<p>S1 Fig. Flow diagram of analyses.</p>

Flow diagram of analyses. aSensitivity analysis was conducted in this population using multiple imputation to account for missing data on the exposure and covariates or removing low-eGFR participations. S2 Fig. Correlation matrix of kidney function markers. Hexbin plot of the relation between eGFRcr, eGFRcys, and eGFRdiff at baseline. (A) Correlation between eGFRcr and eGFRcys. (B) Correlation between eGFRcr and eGFRdiff. (C) Correlation between eGFRcys and eGFRdiff. S3 Fig. Survival curves and proportional hazards assessment. Kaplan-Meyer survival curves using time-scale and scatter plot of the scaled Schoenfeld residuals for eGFRdiff and all cause dementia(A), Alzheimer’s disease(B) and vascular dementia(C). S4 Fig. Nonlinear dose-response relationships. Dose-response relationship between eGFRratio (A to C) and All cause dementia, Alzheimer’s disease, or Vascular dementia. Restricted cubic spline was used to explore nonlinear associations, with three knots fixed at the quartiles for all smooth curves. Green line representing 95% Confidence interval. The HR was derived using Cox proportional hazard regression. Model were adjusted for eGFRcr_cys, Cho, LDL, education, smoking, drinking, physical activities, Townsend deprivation index (TDI), social isolate, hearing, eyesight, diabetes, hypertension, depression, and obesity. S5 Fig. Sensitivity analyses of eGFR measures. Associations between eGFRdiff or eGFRratio(z-score) and incident dementia. (A) Model 1 were adjusted for Cho, LDL, education, smoking, drinking, physical activities, Townsend deprivation index (TDI), social isolate, hearing, eyesight, diabetes, hypertension, depression, and obesity. (B) Model 2 were further adjusted for eGFRcr_cys. S6 Fig. Risk stratification by optimal eGFRdiff cut-off. Kaplan–Meier curves for incident all-cause dementia according to high- and low-risk groups defined by the optimal cut-off value of eGFRdiff (−8.813) derived from maximally selected rank statistics. S7 Fig. Incremental predictive value of eGFRdiff. Time-dependent ROC curves comparing discrimination performance of UKBDRS alone versus UKBDRS combined with eGFRdiff at 5, 10, and 15 years of follow-up. S1 Table. Association between eGFRdiff and dementia after multiple imputation. Hazard ratios (95% CI) for categorical and continuous eGFRdiff with all-cause dementia, Alzheimer’s disease, and vascular dementia following multiple imputation of missing covariates. S2 Table. Association between eGFRdiff and dementia excluding low eGFR participants. Sensitivity analysis showing hazard ratios for eGFRdiff and dementia outcomes after excluding participants with eGFR < 60 ml/min/1.73m². S3 Table. Association between eGFRratio and incident dementia. Hazard ratios for the association between eGFRratio (quartiles and continuous) with all-cause dementia, Alzheimer’s disease, and vascular dementia. S4 Table. Stratified analysis in participants with diabetes. Association between eGFRdiff (categorical and continuous) and incident dementia outcomes among participants with diabetes at baseline. S5 Table. Stratified analysis in participants with hypertension. Association between eGFRdiff (categorical and continuous) and incident dementia outcomes among participants with hypertension at baseline. S6 Table. Stratified analysis in participants with depression. Association between eGFRdiff (categorical and continuous) and incident dementia outcomes among participants with depression at baseline. S7 Table. Negative control analysis using traumatic injury. Association between eGFRdiff and incident traumatic injury (ICD-10 S00–T35, T66–78) as a negative control outcome to assess potential residual confounding. S1 Method. Genetic risk assessment methodology and Survival analysis and model discrimination methods. S1 Code. R Code for analyses of this study. (ZIP)分析流程图。本研究针对该队列开展敏感性分析：针对暴露因素与协变量的缺失数据采用多重插补（multiple imputation）法处理，或剔除估算肾小球滤过率（estimated glomerular filtration rate, eGFR）偏低的参与者。 S2图 肾功能标志物相关矩阵。基线时基于肌酐的估算肾小球滤过率（eGFRcr）、基于胱抑素C的估算肾小球滤过率（eGFRcys）与eGFR差值（eGFRdiff）之间关系的六边形分箱图。(A) eGFRcr与eGFRcys的相关性；(B) eGFRcr与eGFRdiff的相关性；(C) eGFRcys与eGFRdiff的相关性。 S3图 生存曲线与比例风险假设检验。采用时间尺度构建的卡普兰-迈耶（Kaplan-Meyer）生存曲线，以及针对eGFRdiff与全因痴呆(A)、阿尔茨海默病(B)、血管性痴呆(C)的标化斯科恩菲尔德残差散点图。 S4图 非线性剂量反应关系。eGFR比值（eGFRratio）(A~C)与全因痴呆、阿尔茨海默病或血管性痴呆之间的剂量反应关系。采用限制性立方样条（restricted cubic spline）探索非线性关联，所有平滑曲线的3个结点均固定于四分位数位点。绿色线条代表95%置信区间（confidence interval, CI）。风险比（hazard ratio, HR）通过Cox比例风险回归模型计算得到。模型校正了基于肌酐与胱抑素C的估算肾小球滤过率（eGFRcr_cys）、总胆固醇（Cho）、低密度脂蛋白胆固醇（low-density lipoprotein cholesterol, LDL）、受教育程度、吸烟、饮酒、体力活动、汤森剥夺指数（Townsend deprivation index, TDI）、社会隔离状态、听力情况、视力情况、糖尿病、高血压、抑郁及肥胖情况。 S5图 eGFR指标敏感性分析。eGFRdiff或标准化z评分的eGFRratio与新发痴呆的关联。(A) 模型1校正了Cho、LDL、受教育程度、吸烟、饮酒、体力活动、TDI、社会隔离状态、听力情况、视力情况、糖尿病、高血压、抑郁及肥胖情况；(B) 模型2在此基础上进一步校正了eGFRcr_cys。 S6图 基于最优eGFRdiff截断值的风险分层。基于通过最大选择秩和统计得到的eGFRdiff最优截断值(-8.813)划分的高、低风险组，其新发全因痴呆的卡普兰-迈耶生存曲线。 S7图 eGFRdiff的增量预测价值。随访5年、10年、15年时，仅使用UKBDRS与联合使用UKBDRS及eGFRdiff的区分度比较的时变受试者工作特征（Receiver Operating Characteristic, ROC）曲线。 S1表 多重插补后eGFRdiff与痴呆的关联。在对缺失协变量进行多重插补后，分类变量与连续变量形式的eGFRdiff与全因痴呆、阿尔茨海默病及血管性痴呆的风险比（95% CI）。 S2表 剔除低eGFR参与者后eGFRdiff与痴呆的关联。本敏感性分析展示了剔除eGFR<60 ml/min/1.73m²的参与者后，eGFRdiff与各痴呆结局的风险比。 S3表 eGFRratio与新发痴呆的关联。四分位与连续变量形式的eGFRratio与全因痴呆、阿尔茨海默病及血管性痴呆的关联的风险比。 S4表 糖尿病亚组分析。基线时合并糖尿病的参与者中，分类与连续变量形式的eGFRdiff与新发痴呆结局的关联。 S5表 高血压亚组分析。基线时合并高血压的参与者中，分类与连续变量形式的eGFRdiff与新发痴呆结局的关联。 S6表 抑郁亚组分析。基线时合并抑郁的参与者中，分类与连续变量形式的eGFRdiff与新发痴呆结局的关联。 S7表 创伤性阴性对照分析。以创伤性损伤（ICD-10编码S00–T35、T66–78）作为阴性对照结局，评估潜在残余混杂的eGFRdiff与新发创伤性损伤的关联。 S1方法 遗传风险评估方法、生存分析及模型区分度评估方法。 S1代码 本研究分析所用的R代码（压缩包ZIP格式）