The American Journal of Clinical Nutrition 117 (2023) 93 110

Original Research Article

Tommaso Filippini 1,2, Susan Fairweather-Tait3,*, Marco Vinceti 1,4

1. 1 Environmental, Genetic and Nutritional Epidemiology Research Center (CREAGEN), Section of Public Health, Department of Biomedical, Metabolic and Neural Sciences, University of Modena and Reggio Emilia, Modena, Italy; 2 School of Public Health, University of California Berkeley, Berkeley, CA, USA; 3 Norwich Medical School, University of East Anglia, Norwich, UK; 4 Department of Epidemiology, Boston University School of Public Health, Boston, MA, USA

A B S T R A C T

Background: Selenium is an essential trace element with both beneficial and detrimental effects on health depending on dose and chemical form. Currently, there is debate on recommendations for selenium supplementation as a public health measure to improve immune function and reduce infectious disease susceptibility.

Objectives: We performed a systematic review and meta-analysis of experimental studies assessing the effect of selenium supplementation on immunityrelated outcomes in healthy people.

Methods: We undertook a search of published and unpublished studies in literature databases such as PubMed/MEDLINE, Embase, and clinicaltrials.gov up to 17 October, 2022, and performed a meta-analysis comparing the effects on immunity-related outcomes between Se-supplemented versus control arms. Whenever possible we assessed the nonlinear relation using a dose–response approach.

Results: 9 trials were included, 5 in North America, and 4 in Europe, with a duration between 8 and 48 weeks and supplementation of both inorganic and organic selenium forms. Selenium supplementation did not substantially affect immunoglobulin or white blood cell concentrations, and the dose response meta-analysis indicated that an increase in plasma selenium concentrations above 100 μg/L did not further increase IgA levels nor T cells. An inverted Ushaped relation emerged for NK cell count, with a lower number of these cells both below and above 120 μg/L. The only beneficial effect of selenium supplementation was the increased activity for NK lysis, but the available data did not permit dose response analysis. Cytokine levels were substantially unaffected by selenium supplementation.

Conclusions: Although some of the data suggested beneficial effects of selenium supplementation on immune function, the overall picture appears to be inconsistent and heterogeneous due to differences in trial duration and interventions, plus evidence of null and even detrimental effects. Overall, the evidence that we extracted from the literature in this systematic review does not support the need to supplement selenium beyond the recommended dietary intake to obtain beneficial effects on immune function. This trial was registered at PROSPERO (CRD42022312280).

Keywords: selenium, immune function, infectious disease, experimental studies, systematic review, dose response meta-analysis

Introduction

Selenium is an essential trace element with a complex and intriguing relationship with human health, demonstrating both beneficial and detrimental effects, depending on the dose and the chemical species [1]. Selenocysteine is recognized as the 21st amino acid [2] and selenium is an essential component of 25 selenoproteins [3], which participate in a wide variety of physiological processes, including the

regulation of antioxidant response to reactive oxygen species and other physiological properties [1,2,4]. Levels of such functional biomarkers were used by various national and international bodies to suggest dietary recommendations for selenium intake [5 8], that is, both average requirements and dietary reference intakes, ranging from 20 to 75 μg/d depending on the type and amount of selenium-induced proteomic response chosen [1]. Some of these selenoproteins are enzymes that could also be involved in immune function, such as glutathione

Abbreviations: GPX, glutathione peroxidase; KD, Keshan disease; PICOS, Population, Intervention, Comparison, Outcome, Study design; SMD, standardized mean difference.

* Correspondence author. E-mail address: [email protected] (S. Fairweather-Tait).

https://doi.org/10.1016/j.ajcnut.2022.11.007 Received 2 July 2022; Received in revised form 20 October 2022; Accepted 4 November 2022 Available online 19 December 2022 0002-9165/© 2022 American Society for Nutrition. Published by Elsevier Inc. All rights reserved.

T. Filippini et al. The American Journal of Clinical Nutrition 117 (2023) 93 110

Figure 1. Flow chart for study identification in online databases and clinicaltrials.gov registry.

peroxidases, thioredoxin reductases, iodothyronine deiodinases, methionine-R-sulfoxide reductase B1, and selenophosphate synthetase

1. 2 [9]. For this reason and after some laboratory studies [10], selenium deficiency has been suggested as affecting defense against infectious diseases.

In particular, the association between selenium deficiency and adverse health outcomes in humans was originally proposed with the identification of Keshan disease (KD). This disorder is characterized by a severe cardiomyopathy, and it was first recorded in 1935 when it was mainly found in parts of the country where the common feature was low selenium concentrations in soils and foods locally produced [11]. However, some epidemiologic hallmarks of KD could not be explained solely on the basis of selenium deficiency. In particular, seasonal fluctuations in KD incidence suggested involvement of an infectious agent [1,11]. Coxsackievirus was, in fact, detected in the myocardium of KD patients [12], and studies in mice exposed to Coxsackievirus showed that host nutritional deficiency led to viral genome mutations, which rendered benign viruses highly virulent [13]. In addition, other animal and in vitro studies indicated that selenium is able to inhibit the viral replication of Coxsackievirus [14, 15]. These studies illustrate the complexity of selenium interactions in the body and also indicate that specific host nutritional status can alter

viral genotype. In relation to this, it should be noted that other trace elements and vitamins may be implicated in the etiology of KD in relation to both nutritional status and viral infection [16,17], as well as genetic factors such as genetic polymorphisms, including of GPX genes [11].

Overall, interest in the relationship between selenium and the immune system/function has increased over the past years [18,19]. Results from cell and animal models have demonstrated that humoral (adaptive) immunity, such as activation and functions of T and B cells, is affected by the level of selenium exposure; cell-mediated (innate) immunity, including inflammatory signaling capacity and antipathogen activities of macrophages, is also influenced by selenium [9]. However, there are conflicting reports from human trials designed to demonstrate the benefits of selenium supplementation to boost immunity against bacterial and viral pathogens. Given the current interest in the role of nutrition in the immune system [20], we attempted to estimate the intake of selenium that is associated with optimal immune function. We undertook a systematic review of selenium and infectious disease susceptibility, focusing on data extracted from the studies providing the highest level of evidence, namely, randomized controlled trials of selenium supplementation and measures of immune function, performing a dose–response meta-analysis whenever possible.

94

Table 1 Characteristics of included studies

Reference Country Population1 R Blinding Duration2 Intervention Groups Plasma Se levels (μg/ L)

Outcomes Results

Specific response: Ig levels þ plaque forming þ lymphokine synthesis proliferation after mitogen stimulation Specific response: phagocytosis þ intracellular killing þ chemiotaxis

Arvilommi, 1983 (36)

Finland 40 healthy men aged 36 50 y, with no history of cardiovascular, pulmonary, and psychiatric diseases, with low-Se status (<70 μg/L)

20 C 20 T (10þ10)

Mean SD Baseline3: C: 70 9 T: 70 10 End of trial: C: 74 9 T: 169 19

Yes DB 11 wk Sep 1981 to Dec 1981

Se yeast (200 μg/d) þ Se-rich “zwiebackstyle” toast (made with Serich wheat flour)

1. 95

Broome, 2004

(27)

UK, Liverpool 66 (M/F: 33/33) healthy nonsmoking subjects aged 20 47 y not taking medications

Yes4 DB 15 wk (1999 2000)

Sodium selenite (50 or 100 μg/d)

22 C 22 T50 22 T100

Mean Baseline: C: 78.96

Specific response: cellmediated immune response to vaccination with T cells

levels (μg/ L)

with low-Se status (<94.75 μg/ L)

T50: 78.17 T100: 82.12 End of trial: C: 83.5 T50: 91.7 T100: 103.3 (from figure)

and poliovirus antibody production and detection rate Specific response: cytokine levels

1. 96

Hawkes, 2001

(37)

United States, California

11 healthy men aged 26 45 y with weight for height lower than 125% of ideal and without chronic diseases or medication use

Yes DB 99 d Low-Se (13 μg/ d) and high-Se diet (297 μg/d) due to origin of rice and beef staples

6 C 5 T

Mean SD Baseline: Low-Se: 117.65 7.9 High-Se: 105.81 18.95 End of trial: Low-Se: 72.4 9.5 High-Se: 187 23

Specific response: Ig levels þ WBC levels (all and subpopulation) mitogen response þ secondary response (with influenza A and B, and diphtheria) DHT skin response Specific response: Complement components þ NK cell number and activity

1. levels (μg/ L)
2. 97

levels (μg/ L)

1. 98

Hawkes, 2009

(38)

United States, California

42 healthy nonsmoking men aged 18 45 y with selfreported absence of diseases, clinically normal blood count and blood chemistry, and without obesity

Yes DB 48 wk Low-Se vs. High-Se (300 μg/d selenized yeast)

20 C 22 T

Mean SD Baseline: Low-Se: 146 19 High-Se: 142 19 End of trial: Low-Se: 141 18

Specific response: WBC levelsþ DHT Specific response: NK cell number

levels (μg/ L)

High-Se: 228 63

1. 99

Ivory, 2017

(28)

United Kingdom, Norfolk

119 (M/F: 54/65) healthy subjects aged 50 64 y with normal hematology, blood chemistry, blood pressure levels and BMI >18.5 and <35 with low-Se status (Se <110 μg/L)

Yes DB 12 wk Group SeY:

selenized yeast (Se methionine 50, 100, or 200 μg/d)

Group SeO: Serich onions with 50 Se μg/d

20 C

1. 20 T50
2. 21 T100 23 T200

1. 17 CO
2. 18 TO

Mean SD5 Baseline SeY

1. C: 92.0 (11.9) T50: 92.2 13.3 T100: 98.6
2. C: 93.7 16.5 T50: 118.3

Specific response: Proliferating T cells after flu vaccination þcytolytic granules Specific response: NK cells þ cytokines levels

levels (μg/ L)

13.1 T100: 152.0 24.3 T200: 177.4 26.3 Baseline SeO:

1. CO: 93.3 11.5 TO: 97.6 11.5 Week 10 SeO:
2. CO: 94.2 15.0 TO: 106.0

11.9

1. 100

KiremidjianSchumacher, 1994 (23)

United States, New York University Dental Center

32 (M/F: 27/5) healthy subjects aged 24 36 y divided into 2 subtrials:

CL Group: 21 (M/F: 16/5) subjects assessing cytotoxic lymphocytes

NK group: 11 men, assessing NK cells

Yes6 NI 8 wk Sodium selenite

(200 μg/d)

1. 10 C-CL
2. 11 T-CL

1. 5 C-NK
2. 6 T-NK

Mean SE CL group: Baseline

1. C: 133.5
2. C: 133.6

6.2 T: 138.5 5.11 NK group: Baseline C: 122.0 4.0 T: 120.0

Specific response: lymphocytes activity Specific response: NK cell activity

levels (μg/ L)

7.0 End of trial C: 122.0 10.0 T: 114.0 4.0

Peretz, 1991

(25)

Belgium, Brussels

1. 101

22 (M/F: 7/15) elderly institutionalized ( 1 y) subjects aged >65 y with no history of severe disabling diseases (e.g., cancer, cirrhosis, or diabetes)

NI DB 6 mo Selenized yeast

(100 μg/d)

11 C 11 T

Mean SD Baseline: C: 69.5 19.7 T: 66.3 9.5 End of trial: C: 75.01 19.74 T: 130.3 34.7

Specific response: lymphocyte proliferation after mitogen exposure

Roy, 1994 (24) New York University Dental Center

22 (M/F: 12/10) healthy people age 24 36 y

Yes6 NI 8 wk Sodium selenite

(200 μg/d)

11 C 11 T

Wood, 2000

(26)

United States, Southern Arizona

21 (M/F: 13/9) healthy nonsmoking people aged 57 84 y with no history of chronic diseases (e.g.,

Yes SB 6 mo Selenized yeast

(400 μg/d)

8 C 8 T

Mean SE Baseline: C: 128.6 5.5 T: 130.1 7.3 End of trial: C: 143.4 7.4 T: 152.7 6.8

Mean SD Baseline: C: 155

Number of IL-2 receptor sites in peripheral mononuclear cells

Specific response: Total T cells and subgroups Specific response:

Table 1 (continued)

Reference Country Population1 R Blinding Duration2 Intervention Groups Plasma Se levels (μg/ L)

cancer, cardiovascular diseases, or diabetes). Analyzed 16 (8C þ 8T due to 22% attrition rate)

6.0 T: 129 4.8 End of trial C: 153.8 T: 141.3

Outcomes Results

Total WBC and NK cells and activity

1. 102

BMI: body mass index measured as weight (in kg)/height (in m2); C, placebo/control group; ConA, concanavalin A; DHT, delayed-type hypersensitivity; Ig, immunoglobulins; PHA, phytohemagglutinin; PWM, pokeweed mitogen; OKT3, monoclonal anti-human T lymphocyte antibody; T, Se-treated group;

1. 1 Male/female ratio reported whenever possible;
2. 2 Duration in: days (d), weeks (wk) or months (mo);
3. 3 Data reported in Levander et al., 1983 (57);
4. 4 From personal communication with authors;
5. 5 Data reported in previous report Hurst et al., 2010 (58);
6. 6 Included gender, race, age, body weight, height, dietary habits, and history of vitamin intake, tobacco, and alcohol in the randomization process.

Methods

After registering the protocol in PROSPERO (registration no. CRD42022312280), we carried out a systematic review according to the PRISMA guidelines [21].

Study identification and selection

We conducted a search of online literature databases (PubMed/ MEDLINE and EMBASE) and clinicaltrials.gov up to 17 October, 2022, for experimental human studies assessing the association between selenium status and infectious disease susceptibility. We defined the Population, Intervention, Comparison, Outcome, Study design (PICOS) statement as “In healthy adults, what is the susceptibility to infectious diseases in relation to selenium status when assessed in experimental studies investigating the effects of selenium supplementation?.” We used as search keywords terms related to “humans,” “selenium,” or “selenium supplementation,” “infectious disease,” “immune system” or “immunity,” and “trial” or “clinical trial.” Details of database searches are reported in Supplemental Table 1. Inclusion criteria were as follows: being an epidemiologic study in healthy humans, selenium being the only difference between experimental and control groups, and the outcome related to infectious disease susceptibility or immune system. We excluded nonexperimental studies, case reports, reviews, and commentaries.

Risk of bias assessment

We assessed the risk of bias of the included studies using the RoB 2.0 tool of the Cochrane Collaboration [22] using the subsequent 5 domains: 1) bias arising from the randomization process; 2) bias due to deviations from the intended interventions investigating the effect of

assignment to intervention (i.e., intention-to-treat analysis); 3) missing outcome data; 4) bias in measurement of the outcome; and 5) bias in selection of the reported results. For each domain the judgment can be low risk, some concerns, or high risk. Overall, we judged a study at “low risk” of bias if it had low risk in all domains, whereas we judged it as having “some concerns” when at least 1 domain was classified as such, but none was at high risk. A study was considered at “high risk” of bias if there was high risk in at least 1 domain.

Data extraction

We extracted the following data from included reports: first author name, publication year, country, type, and duration of the trial; study population and characteristics; trial design and characteristics (randomization, blinding); dose and duration of selenium supplements, selenium concentrations before and after the intervention, and the difference between intervention and control group at the end of the intervention; and outcomes of interests related to infectious disease and immune function. When data were mentioned but not explicitly reported or only partially available, for example, only in figures and not in a tabular way, we sought to contact study investigators to retrieve the raw data. Whenever possible we extracted mean or median values along with SD, SE, or IQR. When data were available only from figures, we sought to extract mean levels along with SD or SE from figures. About data extraction, we systematically tried to contact study authors to request data when those available in the publication were not enough to include the study in the dose–response meta-analysis. However, authors could not be reached for 2 studies because they were no longer working [23,24]. For 1 study, we did not get any answer despite the availability of email addresses from recent articles [25] and for another study, the corresponding author confirmed that original data

Figure 2. Forest plot of Hedge’s g SMDs for Ig levels, all studies. The area of each gray square is proportional to the inverse of the variance of the estimated SMD, and horizontal lines represent the 95% CI. Black diamonds represent point estimates of overall SMD for each group. The solid vertical line represents null effect. SMD ¼ 0. N, number of participants; Se/C, selenium levels (μg/L) in control group; Se/T, selenium levels (μg/L) in treatment group; SMD, standardized mean difference.

were no longer available [26]. Finally, for 1 study, information was available about study design (e.g., randomization, blinding), but raw data were no longer available [27], but we were able to retrieve original data from a later study for use in the meta-analysis [28]. To perform quantitative analysis, in 5 out of 9 studies, we sought to use data reported in figures [23–27]. For this meta-analysis, we extracted findings for a specific endpoint when at least 2 studies for that endpoint were available. When 3 or more studies for an endpoint were available, we also extracted data about plasma selenium concentrations at the end of the intervention period [20,25,27].

Data analysis

We performed a meta-analysis comparing the higher versus the lowest category, that is, treated versus the control group. We did this through computation of Hedge's g standardized mean differences (SMD) along with their 95% CI, due to heterogeneity in units of measurements of outcomes when at least 2 studies were available for each specific outcome. In addition, whenever possible, we performed dose–response meta-analysis of SMD between selenium levels and parameters of immune function. We used the 1-stage methodology [29, 30], an approach for implementation of dose response meta-analysis based on a weighted mixed effects model and using cubic splines, which enables the pooling of results from all studies when at least 2 levels of exposure are available, as implemented in other fields [31,32]. Having no specific parametric assumption about the shape of the association, we used restricted cubic splines with 3 knots at fixed percentiles (10th, 50th, and 90th) to investigate such association, taking into account both difference in plasma selenium concentrations between the treatment and control groups and final plasma selenium concentrations at the end of the intervention. For each spline, as a reference dose, we used the median value of the set of studies alternatively considered in each specific analysis. All analyses were carried out using “meta” and “drmeta” routines of Stata statistical software (Stata 17.0-SE 2021, StataCorp LLC). For all data analyses we did not use null hypothesis testing and P value cutpoints, following the American Statistician Association guidelines [33] and recent literature in the field [34,35].

Results

Figure 1 shows the flow chart for study retrieval and selection. After the removal of duplicates, we identified 331 records in online databases, and we further excluded 258 records based on title/abstract screening. After full-text evaluation, we eventually included 9 studies in the final analysis [23–28, 36–38], 1 of which was retrieved through citation chasing [37]. Reasons for exclusion after full-text evaluation are reported in detail in Supplemental Table 2.

Overall, the characteristics of the 9 included studies are reported in Table 1. Five studies were carried out in North America/United States [23,24,26,37,38], 2 in the United Kingdom [27,28], 1 in Belgium [25], and 1 in Finland [36]. Overall, they included 370 participants (220 in selenium-treatment groups and 150 in the control groups) with ages ranging from 18 to 64 years in all but 2 studies; 1 was undertaken in elderly institutionalized subjects aged 65 years [25] and 1 in subjects aged 57–84 years [26]. Three studies recruited only male participants [36 38), whereas other 6 studies recruited both male and female participants, although none reporting gender-stratified analysis. The duration of the trials ranged from 8 to 48 weeks. Selenium supplementation occurred in the inorganic form using sodium selenite in 3 trials [23,24,27]. The intervention doses varied with 50 and 100 μg/d

used in 1 trial [27] and 200 μg/d in the remaining 2 trials [23,24]. Conversely, organic selenium was administered in the form of selenized yeast in 5 trials [25,26,28,36,38]. Doses varied from 50 up to 400 μg/d, with some trials having 2 or more intervention groups at increasing doses of selenium. In addition, 2 trials also used Se-rich foods to increase selenium intake in the intervention groups. Specifically, 1 trial administered wheat toast made with Se-rich flour [36] and other Se-enriched onions [28]. Finally, Se-low and Se-high diets were administered to study participants in 1 trial in the control and intervention groups directly providing 3 daily meals made with foods, namely, rice and beef staples, from different geographic origins with very low or very high soil selenium [37].

All studies measured plasma concentrations of selenium before and at the end of the intervention. Baseline plasma concentrations were generally similar in the control and intervention groups, ranging

Figure 3. Dose–response meta-analysis of changes of IgA levels according to plasma selenium difference (A and C) and final levels (B and D) of plasma selenium (μg/L) between selenium-supplemented and control groups at the end of the trials, all studies, N ¼ 3 [28,36,37]. Solid black line represents the effect with variation of SMD (y-axis) according to the plasma selenium levels (x-axis). The curves are designed using restricted cubic spline method using 3 knots at fixed cutpoints (10th, 50th and 90th percentiles) and considering the median value (50th) of such distribution as reference point. The gray area represents 95% CI. The short-dashed line represents the null effect, SMD ¼ 0. SMD, standardized mean difference.

between 70 and 118 μg/L (mean, 110 μg/L) and between 66 and 142 μg/L (mean, 103 μg/L), respectively. Selenium concentrations at the end of the interventions were systematically raised (mean in the selenium-supplemented group, 144 μg/L; range, 92–228 μg/L) but not in the placebo group (mean 108 μg/L; range, 72–153 μg/L). However, in 2 trials, plasma concentrations of selenium did not increase in the intervention groups and were substantially similar to those of the

control groups [23,24]. Finally, in 1 trial [26], selenium supplementation did increase the baseline selenium concentrations (from 129 to 142 μg/L), whereas the control group showed constantly higher concentrations at both baseline and the end of the trial (155 vs. 154 μg/L).

Results of the risk of bias assessment are reported in Supplemental Table 3. Most of the included trials were judged at “low risk” of bias. One study was judged as having “some concerns” due to the single-

1. Figure 4. Forest plot of Hedge’s g SMDs for lymphocyte levels, all studies. The area of each gray square is proportional to the inverse of the variance of the estimated SMD, and horizontal lines represent the 95% CI. Black diamonds represent point estimates of overall SMD for each group. The solid vertical line represents null effect, SMD ¼ 0. N, number of participants; Se/C, selenium levels (μg/L) in control group; Se/T, selenium levels (μg/L) in treatment group; SMD, standardized mean difference.

blind design, although no deviations from the intended intervention were detected [26]. Two additional trials [23,24] were classified as “some concerns” in the randomization process due to the lack of reporting of detailed characteristics of recruited subjects at baseline, hampering the evaluation for this item. In addition, these 2 trials were judged as having a “high risk” of bias due to deviations from the intended interventions because no information about blinding was reported. Furthermore, participants had substantially comparable plasma selenium concentrations in the intervention and control arms at the end of the trial, thus raising questions about the reliability of the study findings. For these reasons, they were judged as having an overall “high risk” of bias.

In Figure 2, the meta-analysis of studies assessing Ig levels showed small-to-null increase for all Ig types due to selenium supplementation (IgA: SMD ¼ 0.13; 95% CI: 0.16, 0.42; IgG: SMD ¼ 0.14; 95% CI:

0.97, 1.25; IgM: SMD ¼ 0.09; 95% CI: 0.50, 0.67). All studies were at low risk of bias and used organic selenium forms, that is, selenized yeast, through diet or foods rich in selenium. In Figure 3 we

reported the dose–response meta-analysis (implemented for IgA levels only). When looking at the plasma selenium difference between treatment and control arms, selenium increase above the median value (40 μg/L) seems to be associated with higher IgA levels. When looking at final plasma selenium concentrations at the end of the trials, such increases can be detected up to 110 μg/L, although further increases in selenium are not associated with any change in IgA levels.

1. Figure 4 shows the SMD for lymphocyte overall levels and sub-
2. Figure 5 presents the dose–response analysis for Tcells and T CD8þ

cells for which a sufficient number of studies were available. When considering both difference and final selenium concentrations, an

1. Figure 5. Dose–response meta-analysis of changes in T cells and T-cytotoxic CD8þ levels according to plasma selenium difference (A and C) and final levels (B and D) of plasma selenium (μg/L) between selenium-supplemented and control groups at the end of the trials, all studies, N ¼ 5 in (A and B) [26–28,37,38], N ¼ 4 in (C and D) [27,28,37,38]. Solid black line represents the effect with variation of SMD (y-axis) according to the plasma selenium levels (x-axis). The curves are designed using restricted cubic spline method using 3 knots at fixed cutpoints (10th, 50th, and 90th percentiles) and considering the median value (50th) of such distribution as reference point. The gray area represents 95% CI. The short-dashed line represents the null effect, SMD 0.

1. Figure 6. Forest plot of Hedge’s g SMDs for NK cells and lysis, all studies. The area of each gray square is proportional to the inverse of the variance of the estimated SMD, and horizontal lines represent the 95% CI. Black diamonds represent point estimates of overall SMD for each group. The solid vertical line represents null effect, SMD ¼ 0. N, number of participants; Se/C, selenium levels (μg/L) in control group; Se/T, selenium levels (μg/L) in treatment group; SMD, standardized mean difference.

increase can benoted until approximately 10 and110μg/L, respectively, above which a plateau is reached, and there is a null increase in total and T CD8þ cells. We observed comparable results when considering studies at low risk of bias only (Supplemental Figure 2).

Figure 6 presents the SMD for NK cell levels and activity, with inverse-to-null effects on NK cells (SMD ¼ 0.14, 95% CI: 0.49, 0.21) but increases in NK lysis (SMD ¼ 0.48, 95% CI: 0.13, 0.82). Results are comparable when considering studies at low risk of bias

1. (Supplemental Figure 3). In the dose–response analysis for NK cells, increases in plasma selenium seem to be associated with an increase in NK cells until 140 μg/L but with further increase an inverse relationship can be noted with a decrease in NK cells (Figure 7). A similar pattern is suggested when looking at differences in selenium concentrations, especially when studies at low risk of bias are considered
2. (Supplemental Figure 4). Supplemental Figure 5 presents SMD for expression or IL-2 re-

ceptors reported in 2 studies indicating no effect of selenium supplementation (either in the form of organic or inorganic selenium), with the same results after exclusion of 1 study at high risk of bias (Supplemental Figure 6). Supplemental Figure 7 shows the effect of selenium supplementation on cytokine levels, namely, IL-10 and IFNgamma from 2 studies, both at low risk of bias, suggesting a slight though imprecise increase (IL-10: SMD ¼ 0.07, 95% CI: 0.18, 0.33; IFN-gamma: SMD ¼ 0.20, 95% CI: 0.05, 0.45). Supplemental Figure 8 shows results of lymphocyte proliferation in either absence or presence of external stimuli (e.g., different mitogens or vaccination) as assessed through total lymphocyte proliferation in all but 1 study assessing T-cell proliferation [28]. All studies are at low risk of bias. Overall, lymphocyte proliferation seems to be unaffected by selenium supplementation. For the analysis including IL-2 receptor, cytokine levels, and lymphocyte proliferation, we could not perform dose–response meta-analysis due to the limited number of studies.

Discussion

This review, to the best of our knowledge, the first to meta-analyze human trials on selenium supplementation and immunologic endpoints using a dose response approach, provides little evidence of the beneficial effects of selenium supplementation on immune function as assessed through the evaluation of levels of immunoglobulins, and white blood cells, particularly lymphocytes and NK cells. In particular, the dose–response meta-analysis indicates that an increase in plasma selenium concentration above 100 μg/L does not further increase IgA levels nor Tcells, as seen at lower levels, suggesting the occurrence of a possible beneficial effect that corresponds to an intake of 70 μg/d, namely, the Dietary Reference Value (DRV) for selenium [5], using a conversion factor of 1.5 as previously suggested [39]. The pattern of association becomes more complex regarding NK cell count, for which an inverted U-shaped relation emerged, with lower numbers of these cells both below and above 120 μg/L (80 μg/d). Such a U-shaped pattern is not unusual for a nutrient, considering that both levels of too-low and too-high exposure may induce adverse effects. The only beneficial effect of selenium emerged for NK lysis, but data did not allow us to test any dose–response relation or the possible presence of a U-shaped curve. The dose–response curve indicated that increased selenium intake at the lowest range of the selenium intake tested in the trials was beneficial, but that achieving intakes above the DRVof ~70 μg/d, as implied by the final blood selenium concentrations of 110 μg/L, does not yield any further beneficial effect. It should also be noted that a recent observational study comparing the effects of IgG against SARS-CoV-2 in subjects with different dietary intakes of selenium or habitual selenium supplementation found no effect of the selenium status on this parameter, which is consistent with the results of the trials [40].

1. Figure 7. Dose–response meta-analysis of changes in NK cell levels according to difference (A) and final (B) levels of plasma selenium (μg/mL) between selenium-supplemented and control groups at the end of the trials, all studies, N ¼ 4 [26,28,37,38]. Solid black line represents the effect with variation of SMD (y-axis) according to the plasma selenium levels (x-axis). The curves are designed using restricted cubic spline method using 3 knots at fixed cutpoints (10th, 50th, and 90th percentiles) and considering the median value (50th) of such distribution as reference point. The gray area represents 95% CI. The short-dashed line represents the null effect, SMD ¼ 0. SMD, standardized mean difference.

Our findings also do not demonstrate that selenium supplementation influences cytokine levels such as IL-10, IFN-gamma, nor IL-2 receptor, contrary to earlier expectations and the working hypothesis driving the trials [18]. Furthermore, results relating to lymphocyte proliferation due to mitogen exposure indicated that selenium supplementation induced, if anything, adverse effects depending on the selenium dose and the mitogen used, although such a relationship could not be investigated through a dose–response approach because the number of relevant studies was low. Previous animal and in vitro studies suggested that selenium may have immunomodulatory effects, including lymphocyte proliferation, antibody concentrations, and cytokine expression and reactivity [2,18], as well as regulation of selenoprotein expression in T cells [41]. However, evidence in humans is controversial, and limited evidence has been provided by human studies [42,43].

It should be noted that the selenium status in the general population appreciably varies across countries. For example, the US selenium exposure tends to be higher than that in other Western countries, particularly European ones. According to the NHANES surveys, serum selenium levels in the US population are generally 130–140 μg/L [44–46], corresponding to a selenium intake of ~86–93 μg/d [1, 39]. Conversely, selenium levels in the European populations are somewhat lower, ranging from 50 to 120 μg/L [5,47,48]. For these reasons and according to our findings, the intake of selenium in these populations can be considered adequate, with no need to increase selenium through supplementation concerning improving immune function. About the presence of a U-shaped pattern with immune function, such an occurrence is not entirely unexpected for selenium, as more generally for micronutrients and other nutrients, for which adverse effects at both low and high exposure levels may occur [49]. Specifically for selenium, a narrow safe range of intake has been already suggested for other health outcomes such as type 2 diabetes. For the latter disease, an excess risk has been identified in human experimental studies where 200 μg/d of selenium supplementation was used in the intervention arms [50], and for increases of blood selenium concentrations approximately above 90 μg/L in nonexperimental studies [32]. Overall, our findings seem to confirm that selenium exposure can be considered beneficial for the immune system until a plasma selenium concentration of ~100 μg/L, whereas higher levels may be associated with null or adverse effects.

Recently, interest has been raised about the possible relationship between selenium intake and COVID-19, under the hypothesis that an impaired selenium status could favor SARS-CoV-2 infection and spread and COVID-19 severity [51]. However, experimental evidence from human trials is lacking as no trials have been performed so far [52], and some evidence from 2 recent Mendelian randomization studies investigating this issue was unable to confirm this possible relationship [53,54]. Similarly, a lack of association was noted for another disease of high public health relevance, HIV infection: despite some evidence from observational studies of a higher risk of infection in subjects with low selenium status [43], findings from experimental studies suggested that increasing selenium intake might delay CD4þ cell decline but not induce viral suppression [55,56].

Our review has limitations that must be acknowledged. First, the low number of trials investigating some of the outcomes hampered the implementation of dose–response meta-analysis by restricting the range of exposure suitable for analysis and decreasing the statistical precision of the estimates, and for many endpoints (e.g., IgG levels, NK lysis, lymphocyte proliferation) even precluded such analysis. For some of the included studies, we extracted numerical data from figures to perform quantitative analysis whenever possible, thus possibly inducing some additional amount of imprecision of the individual estimates. Second, some of the studies we retrieved and included in the analysis were affected by methodological flaws and potential severe sources of bias, including lack of randomization, blinding, and compliance with the intervention, thus considerably reducing the reliability of the results. Concerning selection bias, most of the studies excluded participants with chronic diseases, especially cancer, cardiovascular diseases, or diabetes, and some also included cutoff levels of BMI to exclude subjects with obesity. However, we cannot entirely rule out that some subjects with metabolic disorders could still have been recruited in such studies, somehow limiting the generalization of our findings to the general healthy population. However, all but 2 trials included subjects younger than 65 y, and those including older

participants found consistent results with other studies, thus strengthening the reliability of our analysis.

The strength of this review is that it is the first that systematically addressed the experimental effects of selenium supplementation in humans concerning immunologic endpoints using whenever possible a dose–response approach to assess the level of selenium exposure that could be associated with beneficial or adverse effects on the immune system.

Overall, the assessment and the meta-analysis of the experimental human studies investigating the immunologic effects of selenium administration yielded limited evidence of the beneficial effects of this intervention, and indicated that such beneficial effects were present only in subjects with a low selenium status and occurred only up to intakes of ~70 μg/d, whereas higher intakes were associated with null or even adverse effects.

We are grateful to Jack Dainty, Malcolm Jackson, and Claudio Nicoletti for their valuable suggestions and providing data for the metaanalysis. The authors’ responsibilities were as follows—SF-T: conceived the study with MV and TF; TF: extracted data with the supervision of SF-T and MV, performed the statistical analysis, interpreted the data with SF-T and MV, and prepared the first draft manuscript with substantial contribution of other authors; and all authors: retrieved and selected relevant articles and read and approved the final manuscript. The authors report no conflicts of interest.

Data Availability

Data described in the manuscript, code book, and analytic code will be made available upon request pending an application and the approval of the corresponding author.

Funding

The authors reported no funding received for this study.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https ://doi.org/10.1016/j.ajcnut.2022.11.007.

References

1. [1] M. Vinceti, T. Filippini, L.A. Wise, Environmental selenium and human health: an update, Curr Environ Health Rep 5 (4) (2018) 464 485.
2. [2] S.J. Fairweather-Tait, Y. Bao, M.R. Broadley, R. Collings, D. Ford, J.E. Hesketh, et al., Selenium in human health and disease, Antioxid Redox Signal 14 (7) (2011) 1337–1383.
3. [3] V.M. Labunskyy, D.L. Hatfield, V.N. Gladyshev, Selenoproteins: molecular pathways and physiological roles, Physiol Rev 94 (3) (2014) 739–777.
4. [4] M. Vinceti, T. Filippini, E. Jablonska, Y. Saito, L.A. Wise, Safety of selenium exposure and limitations of selenoprotein maximization: molecular and epidemiologic perspectives, Environ Res 211 (2022), 113092.
5. [5] EFSA Panel on Dietetic Products Nutrition and Allergies (NDA), Scientific opinion on dietary reference values for selenium, EFSA J 12 (10) (2014) 3846.
6. [6] Institute of Medicine, Dietary reference intakes for vitamin C, vitamin E, selenium, and carotenoids, National Academies Press, Washington (DC), 2000.
7. [7] World Health Organization, Vitamin and mineral requirements in human nutrition [Internet], 2nd edn., World Health Organization, Geneva, Switzerland,

2004. Date accessed), https://apps.who.int/iris/handle/10665/42716. (Accessed 17 October 2022).

1. [8] A.P. Kipp, D. Strohm, R. Brigelius-Flohe, L. Schomburg, A. Bechthold, E. Leschik-Bonnet, et al., German Nutrition Society. Revised reference values for selenium intake, J Trace Elem Med Biol 32 (2015) 195 199.
2. [9] J.C. Avery, P.R. Hoffmann, Selenium, selenoproteins, and immunity, Nutrients 10 (9) (2018) 1203.
3. [10] M.A. Beck, Selenium and vitamin E status: impact on viral pathogenicity, J Nutr 137 (5) (2007) 1338 1340.
4. [11] C. Yan, R. Luo, F. Li, M. Liu, J. Li, W. Hua, et al., The epidemiological status, environmental and genetic factors in the etiology of Keshan disease, Cardiovasc Endocrinol Metab 10 (1) (2021) 14–21.
5. [12] L.Q. Ren, X.J. Li, G.S. Li, Z.T. Zhao, B. Sun, F. Sun, Coxsackievirus B3 infection and its mutation in Keshan disease, World J Gastroenterol 10 (22)
6. [13] M.A. Beck, P.C. Kolbeck, Q. Shi, L.H. Rohr, V.C. Morris, O.A. Levander, Increased virulence of a human enterovirus (Coxsackievirus B3) in seleniumdeficient mice, J Infect Dis 170 (2) (1994) 351 357.
7. [14] C. Cermelli, M. Vinceti, E. Scaltriti, E. Bazzani, F. Beretti, G. Vivoli, et al., Selenite inhibition of Coxsackie virus B5 replication: implications on the etiology of Keshan disease, J Trace Elem Med Biol 16 (1) (2002) 41 46.
8. [15] Y. Molin, P. Frisk, N.G. Ilback, Viral RNA kinetics is associated with changes in trace elements in target organs of Coxsackie virus B3 infection, Microbes Infect 11 (4) (2009) 493–499.
9. [16] O.A. Levander, M.A. Beck, Interacting nutritional and infectious etiologies of Keshan disease. Insights from coxsackie virus B-induced myocarditis in mice deficient in selenium or vitamin E, Biol Trace Elem Res 56 (1)
10. [17] A.D. Smith, S. Botero, O.A. Levander, Copper deficiency increases the virulence of amyocarditic and myocarditic strains of coxsackievirus B3 in mice, J Nutr 138 (5) (2008) 849–855.
11. [18] H. Gill, G. Walker, Selenium, immune function and resistance to viral infections, Nutr Diet 65 (Suppl. 3) (2008) S41–S47.
12. [19] Z. Huang, A.H. Rose, P.R. Hoffmann, The role of selenium in inflammation and immunity: from molecular mechanisms to therapeutic opportunities, Antioxid Redox Signal 16 (7) (2012) 705 743.
13. [20] Scientific Advisory Committee on Nutrition, SACN rapid scoping exercise: nutrition, immunity and COVID-19 [Internet], Office for Health Improvement and Disparities, UK, 2020. Date accessed), https://www.gov.uk/government/ publications/sacn-rapid-scoping-exercise-nutrition-immunity-and-covid-19. (Accessed 17 October 2022).
14. [21] M.J. Page, J.E. McKenzie, P.M. Bossuyt, I. Boutron, T.C. Hoffmann, C.D. Mulrow, et al., The PRISMA 2020 statement: an updated guideline for reporting systematic reviews, J Clin Epidemiol 134 (2021) 178–189.
15. [22] J.A.C. Sterne, J. Savovic, M.J. Page, R.G. Elbers, N.S. Blencowe, I. Boutron, et al., RoB 2: a revised tool for assessing risk of bias in randomised trials, BMJ 366 (2019) l4898.
16. [23] L. Kiremidjian-Schumacher, M. Roy, H.I. Wishe, M.W. Cohen, G. Stotzky, Supplementation with selenium and human immune cell functions. II. Effect on cytotoxic lymphocytes and natural killer cells, Biol Trace Elem Res 41 (1 2)
17. [24] M. Roy, L. Kiremidjian-Schumacher, H.I. Wishe, M.W. Cohen, G. Stotzky, Supplementation with selenium and human immune cell functions. I. Effect on lymphocyte proliferation and interleukin 2 receptor expression, Biol Trace Elem Res 41 (1–2) (1994) 103–114.
18. [25] A. Peretz, J. Neve, J. Desmedt, J. Duchateau, M. Dramaix, J.P. Famaey, Lymphocyte response is enhanced by supplementation of elderly subjects with selenium-enriched yeast, Am J Clin Nutr 53 (5) (1991) 1323–1328.
19. [26] S.M. Wood, C. Beckham, A. Yosioka, H. Darban, R.R. Watson, β-Carotene and selenium supplementation enhances immune response in aged humans, Integr Med 2 (2–3) (2000) 85–92.
20. [27] C.S. Broome, F. McArdle, J.A. Kyle, F. Andrews, N.M. Lowe, C.A. Hart, et al., An increase in selenium intake improves immune function and poliovirus handling in adults with marginal selenium status, Am J Clin Nutr 80 (1) (2004) 154 162.
21. [28] K. Ivory, E. Prieto, C. Spinks, C.N. Armah, A.J. Goldson, J.R. Dainty, et al., Selenium supplementation has beneficial and detrimental effects on immunity to influenza vaccine in older adults, Clin Nutr 36 (2) (2017) 407 415.
22. [29] A. Crippa, A. Discacciati, M. Bottai, D. Spiegelman, N. Orsini, One-stage dose-response meta-analysis for aggregated data, Stat Methods Med Res 28 (5)
23. [30] N. Orsini, S.C. Larsson, G. Salanti, Systematic reviews in health research, meta-analysis in context, 3rd edn., John Wiley & Sons Ltd, Oxford, UK, 2022, pp. 258–269. Chapter 14, Dose–response meta-analysis.
24. [31] T. Filippini, M. Malavolti, P.K. Whelton, A. Naska, N. Orsini, M. Vinceti, Blood pressure effects of sodium reduction: dose-response meta-analysis of experimental studies, Circulation 143 (16) (2021) 1542 1567.

1. [32] M.Vinceti, T.Filippini, L.A.Wise,K.J. Rothman, Asystematic review and doseresponse meta-analysis of exposure to environmental selenium and the risk of type 2 diabetes in nonexperimental studies, Environ Res 197 (2021), 111210.
2. [33] R.L. Wasserstein, N.A. Lazar, The ASA statement on p-values: context, process, and purpose, Am Statist 70 (2) (2016) 129 133.
3. [34] N. Berselli, T. Filippini, G. Adani, M. Vinceti, Toxicological risk assessment and multi-system health impacts from exposure, Academic Press, London, UK, 2021, pp. 309–321. Chapter 27, Dismissing the use of P-values and statistical significance testing in scientific research: new methodological perspectives in toxicology and risk assessment.
4. [35] S. Greenland, S.J. Senn, K.J. Rothman, J.B. Carlin, C. Poole, S.N. Goodman, et al., Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations, Eur J Epidemiol 31 (4) (2016) 337–350.
5. [36] H. Arvilommi, K. Poikonen, I. Jokinen, O. Muukkonen, L. R€as€anen, J. Foreman, et al., Selenium and immune functions in humans, Infect Immun 41
6. [37] W.C. Hawkes, D.S. Kelley, P.C. Taylor, The effects of dietary selenium on the immune system in healthy men, Biol Trace Elem Res 81 (3) (2001) 189 213.
7. [38] W.C. Hawkes, A. Hwang, Z. Alkan, The effect of selenium supplementation on DTH skin responses in healthy North American men, J Trace Elem Med Biol 23 (4) (2009) 272 280.
8. [39] M. Haldimann, T.Y. Venner, B. Zimmerli, Determination of selenium in the serum of healthy Swiss adults and correlation to dietary intake, J Trace Elem Med Biol 10 (1) (1996) 31–45.
9. [40] K. Demircan, T.S. Chillon, Q. Sun, R.A. Heller, G.J. Klingenberg,
10. [41] C. Ma, P.R. Hoffmann, Selenoproteins as regulators of T cell proliferation, differentiation, and metabolism, Semin Cell Dev Biol 115 (2021) 54 61.
11. [42] S.J. Fairweather-Tait, T. Filippini, M. Vinceti, Selenium status and immunity, Proc Nutr Soc (2022) 1 7.
12. [43] C.A. Stone, K. Kawai, R. Kupka, W.W. Fawzi, Role of selenium in HIV infection, Nutr Rev 68 (11) (2010) 671 681.
13. [44] M. Laclaustra, S. Stranges, A. Navas-Acien, J.M. Ordovas, E. Guallar, Serum selenium and serum lipids in US adults: National Health and Nutrition Examination Survey (NHANES) 2003 2004, Atherosclerosis 210 (2) (2010) 643–648.
14. [45] X. Wang, Y.A. Seo, S.K. Park, Serum selenium and non-alcoholic fatty liver disease (NAFLD) in U.S. adults: National Health and Nutrition Examination Survey (NHANES) 2011–2016, Environ Res 197 (2021), 111190.

1. [46] S. Moon, H.S. Chung, J.M. Yu, H.J. Yoo, J.H. Park, D.S. Kim, et al., Association between serum selenium level and the prevalence of diabetes mellitus in U.S. population, J Trace Elem Med Biol 52 (2019) 83 88.
2. [47] M. Vinceti, P. Grill, C. Malagoli, T. Filippini, S. Storani, M. Malavolti, et al., Selenium speciation in human serum and its implications for epidemiologic research: a cross-sectional study, J Trace Elem Med Biol 31 (2015) 1 10.
3. [48] T. Urbano, T. Filippini, D. Lasagni, T. De Luca, P. Grill, S. Sucato, et al., Association of urinary and dietary selenium and of serum selenium species with serum alanine aminotransferase in a healthy Italian population, Antioxidants (Basel) 10 (10) (2021) 1516.
4. [49] A.G. Renwick, R. Walker, Risk assessment of micronutrients, Toxicol Lett 180
5. [50] M. Vinceti, T. Filippini, K.J. Rothman, Selenium exposure and the risk of type 2 diabetes: a systematic review and meta-analysis, Eur J Epidemiol 33 (9)
6. [51] J.L. Domingo, M. Marques, The effects of some essential and toxic metals/ metalloids in COVID-19: a review, Food Chem Toxicol 152 (2021), 112161.
7. [52] E. Balboni, F. Zagnoli, T. Filippini, S.J. Fairweather-Tait, M. Vinceti, Zinc and selenium supplementation in COVID-19 prevention and treatment: a systematic review of the experimental studies, J Trace Elem Med Biol 71 (2022), 126956.
8. [53] N. Daniel, E. Bouras, K.K. Tsilidis, D.J. Hughes, Genetically predicted circulating concentrations of micronutrients and COVID-19 susceptibility and severity: a Mendelian randomization study, Front Nutr 9 (2022), 842315.
9. [54] M.K. Sobczyk, T.R. Gaunt, The effect of circulating zinc, selenium, copper and vitamin K1 on COVID-19 outcomes: a Mendelian randomization study, Nutrients 14 (2) (2022) 233.
10. [55] O.M. Guillin, C. Vindry, T. Ohlmann, L. Chavatte, Interplay between selenium, selenoproteins and HIV-1 replication in human CD4 T-lymphocytes, Int J Mol Sci 23 (3) (2022) 1394.
11. [56] B.A. Muzembo, N.R. Ngatu, K. Januka, H.L. Huang, C. Nattadech, T. Suzuki, et al., Selenium supplementation in HIV-infected individuals: a systematic review of randomized controlled trials, Clin Nutr ESPEN 34 (2019) 1 7.
12. [57] O.A. Levander, G. Alfthan, H. Arvilommi, C.G. Gref, J.K. Huttunen, M. Kataja, et al., Bioavailability of selenium to Finnish men as assessed by platelet glutathione peroxidase activity and other blood parameters, Am J Clin Nutr 37 (6) (1983) 887–897.
13. [58] R. Hurst, C.N. Armah, J.R. Dainty, D.J. Hart, B. Teucher, A.J. Goldson, et al., Establishing optimal selenium status: results of a randomized, double-blind, placebo-controlled trial, Am J Clin Nutr 91 (4) (2010) 923–931.