The trace elements copper (Cu), selenium (Se) and zinc (Zn) are essential dietary micronutrients and exert essential roles in the organism (1–3), including proper immune function (2–6). All three elements are central constituents of regulatory or structural proteins such as selenoproteins, zinc proteins or cuproproteins, affecting a wide range of signaling and biochemical pathways, and cell activities (7, 8).

The dietary intakes of several micronutrients varies across the world, with particularly strong differences known for Cu, Se, Fe, I, and Zn (9–11). Moreover, supply might change with future climatic alterations, causing further risks of deficiency (12). Especially during pregnancy, a supply below the Recommended Dietary Allowance (RDA) is of particular concern as it may negatively affect both mother health and child development (13–15). Additional requirements and strong alterations are observed in disease, particularly in critical illness (16, 17). The fast decline in circulating Se and Zn is proposed to constitute a physiological meaningful reaction that may turn problematic for recovery during longer stays in the intensive care unit and during convalescence (18, 19). Accordingly, the issue of substitution or supplementation in pregnancy and disease as an adjuvant measure against low status and developing deficiency has been intensively discussed (20–23). A general supplementation in the preventive setting without diagnosed deficiency, for reducing for example cardiovascular disease or cancer risk, is hitherto not recommended (24), but a personalized correction of severe deficits appears meaningful in subjects with insufficient supply or in severe disease (25–27).

The coronavirus disease 2019 (COVID-2019) pandemic is challenging the intensive care units around the world with many critically diseased patients and a high death toll. Severe disease is associated with strongly altered serum trace element concentrations, notably low concentrations of Se (28–30) and Zn (29–32), and slightly or strongly elevated serum Cu (32, 33). The observed changes are in line with prior reports on the acute phase response in infection or sepsis, where serum Se is established as a negative (34–36), and serum Cu as a positive acute phase reactant (37, 38). Response of Zn seems more complex as it involves strong and dynamic redistribution between the tissues (39). On a population-wide scale, cure rate and death toll have been reported to correlate with soil Se and population biomarkers of Se status in China, where Se availability varies strongly around the country (40, 41).

Considering that different European populations have diverse dietary patterns and habitual intake levels of essential micronutrients, in particular of Se and Zn, we decided to compare trace element status in six European observational studies in relation to COVID-19 mortality. We hypothesized to observe characteristic alterations in serum trace elements in COVID-19 patients across the countries in comparison to healthy European subjects, suitable to derive a population-independent model and relevant thresholds with predictive value for survival.

A total of 551 hospitalized COVID-19 positive patients were enrolled in six studies across six European countries. Patient characteristics are presented in Table 1 . Median age was similar across the countries, with a median above 60 years. Among 551 patients, a total of 87 deaths were recorded. Ten deaths occurred in Belgium, 14 deaths in France, six deaths in Germany, 14 deaths in Ireland, 10 deaths in Italy and 33 deaths in Poland (Figure 1A).

The Cu, Se, and Zn status of COVID-19 positive patients were compared with 2069 healthy adults participating in the cross-sectional EPIC-study (European Prospective Investigation into Cancer and Nutrition). These analyses were conducted in the same laboratory as the other measurements reported here, using the same method. For the reference range the 2.5^th and 97.5^th centiles of the EPIC-reference were calculated which corresponded to 45.9 µg/L and 132.4 µg/L for Se, 640.9 µg/L and 1469.5 µg/L for Zn, and 894.2 µg/L and 2098.9 µg/L for Cu, respectively.

A principal component analysis (PCA) was conducted to assess potential trace element patterns in patients deceased from COVID-19, surviving patients and the reference group ( Figure 1B ). Se and Zn followed a similar pattern, while Cu exhibited a distinct pattern. In Figure 1C , we assessed the distribution of trace elements in each country and in the reference cohort. When assessing correlations, in line with the PCA analysis, Se and Zn correlated positively in the reference cohort (Spearman’s R = 0.261) ( Figure 1D ). The correlation coefficient of Se and Zn in COVID-19 patients was more than two-fold higher than in the reference group, consistently in each country (overall Spearman’s R = 0.564). The correlation of Se and Cu or Cu and Zn were positive overall, but not consistent across all countries.

2.5^th centiles correspond to green solid lines in Figures 2A–C , and overall distribution of the EPIC-reference is displayed as marginal density of the scatter plots. Overall correlations of the trace elements were displayed in scatter plots in Figures 2A–C . According to the 2.5^th centiles of the reference cohort, 10.7% of the samples were Cu deficient, 25.7% were Se deficient, and 30.9% displayed Zn deficiency ( Figure 2D ). Combined deficiencies were most prominent for Se and Zn, as 32.7% of the deficient samples also presented an underlying combined deficiency. In comparison, combined deficiencies were rare (0.3% for a combined Se and Zn deficiency) in the EPIC cohort ( Figures 2E, F ).

In Figure 3A , serum Se concentrations were compared between samples of surviving and deceased patients in each country. Consistent throughout all six countries, Se concentrations were lower in the samples of deceased patients, and the findings were statistically significant, except for Italy, where a trend was observed, however without reaching statistical significance. Similarly, Zn concentrations were significantly lower in the samples of the non-survivors, except for France and Italy ( Figure 3B ).

Serum Cu was significantly lower in samples of non-survivors in France and Germany, but displayed significantly higher concentrations in samples of deceased patients in Poland ( Figure 4A ). Cu/Zn ratio was consistently elevated in deceased patients across four of the countries, except for Italy and France ( Figure 4B ).

An analysis of the country-specific values highlights relatively consistent differences between the surviving and non-surviving patients for serum Se and Zn, but not for Cu. Serum Se concentrations were 19.1-30.4% lower in samples of non-survivors than in survivors, and serum Zn concentrations were 10.5-24.7% lower in samples of the patients who died as compared to those who survived ( Table 2 ).

A pooled analysis based on adjusted restricted cubic spline regression was conducted to assess the relationship between trace element concentrations and death from COVID-19 ( Figure 5 ). The models were further adjusted for age, sex, and country of the patients included (n=549, two missing due to missingness in age). Se and Zn displayed a significant, inverse, non-linear relationship with death from COVID-19 ( Figures 5A, B ). Low Cu was associated with a favorable outcome ( Figure 5C ), while the ratio of Cu to Zn displayed a positive association with death ( Figure 5D ). In order to assess and compare the potential predictive value of each marker, we conducted receiver operating characteristics analyses. As individual predictors, Se had the highest AUC of 0.754, followed by Zn (0.679), and Cu (0.575) ( Figure 5E ). Combined models considering age and sex displayed an AUC of 0.816 for Se, 0.782 for Zn, and 0.769 for Cu ( Figure 5F ). 164 female and 54 male COVID-19 patients with a median (IQR) age of 44 (33,53) were additionally enrolled as outpatients into the AIIDC study in Ireland. When associating trace element concentrations with hospitalization, considering the outpatients as the control group, the results were similar to those with death as an endpoint ( Supplementary Figure 1 ). After adjusting for age and sex, Se and Zn were inversely, nonlinearly associated with hospitalization ( Supplementary Figures 1A, B ). Cu was not associated with hospitalization, but the ratio of Cu to Zn was linearly and positively associated with hospitalization ( Supplementary Figures 1C, D ). Zn had the highest predictive value (AUC of 0.658), followed by Se (0.625) ( Supplementary Figure 1E ). In this analysis, Cu did not perform better than chance (AUC of 0.499). After including patients’ age and sex, the model including Zn had an AUC of 0.790, model with Se had an AUC of 0.789, and the model with Cu displayed an AUC of 0.777 ( Supplementary Figure 1F ).

In this study, we assessed serum Cu, Se and Zn concentrations in COVID-19 patients from separate observational studies conducted in six European countries. A general deficiency in serum Se and Zn in comparison to a reference range from 2069 healthy European adults was observed in the patients, in particular in the non-survivors. Serum Cu concentrations showed no consistent difference, neither in comparison to the reference nor in relation to survival, whereas the Cu/Zn ratio was particularly high in non-survivors, presumably largely attributable to low Zn status.

A parallel regulation of Se and Zn has been observed in the control subjects, much in agreement with another independent sample (n=3834) of the EPIC cohort analyzed recently by a different technology (54). Notably, correlation analyses of the two different sub-cohorts including almost 6000 participants of EPIC yielded identical coefficients for serum Se and Zn, of 0.26 (54) and 0.261 (this study). The almost perfect accordance supports the robustness of this interaction, as it was found in two large subsamples of EPIC in different laboratories using different analytical techniques (inductively coupled plasma tandem mass spectrometry (ICP-MS/MS) versus TXRF). The correlation for serum Se and Zn showed a substantially higher coefficient in the patient samples as compared to EPIC, highlighting common COVID-19-related pathways apparently regulating these two immune-relevant trace elements in the same direction. However, this notion is deduced from single samples only, taken mainly at hospital admission. Our previous time-resolved analyses highlighted low serum Se as part of the acute phase response, causing further selenoprotein P decline with time especially in the non-survivors (28). In comparison, the decline of Zn appears to be a fast reaction and rather a re-distribution (39), as the concentrations tended to normalize with time both in survivors and non-survivors (31).

A regular metabolism and sufficient supply of Se and Zn are of central importance for a full activity of the immune system. The trace element Zn is needed for the function of hundreds of different proteins in our organism, influences the activity and response of immune cells, directly serves signal transduction, and apparently also alters the composition of the microbiota (55–57). Moreover, Zn status is reduced in the elderly, which is associated with a declining activity of the immune system (58). Although this association implies a protective role of Zn against COVID-19 mortality, the results of observational studies are rather ambiguous and support the quest for targeted intervention studies to answer this hypothesis (59, 60). The analysis of an open-label supplementation study with Zn and Se in patients with severe COVID-19 induced acute respiratory distress syndrome (ARDS) supports the hypothesis of positive health effects on the immune system, but results from sufficiently large and randomized controlled intervention studies are lacking (61). The understanding of the molecular effects of Se deficiency and supplemental Se intake on the immune system is more solid, as the Se status directly affects the expression of selenoproteins relevant to the immune system (62–64). The link extents to multiple levels, including quality control of newly synthesized proteins in the ER to prevent secretion of immunogenic misfolded proteins (65, 66), a direct involvement of selenoproteins in intracellular lymphocyte signaling (67–69), and an effect of selenoproteins, particularly GPX4, on immune cell survival and protection against hyperactivation–induced ferroptosis (70, 71). Given that severe Se deficiency is an established risk factor for autoimmune disease (72–74) and appears to influence mortality risk in critical illness (20, 75, 76), the consistent associations observed in this study support discussion of active supplementation of COVID-19 patients with severe deficiency to achieve reference concentrations (19). The extent to which a combined supplementation of Se and Zn in severe COVID-19 has a positive effect on disease progression, immune system function, and survival remains to be investigated. However, based on what is known from supplementation studies, it is unlikely that in patients with diagnosed deficiency, correcting circulating concentrations towards reaching reference ranges will cause adverse side effects in a given patient.

To the best of our knowledge, this is the first multi-cohort study assessing several trace elements across different countries by standardized methodology in one analytical laboratory. The results from two of the six studies have been reported before (Belgium and Germany), where additional biomarkers were determined, verifying the differences also by longitudinal alterations and in different comorbidities (28, 31, 33, 42). Due to reasons of limited sample volumes from some cohorts, differences in the pre-analytical sample preparations, and in view that our prior studies reported consistent results for additional biomarkers of trace element status, this study focused on a single parameter using one robust technique, i.e., total concentrations determined by TXRF analysis.

When compared to other studies, our observations are in line with the findings from several single centers regarding the relatively low concentrations of serum Se and Zn in COVID-19 and in particular in relation to survival (30, 77–81). Beside individual data level analyses, geographical soil trace element content has also been assessed in relation to COVID-19 mortality, yielding congruent results with our study (40, 41). In the USA however, where soil Se content is higher than in European and many Asian and African countries, an association of the case fatality rate of COVID-19 with geography was observed for Zn only, but not for Se (82). Considering that the majority of the US population is Se replete, this finding is indeed reasonable and argues for a potential causal relationship.

Beside serum concentrations, cumulative population level or geographical data, the dietary intake and genetically predicted trace element status has also been assessed in relation to COVID-19 mortality. In contrast to the analytical findings with serum samples, no significant associations were observed for Se, Cu or Zn in genetic analyses based on Mendelian Randomization (MR) studies (83, 84).

Using the EPIC-cohort as reference, serum Se and Zn was lower in the COVID-19 patients analyzed. The relative difference to EPIC was comparable between the countries, despite the relatively high Zn status in Italy, which likely results from locally preferred dietary patterns and food availability (85). In general, vegetarians and vegans have poor supply of Se and Zn, in agreement with the low concentration in soil and accordingly in local plant products (86–88). A diet rich in fish, seafood and meat provides higher amounts of trace elements, as they tend to accumulate in the sea and are usually added to the feed of chicken, pigs or cattle in the farming process (89, 90). Hence, some of the country-specific differences observed may be attributable to the fraction of vegetarians and the frequency of meat and fish-containing meals, however no dietary intake data have been determined in the current study.

Our study has several strengths and limitations. We collected patient samples in six countries of Europe, establishing a relatively large cohort, and conducting a parallel quantitative analysis blinded to clinical data during the measurements and by the same technology. Hereby, the risk of personal bias and technological errors are minimized. Moreover, we were able to relate the findings to a similar analysis of a large sub cohort of healthy European subjects from the EPIC study (50). The congruent results from the PCA for serum Se and Zn support the notion on a representative database and reliable technology (50, 54). The parallel decline of Se and Zn, notably in non-survivors, supports the clinical relevance of the changes for survival prediction. The findings support the notion on the need for conducting micronutrient supplementation to avoid severely declined status and health-compromising deficiencies, as both elements are known for their essential role in the immune system and for convalescence (91, 92).

However, due to the observational study design, we cannot account for reverse causality, i.e., disease severity may have suppressed serum trace element status, with the changes observed potentially not affecting the course of the disease. Another notable limitation concerns the heterogeneity of the studies. Even though all six studies investigated COVID-19 mortality as a main outcome, time point of sampling, clinical care, potential adjuvant treatments, age and sex range of the patients included in the analyses differed between the study sites. This considerable limitation among others precluded on the one hand a reliable adjustment of the analyses for additional clinical parameters, e.g., comorbidities, medication and other potential confounders, so that we cannot account for residual confounding. Yet, the high consistency of the results on the other hand despite these considerable sources of heterogeneity highlights and underlines the robustness of the major results, providing a high generalizability of our findings.

This study supports the notion on a mortality-relevant decline of serum Se and Zn concentrations in COVID-19, irrespective of area of residency in Europe and country specific dietary patterns. As Europe is generally characterized by relatively low Se supply, the findings should not be extrapolated to areas of higher habitual Se intake, from where similar analyses have not yet been reported. As our observational study design cannot address reverse causality, supplementation recommendations are limited to the consented importance of avoiding severe deficiencies, but are not yet based on clinical experience as appropriately designed intervention trials with Se and/or Zn are missing in COVID-19. In case the strong deficits in nonsurvivors contributed to death risk, a personalized substitution for raising the status into the range of survivors would constitute a straightforward, inexpensive and promising adjuvant treatment option, which merits serious consideration as a testable hypothesis.

The datasets presented in this article are not readily available because availability is based on varying local governmental restrictions for each country. R code for statistical analyses conducted to generate the main findings in this study will be made available upon request. Requests to access the datasets should be directed to lutz.schomburg@charite.de.

The studies involving human participants were reviewed and approved by Local Ethics Committee of JPH Ghent and UZ Gent, Institutional review boards of the Strasbourg University Hospitals, Bavaria Ethik-Kommission der Bayerischen Landesärztekammer, Local Irish institutional review boards, ethics committee of Istituto Auxologico Italiano in Milan, Italy, Bioethical Committee at Poznan University of Medical Sciences. The patients/participants provided their written informed consent to participate in this study.

KD: conceptualization, methodology, software, formal analysis, investigation, data curation, writing - original draft, and visualization; TC: conceptualization, methodology, software, formal analysis, investigation, data curation, writing - original draft, and visualization; TB: conceptualization, resources, and writing - review and editing; IB: conceptualization, resources, and writing - review and editing; IC: conceptualization, resources, and writing - review and editing; GDL: conceptualization, resources, and writing - review and editing; SF-K: conceptualization, resources, and writing - review and editing; LF: conceptualization, resources, and writing – review and editing; AG: conceptualization, resources, and writing - review and editing; RH: writing – review and editing and investigation; DH: conceptualization, resources, and writing - review and editing; LI: conceptualization, resources, and writing - review and editing; GK: writing – review and editing, investigation, and methodology; PK: conceptualization, resources, and writing - review and editing; ZK: conceptualization, resources, and writing - review and editing; AL: conceptualization, resources, and writing - review and editing; PM: conceptualization, resources, and writing - review and editing; AM: conceptualization, resources, and writing – review and editing; LP: conceptualization, resources, and writing – review and editing; MP: conceptualization, resources, and writing - review and editing; MR: conceptualization, resources, and writing - review and editing; MS: conceptualization, resources, and writing – review and editing; LV: conceptualization, resources, and writing - review and editing; LS: conceptualization, validation, resources, writing – original draft, supervision, project administration, and funding acquisition. All authors contributed to the article and approved the submitted version.

This work was supported by the Deutsche Forschungsgemeinschaft (DFG), Research Unit FOR-2558 “TraceAge” (Scho 849/6–2) and CRC/TR 296 “Local control of TH action” (LocoTact, P17), and partially by the Italian Ministry of Health (COVENDO study).

The authors would like to thank Vartitér Seher, Gabriele Boehm and Anja Fischbach for excellent technical support. Jozefien De Clercq, Jana Minne and Laura Van der Meulen for their contribution to data collection and processing, the All-Ireland Infectious Diseases (AIID) Cohort Study Group for support, and Rafał Spachacz and Bożena Barczyk from regional hospital in Słupca for excellent cooperation in acquiring and processing of the samples.

LS holds shares of selenOmed GmbH, a company involved in selenium status assessment.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.