Cognitive variation reflects amyloid, tau, and neurodegenerative biomarkers in Alzheimer’s disease

Abstract

In Alzheimer’s disease (AD), the accumulation of neuropathological markers such as amyloid-β plaques, neurofibrillary tangles, and cortical neurodegeneration occurs over many years before overt manifestation of cognitive impairment. There is thus a need for neuropsychological markers that are indicative of pathological changes in the early stages of the disease. Intra-individual cognitive variability (IICV), defined as the variation of an individual’s performance across cognitive domains, is a promising neuropsychological marker measuring heterogeneous changes in cognition that may reflect these early pathological changes. In this study, we comprehensively evaluated the global and regional associations of IICV with positron emission tomography (PET) and magnetic resonance imaging (MRI) measures of AD biomarkers in cognitively normal (CN) and mild cognitive impairment (MCI) participants. We found that higher IICV was robustly associated with increased Aβ, increased tau, decreased brain glucose metabolism, and reduced cortical thickness. Higher IICV was also associated with tau (OR = 2.53, P < .001) and fluorodeoxyglucose (OR = 1.34, P < .001) positivity but not Aβ positivity (OR = 1.15, P = .107). In regional analyses, IICV showed widespread associations with AD biomarkers, with the strongest Aβ and tau effects in the frontal and temporal regions, respectively. The strongest regional cortical thickness effects were found in the entorhinal and parahippocampal cortices. Our findings suggest that IICV may be a useful neuropsychological marker for increased Aβ, and especially increased tau and neurodegeneration that are reflective of emerging AD pathology in individuals without dementia.

Introduction

Alzheimer’s disease (AD) is the most common type of dementia characterized by cognitive impairment, caused by an aggregation of amyloid-β plaques (A) and tau neurofibrillary tangles (T), resulting in neurodegenerative (N) changes such as a reduction in glucose metabolism and lower cortical thickness in the brain [1, 2]. These neuropathological changes lead to deficits in key cognitive domains such as memory, language, and executive function [3]. Accumulation of pathological biomarkers can begin as early as 20 years prior to the onset of cognitive symptoms [4], but when cognitive impairment is detected by common neuropsychological assessments, irreversible levels of neurodegeneration would have likely already occurred [5]. Hence, there is a need to evaluate the utility of other forms of neuropsychological markers to enable the early detection of individuals with increased AD risk.

Intra-individual cognitive variability (IICV), calculated as the variation in performance across cognitive domains measured at a single time point, has been proposed as a promising early neuropsychological marker of increased AD risk [6]. Cognitive decline occurs heterogeneously across domains [7], and increased IICV is considered to reflect subthreshold changes in cognition that may be indicative of early cognitive decline in preclinical AD, providing a more sensitive measure compared to mean performance. It has been shown that IICV predicts future cognitive decline and conversion to dementia when compared to cerebrospinal fluid biomarkers of amyloid and tau, underscoring its utility in predicting disease progression [6, 8, 9]. While the biological mechanisms underlying IICV in dementia are not fully understood, it has been associated with greater atrophy over time, and longitudinal IICV is associated with reduced cerebral blood flow in regions implicated in AD [10, 11]. However, the relationships of IICV with global and regional positron emission tomography (PET) and magnetic resonance imaging (MRI) measures of A/T/N biomarkers in AD are not well-understood. In this study, we aim to comprehensively evaluate the associations between IICV and measurements of amyloid (18F-florbetapir), tau (18F-flortaucipir), and neurodegeneration (18F-fluorodeoxyglucose (FDG) and cortical thickness) in older adults without dementia.

Methods

Participants

Data used in the preparation of this study were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). The ADNI was launched in 2003 as a public–private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of the ADNI has been to test whether serial MRI, PET, other biological markers and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease.

To maximize statistical power, we selected participants for each biomarker analysis based on data availability. We separately evaluated 909 participants (397 CN, 512 MCI) from ADNI2, ADNIGO, and ADNI3 with florbetapir data, 378 participants (252 CN, 126 MCI) with flortaucipir data which only came from the ADNI3 cohort, 1324 participants (595 CN, 729 MCI) with cortical thickness data from all ADNI cohorts, and 1185 participants (394 CN, 791 MCI) with FDG data from all ADNI cohorts. The key demographics are summarized in Table 1.

Table 1.

Key demographics

	Aβ PET sample		Tau PET sample		Cortical thickness sample		FDG PET sample
	CN (n = 397)	MCI (n = 512)	CN (n = 252)	MCI (n = 126)	CN (n = 595)	MCI (n = 729)	CN (n = 394)	MCI (n = 791)
Age, M ± SD	72.22 ± 6.33	71.79 ± 7.37	70.39 ± 6.20	71.88 ± 7.74	72.60 ± 6.09	72.13 ± 7.36	73.60 ± 5.95	72.55 ± 7.52
Education, M ± SD	16.69 ± 2.50	16.22 ± 2.61	16.81 ± 2.24	16.13 ± 2.40	16.44 ± 2.58	15.97 ± 2.81	16.44 ± 2.71	16.06 ± 2.67
Sex, female (%)	169 (42.57)	284 (55.47)	94 (37.30)	74 (58.73)	260 (43.70)	410 (56.24)	196 (49.75)	464 (58.66)
Apolipoprotein ε4 status (0/1/2)	273/110/14	267/191/54	165/79/8	69/42/15	408/171/16	359/289/81	282/103/9	405/301/85
IICV, M ± SD	0.68 ± 0.32	0.72 ± 0.34	0.71 ± 0.34	0.76 ± 0.34	0.68 ± 0.34	0.77 ± 0.36	0.68 ± 0.32	0.76 ± 0.37
Summary florbetapir SUVR, M ± SD	1.12 ± 0.17	1.22 ± 0.23	-	-	-	-	-	-
Meta-temporal flortaucipir SUVR, M ± SD	-	-	1.20 ± 0.12	1.33 ± 0.31	-	-	-	-
Average global cortical thickness, M ± SD	-	-	-	-	2.46 ± 0.12	2.40 ± 0.15	-	-
FDG-ROI score, M ± SD	-	-	-	-	-	-	1.28 ± 0.12	1.22 ± 0.14

Intra-individual cognitive variability (IICV)

IICV was calculated using neuropsychological measures across 3 different cognitive domains: Animal Fluency (language), Trial Making Test (TMT) A and B (executive function), and Rey Auditory Verbal Learning Test (RAVLT) immediate recall and forgetting (memory). These measures were selected due to their routine use in assessing early cognitive changes in AD and administration across all ADNI cohorts.

As previously reported [12], individual raw scores were standardized into z-scores based on the whole sample distribution of all participants without dementia (n = 1978) available for analysis at baseline. Scores for RAVLT forgetting, TMT A and B were reversed after standardization to consistently reflect lower scores as an indication of poorer performance. For each participant, IICV at each timepoint was calculated as the standard deviation of each individual test z-score from the mean test z-score. Consistency between test scores will be reflected with a lower IICV score, and large differences between test scores will be reflected with a higher IICV score.

$${\mathit{IICV}}_{\text{i}}=\sqrt{\frac{{\sum }_{k=1}^K{(Z_{\text{ik}}-A_{\text{i}})}^2}{(K-1)}},A_{\text{i}}=\frac{{\sum }_{k=1}^K{Z}_{\text{ik}}}{K}$$

Neuroimaging acquisition and processing

Florbetapir (18F-AV45) and flortaucipir (18F-AV1451) PET image data were co-registered, averaged, interpolated, and smoothed. To generate Aβ and tau SUVs in cortical regions based on the Desikan-Killany atlas [13], each PET scan was processed with its coregistered structural magnetic resonance images (MRI) with Freesurfer 7.1.1. The processing steps of the florbetapir and florbetapir PET scans have been previously described in greater detail [14, 15]. A cortical summary Aβ SUVR was calculated across 18 cortical regions spanning across the frontal, cingulate, lateral parietal, and lateral temporal areas, normalized against the whole cerebellum SUV. A meta-temporal tau SUVR was calculated across the entorhinal, amygdala, fusiform, inferior, and middle temporal cortical regions, normalized against the inferior cerebellum SUV.

Structural MRI scans were acquired using a 1.5 T scanner for ADNI1 participants and a 3 T scanner for all other participants. Cortical reconstruction and volumetric segmentation were performed using FreeSurfer version 4.3 for 1.5 T data and version 5.1 for 3 T data, and its technical details have been previously described [16, 17]. Global cortical thickness was calculated by averaging thickness measures across all cortical regions, and regional cortical thickness measures were delineated based on the Desikan-Killany atlas [13].

Lastly, acquired FDG PET images were averaged, spatially aligned, interpolated to a standard voxel size, intensity normalized, and smoothed to a common resolution of 8-mm full width at half maximum [18]. Five pre-defined regions of interest (ROIs) were identified from a meta-analysis of FDG-PET studies of AD and MCI patients, and a composite FDG-ROI score was generated by averaging across all five ROIs (i.e., posterior cingulate gyrus, left and right angular gyrus, left and right inferior temporal gyrus) and intensity normalized by dividing by the mean of the top 50% of the pons-vermis reference region [18].

Statistical analysis

We first used Pearson correlation to investigate the relationships between IICV and global Aβ and tau in separate models, followed by multiple linear regression controlling for age, sex, education, APOE ε4 dosage status (0/1/2), and diagnosis (CN or MCI) at baseline. Next, we used logistic regression to investigate whether IICV was associated with amyloid and tau positivity, controlling for the same covariates. Aβ positivity was defined as having a summary score of more than 1.11 [19], and tau positivity was defined as having a meta-temporal summary score of more than 1.80 [20]. Regional florbetapir and flortaucipir SUVs were computed by averaging across the left and right hemispheres across 34 cortical regions of interest [21]. The averaged florbetapir and flortaucipir SUVs were normalized against the whole cerebellum SUV and inferior cerebellum SUV, respectively. Correction for multiple comparisons was performed at a false discovery rate (FDR) of q < 0.05 based on linear regressions conducted across all 34 cortical ROIs for each biomarker [22]. We additionally examined whether log-transforming Aβ and tau SUVR [23] would affect their associations with IICV. Conclusions were unchanged, and we retained the usage of actual SUVR in our analyses.

For neurodegenerative measures, we similarly used Pearson correlation followed by multiple linear regression to investigate the relationship of IICV with global cortical thickness and composite FDG score at baseline across the entire sample, controlling for age, sex, education, APOE ε4 dosage, and diagnosis at baseline. We also evaluated IICV association with FDG positivity with logistic regression, defined as having a FDG score of less than 1.21 [24]. Regional cortical thickness was also averaged across hemispheres and corrected for multiple comparisons at q < 0.05. Regional analyses were not conducted for FDG as data were not available in the cortical regions delineated by the Desikan-Killany atlas. All statistical analyses were performed with R version 4.2.2.

Results

IICV is positively associated with regional amyloid PET

Across individuals without dementia, we found that higher baseline IICV was correlated with higher summary Aβ PET SUVR (r(907) = 0.12, P < 0.001, Fig. 1A) and when controlling for covariates in the multiple regression analysis (β = 0.045, SE = 0.019, P = 0.001). However, logistic regression analysis showed that higher IICV was not associated with Aβ positivity (OR = 1.15, 95% CI = 0.971–1.37, P = 0.107). Across 34 cortical regions of interest, 31 regions were significantly associated with baseline IICV even after correcting for multiple comparisons at an FDR of q < 0.05 (Table 2). The largest estimates were found in the frontal regions such as the frontal pole, medial orbitofrontal cortex, and rostral middle frontal gyrus (Fig. 2A).

Fig. 1.

Baseline IICV is associated with global pathological markers. Pearson correlation plots of IICV with A global amyloid SUVR, B meta-temporal tau SUVR, C global cortical thickness, and D FDG-ROI score

Table 2.

Independent associations of IICV on regional amyloid SUVR, tau SUVR and cortical thickness, adjusted for age, sex, education, APOE ε4 dosage, and diagnosis at baseline

Region of interest	Florbetapir SUVR	Flortaucipir SUVR	Cortical thickness
	Beta	Beta	Beta
Banks of the superior temporal sulcus	0.060**	0.079**	− 0.054***
Caudal anterior cingulate	0.041*	0.034*	− 0.047**
Caudal middle frontal	0.050**	0.071**	− 0.051***
Cuneus	0.047**	0.050**	− 0.019
Entorhinal	0.009	0.053*	− 0.104***
Frontal pole	0.065**	0.026	− 0.043*
Fusiform	0.040**	0.079**	− 0.036*
Inferior parietal	0.056**	0.072**	− 0.065***
Inferior temporal	0.056**	0.074*	− 0.061***
Insula	0.043**	0.037*	− 0.050***
Isthmus cingulate	0.043*	0.049**	− 0.070***
Lateral occipital	0.054**	0.062**	− 0.033**
Lateral orbitofrontal	0.057**	0.032*	− 0.034**
Lingual	0.042**	0.045**	− 0.023*
Medial orbitofrontal	0.063**	0.030*	− 0.023
Middle temporal	0.054**	0.058*	− 0.068***
Paracentral	0.047**	0.029*	− 0.034*
Parahippocampal	0.019	0.054**	− 0.081***
Pars opercularis	0.053**	0.057**	− 0.036**
Pars orbitalis	0.056**	0.021	− 0.047**
Pars triangularis	0.056**	0.034*	− 0.043***
Pericalcarine	0.048**	0.033*	− 0.007
Postcentral	0.041**	0.031*	− 0.036**
Posterior cingulate	0.050*	0.043*	− 0.061***
Precentral	0.034**	0.037**	− 0.042**
Precuneus	0.058**	0.064**	− 0.044***
Rostral anterior cingulate	0.052**	0.025	− 0.046**
Rostral middle frontal	0.062**	0.039*	− 0.036***
Superior frontal	0.052**	0.047**	− 0.045***
Superior parietal	0.058**	0.073**	− 0.045***
Superior temporal	0.047**	0.042*	− 0.044**
Supramarginal	0.049**	0.056**	− 0.063***
Temporal pole	0.034**	0.051*	− 0.059**
Transverse temporal	0.031	0.023	− 0.033

*q < .05, **q < .01, ***q < .001

Fig. 2.

Baseline IICV is associated with regional pathological markers. Independent associations of IICV on A regional amyloid SUVR, B regional tau SUVR, and C regional cortical thickness (q < .05) after controlling for age, education, sex, APOE ε4 dosage, and diagnosis at baseline

IICV is positively associated with regional tau PET

Similarly, higher baseline IICV was significantly correlated with higher tau meta-temporal SUVR (r(376) = 0.14, P = 0.006, Fig. 1B) and in multiple regression analysis (β = 0.081, SE = 0.030, P = 0.007). Logistic regression revealed that a higher IICV was associated with tau positivity in the meta-temporal ROI (OR = 2.53, 95% CI = 1.50–4.35, P < 0.001). Thirty out of 34 cortical regions of interest were also significantly associated with IICV at an FDR of q < 0.05 (Table 2). The largest estimates were found in the temporal regions such as the banks of the superior temporal sulcus, fusiform gyrus, and inferior temporal cortex (Fig. 2B).

IICV is negatively associated with regional cortical thickness and reduced brain glucose metabolism

Higher baseline IICV was likewise found to be significantly correlated with lower global cortical thickness (r(1322) = 0.20, P < 0.001, Fig. 1C) and in multiple regression analysis (β = –0.057, SE = 0.010, P < 0.001). We also found a significant correlation between higher baseline IICV with lower FDG-PET (r(1183) =  − 0.20, P < 0.001, Fig. 1D) and in multiple regression analysis (β =  − 0.062, SE = 0.011, P < 0.001) and greater odds of being FDG positive (OR = 1.34, 95% CI = 1.17–1.54, P < 0.001). In regional cortical thickness analyses, 31 out of 34 cortical regions of interest were significantly associated with IICV at an FDR of q < 0.05 (Table 2), where the largest effects were found in the entorhinal cortex and parahippocampal cortex (Fig. 2C).

Discussion

This study demonstrates that higher variation in intra-individual performance across cognitive domains was associated with a comprehensive array of AD-associated biomarkers in individuals without dementia. Specifically, higher IICV was associated with increased Aβ and tau accumulation and neurodegenerative measures quantified by reduced cortical thickness and brain glucose metabolism. IICV also showed extensive regional cortical associations, most evidently in the form of greater frontal amyloid load, temporal tau burden, and atrophy. These effects were robust to controlling for age, sex, education, APOE ε4 dosage, and clinical diagnosis. Taken together, our findings suggest that greater heterogeneity in performance across cognitive domains in an individual reflects underlying AD-associated pathological changes, supporting its utility as a potential neuropsychological marker of increased AD risk.

Notably, higher IICV was associated with increased odds of tau positivity but not Aβ positivity. These results are consistent with existing studies showing that tau is more strongly associated with cognitive decline as compared to Aβ [25–28]. Since Aβ accumulation starts earlier in the progression of AD, significant build-up would have occurred before cognitive deficits arise [29], and the plateauing of amyloid accumulation may have further suppressed the utility of IICV for ascertaining binary amyloid positivity. Our results indicate that variation in cognitive deficits is also driven comparatively more by tau rather than cerebral Aβ load. IICV appears to be sensitive enough to detect whether individuals are tau-positive, heralding the onset of neurodegeneration, suggesting that IICV may be a useful initial screening criterion for tau-positivity in clinical trials involving tau-targeted therapies.

Despite modest standalone associations of Aβ with cognitive decline in the literature, strong interaction effects between Aβ and tau suggest that Aβ indirectly induces cognitive decline by accelerating tau accumulation [25, 26, 30, 31]. The close association between tau and cognitive decline is likely mediated by neurodegeneration, as neurodegeneration follows closely the direction of tau spread [32, 33]. The formation of NFTs leads to synaptic dysfunction, impairing brain glucose metabolism, and leading not only to cognitive decline [32, 34, 35] but also greater heterogeneity in domain-specific cognitive decline as shown in our study. In line with previously established reports on the predictive value of IICV for longitudinal cognitive decline [6, 8, 9], our findings suggest that higher IICV reflects the substantial accumulation of Aβ, early formation of NFTs in the temporal regions, and downstream cortical thinning in key temporal regions implicated in AD, such as the entorhinal cortex.

While regional analysis revealed that IICV was associated with all biomarkers in the majority of cortical ROIs, the regions with the strongest associations differed. The pathological patterns seen in the progression of AD are described by Aβ accumulation starting in the medial frontal cortical areas [36–38] and the medial temporal regions being most vulnerable to tau accumulation and subsequent neurodegeneration [38–40]. Furthermore, early accumulation of Aβ was found to be strongly associated with a decline in performance on executive function and processing speed, while subsequent tau accumulation is more closely associated with a decline in memory retrieval [41]. As IICV had the strongest associations in the regions where Aβ and tau begin to accumulate, reduced performance in certain cognitive domains resulting in increased variability across cognitive domains likely reflects the concurrent progression of Aβ and tau pathological pathways that may proliferate at different speeds.

This study is limited by the cross-sectional nature of the study and the lack of overlap in participants across all biomarkers, restricting analyses of potential longitudinal and interaction effects. In addition, not all cognitive domains were included in the calculation of IICV, as visuospatial assessments were not administered in the ADNI cohorts. Future studies are needed to investigate whether variation in performance in specific pairs or sets of cognitive domains is more strongly associated with AD biomarkers and disease progression. ADNI has strict inclusion and exclusion criteria, which limits the generalizability of our findings to individuals in the community who are likely to have substantial comorbidities. In addition, more than 90% of the participants are White with relatively high education levels. Given the known contributions of race, ethnicity, and socioeconomic status to cognition, brain health, and risk for dementia [42–44], future studies in more representative and diverse cohorts will be needed to understand how these factors may moderate the relationship between IICV and AD-associated biomarkers. Lastly, participants with MCI included in the study are predominantly of amnestic forms, and our study does not preclude the possibility that cerebrovascular or other non-AD pathologies may contribute to greater IICV [45].

In conclusion, our findings show that IICV has the potential to serve as a neuropsychological marker for increased Aβ, tau, and neurodegeneration across various cortical regions that are reflective of emerging AD pathology. Furthermore, IICV is especially sensitive to tau and neurodegeneration compared to Aβ, congruent with increasing disease severity. This study adds to the small but growing body of evidence that the capturing of heterogeneity in cognitive decline across domains by IICV may have utility as a non-invasive, additional, or alternative neuropsychological marker for the identification of individuals with high AD pathological load who are most at risk of progressing to AD dementia.

Author contribution

KASP: data curation, formal analysis, visualization, and writing—original draft. CHT: conceptualization, methodology, visualization, writing—review and editing, supervision, and project administration.

Funding

This project was supported by the Nanyang Technological University under the URECA Undergraduate Research Programme.

Data collection and sharing for the Alzheimer’s Disease Neuroimaging Initiative (ADNI) is funded by the National Institute on Aging (National Institutes of Health Grant U19AG024904). The grantee organization is the Northern California Institute for Research and Education. In the past, ADNI has also received funding from the National Institute of Biomedical Imaging and Bioengineering, the Canadian Institutes of Health Research, and private sector contributions through the Foundation for the National Institutes of Health (FNIH) including generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics.

Data availability

The data that supports the current study can be requested from https://adni.loni.usc.edu/.

Declarations

Ethics approval and consent to participate

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This research study was conducted retrospectively from data obtained by ADNI.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.