were compared using the Fisher's exact tests. Time from randomization to discharge, 236 to death, to lymphocyte recovery, and to virus clearance time were portrayed by Kaplan-Meier plot and 237 compared using a log-rank test. For comparing serum level of cytokines, anti-SARS-CoV-2 specific 238 antibody and virus copy numbers, mean ± standard error of mean (SEM) is given for continuous variables, 239 median and ranges are given for variables that were not normally distributed. Means were compared using 240 t tests for normally distributed continuous variable. Otherwise, the Mann-Whitney U test was used. All 241 statistical analyses were performed using SPSS (Statistical Package for the Social Science) version 13·0 242 software (SPSS Inc.). P-value < 0.05 (two-tailed) was considered statistically significant. 243

The funder had no role in study design, data collection, data analysis, data interpretation or writing of the 245 report. The corresponding author had full access to all data in the study and final responsibility for the 246 decision to submit the study for publication. 247

Among a total of 58 individuals who were screened for eligibility between February 9 and February 249 28, 2020, 43 patients were randomly assigned to receive ruxolitinib plus SoC treatment (22 patients, 250 ruxolitinib group) or placebo based on SoC treatment (21 patients, control group). Fifteen patients were 251 excluded from the study including ten of them participated in other clinical trials and five of them refused 252 to sign a written informed consent. After randomization, two patients were excluded from the ruxolitinib 253 group since one was found to be ineligible because of persistent humoral immune deficiency post B cell 254 mature antigen (BCMA) targeting chimeric antigen receptor (CAR) T cell therapy, while another 255 withdrew the consent before treatment started ( Figure 1 ). Their data were not included in the analyses. 256

The demographic and clinical characteristics of the patients at baseline are outlined in Table 1 . At 257 baseline, the median age of patients was 63 years (interquartile range [IQR] , 58 to 68 years), ranging from 258 32 years to 75 years, and 58.5% of the patients were male. The median interval time from symptom onset 259 to randomization was 20 days. There was no significant difference between two groups in demographic 260 characteristics, baseline laboratory test results, distribution of ordinal scale scores, or National Early 261

Warning Score 2 (NEWS2) scores at enrollment (Table 1 and Table 2 ). During the study, the use of 262 systemic corticosteroids was balanced between ruxolitinib group (70.0%) and control group (71.4%). The 263

proportion of patients received antivirals were balanced between two groups (90.0 % in the ruxolitinib 264 group versus 90.5% in the control group, Figure 2A) . 269

Eighteen (90%) patients from the ruxolitinib group showed significant improvement on the follow-up 270 chest CT scans at D 14 compared with only 13 (61.9%) patients from the control group (P = 0.0495, Table  271 D 21 between two groups, however, the percentage was numerically higher at D 7 , D 14 and D 21 in the 273 ruxolitinib group than in the control group. A total of 4 patients in control group experienced clinical 274 deterioration. Three of four were transferred to the ICU and required invasive mechanical ventilation. The 275 cumulative improvement rate was compared between the two groups ( Figure 2A) . 276

For secondary endpoint, 3 patients from the control group eventually died of respiratory failure. The 277 28-day overall mortality was 14.3% in the control group, while no patients died in the ruxolitinib group. 278

The median time from randomization to death was 15 days (IQR, 4~19) in the control group. There was 279 no significant difference in the days from randomization to discharge between two groups (Table 3 , P = 280 0.941). The cumulative incidence of death was compared between the two groups ( Figure 2B , P = 0.089). 281

For primary safety endpoints, a total of 16 patients (80%) in the ruxolitinib group and 15 patients 282 (71.4%) in the control group reported adverse events from randomization to D 28 . All adverse events are 283 summarized in Table 4 . There was no significant difference in the total number of adverse events of any 284 grade in hematological, non-hematological toxicities and chemical laboratory abnormalities between the 285 two groups (see Table E1 , in the Online Repository at www.jacionline.org). One patient in the ruxolitinib 286 group developed Grade 3 lymphocytopenia after taking ruxolitinib for 4 days, which improved in two 287 days without interrupting ruxolitinib treatment (see Figure E1A in the Online Repository at 288 www.jacionline.org). Grade 3 hypertension developed in one patient in the ruxolitinib group during the 289 study and was judged by the investigators to be related to the trial medication and was transient and 290

reversible. All serious adverse events, including secondary infection, sepsis, shock, and acute heart failure, 291 only occurred in the control group. To address the concern that whether ruxolitinib has negative influence 292 on SARS-CoV-2 clearance, specific SARS-CoV-2 antibody production and lymphocyte recovery, a total 293 of 17 patient (8 patients in the ruxolitinib group and 9 patients in the control group) who had a positive 294 RT-PCR results at D 0 on the throat swab were serially followed up. Patients in ruxolitinib group had 295 similar median time of virus clearance (13 [IQR 5 -16] days versus 12 [IQR 3-16] days, log-rank test P = 296 0.649, hazard ratio, 1.279; 95% CI, 0.443 to 3.692) compared with patients in control group ( Figure 2C ).

One step RT-ddPCR was used to further evaluate the clearance of SARS-CoV-2, the mean (± SEM) 298 baseline blood viral RNA loads at D 1 in ruxolitinib group were comparable with control group ( Figure 2D , 299 94 ± 26 copies per milliliter versus 102 ± 21, P = 0.565). The viral load at discharge did not differ 300 between the ruxolitinib recipients and those receiving SoC alone ( Figure 2E , P = 0.631). Interestingly, the 301 peak level of anti-IgM of SARS-CoV-2 is profoundly higher in ruxolitinib group than in control group, 302 while no significant difference was found in peak IgG between the two groups (Figure2F-G). A total of 303 21 patients (9 patients in the control group and 12 patients in the ruxolitinib group) were found to have 304 lymphopenia at or after enrollment. Patients in the ruxolitinib group had a significantly shorter median 305 time of recovery from lymphopenia compared to those in the control group. ( We investigated whether ruxolitinib could inhibit cytokines downstream of JAK by assessing the 308 levels of 48 cytokines in the serum of patients who received ruxolitinib and controls. As shown in Figure  309 3A, in control group, the patients' average value of 44 cytokines decreased after SoC therapy while the 310 other four including macrophage inflammatory protein 1α (MIP-1α), granulocyte colony stimulating 311 factor (G-CSF), interferon-α2 (IFN-α2) and interleukin-1α (IL-1α) increased on D 3 . On the contrary, all 312 average values from 48 cytokines decreased in patients from the ruxolitinib group on D 3 . Furthermore, the 313 average fold-change in the ruxolitinib group was 0.466, while it was 0.739 in the control group. The ratios 314 were significantly lower in the ruxolitinib group (P < 0.0001). Moreover, the levels of seven cytokines 315 including interleukin-6 (IL-6), nerve growth factor β (NGF-β), interleukin-12 (IL-12) (p40), macrophage 316 migration inhibitory factor (MIF), MIP-1α, macrophage inflammatory protein 1β (MIP-1β) and vascular 317 endothelial growth factor (VEGF) were markedly decreased in the ruxolitinib group but not in the control 318 group (Figure 3 B-H) . In correlation with the reduction of these cytokines, a significant reduction of 319 hsCRP was observed in the ruxolitinib group but not in control group ( Figure 3I ). While all the three 320 ruxolitinib recipients with fever had rapid resolution of fever in two days, fever resolution took 4 or 5 321 days in two patients with fever in control group. Furthermore, there was no significant difference in the 322 levels of IFN-α2 and IFN-γ, two important cytokines protecting hosts against virus, between the 323 ruxolitinib group and control group ( see Figure E1B -C in the Online Repository at www.jacionline.org). 324

This randomized trial found that ruxolitinib added to SoC treatment was not associated with significantly 326 accelerated clinical improvement in severe patients with COVID-19, although ruxolitinib recipients had a 327 numerically faster clinical improvement compared to the control group. Ruxolitinib recipients showed 328 significantly faster improvement in the chest CT at D 14 compared to the control group (18 (90%) versus 329 13 (61.9%), P = 0.0495). The 28-day mortality was 14.3% in the comparison group. No death or 330 deterioration occurred in ruxolitinib recipients. These data provide a rationale for further trials to 331 determine whether ruxolitinib treatment can reduce the overall incidences of deterioration and death. 332

Patients treated with ruxolitinib showed a significantly shorter lymphocyte recovery than those in control 333 group (5 [IQR 2-7] days versus 8 [IQR 2-11] days, P = 0.033). We assume that a faster recovery from 334 lymphopenia is of clinical relevance since lymphopenia was associated with poor prognosis. A shorter 335 duration of lymphopenia in ruxolitinib recipients was consistent with a higher mean peak level of IgM 336 specific for SARS-CoV-2 in patients treated with ruxolitinib. Among 4 patients from the control group 337 who experienced clinical deterioration, 3 were transferred to the ICU and required invasive mechanical 338 ventilation. Three patients in the control group eventually died of respiratory failure. The demographic 339 and clinical characteristics of the patients were balanced between two groups at enrollment. The use of 340 corticosteroids and antivirals were comparable between the control and ruxolitinib group. Therefore, it 341 was unlikely that the baseline characteristics and treatments of the two groups would affect the endpoints 342 of our study. 343

The present randomized trial also found that ruxolitinib with SoC treatment was well tolerated with 344 low hematological and non-hematological toxicities. All ruxolitinib recipients completed the full course 345 of administration until discharge, while the control group needed more intensive supportive treatments 346 after enrollment due to the deterioration in some cases. The addition of ruxolitinib based on SoC did not increase the risk of adverse events compared to the control group. The overall incidences of adverse 348 events were similar between the two groups. Interestingly, while most of the adverse events occurred at 349 grade 1-2, adverse events at grade 3-4 and serious adverse events were more common in the control group 350 due to the progressive deterioration of COVID-19 in this group. Among all ruxolitinib recipients, only 351 two adverse events at grade 3 occurred and were transient and reversible. There were no unexpected 352 adverse events and previously unknown events in ruxolitinib recipients. One of the major concerns related 353

to the use of ruxolitinib in treatment of COVID-19 is its therapeutic action in reducing systemic 354 inflammation, and potential to unfavorably delay the clearance of viral loads and impair the generation of 355 SARS-CoV-2-specific antibodies. In the current study, there was no significant difference in viral RNA 356 loads or duration as well as IFN-α2 and IFN-γ levels between ruxolitinib recipients and the control group. 357 Interestingly, the mean peak level of IgM specific for SARS-CoV-2 was profoundly higher in the 358 ruxolitinib group compared to the control group, while no significant difference was found in the mean 359 peak IgG against SARS-CoV-2 between the two groups. The favorable side-effect profile observed in the 360 current trial provides a rationale for the initiation of a large-scale clinical trial at the same or higher 361 ruxolitinib dose regimens in efforts to improve outcomes. 362

In the present study, we found that the addition of ruxolitinib to SoC could significantly mitigate 363 exuberant cytokine storm featured in severe COVID-19, which justified the use of ruxolitinib for 364 reduction of systemic inflammation. In two published autopsy reports 33, 34 . The severe immune injury was 365 also involved in other organs without obvious viral inclusions were identified, thus indicating the 366 important role of cytokines storm instead of the direct viral damage to the whole body. The infiltrated 367 immune cells in alveoli were mostly macrophages and monocytes, which was in accordance with our 368 findings on cytokines changes. In particular, the levels of seven cytokines, including IL-6, NGF-β, IL-369 12(p40), MIF, MIP-1α, MIP-1β, and VEGF, were markedly decreased in patients who received 370 ruxolitinib in comparison to control group. Among these cytokines, IL-6 has been reported as a critical 371 cytokine driving pro-inflammatory activity in cytokine-mediated organ dysfunction and tissue damage 35 372 and IL6-directed therapy as the cornerstone of cytokine-based therapy after CAR-T cell therapy 36, 37 . IL-373 12(p40), MIP-1α, and MIP-1β are critical chemokines for the recruitment of activated 374 monocytes/macrophages and other cells to the site of infection [38] [39] [40] . VEGF, which has been reported to 375 recruit monocytes/macrophages, participates in increased capillary permeability syndrome that 376 characterizes some types of viral pneumonia 41 . These results indicate that ruxolitinib may exert its 377 inhibitory effect by targeting multiple critical cytokines rather than any specific cytokine, and these 378 cytokines could be employed as surrogate biomarkers in future ruxolitinib trials. 379

The present study has several limitations. Firstly, the sample size was small due to no eligible 380 patients available at the end of the pandemic at our trial centers and the few endpoints reached statistical 381 significance. The safety profile during this study was favorable, but further testing in larger patient 382 cohorts with different ethnicity or disease status is required. Secondly, there are some limitations related 383 to the ordinal scale that was used to evaluate primary endpoints. Due to the severity of the epidemic, even 384 if the patients had significant improvement in the CT scan as well as clinical symptoms, they still asked 385 for NC oxygen (less than 2L/min) until being discharged from hospital, which may contribute to the non-386 statistically significant p-value of clinical improvement. Finally, critically ill patients or patients with 387 invasive ventilator dependence were not included in this study due to the lack of previous data and our 388 concerns on the unknown safety profile of ruxolitinib treatment in pneumonia. Therefore, our conclusion 389 is confined to patients with severe COVID-19. Nevertheless, this study is the first randomized controlled 390 trial on the use of ruxolitinib in patients with severe COVID-19 based on a novel therapeutic rationale. 391

These findings are hypothesis-generating and require additional larger controlled studies to confirm the 392 possibility of a treatment benefit of ruxolitinib. However, these early data were promising and informative 393 to future trials with ruxolitinib or other JAK1/2 inhibitors. 394

We appreciate eStart Medical Technology. CO., Ltd., Professor & Dr Jie Hou, Meng Chen, Zhongyan 397 Zhang, Xu Ji, Chun Li, etc Control group (N=21) Ruxotlitinib group (N=20) Any Grade

Grade 1-2 Grade 3-4 Any Grade Grade 1-2 Grade 3-4 Hematological adverse events 12 (57.1%) 10 (47.6%) 2 (9.5%) 13 (65.0%) 12 (60.0%) 1 (5.0%) Neutrocytopenia 1 (4.8%) 1 (4.8%) 0 1 (5.0%) 1 (5.0%) 0

Lymphocytopenia 4 (19.0%) 3 (14.3%) 1 (4.8%) 2 (10.0%) 1 (5.0%) 1 (5.0%) Anemia 9 (42.9%) 8 (38.1%) 1 (4.8%) 11 (55.0%) 11 (55.0%) 0

Thrombocytopenia 3 (14.3%) 2 (9.5%) 1 (4.8%) 4 (20.0%) 4 (20.0%) 0

Chemical laboratory abnormalities 7 (33.3%) 7 (33.3%) 0 10 (50.0%) 10 (50.0%) 0 ALT increase 2 (9.5%) 2 (9.5%) 0 7 (35.0%) 7 (35.0%) 0 AST increase 1 (4.8%) 1 (4.8%) 0 3 (15.0%) 3 (15.0%) 0 Alkaline phosphatase increase 1 (4.8%) 1 (4.8%) 0 2 (10.0%) 2 (10.0%) 0 γ-GT increase 2 (9.5%) 2 (9.5%) 0 2 (10.0%) 2 (10.0%) 0 Hypoalbuminemia 3 (14.3%) 3 (14.3%) 0 1 (5.0%) 1 (5.0%) 0 Hypercholesterolemia 4 (19.0%) 4 (19·0%) 0 4 (20.0%) 4 (20.0%) 0 Hypertriglyceridemia 2 (9.5%) 2 (9.5%) 0 0 0 0 Hypokalemia 1 (4.8%) 1 (4.8%) 0 1 (5.0%) 1 (5.0%) 0 Hypochloremia 2 (9.5%) 2 (9.5%) 0 1 (5.0%) 1 (5.0%) 0 Hypocalcemia 2 (9.5%) 2 (9.5%) 0 1 (5.0%) 1 (5.0%) 0 Adverse events 6 (28.6%) 6 (28.6%) 0 7 (35.0%) 7 (35.0%) 0 Headache 0 0 0 1 (5.0%) 1 (5.0%) 0 Dizziness 1 (4.8%). 1 (4.8%) 0 2 (10.0%) 2 (10.0%) 0

Rash 1 (4.8%) 1 (4.8%) 0 2 (10.0%) 2 (10.0%) 0 Nausea 2 (9.5%) 2 (9.5%) 0 2 (10.0%) 2 (10.0%) 0 Decreased appetite 2 (9.5%) 2 (9.5%) 0 1 (5.0%) 1 (5.0%) 0 Hypertension 2 (9.5%) 2 (9·5%) 0 1 (5.0%) 0 1 (5.0%) Serious adverse events 4 (19.0%) 0 4 (19.0%) 0 0 0 Secondary Infection 2 (9.5%) 0 2 (9.5%) 0 0 0 Acute heart failure 2 (9.5%) 0 2 (9.5%) 0 0 0 28 Shock 2 (9.5%) 0 2 (9.5%) 0 0 0 Sepsis 1 (4.8%) 0 1 (4.8%) 0 0 0 *Adverse events that occurred in more than 1 patient after randomization through day 28 are shown. Some patients had more than one adverse 510 event. The proportions of patients with values worse than baseline values are listed here. All deaths were due to respiratory failure. 511 

Early Transmission Dynamics in Wuhan, China, 404 of Novel Coronavirus-Infected Pneumonia

A Novel Coronavirus from Patients with 406 Pneumonia in China

Covid-19: WHO declares pandemic because of "alarming levels" of spread, severity, 408 and inaction

Asymptomatic carrier state, acute 410 respiratory disease, and pneumonia due to severe acute respiratory syndrome coronavirus 2 411 (SARS-CoV-2): Facts and myths

Clinical Characteristics of Coronavirus 413 Disease 2019 in China

Clinical outcome of 55 asymptomatic cases at the 415 time of hospital admission infected with SARS-Coronavirus-2 in Shenzhen

Clinical Characteristics of 138 Hospitalized 418 Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan, China. JAMA 2020

Clinical Features of 85 Fatal Cases of COVID-19 420 from Wuhan: A Retrospective Observational Study

Epidemiological and clinical characteristics 422 of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study

Imaging and Clinical Features of Patients With

CoV-2: A systematic review and meta-analysis

Temporal changes in cytokine/chemokine 430 profiles and pulmonary involvement in severe acute respiratory syndrome

Persistence of lung inflammation and 433 lung cytokines with high-resolution CT abnormalities during recovery from SARS

Plasma inflammatory cytokines and 436 chemokines in severe acute respiratory syndrome

Analysis of serum cytokines in patients with 438 severe acute respiratory syndrome

The cytokine release syndrome (CRS) of severe COVID-440 19 and Interleukin-6 receptor (IL-6R) antagonist Tocilizumab may be the key to reduce the 441 mortality

Pathogenic human coronavirus infections: causes and 443 consequences of cytokine storm and immunopathology

Clinical features of patients infected with 2019 445 novel coronavirus in Wuhan, China

Clinical progression and viral load 447 in a community outbreak of coronavirus-associated SARS pneumonia: a prospective study

COVID-19: combining antiviral 450 and anti-inflammatory treatments

Multiple organ infection and the 452 pathogenesis of SARS

Ruxolitinib treatment for SR-aGVHD in patients with 456 EBV-HLH undergoing allo-HSCT

Ruxolitinib in 458 adult patients with secondary haemophagocytic lymphohistiocytosis: an open-label, single-459 centre, pilot trial

Long-term efficacy and safety 461 of ruxolitinib versus best available therapy in polycythaemia vera (RESPONSE): 5-year follow up 462 of a phase 3 study

Ruxolitinib for the 468 treatment of steroid-refractory acute GVHD (REACH1): a multicenter, open-label, phase 2 trial. 469 Blood 2020. 470 28. National Cancer Institute. Common Terminology Criteria for Adverse Events 471 v5

Ai Tao， Tang Dazhong. COVID-19 pneumonia: CT imaging evolution 474 characteristics of thoracic lesions in recovery stage

CT imaging changes of corona virus disease 476 2019(COVID-19): a multi-center study in Southwest China

A Trial of Lopinavir-Ritonavir in Adults 478 Hospitalized with Severe Covid-19

Pathological findings of COVID-19 483 associated with acute respiratory distress syndrome

The Immunology of Macrophage Activation 487 Syndrome

Tocilizumab for the treatment of chimeric antigen receptor T 489 cell-induced cytokine release syndrome

Management of cytokine release 491 syndrome related to CAR-T cell therapy

Chemokines and Chemokine Receptors 493 in Lymphoid Tissue Dynamics

IL-12p40: an inherently agonistic cytokine

Macrophage inflammatory protein-1

VEGF receptor signalling -in control of 499 vascular function