The Associations of Iron Related Biomarkers with Risk, Clinical Severity and Mortality in SARS-CoV-2 Patients: A Meta-Analysis

The coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is spreading rapidly around the world and has led to millions of infections and deaths. Growing evidence indicates that iron metabolism is associated with COVID-19 progression, and iron-related biomarkers have great potential for detecting these diseases. However, the results of previous studies are conflicting, and there is not consistent numerical magnitude relationship between those biomarkers and COVID-19. Thereby, we aimed to integrate the results of current studies and to further explore their relationships through a meta-analysis. We searched peer-reviewed literature in PubMed, Scopus and Web of Science up to 31 May 2022. A random effects model was used for pooling standard mean difference (SMD) and the calculation of the corresponding 95% confidence interval (CI). I2 was used to evaluate heterogeneity among studies. A total of 72 eligible articles were included in the meta-analysis. It was found that the ferritin levels of patients increased with the severity of the disease, whereas their serum iron levels and hemoglobin levels showed opposite trends. In addition, non-survivors had higher ferritin levels (SMD (95%CI): 1.121 (0.854, 1.388); Z = 8.22 p for Z < 0.001; I2 = 95.7%, p for I2 < 0.001), lower serum iron levels (SMD (95%CI): −0.483 (−0.597, −0.368), Z = 8.27, p for Z < 0.001; I2 = 0.9%, p for I2 =0.423) and significantly lower TIBC levels (SMD (95%CI): −0.612 (−0.900, −0.324), Z = 4.16, p for Z < 0.001; I2 = 71%, p for I2 = 0.016) than survivors. This meta-analysis demonstrates that ferritin, serum iron, hemoglobin and total iron banding capacity (TIBC) levels are strongly associated with the risk, severity and mortality of COVID-19, providing strong evidence for their potential in predicting disease occurrence and progression.


Introduction
Coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1], has become a global pandemic, resulting in 525,646,754 confirmed cases and 6,299,346 deaths as of 31 May 2020 [2,3]. The rapid spread of this disease has put enormous pressure on local medical institutions and their finances. Thus, it is very critical to identify and prevent the spread of COVID-19 early.
The clinical manifestations of COVID-19 are various, from asymptomatic infection to death. Therefore, it is very necessary to find a reliable early biomarker to identify the emergence and progression of disease [3,4]. Numerous studies indicated that iron distribution was closely linked to the onset and progression of COVID-19. In the process of viral pathogen invasion, the defense system is very crucial [5], especially the immune system, which relies on the supply of micronutrients. Iron is not only an important component of micronutrients, but also plays an important role in various fundamental biological processes between human and pathogen, ranging from deoxyribonucleic acid (DNA) synthesis to adenosine triphosphate (ATP) generation [6,7]. Furthermore, some symptoms of COVID-19, such as pneumonia, thrombo-embolism and acute respiratory

Ferritin Level and COVID-19
A total of twenty-nine studies assessed the connections between ferritin levels and the mortality of SARS-CoV-2 patients in this meta-analysis, involving 2131 non-survivors and 7813 survivors. The ferritin levels were significantly higher in the dead patients than that in survivors (SMD (95%CI): 1.121 (0.854, 1.388); Z = 8.22 p for Z < 0.001; I 2 = 95.7%, p for I 2 < 0.001; Table 2

Hepcidin Level and COVID-19
A total of three studies on the relationships between hepcidin and the mortality of COVID-19 were included in this meta-analysis, including 66 non-survivors and 224 survivors. Dead patients had slightly higher hepcidin levels than recovered patients, but this was not significant (SMD (95%CI): 0.447 (−0.287, 1.182); Z = 1.190 p for Z = 0.232; I 2 = 84.8%, p for I 2 =0.001; Table 2, Figure 8). Figure 8. Forest plot of standard mean difference (SMD) with corresponding 95% confidence intervals (CIs) of studies [10,11,13] on hepcidin levels in non-survivors and survivors. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis. The Figure 7. Forest plot of standard mean difference (SMD) with corresponding 95% confidence intervals (CIs) of studies [10,[17][18][19]24,36,37] on hemoglobin levels in COVID-19 cases and controls. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis.
The center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI.

Hepcidin Level and COVID-19
A total of three studies on the relationships between hepcidin and the mortality of COVID-19 were included in this meta-analysis, including 66 non-survivors and 224 survivors. Dead patients had slightly higher hepcidin levels than recovered patients, but this was not significant (SMD (95%CI): 0.447 (−0.287, 1.182); Z = 1.190 p for Z = 0.232; I 2 = 84.8%, p for I 2 =0.001; Table 2, Figure 8). Figure 7. Forest plot of standard mean difference (SMD) with corresponding 95% confidence intervals (CIs) of studies [10,[17][18][19]24,36,37] on hemoglobin levels in COVID-19 cases and controls. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the metaanalysis. The center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI.

Hepcidin Level and COVID-19
A total of three studies on the relationships between hepcidin and the mortality of COVID-19 were included in this meta-analysis, including 66 non-survivors and 224 survivors. Dead patients had slightly higher hepcidin levels than recovered patients, but this was not significant (SMD (95%CI): 0.447 (−0.287, 1.182); Z = 1.190 p for Z = 0.232; I 2 = 84.8%, p for I 2 =0.001; Table 2, Figure 8).  [10,11,13] on hepcidin levels in non-survivors and survivors. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis. The The analysis of COVID-19 risk included four studies totally, with 177 cases and 261 controls. Though higher hepcidin level was found in cases than that in controls, the difference was not significant (SMD (95%CI): 0.750 (−0.805, 2.306); Z = 0.95 p for Z = 0.345; I 2 = 96.40%, p for I 2 < 0.001; Table 3, Figure 9). center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI.
The analysis of COVID-19 risk included four studies totally, with 177 cases and 261 controls. Though higher hepcidin level was found in cases than that in controls, the difference was not significant (SMD (95%CI): 0.750 (−0.805, 2.306); Z = 0.95 p for Z = 0.345; I 2 = 96.40%, p for I 2 < 0.001; Table 3, Figure 9). Figure 9. Forest plot of standard mean difference (SMD) with corresponding 95% confidence intervals (CIs) of studies [10,18,20,24] on hepcidin levels in COVID-19 cases and controls. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis.
The center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI.
The analysis of COVID-19 risk included four studies totally, with 177 cases and 261 controls. Though higher hepcidin level was found in cases than that in controls, the difference was not significant (SMD (95%CI): 0.750 (−0.805, 2.306); Z = 0.95 p for Z = 0.345; I 2 = 96.40%, p for I 2 < 0.001; Table 3, Figure 9). Figure 9. Forest plot of standard mean difference (SMD) with corresponding 95% confidence intervals (CIs) of studies [10,18,20,24] on hepcidin levels in COVID-19 cases and controls. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis.
The center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI.

TIBC, TSAT and the Mortality of COVID-19
Four studies about TIBC were included in the analyses of mortality, including 382 non-survivors and 719 survivors. Analyses about TSAT involved three studies, 281 nonsurvivors and 458 survivors totally. The TIBC level was significantly lower in the death group than in the survivors group (SMD (95%CI): −0.612 (−0.900, −0.324), Z = 4.16, p for Z < 0.001; I 2 = 71%, p for I 2 = 0.016; Table 2, Figure 10). In addition, no significant differences were found in all analyses of TSAT (SMD (95%CI): −0.112 (−0.455, 0.231), Z = 0.64, p for Z = 0.521; I 2 = 59.6%, p for I 2 = 0.084; Table 2, Figure 11).  and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis. The center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI. Figure 11. Forest plot of standard mean difference (SMD) with corresponding 95% confidence intervals (CIs) of studies [8,10,13] on TSAT levels in non-survivors and survivors. The solid diamond and horizontal line represent the study-specific effect and 95%CI, respectively; the size of the grey square is positively correlated with the weight distributed to each study in the meta-analysis. The center of open diamond with the vertical dashed line expresses the pooled SMD, and the width expresses the pooled 95%CI.

Sources of Heterogeneity and Publication Bias
Strong evidence of heterogeneity among studies was documented for the relationships between these iron-related biomarkers and mortality, clinical severity or risk in SARS-CoV-2 patients. The heterogeneity for the between-study was explored through the univariate meta-regression with the covariates of continent, gender, study types, years, ages and numbers of samples in the analysis. Meta-regression indicated that 75.02% of the heterogeneity in the COVID-19 risk analysis in serum iron was explained by continent. In terms of ferritin, publication year and sample size of the death group explained 18.33% and 14.78% of the heterogeneity of the mortality analysis, respectively; study type contributed 21.04% of the heterogeneity of overall analysis; and publication year contributed 20.33% of the heterogeneity of the severe-mild comparison. In addition, publication year and continent explained 46.95% and 46.09% of the heterogeneity of the moderate-mild analysis in ferritin, respectively. Regarding the heterogeneity of hemoglobin and mortality analysis, publication year accounted for 69.79%.

Sources of Heterogeneity and Publication Bias
Strong evidence of heterogeneity among studies was documented for the relationships between these iron-related biomarkers and mortality, clinical severity or risk in SARS-CoV-2 patients. The heterogeneity for the between-study was explored through the univariate meta-regression with the covariates of continent, gender, study types, years, ages and numbers of samples in the analysis. Meta-regression indicated that 75.02% of the heterogeneity in the COVID-19 risk analysis in serum iron was explained by continent. In terms of ferritin, publication year and sample size of the death group explained 18.33% and 14.78% of the heterogeneity of the mortality analysis, respectively; study type contributed 21.04% of the heterogeneity of overall analysis; and publication year contributed 20.33% of the heterogeneity of the severe-mild comparison. In addition, publication year and continent explained 46.95% and 46.09% of the heterogeneity of the moderate-mild analysis in ferritin, respectively. Regarding the heterogeneity of hemoglobin and mortality analysis, publication year accounted for 69.79%.
The results of sensitivity analysis indicated that no individual study had an excessive influence on the pooled measure for all comparisons (Figures S3-S5). Egger's test reported no publication bias in mortality analysis (p = 0.930) or risk analysis (p = 0.129) of serum iron. However, the risk (p = 0.009) and mortality analyses (p < 0.001) of ferritin had significant publication bias. In terms of hemoglobin, Egger's test demonstrated no publication bias in the analyses of mortality (p = 0.700) and risk (p = 0.483). Publication bias in mortality analyses of TIBC (p = 0.077) and TSAT (p = 0.358) was also not found. Moreover, no publication bias was found in the mortality (p = 0.065) and risk analyses (p = 0.699) of hepcidin.     Table 1 and the number orders in the plot were same to the reference list.     [10,[17][18][19][20][21]24,36,37,[42][43][44]72] about risk for ferritin to the heterogeneity. This analysis was based on the relevant data listed in Table 1 and the number orders in the plot were same to the reference list.

Discussion
Our meta-analysis was based on 72 articles, containing 148 studies, of which 85 were about ferritin, 22 were about serum iron, 27 were about hemoglobin, 7 were about hepcidin, 3 were about TSAT and 4 were about TIBC. Mortality analyses included 29 studies on ferritin, 8 on serum iron, 6 on hemoglobin, 3 on hepcidin, 4 on TIBC and 3 on TSAT. Our results indicated that the serum iron and TIBC levels of the deceased were significantly lower than those of the survivors, but higher ferritin levels were found in the deceased. In terms of hemoglobin, hepcidin and TSAT, we did not find an association with death outcome. The risk analyses included 17 articles on ferritin, 7 on serum iron, 7 on hemoglobin and 4 on hepcidin. It was shown that cases had lower serum iron and hemoglobin levels, but higher ferritin levels than controls. There were not significant results in hepcidin analyses. Regarding the analyses of clinical severity, 39 studies on ferritin, 14 on hemoglobin and 7 on serum iron were included. All of them were found to be related to the severity of COVID-19.
According to previous studies, excessive inflammation is a characteristic of COVID-19 [82]. Iron plays an essential role in this process [9]. Serum iron, ferritin, hemoglobin, hepcidin, TIBC and TSAT represent the iron levels in the body; however, some of them have other important physiological functions.
Hepcidin is a main regulatory factor of iron metabolism that is associated with body iron level [83,84]. However, we did not observe significant links between that and COVID-19 onset or progression because of the limited numbers of studies. Differently from us, Denggao Peng et al. [85] classified the groups based on the clinical findings and reported that the hepcidin levels of severe COVID-19 cases were higher than those of non-severe cases. It is common that hepcidin is upregulated after a viral infection, especially for the COVID-19 patients with inflammation [86]. Moreover, thanks to this special change, hepcidin binds to ferroprotein and accelerates its degradation, so that iron uptake decreases and the iron storage in macrophages increases [87], influencing SARS-COV-2.
Ferritin is also recognized as an acute phase reactant of inflammation, influenced by the presence of iron, hepcidin [46,50,[88][89][90] and pro-inflammatory cytokines. Inflammation promotes ferritin synthesis and release in the liver [91,92]. Moreover, ferritin also is the storage form of iron in macrophages, explaining the decrease in serum iron [93][94][95]. Henry et al. [90] found the associations between ferritin and COVID-19 severity. Similarly, we further documented the differences between their associations with COVID-19 risk and mortality. Though serum iron is an important indicator of disease, it cannot accurately represent the iron level because of the various related forms in the human body. Serum iron is essential for both humans and viruses [96][97][98]. In order to deprive SARS-CoV-2 of iron and support immunity, macrophages intake more iron, and the intestinal tract absorbs less, leading to a decrease in serum iron [81,82]. In addition, a study by Ehsani [99] reported a structural similarity between the hepcidin protein and the spiked glycoprotein cytoplasmic tail of SARS-CoV-2. This indicates that SARS-CoV-2 can simulate hepcidin's action, contributing to the decreased serum iron [100]. A decreased hemoglobin level is usually a symbol of anemia, caused by decreased serum iron. TSAT reflects serum iron availability and is frequently used in clinical practice to detect states of iron deficiency or iron overload [101,102].
In meta-analysis, between-study heterogeneity is common. Thus, exploring the sources of between-study heterogeneity is essential. We performed univariate meta-regression, with covariables such as continent, study type, publication year and sample size. The regression results explained part of the heterogeneity in our meta-analysis, but there was still some heterogeneity not being detected. In addition, the Galbraith analysis indicated that high heterogeneity in ferritin resulted from twelve mortality analysis studies and five risk analysis studies. After excluding these studies, low heterogeneity and robust results without small-study effect were documented. However, the final results of ferritin were not changed.
There were some strengths in our meta-analysis: First of all, as far as we know, our study has unified a large number of studies on the associations of iron-related biomarkers with risk, clinical severity and mortality in COVID-19 patients, avoiding inaccurate conclusions of individual studies. More importantly, the random effects were used to estimate the pooled SMD. Thus, it was still possible to draw convincing results though the inconsistent measurement conditions and units for iron-related biomarkers in different studies.
However, our study has several limitations. Firstly, high heterogeneity was found in almost all indicator analyses, but some of analytical results were not explained by metaregression or subgroup analysis. We cannot get a more accurate evaluation for the sources of the heterogeneity due to the lack of corresponding study-level covariates in the reported articles. Moreover, the differences in iron-related biomarkers' reference values between females and males illustrated the importance of gender for our study. However, we could not conduct meta-analysis including gender for the reason that there was not enough information about it.

Conclusions
Our meta-analysis showed that the levels of serum iron and TIBC in dead patients were significantly lower than in survivors, and the ferritin level was higher in death groups than in survivors, whereas the relationship between hemoglobin and mortality was not significant. Moreover, serum iron and hemoglobin levels were lower in cases and negatively correlated with the severity; on the contrary, ferritin level was higher in cases. In addition, no statistically significant results were found in the hepcidin and TSAT levels of the severity and mortality groups. That was possibly due to the limited number of studies.
In conclusion, we found that ferritin, serum iron, hemoglobin and TIBC levels are closely associated with the risk, severity or mortality of COVID-19. These results provide strong evidence for the applications of iron-related biomarkers in the prediction of the COVID-19 occurrence and development. Moreover, lower serum iron and hemoglobin levels could provide clues for explaining the deteriorated process of COVID-19. However, future studies are needed to further confirm these results in future research.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu14163406/s1, Table S1: Quality assessment of studies included in the meta-analysis, Figure S1: Forest plot of standard mean difference (SMD) with corresponding 95% confidence interval (CI) of studies on iron-related biomarkers levels about severity meta-analysis, Figure S2: Forest plot of standard mean difference (SMD) with corresponding 95% confidence interval (CI) of studies on subgroup analysis by publication year, continent and study type, Figure S3: The influence analysis results of iron-related biomarkers levels about mortality meta-analysis, Figure S4: The influence analysis results of iron-related biomarkers levels about risk meta-analysis, Figure S5: The influence analysis results of iron-related biomarkers levels about severity meta-analysis.  Data Availability Statement: The data used in this review come from published articles, all of which are identified in the references. The data used in the meta-analysis have been provided in the tables.

Conflicts of Interest:
The authors declare no conflict of interest.