Association of suboptimal health status with psychosocial stress, plasma cortisol and mRNA expression of glucocorticoid receptor α/β in lymphocyte

Abstract

Suboptimal health status (SHS) has become a new public health challenge in China. This study investigated whether high SHS is associated with psychosocial stress, changes in cortisol level and/or glucocorticoid receptor (GR) isoform expression. Three-hundred eighty-six workers employed in three companies in Beijing were recruited. The SHS score was derived from data collection in the SHS questionnaire (SHSQ-25). The short standard version of the Copenhagen Psychosocial Questionnaire (COPSOQ) was used to assess job-related psychosocial stress. The mean value of the five scales of COPSOQ and distribution of plasma cortisol and mRNA expression of GRα/GRβ between the high level of SHS group and the low level of SHS group were compared using a general linear model procedure. Multiple linear regression analysis was used to analyze the effect of psychosocial stress on SHS. We identified three factors that were predictive of SHS, including “demands at work”, “interpersonal relations and leadership” and “insecurity at work”. Significantly higher levels of plasma cortisol and GRβ/GRα mRNA ratio were observed among the high SHS group. High level of SHS is associated with decreased mRNA expression of GRα. This study confirmed the association between chronic psychosocial stress and SHS, indicating that improving the psychosocial work environment may reduce SHS and then prevent chronic diseases effectively.

• Cortisol
• glucocorticoid receptor
• hypothalamus-pituitary-adrenal
• mRNA expression
• psychosocial stress
• suboptimal health status

Introduction

There have been increases in reports of poor health in the absence of a diagnosable condition, also known as suboptimal health status (SHS) (Yan & Wang, 2008; Yan et al., 2009, 2012). SHS is characterized by ambiguous health complaints, general weakness and lack of vitality. It is deemed that SHS may contribute to the progression or development of chronic disease. Linear two-level models have revealed that SHS is associated with cardiovascular risk factors including increased blood pressure and total cholesterol and decreased HDL cholesterol in either men or women (Yan et al., 2012). In our initial qualitative interview, work stressors such as office competition, high job demands and poor interpersonal relationships are among the most common factors reported to influence general health. According to Kristensen et al., work environment issues are considered to be one of the most important psychosocial stress plaguing contemporary and future societies (Kristensen et al., 2005). Chronic psychosocial stress and consequent physiological dysregulation are increasingly viewed as catalysts of accelerated aging and agitators of disease trajectories (Juster et al., 2010).

One of the major neuroendocrine systems regulating psychological stress is the hypothalamus–pituitary–adrenal (HPA) axis, which produces cortisol as its final hormonal effector. Cortisol displays a robust diurnal profile with peak values occurring between waking and 30–45 min post waking (the cortisol awakening response, CAR) and declines progressively throughout the day to reach a trough at around midnight (diurnal cortisol slope) provided there are no significant external stimuli (Saxbe, 2008). Changes in basal or stimulated cortisol levels are associated with different disease processes (Chrousos, 1998). Recent exposure to a stressor may initially elevate cortisol levels (Dickerson & Kemeny, 2004). However, the findings on the effects of chronic stress on basal cortisol levels are less consistent due to different types of stressors (in terms of type and duration). A recent meta-analysis showed an undisputed positive relation between the CAR and job stress (Chida & Steptoe, 2009).

The sensitivity of a target cell to cortisol not only depends on the concentrations of the cortisol but also on the specific intracellular environment. Cloning of the human glucocorticoid receptor (GR) cDNA has identified two isoforms, namely GRα and GRβ. These two isoforms originated from alternative splicing of the 3′ end of the primary transcript of the GR gene at exon 9 on chromosome 5, resulting in GRβ that has 15 unique amino acids in its carboxyl terminus instead of the last 50 amino acid sequence found in human glucocorticoid receptor α(hGRα). GRα modulates the expression of glucocorticoid (GC)-responsive genes by binding to specific GC-response elements, whereas GRβ does not bind GCs and is transcriptionally inactive (Bamberger et al., 1995; Oakley et al., 1999). Therefore, GRβ seems to function as an inhibitor of GC action and may be an important dominant negative regulator determining GC sensitivity in the target tissues (Bamberger et al., 1996).

In this study, we analyzed work-related psychosocial stress, plasma cortisol and GRα and GRβ mRNA expression in peripheral blood mononuclear cells (PBMC) among workers with high level of SHS. SHS has become a new public health challenge in China (Bi et al., 2014; Meng et al., 2013a; Yan et al., 2009, 2012; Yan & Wang, 2008) and also a global phenomenon of psychosocial stress (Chen et al., 2012; Vitaly et al., 2014). Identifying association of SHS with exposure to chronic stressors and biological stress mediators will not only facilitate finding effective interventions for SHS but will also elucidate the mechanisms mediating the etiologic link between SHS and chronic disease.

Methods

Study participants

From January to April 2012, a cross-sectional study was conducted among workers in urban Beijing. Three random companies whose workers participated in annual physical examination (for at least two consecutive years) during January to April were selected for cluster sampling at the physical examination center of Xuanwu Hospital, Capital Medical University. A sample population of 386 participants (male 189/female 197) from the three companies (staff from Justice Bureau, Insurance Company and University respectively in Beijing) was recruited. Eligible participants had to meet the following inclusion criteria: (1) no history of somatic or psychiatric abnormalities, as confirmed by their medical records; (2) be between 30 and 50 years of age; (3) no history of medication consumption in the previous 2 weeks; and (4) no smoking and no excessive alcohol (≥2 alcoholic drinks daily) or caffeine (≥5 cups of tea/coffee daily) intake. Patients with diagnosis of any specific disease concerning neuroendocrine system, cardiovascular system, respiratory system, genitourinary system, digestive system and blood system were also excluded. The participants signed an informed consent, and the study was approved by the Ethical Committee of the Capital Medical University, Beijing.

Data collection

SHS was measured by the suboptimal health questionnaire (SHSQ-25) including 25 items (Yan et al., 2009). Each subject was asked to rate the frequency with which they suffered from a number of specific complaints in the preceding three months on a five-point Likert-type scale: (1) never or almost never, (2) occasionally, (3) often, (4) very often and (5) always. The raw scores of 1–5 on the questionnaire were recoded as 0 to 4, respectively. SHS scores were calculated for each respondent by summing the ratings for the 25 items. A high score was indicative of a high level of SHS (i.e. poor health status) (Yan & Wang, 2008; Yan et al., 2012).

The friendly short standard version of the Copenhagen Psychosocial Questionnaire (COPSOQ) was used to assess job-related psychosocial stress (Kristensen et al., 2005). The Chinese translation and adaptation of COPSOQ have been tested in different professional people, which showed to have good reliability and validity (Meng et al., 2013b; Shang et al., 2008). The COPSOQ version used in this survey comprises five scales including 34 items, concerning (1) demands at work, (2) work organization and content, (3) interpersonal relations and leadership, (4) insecurity at work and (5) job satisfaction. The other three scales of COPSOQ concerning general health, mental health and vitality were not included in the survey. All scales had a five-point Likert format (except scales measuring insecurity at work), reflecting either intensity or frequency. These scales were transformed into a theoretical range, extending from 0 (do not agree at all) to 100 (fully agree) points. The higher score for scales of “demands at work”, “insecurity at work” and lower score for scales of “work organization and content”, “interpersonal relations and leadership” and “job satisfaction” indicate higher level of job strain.

Blood sample collection

Following an overnight fast, 5 ml of venous blood from each participant was collected in EDTA tubes (Sarstedt, Nümbrecht, Germany) between 7:30 and 8:30 a.m. in a calm state. Blood samples were immediately centrifuged, and plasma was preserved at −80 °C. PBMC were subsequently isolated by Ficoll-Hypaque density gradient centrifugation. Immediately after, 5–8 × 10⁶ cells per ml of TRIzol (Invitrogen, Waltham, MA) were stored at −80 °C until use. In order to analyze the potential influence of the diurnal rhythmicity on cortisol comparison, individual waking time for each participant was also recorded.

Plasma cortisol determination

Plasma cortisol was measured by radioimmunoassay in the Endocrinology Department of the hospital. The intra-assay and inter-assay coefficient of variations of this assay (Sino-uk, Shanghai, China) were <5.5% and <7.5%, respectively. The reference range was 50–280 ng/ml. Quality controls were applied throughout all these assays.

Strategy for the quantification of GRα and GRβ mRNA in PBMC

Total RNA from PBMC was isolated by Trizol, and cDNA was synthesized from total RNA using mouse mammary leukemia virus (MMLV) Reserve Transcriptase Mix (TransGen Biotech, Beijing, China), qPCR was performed with SYBR Green qPCR SuperMix (TransGen Biotech) and the ABI PRISM 7500 fast Real Time PCR System (Applied Biosystems, Waltham, MA). In all reactions, β-actin gene was used as endogenous control in each sample to normalize relative levels of expression. All reagents and PCR conditions are described in the following section “Condition of cDNA synthesis and qPCR”. Similarity and homogeneity of PCR products from samples were confirmed by automated melting curve analysis (SDS 2.6 software, Applied Biosystems), which revealed Ct values of the PCR products. All samples were performed in triplicate. For each experiment, all triplicate values for the samples and standard had SD of <0.3, and the mean values of the triplicates were taken as the final result. The relative amplification efficiencies of the primers were tested and found to be similar. The comparative threshold cycle (2^−ΔCt) method was used to quantify the mRNA of the target genes (Schmittgen & Livak, 2008).

Primers were designed using the software Primer 5.0 (Premier, Palo Alto, CA) based on available sequences (NM_000176.2 and NM_001018077.1) at http://www.ncbi.nlm.nih.gov/nuccore/. The primers used were as follows: GRα-forward: 5′-AAAACTATGGGAGGAAACC-3′; GRα-reverse: 5′-GGAGTGACCAGCCAAGATG-3′; GRβ-forward: 5′-TTCAGAGGCAGGGAGCAAT-3′; GRβ-reverse: 5′-GGAGTGACCAGCCAAGATG-3′; β-actin-forward: 5′-CTCCATC-CTGGCCTCGCTGT-3′; β-actin-reverse: 5′-GCTGTCACCTTC-ACCGTTCC-3′.

Condition of cDNA synthesis and qPCR

RNA quantity and quality were determined by capillary electrophoresis using the lab-on-a-chip technique (Agilent 2100 Bioanalyzer, Waldbronn, Germany). Reverse transcription was performed with 1 μg of the total RNA using 1 μl of MMLV reverse transcriptase mix (TransGen Biotech), 1 µl of 5 × RT buffer and RNase-free water for 50 μl of final reaction volume. Retrotranscription programs consisted of 60 min at 37 °C, followed by enzyme inactivation at 95 °C for 3 min. qPCR was performed with the ABI PRISM 7500 fast real-time PCR system (Applied Biosystems) using 1 μl of cDNA, 0.4 μM of each primer and 10 μl of 2 × SYBR Green qPCR SuperMix (TransGen Biotech) and 8.2 µl of RNase-free water at a final volume of 20 μl. Thermal cycling conditions were: 15 min at 95 °C followed by 40 PCR cycles of denaturing at 95 °C for 10 s, annealing and elongation at 60 °C for 30 s. Fluorescence readings were performed on annealing/elongation steps.

Statistical analysis

To examine the impact of changes in SHS on changes of other variables including psychosocial stress, plasma cortisol and mRNA expression of GRα/β in PBMC, we dichotomized SHS by the median of SHS scores, to define low (<median) and high (≥median) levels of personal SHS. Differences between groups were tested using the general linear model procedure and adjusted for potential confounders such as sex, age, company and time interval (between waking and sampling). Logarithmic transformations were made if data were not normally distributed. The strength and direction of the relationship between psychosocial stress and SHS was examined using Pearson’s correlation analysis. Correlations between cortisol levels with mRNA expression of GR isoforms were analyzed by the Spearman’s rank test. We then used multiple linear regression analysis to calculate the effect of psychosocial stress on personal SHS. A p value of less than 0.05 was considered to be of statistical significance. All p values reported were two-tailed. All statistical analyses were performed using SPSS 13.0 (SPSS Inc., Chicago, IL).

Results

Our analysis was restricted to 382 individuals, with the exception of four participants who were cortisol outliers (>mean + 3SD). The mean age was 42.34 years (SD: 4.50) with women making up 50.8% of the participants. The participants were divided into two groups by the median SHS score of 35: high SHS group (SHS score ≥35) and low SHS group (SHS score <35). The mean SHS scores among the low and high levels of SHS group were 24.49 ± 6.09 and 49.75 ± 9.26, respectively. Table 1 lists the participants’ characteristics according to the groups. The distributions of age and gender were not different between the two groups (p > 0.05). The proportion of white-collar workers among the high score group was significantly higher than that among the low score group (89.6% vs. 74.1%,  = 15.627, p < 0.001).

Table 1. Association between psychosocial stress and SHS.

	High SHS group	Low SHS group
	Mean ± SD	Mean ± S.D.	F (1,376)^a	p Value
Age	42.04 ± 4.12	42.65 ± 4.86	1.324^b	0.186
Gender (N of male/female)	101/92	87/102	1.561^c	0.218
Demands at work	49.90 ± 19.84	30.81 ± 17.69	83.51	<0.001
Work organization and content	45.73 ± 18.11	54.37 ± 19.38	13.85	<0.001
Interpersonal relations and leadership	54.28 ± 17.98	69.07 ± 19.63	51.08	<0.001
Insecurity at work	35.85 ± 26.80	19.88 ± 23.86	48.85	<0.001
Job satisfaction	59.94 ± 18.88	67.91 ± 17.01	16.37	<0.001

^aAdjusted for sex, age and company.

^bt₍₃₈₀₎ value.

^c value.

The Cronbach’s α coefficient of the COPSOQ was 0.874, indicating good internal consistency. Overall, the participants who gave higher SHS score also had a higher level of psychosocial stress than those with lower scores (Table 1). Compared with the low SHS group, scales of “demands at work” and “insecurity at work” were significantly higher among the high SHS group (p < 0.001). The mean value of scales on “work organization and content”, “interpersonal relations and leadership” and “job satisfaction” was higher in the low SHS group than in high SHS group (p < 0.001). The results also indicated that staff reporting that they were highly stressed from their job often showed a higher level of SHS. With regard to the single items of the scale, there were dramatic differences between the high SHS and low SHS groups (p < 0.001), e.g. the mean value of the items “How often do you not have time to complete all your work tasks?” (72.71 ± 24.35 vs. 46.18 ± 27.01, F_{(1, 376)} = 71.691), “Does your work put you in emotionally disturbing situations?” (48.03 ± 25.18 vs. 25.24 ± 22.13, F_{(1, 376)} = 78.750) and “To what extent would you say that your immediate superior is good at work planning?” (50.88 ± 26.20 vs. 72.04 ± 23.88, F_(1,376) = 48.701).

In the bivariate analysis, significant correlations were found between SHS and nearly all of the COPSOQ scales measures (Table 2). High scores of SHS correlated positively with “demands at work” and “insecurity at work”, whereas correlated negatively with “interpersonal relations and leadership”, “work organization and content” and “job satisfaction”. The multivariate linear regression model identified three significant predictors of SHS: “demands at work” (p < 0.001), “interpersonal relations and leadership” (p < 0.001) and “insecurity at work” (p < 0.001; Table 3). These three variables in the model explained 47.5% of the total variance in the SHS measure.

Table 2. Pearson correlation (r₍₃₈₁₎) matrix of psychosocial stress and SHS.

Scales	1	2	3	4	5	6
1. SHS	1.00
2. Demands at work	0.54^a	1.00
3. Work organization and  content	−0.24^a	−0.08	1.00
4. Interpersonal relations  and leadership	−0.47^a	−0.27^a	0.51^a	1.00
5. Insecurity at work	0.43^a	0.26^a	0.03	−0.14^b	1.00
6. Job satisfaction	−0.30^a	−0.41^a	0.31^a	0.37^a	−0.15^b	1.00

^ap < 0.01.

^bp < 0.05.

Table 3. Effect of psychosocial stress on SHS – results of the multivariate linear regression.

Variables	Coefficients	Standardized coefficients	t(373)	p Value
Age	0.104	0.046	0.910	0.364
Gender	4.518	1.894	1.898	0.059
Company	0.565	0.016	0.266	0.792
Demands at work	0.232	0.359	6.236	<0.001
Work organization and  content	−0.044	−0.057	−0.939	0.349
Interpersonal relations  and leadership	−0.217	−0.299	−4.939	<0.001
Insecurity at work	0.174	0.321	6.337	<0.001
Job satisfaction	0.037	0.048	0.807	0.421

RNA electrophoresis and the A260/280 ratio (ranging from 1.80 to 2.22) indicated that both integrity and purity of the isolated RNA samples were acceptable. Reverse transcription PCR (RT-PCR) analysis indicated the presence of specific bands of the correct sizes (308 bp for GRα, 178 bp for GRβ and 268 bp for β-actin). Sequences of resulting PCR products were confirmed to be identical to the reported GRα and GRβ cDNA sequences by direct sequencing (data not shown).

The average time interval between awaking and sampling was 57.22 ± 8.87 min among the participants. There were no significant correlations between cortisol levels and intervals between waking and sampling (Pearson r₍₃₈₁₎ = −0.072, p = 0.163). To clarify whether plasma cortisol level and GRα/GRβ expression in PBMCs assessed by RT-PCR correlated with SHS, the measured values were summarized according to the level of SHS. As listed in Table 4, plasma cortisol levels were significantly higher among the high SHS group than the low SHS group (p < 0.001). The amount of GRβ mRNA expression showed the same distribution in the two groups (p = 0.559). In contrast, the amount of GRα mRNA expression in the high SHS group was significantly lower than those in low SHS group (p = 0.027). As a result, the GRβ/GRα ratio in the high SHS group was significantly greater than that of the other group (p = 0.021). These results showed that GRα mRNA expression was reduced in the high SHS group, which indicated that cortisol may down-regulate the expression of GRα. The bivariate analysis showed that there was significant correlation between the amount of GRα mRNA and plasma cortisol levels in our subjects (r₍₃₈₁₎ = −0.48, p < 0.001). The result indicates that decreased expression of GRα mRNA was correlated to the severity of SHS.

Table 4. Association between SHS and plasma cortisol, mRNA of GRα and GRβ in PBMC.

	High SHS group (n = 193)	Low SHS group (n = 189)	F (1, 375)^a	p Value
Plasma cortisol^b (ng/ml)	178.58 ± 18.25	167.77 ± 12.25	37.49	<0.001
	175.54 (163.42–182.48)	167.51 (157.02–175.98)
mRNA GRα^b	86.79 ± 44.79	100.52 ± 58.71	4.941^c	0.027
	81.57 (53.36–112.41)	92.41 (53.81–128.00)
mRNA GRβ^b	73.77 ± 56.35	73.22 ± 52.39	0.293^c	0.589
	52.10 (28.84–120.82)	55.14 (31.93–121.66)
GRβ/GRα ratio^b	0.93 ± 0.64	0.78 ± 0.47	5.370	0.021
	0.81 (0.49–1.14)	0.61 (0.46–1.11)

^aAdjusted for sex, age, company and time interval (between waking and sampling).

^bData are represented as means ± SD and median (25–75 percentiles).

^cSkewed distributed and analyzed by log-transformed values.

Discussion

With the increasing economic development and urbanization in China, the prevalence of SHS is expected to escalate. Studies on intervention and prognosis of SHS will become increasingly important. We have established a valid instrument (SHSQ-25) for investigating SHS (Yan et al., 2009). It is a self-rated questionnaire of perceived health complaints. The questionnaire includes 25 items covering five subscales including fatigue, cardiovascular system, digestive tract, immune system and mental status (Yan & Wang, 2008; Yan et al., 2009).

One purpose of this study was to identify factors strongly associated with SHS among staff from three companies in Beijing, China. We analyzed the relationship between SHS and psychosocial stress associated with work. We also investigated distribution of circulating cortisol and GR isoforms as physiological mediators of psychosocial stress among different levels of SHS. We found that plasma cortisol was significantly increased, and GRα mRNA expression in PBMCs was significantly decreased, among subjects reporting high levels of SHS.

In this study, the valid and relevant short version of COPSOQ instrument was jointly used to assess psychosocial stress at work. This instrument covers the broad construct of work-related psychological factors by using a multidimensional concept including seven major theories in occupational health psychology (Kristensen et al., 2005). The COPSOQ is applicable in all sectors of the labor market and makes national and international comparisons possible.

We found two psychosocial aspects that were strong predictors of high level of SHS, i.e. “demands at work” and “insecurity at work”. Job demands, in particular quantitative demands, have been studied by many researchers, and the data indicate a consistent association with poor general health, especially with respect to the dimension of emotional exhaustion (Gray-Stanley & Muramatsu, 2011; Sundblad et al., 2008). Job insecurity is the work-related stressor that has been among the least systematically investigated. It reflects the discrepancy between the level of security a person experience and the preferable level of security. The results of the longitudinal study by Dekker & Schaufeli (1995) indicated that job insecurity is associated with symptoms of psychological distress and poor health status. The high quality of “interpersonal relations and leadership” covering the items of predictability, quality of leadership, social support, feedback at work and sense of community is a protective predictor of SHS (Table 3). The perception of support from others is a resource in helping to manage demands at work and is associated with reduced stress levels and burnout scores (Gray-Stanley & Muramatsu, 2011; Skirrow & Hatton, 2007). Special importance is attached to the support given by supervisors. Studies by Gibson et al. (2009) indicated that staff who could entrust themselves openly to their supervisors about work-related or private problems scored significantly lower on the burnout scale.

The observation of significant higher level of plasma cortisol among the high SHS group compared with the low SHS group (p < 0.001) strengthens the evidence that stress is an important related factor for SHS. Increased cortisol secretion might mediate perceived difference of health complaints and SHS. The data also support findings that hypersecretion of cortisol is one of physiological pathways that mediate the effects of chronic psychosocial stress (Vedhara et al., 1999). Lovell’s study revealed differential patterns of cortisol secretion, i.e. driven by elevated evening values, participants in the higher stress group displayed flatter diurnal cortisol slopes and elevated cortisol levels across the day in a young, otherwise healthy sample (Lovell et al., 2011). However, a single-point determination of cortisol in this study could not constitute a representation of disrupt pattern of HPA secretion under SHS.

Because GR is essential for cortisol action, we also investigated GR expression, including GRα and GRβ variants. The data indicate that expression of GRα mRNA in high SHS group was significantly lower than those in low SHS group. There were reports that reduced GRα-regulated expression, and increased nuclear factor kappa B-regulated expression were identified in leukocyte RNA from adults using differential gene expression analysis on microarrays, gene ontology analysis and transcription factor binding motif analyses as biomarkers of exposure to chronic psychosocial stressors (Cole et al., 2007; Miller et al., 2009). The negative correlation between the amount of GRα mRNA and plasma cortisol level in our subjects may be explained by the notion that hypersecretion of cortisol can down-regulate the GR expression (Raison & Miller, 2003). This association between chronic stress and alterations in GC signal transduction was reflected by altered gene expression (Miller et al., 2008). Although no significant difference was observed for expression of GRβ mRNA between the two groups (p = 0.559), the mRNA GRβ/GRα ratio in participants showing a higher level of SHS as a whole, was higher than those in the control group. Increased expression of dominant negative GRβ isoform, a splice variant of the GR, has been implicated in inflammatory response and GC resistance in human beings (Pujols et al., 2007; Schaaf & Cidlowski, 2002).

Our data suggest that psychosocial stress increases plasma cortisol, which subsequently impacts GR expression and splicing, and that psychosocial stress is a major risk factor leading to SHS. However, there are several limitations in our study. A single-point determination of cortisol is not sufficient to be considered a definitive measure of HPA axis activity. Although the short sampling period ensures some conformity in sampling time, individual time of waking for each participant would influence the measured levels of cortisol in a single sample. Furthermore, the association of psychosocial stress with SHS could not be proved as causality in a cross-sectional study, further longitudinal cohort studies merit to confirm our findings. An on-going longitudinal SHS cohort survey (China Sub-optimal Health Cohort Study) will provide a powerful health trial to fulfill this task (Wang et al., 2014).

Conclusion

This study confirmed the association between chronic psychosocial stress and SHS, indicating that improving the psychosocial work environment may reduce SHS and then prevent chronic diseases effectively.

Acknowledgements

We thank all the participants for their enrollment in the study, and we appreciate our colleague Desmond Menon for his support in English editing of this manuscript.

Declaration of interest

There was no conflict of interest in this study.

This study was funded by the National Natural Science Foundation (81102208, 81273170 and 30771193), the Beijing Natural Science Foundation (7133229), the National Science and Technology Support Program (2012BAI37B03), the Importation and Development of High-Caliber Talents Project of Beijing Municipal Institutions (CIT&TCD201304181), Mason Foundation National Medical Program–ANZ–Australia (CT21946), the Australia-China Science and Research Fund (ACSRF06444), the National ECU Industry Collaboration Scheme 2013 (G1001368) and Beijing Municipal Key Laboratory of Clinical Epidemiology Fund Project (2012LCLB01).