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

Cohort name Country Reference Follow‐up 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
EPIC Denmark, Italy, The Netherlands, Spain, Sweden, UK, Germany, Greece Allen et al. (2008) 4.3 yr (median; up to 15.1 yr) Nested Case–Control Public	N ≈ 520,000 Population sampled: general population Excluded: no blood sample, missing information on the date of blood collection or had a history of cancer (except NMSC) n = 2,018 cases: 959 controls: 1,059 (matched for age, time of blood collection) Sex: M Ethnicity: Caucasian Age (yr): 43–76	Data on stage and grade at diagnosis extracted from pathology reports stored at cancer registries or from medical records stored at the treating hospital.	Serum Se (μg/L, geometric mean (95% CI)) Cases: 70.6 (69.7, 71.5) Controls: 71.9 (71.0, 72.7) n, per quintile Q1 < 62: 441 Q2 62–68.5: 391 Q3 68.6–75: 404 Q4 75.1–84.0: 384 Q5 ≥ 84.1: 398	Cases per quintile Q1: 229 Q2: 179 Q3: 192 Q4: 172 Q5: 187	Model 1: crude Model 2: BMI, smoking status, alcohol intake, physical activity, marital status, and education level	Incidence of prostate cancer; RR (95% CI) Model 1 Q1(ref.): 1 Q2: 0.80 (0.60, 1.05) Q3: 0.87 (0.65, 1.16) Q4: 0.85 (0.64, 1.14) Q5: 1.00 (0.74, 1.36) p _trend = 0.48 Model 2 Q1(ref.): 1 Q2: 0.81 (0.61, 1.07) Q3: 0.85 (0.63, 1.14) Q4: 0.82 (0.61, 1.10) Q5: 0.96 (0.70, 1.31) p _trend = 0.25
MEC USA Park et al. (2015) 13.9 yr (mean) Prospective cohort Public	N = 96,896 Population sampled: general population Excluded: not belonging to one of the five racial/ethnic groups, prior prostate cancer, implausible dietary data based on total energy intake or its components, and missing or incomplete data n = 75,216 Sex: M Ethnicity: 26% White, 12% African American, 7% Native Hawaiian, 32% Japanese American, 23% Latino Age (yr): 45–75	Cases identified by linkage to the Surveillance, Epidemiology, and End Results (SEER) cancer registries covering the states of Hawaii and California.	Se intake assessed through a SFFQ Quintiles cut‐points, μg/1,000 kcal Q1 < 44 Q2 44–49.4 Q3 49.5–54.3 Q4 54.4–60 Q5 ≥ 60.1 n per quintile: NR	Total cases: 7,115 Cases per quintile: NR	Age at cohort entry, race/ethnicity, family history of prostate cancer, BMI, smoking status, education, history of diabetes, physical activity, alcohol consumption, calcium intake, legume intake, lycopene intake.	Incidence of prostate cancer; RR (95% CI) Q1(ref): 1 Q2: 1.10 (0.96, 1.26) Q3: 1.10 (0.95, 1.27) Q4: 0.98 (0.84, 1.16) Q5: 1.01 (0.84, 1.20) p _trend = 0.71
DCH cohort Denmark Outzen et al. (2021) Up to 19 yr Nested Case–Control Mixed	N = 27,178 Population sampled: general population, aged 50–64 yr, without prior diagnosis of cancer Excluded: lack of toenail sample, very high toenail Se concentration, very low toenail sample mass, missing data, incomplete case–control pairs n = 2,320 Cases: 1,160 Controls: 1,160 Sex: M Ethnicity: Caucasian Age (yr): 50–64	Data on cancer occurrence obtained through record linkage to the Danish Cancer Registry.	Toenail selenium (μg/g, median (P5, P95): 0.510 (0.394, 0.717) n, per quintile Q1 ≤ 0.447: 481 Q2 0.447–0.488: 441 Q3 0.488–0.533: 474 Q4 0.533–0.599: 461 Q5 ≥ 0.599: 463 Plasma selenoprotein P (n = 993; mg/L, median (P5, P95): 5.5 (3.5, 8.0) n, per quintile Q1 ≤ 4.4: 399 Q2 4.4–5.2: 425 Q3 5.2–5.8: 408 Q4 5.8–6.7: 382 Q5 > 6.7: 372	Cases per quintile of toenail Se Q1: 247 Q2: 211 Q3: 239 Q4: 232 Q5: 231 Cases per quintile of plasma selenoprotein P Q1: 200 Q2: 226 Q3: 209 Q4: 184 Q5: 174	BMI, smoking status, education, participation in sport	Incidence of prostate cancer, OR (95% CI) By toenail Se Q1(ref): 1 Q2: 0.87 (0.67, 1.12) Q3: 0.97 (0.75, 1.27) Q4: 0.98 (0.74, 1.30) Q5: 0.95 (0.72, 1.26) p _trend = 0.88 By plasma selenoprotein P Model 1 Q1(ref): 1 Q2: 1.13 (0.86, 1.49) Q3: 1.03 (0.78, 1.37) Q4: 0.90 (0.68, 1.20) Q5: 0.83 (0.61, 1.13) p _trend = 0.11
Netherlands Cohort Study The Netherlands Geybels et al. (2013) 17.3 yr Prospective case‐cohort Private	N = 58,279 Population sampled: general population, aged 55–69 yr Excluded: prevalent cancer other than skin cancer at baseline; incomplete/inconsistent dietary questionnaire n = 2,074 Cases: 898 Sub‐cohort: 1,176 Sex: M Ethnicity: Caucasian Age (yr) Cases: 62.1 (4.1) Sub‐cohort: 61.3 (4.2)	Cases of advanced prostate cancer (International Union Against Cancer (UICC) stage III/IV) identified by annual record linkage to the Netherlands Cancer Registry and the Dutch Pathology Registry	Toenail Se (μg/g, mean (SD)) Cases: 0.527 (0.169) Sub‐cohort: 0.550 (0.129) *person‐years, per quintile for sub‐cohort** Q1 ≤ 0.469: 3,203 Q2 0.469–0.515: 3,283 Q3 0.515–0.560: 3,336 Q4 0.560–0.617: 3,449 Q5 > 0.617: 3,375 *based on the distribution among sub‐cohort members.	Total = 898 n, per quintile Q1: 261 Q2: 214 Q3: 178 Q4: 130 Q5: 115	Model 1: age Model 2: age, first‐degree family history of prostate cancer, smoking status, duration of smoking, and frequency of smoking	HR (95% CI) for advanced prostate cancer Model 1 Q1(ref): 1 Q2: 0.78 (0.59, 1.02) Q3: 0.63 (0.48, 0.83) Q4: 0.47 (0.35, 0.63) Q5: 0.39 (0.29, 0.53) p _trend < 0.001 Model 2 Q1(ref): 1 Q2: 0.75 (0.57, 1.00) Q3: 0.59 (0.44, 0.79) Q4: 0.43 (0.31, 0.58) Q5: 0.37 (0.27, 0.51) p _trend < 0.001
ULSAM Sweden Grundmark et al. (2011) 26.5 yr (mean) Prospective cohort Mixed	N = 2,322 Population Sampled: general male population residents in Uppsala Excluded: missing serum Se n = 2,045 Sex: M Ethnicity: Caucasian Age (yr): 50	Prostate cancer cases ascertained via linkage with the nationwide Population Register, the Cancer Register, the Hospital Discharge Register and the Causes of Death Register. Cases confirmed by reviewing the medical records	n, per tertile of serum Se (μg/L) T1 ≤ 70: 759 T2 70.1–81: 653 T3 > 81: 633	n, per tertile T1: 84 T2: 65 T3: 59	unadjusted	Incidence of prostate cancer, RR (95% CI) T1(ref): 1 T2: 0.89 (0.65, 1.24) T3: 0.83 (0.60, 1.16)
PCPT USA and Canada Kristal et al. (2010) 9 yr Nested case–control Public	N = 18,880 Population sampled: men with PSA levels ≤ 3 ng/mL and normal DRE, participating in PCPT trial Excluded: missing end‐of‐study biopsy, BMI, dietary intake or Gleason scores; had prostatectomy for reasons other than cancer; cases diagnosed on or after the trial end‐date; unreliable dietary information n = 9,559 Cases: 1,703 Controls: 7,856 Sex: M Ethnicity: ~94% White Age (yr) ≥ 55	Adenocarcinoma identified through biopsies, consisting of a minimum of 6 core samples, reviewed by the pathologist at the local study site and a central pathology laboratory	Se intake assessed through a nutritional supplement questionnaire n, per supplemental intake (μg/d) category For Gleason score 2–7 C1 < 10: 5,351 C2 10–30: 2,947 C3 > 30: 1,134 For Gleason score 8–10 C1 < 10: 4,559 C2 10–30: 2,466 C3 > 30: 958	n, for Gleason score 2–7 C1: 870 C2: 514 C3: 192 n, for Gleason score 8–10 C1: 78 C2: 33 C3: 16	age, race/ethnicity, family history of prostate cancer in first‐degree relatives, treatment arm, and BMI	Incidence of prostate cancer, OR (95% CI) For Gleason score 2–7 C1(ref): 1 C2: 1.08 (0.96, 1.22) C3: 1.06 (0.89, 1.25) For Gleason score 8–10 C1(ref): 1 C2: 0.80 (0.53, 1.21) C3: 1.00 (0.58, 1.73)
PLCOCS USA Peters et al. (2007) Up to 8 yr Nested case–control Public	N = 38,352 Population sampled: participants from the screening arm for the PLCO trial, aged 55–74 yr Excluded: history of cancer (other than NMSC), unable to be contacted, ethnic or racial background other than non‐Hispanic white, missing data n = 1,603 cases: 879 controls: 724 (matched for age, time since initial screening, race, year of blood draw) Sex: M Ethnicity: non‐Hispanic white Age (yr): 55–74	Cases of adenocarcinoma of prostate were identified. PSA measured at entry and annually for 5 yr and DRE at entry and annually for 3 yr; men with PSA levels > 4 ng/mL or suspicious DRE referred to their medical care providers for prostate cancer diagnosis. Follow‐up for recent diagnosis of cancer carried out by annual mailed questionnaires and through searches of the National Death Index; confirmed against death certificates and medical or pathologic records	Serum Se (ng/mL, median (range)) n, per quartile* Q1 113.7 (50.5 to < 126.8): 414 Q2 135.3 (≥ 126.8 to < 141.9): 409 Q3 149.4 (≥ 141.9 to < 158): 418 Q4 170.4 (≥ 158 to 253): 362 *Based on the distribution among controls	n, per quartile Q1: 195 Q2: 189 Q3: 198 Q4: 142		Incidence of prostate cancer, OR (95% CI) Q1(Ref): 1 Q2: 0.95 (0.72, 1.27) Q3: 1.13 (0.85, 1.51) Q4: 0.84 (0.62, 1.14) p _trend = 0.70
PHS USA Li et al. (2004) 13 yr Nested case–control Public	N = 22,071 Population sampled: male physicians Excluded: history of myocardial infarction, stroke, transient ischemic attack, unstable angina; cancer (except for NMSC); renal or liver disease, peptic ulcer, gout; use of platelet‐active agents, vitamin A, or β‐carotene supplements n = 1,163 cases: 586 controls: 577 Sex: M Ethnicity: majority Caucasian (94%) Age (yr): 40–84	Cases of prostate cancer self‐reported; confirmed against hospital records and pathology reports by study physicians from the End Point Committee Stage	Plasma Se (ppm, median (range)) n, per quintileError! Bookmark not defined. Q1 0.09 (0.06–0.09): 236 Q2 0.10 (0.09–0.10): 253 Q3 0.11 (0.10–0.11): 217 Q4 0.12 (0.11–0.12):245 Q5 0.13 (0.12–0.19): 212 Based on the distribution among controls	n, per quintile Q1: 121 Q2: 137 Q3: 105 Q4: 127 Q5: 96	Age at baseline, smoking status, and duration of follow‐up (duration of follow‐up for case subjects was number of years between baseline and diagnosis; duration of follow‐up for control subjects was the same as that for corresponding case subjects).	Incidence of prostate cancer, OR (95% CI) Q1(ref): 1 Q2: 1.13 (0.79, 1.61) Q3: 0.88 (0.61, 1.28) Q4: 1.02 (0.71, 1.45) Q5: 0.78 (0.54, 1.13) p _trend = 0.16
CARET USA Goodman et al. (2001) 4.7 yr (mean) Nested case–control Public	N = 18,306 Population sampled: asbestos workers and heavy smokers participating to CARET trial n = 691 cases: 235 controls: 456 (matched for randomisation year, age group, smoking status, treatment arm, year of blood draw Sex: M Ethnicity: 91% White, 6% Black, 3% other/unknown Age (yr): 45–74	Self‐reported prostate cancer; confirmed against medical records and pathology reports	*Serum Se quartiles (μg/dl)** Q1: 5.07–10.12 Q2: 10.13–11.25 Q3: 11.26–12.59 Q4: 12.60–21.96 Based on the distribution among controls n per quartile:* NR	n, per quartile: NR		Incidence of prostate cancer, OR (95% CI) Q1 (ref): 1 Q2: 0.85 (0.53, 1.35) Q3: 1.08 (0.69, 1.71) Q4: 1.02 (0.65, 1.60) p _trend = 0.69
HHP USA Nomura et al. (2000) 12.4 yr (mean) Nested case–control Public	N = 9,345 Population sampled: Japanese American men Excluded: history of cancer prior to baseline n = 498 Cases: 249 Controls: 249 (matched on age, smoking status, time of examination) Sex: M Ethnicity: Asian Age (yr): 45–68	Prostate cancer cases identified through discharge records of hospitals and linkage with the Hawaii Tumour Registry	n per quartile of serum Se (ng/mL) Q1 < 119.3: 137 Q2 119.3 to < 130.6: 127 Q3 130.6 to < 147.2: 134 Q4 ≥ 147.2: 100 *Based on the distribution among controls	n, per quartile Q1: 75 Q2: 64 Q3: 72 Q4: 38		Incidence of prostate cancer, OR (95% CI) Q1(ref): 1 Q2: 0.9 (0.5, 1.4) Q3: 1.0 (0.6, 1.6) Q4: 0.5 (0.3, 0.9) p _trend = 0.2
CLUE II USA Helzlsouer et al. (2000) 7 yr Nested case–control Public	N = 10,456 Population sampled: general male population residents of Washington county Excluded: NR n = 350 cases: 117 controls: 233 (matched for age, race, date of participation in the CLUE II program, size of toenail clipping) Sex: M Ethnicity: NR Age (yr): > 45	Cases of prostate cancer identified by linkage with Washington County Cancer Registry and the Maryland Cancer Registry (since 1995)	n, per quintile of toenail Se (ppm) Q1: < 0.69: 77 Q2: 0.69–0.75: 68 Q3: 0.75–0.81: 67 Q4: 0.81–0.91: 71 Q5: > 0.91: 67 *Based on the distribution among controls	n, per quintile Q1: 32 Q2: 20 Q3: 21 Q4: 24 Q5: 20	BMI at age 21 years, education, and hours since last meal	Incidence of prostate cancer, OR (95% CI) Q1(ref): 1 Q2: 0.41 (0.18, 0.93) Q3: 0.55 (0.26, 1.17) Q4: 0.66 (0.33, 1.33) Q5: 0.38 (0.17, 0.85)
HPFS USA Yoshizawa et al. (1998) 7 yr Nested case–control Public	N = 51,529 Population sampled: male health professionals, aged 40–75 yr Excluded: energy intake < 800 or > 4,200 kcal/d, incomplete questionnaire, cases occurring in first 2 yr n = 362 Cases: 181 Controls: 181 (matched for age, smoking status, date of toenail return) Sex: M Ethnicity: majority Caucasian Age (yr, median): 64	Self‐reported incident cases of prostate cancer identified via biannual questionnaires, confirmed through a review of histopathologic reports from medical records	n per quintile (median (range)) of toenail Se (ppm) Q1: 0.66 (0.53–0.73): 89 Q2: 0.76 (0.73–0.79): 71 Q3: 0.82 (0.79–0.85): 66 Q4: 0.88 (0.85–0.94): 71 Q5: 1.14 (0.94–7.09): 65	n, per quintile Q1: 54 Q2: 34 Q3: 29 Q4: 36 Q5: 28	Model 1: crude Model 2: quintiles of lycopene, saturated fat, and calcium, for family history of prostate cancer (binary), for body mass index (quintiles), and for vasectomy (binary). Model 3 (n = 354): model 2 + region (soil selenium content high, medium, low)	Incidence of prostate cancer, OR (95% CI) Model 1 Q1(ref): 1 Q2: 0.57 (0.29, 1.12) Q3: 0.53 (0.28, 1.01) Q4: 0.67 (0.34, 1.32) Q5: 0.49 (0.25, 0.96) p _trend = 0.11 Model 2 Q1(ref): 1 Q2: 0.62 (0.29, 1.35) Q3: 0.35 (0.16, 0.78) Q4: 0.80 (0.35, 1.80) Q5: 0.39 (0.18, 0.84) p _trend = 0.05 Model 3 Q1(ref): 1 Q2: 0.59 (0.27, 1.30) Q3: 0.35 (0.16, 0.78) Q4: 0.76 (0.34, 1.73) Q5: 0.35 (0.16, 0.78) p _trend = 0.03
ATBC study Finland Hartman et al. (1998) 7 y (median) Prospective case‐cohort Public	N = 29,133 Population sampled: smokers, aged 50–69 yr, participating in the ATBC trial Excluded: alcoholics, cirrhosis of the liver, severe angina with exertion, chronic renal insufficiency, previously diagnosed with cancer, use of vitamins E or A or β‐carotene supplements, receiving anticoagulant therapy n = 29,133 cases: 317 controls: 28,816 Sex: M Ethnicity: Caucasian Age (yr): cases: 60.9 ± 5.1 controls: 57.2 ± 5.1	Prostate cancer cases identified through the Finnish Cancer Registry and the Register of Causes of Death; confirmed against medical records	Se intake (μg/d) assessed through a SFFQ Cases: 93.9 ± 40.2 Controls: 95.9 ± 36.5 Quartiles of Se intake (n = 190) Q1: < 71.52 Q2: 71.52–89.12 Q3: 89.13–111.05 Q4: > 111.05 n per quartile: NR	n, per quartile: NR	Age, BPH, living in an urban area, β‐carotene intervention, and total energy (dietary factors)	Incidence of prostate cancer, RR (95% CI) Q1(ref): 1 Q2: 1.09 (0.71, 1.68) Q3: 0.97 (0.59, 1.60) Q4: 1.27 (0.70, 2.20) p _trend = 0.49
MCHES Finland Knekt et al. (1990) 10 yr (median) Nested case–control Public	N = 39,268 Population sampled: general population, aged 15–99 yr Excluded: history of cancer n = 102 Cases: 51 Controls: 51 (matched for sex, age and municipality) Sex: M Ethnicity: Caucasian Age (yr): 15–99	Prostate cancer cases identified through the nationwide Finnish Cancer Registry.	Serum Se (μg/L) Cases: 59.6 ± 19.4 Controls: 58.3 ± 14.8 Quintiles Q1: < 0.49 Q2: 49–57 Q3: 58–66 Q4: 67–77 Q5: ≥ 78 n per quintile: NR	n, per quintile: NR	Smoking	Incidence of prostate cancer, RR (95% CI NR) Q1(ref): 1 Q2: 1.37 Q3: 0.85 Q4: 0.87 Q5: 1.15 p _trend = 0.707
Seattle firms USA Coates et al. (1988) 10 yr Nested case–control Public	N = 6,167 Population sampled: employees from two Seattle firms Excluded: NR n = 37 cases: 13 controls: 24 (matched for employer, age, sex, race, date of blood draw) Sex: M Ethnicity: ~70% white Age (yr): > 18	Names and birthdates of the employees who developed prostate cancer were matched against the records of the Cancer Surveillance System.	Serum Se tertiles (μg/L, n = 241) T1: 98–148 T2: 149–170 T3: 171–240 n per tertile: NR	n, per tertile: NR		Incidence of prostate cancer, RR (95% CI: NR) T1(ref): 1 T2: 0.2 T3: 0.3 p _trend = 0.18

Cohort name

Country

Reference

Follow‐up

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

EPIC

Denmark, Italy, The Netherlands, Spain, Sweden, UK, Germany, Greece

Allen et al. (2008)

4.3 yr (median; up to 15.1 yr)

Nested Case–Control

Public

N ≈ 520,000

Population sampled: general population

Excluded: no blood sample, missing information on the date of blood collection or had a history of cancer (except NMSC)

n = 2,018

cases: 959

controls: 1,059 (matched for age, time of blood collection)

Sex: M

Ethnicity: Caucasian

Age (yr): 43–76

Data on stage and grade at diagnosis extracted from pathology reports stored at cancer registries or from medical records stored at the treating hospital.

Serum Se (μg/L, geometric mean (95% CI))

Cases: 70.6 (69.7, 71.5)

Controls: 71.9 (71.0, 72.7)

n, per quintile

Q1 < 62: 441

Q2 62–68.5: 391

Q3 68.6–75: 404 Q4 75.1–84.0: 384

Q5 ≥ 84.1: 398

Cases per quintile

Q1: 229

Q2: 179

Q3: 192 Q4: 172

Q5: 187

Model 1: crude

Model 2: BMI, smoking status, alcohol intake, physical activity, marital status, and education level

Incidence of prostate cancer; RR (95% CI)

Model 1

Q1(ref.): 1

Q2: 0.80 (0.60, 1.05)

Q3: 0.87 (0.65, 1.16) Q4: 0.85 (0.64, 1.14)

Q5: 1.00 (0.74, 1.36)

p _trend = 0.48

Model 2

Q1(ref.): 1

Q2: 0.81 (0.61, 1.07)

Q3: 0.85 (0.63, 1.14) Q4: 0.82 (0.61, 1.10)

Q5: 0.96 (0.70, 1.31)

p _trend = 0.25

MEC

USA

Park et al. (2015)

13.9 yr (mean)

Prospective cohort

Public

N = 96,896

Population sampled: general population

Excluded: not belonging to one of the five racial/ethnic groups, prior prostate cancer, implausible dietary data based on total energy intake or its components, and missing or incomplete data

n = 75,216

Sex: M

Ethnicity: 26% White, 12% African American, 7% Native Hawaiian, 32% Japanese American, 23% Latino

Age (yr): 45–75

Cases identified by linkage to the Surveillance, Epidemiology, and End Results (SEER) cancer registries covering the states of Hawaii and California.

Se intake assessed through a SFFQ

Quintiles cut‐points, μg/1,000 kcal

Q1 < 44

Q2 44–49.4

Q3 49.5–54.3

Q4 54.4–60

Q5 ≥ 60.1

n per quintile: NR

Total cases: 7,115

Cases per quintile: NR

Age at cohort entry, race/ethnicity, family history of prostate cancer, BMI, smoking status, education, history of diabetes, physical activity, alcohol consumption, calcium intake, legume intake, lycopene intake.

Incidence of prostate cancer; RR (95% CI)

Q1(ref): 1

Q2: 1.10 (0.96, 1.26)

Q3: 1.10 (0.95, 1.27)

Q4: 0.98 (0.84, 1.16)

Q5: 1.01 (0.84, 1.20)

p _trend = 0.71

DCH cohort

Denmark

Outzen et al. (2021)

Up to 19 yr

Nested Case–Control

Mixed

N = 27,178

Population sampled: general population, aged 50–64 yr, without prior diagnosis of cancer

Excluded: lack of toenail sample, very high toenail Se concentration, very low toenail sample mass, missing data, incomplete case–control pairs

n = 2,320

Cases: 1,160

Controls: 1,160

Sex: M

Ethnicity: Caucasian

Age (yr): 50–64

Data on cancer occurrence obtained through record linkage to the Danish Cancer Registry.

Toenail selenium (μg/g, median (P5, P95): 0.510 (0.394, 0.717)

n, per quintile

Q1 ≤ 0.447: 481

Q2 0.447–0.488: 441

Q3 0.488–0.533: 474

Q4 0.533–0.599: 461

Q5 ≥ 0.599: 463

Plasma selenoprotein P (n = 993; mg/L, median (P5, P95): 5.5 (3.5, 8.0)

n, per quintile

Q1 ≤ 4.4: 399

Q2 4.4–5.2: 425

Q3 5.2–5.8: 408

Q4 5.8–6.7: 382

Q5 > 6.7: 372

Cases per quintile of toenail Se

Q1: 247

Q2: 211

Q3: 239

Q4: 232

Q5: 231

Cases per quintile of plasma selenoprotein P

Q1: 200

Q2: 226

Q3: 209

Q4: 184

Q5: 174

BMI, smoking status, education, participation in sport

Incidence of prostate cancer, OR (95% CI)

By toenail Se

Q1(ref): 1

Q2: 0.87 (0.67, 1.12)

Q3: 0.97 (0.75, 1.27)

Q4: 0.98 (0.74, 1.30)

Q5: 0.95 (0.72, 1.26)

p _trend = 0.88

By plasma selenoprotein P

Model 1

Q1(ref): 1

Q2: 1.13 (0.86, 1.49)

Q3: 1.03 (0.78, 1.37)

Q4: 0.90 (0.68, 1.20)

Q5: 0.83 (0.61, 1.13)

p _trend = 0.11

Netherlands Cohort Study

The Netherlands

Geybels et al. (2013)

17.3 yr

Prospective case‐cohort

Private

N = 58,279

Population sampled: general population, aged 55–69 yr

Excluded: prevalent cancer other than skin cancer at baseline; incomplete/inconsistent dietary questionnaire

n = 2,074

Cases: 898

Sub‐cohort: 1,176

Sex: M

Ethnicity: Caucasian

Age (yr)

Cases: 62.1 (4.1)

Sub‐cohort: 61.3 (4.2)

Cases of advanced prostate cancer (International Union Against Cancer (UICC) stage III/IV) identified by annual record linkage to the Netherlands Cancer Registry and the Dutch Pathology Registry

Toenail Se (μg/g, mean (SD))

Cases: 0.527 (0.169)

Sub‐cohort: 0.550 (0.129)

person‐years, per quintile* for sub‐cohort

Q1 ≤ 0.469: 3,203

Q2 0.469–0.515: 3,283

Q3 0.515–0.560: 3,336

Q4 0.560–0.617: 3,449

Q5 > 0.617: 3,375

*based on the distribution among sub‐cohort members.

Total = 898

n, per quintile

Q1: 261

Q2: 214

Q3: 178

Q4: 130

Q5: 115

Model 1: age

Model 2: age, first‐degree family history of prostate cancer, smoking status, duration of smoking, and frequency of smoking

HR (95% CI) for advanced prostate cancer

Model 1

Q1(ref): 1

Q2: 0.78 (0.59, 1.02)

Q3: 0.63 (0.48, 0.83)

Q4: 0.47 (0.35, 0.63)

Q5: 0.39 (0.29, 0.53)

p _trend < 0.001

Model 2

Q1(ref): 1

Q2: 0.75 (0.57, 1.00)

Q3: 0.59 (0.44, 0.79)

Q4: 0.43 (0.31, 0.58)

Q5: 0.37 (0.27, 0.51)

p _trend < 0.001

ULSAM

Sweden

Grundmark et al. (2011)

26.5 yr (mean)

Prospective cohort

Mixed

N = 2,322

Population Sampled: general male population residents in Uppsala

Excluded: missing serum Se

n = 2,045

Sex: M

Ethnicity: Caucasian

Age (yr): 50

Prostate cancer cases ascertained via linkage with the nationwide Population Register, the Cancer Register, the

Hospital Discharge Register and the Causes of Death Register. Cases confirmed by reviewing the medical records

n, per tertile of serum Se (μg/L)

T1 ≤ 70: 759

T2 70.1–81: 653

T3 > 81: 633

n, per tertile

T1: 84

T2: 65

T3: 59

unadjusted

Incidence of prostate cancer, RR (95% CI)

T1(ref): 1

T2: 0.89 (0.65, 1.24)

T3: 0.83 (0.60, 1.16)

PCPT

USA and Canada

Kristal et al. (2010)

9 yr

Nested case–control

Public

N = 18,880

Population sampled: men with PSA levels ≤ 3 ng/mL and normal DRE, participating in PCPT trial

Excluded: missing end‐of‐study biopsy, BMI, dietary intake or Gleason scores; had prostatectomy for reasons other than cancer; cases diagnosed on or after the trial end‐date; unreliable dietary information

n = 9,559

Cases: 1,703

Controls: 7,856

Sex: M

Ethnicity: ~94% White

Age (yr) ≥ 55

Adenocarcinoma identified through biopsies, consisting of a minimum of 6 core samples, reviewed by the pathologist at the local study site and a central pathology laboratory

Se intake assessed through a nutritional supplement questionnaire

n, per supplemental intake (μg/d) category

For Gleason score 2–7

C1 < 10: 5,351

C2 10–30: 2,947

C3 > 30: 1,134

For Gleason score 8–10

C1 < 10: 4,559

C2 10–30: 2,466

C3 > 30: 958

n, for Gleason score 2–7

C1: 870

C2: 514

C3: 192

n, for Gleason score 8–10

C1: 78

C2: 33

C3: 16

age, race/ethnicity, family

history of prostate cancer in first‐degree relatives, treatment arm, and BMI

Incidence of prostate cancer, OR (95% CI)

For Gleason score 2–7

C1(ref): 1

C2: 1.08 (0.96, 1.22)

C3: 1.06 (0.89, 1.25)

For Gleason score 8–10

C1(ref): 1

C2: 0.80 (0.53, 1.21)

C3: 1.00 (0.58, 1.73)

PLCOCS

USA

Peters et al. (2007)

Up to 8 yr

Nested case–control

Public

N = 38,352

Population sampled: participants from the screening arm for the PLCO trial, aged 55–74 yr

Excluded: history of cancer (other than NMSC), unable to be contacted, ethnic or racial background other than non‐Hispanic white, missing data

n = 1,603

cases: 879

controls: 724 (matched for age, time since initial screening, race, year of blood draw)

Sex: M

Ethnicity: non‐Hispanic white

Age (yr): 55–74

Cases of adenocarcinoma of prostate were identified. PSA measured at entry and annually for 5 yr and DRE at entry and annually for 3 yr; men with PSA levels > 4 ng/mL or suspicious DRE referred to their medical care providers for prostate cancer diagnosis. Follow‐up for recent diagnosis of cancer carried out by annual mailed questionnaires and through searches of the National Death Index; confirmed against death certificates and medical or pathologic records

Serum Se (ng/mL, median (range))

n, per quartile*

Q1 113.7 (50.5 to < 126.8): 414

Q2 135.3 (≥ 126.8 to < 141.9): 409

Q3 149.4 (≥ 141.9 to < 158): 418 Q4 170.4 (≥ 158 to 253): 362

*Based on the distribution among controls

n, per quartile

Q1: 195

Q2: 189

Q3: 198 Q4: 142

Incidence of prostate cancer, OR (95% CI)

Q1(Ref): 1

Q2: 0.95 (0.72, 1.27)

Q3: 1.13 (0.85, 1.51) Q4: 0.84 (0.62, 1.14)

p _trend = 0.70

PHS

USA

Li et al. (2004)

13 yr

Nested case–control

Public

N = 22,071

Population sampled: male physicians

Excluded: history of myocardial infarction, stroke, transient ischemic attack, unstable angina; cancer (except for NMSC); renal or liver disease, peptic ulcer, gout; use of platelet‐active agents, vitamin A, or β‐carotene supplements

n = 1,163

cases: 586

controls: 577

Sex: M

Ethnicity: majority Caucasian (94%)

Age (yr): 40–84

Cases of prostate cancer self‐reported; confirmed against hospital records and pathology reports by study physicians from the End Point Committee

Stage

Plasma Se (ppm, median (range))

n, per quintile*Error! Bookmark not defined.

Q1 0.09 (0.06–0.09): 236

Q2 0.10 (0.09–0.10): 253

Q3 0.11 (0.10–0.11): 217

Q4 0.12 (0.11–0.12):245

Q5 0.13 (0.12–0.19): 212

*Based on the distribution among controls

n, per quintile

Q1: 121

Q2: 137

Q3: 105

Q4: 127

Q5: 96

Age at baseline, smoking status, and duration of follow‐up (duration of follow‐up for case subjects was number of years between baseline and diagnosis; duration of follow‐up for control subjects was the same as that for corresponding case subjects).

Incidence of prostate cancer, OR (95% CI)

Q1(ref): 1

Q2: 1.13 (0.79, 1.61)

Q3: 0.88 (0.61, 1.28)

Q4: 1.02 (0.71, 1.45)

Q5: 0.78 (0.54, 1.13)

p _trend = 0.16

CARET

USA

Goodman et al. (2001)

4.7 yr (mean)

Nested case–control

Public

N = 18,306

Population sampled: asbestos workers and heavy smokers participating to CARET trial

n = 691

cases: 235

controls: 456 (matched for randomisation year, age group, smoking status, treatment arm, year of blood draw

Sex: M

Ethnicity: 91% White, 6% Black, 3% other/unknown

Age (yr): 45–74

Self‐reported prostate cancer; confirmed against medical records and pathology reports

Serum Se quartiles* (μg/dl)

Q1: 5.07–10.12

Q2: 10.13–11.25

Q3: 11.26–12.59

Q4: 12.60–21.96

*Based on the distribution among controls

n per quartile: NR

n, per quartile: NR

Incidence of prostate cancer, OR (95% CI)

Q1 (ref): 1

Q2: 0.85 (0.53, 1.35)

Q3: 1.08 (0.69, 1.71)

Q4: 1.02 (0.65, 1.60)

p _trend = 0.69

HHP

USA

Nomura et al. (2000)

12.4 yr (mean)

Nested case–control

Public

N = 9,345

Population sampled: Japanese American men

Excluded: history of cancer prior to baseline

n = 498

Cases: 249

Controls: 249 (matched on age, smoking status, time of examination)

Sex: M

Ethnicity: Asian

Age (yr): 45–68

Prostate cancer cases identified through discharge records of hospitals and linkage with the Hawaii Tumour Registry

n per quartile of serum Se (ng/mL)

Q1 < 119.3: 137

Q2 119.3 to < 130.6: 127

Q3 130.6 to < 147.2: 134

Q4 ≥ 147.2: 100

*Based on the distribution among controls

n, per quartile

Q1: 75

Q2: 64

Q3: 72

Q4: 38

Incidence of prostate cancer, OR (95% CI)

Q1(ref): 1

Q2: 0.9 (0.5, 1.4)

Q3: 1.0 (0.6, 1.6)

Q4: 0.5 (0.3, 0.9)

p _trend = 0.2

CLUE II

USA

Helzlsouer et al. (2000)

7 yr

Nested case–control

Public

N = 10,456

Population sampled: general male population residents of Washington county

Excluded: NR

n = 350

cases: 117

controls: 233 (matched for age, race, date of participation in the CLUE II program, size of toenail clipping)

Sex: M

Ethnicity: NR

Age (yr): > 45

Cases of prostate cancer identified by linkage with Washington County Cancer Registry and the Maryland Cancer Registry (since 1995)

n, per quintile of toenail Se (ppm)

Q1: < 0.69: 77

Q2: 0.69–0.75: 68

Q3: 0.75–0.81: 67

Q4: 0.81–0.91: 71

Q5: > 0.91: 67

*Based on the distribution among controls

n, per quintile

Q1: 32

Q2: 20

Q3: 21

Q4: 24

Q5: 20

BMI at age 21 years, education, and hours since last meal

Incidence of prostate cancer, OR (95% CI)

Q1(ref): 1

Q2: 0.41 (0.18, 0.93)

Q3: 0.55 (0.26, 1.17)

Q4: 0.66 (0.33, 1.33)

Q5: 0.38 (0.17, 0.85)

HPFS

USA

Yoshizawa et al. (1998)

7 yr

Nested case–control

Public

N = 51,529

Population sampled: male health professionals, aged 40–75 yr

Excluded: energy intake < 800 or > 4,200 kcal/d, incomplete questionnaire, cases occurring in first 2 yr

n = 362

Cases: 181

Controls: 181 (matched for age, smoking status, date of toenail return)

Sex: M

Ethnicity: majority Caucasian

Age (yr, median): 64

Self‐reported incident cases of prostate cancer identified via biannual questionnaires, confirmed through a review of histopathologic reports from medical records

n per quintile (median (range)) of toenail Se (ppm)

Q1: 0.66 (0.53–0.73): 89

Q2: 0.76 (0.73–0.79): 71

Q3: 0.82 (0.79–0.85): 66

Q4: 0.88 (0.85–0.94): 71

Q5: 1.14 (0.94–7.09): 65

n, per quintile

Q1: 54

Q2: 34

Q3: 29

Q4: 36

Q5: 28

Model 1: crude

Model 2: quintiles of lycopene, saturated fat, and calcium, for family history of prostate cancer (binary), for body mass index (quintiles), and for vasectomy (binary).

Model 3 (n = 354): model 2 + region (soil selenium content high, medium, low)

Incidence of prostate cancer, OR (95% CI)

Model 1

Q1(ref): 1

Q2: 0.57 (0.29, 1.12)

Q3: 0.53 (0.28, 1.01)

Q4: 0.67 (0.34, 1.32)

Q5: 0.49 (0.25, 0.96)

p _trend = 0.11

Model 2

Q1(ref): 1

Q2: 0.62 (0.29, 1.35)

Q3: 0.35 (0.16, 0.78)

Q4: 0.80 (0.35, 1.80)

Q5: 0.39 (0.18, 0.84)

p _trend = 0.05

Model 3

Q1(ref): 1

Q2: 0.59 (0.27, 1.30)

Q3: 0.35 (0.16, 0.78)

Q4: 0.76 (0.34, 1.73)

Q5: 0.35 (0.16, 0.78)

p _trend = 0.03

ATBC study

Finland

Hartman et al. (1998)

7 y (median)

Prospective case‐cohort

Public

N = 29,133

Population sampled: smokers, aged 50–69 yr, participating in the ATBC trial

Excluded: alcoholics, cirrhosis of the liver, severe angina with exertion, chronic renal insufficiency, previously diagnosed with cancer, use of vitamins E or A or β‐carotene supplements, receiving anticoagulant therapy

n = 29,133

cases: 317

controls: 28,816

Sex: M

Ethnicity: Caucasian

Age (yr):

cases: 60.9 ± 5.1

controls: 57.2 ± 5.1

Prostate cancer cases identified through the Finnish Cancer Registry and the Register of Causes of Death; confirmed against medical records

Se intake (μg/d) assessed through a SFFQ

Cases: 93.9 ± 40.2

Controls: 95.9 ± 36.5

Quartiles of Se intake (n = 190)

Q1: < 71.52

Q2: 71.52–89.12

Q3: 89.13–111.05

Q4: > 111.05

n per quartile: NR

n, per quartile: NR

Age, BPH, living in an urban area, β‐carotene intervention, and total energy (dietary factors)

Incidence of prostate cancer, RR (95% CI)

Q1(ref): 1

Q2: 1.09 (0.71, 1.68)

Q3: 0.97 (0.59, 1.60)

Q4: 1.27 (0.70, 2.20)

p _trend = 0.49

MCHES

Finland

Knekt et al. (1990)

10 yr (median)

Nested case–control

Public

N = 39,268

Population sampled: general population, aged 15–99 yr

Excluded: history of cancer

n = 102

Cases: 51

Controls: 51 (matched for sex, age and municipality)

Sex: M

Ethnicity: Caucasian

Age (yr): 15–99

Prostate cancer cases identified through the nationwide Finnish Cancer Registry.

Serum Se (μg/L)

Cases: 59.6 ± 19.4

Controls: 58.3 ± 14.8

Quintiles

Q1: < 0.49

Q2: 49–57

Q3: 58–66

Q4: 67–77

Q5: ≥ 78

n per quintile: NR

n, per quintile: NR

Smoking

Incidence of prostate cancer, RR (95% CI NR)

Q1(ref): 1

Q2: 1.37

Q3: 0.85

Q4: 0.87

Q5: 1.15

p _trend = 0.707

Seattle firms

USA

Coates et al. (1988)

10 yr

Nested case–control

Public

N = 6,167

Population sampled: employees from two Seattle firms

Excluded: NR

n = 37

cases: 13

controls: 24 (matched for employer, age, sex, race, date of blood draw)

Sex: M

Ethnicity: ~70% white

Age (yr): > 18

Names and birthdates of the employees who developed prostate cancer were matched against the records of the Cancer Surveillance System.

Serum Se tertiles (μg/L, n = 241)

T1: 98–148

T2: 149–170

T3: 171–240

n per tertile: NR

n, per tertile: NR

Incidence of prostate cancer, RR (95% CI: NR)

T1(ref): 1

T2: 0.2 T3: 0.3

p _trend = 0.18

ATBC: Alpha‐Tocopherol, Beta‐Carotene Cancer Prevention; BMI: body mass index; Cx: category x; CARET: Carotene and Retinol Efficacy Trial; CLUE II: Campaign Against Cancer and Heart Disease II; CI: confidence interval; DCH: Diet, Cancer and Health; DRE: digital rectal examination; EPIC: European Prospective Investigation into Cancer and Nutrition; HHP: Honolulu Heart Program; HPFS: Health Professionals Follow‐Up Study; HR: hazard ratio; M: males; MCHES: Mobile Clinic Health Examination Survey; MEC: The Multiethnic cohort; NR: not reported; OR: odds ratio; PCPT: Prostate Cancer Prevention Trial; PH: benign prostatic hyperplasia; PHS: Physicians' Health Study; PLCOCS: Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; PSA: prostate specific antigen; Qx: quintile/quartile x; RR: relative risk; SD: standard deviation; SFFQ: semi‐quantitative food frequency questionnaire; Tx: tertile x; ULSAM: Uppsala Longitudinal Study of Adult Men; UPI: unique personal identification; USA: United States of America; yr: year.

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