Classifying anatomical subtypes of subjective memory impairment

Abstract

We aimed to categorize subjective memory impairment (SMI) individuals based on their patterns of cortical thickness and to propose simple models that can classify each subtype. We recruited 613 SMI individuals and 613 age- and gender-matched normal controls. Using hierarchical agglomerative cluster analysis, SMI individuals were divided into 3 subtypes: temporal atrophy (12.9%), minimal atrophy (52.4%), and diffuse atrophy (34.6%). Individuals in the temporal atrophy (Alzheimer's disease–like atrophy) subtype were older, had more vascular risk factors, and scored the lowest on neuropsychological tests. Combination of these factors classified the temporal atrophy subtype with 73.2% accuracy. On the other hand, individuals with the minimal atrophy (non-neurodegenerative) subtype were younger, were more likely to be female, and had depression. Combination of these factors discriminated the minimal atrophy subtype with 76.0% accuracy. We suggest that SMI can be largely categorized into 3 anatomical subtypes that have distinct clinical features. Our models may help physicians decide next steps when encountering SMI patients and may also be used in clinical trials.

Introduction

Subjective memory impairment (SMI), subjective complaints of memory decline without objective evidence of impairment, is generally thought to represent subtle changes in memory that fall below detectable thresholds on common cognitive tests (Jessen et al., 2014). Whether individuals with SMI are at risk of progression to Alzheimer's disease (AD) dementia remains controversial. Some reports concluded that SMI patients have characteristics of the “worried well” population (Flicker et al., 1993, Gino et al., 2010), whereas others found that SMI patients are at risk of AD dementia (Mitchell et al., 2014). This disparity is probably due to the fact that SMI is a heterogeneous classification that includes not only preclinical AD (van Norden et al., 2008, Peter et al., 2014, Saykin et al., 2006, Striepens et al., 2010), but also various conditions that can affect cognition, such as depression and anxiety (Jessen et al., 2007, Slavin et al., 2010). Because of the high prevalence of SMI in older adults patients (26%–80%) (Jessen et al., 2011, Slavin et al., 2010, Stewart et al., 2011), it is important for clinicians to know who should be given further analysis in this heterogeneous group.

Cortical atrophic patterns can offer evidence of underlying pathology in SMI. According to temporal sequences of biomarker changes, cortical atrophy emerges years before clinical symptom onset (Jack et al., 2013). AD-like atrophy in SMI individuals can be a sign of underlying AD processes (Peter et al., 2014, Schultz et al., 2015), whereas SMI with no atrophy can be explained by psychiatric conditions, such as depression. Indeed, previous reports indicated that patients who did not have signs of neurodegeneration continued on a benign course (Vos et al., 2013). Although cerebrospinal fluid (CSF) biomarkers or amyloid positron emission tomography (PET) can identify patients who are undergoing AD pathophysiological process (Dickerson et al., 2013, Nettiksimmons et al., 2010, Perrotin et al., 2012), these tests are not widely used because of their invasiveness and high cost, and facility limitations.

Herein, we categorized SMI patients based on their cortical atrophic patterns using 3-dimensional magnetic resonance imaging (MRI) and compared demographics and cognitive function between the patient subgroups. Based on our clinical findings, we developed brief, inexpensive, and noninvasive models to discriminate (1) SMI individuals with an AD-like atrophic pattern; and (2) SMI individuals without neurodegeneration.