Claude AI May 2026

TL;DR

• Selenium has the strongest, most uniquely "selenium-only" evidence for: thyroid autoimmunity (Hashimoto's TPOAb reduction — Huwiler Thyroid 2024 SMD −0.96; Graves' orbitopathy QoL — Marcocci NEJM 2011), Keshan-disease cardiomyopathy, Kashin-Beck osteoarthropathy, mercury/methylmercury detoxification (Hg–Se equimolar binding to SELENOP), HIV viral-load suppression in Se-deficient patients (Hurwitz 2007), male-fertility/sperm parameters, and cancer prevention in selenium-deficient populations (Linxian "Factor D"; NPC trial low-tertile subgroup). Vitamin D has no equivalent evidence base in any of these.
• Selenium is null or harmful in selenium-replete Western populations. SELECT (n=35,533) showed no prostate-cancer benefit and the vitamin E arm showed +17% prostate cancer risk; Stranges Ann Intern Med 2007 (NPC top-tertile) HR 1.55 for T2D; SISPCT (n=1,089) 28-d mortality 28.3% selenium vs 25.5% placebo (NS). The dose-response is U-shaped, with optimal serum Se ~110–130 µg/L and diabetes risk rising above ~140 µg/L (Laclaustra Environ Health Perspect 2009).
• Practical bottom line: 100–200 µg/d Se-enriched yeast or SeMet is justified ONLY for individuals with baseline serum Se <100 µg/L (most of the UK, parts of Europe, New Zealand, Se-deficient China) or for specific indications (Hashimoto's, mild Graves' orbitopathy, HIV in deficient patients, pre-conception/pregnancy in Se-deficient areas). Do NOT routinely supplement Se-replete US adults. Practical ceiling: EFSA 2023 UL = 255 µg/d (lowered from 300, alopecia endpoint from SELECT); US IOM UL = 400 µg/d.

Key Findings

1. Biochemical foundation — why selenium is uniquely positioned

Selenium is incorporated as selenocysteine (the 21st amino acid) into ~25 human selenoproteins via UGA-recoding through SECIS elements. The functional families are:

• Glutathione peroxidases (GPX1–4, GPX6) — H₂O₂ and lipid-peroxide reduction; GPX4 is the gatekeeper against ferroptosis.
• Thioredoxin reductases (TXNRD1–3) — redox homeostasis, DNA synthesis support.
• Iodothyronine deiodinases (DIO1–3) — peripheral T4→T3 conversion (selenium's signature endocrine role).
• Selenoprotein P (SELENOP/SEPP1) — 10 selenocysteine residues; hepatic Se transporter; ALSO the major plasma Hg/Cd sequestering protein.
• SELENON (SEPN1) — mutations cause congenital myopathy.
• SELENOK, SELENOS — ER stress and inflammation regulation.

Vitamin D signals through one nuclear receptor (VDR) and one hydroxylated metabolite. Selenium acts through ~25 enzymes spanning antioxidant defense, thyroid hormone activation, and heavy-metal detoxification — domains where vitamin D has no enzymatic role.

2. Thyroid health — the strongest domain where selenium beats vitamin D

Hashimoto's thyroiditis (HT): - Huwiler et al. Thyroid 2024 meta-analysis (35 RCTs, 42 cohorts, 2,358 participants): selenium reduced TPOAb SMD −0.96 (95% CI −1.36 to −0.56); reduced TSH in patients not on levothyroxine; reduced MDA (oxidative stress). Adverse events comparable to placebo (OR 0.89, 95% CI 0.46–1.75). Certainty: moderate. - Hu et al. Medicine 2023 (21 RCTs, 1,610 patients): TPOAb reduced at 3 mo (SMD −0.46, p=0.001) and 6 mo (SMD −0.80, p=0.008); TgAb reduced at 3 mo (SMD −0.46) but not at 6 mo. - Founding RCT: Gärtner et al. JCEM 2002 (200 µg/d sodium selenite, 9 mo) — TPOAb fell 36.4% vs 12% placebo. - Mechanism: GPX3 and TXNRD protect thyrocytes from H₂O₂ generated during thyroid hormone synthesis; DIO1/DIO2 are selenoproteins required for T4→T3 conversion.

Graves' ophthalmopathy (mild):

• Marcocci et al. NEJM 2011;364:1920–1931 (EUGOGO; n=159; sodium selenite 100 µg BID = 200 µg/d × 6 mo vs pentoxifylline vs placebo):
  • QoL improvement: selenium 61% vs placebo 36% (p<0.001)
  • Ocular involvement improved: 61% vs 35% (p=0.01)
  • GO progression: 7% vs 26% (p=0.01)
  • Benefits persisted at 12 months (6 months post-treatment): appearance subscale 84% (95% CI 67–95); visual functioning 72% (53–86); overall QoL 81% (63–93)
• Vitamin D has no RCT with comparable Graves'-orbitopathy outcomes.
• Caveat: Recent work in Se-sufficient Korean patients suggests baseline-Se modification; the Marcocci effect likely requires marginal Se deficiency.

Graves' hyperthyroidism (gland): Kahaly et al. JCEM 2017 RCT — selenium adjunctive to antithyroid drug did NOT improve remission rates. Selenium's eye-disease effect does not extend to glandular Graves'.

3. Cancer — large effect in Se-deficient populations; NULL in Se-replete

Linxian General Population Trial — Blot JNCI 1993; n=29,584 in Se-deficient rural China:

• "Factor D" = 50 µg selenium-yeast + 30 mg α-tocopherol + 15 mg β-carotene daily × 5.25 y.
• Total mortality RR 0.91 (95% CI 0.84–0.99, p=0.03)
• Total cancer mortality RR 0.87 (95% CI 0.75–1.00)
• Stomach cancer mortality RR 0.79 (95% CI 0.64–0.99) — 21% reduction
• 10-year follow-up (Qiao JNCI 2009;101:507): all-cause HR 0.95 (0.91–0.99, p=0.009); gastric cancer HR 0.89 (0.79–1.00, p=0.043); effects concentrated in subjects <55 y at randomization.

Qidong liver cancer trial: Sodium selenite-fortified salt (50–80 µg/d) reduced primary liver cancer incidence 35.1% over 8 years in a selenium-deficient region; benefit lost after withdrawal.

NPC trial — Clark JAMA 1996 (n=1,312 US, 200 µg/d Se-yeast × 4.5 y mean):

• Primary endpoint (skin cancer): missed.
• Secondary endpoints (original 1996 report): total cancer incidence RR 0.63; total cancer mortality RR 0.50; prostate cancer RR 0.37; colorectal RR 0.42; lung RR 0.54.
• Duffield-Lillico BJU Int 2003: benefit concentrated in baseline plasma Se <122 ng/mL (lowest tertile); no benefit (and possibly harm) in top tertile.

SELECT — Lippman JAMA 2009 / Klein JAMA 2011 (n=35,533; selenium 200 µg/d L-selenomethionine ± vitamin E 400 IU; median 5.5 y supplementation, 7 y follow-up):

• NULL for prostate cancer: HR 1.04 (99% CI 0.87–1.24) for selenium; HR 1.13 (99% CI 0.95–1.35) for vitamin E (significant on extended follow-up, +17% risk).
• Absolute increase per 1,000 person-years: 1.6 (vit E), 0.8 (Se), 0.4 (Se+E) vs placebo.
• No effect on lung, colorectal, total cancer, all-cause mortality, or CVD events.
• Population mean baseline Se ~135 µg/L (Se-replete); SeMet vs Se-yeast formulation difference may matter.

Reconciliation: Selenium prevents cancer in selenium-deficient populations and at the lowest-Se tertile, but provides no benefit (and possibly slight harm in some subgroups) in Se-replete people.

4. Cardiovascular disease

• Keshan disease: endemic dilated cardiomyopathy in the selenium-deficient soil belt of China + Coxsackievirus B3 cofactor. Sodium selenite supplementation reduced incidence dramatically; no vitamin D analog exists.
• Beck Nature Medicine 1995 (Beck MA, Shi Q, Morris VC, Levander OA): Selenium-deficient mice convert benign coxsackievirus B3/0 into virulent myocarditic phenotype via 6 nucleotide mutations matching virulent strains; once mutated, virus causes disease even in Se-adequate hosts. Same phenomenon shown for influenza A and replicated with vitamin E deficiency.
• Primary prevention RCTs — Cochrane Rees 2021 (CD009671): No statistically significant effect of selenium on all-cause mortality (RR 0.97, 95% CI 0.88–1.08) or CVD mortality.
• Observational meta — Kuria Crit Rev Food Sci Nutr 2021: Higher selenium status associated with reduced CVD mortality RR 0.69 (95% CI 0.57–0.84) — but causal interpretation limited by confounding.
• Flores-Mateo AJCN 2006 CHD meta: inverse association between blood/toenail Se and CHD in observational studies; trial evidence weak.

5. Immune function / viral infections

HIV — selenium-specific RCT evidence:

• Hurwitz BE et al. Arch Intern Med 2007;167(2):148–154 (n=262 HIV+; 200 µg/d Se-yeast × 9 mo): significant suppression of HIV-1 viral burden; CD4 Δ = +27.9 ± 150.2 cells/µL in selenium arm; concluded selenium "can suppress the progression of HIV-1 viral burden and provide indirect improvement of CD4 count."
• Rwanda RCT (Kamwesiga 2015, n=300, 200 µg/d × 24 mo): No significant overall effect on CD4 decline or ART initiation — benefit appears restricted to severely deficient subgroups.
• Botswana (Baum 2013): Selenium alone null; selenium + multivitamin reduced HIV progression significantly.

COVID-19:

• Moghaddam/Schomburg Nutrients 2020;12:2098 (n=33 hospitalized German COVID patients, 166 serial sera): Mean serum Se 50.8 µg/L (severely deficient); 43.4% had Se <45.7 µg/L; 39.2% had SELENOP <2.56 mg/L. SELENOP-based mortality prediction AUC = 0.76; optimal cutpoint 3.1 mg/L (sensitivity 91.2%, specificity 50.8%). No definitive supplementation RCT has been completed.

Influenza: Se-deficient mice infected with influenza A/Bangkok/1/79 develop more severe pneumonitis and the virus acquires HA/NA/M-gene mutations rendering it more virulent in Se-adequate hosts.

6. Sepsis / critical illness — DISAPPOINTING for high-dose IV selenium

• SISPCT — Bloos JAMA Intern Med 2016 (n=1,089 severe sepsis/septic shock; 33 German ICUs; sodium selenite 1,000 µg loading + 1,000 µg/d continuous ≤21 d): 28-day mortality 28.3% selenium vs 25.5% placebo (NS). No benefit on any secondary outcome. 2025 secondary analysis by sepsis phenotype (Bloos, J Intensive Care): still no signal across α/β/γ/δ phenotypes.
• REDOXS — Heyland NEJM 2013 (n=1,223): Glutamine + antioxidants (incl. Se) — no mortality benefit; possible harm in renal-dysfunction subgroup (OR 3.39 for antioxidants in baseline renal dysfunction).
• Earlier meta-analyses (Manzanares 2012) suggested mortality reduction (RR 0.83); signal disappeared after SISPCT/REDOXS in the 2016 Manzanares update: "high-dose IV selenium does not improve clinical outcomes in the critically ill."
• Practical implication: Do not pursue high-dose pharmacologic IV selenium in sepsis.

7. Male fertility / sperm quality

• Zhao network meta-analysis 2022 (PMC8898892; 23 RCTs, n=1,917): Selenium improved sperm motility, concentration, and morphology vs placebo (L-carnitine ranked first; selenium close behind).
• Salas-Huetos Adv Nutr 2018 systematic review: significant beneficial effect of selenium on sperm concentration, motility, and morphology.
• Combined Se + vitamin E meta 2025 (8 RCTs, n=736): sperm vitality +23.24% (95% CI 16.66–29.82, p<0.00001); normal morphology +0.49% (95% CI 0.16–0.81, p=0.003); motility +13.56%; mixed effects on count/semen volume. Typical doses: 100–200 µg/d × 12–26 wk.
• Vitamin D has weaker and more inconsistent sperm-quality evidence.

8. Pregnancy outcomes

• Xu Preeclampsia meta-analysis Biol Trace Elem Res 2016 (13 obs + 3 RCTs): PE patients had lower blood Se (mean difference −6.47 µg/L, 95% CI −11.24 to −1.7, p=0.008); RCT RR 0.28 (95% CI 0.09–0.84, p=0.02) for selenium reducing PE incidence.
• 2022 update (26 studies, 1,855 PE cases): SMD −0.85 for serum Se in PE vs healthy pregnancy.
• Tara Taiwan J Obstet Gynecol 2010 Iranian RCT: 60 µg/d Se reduced PE incidence in low-Se pregnant women.

9. Cognitive / neurodegenerative

• Zhou Alzheimer meta 2023 (48 studies, n=5,334): AD patients have lower selenium than controls (SMD −0.41, 95% CI −0.64 to −0.17, p<0.001); RBC-Se difference even larger (SMD −1.54, 95% CI −2.97 to −0.12).
• Tamtaji 2022 systematic review (6 RCTs in MCI/AD): selenium supplementation improved GPX activity and some cognitive tests in MCI; effects in established AD inconsistent.
• Brazil-nut intervention studies in MCI show MMSE gains with concurrent rise in serum Se and GPX.
• Mechanism: brain preferentially retains selenium during deficiency; selenoproteins (especially GPX4 and SELENOP via the apoER2 brain receptor) protect against ferroptosis and Aβ/tau-associated oxidative damage.

10. Mercury / heavy-metal detoxification — SELENIUM-UNIQUE domain

• Hg has association constant Kₐ ~10⁴⁵ for selenium vs ~10³⁹ for sulfur — selenium is the strongest known physiologic mercury chelator.
• Yoneda & Suzuki Toxicol Appl Pharmacol 1997 (PMID 9144444): Hg and Se form equimolar (Hg–Se)ₙ complexes that bind to SELENOP at 1:1 Se:Hg.
• Suzuki PMID 9070209: Each SELENOP molecule (with 10 Sec residues) sequesters >1,000 (Hg–Se) units in serum.
• Liu Toxicol Appl Pharmacol 2018 (PMID 29704998): In methylmercury-poisoned rats, Se treatment increased SELENOP-bound Hg from 64.4% to 73% of total serum Hg, facilitating tissue extraction and redistribution for elimination.
• Se:Hg molar ratio >1 is the basis for revised FDA/EPA fish-advisory math (but molar ratio alone is insufficient — high-Hg fish still warrant limits).
• Vitamin D has no role in heavy-metal detoxification.

11. Kashin-Beck disease (KBD)

• Endemic osteoarthropathy in regions deficient in BOTH selenium and iodine (Tibet, NE→SW China, Siberia, North Korea).
• Moreno-Reyes NEJM 1998 (575 Tibetan children): iodine deficiency was the dominant risk factor; low Se alone was not (adjusted analysis).
• Moreno-Reyes AJCN 2003 RCT (n=324 children): once iodine corrected, selenium had no further effect on established KBD. Selenium may still help prevention, but iodine is the primary lever. Selenium-iodine interaction (DIO selenoproteins) is the mechanistic bridge.

12. Type 2 diabetes — PARADOXICAL HARM SIGNAL

• Stranges Ann Intern Med 2007 (NPC post-hoc): 200 µg/d Se-yeast increased T2D incidence HR 1.55, concentrated in baseline plasma Se >121.6 ng/mL (top tertile).
• SELECT (Lippman 2009): Small non-significant T2D excess in selenium arm (~7% RR).
• Vinceti meta-analysis Eur J Epidemiol 2018 (5 RCTs + 50 obs): RR for diabetes with Se supplementation 1.11 (95% CI 1.01–1.22); observational dose-response shows RR 3.6 (95% CI 1.4–9.4) for plasma Se 140 µg/L vs <45 µg/L.
• Vinceti Environ Res 2021 update: Nonlinear, risk rises above ~80 µg/d intake or ~120 µg/L blood Se; whole-blood/plasma/serum Se of 160 µg/L corresponds to RR 1.96 (95% CI 1.27–3.03) vs 90 µg/L.
• Implication: Do not supplement Se in Se-replete individuals; T2D risk is real and dose-dependent.

13. Other domains (briefer)

• Rheumatoid arthritis (Yu J Trace Elem Med Biol 2025 meta): Selenium reduced joint pain VAS (MD −12.68, 95% CI −19.08 to −6.28) and Ritchie index (MD −1.13); effects on CRP/ESR less robust; older Belgian RCT (Peretz) and German trial showed adjunctive benefit on top of standard care.
• Asthma, IBD: Mechanistic plausibility; selenium nanoparticle studies; no definitive RCTs.
• Skin cancer (BCC/SCC): NPC primary endpoint failed; selenium does NOT prevent nonmelanoma skin cancer.
• Hearing loss, depression: Limited observational data only.

14. The vitamin D vs. selenium comparison — explicit map

Condition	Selenium evidence	Vitamin D evidence	Winner
Hashimoto's TPOAb reduction	Huwiler 2024 meta SMD −0.96	Weak, mixed	Selenium
Graves' orbitopathy	Marcocci NEJM 2011 RCT (61% vs 36% QoL)	None comparable	Selenium
Keshan-disease cardiomyopathy	Definitive	None	Selenium-only
Kashin-Beck disease	Adjunct to iodine	None	Selenium-only domain
Mercury detoxification	Hg–Se equimolar SELENOP binding	None	Selenium-only
Coxsackie/influenza viral virulence	Beck mouse work (NM 1995)	None	Selenium-only
HIV viral suppression	Hurwitz 2007 RCT (CD4 Δ +27.9)	Limited, mostly null	Selenium in deficient populations
Sperm quality	Network meta positive	Weaker	Selenium
Preeclampsia	RCT pooled RR 0.28	Mostly null large RCT	Selenium
Cancer prevention (Se-deficient)	Linxian, NPC low-tertile	VITAL null overall	Selenium in deficient populations
Cancer prevention (Se-replete US)	SELECT null/harm	VITAL null	Tie (both null)
CVD primary prevention	Cochrane null	VITAL null	Tie
All-cause mortality	Mixed/null	Zhang BMJ 2019 (52 RCTs, n=75,454) RR 0.98 (0.95–1.02), NS; Zheng PLoS One 2013 long-FU subgroup RR 0.94 (0.90–0.98) — modest 6% reduction with >3-yr follow-up	Vitamin D modest edge
Bone/fracture	None	RDA-level D + calcium reduce hip fracture in deficient elderly	Vitamin D
Respiratory infections	Marginal	Martineau BMJ 2017 IPD (25 RCTs, n=10,933): AOR 0.88 (0.81–0.96, NNT=33); deficient <25 nmol/L AOR 0.30 (0.17–0.53)	Vitamin D
Multiple sclerosis	None	Strong MR evidence	Vitamin D
Type 2 diabetes	HARM signal (RR 1.11)	Tobias cholecalciferol-only meta (Tromsø, D2d, VITAL, FIND): HR 0.89 (95% CI 0.80–0.99, I²=0%), 11% RRR in progression from prediabetes; IPD of 3 prediabetes RCTs (D2d, Tromsø, DPVD; n=4,190) showed 15% RRR	Vitamin D clearly
Autoimmune disease incidence	Limited	Hahn BMJ 2022 VITAL substudy (n=25,871; 2,000 IU/d × 5.3 y): confirmed-AD HR 0.78 (95% CI 0.61–0.99, p=0.05) — 22% reduction	Vitamin D
Depression	Limited	Modest support	Vitamin D

Synergy candidates: Both essential for immunity; combined deficiency synergistically worsens autoimmune thyroid disease, COVID severity, and certain cancers. Small Se+D₃ pilot RCTs in autoimmune thyroid disease show additive TPOAb reduction beyond either alone.

15. Practical specifics

Forms (head-to-head):

• Selenomethionine (SeMet): ~90% absorbed (Swanson 1991; IOM 2000); urinary excretion ~60% (Burk CEBP 2006). Raises plasma Se highest because nonspecifically incorporated into albumin/methionine pool. Form used in SELECT.
• Sodium selenite/selenate: Inorganic; absorption ~50%; urinary excretion ~41%; preferentially enters functional selenoprotein synthesis. Used in Marcocci 2011 Graves' RCT and Gärtner 2002 Hashimoto's RCT.
• Selenium-enriched yeast: Mostly SeMet + methylselenocysteine (MSC) + selenocystine + inorganic. Used in NPC and Linxian. PRECISE pilot (Rayman Ann Intern Med 2011, n=501 UK elderly, baseline plasma Se 88.8 ng/g): dose-dependent rise to ~130 ng/g (100 µg/d), ~150 ng/g (200 µg/d), ~165 ng/g (300 µg/d) at 6 months.
• Methylselenocysteine (MSC): Yields methylselenol — most pro-apoptotic / anti-cancer metabolite in vitro; raises plasma Se less because not incorporated into the protein pool.

Optimal status (Hurst AJCN 2010; Xia AJCN 2010; Hoeflich/Schomburg Br J Nutr 2010):

• GPx3 plateau: plasma Se ~70–90 ng/mL
• SELENOP plateau: plasma Se ~110–125 ng/mL (~105 µg/d total intake)
• U-shaped sweet spot for general health: ~110–130 µg/L
• T2D risk inflection (NHANES, Laclaustra 2009): lowest risk at ~122 µg/L; rises above

RDA / UL / toxicity:

• US RDA (IOM 2000): 55 µg/d adults; pregnancy 60; lactation 70. EAR 45 µg/d.
• US UL: 400 µg/d (IOM 2000; NOAEL 800, LOAEL 910 from Yang 1989 Enshi China data, UF=2).
• EFSA 2023 UL: 255 µg/d (lowered from 300; alopecia endpoint from SELECT, LOAEL 330 µg/d, UF=1.3).
• Selenosis (hair loss, nail brittleness, garlic breath, peripheral neuropathy): intake ≥910 µg/d or whole-blood Se ≥1000 µg/L (Yang 1989; no cases below 1000 µg/L in 228 individuals).

Geographic baseline serum Se:

• Finland — fertilizer Se policy (Varo AJCN 1988; Aro 1995; Alfthan JTEMB 2015): pre-1984 mean intake ~25–40 µg/d, serum Se ~70 ng/mL; post-1984 sodium-selenate fertilizer enrichment → intake ~90–110 µg/d, serum Se ~110–120 ng/mL; stable plasma plateau ~1.40 µmol/L by 2015. Largest national-scale natural experiment in Se public health.
• UK (Rayman Br J Nutr 2008): intake fell from ~60 µg/d (1974) to ~30–40 µg/d (mid-1990s) after shift from North American to UK wheat.
• New Zealand Otago (Thomson Eur J Clin Nutr 1996): whole-blood Se ~60 ng/mL in 1970s–80s — among the lowest globally; rose to ~80 ng/mL by 1990s.
• US Eastern Seaboard (NPC baseline tertiles): low <106 ng/mL; high >123 ng/mL.
• China Keshan/KBD belts: very low.
• Food sources: Brazil nuts 68–91 µg per nut (huge variability — Brazilian-Amazon nuts much higher than Bolivian); 1–2 nuts/d suffices for most. Other major sources: seafood (tuna, sardines), organ meats, sunflower seeds.

Interactions:

• Iodine: synergistic in thyroid; selenium without iodine repletion can worsen hypothyroidism in severe iodine deficiency (selenium activates DIO and accelerates iodine consumption).
• Vitamin E: synergistic antioxidant; both deficient = enhanced Coxsackie virulence.
• Mercury, cadmium, arsenic, silver: sequestered by SELENOP-bound complexes.
• Statins: may blunt GPx response in some sub-analyses.

Details — research gaps and ongoing work

• Mendelian randomization on selenium remains limited; current MR signals modestly suggest a causal effect on some cancers but are inconsistent.
• Selenoprotein SNPs: GPX1 Pro198Leu, SELENOP rs7579, SEP15 1125G/A, SELENOS −105G/A modulate disease risk; could explain heterogeneity in trial responses.
• Personalized Se+SELENOP-guided supplementation is the future direction (Schomburg group, Berlin).
• Stratified trials (supplementing only those with baseline Se <100 µg/L) are needed and largely missing.
• COVID-19: no definitive supplementation RCT — only observational SELENOP/mortality data.
• PRECISE-Denmark long-term mortality trial (Cold 2018, n=491, up to 19-y follow-up) found no overall mortality benefit and a non-significant signal for increased mortality at 300 µg/d — adding caution to long-term high-dose use in Se-replete populations.

Recommendations (staged, decision-ready for VitaminDWiki)

Stage 1 — Recommend Se status assessment and 100–200 µg/d supplementation in:

• Hashimoto's thyroiditis patients with TPOAb >500 IU/mL or symptomatic on stable levothyroxine. Test SELENOP or serum Se; if <100 µg/L, give 100–200 µg/d Se-yeast or sodium selenite × 6 mo; retest TPOAb and Se status.
• Mild Graves' orbitopathy: sodium selenite 100 µg BID × 6 mo per the Marcocci protocol (Level 1 evidence).
• HIV patients in resource-limited settings or with documented serum Se <85 µg/L: 200 µg/d Se-yeast (Hurwitz protocol).
• Male infertility workup with documented low Se: 100–200 µg/d × 3–6 mo.
• Pregnancy in known Se-deficient regions (UK, parts of Europe, NZ): 60–100 µg/d.

Stage 2 — Do NOT routinely recommend Se supplementation in:

• US adults without measured deficiency (mean US plasma Se ~135 µg/L is already at or above optimal).
• Established T2D, prediabetes, or strong family history (T2D risk signal at >120 µg/L; Vinceti 2018 RR 1.11 for supplementation).
• Sepsis/critical illness (SISPCT, REDOXS null).
• Generic cancer prevention in Se-replete populations (SELECT null; vitamin E arm caused harm).

Stage 3 — Combine selenium with vitamin D when:

• Autoimmune thyroid disease (both have signal; combination plausibly additive based on small pilots).
• COVID-19 / respiratory infection prophylaxis in deficient populations (vitamin D Martineau IPD AOR 0.88; selenium plausible adjunct).
• Pregnancy in deficient regions (both micronutrients address preeclampsia and immune risk).

Benchmarks that should change the recommendation: - Baseline serum Se <90 µg/L → supplement. - Serum Se >140 µg/L → do NOT supplement (diabetes risk). - TPOAb >500 IU/mL on stable levothyroxine → consider 6-month Se trial. - Use EFSA 2023 UL of 255 µg/d as the practical ceiling, not the older 400 µg/d. - For long-term (>3 y) high-dose (300 µg/d) use, exercise caution given PRECISE-Denmark mortality signal.

Caveats

• The U-shaped curve is the dominant theme. Unlike vitamin D — where most US adults are deficient and benefit from supplementation — selenium supplementation in Se-replete Americans is unjustified and potentially harmful (T2D, possibly high-grade prostate cancer in some subgroups). The "more is better" intuition is wrong for selenium.
• SELECT used L-selenomethionine; NPC and Linxian used Se-yeast. Some experts argue Se-yeast has qualitatively different effects via MSC-derived methylselenol, particularly for cancer prevention. This is a real, unresolved scientific question that the field has not fully addressed.
• Most "selenium prevents disease X" headlines come from observational studies in heterogeneous populations; the RCT evidence is far more restrained, and effects appear to require baseline deficiency.
• Selenium-iodine interaction: Never supplement selenium aggressively in severely iodine-deficient individuals without iodine repletion — selenium can accelerate iodine consumption via DIO activation and worsen hypothyroidism.
• Mercury/Se:Hg molar ratio: The simplified "ratio >1 = safe" math is incomplete; high-Hg fish should still be limited even when Se-rich, and tubulin-binding mechanisms of Hg toxicity are not fully neutralized by Se.
• Coxsackie/influenza viral-mutation work is exclusively rodent; human extrapolation remains inferential — biologically plausible but not proven in humans.
• Marcocci's Graves' orbitopathy finding may not generalize to Se-sufficient populations; recent Korean data suggest the effect requires marginal deficiency.
• High-dose IV selenium in sepsis is dead — SISPCT and REDOXS were definitive; do not resurrect this paradigm without major mechanistic justification.