A total of 775 patients were included in our final analysis, of which 490 patients were from Wuhan and 285 patients were from Sichuan. Eventually, 238 (30.7%) patients received systemic corticosteroids treatment including 201 (41.0%) patients from Wuhan and 37 (13.0%) from Sichuan. Excluding the patients with ARDS separately collected from the 2 centres in Wuhan, the overall mortality rate was 5.4%, while the mortality rate in Wuhan and Sichuan was 7.7% and 1.3%, respectively.
Table 1 and 2 summarized the demographic characteristics and baseline blood test results between the patients with and without corticosteroids treatment. We found that, compared to the patients without corticosteroids treatment, patients with corticosteroids treatment were older patients (57±16 vs. 51±17, P<0.001) and there were more male patients (59.1% vs. 44.5%, P<0.001), more patients with symptoms (e.g. dyspnoea: 34.4% vs. 17.1%, P<0.001) and comorbidities (50.0% vs. 41.5%, P=0.030), and more severe patients (50.8% vs. 18.6%, P<0.001) and ARDS patients (26.9% vs. 5.8%, P<0.001). The patients in the corticosteroids treatment had significantly lower oxygenation (SpO2/FiO2: 194±96 vs. 303±81, P<0.001), blood lymphocytes (14.9±11.5% vs. 25.0±13.2%, P<0.001), and T cells (560±331 vs. 876±422, P<0.001) and its subsets (CD4 cells: 339±209 vs. 545±284, P<0.001; CD8 cells: 207±170 vs. 305±175, P<0.001), but higher neutrophils (76.6±16.8% vs. 64.9±14.3%, P<0.001), D-dimer (6.04±19.70mg/L vs. 2.39±9.31mg/L, P=0.015), and creatine kinase (CK) (155±365 vs. 98±126, P=0.036) than patients without corticosteroids treatment. Although significant differences could also be seen in RR, total bilirubin (TB), alanine aminotransferase (ALT), aspartate transaminase (AST), albumin (ALB), and blood urea nitrogen (BUN), however, they all fell into the normal ranges.
Due to the inconsistent baseline characteristics between patients with and without corticosteroids treatment, we tried to investigate whether these significant differences were derived from the different disease severities in each group. We found that almost all of these characteristics were significantly different between the two groups (Table S1). Therefore, we further compared the outcomes after stratification of the patients into different disease severities.
In total patients, compared with patients without corticosteroids treatment, patients with corticosteroids treatment had significantly higher mortality (19.3% vs. 3.7%, P<0.001) but lower rate of virus clearance (43.2% vs. 66.7%, P<0.001). (Table 3) After stratification of disease severities, corticosteroids treatment was associated with higher mortality in patients with severe-to-critical diseases (37.6% vs. 16.5%, P<0.001), while no significant difference was found in the patients with mild-to-moderate diseases (0% vs. 0.5%, P=0.302). However, there were no changes in the rate of virus clearance after stratification (mild-to-moderate: 47.0% vs. 67.4%, P<0.001; severe-to-critical: 37.7% vs. 63.2%, P=0.002). In patients with ARDS, no differences (mortality: 46.8% vs. 29.1%, P=0.090; rate of virus clearance: 38.1% vs. 62.5%, P=0.406) were shown in the two groups. (Table 3)
When compared the treatment outcomes of corticosteroids in patients from different regions, we found a similar pattern in patients from Wuhan, while in patients from Sichuan, no significant differences were shown in either disease severity groups. (Table 3)
Significantly longer hospital LOS (18.7±12.0 vs. 15.0±8.9, P<0.001) and virus clearance time (11.9±9.2 vs. 9.4±8.3, P=0.011) were found in patients with corticosteroids treatments but not in the subgroup of severe-to-critical disease severities (hospital LOS: 18.3±13.4 vs. 15.0±10.3, P=0.051; virus clearance time: 11.5±10.5 vs. 10.5±8.9, P=0.626), which was also true in patients from Wuhan and Sichuan. (Table 3)
In terms of the improvement of oxygenation and blood lymphocytes over time (24h vs. 48h vs. 72h vs. 7d), we found that corticosteroids could not improve oxygenation (SpO2/FiO2) except for patients with ARDS (160±68 vs. 230±109 vs. 265±123 vs. 216±37, P=0.044), neither could it increase blood lymphocytes except for patients with severe-to-critical diseases (9.2±8.9% vs. 13.8±7.8% vs. 15.3±11.2% vs. 16.0±10.9%, P=0.008) and ARDS (6.8±7.8% vs. 14.7±10.4% vs. 12.4±12.6% vs. 13.3±11.1%, P=0.050). (Table S2 and Fig. 1 and 2) However, despite the improvement of oxygenation and blood lymphocytes after corticosteroids treatment, the levels were still significantly lower than those in patients without corticosteroids treatment. (Table S2 and Fig. 1 and 2)
Risk factors for death associated with corticosteroids treatment
Table 4 showed the potential factors associated with death in patients with corticosteroids treatment. We found that the factors with significant differences were similar to those demographics and characteristics identified as significantly different between the patients with and without corticosteroids treatment. Compared to survivals, patients who died with corticosteroids treatment were older (72±13 vs. 54±15, P<0.001) and there were more patients with symptoms (e.g. dyspnoea: 66.7% vs. 27.9%, P<0.001) and comorbidities (74.4% vs. 45.6%, P=0.001), and more severe patients (100.0% vs. 39.7%, P<0.001) and ARDS patients (67.4% vs. 21.9%, P<0.001). Interestingly, we also noticed more patients who received corticosteroids treatment within 48 hours of admission in the group of death (79.2% vs. 51.9%, P=0.012). The patients in the corticosteroids treatment had significantly lower oxygenation (SpO2/FiO2: 132±45 vs. 226±99, P<0.001), blood lymphocytes (5.2±4.0% vs. 17.9±11.4%, P<0.001), and T cells (341±237 vs. 597±331, P=0.016), but higher neutrophils (91.0±5.1% vs. 72.2±16.9%, P<0.001), D-dimer (17.53±43.76mg/L vs. 3.73±8.54mg/L, P=0.006), TB (20.4±17.1μmol/L vs. 12.2±7.4μmol/L, P=0.001), ALT (54.8±56.2U/L vs. 34.4±24.4U/L, P=0.011), AST (67.6±42.1U/L vs. 31.7±16.3U/L, P<0.001), CK (445±802U/L vs. 93±91U/L), and BUN (11.54±9.10mmol/L vs. 5.43±4.14mmol/L, P<0.001) than patients without corticosteroids treatment upon admission.
After the correlation matrix analysis of the above potential risk factors (Table S3), we finally chose the variates with less internal correlations in our multivariate logistic regression model, which included age, blood lymphocytes, D-dimer, ALT, CK, and the percentage of comorbidities and use of corticosteroids within 48 hours of admission. The multivariate logistic regression showed that blood lymphocytes (odds ratio (OR) 0.792, 95% confidence interval (CI) 0.672-0.932, P=0.005) and CK (OR 1.006, 95%CI 1.000-1.012, P=0.038) were independent risk factors associated with death in patients with corticosteroids treatment. (Table 5) The ROC curves identified a cut-off value of 10.45% for blood lymphocytes (AUC 0.863, 95%CI 0.778-0.947, P<0.001) with a sensitivity and specificity of 90.91% and 70.75%, and a cut-off value of 239.50 U/L for CK (AUC 0.672, 95%CI 0.504-0.840, P=0.023) with a sensitivity and specificity of 44.44% and 94.05%. (Figure 3 and Table 5)