Tocilizumab and steroid treatment in patients with severe Covid-19 pneumonia.

Introduction Coronavirus disease 2019 (COVID-19) can lead to respiratory failure due to severe immune response. Treatment targeting this immune response might be beneficial but there is limited evidence on its efficacy. The aim of this study was to determine if early treatment of patients with COVID-19 pneumonia with tocilizumab and/or steroids was associated with better outcome. Methods This observational single-center study included patients with COVID-19 pneumonia who were not intubated and received either standard of care (SOC, controls) or SOC plus early (within 3 days from hospital admission) anti-inflammatory treatment. SOC consisted of hydroxychloroquine 400mg bid plus, in those admitted before March 24th, also darunavir/ritonavir. Anti-inflammatory treatment consisted of either tocilizumab (8mg/kg intravenously or 162mg subcutaneously) or methylprednisolone 1 mg/kg for 5 days or both. Failure was defined as intubation or death, and the endpoints were failure-free survival (primary endpoint) and overall survival (secondary) at day 30. Difference between the groups was estimated as Hazard Ratio by a propensity score weighted Cox regression analysis (HROW). Results Overall, 196 adults were included in the analyses. They were mainly male (67.4%), with comorbidities (78.1%) and severe COVID-19 pneumonia (83.7%). Median age was 67.9 years (range, 30-100) and median PaO2/FiO2 200 mmHg (IQR 133-289). Among them, 130 received early anti-inflammatory treatment with: tocilizumab (n=29, 22.3%), methylprednisolone (n=45, 34.6%), or both (n=56, 43.1%). The adjusted failure-free survival among tocilizumab/methylprednisolone/SOC treated patients vs. SOC was 80.8% (95%CI, 72.8-86.7) vs. 64.1% (95%CI, 51.3-74.0), HROW 0.48, 95%CI, 0.23-0.99; p=0.049. The overall survival among tocilizumab/methylprednisolone/SOC patients vs. SOC was 85.9% (95%CI, 80.7-92.6) vs. 71.9% (95%CI, 46-73), HROW 0.41, 95%CI: 0.19-0.89, p=0.025. Conclusion Early adjunctive treatment with tocilizumab, methylprednisolone or both may improve outcomes in patients with COVID-19 pneumonia.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 Introduction 78 Pandemic coronavirus disease 2019  caused by SARS-CoV-2 coronavirus has 79 recently emerged [1]. Although most of the infected patients will remain asymptomatic or 80 develop mild symptoms, up to 20% may develop severe disease with pneumonia and 81 respiratory failure [2]. Oxygen administration is the cornerstone of supportive treatment and 82 is required in approximately 15% of cases, while invasive mechanical ventilation is necessary 83 in up to 5-7% of severe cases [3][4][5]. Since mortality in patients with invasive ventilation can 84 be very high, halting the progression from moderate to severe respiratory failure should 85 reduce the mortality in . 86 At the beginning of COVID-19 pandemics, based on the experience with previous studies in 87 viral pneumonia, including SARS-CoV and MERS, the use of steroids was discouraged, 88 mainly due to undocumented benefit and fearing potential increase in viral proliferation and 89 side effects [7,8]. However, with the increasing knowledge about COVID-19, a biphasic 90 model of the disease has been proposed [9,10]. According to this model, the first phase is 91 caused directly by viral replication, while in the second phase, the symptoms and respiratory 92 failure are due to inflammatory response, and could be treated with agents which reduce 93 inflammation, such as corticosteroids, or inhibitors of pro-inflammatory interleukins and 94 Janus kinase (JAK) [9][10][11]. 95 Indeed, some real life experiences in COVID-19 patients showed that the use of anti-96 inflammatory treatments might be beneficial [12]. In fact, short-term steroid therapy was 97 associated with lower mortality in 201 patients with acute respiratory distress syndrome 98 (ARDS) [13]. Additionally, following the data on presence of inflammatory cytokine storm in 99 severe COVID-19, tocilizumab use has been advocated. This monoclonal antibody, which 100 binds to interleukin 6 (IL-6) receptor and blocks the IL-6 mediated inflammatory response, is 101 approved for treatment of rheumatologic disorders and cytokine-release syndrome associated 102 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 6 with Chimeric Antigen Receptor T-cell (CAR-T) administration. It was reported to reduce 103 COVID-19-associated inflammation, and was approved in China for this indication [12,14]. 104 Based on the first evidences, we formulated the hypothesis of potential benefit of anti-105 inflammatory treatment, and progressively modified our therapeutic approach to  We started using tocilizumab in patients with respiratory failure, and subsequently, we 107 introduced into our protocol early administration of methylprednisolone treatment, followed 108 in more severe cases by the administration of tocilizumab. 109 We hypothesized that outcomes such as no need for intubation and survival of patients with 110 severe COVID-19 pneumonia in whom tocilizumab and/or methylprednisolone were 111 administered in addition to standard of care (SOC) could be better than in those who received 112 only SOC. 113 114

Patients and methods 115
Setting and patients 116 In this observational single-center study, adult patients admitted to the San Martino 117 University Hospital, Genova, Italy, for COVID-19 pneumonia were included as cases if 118 treated with tocilizumab and/or methylprednisolone, not intubated, not treated with remdesivir 119 and not pregnant. The outcomes of patients treated with tocilizumab/methylprednisolone were 120 compared to data from consecutive patients admitted to our hospital for COVID-19 121 pneumonia who received only SOC, mainly because they were admitted before the routine 122 use of tocilizumab/methylprednisolone (control group). 123 All patients provided an informed consent for treatment with off label agents according to the 124 local protocol approved by Hospital Authorities, for data collection and for participation in 125 the study. The study was carried out in accordance with the principles of the Declaration of 126 Helsinki and approved by the Regional Ethic Committee (N. CER Liguria 114/2020-ID 127 10420). 128 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 Data collection and definitions 129 Data were collected from hospital information system by a standard based automatic 130 procedure and stored in an online database with pseudo-anonymization features suitable for 131 secondary use of clinical data [15]. Controls were identified through this prospectively 132 collected database of hospital-admitted COVID-19 patients. 133 Patients who had any of the following features at the time of, or after, admission were 134 classified as having severe pneumonia: (1) respiratory distress (≥30 breaths per min); or (2) 135 oxygen saturation at rest ≤ 93%; or (3) ratio of partial pressure of arterial oxygen to fractional 136 concentration of oxygen inspired air (PaO 2 /FiO 2 ) ≤ 300 mm Hg; or (4) severe disease 137 complications (e.g., respiratory failure, requirement of mechanical ventilation, septic shock, 138 or non-respiratory organ failure) [7,14,15]. 139 Systemic inflammation was defined as presence at baseline of one of the following: fever > 140 38°C, C-reactive protein (CRP) 10 times above the upper limit of normal (ULN) of 5 mg/dl, 141 ferritin 2 times above the ULN (400 µg/L), or IL-6 10 times above the ULN of 3.4 ng/L. The 142 first results obtained at hospital admission, and in any case not later than within day 3 of 143 admission were considered. 144 Adverse events possibly or probably related to steroid and tocilizumab treatment, such as 145 neutropenia, anemia, thrombocytopenia, increase in alanine aminotransferase (ALT) levels, 146 (UNL < 40 U/L), microbiologically documented infections and allergic reactions were 147 evaluated for both treated patients and controls. Adverse events were collected up to the last 148 available follow up from starting of tocilizumab and/or methylprednisolone in treatment 149 group and from hospital admission in the control group. The common terminology criteria for 150 adverse event (CTCAE) version 5.0 was used. 151 Standard of care and treatment procedures 152 The diagnosis of COVID-19 was made with a positive RT-PCR assay performed on nasal 153 swab or broncoalveolar lavage fluid according to World Health Organization interim guidance 154 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. The landmark analysis was applied in order to minimize the potential immortal time bias that 178 can arise in non-randomized studies and is related to the fact that patients treated after a 179 longer time from admission must have not experienced the event up to that time, and that 180 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.22.20133413 doi: medRxiv preprint 9 patients with a very early event (e.g. death) were more likely assigned to the untreated group. 181 This is a conservative analysis which reduced the risk that the treatment choice was motivated 182 by the patient's disease course. Therefore, day 3 from hospital admission was set as a 183 landmark time point: those who died, were intubated or discharged from the hospital before 184 day 3 were excluded, while patients were included in the tocilizumab/methylprednisolone 185 treatment group if the treatment was started within 3 days from hospital admission (see Figure  186 1). 187 To minimize baseline differences between treated and untreated patients a propensity score-188 based analysis was performed. Propensity score (PS) was derived by a logistic regression 189 model including the following baseline variables: age, gender, presence of comorbidities and 190 week of treatment start, ratio of partial pressure of arterial oxygen to fractional concentration 191 of oxygen inspired air (PaO 2 /FiO 2 ), non-invasive ventilation (NIV), time from symptoms 192 onset to hospital admission, IL-6, ferritin, C-reactive protein (CRP) and d-dimer serum levels. 193 Positivity assumption of PS was checked after the calculation. For each patient, the overlap 194 weight (OW) based on PS was calculated [19]. To assess the balance of covariate distribution 195 between the two groups, Cohen's standardized mean differences were calculated between the 196 two groups in the original samples and after weighting. An absolute value of difference < 0.10 197 was considered an acceptable balance. 198 The OW-weighted Cox proportional hazard regression model was used to calculate the 199 adjusted hazard-ratio (HR OW ) of tocilizumab/methylprednisolone/SOC vs SOC patients. 200 Weighted cumulative probability of failure or death was calculated by mean of Kaplan-Meier 201 (KM) survival curves. 202 To define risk factors associated with outcomes and to compare the three treatment groups, 203 adjusted HRs were estimated by a multivariable Cox proportional hazard regression model. 204 The same baseline variables used in the calculation of PS were considered for the 205 multivariable analysis. To avoid overfitting, only those characteristics who showed a p-value 206 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.22.20133413 doi: medRxiv preprint ≤ 0.15 at univariable analysis and after inclusion in the multivariable model were considered, 207 with age and gender forced into the model. For a better interpretation and to avoid the 208 influence of outliers on estimation, the IL-6, ferritin, CRP and d-dimer were log-transformed 209 before the analysis due to the highly skewed distribution. All results were reported as HR with 210 95% confidence interval (95%CI). 211 A subgroup analysis was performed to assess if the treatment effect of tocilizumab versus 212 methylprednisolone on primary outcome was different between subgroups defined according 213 to categorized baseline variables. An interaction test was used to assess a different treatment 214 effect in subgroups. 215 The sensitivity analysis was pre-planned, and the comparison between 216 tocilizumab/methylprednisolone/SOC and SOC patients was reassessed excluding from the 217 SOC group the patients that received tocilizumab or methylprednisolone after 3 days from 218 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Failure rate 235
In univariable and multivariable analyses, older age, male gender, higher baseline 236 inflammatory markers, especially IL-6, and PaO 2 /FiO 2 < 100 mmHg were identified as risk 237 factors for failure (Table 2). 238 Differences between the two groups were consistently reduced after OW. OW weighted 239 characteristics of two groups of patients are shown in Table 3. were younger and with fewer comorbidities but with comparable inflammatory markers, the 255 frequency of low PaO 2 /FiO 2 and NIV (Table 4). 256 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020.  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. In this observational study in not intubated patients with mainly severe COVID-19 305 pneumonia, the early addition of tocilizumab and/or methylprednisolone to SOC resulted in 306 adjusted failure-free survival of 86.5% and 80.8% at day 14 and 30, which was, respectively, 307 10.7% and 16.7% higher than in SOC patients. 308 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . However, considering the rapid widespread increase in severe cases of COVID-320 19 worldwide, the availability and the cost of anti-IL-6 treatment might limit its use. 321 Therefore, at the peak of COVID-19 epidemics in our city, we implemented the early use of 322 corticosteroids and the use of subcutaneous tocilizumab if intravenous formulation was not 323 readily available. We acknowledge that the benefit of subcutaneous formulation might be 324 lower and slower than in case of intravenous drug, but no standard intravenous dose for 325 COVID-19 has been established, as one study used 400 mg and the other the rage of doses 326 from 80 mg to 600 mg, irrespective of the patients' weight [12,22]. 327 The first study that reported the impact of steroid treatment in COVID-19, showed that it was 328 administered to 30.8% of patients, mainly in case of more severe disease, and was associated 329 with a significant reduction of the risk of death in patients with ARDS (HR=0.38) [13]. While 330 other cohorts reported steroid use in approximately 44% of severely ill patients, they did not 331 analyze their influence on outcome [20,23]. Subsequently, an observational study reported a 332 benefit of early administration of steroid therapy on composite outcome of ICU admission, 333 mechanical ventilation or death at 14 days (34.9% vs. 54.3%) and overall survival (86.4% vs. 334 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 15 73.7%) [24]. In addition to these data, we were able to demonstrate that, after minimizing as 335 much as possible the differences between the groups through OW adjustment, the outcome of 336 patients was better in case of early treatment with tocilizumab and/or methylprednisolone. 337 Indeed, in SOC group the rate of failure at day 14 of 24.2% was very similar to what reported 338 in other cohorts with severe pneumonia (24.9%), while it was reduced to 13.5% in our 339 tocilizumab/methylprednisolone/SOC group [20]. Our data show that this benefit was also 340 present with month-long follow up (overall HR ow of 0.48), which is important in establishing 341 long term prognosis of these patients. Moreover, the benefit of early 342 tocilizumab/methylprednisolone was also noted on overall survival, both at 14 and 30 days 343 (respectively, 92.7% vs. 78.2% and 85.9% vs. 71.9%). Compared to the study with steroid use 344 only, the 14-day survival was higher in our cohort, providing background for the hypothesis 345 that combined tocilizumab/steroid treatment might be warranted [24]. Finally, preliminary 346 results of the RECOVERY trial reported higher survival rate in patients with severe COVID-347 19 pneumonia treated with dexamethasone, although the detailed results are not available yet. 348 Interestingly, our observational study, which showed the effectives of early anti-inflammatory 349 treatment in a cohort of non-intubated patients with severe COVID-19 pneumonia, 350 documented that these patients were treated at a median time of 8 days after the onset of 351 symptoms, which is exactly the time of initial cytokine storm, and therefore might be optimal 352 for the effect of anti-inflammatory treatment. 353 Consistent with other studies, we identified older age, high IL-6 levels and poor respiratory 354 function as independent predictors of failure, with possible impact also of CRP and d-dimer 355 levels [25]. Possibly due to a limited sample size, we were unable to document which of three 356 treatment groups provided most benefit, and if there were predictors of better response to 357 tocilizumab/methylprednisolone compared to methylprednisolone alone in any subset of 358 patients. However, the rate of failure-free survival was the highest in the combination 359 treatment group. In addition, based on our sensitivity analyses, adding the anti-inflammatory 360 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.22.20133413 doi: medRxiv preprint treatment later after hospital admission might still provide some clinical benefit. In fact, 361 including in the SOC group patients who received anti-inflammatory treatment later during 362 the infectious course possibly reduced the difference between the study arms, highlighting the 363 benefit of an early anti-inflammatory treatment. 364 The limitations of this study include the non-randomized design, yet the inclusion of 365 consecutive patients using the same SOC but not treated with tocilizumab or 366 methylprednisolone, and adjustment for the outcome-associated variables, allowed to notice 367 the improvement in patient outcomes. Nonetheless, it is possible that some benefit observed 368 was partially due to general improvements in patient clinical care that occur with time. 369 Additionally, this being a single-center experience might limit the applicability to other 370 settings, since our hospital managed to rapidly increase the capacity for hospitalization and 371 ventilation support, potentially improving general patient care. However, the adjustment for 372 the differences between patient groups through propensity score and conservative approach 373 with the use of landmark analysis were directly at minimizing the risk associated with an 374 absence of randomization. Finally, we believe that the rate of failure observed in this study of 375 7.3% at 14 days in those with severe COVID-19 treated with SOC and 376 tocilizumab/methylprednisolone might help to better define the expected rate of response and 377 calculate the number of patients needed to include in the studies assessing various treatment 378 options. 379

Conclusion 380
In conclusion, the negative impact of immune response in COVID-19 might be mitigated by 381 early administration of anti-inflammatory therapy with tocilizumab, methylprednisolone or 382 both. Randomized studies are warranted to establish the best treatment options, their timing 383 and limitations. 384 385 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020 Acknowledgments 386 We would like to thank all the patients and the hospital staff, with particular mention of Mrs 387 Enrica Lombardi, who helped us to get through these difficult weeks. 388 This study was supported by the efforts of all members of GECOVID group. 389 390 Funding 391 No funding or sponsorship was received for this study or publication of this article. 392 Independently of the study, Roche provided intravenous tocilizumab free of charge for the 393 Ligurian Region during the first weeks of COVID-19 epidemics. 394 Some of the patients were included also in the observational Italian national AIFA study on 395 intravenous tocilizumab. 396 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 Tables 474 Table 1  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020 22 CRP, C reactive protein; IL-6, interleukin 6; NIV, non invasive ventilation; PaO 2 /FiO 2 , ratio 477 of partial pressure of arterial oxygen to fractional concentration of oxygen inspired air; SOC, 478 standard of care. 479 (IQR) . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020  PaO 2 /FiO 2 , ratio of partial pressure of arterial oxygen to fractional concentration of oxygen 483 inspired air. 484 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.22.20133413 doi: medRxiv preprint *None of the following single comorbidities resulted significant in multivariable analysis: 485 hypertension, diabetes, cancer, obesity, ischemic heart failure, and only ischemic hear disease 486 was significant in univariate analysis. 487 488 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020   . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020  1.20 (0.78-1.83) 0.41 -. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 26, 2020. PaO 2 /FiO 2 , ratio of partial pressure of arterial oxygen to fractional concentration of oxygen 503 inspired air. 504 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10.1101/2020.06.22.20133413 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 26, 2020. . https://doi.org/10. 1101/2020