key: cord-0773743-fd7ovpgs
authors: Maulin, Laurence; Martinez, Stéphanie
title: Corticosteroids in patients hospitalised for COVID-19 pneumonia who require oxygen: observational comparative study using routine care data
date: 2020-12-08
journal: Clin Microbiol Infect
DOI: 10.1016/j.cmi.2020.11.035
sha: ee5fb5be202d23fe2410da3bbbabae563a592627
doc_id: 773743
cord_uid: fd7ovpgs

OBJECTIVE: To assess the effectiveness of corticosteroids on outcomes of patients with COVID-19 pneumonia requiring oxygen without mechanical ventilation. METHODS: We used routine care data from 51 hospitals in France and Luxembourg to assess the effectiveness of corticosteroids at 0.8 mg/kg/day eq. prednisone (CTC group) versus standard of care (no-CTC group) among adults 18 to 80 years old with confirmed COVID-19 pneumonia requiring oxygen without mechanical ventilation. The primary outcome was intubation or death by day 28. In our main analysis, characteristics of patients at baseline (i.e., time when patients met all inclusion criteria) were balanced by using propensity-score inverse probability of treatment weighting. RESULTS: Among the 891 patients included in the analysis, 203 were assigned to the CTC group. Use of corticosteroids was not significantly associated with risk of intubation or death by day 28 (weighted hazard ratio [wHR] 0.92, 95% CI 0.61 to 1.39) or cumulative death rate (wHR 1.03, 95% CI 0.54 to 1.98). However, use of corticosteroids was associated with reduced risk of intubation or death by day 28 in the prespecified subgroups of patients requiring oxygen ≥ 3 L/min (wHR 0.50, 95% CI 0.30 to 0.85) or C-reactive protein level ≥ 100 mg/L (wHR 0.44, 95%CI 0.23 to 0.85). Number of hyperglycaemia events was higher for patients with than without corticosteroids, but number of infections was similar. CONCLUSIONS: We found no association between the use of corticosteroids and intubation or death in the broad population of patients 18 to 80 years old with COVID-19 hospitalized in non-intensive care unit settings. However, the treatment was associated with reduced risk of intubation or death for patients with ≥ 3 L/min oxygen or C-reactive protein level ≥ 100 mg/L at baseline. Further research need to confirm the right timing of corticosteroids for patients with COVID-19 requiring oxygen only.

To assess the effectiveness of corticosteroids on outcomes of patients with COVID-19 pneumonia requiring oxygen without mechanical ventilation.

We used routine care data from 51 hospitals in France and Luxembourg to assess the effectiveness of corticosteroids at 0.8 mg/kg/day eq. prednisone (CTC group) versus standard of care (no-CTC group) among adults 18 to 80 years old with confirmed COVID-19 pneumonia requiring oxygen without mechanical ventilation. The primary outcome was intubation or death by day 28. In our main analysis, characteristics of patients at baseline (i.e., time when patients met all inclusion criteria) were balanced by using propensity-score inverse probability of treatment weighting.

Among the 891 patients included in the analysis, 203 were assigned to the CTC group. Use of corticosteroids was not significantly associated with risk of intubation or death by day 28 (weighted hazard ratio [wHR] 0.92, 95% CI 0.61 to 1.39) or cumulative death rate (wHR 1.03, 95% CI 0.54 to 1.98). However, use of corticosteroids was associated with reduced risk of intubation or death by day 28 in the prespecified subgroups of patients requiring oxygen ≥ 3 L/min (wHR 0.50, 95% CI 0.30 to 0.85) or C-reactive protein level ≥ 100 mg/L (wHR 0.44, 95%CI 0.23 to 0.85). Number of hyperglycaemia events was higher for patients with than without corticosteroids, but number of infections was similar.

We found no association between the use of corticosteroids and intubation or death in the broad population of patients 18 to 80 years old with COVID-19 hospitalized in non-intensive

Coronavirus disease 2019 pneumonia is associated with a hyper-inflammatory phase, which is deemed responsible for the clinical worsening of many patients [1] . At the onset of the COVID-19 pandemic, guidance regarding corticosteroids for patients without acute respiratory distress syndrome was mixed [2] . Only evidence from small observational studies, with contrasting results, was available and practices varied widely across the world [3] [4] [5] [6] [7] [8] .

In July 2020, the RECOVERY trial showed that dexamethasone reduced 28-day mortality for patients receiving oxygen with and without invasive mechanical ventilation [9] . The efficacy of corticosteroids in critically ill patients was confirmed in subsequent trials and in a metaanalysis [10] [11] [12] . However, for patients with less severe disease, the trial highlighted a heterogeneity of the treatment effect, with a lower incidence of death among patients receiving oxygen but not among those without oxygen at randomization. We hypothesize that this relationship is a continuum and that the more severe the disease, the more effective the treatment. Differences in severity at baseline among the broad population of patients receiving oxygen only could explain the discordance between studies investigating the use of corticosteroids for COVID-19 [13] [14] [15] [16] .

Our aim was to bring additional data to better understand the effectiveness of corticosteroids in COVID-19 pneumonia, exploring a potential heterogeneity of the treatment effect in the broad population of patients with COVID-19 receiving oxygen without mechanical ventilation.

We retrospectively used data collected from routine care to emulate a target trial aimed at assessing the efficacy and safety of corticosteroids in patients hospitalised with COVID-19, requiring oxygen and with an inflammatory syndrome [17] .

Our study involved the internal medicine or infectious disease wards from 51 hospitals in France and Luxembourg part of a network coordinated by REACTing (INSERM).

Physicians performed a patient-by-patient screening of all patients hospitalised between March 1 and May 1, 2020. First, they identified all consecutive patients aged 18 to 80 years old who had PCR-confirmed SARS-CoV-2 infection. Second, they assessed whether these patients had an inflammatory syndrome with C-reactive protein (CRP) level ≥ 40 mg/L and required oxygen by mask or nasal prongs during hospitalization. Third, they excluded patients 1) discharged from the intensive care unit (ICU) to standard care or requiring immediate admission to the ICU (including patients requiring non-invasive ventilation); 2) with a contraindication to corticosteroids; 3) who received corticosteroids or anti-interleukin agents before baseline; 4) with severe comorbidities (body mass index <16 kg/m 2 , renal diseases requiring dialysis, chronic heart failure New York Heart Association grade IV, liver cirrhosis Child grade C, or long-term oxygen therapy); 5) with decision to limit active treatments before baseline; and 6) included in the DISCOVERY trial (NCT04315948).

The study received approval by the IRB of the Henri-Mondor Hospital (AP-HP), France (no. 00011558) and by the national ethics committee of Luxembourg (no. 0620-101). The study was based on data from routine care already collected at the time of the study. All patients were informed that their hospital data would be used for research purposes and could refuse this use.

We compared two treatment strategies: receiving at least one dose of corticosteroids at ≥0.8 mg/kg/day eq. prednisone or ≥0.4 mg/kg/day eq. prednisone if co-administered with lopinavir-ritonavir (CTC group) versus the standard of care (no-CTC group). These values were chosen to account for dose rounding by physician and for the drug-drug interaction between ritonavir and steroids (i.e., patients receiving lopinavir-ritonavir require a lower dose of corticosteroids to obtain an effect) [18] . Standard of care consisted of supportive therapy and treating the symptoms to prevent respiratory failure. No systematic antibiotic prophylaxis or antiviral agents were provided to patients in the study centres. Patients in the CTC group could start corticosteroids within a prespecified "grace period" of 5 days after baseline [17] .

Our causal contrast of interest was the per-protocol effect; thus, patients receiving a lower dose of corticosteroids or starting corticosteroids after 5 days were dropped from the analysis.

A specific sensitivity analysis mimicking an intention-to-treat analysis was performed in which all patients eligible for the study were analysed. For safety outcomes, all patients who received corticosteroids, whatever the dose and timing, were considered in the CTC group.

The definition of group assignment based on patients' observational data is reported in Supplementary material 1.

The start of follow-up (baseline or time zero) for each individual was the time all eligibility criteria were checked. All patients were followed up from baseline until the occurrence of one of the following: 1) death, 2) loss to follow-up, or 3) end of follow-up, which occurred at least 28 days after baseline.

The primary outcome was intubation or death by day 28. Secondary outcomes were death from any cause by day 28, weaning from oxygen on day 28, and discharge from hospital to J o u r n a l P r e -p r o o f home/rehabilitation on day 28. All adverse events (e.g., infection, thromboembolic events, etc.) were abstracted from electronic health records in free text and independently re-coded by four physicians (XL, FG, TP and MM). In case of disagreement, consensus was obtained.

A non-parsimonious multivariable logistic regression model was constructed to estimate the probability of receiving corticosteroids given baseline covariates (i.e., the propensity score [PS] ). Variables of the PS model were pre-specified before any outcome analyses and are detailed in Supplementary material 2. Standardised differences were examined to assess balance, with a threshold of 10% to indicate clinically meaningful imbalance [19] .

Inverse probability of treatment weighting (IPTW) was used to balance the two groups on potential confounders at baseline. Stabilized weights were used to reduce the variability of weights and standard errors of estimated treatment effects [20] . Cox proportional-hazards models were used to compute IPTW hazard ratios (wHRs) with 95% confidence intervals (CIs). IPTW estimates of the relative risk (wRR) were computed for binary outcomes.

Outcomes are presented for the total population and for predefined subgroups: those with ≥3 L/min oxygen at baseline [21] , CRP level ≥ 100 mg/L at baseline [22] , and a high dose of corticosteroids (≥120 mg) at baseline [23] . We also performed two post-hoc subgroup analyses exploring the treatment effect by time since symptom onset ≤ 7 or > 7 days [9] and among patients with both ≥3 L/min oxygen and CRP level ≥ 100 mg/L at baseline. In each subgroup, we recalculated the PS to balance the differences in baseline variables between treatment groups.

To account for immortal time bias, all patients in the no-CTC group who met the primary outcome (intubation or death) during the grace period were randomly assigned to one of the J o u r n a l P r e -p r o o f two groups, given that their observational data were compatible with both groups at the time of the event [17] (Supplementary material 3) .

We aimed for the maximal sample size achievable while accounting for logistical feasibility and costs. We supplemented our results with an "a posteriori" power calculation with an alpha risk of 0.05, based on the number of events observed in our data [24, 25] .

Missing baseline and outcome variables were handled by multiple imputations by chained equations by using the other variables available. All statistical analyses were performed with R v3.6.1 or later (The R Foundation for Statistical Computing, https://www.R-project.org/).

Among the 965 patients eligible for analysis, 194 received corticosteroids at ≥ 0.8 mg/kg/d eq. prednisone (or 0.4 mg/kg/d eq. prednisone if co-administered with lopinavir-ritonavir) within 5 days from baseline; 697 did not receive corticosteroids, 28 received corticosteroids at a lower dose, and 46 received corticosteroids after 5 days from baseline.

In all, 107 patients who did not receive corticosteroids met the primary outcome (intubation or death) during the grace period. To account for immortal time bias, we randomly assigned these 107 patients to the CTC or no-CTC group [17] . Thus, we compared 203 participants in the CTC group to 688 in the no-CTC group (Figure 1 and supplementary material 3) . 

Among the 891 patients included in the main analysis, 78 had follow-up < 28 days (70 were discharged in good health status) and 63 died before day 28. Median follow-up for surviving patients was 80 days (IQR, 38 to 94).

In the unweighted sample, by day 28, 18.0% (n=36) and 19.4% (n=131) of patients in the CTC and no-CTC groups had been intubated or died (HR 0.89, 95% CI 0.62 to 1.28). In total, 8.5% (n=17) and 6.9% (n=46) of patients died in the CTC and no-CTC groups (HR 1.23, 95% CI 0.70 to 2.13). On day 28, 79.9% (n=148) and 83.6% (n=522) of patients, respectively, were weaned from oxygen (RR 0.96, 95% CI 0.88 to 1.03). Furthermore, 81.6% (n=151) and 84.4% (n=529), respectively, were discharged to home/rehabilitation (RR 0.97, 95% CI 0.90 to 1.04). From the number of events observed in our data, our sample allowed for 45% to 70% power to detect an HR of 0.7 to 0.6, consistent with results for patients ≤ 70 years old in the RECOVERY trial.

In the IPTW analyses, on day 28, 20.1% and 20.9% of patients in the CTC and no-CTC groups had been intubated or died (weighted HR [wHR] 0.92, 95% CI 0.61 to 1.39) ( Figure   2 ) and 8.7% and 8.2% died, respectively (wHR 1.03, 95% CI 0.54 to 1.98) (Figure 3) ( Table   2 ). The two groups did not differ in oxygen weaning (wRR 0.98, 95% CI 0.89 to 1.07) or discharge to home/rehabilitation (wRR 1.00, 95% CI 0.92 to 1.09) ( Table 2) .

Results for subgroups are presented in Figure 4 and Supplementary material 8. 

We report a multicentre study that used real-world data to emulate a target trial on the effectiveness of corticosteroids for patients with COVID-19 pneumonia in non-ICU settings.

We found no difference in risk of intubation or death for patients who received and did not receive corticosteroids, although the result was compatible with a 40% reduction in risk.

However, rate of intubation or death was lower for patients receiving corticosteroids with ≥ 3 L/min oxygen or with CRP level >100 mg/L at baseline.

In all, our results are consistent with the hypothesis that that the more severe the disease (i.e., requiring higher oxygen flow and with a more severe inflammatory syndrome), the more effective the treatment. Difference in between our results and those from the RECOVERY trial may be due to the fact that we included a large number of patients who had limited requirements for oxygen (e.g., in the no-CTC group, 18% of patients received ≤1 L/min oxygen at baseline) and whose condition may be closer to the subgroup of patients who were not under oxygen at the time of randomization in the RECOVERY trial. Second, to maximize comparability between the two groups, we also chose to exclude patients with severe chronic conditions. This could explain the difference in death rates by 28 days in the no-CTC groups (8.2% in our study and 25% in the RECOVERY trial).

In all, our findings question the benefit/risk balance of corticosteroids in patients with the least severe disease. Indeed, although we found no increase in secondary infections, hyperglycaemia occurrence was higher in patients with than without corticosteroids.

Our study is based on the analysis of a large number of consecutive patients and provides insight into the management of COVID-19 during the study period in France and

Luxembourg. Second, all safety data were reviewed in duplicate and independently by several clinicians, which strengthens our confidence in the absence of any warning signal for using corticosteroids in COVID-19 pneumonia in terms of secondary bacterial or fungal infection.

Our study has several limitations. First, despite the use of robust methods to draw causal inferences, our study is observational, and potential unmeasured confounders may bias our results [26] . Second, our study dealt with a heterogeneous prescription of corticosteroids in terms of drugs, time of start, dose, administration mode and duration. Third, we could not account for duration of corticoid prescription. Indeed, we may have observed only 3 days of corticosteroid treatment for a given patient because an event occurred on the fourth day. Only trials in which the dose and duration are specified before treatment, in "intent-to-treat", can answer this question. Fourth, because we used data from a large number of hospitals, we could not standardize the process of identifying patients eligible to participate in the study.

This situation may have led to missing some patients. Fifth, our sample was limited to the number of eligible patients available at the time of analysis and we cannot rule out that our findings are due to a lack of power

We found no association of the use of corticosteroids and intubation or death in a broad sample of patients 18 to 80 years old hospitalised for COVID-19 pneumonia in non-ICU settings in France. However, use of corticosteroids was associated with reduced rate of intubation or death for patients receiving oxygen ≥ 3 L/min or CRP level ≥ 100 mg/L at baseline. These data support the hypothesis that that the more severe the disease, the more effective the treatment and call for discussing the treatment for patients requiring low oxygen flow. Dissemination to related patient and public communities: We plan to issue a press release on official publication of this manuscript and disseminate our findings through social media outlets, to ensure the results of the study have broad public outreach.

The lead author (X. Lescure) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.

Contributorship statement: V-TT, EP, PR and F-XL conceived the study; MM, FB, OB, TP, LG, FG and F-XL participated in data collection; V-TT, EP and PR performed the statistical analyses; MM, OB, TP, FG and F-XL reviewed the data for the adverse events; MM, LG, and F-XL provided administrative and logistic support; V-TT, MM and F-XL wrote the first draft of the manuscript; All members of the writing committee contributed to the writing of the manuscript. All members of the writing committee meet the ICMJE criteria. All members of the writing committee agree with the manuscript results and conclusions. F-XL is the guarantor, had full access to the data in the study, and takes responsibility for the integrity of the data and the accuracy of the data analysis. All collaborators are listed in the supplementary appendix. Table 1 : Demographic and clinical characteristics of the patients in the corticosteroids (CTC) and no-CTC groups at baseline. The number in brackets in the first column corresponds to the actual quantity of data available for the corresponding variables before imputation of missing baseline data by multiple imputations by chained equations using the other variables available. Table 1 : Demographic and clinical characteristics of the patients in the corticosteroids (CTC) and no-CTC groups at baseline. The number in brackets in the first column corresponds to the actual quantity of data available for the corresponding variables before imputation of missing baseline data by multiple imputations by chained equations using the other variables available. Results are presented as % (absolute number) unless stated otherwise.

Abbreviations: IQR, interquartile range; NYHA, New York Heart Association; BMI, body mass index; ACEIs, angiotensin-converting enzyme inhibitors; ARBs, angiotensin receptor blockers.

* 127 patients did not have a CT scan at admission; **Corresponds to the data for the 194 patients who received corticosteroids within 5 days of eligibility, with at least 0.8 mg/kg/day eq. prednisone or 0.4 mg/kg/day eq. prednisone if co-administered with lopinavir-ritonavir.

J o u r n a l P r e -p r o o f *missing data were managed by using multiple imputations by chained equations.

J o u r n a l P r e -p r o o f 20 

The trinity of COVID-19: immunity, inflammation and intervention

Infectious Diseases Society of America Guidelines on the Treatment and Management of Patients with COVID-19

Early Short Course Corticosteroids in Hospitalized Patients with COVID-19

Systemic Corticosteroids and Mortality in Severe and Critical COVID-19 Patients in Wuhan

Corticosteroid therapy for patients with COVID-19 pneumonia: a before-after study

Adjuvant corticosteroid therapy for critically ill patients with COVID-19

Covid-19 in Immune-Mediated Inflammatory Diseases -Case Series from New York

Clinical findings in a group of patients infected with the 2019 novel coronavirus (SARS-Cov-2) outside of Wuhan, China: retrospective case series

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

Effect of Hydrocortisone on Mortality and Organ Support in Patients With Severe

Corticosteroid Domain Randomized Clinical Trial

Effect of Dexamethasone on Days Alive and Ventilator-Free in Patients With Moderate or Severe Acute Respiratory Distress Syndrome and COVID-19: The CoDEX Randomized Clinical Trial

Methylprednisolone as Adjunctive Therapy for Patients Hospitalized With COVID-19 (Metcovid): A Randomised, Double-Blind, Phase IIb, Placebo-Controlled Trial

Efficacy of corticosteroid treatment for hospitalized patients with severe COVID-19: a multicenter study

A controlled trial of methylprednisolone in adults hospitalized with COVID-19 pneumonia

Low-to-moderate dose corticosteroids treatment in hospitalized adults with COVID-19

Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available

Drug-drug interactions in the treatment of HIV infection: focus on pharmacokinetic enhancement through CYP3A inhibition

Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples

Marginal structural models and causal inference in epidemiology

Effect of Tocilizumab vs Usual Care in Adults Hospitalized With COVID-19 and Moderate or Severe Pneumonia: A Randomized Clinical Trial

Longitudinal changes of inflammatory parameters and their correlation with disease severity and outcomes in patients with COVID-19 from Wuhan

Standardised nomenclature for glucocorticoid dosages and glucocorticoid treatment regimens: current questions and tentative answers in rheumatology

Simple, defensible sample sizes based on cost efficiency

Current sample size conventions: flaws, harms, and alternatives

Scientific Euphemisms Do Not Improve Causal Inference From Observational Data

with at least 0.8 mg/kg/d eq. prednisone or 0.4 mg/kg/d eq. prednisone if co-administrated with lopinavir-ritonavir within 5 days from eligibility 771 did not receive corticosteroids with at least 0.8 mg/kg/d eq. prednisone or 0.4 mg/kg/d eq. prednisone if coadministrated with ritonavir within 5 days from eligibility 965 patients were included in the analysis 28 received corticosteroids less than 0.8 mg/kg/d eq. prednisone or less than 0.4 mg/kg/d eq. prednisone if co-administrated with lopinavir-ritonavir 697 did not receive corticosteroids 46 received corticosteroids after 5 days from eligibility 203 patients were analyzed in the CTC group (per protocol) 688 patients were analyzed in the no CTC group (per protocol) 9 patients who met the primary outcome before 5 days were randomly assigned in the CTC group*