Remember, a long time has passed and a lot of people have died while Evidence-Based Medicine was facing the wrong way.
And asking the wrong question. "What new treatment will save more lives in the ICU?" is an important question, but one with few answers and no great ones - "What can stop people who catch SARS-CoV-2 coming to the ICU?" is a better one in a pandemic, and one that might also lead to better treatment protocols.
Selenium reduces COVID-19 mortality: A Bradford Hill analysis
Selenium reduces COVID-19 mortality: A Bradford Hill analysis
1) Strength of association. Very Strong.
1) On inspection of the Hubei data, it is notable that the cure rate in Enshi city, at 36.4%, was much higher than that of other Hubei cities, where the overall cure rate was 13.1% (Supplemental Table 1); indeed, the Enshi cure rate was significantly different from that in the rest of Hubei (P < 0.0001). Enshi is renowned for its high selenium intake and status [mean ± SD: hair selenium: 3.13 ± 1.91 mg/kg for females and 2.21 ± 1.14 mg/kg for males]—compare typical levels in Hubei of 0.55 mg/kg (10)—so much so that selenium toxicity was observed there in the 1960s. Selenium intake in Enshi was reported as 550 µg/d in 2013.
Similar inspection of data from provinces outside Hubei shows that Heilongjiang Province in northeast China, a notoriously low-selenium region in which Keshan is located, had a much higher death rate, at 2.4%, than that of other provinces (0.5%; P < 0.0001). The selenium intake was recorded as only 16 µg/d in a 2018 publication, while hair selenium in the Songnen Plain of Heilongjiang was measured as only 0.26 mg/kg (Supplemental Table 2).
Finally, we found a significant association between cure rate and background selenium status in cities outside Hubei (R2 = 0.72, F test P < 0.0001; Figure 1, Supplemental Table 2).[2]
![Correlation between COVID-19 cure rate in 17 cities outside Hubei, China, on 18 February, 2020 and city population selenium status (hair selenium concentration) analyzed using weighted linear regression (mean ± SD = 35.5 ± 11.1, R2 = 0.72, F test P < 0.0001). Each data point represents the cure rate, calculated as the number of cured patients divided by the number of confirmed cases, expressed as a percentage. The size of the marker is proportional to the number of cases.]()
2) Serum samples (n = 166) from COVID-19 patients (n = 33) were collected consecutively and analyzed for total Se by X-ray fluorescence and selenoprotein P (SELENOP) by a validated ELISA. Both biomarkers showed the expected strong correlation (r = 0.7758, p < 0.001), pointing to an insufficient Se availability for optimal selenoprotein expression. In comparison with reference data from a European cross-sectional analysis (EPIC, n = 1915), the patients showed a pronounced deficit in total serum Se (mean ± SD, 50.8 ± 15.7 vs. 84.4 ± 23.4 µg/L) and SELENOP (3.0 ± 1.4 vs. 4.3 ± 1.0 mg/L) concentrations. A Se status below the 2.5th percentile of the reference population, i.e., [Se] < 45.7 µg/L and [SELENOP] < 2.56 mg/L, was present in 43.4% and 39.2% of COVID samples, respectively.
The Se status was significantly higher in samples from surviving COVID patients as compared with non-survivors (Se; 53.3 ± 16.2 vs. 40.8 ± 8.1 µg/L, SELENOP; 3.3 ± 1.3 vs. 2.1 ± 0.9 mg/L), recovering with time in survivors while remaining low or even declining in non-survivors.[3]
2) Serum samples (n = 166) from COVID-19 patients (n = 33) were collected consecutively and analyzed for total Se by X-ray fluorescence and selenoprotein P (SELENOP) by a validated ELISA. Both biomarkers showed the expected strong correlation (r = 0.7758, p < 0.001), pointing to an insufficient Se availability for optimal selenoprotein expression. In comparison with reference data from a European cross-sectional analysis (EPIC, n = 1915), the patients showed a pronounced deficit in total serum Se (mean ± SD, 50.8 ± 15.7 vs. 84.4 ± 23.4 µg/L) and SELENOP (3.0 ± 1.4 vs. 4.3 ± 1.0 mg/L) concentrations. A Se status below the 2.5th percentile of the reference population, i.e., [Se] < 45.7 µg/L and [SELENOP] < 2.56 mg/L, was present in 43.4% and 39.2% of COVID samples, respectively.
The Se status was significantly higher in samples from surviving COVID patients as compared with non-survivors (Se; 53.3 ± 16.2 vs. 40.8 ± 8.1 µg/L, SELENOP; 3.3 ± 1.3 vs. 2.1 ± 0.9 mg/L), recovering with time in survivors while remaining low or even declining in non-survivors.[3]
2) Consistency - Strong
All epidemiological data about selenium and COVID-19 is consistent in direction and effect size. However, tests that could be done comparing COVID-19 risk in high and low selenium regions of Brazil and the USA would establish consistency further.
All epidemiological data about selenium and COVID-19 is consistent in direction and effect size. However, tests that could be done comparing COVID-19 risk in high and low selenium regions of Brazil and the USA would establish consistency further.
3) Specificity - Very Strong
Selenium has much weaker or less consistent associations with other diseases, except those caused by other RNA viruses, e.g. when risk of hepatocellular cancer in viral hepatitis patients is compared with risk of osteoporosis.[4, 5]
4) Temporality - Strong
Prospective ecological comparisons are temporal by design.[2] In the German study, the temporal association between low serum selenium levels and COVID-19 symptom severity was closely tracked.[3]
![Nutrients 12 02098 g003 550]()
5) Dose-response gradient - Very Strong
Selenium has much weaker or less consistent associations with other diseases, except those caused by other RNA viruses, e.g. when risk of hepatocellular cancer in viral hepatitis patients is compared with risk of osteoporosis.[4, 5]
4) Temporality - Strong
Prospective ecological comparisons are temporal by design.[2] In the German study, the temporal association between low serum selenium levels and COVID-19 symptom severity was closely tracked.[3]
5) Dose-response gradient - Very Strong
A strong, consistent dose-response is seen, even at levels where the risk of selenium toxicity exists, and despite the fact that toxic levels of soil selenium are often a legacy of industrial pollution in China.[2]
6) Plausibility - Very High
Reading references 2 and 3, as well as this review of the evidence written before reference 2 was published, should be persuasive.[6] See also ref 17 for antiviral effects. This one goes up to 11.
7) Coherence - Very High
Selenium is well-studied and nothing in its story seems to contradict the idea that higher intakes will protect against COVID-19 mortality and reduce the severity of disease.
Dexamethasone, a drug which can reduce COVID-19 mortality in the ICU, enhances 1α,25-dihydroxyvitamin D3 effects by increasing vitamin D receptor transcription.[7]
Selenium sufficiency is essential for the function of vitamin D in peripheral blood monocytes.[8] Vitamin D status also correlates with COVID-19 survival.[9]
6) Plausibility - Very High
Reading references 2 and 3, as well as this review of the evidence written before reference 2 was published, should be persuasive.[6] See also ref 17 for antiviral effects. This one goes up to 11.
7) Coherence - Very High
Selenium is well-studied and nothing in its story seems to contradict the idea that higher intakes will protect against COVID-19 mortality and reduce the severity of disease.
Dexamethasone, a drug which can reduce COVID-19 mortality in the ICU, enhances 1α,25-dihydroxyvitamin D3 effects by increasing vitamin D receptor transcription.[7]
Selenium sufficiency is essential for the function of vitamin D in peripheral blood monocytes.[8] Vitamin D status also correlates with COVID-19 survival.[9]
8) Experiment - Weak (Insufficient)
This is an area of neglect, but overlaps with the next section as there are several trials of selenium supplementation in other viral diseases, and animal experiments in analogous conditions, and many mechanistic experiments that are non-specific. However the interaction between SARS-CoV-2 and selenoproteins has been confirmed by experiment.[10]
9) Analogy - Strong
Selenium intake is protective, and selenium supplementation has been useful, in other viral illnesses.
However, the protective effect of high selenium intakes before infection in epidemiology appears stronger than the protective effect of selenium as a late intervention in disease.[6, 11]
This is an area of neglect, but overlaps with the next section as there are several trials of selenium supplementation in other viral diseases, and animal experiments in analogous conditions, and many mechanistic experiments that are non-specific. However the interaction between SARS-CoV-2 and selenoproteins has been confirmed by experiment.[10]
9) Analogy - Strong
Selenium intake is protective, and selenium supplementation has been useful, in other viral illnesses.
However, the protective effect of high selenium intakes before infection in epidemiology appears stronger than the protective effect of selenium as a late intervention in disease.[6, 11]
10)Risk - Weak, Well-Established
We can add the most relevant of extra questions to any given set of criteria - "strength of the alternative hypothesis" would be a good one for any lipid hypothesis.
Bradford Hill stated that some interventions are easier to justify than others.
[Edit: 1/09/20] There is also very good evidence that intravenous high dose selenite is safe in the ICU setting.
Totally 19 RCTs involving 3341 critically ill patients were carried out in which 1694 participates were in the selenium supplementation group, and 1647 in the control. The aggregated results suggested that compared with the control, intravenous selenium supplement as a single therapy could decrease the total mortality (RR = 0.86, 95% CI: 0.78–0.95, P = .002, TSA-adjusted 95% CI = 0.77–0.96, RIS = 4108, n = 3297) and may shorten the length of stay in hospital (MD −2.30, 95% CI −4.03 to −0.57, P = .009), but had no significant treatment effect on 28-days mortality (RR = 0.96, 95% CI: 0.85–1.09, P = .54) and could not shorten the length of ICU stay (MD −0.15, 95% CI −1.68 to 1.38, P = .84) in critically ill patients.[13]
This, and an earlier analysis which found less benefit, did not single out viral illnesses as a subgroup - this is only evidence for safety - but the earlier analysis did find a) slightly lower mortality in trials without an initial bolus dose, b) no increased risk in patients with renal disease.[14]
I will hypothesize briefly on selenium increasing mortality at 300 mcg/day in the Danish intervention study, a dose far too low to cause selenosis. This trial used selenium yeast, which supplies selenium as selenomethionine. This is somewhat more effective at raising selenoprotein activity than selenite over the short-term (Brazil nuts are more effective than both). However, if selenomethionine intake is in excess of the body's ability to convert it to selenocysteine by transmethylation, and especially if methionine intake is low, then selenomethionine will be incorporated into proteins instead of methionine, causing these proteins to lose some functionality. Because mammals cannot synthesize methionine de novo, but can synthesize selenocysteine, this risk does not exist with selenite.[15] (Selenite itself also plays a role in signalling that might be critical in SARS-CoV-2 infections.) If much of the selenium in high-selenium regions is present as inorganic selenium in drinking water, and selenocysteine in the flesh of animals that have processed the extra selenomethionine in local plants, this explains a higher selenium tolerance than seen in the Danish study.
(The conventional signs of selenosis result from selenocysteine replacing cysteine in proteins, and the relative weakness of the Se-Se bond compared with the S-S bond.)
It makes sense to me that selenomethionine, very useful as it will increase selenoprotein levels quickly if you don't have much time, should be replaced with sodium selenite for long-term coverage.
Brazil nuts are a variable quantity, a sample of nuts sold in NZ in 2008 had an average of 19 mcg per nut and increased selenoprotein levels more than selenomethionine.[16]
Plasma selenium increased by 64.2%, 61.0%, and 7.6%; plasma GPx by 8.3%, 3.4%, and -1.2%; and whole blood GPx by 13.2%, 5.3%, and 1.9% in the Brazil nut, selenomethionine, and placebo groups, respectively. Change over time at 12 wk in plasma selenium (P < 0.0001 for both groups) and plasma GPx activity in the Brazil nut (P < 0.001) and selenomethionine (P = 0.014) groups differed significantly from the placebo group but not from each other. The change in whole blood GPx activity was greater in the Brazil nut group than in the placebo (P = 0.002) and selenomethionine (P = 0.032) groups.
[Edit 02/09/20] - thanks to Mike Angell for this link; while all selenium sources are probably protective against death and ongoing harm from COVID-19, only selenite is likely to have an additional antiviral effect, and has low toxicity.[17]
![]()
We can add the most relevant of extra questions to any given set of criteria - "strength of the alternative hypothesis" would be a good one for any lipid hypothesis.
Bradford Hill stated that some interventions are easier to justify than others.
On fair evidence we might take action on what appears to be an occupational hazard, e.g. we might change from a probably carcinogenic oil to a non-carcinogenic oil in a limited environment and without too much injustice if we are wrong. But we should need very strong evidence before we made people burn a fuel in their homes that they do not like or stop smoking the cigarettes and eating the fats and sugar that they do like. In asking for very strong evidence I would, however, repeat emphatically that this does not imply crossing every ‘t’, and swords with every critic, before we act.
With nutrient intakes there is often an identifiable risk, with a J-shaped curve. With selenium the risk is selenosis, which is a condition that requires chronic high exposure (I have given myself mild selenosis with around 900mcg selenium a day and it was not a terrible condition to experience and was reversible). There could be other risks. Luckily we have an experiment that tells us where the limit is.
In a low selenium country, like New Zealand or Denmark, you don't want to take more than 200mcg of extra selenium long term.[12] Pity the low dose arms here weren't retained in the intervention.
![fx1]()
During 6871 person-years of follow-up, 158 deaths occurred. In an intention-to-treat analysis, the hazard ratio (95% confidence interval) for all-cause mortality comparing 300 µg selenium/d to placebo was 1.62 (0.66, 3.96) after 5 years of treatment and 1.59 (1.02, 2.46) over the entire follow-up period. The 100 and 200 µg/d doses showed non-significant decreases in mortality during the intervention period that disappeared after treatment cessation. Although we lacked power for endpoints other than all-cause mortality, the effects on cancer and cardiovascular mortality appeared similar.
Howsoever that may be, taking extra selenium above 200mcg per day may yet be advised if one becomes ill with Covid, but an inorganic salt of selenium like sodium selenite (which is anyhow probably safer than the selenomethionine form long-term, as I'll discuss below) is preferable, according to the selenovirus expert, Ethan Will Taylor.
(this video link does not show in the mobile version of this post but can be reached through the web view option at the bottom)
With nutrient intakes there is often an identifiable risk, with a J-shaped curve. With selenium the risk is selenosis, which is a condition that requires chronic high exposure (I have given myself mild selenosis with around 900mcg selenium a day and it was not a terrible condition to experience and was reversible). There could be other risks. Luckily we have an experiment that tells us where the limit is.
In a low selenium country, like New Zealand or Denmark, you don't want to take more than 200mcg of extra selenium long term.[12] Pity the low dose arms here weren't retained in the intervention.
During 6871 person-years of follow-up, 158 deaths occurred. In an intention-to-treat analysis, the hazard ratio (95% confidence interval) for all-cause mortality comparing 300 µg selenium/d to placebo was 1.62 (0.66, 3.96) after 5 years of treatment and 1.59 (1.02, 2.46) over the entire follow-up period. The 100 and 200 µg/d doses showed non-significant decreases in mortality during the intervention period that disappeared after treatment cessation. Although we lacked power for endpoints other than all-cause mortality, the effects on cancer and cardiovascular mortality appeared similar.
Howsoever that may be, taking extra selenium above 200mcg per day may yet be advised if one becomes ill with Covid, but an inorganic salt of selenium like sodium selenite (which is anyhow probably safer than the selenomethionine form long-term, as I'll discuss below) is preferable, according to the selenovirus expert, Ethan Will Taylor.
(this video link does not show in the mobile version of this post but can be reached through the web view option at the bottom)
[Edit: 1/09/20] There is also very good evidence that intravenous high dose selenite is safe in the ICU setting.
Totally 19 RCTs involving 3341 critically ill patients were carried out in which 1694 participates were in the selenium supplementation group, and 1647 in the control. The aggregated results suggested that compared with the control, intravenous selenium supplement as a single therapy could decrease the total mortality (RR = 0.86, 95% CI: 0.78–0.95, P = .002, TSA-adjusted 95% CI = 0.77–0.96, RIS = 4108, n = 3297) and may shorten the length of stay in hospital (MD −2.30, 95% CI −4.03 to −0.57, P = .009), but had no significant treatment effect on 28-days mortality (RR = 0.96, 95% CI: 0.85–1.09, P = .54) and could not shorten the length of ICU stay (MD −0.15, 95% CI −1.68 to 1.38, P = .84) in critically ill patients.[13]
This, and an earlier analysis which found less benefit, did not single out viral illnesses as a subgroup - this is only evidence for safety - but the earlier analysis did find a) slightly lower mortality in trials without an initial bolus dose, b) no increased risk in patients with renal disease.[14]
I will hypothesize briefly on selenium increasing mortality at 300 mcg/day in the Danish intervention study, a dose far too low to cause selenosis. This trial used selenium yeast, which supplies selenium as selenomethionine. This is somewhat more effective at raising selenoprotein activity than selenite over the short-term (Brazil nuts are more effective than both). However, if selenomethionine intake is in excess of the body's ability to convert it to selenocysteine by transmethylation, and especially if methionine intake is low, then selenomethionine will be incorporated into proteins instead of methionine, causing these proteins to lose some functionality. Because mammals cannot synthesize methionine de novo, but can synthesize selenocysteine, this risk does not exist with selenite.[15] (Selenite itself also plays a role in signalling that might be critical in SARS-CoV-2 infections.) If much of the selenium in high-selenium regions is present as inorganic selenium in drinking water, and selenocysteine in the flesh of animals that have processed the extra selenomethionine in local plants, this explains a higher selenium tolerance than seen in the Danish study.
(The conventional signs of selenosis result from selenocysteine replacing cysteine in proteins, and the relative weakness of the Se-Se bond compared with the S-S bond.)
It makes sense to me that selenomethionine, very useful as it will increase selenoprotein levels quickly if you don't have much time, should be replaced with sodium selenite for long-term coverage.
Brazil nuts are a variable quantity, a sample of nuts sold in NZ in 2008 had an average of 19 mcg per nut and increased selenoprotein levels more than selenomethionine.[16]
Plasma selenium increased by 64.2%, 61.0%, and 7.6%; plasma GPx by 8.3%, 3.4%, and -1.2%; and whole blood GPx by 13.2%, 5.3%, and 1.9% in the Brazil nut, selenomethionine, and placebo groups, respectively. Change over time at 12 wk in plasma selenium (P < 0.0001 for both groups) and plasma GPx activity in the Brazil nut (P < 0.001) and selenomethionine (P = 0.014) groups differed significantly from the placebo group but not from each other. The change in whole blood GPx activity was greater in the Brazil nut group than in the placebo (P = 0.002) and selenomethionine (P = 0.032) groups.
[Edit 02/09/20] - thanks to Mike Angell for this link; while all selenium sources are probably protective against death and ongoing harm from COVID-19, only selenite is likely to have an additional antiviral effect, and has low toxicity.[17]
All scientific work is incomplete - whether it be observational or experimental. All scientific work is liable to be upset or modified by advancing knowledge. That does not confer upon us a freedom to ignore the knowledge we already have, or to postpone the action that it appears to demand at a given time.
Austin Bradford Hill, 1965.
References:
[1] Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58(5):295-300.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1898525/pdf/procrsmed00196-0010.pdf[2] Jinsong Zhang, Ethan Will Taylor, Kate Bennett, Ramy Saad, Margaret P Rayman, Association between regional selenium status and reported outcome of COVID-19 cases in China, The American Journal of Clinical Nutrition, Volume 111, Issue 6, June 2020, Pages 1297–1299, https://doi.org/10.1093/ajcn/nqaa095
[3] Moghaddam, A.; Heller, R.A.; Sun, Q.; Seelig, J.; Cherkezov, A.; Seibert, L.; Hackler, J.; Seemann, P.; Diegmann, J.; Pilz, M.; Bachmann, M.; Minich, W.B.; Schomburg, L. Selenium Deficiency Is Associated with Mortality Risk from COVID-19. Nutrients 2020, 12, 2098.
[4] Yu MW, Horng IS, Hsu KH, Chiang YC, Liaw YF, Chen CJ. Plasma selenium levels and risk of hepatocellular carcinoma among men with chronic hepatitis virus infection. Am J Epidemiol. 1999;150(4):367-374. doi:10.1093/oxfordjournals.aje.a010016
[4] Yu MW, Horng IS, Hsu KH, Chiang YC, Liaw YF, Chen CJ. Plasma selenium levels and risk of hepatocellular carcinoma among men with chronic hepatitis virus infection. Am J Epidemiol. 1999;150(4):367-374. doi:10.1093/oxfordjournals.aje.a010016
[5] Wang, Y., Xie, D., Li, J. et al. Association between dietary selenium intake and the prevalence of osteoporosis: a cross-sectional study. BMC Musculoskelet Disord 20, 585 (2019). https://doi.org/10.1186/s12891-019-2958-5
[6] Bermano, G., Méplan, C., Mercer, D., & Hesketh, J. (2020). Selenium and viral infection: Are there lessons for COVID-19? British Journal of Nutrition, 1-37. doi:10.1017/S0007114520003128
https://www.cambridge.org/core/journals/british-journal-of-nutrition/article/selenium-and-viral-infection-are-there-lessons-for-covid19/BE3AC78D5C92725BE83C4E474ECBB548
[7] Hidalgo AA, Deeb KK, Pike JW, Johnson CS, Trump DL. Dexamethasone enhances 1alpha,25-dihydroxyvitamin D3 effects by increasing vitamin D receptor transcription. J Biol Chem. 2011;286(42):36228-36237. doi:10.1074/jbc.M111.244061
[8] Schütze N, Fritsche J, Ebert-Dümig R, et al. The selenoprotein thioredoxin reductase is expressed in peripheral blood monocytes and THP1 human myeloid leukemia cells--regulation by 1,25-dihydroxyvitamin D3 and selenite. Biofactors. 1999;10(4):329-338. doi:10.1002/biof.5520100403
[9] Martín Giménez, V.M., Inserra, F., Ferder, L. et al. Vitamin D deficiency in African Americans is associated with a high risk of severe disease and mortality by SARS-CoV-2. J Hum Hypertens (2020). https://doi.org/10.1038/s41371-020-00398-z
[10] Wang, Y et al. SARS-CoV-2 suppresses mRNA expression of selenoproteins associated with ferroptosis, ER stress and DNA synthesis. Preprint, 2020/07/31. 10.1101/2020.07.31.230243
https://www.researchgate.net/publication/343365020_SARS-CoV-2_suppresses_mRNA_expression_of_selenoproteins_associated_with_ferroptosis_ER_stress_and_DNA_synthesis[11] Steinbrenner H, Al-Quraishy S, Dkhil MA, Wunderlich F, Sies H. Dietary selenium in adjuvant therapy of viral and bacterial infections. Adv Nutr. 2015;6(1):73-82. Published 2015 Jan 15. doi:10.3945/an.114.007575
[12] Rayman MP, Winther KH, Pastor-Barriuso R, et al. Effect of long-term selenium supplementation on mortality: Results from a multiple-dose, randomised controlled trial. Free Radic Biol Med. 2018;127:46-54. doi:10.1016/j.freeradbiomed.2018.02.015
[13] Zhao Y, Yang M, Mao Z, et al. The clinical outcomes of selenium supplementation on critically ill patients: A meta-analysis of randomized controlled trials. Medicine (Baltimore). 2019;98(20):e15473. doi:10.1097/MD.0000000000015473
[14] Manzanares W, Lemieux M, Elke G, Langlois PL, Bloos F, Heyland DK. High-dose intravenous selenium does not improve clinical outcomes in the critically ill: a systematic review and meta-analysis. Crit Care. 2016;20(1):356. Published 2016 Oct 28. doi:10.1186/s13054-016-1529-5
[15] Xu X-M, Carlson BA, Mix H, Zhang Y, Saira K, Glass RS, et al. (2007) Biosynthesis of Selenocysteine on Its tRNA in Eukaryotes. PLoS Biol 5(1): e4. https://doi.org/10.1371/journal.pbio.0050004
[16] Thomson CD, Chisholm A, McLachlan SK, Campbell JM. Brazil nuts: an effective way to improve selenium status. Am J Clin Nutr. 2008;87(2):379-384. doi:10.1093/ajcn/87.2.379
https://academic.oup.com/ajcn/article/87/2/379/4633360
[17] Kieliszek M, Lipinski B. Selenium supplementation in the prevention of coronavirus infections (COVID-19) [published online ahead of print, 2020 May 24]. Med Hypotheses. 2020;143:109878. doi:10.1016/j.mehy.2020.109878
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7246001/