Estimates of age-related memory decline are inflated by unrecognized Alzheimer's disease

Abstract

Cognitive decline is considered an inevitable consequence of aging; however, estimates of cognitive aging may be influenced negatively by undetected preclinical Alzheimer's disease (AD). This study aimed to determine the extent to which estimates of cognitive aging were biased by preclinical AD. Cognitively normal older adults (n = 494) with amyloid-β status determined from positron emission tomography neuroimaging underwent serial neuropsychological assessment at 18-month intervals over 72 months. Estimates of the effects of age on verbal memory, working memory, executive function, and processing speed were derived using linear mixed models. The presence of preclinical AD and clinical progression to mild cognitive impairment or dementia during the study were then added to these models as covariates. Initially, age was associated with decline across all 4 cognitive domains. With the effects of elevated amyloid-β and clinical progression controlled, age was no longer associated with decline in verbal or working memory. However, the magnitude of decline was reduced only slightly for executive function and was unchanged for processing speed. Thus, considered together, the results of the study indicate that undetected preclinical AD negatively biases estimates of age-related cognitive decline for verbal and working memory.

Introduction

Cognitive aging is a lifespan concept that describes changes in cognition that occur as individuals progress from young to very old age. In healthy older adults (i.e., >60 years old), cognitive aging is reported to involve decline across multiple domains of cognition, including episodic memory, working memory, executive function, and processing speed (Ritchie et al., 2016, Salthouse, 2010, Schaie, 2005, Wilson et al., 2015). However, emerging evidence suggests that at least some age-related cognitive decline may actually reflect the influence of biological processes, such as neurodegenerative disease, that are common in older adults but which cannot be considered to be normal aging (DeCarlo et al., 2014, Hofer and Sliwinski, 2001, Spiro and Brady, 2011). A challenge for models of cognitive aging is therefore to determine the nature and extent to which such processes may affect estimates of age-related cognitive decline.

It is now recognized that a high proportion of adults aged over 60 years, who show no impairment in cognition, mood, or behavior, harbor signs of Alzheimer's disease (AD) neuropathological changes, such as abnormal accumulation of amyloid-β (Aβ+), and as such can be considered to be in the preclinical stage of the disease (Sperling et al., 2011). For example, postmortem studies show that approximately 60%–98% of older adults aged over 60 years have evidence of AD neuropathological changes at death, despite having had no signs of cognitive impairment during life (Bennett et al., 2012, Monsell et al., 2014). Results from in vivo studies using cerebrospinal fluid and neuroimaging biomarkers also show that Aβ+ occurs in approximately 30% of cognitively normal older adults (Jack et al., 2014, Jansen et al., 2015, Vos et al., 2013). Importantly, for models of cognitive aging, the proportion of cognitively normal older adults with Aβ+ increases with age, from 10% at age of 50 years to 44% at age of 90 years (Jansen et al., 2015). As such, the potential for Aβ+ to influence estimates of cognitive aging also increases with age.

Given the relationship between the prevalence of preclinical AD and age, it is possible that psychological studies of cognitive aging have included individuals with preclinical AD in their samples of older adults. Review of the inclusion criteria of large studies of cognitive aging indicates that most aging studies exclude individuals with a clinical diagnosis of dementia from enrollment, and many also use brief cognitive screening tools such as the Mini-Mental State Examination or the Montreal Cognitive Assessment to exclude individuals with cognitive impairment from their samples (Harrington et al., 2017b). However, while such tests will identify individuals with dementia, they are insensitive to the mild cognitive difficulties that characterize the preclinical phase of the disease (Lacy et al., 2015, Spencer et al., 2013). For example, a recent meta-analysis demonstrated that Aβ+ in cognitively normal older adults is associated with a subtle decline (Cohen's d = −0.24 to −0.30) in global cognition, visuospatial function, processing speed, executive function, and episodic memory (Baker et al., 2016). In addition, few aging studies have assessed whether any of their participants progress to dementia during or at some short time after the study period, an event indicating that a neurodegenerative disease process was present during any earlier cognitive assessment (Sliwinski et al., 1996). Consequently, it is possible that the purportedly healthy samples of older adults from cognitive aging studies to date have included a substantial proportion of individuals with preclinical AD, and therefore, the effects of age on cognition may have been overestimated.

The aim of this study was to determine the influence of preclinical AD and progression to mild cognitive impairment (MCI) or dementia on estimates of age-related cognitive decline in a sample of cognitively normal older adults who have undergone objective assessment of physical and cognitive health before entry into the study. The first hypothesis was that older age would be associated with decline across all areas of cognition when Aβ+ and progression were not controlled. The second hypothesis was that the effect of age on cognition would be reduced when the effects of Aβ+ and progression were taken into account.