Characterizing mild cognitive impairment to predict incident dementia in adults with bipolar disorder: What should the benchmark be?

<b>Objective:</b> Although mild cognitive impairment (MCI) is generally considered a risk state for dementia, its prevalence and association with dementia are impacted by the number of tests and cut-points used to assess cognition and define “impairment,” and sources of norms. Here, we investigate how these methodological variations impact estimates of incident dementia in adults with bipolar disorder (BD), a vulnerable population with pre-existing cognitive deficits and increased dementia risk. <b>Method</b>: Neuropsychological data from 148 adults with BD and 13,610 healthy controls (HC) were drawn from the National Alzheimer’s Coordinating Center. BD participants’ scores were standardized against published norms and again using regression-based norms generated from HC within the same catchment area as individual BD patients (“site-specific norms”), varying the number of within-domain tests (one vs. two) and the cut-points (−1 vs. −1.5 <i>SD</i>) used to operationalize MCI. <b>Results</b>: Site-specific norms were more sensitive to incident dementia (88.6%–94.3%) than published norms (74.3%–88.6%), but only when using a “single test” definition of impairment. Specificity (22.1%–74.3%), accuracy (37.8%–68.9%), and positive predictive values (26.1%–38.3%) were overall poor. Applying a “single test” definition of impairment resulted in better negative predictive values using site-specific (92.3%–93.3%) than published norms (83.6%–86.2%), and a substantial increase in relative risk of incident dementia relative to published norms. <b>Conclusions</b>: Neuropsychologists should define “impairment” as scores below −1.0 or −1.5 <i>SD</i> on at least two within-domain measures when using published norms to interpret cognitive performance in adults with BD.

<b>研究目标：</b> 尽管轻度认知障碍（mild cognitive impairment, MCI）通常被视为痴呆的风险状态，但其患病率以及与痴呆的关联，会受到用于评估认知、定义“认知损害”的测验数量、分界值以及常模来源的影响。本研究旨在探讨这些方法学差异，如何影响双相情感障碍（bipolar disorder, BD）成人患者的新发痴呆评估结果；该人群本身已存在认知缺陷，且痴呆风险更高，属于易感群体。<b>研究方法：</b> 本研究从美国国家阿尔茨海默病协调中心（National Alzheimer’s Coordinating Center）获取了148名BD成人患者与13610名健康对照（healthy controls, HC）的神经心理学数据。研究采用两种常模对BD患者的测验分数进行标准化：一是已发表的常模，二是基于与BD患者同一场域内健康对照生成的回归常模（即“场域特异性常模”，site-specific norms）；同时变化域内测验数量（1个vs.2个）以及用于定义MCI的分界值（−1倍标准差vs.−1.5倍标准差，<i>SD</i>）。<b>研究结果：</b> 相较于已发表常模（74.3%~88.6%），场域特异性常模对新发痴呆的检出敏感性更高（88.6%~94.3%），但该优势仅在采用“单测验”损害定义时成立。特异性（22.1%~74.3%）、准确率（37.8%~68.9%）以及阳性预测值（26.1%~38.3%）整体表现较差。当采用“单测验”损害定义时，场域特异性常模的阴性预测值（92.3%~93.3%）优于已发表常模（83.6%~86.2%），且相较于已发表常模，新发痴呆的相对风险显著升高。<b>研究结论：</b> 神经心理学家在采用已发表常模解读BD成人患者的认知表现时，应将“认知损害”定义为：至少在2项域内测验中得分低于−1.0倍或−1.5倍标准差（<i>SD</i>）。