Trial characteristics
The searches identified 605 relevant articles. Of these, 43 trials were eligible for inclusion according to our criteria for considering studies for this meta-analysis[16, 17, 21-63] (Figure 1). Forty-two trials were retrospective case series (RCS), and one trial was a randomized controlled trial (RCT). There are 15 RCT protocols which were not included, still no results (Supplement). A total of 6603 patients with COVID-19 were included in the analyses. Among the 43 included trials, 11 were multicenter trials, and 22 were single-center trials. Twenty-six trials were conducted in Wuhan, China, 14 were conducted outside of Wuhan, China, and the other 3 trials were conducted outside of China. Eleven studies performed analyses by severity; 3 trials divided patients into the ICU or non-ICU groups, and 8 trials divided patients into the severe or nonsevere groups. Viral clearance was compared in 4 trials. The effect of mortality was analyzed in 7 trials. Most of the trials indicated that 40-80 mg methylprednisolone was used once or twice per day, ranging from 4-15 days. Antibiotics were not administered in three trials, 1 trial had no antibiotics data, and 39 trials administered antibiotics. The most common complications were ARDS, acute coagulopathy, acute liver injury and acute kidney injury. The characteristics of the 43 included trials are summarized in Table 1.
The proportion of corticosteroid treatments
The proportion of COVID-19 patients treated with corticosteroids vs. those that were not is described in all 43 included trials (n=6603 patients). The meta-analysis demonstrated that the proportion of COVID-19 patients treated with corticosteroids was significantly lower than that of patients who were not treated with corticosteroids (34.11% vs. 64.39%, 2252 vs. 4351 OR: 0.33, 95% CI: 0.17-0.61, P <0.01; Figure 2). There was evidence of significant heterogeneity between trials (P <0.01, I2 = 98%).
Comparing the proportion of severe patients to the proportion of nonsevere patients administered corticosteroids
The proportion of severe patients treated with corticosteroids was 29.95% (n=254) vs. 21.09% (n=382) of nonsevere patients treated with corticosteroids in 11 trials (n=2635 patients). The meta-analysis demonstrated no significant difference in the proportions of severe and nonsevere patients administered corticosteroids (OR: 2.36, 95% CI: 0.79-7.04, P =0.12; Figure 3). There was evidence of significant heterogeneity between trials (P <0.01, I2 = 94%).
The effect of corticosteroid use on viral clearance
We evaluated the viral clearance time in patients treated with corticosteroids and compared it to that of patients treated without corticosteroids using a random-effects model (Figure 4). The pooled estimates showed that corticosteroid treatment significantly delayed the viral clearance time (WMD: 2.96, 95% CI: 0.59-5.34, P < 0.01; I2 =77%). However, there was significant heterogeneity between studies.
The effect of corticosteroid use on mortality
The mortality of COVID-19 patients treated with corticosteroids for 4-15 days was described in 7 trials (n=724 patients). The meta-analysis demonstrated no significant difference in the use of corticosteroids between COVID-19 patients who died and those who survived (overall OR: 1.83, 95% CI 0.87-3.84, P = 0.11; Figure 5). There was evidence of significant heterogeneity between trials (P = 0.0006, I2 = 75%).
Subgroup and sensitivity analysis
Subgroup analyses stratified by region indicated that the proportion of COVID-19 patients administered corticosteroids was significantly lower than that of patients who were not administered corticosteroids in Wuhan, China (OR: 0.45, 95% CI: 0.22-0.91, P < 0.00001; I2 = 98%, Figure 2), outside of Wuhan (OR: 0.12, 95% CI: 0.03-0.43, P < 0.00001; I2 =98%, Figure 2) and outside of China (OR: 1.98, 95% CI: 0.02-158.5, P < 0.00001; I2 = 98%, Figure 2).
Subgroup analyses were also stratified by severity. Patients in the ICU or who were identified as severe or critical were collectively considered the “severe” group, while those not in the ICU as well as mild and common COVID-19 patients, were considered the “nonsevere” group. The subgroup analysis indicated that the proportion of patients treated with corticosteroids was significantly higher among ICU patients than among non-ICU patients (OR: 5.45 95% CI: 9.30, 2.93 P =0.0002; I2 = 59%; Figure 3), but there was no significant difference in the proportion of critical and severe patients versus mild and common level patients treated with corticosteroids (OR: 1.42 95% CI: 0.37, 5.41 P =0.61; I2 = 96%; Figure 3).
Assessment of Study Quality
The level of evidence for each trial was graded from 1 to 5 according to the Jadad quality score (Table). For publication bias, the shape of the funnel plot showed obvious asymmetry for trials investigating the proportion of corticosteroid use for COVID-19 patients, regardless of region or severity ( Supplemental Figures A,B), but slight asymmetry for trials investigating the effect of viral clearance (Supplemental Figures C) and mortality (Supplemental Figures D).