厦门医学论文发表-痴呆症患者跌倒和其他严重不良事件与抗高血压药物之间的关联：一项观察性队列研究

藤原武，康斯坦丁诺斯·科希亚里斯，蔡婷，王丽娟，李约瑟夫，莎拉·莱-弗鲁里，阿米塔瓦·班纳吉，安德鲁·克莱格，痴呆症患者跌倒和其他严重不良事件与抗高血压药物之间的关联：一项观察性队列研究

藤原武，康斯坦丁诺斯·科希亚里斯，蔡婷，王丽娟，李约瑟夫，莎拉·莱-弗鲁里，阿米塔瓦·班纳吉，安德鲁·克莱格，

抽象

背景

与降低痴呆症患者血压水平相关的益处和风险平衡仍然存在争议，因为缺乏证据表明与抗高血压治疗相关的可能危害。我们检查了痴呆症患者与非痴呆症患者相比，抗高血压药物与严重不良事件之间的关联。

方法和发现

这是一项回顾性分析，使用了 1998 年至 2018 年间具有全国代表性的英国全科医学人群，来自电子健康记录（临床实践研究数据链、CPRD、GOLD）。符合条件的个体年龄为 ≥40 岁，收缩压为 130-179 mmHg，并且之前未接受过抗高血压治疗。痴呆症的诊断基于电子健康记录中的临床代码。如果个体在 12 个月的暴露期内至少服用了一种抗高血压药物，则他们被分配到暴露组。暴露期间未开具任何抗高血压药物的患者被分配到对照组。主要结局是随访期 10 年内首次住院或跌倒死亡。次要结局是首次住院或因低血压、晕厥和骨折而死亡。在 1,219,732 人的人口中，有 23,510 人患有痴呆症。在12个月的暴露期内，4,062/23,510名（17.3%）痴呆患者和142,385/1,196,222名（11.9%）无痴呆症患者新开具了抗高血压药物。在调整倾向评分和既往相关结局史的主要分析中，降压治疗与住院或跌倒死亡风险略有增加相关（调整后风险比 [aHR] 1.15,95% 置信区间 [CI] 1.08， 1.22）、低血压（aHR 1.51,95%CI 1.29， 1.78）、晕厥（aHR 1.34,95%CI 1.11， 1.61），但不是骨折（aHR 1.05,95%CI 0.96,1.15），在痴呆患者中。这些发现在不同的分析方法中是一致的，包括多变量调整、倾向评分匹配和逆概率治疗加权。在没有痴呆的个体中，抗高血压治疗与严重不良事件之间的关联相似，住院或跌倒死亡的风险略有增加（aHR 1.07,95%CI 1.05,1.10）。然而，与痴呆患者（每年每10,000人中有47例，95%CI 26,70）相比，治疗相关的绝对跌倒风险显著高于无降压治疗的患者（每年每10,000人中有14例，95%CI 10， 18）。与未接受抗高血压治疗的个痴呆体相比，接受抗高血压治疗的个体发生低血压和晕厥的绝对风险也更高。主要局限性是基于电子健康记录中编码条目的痴呆诊断中可能存在无法测量的混杂和异质性。

结论

抗高血压治疗与有和没有痴呆的个体发生严重不良事件的风险增加有关，然而，痴呆个体的绝对伤害风险增加了一倍多。这些数据表明，临床医生、患者及其护理人员在开始新的抗高血压药物之前应考虑这些风险，尤其是在痴呆的情况下。

作者总结

为什么要进行这项研究？

抗高血压药物已被证明可以降低所有年龄组的心血管疾病发病率和死亡率。

先前的研究表明，抗高血压治疗与严重不良事件的风险增加有关，尤其是在老年人和严重虚弱的人中。

然而，缺乏关于抗高血压治疗对痴呆患者有害的确凿证据。

研究人员做了什么并发现了什么？

这是一项回顾性分析，使用了 1998 年至 2018 年间具有全国代表性的英国全科人群，来自电子健康记录。

在总共 1,219,732 名患者（23,510 名痴呆患者）中，抗高血压治疗与患有和不患有痴呆症的个体因跌倒、低血压和晕厥而住院或死亡的风险增加有关。

痴呆症患者接受抗高血压治疗的绝对伤害风险显着更高。

这些发现意味着什么？

这些发现表明，临床医生、患者及其护理人员在为痴呆患者开始使用新的降压药物时应考虑严重不良事件的风险。

局限性在于基于电子健康记录中的编码诊断的痴呆诊断中可能存在无法测量的混杂和异质性。

数字

图2Table 3Table 4Table 1Fig 1Table 2图2Table 3Table 4Table 1Fig 1Table 2

引文： Fujiwara T， Koshiaris C， Cai T， Wang A， Lee J， Lay-Flurrie S， et al. （2025） 痴呆症患者跌倒和其他严重不良事件与抗高血压药物之间的关联：一项观察性队列研究。公共科学图书馆医学 22（9）： 电子邮件 1004731。 https://doi.org/10.1371/journal.pmed.1004731

学术编辑： Carol Brayne，剑桥大学，大不列颠及北爱尔兰联合王国

收到： 2025 年 1 月 6 日;接受： 2025 年 8 月 22 日;发表： 9月 17， 2025

版权所有： © 2025 Fujiwara et al.这是一篇根据知识共享署名许可条款分发的开放获取文章，允许在任何媒体上不受限制地使用、分发和复制，前提是注明原作者和来源。

数据可用性： 数据是通过 CPRD 机构许可证获得的。由于数据控制者（临床实践研究数据链 [CPRD]）对数据的使用施加了限制，作者无法公开共享数据。这些要求需要访问数据的个人在 CPRD 独立科学咨询委员会批准的协议中指定。该协议必须作为数据控制者 （CPRD） 和数据处理者（作者）之间数据共享协议的一部分包含在内。数据可从 CPRD（通过 enquiries@cprd.com 联系）获得，供符合访问这些机密数据标准的研究人员使用。本分析中使用的医院事件统计数据经 NHS Digital（通过 enquiries@nhsdigital.nhs.uk 联系）许可重复使用，NHS Digital 保留该数据的版权。此分析中使用的变量的完整代码列表可以在 https://github.com/jamessheppard48/STRATIFY-BP/tree/Causal-inference-project 中找到。

资金： TF 获得了 SENSHIN 医学研究基金会的资助。JPS 和 CK 通过亨利·戴尔爵士奖学金（参考：211182/Z/18/Z）获得惠康信托基金/皇家学会的资助。这项研究部分由惠康信托基金资助（参考：211182/Z/18/Z）。FDRH 得到了 NIHR 应用研究合作 （ARC） 牛津泰晤士河谷的部分支持。RJM 得到了 NIHR 牛津和泰晤士河谷应用研究联盟的支持。资助者在研究设计、数据收集和分析、发表决定或手稿准备方面没有任何作用。

利益争夺： 作者表示，他们没有潜在的利益冲突。

缩写：aHR、 调整后的风险比;BMI， 体重指数;BP， 血压;词 置信区间;CPRD， 临床实践研究数据链;CVD， 心血管疾病;高密度脂蛋白， 高密度脂蛋白;RR， 相对风险;贴片， 标准化均值差

介绍

药物治疗降低血压（BP）可降低心血管疾病（cardiovascular disease， CVD）的发病率和死亡率[1,2]。所有年龄组的治疗获益基本一致[3]。事实上，在过去10年中，严格控制血压对降低活跃老年人CVD事件风险的有益作用已被证明[4,5]。最近的证据表明，虚弱患者仍可能从降压治疗中获益[6,7]，一项系统评价表明，即使是虚弱的老年人也可能从降压中获益[8]。治疗老年人的高血压是初级保健医生的一项重要任务。然而，应谨慎行事，因为老年人发生严重不良事件的风险增加，尤其是衰弱进展的患者[9]。最近的高血压指南建议，对老年人必须仔细评估降压治疗的益处和危害平衡[10\u201212]，尽管这些建议主要基于专家意见和低级别证据，因为在老年人中进行降压治疗的试验很少，尤其是虚弱和痴呆患者。

痴呆症的全球健康负担越来越大，预计到2050年，全球受影响的人数将从2019年的5700万增加到1.53亿以上，增加近两倍[13]。高血压是痴呆的可改变危险因素之一[14]，在痴呆患者中很常见[15]，但对于降低痴呆患者血压水平的获益和风险平衡一直存在争议。最近一项随机对照试验的证据表明，强化降压可有效降低高血压患者发生全因性痴呆的风险，这表明抗高血压治疗具有潜在的预防作用[16]。一些研究表明，降压药物可增加痴呆患者的脑血流量[17,18]，而另一项观察性研究表明，在接受抗高血压药物治疗的患者中，日间收缩压低与轻度认知障碍或痴呆患者的认知能力下降进展更大相关[19]。痴呆已被证明与CVD事件风险增加有关[20,21]，但迄今为止尚无直接证据表明降压药物可以降低痴呆患者的CVD风险。痴呆患者使用降压药可能会增加发生严重不良事件的风险，因为存在多种疾病和多重用药[22,23]。此外，痴呆患者更容易身体虚弱[24]。一项离散选择实验表明，临床医生倾向于改变或停用不必要的降压药物，以降低痴呆患者跌倒的风险[25]。然而，迄今为止，缺乏关于抗高血压治疗对痴呆症患者可能造成的危害的经验证据。

因此，我们着手确定与非抗高血压患者相比，抗高血压治疗是否与痴呆患者发生严重不良事件的风险增加有关。

方法

本研究是使用观察性常规收集数据 （RECORD） 指南（S1 表）进行的研究的转载进行的。临床实践研究数据链 （CPRD） 已获得全球伦理批准，可用于研究目的使用匿名电子健康记录，但须经其独立科学咨询委员会批准研究方案。本研究的方案于 2019 年 2 月获得前瞻性批准（ISAC 方案编号 19_042），并在 S1 方案中提供。

学习环境

这是一项回顾性观察性队列研究。我们使用了CPRD GOLD中保存的英国初级保健电子健康记录数据（基于使用视力保健电子健康记录软件的实践数据;Meddbase 软件——医疗管理系统，英国伦敦）。CPRD GOLD覆盖了来自英国968家诊所的2000多万个人，所纳入的人群在年龄、性别和种族方面广泛代表了英国人口[26]。

研究人群

符合条件的个人是：（1） 年龄在 ≥40 岁;（2） 暴露期前符合条件的第一收缩压水平在 130 至 179 mmHg [12] 之间;（3） 在研究开始日期之前未接受过任何抗高血压药物;及 （4） 在 1998 年 1 月 1 日至 2018 年 12 月 31 日期间在 CPRD GOLD 注册的人。暴露期和随访期是相对于每个人的队列进入日期定义的，1 年的暴露评估期，然后是长达 10 年的随访期。该设计反映了跨年龄组成年人口的代表性样本，并侧重于抗高血压治疗的新用户，从而可以更清晰地评估治疗相关风险。在个体之间统一应用固定的随访时间，以确保暴露组之间绝对风险估计的可比性。排除标准是：（1）没有血压测量记录;（2）符合条件的收缩压≥180mmHg，因为血压高于此水平的患者被认为需要服用降压治疗处方，无论严重不良事件风险如何[12]。个人的特征是根据随访期开始前任何时间点记录的信息确定的。个体在研究结束日期退出研究，当时他们从注册的 CPRD 实践中转出、死亡或经历感兴趣的特定结果（S1 图）。

痴呆的定义

痴呆症的诊断基于痴呆症的临床代码。我们使用所有病理分类定义全因痴呆，包括阿尔茨海默病、血管性痴呆和其他类型的痴呆（根据 S2 表列出的 ICD-9 和 ICD-10 代码定义）。不包括患有轻度认知障碍的个体，因为我们无法在数据集中一致地识别出患有轻度认知障碍的个体。

暴露

暴露是英国国家处方集（S3 表）中定义的任何抗高血压药物的处方。如果个人在随访开始日期之前的 12 个月内至少服用了一种抗高血压药物，则他们被分配到暴露组。基线时的药物由该开始日期之前的最新处方定义。在此期间没有开出任何抗高血压药物的人被分配到对照组。

结果

主要结局是随访后 10 年内首次因跌倒住院或死亡。次要结局是首次住院或因低血压、晕厥或骨折而死亡。结果是从 ICD-9 和 ICD-10 代码中捕获的，作为基本住院病发统计中的主要入院原因和国家统计局死亡证明上的主要死因。

缺失变量的多重插补

本研究中缺失变量的百分比如下：体重指数 （BMI） 15.2%、种族 62.7%、吸烟状况 4.8%、饮酒 14.8%、剥夺 0.08%、总胆固醇 53.9% 和高密度脂蛋白 （HDL） 胆固醇 65.5%。1,219,732 例病例中总共有 110,666 例 （9.1%） 是完全病例。使用具有链式方程的多个插补，使用 R 中的“小鼠”包来插补缺失数据。我们为患有和不患有痴呆症的个体组创建了10个插补数据集，并分别分析了每个插补数据集中的治疗效果[27]。这些估计值及其标准误差使用鲁宾规则进行组合[28]。每个分析的感兴趣结果都包含在单独的插补模型中。我们还对完整病例的子集进行了敏感性分析。

协变量

协变量包括人口统计学（年龄、性别、种族和多重剥夺指数）、临床特征（收缩压和舒张压、BMI、总胆固醇、高密度脂蛋白胆固醇、吸烟状况和饮酒量）、既往病史（中风、短暂性脑缺血发作、心肌梗塞、心力衰竭、外周血管疾病、冠状动脉旁路移植术、心绞痛、慢性肾病、糖尿病、心房颤动和癌症）、其他处方药（他汀类药物、 抗血栓药[包括抗血小板药和抗凝药]、阿片类药物、催眠药/抗焦虑药、抗抑郁药和抗胆碱能药物）、衰弱（使用经验证的电子衰弱指数，合并36项缺陷）[29]、原发性CVD风险（基于QRisk2评分，QRisk2评分不可用，则替换为总体均值）[30]，以及既往有相关结局的病史。这些协变量是根据临床治疗指南和专家意见先验选择的[12]。

使用 2015 年英国多重剥夺指数 （https://www.gov.uk/government/statistics/english-indices-of-deprivation-2015） 评估剥夺。2015年多重剥夺指数是衡量相对剥夺的官方指标，全面概述了英格兰各地区的社会经济剥夺情况[31]。在本研究中，多重剥夺指数分数分为五分位数，多重剥夺指数分数为 5 表示处于最高五分位数（最被剥夺）的个体。

倾向评分生成

为了解释两组之间基线特征的差异，我们分别对患有和不患有痴呆症的个体进行了倾向评分分析。在这两组中的每一组中，我们使用多变量逻辑回归在每个插补数据集中生成倾向评分，以抗高血压药物处方作为结果，并包括所有基线协变量（列在上一节以及 S4 和 S5 表中）。对连续变量进行分类以解释与结果的非线性关联（探索了样条曲线和分数多项式的使用，但导致了模型收敛问题）。对于匹配分析，使用1：1的最近邻匹配方法，卡尺限制为0.2。估计匹配前后所有基线协变量的标准化平均差 （SMD），以评估匹配前和匹配后的协变量平衡。我们认为协变量的SMD为<0.1，这是可接受的平衡指标[32]。

主要分析

严重不良事件的累积发生率采用Kaplan-Meier法估计。对于主要分析，在 Cox 回归模型中调整了倾向评分和感兴趣结果的既往史，以检查抗高血压治疗与严重不良事件之间的关联。对于二次分析，我们使用了：（1）针对倾向评分模型中包含的相同变量进行了调整的Cox回归模型;（2）倾向评分匹配多因素调整Cox回归模型;（3）逆概率处理加权模型。这三个模型还包括感兴趣结果的先前历史记录作为调整变量。逆概率治疗权重计算为暴露组个体倾向评分的倒数和对照组个体的倾向评分的倒数（1—倾向评分）[33,34]。应用逆概率治疗加权模型生成加权队列[35]。这四种方法有望提供相似的结果，并用于证明结果的稳健性。所有四个模型分别针对患有和不患有痴呆症的个体进行拟合，以估计痴呆症抗高血压治疗与每种结局之间的特异性关联。检查舍恩菲尔德残差以检查比例风险假设。绝对风险差异定义为假设接受抗高血压治疗的人群中每种结局的风险减去假设未接受抗高血压治疗的每种结局的风险，使用来自Cox回归模型的治疗效果估计。该风险差异估计并报告为每10,000例患者每年的事件数[36]，每个插补数据集使用200次引导迭代获得95%置信区间（confidence interval， CI）。需要伤害的数量是根据绝对风险差计算的。

Subgroup and sensitivity analyses

Serious adverse event risks by number of antihypertensive drugs used were calculated for each outcome, by dementia status. The number of antihypertensive drugs were categorised into 0, 1, 2, or ≥3, and serious adverse event risks were calculated for each outcome in both groups using the number of antihypertensive drugs used zero as the reference category, using propensity score adjustment to control for confounding.

To assess whether there was a linear trend in the risk of adverse events across increasing numbers of antihypertensive medications, we additionally conducted a trend analysis by modelling the number of antihypertensive drugs as an ordered categorical variable in the Cox models. This approach allowed us to formally test for a linear increase in hazard with increasing medication count, after adjusting for imputed propensity scores and prior history of the outcome. Trend tests were conducted separately in individuals with and without dementia using pooled estimates from multiply imputed datasets.

During the study period, the use of diagnostic codes for dementia increased due to two factors: (1) the introduction of the NHS England national enhanced service for dementia diagnosis in 2013/14, and (2) the introduction of a dementia register in the Quality and Outcomes Framework incentivization scheme introduced in 2014/15. These factors may have affected the specificity of clinical codes for dementia, and the association between the risk of serious adverse events with antihypertensive medication and dementia status in this study. Therefore, sensitivity analyses were performed to assess whether the association between the risk of serious adverse events with antihypertensive medication and dementia status differed depending on whether the follow-up period started before or after April 1, 2014.

Results

Population characteristics

From a total of 16,071,111 registered individuals, 1,219,732 fulfilled the eligibility criteria (S2 Fig). Among them, 23,510 (1.9%) had dementia. In the 12-month exposure period, 4,062 (17.3%) of the 23,510 individuals with dementia and 142,385 (11.9%) of the 1,196,222 individuals without dementia started at least one antihypertensive medication, respectively, and were included in the exposure groups. Before propensity score matching, there were differences in almost all baseline characteristics between individuals in the exposure group and those in the control group in both individuals with dementia and those without (Table 1). One-to-one matching by propensity score analysis resulted in 3,795 matched individuals with dementia and 131,199 matched those without between the exposure and control groups, respectively. After matching, the SMD for all variables included in propensity score was reduced to below 0.1, indicating effective matching (Tables 1 and S6).

thumbnailDownload:

PPTPowerPoint slide

PNGlarger image

TIFForiginal image

Table 1. Baseline characteristics of the study population before and after propensity score matching, with standardised mean differences between exposure and control groups.

https://doi.org/10.1371/journal.pmed.1004731.t001

To further assess the appropriateness of propensity score methods, we examined the distribution of propensity scores in individuals with and without dementia (S3 Fig).

Primary outcome

During a median follow-up of 6.3 years (interquartile range 2.7, 10.0) years, a total of 8,314 individuals with dementia (35.4%) and 87,785 those without (7.3%) were hospitalised or died following a fall. The cumulative incidence of falls was higher in individuals with dementia compared to those without, and in the exposure group compared to the control group (Fig 1A). In the primary analyses using propensity score adjustment, antihypertensive treatment was associated with an increased risk of hospitalisation or death from falls in both individuals with dementia (adjusted hazard ratio [aHR] 1.15, 95% CI 1.08, 1.22) and in those without (aHR 1.07, 95%CI 1.05, 1.10) (Table 2). Analyses using multivariable adjustment, propensity score matching, and inverse probability treatment weighting showed similar results. While the interaction was statistically significant in three of the four analytical approaches, it was not significant in the inverse probability treatment weighting model. The absolute risk difference of falls with antihypertensive treatment was significantly higher in individuals with dementia (absolute risk difference 47 per 10,000 individuals per year, 95%CI 26, 70, equivalent to a number of needed to harm of 213 per year) compared to those without (absolute risk difference 14 per 10,000 individuals per year, 95%CI 10, 18, equivalent to a number of needed to harm of 714 per year) (Fig 2). The results were similar when analyses were undertaken using complete case only (S7 Table; S4 and S5 Figs).

thumbnailDownload:

PPTPowerPoint slide

PNGlarger image

TIFForiginal image

Fig 1. Cumulative risk of serious adverse events by antihypertensive medication according to categories of dementia and drug exposure status.

显示了四个分类组严重不良事件累积发生率的 Kaplan-Meier 曲线。（A、B、C 和 D）分别显示跌倒、低血压、晕厥和骨折的累积发生率。每条实线表示每个严重不良事件的累积发生率，其周围区域表示 95% 置信区间。对于图中报告的所有 p 值，使用的统计检验是对数秩检验。

https://doi.org/10.1371/journal.pmed.1004731.g001

thumbnail下载：

PPT的PowerPoint 幻灯片

巴布亚新几内亚大图

蒂夫原图

表 2. 插补数据集中抗高血压治疗的住院或跌倒死亡风险与痴呆状态之间的关联。

https://doi.org/10.1371/journal.pmed.1004731.t002

缩略图下载：

PPT的PowerPoint 幻灯片

巴布亚新几内亚大图

蒂夫原图

图2. 与抗高血压治疗相关的严重不良事件风险的差异（按痴呆状态划分）。

根据倾向得分调整的模型。绝对风险差异被描述为每年每 10,000 名患者的额外事件。缩写：CI，置信区间;HR，风险比。

https://doi.org/10.1371/journal.pmed.1004731.g002

次要结果

在痴呆患者中，共有 1,005 人 （4.3%） 出现严重低血压，860 人 （3.7%） 出现晕厥，4,125 人 （17.5%） 出现骨折，每种情况都需要住院或与死亡有关。在没有痴呆症的患者中，9,634 人 （0.8%） 出现严重低血压，13,766 人 （1.2%） 出现晕厥，73,155 人 （6.1%） 出现骨折。这些严重不良事件的累积发生率在痴呆患者和暴露组中更高（图1B-1D）。在痴呆患者和非骨折患者中，抗高血压治疗与低血压和晕厥住院或死亡风险增加相关，但无骨折：在痴呆患者中，低血压（aHR 1.51,95%CI 1.29,1.78）、晕厥（aHR 1.34,95%CI 1.11,1.61）和骨折（aHR 1.05,95%CI 0.96,1.15）;在没有痴呆、低血压（aHR 1.70,95%CI 1.60,1.79）、晕厥（aHR 1.25,95%CI 1.18,1.31）和骨折（aHR 1.00,95%CI 0.98,1.02）的个体中（表 3）。与没有痴呆的人相比，患有的个体低血压和晕厥的绝对风险差异更高（图 2）。仅使用完整病例进行分析时，也观察到了类似的结果（S7 表;S4 和 S5 图）。

thumbnail下载：

PPT的PowerPoint 幻灯片

巴布亚新几内亚大图

蒂夫原图

表 3. 插补数据集中每种结果开始使用抗高血压药物的风险比。

https://doi.org/10.1371/journal.pmed.1004731.t003

亚组分析

在患有和不患有痴呆症的个体中，跌倒和低血压的风险随着降压药物数量的增加而增加（表 4）。无论痴呆诊断如何，骨折风险与抗高血压药物的数量之间没有显着关联。仅使用完整案例进行分析时，结果相似（S8 表）。

thumbnail下载：

PPT的PowerPoint 幻灯片

巴布亚新几内亚大图

蒂夫原图

Table 4. Hazard ratios by the number of antihypertensive medications use for each outcome in imputation datasets.

https://doi.org/10.1371/journal.pmed.1004731.t004

Sensitivity analyses

Of all study individuals, 1,107,593 (90.8%) started their follow-up periods before April 1, 2014, while 112,139 (9.2%) started them after that date. Among these, 23,206 (2.1%) and 304 (0.27%) were diagnosed with dementia, respectively.

Among those who started their follow-up periods after April 1, 2014, 77 (25.3%) with dementia and 1,647 (1.5%) without dementia experienced hospitalisation or died following a fall. Antihypertensive treatment was associated with an increased risk of hospitalisation or death from a fall in individuals without dementia (aHR 1.25, 95%CI 1.08, 1.46), but not in those with dementia (aHR 1.15, 95%CI 0.59, 2.27) (S9 Table).

Discussion

In this nationwide electronic health record data-based observational study of 1.2 million previously untreated individuals, antihypertensive treatment was associated with an increased risk of hospitalisation or death from falls, hypotension, and syncope in both individuals with and without dementia, with comparable relative risks (RRs) observed between the two groups. However, the absolute risk of harm with antihypertensive treatment was significantly higher in individuals with dementia than in those without dementia. These findings indicate that the absolute risk of serious adverse events with antihypertensive drugs differs by dementia status, and that careful consideration of the priorities of individuals with dementia is necessary when starting new antihypertensive medications in this group.

There is inconsistent evidence regarding the association of serious adverse events with antihypertensive therapy in dementia. In an observational study of 160 individuals living with dementia in group dwellings (mean age 84 ± 7 years, 80% female), antihypertensive medication use was not a significant risk factor for falls [37]. On the other hand, in the Hypertension in Dementia study of 180 individuals with diagnosed hypertension and dementia (mean age 82 ± 6 years, 70% female, 87% living their own home), a total of 214 falls (a rate of 2,760 falls per 1,000 patient-years) were observed during the 6-month follow-up, in individuals prescribed a median of one antihypertensive medication [38]. These data could not determine the association between serious adverse events by antihypertensive therapy and dementia because of very limited sample size, and lack of a non-treated control group. In contrast, our analyses examined a large population from general practice.

There are some possible mechanisms by which antihypertensive treatment could have adverse effects in individuals with dementia, such as orthostatic BP changes [39], accumulation of anticholinergic effects from antihypertensive drugs [40], drug-induced delirium [41], and drug interaction [42]. Frailty, social, environmental, and cognitive factors could also explain the higher absolute risk of falls in individuals with dementia [43–45]. Furthermore, our findings suggest that dementia itself may amplify the risk of serious adverse events associated with antihypertensive treatment. In our cohort, individuals with dementia had a markedly higher absolute risk of falls compared to those without dementia, despite similar RRs, supporting the notion that dementia-related vulnerability may enhance the harms of treatment.

There are very limited data regarding the benefits and harms of antihypertensive medications use in individuals with dementia. In a double-blind randomised controlled trial, the calcium channel blocker, nilvadipine, increased cerebral blood flow in individuals with mild-to-moderate Alzheimer’s disease [17]. Additionally, another open-label randomised controlled trial demonstrated that in older hypertensive individuals with Alzheimer’s disease, the angiotensin Ⅱ receptor blockers, telmisartan, increased regional cerebral blood flow, including the superior parietal lobe which was the most severely affected region in Alzheimer’s disease, compared to the calcium channel blocker, amlodipine [18]. One randomised controlled trial investigated the effects of the calcium channel blocker, nimodipine, in individuals with subcortical vascular dementia [46]. This study showed placebo was associated with increased risks of cerebrovascular disease (RR 2.48, 95%CI 1.23, 4.98) and CVD (RR 2.26, 95% CI 1.11, 4.60) compared with nimodipine treatment group. However, the favourable effects of nimodipine did not depend on the BP-lowering effect of nimodipine, and the original aim of this study was not to test the efficacy of nimodipine in the secondary prevention of vascular diseases. Furthermore, a recent randomised controlled trial among long-term care residents with moderate-to-severe dementia suggested that discontinuation of antihypertensive treatment tended to increase the risk of serious adverse events, although this trial was not powered on clinical outcome events and had to be stopped early, precluding any definitive conclusions from being drawn [47]. The potential benefit of BP-lowering effects may be modified by the degree of cognitive decline. In the Longitudinal Ageing Study Amsterdam, lower diastolic BP was associated with higher all-cause mortality risk in individuals with cognitive dysfunction [48]. In a cohort study of individuals with diagnosed mild cognitive impairment or dementia, lower daytime systolic BP was associated with greater progression of cognitive decline among individuals with antihypertensive drug therapy [19].

While our previous observational study, which included the same individuals as the present analysis, demonstrated a similar association between antihypertensive treatment and serious adverse events [9], a meta-analysis of randomised controlled trials reported no significant association between antihypertensive treatment and fall risk [49]. This discrepancy may be partly explained by differences in study population, particularly the exclusion of frailer individuals and people with dementia from many clinical trials, in contrast to our real-world cohort. It may also reflect the influence of residual confounding, such as confounding by indication, which is more likely to affect observational data [50].

Recent guidelines for treatment of hypertension recommend an individualised approach for people with dementia [10–12]. Some antihypertensive drugs for individuals with dementia may have beneficial effects for cerebral blood flow [17,18], but there is no definitive evidence to date for benefit from antihypertensive therapy on CVD outcomes in individuals with dementia [51,52], because most trial participants were relatively fit, people with dementia were excluded by design [4,5,53]. Our findings suggest that, in line with previous literature [54–57], the cumulative event rates of falls, hypotension, syncope, and fractures and absolute risk differences of them were higher in individuals with dementia than in those without. Various factors, such as advanced age [58,59], sleep-related disorders [60,61], depression [62], autonomic dysfunction [63], socioeconomic factors [43], and quality of care could affect the association between the increase of harmful events and dementia [64]. Our current findings extend these previous findings and suggest that new prescriptions of antihypertensive drugs might increase the risk of harmful events in individuals with dementia. Furthermore, some individuals with dementia may have different health goals when considering preventive therapy, for example, maximising quality of life and avoiding hospitalisation due to serious adverse events may be more important than preventing fatal CVD events [65]. Previous work found that clinicians encountered numerous challenges when optimising prescribing for individuals with dementia, including decisions about stopping medication [66]. Our findings can inform such clinical decision-making and clinicians managing hypertension in individuals with dementia should consider the balance of risks and benefits. Clinicians should pursue patient-centred care, where the patient’s goals and wishes are prioritised in agreeing the treatment.

Formal interaction tests between dementia status and antihypertensive treatment were statistically significant in three of the four analytical models, suggesting a modest but statistically detectable difference in RR by dementia status. However, the magnitude of effect modification was small, and given the large sample size, these findings should be interpreted cautiously as they may have limited clinical relevance.

Strengths of this study include the large general practice population-based sample and the examination of how the association between antihypertensive therapy and serious adverse events differ by dementia status. Despite the strengths of the study, it should be interpreted within the context of its potential limitations. First, while antihypertensive treatment is known to reduce the risk of cardiovascular mortality, our study was not designed to evaluate the potential cardiovascular benefits of antihypertensive treatment in individuals with dementia. As such, our findings should not be interpreted as a comprehensive risk-benefit assessment of antihypertensive therapy in this population. Second, dementia might have been underdiagnosed in this study population, which might have affected the study results. The association between serious adverse events risk with antihypertensive medication and dementia status diagnosed after April 2014 could not be reliably assessed in the present study due to very limited number of dementia cases after this date. Furthermore, although CPRD GOLD is a widely validated and high-quality UK primary care database, the accuracy and completeness of diagnostic coding for dementia and related comorbidities may be imperfect. Previous validation studies have demonstrated acceptable levels of diagnostic accuracy for these conditions; however, under-ascertainment or misclassification remains possible and may have influenced our study findings [26,67,68]. Third, the study included individuals with different types of dementia, based on coded diagnoses in the electronic health record, rather than careful cognitive testing. This heterogeneity of dementia diagnosis could also affect the results, as previous study showed that vascular dementia, mixed dementia, and dementia in other diseases were associated with increased risk of falls compared to Alzheimer’s disease [43]. In addition, this study did not include individuals with mild cognitive impairment who may also be at risk of serious adverse events. Fourth, we used an “intention-to-treat approach” and did not account for individuals who developed new dementia during the observation period or who initiated antihypertensive treatment in the control group (28.9% during follow-up). However, the median treatment duration was 6 years in the exposure group versus 0 years in the control group, suggesting that our effect estimates likely reflect conservative estimates of the true treatment-associated risk. More complex analyses using time-varying covariates could theoretically refine exposure classification [69]; however, integrating time-varying exposure into an analysis already involving multiple imputation and propensity score adjustment across a large, real-world dataset poses considerable methodological challenges, which are not well understood in the literature. Moreover, this approach likely yields conservative risk estimates, as control individuals who initiated treatment during follow-up were not reclassified, which may have attenuated observed treatment effects. In addition, time-varying exposure in observational data is highly susceptible to confounding by indication, as treatment initiation during follow-up may reflect worsening clinical status, thereby introducing bias that is difficult to fully address. In addition, when exposure status changes over time and may itself be influenced by prior outcomes, fixed-exposure models may be particularly vulnerable to structural confounding. While causal methods such as G-methods have been developed to address this issue [70], their implementation would require further assumptions and complexity that go beyond the scope of the current study. Furthermore, by treating exposure as fixed over time, our analysis may be subject to additional unmeasured confounding arising from changes in treatment status that are not captured in our model. We also note that we did not evaluate subsequent treatment changes, including discontinuation or dose modification of antihypertensive drugs, during the follow-up period in the exposure group. Fifth, we could not assess adherence and persistence of antihypertensive medications during the observation period. Some people with dementia have poor adherence to antihypertensive medication [71], while others might have better adherence than those without dementia due to support from the community and carers [72], which might lead to an underestimation or overestimation of the potential association between antihypertensive treatment and serious adverse events. In addition, although the dosage of antihypertensive medications is associated with the risk of serious adverse events [73], we could not assess this relationship in our study. Sixth, our findings may not be generalisable to individuals with advanced dementia, given evidence from the previous study linking sedentary behaviour to decreased cognitive function [74]. Seventh, although our propensity score analyses were successful in balancing the groups based on known confounding variables, we cannot rule out the presence of unmeasured confounding [49,75]. Nevertheless, the relative comparison between individuals with and without dementia remains valid, since the treatment effects in both individuals with and without dementia would be subject to the same potential confounders. Eighth, while we took steps to minimise the impact of missing data through appropriate imputation methods, a high proportion of individuals required imputation for a limited number of variables. Our analysis showed only a slight difference between the imputation datasets and the complete-case dataset, indicating that the imputation might introduce minimal bias. However, the findings should still be interpreted with caution due to the presence of missing variables. Ninth, while the CPRD Gold database is representative of the UK population in terms of ethnicity [26], potential underreporting of dementia among minority ethnic groups may result in an overrepresentation of White individuals in our dataset. Tenth, we did not use a Fine-Gray competing risks model because our primary aim was to estimate cause-specific hazards rather than cumulative incidence, and individuals who died from unrelated causes were censored accordingly. Moreover, previous analyses using the same dataset have shown that results from Cox and Fine-Gray models were largely consistent [9], supporting the robustness of our approach. Finally, our findings may not be generalisable to different populations or ethnic groups.

Taken together, in previously untreated individuals with elevated systolic BP, this study demonstrated the increased absolute risk of serious adverse events associated with antihypertensive treatment in individuals with dementia compared to those without dementia. When prescribing antihypertensive medication for individuals with dementia, careful consideration of potential adverse events is needed. Clinicians, patients, and their carers should carefully consider the balance between benefits and harms of antihypertensive treatment for individuals with dementia.

Supporting information

ISAC Protocol.

Showing 1/15: pmed.1004731.s001.docx

Skip to figshare navigation

Sorry we could not load your data.

1 / 15

下载

无花果分享

S1 协议。 ISAC 协议。

https://doi.org/10.1371/journal.pmed.1004731.s001

（S1_Protocol.DOCX）

S1 表。 RECORD 语句。

https://doi.org/10.1371/journal.pmed.1004731.s002

（S1_Table.DOCX）

S2 表。 用于定义痴呆症的 CPRD GOLD 医疗代码。

CPRD 表示临床实践研究数据链。

https://doi.org/10.1371/journal.pmed.1004731.s003

（S2_Table.DOCX）

S3 表。 分析中包含的药物类型和相应的英国国家处方集标题。

https://doi.org/10.1371/journal.pmed.1004731.s004

（S3_Table.DOCX）

S4 表。 插补数据集中的倾向评分模型。

血压表示血压;CI，置信区间;舒张压，舒张压;FI，衰弱指数;HDL，高密度脂蛋白;IMD，多重剥夺指数;收缩压，收缩压。

https://doi.org/10.1371/journal.pmed.1004731.s005

（S4_Table.DOCX）

S5 表。 完整案例数据集中的倾向评分模型。

血压表示血压;CI，置信区间;舒张压，舒张压;FI，衰弱指数;HDL，高密度脂蛋白;IMD，多重剥夺指数;收缩压，收缩压。

https://doi.org/10.1371/journal.pmed.1004731.s006

（S5_Table.DOCX）

S6 表。 完整案例数据集中倾向评分匹配之前或之后研究人群的基线特征。

数据是平均值±标准差或数字（百分比）。标准化平均差 （SMD） 为 <0.1 表明倾向评分匹配后有足够的可变平衡。*高剥夺表示多重剥夺得分为 5（最剥夺）。†噻嗪类药物包括噻嗪类利尿剂。‡其他抗高血压药包括中枢作用的降压药、直接肾素抑制剂和血管扩张剂。ACE表示血管紧张素转换酶;ARB、血管紧张素II.受体阻滞剂;eFI，电子衰弱指数;SMD，标准化均值差。

https://doi.org/10.1371/journal.pmed.1004731.s007

（S6_Table.DOCX）

S7 表。 完整病例数据集中每种结果开始使用抗高血压药物的风险比。

暴露组和对照组分别表示有降压处方和没有降压处方。CI表示置信区间;IPTW，治疗加权的逆概率。

https://doi.org/10.1371/journal.pmed.1004731.s008

（S7_Table.DOCX）

S8 表。 完整病例数据集中每种结果使用的抗高血压药物数量的风险比。

显示了根据倾向评分调整的 Cox 回归分析结果。服用三种以上降压药物的痴呆患者人数如下：3种药物，n=150;4 种药物，n = 43;5 种药物，n = 9;6种药物，n = 3;7 种药物，n = 0。服用三种以上降压药的无痴呆症患者人数如下：3种药物，n=4,001;4 种药物，n = 1,133;5 种药物，n = 257;6 种药物，n = 42;7 种药物，n = 2。

https://doi.org/10.1371/journal.pmed.1004731.s009

（S8_Table.DOCX）

S9 表。 随访期开始日期之前开始服用降压药物治疗跌倒的风险比。

暴露组和对照组分别表示有降压处方和没有降压处方。CI表示置信区间;IPTW，逆概率处理加权。

https://doi.org/10.1371/journal.pmed.1004731.s010

(S9_Table.DOCX)

S1 Fig. Definition of time periods used to define the cohort and follow-up periods.

Individuals were eligible for cohort entry if they met the following criteria: (1) aged ≥40 years old; (2) with qualifying first systolic BP levels of between 130 − 179 mmHg prior to the exposure period; (3) not having received any antihypertensives prior to the study start date; and (4) who were registered between 1st January 1998 and 31st December 2018 in CPRD GOLD.

https://doi.org/10.1371/journal.pmed.1004731.s011

(S1_Fig.DOCX)

S2 Fig. Flow diagram showing selection of patient records for inclusion in the study.

https://doi.org/10.1371/journal.pmed.1004731.s012

(S2_Fig.DOCX)

S3 Fig. Distribution of propensity scores by treatment status in individuals with and without dementia.

(A) the distribution of propensity scores among individuals with dementia, and (B) the distribution among those without dementia.

https://doi.org/10.1371/journal.pmed.1004731.s013

(S3_Fig.DOCX)

S4 Fig. Cumulative risk of serious adverse events by antihypertensive medication according to categories of dementia and drug exposure status in the complete-case dataset.

Kaplan–Meier curves of the cumulative incidence of serious adverse events by the four categorical groups in complete case are shown. S3A, S3B, S3C, and S3D Fig shows the cumulative incidence of falls, hypotension, syncope, and fracture, respectively. Each solid line indicates cumulative incident rate for each serious adverse event and its around area indicates the 95% confidence interval.

https://doi.org/10.1371/journal.pmed.1004731.s014

(S4_Fig.DOCX)

S5 Fig. Differences in the risk of serious adverse events associated with antihypertensive treatment by dementia status in the complete-case dataset.

Models adjusted for propensity score using complete cases only. Absolute risk differences were described as additional events per 10,000 patients per year. CI indicates confidence interval; and HR, hazard ratio.

https://doi.org/10.1371/journal.pmed.1004731.s015

(S5_Fig.DOCX)

Acknowledgments

We would like to acknowledge all investigators involved in the study. We thank Richard Stevens, medical statistician at Nuffield Department of Primary Care Health Sciences, University of Oxford, for giving us important contributions to creating the programme of STRATIFY project. This work uses data provided by patients and collected by the NHS as part of their care and support. We are very grateful to all those patients who permit their anonymised routine NHS data to be used for this approved research. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.

References

1.Ettehad D, Emdin CA, Kiran A, Anderson SG, Callender T, Emberson J, et al. Blood pressure lowering for prevention of cardiovascular disease and death: a systematic review and meta-analysis. Lancet. 2016;387(10022):957–67. pmid:26724178

View ArticlePubMed/NCBIGoogle Scholar

2.降血压治疗试验者的合作。药物降压用于不同血压水平心血管疾病的一级和二级预防：个体参与者水平的数据荟萃分析。柳叶 刀。2021;397(10285):1625–36.PMID：33933205

查看文章考研/NCBI谷歌学术

3.降血压治疗试验者的合作。降血压药物治疗预防心血管疾病和死亡的年龄分层和血压分层效果：个体参与者水平的数据meta分析。柳叶 刀。2021;398(10305):1053–64.PMID：34461040

查看文章考研/NCBI谷歌学术

4.威廉姆森 JD、Supiano MA、Applegate WB、Berlowitz DR、Campbell RC、Chertow GM 等。≥75 岁成年人的强化与标准血压控制和心血管疾病结果：一项随机临床试验。美国医学会。2016;315(24):2673–82.PMID：27195814

查看文章考研/NCBI谷歌学术

5.张 W、张 S、邓 Y、吴 S、任 J、孙 G 等。老年高血压患者强化血压控制试验。N Engl J Med. 2021;385(14):1268–79.PMID：34491661

查看文章考研/NCBI谷歌学术

6.王 Z、杜 X、华 C、李 W、张 H、刘 X 等。虚弱对强化血压控制有效性和安全性的影响：SPRINT 试验的事后分析。流通。2023;148(7):565–74.PMID：37401465

查看文章考研/NCBI谷歌学术

7.Chen L， You S， Ee N， Rockwood K， Ward DD， Woodward M. 虚弱对老年人抗高血压治疗的影响。高血压。2025;82:509–19.

查看文章谷歌学术

8.Bogaerts JMK、von Ballmoos LM、Achterberg WP、Gussekloo J、Streit S、van der Ploeg MA 等。我们是否同意老年人抗高血压药物治疗的目标：心血管疾病一级预防指南的系统评价。年龄老化。2022;51（1）：AFAB192。PMID：34718378

查看文章考研/NCBI谷歌学术

9.谢泼德 JP、科希亚里斯 C、史蒂文斯 R、莱-弗鲁里 S、班纳吉 A、贝洛斯 BK 等。按年龄和虚弱划分的抗高血压治疗与严重不良事件之间的关联：一项队列研究。公共科学图书馆医学 2023;20（4）：e1004223。PMID：37075078

查看文章考研/NCBI谷歌学术

10.Whelton PK、Carey RM、Aronow WS、Casey DE Jr、Collins KJ、Dennison Himmelfarb C 等人 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA 成人高血压预防、检测、评估和管理指南：执行摘要：美国心脏病学会/美国心脏协会临床实践指南工作组的报告。高血压。2018;71(6):1269–324.PMID：29133354

查看文章考研/NCBI谷歌学术

11.梅村 S、有马 H、有马 S、浅山 K、Dohi Y、广冈 Y 等。日本高血压学会高血压管理指南 （JSH 2019）。高血压研究 2019 年;42(9):1235–481.PMID：31375757

查看文章考研/NCBI谷歌学术

12.Mancia G、Kreutz R、Brunstrom M、Burnier M、Grassi G、Januszewicz A 等人 2023 年 ESH 动脉高血压管理指南。J 高血压。2023;41:1874–2071.

查看文章谷歌学术

13.GBD 2019 痴呆症预测合作者。2019 年全球痴呆症患病率的估计和 2050 年的预测患病率：2019 年全球疾病负担研究的分析。柳叶刀公共卫生。2022;7（2）：e105–25。PMID：34998485

查看文章考研/NCBI谷歌学术

14.利文斯顿 G、亨特利 J、索默拉德 A、艾姆斯 D、巴拉德 C、班纳吉 S 等。痴呆症预防、干预和护理：柳叶刀委员会 2020 年报告。柳叶 刀。2020;396(10248):413–46.PMID：32738937

查看文章考研/NCBI谷歌学术

15.Clague F、Mercer SW、McLean G、Reynish E、Guthrie B. 痴呆症患者的合并症和多重用药：对初级保健数据进行基于人群的大型横断面分析的见解。年龄老化。2017;46:33–9.

查看文章谷歌学术

16.He J， Zhao C， Zhong S， Ouyang N， Sun G， Qiao L， et al.未控制的高血压患者的血压降低和全因痴呆：一项开放标签、盲法终点、整群随机试验。国家医学 2025 年;31(6):2054–61.PMID：40258956

查看文章考研/NCBI谷歌学术

17.de Jong DLK、de Heus RAA、Rijpma A、Donders R、Olde Rikkert MGM、Günther M 等。尼伐地平对阿尔茨海默病患者脑血流量的影响。高血压。2019;74(2):413–20.PMID：31203725

查看文章考研/NCBI谷歌学术

18.Kume K， Hanyu H， Sakurai H， Takada Y， Onuma T， Iwamoto T. 替米沙坦对阿尔茨海默病高血压患者认知和局部脑血流量的影响。Geriatr Gerontol 国际 2011;12:207–14.

查看文章谷歌学术

19.Mossello E、Pieraccioli M、Nesti N、Bulgaresi M、Lorenzi C、Caleri V 等。低血压对接受降压药物治疗的认知障碍老年患者的影响。JAMA 实习医学 2015 年;175(4):578–85.PMID：25730775

查看文章考研/NCBI谷歌学术

20.O'Donnell M、Teo K、Gao P、Anderson C、Sleight P、Dans A 等。认知障碍以及心血管事件和死亡的风险。欧洲心脏杂志 2012 年;33(14):1777–86.PMID：22551598

查看文章考研/NCBI谷歌学术

21.An J， Li H， Tang Z， Zheng D， Guo J， Liu Y， et al.20 年随访期间的认知障碍以及全因和心血管疾病死亡风险：BLSA 的结果。J Am 心脏协会 2018 年;7（15）：e008252。PMID：30371231

查看文章考研/NCBI谷歌学术

22.Tinetti ME、Speechley M、Ginter SF。居住在社区的老年人跌倒的危险因素。N Engl J Med. 1988 年;319(26):1701–7.PMID：3205267

查看文章考研/NCBI谷歌学术

23.Lau DT、Mercaldo ND、Harris AT、Trittschuh E、Shega J、Weintraub S. 社区痴呆症老年人的多重用药和潜在的不适当药物使用。阿尔茨海默病协会。2010;24(1):56–63.PMID：19561441

查看文章考研/NCBI谷歌学术

24.罗伯逊 DA，Savva 总经理，肯尼 RA。虚弱和认知障碍——证据和因果机制的综述。老化研究修订版 2013;12(4):840–51.PMID：23831959

查看文章考研/NCBI谷歌学术

25.拉古南丹 R、霍华德 K、伊洛马基 J、希尔默 SN、格吉迪奇 D、贝尔 JS。临床医生、护理人员和痴呆症患者对取消抗高血压药物处方的偏好：离散选择实验。年龄老化。2023;52（8）：阿法德153。PMID：37596920

查看文章考研/NCBI谷歌学术

26.Herrett E、Gallagher AM、Bhaskaran K、Forbes H、Mathur R、van Staa T 等。数据资源概况：临床实践研究数据链 （CPRD）。国际流行病学杂志。2015;44(3):827–36.PMID：26050254

查看文章考研/NCBI谷歌学术

27.斯特恩 JAC、怀特 IR、卡林 JB、斯普拉特 M、罗伊斯顿 P、肯沃德 MG 等。流行病学和临床研究中缺失数据的多重插补：潜力和陷阱。英国医学杂志。2009;338：b2393。PMID：19564179

查看文章考研/NCBI谷歌学术

28.鲁宾数据库。调查中无响应的多重插补。约翰·威利父子。1987.

29.克莱格 A、贝茨 C、杨 J、瑞安 R、尼科尔斯 L、蒂尔 EA 等。使用常规初级保健电子健康记录数据开发和验证电子虚弱指数。年龄老化。2016;45:353–60.

查看文章谷歌学术

30.Hippisley-Cox J、Coupland C、Vinogradova Y、Robson J、Minhas R、Sheikh A 等。预测英格兰和威尔士的心血管风险：QRISK2 的前瞻性推导和验证。英国医学杂志。2008;336(7659):1475–82.PMID：18573856

查看文章考研/NCBI谷歌学术

31.Lloyd CD、Norman PD、McLennan D. 1971-2020 年英格兰的剥夺。应用肛门政策。2023;16:461–84.

查看文章谷歌学术

32.豪库斯 JS，刘易斯 RJ。倾向分数。美国医学会。2015;314(15):1637–8.PMID：26501539

查看文章考研/NCBI谷歌学术

33.Chesnaye NC、Stel VS、Tripepi G、Dekker FW、Fu EL、Zoccali C 等。观察性研究中治疗加权的逆概率简介。临床肾脏 J. 2021;15(1):14–20.PMID：35035932

查看文章考研/NCBI谷歌学术

34.Yiu ZZN、Mason KJ、Hampton PJ、Reynolds NJ、Smith CH、Lunt M. 使用观察数据比较苏金单抗和优特克单抗治疗银屑病有效性的随机试验复制：英国皮肤科医生协会生物制剂和免疫调节剂注册的一项研究。美国医学会皮肤醇。2021;157:66–73.

查看文章谷歌学术

35.库斯 T、沃克 AM、格林 RJ、陈 KA、加齐亚诺 JM、伯杰 K 等。非均匀效应条件下多变量logistic回归、倾向匹配、倾向调整和基于倾向加权的结果。Am J 流行病学。2006;163(3):262–70.PMID：16371515

查看文章考研/NCBI谷歌学术

36.奥斯汀电脑版。绝对风险降低和需要治疗的数量可以从事件发生时间结局的调整生存模型中获得。临床流行病学杂志。2010;63(1):46–55.PMID：19595575

查看文章考研/NCBI谷歌学术

37.Pellfolk T， Gustafsson T， Gustafson Y， Karlsson S. 居住在集体住宅中的痴呆症居民跌倒的危险因素。国际老年精神病学家。2009;21:187–94.

查看文章谷歌学术

38.威尔士 TJ、戈登 AL、格拉德曼 JRF。痴呆症患者高血压的治疗：一项多中心前瞻性观察队列研究。J Am Med Dir 协会 2019 年;20(9):1111–5.PMID：31109906

查看文章考研/NCBI谷歌学术

39.Isik AT、Dost FS、Yavuz I、Ontan MS、Ates Bulut E、Kaya D. 路易体痴呆患者的直立性低血压：前瞻性研究的荟萃分析。Clin Auton Res. 2023;33(2):133–41.PMID：36862320

查看文章考研/NCBI谷歌学术

40.Green AR、Reifler LM、Bayliss EA、Weffald LA、Boyd CM。导致患有轻度认知障碍、痴呆症和多种慢性疾病的老年人的抗胆碱能负担和跌倒或跌倒相关伤害风险的药物：一项回顾性队列研究。药物老化。2019;36(3):289–97.PMID：30652263

查看文章考研/NCBI谷歌学术

41.Marvanova M. 药物引起的认知障碍：心血管药物的作用。精神健康克林。2016;6:201–6.

查看文章谷歌学术

42.Ruangritchankul S、皮尔 NM、汉贾尼 LS、格雷 LC。患有痴呆症的老年人的药物相关问题。公共科学图书馆一号。2020;15（7）：e0236830。PMID：32735592

查看文章考研/NCBI谷歌学术

43.夏尔马 S、穆勒 C、斯图尔特 R、委罗内塞 N、范坎普福特 D、小柳 A 等。跌倒和骨折导致痴呆症患者住院的预测因素：一项具有代表性的队列研究。J Am Med Dir 协会 2018 年;19(7):607–12.PMID：29752159

查看文章考研/NCBI谷歌学术

44.奥科耶 SM、法比乌斯 CD、里德 L、沃尔夫 JL。患有和不患有痴呆症的老年人跌倒的预测因素。阿尔茨海默痴呆症。2023;19(7):2888–97.PMID：36633222

查看文章考研/NCBI谷歌学术

45.葛 ML、Chu NM、Simonsick EM、Kasper JD、薛 Q-L。社区老年人身体虚弱和认知障碍的发病顺序以及反复跌倒的风险。J Am Med Dir 协会 2023 年;24（4）：482-488.e4。PMID：36852758

查看文章考研/NCBI谷歌学术

46.潘托尼 L、德尔塞尔 T、索格利安 AG、阿米戈尼 S、斯帕达里 G、比内利 D 等。尼莫地平治疗皮质下血管性痴呆的疗效和安全性：一项随机安慰剂对照试验。中风。2005;36(3):619–24.PMID：15692125

查看文章考研/NCBI谷歌学术

47.Bogaerts JMK、Gussekloo J、de Jong-Schmit BEM、Le Cessie S、Mooijaart SP、van der Mast RC 等。停止抗高血压治疗对疗养院痴呆症居民 （DANTON） 神经精神症状和生活质量的影响：一项多中心、开放标签、盲法结果、随机对照试验。年龄老化。2024;53（7）：AFAE133。PMID：38970547

查看文章考研/NCBI谷歌学术

48.Post Hospers G、Smulders YM、Maier AB、Deeg DJ、Muller M. 老年人群血压与死亡风险之间的关系：实际年龄和生物学年龄的作用。医学实习生杂志 2015 年;277(4):488–97.PMID：25041041

查看文章考研/NCBI谷歌学术

49.阿尔巴斯里 A、哈特尔 M、科希亚里斯 C、邓尼根 A、帕克斯顿 B、福克斯 SE 等。降压治疗与不良事件的关联：系统评价和荟萃分析。英国医学杂志。2021;372：n189。PMID：33568342

查看文章考研/NCBI谷歌学术

50.范登布鲁克 JP.观察性研究何时与随机试验一样可信？柳叶 刀。2004;363(9422):1728–31.PMID：15158638

查看文章考研/NCBI谷歌学术

51.北顺 LC、哈里森 JK、哈伍德 RH、罗宾逊 TG、格拉德曼 JRF、康罗伊 SP。治疗老年痴呆高血压的证据：系统评价。J 哼哼高血压。2014;28(5):283–7.PMID：24196416

查看文章考研/NCBI谷歌学术

52.van der Wardt V、Logan P、Conroy S、Harwood R、Gladman J. 痴呆症患者的抗高血压治疗。J Am Med Dir 协会 2014 年;15(9):620–9.PMID：24755477

查看文章考研/NCBI谷歌学术

53.贝克特 NS、彼得斯 R、弗莱彻 AE、斯泰森 JA、刘 L、杜米特拉斯库 D 等。80岁或以上患者高血压的治疗。N Engl J Med. 2008;358(18):1887–98.PMID：18378519

查看文章考研/NCBI谷歌学术

54.范多恩 C、格鲁伯-巴尔迪尼 AL、齐默尔曼 S、赫贝尔 JR、Port CL、鲍姆加滕 M 等。痴呆症是疗养院居民跌倒和跌倒受伤的危险因素。J Am Geriatr Soc. 2003 年;51(9):1213–8.PMID：12919232

查看文章考研/NCBI谷歌学术

55.安德森 M、汉森 O、明顿 L、巴拉德 CG、隆多斯 E.路易体痴呆患者直立性挑战后的低血压持续时间延长。国际老年精神病学杂志。2008;23:192–8.

查看文章谷歌学术

56.Mossello E、Ceccofiglio A、Rafanelli M、Riccardi A、Mussi C、Bellelli G 等。不明原因痴呆跌倒的鉴别诊断：“晕厥和痴呆”登记结果。Eur J 实习医学 2018 年;50:41–6.PMID：29398249

查看文章考研/NCBI谷歌学术

57.王 H-K、洪 C-M、林 S-H、Tai Y-C、Lu K、Liliang P-C 等。痴呆患者髋部骨折的风险增加：一项全国性人群研究。BMC 神经学 2014;14：175。PMID：25213690

查看文章考研/NCBI谷歌学术

58.Bueno-Cavanillas A， Padilla-Ruiz F， Jiménez-Moleón JJ， Peinado-Alonso CA， Gálvez-Vargas R. 根据外在和内在诱发原因，老年人跌倒的危险因素。欧洲流行病学杂志。2000;16(9):849–59.PMID：11297228

查看文章考研/NCBI谷歌学术

59.Saedon NI、Pin Tan M、Frith J.体位性低血压的患病率：系统评价和荟萃分析。J Gerontol A Biol Sci Med Sci. 2020;75:117–22.

查看文章谷歌学术

60.布利怀斯 DL。阿尔茨海默病和其他痴呆症的睡眠障碍。临床基石。2004;6 增刊 1A：S16-28。PMID：15259536

查看文章考研/NCBI谷歌学术

61.梅利诺 G、皮亚尼 A、吉利 GL、坎塞利 I、里纳尔迪 A、巴罗塞利 A 等。白天嗜睡与意大利老年人的痴呆和认知能力下降有关：一项基于人群的研究。睡眠医学 2010 年;11(4):372–7.PMID：20219426

查看文章考研/NCBI谷歌学术

62.Kallin K， Gustafson Y， Sandman PO， Karlsson S. 老年护理机构中与老年人认知障碍人群跌倒相关的因素：一项基于人群的研究。Am J 老年精神病学。2005;13(6):501–9.PMID：15956270

查看文章考研/NCBI谷歌学术

63.艾伦·莱姆。痴呆综合征自主神经功能障碍的诊断和治疗。Curr Treat Options Neurol. 2019 年;21(8):38.PMID：31290049

查看文章考研/NCBI谷歌学术

64.Shippee TP、Parikh RR、Baker ZG、Bucy TI、Ng W、Jarosek S. 患有阿尔茨海默病和相关痴呆症的居民在疗养院生活质量方面的种族差异。J 老龄化健康。2023.

查看文章谷歌学术

65.库鲁斯基 K、吉尔 A、纳加纳坦 G、乌普舒尔 R、贾基梅宁 RL、沃奇斯 WP。一项关于患有多种疾病的老年人、他们的家庭医生和非正式护理人员之间护理目标一致性的定性描述性研究。BMC 家庭实践。2013;14:133.PMID：24010523

查看文章考研/NCBI谷歌学术

66.格林 AR、李 P、里夫 E、沃尔夫 JL、陈 CCG、克鲁赞 R 等。临床医生对痴呆症和共存疾病患者最佳处方的障碍和推动因素的看法。J Am Board Fam Med. 2019 年;32(3):383–91.PMID：31068402

查看文章考研/NCBI谷歌学术

67.麦吉尼斯 LA、沃伦-加什 C、摩尔豪斯 LR、托马斯 SL。英国常规收集的电子健康记录中痴呆诊断的有效性：一项系统评价。药物表效药 Saf.2019;28(2):244–55.PMID：30667114

查看文章考研/NCBI谷歌学术

68.Morgan A、Sinnott SJ、Smeeth L、Minassian C、Quint J. 英国初级和二级保健数据集中中风记录的一致性：一项基于人群的队列研究。BJGP 公开赛。2021;5（2）：北京石油公司.2020.0117.PMID：33234512

查看文章考研/NCBI谷歌学术

69.罗宾斯 JM，格陵兰 S，胡 FC。估计时变暴露对重复二元结果的边际平均值的因果影响。J Am Stat 协会 1999 年;94:687–700.

查看文章谷歌学术

70.丹尼尔 RM、库森斯 SN、德斯塔沃拉 BL、肯沃德 MG、斯特恩 JAC。处理时间相关混杂的方法。统计医学 2013 年;32(9):1584–618.PMID：23208861

查看文章考研/NCBI谷歌学术

71.Poon I、Lal LS、福特 ME、英国博朗。患有高血压和痴呆症的退伍军人在药物使用方面的种族/民族差异：一项全国队列研究。安·药剂师。2009;43(2):185–93.PMID：19193586

查看文章考研/NCBI谷歌学术

72.Arlt S、Lindner R、Rösler A、von Renteln-Kruse W. 痴呆患者对药物的依从性：预测因素和改善策略。药物老化。2008;25:1033–47.

查看文章谷歌学术

73.Lipsitz LA、Habtemariam D、Gagnon M、Iloputaife I、Sorond F、Tchalla AE 等。重新审视降压药物对老年跌倒的影响。高血压。2015;66(1):183–9.PMID：25941341

查看文章考研/NCBI谷歌学术

74.法尔克 RS、戴维斯 JC、刘-安布罗斯 T.久坐行为与认知功能之间有什么关联？系统评价。Br J 体育医学 2017 年;51(10):800–11.PMID：27153869

查看文章考研/NCBI谷歌学术

75.Brookhart MA、Wyss R、Layton JB、Stürmer T. 非实验研究中混杂控制的倾向评分方法。循环心血管定性结果。2013;6(5):604–11.PMID：24021692

查看文章考研/NCBI谷歌学术