key: cord-1049799-mabo1yns
authors: Liu, Zhibo; Li, Xia; Fan, Guohui; Zhou, Fei; Wang, Yeming; Huang, Lixue; Yu, Jiapei; Yang, Luning; Shang, Lianhan; Xie, Ke; Xu, Jiuyang; Huang, Zhisheng; Gu, Xiaoying; Li, Hui; Zhang, Yi; Wang, Yimin; Huang, Zhenghui; Cao, Bin
title: Low-to-moderate dose corticosteroids treatment in hospitalized adults with COVID-19
date: 2020-09-29
journal: Clin Microbiol Infect
DOI: 10.1016/j.cmi.2020.09.045
sha: b5631ffae51375f8c142284d358adf0045cdfca0
doc_id: 1049799
cord_uid: mabo1yns

OBJECTIVES: Use of corticosteroids is common in the treatment of coronavirus disease 2019, but the clinical effectiveness was controversial. We aimed to investigate the association of corticosteroids therapy with clinical outcomes of hospitalized COVID-19 patients. METHODS: In this single center, retrospective cohort study, adult patients with confirmed coronavirus disease 2019 and dead or discharged between December 29, 2019 and February 15, 2020 were studied. 1:1 propensity score matchings were performed between patients with or without corticosteroids treatment. Multivariable COX proportional hazards model was used to estimate the association between corticosteroids treatment and in-hospital mortality by taking corticosteroids as a time-varying covariate. RESULTS: Among 646 patients, in-hospital death rate was higher in 158 patients with corticosteroids administration (72/158, 45.6% vs 56/488, 11.5%, p<0.0001). After propensity score-match analysis, no significant differences were observed in in-hospital death between patients with and without corticosteroids treatment (47/124, 37.9% vs 47/124, 37.9%, p=1.000). When patients received corticosteroids before they required nasal high-flow oxygen therapy or mechanical ventilation, in-hospital death rate was lower than that in patients who were not administered with corticosteroids (17/86, 19.8% vs. 26/86, 30.2%, log rank p=0.0102), whereas time from admission to clinical improvement was longer (13 [IQR, 10∼17] days vs 10 [IQR, 8∼13] days; p＜0.001). Using Cox proportional hazards regression model accounting for time varying exposures in matched pairs, corticosteroid therapy was not associated with mortality difference (HR=0.98, 95%CI: 0.93-1.03, p=0.4694). CONCLUSIONS: Corticosteroids use in COVID-19 patients may not be associated with in-hospital mortality.

Objectives Use of corticosteroids is common in the treatment of coronavirus disease 2 2019, but the clinical effectiveness was controversial. We aimed to investigate the 3 association of corticosteroids therapy with clinical outcomes of hospitalized 4 COVID-19 patients.

Methods In this single center, retrospective cohort study, adult patients with confirmed 6 coronavirus disease 2019 and dead or discharged between December 29, 2019 and 7 February 15, 2020 were studied. 1:1 propensity score matchings were performed 8 between patients with or without corticosteroids treatment. Multivariable COX 9 proportional hazards model was used to estimate the association between 10 corticosteroids treatment and in-hospital mortality by taking corticosteroids as a 11 time-varying covariate.

Results Among 646 patients, in-hospital death rate was higher in 158 patients with 13 corticosteroids administration (72/158, 45.6% vs 56/488, 11.5%, p<0.0001). After 14 propensity score-match analysis, no significant differences were observed in 15 in-hospital death between patients with and without corticosteroids treatment (47/124, 16 37.9% vs 47/124, 37.9%, p=1.000). When patients received corticosteroids before they 17 required nasal high-flow oxygen therapy or mechanical ventilation, in-hospital death 18 rate was lower than that in patients who were not administered with corticosteroids 19 (17/86, 19.8% vs. 26/86, 30 .2%, log rank p=0.0102), whereas time from admission to Since December 2019, the SARS-CoV-2 pandemic occurred and emerged as a 2 major global health threat [1, 2] . Characteristics of viral associated disease, 3 coronavirus disease 2019 , has been described [3] [4] [5] [6] . In absence of 4 effective antiviral drugs [7, 8] , supportive and adjuvant treatments are of importance.

Systemic corticosteroids treatment was commonly used among critically ill 6 patients with viral pneumonia, including severe acute respiratory syndrome (SARS) 7 and middle east respiratory syndrome (MERS), but the effect was controversial. 8 Previous studies showed that systemic corticosteroids did not improve outcome and 9 delayed clearance of viral RNA or increased side effect complications [9] [10] [11] were also excluded if they received corticosteroid dosage higher than recommended Data were collected from electronic medical records using a standardised data 10 collection form. All data were checked by two physicians (ZL and FZ) and a third 11 researcher (GF) adjudicated any difference in interpretation between the two primary 12 reviewers.

The study was approved by the Research Ethics Commission of Jinyintan Hospital 14 (KY-2020-01.01) and the informed consent was waived by the Ethics Commission as 15 described previously [6, 20] . intravenous route at least one dose in-hospital. According to expert consensus 20 statement in China [19] , the recommending dose of corticosteroid was low-to-moderate 21 (≤0·5-1 mg/kg per day methylprednisolone or equivalent). Low-to-moderate dose was defined as below and equal to 80mg per day methylprednisolone or equivalent.

High-dose was defined as ＞80 mg per day methylprednisolone or equivalent. covariates. Based on the propensity score, we performed a 1:1 match to match patients 10 who were not administered with corticosteroids to those who were, on a range ± 11 0.0001 to ±0.1. The match started with the range of ±0.0001, and those who were 12 matched were extracted from database and excluded from the following ranges. If more 13 than 2 patients who were not administered with corticosteroids were detected, only one 14 of them selected randomly. Variables involved in the propensity score estimation 15 included age, gender, lymphocyte count, disease severity status and antiviral treatment.

As sensitivity analysis, we performed PS-matched analysis in two subgroups: patients 17 not requiring advanced oxygen therapy, and patients requiring advanced oxygen 18 therapy.

Kaplan-Meier curves and log-rank tests were used to display and test the 20 differences of in-hospital death rates between groups with or without corticosteroids 21 treatment. In multivariable COX proportional hazard model, only variables with a P value < .05 in univariable analysis or a presumptive association with the event were 1 included to avoid overfitting. Moreover, considering patients were treated with 2 corticosteroids at different times after admission, the treatment was taken as 3 time-varying covariate.

A 2-sided α less than 0.05 was considered statistically significant. Statistical 5 analyses were conducted using SAS software, version 9.4 (SAS Institute). (Table 1) .

Among 646 patients included, the median age was 57 (IQR, 47~67) years, and 17 49.7% (321/646) were males. Severe or critically ill patients (116/158, 73.4%) more 18 frequently received corticosteroids treatment (Table 1) .

After the propensity score match, a total of 124 propensity score-matched pairs 20 were selected. No significant differences were observed after match and characteristics 21 for patients with or without corticosteroids were well balanced. Other details of patient characteristics were shown in Table 1 Figure 2A ). ICU admission and median ICU length of stay were 19 similar in two matched groups (Table 2) . Figure 1B) . There is no evidence that higher dose corticosteroid is better than lower dose.

After the release of RECOVERY trial, corticosteroids were soon accepted as the should also be taken as essential outcomes.

There were some limitations in the study. Firstly, as an observational study, 21 baseline characteristics differed greatly between patients received corticosteroids and who didn't. Propensity score-matched analysis was conducted to remedy the 1 imbalance but with exclusion of a quite large proportion (about 20%) of patients. The 2 sample size of was small, and the lack of difference might be due to limited power.

Secondly, the conclusion of the study is based on retrospective data. As there was no 4 placebo, we could not match for time of corticosteroids use when calculating 5 propensity score because the time point of "corticosteroids initiation" didn't exist in 6 patients without corticosteroids administration, in fact, time varying adjustment was 7 conducted to deal with this. Thirdly, some parameters, for example hospital length of 8 stay,are only described here, since as outcome measure, they would be affected by 9 survivor bias As a retrospective study, side effects of corticosteroids, such as 10 hypertension and hyperglycemia, were missed. The timing of viral RNA sampling 11 was sparse and not accurate, we report it here for completeness but the evidence is not 12 conclusive.

In conclusions, corticosteroids use in COVID-19 patients was not associated with J o u r n a l P r e -p r o o f Time from admission to corticosteroids treatment (days) --3 (2, 5) --3 (2, 5) Table 3 . Adjusted hazard ratios of risk factors associated with death for matched cases administered with corticosteroids.

Note. Adjusted hazard ratios were expressed as adjusted hazard ratios (95% confidence intervals), estimated by COX proportional hazard model. 

Hospital length of stay (days)

Duration of detectable RNA after illness onset (days) * 19 (15, 23) 19 (14, 23) 19 (16, 23) 0

Duration of detectable RNA after admission (days) *

Data are median (IQR) or n (%). p values were calculated by Mann-Whitney U test, χ 2 test, or Fisher's exact test, as appropriate. ICU=intensive care unit

* The date of viral negative was available in survival patients, because the virus was continuously detectable until death in non-survivors