2. Association between Selenium Level and Chronic Liver Diseases
First, we included all the individuals to calculate the average selenium status in patients and healthy controls using random-effects models. We found an adverse association between body selenium status and chronic liver diseases (SMD = −1.70, 95% CI: −2.30 to −1.11, n = 137), but the heterogeneity was high (I2 = 99.5%, p < 0.001). Due to discrepancies in the literature, we performed subgroup analysis based on the severity of chronic liver diseases. We found that fatty liver disease patients had an equivalent level of selenium to healthy controls (SMD = 1.06, 95% CI: −1.78 to 3.89, n = 11), whereas patients with hepatitis (SMD = −1.78, 95% CI: −2.22 to −1.34, n = 44), liver cirrhosis (SMD = −2.06, 95% CI: −2.48 to −1.63, n = 57), and liver cancer (SMD = −2.71, 95% CI: −3.31 to −2.11, n = 25) had a lower selenium level than healthy controls, regardless of the baseline of selenium level in the body. The funnel plot of hepatitis, liver cirrhosis, and liver cancer articles all appeared symmetric, and egger’s test detected no publication bias. Although the funnel plot of fatty liver diseases showed bias, no statistically significant publication bias was found by Egger’s test. There was no presence of publication bias for the studies.
Additionally, we performed a further subgroup analysis according to the types of diseases, sample sources, different regions, years of publication, and the reference blood selenium levels (optimal or suboptimal).
For fatty liver diseases, stratified analyses by types of diseases showed that, compared with the controls, alcoholic fatty liver disease patients (SMD = −1.29, 95% CI: −2.08 to −0.50, n = 3) had significantly lower selenium levels, while simple fatty liver disease patients had no difference (SMD = −0.51, 95% CI: −0.90 to −0.12, n = 4) and NAFLD patients had a higher level (SMD = 4.39, 95% CI: −0.55 to 9.34, n = 4), although not significantly. Stratified analyses by study regions showed that fatty liver patients had a significantly lower selenium level in Europe than controls, but no difference in the USA and a higher level in Asia. Stratified analyses by sample sources showed that fatty liver disease patients had a significantly lower hair selenium level than controls, but no difference in whole blood, serum, and plasma. Stratified analyses by study design showed that fatty liver disease patients had a significantly lower selenium level than controls in case-control studies, but no difference in cross-sectional studies. Stratified analyses by year of publication showed that fatty liver disease patients had a significantly lower selenium level than controls in studies carried out before 1990, but no difference in studies carried out after 1991. Subgroup analysis stratified by the mean baseline blood selenium level in the normal control group showed that, interestingly, when in an optimal selenium level (70–150 µg/L), fatty liver disease patients (3 in alcohol liver disease and 1 in NAFLD) had a significantly lower blood selenium level than controls (SMD = −0.92, 95% CI: −1.87 to 0.02, n = 4), but in a suboptimal selenium level (>150 µg/L), there was no significant difference between NAFLD patients and controls (SMD = 5.84, 95% CI: −3.29 to 14.97, n = 3).
For hepatitis, stratified analyses by types of diseases showed that, compared with controls, viral hepatitis patients (SMD = −1.88, 95% CI: −2.42 to −1.35, n = 34) and alcoholic hepatitis patients (SMD = −1.36, 95% CI: −1.94 to −0.77, n = 9) had a significantly lower selenium level. Stratified analyses by study regions showed that hepatitis patients had a significantly lower selenium level in Europe and Asia than controls, while no difference was found in Africa. Further subgroup analysis stratified by sample sources showed that hepatitis patients had a significantly lower selenium level than controls in each sample. Subgroup analysis stratified by study design showed that hepatitis patients had a significantly lower selenium level than controls in each type of study. Subgroup analysis stratified by year of publication showed that hepatitis patients had a significantly lower selenium level than controls in most studies, except for in studies carried out from 1991 to 2000. Subgroup analysis stratified by the mean baseline blood selenium level in the normal control group showed that whether in an optimal (70–150 µg/L (SMD = −1.22, 95% CI: −1.62 to −0.82, n = 24)) or suboptimal (<70 µg/L (SMD = −2.38, 95% CI: −3.57 to −1.19, n = 4) or >150 µg/L (SMD = −2.65, 95% CI: −4.04 to −1.26, n = 12)) selenium level, hepatitis patients had a significantly lower blood selenium level than controls.
For liver cirrhosis, stratified analyses by types of diseases showed that alcoholic cirrhosis patients (SMD = −2.45, 95% CI: −2.99 to −1.90, n = 21) and other cirrhosis patients (SMD = −2.41, 95% CI: −2.94 to −1.89, n = 27) had a significantly lower selenium level than controls, but no difference in primary liver cirrhosis (SMD = 0.90, 95% CI: −1.98 to 3.79, n = 9). Stratified analyses by study regions showed that liver cirrhosis patients had a significantly lower selenium level in Europe, the USA, and Asia than controls. Stratified analyses by sample sources showed that liver cirrhosis patients had a significantly lower selenium level than controls in most samples, except for the liver. Subgroup analysis stratified by study design showed that liver cirrhosis patients had a significantly lower selenium level than controls in each type of study. Stratified analyses by year of publication showed that liver cirrhosis patients had a significantly lower selenium level than controls in most studies, except for in studies carried out before 1990. Subgroup analysis stratified by the mean baseline blood selenium level in the normal control group showed that whether in an optimal (70–150 µg/L) or suboptimal (<70 µg/L or >150 µg/L) selenium level, liver cirrhosis patients had a significantly lower blood selenium level than controls.
For liver cancer, almost all studies of liver cancer have only included patients with hepatocellular carcinoma, so we did not perform subgroup analysis by the types of liver cancer. Stratified analyses by study regions showed that liver cancer patients had a significantly lower selenium level in Europe, the USA, and Asia than controls. Subgroup analysis stratified by sample sources showed that liver cancer patients had a significantly lower selenium level than controls in each sample. Subgroup analysis stratified by study design showed that liver cancer patients had a significantly lower selenium level than controls in each type of study. Subgroup analysis stratified by year of publication showed that liver cancer patients had a significantly lower selenium level than controls in each publication year. Subgroup analysis stratified by the mean baseline blood selenium level in the normal control group showed that whether in an optimal (70–150 µg/L) or suboptimal (<70 µg/L or >150 µg/L) selenium level, liver cancer patients had a significantly lower blood selenium level than controls.