Introduction

Selenium, a crucial trace element in the human body, has attracted significant attention from the medical community since its discovery by Swedish scientists in 1817. After numerous generations of scientific inquiry, it has become evident that the biological impact of selenium is mediated primarily through selenoproteins1. These proteins play pivotal roles in numerous of physiological and pathological processes within the body. They contribute to the maintenance of the endogenous antioxidant system2, influence innate and adaptive immunity3, and exhibit intricate connections with brain function, cardiovascular health, thyroid hormone metabolism, diabetes, and cancer4,5.

Given the crucial biological roles of the selenium, extensive research has been conducted to elucidate the correlation between selenium deficiency, which is resulting from insufficient dietary intake, pregnancy, diminished physical capabilities, advanced age, obesity6, and the development of various pathological conditions. The renowned ailments stemming from selenium deficiency, often referred to as geochemical disorders, include Keshan disease and Kashin-Beck disease7. Furthermore, selenium deficiency is linked to an increased incidence of muscle necrosis, hypothyroidism, cardiovascular and cerebrovascular disorders, male infertility, and various forms of cancer8. China, with over 105 million individuals potentially facing health risks due to selenium deficiency, is significantly deficient in this essential micronutrient and warrants substantial attention9. Regrettably, excessive selenium intake and high serum selenium doses can potentially lead to adverse health effects7. Cases of human selenium poisoning have been well-documented in the regions with high selenium concentrations, including Punjab10, India; Enshi, Hubei Province, China11; and Venezuela, South America. Clinical manifestations of selenium toxicity encompass symptoms, such as garlic-scented breath, dermatitis, hair and nail loss, acute respiratory distress, myocardial infarction, and renal failure8. However, due to limited dose–response relationships between selenium and specific health outcomes, consensus on the optimal selenium intake remains elusive12. Therefore, determining the optimal selenium levels is crucial for maintaining health and longevity.

In recent years, there has been a growing scholarly focus on epidemiological research examining the association between selenium levels and both disease and long-term health outcomes. Studies have indicated that elevated whole blood selenium levels in American adults correlate with increased insulin levels and insulin resistance13, whereas populations in China with high selenium levels show a higher prevalence of type 2 diabetes mellitus (T2DM)14, highlighting the detrimental effects of excessive selenium exposure. Conversely, a Dutch cohort study15 demonstrated a concurrence between elevated serum selenium concentration and a reduced risk of all-cause and/or cardiovascular relative mortality. These findings align with results from the Chinese Dongfeng-Tongji cohort16, selenium-rich population17 and elderly participants18, suggesting potential health benefits of elevated selenium levels. While definitive conclusions cannot be drawn solely from these studies, they underscore the critical role of selenium in health outcomes and underscore the need to establish optimal selenium concentrations. Furthermore, higher serum selenium concentrations have been linked to lower risks of all-cause and/or cardiovascular mortality in specific subgroups, including those with type 2 diabetes19, end-stage renal disease20, systemic inflammatory response syndrome21, Hashimoto’s thyroiditis22,23, or nonalcoholic fatty liver disease24. The consistency of these findings suggests that selenium may offer maximal benefits in targeted populations.

Therefore, employing serum selenium levels as indicators of systemic selenium status and mortality as the primary endpoint, we utilized a nationally representative sample from the NHANES general population to investigate the association between selenium and long-term health outcomes. Additionally, we segmented the population into distinct subgroups to explore the potential maximal benefits or adverse impacts of selenium in specific demographic categories.

Results

Baseline characteristics

The study included 5449 individuals (Fig. 1), of whom 50.8% were male, with an average age of 48 years. Of these individuals, 2361 (42.9%) were non-Hispanic whites, 4164 (75.8%) had completed high school education or above, 53.8% had never smoked, and 2210 (40.2%) reported a history of heavy alcohol consumption. Table 1 displays the baseline characteristics of the population, categorized by the quartile of the serum selenium concentration. Individuals with higher serum selenium concentrations are more likely to be older, non-Hispanic white women, who currently have higher household incomes and are currently more likely to smoke than the population with lower serum selenium concentrations. Greater serum selenium concentrations have been linked to greater systolic and diastolic blood pressure in terms of physical health. In addition, they are less likely to have had a stroke in the past and are more likely to have a history of high blood pressure and cholesterol.

Table 1 Baseline characteristics of participants according to serum selenium.
Full size table
Fig. 1
figure 1

The participants selection flow-chart. For follow-up studies, researchers considered eligible must have both age, gender, race information and serum selenium concentrations. In addition, they should have no history of cancer at baseline, eligibility status for mortality follow-up, and no disappearance during follow-up. Finally, a total of 5499 participants were included in the study.

Full size image

Serum selenium concentration and cause-specific mortality

A total of 837 fatalities were recorded up to December 31, 2019. Among them, 252 individuals died from cardiovascular disease, 179 from cancer, and 585 from noncardiovascular disease, which included chronic lower respiratory diseases, accidents, cerebrovascular diseases, Diabetes mellitus, and Alzheimer’s disease. The Cox proportional hazard models, after being fully adjusted for multiple variables, demonstrated a statistically significant association between the serum selenium concentration (considered a continuous variable) and mortality related to cardiovascular disease (CVD) and cancer (Table 2). Specifically, the risk of mortality from CVD decreased by 76%, and the risk of mortality from cancer decreased by 77% for every unit increase in logarithmically converted serum selenium concentration. Upon categorizing serum selenium into quartiles, a distinct pattern emerged. Compared with the lowest quartile (Q1, < 119.9 μg/L), the second quartile (Q2, 119.9–130.4 μg/L) was associated with decreased all-cause and non-CVD mortality (HR and 95% CI were 0.61, 0.45–0.83 and 0.59, 0.42–0.83, respectively). The third quartile (Q3, 130.5–142.0 μg/L) was associated with lower rates of mortality from CVD (HR and 95% CI were 0.49, 0.32–0.76), whereas the second and fourth quartiles (Q4, 142.1–313.0 μg/L) were associated with lower rates of mortality from cancer (HR and 95% CI were 0.53 (0.29–0.97) and 0.52 (0.29–0.92), respectively). Furthermore, a trend (p for trend = 0.056) that approached significance was observed (Table 2, Fig. 2). Survival analysis, stratified by serum selenium quartiles, displayed in Kaplan‒Meier survival curves (Fig. 3), revealed lower all-cause and CVD-related mortality among individuals with higher serum selenium levels than among those with lower levels (log-rank test, p-values 0.005 and 0.015, respectively). A nonlinear association between the serum selenium concentration and mortality from all causes, CVD-related causes, and non-CVD-related causes was revealed via martingale residual testing (Fig. S1). In addition, we performed restricted cubic spline analysis on the fully adjusted models, as depicted in Fig. S2, confirming the observed nonlinear relationship. The results of the segmented regression (Table S1) revealed that all-cause and non-CVD-related mortality rates decreased by 3% and 4%, respectively, with each standard deviation increasing when the blood selenium content was ≤ 129.82 μg/L. Similarly, a one-standard deviation increases in the serum selenium concentration to ≤ 129.08 μg/L reduced CVD mortality by 4%. A significant correlation was identified in subgroup analyses stratified by age (≤ 60 years old, > 60 years old), sex (male, female), smoking status (never smoker, former smoker, current smoker) (Table 3). Specifically, in people who had never smoked and had no history of stroke, higher serum selenium concentrations were associated with a lower all-cause mortality rate when they were ≤ 129.82 μg/L. Among males, current smokers, individuals with moderate family income, and those without a high cholesterol or stroke history, higher serum selenium levels were associated with lower mortality from non-CVD. Moreover, in the female population and among individuals with a history of smoking (Table S2), higher serum selenium levels were correlated with lower mortality from CVD. Notably, significant interactions were found between serum selenium levels and smoking, PIR, hypertension, and a history of high cholesterol in relation to all-cause mortality. Similarly, notable interactions were observed between serum selenium levels and smoking, PIR, high cholesterol, and history of stroke concerning non-CVD-related mortality. Furthermore, our findings indicated that there was no association between the serum selenium concentration and cause-specific mortality in the nonsmoking population. However, when the serum selenium concentration exceeded 129.82 μg/L, a positive correlation was detected between the serum selenium concentration and all-cause mortality and non-CVD mortality. Finally, the link between blood selenium levels and cause-specific mortality remained unchanged after sensitivity analysis that corrected for total dietary selenium intake and examination session (Table S3).

Table 2 HRs (95% CIs) for cause-specific mortality according to serum selenium concentration among participants8.
Full size table
Fig. 2
figure 2

Multivariate cox regression models of serum selenium concentrations with cause-specific mortality. Cox multivariate analysis of serum selenium with (A) all-cause mortality, (B) CVDs-related mortality, (C) cancer mortality and (D) non-CVDs-related mortality Abbreviations: DBP Diastolic blood pressure, SBP Systolic blood pressure, PIR Family income-to-poverty ratio, HR Hazard ratio, CVDs Cardiovascular diseases, Q quartiles.

Full size image
Fig. 3
figure 3

Kaplan–Meier curves for cause-specific mortality by serum selenium quartile groups. Among the four subgroups of serum selenium, the survival rates for (A) all-cause mortality, (B) CVDs-related mortality, (C) cancer mortality and (D) non-CVDs-related mortality were significantly different according to the log-rank test (p < 0.05). Abbreviations: CVDs Cardiovascular diseases, Q quartiles.

Full size image
Table 3 Subgroup analysis for the association between serum selenium concentrations and cause-specific mortality.
Full size table

Discussion

In this prospective study involving American adults, we found that a linear negative correlation was observed between the serum selenium concentration and cancer mortality, but there was a nonlinear relationship between the serum selenium concentration and all-cause, CVD-related, and non-CVD related mortality. Higher serum selenium levels (range ≤ 129.82 μg/L) were associated with lower rates of all-cause and non-CVD-related mortality. Similarly, in cases where the serum selenium concentration was ≤ 129.08 μg/L, higher levels were associated with lower CVD-related mortality. However, this association ceases to be significant when serum selenium concentration exceeds the specified threshold range. All the results remained consistent across the subgroup analyses. Notably, we observed significant interactions between the serum selenium concentration and smoking, PIR, history of hypertension, and high cholesterol in terms of all-cause mortality. Specifically, when the serum selenium concentration exceeded 129.82 μg/L, in the nonsmoking population, there was no association between the serum selenium concentration and cause-specific mortality rates, while higher serum selenium levels were associated with a greater risk of overall mortality in the group of current smokers.

The complex biological functions of serum selenium have garnered significant interest among researchers. Elucidating its association with human health outcomes remains a formidable challenge. In studies of the general population, the relationship between serum selenium levels and cause-specific mortality is intricate15,25,26. This complexity highlights the necessity for a more comprehensive elucidation of the dose‒response relationship between these two factors. Simple linear correlations may not be sufficient to explain the aforementioned outcomes. In our study, it is noteworthy that the previously observed negative correlation between serum selenium concentration and specific mortality rates lost its statistical significance once the serum selenium concentration surpassed a particular threshold. In addition, serum selenium has a significant protective effect against cause-specific mortality within distinct populations, including subgroups with diabetes19, hypertension27, kidney disease20, and liver disease. However, we detected a positive correlation between the serum selenium concentration and overall mortality rate, specifically among the current smoking population, when the serum selenium concentration surpassed the established range. Consequently, it is imperative to investigate the relationships between serum selenium and both all-cause mortality and cause-specific mortality in the general population and to identify the populations that are most benefitted or at risk.

In the present study, we observed a linear negative correlation between the serum selenium concentration and cancer mortality in the general population, despite a p trend of 0.056. This finding is consistent with a 9-year longitudinal epidemiological research on aging caused by vascular disease28. Furthermore, a prospective study conducted in Poland revealed that higher serum selenium levels significantly improved the ten-year survival rate for patients with breast cancer29. In the Cancer Nutrition Prevention Trial involving 1312 participants, it was also established that exogenous selenium supplementation could reduce cancer mortality by 52%30. Elevated oxidative stress and impaired immune function in cancer patients may partially account for this observation. Selenium inhibits the Nrf2/Prx1 pathway, reducing tumor cell detoxification capacity and thereby enhancing drug cytotoxicity31,32,33. Moreover, selenium helps maintain normal cellular redox balance, facilitating drug detoxification while minimizing damage to healthy tissues. Additionally, selenium enhances the expression of p53-dependent DNA repair proteins, aiding in the repair of DNA damage in normal cells33.

A comprehensive cohort study involving the general population in the Netherlands revealed an association between elevated serum selenium concentrations and decreased all-cause mortality15. Similarly, a meta-analysis encompassing 25,667 subjects demonstrated that lower selenium levels were correlated with an elevated risk of all-cause mortality in the broader population34. In contrast, a 15-year follow-up study involving 1103 individuals in the Chinese population did not reveal any significant associations between total mortality and serum selenium levels25. The aforementioned studies included representative samples from the general population, and the disparities in their findings underscore the limitations of establishing a linear correlation between serum selenium levels and mortality within the general population. Given the varying selenium statuses worldwide, it is plausible that a nonlinear relationship may more accurately reflect the actual situation35. Our study revealed a correlation between elevated serum selenium concentrations and decreased all-cause mortality, non-CVD-related mortality (range ≤ 129.82 μg/L), and CVD-related mortality (range ≤ 129.08 μg/L), aligning with prior research36, in which the safety range of the serum selenium level is less than 130 μg/L. Appropriate selenium levels can enhance the proliferation of activated T cells, increase natural killer cell activity, and augment cytotoxicity mediated by lymphocytes2, thereby bolstering the body’s defense against external stressors. Additionally, adequate selenium levels can promote synthesis of selenoproteins involved in genomic maintenance. For instance, GPX1 inhibits DNA breaks induced by ultraviolet and ionizing radiation, while SELENOH shields cells from genotoxic agents that elevate oxidative stress37. However, this associations lost significance when serum selenium concentration surpassed the established limit, suggesting that excessively high serum selenium concentrations do not continue to reduce the risk of mortality. In the Selenium and Vitamin E Cancer Prevention Trial (SELECT), participants administered selenium supplements were tracked over 5.5 and 8 years38,39, and no notable improvement in all-cause mortality was found. Similarly, a 7.6-year follow-up in a cancer nutrition prevention study failed to establish a significant correlation between selenium supplementation and any cardiovascular disease endpoint30. Notably, a randomized double-blind controlled trial conducted in the Danish population revealed that the administration of 300 μg/day selenium supplements for five consecutive years resulted in an increased 10-year all-cause mortality in regions characterized by low selenium levels26. The above studies provide further evidence of our discovery regarding the intricate relationship between the serum selenium concentration and cause-specific mortality. Adequate selenium concentrations are beneficial for human health; however, beyond a certain threshold (in this study, 130 μg/L), this beneficial effect diminishes. Studies have shown that high levels of selenium, after participating in the synthesis of various selenoproteins, may influence protein function through residual low-molecular-weight selenium compounds and active selenium intermediates40. These compounds can modulate oxidation-sensitive cysteine residues in proteins or disrupt selenoprotein-dependent antioxidant systems. For example, selenocysteine can metabolize into selenol/selenate, engaging in intracellular redox reactions that generate superoxide radicals and react with thiols/diseleides to form selenylsulfides/disulfides26. The latter processes can lead to protein aggregation, transcription factor inactivation, disruption of redox-regulated cellular signaling pathways, endoplasmic reticulum stress, and consequent cellular damage41. Selenium deficiency results in a marked reduction in glutathione peroxidase activity, consequently elevating oxidative stress42. Conversely, elevated selenium levels induce DNA damage and cytotoxicity by augmenting reactive oxygen species (ROS) production43. Furthermore, excessive selenium intake diminishes nitric oxide bioavailability and exacerbates ROS production, culminating in endothelial dysfunction44.

In further exploration of the relationships between the serum selenium concentration and all-cause mortality, as well as cause-specific mortality, we identified no correlation between the serum selenium concentration and cause-specific mortality in nonsmokers. Notably, when the serum selenium concentration exceeded 129.82 μg/L, elevated levels were associated with an increased risk of all-cause mortality among current smokers. Consistent with prior research45,46,47, our study aligns with the observation that nonsmokers present significantly higher serum selenium concentrations than current smokers (nonsmokers vs. current smokers, 131.5 μg/L vs. 128.7 μg/L). This discrepancy might result in a swift attainment of a plateau in selenoprotein concentration among nonsmokers, potentially diminishing its effectiveness in mitigating the risk of mortality. Furthermore, a population-based study revealed higher levels of serum selenium glutathione peroxidase, glutathione reductase, and extracellular superoxide dismutase activities in nonsmokers than in current smokers. This disparity may partially counteract the antioxidant effect of serum selenium in vivo, resulting in an inconspicuous protective effect of serum selenium in this population. In the case of current smokers, research has indicated that cadmium in cigarettes not only diminishes the bioavailability of selenium48 but also hastens excretion selenium45. Elevated selenium levels in the body, under such circumstances, may lead to selenium excess and subsequent adverse events. Notably, smoking also increases the content of arsenic, an element known for its synergistic toxic effect with selenium49. This interaction amplifies the toxicity of methylated selenium by obstructing the detoxification process, thereby resulting in adverse outcomes.

Our findings suggest the following strengths: First, in light of regional and dietary differences, we opted against using dietary selenium intake as the primary study indicator. Instead, we selected serum selenium as it is the most representative measure of internal selenium levels50. Additionally, this study employed ICPDRC-MS to measure serum selenium levels, facilitating cross-comparison with other studies and providing insights into its clinical significance. Furthermore, this study is the first to comprehensively elucidate the impact of smoking status on the association between serum selenium and cause-specific mortality, suggesting that selenium supplementation may not be universally applicable. However, our study has limitations that should be considered. Due to inherent constraints in the database design, we only collected information on baseline serum selenium, thus we were unable to assess the impact of changes in serum selenium during follow-up on cause-specific mortality, potentially resulting in survival bias. Moreover, self-reported data on health status and smoking and drinking status may be subject to memory bias.

Overall, this study reveals an important threshold range exists for the protective effect of serum selenium concentration for cause-specific mortality. Elevated serum selenium levels were correlated with reduced all-cause mortality and non-cardiovascular disease-related mortality when serum selenium concentrations were ≤ 129.82 μg/L. Additionally, increased serum selenium levels were linked to decreased cardiovascular disease-related mortality when serum selenium concentrations were ≤ 129.08 μg/L. However, this correlation diminishes when serum selenium concentrations surpass the specified range. Moreover, the study further emphasizes the impact of smoking on this association, specifically highlighting the elevated mortality risk correlated with higher serum selenium levels in current smokers. Therefore, based on our observation, the narrow therapeutic window of selenium warrants careful consideration. It is recommended that selenium supplementation be approached with heightened caution, considering regional, ethnic, and individual variability. Individuals exhibiting low baseline selenium levels may derive benefits from supplementation, whereas those with elevated selenium levels could face potential health risks, particularly in smokers. Regular monitoring of serum selenium concentrations during supplementation is imperative. Determining the optimal serum selenium levels and establishing safe supplementation dosages remain critical issues that require prompt resolution.

Methods

Study population

The National Health and Nutrition Survey (NHANES) is a comprehensive research endeavor conducted by the National Center for Health Statistics, employing a stratified, clustered, multistage probability sampling methodology. Its overarching objective is to gauge disease incidence, ascertain disease risk factors, and investigate the intricate interplay between nutritional status, health promotion, and disease prevention within the United States. The NHANES interview included demographic, socioeconomic, dietary, and health-related questions. The examination component consists of medical, dental, and physiological measurements. Details have been released elsewhere51.

For this investigation, we utilized data from one NHANES cycle from 2003–2004, as well as three cycles spanning from 2011 to 2016. Initially, a cohort of 22,089 participants aged 20 years and older was selected for analysis. To further refine the experiment, additional criteria were taken into account, including the determination of death status, the presence of cancer at baseline, and the availability of complete serum selenium level data for each participant. Ultimately, a cohort comprising 5499 participants met the inclusion criteria for the follow-up study. The complete process of population inclusion is illustrated in Fig. 1. Every participant in the NHANES provided written informed consent upon their inclusion in the study.

Measurement of serum selenium concentrations

The participants’ blood was drawn into prescreened vacuum blood collection tubes, allowed to coagulate for 30–40 min, and then centrifuged at 2400 rpm for 10 min. A serum separator was used to extract serum from the blood clot, which was then transferred into a polyethylene vial. Serum samples were preserved at − 70 °C and subsequently transported to the National Center for Environmental Health for testing. Serum selenium levels were assessed via inductively coupled plasma‒dynamic reaction cell‒mass spectrometry (ICP‒DRC‒MS), and the lower detection limit was 4.5 μg/L. Values falling below the lower detection limit were substituted with the square root of 4.5 divided by 2. For detailed information regarding data quality control, processing, and editing, comprehensive resources can be accessed on the NHANES website.

Ascertainment of mortality

Mortality status was ascertained via National Center for Health Statistics (NCHS) 2019 Public-Use Linked Mortality Files. The mortality data of all participants were tracked until December 31, 2019. Each eligible participant was assigned a vital status code. All-cause mortality, cardiovascular related mortality, cancer mortality, and noncardiovascular relative mortality were identified via the International Classification of Diseases, Tenth Revision (ICD-10).

Ascertainment of smoking status

Information on smoking status, alcohol intake, and medical condition was collected through standardized questionnaires. If an individual has smoked fewer than 100 cigarettes in their lifetime, they are classified as a ‘Never smoker’. If they answer ‘not at all’ to the question ‘Do you now smoke cigarettes?’, they are classified as a ‘former smoker’. However, if the response is ‘some days’ or ‘every day’ to the question ‘Do you now smoke cigarettes?’, they are classified as a ‘current smoker’.

Assessment of covariates

Information on age, sex, race, education level, and family income were obtained from demographic documents. Race was classified as Mexican American, Non-Hispanic White, Non-Hispanic Black or Other. Education level was categorized as less than high school, high school or more than high school. The family income‒to‒poverty ratio (PIR) was divided into 0–1.3, 1.3–3.5 and ≥ 3.5. Blood pressure, weight, and height measurements were conducted at the mobile examination center. Body mass index (BMI) was calculated by dividing weight in kilograms by the square of height in meters. BMI was then categorized into three groups: normal (≤ 25 kg/m2), overweight (25–30 kg/m2), and obese (≥ 30 kg/m2). Drinking status was grouped into nondrinker, low-to-moderate drinker (defined as, 2 drinks/day in men and, 1 drink/day in women), or heavy drinker (defined as ≥ 2 drinks/day in men and ≥ 1 drink/day in women). Dietary selenium intake data were assessed through a 24-h recall survey. In this experiment, the average dietary intake data for the first 24-h recall were used.

Statistical analysis

In our analysis, we took into account the weights derived from the NHANES stratified, clustered, and multistage probability complex sampling methods. These weights include sample weights, strata, and primary sampling units. Continuous variables with a normal distribution are represented the means (SDs), whereas nonnormally distributed data are represented as medians (interquartile ranges). Normally distributed and nonnormally distributed data were compared via one-way analysis of variance and the Kruskal‒Wallis test, respectively. The constituent variables are expressed as percentages, and differences between groups were compared via the Pearson chi-square test. The serum selenium concentration was categorized into four groups based on quartiles: Q1(< 119.9 μg/L), Q2(119.9–130.4 μg/L), Q3(130.5–142.0 μg/L), and Q4(142.1–313.0 μg/L). Additionally, the serum selenium concentration was analyzed as a continuous variable after natural logarithm conversion. The Cox proportional hazard model was employed to estimate the hazard ratio (HR) and its 95% confidence interval (CI), elucidating the correlation between the serum selenium concentration and both all-cause and cause-specific mortality. For each of the four follow-up outcomes, we constructed separate models, incorporating variables that exhibited significant differences in the Cox multivariable analysis. Linear trend detection was conducted by treating the median of categorical variables as continuous variables.

In addition, Kaplan‒Meier survival analysis and the log-rank test were employed to assess differences between various groups based on the serum selenium concentration and the four outcomes. For each outcome, a restricted cubic spline regression was executed to investigate the dose‒response relationship between the serum selenium concentration and specific mortality, with complete adjustments made in the model. Based on the Cox proportional hazards model, the Martingale residual test was employed to identify any nonlinear relationship between risk and the serum selenium concentration. If a nonlinear relationship was detected, piecewise regression was performed to investigate differences in the correlation at the threshold point, and the likelihood ratio test was performed on the model. Stratified analyses were performed by age (≤ 60 years, > 60 years), sex (male, female), smoking status (never smoker, former smoker, current smoker), PIR (0–1.3, 1.3–3.5, > 3.5), hypertension (yes, no), stroke (yes, no) and / or high cholesterol (yes, no). Finally, a sensitivity analysis was conducted to assess the robustness of the research findings. All calculations and visualizations were performed via R version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria; URL: https://www.R-project.org/). The threshold for statistical significance was set at 0.05 (two-tailed).