Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Apr 20.
Published in final edited form as: Sci Total Environ. 2021 Nov 22;818:151848. doi: 10.1016/j.scitotenv.2021.151848

Long-term Association of Serum Selenium Levels and the Diabetes Risk: Findings from A Case-Control Study Nested in the Prospective Jinchang Cohort

Zhiyuan Cheng 1, Yuanyuan Li 2, Jamie L Young 3, Ning Cheng 5, Chenhui Yang 2, George D Papandonatos 6, Karl T Kelsey 7, John Pierce Wise Sr 3,4, Shi Kunchong 7, Tongzhang Zheng 7, Simin Liu 7, Yana Bai 8
PMCID: PMC8909917  NIHMSID: NIHMS1772915  PMID: 34822883

Abstract

An increasing body of evidence implicates high levels of selenium intake in the development of diabetes, although prospective studies remain sparse. We conducted a nested case-control study of 622 diabetes incident cases and 622-age, sex, and follow-up time-matched controls in the prospective Jinchang cohort of 48,001 participants with a median of 5.8 years of follow-up. Inductively coupled plasma mass spectrometry (ICP-MS) was used to measure all 622 case-control pairs’ baseline serum levels of selenium (Se), which were then categorized into quartiles based on the frequency distribution among the controls. Multivariable adjusted conditional logistic regression and restricted cubic splines (RCS) models were applied to evaluate independent odds ratios (OR) as estimates for relative risks (RR) of diabetes according to quartiles (Q) of selenium levels. Compared to the lowest quartile (Q1 as reference), significantly greater diabetes risks (with 95% confidence interval) were observed in Q3 (OR=1.62, 1.17–2.35) and Q4 (OR=1.79, 1.21–2.64). Sub-analyses showed these increased risks of diabetes by serum levels of Se. appeared to differ by sex, age, BMI status, history of hypertension, and dyslipidemia. Further, application of RSC models showed that serum Se levels between 95 and 120 μg/L were significantly and positively associated with diabetes risk whereas no apparent relation exists when Se levels were under 95 μg/L in this cohort population.

Keywords: Selenium, diabetes, Incidence, Risk factor, A case-control study nested in prospective cohort

Introduction

Diabetes, a chronic metabolic disease accounting for over 5 million deaths annually and countless complications every year (Cho et al., 2018; Collaborators, 2020; Lin et al., 2020; Williams et al., 2020) including cardio-cerebrovascular diseases, renal failure, diabetic retinopathy, and nerve damage (Emerging Risk Factors et al., 2010; Kim et al., 2019; Li et al., 2020; Winocour, 2018). In 2017 alone, over 850 billion U.S. dollars was spent on diabetes related healthcare complications. Unfortunately, the prevalence of diabetes is growing at epidemic proportions. From 1980–2019 the global prevalence of diabetes increased from 4.7 (108 million people) to 9.3% (422 million people) and by the year 2045 over 690 million people are estimated to have diabetes (Balakumar et al., 2016; Saeedi et al., 2019). This trend has been observed across very race/ethnicity and region around the world (Davis et al., 2018; Geiss et al., 2014; Meeks et al., 2016; Ramachandran et al., 2010; Wang et al., 2017; Xu et al., 2018). Although it is well established that family history, obesity, hypertension, dyslipidemia, physical in activities, and poor dietary quality are major risk factors for diabetes may greatly contributing to diabetes prevention (Danaei et al., 2011; Hu et al., 2001), however, these factors alone may not explain the rapid increase in the prevalence of the disease worldwide (Alberti et al., 2007). Recently growing evidence from both epidemiological and mechanistic studies also suggests that exposures to environmental chemicals, such as trace elements, may play a role in diabetes development (Gong et al., 2020; Khan and Awan, 2014; Kuo and Navas-Acien, 2015; Liu et al., 2005; Song et al., 2004).

Selenium (Se) has traditionally been regarded as an essential trace element, once believed to exert beneficial effects resulting in improved health, particularly for chronic diseases. Studies focused on its prospective effects on the incident of diabetes was still urgently need. Daily intake of Se dietary supplements was commercially promoted in most countries, particularly after the Nutritional Prevention of Cancer (NPC) considered Se as a potential chemo-preventive agent for colorectal and prostate cancers (Clark et al., 1996). However, previous studies performed by Whanger P et al. (Whanger et al., 1996) and Vinceti M et al. (Vinceti et al., 2001) suggested Se has a limited therapeutic window with high between and within-person variation. Moreover, concerns have also been raised about the potential carcinogenicity of selenite, selenate, selenium sulfide and selenomethionine (Fürnsinn et al., 1996; Rayman and Stranges, 2013; Satyanarayana et al., 2006). In the 8-years follow-up of the NPC study (Clark et al., 1996), for example, Stranges et al. reported a significantly increased risk of diabetes among those in the highest tertile of baseline plasma Se levels in comparison with the lowest tertile (HR: 2.70, 95% CI: 1.30–5.61) (Stranges et al., 2007). The beneficial effects of Se was once focused on the antioxidant and anti-inflammation capacity of selenoproteins, including glutathione peroxidases (GPx) and thioredoxin reductases (TrxR) (Steinbrenner and Sies, 2009). Subsequent studies have further demonstrated that abnormal expression of selenoproteins can cause diabetes (McCann and Ames, 2011; Rayman, 2012), metabolic syndromes (Kuzuya et al., 2008), obesity, and even cardio-cerebrovascular diseases (Alanne et al., 2007; Dora et al., 2010; Hamanishi et al., 2004) in human. Recent observational studies of diverse populations also showed direct relation between intake of Se supplements or dietary Se intake, may play a role in the development of diabetes (Kohler et al., 2018; Laclaustra et al., 2009; Li et al., 2017). Although findings have not been consistent (Gao et al., 2014; Simic et al., 2017) and potential confounding and selection biases may have affected in majority of the previous cross-sectional studies.

Recent studies have found a positive association between selenium and diabetes risk from experimental to epidemiologic studies, however, most of them were conducted among prevalent cases, it is still hard to distinguish whether Se is the cause or the result of diabetes status (Liao et al., 2020; Lin and Shen, 2021; Rath et al., 2021; Vinceti et al., 2021a; Vinceti et al., 2018; Vinceti et al., 2021b). As yet, there remain few prospective studies that have directly investigate the long-term relation between serum Se levels and diabetes incidence.

We therefore conducted a nested case-control study based on the world’s largest occupational heavy metal exposure cohort. This study focused on exploring the long-term association between serum Se levels and the risk of developing diabetes in different subgroups population, and to identify the safety range of Se for diabetes precaution. Nonlinear regressions were further conducted to explore the dose-response between selenium levels and increased risk of developing diabetes.

Methods

Study population

The Jinchang Cohort Study, an ongoing prospective cohort study in Gansu Province, China was established on the basis of a biannual free medical examination for all employees by the Jinchuan Nonferrous Metals Corporation (JNMC) (Bai et al., 2017). The baseline survey of the Jinchang Cohort was established through a cross-sectional biannual medical exam conducted from January 2011 to June 2013. All 48,001 JNMC employees and retirees completed a 4-parts of the health examination, including in-person epidemiological interviews, physical exams, laboratory tests, and donation of bio-samples (blood and urine) with a mean age of 47.2±13.9 years (range from 18 to 91). The standardized and structured questionnaires were used by uniformly trained interviewers to collect information on demographic, lifestyle factors (physical activity, dietary and nutrient intakes history, and etc.), occupational exposure history, past medical history, family cancer and chronic disease history, female reproduction history, etc. The phase I follow-up study was initiated in February 2013 and continued through December 2015. The phase II follow-up study was initiated in June 2016 and continued through December 2018. The phase III follow-up were stated in August 2020 and will be finished in August, 2022. Information was collected from a total of 37,040 follow-up participants with a follow-up rate of 81.09% and a mean follow-up duration of 5.78±0.83 years. The information including epidemiology data, health examination, bio-indicators testing, and bio-sampling were kept same with the baseline database. All enrolled participants were required to eat their regular diet (light diet without high-carbohydrate or high-fat intake) for two continuous days before health examination and fast at least 8 hours before blood donation. The biochemical examinations were performed at the Workers’ Hospital of the Jinchuan Company. Blood biochemical indicators, including serum uric acid (SUA), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), fasting plasma glucose (FPG) were automatically tested by biochemical analyzers (Hitachi 7600–020).

Study design and covariates assessment

All participants of the case population involved in our study were newly diagnosed as diabetes patients at follow-up phases. Since comprehensive physical examinations, biochemistry tests (e.g. fasting plasma glucose (FPG), glucose tolerance testing, glycosylated hemoglobin), inpatient history (especially for admitting diagnosis and discharge diagnosis from the department of endocrinology), medical health insurance data (including drug prescriptions for diabetes), and epidemical surveys were conducted for the entire cohort population at baseline (48,001/48,001). It provides us with the capacity to detect all diabetes cases (even for the pre-diabetes patients) at baseline, which ensuring that all incident cases were newly onset throughout the follow-up phases. Thus, any bias induced by the early effect of undetected early diabetes at an early stage of follow-up was minimized to the greatest extent possible. The diagnosed diabetes incident patients were identified as those who did not have diabetes nor pre-diabetes (5.6 <FPG<7.0 mmol/L) at baseline, while dynamically monitoring as the diabetes cases at any of the follow-up phases [with self-reporting history of diabetes, or fasting plasma glucose (FPG) >7.0 mmol/l, or glucose tolerance test >11.0 mmol/l, or with explicit inpatient medical history, or diabetes pharmacotherapy history (e.g. sulfonylurea, DPP4 inhibitor, α-GLP-1 receptor stimulant, glucosidase inhibitor, thiazolidinedione, and SGLT2 inhibitors)].

The including and excluding criteria for the ascertainment of study population were showed as follow: First, 3,837 diabetes patients at baseline were initially excluded from the baseline database. To minimize the early effect caused by undetected diabetes or pre-diabetes patient, a total of 8,872 pre-diabetes patients were further excluded which results 35,292 diabetes and pre-diabetes free population at baseline. By tracing each participant with unique ID number between baseline data and follow-up data, a total of 1,182 diabetes cases were collected during the phase I, II, and III follow-up period. Next, 622 diabetes incident cases were randomly selected from the total diabetes population to serve as the diabetes case group. The detailed including and excluding criteria of case group was showed in figure 1. An equal number of participants from the population-based control group were pair-matched to the diabetes case group on the basis of gender, age (±2 years-old), date of blood sample collection (±1 month) and occupation (worker, technician, logistical support, and manager). Except for the ascertainment of diabetes incident, All data, including blood biochemical indicators, epidemiological questionnaire, serum heavy metal concentration reported and analyzed in this study came from the baseline data (January 2011 to June 2013). Overall, the baseline blood samples were collected from a total of 622 clinically-confirmed diabetes incident cases and 622 pair-matched, population-based controls.

Figure 1.

Figure 1.

Open in a new tab

Flowchart of including and excluding criteria for the study population

All appropriate ethical and regulatory permissions were obtained before the initiation of this study. This study was approved by the Ethics Committees of Lanzhou University and the Jinchuan Nonferrous Metal Corporation, conforming to the ethical principles of the Declaration of Helsinki 2008 (sixth revision). All participants signed written informed consent before enrolling into the study.

Sample collection and quality control

Blood samples of both case and control groups for testing were collected at baseline. All 1,244 blood samples were randomly separated into 13 batches for testing, each batch containing 45 randomly chosen blood samples (0.5 cc) from the diabetes case group and 45 randomly chosen blood samples (0.5 cc) from the control group. Two internal standardized samples (0.5 cc, Ⅳ-ICPMS-71D, INORGANIC), two mix-up serum samples (0.5 cc, Ⅳ-ICPMS-28 INORGANIC) and two multi-element calibration standard samples-4 (0.5 cc, Multi-Element Calibration Standard-4, Agilent Technologies) were randomly put into each batch for quality control and coefficient of variation (CV) calculation. All internal standardized samples, mix-up serum samples and multi-element calibration standard samples were purchase from the same batch, on same date and stored at −20°C for further use. After recording the identification number of each blood sample and quality control sample, a batch of 96 samples was encased in dry ice and air-transported to the School of Public Health at Huazhong University of Science and Technology for further measurement of serum selenium. A double-blind method was strictly performed during the whole processes for the laboratory testing and transportation.

Before the initiation of the laboratory serum test, at least three rounds of quality control training were conducted for all lab technicians. Only the technicians that could produce consistently accurate results were included in the research team. Furthermore, to take into account the possibility of individual bias caused by different technicians among 13 batches of testing, only one specific technician was assigned to perform the laboratory work throughout the whole processes of measuring selenium serum levels. During the serum testing procedure, one batch of serum sample (96 samples) was divided into four groups for further testing. Additional laboratory quality control was conducted for each group of serum sample. More specifically, 24 cryogenic vials were randomly mixed up with one procedure blank, one standard reference material, one parallel sample within batches, and one spiked sample for testing.

Selenium Analysis

Each serum sample (90 μl) was normalized to room temperature followed by decomposition with 900 μl digestion solution and 810 μl ultra-pure water (MilliQ-Element, Merck) in a 2 ml Eppendorf (EP) tube (Eppendorf 250 Safe-Lock Tubes). The digestion solution (2% nitric acid diluent) was prepared as follow: 10 ml of 60% ultrapure nitrate (Fisher, ppt grade) + 3 ml n-butyl alcohol (SIGMA) + 100 mg Triton-X100 (SIGMA), then brought to 500 ml volume with Milli-Q deionized water (MilliQ-Element, Merck). Digested samples were vortexed for 20 seconds, placed in an ultrasonic bath for 20 minutes at 42 °C, and then centrifuged for 10 minutes at 15,000 rpm.. Pipetted 1.5mL supernatant for further metal levels assessed using a quadrupole Inductively Coupled Plasma Mass Spectrometry (ICP-MS, Agilent 7900, Thermo Fisher Scientific) equipped with an ESI SC-2 autosampler (Elemental Scientific, Inc.) for sample injection. During the sample injection, two internal standards and two mix-up standards (including 28 metals) were double-blinded injected and tested for drift correction and accuracy improvement. The ICP-MS was pre-heated for at least 30 minutes before every measurement, and each sample were tested for three times which presented as a mean value (μg/L). Any measurement results less than the intercept concentration were uniformly considered as a non-detected value.

Statistical methods

Conditional logistic regression models were used to analyze the association between each individual heavy metal level and the risk of developing diabetes. Multivariable adjusted odds ratios (OR) with 95% confidence interval (CI) were used to assess the risk of diabetes among different serum selenium stratums. For all analysis, selenium concentrations were categorized into quartiles based on the frequency distribution among the controls: quartile 1 (Q1) <85.45 μg/L, Q2 85.45–92.51 μg/L, Q3 95.52–103.43 μg/L, Q4 ≥103.44 μg/L. This stratification of selenium was applied in all sub-analyses among different genders, ages, BMI status, history of hypertension, and dyslipidemia.

After multiple adjustments for the conditional logistic regression models, confounders that changed the adjusted OR by less than 10% (stepwise method) were excluded from the final analysis. As a result, education level, household monthly income per capita, and tea consumption were excluded. The following confounders were included in the final regression model 1 for adjustment: age at diagnosis (<40, 40–49, 50–59, 60–69, ≥70), gender (male/female), physical exercise (no, occasionally, often), diabetes family history (yes/no), BMI (<18.5, 18.6–23.9, 24.0–27.9, ≥28.0), hypertension status [normal (<120/80 mmHg), elevated (120–129/80 mmHg), stage I (130–139/80–89 mmHg), stage II (≥140/90 mmHg)], smoking index, pack-year (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, kilogram-year (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), TC (<1.70 mmol/L, 1.70–2.25 mmol/L,≥2.20 mmol/L), HDL-C (<1.04 mmol/L, 1.05–1.55 mmol/L, ≥1.55 mmol/L), and LDL-C (<2.60 mmol/L, 2.60–4.11 mmol/L, ≥4.12 mmol/L). On the basis of the adjusted confounding factors in model 1, additional adjustments of serum heavy metals and trace elements which may alter diabetes status were performed in regression model 2. Further adjusted serum levels of heavy metals and trace elements were all treated as continues variables (μg/L) which including: Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic and Magnesium (Chen et al., 2009; Dubey et al., 2020). Generalized linear mixed models was applied to perform the interaction tests in all sub-group analyses.

To quantitively adjusted for the potential biases due to smoking and alcohol intake, the smoking index (pack-year) and alcohol index (Kilogram-year) were calculated for further statistical correction. The smoking index (pack-year) was calculated as the number of smoking packets per day × years of smoking; The alcohol index (Kilogram-year) was calculated as volume of daily alcohol beverage × years of drinking × coefficient of alcohol content. In detail, coefficient of alcohol content was defined as the average alcoholicity of each alcoholic beverages type (liquor 54%, wine 12.0%, and beer 3.0%).

To account for the possibility that the association between serum selenium levels and the incidence of diabetes may result in a non-linear correlation, Multivariable adjusted restricted cubic splines (RCS) models with 4 knots were applied to explorer the non-linear association between serum molybdenum concentration and the risk of diabetes. The determination of knots number was ascertained by comparing the criterions between Bayesian and Akaike information. The RCS models were adjusted by the potential confounding factors listed above. The conditional logistic regressions were performed by SAS program, version 9.4 (SAS Institute Inc. NC, USA). The RCS models were performed by R software (R Foundation for Statistical Computing), version 4.0.4 (package ‘rms’, ‘Hmisc’, and ‘tidyverse’). All statistical analyses were set as two-sided, and ρ<0.05 was set as the significance level for all tests.

Results

Table 1 shows the selected demographic characteristics of the study population at baseline, stratified by case-control. All demographic and biological variables are presented as number of cases (n) and frequency (%). As expected, non-significant differences were observed in age distribution, education level, smoking index, alcohol index, and exercise quantity between the diabetes case and control group. In the diabetes case group the frequency of patients with a BMI above 24.0 (68.81% vs 49.63%), hypertension stage I (30.39% vs 22.35%) and stage II (23.15% vs 9.65%), and with diabetes family history (20.50% vs 13.41%) was significantly greater when compared with controls, respectively.

Table 1.

Demographic characteristics of cohort baseline population, stratified by case-control

Demographic Baseline data, n (%)
diabetes cases Control Sum
n1=622 (freq %) n2=622 (freq %) N=1,244 (freq %)
Age (years), n (%)
 <40 56 (9.00) 53 (8.52) 109 (8.76)
 40~ 207 (33.28) 222 (35.69) 429 (34.49)
 50~ 180 (28.94) 160 (25.72) 340 (27.33)
 60~ 123 (19.77) 125 (20.10) 248 (19.94)
 ≥70 56 (9.00) 62 (9.97) 118 (9.49)
Education, n (%)
 Uneducated 24 (3.86) 20 (3.22) 44 (3.54)
 High school or less 474 (76.21) 467 (75.08) 941 (75.64)
 Undergraduate 124 (19.94) 135 (21.70) 259 (20.82)
 Graduated or above 0 (0.00) 0 (0.00) 0 (0.00)
Smoking index (pack-year), n (%) *
 No 282 (45.34) 275 (44.21) 557 (44.77)
 0~6.62 39 (6.27) 45 (7.23) 84 (6.75)
 6.63~14.07 68 (10.93) 75 (12.06) 143 (11.50)
 14.08~23.28 60 (9.65) 77 (12.38) 137 (11.01)
 ≥23.29 173 (27.81) 150 (24.12) 323 (25.96)
Alcohol index (kilogram-year), n (%) †
 No 413 (66.40) 438 (70.42) 851 (68.41)
 0~72.26 27 (4.34) 28 (4.50) 55 (4.42)
 72.27~167.53 50 (8.04) 45 (7.23) 95 (7.64)
 167.54~380.11 61 (9.81) 51 (8.20) 112 (9.00)
 ≥380.12 71 (11.41) 60 (9.65) 131 (10.53)
Exercise, n (%) ‡
 No 67 (10.77) 73 (11.74) 140 (11.25)
 Occasionally 218 (35.05) 247 (39.71) 465 (37.38)
 Often 337 (54.18) 302 (48.55) 639 (51.37)
BMI Categories, n (%)
 ≤18.5 11 (1.77) 26 (4.18) 37 (2.97)
 18.5~ 183 (29.42) 306 (49.20) 489 (39.31)
 24.0~ 284 (45.66) 230 (36.98) 514 (41.32)
 ≥28 144 (23.15) 60 (9.65) 204 (16.40)
Hypertension status, n (%)
Normal (<120/80 mmHg) 133 (21.38) 250 (40.19) 383 (45.42)
Elevated (120–129/<80 mmHg) 177 (28.46) 174 (27.97) 143 (17.33)
Stage I (130–139/80–89 mmHg) 189 (30.39) 139 (22.35) 120 (14.55)
Stage II (>=140/90 mmHg) 123 (19.77) 59 (9.49) 179 (21.70)
T2MD family history, n (%)
Yes 128 (20.50) 83 (13.41) 211 (16.38)
No 494 (79.50) 539 (86.59) 1,033 (83.05)
Open in a new tab
*

: Smoking index was categorized by the quartile distribution of pack-year based on the distribution of diabetes-free population at baseline.

†

: Alcohol index was categorized by the quartile distribution of alcohol intake weight based on the distribution of the diabetes-free population at baseline.

‡

: Exercise last more than 30 minutes was defined as a valid exercise; Occasionally exercise was defined as exercise less than 3 times per week; Often exercise is defined as at least 3 times per week, each time lasts more than 30 minutes.

Table 2 shows the means (± standard deviation), selected percentiles, and geometric means with 95% CI of serum selenium between case and control groups. The adjusted covariance significant test was used to test the difference of geometric means between case and control groups. Significantly higher geometric means and all selected percentiles of selenium (ρ<0.05) were observed in the diabetes incident case group when compared with controls.

Table 2.

Median, percentiles and geometric means with 95% confidence intervals of Se serum concentration (ng/ml) at baselinea

Group n Mean±Sth. Median 5% 25% 75% 95% CV (%) b Geometric Means (95% CI) ρ− Value c
Case 622 97.71±15.49 96.41 75.58 86.74 106.34 124.95 15.85 96.21 (95.03–97.41) <0.05
Control 622 94.33±16.70 92.52 71.45 84.45 103.44 121.81 17.70 92.99 (91.76–94.22)
Open in a new tab

Table Footnote

a

: The limit of detection (LOD) for Se was 0.0736 ng/ml.

b

: CV represented coefficient of variation, calculated as: standard deviation /means*100%.

c

: ρ represent the covariance significant test for geometric means, adjusted by: Age at diagnosis (<40, 40–49, 50–59, 60–69, ≥70), gender (male/female), BMIs (≤18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), Hypertension status (normal, elevated, stage I, stage II).

Table 3 shows the risk of diabetes incident associated with selenium concentration for males and females, respectively. A non-significant increased risk of diabetes was observed in the female population when the Q4 selenium stratum was compared with Q1 stratum (OR: 1.33, 95% CI: 0.42–4.19). However, statistically increased ORs were observed in the male and total populations, with covariate adjusted ORs in male population of 1.65 (95% CI: 1.02–2.84) and 1.74 (95% CI: 1.09–3.06) when the Q3 and Q4 stratums of selenium was compared with lowest quartile. The covariates adjusted ORs (model 2) in the total population was 1.62 (95% CI: 1.05–2.51) and 1.64 (95% CI: 1.02–2.65) when the Q3 and Q4 stratums of selenium was compared with Q1 stratum. The RCS results also demonstrated a similar result with a significant increase of diabetes risk observed only in the males within serum selenium levels between 95 μg/L to 120 μg/L, and in the total population with serum selenium level between 95 μg/L to 120 μg/L (Figure 2, sub-figure 2A–2C).

Table 3.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by gender

Gender subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
Male Q1 83 (18.20) 104 (22.66) 1.00 1.00 1.00
Q2 87 (19.08) 111 (24.18) 1.00 (0.67–1.51) 1.24 (0.77–1.99) 1.33 (0.76–2.33)
Q3 129 (28.29) 120 (26.14) 1.35 (0.91–1.99) 1.58 (1.00–2.49) 1.65 (1.02–2.84)
Q4 157 (34.43) 124 (27.02) 1.69 (1.13–2.52) 1.85 (1.15–2.97) 1.74 (1.09–3.06)
ρ for trend: 0.01 ρ for trend: 0.08 ρ for trend: 0.31
Female Q1 40 (24.10) 51 (31.29) 1.00 1.00 1.00
Q2 39 (23.49) 45 (27.61) 1.11 (0.61–2.01) 1.12 (0.55–2.28) 0.98 (0.41–2.36)
Q3 47 (28.31) 35 (21.47) 1.62 (0.90–2.92) 1.93 (0.94–3.94) 1.76 (0.76–4.09)
Q4 40 (24.10) 32 (19.63) 1.55 (0.79–3.06) 1.69 (0.79–3.77) 1.33 (0.42–4.19)
ρ for trend: 0.32 ρ for trend: 0.32 ρ for trend:0.59
Total Q1 123 (19.77) 155 (24.92) 1.00 1.00 1.00
Q2 126 (20.26) 156 (25.08) 1.05 (0.75–1.47) 1.26 (0.86–1.85) 1.29 (0.82–2.03)
Q3 176 (28.30) 155 (24.92) 1.42 (1.03–1.96) 1.62 (1.17–2.35) 1.62 (1.05–2.51)
Q4 197 (31.67) 156 (25.08) 1.70 (1.21–2.38) 1.79 (1.21–2.64) 1.64 (1.02–2.65)
ρ for trend: <0.01 ρ for trend: 0.12 ρ for trend:0.20
Open in a new tab

Note: The interaction test between gender and serum Se stratums: F=1.97, ρ=0.06.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by age at diagnosis (<30, 30–39, 40–49, 50–59, 60–69, ≥70), BMIs (<18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), triglyceride (<1.80 mmol/L, 1.80~2.20 mmol/L, ≥2.20 mmol/L), HDL-C(<1.04 mmol/L, 1.04–1.55 mmol/L, ≥1.55 mmol/L), LDL-C (<3.37 mmol/L, 3.37–4.11 mmol/L, ≥4.12 mmol/L), and hypertension status (normal, elevated, stage I, stage II).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

Figure 2. Sub-analyses of diabetes risk and baseline serum selenium level based on the RCS model, stratified by gender, age, and BMI.

Figure 2.

Open in a new tab

Note: 1) Totally 9 figures were contained in figure 2, the detailed subgroups were listed as follow from left to right: 1) up row: males, females, total study population, 2) middle row: age<45 years-old, age between 45–64 years-old, age≥65 years-old, 3) down row: BMI<24.0, BMI between 24.0–27.9, BMI≥28.

2) Chi-square tests, ρ values for the non-linearity between serum selenium and diabetes ORs were listed as follow: 2A (F=6.07, ρ=0.04), 2B (F=5.62, ρ=0.04), 2C (F=5.26, ρ=0.07), 2D (F=3.54, ρ=0.10), 2E (F=4.79, ρ=0.08), 2F (F=5.89, ρ=0.05), 2G (F=11.82, ρ<0.01), 2H (F=2.11, ρ=0.18), 2I (F=0.89, ρ=0.51)

Table 4 shows sub-group analysis of the association between serum selenium levels and diabetes risk among different age groups. No significant increase in risk of developing diabetes was observed in the population under 45 years-old. However, significantly increased adjusted OR of 1.49 (95%CI: 1.01–2.12) was observed in the population between 45–64 years old when Q4 was compared with the reference group. A significantly increased covariates-adjusted OR of 2.91 (95%CI: 1.06–7.97) was observed in the population over 65 years-old when the Q3 of selenium was compared with Q1. The RCS dose-respond analysis demonstrated that the risk of diabetes was significantly increased when the selenium serum level was between 95 μg/L to 110 μg/L in the population under 45 years old, and between 100 μg/L to 120 μg/L in the 45 to 64 years old population (Figure 2, sub-figure 2D–2F).

Table 4.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by age

Age subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
<45 Q1 23 (18.40) 31 (22.79) 1.00 1.00 1.00
Q2 27 (21.60) 45 (33.09) 0.85 (0.39–1.85) 0.98 (0.33–2.92) 0.68 (0.28–1.40)
Q3 41 (32.80) 33 (24.26) 1.77 (0.86–4.13) 2.33 (0.63–6.56) 1.39 (0.53–3.63)
Q4 34 (27.20) 27 (19.85) 2.35 (0.94–5.90) 2.45 (0.70–6.55) 1.13 (0.32–2.75)
ρ for trend: 0.02 ρ for trend: 0.11 ρ for trend:0.27
45–64 Q1 74 (19.63) 80 (21.98) 1.00 1.00 1.00
Q2 77 (20.42) 90 (24.73) 0.80 (0.50–1.28) 1.09 (0.68–1.75) 0.90 (0.53–1.54)
Q3 97 (25.73) 98 (26.92) 1.00 (0.65–1.54) 1.30 (0.82–2.05) 1.12 (0.66–1.90)
Q4 129 (34.22) 96 (26.37) 1.49 (0.94–2.37) 1.74 (1.10–2.75) 1.49 (1.01–2.12)
ρ for trend: 0.01 ρ for trend: 0.04 ρ for trend:0.15
>=65 Q1 26 (21.67) 44 (36.07) 1.00 1.00 1.00
Q2 22 (18.33) 21 (17.21) 2.01 (0.91–4.44) 2.17 (0.76–6.20) 1.71 (0.59–4.92)
Q3 38 (31.67) 24 (19.67) 3.26 (1.44–7.36) 4.60 (1.50–8.02) 2.91 (1.06–7.97)
Q4 34 (28.33) 33 (27.05) 1.84 (0.89–3.80) 1.55 (0.60–4.00) 1.36 (0.95–3.68)
ρ for trend: 0.83 ρ for trend: 0.50 ρ for trend:0.68
Open in a new tab

Note: The interaction test between age-group and serum Se stratums: F=1.91, ρ=0.035.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by gender (male/female), BMIs (<18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), triglyceride (<1.70 mmol/L, 1.70~2.25 mmol/L, >2.20 mmol/L), high-density lipoprotein (<1.04 mmol/L, 1.04–1.55 mmol/L, >1.55 mmol/L), low-density lipoprotein (<2.60 mmol/L, 2.60–4.11 mmol/L, >4.12 mmol/L), and hypertension status (normal, elevated, stage I, stage II).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

The sub-analysis of different BMI groups is shown in table 5. The OR was not significantly increased in the population with a BMI ≥ 28. However, in the populations with a BMI under 28, a significant increased adjusted OR was observed. An OR of 2.39 (95%CI: 1.15–4.05) was observed in the population with a BMI between 24.0–27.9 population when the Q4 selenium was compared with Q1 selenium. In the population with a BMI under 24, both Q3 (OR: 2.06, 95%CI: 1.07–3.97) and Q4 (OR: 2.21, 95%CI: 1.07–4.03) had significantly increased adjusted ORs in comparison with the lowest quartile of serum selenium. Similar results were also observed in the RCS dose-response analysis, lower serum selenium level (<95 μg/L) was negatively correlated with the risk of diabetes in the population with BMI’s under 24. No variation of adjusted risk was found in the obese population. (Figure 2, sub-figure 2G–2I).

Table 5.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by BMI categories

BMI subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
<24.0 Q1 32 (16.49) 82 (24.70) 1.00 1.00 1.00
Q2 26 (13.40) 98 (29.52) 0.68 (0.38–1.23) 0.63 (0.34–1.19) 0.61 (0.30–1.22)
Q3 66 (34.02) 75 (22.59) 2.26 (1.33–3.81) 3.07 (1.24–5.81) 2.06 (1.07–3.97)
Q4 70 (36.08) 77 (23.19) 3.33 (2.18–3.92) 4.25 (1.28–7.94) 2.21 (1.12–4.03)
ρ for trend: <0.01 ρ for trend: <0.01 ρ for trend: <0.01
24.0–27.9 Q1 54 (19.01) 57 (24.78) 1.00 1.00 1.00
Q2 66 (23.24) 45 (19.57) 2.20 (0.97–5.48) 2.50 (0.79–4.86) 1.46 (0.78–2.73)
Q3 74 (26.06) 65 (28.26) 1.11 (0.47–2.63) 1.63 (0.49–5.45) 1.21 (0.67–2.19)
Q4 90 (31.69) 63 (27.39) 2.37 (1.12–6.11) 3.74 (1.09–6.85) 2.39 (1.15–4.05)
ρ for trend: 0.87 ρ for trend: 0.50 ρ for trend: <0.01
>=28 Q1 37 (25.69) 16 (26.67) 1.00 1.00 1.00
Q2 34 (23.61) 13 (21.67) 1.59 (0.20–2.69) 1.06 (0.40–2.78) 1.00 (0.34–2.93)
Q3 36 (25.00) 15 (25.00) 1.79 (0.45–2.41) 1.37 (0.54–3.51) 1.14 (0.39–3.07)
Q4 37 (25.69) 16 (26.67) 1.79 (0.44–2.29) 1.02 (0.41–2.57) 0.86 (0.30–2.53)
ρ for trend: 0.78 ρ for trend: 0.94 ρ for trend: 0.84
Open in a new tab

Note: The interaction test between BMI group and serum Se stratums: F=9.01, ρ<0.0001.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by age at diagnosis (<30, 30–39, 40–49, 50–59, 60–69, ≥70), gender (male/female), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), triglyceride (<1.70 mmol/L, 1.70~2.25 mmol/L, >2.20 mmol/L), high-density lipoprotein (<1.04 mmol/L, 1.04–1.55 mmol/L, >1.55 mmol/L), low-density lipoprotein (<2.60 mmol/L, 2.60–4.11 mmol/L, >4.12 mmol/L), and hypertension status (normal, elevated, stage I, stage II).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

Appendix table 1 demonstrate the comparison of demographic variables between random sampled 622 cases group and total 1,182 incident diabetes cases, the results showed that the random sampling from total diabetes incident cases did not bring any potential bias caused by major risk factors of diabetes which including: age, smoking, alcohol drinking, exercise, BMI, diabetes family history, and etc. Appendix tables 2–6 showed that LDL-C, TC, TG, HDL-C, and SUA were all significantly higher in the diabetes incident cases when compared with controls, as expected (Appendix Table 1). A significantly increased risk of developing diabetes was observed in the blood pressure elevated population (OR: 1.93, 95% CI: 1.00–4.06) and hypertension population (OR: 1.63, 95% CI: 1.05–2.72) when the Q4 was compared with lowest quartile (Appendix Table 2). An OR of 2.20 (95%CI: 1.04–5.04) was observed when the Q3 compared with Q1 among the population with higher TG levels (Appendix Table 3). Among moderate to high levels of HDL-C subgroups, significantly increased risk of developing diabetes was found when the Q4 compared with the Q1 (Appendix Table 4), while a non-significantly increased risk of developing diabetes was discovered among all LDL-C subgroups (Appendix Table 5).

Discussion

In this nested case-control study, high baseline serum selenium levels were significantly associated with increased risk of diabetes in 5.8 years of follow-up. The observed increased incidence risk of diabetes was stronger in men (Q4 vs Q1 OR:1.74, 95%CI: 1.09–3.06), participants over 45 years-old (age 45–64, OR:1.49, 95%CI: 1.01–2.12; age over 65, OR: 2.91, 95%CI:1.06–7.97), non-obese (BMI<24.0, OR: 2.21, 95%CI: 1.12–4.03; BMI 24.0–27.9, OR: 2.39, 95%CI: 1.15–4.05), and participants’ blood pressure beyond normal (OR of elevated stage: 1.93, 95%CI: 1.00–4.06; OR of hypertension: 1.63, 95% CI: 1.05–2.72). These observed associations all demonstrated a clear non-linear, dose-response between the incidence risk of diabetes and serum selenium levels, and these significantly increased risks of diabetes was mainly observed between serum selenium level ranging from 95 μg/L to 120 μg/L. Although previous studies have correlated a variety of risk factors with the prevalence of diabetes, the majority of the these traditional risk factors are based on physical, physiological or lifestyles elements and not internal measures of possible exposures. Therefore, our study is one of the first prospective studies to explore the epidemiological etiology of diabetes associated with internal selenium levels.

The results of this study are in agreement with recent cross-sectional studies relating high-level blood selenium concentration to the diabetes risk. Two cross-sectional surveys from National Health and Nutrition Examination Survey (NHANES) (Bleys et al., 2007; Laclaustra et al., 2009) that also reported significantly increased risk of diabetes associated with high selenium status or dietary selenium supplementary among United States adults. Both NHANES studies together covered more than 9,000 people and all found higher selenium concentrations were correlated with diabetes risks with an OR (95% CI) of 1.57 (1.16, 2.13) (Bleys et al., 2007) and 7.64 (3.34, 17.46) (Laclaustra et al., 2009) when the highest serum selenium quartile was compared to lowest quartile. An Italian-based selenium supplemental intervention study conducted by Stranges et al (Stranges et al., 2007) also found similar results. After 8 years of selenium supplementation (200 μg/d) an adjusted hazard ratio (HR) of diabetes was observed (HR: 1.55, 95%CI: 1.03–2.33) when compared with the control placebo group, and a HR of 2.70 (95%CI: 1.30–5.61) was discovered when the highest selenium quartile was compared with lowest (HR: 2.70, 95%CI: 1.30–5.61). In a Chinese-based cross-sectional, Zhang T et al. (Zhang et al., 2017) showed serum selenium concentrations were substantially higher (ρ< 0.001) in diabetic patients when compared to non-diabetic patients, and a significantly increased OR was also observed (OR: 2.69, 95%CI: 1.31–3.49) when patients in the highest selenium quartile were compared with those in the lowest quartile. Another European population-based nutritional supplemental intervention study (Antioxidant Vitamins and Minerals, SU.VI.MAX) found a positive association between plasma selenium concentration and fasting glucose (Czernichow et al., 2006).

In contrast, some studies have shown no association between blood levels of selenium and diabetes development. For example, two studies from the Nord-Trøndelag Health Survey (HUNT-3) in Norway found no overall association between selenium and diabetes risk (Hansen et al., 2017; Simic et al., 2017). Furthermore, in a randomized, placebo-controlled trial involving more than 35,000 North American males (The Selenium and Vitamin E Cancer Prevention Trial, SELECT), selenium supplementation (200 μg/d) showed a nonsignificant increased risk of diabetes when compared with the placebo group (relative risk: 1.07, 95%CI: 0.94–1.22). Despite the potential biases (older age, laboratorial testing, metabolic pathway of selenium, and etc.) involved in the controversial studies listed above, the underling mechanism for the differential findings between positive and negative health effects of selenium are still unclear.

These observed differences of diabetes risk associated with selenium may, in part, be explained by the large variation in serum selenium concentration across studies conducted in various regions and in different races (Norwegian: 100–105 μg/L (Hansen et al., 2017; Simic et al., 2017); Italian: selenium intervention group: around 175 μg/L (Stranges et al., 2007); NHANES, 2003–2004: 137.1 μg/L (Laclaustra et al., 2009); NHANES, Phase III: 126.8 μg/L (Bleys et al., 2007)). The studies with positive associations listed above were all conducted in populations with high selenium levels (above the Norwegian range of 100–105 μg/L) when compared with those studies with no association. Therefore, suggesting that serum selenium levels above 100–105 mg/ml may increase the risk of developing diabetes, however, another study performed by Stranges S (Stranges et al., 2011) et al. showed a positive association between serum selenium levels at relatively lower concentrations (77.5±18.4 μg/L) and diabetes incidence risk. Similarly, in our study a mean serum selenium level of 95.87±16.17 μg/L, which is less than the reported levels in the Norwegian study, was associated with a significant increase in diabetes risk in the highest selenium quartile when compared with the lowest quartile. Additionally, a substantially higher geometric mean of selenium (96.21 μg/L) was also observed in the cases group when compared with controls (92.99 μg/L). Thus, it remains unclear whether there is a universal threshold of selenium concentration that alters diabetes risk or if it is dependent on ethnicity and/or geographic region. Nonetheless, our non-linear, dose-response correlation may support the perspective that baseline selenium serum concentration ranging from 95 μg/L to 120 μg/L may increase the risk of diabetes among this cohort population. Given the inconsistent evidence listed above, the association between internal selenium level and the diabetes risk are more tentative than corroborative.

An increased risk of diabetes associated with higher serum levels of selenium is biologically plausible. Fürnsinn C et al. (Fürnsinn et al., 1996) showed selenite and selenate can trigger the transport of glucose through an unspecific reaction to toxic or osmotic stress in conjunction with a catabolic response. In a study performed by Satyanarayana S et al. (Satyanarayana et al., 2006), high dietary selenium intake stimulated the release of glucagon and subsequent hyperglycemia in an Albino rat model. Abnormal expression of the selenium-containing protein, glutathione peroxidase 1 (GPx-1) is another possible mechanism linking high selenium exposure to the development of diabetes (Rayman and Stranges, 2013). In mammals, Gpx-1 activity is highly dependent on blood selenium concentrations. Animal experiments found that activation of Gpx-1 downregulated the insulin signal pathway while the overexpression of Gpx-1 resulted in obesity and insulin resistance (Allan et al., 1999; Hamanishi et al., 2004; McClung et al., 2004). Additionally, such selenomethionine enzymes are responsible for the creation of the metabolite methaneselenol which has been shown to mediate overproduction of reactive oxygen species and subsequent oxidative damage in pancreatic β cells (Rayman and Stranges, 2013). Another selenium-cored, liver-derived secretory protein, selenoprotein P (SeP) may be responsible for the insulin resistance. Selenoprotein P is an extracellular protein that harboring approximately 10 selenium-based selenocysteines which mainly respond as the peroxidase for glutathione, dithiothreitol, cysteine, and homocysteine (Saito et al., 1999). Misu et al. found significant in vivo evidence that SeP was positively correlated with the increment of serum glucose, FPG, and HbA1c (Misu et al., 2010; Tsutsumi and Saito, 2020). The underline mechanism may associate with the SeP-mediated disruption via adenosine monophosphate activated protein kinase (AMPK)-independent pathways in the insulin signal cascade at various levels between hepatocytes and myocytes. Overall, the metabolic mechanisms of selenium-associated diabetes are not yet fully understood. Thus, further molecular biology, toxicology, and prospective studies involving the mechanism of selenium correlated with diabetes are warranted. Large population-based selenium intervention studies are also needed to determine the safety threshold of selenium and to better understand selenium heterogeneity among different demographic including regions, races, ages, and genders. Given the occupational feature of this cohort population, it is plausible that serum Se level may affected via occupation exposure. However, based on the report from World Health Organization (International Programme on Chemical et al., 1987), the selenium intake pathway may mainly from dietary and water. While other exposure pathway including soil, rocks (usually < 1mg/kg), and air, have large variations in different regions. Moreover, other industrial sources of selenium intake pathway including burning of fossil fuels (coal: 2.8 mg/kg to 10.65 mg/kg), production of copper, lead, zinc, phosphate, and uranium may not be applicable in our study, since the Jinchang cohort population was not involved in any of the above-mentioned products. But still, the industrial sources of selenium exposure in this cohort population are plausible but must be validated by the future environmental monitoring data.

The prospective design of this study was a major strength, resulting in the confirmation that serum selenium levels are associated with increased risk of diabetes. Based on this large-population cohort study, efficient data was collected from incident patients with diabetes to further explorer the true nature connecting selenium exposure to the risk of diabetes. Meanwhile, further testing for more newly diagnosed diabetes patients is still being performed during the writing of this article and we anticipate the results will increase the reliability of our study. Moreover, since the Jinchang Cohort is an ongoing cohort study, the phase III follow-up is still ongoing (December 2020 to December 2022) and will provide further information about the health effects of selenium exposure and disease development. Although we were unable to obtain detailed nutrition intake history, such as different food sources of antioxidants or daily dietary selenium, a dietary habits questionnaire was used in our study to determine fruit intake, vegetable intake, total carbohydrate intake, red/smoked/salted meat intake, etc. There were no significant differences in dietary patterns found during our pre-study between the cases and controls groups. Antioxidant and oxidative stress levels were not measured in our study; therefore, future work will involve these measurements in order to better understand their potential roll in the development of selenium-associated diabetes.

Conclusion

The results of this nested case-control study are in line with previous cross-sectional and experimental studies that high serum levels of Se are positively associated with the increasement of diabetes incident risk. These increased risks of diabetes by serum levels of Se. appeared to differ by sex, age, BMI status, history of hypertension, and dyslipidemia. Furthermore, Se levels between 95 and 120 μg/L would increase the diabetes incident risk in this cohort population.

The highlights.

Growing interest in the function of trace elements in the development of diabetes.

High serum level of selenium at baseline could increase the incidence risk of diabetes.

Se levels between 95 and 120 μg/L would increase the diabetes incident risk in this cohort population.

Population with specific features may be more vulnerable to selenium exposure.

Acknowledgments

We would like to sincerely thank all of the participants and researchers from Jinchang Nonferrous Metal Company, the Worker’s Hospital of JNMC, Huazhong University of Science and Technology, Brown University, and Lanzhou University in the Jinchang Cohort Study for their contributions, collaboration, and enthusiasm.

Funding

This work was supported in part by grants: Lanzhou University (2018ldbrzd008), National Natural Science Foundation of China (No. 81673248), and the Foundation for the National Institutes of Health (U.S.) (NIH R01ES029082). The funding organizations had no role in the design, analysis, and conduct of the study.

Appendix

Appendix Table 1.

Comparison of demographic data between random sampled diabetes cases and total diabetes incident cases

Demographic Baseline data, n (%)
Random cases Total incident diabetes cases
n1=622 (freq %) n2=1,182 (freq %%)
Age (years), n (%)
 <40 56 (9.00) 102 (8.66)
 40~ 207 (33.28) 422 (35.74)
 50~ 180 (28.94) 309 (26.17)
 60~ 123 (19.77) 226 (19.11)
 ≥70 56 (9.00) 122 (10.32)
Education, n (%)
 Uneducated 24 (3.86) 30 (2.51)
 High school or less 474 (76.21) 877 (74.18)
 Undergraduate 124 (19.94) 216 (18.31)
 Graduated or above 0 (0.00) 59 (5.00)
Smoking index (pack-year), n (%) *
 No 282 (45.34) 523 (44.26)
 0~6.62 39 (6.27) 90 (7.65)
 6.63~14.07 68 (10.93) 105 (8.88)
 14.08~23.28 60 (9.65) 141 (11.91)
 ≥23.29 173 (27.81) 323 (27.29)
Alcohol index (kilogram-year), n (%) †
 No 413 (66.40) 825 (69.82)
 0~72.26 27 (4.34) 55 (4.62)
 72.27~167.53 50 (8.04) 79 (6.71)
 167.54~380.11 61 (9.81) 91 (7.73)
 ≥380.12 71 (11.41) 131 (11.12)
Exercise, n (%) ‡
 No 67 (10.77) 98 (8.29)
 Occasionally 218 (35.05) 443 (37.51)
 Often 337 (54.18) 641 (54.2)
BMI Categories, n (%)
 ≤18.5 11 (1.77) 30 (2.56)
 18.5~ 183 (29.42) 321 (27.18)
 24.0~ 284 (45.66) 565 (47.77)
 ≥28 144 (23.15) 266 (22.49)
Hypertension status, n (%)
Normal (<120/80 mmHg) 133 (21.38) 250 (21.16)
Elevated (120–129/<80 mmHg) 177 (28.46) 340 (28.74)
Stage I (130–139/80–89 mmHg) 189 (30.39) 333 (28.16)
Stage II (>=140/90 mmHg) 123 (19.77) 259 (21.95)
diabetes family history, n (%)
Yes 128 (20.50) 231 (19.52)
No 494 (79.50) 951 (80.48)
Open in a new tab
*

: Smoking index was categorized by the quartile distribution of pack-year based on the distribution of diabetes-free population at baseline.

†

: Alcohol index was categorized by the quartile distribution of alcohol intake weight based on the distribution of the diabetes-free population at baseline.

‡

: Exercise last more than 30 minutes was defined as a valid exercise; Occasionally exercise was defined as exercise less than 3 times per week; Often exercise is defined as at least 3 times per week, each time lasts more than 30 minutes.

Appendix Table 2.

Median, percentiles and geometric means with 95% confidence intervals of different biochemical criterion at baseline population

Blood biochemical items Group n Mean±Sth. Median 5% 25% 75% 95% Geometric Means (95%CI) ρ− Value *
Total Cholesterol (mmol/L) Case 622 4.93±0.97 4.89 3.44 4.29 5.52 6.50 4.83 (4.76–4.91) <0.05
Control 622 4.69±0.89 4.64 3.40 4.12 5.17 6.20 4.61 (4.54–4.68)
Triglyceride (mmol/L) Case 622 2.44±1.82 1.90 0.89 1.39 2.88 5.96 2.04 (1.96–2.14) <0.05
Control 622 1.72±1.17 1.41 0.70 1.04 2.04 3.73 1.49 (1.43–1.55)
High-density Lipoprotein cholesterol (mmol/L) Case 622 1.26±0.34 1.22 0.84 1.03 1.43 1.78 1.22 (1.20–1.24) <0.05
Control 622 1.35±0.42 1.31 0.84 1.09 1.54 2.02 1.30 (1.27–1.33)
Low-density Lipoprotein cholesterol (mmol/L) Case 622 2.82±1.02 2.74 1.26 2.04 3.58 4.14 2.61 (2.53–2.70) 0.72
Control 622 2.75±0.92 2.75 1.23 2.13 3.37 4.18 2.58 (2.50–2.66)
Serum Uric Acid (μmol/L) Case 622 351.0±84.78 347.00 228.00 293.00 398.00 492.00 341.0 (334.6–347.6) <0.05
Control 622 325.92±71.97 322.50 220.00 275.00 368.00 463.00 318.1 (312.5–323.7)
Open in a new tab
*

: ρ represent the covariance significant test for geometric means, adjusted by: Age at diagnosis (<40, 40–49, 50–59, 60–69, ≥70), gender (male/female), BMIs (≤18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), hypertension status (normal, elevated, stage I, stage II).

Appendix Table 3.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by hypertension status

Hypertension subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
Normal (<120/80 mmHg) Q1 35 (19.77) 43 (24.71) 1.00 1.00 1.00
Q2 33 (18.64) 50 (28.74) 1.01 (0.43–2.23) 0.77 (0.40–1.50) 1.09 (0.53–2.26)
Q3 55 (31.07) 43 (24.71) 1.97 (0.86–3.25) 1.70 (0.89–3.25) 1.24 (0.58–2.64)
Q4 54 (30.51) 38 (24.84) 2.21 (1.21–3.52) 1.85 (0.94–3.66) 1.31 (0.68–2.90)
ρ for trend: 0.01 ρ for trend: 0.01 ρ for trend: 0.63
Elevated (120–129/80 mmHg) Q1 32 (16.93) 30 (21.58) 1.00 1.00 1.00
Q2 38 (20.11) 28 (20.14) 1.98 (0.47–4.34) 1.46 (0.67–3.20) 0.67 (0.32–1.43)
Q3 63 (33.33) 39 (28.06) 1.46 (0.47–4.51) 1.74 (0.84–3.62) 1.48 (0.72–3.04)
Q4 56 (29.63) 42 (30.22) 1.83 (0.49–5.84) 1.30 (0.63–2.66) 1.93 (1.00–4.06)
ρ for trend: 0.92 ρ for trend: 0.73 ρ for trend: 0.01
Hypertension (≥130/80 mmHg) Q1 56 (21.88) 82 (26.54) 1.00 1.00 1.00
Q2 55 (21.48) 78 (25.24) 1.03 (0.64–1.68) 1.26 (0.74–2.14) 1.21 (0.61–2.38)
Q3 58 (22.66) 73 (23.62) 1.16 (0.72–1.89) 1.34 (0.79–2.28) 1.27 (0.71–2.39)
Q4 87 (33.98) 76 (24.60) 1.68 (1.06–2.65) 1.74 (1.08–2.90) 1.63 (1.05–2.72)
ρ for trend: 0.04 ρ for trend: 0.21 ρ for trend: 0.42
Open in a new tab

Note: The interaction test between hypertension status and serum Se stratums: F=2.62, ρ=0.003.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by age at diagnosis (<30, 30–39, 40–49, 50–59, 60–69, ≥70), gender (male/female), BMIs (<18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), triglyceride (<1.70 mmol/L, 1.70~2.25 mmol/L, >2.20 mmol/L), high-density lipoprotein (<1.04 mmol/L, 1.04–1.55 mmol/L, >1.55 mmol/L), low-density lipoprotein (<2.60 mmol/L, 2.60–4.11 mmol/L, >4.12 mmol/L).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

Appendix Table 4.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by triglyceride levels

TG subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
Low (<1.70 mmol/L) Q1 57 (22.71) 105 (26.79) 1.00 1.00 1.00
Q2 52 (20.72) 97 (24.74) 1.03 (0.53–2.01) 1.08 (0.66–1.79) 0.95 (0.53–1.70)
Q3 74 (29.48) 109 (27.81) 0.87 (0.46–1.62) 1.37 (0.86–2.19) 1.20 (0.71–2.05)
Q4 68 (27.09) 81 (20.66) 1.23 (0.65–2.33) 1.82 (1.10–3.01) 1.66 (0.96–2.83)
ρ for trend: 0.61 ρ for trend: 0.04 ρ for trend: 0.10
Median (1.70~2.25 mmol/L) Q1 22 (16.42) 21 (19.63) 1.00 1.00 1.00
Q2 32 (23.88) 28 (26.17) 1.09 (0.50–2.39) 1.23 (0.50–3.07) 0.84 (0.28–2.57)
Q3 34 (25.37) 24 (22.43) 1.35 (0.61–2.99) 1.69 (0.69–4.18) 2.00 (0.63–6.34)
Q4 46 (34.33) 34 (31.78) 1.29 (0.61–2.72) 1.74 (0.74–4.13) 1.37 (0.44–3.62)
ρ for trend: 0.24 ρ for trend: 0.39 ρ for trend: 0.40
High (≥2.25 mmol/L) Q1 44 (18.57) 29 (23.58) 1.00 1.00 1.00
Q2 42 (17.72) 31 (25.20) 0.89 (0.46–1.73) 0.99 (0.48–2.03) 1.18 (0.52–2.67)
Q3 68 (28.69) 22 (17.89) 2.04 (1.04–3.99) 2.58 (1.26–5.28) 2.20 (1.04–5.04)
Q4 83 (35.02) 41 (33.33) 1.33 (0.73–2.43) 1.46 (0.76–2.79) 1.14 (0.53–2.49)
ρ for trend: 0.19 ρ for trend: 0.23 ρ for trend: 0.94
Open in a new tab

Note: The interaction test between triglyceride categories and serum Se stratums: F=7.09, ρ<0.0001.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by age at diagnosis (<30, 30–39, 40–49, 50–59, 60–69, ≥70), gender (male/female), BMIs (<18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), high-density lipoprotein (<1.04 mmol/L, 1.04–1.55 mmol/L, >1.55 mmol/L), low-density lipoprotein (<2.60 mmol/L, 2.60–4.11 mmol/L, >4.12 mmol/L), and hypertension status (normal, elevated, stage I, stage II).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

Appendix Table 5.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by HDL-C levels

HDL-C subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
Low (<1.04 mmol/L) Q1 39 (24.68) 32 (26.67) 1.00 1.00 1.00
Q2 38 (24.05) 34 (28.33) 0.92 (0.48–1.77) 0.93 (0.45–1.93) 0.77 (0.33–1.77)
Q3 46 (29.11) 22 (18.33) 1.72 (0.86–3.42) 2.34 (1.08–5.10) 2.11 (0.97–5.08)
Q4 35 (22.15) 32 (26.67) 0.88 (0.46–1.75) 0.87 (0.42–1.83) 0.69 (0.29–1.65)
ρ for trend: 0.95 ρ for trend: 0.86 ρ for trend: 0.81
Medium (1.04~1.55 mmol/L) Q1 66 (18.03) 81 (23.08) 1.00 1.00 1.00
Q2 71 (19.40) 77 (21.94) 1.13 (0.72–1.79) 1.35 (0.82–2.23) 1.32 (0.75–2.33)
Q3 107 (29.23) 100 (28.49) 1.31 (0.86–2.01) 1.50 (0.95–2.38) 1.41 (0.84–2.36)
Q4 122 (33.33) 93 (26.50) 1.61 (1.06–2.46) 1.67 (1.05–2.66) 1.66 (1.05–2.76)
ρ for trend: 0.10 ρ for trend: 0.37 ρ for trend: 0.39
High (≥1.55 mmol/L) Q1 18 (18.37) 42 (27.81) 1.00 1.00 1.00
Q2 17 (17.35) 45 (29.80) 0.88 (0.40–1.93) 0.63 (0.25–1.59) 0.71 (0.25–1.99)
Q3 23 (23.47) 33 (21.85) 1.63 (0.76–3.50) 1.42 (0.59–3.41) 1.60 (0.53–4.86)
Q4 40 (40.82) 31 (20.53) 3.01 (1.46–6.21) 3.05 (1.27–7.32) 3.17 (1.13–8.89)
ρ for trend: <0.01 ρ for trend: <0.01 ρ for trend: <0.01
Open in a new tab

Note: The interaction test between HDL-C categories and serum Se stratums: F=3.51, ρ<0.0001.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by age at diagnosis (<30, 30–39, 40–49, 50–59, 60–69, ≥70), gender (male/female), BMIs (<18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), triglyceride (<1.70 mmol/L, 1.70~2.25 mmol/L, >2.20 mmol/L), high-density lipoprotein (<1.04 mmol/L, 1.04–1.55 mmol/L, >1.55 mmol/L), and hypertension status (normal, elevated, stage I, stage II).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

Appendix Table 6.

Multi-variables adjusted odds ratio of diabetes incident associated with Se stratums, sub-categorized by LDL-C levels

LDL-C subgroups Stratums Incidence cases, n Control, n OR*(95% CI) Crude OR† (95% CI) Model 1 OR‡ (95% CI) Model 2
Low (<2.59 mmol/L) Q1 68 (24.46) 81 (30.00) 1.00 1.00 1.00
Q2 52 (18.71) 56 (20.74) 1.03 (0.53–1.99) 1.08 (0.49–2.36) 1.22 (0.65–2.29)
Q3 80 (28.78) 68 (25.19) 1.33 (0.72–2.42) 1.43 (0.70–2.94) 1.57 (0.89–2.76)
Q4 78 (28.06) 65 (24.07) 1.54 (0.85–2.78) 1.46 (0.74–2.88) 1.53 (0.84–2.79)
ρ for trend: 0.25 ρ for trend: 0.46 ρ for trend: 0.50
Medium (2.59~3.34 mmol/L) Q1 22 (14.38) 37 (19.79) 1.00 1.00 1.00
Q2 32 (20.92) 53 (28.34) 1.02 (0.51–2.02) 1.04 (0.49–2.22) 0.89 (0.38–2.08)
Q3 49 (32.03) 56 (29.95) 1.47 (0.77–2.83) 1.74 (0.85–3.58) 1.53 (0.67–3.49)
Q4 50 (32.68) 41 (21.93) 2.05 (1.05–4.01) 1.65 (0.77–3.50) 1.75 (0.75–4.10)
ρ for trend: 0.02 ρ for trend: 0.19 ρ for trend: 0.08
High (>=3.34 mmol/L) Q1 33 (17.28) 37 (22.42) 1.00 1.00 1.00
Q2 42 (21.99) 47 (28.48) 1.00 (0.54–1.88) 1.11 (0.55–2.25) 1.18 (0.51–2.72)
Q3 47 (24.61) 31 (18.79) 1.70 (0.89–3.27) 1.85 (0.89–3.84) 1.87 (0.77–4.53)
Q4 69 (36.13) 50 (30.30) 1.55 (0.85–2.80) 1.90 (0.96–3.76) 1.83 (0.81–4.12)
ρ for trend: 0.12 ρ for trend: 0.09 ρ for trend: 0.24
Open in a new tab

Note: The interaction test between LDL-C categories and serum Se stratums: F=1.85, ρ=0.042.

*

: Represent the crude odds ratios

†

: Odds ratios were adjusted by age at diagnosis (<30, 30–39, 40–49, 50–59, 60–69, ≥70), gender (male/female), BMIs (<18.5, 18.5–23.9, 24.0–27.9, ≥28.0), family history of diabetes (yes or no), physical exercise (no, occasionally, often), smoking index, (no smoking, 0–6.3, 6.4–14.0, 14.1–23.2, ≥23.3), lifetime total alcohol intake, (no drinking, 0–72.2, 72.3–167.5, 167.5–380.1, ≥380.1), triglyceride (<1.80 mmol/L, 1.80~2.20 mmol/L, ≥2.20 mmol/L), high-density lipoprotein (<1.04 mmol/L, 1.04–1.55 mmol/L, ≥1.55 mmol/L), and hypertension status (normal, elevated, stage I, stage II).

‡

: Odds ratios in model 2 were further adjusted by serum level (μg/L) of Nickel, Cobalt, Copper, Zinc, Cadmium, Mercury, Chromium, Arsenic, and Magnesium (continues variables) on the basis of adjusted confounding factors in model 1.

Footnotes

Conflict of interest,

The author(s) declare(s) that there is no potential conflict of interest with respect to the research, authorship, and/or publication in this paper.

Consent for publication

This study does not contain any personal privacy information in any form whatsoever (including any individual details, images or videos). All the information provided by the participants was anonymized during the entire data analysis, paper writing and review process. The author assumes full responsibility for any release or disclosure related to this paper.

Availability of data and materials

The datasets generated during the current study are not publicly available due to the protection of privacy of all enrolled participants. However, qualified scientific and medical researchers can request the data that underlie the results reported in this article. Proposals for data will be evaluated and approved by corresponding author in her sole discretion. All approved researchers must sign a data access agreement before accessing the data by de-identified all participants’ information.

References

  1. Alanne M, Kristiansson K, Auro K, Silander K, Kuulasmaa K, Peltonen L, et al. Variation in the selenoprotein S gene locus is associated with coronary heart disease and ischemic stroke in two independent Finnish cohorts. Hum Genet 2007; 122: 355–65. [DOI] [PubMed] [Google Scholar]
  2. Alberti KG, Zimmet P, Shaw J. International Diabetes Federation: a consensus on Type 2 diabetes prevention. Diabet Med 2007; 24: 451–63. [DOI] [PubMed] [Google Scholar]
  3. Allan CB, Lacourciere GM, Stadtman TC. Responsiveness of selenoproteins to dietary selenium. Annu Rev Nutr 1999; 19: 1–16. [DOI] [PubMed] [Google Scholar]
  4. Bai Y, Yang A, Pu H, Dai M, Cheng N, Ding J, et al. Cohort Profile: The China Metal-Exposed Workers Cohort Study (Jinchang Cohort). Int J Epidemiol 2017; 46: 1095–1096e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Balakumar P, Maung UK, Jagadeesh G. Prevalence and prevention of cardiovascular disease and diabetes mellitus. Pharmacol Res 2016; 113: 600–609. [DOI] [PubMed] [Google Scholar]
  6. Bleys J, Navas-Acien A, Guallar E. Serum selenium and diabetes in U.S. adults. Diabetes Care 2007; 30: 829–34. [DOI] [PubMed] [Google Scholar]
  7. Chen YW, Yang CY, Huang CF, Hung DZ, Leung YM, Liu SH. Heavy metals, islet function and diabetes development. Islets 2009; 1: 169–76. [DOI] [PubMed] [Google Scholar]
  8. Cho NH, Shaw JE, Karuranga S, Huang Y, da Rocha Fernandes JD, Ohlrogge AW, et al. IDF Diabetes Atlas: Global estimates of diabetes prevalence for 2017 and projections for 2045. Diabetes Res Clin Pract 2018; 138: 271–281. [DOI] [PubMed] [Google Scholar]
  9. Clark LC, Combs GF Jr., Turnbull BW, Slate EH, Chalker DK, Chow J, et al. Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial. Nutritional Prevention of Cancer Study Group. JAMA 1996; 276: 1957–63. [PubMed] [Google Scholar]
  10. Collaborators GBDRF. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet 2020; 396: 1223–1249. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Czernichow S, Couthouis A, Bertrais S, Vergnaud A-C, Dauchet L, Galan P, et al. Antioxidant supplementation does not affect fasting plasma glucose in the Supplementation with Antioxidant Vitamins and Minerals (SU.VI.MAX) study in France: association with dietary intake and plasma concentrations. Am J Clin Nutr 2006; 84: 395–399. [DOI] [PubMed] [Google Scholar]
  12. Danaei G, Finucane MM, Lu Y, Singh GM, Cowan MJ, Paciorek CJ, et al. National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2·7 million participants. The Lancet 2011; 378: 31–40. [DOI] [PubMed] [Google Scholar]
  13. Davis WA, Peters KE, Makepeace A, Griffiths S, Bundell C, Grant SFA, et al. Prevalence of diabetes in Australia: insights from the Fremantle Diabetes Study Phase II. Intern Med J 2018; 48: 803–809. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Dora JM, Machado WE, Rheinheimer J, Crispim D, Maia AL. Association of the type 2 deiodinase Thr92Ala polymorphism with type 2 diabetes: case-control study and meta-analysis. Eur J Endocrinol 2010; 163: 427–34. [DOI] [PubMed] [Google Scholar]
  15. Dubey P, Thakur V, Chattopadhyay M. Role of Minerals and Trace Elements in Diabetes and Insulin Resistance. Nutrients 2020; 12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Emerging Risk Factors C, Sarwar N, Gao P, Seshasai SR, Gobin R, Kaptoge S, et al. Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet 2010; 375: 2215–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fürnsinn C, Englisch R, Ebner K, Nowotny P, Vogl C, Waldhäusl W. Insulin-like vs. non-insulin-like stimulation of glucose metabolism by vanadium, tungsten, and selenium compounds in rat muscle. Life Sciences 1996; 59: 1989–2000. [DOI] [PubMed] [Google Scholar]
  18. Gao H, Hagg S, Sjogren P, Lambert PC, Ingelsson E, van Dam RM. Serum selenium in relation to measures of glucose metabolism and incidence of Type 2 diabetes in an older Swedish population. Diabet Med 2014; 31: 787–93. [DOI] [PubMed] [Google Scholar]
  19. Geiss LS, Wang J, Cheng YJ, Thompson TJ, Barker L, Li Y, et al. Prevalence and incidence trends for diagnosed diabetes among adults aged 20 to 79 years, United States, 1980–2012. JAMA 2014; 312: 1218–26. [DOI] [PubMed] [Google Scholar]
  20. Gong JH, Lo K, Liu Q, Li J, Lai S, Shadyab AH, et al. Dietary Manganese, Plasma Markers of Inflammation, and the Development of Type 2 Diabetes in Postmenopausal Women: Findings From the Women’s Health Initiative. Diabetes Care 2020; 43: 1344–1351. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hamanishi T, Furuta H, Kato H, Doi A, Tamai M, Shimomura H, et al. Functional variants in the glutathione peroxidase-1 (GPx-1) gene are associated with increased intima-media thickness of carotid arteries and risk of macrovascular diseases in japanese type 2 diabetic patients. Diabetes 2004; 53: 2455–60. [DOI] [PubMed] [Google Scholar]
  22. Hansen AF, Simic A, Asvold BO, Romundstad PR, Midthjell K, Syversen T, et al. Trace elements in early phase type 2 diabetes mellitus-A population-based study. The HUNT study in Norway. J Trace Elem Med Biol 2017; 40: 46–53. [DOI] [PubMed] [Google Scholar]
  23. Hu FB, Manson JE, Stampfer MJ, Colditz G, Liu S, Solomon CG, et al. Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N Engl J Med 2001; 345: 790–7. [DOI] [PubMed] [Google Scholar]
  24. International Programme on Chemical S, United Nations Environment P, International Labour O, World Health O. Selenium. World Health Organization, Geneva, 1987. [Google Scholar]
  25. Khan AR, Awan FR. Metals in the pathogenesis of type 2 diabetes. J Diabetes Metab Disord 2014; 13: 16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kim EJ, Ha KH, Kim DJ, Choi YH. Diabetes and the Risk of Infection: A National Cohort Study. Diabetes Metab J 2019; 43: 804–814. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kohler LN, Florea A, Kelley CP, Chow S, Hsu P, Batai K, et al. Higher Plasma Selenium Concentrations Are Associated with Increased Odds of Prevalent Type 2 Diabetes. J Nutr 2018; 148: 1333–1340. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kuo CC, Navas-Acien A. Commentary: Environmental chemicals and diabetes: which ones are we missing? Int J Epidemiol 2015; 44: 248–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Kuzuya M, Ando F, Iguchi A, Shimokata H. Glutathione peroxidase 1 Pro198Leu variant contributes to the metabolic syndrome in men in a large Japanese cohort. Am J Clin Nutr 2008; 87: 1939–44. [DOI] [PubMed] [Google Scholar]
  30. Laclaustra M, Navas-Acien A, Stranges S, Ordovas JM, Guallar E. Serum selenium concentrations and diabetes in U.S. adults: National Health and Nutrition Examination Survey (NHANES) 2003–2004. Environ Health Perspect 2009; 117: 1409–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Li JQ, Welchowski T, Schmid M, Letow J, Wolpers C, Pascual-Camps I, et al. Prevalence, incidence and future projection of diabetic eye disease in Europe: a systematic review and meta-analysis. Eur J Epidemiol 2020; 35: 11–23. [DOI] [PubMed] [Google Scholar]
  32. Li XT, Yu PF, Gao Y, Guo WH, Wang J, Liu X, et al. Association between Plasma Metal Levels and Diabetes Risk: a Case-control Study in China. Biomed Environ Sci 2017; 30: 482–491. [DOI] [PubMed] [Google Scholar]
  33. Liao XL, Wang ZH, Liang XN, Liang J, Wei XB, Wang SH, et al. The Association of Circulating Selenium Concentrations with Diabetes Mellitus. Diabetes Metab Syndr Obes 2020; 13: 4755–4761. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lin J, Shen T. Association of dietary and serum selenium concentrations with glucose level and risk of diabetes mellitus: A cross sectional study of national health and nutrition examination survey, 1999–2006. J Trace Elem Med Biol 2021; 63: 126660. [DOI] [PubMed] [Google Scholar]
  35. Lin X, Xu Y, Pan X, Xu J, Ding Y, Sun X, et al. Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025. Sci Rep 2020; 10: 14790. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Liu S, Song Y, Ford ES, Manson JE, Buring JE, Ridker PM. Dietary calcium, vitamin D, and the prevalence of metabolic syndrome in middle-aged and older U.S. women. Diabetes Care 2005; 28: 2926–32. [DOI] [PubMed] [Google Scholar]
  37. McCann JC, Ames BN. Adaptive dysfunction of selenoproteins from the perspective of the triage theory: why modest selenium deficiency may increase risk of diseases of aging. FASEB J 2011; 25: 1793–814. [DOI] [PubMed] [Google Scholar]
  38. McClung JP, Roneker CA, Mu W, Lisk DJ, Langlais P, Liu F, et al. Development of insulin resistance and obesity in mice overexpressing cellular glutathione peroxidase. Proc Natl Acad Sci U S A 2004; 101: 8852–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Meeks KAC, Freitas-Da-Silva D, Adeyemo A, Beune EJAJ, Modesti PA, Stronks K, et al. Disparities in type 2 diabetes prevalence among ethnic minority groups resident in Europe: a systematic review and meta-analysis. Intern Emerg Med 2016; 11: 327–340. [DOI] [PubMed] [Google Scholar]
  40. Misu H, Takamura T, Takayama H, Hayashi H, Matsuzawa-Nagata N, Kurita S, et al. A liver-derived secretory protein, selenoprotein P, causes insulin resistance. Cell Metab 2010; 12: 483–95. [DOI] [PubMed] [Google Scholar]
  41. Ramachandran A, Wan Ma RC, Snehalatha C. Diabetes in Asia. The Lancet 2010; 375: 408–418. [DOI] [PubMed] [Google Scholar]
  42. Rath AA, Lam HS, Schooling CM. Effects of selenium on coronary artery disease, type 2 diabetes and their risk factors: a Mendelian randomization study. Eur J Clin Nutr 2021. [DOI] [PubMed] [Google Scholar]
  43. Rayman MP. Selenium and human health. Lancet 2012; 379: 1256–68. [DOI] [PubMed] [Google Scholar]
  44. Rayman MP, Stranges S. Epidemiology of selenium and type 2 diabetes: can we make sense of it? Free Radic Biol Med 2013; 65: 1557–1564. [DOI] [PubMed] [Google Scholar]
  45. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N, et al. Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from the International Diabetes Federation Diabetes Atlas, 9(th) edition. Diabetes Res Clin Pract 2019; 157: 107843. [DOI] [PubMed] [Google Scholar]
  46. Saito Y, Hayashi T, Tanaka A, Watanabe Y, Suzuki M, Saito E, et al. Selenoprotein P in human plasma as an extracellular phospholipid hydroperoxide glutathione peroxidase. Isolation and enzymatic characterization of human selenoprotein p. J Biol Chem 1999; 274: 2866–71. [DOI] [PubMed] [Google Scholar]
  47. Satyanarayana S, Sekhar JR, Kumar KE, Shannika LB, Rajanna B, Rajanna S. Influence of selenium (antioxidant) on gliclazide induced hypoglycaemia/anti hyperglycaemia in normal/alloxan-induced diabetic rats. Mol Cell Biochem 2006; 283: 123–7. [DOI] [PubMed] [Google Scholar]
  48. Simic A, Hansen AF, Asvold BO, Romundstad PR, Midthjell K, Syversen T, et al. Trace element status in patients with type 2 diabetes in Norway: The HUNT3 Survey. J Trace Elem Med Biol 2017; 41: 91–98. [DOI] [PubMed] [Google Scholar]
  49. Song Y, Manson JE, Buring JE, Liu S. Dietary magnesium intake in relation to plasma insulin levels and risk of type 2 diabetes in women. Diabetes Care 2004; 27: 59–65. [DOI] [PubMed] [Google Scholar]
  50. Steinbrenner H, Sies H. Protection against reactive oxygen species by selenoproteins. Biochim Biophys Acta 2009; 1790: 1478–85. [DOI] [PubMed] [Google Scholar]
  51. Stranges S, Galletti F, Farinaro E, D’Elia L, Russo O, Iacone R, et al. Associations of selenium status with cardiometabolic risk factors: an 8-year follow-up analysis of the Olivetti Heart study. Atherosclerosis 2011; 217: 274–8. [DOI] [PubMed] [Google Scholar]
  52. Stranges S, Marshall JR, Natarajan R, Donahue RP, Trevisan M, Combs GF, et al. Effects of long-term selenium supplementation on the incidence of type 2 diabetes: a randomized trial. Ann Intern Med 2007; 147: 217–23. [DOI] [PubMed] [Google Scholar]
  53. Tsutsumi R, Saito Y. Selenoprotein P; P for Plasma, Prognosis, Prophylaxis, and More. Biol Pharm Bull 2020; 43: 366–374. [DOI] [PubMed] [Google Scholar]
  54. Vinceti M, Bonaccio M, Filippini T, Costanzo S, Wise LA, Di Castelnuovo A, et al. Dietary selenium intake and risk of hospitalization for type 2 diabetes in the Moli-sani study cohort. Nutr Metab Cardiovasc Dis 2021a; 31: 1738–1746. [DOI] [PubMed] [Google Scholar]
  55. Vinceti M, Filippini T, Rothman KJ. Selenium exposure and the risk of type 2 diabetes: a systematic review and meta-analysis. Eur J Epidemiol 2018; 33: 789–810. [DOI] [PubMed] [Google Scholar]
  56. Vinceti M, Filippini T, Wise LA, Rothman KJ. A systematic review and dose-response meta-analysis of exposure to environmental selenium and the risk of type 2 diabetes in nonexperimental studies. Environ Res 2021b; 197: 111210. [DOI] [PubMed] [Google Scholar]
  57. Vinceti M, Wei ET, Malagoli C, Bergomi M, Vivoli G. Adverse health effects of selenium in humans. Rev Environ Health 2001; 16: 233–51. [DOI] [PubMed] [Google Scholar]
  58. Wang L, Gao P, Zhang M, Huang Z, Zhang D, Deng Q, et al. Prevalence and Ethnic Pattern of Diabetes and Prediabetes in China in 2013. JAMA 2017; 317: 2515–2523. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Whanger P, Vendeland S, Park YC, Xia Y. Metabolism of subtoxic levels of selenium in animals and humans. Ann Clin Lab Sci 1996; 26: 99–113. [PubMed] [Google Scholar]
  60. Williams R, Karuranga S, Malanda B, Saeedi P, Basit A, Besancon S, et al. Global and regional estimates and projections of diabetes-related health expenditure: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract 2020; 162: 108072. [DOI] [PubMed] [Google Scholar]
  61. Winocour PH. Diabetes and chronic kidney disease: an increasingly common multi-morbid disease in need of a paradigm shift in care. Diabet Med 2018; 35: 300–305. [DOI] [PubMed] [Google Scholar]
  62. Xu G, Liu B, Sun Y, Du Y, Snetselaar LG, Hu FB, et al. Prevalence of diagnosed type 1 and type 2 diabetes among US adults in 2016 and 2017: population based study. BMJ 2018; 362: k1497. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Zhang H, Yan C, Yang Z, Zhang W, Niu Y, Li X, et al. Alterations of serum trace elements in patients with type 2 diabetes. J Trace Elem Med Biol 2017; 40: 91–96. [DOI] [PubMed] [Google Scholar]

RESOURCES