Leisure-Time Television Viewing and Computer Use, Family History, and Incidence of Dementia

Abstract

Introduction

Time spent on screen-based sedentary activities is significantly associated with dementia risk, however, whether the associations vary by family history (FHx) of dementia is currently unknown. We aimed to examine independent associations of two prevalent types of screen-based sedentary activities (television [TV] viewing and computer use) with dementia and assess the modifying effect of FHx.

Methods

We included 415,048 individuals free of dementia from the UK Biobank. Associations of TV viewing, computer use, and FHx with dementia risk were determined using Cox regression models. We estimated both multiplicative- and additive-scale interactions between TV viewing and computer use and FHx.

Results

During a median follow-up of 12.6 years, 5,549 participants developed dementia. After adjusting for potential confounding factors, we observed that moderate (2–3 h/day; hazard ratio [HR] 1.13, 95% confidence interval 0.03–1.23) and high (>3 h/day; 1.33, 1.21–1.46) TV viewing was associated with a higher dementia risk, compared with low (0–1 h/day) TV viewing. Using restricted cubic spline models, the relationship of TV viewing with dementia was nonlinear (relative to 0 h/day; p for nonlinear = 0.005). We found that >3 h/day of TV viewing was associated with a 42% (1.42, 1.18–1.71) higher dementia risk in participants with FHx while a 30% (1.30, 1.17–1.45) in those without FHx. For computer use, both low (0 h/day; 1.41, 1.33–1.50) and high (>2 h/day; 1.17, 1.05–1.29) computer use were associated with elevated dementia risk, compared with moderate (1–2 h/day) computer use. We observed a J-shaped relationship with dementia (relative to 2 h/day; p for nonlinear <0.001). Compared with 1–2 h/day of computer use, the HRs of dementia were 1.46 (1.29–1.65) and 1.10 (0.90–1.36) for 0 h/day and >2 h/day of computer use in participants with FHx, respectively, while the corresponding HRs were 1.40 (1.30–1.50) and 1.19 (1.06–1.33) in those without FHx. We observed a positive additive interaction (RERI 0.29, 0.06–0.53) between computer use and FHx, while little evidence of interaction between TV viewing and FHx.

Conclusions

The time spent on TV viewing and computer use were independent risk factors for dementia, and the adverse effects of computer use and FHx were additive. Our findings point to new behavioral targets for intervention on preventing an early onset of dementia, especially for those with FHx.

Introduction

Dementia, characterized by a progressive and unrelenting impairment of cognitive faculties [1], has been a major public health concern posing a substantial burden to patients, their families, and the society [2–4]. It is estimated that 152 million individuals would have dementia by 2050 worldwide [5]. Both genetic factors (e.g., susceptibility gene loci) and environmental factors (e.g., physical activity) play an essential role in determining risk of dementia, individually or in interaction with each other [6]. Given the lack of a cure, efforts to develop preventive strategies that targets potential risk factors are a public health priority.

In the past few decades, it has been observed that there is a shift in the activity profile of individuals with physical activity and sleep being partially replaced by screen-based sedentary activities, which are potential neurogenic stress components due to their neurophysiological effects, such as reduced cerebral blood flow, inflammation, and synaptic decline, contributing to various effects on cognitive functions [7, 8]. According to the UK Chief Medical Officers’ Physical Activity Guidelines, screen-based sedentary activities such as television (TV) viewing and computer use are among the most common sedentary behaviors [9]. Although several population-based studies have associated time spent on TV viewing and computer use with cognitive performance [10–12], including verbal memory, executive function, processing speed as well as global cognitive function, only a few studies have investigated the association of TV viewing with dementia risk [13–16]. Most importantly, family history (FHx) of dementia is a risk factor independently associated with dementia among asymptomatic subjects, which reflects the inherited and shared environmental predisposition to dementia. However, no previous research on the role of different types of screen-based sedentary activities in dementia has considered the influence of individuals’ FHx of dementia. A previous study has suggested a twofold higher lifetime risk of dementia among those with FHx (20%) than the general population (10%) [17]. As such, the presence of these two dementia risk markers can help identify asymptomatic subjects at elevated dementia risk and may be useful for guiding primary prevention therapy decisions.

Therefore, we aimed to prospectively explore the associations between time spent on screen-based sedentary activities (including TV viewing and computer use) and the risk of all-cause dementia in the UK Biobank. In addition, we tested possible interactions of FHx with TV watching and computer use in relation to all-cause dementia risk.

Methods

Study Population

The UK Biobank study is a large-scale population-based prospective cohort study, which recruited more than 500,000 adults aged 37–73 years at baseline from the general population [18]. Briefly, participants attended 1 of the 22 assessment centers between April 2007 and December 2010 across Scotland, England, and Wales. At baseline, all participants completed a touch-screen questionnaire to collect information on socioeconomic status, lifestyle factors, personal medical history of chronic disease, family history disease, and so on. Physical measurements, including height, weight, and waist circumference, were obtained by trained interviewers. Information on all available UK Biobank resources can be found online (http://www.ukbiobank.ac.uk/resources/). In this study, we excluded participants with incomplete data on TV viewing, computer use, and ethnicity (N = 76,231) and those with prevalent dementia (N = 190) at baseline. Given the possibility that underlying dementia may cause changes in behaviors in advance of diagnosis, we excluded incident dementia cases that occurred in the first 5 years from baseline assessment (N = 7,992). We also excluded participants with the total time spent on TV viewing, computer use, sleep, and physical activity exceeding 24 h/day (N = 3,180). The final analysis included a total of 415,048 participants.

The UK Biobank has been approved by the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland, and the North West Multicenter Research Ethics Committee. All participants have electronically signed consent forms. This research was conducted under the UK Biobank Application Number 44430. This study was also approved by the Ethical Committee of Peking University (Beijing, China).

Exposure Assessment

At baseline, time spent on leisure-time TV viewing and computer use (both are nonoccupational) were assessed using a self-reported, touch-screen questionnaire. For TV viewing, participants were asked: “In a typical day, how many hours do you spend watching TV?” For leisure-time computer use, they were also asked: “In a typical day, how many hours do you spend using the computer? (Do not include using a computer at work).” If the time participant spent on TV viewing and computer use varies a lot, then the participant was asked to give the average time for a 24-h day in the last 4 weeks. The following checks were performed for the answers: (1) if the answer <0 or >24 then rejected; (2) if answer >6 then asked the participant to confirm. We treated participants that answered “do not know” or “prefer not to answer” as missing; while the answer “less than an hour a day” was replaced by 0.5 h/day, the median between 0 and 1 h/day. TV viewing was divided into three categories including low (0–1 h/day), moderate (2–3 h/day), and high (>3 h/day), while computer use was categorized into low (0 h/day), moderate (0.5–2 h/day), and high (>2 h/day), which were consistent with previous studies [12, 19, 20].

For each participant, the first-degree family members’ dementia history was assessed by the baseline questionnaire in the UK Biobank through a series of questions, which asked participant if their father, mother, or siblings (only for blood relations) ever had “Alzheimer’s disease/dementia.” In the current study, a participant was classified as “having a FHx of dementia” if they reported at least one first-degree relatives had dementia.

Incident Dementia

All-cause dementia within the UK Biobank study was ascertained using hospital inpatient records containing data on admissions and diagnoses obtained from the Hospital Episode Statistics for England, the Scottish Morbidity Record Data for Scotland, and the Patient Episode Database for Wales. We also detected additional cases through linkage to death register data from the National Health Service Digital for England and Wales and the Information and Statistics Division for Scotland. According to the International Classification of Diseases (ICD) coding system, prevalent and incident dementia were identified as having a primary/secondary diagnosis (hospital records) or underlying/contributory cause of death (death register) using ICD-9 and ICD-10 codes for dementia (online suppl. eTable 1; for all online suppl. material, see https://doi.org/10.1159/000531237) [21]. Additionally, self-reported dementia cases at baseline were classified as prevalent cases.

Covariates

In the UK Biobank, covariates were documented including age (continuous) (years), sex (male or female), ethnicity (White European, mixed, South Asian, Black, or other), UK Biobank assessment center, Townsend deprivation index (TDI, continuous), highest education level (lower secondary, upper secondary, higher, vocational, other, or missing), alcohol consumption (continuous) (grams/day), smoking status (current, former, never, or missing), body mass index (BMI, continuous) (kg/m²), systolic blood pressure (SBP, continuous) (mm Hg), physical activity (continuous) (metabolic equivalent of task, minutes/week), healthy diet score (0, 1, 2, 3, 4, 5, or missing), use of antihypertensive drugs (yes, no, or missing), use of lipid-lowering drugs (yes, no, or missing), use of insulin (yes, no, or missing), family history of dementia (yes, no, or missing). We used the TDI to estimate the regional socioeconomic status in the UK. BMI was calculated as weight in kilograms divided by height in meters square. We used MET obtained from the short International Physical Activity Questionnaire to measure physical activity. The healthy diet score was established with reference to the American Heart Association guidelines [22–24]. As a result, five healthy diet factors were identified using the median level of consumption: vegetable intake ≥4 tablespoons/day, fruit intake ≥3 pieces/day, fish intake ≥ twice/week, unprocessed red meat intake ≤ twice/week, and processed meat intake < twice per week. We gave one point for each favorable diet factor and calculated a healthy diet score of 0–5 using the sum of the corresponding score of each diet factor. For mental health, we asked if the participants have ever seen a psychiatrist or general practitioner for nerves, anxiety, tension, or depression, with responses of “yes” or “no.” For occupation, the information on current employment status was collected, including “in paid employment or self-employed,” “doing unpaid or voluntary work,” “unemployed,” “retired,” “looking after home and/or family,” or “full or part-time student.”

Statistical Analysis

Percentages and means (standard deviations) of baseline characteristics were described across incident dementia status for categorical and continuous variables, respectively. Survival time in person-year for each participant was calculated from the date of baseline assessment until the date of the first diagnosis of dementia, loss to follow-up, death, or the last date of hospital admission (March 31, 2017, for England; October 31, 2016, for Scotland; and February 29, 2016, for Wales), whichever came first. Cox regression was used to estimate the hazard ratios (HRs) and 95% confidence interval (CI) for the association of FHx and time spent on TV viewing and computer use with incident dementia, with age as the underlying time scale.

We constructed three multivariable models. In model 1, we adjusted for age, sex, ethnicity, and UK Biobank assessment center. In model 2, we further adjusted for TDI, highest education level, alcohol consumption, smoking status, BMI, SBP, physical activity, healthy diet score, use of antihypertensive drugs, use of lipid-lowering drugs, use of insulin, and FHx of dementia (only in analysis for TV viewing and computer use). In model 3, we further mutually adjusted for TV viewing and computer use on the basis of model 2. When the covariates were missing, a missing indicator category was coded for categorical variables and mean values were imputed for continuous variables. Population attributable fractions were calculated to estimate proportional risk reductions in dementia that would occur if time spent on TV viewing and computer use were restricted to the group with lowest risk, respectively, assuming causality.

We further evaluated whether FHx might modify the associations. We conducted a stratified analysis according to the FHx and explored the potential interactions with long-term dementia risk. We used the likelihood test to examine the multiplicative interaction by setting variable cross-product terms of FHx with TV viewing or computer use in the models. In terms of the additive interaction, we estimated relative excess risk due to interaction, and the attributable proportion due to interaction [25, 26]. When relative excess risk due to interaction is positive, it indicates super-additive interaction (i.e., increased risk due to the additive interaction). The attributable proportion due to interaction is the proportion of the risk due to the interaction in the doubly exposed group. These measures were used to investigate whether the risk due to having both risk factors is greater than the sum of the risks due to each condition.

We also conducted several sensitivity analyses to evaluate the robustness of our findings. Since central obesity as well as mental status might influence dementia risk [27], we first further adjusted for waist circumference, and nerves, anxiety, tension, or depression in the models. Various social supports might have different effects on the risk of dementia [28]; therefore, we also conducted a sensitivity analysis with further adjustment for frequency of friend/family visits, leisure/social activities, and frequency of confiding. In addition, we conducted the Fine‐Grey subdistribution hazard model to statistically account for the competing risk of death. Further, we excluded incident dementia cases that occurred in the first 8 years from baseline assessment to dementia diagnosis to minimize the reverse causality of observed association [29]. To rule out the influence of baseline chronic conditions, we further excluded participants with prevalent diabetes, cardiovascular diseases, and cancer. In addition, we also performed a sensitivity analysis excluding participants with imputed covariates. To investigate if the effects are consistent across multiple groups, we performed stratified analyses by age and occupation. All analyses were performed using Stata version 16.0 (StataCorp). All values were 2-sided and a p value <0.05 was regarded as statistically significant.

Results

The baseline characteristics of the participants by incident dementia status are shown in Table 1. Compared with those without incident dementia, participants who had incident dementia were older and had a higher proportion of men and white race. In addition, those with incident dementia tended to be current smokers, with lower education level, TDI and metabolic equivalent of task, and higher BMI and SBP. We observed higher percentage of FHx (24.1%) among those with incident dementia compared with those without incident dementia (13.3%). Furthermore, mean (standard deviation) estimates of time spent on TV viewing and computer use were 3.4 (1.9) m/s and 0.8 (1.2) h/day among participants with incident dementia, and the corresponding values were 2.7 (1.6) and 1.1 (1.4) for those without incident dementia, respectively. Baseline characteristics of participants by FHx and time spent on TV viewing and computer use are described in online supplementary eTables 2–4.

Table 1.

Baseline characteristics of participants by incident dementia

Characteristic	All participants (N = 415,048)	Incident dementia
		no (n = 409,499)	yes (n = 5,549)
Age, years	56.2 (8.1)	56.1 (8.1)	64.3 (4.7)
Male sex	45	45	51.7
White race	94.6	94.6	96.2
College or university degree	33	33.2	20.1
Townsend deprivation index	−1.3 (3.1)	−1.3 (3.1)	−1.0 (3.3)
BMI, kg/m2	27.4 (4.7)	27.4 (4.7)	27.8 (4.9)
MET, min per week	2,591.1 (2,317.1)	2,590.4 (2,315.9)	2,643.8 (2,401.7)
Alcohol consumption, g/day	14.8 (18.1)	14.8 (18.0)	14.0 (20.7)
SBP, mm Hg	137.5 (18.1)	137.4 (18.0)	143.7 (18.8)
Current smoker	10.2	10.2	10.6
Healthy diet score
0–1	15.6	15.6	14.1
2–3	50.6	50.7	49.8
4–5	33.8	33.8	36.1
Cholesterol-lowering drug use	22.7	22.3	48.2
Antihypertensive drug use	15.7	15.5	25.9
Insulin use	6	5.9	11.8
Diabetes	7.3	7.2	16.2
Vascular disease	28.9	28.6	49.2
Cancer	10	9.9	13.2
FHx of dementia	13.4	13.3	24.1
TV viewing, h/day	2.8 (1.6)	2.7 (1.6)	3.4 (1.9)
Computer use, h/day	1.1 (1.4)	1.1 (1.4)	0.8 (1.2)

Continuous and categorical variables are presented as means (SDs) and percentages (%), respectively.

BMI, body mass index; MET, metabolic equivalent of task; SBP, systolic blood pressure; SD, standard deviation.

During a median follow-up of 12.6 years (5,137,559 person-years), we documented 5,549 incident dementia cases. Compared with 0–1 h/day of TV viewing, 2–3 h/day and >3 h/day TV viewing was associated with a 19% (95% CI: 10–30%) and 59% (46–73%) higher risk of incident dementia in the model 1, respectively. After further adjustment for TDI, education, alcohol consumption, smoking status, BMI, SBP, physical activity, healthy diet score, use of antihypertensive drugs, lipid-lowering drugs and insulin, FHx of dementia, and computer use (model 3), the corresponding HRs of dementia were 1.13 (1.03–1.23) and 1.33 (1.21–1.46) (p trend <0.001) (Table 2). Restricted cubic spline models indicated a nonlinear association between TV viewing and dementia, where the risk increased markedly with >3 h/day of TV viewing (p for nonlinearity = 0.005) (Fig. 1). For computer use, the HRs of dementia for 0 h/day and >2 h/day of computer use were 1.63 (1.54–1.72) and 1.21 (1.09–1.33) compared with 1–2 h/day of computer use in the model 1, respectively. After multivariable adjustment (model 3), the risk of dementia (HR [95% CI]) in 0 h/day and >2 h/day of computer use was 1.41 (1.33–1.50) and 1.17 (1.05–1.29), respectively (Table 2). Using restricted cubic spline models, we found that there was a J-shaped association between computer use and dementia, where both low computer use and high computer use conferred greater dementia risk than did moderate computer use (2 h/day) (p for nonlinear <0.001) (Fig. 1). Assuming causality, around 15% of dementia could be prevented if all participants adhered to low TV viewing (0–1 h/day) or moderate computer use (1–2 h/day) (Table 3).

Table 2.

Associations of TV viewing and computer use with incident dementia

	Screen-based sedentary activities			p for trend
	low	moderate	high
TV viewing
Range, h/day	0–1	2–3	>3
Participants, n	87,263	209,863	117,922
Incidence rates per 1,000 person-years (95% CI)	0.62 (0.57, 0.67)	0.92 (0.89, 0.96)	1.70 (1.64, 1.77)
Model 1a	(Reference)	1.19 (1.10, 1.30)	1.59 (1.46, 1.73)	<0.001
Model 2b	(Reference)	1.13 (1.03, 1.23)	1.34 (1.22, 1.47)	<0.001
Model 3c	(Reference)	1.13 (1.03, 1.23)	1.33 (1.21, 1.46)	<0.001
Computer use
Range, h/day	0	1–2	>2
Participants, n	113,843	260,897	40,308
Incidence rates per 1,000 person-years (95% CI)	1.81 (1.74, 1.88)	0.78 (0.75, 0.81)	0.95 (0.86, 1.03)
Model 1a	1.63 (1.54, 1.72)	(Reference)	1.21 (1.09, 1.33)	<0.001
Model 2b	1.44 (1.35, 1.53)	(Reference)	1.16 (1.05, 1.29)	<0.001
Model 3c	1.41 (1.33, 1.50)	(Reference)	1.17 (1.05, 1.29)	<0.001

Data for models given as HR (95% CI).

HR, hazard ratio; CI, confidence interval.

^aAdjusted for age (continuous) (years), sex (male or female), ethnicity (White European, mixed, South Asian, Black, or other), and UK Biobank assessment center.

^bFurther adjusted for Townsend deprivation index (continuous), highest education level (lower secondary, upper secondary, higher, vocational, other, or missing), alcohol consumption (continuous) (grams), smoking status (current, former, never, or missing), BMI (continuous) (kg/m²), SBP (continuous) (mm Hg), physical activity (continuous) (metabolic equivalent of task [minutes]), healthy diet score (0, 1, 2, 3, 4, 5, or missing), use of antihypertensive drugs (yes, no, or missing), use of lipid-lowering drugs (yes, no, or missing), use of insulin (yes, no, or missing), FHx of dementia (yes, no, or missing).

^cFurther mutual adjustment of the two exposure variables (TV viewing and computer use).

Fig. 1.

Cubic spline models representing trends of associations between continuous variables of TV viewing (relative to 0 h/day of TV viewing), computer use (relative to 2 h/day of computer use), and dementia risk. The model was adjusted for age (continuous) (years), sex (male or female), ethnicity (White European, mixed, South Asian, Black, or other), UK Biobank assessment center, Townsend deprivation index (continuous), highest education level (lower secondary, upper secondary, higher, vocational, other, or missing), alcohol consumption (continuous) (grams), smoking status (current, former, never, or missing), BMI (continuous) (kg/m²), SBP (continuous) (mm Hg), physical activity (continuous) (metabolic equivalent of task [minutes]), healthy diet score (0, 1, 2, 3, 4, 5, or missing), use of antihypertensive drugs (yes, no, or missing), use of lipid-lowering drugs (yes, no, or missing), use of insulin (yes, no, or missing), FHx of dementia (yes, no, or missing), with mutual adjustment of the two exposure variables (TV viewing and computer use).

Table 3.

Multivariable-adjusted HRs (95% CI) and PAR% (95% CI) of time spent in TV viewing and computer use for dementia events by categories of FHx

	HR (95% CI)	Risk difference, %	PAR% (95% CI)
TV viewing, h/day
All participants	1.09 (1.07, 1.10)	9	15.90 (9.61, 21.75)
FHx+	1.09 (1.05, 1.12)	9	18.83 (6.14, 29.80)
FHx−	1.09 (1.07, 1.11)	9	14.85 (0.75, 21.64)
Computer use, h/day
All participants	0.96 (0.94, 0.98)	−4	14.57 (12.28, 16.79)
FHx+	0.94 (0.89, 0.99)	−6	14.20 (9.88, 18.32)
FHx−	0.96 (0.94, 0.99)	−4	14.67 (11.97, 17.29)

Data for models given as HR (95% CI).

HR, hazard ratio; CI, confidence interval; PAR%, population attributable risk percent.

The model was adjusted for age (continuous) (years), sex (male or female), ethnicity (White European, mixed, South Asian, Black, or other), UK Biobank assessment center, Townsend deprivation index (continuous), highest education level (lower secondary, upper secondary, higher, vocational, other, or missing), alcohol consumption (continuous) (grams), smoking status (current, former, never, or missing), BMI (continuous) (kg/m²), SBP (continuous) (mm Hg), physical activity (continuous) (metabolic equivalent of task [minutes]), healthy diet score (0, 1, 2, 3, 4, 5, or missing), use of antihypertensive drugs (yes, no, or missing), use of lipid-lowering drugs (yes, no, or missing), use of insulin (yes, no, or missing), with mutual adjustment of the two exposure variables (TV viewing and computer use).

In the model 1 adjusting for age, sex, ethnicity, assessment center, and UK Biobank assessment center, we observed a significant association of FHx with the risk of incident dementia. The FHx was associated with a 63% higher risk of incident dementia (95% CI: 43–76%) (online suppl. eTable 5). The association remained significant in the multivariable-adjusted model: the HR of dementia was 1.70 (1.49–1.81) for participants with FHx, compared with those without FHx (online suppl. eTable 5).

Among participants without FHx, compared with 0–1 h/day of TV viewing, the 10-year absolute risk of dementia increased by 0.26% and 0.98% for 2–3 h/day and >3 h/day of TV viewing, respectively. Likewise, the rates were increased by 0.53% and 1.64% for 2–3 h/day and >3 h/day of TV viewing in the group with FHx (Table 4). When compared with the group of 1–2 h/day computer use, 0 h/day and >2 h/day of computer use were associated with a 10-year absolute risk elevation in dementia of 0.94% and 0.16% among participants without FHx, respectively. The rates were increased by 1.67% and 0.22% in 0 h/day and >2 h/day of computer use in the group with FHx, respectively (Table 4). In addition, based on the Kaplan-Meier curves of cumulative incidence for dementia during follow-up, the participants with FHx who were classified as high TV viewing or low computer use had the highest cumulative incidence in TV viewing and computer use groups, respectively (Fig. 2).

Table 4.

Risk of incident dementia according to TV viewing, computer use, and FHx of dementia

	FHx−			FHx+
	low	moderate	high	low	moderate	high
TV viewing
Cases/person-years	509/950,218.8	1,819/2,264,326	1,885/1,243,969	164/136,089.5	587/337,909.3	585/205,047.3
10-year absolute risk (95% CI)	0.54 (0.49, 0.58)	0.80 (0.77, 0.84)	1.52 (1.45, 1.58)	1.21 (1.02, 1.39)	1.74 (1.60, 1.88)	2.85 (2.62, 3.08)
Absolute risk elevation	(Reference)	0.26%	0.98%	(Reference)	0.53%	1.64%
Computer use
Cases/person-years	1,974/1,229,380	1,879/2,797,792	360/431,340.5	573/180,774.9	654/434,752.9	109/63,518.39
10-year absolute risk (95% CI)	1.61 (1.53, 1.68)	0.67 (0.64, 0.70)	0.83 (0.75, 0.92)	3.17 (2.91, 3.43)	1.50 (1.39, 1.62)	1.72 (1.39, 2.04)
Absolute risk elevation	0.94%	(Reference)	0.16%	1.67%	(Reference)	0.22%

CI, confidence interval.

Fig. 2.

Cumulative incidence of dementia events across TV viewing, computer use, and FHx groups of UK Biobank participants. All p values for the log-rank test of six groups were <0.001.

We further assessed the interaction between TV viewing and computer use and FHx and the risk of dementia. For TV viewing, the multiplicative interaction was not insignificant (p = 0.901), and there was no evidence of additive interaction between TV viewing and FHx with a relative excess risk due to interaction of 0.18 (−0.04 to −0.39). Compared with 0–1 h/day of TV viewing, >3 h/day of TV viewing was associated with 42% (18–71%) higher hazard of dementia among participants with FHx, and 30% (17–45%) among those without FHx (Fig. 3). For computer use, we did not observe significant multiplicative interaction with FHx (p = 0.517). However, all indices of the additive interaction were statistically significant with positive values: the relative excess risk due to interaction was 0.29 (0.06–0.53), and the attributable proportions due to interaction were 21% (6–36%) (online suppl. eTable 6). The HR of dementia in 0 h/day and >2 h/day of computer use compared with 1–2 h/day of computer use was 1.40 (1.30–1.50) and 1.19 (1.06–1.33) among those without FHx while the corresponding estimates increased to 1.46 (1.29–1.65) and 1.10 (0.90–1.36) in those with FHx, respectively (Fig. 3).

Fig. 3.

Association of TV viewing and computer use with incident dementia stratified by family history of dementia. The model was adjusted for age (continuous) (years), sex (male or female), ethnicity (White European, mixed, South Asian, Black, or other), UK Biobank assessment center, Townsend deprivation index (continuous), highest education level (lower secondary, upper secondary, higher, vocational, other, or missing), alcohol consumption (continuous) (grams), smoking status (current, former, never, or missing), BMI (continuous) (kg/m²), SBP (continuous) (mm Hg), physical activity (continuous) (metabolic equivalent of task [minutes]), healthy diet score (0, 1, 2, 3, 4, 5, or missing), use of antihypertensive drugs (yes, no, or missing), use of lipid-lowering drugs (yes, no, or missing), use of insulin (yes, no, or missing), FHx of dementia (yes, no, or missing), with mutual adjustment of the two exposure variables (TV viewing and computer use). p values for multiplicative interactions between FHx and TV viewing and between FHx and computer use were 0.208 and 0.040, respectively. Rates are per 100,000 person-years.

In the sensitivity analyses, we found that the multivariable-adjusted HRs for dementia events by categories of TV viewing, computer use, and FHx were appreciably unchanged with further adjustment for waist circumference, and nerves, anxiety, tension, or depression. Furthermore, the results were similar if we further adjusted for social activities, including frequency of friend/family visits, leisure/social activities, and frequency of confiding (online suppl. eTables 7–11). In the Fine‐Grey subdistribution hazard model, the results remained unchanged (online suppl. eTable 12). To prove that the screen-based sedentary behaviors may be risk factors of dementia instead of symptoms, we further excluded participants with less than 8 years of follow-up and obtained similar results (online suppl. eTable 13). While the association of TV viewing and computer use with dementia was modestly strengthened after excluding participants with chronic conditions or those with missing covariates (online suppl. eTables 14–15). In the age-stratified analysis, we observed that people with low computer use in midlife (<55 years old) have a weaker positive association with dementia relative to someone older (≥55 years old); while the effect of high TV viewing is stronger at younger age than older age (online suppl. eTable 16). In the occupation-stratified analysis, the results showed a stronger adverse effect of high TV viewing on dementia among participants who were employed than those unemployed, while the effect of low computer use was more pronounced among unemployed participants than that of employed participants (online suppl. eTable 17).

Discussion

In this large-scale prospective cohort study, we evaluated the independent associations between two screen-based sedentary activities, TV viewing, and computer use, and the risk of incident dementia. The results showed that TV viewing was associated with an elevated risk of dementia, whereas computer use and dementia had a J-shaped relationship. In addition, the adverse effects of computer use and FHx were additive for dementia risk assessment, providing new behavioral targets for intervention to prevent dementia, particularly for those with FHx.

Despite the hypothesis that excessive TV viewing could contribute to the development of dementia have been proposed nearly 30 years ago [30], the current evidence base is deemed to be too limited to draw any conclusions. Our results have validated the conclusion of a previous study in the UK biobank [15] and are in accordance with a case-control study (135 cases and 331 controls) which suggests that each hour increase in TV viewing in middle-adulthood corresponded to a 1.3 times greater risk of Alzheimer’s disease [13]. Another prospective study has investigated the incidence of dementia using the data from the São Paulo Aging and Health (SPAH) cohort, indicating that TV did not show any positive or negative effect on the risk of dementia [14]. A possible reason why our results differ from theirs is the differences in the populations studied. The SPAH study enrolled older (≥65 years) adults with less formal education (89% with 0–3 years of schooling) for whom watching TV was the main leisure activity, resulting in fewer opportunities to be exposed to highly cognitive stimulus that would compensate the passiveness of watching TV [14]. Furthermore, these findings of the SPAH study with relatively small sample size (N = 1,243) and short follow-up duration (2 years) highlighted the need for future research to further explore the relationship between TV viewing and dementia [14]. We examined the association between TV viewing and incidence of dementia among >400,000 middle-aged or older adults and found a nonlinear association between TV viewing time and the risk of dementia during a comparatively longer follow-up time, suggesting a recommend level of ≤3 h/day of TV viewing for the prevention of dementia. The direction of our estimate for TV viewing and dementia is in line with several prospective population-based cohorts showing that TV viewing may be related to adverse mental health and cognitive decline [10–12]. One possible explanation underlying the observations is that TV spectators are passively exposed to a mass of successive and rapid stimuli, a considerable part of which is stressogenic, and generally there is no opportunity for a subsequent release of tension [12]. Stress, via the glucocorticosteroids which it induces, has been shown to damage neurons in the hippocampus, a brain region involved in memory processes [31]. In addition, more TV viewing is associated with lower gray matter volume, a marker of brain aging, and thus increases the risk of dementia [32, 33].

Computer use is considered one of cognitively stimulating activities and has influence on cognitive function [34]. Despite growing evidence that cognitive, intellectual, or mentally stimulating activities could decrease risk of dementia [16, 35–39], few studies have focused on the effect of leisure computer use on long-term dementia risk [15]. Several studies have evaluated the association between computer use and incident mild cognitive impairment (MCI) [40, 41]. For example, the population-based Mayo Clinic Study of Aging (MCSA) (N = 1,929) showed that compared with using computer 2–3 times a month or less, using computer at least 1–2 times per week was associated with a 26% lower risk of incident MCI among participants aged ≥70 years [40]. In addition, another prospective study from the MCSA (N = 2,000) indicated that using computer was associated with a decreased risk of incident MCI regardless of timing (the HR was 0.52 in midlife, 0.70 in late life, 0.63 in midlife, and late life, respectively). Different from previous findings, we observed a J-shaped association between time spent on computer use and dementia, with the lowest HR observed at 2 h/day of computer use. Since our findings suggest that excessive TV viewing is an important risk factor for dementia while moderate computer use had a protective effect. Thus, the problems of sedentariness may not only be attributed to a lack of movement but also to the stimulation provided by the replacing activities. Moderate computer use would trigger neural circuits, build resilience, and protect the brain due to its nature of active cognitive stimulation [42]. Lack of daily computer use is associated with smaller brain volume in regions that are integral to memory function and known to be involved early with Alzheimer’s pathology and conversion to dementia [43]; while excessive computer use also has adverse impacts on circadian physiology, alertness, and cognitive performance levels due to the spectral profile of light emitted by computer screens [44]. Taken together, our results indicate that future educational and public health programs should include information on the adverse effect of excessive TV viewing on dementia risk and encourage adults to use computer moderately in their leisure time.

TV viewing and computer use are cognitive-related behavioral factors, while FHx loosely captures the causal contribution of genetics, diet, and lifestyle to risk of dementia. Thus, it is essential to investigate to what extent the risk of dementia associated with TV viewing and computer use can be modified by FHx. In a small prior cohort study in Minnesota (N = 1,929), the data pointed toward the lowest risk of incident MCI for Apolipoprotein E (APOE) ɛ4 noncarriers who engaged in computer activities and toward the highest risk of incident MCI for APOE ɛ4 carriers who did not engage in computer activities [40]. However, they only considered single susceptibility locus (i.e., APOE) instead of FHx, which might better reflect the overall genetic susceptibility to dementia. In the current study, we observed positive additive interactions of computer use with FHx of dementia in regard to the risk of incident dementia: the strength of this joint association was higher than what was observed with either risk factor alone. It is also worth noting that approximately one-fifth of dementia could be averted if limited computer use time to 1–2 h/day, even after accounting for FHx, assuming causality. The PAR% may be relatively larger in a population with FHx, despite the wide and overlapping CIs. Therefore, in terms of biological plausibility and from the public health perspective, our findings may be useful for informing dementia prevention strategies among asymptomatic subjects, highlighting that individuals with FHx should pay more attention on monitoring the time spent on computer use. Existing evidence has suggested potential mechanisms for the significant interactions. In addition to genetic factors, FHx was related to risk factors influencing the risk of dementia, such as physical and social activity, lipids, obesity, blood pressure, smoking status, and alcohol intake [45]. Notably, these traits were also associated with computer use. Therefore, we assumed that computer use and FHx might have additive effects on the risk of dementia through overlapped biological mechanisms related to these aforementioned traits to a certain extent. To date, no studies have investigated the interactions between TV viewing and computer use and FHx on dementia risk. Additional researches are warranted to validate our findings and further explore the extent to which individuals with FHx would benefit from adopting recommended levels of computer use in comparison with those without FHx.

Our study has several major strengths, including a large sample size and a wide range of information on demographic factors, lifestyles, and dietary patterns, which enabled us to rule out the influence of potential confounders. In addition, we performed several sensitivity analyses and found the results were largely unchanged, suggesting the robustness of our findings. Furthermore, we for the first time assessed the interactions between TV viewing and computer use and FHx on dementia risk. The novel finding on the additive interactions are highly relevant in the context of the current landscape of primary dementia prevention. However, several limitations of this study should also be noted. First, since the UK Biobank study was not designed to represent the general UK population, it must be cautious to extend our findings to a wider population in the UK. However, the multiethnic nature of our population does enhance the application to diverse populations. Second, the information on TV viewing, computer use, and FHx was self-reported by participants at baseline, which might have resulted in recall bias. Third, our analyses relied on single assessments of TV viewing and computer use at enrollment while failing to consider changes in these exposures over the follow-up period. Fourth, residual confounding might still exist despite our careful adjustment for these covariates at baseline in the analyses. Fifth, we could not identify additional cases of dementia through the collection of codes for dementia prescription drugs, which may lead to missed diagnoses. Sixth, we did not consider the number of family members with history of dementia, which may reflect a dose-response relationship. Finally, the observed associations could not represent causality. However, given that dementia has a long prodromal phase, we further excluded incident dementia cases that occurred in the first 8 years of follow-up in sensitivity analyses to limit the possibility of reverse causality.

In conclusion, TV viewing and computer use are independent risk factors for incident dementia, and the adverse effects of computer use and FHx were additive for dementia risk assessment. On the basis of the findings of this study, more specific public health guidance could be indicated differentiating between types of screen-based sedentary activities. Further research is recommended to confirm these results. Overall, the research adds to the growing body of evidence linking TV viewing and computer use, especially among those with FHx of dementia, to increased risk of an early onset of dementia event.

Statement of Ethics

The UKB study was approved by the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland and the North West Multicenter Research Ethics Committee. This research was conducted under the UK Biobank Application Number 44430. All participants gave written informed consent. This UKB study was also approved by the Ethical Committee of Peking University (Beijing, China) (IRB00001052-22186). Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Conflict of Interest Statement

The authors have no conflicts of interest to disclose.

Funding Sources

The study was supported by grants from the National Key R and D Program of China (2020YFC2003401) and High-performance Computing Platform of Peking University. The funders had no role in the study design, data collection, data analysis and interpretation, writing of the report, or the decision to submit the article for publication.

Author Contributions

Z.Z. and T.H. are the guarantors for the study, designed the research, wrote the paper, performed the data analysis, and had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to the statistical analysis, critically reviewed the manuscript during the writing process, and approved the final version to be published.

Funding Statement

The study was supported by grants from the National Key R and D Program of China (2020YFC2003401) and High-performance Computing Platform of Peking University. The funders had no role in the study design, data collection, data analysis and interpretation, writing of the report, or the decision to submit the article for publication.

Data Availability Statement

UKB data are available in a public, open access repository. This research has been conducted using the UKB Resource under Application Number 44430. The UKB data are available on application to the UK Biobank (www.ukbiobank.ac.uk/).