When does cognitive decline begin? A systematic review of change point studies on accelerated decline in cognitive and neurological outcomes preceding mild cognitive impairment, dementia, and death

Abstract

Older adults who ultimately develop dementia experience accelerated cognitive decline long before diagnosis. A similar acceleration in cognitive decline occurs in the years before death as well. To evaluate preclinical and terminal cognitive decline, past researchers have incorporated change points in their analyses of longitudinal data, identifying point estimates of how many years prior to diagnosis or death that decline begins to accelerate. The current systematic review aimed to summarize the published literature on preclinical and terminal change points in relation to mild cognitive impairment (MCI), dementia, and death, identifying the order in which cognitive and neurological outcomes decline and factors that modify the onset and rate of decline. A systematic search protocol yielded 35 studies, describing 16 longitudinal cohorts, modeling change points for cognitive and neurological outcomes preceding MCI, dementia, or death. Change points for cognitive abilities ranged from 3-7 years prior to MCI diagnosis, 1-11 years prior to dementia diagnosis, and 3-15 years before death. No sequence of decline was observed preceding MCI or death, but the following sequence was tentatively accepted for Alzheimer's disease: verbal memory, visuospatial ability, executive functions and fluency, and last, verbal IQ. Some of the modifiers of the onset and rate of decline examined by previous researchers included gender, education, genetics, neuropathology, and personality. Change point analyses evidence accelerated decline preceding MCI, dementia, and death, but moderators of the onset and rate of decline remain ambiguous due to between-study modeling differences, and coordinated analyses may improve comparability across future studies. (PsycINFO Database Record

Figure 1

Flowchart of Systematic Review

Figure 2. Change Point Estimates with 95% Confidence Intervals for MCI Studies by Outcome Type

Note. All change point estimates derive from the Oregon Brain Aging Study (Buracchio et al., 2010; Carlson, 2008; Howieson et al., 2008; Silbert et al., 2012), except for both Global Cognition change points, which derive from the Religious Orders Study and Memory and Aging Project (Wilson et al., 2011). CSF = Cerebrospinal Fluid; LM = Logical Memory; MCI = Mild Cognitive Impairment; WMH = White Matter Hyperintensities. Lower end confidence intervals for Gait Speed and WMH were unknown. All change points for motor outcomes come from the full sample (Men and Women; Buracchio et al., 2010).

Figure 3. Change Point Estimates with 95% Confidence/Credible Intervals for Alzheimer’s Disease Studies by Cognitive Construct

Note. ADC NDI = Alzheimer’s Disease Centers Neuropsychological Database Initiative; ADRC = Washington University School of Medicine Alzheimer’s Disease Research Center; AL = Associate Learning; AMNART = American Version of the Nelson Adult Reading Test; BAS = Bronx Aging Study; BLSA = Baltimore Longitudinal Study of Aging; BSRT = Buschke Selective Reminding Test; Comp. = Composite; FCSRT = Free and Cued Selective Reminding Test; H70 = Gerontological and Geriatric Population Study; KP = Kungsholmen Project; LM = Logical Memory; MAP = Memory and Aging Project; MMSE = Mini Mental Status Examination; ROS = Religious Orders Study; TMT = Trail Making Test; WMS = Wechsler Memory Scale. For the Bronx Aging Study, the only change point listed above is for the AD-only sample (Hall et al., 2000). One study (Johnson et al., 2009) reported a change point for a Verbal Memory factor, and separate change points for the tests that composed this Verbal Memory factor. The change points included in the figure above came from the individual tests and not the factor. Another study (Grober et al., 2008) provided two change points for the Free and Cued Selective Reminding Test. Only the first, earlier change point is displayed on the figure. Two studies reported change points for the ROS/MAP cohort (Li et al., 2015; Wilson et al., 2011), and only the change points from the study that reported estimates for multiple constructs are displayed above (Wilson et al., 2011).

Figure 4. Change Point Estimates with 95% Confidence/Credible Intervals for Terminal Decline Studies by Cognitive Construct

Note. BAS = Bronx Aging Study; BSRT = Buschke Selective Reminding Test; CC75C = Cambridge City over 75 Cohort Study; CLS = Canberra Longitudinal Study; H70 = Gerontological and Geriatric Population Study; MMSE = Mini Mental Status Examination; MoVIES = Monongahela Valley Independent Elders Survey; OCTO-Twin = Origins of Variance in the Old-old Study; MAP = Memory and Aging Project; ROS = Religious Orders Study; VLS = Victoria Longitudinal Study. Three studies (Muniz-Terrera et al., 2013; Muniz Terrera et al., 2014; van den Hout et al., 2013) all reported change points for the MMSE using CC75C data, but only the change point for one study (Muniz Terrera et al., 2014) is displayed above. Eight studies (Boyle et al., 2013; Wilson et al., 2003, 2007, 2010, 2012a, 2012b, 2015; Yu et al., 2013) provided change points for a Global Cognition Summary Measure using data from the ROS and/or MAP, but only the change point from one study (Wilson et al., 2015) is displayed above. Three studies provided estimates for ROS data based on cognitive construct (Wilson et al., 2003, 2012a, 2012b), and only the change points from one study (Wilson et al., 2003) is displayed above.