Study Selection
In total, we identified 6,717 articles from the three databases. Three RCTs records (DEXA-COVID 19 (The efficacy of dexamethasone treatment for patients with ARDS caused by COVID-19; NCT04325061); COVID STEROID (The hydrocortisone for COVID-19 and severe hypoxia; NCT04348305); Steroids-SARI (Glucocorticoid therapy for COVID-19 critically ill patients with severe acute respiratory failure; NCT04244591)) were searched by related meta-analysis [13] to obtain the prospective data; the other two unpublished records [18, 19] were identified by medRxiv website. There were 4,508 records left after removing the duplicates. Furthermore, we identified 106 studies after preliminary screening by title or abstract. Two OS [20, 21] were not in the range of 95% CIs after sensitivity analysis, and so the two articles were deleted. Finally, our meta-analysis included 11 RCTs and 46 OS enrolling 7,893 and 4,1696 patients (Figure 1).
Study Characteristics
There were 11 RCTs [9-12, 22-25] (DEXA-COVID 19, NCT04325061; COVID STEROID, NCT04348305; Steroids-SARI, NCT04244591) and 46 OS [18, 19, 26-69] that reported an association between COVID-19 patients’ mortality and corticosteroid therapy. There were five RCTs [10, 22, 23] (DEXA-COVID 19, NCT04325061; COVID STEROID, NCT04348305; Steroids-SARI, NCT04244591) and 10 OS [30, 36, 42, 43, 45, 51, 53, 55, 59, 60] that used high doses of corticosteroid (>15 mg/d De), and 6 OS [33, 34, 36, 56-58] that administered pulse dose corticosteroid. There were two RCTs [9, 11] and 20 OS [18, 19, 32, 35, 39, 43, 44, 48, 49, 52, 54, 55, 57, 59, 61, 64, 66-69] that included severe or critical COVID-19 patients. Table 1 and Table S3 present the characteristics of literature included.
Risk of Bias Assessment
Figure S1 presents the RCTs of the Cochrane Collaboration bias risk evaluation tool. Four RCTs [10, 12, 22, 25] were high-risk bias because of performance bias. Three RCTs [9, 11, 24] were low-risk bias, and one trial [23] had unclear risk bias. According to the NOS, all 46 eligible OS [18, 19, 26-69] were greater than or equal to seven points, indicating a low-risk bias. Table S4 reports the specific contents of risk bias in the included OS.
Corticosteroid Effects on Outcomes
Figure 2 shows the preliminary results of RCTs. The results showed that the corticosteroid use did not reduce the in-hospital mortality significantly (OR, 0.88; 95% CI, 0.74–1.05; I2=66.9%; evidence rank, moderate). Considering the effect of confounding factors on mortality of OS, we listed the adjusted and unadjusted OR separately in the forest plot. The adjusted or unadjusted OS results were consistent with the RCTs (OR, 0.91; 95% CI, 0.74–1.11; I2=85.3%) (Figure S2). Figures 3 and S3 show the secondary outcomes of the RCTs. Corticosteroid administration did reduce the need for MV (OR, 0.67; 95% CI, 0.51–0.90; I2=7.5%; evidence rank, moderate) and did not statistically increase the serious adverse events (OR, 0.84; 95% CI, 0.30–2.37; I2=33.3%; evidence rank, moderate) among COVID-19 patients.
Subgroup Analysis with Mortality
We analyzed subgroups to explore the sources of high heterogeneity, according to severe or critical COVID-19 patients and corticosteroid type and dose. Figures S4 and S5 show that corticosteroid therapy did not result in a significantly lower hospital mortality in severe COVID-19 patients (RCT: OR, 0.66; 95% CI, 0.39–1.13; I2=28.9%; OS: OR, 1.09; 95% CI, 0.73–1.63; I2=88.8%). We also did a subgroup analysis on patients with corticosteroid dose. It was found that the low- or high-dose corticosteroids therapy was not associated with any significant positive impact on hospital mortality (low: OR, 0.91; 95% CI, 0.78–1.06; I2=63.4%; high: OR, 0.90; 95% CI, 0.43–1.89; I2=74.7%) (Figure S6). We also analyzed the effect of corticosteroid types on mortality (De: OR, 0.97; 95% CI, 0.77–1.23; I2=43.6%; Hy: OR, 0.82; 95% CI, 0.37–1.82; I2=58.6%; Me: OR, 0.64; 95% CI, 0.27–1.52; I2=85.2%) (Figure S7). We found that the pulse dose corticosteroid treatment improved survival in OS (pulse: OR, 0.52; 95% CI, 0.39–0.70; I2=2.7%; low: OR, 0.86; 95% CI, 0.71–1.06; I2=63.2%; high: OR, 1.37; 95% CI, 0.80–2.33; I2=82.2%) (Figure 4). The methylprednisolone use also lowered the mortality (Me: OR, 0.69; 95% CI, 0.52–0.92; I2=66.7%; Pre: OR, 0.94; 95% CI, 0.54–1.65; I2<0.01%; Hy: OR, 1.36; 95% CI, 0.93–1.99; I2<0.01%; De: OR, 0.49; 95% CI, 0.15–1.53; I2=93.0%) (Figure 5).
Sensitivity Analyses
Due to the high heterogeneity of our results, we conducted a sensitivity analysis to evaluate the impact of any single study on the pooled OR and 95% CI by omitting one study at a time. We found that these RCTs and OS results were robust and reliable (Figures S8 A and B).
Publication Bias Assessment
We performed funnel plots (Figures S9 A and B) and Begg’s regression tests to examine the publication bias of the included studies. There was no significant publication bias in RCTs (P=0.876) and OS (P=0.425).