Scientific opinion on the tolerable upper intake level for selenium

View full-text article in PMC

. 2023 Jan 20;21(1):e07704. doi: 10.2903/j.efsa.2023.7704

Study name Country Reference Follow‐up Study design Funding	Original cohort (N total) Exclusion criteria Study population (n, sex and age at baseline^(a))	Ascertainment of outcome	Exposure groups^(a) n/person‐years	Incident cases	Model covariates	Results
Hortega Study Spain Galan‐Chilet et al. (2017) 13.2 yr (median) Prospective cohort Public	N = 1,502 Population sampled: adults in the catchment area of the Rio Hortega University Hospital Excluded: missing Se data, smoking or covariate variables; prevalent diabetes; loss to follow up. n = 1,234 Sex (% F): 51 Ethnicity: Caucasian Age (yr): 15–85	Incident cases of T2DM identified based on review of health records (including primary care, hospitalisation and mortality records) when: diagnosed with T2DM by a physician; using diabetes medications; with fasting plasma glucose > 126 mg/dl; or with HbA1c > 6.5%	N per tertile of plasma Se (μg/L) T1 < 76.3: 431 T2 76.3–94.2: 376 T3 ≥ 94.2: 397	n, per tertile: T1: 19 T2: 30 T3: 26	Model 1: unadjusted Model 2: Age, gender and education Model 3: Age, gender, education, urine cotinine levels, smoking status and alcohol intake	HR (95% CI) for T2DM Model 1 T1: 1 T2: 1.78 (0.98, 3.26) T3: 1.78 (0.97, 3.28) p lineal: 0.16 Model 2 T1: 1 T2: 1.83 (1.00, 3.35) T3: 1.86 (1.01, 3.41) *p lineal:* 0.13 Model 3 T1: 1 T2: 1.76 (0.96, 3.22) T3: 1.80 (0.98, 3.31) *p lineal:* 0.15
ULSAM Sweden Gao et al. (2014) 10–27 yr Prospective cohort Mixed	N = 2,322 Population sampled: general male population, aged 50 yr Excluded: missing Se, smoking or covariate variables; prevalent diabetes; loss to follow up. n = 1,925 (n = 1,539 at 60 yr; n = 1,024 at 70 yr; n = 656 at 77 yr) Sex: M Ethnicity: Caucasian Age (yr): 49.1–50.2	Incident cases of T2DM defined as elevated fasting glucose levels and/or use of anti‐diabetic medicine at follow up visits. Elevated glucose levels defined as FPG ≥ 6.1 mmol/l at 60 yr and FPG ≥ 7.0 mmol/l at 70 yr and 77 yr	N per tertile of plasma Se (μg/L), at baseline/60 yr/70 yr/77 yr T1 61.4 ± 6.8: 670/516/326/206 T2 75.1 ± 3.1: 620/504/347/216 T3 91.1 ± 10.2: 635/519/351/234	n, per tertile at 60 yr/70 yr/77 yr T1: 15/24/27 T2: 15/33/34 T3: 23/31/30	Age at baseline, BMI, cigarette smoking, leisure time physical activity and education level	OR (95% CI) for T2DM At 60 yr T1: 1 T2: 0.91 (0.43, 1.95) T3: 1.28 (0.64, 2.61) p for trend : 0.437 At 70 yr T1: 1 T2: 1.31 (0.73, 2.36) T3: 1.25 (0.68, 2.27) p _trend: 0.497 At 77 yr T1: 1 T2: 1.16 (0.65, 2.08) T3: 0.97 (0.54, 1.75) p _trend: 0.880
NHS & HPFS USA Park et al. (2012) NHS 26 yr HPFS 22 yr Prospective cohort Public	NHS N = 121,700 HPFS N = 51,529 Population sampled: registered nurses (NHS) and male health professionals (HPFS) Excluded: no toenail Se data; missing covariate information; free of prevalent T2DM or heart disease at baseline (defined as date of toenail sampling) NHS n = 3,630 HPFS n = 3,535 NHS Sex: F HPFS Sex: M Ethnicity: Majority Caucasian Age (yr): NHS: 30–55 HPFS: 40–75	Self‐reported incident cases of T2DM identified via biennial questionnaires; confirmed by a validated supplementary questionnaire. Cases before 1998 defined by National Diabetes Data Group criteria^(b) and cases after 1998 defined by American Diabetes Association criteria (1997).	N per quintile of toenail Se (μg/g) NHS Q1 0.62 ± 0.04: 730 Q2 0.70 ± 0.02: 727 Q3 0.75 ± 0.02: 729 Q4 0.82 ± 0.02: 714 Q5 0.96 ± 0.10: 730 HPFS Q1 0.66 ± 0.05: 699 Q2: 0.76 ± 0.02: 700 Q3: 0.82 ± 0.02: 726 Q4: 0.90 ± 0.03: 706 Q5: 1.07 ± 0.12: 704	780 (NHS & HPFS) N cases per quintile NR	Model 1: age, sex, future case–control status. Model 2: Model 1 + geographic region, smoking, alcohol intake, physical activity, BMI, Se supplement use, and multivitamin use. Model 3: Model 2 + consumption of total energy, the ratio of polyunsaturated to saturated fats, trans fat, whole grains, and coffee.	HR (95% CI) for T2DM NHS Model 1 Q1: 1 Q2: 0.88 (0.65, 1.18) Q3: 0.92 (0.69, 1.23) Q4: 0.85 (0.63, 1.14) Q5: 0.96 (0.72, 1.28) p _trend: 0.8 Model 2 Q1: 1 Q2: 0.82 (0.61, 1.11) Q3: 0.81 (0.60, 1.09) Q4: 0.70 (0.51, 0.95) Q5: 0.80 (0.59, 1.07) p _trend: 0.1 HPFS Model 1 Q1: 1 Q2: 1.07 (0.78, 1.48) Q3: 0.78 (0.52, 1.06) Q4: 0.74 (0.54, 1.13) Q5: 0.70 (0.49, 1.00) p _trend: 0.01 Model 2 Q1: 1 Q2: 1.08 (0.78, 1.50) Q3: 0.78 (0.54, 1.13) Q4: 0.78 (0.54, 1.13) Q5: 0.75 (0.51, 1.10) p _trend: 0.047 Results for Model 3 were similar
ORDET Italy Stranges et al. (2010) 16 y (median) Prospective cohort Mixed	N = 10,786 Population sampled: general population Excluded: T2DM at baseline; missing data; ratio of total energy intake to basal metabolic rate at either extreme of the distribution (cut‐offs 0.5 and 99.5 percentiles); death during follow‐up from causes other than T2DM; loss to follow‐up. n = 7,182 Sex: F Ethnicity: Caucasian Age (yr): 34–70	Incident cases self‐reported via telephone interviews (diagnosis of T2DM from a physician or taking medications for treatment of diabetes) or identified via prescription for insulin or oral hypoglycaemic medication (by linkage with regional prescription drug database) or identified via hospital discharge record (by linkage with medical discharge records)	*N per quintile of Se intake (by SFFQ) (μg/day, mean)** Q1 41.7: 1,437 Q2 50.2: 1,436 Q3 55.7: 1,437 Q4 62.0: 1,436 Q5 75.1: 1,436 *Adjusted for total energy intake	n, per quintile Q1: 32 Q2: 42 Q3: 45 Q4: 55 Q5: 79	Model 1: age, education, menopausal status Model 2: age, education, menopausal status, BMI, smoking, alcohol intake, energy intake, saturated/polyunsaturated fatty acids ratio, animal proteins, total carbohydrates, and weight change between baseline and follow‐up examinations	OR (95% CI) for T2DM Model 1 Q1: 1 Q2: 1.31 (0.82, 2.09) Q3: 1.38 (0.87, 2.19) Q4: 1.74 (1.12, 2.72) Q5: 2.64 (1.73, 4.01) p _trend: < 0.001 Model 2 Q1: 1 Q2: 1.42 (0.87, 2.34) Q3: 1.43 (0.86, 2.38) Q4: 1.65 (0.98, 2.78) Q5: 2.39 (1.32, 4.32) p _trend: 0.005 OR (95% CI), per 10 μg/day increase in Se Model 1: 1.29 (1.17, 1.41) Model 2: 1.29 (1.10, 1.52) The linearity of the relationship between selenium intake and risk of diabetes confirmed in spline regression models (not shown)
ORDET Italy Vinceti et al. (2015b) 16 y (median) Nested case–control Mixed	N = 10,786 Population sampled: general population Excluded: history of cancer, bilateral ovariectomy, chronic or acute liver disease or who had received hormone therapy in the 3 mo before recruitment; baseline T2DM; loss to follow up; death during follow‐up from causes other than T2DM n = 621 cases: 226 controls: 395 (matched for age) Sex: F Ethnicity: Caucasian Age (yr): 35–70	Incident cases self‐reported via telephone interviews (diagnosis of T2DM from a physician or taking medications for treatment of diabetes) or identified via prescription for insulin or oral hypoglycemic maedication (by linkage with regional prescription drug database) or identified via hospital discharge record (by linkage with medical discharge records).	N per category of toenail Se (μg/g) Q1 < 0.54: 190 Q2 ≥ 0.54–< 0.59: 150 Q3 ≥ 0.59–< 0.66: 147 Q4 ≥ 0.66: 134	n, per category Q1: 83 Q2: 58 Q3: 47 Q4: 38	Model 1: BMI Model 2: BMI, education, smoking, coffee and alcohol consumption	OR (95% CI) for T2DM Model 1 Q1: 1 Q2: 1.04 (0.59, 1.84) Q3: 0.67 (0.36, 1.24) Q4: 1.08 (0.45, 2.58) p _trend: 0.275 Model 2 Q1: 1 Q2: 1.09 (0.61, 1.96) Q3: 0.71 (0.38, 1.34) Q4: 1.14 (0.46, 2.80) p _trend: 0.362
Moli‐sani cohort Italy Vinceti et al. (2021) 8.2 yr (median) Prospective cohort Mixed	N = 24,325 Population sampled: general population Excluded: previous diagnosis of T2DM and/or treated with hypoglycaemic drugs; missing data; implausible energy intakes (4,000 kcal/day in men and > 3,500 kcal/day in women); unreliable dietary or medical questionnaires; lost to follow‐up n = 21,335 Sex (% F): 53 Ethnicity: Caucasian Age (yr): ≥35	Hospitalisation for T2DM, ascertained through direct linkage with the regional hospital discharge registry	Intake Se estimate (by SFFQ) (μg/day, median (SD)) F/M Q1: 38.8 (6.5)/43.6 (6.6) Q2: 51.2 (3.0)/55.9 (3.0) Q3: 61.7 (3.5)/67.1 (3.7) Q4: 78.6 (13.8)/87.8 (14.5) n/person‐years, sex combined: Q1: 5,333/42,481 Q2: 5,332/43,359 Q3: 5,334/43,757 Q4: 5,336/44,826	n, per quartile Q1: 33 Q2: 29 Q3: 30 Q4: 43	Model 1: age and sex Model 2: Age, sex, and energy intake. Model 3: Model 2 + education, housing, place of residence, smoking status, and physical activity. Model 4: Model 3+ Mediterranean diet score Model 5: Model 4 + BMI	HR (95% CI) for T2DM Model 1 Q1: 1 Q2: 0.94 (0.57, 1.55) Q3: 0.99 (0.60, 1.63) Q4: 1.44 (0.91, 2.29) Model 2 Q1: 1 Q2: 1.06 (0.63, 1.77) Q3: 1.21 (0.71, 2.08) Q4: 1.96 (1.12, 3.44) Model 3 Q1: 1 Q2: 1.07 (0.64, 1.79) Q3: 1.23 (0.71, 2.10) Q4: 1.99 (1.14, 3.49) Model 4 Q1: 1 Q2: 1.05 (0.63, 1.77) Q3: 1.20 (0.69, 2.07) Q4: 1.92 (1.09, 3.40) Model 5 Q1: 1 Q2: 1.01 (0.60, 1.70) Q3: 1.13 (0.66, 1.96) Q4: 1.75 (0.99, 3.10)
DFTJ cohort China Yuan et al. (2018) 4.6 yr Nested case–control Private	N = 27,009 Population sampled: retired employees of the Dongfeng Motor Corporation Excluded: baseline cardiovascular disease or cancer; insufficient blood samples; diabetes at baseline n = 2,078 Cases: 1,039 Controls: 1,039 (matched by age) Sex (% F): 55 Ethnicity: Asian Age (yr, mean): 63	Incident T2DM identified if FPG ≥ 7.0 mmoL/L; or haemoglobin A1c (HbA1c) ≥ 6.5%; or self‐reported physician diagnosis of diabetes or use of anti‐diabetic medication during the follow‐up visits	N per category of plasma Se (μg/L) Q1 < 54.10: 470 Q2: 54.10–61.71: 488 Q3: 61.71–72.16: 587 Q4: > 72.16: 533	n, per category Q1: 210 Q2: 229 Q3: 326 Q4: 274	BMI, smoking status, drinking status, education, physical activity, hypertension, hyperlipidemia, family history of diabetes, and eGFR For subgroups analyses: age, sex, BMI, smoking status, alcohol intake status, education, physical activity, hypertension, hyperlipidemia, family history of diabetes, and eGFR	OR (95% CI) for T2DM Q1: 1 Q2: 1.08 (0.80, 1.46) Q3: 1.45 (1.09, 1.93) Q4: 1.27 (0.93, 1.74) p _trend: 0.05 < 65 yr Q1: 1 Q2: 1.06 (0.74, 1.53) Q3: 2.00 (1.42, 2.83) Q4: 1.36 (0.95, 1.94) p _trend: 0.02 ≥ 65 yr Q1: 1 Q2: 1.15 (0.73, 1.80) Q3: 1.40 (0.90, 2.18) Q4: 1.44 (0.92, 2.24) p _trend: 0.08 Males Q1: 1 Q2: 1.03 (0.69, 1.55) Q3: 1.34 (0.90, 2.00) Q4: 1.39 (0.93, 2.07) p for trend : 0.06 Females Q1: 1 Q2: 1.14 (0.77, 1.68) Q3: 1.64 (1.13, 2.39) Q4: 1.52 (1.03, 2.23) p _trend: 0.02
EPIC – Potsdam Study Germany Cabral et al. (2021) 6.6 yr Nested‐case cohort Public	N = 27,548 Population sampled: general population, aged 35–64 yr Excluded: insufficient/no serum; unclear disease status; prevalent T2DM, cancer, myocardial infarction or stroke at baseline; incomplete follow‐up information n = 2,741 Cases: 705 Random sub‐cohort: 2,090 (overlap: 54) Sex (% F): 62.55 Ethnicity: Caucasian Age (yr, median): 49	Self‐reported incident cases (diabetes diagnosis, diabetes‐relevant medication, or dietary treatment due to diabetes) every 2 years; confirmed by questionnaires to the diagnosing physician	Serum Se (μg/L, median (IQR)) 80.0 (19.1)	705 cases	Model 1: age, sex, educational attainment, BMI, WC, smoking status, physical activity, alcohol intake, vitamin and mineral preparations, hypertension, lipid‐lowering medication, and Mediterranean diet score. Model 2: model 1 + serum trace elements Manganese, Iron, Copper, Zinc, and Iodine	HR (95%) for T2DM, per SD increase in serum Se Model 1: 1.26 (1.12, 1.41) Model 2: 1.19 (1.06, 1.34)
CSPPT Zhang et al. (2019) China 4.5 yr Prospective Cohort (sub‐sample of a trial) Public	N = 20,702 Population sampled: Hypertensive patients, aged 45–75 yr Excluded: history of CVD; diabetic at baseline n = 2,367 (sub‐sample from CSPPT trial) Sex (% F): 53 Ethnicity: Asian Age (yr): 61.4 ± 7.6	Physician‐diagnosed diabetes or use of glucose‐lowering drugs, or new onset (FPG ≥ 126.0 mg/dL) at the exit visit.	Plasma Se (μg/L) 84.8 (21.1) n per quartile of plasma Se Q1 < 71.0: 592 Q2 71.0–< 82.3: 591 Q3 82.3‐< 94.8: 592 Q4 ≥ 94.8: 592	n cases, per quartile Q1 (Ref): 63 Q2: 63 Q3: 58 Q4: 86	Model 1: Unadjusted Model 2: age, sex, study center, study treatment group, BMI, MTHFR C677 T genotype, smoking, alcohol drinking, family history of diabetes, SBP, fasting glucose, total cholesterol, HDL, triglycerides, creatinine, folate at baseline, and time‐averaged SBP during treatment period.	OR (95% CI) for T2DM Model 1 Q1 (Ref): 1 Q2: 1.00 (0.69, 1.45), p = 0.992 Q3: 0.91 (0.63, 1.33), p = 0.632 Q4: 1.43 (1.01, 2.02), p = 0.045 Model 2 Q1 (Ref): 1 Q2: 0.99 (0.67, 1.47), p = 0.968 Q3: 0.87 (0.59, 1.30), i = 0.509 Q4: 1.29 (0.89, 1.89), p = 0.184
Jinchang Cheng et al. (2022) China 5.8 yr Nested case–control Public	N = 48,001 Population sampled: employees of the Jinchuan Nonferrous Metals Corporation Excluded: diabetes and pre‐diabetes patients at baseline n = 1,244 Cases: 622 Controls: 622 (matched according to age, sex and follow‐up time) Sex (% F) = 26 Ethnicity = Asian Age (yr): 47.2 ± 13.9	Self‐reported T2DM, or FPG > 7.0 mmol/L, or glucose tolerance test > 11.0 mmol/L, or with explicit inpatient medical history, or diabetes pharmacotherapy history at any follow up biannual examination	n per quartile of Serum Se (μg/L) Q1 < 85.45: 278 Q2 85.45–92.51: 282 Q3 92.52–103.43: 331 Q4 ≥ 103.44: 353	n cases, per quartile Q1 (Ref.): 123 Q2: 126 Q3: 176 Q4: 197	Model 1: unadjusted Model 2: age at diagnosis, family history of diabetes, physical exercise, smoking index, lifetime total alcohol intake, triglyceride, high‐density lipoprotein, low‐density lipoprotein, and hypertension status. Model 3: Model 2 + y serum Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium	OR (95% CI) for T2DM Model 1 Q1 (Ref): 1 Q2: 1.05 (0.75, 1.47) Q3: 1.42 (1.03, 1.96) Q4: 1.70 (1.01, 2.02) p_trend = < 0.01 Model 2 Q1 (Ref): 1 Q2: 1.26 (0.86, 1.85) Q3: 1.62 (1.17, 2.35) Q4: 1.79 (1.21, 2.64) p_trend = 0.12 Model 3 Q1 (Ref): 1 Q2: 1.29 (0.82, 2.03) Q3: 1.62 (1.05, 2.51) Q4: 1.64 (1.02, 2.65) p _trend = 0.20

Study name

Country

Reference

Follow‐up

Study design

Funding

Original cohort (N total)

Exclusion criteria

Study population (n, sex and age at baseline^(a))

Ascertainment of outcome

Exposure groups^(a)

n/person‐years

Incident cases

Model covariates

Results

Hortega Study

Spain

Galan‐Chilet et al. (2017)

13.2 yr (median)

Prospective cohort

Public

N = 1,502

Population sampled: adults in the catchment area of the Rio Hortega University Hospital

Excluded: missing Se data, smoking or covariate variables; prevalent diabetes; loss to follow up.

n = 1,234

Sex (% F): 51

Ethnicity: Caucasian

Age (yr): 15–85

Incident cases of T2DM identified based on review of health records (including primary care, hospitalisation and mortality records) when: diagnosed with T2DM by a physician; using diabetes medications; with fasting plasma glucose > 126 mg/dl; or with HbA1c > 6.5%

N per tertile of plasma Se (μg/L)

T1 < 76.3: 431

T2 76.3–94.2: 376

T3 ≥ 94.2: 397

n, per tertile: T1: 19

T2: 30

T3: 26

Model 1: unadjusted

Model 2: Age, gender and education

Model 3: Age, gender, education, urine cotinine levels, smoking status and alcohol intake

HR (95% CI) for T2DM

Model 1

T1: 1

T2: 1.78 (0.98, 3.26)

T3: 1.78 (0.97, 3.28)

p lineal: 0.16

Model 2

T1: 1

T2: 1.83 (1.00, 3.35)

T3: 1.86 (1.01, 3.41)

p lineal: 0.13

Model 3

T1: 1

T2: 1.76 (0.96, 3.22)

T3: 1.80 (0.98, 3.31)

p lineal: 0.15

ULSAM

Sweden

Gao et al. (2014)

10–27 yr

Prospective cohort

Mixed

N = 2,322

Population sampled: general male population, aged 50 yr

Excluded: missing Se, smoking or covariate variables; prevalent diabetes; loss to follow up.

n = 1,925

(n = 1,539 at 60 yr;

n = 1,024 at 70 yr;

n = 656 at 77 yr)

Sex: M

Ethnicity: Caucasian

Age (yr): 49.1–50.2

Incident cases of T2DM defined as elevated fasting glucose levels and/or use of anti‐diabetic medicine at follow up visits. Elevated glucose levels defined as FPG ≥ 6.1 mmol/l at 60 yr and FPG ≥ 7.0 mmol/l at 70 yr and 77 yr

N per tertile of plasma Se (μg/L), at baseline/60 yr/70 yr/77 yr

T1 61.4 ± 6.8: 670/516/326/206

T2 75.1 ± 3.1: 620/504/347/216

T3 91.1 ± 10.2: 635/519/351/234

n, per tertile at 60 yr/70 yr/77 yr

T1: 15/24/27

T2: 15/33/34

T3: 23/31/30

Age at baseline, BMI, cigarette smoking, leisure time physical activity and education level

OR (95% CI) for T2DM

At 60 yr

T1: 1

T2: 0.91 (0.43, 1.95)

T3: 1.28 (0.64, 2.61)

p for trend : 0.437

At 70 yr

T1: 1

T2: 1.31 (0.73, 2.36)

T3: 1.25 (0.68, 2.27)

p _trend: 0.497

At 77 yr

T1: 1

T2: 1.16 (0.65, 2.08)

T3: 0.97 (0.54, 1.75)

p _trend: 0.880

NHS & HPFS

USA

Park et al. (2012)

NHS 26 yr

HPFS 22 yr

Prospective cohort

Public

NHS N = 121,700

HPFS N = 51,529

Population

sampled: registered nurses (NHS) and male health professionals (HPFS)

Excluded: no toenail Se data; missing covariate information; free of prevalent T2DM or heart disease at baseline (defined as date of toenail sampling)

NHS n = 3,630

HPFS n = 3,535

NHS Sex: F

HPFS Sex: M

Ethnicity: Majority Caucasian

Age (yr):

NHS: 30–55

HPFS: 40–75

Self‐reported incident cases of T2DM identified via biennial questionnaires; confirmed by a validated supplementary questionnaire. Cases before 1998 defined by National Diabetes Data Group criteria^(b) and cases after 1998 defined by American Diabetes Association criteria (1997).

N per quintile of toenail Se (μg/g)

NHS

Q1 0.62 ± 0.04: 730

Q2 0.70 ± 0.02: 727

Q3 0.75 ± 0.02: 729

Q4 0.82 ± 0.02: 714

Q5 0.96 ± 0.10: 730

HPFS

Q1 0.66 ± 0.05: 699

Q2: 0.76 ± 0.02: 700

Q3: 0.82 ± 0.02: 726

Q4: 0.90 ± 0.03: 706

Q5: 1.07 ± 0.12: 704

780 (NHS & HPFS)

N cases per quintile NR

Model 1:

age, sex, future case–control status.

Model 2:

Model 1 + geographic region, smoking, alcohol intake, physical activity, BMI, Se supplement use, and multivitamin use.

Model 3:

Model 2 + consumption of total energy, the ratio of polyunsaturated to saturated fats, trans fat, whole grains, and coffee.

HR (95% CI) for T2DM

NHS

Model 1

Q1: 1

Q2: 0.88 (0.65, 1.18)

Q3: 0.92 (0.69, 1.23)

Q4: 0.85 (0.63, 1.14)

Q5: 0.96 (0.72, 1.28)

p _trend: 0.8

Model 2

Q1: 1

Q2: 0.82 (0.61, 1.11)

Q3: 0.81 (0.60, 1.09)

Q4: 0.70 (0.51, 0.95)

Q5: 0.80 (0.59, 1.07)

p _trend: 0.1

HPFS

Model 1

Q1: 1

Q2: 1.07 (0.78, 1.48)

Q3: 0.78 (0.52, 1.06)

Q4: 0.74 (0.54, 1.13)

Q5: 0.70 (0.49, 1.00)

p _trend: 0.01

Model 2

Q1: 1

Q2: 1.08 (0.78, 1.50)

Q3: 0.78 (0.54, 1.13)

Q4: 0.78 (0.54, 1.13)

Q5: 0.75 (0.51, 1.10)

p _trend: 0.047

Results for Model 3 were similar

ORDET

Italy

Stranges et al. (2010)

16 y (median)

Prospective cohort

Mixed

N = 10,786

Population sampled: general population

Excluded: T2DM at baseline; missing data; ratio of total energy intake to basal metabolic rate at either extreme of the distribution (cut‐offs 0.5 and 99.5 percentiles); death during follow‐up from causes other than T2DM; loss to follow‐up.

n = 7,182

Sex: F

Ethnicity: Caucasian

Age (yr): 34–70

Incident cases self‐reported via telephone interviews (diagnosis of T2DM from a physician or taking medications for treatment of diabetes) or identified via prescription for insulin or oral hypoglycaemic medication (by linkage with regional prescription drug database) or identified via hospital discharge record (by linkage with medical discharge records)

N per quintile of Se intake* (by SFFQ) (μg/day, mean)

Q1 41.7: 1,437

Q2 50.2: 1,436

Q3 55.7: 1,437

Q4 62.0: 1,436

Q5 75.1: 1,436

*Adjusted for total energy intake

n, per quintile

Q1: 32

Q2: 42

Q3: 45

Q4: 55

Q5: 79

Model 1:

age, education, menopausal status

Model 2:

age, education, menopausal status, BMI, smoking, alcohol intake, energy intake, saturated/polyunsaturated fatty acids ratio, animal proteins, total carbohydrates, and weight change between baseline and follow‐up examinations

OR (95% CI) for T2DM

Model 1

Q1: 1

Q2: 1.31 (0.82, 2.09)

Q3: 1.38 (0.87, 2.19)

Q4: 1.74 (1.12, 2.72)

Q5: 2.64 (1.73, 4.01)

p _trend: < 0.001

Model 2

Q1: 1

Q2: 1.42 (0.87, 2.34)

Q3: 1.43 (0.86, 2.38)

Q4: 1.65 (0.98, 2.78)

Q5: 2.39 (1.32, 4.32)

p _trend: 0.005

OR (95% CI), per 10 μg/day increase in Se

Model 1: 1.29 (1.17, 1.41)

Model 2: 1.29 (1.10, 1.52)

The linearity of the relationship

between selenium intake and risk of diabetes confirmed in spline regression models (not shown)

ORDET

Italy

Vinceti et al. (2015b)

16 y (median)

Nested case–control

Mixed

N = 10,786

Population sampled: general population

Excluded:

history of cancer, bilateral ovariectomy, chronic or acute liver disease or who had received hormone therapy in the 3 mo before recruitment; baseline T2DM; loss to follow up; death during follow‐up from causes other than T2DM

n = 621

cases: 226

controls: 395 (matched for age)

Sex: F

Ethnicity: Caucasian

Age (yr): 35–70

Incident cases self‐reported via telephone interviews (diagnosis of T2DM from a physician or taking medications for treatment of diabetes) or identified via prescription for insulin or oral hypoglycemic maedication (by linkage with regional prescription drug database) or identified via hospital discharge record (by linkage with medical discharge records).

N per category of toenail Se (μg/g)

Q1 < 0.54: 190

Q2 ≥ 0.54–< 0.59: 150

Q3 ≥ 0.59–< 0.66: 147

Q4 ≥ 0.66: 134

n, per category

Q1: 83

Q2: 58

Q3: 47

Q4: 38

Model 1: BMI

Model 2:

BMI, education, smoking, coffee and alcohol consumption

OR (95% CI) for T2DM

Model 1

Q1: 1

Q2: 1.04 (0.59, 1.84)

Q3: 0.67 (0.36, 1.24)

Q4: 1.08 (0.45, 2.58)

p _trend: 0.275

Model 2

Q1: 1

Q2: 1.09 (0.61, 1.96)

Q3: 0.71 (0.38, 1.34)

Q4: 1.14 (0.46, 2.80)

p _trend: 0.362

Moli‐sani cohort

Italy

Vinceti et al. (2021)

8.2 yr (median)

Prospective cohort

Mixed

N = 24,325

Population sampled: general population

Excluded: previous diagnosis of T2DM and/or treated with hypoglycaemic drugs; missing data; implausible energy intakes (4,000 kcal/day in men and > 3,500 kcal/day in women); unreliable dietary or medical questionnaires; lost to follow‐up

n = 21,335

Sex (% F): 53

Ethnicity: Caucasian

Age (yr): ≥35

Hospitalisation for T2DM, ascertained through direct linkage with the regional hospital discharge registry

Intake Se estimate (by SFFQ) (μg/day, median (SD))

F/M

Q1: 38.8 (6.5)/43.6 (6.6)

Q2: 51.2 (3.0)/55.9 (3.0)

Q3: 61.7 (3.5)/67.1 (3.7)

Q4: 78.6 (13.8)/87.8 (14.5)

n/person‐years,

sex combined:

Q1: 5,333/42,481

Q2: 5,332/43,359

Q3: 5,334/43,757

Q4: 5,336/44,826

n, per quartile

Q1: 33

Q2: 29

Q3: 30

Q4: 43

Model 1:

age and sex

Model 2:

Age, sex, and energy intake.

Model 3:

Model 2 + education, housing, place of residence, smoking status, and physical activity.

Model 4:

Model 3+ Mediterranean diet score

Model 5:

Model 4 + BMI

HR (95% CI) for T2DM

Model 1

Q1: 1

Q2: 0.94 (0.57, 1.55)

Q3: 0.99 (0.60, 1.63)

Q4: 1.44 (0.91, 2.29)

Model 2

Q1: 1

Q2: 1.06 (0.63, 1.77)

Q3: 1.21 (0.71, 2.08)

Q4: 1.96 (1.12, 3.44)

Model 3

Q1: 1

Q2: 1.07 (0.64, 1.79)

Q3: 1.23 (0.71, 2.10)

Q4: 1.99 (1.14, 3.49)

Model 4

Q1: 1

Q2: 1.05 (0.63, 1.77)

Q3: 1.20 (0.69, 2.07)

Q4: 1.92 (1.09, 3.40)

Model 5

Q1: 1

Q2: 1.01 (0.60, 1.70)

Q3: 1.13 (0.66, 1.96)

Q4: 1.75 (0.99, 3.10)

DFTJ cohort

China

Yuan et al. (2018)

4.6 yr

Nested case–control

Private

N = 27,009

Population sampled: retired employees of the Dongfeng Motor Corporation

Excluded: baseline cardiovascular disease or cancer; insufficient blood samples; diabetes at baseline

n = 2,078

Cases: 1,039

Controls: 1,039 (matched by age)

Sex (% F): 55

Ethnicity: Asian

Age (yr, mean): 63

Incident T2DM identified if FPG ≥ 7.0 mmoL/L; or haemoglobin A1c (HbA1c) ≥ 6.5%; or self‐reported physician diagnosis of diabetes or use of anti‐diabetic medication during the follow‐up visits

N per category of plasma Se (μg/L)

Q1 < 54.10: 470

Q2: 54.10–61.71: 488

Q3: 61.71–72.16: 587

Q4: > 72.16: 533

n, per category

Q1: 210

Q2: 229

Q3: 326

Q4: 274

BMI, smoking status, drinking status, education, physical activity, hypertension, hyperlipidemia, family history of diabetes, and eGFR

For subgroups analyses: age, sex, BMI, smoking status, alcohol intake status, education, physical activity, hypertension, hyperlipidemia, family history of diabetes, and eGFR

OR (95% CI) for T2DM

Q1: 1

Q2: 1.08 (0.80, 1.46)

Q3: 1.45 (1.09, 1.93)

Q4: 1.27 (0.93, 1.74)

p _trend: 0.05

< 65 yr

Q1: 1

Q2: 1.06 (0.74, 1.53)

Q3: 2.00 (1.42, 2.83)

Q4: 1.36 (0.95, 1.94)

p _trend: 0.02

≥ 65 yr

Q1: 1

Q2: 1.15 (0.73, 1.80)

Q3: 1.40 (0.90, 2.18)

Q4: 1.44 (0.92, 2.24)

p _trend: 0.08

Males

Q1: 1

Q2: 1.03 (0.69, 1.55)

Q3: 1.34 (0.90, 2.00)

Q4: 1.39 (0.93, 2.07)

p for trend : 0.06

Females

Q1: 1

Q2: 1.14 (0.77, 1.68)

Q3: 1.64 (1.13, 2.39)

Q4: 1.52 (1.03, 2.23)

p _trend: 0.02

EPIC – Potsdam Study

Germany

Cabral et al. (2021)

6.6 yr

Nested‐case cohort

Public

N = 27,548

Population sampled: general population, aged 35–64 yr

Excluded: insufficient/no serum; unclear disease status; prevalent T2DM, cancer, myocardial infarction or stroke at baseline; incomplete follow‐up information

n = 2,741

Cases: 705

Random sub‐cohort: 2,090 (overlap: 54)

Sex (% F): 62.55

Ethnicity: Caucasian

Age (yr, median): 49

Self‐reported incident cases (diabetes diagnosis, diabetes‐relevant medication, or dietary treatment due to diabetes) every 2 years; confirmed by questionnaires to the diagnosing physician

Serum Se (μg/L, median (IQR))

80.0 (19.1)

705 cases

Model 1: age, sex, educational attainment, BMI, WC, smoking status, physical activity, alcohol intake, vitamin and mineral preparations, hypertension, lipid‐lowering medication, and Mediterranean diet score.

Model 2: model 1 + serum trace elements Manganese, Iron, Copper, Zinc, and Iodine

HR (95%) for T2DM, per SD increase in serum Se

Model 1: 1.26 (1.12, 1.41)

Model 2: 1.19 (1.06, 1.34)

CSPPT

Zhang et al. (2019)

China

4.5 yr

Prospective Cohort

(sub‐sample of a trial)

Public

N = 20,702

Population sampled:

Hypertensive patients, aged 45–75 yr

Excluded: history of CVD; diabetic at baseline

n = 2,367 (sub‐sample from CSPPT trial)

Sex (% F): 53

Ethnicity: Asian

Age (yr): 61.4 ± 7.6

Physician‐diagnosed diabetes or use of glucose‐lowering drugs, or new onset (FPG ≥ 126.0 mg/dL) at the exit visit.

Plasma Se (μg/L)

84.8 (21.1)

n per quartile of plasma Se

Q1 < 71.0: 592

Q2 71.0–< 82.3: 591

Q3 82.3‐< 94.8: 592

Q4 ≥ 94.8: 592

n cases, per quartile

Q1 (Ref): 63

Q2: 63

Q3: 58

Q4: 86

Model 1: Unadjusted

Model 2: age, sex, study center, study treatment group, BMI, MTHFR C677 T genotype, smoking, alcohol drinking, family history of diabetes, SBP, fasting glucose, total cholesterol, HDL, triglycerides, creatinine, folate at baseline, and time‐averaged SBP during treatment period.

OR (95% CI) for T2DM

Model 1

Q1 (Ref): 1

Q2: 1.00 (0.69, 1.45), p = 0.992

Q3: 0.91 (0.63, 1.33), p = 0.632

Q4: 1.43 (1.01, 2.02), p = 0.045

Model 2

Q1 (Ref): 1

Q2: 0.99 (0.67, 1.47), p = 0.968

Q3: 0.87 (0.59, 1.30), i = 0.509

Q4: 1.29 (0.89, 1.89), p = 0.184

Jinchang

Cheng et al. (2022)

China

5.8 yr

Nested case–control

Public

N = 48,001

Population sampled: employees of the Jinchuan Nonferrous Metals Corporation

Excluded: diabetes and pre‐diabetes patients at baseline

n = 1,244

Cases: 622

Controls: 622 (matched according to age, sex and follow‐up time)

Sex (% F) = 26

Ethnicity = Asian

Age (yr): 47.2 ± 13.9

Self‐reported T2DM, or FPG > 7.0 mmol/L, or glucose tolerance test > 11.0 mmol/L, or with explicit inpatient medical history, or diabetes pharmacotherapy history at any follow up biannual examination

n per quartile of Serum Se (μg/L)

Q1 < 85.45: 278

Q2 85.45–92.51: 282

Q3 92.52–103.43: 331

Q4 ≥ 103.44: 353

n cases, per quartile

Q1 (Ref.): 123

Q2: 126

Q3: 176

Q4: 197

Model 1: unadjusted

Model 2: age at diagnosis, family history of diabetes, physical exercise, smoking index, lifetime total alcohol intake, triglyceride, high‐density lipoprotein, low‐density lipoprotein, and hypertension status.

Model 3: Model 2 + y serum Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium

OR (95% CI) for T2DM

Model 1

Q1 (Ref): 1

Q2: 1.05 (0.75, 1.47)

Q3: 1.42 (1.03, 1.96)

Q4: 1.70 (1.01, 2.02)

p_trend = < 0.01

Model 2

Q1 (Ref): 1

Q2: 1.26 (0.86, 1.85)

Q3: 1.62 (1.17, 2.35)

Q4: 1.79 (1.21, 2.64)

p_trend = 0.12

Model 3

Q1 (Ref): 1

Q2: 1.29 (0.82, 2.03)

Q3: 1.62 (1.05, 2.51)

Q4: 1.64 (1.02, 2.65)

p _trend = 0.20

BMI: body mass index; CI: confidence interval; CSPPT: China Stroke Primary Prevention Trial; CVD: cardiovascular disease; DFTJ: Dongfeng‐Tongji cohort; eGFR: estimated glomerular filtration rate; EPIC: European Prospective Investigation into Cancer and Nutrition; F: females; FFQ: Food Frequency Questionnaire; FPG: fasting blood glucose; HbA1c: glycated haemoglobin; HDL: high‐density lipoprotein; HPFS: Health Professional Study; HR: hazard ratio; ICP‐MS: Inductively coupled plasma mass spectrometry; IQR: interquartile range; M: males; MTHFR: Methylenetetrahydrofolate reductase; NHS: Nurses' Health study; OR: odds ratio; ORDET: HORmones and Diet in the ETioliogy of Breast Cancer study; Qx, quartile/quintile; SBP: systolic blood pressure; SD: standard deviation; Se: selenium; SFFQ: semi‐0quantitative food frequency questionnaire; T2DM: type 2 diabetes mellitus; ULSAM: The Uppsala Longitudinal Study of Adult Men; yr: year.

(a) Mean ± SD (range), unless specified otherwise.

(b): National Diabetes Data Group. Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes 1979;28:1039–57.