Impact of previous exposure to systemic corticosteroids on unfavorable outcome in patients hospitalized for COVID-19

The impact of prior exposure to systemic corticosteroids on COVID-19 severity in patients hospitalized for a SARS-CoV-2 pneumonia is not known. The present study was designed to answer to this question. The population study was the Covid-Clinic-Toul cohort which records data about all hospitalized patients with a positive reverse transcriptase polymerase chain reaction for a SARS-CoV-2 infection at Toulouse University hospital, France. Exposure to systemic corticosteroids was assessed at hospital admission. A propensity score (PS) according to corticosteroid exposure was calculated including comorbidities, clinical, radiological and biological variables that impact COVID-19 severity. The primary outcome was composite, including admission to intensive care unit, need of mechanical ventilation and death occurring during the 14 days after hospital admission. Logistic regression models adjusted for the PS (overlap weighting) provided odds ratios (ORs) and their 95% confidence intervals (95% CIs). Overall, 253 patients were included in the study. Median age was 64 years, 140 patients (59.6%) were men and 218 (86.2%) had at least one comorbidity. Seventeen patients (6.7%) were exposed to corticosteroids before hospital admission. Chronic inflammatory disease (n = 8) was the most frequent indication. One hundred and twenty patients (47.4%) met the composite outcome. In the crude model, the OR of previous exposure to systemic corticosteroids was 1.64; 95% CI: 0.60–4.44. In the adjusted model, it was 1.09 (95% CI: 0.65–1.83). Overall, this study provide some evidences for an absence of an increased risk of unfavorable outcome with previous exposure to corticosteroids in the general setting of patients hospitalized for COVID-19.


Background
Corticosteroid-based therapy is used to treat patients with severe coronavirus disease 2019 (COVID-19) to reduce inflammatory lung injury, notably when a major cytokine reaction is responsible for clinical worsening [1][2][3][4]. A recent prospective meta-analysis of clinical trials conducted by the World Health Organization (WHO) Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group showed that administration of systemic corticosteroids in critically ill patients with COVID-19, compared with usual care or placebo, was associated with lower 28-day all-cause mortality [5]. Moreover, exposure to systemic corticosteroids before COVID-19, responsible for immunosuppression, has been hypothesized to be associated with severe forms of COVID-19. Systemic exposure to corticosteroids has been associated with an increased risk of hospitalization in patients with rheumatic diseases (≥10 mg/day in prednisone-equivalent dosage, OR: 2.05; 95% CI: 1.06-3.96) [6]. A major increased risk of severe COVID-19 (defined by admission in intensive care unit -ICU, need of mechanical ventilation or death) was found in patients with chronic inflammatory bowel disease (OR: 6.9; 95% CI: 2.3-20.5) [7]. These studies were focused on patients with some autoimmune diseases and were not adjusted for clinical, biological and radiological markers of COVID-19 severity. The impact of previous exposure to corticosteroids in the general setting of patients hospitalized for COVID-19 is unknown.
We aimed to assess the impact of prior exposure to systemic corticosteroids on COVID-19 severity in patients hospitalized for reverse transcriptase polymerase chain reaction (RT-PCR)-proven SARS-CoV-2 infection.

Study population
The study was conducted within the Covid-Clinic-Toul cohort which records data about all hospitalized patients with a positive RT-PCR for a SARS-CoV-2 infection at Toulouse University hospital, France [8,9]. This cohort has been approved by institutional review board (n°RnIPH 2020-31), in accordance with French law. All patients, or their representatives, were informed by a letter given at admission to hospital and/or sent to their place of residency. Exclusion criterion was opposition to data collection. We selected the patients included up to April 20, 2020 and with a chest computed tomography (CT) scan at admission. Patients from April 1st were prospectively included.

Exposure
Exposure to systemic corticosteroid at hospital admission was assessed by physicians and then extracted from electronic medical records. Drug, dosage, duration of use (categorized as short-term exposure for < 7 days vs ≥ 7 days) and indication were described.

Outcome
The primary outcome was composite, including admission to ICU, need of mechanical ventilation and death occurring during the 14 days after hospital admission.

Statistical analyses
For descriptive analyses, continuous variables were expressed by mean and standard deviation or median and interquartile range (IQR) depending on their distribution, and categorical variables by percentages. For comparatives analyses, multiple imputation (n = 5) was used to handle missing values [10]. A propensity score (PS) was calculated based on the covariables listed above [11]. Analyses were conducted using logistic regression models and were adjusted with overlap weighting (OW) on the PS, providing odds ratios (ORs) and their 95% confidence intervals (95% CIs) [12]. OW allows adjustment in population with large differences in covariables, by emphasizing the target population with the most overlap in observed characteristics between exposed and unexposed patients, and down-weighting the tails units. Exposed patients were weighted by the probability of not receiving corticosteroids (1 − PS) and unexposed patients were weighted by the probability of receiving corticosteroids (PS) [13]. Statistical analyses were performed using SAS V9.4™ (SAS institute, Cary, NC, USA).

Study population
Overall, 253 patients were included in the study. Characteristics of the patients are presented in Table 1. Median age was 64 years (IQR: 54-76), 140 patients (59.6%) were men, and 218 (86.2%) presented at least one comorbidity. The median duration of symptoms at the time of admission to hospital was 7 days (IQR: 4-10).

Comparative analyses
In the crude model, the OR of exposure to systemic corticosteroid at the time of admission to hospital with outcome occurrence was 1.64; 95% CI: 0.60-4.44. The PS distribution is presented in Supplementary Fig. S1. The PS was efficient in establishing balance for each covariable (data not shown). In the adjusted model with OW on the PS, the OR was 1.09 (95% CI: 0.65-1.83).

Discussion
In this study, we found a trend for an increased risk of poor outcome in COVID-19 hospitalized patients in case of previous exposure to corticosteroid before hospital admission. However, after adjustment for potential confounders, we found no evidence for an increased risk. This result differs from the two previously quoted studies in patients with rheumatic disease or chronic inflammatory bowel disease [6,7]. However, these were other settings, with no adjustment for other COVID-19 severity markers at the time of admission. Moreover, in the study conducted in patients with chronic inflammatory bowel disease, comorbidities were included quantitatively [7]. However, some comorbidities are expected to be more related with both exposure to corticosteroids and disease severity. That's why we included each comorbidity in the PS calculation. Of note, patients exposed to corticosteroids in our cohort had more frequently chronic lung disease, hypertension, and cause of immunosuppression only. This study conducted in the Covid-Clinic-Toul cohort presented strengths. The cohort is a clinical cohort with most of the data prospectively collected. Exposure to systemic corticosteroids was exhaustively assessed for each patient. Missing values were very rare and handled by multiple imputation. Adjustment using OW on the PS provided risk estimation by minimizing confusion bias due to important differences in the characteristics between exposed and unexposed patients.
Our study had several limitations. Data were restricted to a single hospital center. Only 6.7% (n = 17) of the patients were exposed to corticosteroids before hospital admission. This low sample size limited the interpretation of the results. Therefore, we could have only detected a major effect of corticosteroids on unfavorable outcome. Subgroup analyses by corticosteroids dosage, duration of exposure, and indications cannot be conducted due to this low number of exposed patients. Of note, 8 of the exposed patients (47%) in our cohort had a daily prednisone equivalent dosage < 10 mg, which was not associated with an increased rate of hospitalization in the study in patients with rheumatic disease [6]. Because previous exposure to systemic corticosteroids is strongly associated with their indication and because subgroups analyses were not possible, results need to be interpreted cautiously. Finally, we cannot exclude the presence of unmeasured confounding factors like smoking. However, these factors are certainly related to comorbidities included in the PS and it is not clear whether they impact COVID-19 severity.

Conclusion
Overall, this study provide some evidences for an absence of an increased risk of unfavorable outcome with previous exposure to corticosteroids in the general setting of patients hospitalized for COVID-19.