Demographic information of survivors and non-survivors
A total of 198 patients confirmed severe COVID-19 were enrolled in this study. The median age of patients and the gender distribution between two groups (survivor group and non-survivor group) was basically the same. Majority of the included patients in both two groups were with comorbidity and more than half of the patients had at least one underlying disease. The ranking of the underlying disease was hypertension (40.0%), diabetes (16.7%), chronic respiratory disease (5.7%), cardiovascular disease (5.2%) and so on. Among the underlying disease, the percentage of malignant disease of patients in the non-survivor group is higher than it in the other group. The most common symptoms on admission were fever and cough, followed by fatigue and sputum production in both two groups. During the clinical treatment, the most intense level of oxygen support was recorded. Patients in survivor group were mostly under oxygen therapy by nasal cannula in survivors compared to patients in non-survivor group (47% vs 0, P <0.05). The proportion of patients under invasive ventilation (IMV) in non-survivors was significantly higher than that in the other group (53.8% vs 2.2%, P <0.05). More than 70% of the patients received antivirals, and Lopinavir/Ritonavir use differed significantly between non-survivors and survivors (6.7% vs 100.0%, P <0.05). According to the CURB-65 score, the proportion of patients with different grade showed significant differences between two groups. Systematic glucocorticoids use differed significantly between non-survivors and survivors (66.7% vs 22.5%, P=0.002). At the end of the observing period, 185 (93.4%) patients had been discharged and 13 (6.6%) patients had died. Demographic information showed in Table 1.
Initial laboratory findings among survivors and non-survivors
Hematologic profile on admission varied among patients which hinted with different outcomes. Lymphocytes (LYM), monocytes (MONO) and platelets (PLT) were dramatically decreased in non-survivors with significant differences compared to the survivors (LYM: 10.9% vs 24.5%, P <0.05; MONO: 5.9% vs 8.7%, P <0.05; PLT: 145[76, 241] ×109/L vs 232[171, 305] ×109/L, P <0.05). Moreover, compared to survivors, the counts of eosinophils (EOS) and basophils (BASO) in non-survivors were too low to be detected and to analysis. In contrast with survivors, level of neutrophils (NEU) and the ratio of neutrophils-to-lymphocytes ratio (NLR) were significantly elevated in non-survivors (NEU: 83.6% vs 65.5%, P <0.05; NLR: 8.43 vs 3.05, P <0.05). The percentage alteration was in consistent with the absolute counts change for each analyzed peripheral blood cells. Moreover, compared with survivors, levels of C-reactive protein (CRP), IL-6, and serum ferritin were significantly increased in non-survivors with statistical differences. Regarding the coagulation parameters, the prolonging of PT and increased levels of D-dimer were significantly higher in non-survivors compared to survivors (Table 2; Figure 1).
Effects on clinical characteristics of different EOS levels in COVID-19 patients
According to the level of circulating EOS counts, COVID-19 patients were divided into two groups: low EOS group (< 0.02×109/L) and normal EOS group (≥ 0.02×109/L). The media age and gender distribution between the two groups without significant differences (age: 64[53, 71] yrs vs 61[47, 69] yrs, P=0.126; gender (male): 52.6% vs 48.3%, P=0.561). Fever of patients in the low EOS group were higher compared with normal EOS group with significant differences, rather than other symptoms (38.9[38.4, 39.0] °C vs 38.5[38.0, 39.0] °C, P=0.011). The severity of disease was evaluated by CURB-65 score. The percentage of CURB-65 score of patients between the two groups was differed with significant differences. The proportion of patients in low EOS group who used glucocorticoids was significantly higher than that of normal EOS group (44.8% vs 12.5%, P< 0.05). Duration of viral shedding in two groups showed no significant difference. The time from illness onset to discharge in low EOS group significantly longer than the other group (28[23, 35] d vs 25[20, 32] d, P=0.046). Moreover, death rate in the low EOS group was significantly higher and no patient death in normal EOS group (16.7% vs 0, P< 0.05) (Table 3).
Correlation networks analysis for peripheral blood cells
On admission, we observed a positive correlation between the counts of NEU and MONO (r =0.549), NEU and PLT (r =0.530) in non-survivors, but not in survivors. When at end-hospitalization, strong correlations between counts of NEU and MONO (r =0.771), NEU and EOS (r =0.735), NEU and BASO (r =0.623) were observed in non-survivors. Furthermore, non-survivors showed similar positive strong correlations between the counts of EOS and BASO (r =0.562) (Figure 2).
Patients who received glucocorticoids therapy showed a negative correlation between the counts of WBC and LYM (r =-0.265), but a positive correlation between WBC and LYM was observed in patients without glucocorticoids therapy (r =0.531). After glucocorticoids treatment, the counts of EOS negatively correlated with NEU (r =- 0.288), but no correlation was observed before the treatment (r = 0.058). A similar correlation was observed between the counts of LYM and EOS in patients received glucocorticoids therapy (r =0.454), but no correlation was found before the treatment (r =0.020) (Figure 3).