Associations between electronic devices use and common mental traits: A gene–environment interaction model using the UK Biobank data

1 INTRODUCTION

Mental disorders account for a large part of the global disease burden, accounting for 32.4% of the life years of disabled persons (YLDs) and 13.0% of the disability-adjusted life years (DALYs).1 Anxiety, depression, cigarette smoking and alcohol drinking are four common mental traits. Based on the Global Burden of Disease, Injuries, and Risk Factors Study 2017 (GBD 2017), depression was the third leading cause of disability worldwide.2 Depression and anxiety perplexed about 2.64 million and about 2.84 million people worldwide, respectively.2 A recent meta-analysis found that the risk ratio (RR) of frailty based on cigarette smoking was 1.22, increasing the risk of disability.3 In addition, Salonsalmi et al. conducted a cohort study with an average follow-up time of 8 years and found that adverse alcohol drinking was associated with disability retirement in the working population.4 Due to the shortage of medical resources, the probability of patients receiving treatment and the treatment rate were very low.5, 6

Previous studies have demonstrated the important role of genetic variants in the development of common mental traits. Five twin studies suggested familial aggregation of depression was mostly caused by additive genetic effects (heritability of liability = 37%).7 Meta-analysis of twin studies suggested that approximately 30%–60% of anxiety disorder8 could be explained by genetic effects. Using 7694 MZ and DZ twin pairs of Dutch population, Hamdi et al. observed that the twin-based heritability of alcohol dependence was estimated at 60%.9 A recent genome-wide association study (GWAS) of up to 1.2 million people identified 406 gene loci associated with tobacco use or/and alcohol abuse.10 A GWAS meta-analysis identified 44 independent loci associated with major depression disorder (MDD) in 135 458 cases and 344 901 controls and demonstrated the shared portion of biological etiology between MDD and schizophrenia.11

Most mental traits are caused by a variety of factors, including causal genes, environmental factors and gene–environment interactions.12 Traditional genetic association studies usually only consider the role of a single genetic variants and do not consider its potential role with other factors such as environmental factors, which may lead to an underestimate of the role of heredity.13 So, a genome-wide gene–environment interaction study (GWEIS) was used in the study, which is designed on the notion that individuals respond differently to environmental stimuli based on their genotype.14 Previous studies demonstrated that GWEIS could help to uncover new genetic risk transitions and elucidate potential biological pathways.14, 15 For instance, Cui et al. conducted a GWEIS between platinum-based chemotherapy response, drug toxicity and gene and demonstrated the interactions could predict the drug response and toxicity of platinum chemotherapy in non-small-cell lung cancer patients.16

In recent years, computers, tablets, smartphones and other electronic devices are taking up more and more time in people's daily life.17 Therefore, researchers investigated the effect of electronic devices use on health in different countries and regions. They found that high frequency and prolonged use of electronic devices can increase the burden on the body and increase the risk of adverse outcomes, such as hypertension, endocrine or metabolic disorder, stroke and respiratory diseases.18, 19 Regarding electronic devices use and mental traits, several cross-sectional studies have reported significant associations of electronic screen use with anxiety and depression.19-21 Comparing screen use such as TV watching and computer using with other sedentary behaviours in older adults, the study results showed that women who watched TV exhibited higher levels of depressive symptoms in other behaviours.21

We aimed to explore the effects of electronic devices use on mental health in the European population and to investigate its underlying genetic mechanisms in this study. We estimated the association of electronic devices with anxiety, depression, cigarette smoking and alcohol drinking through a linear regression model. GWEIS was then applied to investigate the interaction effects of genes and electronic devices use on the risks of four common mental traits.

2 MATERIALS AND METHODS 2.1 UK Biobank cohort

The UK Biobank health resource is an open access resource. (http://www.ukbiobank.ac.uk/abbioban-uk/), which included health, hospital records and genetic data from 502 656 participants aged 40–69 in 2006 and 2010. The large prospective study collected a series of reliable data through questionnaires, interviews and physical measures from 22 assessment centres. It was approved by UK Biobank's research ethics committee and Human Tissue Authority Research Tissue bank and got the participants' electronic signed consent. All participants agreed to use their anonymous data and samples to conduct any health-related studies and to reconnect for further sub-studies.22

2.2 Phenotypes definition

Three indicators of electronic devices use were included in our study, including TV watching, computer using and computer playing. TV watching was defined from the response of the question ‘In a typical DAY, how many hours do you spend watching TV?’. Computer using was defined from the response of the question ‘In a typical DAY, how many hours do you spend using the computer?’. And computer playing was measured through the question ‘Do you play computer games?’.

For mental traits, anxiety score and depression score were defined according to Generalized Anxiety Disorder (GAD)-7 and Patient Health Questionnaire (PHQ)-9.23 PHQ-923 is a total score (0–27) classification algorithm for screening and measuring depression severity, focusing on nine depressive symptoms and signs. And GAD-723 is also a classification algorithm with a total score of 0–21 to screen and measure anxiety severity, focusing on seven symptoms and signs of anxiety. More details were provided in the Supporting Information.

Cigarette smoking and alcohol drinking of each individual were defined by the amount of smoking and drinking they reported, respectively. We coded smoke cigarette smoking as 0 if ever-smoking status was also 0; otherwise, we used the maximum number of reported past or current cigarettes (or pipes/cigars) consumed per day. Similarly, we coded alcohol drinking as 0 if ever-drinking status was also 0; otherwise, we added up the average amount of different types of alcohol per week for these respondents. Those who reported drinking less than once a week were asked the average amount of different types of alcohol per month. For these subjects, we added up the total amount of alcohol consumed each month and divided it by 4 to get about the amount consumed per week. All variables were standardized to have mean 0 and variance 1 before further analysis. Detailed information of sample size, age and sex of each phenotype was provided in Table 1.

TABLE 1. Basic characteristics of the UK Biobank individuals included in this study TV watching Computer using Computer playing Anxiety score N 139 625 29 763 119 137 Sex (female) 78 412 15 214 68 027 Age (SD) 56.01 (7.72) 55.72 (7.41) 56.08 (7.66) Depression score N 138 976 29 636 118 615 Sex (female) 78 044 15,151 67,747 Age (SD) 56.02 (7.72) 55.74 (7.41) 56.09 (7.66) Alcohol drinking N 359 803 37 497 312 740 Sex (female) 183 495 17 507 163 351 Age (SD) 56.69 (8.04) 55.97 (7.48) 56.86 (7.97) Cigarette smoking N 383 742 38 599 330 985 Sex (female) 210 886 19 440 185 704 Age (SD) 56.53 (8.07) 55.7 (7.51) 56.73 (8.00) Note: Age was described as mean ± standard deviation. 2.3 Statistical analysis between electronic devices use and mental traits

We firstly estimated the relationship among electronic devices use, mental traits and the covariates through Spearman correlation analyses. And then a multivariable linear regression model was used to estimate the association between electronic devices use and mental traits. The instrumental variables were electronic devices use including TV watching, computer using and playing, and the outcome variables were mental traits including anxiety score, depression score, cigarette smoking and alcohol drinking. The outcome variables were adjusted for sex, age, educational attainment, economic Townsend deprivation index (TDI) and the first 10 principal components of population structure. To investigate the role of sex, the associations were also tested in women and men. And to explore the independent effect of electronic devices use on anxiety, we explored the associations between anxiety and electronic devices use with the other mental traits as covariates. Same analyses were performed for the other mental traits. Beta coefficient (B) with 95% confidence intervals (CI) and P-value were calculated. All statistical analyses were conducted by R 3.5.3 (https://www.r-project.org/).

2.4 Mendelian randomization analyses

The independent lead variants associated with electronic devices use (including TV watching, computer using and playing) at P < 5.00 × 10−8 were used as instrumental variables in the main inverse-variance-weighted (IVW) fixed effects meta analyses model of Mendelian randomization (MR). We conducted two-sample MR analyses through MR-base platform (http://app.mrbase.org/),24 and two-sided P < 0.05 was considered as statistically significant. Considering the pleiotropy in MR analyses, MR-Egger and weighted median model were tested. Detailed information about MR analysis model were introduced in the previous study.24 And the GWAS of electronic devices use and mental traits were provided in the Supporting Information.

2.5 UK Biobank genotyping, imputation and quality control

In UK Biobank, 488 377 participants were genotyped by either the UK BiLEVE array or the UK Biobank axiom array under favourable quality control. Details of the array design, genotyping and quality control procedures have been described elsewhere.25 These genotypes were subsequently imputed to the Haplotype Reference Consortium (HRC) reference panel26 (Version 1.1) and UK10K and 1000 Genomes project reference panels.25 We removed the participants whose report has inconsistencies between self-reported gender and genetic gender, or who were genotyped but not imputed, and who withdraw their consents. Additionally, individuals were restricted to only ‘white British’ based on self-reported ethnicity (UK Biobank field ID: 21000). UK used an estimator implemented in the KING software to screen out genetically related individuals.27

2.6 Genome-wide gene–environmental interaction analysis

GWEIS was conducted to estimate the interaction effects between single nucleotide polymorphism (SNP) and electronic devices on the risks of the four mental traits in UK Biobank cohort. The outcomes variables including anxiety score, depression score, cigarette smoking and alcohol drinking were adjusted by age, sex and the first 10 principal components of population structure. We used the generalized linear regression model of PLINK 2.0.28 Additional quality control filters were used to select high-quality SNPs: The SNPs with low call rates (<0.90), low Hardy–Weinberg equilibrium exact test P-values (<0.001) or low minor allele frequencies (MAFs; <0.01) were excluded.28 A significance threshold was set at P = 4.17 × 10−9 for genome-wide gene–environment interaction effects based on Bonferroni correction approach (5 × 10−8/12 = 4.17 × 10−9). Detailed information of GWEIS approach was provided in the Supporting Information. Visualization of all results was done using R software. Circular Manhattan plots were generated using the ‘CMplot’ R script (https://github.com/YinLiLin/R-CMplot).

3 RESULT 3.1 Descriptive characteristics of study samples

Table 1 show the descriptive characteristics of study sample.

3.2 Association between electronic devices use and mental traits

Table S1 shows that sex, age, anxiety score, depression score, cigarette smoking, alcohol drinking, TV watching, computer using, computer playing, TDI and educational attainment were associated with each other, except the association between using computer and age, anxiety score and TV watching.

Significant associations were observed between TV watching and mental traits (all P < 1.0 × 10−9), including anxiety score (B = 0.051), depression score (B = 0.094), cigarette smoking (B = 0.100) and alcohol drinking (B = 0.042). Computer using appeared to be associated with anxiety score (B = 0.043, P < 1.0 × 10−9), depression score (B = 0.078, P < 1.0 × 10−9) and cigarette smoking (B = 0.046, P < 1.0 × 10−9). Computer playing was also found to be associated with anxiety score (B = 0.017, P = 1.19 × 10−9), depression score (B = 0.053, P < 1.0 × 10−9), cigarette smoking (B = 0.058, P < 1.0 × 10−9) and alcohol drinking (B = 0.010, P < 1.0 × 10−9).

As Figure 1 shows, the associations between electronic devices use and mental traits were also significant in women and men, except the association between the frequency of alcohol drinking and computer using (P > 0.05 in women and P = 1.71 × 10−5 in men). In addition, we also explore the effects of electronic devices use on mental traits in men and women, such as the association between the frequency of cigarette smoking and TV watching (B = 0.096, P < 1.0 × 10−9 in women and B = 0.111, P < 1.0 × 10−9 in men) and the association between depression score and computer playing (B = 0.056, P < 1.0 × 10−9 in women and B = 0.050, P < 1.0 × 10−9 in men).

As Table 2 shows, cigarette smoking and depression score are also associated with electronic devices use even after adjusting the other mental traits. And for TV watching, it was significantly and independently associated with mental traits. More additional association details were provided in Tables 2 and S1.

image

Association between electronic devices use and mental traits based on sex using the UK population data. Notes: The x-axis refers to beta coefficient (B). The y-axis represents the instrumental variables. Points display the Beta and 95% CIs (error bars) of Beta. Except for special notes (P > 0.05), all other results showed P < 0.05

TABLE 2. Association between electronic devices and mental traits TV watching Computer using Computer playing Beta P Beta P Beta P Anxiety score Model 0 0.051 <1.00 × 10−09 0.043 <1.00 × 10−09 0.017 1.19 × 10−09 Model 1 0.041 <1.00 × 10−09 0.034 5.74 × 10−07 0.014 3.97 × 10−05 Model 2 −0.012 2.17 × 10−02 −0.010 6.47 × 10−02 −0.013 8.27 × 10−07 Depression score Model 0 0.094 <1.00 × 10−09 0.078 <1.00 × 10−09 0.053 <1.00 × 10−09 Model 1 0.076 <1.00 × 10−09 0.066 <1.00 × 10−09 0.042 <1.00 × 10−09 Model 2 0.064 <1.00 × 10−09 0.057 <1.00 × 10−09 0.037 <1.00 × 10−09 Alcohol drinking Model 0 0.042 <1.00 × 10−09 0.007 1.51 × 10−01 0.010 <1.00 × 10−09 Model 1 0.031 <1.00 × 10−09 0.010 9.24 × 10−02 0.017 2.51 × 10−08 Model 2 0.019 2.87 × 10−09 0.001 9.30 × 10−01 0.006 6.55 × 10−02 Cigarette smoking Model 0 0.100 <1.00 × 10−09 0.046 <1.00 × 10−09 0.058 <1.00 × 10−09 Model 1 0.081 <1.00 × 10−09 0.043 <1.00 × 10−09 0.055 <1.00 × 10−09 Model 2 0.072 <1.00 × 10−09 0.030 4.15 × 10−06 0.050 <1.00 × 10−09 Notes: Model 0 was adjusted the sex, age, educational attainment, Townsend deprivation index and the first 10 principal components of population structure. For anxiety score and depression score, Model 1was adjusted Model 0 + the frequency of alcohol drinking and cigarette smoking; for the frequency of alcohol drinking and cigarette smoking, Model 1was adjusted Model 0 + anxiety score and depression score. Model 2: Model 0 + the other three mental traits. 3.3 Causal relationship between electronic devices use and mental traits

A series of MR analyses was performed to test the hypothesis that increased TV watching, computer using and playing are causal risk factors for mental traits. Using the IVW-MR fixed effects approach, the results showed that a causal effect was estimated between alcohol and all the electronic devices use, including TV watching (B = −0.049, P = 8.61 × 10−3), computer using (B = −0.055, P = 2.21 × 10−3) and computer playing (B = −0.089, P = 0.040). Cigarette smoking and depression were significantly associated with TV watching (B = 0.406, P = 5.03 × 10−15 for cigarette smoking and B = 0.527, P = 0.023 for depression) and computer using (B = 0.546, P = 1.69 × 10−5 for cigarette smoking and B = 0.657, P = 8.58 × 10−3 for depression). The association between anxiety and computer playing was also significant (B = −1.354, P = 1.94 × 10−3) (Table 3). Most relationships between mental traits were consistent with the pleiotropy analyses. More information was provided in Table S2.

TABLE 3. Mendelian randomization (MR) estimates of electronic devices on mental traits Outcome Instrument N SNPs Beta SE P Anxiety TV watching 85 0.191 0.188 0.310 Cigarette smoking TV watching 109 0.406 0.052 5.03 × 10−15 Alcohol drinking TV watching 110 −0.055 0.018 2.21 × 10−03 Depression TV watching 79 −0.527 0.232 0.023 Anxiety Computer using 26 −0.336 0.259 0.194 Alcohol drinking Computer using 82 −0.049 0.019 8.61 × 10−03 Cigarette smoking Computer using 83 −0.051 0.053 0.338 Depression Computer using 59 −0.657 0.250 8.58 × 10−03 Anxiety Computer playing 40 −1.354 0.437 1.94 × 10−03 Cigarette smoking Computer playing 46 0.546 0.127 1.69 × 10−05 Alcohol drinking Computer playing 47 −0.089 0.043 0.040 Depression Computer playing 34 −0.896 0.576 0.120 Note: Summary MR estimates of the causal associations between electronic devices use and mental traits using inverse-variance-weighted fixed effects method. 3.3.1 Interactions analysis results of gene × electronic devices use

GWEIS identified multiple candidate loci interacting with TV watching on mental traits, such as TDRP (rs1448088, P = 1.20 × 10−8) for anxiety score, CDH13 (rs76087704, P = 1.59 × 10−9) for depression score and KCNQ3 (rs17653044, P = 8.77 × 10−11) for alcohol drinking (Table 4). More additional details were provided in Figure S1 and Tables S3S7.

GWEIS identified multiple candidate loci with computer using on mental traits, such as TEX19 (rs72856174, P = 1.18 × 10−9) for anxiety score, ENOX1 (rs142337497, P = 4.29 × 10−10) for depression score and PRKCE (rs56181965, P = 9.64 × 10−10) for cigarette smoking (Table 4). More additional details were provided in Figure S2 and Tables S8–S11.

GWEIS identified multiple candidate loci with computer playing on mental traits, such as CAC15A (rs139070788, P = 5.01 × 10−14) for anxiety score and FRMD4A (rs56227933, P = 7.42 × 10−11) for depression score (Table 4). More additional details were provided in Figure S3 and Tables S12–S15. And we provided the boxplot of the top significant results in Figures S4–S6.

TABLE 4. Top significant results from GWEIS approach TV watching Computer using Computer playing Gene SNP P Gene SNP P Gene SNP P Anxiety score TDRP rs1448088 1.20 × 10−8 TEX19 rs72856174 1.18 × 10−9 CDC25A rs139070788 5.01 × 10−14 TDRP rs1900696 2.25 × 10−8 TEX19 rs72856167 1.46 × 10−9 SMARCC1 rs150488176 7.20 × 10−14 Depression score CDH13 rs76087704 1.59 × 10−09 ENOX1 rs9525780 8.61 × 10−10 FRMD4A rs56227933 7.42 × 10−11 YIPF1 rs115493768 2.97 × 10−09 ENOX1 rs9533497 8.99 × 10−10 ATP8B1 rs144512701 1.03 × 10−10 Cigarette smoking PDE4D rs79364791 6.00 × 10−09 PRKCE rs56181965 9.64 × 10−10 SIPA1L1 rs35589413 4.21 × 10−08 PDE4D rs186947563 6.63 × 10−09 FREM3 rs147934227 1.45 × 10−09 SIPA1L1 rs56180634 4.31 × 10−08 Alcohol drinking GSTA3 rs562858885 6.80 × 10−11 CNKSR3 rs9479856 4.11 × 10−08 DENND1A rs113977954 2.08 × 10−08 KCNQ3 rs17653044

留言 (0)

沒有登入
gif