Amyloid Positivity as a Risk Factor for Memory Decline and Lower Memory Performance as an Indicator of Conversion to Amyloid Positivity: Chicken and Egg

Cortical amyloid-β (Aβ) pathology is a defining characteristic of Alzheimer’s disease (AD) dementia. However, unimpaired individuals also exhibit a higher prevalence of amyloid positivity with age: about 30% of individuals over 60 years of age have amyloid levels within the range observed in AD dementia. While amyloid-positive individuals can remain cognitively normal for more than a decade (1), longitudinal studies have shown that amyloid burden is associated with steeper declines in cognition among cognitively normal individuals (2). Moreover, the risk of developing dementia is higher for individuals with elevated amyloid burden even after accounting for neurodegeneration and other forms of neuropathology, including tau pathology. Such findings suggest an association between amyloid and development of cognitive deficits, but antiamyloid therapies have failed to demonstrate cognitive benefits in phase III trials. One explanation for these failures is that antiamyloid interventions were administered too late in the disease process, after there was irreversible neuronal damage (3). The hypothesis that therapeutic intervention will be most effective if administered before clinical symptoms or even before amyloid plaque build-up has highlighted the importance of identifying risk factors for amyloid accumulation.

In this issue of Biological Psychiatry, Elman et al. (4) address this question by investigating factors associated with future conversion to amyloid positivity among individuals without dementia. Most previous studies that examined cognition in relation to amyloid burden operated within the National Institute on Aging–Alzheimer’s Association research framework that cognitive declines follow elevated amyloid burden (5), and therefore investigated cognitive performance as a dependent variable. Elman et al. (4) posed the “inverse” question of whether memory performance could be used to explain conversion to amyloid positivity within about 3 years, studying a sample of 292 individuals from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), 138 of whom had a diagnosis of mild cognitive impairment and 154 of whom were cognitively normal. The authors determined amyloid status based on cerebrospinal fluid (CSF) Aβ or amyloid positron emission tomography imaging. All participants were amyloid negative at baseline, with 13.7% becoming amyloid positive at follow-up. Using logistic regression and Cox proportional hazards models, Elman et al. (4) showed that 1 standard deviation difference in baseline memory performance, as measured by the Preclinical Alzheimer Cognitive Composite or the ADNI memory factor score, conferred a 1.5- to 1.6-fold risk of becoming amyloid positive, adjusting for APOE ε4 status, age, education, and duration of follow-up. The authors further demonstrated that higher CSF phosphorylated tau and lower CSF Aβ levels were also associated with conversion to amyloid positivity (Table 1). These findings suggest that lower memory performance is associated with a higher risk of becoming amyloid positive, even after adjusting for CSF phosphorylated tau and subthreshold levels of Aβ. Consistent with other studies, this study highlights APOE ε4 positivity as a major risk factor for amyloid accumulation (1,6). Unexpectedly, higher education was associated with subsequent amyloid positivity. As noted by the authors, this may indicate that individuals with lower education were already amyloid positive at baseline and were excluded from these analyses, or that well-educated individuals with memory concerns are more likely to enroll in the ADNI study.

Table 1.

Summary of Models Investigating Associations With Conversion to Amyloid Positivity

	Independent Variable
	Age	APOE ε4+	Education	Length of Follow-up	ADNI_MEM	PACC	CSF p-Tau+	CSF Aβ	CSF p-Tau
Logistic Regression Models	NS	↑	↑	↑
	NS	↑	↑	↑	↑
	NS	↑	↑	↑		↑
	NS	↑	↑	↑	↑		NS
	NS	↑	↑	↑		↑	NS
	NS	NS	↑	↑	↑			↓	↑
	NS	NS	↑	↑		NS		↓	↑
Cox Proportional Hazards Models	NS	↑	↑		↑
	NS	↑	↑			↑
	NS	NS	NS	NS				↓	↑
	NS	NS	NS			↑		↓	↑

Each row corresponds to a separate model. Arrows indicate the directions of the statistically significant associations of the independent variables with amyloid positivity. Cells are left blank for variables not included in the models. We indicate CSF Aβ level in this table, whereas the results reported in Elman et al. (4) are based on CSF Aβ abnormality, which is given by negating CSF Aβ level.

Aβ, amyloid-β; ADNI_MEM, Alzheimer’s Disease Neuroimaging Initiative memory factor score; CSF, cerebrospinal fluid; NS, not significant; PACC, Preclinical Alzheimer Cognitive Composite; p-tau, phosphorylated tau.

The current findings extend an accumulating literature investigating amyloid levels below the current clinical positivity thresholds (7). Previous studies have demonstrated associations of such subthreshold amyloid levels with concurrent and prospective cognitive performance as well as tau pathology (8). As Elman et al. (4) point out, it is possible that some individuals who are designated as “amyloid negative” may actually have amyloid pathology but not be classified as such because of the relatively high threshold for positivity used. An implication of this view is that the cognitive associations reported in this paper may be due to existing amyloid pathology. To test this, the authors included continuous CSF Aβ levels as an additional independent variable in the models. Supporting the idea of existing baseline amyloid pathology within this sample, subthreshold CSF Aβ level was associated with future amyloid positivity, but interestingly, so was baseline memory performance even after adjusting for baseline CSF Aβ level. These findings suggest that subtle cognitive changes may co-occur with, or even precede, initial amyloid accumulation, implying that the current hypothetical models positing a distinct sequence of changes in AD biomarkers in relation to cognition may be overly simplified. This study, along with a recent study showing that cognitive changes may be predictive of future amyloid accumulation and neurodegeneration (9), suggests a more complex temporal ordering of changes.

The authors did not investigate whether concomitant neurodegenerative changes, such as cortical atrophy or changes in glucose metabolism, can account for the observed associations between cognition and amyloid pathology. Such brain changes may constitute common drivers of both cognitive decline and amyloid accumulation. Further research taking into account neurodegeneration will help elucidate the independent associations between cognition and future amyloid status.

It is worth noting that the results presented by Elman et al. (4) are based on statistical inference at the group level to demonstrate associations between cognitive performance and future amyloid positivity. Individual-level prediction is a more difficult problem than demonstrating population-level associations and requires validation in independent samples. While the current findings can be used to design better enrichment strategies for clinical trials targeted at individuals with higher risk of becoming amyloid positive, individual-level prediction will be important for making treatment decisions if and when antiamyloid therapies are available for use in asymptomatic populations. Extension of this work will be essential to demonstrate the sensitivity and specificity of baseline cognitive performance in predicting conversion to amyloid positivity on an individual basis.

The hypothesis that antiamyloid intervention can be beneficial for cognitive performance if administered early, while individuals are still cognitively normal but amyloid positive, is currently being tested in the A4 study [Anti-Amyloid Treatment in Asymptomatic Alzheimer’s Disease (10)]. Therapeutic intervention even earlier, among cognitively normal older individuals who are amyloid negative but who are predicted to be at increased risk for amyloid accumulation, will be tested in the A3 study (Ante-Amyloid Prevention of Alzheimer’s Disease). The findings presented by Elman et al. (4) provide important information for such clinical trials going forward, as they can inform sample enrichment to enroll individuals who are more likely to become amyloid positive.

The study by Elman et al. (4) provides evidence for the existence of cognitive differences before substantial amyloid accumulation, highlighting the importance of understanding the impact of subthreshold levels of amyloid on cognitive performance and the importance of studying factors other than amyloid that may be contributing to such cognitive differences. This work should motivate greater emphasis on the earliest cognitive changes that may occur in association with the AD pathophysiologic process and raises additional questions about the temporal ordering of cognitive decline and the initial stages of the development of AD-associated neuropathology. So, which comes first—the chicken or the egg? Emerging evidence suggests that subtle cognitive decline may precede amyloid accumulation, but further research is needed to determine the direct causal links, if any, between cognitive decline and amyloid pathology.