Detection of Tau-PET Positivity in Clinically Diagnosed Mild Cognitive Impairment with Multidimensional Features

Abstract

Background:

The way to evaluate brain tau pathology in vivo is tau positron emission tomography (tau-PET) or cerebrospinal fluid (CSF) analysis. In the clinically diagnosed mild cognitive impairment (MCI), a proportion of tau-PET are negative. Interest in less expensive and convenient ways to detect tau pathology in Alzheimer’s disease has increased due to the high cost of tau-PET and the invasiveness of lumbar puncture, which typically slows down the cost and enrollment of clinical trials.

Objective:

We aimed to investigate one simple and effective method in predicting tau-PET status in MCI individuals.

Methods:

The sample included 154 individuals which were dichotomized into tau-PET (+) and tau-PET (–) using a cut-off of >1.33. We used stepwise regression to select the unitary or combination of variables that best predicted tau-PET. The receiver operating characteristic curve was used to assess the accuracy of single and multiple clinical markers.

Results:

The combined performance of three variables [Alzheimer’s Disease Assessment Scale-Cognitive Subscale 13 (ADAS-Cog13), Mini-Mental State Examination (MMSE), ADNI-Memory summary score (ADNI-MEM)] in neurocognitive measures demonstrated good predictive accuracy of tau-PET status [accuracy = 85.7%, area under the curve (AUC) = 0.879]. The combination of clinical markers model (APOE ɛ4, neurocognitive measures and structural MRI imaging of middle temporal) had the best discriminative power (AUC = 0.946).

Conclusion:

As a noninvasive test, the combination of APOE ɛ4, neurocognitive measures and structural MRI imaging of middle temporal accurately predicts tau-PET status. The finding may provide a non-invasive, cost-effective tool for clinical application in predicting tau pathology among MCI individuals.

Keywords

Alzheimer’s disease Apolipoprotein E4 magnetic resonance imaging mild cognitive impairment positron-emission tomography

INTRODUCTION

Alzheimer’s disease (AD), the most common cause of dementia, is a progressive neurodegenerative disease, and pathologically defined as amyloid-β (Aβ) plaques, tau protein neurofibrillary tangles, and neurodegeneration [1]. Clinical stages can range from cognitively normal (CN) to mild cognitive impairment (MCI) and dementia [2]. At present, there is no definite method to treat AD. However, early diagnosis and interventions in AD have multiple benefits, including longer survival, improved mental health, and lower healthcare costs [3]. Therefore, as an intermediate stage between CN and AD, MCI is necessary to identify before the onset of dementia [4].

Now, it is generally accepted that positron emission tomography (PET) and cerebrospinal fluid (CSF) are the most effective methods to detect and track the pathological change in the brain [5]. Aβ-PET and tau-PET make it possible to quantify the pathophysiological presence of AD [6], and the analysis of CSF biomarkers can be used to identify protein types [7]. Despite these advances, not everyone can afford the costly PET scans, and applications in clinical and experimental environmental are limited. Although lumbar puncture is very safe, invasive operation may bring pain to patients and there is still reluctance from patients and professionals to collect CSF biological samples. For Aβ-PET, a cost-effective clinical substitute has been proposed. For example, the combination of APOE ɛ4, age, and the Alzheimer’s Disease Assessment Scale-Cognitive Subscale 13 (ADAS-Cog13) can accurately predict the state of amyloid-PET [8]. Other studies have systematically compared the models predicting Aβ-PET positive and found an effective combination of demographics, APOE ɛ4, cognition, structural MRI, and plasma biomarkers [9]. As one of the pathological features of AD, tau accumulation is closely related to synaptic loss, neurodegeneration, and cognitive impairment [10]. It is worth noting that compared with Aβ, the pathology of tau protein is more linked to the symptoms of AD [11]. However, in the clinically diagnosed MCI, a proportion of tau-PET are negative [12]. So far, there is still a lack of research on using clinically available minimally invasive biomarkers to accurately identify and predict tau-PET status. Therefore, there is an urgent need to find a simple and popularized alternative operation to predict tau-PET positive in MCI individuals.

Demographics, especially APOE ɛ4 genotype, cognitive performances, and structural MRI have consistently been shown to be associated with AD spectrum [8]. Herein, we performed cross-sectional observations of individuals with MCI from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort stratified into tau pathological status using tau-PET in order to determine which combinations of AD-related demographic, neurocognitive, and available structural MRI measures are better predictors.

METHODS

Study design

Data used in the preparation of the article were obtained from the ADNI database (http://adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership. Its main goal is to test whether the combination of neuroimaging, clinical, biological, and neuropsychological markers can measure MCI and early AD progression. The ADNI study was approved by the institutional review boards of all participating centers (https://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf).

Cohort

Subjects of this study were ADNI participants with known tau-PET status and with demographics data (age, sex, years of education, APOE ɛ4 status), clinical assessments, CSF biomarkers, and structural MRI. To compare the predictive power of CSF biomarkers, we regard it as subgroup including demographics and CSF biomarkers. All participants were MCI, which was defined by ADNI as that they complained about objective memory impairment, but did not show impairment in other cognitive areas, demonstrated preserved activities of daily living and were not demented. This definition is supported by a Mini-Mental State Examination (MMSE) score of 24 to 30 and a Clinical Dementia Rating score (CDR) of 0.5. Suffering from MCI excluding aging, small vessel disease, stress, depression, drugs, demyelinating disease, or unknown/uncertain etiology. The primary focus of this study was to assess tau-PET positivity from single time-point observations of demographics, neurocognitive measures, structural MRI, and CSF biomarkers. Therefore, a cross-sectional study design was used. The data collection deadline was March 3, 2020.

PET tau status

Whole brain anatomical T1-MPRAGE or IR-SPGR MRI (Slice thickness 1–1.2 mm; TR, 2300–3000 ms; TE 2.9–3.5 ms; FoV, 256×256 cm²) and ¹⁸F-flortaucipir (AV-1451) tau-PET were acquired. A detailed description of tau-PET image acquisition and processing can be found at http://adni.loni.usc.edu/datasamples/pet/. All T1-weighted images were processed in the FreeSurfer (V.6.0) recon-all pipeline, including registration to standard space, intensity normalization, brain extraction, tissue type classification, surface reconstruction, and probabilistic anatomical labelling [13]. Cortical tau depositions were assessed by ¹⁸F-AV1451 PET imaging using the PETSurfer tool in FreeSurfer [14]. All ¹⁸F-AV1451 scans were downloaded in the most fully preprocessed format available on LONI (https://ida.loni.usc.edu/; series description: AV1451 Coreg, Avg, Std Img and Vox Siz, Uniform Resolution) and were coregistered to the corresponding anatomical MRI. The mean standard uptake value ratio (SUVR) was calculated relative to a reference region. SUVR images were normalized by mean inferior cerebellar gray matter. To perform surface-based analysis, the results were sampled onto the FreeSurfer fsaverage surface, halfway between the white and pial surface via the individual surface. The results were spatially smoothed on the surface using a 5 mm FWHM Gaussian kernel. We calculated a composite meta region of interest (ROI) using the standardized uptake value ratio in FreeSurfer-derived atlas regions (Desikan-Killiany Atlas), extracted for each participant for tau-PET. We calculated composite meta-ROIs for tau-PET. As for tau-PET meta-ROIs which captures a broad dynamic range across the normal to pathological aging to AD dementia spectrum, the composite region was made up of amygdala, entorhinal cortex, fusiform, parahippocampal, and inferior temporal and middle temporal gyri [15]. The AD-signature ROI proposed by Jack et al. contains absorption of key regions susceptible to tau accumulation, which may be more sensitive to detection of early tau pathology [16]. The meta-ROIs SUVR threshold of tau-PET was 1.33 [17].

Demographics data and APOE ɛ4 genotype

Age at tau-PET imaging, sex, and years of education were included as demographic characteristics of each participant. APOE genotypes were tested using standard polymerase chain reaction methods as described at http://www.adni-info.org. Study participants were stratified into APOE ɛ4(+/+), APOE ɛ4(+/–), and APOE ɛ4(–/–) carriers according to the number of ɛ4 alleles that they had.

Neurocognitive measures

Subjects in the main cohort underwent neurocognitive measures at the time of tau-PET in this study include the MMSE based on a 30-point questionnaire, ADAS-Cog13 items, the CDR Sum of Boxes, Montreal Cognitive Assessment (MoCA), Functional Assessment Questionnaire (FAQ) to assess the global cognition, the ADNI-EF to evaluate executive function, the Everyday Cognition (ECog) questionnaire to examine cognitive complaints, the Rey Auditory Verbal Learning Test (RAVLT), and ADNI-MEM to evaluate memory function.

MRI imaging

The imaging data obtained from the ADNI dataset were fully preprocessed using a standardized pipeline; the image acquisition details are provided 173 elsewhere (http://adni.loni.usc.edu/) and are summarized briefly below.

These scans were performed using a 3.0T MRI scanner with T1-weighted imaging parameters. Details parameters are provided on the ADNI website (http://adni.loni.usc.edu/). The ADNI database provides a standardized MRI data sets. These MRI images were automatically quantified using FreeSurfer (version 5.1). The ADNI staff members have finished MRI Image processing work. According to the labels on Jacob’s atlas defined by FreeSurfer, 107 regions of interest (ROIs) were divided [18]. Combined with the results of existing literature, we selected ROIs related to hippocampus, entorhinal, temporal pole, middle temporal, inferior temporal and fusiform, which may significantly affect the progression of MCI [19].

CSF data

On the multiplex xMAPLuminex platform, total tau (t-tau), CSF Aβ_1 - 42, and phosphorylated-tau at threonine 181 (p-tau) were detected by the Innogenetics INNO-BIA AlzBio3 immunoassay kit [20].

Statistical analysis

Demographic, clinical and other characteristic differences between tau-PET (+) and tau-PET (–) participants were described using two-tailed Student t test and the chi-square (χ²) test for continuous and categorical variables, respectively. If normality or homogeneity of variance is not satisfied in each group, Wilcoxon rank sum test was used. Statistical analysis was performed using IBM SPSS Statistics 26. We considered the p value <0.05 as statistically significant.

We used stepwise regression to select the unitary or combination of variables that best predicted tau-PET from different data types. Receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive value of each marker or marker combination selected for tau-PET status. For each ROC curve, we also calculated the area under the curve (AUC) value, as well as sensitivity, specificity, and accuracy. Generally, if the AUC of a trial is not less than 0.70, it is acceptable in terms of clinical efficacy [8]. Select the best model and draw a nomogram to judge the relative importance of the prediction variables. Finally, the clinical diagnostic value of the best model was evaluated by decision curve analysis (DCA). The above analysis is realized by R statistical software (version 4.1.2).

RESULTS

Participant characteristics

We included 154 participants who were clinically diagnosed with MCI, of whom 84 (17 tau-PET (+) and 67 tau-PET (–)) were in the main cohort. The demographic and cognitive measurement characteristics of participants in the main cohort are shown in Table 1. The MRI features are summarized in Table 2. And participant characteristics of CSF sub-cohort are shown in Table 3. In brief, the frequency of APOE ɛ4 allele in tau-PET (+) was higher than that in tau-PET (–) in both the main cohort and sub-cohort. Within the CSF sub-cohort, tau-PET (+) had greater CSF Aβ_42/40 compared to tau-PET (–).

Table 1

The demographic and neurocognitive measurement characteristics of participants in the main cohort

	Tau-PET (+) MCI	Tau-PET (–) MCI	p
	(N = 17)	(N = 67)
Female, n (%)	10 (58.82%)	25 (37.31%)	0.17
APOEɛ4 (+/+), n (%)	13 (76.47%)	25 (37.31%)	<0.01^*
Years of education	16.18±2.04	16.52±2.46	0.555
Age	69.95±6.38	72.39±8.23	0.196
ADAS-Cog13	21.33±7.33	12.78±4.54	<0.01^*
ADNI-EF	0.04±0.97	0.67±0.85	0.022^*
ADNI-MEM	–0.02±0.68	0.46±0.53	0.014^*
CDR-SB	1.91±1.15	1.32±0.99	0.064
EcogPtTotal	1.86±0.6	1.89±0.67	0.862
FAQ	5.35±4.05	2.73±4.19	0.026^*
MMSE	26.29±2.76	28.31±1.49	<0.01^*
MoCA	21.41±4.15	23.82±2.77	0.035^*
RAVLT_forgetting	2.94±8.22	5.01±2.56	0.319
RAVLT_immediate	32.24±11.73	38.73±9.79	0.047^*
RAVLT_learning	3.18±2.72	4.96±2.49	0.022^*
RAVLT_perc_forgetting	41.67±117.44	54.68±27.91	0.656

Results are mean±(SD). p value assessed using two-tailed Student t test for each variable except female and APOE ɛ4, where chi-square (χ²) test was performed. ^*p value statistically significant. MCI, mild cognitive impairment; APOE, Apolipoprotein E; ADAS-Cog13, Alzheimer’s Disease Assessment Scale-Cognitive Subscale 13; ADNI-EF, ADNI-Executive function summary score; ADNI-MEM, ADNI-Memory summary score; CDR-SB, Clinical Dementia Rating Sum of Boxes; FAQ, Functional Assessment Questionnaire; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; RAVLT, Rey Auditory Verbal Learning Test.

Table 2

The MRI imaging characteristics of participants in the main cohort

	Tau-PET (+) MCI	Tau-PET (–) MCI	p
	(N = 17)	(N = 67)
Volumes of right hippocampus	3413.16±574.3	3813.59±539.46	0.016^*
Volumes of left hippocampus	3301.26±493.59	3692.61±499.71	<0.01^*
Volumes of right entorhinal	1863.41±489.95	2015.79±429.14	0.253
Surface area of right entorhinal	399.18±63.34	389.91±64.94	0.597
Thickness average of right entorhinal	3.2±0.42	3.55±0.46	<0.01^*
Cortical thickness standard deviation of right entorhinal	0.85±0.12	0.76±0.15	0.011^*
Volumes of left entorhinal	1848.35±481.33	1984.63±386.67	0.291
Surface area of left entorhinal	431.47±69.67	427.31±75.54	0.831
Thickness average of left entorhinal	3.08±0.41	3.44±0.49	<0.01^*
Cortical thickness standard deviation of left entorhinal	0.85±0.1	0.78±0.16	0.04^*
Volumes of right temporal pole	2394.53±520.8	2445.18±421.45	0.714
Surface area of right temporal pole	436.94±60.9	439.76±56.2	0.864
Thickness average of right temporal pole	3.52±0.4	3.73±0.35	0.065
Cortical thickness standard deviation of right temporal pole	0.72±0.12	0.65±0.13	0.054
Volumes of left temporal pole	2416.59±461.17	2382.09±403.11	0.78
Surface area of left temporal pole	437.35±45.27	450.49±48.65	0.302
Thickness average of left temporal pole	3.52±0.39	3.61±0.38	0.441
Cortical thickness standard deviation of left temporal pole	0.73±0.13	0.69±0.13	0.385
Volumes of right middle temporal	9813.18±1402.46	10968.33±1441.05	<0.01^*
Surface area of right middle temporal	3003.41±314.64	3217.97±387.28	0.023^*
Thickness average of right middle temporal	2.71±0.21	2.86±0.16	0.013^*
Cortical thickness standard deviation of right middle temporal	0.64±0.08	0.63±0.08	0.902
Volumes of left middle temporal	9271.59±1264	10186.39±1678.58	0.019^*
Surface area of left middle temporal	2725.76±255.89	2967.78±407.44	<0.01^*
Thickness average of left middle temporal	2.74±0.2	2.82±0.19	0.178
Cortical thickness standard deviation of left middle temporal	0.66±0.08	0.64±0.06	0.305
Volumes of right inferior temporal	9337.82±1352.36	9985.61±1631.95	0.102
Surface area of right inferior temporal	2882.59±309.52	2970.4±430.01	0.345
Thickness average of right inferior temporal	2.74±0.22	2.86±0.14	0.035^*
Cortical thickness standard deviation of right inferior temporal	0.72±0.1	0.7±0.08	0.361
Volumes of left inferior temporal	9233.24±1396.84	10845.87±1708.87	<0.01^*
Surface area of left inferior temporal	2885.94±319.66	3189.78±461.88	<0.01^*
Thickness average of left inferior temporal	2.7±0.2	2.86±0.15	<0.01^*
Cortical thickness standard deviation of left inferior temporal	0.7±0.06	0.7±0.07	0.862
Volumes of right fusiform	9157.94±1924.29	9413.93±1338.54	0.61
Surface area of right fusiform	2830.71±362	2893.49±346.35	0.526
Thickness average of right fusiform	2.78±0.32	2.84±0.17	0.477
Cortical thickness standard deviation of right fusiform	0.7±0.06	0.69±0.08	0.662
Volumes of left fusiform	9300.18±1567.96	9525.82±1363.52	0.592
Surface area of left fusiform	2929.53±314.50	2928.16±341.1	0.988
Thickness average of left fusiform	2.71±0.27	2.84±0.17	0.076
Cortical thickness standard deviation of left fusiform	0.68±0.07	0.69±0.09	0.81

Table 3

Participant characteristics in CSF sub-cohort

	Tau-PET (+) MCI	Tau-PET (–) MCI	p
	(N = 17)	(N = 53)
Demographics
Female, n (%)	11 (64.70%)	21 (39.62%)	0.47
APOEɛ4 (+/+), n (%)	13 (76.47%)	19 (35.85%)	<0.01^*
Age	69.64±6.31	71.93±8.28	0.33
Years of education	16.29±2.14	16.30±2.64	0.89
CSF biomarker
Aβ₄₀	17890.00±6033.84	17554.53±5417.73	0.89
Aβ₄₂	575.99±262.47	1166.41±622.75	<0.01^*
p-tau	40.59±20.25	22.08±11.20	<0.01^*
t-tau	377.57±153.36	233.40±91.86	<0.01^*
Aβ_42/40	0.33±0.01	0.07±0.03	<0.01^*

Results are mean±(SD). p value assessed using two-tailed Student t test for age, years of education, Aβ₄₀ and Aβ_42/40; chi-square (χ²) test for female and APOE ɛ4; Wilcoxon rank sum test for Aβ₄₂, p-tau, t-tau. ^*p value statistically significant. MCI, mild cognitive impairment; APOE, Apolipoprotein E; CSF, cerebrospinal fluid; Aβ, amyloid-β; p-tau, phosphorylated tau; t-tau, total tau.

Compared to tau-PET (–), tau-PET (+) showed poor performance in the global cognitive scale and subscales such as memory, executive ability, and learning. To be specific, the global cognitive performance was lower ADAS-Cog13, MMSE, MoCA, and greater FAQ. In each cognitive domain, participants of tau-PET (+) performed worse in executive function (ADNI-EF), had lower memory (ADNI-MEM, RAVLT-immediate) and learning (RAVLT-learning) function score.

Significant differences in 8 of the ROIs were found between tau-PET (–) and tau-PET (+) individuals. We found that compared with tau-PET (–), tau-PET (+) individuals had smaller volumes of bilateral hippocampus, bilateral middle and left inferior temporal areas, and smaller thickness average of bilateral entorhinal cortex, right middle and bilateral inferior temporal cortex areas. The surface area of bilateral middle and left inferior temporal was also smaller.

Predictive data modality

The variable combination from each data modality with the highest AUC was selected, as shown in Table 4. Confidence intervals for these variables and model performance also reported. In structural MRI modality, although the combination of variables selected by stepwise regression algorithm had the largest AUC, it selected too many variables which was limited in practical work. Then we selected the best combination of predictive variables from each structural MRI region.

Table 4

Diagnostic values of different biomarkers and clinical assessments in predicating tau-PET positivity in MCI

Cohort	Variable types	Variables	AUC	95% CI	Sensitivity (%)	Specificity (%)	Accuracy (%)
Main cohort	Demographics	APOEɛ4	0.757	0.646–0.867	94.1	50.7	59.5
	Neurocognitive measures	ADAS-Cog13 + MMSE + ADNI-MEM	0.879	0.772–0.985	76.5	88.1	85.7
	MRI imaging
	Hippocampus	Volumes of left hippocampus	0.723	0.578–0.869	58.8	83.6	78.6
	Entorhinal	Cortical thickness standard deviation of right entorhinal + Thickness average of left entorhinal	0.767	0.647–0.888	94.1	56.7	64.3
	Temporal pole	Thickness average of right temporal pole + Volumes of left temporal pole + Surface area of left temporal pole + Thickness average of left temporal pole	0.788	0.633–0.942	64.7	88.1	83.3
	Middle temporal	Thickness average of right middle temporal + Volumes of left middle temporal + Surface area of left middle temporal	0.835	0.733–0.937	82.4	70.1	72.6
	Inferior temporal	Surface area of right inferior temporal + Surface area of left inferior temporal + Thickness average of left inferior temporal	0.822	0.700–0.944	70.6	86.6	83.3
	Fusiform	Volumes of left fusiform + Surface area of left fusiform + Thickness average of left fusiform	0.755	0.609–0.901	52.9	95.5	86.9
CSF sub-cohort	Demographics	APOEɛ4 + gender	0.775	0.657–0.892	70.6	73.6	72.9
	CSF biomarkers	Aβ₄₀ + p-tau + t-tau + Aβ_42/40	0.922	0.861–0.984	94.1	81.1	84.3

The area under curve of receiver operating characteristic, accuracy, sensitivity, and 95% confidence interval of the best combination of predictive tau-PET positive variables selected in the main cohort and CSF sub-cohort under each data modality. APOE, Apolipoprotein E; ADAS-Cog13, Alzheimer’s Disease Assessment Scale-Cognitive Subscale 13; MMSE, Mini-Mental State Examination; ADNI-MEM, ADNI-Memory summary score; CSF, cerebrospinal fluid; p-tau, phosphorylated tau; t-tau, total tau; AUC, area under curve of receiver operating characteristic; CI, confidence intervals.

The combined performance of four variables in cerebrospinal fluid biomarkers demonstrated the best predictive accuracy of tau-PET status (accuracy = 84.3%; AUC = 0.922), followed by neurocognitive measures using the combination of three variables (ADAS-Cog13, MMSE, ADNI-MEM) (accuracy = 85.7%; AUC = 0.879). Structural MRI in the middle temporal (thickness average of right, volumes of left, surface area of left) also showed high accuracy (accuracy = 72.6%; AUC = 0.835). Specific ROC curve is shown in Fig. 1A and 1B.

Fig. 1

Receiver operating characteristic (ROC) analysis of each data modality. ROC was performed in each data modality to select a combination of variables for the positive prediction of tau-PET in the main cohort (A) and CSF sub-cohort (B). A) ROC curve of APOE ɛ4, neurocognitive measures (ADAS-Cog13, MMSE, ADNI-MEM combined) and MRI (thickness average of right middle temporal, volumes of left middle temporal, surface area of left middle temporal combined). B) ROC curve of demographics (APOE ɛ4, gender combined) and CSF (Aβ₄₀, p-tau, t-tau, Aβ_42/40 combined).

ROC curves for individual measurements of selected variables in each model are shown in Fig. 2A and 2B. Of these modalities, the best AUC was recorded by univariate CSF Aβ_42/40 (AUC = 0.875), trailed by ADAS-COg13 (AUC = 0.818). The volumes of left middle temporal, surface area of left middle temporal, gender, CSF Aβ₄₀ did not show statistically significant predictive accuracy for tau-PET status (AUC <0.70).

Fig. 2

Receiver operating characteristic (ROC) analysis for selected individual variable in each data modality. ROC was an individual variable selected variables in each data modality with positive predictions of tau-PET in the main cohort (A) and CSF sub-cohort (B). A) ROC curve of APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM, thickness average of right middle temporal, volumes of left middle temporal and surface area of left middle temporal. B) ROC curve of APOE ɛ4, gender, Aβ₄₀, p-tau, t-tau, and Aβ_42/40.

The ROC curves of demographics (APOE ɛ4 genotypes alone or combined sex) and combined multi-marker (models 1, 2, 3, and 4) are demonstrated in Fig. 3A and 3B. Model 1 was selected for combined APOE ɛ4 and neurocognitive measures (ADAS-Cog13, MMSE, ADNI-MEM) in predicting tau-PET status, model 2 for combined APOE ɛ4 and structural MRI in the middle temporal (thickness average of right, volumes of left, surface area of left), model 3 for combined APOE ɛ4, neurocognitive measures, and structural MRI in the middle temporal. Model 4 was selected for combined demographics (APOE ɛ4, gender) and CSF biomarkers (Aβ₄₀, p-tau, tau, Aβ_42/40) in predicting tau-PET status in subgroup. The results of the corresponding ROC curve analysis are summarized in Table 5. The AUC, sensitivity, and specificity for APOE ɛ4 genotype were 0.757, 94.1%, and 50.7%. The AUC of models 1, 2, and 3 were significantly higher than APOE ɛ4 genotype. The AUC of model 3 was the highest (AUC = 0.946) with both high sensitivity (88.2%) and high specificity (88.1%). In the subgroup, model 4 (demographics and CSF biomarkers) also has good predictive performance. The AUC, sensitivity, and specificity for model 4 were 0.936, 94.1%, and 83.0%.

Fig. 3

Receiver operating characteristic (ROC) analysis of demographic combined with other model combinations. ROC was performed in demographic combined with other model combinations to select a combination of screening variables for positive prediction of tau-PET in the main cohort (A) and CSF sub-cohort (B). A) Model 1: ROC curve of APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM combined. Model 2: ROC curve of APOE ɛ4, thickness average of right middle temporal, volumes of left middle temporal and surface area of left middle temporal combined. Model 3: ROC curve of APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM, thickness average of right middle temporal, volumes of left middle temporal and surface area of left middle temporal combined. B) Model 4: ROC curve of APOE ɛ4, gender, Aβ₄₀, p-tau, t-tau and Aβ_42/40 combined.

Table 5

Diagnostic values of multiple indicators combined in predicating tau-PET positivity in MCI

Model	AUC	95% CI	Sensitivity (%)	Specificity (%)	Accuracy (%)
Model 1	0.924	0.863–0.986	76.5	94.0	90.5
Model 2	0.868	0.756–0.981	70.6	73.6	72.9
Model 3	0.946	0.894–0.997	88.2	88.1	88.1
Model 4	0.936	0.881–0.990	94.1	83.0	85.7

Model 1 means the combination of APOE ɛ4, ADAS-Cog13, MMSE, and ADNI-MEM. Model 2 means the combination of APOE ɛ4, thickness average of right middle temporal, volumes of left middle temporal and surface area of left middle temporal. Model 3 means the combination of APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM, thickness average of right middle temporal, volumes of left middle temporal and surface area of left middle temporal. Model 4 means the combination of APOE ɛ4, gender, Aβ₄₀, p-tau, t-tau, and Aβ_42/40. AUC, area under curve of receiver operating characteristic; CI, confidence intervals.

Development of a nomogram for the diagnosis of tau-PET status

A nomogram for risk assessment of tau-PET (+) was further established by using the variables of the model 3 with the best AUC (Fig. 4), including APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM, and structural MRI in the middle temporal (thickness average of right, volumes of left, surface area of left). In a nomogram, values for a particular patient were positioned along each variable axis, and a vertical line was drawn on the point axis to obtain a score for each variable. Add the score for each variable to obtain the total score, which was shown on the main branch at the bottom of the nomogram. We would then obtain the probability by plotting a vertical line from the total score to the risk axis. DCA results showed that model 3 had significant positive net benefits in risk prediction, which confirmed its good clinical application value (Fig. 5).

Fig. 4

Nomogram predicting tau-PET in the main cohort. Each variable in the figure, such as APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM, thickness average of right middle temporal, volumes of left middle temporal and surface area of left middle temporal in this study. correspond to the points at the top, and the sum of the scores of all variables is total points. Different total points correspond to the occurrence probability of tau-PET positive at the bottom.

Fig. 5

Decision curve analysis (DCA) for the prediction model 3 in the main cohort. The y-axis represents the net benefits, while the x-axis represents the threshold probability. The red solid line represents the probability derived from the prediction model, while the gray line presents the assumption that no patients were diagnosed as tau-PET positive. The horizontal black line represents the assumption that all patients were diagnosed as tau-PET positive.

DISCUSSION

The aim of this study was to find a simple and effective method to predict tau-PET status in MCI individuals by stepwise regression. We first determined which combinations of variables among the four modalities (demographic, structural MRI, neurocognitive measures, and CSF biomarkers) accurately predicted tau-PET status, and then examined the discriminative power of the individual variables selected from each combination. Finally, the best model of predictor variables is determined. The major findings were as follows: 1) CSF biomarkers were the most accurate biomarkers for predicting tau-PET status in clinically diagnosed MCI individuals. 2) In comparison to CSF biomarkers, neurocognitive assessment that combined ADAS-Cog13, MMSE, and ADNI-MEM also had better predictive value in predicting tau-PET status. 3) Single marker analysis showed that APOE ɛ4 genotype, ADAS-Cog13, or thickness average of right middle temporal are also accurate predictors of tau-PET status. IV) Combining APOE ɛ4, ADAS-Cog13, MMSE, ADNI-MEM, and structural MRI in the middle temporal (thickness average of right, volumes of left, surface area of left) was determined to be the best predictive model.

CSF p-tau and tau-PET imaging are biomarker for detecting phosphorylated-tau in vivo. Tau-PET makes tau protein detection in vivo basically consistent with autopsy results [21]. In this study, through the analysis of different types of data, it was found that the combination of Aβ₄₀, p-tau, t-tau, and Aβ_42/40 can accurately predict tau-pet status, and its AUC is the highest, and univariate analysis of CSF biomarkers found that Aβ_42/40, but not p-tau, was the strongest predictor of tau-PET status. This is consistent with the literature on the association of Aβ and tau pathology, with Aβ promoting neocortical spread of tau pathology [22], and the spread of this Aβ-dependent tau pathology is the driving factor behind cognitive decline [23]. CSF p-tau with higher sensitivity and tau-PET imaging with persistent accumulation may capture distinct states of tau pathology [21, 24]. CSF p-tau, as a marker to identify the “disease state”, is elevated early in the pathophysiological process of AD [25], and tau-PET imaging, as the signal accumulates, is a powerful marker of “disease stage” [21, 26]. A recent TRIAD cohort study showed a low correlation between CSF p-tau₁₈₁, p-tau₂₁₇, p-tau₂₃₁, p-tau₂₃₅, and tau-PET in the cerebral cortex [27]. Another study using stable isotope labeling kinetics in humans showed that soluble p-tau production was positively correlated with amyloid-PET signaling but did not change when tau PET was elevated [28]. Therefore, CSF p-tau may be less accurate than Aβ_42/40 in differentiating tau-PET at the MCI stage, as shown in the present study.

However, the acquisition of CSF and tau-PET has certain limitations in clinical practice. This study demonstrates that the neurocognitive assessment approach has acceptable discriminatory properties and are superior to other methods alone, including neuroimaging, demographics, and genetics. As a microtubule-stabilizing protein, tau protein has a reduced ability to stabilize microtubules in pathological conditions, resulting in impaired axonal transport, which can further lead to synaptic loss and neurodegeneration, resulting in massive network damage and impaired cognitive function [29]. Memory and executive domains have been widely demonstrated to be sensitive indicators of early cognitive decline in AD [30]. Previous studies have shown that APOE ɛ4 allele status and MMSE score can affect tau-PET status and are significantly correlated with tau-PET level [31]. A retrospective longitudinal study showed that when Aβ-PET, tau-PET, and hippocampal volume were analyzed in a single statistical model, only tau-PET was an important predictor of overall cognitive decline. However, in this study, the best model obtained in this study shows that the combined ADNI-MEM reflecting memory function, ADAS-Cog13, and MMSE reflecting global cognition had strong predictive ability. In each cognitive domain, tau-PET had the greatest negative impact on the test of episodic memory and executive function [32]. This study also expressed similar results. Compared with tau-PET (–), tau-PET (+) had lower execution, memory, and overall cognitive scores. More and more evidence also show that the increase of tau-PET signal may reflect the overall accumulation of brain pathology, which is related to subsequent brain atrophy and cognitive decline [33]. Among them, ADAS-Cog13 can have the best predictive performance, which suggests important clinical applications. Compared with MMSE, ADAS-Cog 13 increased the attentional component and executive function of the number-cancellation item and delayed recall task, thereby providing a more comprehensive reflection of the global cognitive state, which explains the better association with tau. Therefore, choosing a single indicator to predict tau-PET in MCI individuals, ADAS-Cog 13 may be a cost-effective alternative.

It is well known that structural MRI, as a simple and easy test, is the most commonly used technique to identify brain atrophy associated with AD. The entorhinal cortex plays a central role in the early emergence of tau, but it may be the inferior temporal cortex, a key region for rapid tau accumulation in preclinical AD [34]. Evidence from autopsy suggests that the medial temporal lobe is the site of early tau accumulation. Interestingly, the accumulation of tau pathology was directly associated with cognitive dysfunction in the medial temporal brain region, regardless of the presence or absence of Aβ [35]. There is a local association between tau and atrophy [36]. Cortical thickness alone is not accurate enough to measure atrophy, so volumes, surface area, thickness average was also segmented in this study to investigate different atrophy patterns. We found that thickness average of right middle temporal, volumes of left middle temporal, surface area of left middle temporal could significantly distinguish tau-PET states and were better predictors. When structural MRI imaging of middle temporal and the presence of APOE ɛ4 status were combined, the prediction performance was substantially enhanced. Studies have shown that the introduction of APOE ɛ4 leads to deterioration of tau pathology and neurodegeneration [37]. Greater tau pathology may be responsible for the medial temporal neurodegeneration observed in APOE ɛ4 carriers [38].

The most powerful genetic risk factor for sporadic AD is known to be the APOE ɛ4 allele, and the risk rises in a dose-dependent manner [39]. Homozygous ɛ4 carriers had a more extensive pattern of cerebral tau pathology than heterozygous ɛ4 carriers [40]. Studies have demonstrated that APOE ɛ4 is essential for enhanced tau phosphorylation in both neuronal and astrocytic cells. Tau phosphorylation in neurons may be accelerated by astrocyte APOE ɛ4, and neuronal APOE ɛ4 may inhibit p-tau transport to neuroastrocytes [41]. Additionally, APOE ɛ4 was significantly associated with tau burden at all disease stages, with the strongest association in CN and MCI subjects [42]. Our study showed similar results, with a significantly higher proportion of homozygous ɛ4 carriers in tau-PET (+) than in tau-PET (–), and APOE ɛ4 genotype is useful predictors of tau positivity in MCI. However, the value of gender in predicting tau positivity is debatable.

The combination of APOE ɛ4 genotype, neurocognitive measures, and structural MRI imaging of the middle temporal multi-clinical marker model improved tau status prediction accuracy and expanded the knowledge in this field. In addition, we present a nomogram for predicting tau-PET status in clinically diagnosed individuals with MCI. It facilitates rapid calculation of the risk of tau-PET positivity based on the point of the predictor variable in clinical practice. The nomogram is non-invasive and inexpensive. It has been shown that tau-PET over Aβ-PET is a promising tool for predicting cognitive change and may support the prognostic course of AD in preclinical and prodromal phases [43]. Therefore, our model may be useful for identifying patients with MCI with tau pathology to consider the prognostic course and treatment.

Our study has several limitations. First, we used stepwise regression algorithm to screen variables and some combinations of variables, which has high predictive efficiency and has certain clinical significance. However, it does not exclude the existence of other variable combinations with similar or higher predictive performance. While we identified ADAS-Cog13, MMSE, and ADNI-Memory summary score with good predictive accuracy of tau-PET status, this combination of cognitive test may not be feasible or practical in clinical practice. The components of the nomogram are also not routinely performed in clinical practice. Second, we only analyzed the data of cognitive scores, CSF biomarkers, demographic indicators, and structural MRI in the ADNI database. Future studies can further investigate plasma biomarkers with unified specifications and clinical application prospects. Third, we have included all MCI participants with tau-PET imaging in ADNI in the analysis, but the sample size is relatively small due to the small amount of cerebrospinal fluid data only used as a sub-cohort study, which may lead to bias. Fourth, the study used tau-PET as the gold standard for brain tau-positivity, rather than the true gold standard for neuropathology. This prediction model can predict tau-PET status well, but the prediction of tau deposition progression by stages needs further study.

In conclusion, the combination of APOE ɛ4, neurocognitive measures (ADAS-Cog13, MMSE, ADNI-MEM) and structural MRI in the middle temporal (thickness average of right, volumes of left, surface area of left) accurately predicted tau-PET status in individuals with clinically diagnosed MCI. This finding provides a noninvasive, cost-effective, and time-saving tool that provides an attractive alternative assessment for the prediction of tau-PET status which can be used to predict cognitive decline in individuals with MCI. The model may accelerate the recruitment of individuals with potential risk of tau-PET positivity, thereby facilitating clinical drug development for AD.

Footnotes

ACKNOWLEDGMENTS

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through 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. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

FUNDING

The study was funded by Shandong Provincial key research and development project (2018GSF118235) and Chinese National Natural Science Foundation (no. 81571234).

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

Data sharing is not applicable to this article as no datasets were generated or analyzed during this study.

References

DeTure

, Dickson

(2019) The neuropathological diagnosis of Alzheimer’s disease. Mol Neurodegener 14, 32.

Scheltens

, De Strooper

, Kivipelto

, Holstege

, Chetelat

, Teunissen

, Cummings

, van der Flier

(2021) Alzheimer’s disease. Lancet 397, 1577–1590.

Bruandet

, Richard

, Bombois

, Maurage

, Deramecourt

, Lebert

, Amouyel

, Pasquier

(2009) Alzheimer disease with cerebrovascular disease and vascular dementia: Clinical features and course compared with Alzheimer disease. J Neurol Neurosurg Psychiatry 80, 133–139.

Sperling

, Aisen

, Beckett

, Bennett

, Craft

, Fagan

, Iwatsubo

, Jack

, Kaye

, Montine

, Park

, Reiman

, Rowe

, Siemers

, Stern

, Yaffe

, Carrillo

, Thies

, Morrison-Bogorad

, Wagster

, Phelps

(2011) Toward defining the preclinical stages of Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 280–292.

Hansson

, Seibyl

, Stomrud

, Zetterberg

, Trojanowski

, Bittner

, Lifke

, Corradini

, Eichenlaub

, Batrla

, Buck

, Zink

, Rabe

, Blennow

, Shaw

, Swedish Bio Fsg, Alzheimer’s Disease Neuroimaging Initiative (2018) CSF biomarkers of Alzheimer’s disease concord with amyloid-beta PET and predict clinical progression: A study of fully automated immunoassays in BioFINDER and ADNI cohorts. Alzheimers Dement 14, 1470–1481.

Schilling

, Zimmer

, Shin

, Leuzy

, Pascoal

, Benedet

, Borelli

, Palmini

, Gauthier

, Rosa-Neto

(2016) Imaging Alzheimer’s disease pathophysiology with PET. Dement Neuropsychol 10, 79–90.

Holtzman

(2011) CSF biomarkers for Alzheimer’s disease: Current utility and potential future use. Neurobiol Aging 32, S4–S9.

, Ng

, Gao

, Kong

, Guan

, Yu

, Alzheimer’s Disease Neuroimaging Initiative (2019) The combination of apolipoprotein E4, age and Alzheimer’s Disease Assessment Scale-Cognitive Subscale improves the prediction of amyloid positron emission tomography status in clinically diagnosed mild cognitive impairment. Eur J Neurol 26, 733–e53.

Tosun

, Veitch

, Aisen

, Jack

Jr. , Jagust

, Petersen

, Saykin

, Bollinger

, Ovod

, Mawuenyega

, Bateman

, Shaw

, Trojanowski

, Blennow

, Zetterberg

, Weiner

(2021) Detection of beta-amyloid positivity in Alzheimer’s Disease Neuroimaging Initiative participants with demographics, cognition, MRI and plasma biomarkers. Brain Commun 3, fcab008.

10.

Selkoe

(1991) The molecular pathology of Alzheimer’s disease. Neuron 6, 487–498.

11.

Jack

Jr. , Bennett

, Blennow

, Carrillo

, Dunn

, Haeberlein

, Holtzman

, Jagust

, Jessen

, Karlawish

, Liu

, Molinuevo

, Montine

, Phelps

, Rankin

, Rowe

, Scheltens

, Siemers

, Snyder

, Sperling R, Contributors (2018) NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement 14, 535–562.

12.

Ossenkoppele

, Leuzy

, Cho

, Sudre

, Strandberg

, Smith

, Palmqvist

, Mattsson-Carlgren

, Olsson

, Jögi

, Stormrud

, Ryu

, Choi

; Alzheimer’s Disease Neuroimaging Initiative; PREVENT-AD research group; Boxer

, Gorno-Tempini

, Miller

, Soleimani-Meigooni

, Iaccarino

, La Joie

, Borroni

, Klein

, Pontecorvo

, Devous

Sr , Villeneuve

, Lyoo

, Rabinovici

, Hansson

(2021) The impact of demographic, clinical, genetic, and imaging variables on tau PET status. Eur J Nucl Med Mol Imaging 48, 2245–2258.

13.

Fischl

, Salat

, Busa

, Albert

, Dieterich

, Haselgrove

, van der Kouwe

, Killiany

, Kennedy

, Klaveness

, Montillo

, Makris

, Rosen

, Dale

(2002) Whole brain segmentation: Automated labeling of neuroanatomical structures in the human brain. Neuron 33, 341–355.

14.

Greve

, Svarer

, Fisher

, Feng

, Hansen

, Baare

, Rosen

, Fischl

, Knudsen

(2014) Cortical surface-based analysis reduces bias and variance in kinetic modeling of brain PET data. Neuroimage 92, 225–236.

15.

Jack

Jr. , Wiste

, Weigand

, Therneau

, Lowe

, Knopman

, Gunter

, Senjem

, Jones

, Kantarci

, Machulda

, Mielke

, Roberts

, Vemuri

, Reyes

, Petersen

(2017) Defining imaging biomarker cut points for brain aging and Alzheimer’s disease. Alzheimers Dement 13, 205–216.

16.

Maass

, Landau

, Baker

, Horng

, Lockhart

, La Joie

, Rabinovici

, Jagust

, Alzheimer’s Disease Neuroimaging Initiative (2017) Comparison of multiple tau-PET measures as biomarkers in aging and Alzheimer’s disease. Neuroimage 157, 448–463.

17.

Jack

Jr. , Wiste

, Therneau

, Weigand

, Knopman

, Mielke

, Lowe

, Vemuri

, Machulda

, Schwarz

, Gunter

, Senjem

, Graff-Radford

, Jones

, Roberts

, Rocca

, Petersen

(2019) Associations of amyloid, tau, and neurodegeneration biomarker profiles with rates of memory decline among individuals without dementia. JAMA 321, 2316–2325.

18.

Bryant

, Giovanello

, Ibrahim

, Chang

, Shen

, Peterson

, Zhu

, Alzheimer’s Disease Neuroimaging Initiative (2013) Mapping the genetic variation of regional brain volumes as explained by all common SNPs from the ADNI study. PLoS One 8, e71723.

19.

Desikan

, Sabuncu

, Schmansky

, Reuter

, Cabral

, Hess

, Weiner

, Biffi

, Anderson

, Rosand

, Salat

, Kemper

, Dale

, Sperling

, Fischl

, Alzheimer’s Disease Neuroimaging Initiative (2010) Selective disruption of the cerebral neocortex in Alzheimer’s disease. PLoS One 5, e12853.

20.

Shaw

, Vanderstichele

, Knapik-Czajka

, Clark

, Aisen

, Petersen

, Blennow

, Soares

, Simon

, Lewczuk

, Dean

, Siemers

, Potter

, Lee

, Trojanowski

, Alzheimer’s Disease Neuroimaging Initiative (2009) Cerebrospinal fluid biomarker signature in Alzheimer’s disease neuroimaging initiative subjects. Ann Neurol 65, 403–413.

21.

Scholl

, Maass

, Mattsson

, Ashton

, Blennow

, Zetterberg

, Jagust

(2019) Biomarkers for tau pathology. Mol Cell Neurosci 97, 18–33.

22.

Busche

, Hyman

(2020) Synergy between amyloid-beta and tau in Alzheimer’s disease. Nat Neurosci 23, 1183–1193.

23.

Lockhart

, Scholl

, Baker

, Ayakta

, Swinnerton

, Bell

, Mellinger

, Shah

, O’Neil

, Janabi

, Jagust

(2017) Amyloid and tau PET demonstrate region-specific associations in normal older people. Neuroimage 150, 191–199.

24.

Mattsson-Carlgren

, Leuzy

, Janelidze

, Palmqvist

, Stomrud

, Strandberg

, Smith

, Hansson

(2020) The implications of different approaches to define AT(N) in Alzheimer disease. Neurology 94, e2233–e2244.

25.

Mattsson-Carlgren

, Janelidze

, Bateman

, Smith

, Stomrud

, Serrano

, Reiman

, Palmqvist

, Dage

, Beach

, Hansson

(2021) Soluble P-tau217 reflects amyloid and tau pathology and mediates the association of amyloid with tau. EMBO Mol Med 13, e14022.

26.

Guo

, Huang

, Shen

, Chen

, Hu

, Wang

, Tan

, Yu

, Alzheimer’s Disease Neuroimaging Initiative (2021) Characterization of Alzheimer’s tau biomarker discordance using plasma, CSF, and PET. Alzheimers Res Ther 13, 93.

27.

Therriault

, Vermeiren

, Servaes

, Tissot

, Ashton

, Benedet

, Karikari

, Lantero-Rodriguez

, Brum

, Lussier

, Bezgin

, Stevenson

, Rahmouni

, Kunach

, Wang

, Fernandez-Arias

, Socualaya

, Macedo

, Ferrari-Souza

, Ferreira

PCL

, Bellaver

, Leffa

, Zimmer

, Vitali

, Soucy

, Triana-Baltzer

, Kolb

, Pascoal

, Saha-Chaudhuri

, Gauthier

, Zetterberg

, Blennow

, Rosa-Neto

(2023) Association of phosphorylated tau biomarkers with amyloid positron emission tomography vs tau positron emission tomography. JAMA Neurol 80, 188–199.

28.

Sato

, Barthelemy

, Mawuenyega

, Patterson

, Gordon

, Jockel-Balsarotti

, Sullivan

, Crisp

, Kasten

, Kirmess

, Kanaan

, Yarasheski

, Baker-Nigh

, Benzinger

TLS

, Miller

, Karch

, Bateman

(2018) Tau kinetics in neurons and the human central nervous system. Neuron 97, 1284–1298.e1287.

29.

Menkes-Caspi

, Yamin

, Kellner

, Spires-Jones

, Cohen

, Stern

(2015) Pathological tau disrupts ongoing network activity. Neuron 85, 959–966.

30.

Palmqvist

, Tideman

, Cullen

, Zetterberg

, Blennow

, Alzheimer’s Disease Neuroimaging I, Dage

, Stomrud

, Janelidze

, Mattsson-Carlgren

, Hansson

(2021) Prediction of future Alzheimer’s disease dementia using plasma phospho-tau combined with other accessible measures. Nat Med 27, 1034–1042.

31.

Tan

, Yang

, Wang

, Xu

, Tan

, Zuo

, Dong

, Tan

, Yu

, Alzheimer’s Disease Neuroimaging I (2020) PET amyloid and tau status are differently affected by patient features. J Alzheimers Dis 78, 1129–1136.

32.

Aschenbrenner

, Gordon

, Benzinger

TLS

, Morris

, Hassenstab

(2018) Influence of tau PET, amyloid PET, and hippocampal volume on cognition in Alzheimer disease. Neurology 91, e859–e866.

33.

Bejanin

, Schonhaut

, La Joie

, Kramer

, Baker

, Sosa

, Ayakta

, Cantwell

, Janabi

, Lauriola

, O’Neil

, Gorno-Tempini

, Miller

, Rosen

, Miller

, Jagust

, Rabinovici

(2017) Tau pathology and neurodegeneration contribute to cognitive impairment in Alzheimer’s disease. Brain 140, 3286–3300.

34.

Insel

, Young

, Aisen

, Johnson

, Sperling

, Mormino

, Donohue

(2023) Tau positron emission tomography in preclinical Alzheimer’s disease. Brain 146, 700–711.

35.

Therriault

, Pascoal

, Sefranek

, Mathotaarachchi

, Benedet

, Chamoun

, Lussier

, Tissot

, Bellaver

, Lukasewicz

, Zimmer

, Saha-Chaudhuri

, Gauthier

, Rosa-Neto

, Alzheimer’s Disease Neuroimaging Initiative (2021) Amyloid-dependent and amyloid-independent effects of Tau in individuals without dementia. Ann Clin Transl Neurol 8, 2083–2092.

36.

Timmers

, Ossenkoppele

, Wolters

, Verfaillie

SCJ

, Visser

, Golla

SSV

, Barkhof

, Scheltens

, Boellaard

, van der Flier

, van Berckel

BNM

(2019) Associations between quantitative [(18)F]flortaucipir tau PET and atrophy across the Alzheimer’s disease spectrum. Alzheimers Res Ther 11, 60.

37.

Wadhwani

, Affaneh

, Van Gulden

, Kessler

(2019) Neuronal apolipoprotein E4 increases cell death and phosphorylated tau release in Alzheimer disease. Ann Neurol 85, 726–739.

38.

Therriault

, Benedet

, Pascoal

, Mathotaarachchi

, Chamoun

, Savard

, Thomas

, Kang

, Lussier

, Tissot

, Parsons

, Qureshi

MNI

, Vitali

, Massarweh

, Soucy

, Rej

, Saha-Chaudhuri

, Gauthier

, Rosa-Neto

(2020) Association of Apolipoprotein E epsilon4 with medial temporal tau independent of amyloid-beta. JAMA Neurol 77, 470–479.

39.

Saunders

, Jenkins

, Blennow

, Ritchie

, Muniz-Terrera

(2022) Interactions between apolipoprotein E, sex, and amyloid-beta on cerebrospinal fluid p-tau levels in the European prevention of Alzheimer’s dementia longitudinal cohort study (EPAD LCS). EBioMedicine 83, 104241.

40.

Therriault

, Benedet

, Pascoal

, Mathotaarachchi

, Savard

, Chamoun

, Thomas

, Kang

, Lussier

, Tissot

, Soucy

, Massarweh

, Rej

, Saha-Chaudhuri

, Poirier

, Gauthier

, Rosa-Neto

, Alzheimer’s Disease Neuroimaging Initiative (2021) APOEepsilon4 potentiates the relationship between amyloid-beta and tau pathologies. Mol Psychiatry 26, 5977–5988.

41.

Huang

, Zhang

, Cao

, Yu

, Pei

(2022) Chimeric cerebral organoids reveal the essentials of neuronal and astrocytic APOE4 for Alzheimer’s tau pathology. Signal Transduct Target Ther 7, 176.

42.

Stage

, Risacher

, Lane

, Gao

, Nho

, Saykin

, Apostolova

, Alzheimer’s Disease Neuroimaging Initiative (2022) Association of the top 20 Alzheimer’s disease risk genes with [(18)F]flortaucipir PET. Alzheimers Dement (Amst) 14, e12308.

43.

Ossenkoppele

, Smith

, Mattsson-Carlgren

, Groot

, Leuzy

, Strandberg

, Palmqvist

, Olsson

, Jogi

, Stormrud

, Cho

, Ryu

, Choi

, Boxer

, Gorno-Tempini

, Miller

, Soleimani-Meigooni

, Iaccarino

, La Joie

, Baker

, Borroni

, Klein

, Pontecorvo

, Devous

Sr. , Jagust

, Lyoo

, Rabinovici

, Hansson

(2021) Accuracy of tau positron emission tomography as a prognostic marker in preclinical and prodromal Alzheimer disease: A head-to-head comparison against amyloid positron emission tomography and magnetic resonance imaging. JAMA Neurol 78, 961–971.