Association of COVID-19 mortality with serum selenium, zinc and copper: Six observational studies across Europe

Abstract

Introduction: Certain trace elements are essential for life and affect immune system function, and their intake varies by region and population. Alterations in serum Se, Zn and Cu have been associated with COVID-19 mortality risk. We tested the hypothesis that a disease-specific decline occurs and correlates with mortality risk in different countries in Europe.

Methods: Serum samples from 551 COVID-19 patients (including 87 non-survivors) who had participated in observational studies in Europe (Belgium, France, Germany, Ireland, Italy, and Poland) were analyzed for trace elements by total reflection X-ray fluorescence. A subset (n=2069) of the European EPIC study served as reference. Analyses were performed blinded to clinical data in one analytical laboratory.

Results: Median levels of Se and Zn were lower than in EPIC, except for Zn in Italy. Non-survivors consistently had lower Se and Zn concentrations than survivors and displayed an elevated Cu/Zn ratio. Restricted cubic spline regression models revealed an inverse nonlinear association between Se or Zn and death, and a positive association between Cu/Zn ratio and death. With respect to patient age and sex, Se showed the highest predictive value for death (AUC=0.816), compared with Zn (0.782) or Cu (0.769).

Discussion: The data support the potential relevance of a decrease in serum Se and Zn for survival in COVID-19 across Europe. The observational study design cannot account for residual confounding and reverse causation, but supports the need for intervention trials in COVID-19 patients with severe Se and Zn deficiency to test the potential benefit of correcting their deficits for survival and convalescence.

Keywords: SARS-CoV-2; biomarker; mortality; nutrition; trace elements.

Figure 1

Trace element patterns and correlations. (A) Map displaying the countries contributing to this study. (B) Biplot of principal component analysis displaying patterns of trace elements, and Cu/Zn ratio in relation to clusters of alive, deceased and healthy subjects. (C) Ridgeline plot displaying the distribution of each trace element in each country as well as the healthy EPIC reference. (D) Spearman’s R for correlation between trace elements in each country. Solid circular green lines represent the correlations in the reference group, i.e. a representative sample of the EPIC-Study.

Figure 2

Trace element status in relation to each other and EPIC-reference cohort. (A–C) Scatter plots displaying the correlation between trace elements in COVID-19 positive patients. Yellow points display deceased patients, grey points display alive patients. Point shapes differ according to each country. For each trace element, the distribution of the healthy EPIC-cohort (n=2069) was plotted as marginal density plots. Green lines correspond to the 2.5^th lowest centile in the EPIC-cohort. Black line is computed using linear regression, gray shadows correspond to 95% confidence intervals. Correlation coefficient (R) was computed with the Spearman’s R. (D) Deficiencies in trace elements (below 2.5^th centile in EPIC) are shown. Venn diagram shows overlapping/combined deficiencies. (E) Deficiencies in trace elements are shown in the EPIC cohort. (F) Proportions of deficiency in each trace element as well as a combined Se and Zn deficiency are displayed.

Figure 3

Selenium and Zn status in samples of deceased and alive patients. (A) Serum Se concentrations were lower in samples of non-survivors than in survivors across all countries. (B) Serum Zn concentrations were also relatively low in the non-survivors across all countries. The green line denotes the median value as determined in EPIC, and broken lines indicate the 2.5th and 97.5th centiles. Results from Wilcoxon-Rank-sum testing are indicated as p-values.

Figure 4

Cu and Cu/Zn status in samples of deceased and alive patients. (A) Serum Cu concentrations showed inconsistent differences between samples of deceased and alive patients across all countries. (B) The serum Cu/Zn ratios was relatively higher in non-survivors than in survivors in all the countries with France as the exception. The green line denotes the median value as determined in EPIC, and broken lines indicate the 2.5^th and 97.5^th centiles. Results from Wilcoxon-Rank-sum testing are indicated as p-values.

Figure 5

Adjusted restricted cubic spline regression in a pool of all patients from all countries. (A–D) Restricted cubic spline regression with three knots at 10^th, 50^th and 90^th centiles were conducted to assess the relationship between trace elements and odds ratio (OR) of death. All analyses were adjusted for age, sex, and country of the samples. P for nonlinearity was calculated by comparing nested linear regression models to the restricted cubic splines by likelihood ratio X² test. (E) Se, Zn and Cu were evaluated as individual predictors in relation to death using the restricted cubic spline model. Areas under the curves (AUC) are plotted for each predictor. (F) Models were further adjusted for age and sex.