Although corticosteroid therapy has become a standard of care for patients with COVID-19 pneumonia, the efficacy and safety of its use remains controversial. The current widespread use of systemic corticosteroids in this clinical setting relies mostly on the results of the RECOVERY study, which showed that dexamethasone treatment reduced deaths by one-fifth in patients with COVID-19 pneumonia who needed supplemental oxygen.1 The results of three meta-analyses supported this finding,2–4 but it must be mentioned that the weight of the RECOVERY trial was more than 50 per cent in all of them and the magnitude of the effect was modest (OR ranging from 0.70 to 0.88). In this study, we sought to evaluate the effect of corticosteroids in preventing an unfavorable outcome in patients admitted to hospital with COVID-19 pneumonia.

Data were obtained from two different cohorts of patients admitted with COVID-19 pneumonia during the COVID-19 spring 2020 outbreak. The first cohort was composed of 1292 patients admitted to the eight Galician tertiary hospitals and the second was composed of 1548 patients admitted to four other Spanish hospitals. The study was approved by the Biomedical Research Ethics Committee of La Fe University and Polytechnic Hospital (2020-122-1) and by the Ethics Committee of Galicia (Cod. 2020/239). Data (demographic, clinical, laboratory) were collected on admission or during the first 24h of hospitalization.

Baseline characteristics were compared between patients with and without corticosteroid treatment. Data were reported as a percentage for categorical variables and as mean±standard deviation (SD) or median with interquartile range (IQR, 25–75%) for continuous variables. Categorical variables were compared using Fisher's exact test or Pearson's chi-square test, as appropriate, and continuous variables were compared using the Mann–Whitney U test or Student's t test.

To minimize the effect of a corticosteroid treatment selection bias and to control for potential confounding factors, a propensity score matching (PSM) procedure was performed. The primary outcome was a composite of in-hospital death and orotracheal intubation and the secondary outcome measure was mortality. The model included six pre-selected baseline variables that might have affected patient assignment to a corticosteroid or non-corticosteroid group, as well as clinical outcomes; specifically: room-air SpO2 for assessing the severity of respiratory failure, level of CRP for indicating systemic inflammatory response syndrome, age and sex as basic characteristics, comorbidities (using the Charlson index), and region (Galicia vs. rest of Spain). The treatment and control pairs were matched via the nearest neighbor matching approach and the caliper was set at 0.25. PSM was performed using the R package MatchIt version 3.6.2. A Cox regression model was conducted to examine the impact of corticosteroid treatment on patient outcomes. In order to examine the influence of early (<10 days from symptoms onset) versus late (≥10 days) initiation of corticosteroid therapy, a boosted multinomial logistic regression analysis was used to calculate propensity scores and to allow the comparison of three therapeutic conditions (early, late and no use of corticosteroids). Two-sided p values<0.05 were considered statistically significant. Statistical analyses were done using the SPSS version 22.0 (IBM SPSS, Armonk, NY).

Among the 2840 screened cases, 187 were excluded for death or orotracheal intubation within 24h after hospital admission, 301 were excluded for having received tocilizumab and 772 for absence of relevant data. Among the remaining 1580 patients, 464 were given corticosteroids. Nine hundred and twenty-eight patients were matched into 464 pairs. Clinical and laboratory characteristics in the cohort of patients exposed and not exposed to corticosteroids, before and after PSM, are shown in Table 1. Patients not receiving corticosteroids therapy were younger, less frequently male, had fewer comorbidities, better baseline SpO2 and lower CRP on admission. No significant differences were observed after matching, and characteristics for patients with or without corticosteroids were well balanced.

In the crude analysis, 122 patients (26.3%) corticosteroid-treated patients reached the composite outcome, compared to 122 patients (10.9%) who did not receive this therapy. This translated into an increased risk of death or need for mechanical ventilation (HR 1.91; 95% CI: 1.38–2.61). However, among 464 PSM pairs, there was no difference in the composite outcome (HR 1.64; 95% CI: 0.93–2.92).

With regard to the secondary outcome measure, 93 patients (20.0%) corticosteroid-treated patients died, in contrast to 98 patients (8.7%) in the non-corticosteroid group (HR 1.47; 95% CI: 1.02–2.10). Among the propensity score matched sample, mortality did not differ between corticosteroid-treated and non-corticosteroid-treated patients (HR 1.33; 95% CI: 0.94–2.28). These results are summarized in Table 2.

Due to data availability, only patients from the Galician cohort (n=1014) were included in the boosted multinomial logistic regression analysis. Early start was associated with an increased risk of death (OR 1.52; 95% CI: 1.03–2.23) whereas this association was not found among patients with late initiation (OR 0.92; 95% CI: 0.61–1.38).

These results are in concordance with those obtained in other PSM studies that found no impact of corticosteroids on COVID-19 pneumonia outcome.5–10 An increased mortality was observed in another report11 and two additional PSM studies concluded that corticosteroid therapy was associated with lower mortality,12,13 although they included critically ill patients, the specific population that could benefit the most from this therapy according to the results of the RECOVERY trial.1 Considering all the available evidence, the role of corticosteroids remains controversial in the clinical context of non-critically ill patients who need supplemental oxygen. It seems very plausible that the type, the dose, the duration and the timing of prescription (early start may increase viral replication) could influence the outcome, and the available literature has not accounted for these relevant factors. Moreover, it could be hypothesized that COVID-19 patients may suffer different histopathological patterns of lung injury, either corticosteroid-resistant (acute lung injury and diffuse alveolar damage) or steroid-responsive (i.e. organizing pneumonia), but more research is needed to investigate this possibility.

In accordance with the results of the RECOVERY trial,1 we found a harmful effect of early (<10 days) initiation of corticosteroids, hypothetically reflecting that anti-inflammatory therapy might not be advisable during the initial, replicative, phase of the process.

The strengths of this study are the well-defined and homogenous population included, the large size, the multicenter design and the careful minimization of confounding variables related to propensity of corticosteroid treatment through a PSM analysis. The main limitations include the retrospective design, possible residual selection bias (the variables used in the PSM do not represent all of the factors that could influence an adverse prognosis, and the Charlson index is a somewhat crude measure of comorbidities) and lack of information regarding timing, type and dosages of corticosteroids.

In conclusion, we found no association between the use of corticosteroids and intubation or death in patients hospitalized with COVID-19 pneumonia. Early initiation (<10 days from symptom onset) of corticosteroids might increase the risk of mortality and this therapeutic decision must be taken with caution, weighing risks and benefits.