Introduction

As of 27 September 2021, the outbreak of coronavirus disease 2019 (COVID-19) has affected more than 200 countries, with 231,703,120 confirmed cases and 4,746,620 deaths globally [1]. The disease is caused by infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which results in a large number of severe/critical ill patients who require rigorous management in intensive-care units (ICUs) [2,3,4]. Until now, there has been no consensus on an effective method to eradicate SARS-CoV-2. Prompt recognition and supportive care for potentially severe/critical ill patients are the mainstay treatments to save lives.

Our previous study [5] showed that the counts of lymphocytes, T cell subsets, and eosinophils decreased markedly in severely and fatally ill patients. Non-survivors maintained high levels of, or showed an upward trend in, neutrophil (Neu) counts, interleukin-6 (IL-6), procalcitonin (PCT), serum amyloid A protein (SAA), and C-reactive protein (CRP) levels, while levels of these markers held stable or showed a downward trend in survivors. In addition, studies from other research groups have also investigated the correlation between abnormal immune parameters, including white blood cells (WBC), lymphocytes (Lym), and eosinophil (Eos) counts, infection-related variables, serum inflammatory-cytokine levels, and severity or mortality of the disease [5,6,7]. Indeed, identifying early and sensitive indicators representative of innate and adaptive immune responses to COVID-19 may help predict the disease progression and potential fatal outcomes.

The evidence of immune abnormalities associated with disease severity and mortality in COVID-19 patients has been widely reported in many published observation clinical studies. However, these studies presented a significant heterogeneity in demographic characteristics, genetic features, and therapeutic approaches before hospital admission. Although previous systematic meta-analyses provided evidence of immune signatures in patients with COVID-19 in the early phase of the disease outbreak [8,9,10,11], a number of studies have emerged that offer updated data on the immune abnormality associated with poor clinical outcomes [12,13,14,15,16]. Therefore, we aimed to obtain updated, comprehensive evidence of the immune index alongside hematological, biochemical, inflammatory, and coagulation parameters in either a severity or mortality cohort to present the interplay between impaired immune responses and multi-system abnormality contributing to disease progression.

Materials and Methods

Search Strategy and Selection Criteria

This systematic review was conducted according to the Preferred Reporting in Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We previously registered this meta-analysis in PROSPERO, and the study registration number is CRD42020196272. We searched seven databases, PubMed, Scopus, Ovid, Willey, Web of Science, Cochrane Library and the China National Knowledge Infrastructure (CNKI), using the advanced search mode in the field “Title/Abstract,” the search terms [“COVID-19” OR “SARS-COV-2”] AND [“biomarkers” OR “predictors” OR “parameters”] AND [“severity” OR “mortality”], from January 1, 2020 to August 20, 2021, without any language restrictions. After removing duplicates by Endnote, two reviewers independently assessed the title, abstract, and full text of each article to identify eligibility. Any disagreements were solved by a discussion with a third reviewer to reach a consensus. We included observational studies that consisted of two groups: (a) patients with different severities of COVID-19 and/or (b) patients who died from COVID-19 compared to those who survived. Articles with computable data about immune-related variables, including immunological and/or hematological, coagulation, inflammatory, and biochemical variables, were included in the current meta-analysis. The following results were excluded: reviews and meta-analysis, case reports, editorials, preprints, correspondences and letters, data papers, notes, comments, news, short surveys, erratums and retractions, guidelines, and mathematical models. Moreover, we used the Newcastle–Ottawa Scale (NOS) to evaluate the quality of each included publication.

Data Collection

Based on the classifications of the COVID-19 Diagnosis and Treatment Guideline in China (Interim version 8) [17], the severity of disease was classified as four types: mild, moderate, severe and critical. As the originally reported clinical groups were highly diverse among the included publications, we attempted to combine them into two groups, severe COVID-19 and non-severe COVID-19, for further meta-analysis. The strategy for this combination was as follows: (1) groups consisting of severe or critical cases, cases treated in ICUs, aggravations, refractory disease, and ARDS cases; and (2) groups consisting of non-severe, mild, moderate, common, ordinary, or general cases, cases not treated in ICUs, no aggravations, and cases without ARDS were placed into the non-severe COVID-19 group. Raw published/publicly available data were extracted, verified in duplication, and combined into a single database. In order to present the detailed characteristics of included studies, we extracted basic information of each study, which included the first author, year of publication, country and region, language, original reported groups, combined groups, average age, gender, and sample size of the “case and control groups.” We defined “severe and non-survivors” as “case groups” and “non-severe and survivors” as “control groups.” We also described the collected parameters in each study, including immunological and hematological parameters that are closely associated with immune function, and a few indexes reflecting infection, coagulation, and biochemical status. The final item was the quality score of studies, evaluated by the Newcastle–Ottawa Scale, with a higher score meaning higher quality.

Statistical Analysis

All analyses were performed using R software version 3.6.2 (package: meta/metafor; R Project for Statistical Computing, https://www.r-project.org). We divided studies into two separate cohorts for analysis: a severity cohort and a mortality cohort. For the meta-analysis, we transformed the format of laboratory variables presented as “median [interquartile range (IQR)]” into that of “mean [standard deviation (SD)]” [18, 19]. The value of “mean (SD)” of each included variable in the combined groups was calculated with the raw data from the originally reported groups using the formula proposed by Zhang et al. [20]. Standardized mean differences (SMDs) and 95% confidence intervals (95%CIs) were calculated as the primary metrics for each laboratory variable. Laboratory data was pooled whenever two or more publications reported a given variable. We quantified the variations in observed laboratory variables across studies attributable to heterogeneity using the I2 statistic, a metric ranging from 0% (indicating that all the heterogeneity was spurious) to 100% (indicating that all the heterogeneity was “real” and required further examination or explanation). To probe the sources of heterogeneity, we conducted a meta-regression analysis with three potential factors: the approach of combining disease severity, age, and region. The included variables that presented high heterogeneity (I2 > 50%) and were reported by an adequate number of studies (n ≥ 10) were applied to the analysis. In addition, the robustness of the results was applied by performing leave-one-out sensitivity analysis. The funnel plot method was used to test the publication bias.

Results

Figure 1 shows the flow diagram of selecting studies according to the PRISMA guidelines. We identified a total of 8552 records by searching seven databases. After removing duplicates, we screened the title and abstract of 5461 articles and excluded ineligible study designs (n = 2061) and unrelated to the topic (n = 1782). Then, we assessed 1618 full-text articles and excluded 1473 publications, mainly owing to no targeted groups (n = 654) and lacking of available and computable laboratory data (n = 819). Ultimately, we included 145 eligible publications in the systematic review and meta-analysis [5, 21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164]. Among the included studies, 91 ones were from China; and 54 studies were from America, Pakistan, Japan, Italy, France, Turkey, Korea, UK, Saudi Arabia, Egypt, India, Serbia, Greece, Libya, Spain, Iran, Mexico, Poland, Germany, and the Netherlands. All studies reported that laboratory variables were measured on admission or early during the hospitalization. There were 137 studies published in English and 8 studies published in Chinese. The characteristics of the included studies are presented in Table 1. Detailed results of the quality assessment of the included studies are presented in Fig. E1.

Fig. 1
figure 1

PRISMA flowchart of the study selection process

Full size image
Table 1 Characteristics of included studies
Full size table

Immunological Results

A total of 26 immunological variables were included for comparisons between patients with severe and those with non-severe COVID-19, including IL-1β, IL-1Ra, IL-2, IL-2R, IL-4, IL-6, IL-8, IL-10, IL-18, tumor necrosis factor-alpha (TNF-α), interferon-γ (IFN-γ), CD3-positive T-lymphocyte absolute count (CD3+ T[ab]), CD3+ T percentage (CD3+T[%]), CD4+T(ab), CD4+T(%), CD8+T(ab), CD8+T(%), CD4+T(ab)/CD8+T(ab) ratio, B-lymphocyte absolute count (B cell[ab]), Natural-killer cell absolute count (NK[ab]), immunoglobulin A (IgA), IgM, IgG, IgE, C3(Complement 3), and C4. Of these, IL-1β, IL-2, IL-2R, IL-4, IL-6, IL-8, IL-10, TNF-α, IFN-γ, CD3+ T(ab), CD3+ T(%), CD4+ T(ab), CD4+T(%), CD8+ T(ab), CD8+T(%), CD4+T(ab)/CD8+T(ab) ratio, B cell(ab), NK cell(ab), IgA, IgM, IgG, C3, and C4 were available for comparisons between non-survivors and survivors infected with SARS-CoV-2. The summarized results are presented in Fig. 2. The detailed forest plots are presented in Fig. E2.

  1. 1.

    Severe Versus Non-severe COVID-19

Fig. 2
figure 2

Summary result of the comparison of immunological parameters between patients with severe COVID-19 and non-severe COVID-19 (A), and between non-survivors and survivors with COVID-19 (B)

Full size image

IL-1β, IL-1Ra, IL-2R, IL-4, IL-6, IL-8, IL-10, IL-18, TNF-α, IFN-γ, IgA, and IgG were significantly increased in patients with severe versus those with non-severe COVID-19 (IL-1β = 0.13 [95%CI, 0.03 to 0.24], P = 0.0121, I2 = 39.1%; IL-1Ra = 0.71 [95%CI, 0.45 to 0.98], P < 0.001, I2 = 28.6%; IL-2R = 1.05 [95%CI, 0.65 to 1.44], P < 0.0001, I2 = 89%; IL-4 = 0.53 [95%CI, 0.11 to 0.95], P = 0.014, I2 = 92.3%; IL-6 = 1.07 [95%CI, 0.88 to 1.25], P < 0.001, I2 = 91.2%; IL-8 = 0.69 [95%CI, 0.45 to 0.94], P < 0.0001, I2 = 69.5%; IL-10 = 0.91 [95%CI, 0.61 to 1.20], P < 0.001, I2 = 92.9%; IL-18 = 0.71 [95%CI, 0.37 to 1.05], P < 0.001, I2 = 63%; TNF-α = 0.28 [95%CI, 0.05 to 0.51], P = 0.0186, I2 = 87.3%; IFN-γ = 0.44 [0.07; 0.81], P = 0.0196, I2 = 89.2%; IgA = 0.18 [95%CI, 0.07 to 0.29], P < 0.001, I2 = 24.2%; IgG = 0.11 [95%CI, 0.01 to 0.22], P = 0.0335, I2 = 46.6%); whereas CD3+ T(ab), CD3+ T(%), CD4+ T(ab), CD4+ T(%), CD8+ T(ab), CD8+ T(%), Total B cell(ab), NK cell(ab), and IgM were significantly decreased in patients with severe versus those with non-severe COVID-19 (CD3+ T(ab) =  −1.06 [95%CI, −1.24 to −0.89], P < 0.001, I2 = 77.6%; CD3+ T(%) =  −0.58[95%CI, −0.87 to −0.29], P < 0.001, I2 = 78.2%; CD4+ T(ab) =  −1.09 [95%CI, −1.29 to −0.89], P < 0.001, I2 = 86.9%; CD4+ T(%) =  −0.21[95%CI, −0.32 to −0.09], P < 0.001, I2 = 37.9%; CD8+T(ab) =  −1.00 [95%CI, −1.20 to −0.80], P < 0.001, I2 = 86.3%; B cell(ab) =  −0.70 [95%CI, −1.02 to −0.38], P < 0.001, I2 = 87.4%; NK cell(ab) =  −0.56 [95%CI, −0.79 to −0.33], P < 0.001, I2 = 78.1% and IgM =  −0.21 [95%CI, −0.32 to −0.11], P < 0.001, I2 = 26.1%). There were no differences in IL-2, CD8+ T(%), CD4+T/CD8 + T ratio, C3, C4, and IgE between the two groups.

  1. 2.

    Non-survivors Versus Survivors of COVID-19

IL-1β, IL-2R, IL-6, IL-8, IL-10, TNF-α and CD4+T/CD8+T ratio, IgA, and IgG were significantly increased in non-survivors versus survivors of COVID-19 (IL-1β = 0.72 [95%CI, 0.48 to 0.96], P < 0.001, I2 = 40.6%; IL-2R = 1.44 [95%CI, 1.04 to 1.83], P < 0.001, I2 = 84.3%; IL-6 = 1.13 [95%CI, 0.99 to 1.27], P < 0.001, I2 = 83.6%; IL-8 = 1.02 [95%CI, 0.70 to 1.35], P < 0.001, I2 = 81.6%; IL-10 = 1.19 [95%CI, 1.08 to 1.30], P < 0.001, I2 = 49.1%; TNF-α = 0.68 [95%CI, 0.38 to 0.97], P < 0.001, I2 = 83.8%; CD4+T/CD8+T ratio = 0.71[95%CI, 0.30 to 1.11], P = 0.0007, I2 = 94%; IgA = 0.36 [95%CI, 0.23 to 0.49], P < 0.001, I2 = 0%; IgG = 0.39[95%CI, 0.26 to 0.52], P < 0.001, I2 = 27%). CD3+ T(ab), CD4+ T(ab), CD8+ T(ab), CD8+ T(%), B cell(ab), NK cell(ab) and C3 were significantly decreased in non-survivors versus survivors (CD3+ T(ab) =  −1.51 [95%CI, −1.89 to −1.13], P < 0.001, I2 = 94%; CD4+ T(ab) =  −1.12[95%CI, −1.45 to −0.80], P < 0.001, I2 = 94%; CD8+T(ab) =  −1.18[95%CI, –1.57 to −0.78], P < 0.001, I2 = 96%; CD8+ T(%) =  −0.62 [95%CI, −0.79 to −0.45], P < 0.001, I2 = 23%; B cell(ab) =  −0.35 [95%CI, −0.67 to −0.02], P = 0.0367, I2 = 87.7%; NK cell(ab) =  −0.61 [95%CI, −0.83 to −0.40], P < 0.001, I2 = 54%; C3 =  −0.63[95%CI, -1.02 to −0.24], P = 0.0014, I2 = 83%). There were no differences in IL-2, IL-4, IFN-γ, CD3+ T(%), CD4+ T(%), C4, and IgM between the two groups.

Hematological Results

Eleven hematological variables, including WBC, neutrophil (Neu), lymphocyte (Lym), eosinophil (Eos), monocyte (Mono), basophil (Bas) absolute counts and platelet (PLT), hemoglobin (HB), neutrophil/lymphocyte ratio(NLR), lymphocyte/monocyte ratio (LMR), and platelet/lymphocyte ratio (PLR), were included in the meta-analysis for comparisons between patients with severe and non-severe COVID-19. All hematological parameters were available for comparisons between non-survivors and survivors of COVID-19. The summarized results are presented in Fig. 3. The detailed forest plots are presented in Fig. E3.

  1. 1.

    Severe Versus Non-severe COVID-19

Fig. 3
figure 3

Summary result of the comparison of hematological parameters between patients with severe COVID-19 and non-severe COVID-19 (A), and between non-survivors and survivors with COVID-19 (B)

Full size image

WBC, Neu, NLR, and PLR counts were significantly increased in patients with severe versus those with non-severe COVID-19 (WBC = 0.48 [95%CI, 0.37 to 0.59], P < 0.001, I2 = 83.7%; Neu = 0.73 [95%CI, 0.63 to 0.84], P < 0.001, I2 = 80.2%; NLR = 0.95 [95%CI, 0.70 to 1.20], P < 0.001, I2 = 87%; PLR = 0.47 [95%CI, 0.27 to 0.68], P < 0.001, I2 = 77.3%), whereas Lym, Mono, LMR, Eos, PLT and HB were significantly decreased in patients with severe versus those with non-severe COVID-19 (Lym =  −0.74 [95%CI, −0.87 to −0.61], P < 0.001, I2 = 89.9%; Mono =  −0.10 [95%CI, −0.21 to −0.00], P = 0.0465, I2 = 50.2%; Eos =  −0.39 [95%CI, −0.55 to −0.22], P < 0.001, I2 = 79.1%; LMR =  −0.94 [95%CI, −1.05 to −0.83], P < 0.001, I2 = 46.8%; PLT =  −0.27 [95%CI, −0.41 to −0.133], P < 0.001, I2 = 86.1%; HB =  −0.21 [95%CI, −0.36 to −0.06], P = 0.006, I2 = 83.7%) There was no difference in the Bas count between the two groups.

  1. 2.

    Non-survivors Versus Survivors of COVID-19

Similarly, WBC, Neu, NLR, and PLR were significantly increased in non-survivors versus survivors of COVID-19 (WBC = 0.74 [95%CI, 0.62 to 0.86], P < 0.001, I2 = 90.4%; Neu = 0.96 [95%CI, 0.82 to 1.10], P < 0.001, I2 = 89.9%; NLR = 0.45 [95%CI, 0.08 to 0.82], P = 0.0169, I2 = 97.4%; PLR = 0.44 [95%CI, 0.02 to 0.86], P = 0.038, I2 = 84.1%), whereas Lym, Eos, LMR, PLT and HB were significantly decreased in non-survivors versus survivors (Lym =  −0.70 [95%CI, −0.83 to −0.56], P < 0.001, I2 = 92.5%; Eos =  −0.65 [95%CI, −0.78 to −0.53], P < 0.001, I2 = 50.2%; LMR =  −2.15 [95%CI, −3.77 to −0.54], P = 0.009, I2 = 98%; PLT =  −0.35 [95%CI, −0.43 to −0.26], P < 0.001, I2 = 75%; HB =  − 0.41 [95%CI, −0.61 to −0.21], P < 0.001, I2 = 95.3%). There were no differences in the Mono and Bas count between the two groups.

Other Abnormal Clinical Parameters Deriving from Immune Dysfunction

Beyond immunological and hematological cells, cytokines, antibodies and complements, there are some other laboratory parameters that are related to immune dysfunction and reflect the progression of COVID-19 which have been reported in previous studies [165,166,167,168]. In the current study, we simultaneously included coagulation parameters (including prothrombin time(PT), activated partial thromboplastin time(APTT), D-dimer and fibrinogen (FIB)), inflammatory parameters (containing C-reactive protein(CRP), procalcitonin(PCT), erythrocyte sedimentation rate(ESR), serum amyloid A(SAA)) and ferritin, biochemical parameters (including cardiac function related ones such as creatine kinase(CK), cardiac troponin I(cTnI), myoglobin (MYO), lactate dehydrogenase (LDH), liver function related ones such as aspartate aminotransferase(AST), alanine aminotransferase (ALT), total bilirubin(TBIL) and kidney function related ones such as creatinine(CRN), albumin(ALB), blood urea nitrogen(BUN). The summarized results are presented in Fig. 4.

Fig. 4
figure 4

Summary result of the comparison of coagulation, inflammatory and biochemical parameters between patients with severe COVID-19 and non-severe COVID-19 (A), and between non-survivors and survivors with COVID-19 (B)

Full size image

Coagulation Results

Four coagulation variables, namely prothrombin time (PT), activated partial thromboplastin time (APTT), D-dimer and fibrinogen (FIB), were included in this study. All coagulation variables were available for comparisons between non-survivors and survivors of COVID-19. The detailed forest plots are presented in Fig. E4.

  1. 1.

    Severe Versus Non-severe COVID-19

The included four coagulation variables were significantly increased in patients with severe versus those with non-severe COVID-19 (PT = 0.60 [95%CI, 0.46 to 0.75], P < 0.001, I2 = 76.7%; APTT = 0.40 [95%CI, 0.13 to 0.67], P = 0.0033, I2 = 91.5%; D-dimer = 0.79 [95%CI, 0.65 to 0.93], P < 0.001, I2 = 86.8%; FIB = 0.62 [95%CI, 0.43 to 0.81], P < 0.001, I2 = 79.1%).

  1. 2.

    Non-survivors Versus Survivors of COVID-19

Similarly, all the four coagulation variables were significantly increased in non-survivors versus survivors of COVID-19 (PT = 0.77 [95%CI, 0.55 to 0.98], P < 0.001, I2 = 90.1%; APTT = 0.26 [95%CI, 0.04 to 0.48], P = 0.0187, I2 = 86%; D-dimer = 0.94 [95%CI, 0.80; 1.09], P < 0.001, I2 = 91.8%). However, there was no difference in FIB between the two groups.

Inflammatory Results

Five inflammatory variables, C-reactive protein (CRP), procalcitonin (PCT), erythrocyte sedimentation rate (ESR), serum amyloid A (SAA) and ferritin, were included for comparisons between patients with severe and those with non-severe COVID-19, Of these, CRP, PCT, ESR and ferritin were available for comparisons between non-survivors and survivors infected with SARS-CoV-2. The detailed forest plots are presented in Fig. E5.

  1. 1.

    Severe Versus Non-severe COVID-19

Levels of all five inflammatory variables were significantly increased in patients with severe versus those with non-severe COVID-19 (CRP = 1.09 [95%CI, 0.96 to 1.22], P < 0.001, I2 = 88.4%; PCT = 0.76 [95%CI, 0.36 to 1.15], P < 0.001, I2 = 97.4%; ESR = 0.68 [95%CI, 0.44 to 0.93], P < 0.001, I2 = 90.4%; ferritin = 0.83 [95%CI, 0.63 to 1.02], P < 0.001, I2 = 86.3%; SAA = 1.13 [95%CI, 0.71 to 1.56], P < 0.001, I2 = 90.8%).

  1. 2.

    Non-survivors Versus Survivors of COVID-19

CRP, PCT, and ferritin were significantly increased in non-survivors versus survivors of COVID-19 (CRP = 1.08 [95%CI, 0.95 to 1.21], P < 0.001, I2 = 91.2%; PCT = 1.00 [95%CI, 0.83 to 1.17], P < 0.001, I2 = 92%; ferritin = 0.78 [95%CI, 0.62 to 0.93], P < 0.001, I2 = 89.9%). There was no difference in ESR between the two groups.

Biochemical Results

Ten biochemical variables, namely creatine kinase (CK), cardiac troponin I (cTnI), myoglobin (MYO), lactate dehydrogenase (LDH), aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin (TBIL), creatinine (CRN), albumin (ALB), and blood urea nitrogen (BUN), were included in this study. All biochemical variables were available for comparisons between non-survivors and survivors of COVID-19. The detailed forest plots are presented in Fig. E6.

  1. 1.

    Severe Versus Non-severe COVID-19

CK, cTnI, MYO, LDH, AST, ALT, TBIL, CRN, and BUN were significantly increased in patients with severe versus those with non-severe COVID-19 (CK = 0.62 [95%CI, 0.41 to 0.82], P < 0.001, I2 = 88.2%; cTnI = 0.54 [95%CI, 0.33 to 0.76], P < 0.001, I2 = 82.9%; MYO = 0.77 [95%CI, 0.51 to 1.06], P < 0.001, I2 = 81.5%; LDH = 1.19 [95%CI, 1.02 to 1.35], P < 0.001, I2 = 88%; AST = 0.71 [95%CI, 0.55 to 0.87], P < 0.001, I2 = 88.8%; ALT = 0.39 [95%CI, 0.31 to 0.47], P < 0.001, I2 = 54.1%; TBIL = 0.36 [95%CI, 0.20 to 0.52],P < 0.001, I2 = 78.9%; CRN = 0.41 [95%CI, 0.29 to 0.52], P < 0.001, I2 = 75.8%; BUN = 0.69 [95%CI, 0.53 to 0.85], P < 0.001, I2 = 80.5%;), whereas ALB was significantly decreased in patients with severe versus non-severe COVID-19 (ALB =  −0.96 [95%CI, −1.08 to −0.83], P < 0.001, I2 = 65.9%).

  1. 2.

    Non-survivors Versus Survivors of COVID-19

Similarly, CK, cTnI, MYO, LDH, AST, ALT, TBIL, CRN, and BUN were significantly increased in non-survivors versus survivors of COVID-19 (CK = 0.79 [95%CI, 0.57 to 1.02], P < 0.001, I2 = 92.3%; cTnI = 1.24 [95%CI, 1.01 to 1.47], P < 0.001, I2 = 92.5%; MYO = 2.67 [95%CI, 1.57 to 3.77], P < 0.001, I2 = 97.3%; LDH = 1.14 [95%CI, 0.94 to 1.34], P < 0.001, I2 = 94.8%; AST = 0.66 [95%CI, 0.53; 0.80], P < 0.001, I2 = 89.9%; ALT = 0.27 [95%CI, 0.10 to 0.43], P = 0.013, I2 = 93.3%; TBIL = 0.47 [95%CI, 0.32 to 0.63], P < 0.001, I2 = 83.7%; CRN = 0.61 [95%CI, 0.37 to 0.84], P < 0.001, I2 = 96%; BUN = 1.07 [95%CI, 0.88 to 1.26], P < 0.001, I2 = 88.1%). In contrast, ALB was significantly decreased in patients with severe versus non-severe COVID-19 (ALB =  −0.86 [95%CI, –1.03 to −0.70], P < 0.001, I2 = 90.9%).

Publication Bias

Funnel plots are shown in Figs. E7 and E8. In severe and non-severe patients of COVID-19, the obvious publication bias was presented in B cell (ab), NK cell (ab), IL-1β, IL-4, IL-6, IL-10, TNF-α, NLR, CRP, D-dimer, and cTnI. In contrast, in non-survivors and survivors of COVID-19, obvious publication bias was present in IL-6, IL-8, IL-10, TNF-α, PLT, HB, CRP, Ferritin, ALT, and ALB. Many factors may have led to the publication bias, such as not enough amounts of originally included studies, different characteristics, and the wide ranges of the parameter results.

Sensitivity Analysis

Results of the sensitivity analysis, using the leave-to-out method, showed that most parameters presented good reliability and stability. However, there were also some parameters showed high sensitivity. Detailed results of each parameter are shown in Figs. E9, E10, E11, E12, and E13.

Investigation of Heterogeneity

A majority of included variables in the current review presented significant heterogeneity (I2 > 50%). The heterogeneity might have come from various factors, such as demographic and clinical characteristics of included patients, time of the symptom onset and laboratory parameters measured, and treatment intervention before the admission. Therefore, we conducted a meta-regression analysis with three potential factors, including the approach of combining disease severity, age, and region, to identify the sources of heterogeneity. The included variables presenting high heterogeneity (I2 > 50%) and reported by an adequate number of studies (n ≥ 10) were applied to the analysis. Regarding the approach of combing disease severity, we identified four subgroups in our severe group according to the originally reported disease severity: severe and critical (severe/critical), severe alone, critical alone, and other. The findings showed that the potential heterogeneity of 16 of 39 variables, including CD3+T(%), B cell(ab), NK(ab), IL-4, IL-6, IL-8, Lym, Eos, HB, NLR, CRP, Ferritin, LDH, ALB, CRN, and BUN, were related to the originally reported disease severity. The detailed results are presented in Table E1. Second, based on the available average age of severely ill patients and non-survivors of COVID-19, we classified the studies into six subgroups (average age ≤ 18 years(y), 30 ~ 49y, 50 ~ 59y, 60 ~ 69y, 70 ~ 79y, and ≥ 80y). In severe patients, the findings showed that the potential heterogeneity of 13 of 38 variables, including CD3+T(%), IL-8, PLT, HB, ESR, Ferritin, APTT, PT, FIB, cTnI, ALB, CRN, and BUN, were related to the different ages of the patients in the included studies. The detailed results are presented in Table E2. Similarly, the potential heterogeneity of 14 of 30 variables in non-survivors, including CD3+T(%), CD4+T(ab), CD8+T(ab), Neu, PLT, CRP, PCT, Ferritin, D-dimer, cTnI, AST, ALT, TBIL, and CRN, were related to the ages of the patients in different included studies. The detailed results are presented in Table E3. Moreover, we divided the four subgroups according to the continents (Asia, Europe, North America and Africa). In severe groups, the findings showed that the potential heterogeneity of 28 of 30 variables, including CD3+T(ab), CD3+T(%), CD4+T(ab), CD8+T(ab), B cell(ab), NK(ab), IL-4, IL-6, IL-8, TNF-α, WBC, Neu, Lym, Eos, HB, NLR, PCT, Ferritin, APTT, PT, FIB, CK, cTnI, LDH, AST, ALT, TBIL, and CRN, were related to the region. The detailed results are presented in Table E4. In non-survivors, the findings showed that the potential heterogeneity of 25 of 27 variables, including CD4+T/CD8+T ratio, IL-6, IL-8, TNF-α, WBC, Neu, Lym, PLT, HB, NLR, CRP, PCT, Ferritin, APTT, D-dimer, FIB, CK, cTnI, LDH, AST, ALT, TBIL, ALB, CRN and BUN, were related to the region. The detailed results are presented in Table E5. We considered the major source of heterogeneity as the regional differences among our included studies, while the approach of combining disease severity and the age of patients partially contributed to the marked heterogeneity observed.

Discussion

In the current updated meta-analysis, our synthetic results of 145 included studies identified a hypercytokinemia profile, including IL-1β, IL-1Ra, IL-2R, IL-4, IL-6, IL-8, IL-10, IL-18, TNF-α, and IFN-γ, which was associated with increased severity and mortality in patients with COVID-19 infection. By contrast, patients with non-severe COVID-19 and survivors exhibited functional innate and adaptive immune responses, presenting by higher levels of eosinophils, lymphocytes, monocytes, B cells, NK cells, T cells and its subset CD4+ T, and CD8+T. Furthermore, in line with an elevated concentration of proinflammatory cytokines, augmented information (indicated by increased WBC, Neu, NLR, PLR, PCT, ESR, CRP, ferritin, or SAA), coagulation dysfunction (indicated by abnormal D-dimer, FIB, APTT and PT) as well as myocardial/liver/renal injury (indicated by elevated CK, cTnI, MYO, LDH, ALT, AST, TBIL, ALB, CRN, and BUN) were the main clinical abnormalities of patients with COVID-19 infection in the severe and fatal cohort.

SARS-CoV-2 infection can initiate a potent immune response, which includes innate immune activation and antiviral immune responses [169, 170]. However, the transition between innate and adaptive immune responses is the core of determining the clinical outcomes and prognosis of COVID-19 infection [171]. Early immune responses against COVID-19 primarily play a protective role in viral clearance, whereas exacerbated and dysregulated immune responses, otherwise known as the “cytokine storm,” can cause tissue damage contributing to poor disease outcomes [172]. An overreactive immune response releases excess pro-inflammatory cytokines and chemokines of which has been well documented [173]. Of these elevated pro-inflammatory cytokines, IL-6 is the most investigated and is a key driver of cytokine dysregulation, which is responsible for the hyper-inflammation in lungs in patients infected with COVID-19 [174]. A recent meta-analysis showed that the anti-IL-6 agent (Tocilizumab) was associated with a lower relative risk of mortality in patients with COVID-19 infection [175]. Other cytokines, such as IL-8 and IL-10, were also proposed to that play a significant role in the inflammatory cascade [176, 177]. We identified an updated abnormal cytokine profile, including IL-1β, IL-1Ra, IL-2R, IL-4, IL-6, IL-8, IL-10, IL-18, TNF-α, and IFN-γ, relating to severe COVID-19 infection and fatality. It is well known that cytokine storm and the subsequent inflammation cascade relay on a complex cytokine network. Our synthesis results offer updated evidence on revealing the structure of cytokine networks related to the poor clinical outcomes, which helps clarify the underlying complex inflammatory pathways, so we can target new treatment agents.

Current management of COVID-19 is supportive, and respiratory failure from acute respiratory distress syndrome (ARDS) is the leading cause of mortality [178]. Hyperinflammation is the prominent feature of patients with ARDS and those non-survivors. Our previous longitudinal study of 548 revealed that patients who died from COVID-19 infection commonly showed an upward trend for neutrophils, IL-6, and C-reactive protein [5]. Other inflammatory parameters, including WBC, PCT, ESR, and SAA, were also proposed as the predictors of fatality. Our synthesis agrees with the findings of previous studies [179,180,181,182,183]. All patients with COVID-19, regardless of the severity, should be screened for hyperinflammation as precaution for potential ARDS once increases in these indicators are detected. Identification of the early signs of ARDS is critical for early intervention (such as low tidal volumes and prone ventilation) to improve oxygenation and lung compliance. Currently, the rates of bacterial/fungal co-infection reported in patients with COVID-19 appear to be low. Timothy et al. included nine studies and found that only 8% (62/806) of cases of bacterial/fungal co-infection were reported [184]. Nevertheless, our data observed that an increased infectious parameter profile detected on admission was strongly associated with poor clinical outcomes, which suggested that prompt antibiotic therapy should be considered after a comprehensive infectious assessment. Additionally, a combined assessment of using abnormal inflammatory parameters and increased cytokine levels might better identify the subgroup of patients for whom immunosuppression could improve mortality. Beneficial anti-inflammatory effects should be weighed against the potentially detrimental effects of inhibiting anti-viral immunity, thereby delaying virus clearance and perpetuating illness [185].

In addition, we observed substantial decreases in B cells, NK cells, T cells, and its subsets, including CD4+T cells and CD8+ T cells in patients with severe disease, compared to those with non-severe disease. We also found that decreased CD3+T, CD4+T, CD8+ T cells, and higher ratio of CD4+ to CD8+ T cells were associated with a fatal outcome. Our findings were in line with the results from a recent meta-analysis targeting lymphocytes and their subset counts [186] as well as observations from clinical practice. However, the underlying mechanism of observed lymphopenia in severe or fatal COVID-19 patients remains unclear. Based on the current evidence, it is proposed that lymphopenia be relating to the following causes: (1) suppression by cytokine mediation; (2) T cells infected by the virus; (3) T cell exhaustion (4) T cell expansion interfered with by the virus; and (5) organ inflation. Furthermore, our data supported that eosinopenia was associated with both severe disease and a fatal outcome. Our previous study suggested that dynamic changes in blood eosinophil counts might predict COVID-19 progression and recovery [5]. However, the pathophysiology for eosinopenia in COVID-19 remains unclear but is likely multifactorial [187], involving (1) reduced expression of adhesion/chemokine/cytokine, (2) direct eosinophil apoptosis, (3) blockade of eosinophilopoiesis, and (4) inhibition of eosinophil egress from the bone marrow. The finding that eosinophil levels improved in patients before discharge might serve as an indicator of improving clinical status.

The presence of the hypercoagulable state in patients with COVID-19 was another marked clinical feature of patients with increased mortality and a more severe form of the disease. The underlying pathophysiology mechanism was also associated with impaired immune responses [188]. SARS-CoV-2 infects host endothelial cells through ACE2 (an integral membrane protein) [189]. Patients with COVID-19 tend to exhibit greater numbers of ACE2-positive endothelial cells [190]. Therefore, vascular endothelial injury is commonly presented in patients with COVID-19. Vascular endothelial injury caused by COVID-19 infection would lead to the formation of microvascular microthrombi, which would trigger active tissue factor expression on macrophages and endothelial cells [191] Elevated tissue factor levels alongside local hypoxia from COVID-19 induced ARDS create a positive thromboinflammatory feedback loop, also known as a cytokine storm [191] The strong interaction between coagulation cascade activation and the cytokine storm might be responsible for the increased incidence of thrombotic events and aggressive inflammatory reactions. Based on our meta-analysis, increased APTT, PT, D-dimer, and FIB were identified as the indicators of coagulation dysfunction contributing to the unfavorable clinical outcomes. Simultaneously increased coagulation parameters and immune index might imply the interplay between overreactive immune responses and coagulation dysfunction which might serve as a more sensitive predicted index of a poor prognosis of COVID-19 infection. Additionally, we also identified several abnormal biochemical parameters representative of myocardial, liver, or renal injury in the severe and non-survivors cohort, such as CK, cTnI, MYO, LDH, ALT, AST, TBIL, ALB, CRN, and BUN. Although the pathophysiological mechanisms underlying myocardial/liver/renal injury by COVID-19 are not well-known so far, innate dysfunction and adaptive immune systems driving the cytokine storm seem to play a role in non-pulmonary organ damage [192,193,194,195,196], particularly those with comorbidities of cardiovascular, liver, and renal diseases.

The purpose of this meta-analysis is two-fold. First, to provide robust evidence of identifying a series of abnormal immunological indicators early to distinguish patients with poor clinical outcomes and to offer valuable information for exploring the underlying mechanism of COVID-19 progression. Second, to draw a picture of the interaction between immune abnormality and other body system dysfunction, including coagulation, inflammation, and non-pulmonary function. However, our meta-analysis has limitations. In line with the heterogeneity that characterized these observational studies [197, 198], a majority of included variables presented large I2 values, indicating significant variations in terms of outcomes observed. Although we attempted to manage this by performing subgroup analysis and meta-regression by disease severity, the age of included patients, and genetic characteristics, the results could not fully explain the source of heterogeneity. We were confined by the methodologies of the studies included, as well as the heterogeneity in characteristics of included patients, such as comorbidities, the therapeutic approach before hospital admission, and the time of symptom onset, which were not provided in the included studies. However, the observed heterogeneity did not impair our main conclusion that severe COVID-19 and mortality were associated with significant abnormalities in the immunological, hematological, coagulation, inflammatory, and biochemical variables. What the heterogeneity suggests is that these abnormalities might show some variation from one country to another, from one city to another, and from one clinical setting to another.

Conclusions

The currently updated meta-analysis primarily identified a hypercytokinemia profile with the severity and mortality of COVID-19, containing IL-1β, IL-1Ra, IL-2R, IL-4, IL-6, IL-8, IL-10, IL-18, TNF-α, and IFN-γ. Impaired innate and adaptive immune responses, reflected by decreased eosinophils, lymphocytes, monocytes, B cells, NK cells, T cells and their subtype CD4+ and CD8+ T cells, and augmented inflammation, coagulation dysfunction, and nonpulmonary organ injury, were marked features of patients with a poor prognosis. Given the strong interplay between immune response dysfunction, aggressive inflammation, coagulation abnormality, and nonpulmonary organ injury, parameters of immune response dysfunction combined with either inflammatory, coagulated, or nonpulmonary organ injury indicators may be more sensitive to predict outcomes in severe patients versus non-survivors.