A total of 3009 citations was identified through electronic database searches. Of these, 1021 were duplicate reports. After screening titles and abstracts, 1694 articles were excluded and 294 full-text articles were retrieved for more assessment. Finally, 17 articles (22 studies) were found to be eligible for this meta-analysis. We summarized the process of study identification and selection in Fig. 1.
All included articles were conducted in China (12-25) except one that was performed in Singapore (26). Twenty-three studies included 3396 (ranging from 12 to1099) patients who were 720 in severe and 2676 in non-severe groups. The number of studies with different laboratory tests are as following: twenty-two studies were on lymphocyte, 21 on WBC, 18 on neutrophil and CRP, 17 on platelet, 15 on ALT, 14 on AST, Cr, and CK, 12 on albumin, PCT, and D-dimer, 11 on LDH, 10 on monocyte, 9 on hemoglobin, TBIL, BUN, ESR, sodium, and potassium, 8 on PT, 6 on IL-6, 5 on eosinophil and troponin I, 4 on fibrinogen and glucose, and 3 on myoglobin. Most studies clearly stated that data were basic (on admission / before treatment) laboratory test results (12-14, 16, 17, 19-26). The study characteristics of these 23 included articles are presented in Table 1.
Main outcomes
The mean difference forest plots of laboratory features in severe vs. non-severe hospitalized patients with COVID-19 is shown in Appendix 2: 2a-2h.
CBC with differential counts of WBC
Using random-effects model, our meta-analyses showed a significant decrease in the WMD of lymphocyte [WMD= -0.43× 109 per L; 95 % CI, -0.56, -0.30, P< 0.001; I2= 90.1% (with 22 studies)], monocyte [WMD= -0.06× 109 per L; 95 % CI, -0.12, -0.01, P= 0.032; I2= 71.8% (with 10 studies)], and eosinophil [WMD= -0.03×109 per L; 95 % CI, -0.05, -0.00, P= 0.037; I2= 86.1% (with 5 studies)], hemoglobin [WMD= -5.94 g/L; 95 % CI, -8.23, -3.64, P< 0.001; I2= 0.0% (with 9 studies)], platelet [WMD= -27.97× 109 per L; 95 % CI, -39.60, -16.35, P< 0.001; I2= 55.8% (with 17 studies)], and increased in the WMD of neutrophil [WMD= 0.74 ×109 per L; 95 % CI, 0.16, 1.33, P= 0.013; I2= 74.6% (with 18 studies)], in the severe group compared with the non-severe group. However, no significant differences were found in WBC [WMD= 0.55×109 per L; 95 % CI, -0.09, 1.19, P= 0.094; I2= 81.3% (with 21 studies)] between the two groups were observed (Appendix 2a, Fig: A-G).
Laboratory tests for liver and kidney function
The results indicated a significant decrease in the WMD of albumin [WMD= -4.20 g/L; 95 % CI, -5.99, -2.41, P< 0.001; I2= 73.9% (with 12 studies)], and increased in the WMD of ALT [WMD= 6.65 U/L; 95 % CI, 4.21, 9.09, P< 0.001; I2= 0.0% (with 15 RCTs)], AST [WMD= 11.91 U/L; 95 % CI, 8.29, 15.53, P< 0.001; I2= 46.1% (with 14 studies)], TBIL [WMD= 0.08 mg/dL; 95 % CI, 0.03, 0.14, P= 0.005; I2= 0.0% (with 9 studies)], BUN [WMD= 2.34 mg/dL; 95 % CI, 0.66, 4.03, P= 0.006; I2= 39.3% (with 9 studies)], and Cr [WMD= 0.08 mg/dL; 95 % CI, 0.03, 0.12, P< 0.001; I2= 0.0% (with 14 studies)] in the severe group compared with the non-severe group (Appendix 2b, Fig: A-F).
Myocardial enzymes and Myoglobin
The pooled findings showed no significant differences between the two groups of COVID-19 patients on myocardial enzymes and myoglobin, including CK [WMD= -3.01 U/L; 95 % CI, -12.91, 6.90, P=552; I2= 51.7% (with 14 studies)], troponin I [SMD= 0.27; 95 % CI, -0.14, 0.67, P= 0.193; I2= 77.3% (with 5 studies)], and myoglobin [WMD= 8.11 ng/mL; 95 % CI, -6.10, 22.33, P=0.263; I2= 73.2% (with 3 studies)] (Appendix 2c, Fig: A-C).
Inflammatory markers
Our findings of inflammatory markers showed a significant increase in the WMD of ESR [WMD= 27.67 mm/h; 95 % CI, 22.94, 32.40, P< 0.001; I2= 22.3% (with 9 studies)], CRP [WMD= 36.61 mg/L; 95 % CI, 24.40, 48.82, P< 0.001; I2= 91.9% (with 18 studies)], LDH [WMD= 102.15 U/L; 95 % CI, 72.76, 131.53, P< 0.001; I2= 50.3% (with 11 studies)], and PCT [WMD= 0.03 ng/mL; 95 % CI, 0.00, 0.06, P= 0.043; I2= 41.1% (with 12 studies)] in the severe group compared with the non-severe group. While no significant changes were observed in IL-6 [SMD= 0.54; 95 % CI, -0.37, 1.45, P= 0.243; I2= 95.5% (with 6 studies)] between the the two groups (Appendix 2d, Fig: A-E).
Serum electrolytes
The pooled results of serum electrolytes among severe compared to non-severe patients indicated a significant decrease in the WMD of sodium [WMD= -1.95 mmol/L; 95 % CI, -2.87, -1.03, P< 0.001; I2= 75.5% (with 9 studies)], but non-significant difference on potassium [WMD= -0.07 mmol/L; 95 % CI, -0.18, 0.04, P= 0.206; I2= 34.3% (with 9 studies)] (Appendix 2e, Fig: A and B).
Laboratory tests for coagulation functions
Pooled findings on laboratory tests for coagulation functions showed a significant increase in the WMD of fibrinogen [WMD= 0.80 g/L; 95 % CI, 0.32, 1.28, P= 0.001; I2= 82.2% (with 4 studies)], PT [WMD= 0.63 second; 95 % CI, 0.27, 0.99, P= 0.001; I2= 69.2% (with 8 studies)], and D-dimer [WMD= 0.18 mg/L; 95 % CI, 0.10, 0.27, P< 0.001; I2= 99.3% (with 12 studies)] in severe vs. non-severe hospitalized patients (Appendix 2f, Fig: A-C).
Glucose level
We found a significant increase in glucose levels among the severe patients [WMD= 12.43 second; 95 % CI, 1.95, 22.91, P= 0.020; I2= 0.0% (with 4 studies)] when compared with non-severe patients (Appendix 2g, Fig: A).
Combined markers
The pooled findings on the new combined markers showed a significant increase in the SMD of NLR [SMD= 0.23; 95 % CI, 0.08, 0.37, P= 0.002; I2= 14.6% (with 18 studies)] and a decreases in LCR [SMD= -8.12; 95 % CI, -10.05, -6.18, P= 0.001; I2= 98.6% (with 18 studies)], LeCR [SMD= -1.47; 95 % CI, -2.13, -0.80, P= 0.001; I2= 94.6% (with 17studies)], and (LeIR) [SMD= -0.99; 95 % CI, -1.98, -0.00, P= 0.049; I2= 93.6% (with 5 studies)] in severe vs. non-severe hospitalized patients infected by COVID-2019 (Appendix 2h, Fig: A-D).
Sensitivity analysis
We evaluated the effect of each study on the strength of the pooled WMDs or SMDs by excluding each study from the meta-analysis. We found no significant differences between the pre- and post-sensitivity pooled effect sizes for lymphocyte, hemoglobin, platelet, neutrophil, albumin, ALT, AST, TBIL, BUN, Cr, CK, troponin I, myoglobin, ESR, CRP, LDH, IL-6, sodium, fibrinogen, PT, D-dimer, glucose level, NLR, LCR, and LeCR. However after omitting Liu (Spngqiao) (a) et al. (22), the study on monocyte, (WMD= -0.04, 95%CI: -0.10, 0.01), Liu (Spngqiao) et al. (b) (22), the study on eosinophil (WMD= -0.02, 95%CI: -0.06, 0.01), Guan et al (18), the study on WBC (WMD=0.65, 95%CI: 0.02, 1.27), Liu (Yanli) et al. (22), the study on PCT (WMD=0.02, 95%CI: -0.004, 0.05), Huang et al. (16), the study on potassium (WMD=-0.10, 95%CI: -0.18, -0.02), and Deng et al. (b) (17) the study on LeIR (SMD=-0.52, 95%CI: -1.35, 0.31), we found significant differences between pre- and post-sensitivity pooled effect sizes.
Publication bias
The Egger’s regression and Begg’s rank correlation tests were performed to detect potential publication bias. These indicated no significant publication bias for lymphocyte, monocyte, eosinophil, hemoglobin, platelet, neutrophil, albumin, AST, ALT, TBIL, BUN, Cr, CK, troponin I, myoglobin, ESR, LDH, PCT, IL-6, potassium, fibrinogen, PT, D-dimer, glucose, NLR, LeCR, and LeIR. Because there was evidence of publication bias on WBC [Egger (p< 0.01), Begg (P=0.13)], CRP [Egger (p< 0.01), Begg (P< 0.01)], sodium [Egger (p= 0.02), Begg (P= 0.29)], and LCR [Egger (p< 0.01), Begg (P< 0.01)], we conducted the non-parametric method (Duval and Tweedie) to estimate the findings of censored studies. There were no significant differences between before and after including censored studies for WBC, sodium, and LCR but not for CRP [before (WMD= 36.61 mg/L; 95 % CI, 24.40, 48.82) and after (WMD=12.21 mg/L; 95 % CI, -1.28, 25.84)].