Supplementary Material for: Leisure-time television viewing and computer use, family history, and incidence of dementia

Background & Objectives: Time spent on screen-based sedentary activities are 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 [CI] 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 hour/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 (0h/day; 1.41, 1.33-1.50) and high (>2h/day; 1.17, 1.05-1.29) computer use were associated with elevated dementia risk, compared with moderate (1-2h/day) computer use. We observed a J-shaped relationship with dementia (relative to 2 hour/day; P for nonlinear < 0.001). Compared with 1-2h/day of computer use, the HRs of dementia were 1.46 (1.29-1.65) and 1.10 (0.90-1.36) for 0h/day and >2h/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 adverse effects of time spent on TV viewing or computer use and FHx were independent on dementia risk. Our findings point to new behavioral targets for intervention on preventing an early onset of dementia, especially for those with FHx.

研究背景与研究目的：屏幕久坐行为时长与痴呆风险显著相关，但目前尚不清楚该关联是否因痴呆家族史（family history, FHx）而异。本研究旨在探讨两种常见屏幕久坐行为——电视（television, TV）观看与电脑使用——与痴呆风险的独立关联，并评估痴呆家族史的调节效应。
研究方法：本研究纳入英国生物样本库（UK Biobank）中415048名无痴呆病史的参与者。采用Cox回归模型分析电视观看时长、电脑使用时长以及痴呆家族史与痴呆风险的关联，并分别估算电视观看、电脑使用与痴呆家族史在相乘尺度和相加尺度上的交互作用。
研究结果：中位随访12.6年期间，共有5549名参与者罹患痴呆。校正潜在混杂因素后，与低量电视观看（0~1小时/天）相比，中量（2~3小时/天；风险比（hazard ratio, HR）=1.13，95%置信区间（confidence interval, CI）=0.03~1.23）与高量（>3小时/天；HR=1.33，95%CI=1.21~1.46）的电视观看均与更高的痴呆风险相关。通过限制性立方样条模型分析发现，电视观看时长与痴呆风险的关联呈非线性（以0小时/天为参照；非线性检验P=0.005）。在携带痴呆家族史的参与者中，每日电视观看时长>3小时与痴呆风险升高42%（HR=1.42，95%CI=1.18~1.71）相关，而在无痴呆家族史的参与者中，该关联为风险升高30%（HR=1.30，95%CI=1.17~1.45）。
对于电脑使用而言，与中量使用（1~2小时/天）相比，低量使用（0小时/天；HR=1.41，95%CI=1.33~1.50）与高量使用（>2小时/天；HR=1.17，95%CI=1.05~1.29）均与痴呆风险升高相关。限制性立方样条分析显示电脑使用时长与痴呆风险呈J型关联（以2小时/天为参照；非线性检验P<0.001）。在携带痴呆家族史的参与者中，与每日1~2小时电脑使用相比，0小时/天与>2小时/天的电脑使用对应的痴呆HR分别为1.46（95%CI=1.29~1.65）与1.10（95%CI=0.90~1.36）；而在无痴呆家族史的参与者中，对应的HR分别为1.40（95%CI=1.30~1.50）与1.19（95%CI=1.06~1.33）。本研究发现电脑使用与痴呆家族史存在正向相加尺度交互作用（交互作用相对超额危险度（relative excess risk of interaction, RERI）=0.29，95%CI=0.06~0.53），而未观察到电视观看与痴呆家族史存在明显交互作用。
研究结论：电视观看或电脑使用的久坐时长与痴呆家族史对痴呆风险的不良影响相互独立。本研究结果为痴呆早期预防的行为干预提供了新的靶向目标，尤其针对携带痴呆家族史的人群。