High-dose intravenous selenium does not improve clinical outcomes in the critically ill: a systematic review and meta-analysis

Background Selenium (Se) is an essential trace element with antioxidant, anti-inflammatory, and immunomodulatory effects. So far, several randomized clinical trials (RCTs) have demonstrated that parenteral Se may improve clinical outcomes in intensive care unit (ICU) patients. Since publication of our previous systematic review and meta-analysis on antioxidants in the ICU, reports of several trials have been published, including the largest RCT on Se therapy. The purpose of the present systematic review was to update our previous data on intravenous (IV) Se in the critically ill. Methods We searched MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials. We included RCTs with parallel groups comparing parenteral Se as single or combined therapy with placebo. Potential trials were evaluated according to specific eligibility criteria, and two reviewers abstracted data from original trials in duplicate independently. Overall mortality was the primary outcome; secondary outcomes were infections, ICU length of stay (LOS), hospital LOS, ventilator days, and new renal dysfunction. Results A total of 21 RCTs met our inclusion criteria. When the data from these trials were aggregated, IV Se had no effect on mortality (risk ratio [RR] 0.98, 95 % CI 0.90–1.08, P = 0.72, heterogeneity I 2 = 0 %). In addition, when the results of ten trials in which researchers reported on infections were statistically aggregated, there was no significant treatment effect of parenteral Se (RR 0.95, 95 % CI 0.88–1.02, P = 0.15, I 2 = 0 %). There was no positive or negative effect of Se therapy on ICU and hospital LOS, renal function, or ventilator days. Conclusions In critically ill patients, IV Se as monotherapy does not improve clinical outcomes. Electronic supplementary material The online version of this article (doi:10.1186/s13054-016-1529-5) contains supplementary material, which is available to authorized users.

Background Selenium (Se) is an essential trace element with antiinflammatory and immunomodulatory properties, and it is currently considered the cornerstone of the antioxidant defense system [1,2]. Over the past 30 years, a large number of basic and clinical studies have revealed the crucial role of Se in the maintenance of immune, metabolic, endocrine, and cellular homeostasis, which is attributed to its presence in selenoproteins, as the 21st amino acid selenocysteine [3]. These selenoenzymes are involved in redox signaling, antioxidant defense, thyroid hormone metabolism, and immune responses [4]. Thus far, critical illness with systemic inflammation and multiple organ dysfunction syndrome was deemed to be associated with an early reduction in plasma/serum Se and glutathione peroxidase (GPx) activity, where both parameters correlate inversely with the severity of illness and clinical outcome [5,6].
Over the last two decades, researchers in several randomized clinical trials (RCTs) have evaluated the role of parenteral inorganic selenocompounds such as sodium selenite or selenious acid, either as a single-agent strategy or in combination with other antioxidant micronutrients (antioxidant cocktails) and using different dose regimens in critically ill patients with systemic inflammation. These studies have shown beneficial results in terms of reduction of infections, mortality, and other relevant clinical outcomes in the critically ill.
In 2005, the authors of the first comprehensive systematic review and meta-analysis on antioxidant nutrients in the critically ill [7] demonstrated that Se supplementation could be associated with a reduction in mortality, while nonselenium antioxidants had no effect on mortality. More recently, authors of other systematic reviews and meta-analyses [8][9][10][11][12][13][14] on Se therapy in intensive care unit (ICU) patients found that pharmaconutrition with parenteral Se monotherapy may be able to significantly reduce mortality in patients with sepsis, particularly when an intravenous (IV) bolus was provided and daily doses higher than 500 μg were administered. In addition, Se substitution was more effective in those patients with higher risk of death [13].
Nonetheless, since 2013, several studies of the effects of parenteral Se supplementation as single or combined therapy have been published [7,[14][15][16][17]. The REducing Deaths due to Oxidative Stress (REDOXS) trial [16] investigators were unable to find a therapeutic benefit of a combined Se supplementation regimen (300 μg enteral plus 500 μg parenteral). The most recent and largest RCT on Se monotherapy in severe sepsis and septic shock, the Sodium Selenite and Procalcitonin Guided Antimicrobial Therapy in Severe Sepsis (SISPCT) study [17], further demonstrated that high-dose IV sodium selenite was not associated with improved survival. Therefore, with the aim of elucidating the overall efficacy of parenteral Se as single or combined therapy (antioxidant cocktails) in adult critically ill patients, we performed an update of our previous systematic review and meta-analysis of the literature.

Study identification
We conducted a systematic review of the literature published between 1980 and 2015 using the computerized databases of MEDLINE, Embase, the Cochrane Controlled Trials Register, and the Cochrane Database of Systematic Reviews. Text words or MeSH headings containing "randomized," "blind," "clinical trial," "parenteral," "intravenous," "selenium," "sodium selenite," "selenious acid," "antioxidant cocktails," "critical illness," and "critically ill" were used without any language restriction. We also reviewed our personal files and comprehensive reviews for additional original studies.

Study selection criteria
Original studies were included if they met the following criteria: 1. Randomized controlled trial study design with a parallel group 2. A population of critically ill adult patients (>18 years old), defined as patients admitted to an ICU (If the study population was unclear, we considered a mortality rate higher than 5 % in the control group to be consistent with critical illness.) 3. The administration of parenteral Se in the intervention arm (either with or without an initial bolus) as a single-agent strategy or in combination with other antioxidant micronutrients compared with a control group with a placebo 4. The evaluation of clinically relevant outcomes such as mortality, infectious complications, ICU or hospital length of stay (LOS), length of mechanical ventilation (MV), and new renal dysfunction including the requirement of renal replacement therapy Clinical studies that reported only biochemical, metabolic, or immunologic results were excluded.
All original studies were evaluated and abstracted in duplicate independently by two reviewers using a data abstraction form that had been used previously [18]. A discussion was held and consensus was obtained between the reviewers when a disagreement occurred. When additional data were needed, we attempted to contact the authors of the published article. We scored the methodological quality of the original trials, aiming to obtain a score between 0 and 14, with the following high-quality criteria: (1) the extent to which randomization was concealed, (2) intention-to-treat (ITT)-based analysis, (3) extent of blinding, (4) baseline comparability of groups, (5) extent of follow-up, (6) description of treatment protocols and cointerventions in both arms, and (7) definition of clinical outcomes [18].
We designated studies as level I if all of the following criteria were fulfilled: concealed randomization, blinded outcome adjudication, and an ITT analysis, which are the strongest methodological tools to reduce bias. A study was considered as level II if any one of the abovedescribed characteristics was unfulfilled.

Data analysis
The primary outcome was overall mortality. Hospital mortality, when available, was used for the statistical analysis. If not reported, we used ICU mortality or 28day mortality. When not specified, mortality was assumed to be hospital mortality. Secondary outcomes included infections, hospital and ICU LOS, MV days, and new renal dysfunction as defined by the authors of the original articles. We used the definition of infections employed by each author. The data from all trials reporting the specific outcome were combined to calculate the pooled risk ratio (RR) for mortality and infections, and pooled weighted mean difference (WMD) for LOS, both with 95 % CIs. All analyses were conducted using Review Manager (RevMan) 5.3 software, except for the test for asymmetry. Pooled RRs were calculated using the Mantel-Haenszel estimator, and WMDs were estimated by the inverse variance approach. The random effects model of DerSimonian and Laird [19] was used to estimate variances for the Mantel-Haenszel and inverse variance estimators. When possible, studies were aggregated on an ITT basis. Heterogeneity in the data was tested by a weighted Mantel-Haenszel chi-square test and quantified by using the I 2 statistics implemented in RevMan 5.3 software. Differences between subgroups were analyzed using the test of subgroup differences described by Deeks et al. [20], and the results were expressed using the P values. Funnel plots were generated to assess the possibility of publication bias, and the Egger regression test was used to measure funnel plot asymmetry [21]. Asymmetry was calculated using Comprehensive Meta-Analysis 3.0 statistical software (Biostat Inc., Englewood, NJ, USA). P values <0.05 and <0.10 were considered as statistically significant and indicators of a trend, respectively.

A priori hypothesis testing
Significant differences in the protocols of the original studies were expected. Thus, several prespecified hypothesisgenerating subgroup analyses were performed to identify potentially more beneficial treatment strategies. First, we compared the results of trials in which investigators administered parenteral Se as monotherapy with studies in which researchers provided parenteral Se in antioxidant cocktails. Based on previous RCTs showing a beneficial effect of an initial loading dose, those RCTs using an initial loading dose as an IV bolus of Se were then compared with trials those that did not. In addition, because researchers in previous trials found that daily doses higher than 500 μg were associated with better outcomes, we compared the results between three subgroups having different daily doses: lower than 500 μg, equal to 500 μg, and greater than 500 μg. Moreover, on the basis of a possibly larger treatment effect in patients with higher risk of death, we compared studies including patients with higher mortality vs. those with lower mortality. Mortality was considered to be high or low based on whether it was greater or less than the mean control group mortality of all the trials. Additionally, we postulated that trials with lower quality (level II studies) might demonstrate a greater treatment effect than those trials with higher quality (level I studies). Furthermore, as current evidence showed benefits in terms of reduction in mortality in septic patients, the results of RCTs performed only with patients with sepsis were compared with RCTs performed with heterogeneous patient populations (nonsepsis studies). We also assessed the effect of Se in soils according to the geographical region where the trial was conducted. For this purpose, we compared RCTs performed in deprived regions (Europe, South America, and Asia) versus trials performed in nondeprived regions (North America). Finally, given the interaction between Se and procalcitonin (PCT) in the SISPCT study [17], we conducted a sensitivity analysis excluding the PCT guidance group of patients.

Study identification and selection
A total of 41 relevant citations were identified in the search of computerized bibliographic databases and a review of reference lists in related articles. Of these, we excluded 20 for the following reasons: 8 trials did not include ICU patients (mostly surgery patients) [22][23][24][25][26][27][28][29]; 1 study did not evaluate clinical outcomes [30]; 1 study compared high-dose with low-dose Se [31]; 3 articles were duplicates [32][33][34]; 4 articles were systematic reviews; 1 trial was published as an abstract [35], and we were unable to obtain the data from the authors to complete our data abstraction process; 1 study was not an RCT [36]; and in 1 trial Se was not given intravenously [37].
Ultimately, 21 studies [14-17, 38-54] met our inclusion criteria and were included; they comprised a total of 4044 patients (Tables 1 and 2). The reviewers reached 100 % agreement for the inclusion of the trials. The mean methodological score of all trials was 9 of a maximum possible score of 14 (range 4-13). Randomization was concealed in 9 (43 %) of 21 trials; ITT analysis was performed in 14 (67 %) of 21 trials; and double-blinding was done in 7 (33 %) of 21 of the studies. There were 6 level I studies and 15 level II studies. The details of the methodological quality of the individual trials are shown in Table 1.

Primary outcome: mortality
When the results of the 21 trials in which researchers reported mortality were aggregated, no statistically significant difference was found between Se supplementation and placebo (RR 0.98, 95 % CI 0.90-1.08, P = 0.72, heterogeneity I 2 = 0 %) (Fig. 1). In the sensitivity analysis, after excluding the PCT guidance group of the Bloos et al. study, we found that the new RR was 0.95 (95 % CI 0.79-1.15, P = 0.63, I 2 = 35 %) (Additional file 1: Figure S1).

Secondary outcomes Overall effect on new infectious complications
When data from ten studies were included in the metaanalysis, no significant effect of Se supplementation on infections was found (RR 0.95, 95 % CI 0.88-1.02, P = 0.15, I 2 = 0 %) (Fig. 2).
Overall effect on ICU and hospital length of stay and ventilator days When ten RCTs in which researchers reported ICU LOS were statistically aggregated, there were no significant differences between the groups (WMD 0.32, 95 % CI −0.80

Overall effect on new renal dysfunction
After aggregating the data from eight RCTs in which researchers reported new renal dysfunction, Se therapy was not associated with a significant reduction in the incidence of renal dysfunction (RR 0.79, 95 % CI 0.57-1.08, P = 0.14,

Subgroup analyses PN selenium monotherapy vs. combined therapy
The RR associated with parenteral Se monotherapy for mortality was 0.91 (95 % CI 0.79-1.04, P = 0.16; I 2 = 12 %) ( Fig. 1), compared with 1.08 for studies in which researchers used Se as part of combination therapy (95 % CI 0.93-1.25, P = 0.33). The test of subgroup differences was not significant (P = 0.10). There was no effect on infections with either Se monotherapy or combined therapy (test for subgroup differences, P = 0.59) (Fig. 2).

PN selenium loading dose vs. no loading dose
There was no treatment effect difference in mortality between studies in which researchers used a parenteral loading dose of Se as an IV bolus (RR 0.90, 95 % CI 0.75-1.08, P = 0.30, I 2 = 18 %) and those not using a loading dose (RR 1.01, 95 % CI 0.90-1.14, P = 0.83) (Fig. 3). The test for subgroup differences was not statistically significant (P = 0.30). There was also no effect on infectious complications in studies in which researchers used a parenteral loading dose (RR 0.99, 95 % CI 0.90-1.09, P = 0.84, I 2 = 0 %), whereas there was a significant reduction on infections in studies without a parenteral Se loading dose (RR 0.87, 95 % CI 0.77-0.99, P = 0.03, I 2 = 0 %) (Fig. 4); the test for subgroup differences was not significant (P = 0.11).

PN selenium high dose vs. low dose
There were no significant treatment effect differences in mortality (test for subgroup differences P = 0.89) and infections (test for subgroup differences P = 0.52) when we compared studies with high-dose vs. low-dose parenteral Se supplementation (data not shown).

Higher vs. lower mortality
The mean hospital mortality rate (or ICU mortality when hospital mortality was not reported) in the control group of all the trials was 33 %. After aggregating 11 studies with a higher mortality rate, we found that IV Se did not have an effect on mortality (RR 0.94, 95 % CI 0.84-1.05, P = 0.26, I 2 = 17 %). In addition, IV Se did not have an effect on mortality in eight studies with a lower mortality (RR 1.08, 95 % CI 0.90-1.08, P = 0.24, I 2 = 0 %). The test for subgroup differences was not significant (P = 0.17) (see Additional file 2: Figure S2) In addition, there were no significant treatment effect differences in infections (test for subgroup differences P = 0.35) when we compared studies with high vs. low mortality rates in the control group (see Additional file 2: Figure S3).

Study quality on outcomes
When we evaluated study quality on outcomes, a statistically significant effect of IV Se on the reduction of mortality was found in the low-quality trials (RR 0.79, 95 % CI 0.66-0.94, P = 0.007, I 2 = 0 %), whereas trials with higher methodology scores did not show any significant effect (RR 1.06, 95 % CI 0.95-1.19, P = 0.27, heterogeneity I 2 = 0 %). The overall tests for significance revealed statistically significant differences between these subgroups (P = 0.004) (see Additional file 3: Figure S4). Neither high-nor low-quality trials demonstrated any effect on infections (test for subgroup differences P = 0.72) (see Additional file 3: Figure S5).

Effect of geographic representation of study patients on outcomes
We found no significant treatment effect differences in mortality (test for subgroup differences P = 0.33) and infections (test for subgroup differences P = 0.57) when we compared studies in which researchers administered IV Se in North America vs. other geographic regions (see Additional file 4: Figs. S6 and S7).

Publication bias
There was no indication that publication bias influenced the observed aggregated results. In fact, funnel plots were created for each study outcome, and the tests of asymmetry showed a trend for mortality (P = 0.08), although the data were not significant for overall infections (P = 0.19), hospital LOS (P = 0.50), or MV days (P = 0.37). However, the test for asymmetry was significant for ICU LOS (P = 0. 047).

Discussion
This updated systematic review and meta-analysis of the effects of parenteral Se as single or combined therapy on

PN Selenium Combined
Berger 1998     [16,17], the main finding of our meta-analysis is that there was a lack of treatment effect when critically ill patients were treated with IV Se as single or combined therapy. In fact, we were unable to demonstrate any effect of IV Se supplementation on mortality or any significant effect on infections, ventilator days, or ICU and hospital LOS. Furthermore, our a priori defined subgroup analyses did not show any treatment effects on mortality. However, we found a significant effect of Se therapy on infectious complications in those studies without an initial IV loading dose, as well as a similar effect in trials conducted in nonseptic patients.
Over the last few years, different meta-analyses of Se supplementation in the critically ill have been published [8][9][10][11][12][13]. The differences with our present review are due largely to the variations in the studies included in the reviews, as our systematic review and meta-analysis is the first to include the two largest trials done to date on Se therapy in the critically ill.
Our present findings do not support the concept of pharmaconutrition by which micronutrients such as Se are provided in high (i.e., supraphysiological) doses in order to derive a pharmacological effect. Conversely to previous findings regarding IV high-dose Se monotherapy [8,9], we did not find an overall effect of Se on infectious complications in the critically ill, although parenteral Se without an initial bolus significantly reduced infections. Nonetheless, the overall point estimate on infections of Se monotherapy is primarily and largely influenced by the Bloos et al. [17] study, the largest trial (n = 1089) (n= 1089) on pharmaconutrition with high-dose Se monotherapy and PCT-guided antibiotic therapy in patients with severe sepsis and septic shock. After giving an initial IV loading dose of 1000 μg sodium selenite followed by a continuous infusion of 1000 μg sodium selenite daily for no longer than 21 days, Bloos et al. [17] found that secondary infections were similar in both groups of patients. Interestingly enough, the SISPCT study [17] demonstrated that high-dose Se supplementation had no therapeutic benefit in septic patients, although plasma Se depletion at baseline was restored to the normal range already by treatment day 1, suggesting that correction of plasma Se concentration may have no beneficial value. According to this study, plasma Se levels in the Se or placebo groups were not affected by allocation to the PCT guidance or non-PCT   guidance group. Our sensitivity analysis showed that, after excluding the PCT guidance arm, the new RR was 0.95 (previous RR 0.98), which confirms that excluding the PCT arm of the Bloos et al. study [17] did not affect the overall result of our analysis.
In contrast to previous knowledge, we were unable to find beneficial effects with daily doses higher than 500 μg or with providing an initial loading dose as an IV bolus (usually 1000-2000 μg in 30 minutes to 2 h). However, the absence of a significant test of subgroup differences weakens any inferences drawn from this subgroup analysis, but it likely shows the ineffectiveness of employing a loading dose with an aim of improving outcomes. So far, it has been proposed that a loading dose of 1000-2000 μg Se as pentahydrate sodium selenite has prooxidant and cytotoxic effects [1,53,55]. In a previous pharmacokinetic study [31], it was demonstrated that an initial IV bolus of 2000 μg followed by a continuous infusion of 1600 μg/day was the most effective dose for returning serum Se to physiologic levels and safely maximizing extracellular GPx activity and therefore the antioxidant capacity in critically ill patients. According to current knowledge, a very high Se concentration may be able to produce an inhibition of nuclear factor-κB binding to DNA, controlling gene expression and the synthesis of proinflammatory cytokines [56,57], and also may be able to induce apoptosis and cytotoxicity in activated proinflammatory cells [58]. In addition,    [59] demonstrated that an IV Se bolus improved hemodynamic status, decreased inflammation biomarkers, and reduced mortality. Meanwhile, contrary to our present data, in 2012 we found that a parenteral loading dose showed a trend toward reduction in mortality, whereas studies that did not use a bolus-loading dose did not show any effect on mortality. Similarly, Huang et al. [9], after aggregating nine RCTs on Se monotherapy, demonstrated that an IV bolus was associated with a significant reduction in mortality (RR 0.73, 95 % CI 0.58-0.94, P = 0.01). However, neither of the previous meta-analyses considered the SISPCT study [17].
So far, most clinical studies using Se at low doses have been underpowered and have involved Se administered in a cocktail approach. Thus, positive results in those trials cannot be clearly attributed solely to Se supplementation. Notwithstanding this, according to the concept of nutrient replacement, by which micronutrient substitution is aimed at replenishing losses and target restoration of physiological function [60], Se must be supplemented at standard doses by the enteral (77-100 μg/day) or the parenteral (100-400 μg/day) route [61] because the results of our meta-analysis do not refute previously recommended Se substitution doses.
Despite earlier results of the Angstwurm study [41], which demonstrated that IV high-dose Se substitution in septic patients with an Acute Physiology and Chronic Health Evaluation II score higher than 15 significantly reduced the requirements of renal replacement therapy (P = 0.035), the post hoc analysis of the REDOXS [16] study demonstrated that patients with renal failure might have a worse outcome when treated with highdose antioxidants. In fact, Heyland and coworkers [16] demonstrated that both glutamine and antioxidants appeared harmful in patients with baseline renal dysfunction, showing a higher 28-day mortality (OR 3.39, 95 % CI 1.41-8.17, and 3.07, 95 % CI 1.24-7.59, for antioxidants alone and glutamine plus antioxidants, respectively). Nonetheless, in the recently published SISPCT study [17], researchers did not find any risk of increased harm in patients with baseline renal failure. In addition, in the present study, we were unable to find any deleterious effect of Se therapy on renal function in the critically ill.
Thus, the current understanding of why there is a lack of therapeutic effect of IV Se therapy in critically ill patients and patients with severe sepsis remains unclear.   Notwithstanding this, Se therapy could show benefits in other patient populations that were not considered in our meta-analysis. In fact, it is currently known that circulating Se levels significantly decrease in the perioperative period of cardiac surgery [62]. Also, in a nonrandomized interventional trial, Stoppe et al. [63] demonstrated that high-dose sodium selenite therapy as a pharmaconutrient strategy was effective in preventing the decrease of Se levels and that clinical outcomes may be superior in supplemented patients compared with a historical control group. The SodiUm SeleniTe Administration IN Cardiac Surgery (SUSTAIN CSX®trial, ClinicalTrials.gov identifier NCT02002247), an RCT aimed at evaluating the effects of perioperative high-dose Se supplementation in high-risk cardiac surgical patients undergoing complicated open heart surgery, is currently recruiting participants [64].
According to current evidence derived from recent trials and our meta-analysis, the updated version of Canadian Clinical Guidelines [65] recommended not using IV Se alone or in combination with other antioxidants in critically ill patients, which means that this strategy has recently been downgraded.
The strength of our meta-analysis is based on the fact that we used several methods to reduce bias (comprehensive literature search, duplicate data abstraction, comprehensive search strategy using specific criteria, and including non-English-language articles), we contacted trial authors to obtain additional data and refine our analysis, and we ultimately focused on clinically important primary outcomes in ICU patients. In addition, given the wide variety of clinical diagnoses and the heterogeneous population of ICU patients included in this systematic review (sepsis, septic shock, trauma, pancreatitis, surgical ICU patients), the results and conclusions may be applied to a broad and heterogeneous group of critically ill patients. While having a fairly large overall sample size and low heterogeneity, which makes the estimate quite robust, our subgroup analyses are limited by the small number of trials.

Conclusions
In this updated systematic review and meta-analysis, we found that parenteral Se as single or combined therapy with other antioxidant micronutrients had no effect on mortality, infections, renal function, ICU and hospital