Modifiable Factors Associated with the Longitudinal Increase and Spatial Extent of Tau Pathology in Alzheimer Disease

Visual Abstract

Abstract

There are 14 modifiable factors that are associated with a significantly lower risk of dementia. We tested the interactive effect of modifiable factors, genetic determinants, and initial pathologic burden on the spatial progression and local amplification of tau pathology. Methods: In total, 162 amyloid-positive individuals were included, for whom longitudinal [¹⁸F]AV-1451 PET scans, baseline information on global amyloid burden, ApoE4 status, body mass index (BMI), hypertension, education, neuropsychiatric symptom severity, and demographic information were available in the Alzheimer Disease Neuroimaging Initiative. All [¹⁸F]AV-1451 scans were intensity-standardized (reference: inferior cerebellum), z-transformed (control sample: 147 amyloid-negative subjects), thresholded (z score, >1.96), and converted to volume maps. Longitudinal tau changes were then assessed in terms of tau spatial extent (i.e., newly affected volume at follow-up) and tau level rise (i.e., tau increase in previously affected volume). These 2 measures were entered as dependent variables in linear mixed-effects models, including baseline modifiable risk factors (BMI, education, hypertension, neuropsychiatric symptom severity), global amyloid, tau volume or tau burden, ApoE4 status, clinical stage, sex, and age as predictors. Next, we tested the interactive effects between baseline amyloid or tau burden with the 4 modifiable factors on tau extent or tau level rise, respectively. Results: Greater tau extent was linked to higher BMI (β = 0.002; 95% CI, 0.0003–0.003), ApoE4 status (β = 0.024; 95% CI, 0.001–0.046), and baseline tau volume (β = 0.207; 95% CI, 0.107–0.308) across groups. In terms of tau level rise, we observed that absence of hypertension (β = 0.295; 95% CI, −0.477 to −0.114), dementia group (β = 0.305; 95% CI, 0.088–0.522), and BMI (β = 0.011; 95% CI, 0.00004–0.022) were linked to increased tau burden. A load-dependent effect of baseline amyloid and tau volume/burden was found for both tau extent (β = −0.005; 95% CI, −0.008 to −0.002) and tau level rise (β = −0.003; 95% CI, −0.005 to −0.001). Higher amyloid and BMI (β = 0.001; 95% CI, 0.0004–0.001) and lower education and higher tau burden (β = −0.035; 95% CI, −0.064 to −0.006) were linked to greater tau level rise. Conclusion: Education, BMI, and hypertension differentially influence tau’s spatial extent and increase by its interaction with initial pathologic burden. Timely modification of these factors may overall slow tau progression.

Nearly half of dementia cases may be preventable by modifying 14 risk factors (1). These factors could thus influence the onset and progression of Alzheimer disease (AD). PET imaging shows tau pathology is more closely tied to symptoms than amyloid (2), yet amyloid remains the strongest predictor of tau progression (3). However, little is known about how potentially modifiable risk factors interact with baseline pathology burden and subsequent tau spreading. Prior studies, for instance, suggest that higher education (4), neuropsychiatric symptoms (5), higher body mass index (BMI) (6), and hypertension (7) are associated with greater AD pathology.

In this study, using data availability of the Alzheimer Disease Neuroimaging Initiative (ADNI) 3 cohort, we focused on examining the association between tau pathology spread and 4 risk factors (education, neuropsychiatric symptom severity [NSS], BMI, hypertension), which are potentially modifiable through lifestyle and pharmacologic interventions. Specifically, we tested the contribution of these factors to the spatial progression of tau pathology (i.e., tau spatial extent) and the local increase in tau burden (i.e., tau level rise), anticipating that healthier lifestyle profiles would slow tau progression (e.g., lower BMI, higher education, and no hypertension or neuropsychiatric symptoms), whereas baseline amyloid and tau burden and ApoE4 carriership (i.e., established contributors) would be inversely related with tau spatial extent and level rise. We further expected that protective effects would be driven by an interaction with lower baseline amyloid burden, rather than baseline tau burden, presuming that amyloid burden is the key driver of tau spread.

MATERIALS AND METHODS

Participants

Data for this study were retrieved from the ADNI (adni.loni.usc.edu) in July 2024. For the current analysis, the key inclusion criteria included at least 2 evaluable [¹⁸F]AV-1451 PET scans; baseline amyloid PET scan within 3 mo of the baseline tau PET acquisition and confirmed amyloid positivity; ApoE4 carriership information; and available baseline information on BMI, hypertension, education, and neuropsychiatric inventory (NPI) in addition to demographic information. This resulted in 162 individuals fulfilling the criteria, who were grouped according to the diagnostic category at the baseline tau PET neuroimaging session, resulting in 77 cognitively unimpaired (CU), 55 patients with mild cognitive impairment (MCI) (8), and 30 patients with AD (9). More detailed information on the diagnostic grouping can be found in the supplemental materials (available at http://jnm.snmjournals.org). Demographic characteristics of the respective groups are provided in Table 1. In addition, an amyloid-negative cognitively normal group was included, for which 2 tau PET scans and a baseline amyloid scan were available (n = 147; mean age (y), 71.63 ± 6.32 y; mean education (y), 16.92 ± 2.38; mean AD Assessment Scale Cognitive 13 score, 7.56 ± 4.28; 67 men and 80 women; ApoE4 (45 yes, 100 no, 2 missing). This group was used as reference group for the z-transformation of the tau PET scans. Ethics committee approval was obtained at each of the participating centers of ADNI. Participants’ informed consent was obtained from each participating centers.

TABLE 1.

Demographic Characteristics of the Study Cohort

Demographics	CU positive	MCI	AD	Group differences	P
n	77	55	30
Baseline age (y)	74.67 (7.07)	73.50 (7.10)	75.92 (8.83)	H(2, N = 162) = 2.53	0.282
Sex				χ2(2, N = 162) = 7.39	0.025
M	28	30	20
F	49	25	10
ApoE4				χ2(2, N = 162) = 9.31	0.010
No	34	12	9
Yes	43	43	21
ADAS13 (points)	8.97 (4.91)	16.66 (6.46)	29.06 (8.74)	H(2, N = 162) = 89.20	<0.001
Education (y)	16.53 (2.29)	16.16 (2.54)	15.73 (2.49)	H(2, N = 162) = 2.70	0.259
BMI	28.79 (6.22)	27.32 (6.86)	26.88 (5.54)	H(2, N = 162) = 4.28	0.118
Hypertension				χ2(3, N = 162) = 10.67	0.031
Ideal	17	22	3
Intermediate	30	19	14
Poor	30	14	13
NPI (score)	1.09 (2.41)	3.40 (5.20)	7.70 (6.56)	H(2, N = 162) = 42.74	<0.001
Global amyloid load (centiloids)	65.56 (35.92)	76.49 (31.66)	102.47 (35.97)	H(2, N = 162) = 20.82	<0.001
Mean baseline tau load (z score)	2.54 (.35)	3.14 (1.09)	4.07 (1.97)	H(2, N = 162) = 34.52	<0.001
Mean baseline volume (L)	0.09 (0.11)	0.27 (0.31)	0.49 (0.42)	H(2, N = 162) = 31.73	<0.001

ADAS13 = AD Assessment Scale Cognitive 13 score.

Mean and SD are provided for continuous and distributions for ordinal variables. Significant group differences were tested using a Kruskal–Wallis test for continuous and χ² test for ordinal variables.

Risk Factors

Education

Educational attainment comprised years of formal school education in addition to following higher education (i.e., college, university).

BMI

Height (inches or centimeters) and weight (pounds or kilograms) were converted to meters and kilograms, respectively. BMI was calculated as (weight in kilograms)/(height in meters)².

Hypertension

Hypertension information was derived based on the systolic blood pressure (SBP) and diastolic blood pressure (DBP) assessed at the respective neuroimaging baseline visit. The following groups were established using the cutoff values based on the recommendations of the American Heart Association: <120 mm Hg/<80 mm Hg is ideal; SBP of 120–139 mm Hg or DBP of 80–89 mm Hg is intermediate; SBP of ≥140 mm Hg or DBP ≥90 mm Hg is poor.

NPI

The NPI entails 10 behavioral areas (e.g., depression, apathy, anxiety, etc.) and 2 neurovegetative areas (e.g., sleep, eating disorder). Symptoms need to be present for at least 4 wk, and scoring is based on frequency ranging from 1 to 4 and severity ranging from 1 to 3. A total NPI score is then calculated by adding the scores (frequency by severity) of the 12 domains, which was used in this study to approximate NSS.

PET Image Processing

Preprocessed tau PET images provided by ADNI were used, including coregistration to the individual MRI, normalization to MNI space and intensity-normalization using the inferior cerebellar gray. The respective tau PET scans (baseline and follow-up [FU]) were z-transformed using the mean of the baseline and first FU tau PET scans of the amyloid-negative control group to control for repeated measurement effects. Information on global amyloid load at baseline was extracted based on the centiloid information provided in the data sheet “UCBERKELEY_AMY_ 6MM.” We used centiloids as a tracer-independent measure for global amyloid burden as either [¹⁸F]florbetapir or [¹⁸F]florbetaben PET scans were available.

Volumetric Approach

We performed a volumetric approach to quantify the spatial extent and level rise (increase) of tau pathology over time within the cortical gray matter. First, we thresholded the tau z-maps at a z score of greater than 1.96 (1-tailed P < 0.025) and subsequently created binarized volume maps comprising only voxels above this threshold within the cortical gray matter mask. Then, we defined 2 respective volumes for each FU time point (FU_tx):

1. Newly affected volume = binarized z-map FU_tx − binarized z-map of baseline tau scan; FU_tx volume – baseline volume. Consequently, the newly affected volume reflects all regions that were not affected at baseline but only at FU. The number of voxels were then translated into volume in liters (voxel number × (voxel size of 1.5 mm)³/1,000,000). This measure was used as proxy for the spatial extent of tau pathology over time.

2. More affected volume = regions of spatial overlap at FU_tx with previous time point; FU_{Volume_tx} ∩ FU_{Volume_tx-1}. This volume was used to quantify the increase in tau pathology over time in regions that had already been affected at the previous time point. The z score change in the more affected volume was extracted by taking the mean difference of FU_tx and baseline FU within the more affected volume between the 2 time points. The z score changes were used to quantify tau level rise, which represents the increase of tau burden over time.

Additionally, the baseline volume and mean z score were extracted for each individual.

Statistical Analyses

Correlation Analyses and Group Comparisons to Test Baseline Associations

To investigate the baseline relationships among variables of interest—BMI, educational attainment, NSS, baseline centiloids, and mean baseline tau z score—we used partial Spearman correlations (ρ) (1-tailed), adjusted for age and sex. Due to nonnormality of the data, we used rank-based ANCOVA to compare the hypertension and ApoE4 groups in terms of centiloids, baseline tau burden/volume, BMI, education, and NSS, corrected for age and sex. A χ² test was used to compare the distribution of ApoE4 and hypertension severity. Correction for multiple comparisons was performed using Benjamini–Hochberg method (false discovery rate) within each of the tested models (Spearman correlations and the 2 rank-based ANCOVA for ApoE4 and hypertension severity, separately).

Linear Mixed Model (LMM) Effects

To examine the effects of the (non-)modifiable factors to the spatial extent and level rise of tau pathology, we evaluated LMM effects in SPSS 28 (IBM Corp.). The models included repeated measures of either the newly affected volume or increase in tau burden over time and an unstructured covariance matrix to account for individual intercepts and slopes. Model parameters were estimated using the restricted maximum likelihood, which provides less biased estimates of variance components.

The first LMM included the newly affected volume (i.e., tau spatial extent) of up to 5 y of follow-up data (model 1: tau spatial extent). The following continuous fixed effects were introduced: age at baseline, time difference in months from the baseline tau scan acquisition, baseline centiloids, baseline tau volume, education years, BMI, and NPI total score. Ordinal fixed effects comprised sex (F/M), ApoE4 carriership (yes/no), group status (CU, MCI, AD), and hypertension (ideal, intermediate, poor). In addition, the model included 2 random effects (subject and time) that allowed individual intercepts and slopes using an unstructured covariance matrix. Given our a priori hypotheses, we subsequently added the following interaction terms to the same model: baseline centiloids and baseline tau volume with the 4 risk factors, in addition to the interaction between hypertension severity and ApoE4 status given the previously reported interaction between these 2 variables on amyloid burden (10).

The second LMM included the z score change over time in months from the baseline tau scan acquisition in the more affected volume as a dependent variable (model 2: tau level rise). The fixed effects remained the same as in model 1, except that we included baseline tau burden instead of baseline tau volume as fixed effects since we focused on tau intensity in this model. Given that subjects did not significantly vary in their intercept values (tau burden at baseline), we only introduced a random slope. Next, we tested the same interaction effects as in model 1, except that we used tau burden instead of tau volume for the interactions in model 2.

We further performed 2 additional sensitivity analyses: First, we tested whether similar effects were found when using the mean z score in the already affected volume rather than the difference in z score over time in months from the baseline tau scan acquisition (i.e., tau level rise). To do so, we used the mean z score in the affected volume as a dependent variable and employed the same fixed and random effects as in model 2. Second, we ran the abovementioned analyses in individuals (n = 89) that had at least 2 FUs available to avoid the influence of singular measurements on the longitudinal analyses.

Normal distribution for the models’ residuals was tested. Significance level was set at a P value of less than 0.05. As our aim was to examine the association of (non-)modifiable factors with either tau spatial extent or tau level rise, rather than comparing the effects between the 2 measures of interest, no correction for the number of LMMs was applied. Visualization of results was performed in R Studio using ggplot (11).

Group Comparison of Yearly Tau Spatial Extent and Tau Level Rise

Finally, we computed the mean of the yearly increase in the newly affected volume (tau spatial extent) and increase in tau burden (tau level rise) for every individual (i.e., mean tau measure of interest/time interval between FU visits in years). We then compared the yearly mean increase in spatial extent and level rise between the 3 groups using the Kruskal–Wallis test given the nonnormality of the variable of interest.

RESULTS

Group Characteristics

The 3 amyloid-positive groups (CU, MCI, AD) of interest did not differ significantly in terms of age, education, and BMI, but in the remaining variables (sex, ApoE4 carriership, hypertension, AD Assessment Scale Cognitive 13 score, NPI score, centiloids, baseline tau burden, and volume). The mean ranks per group were related to AD severity (CU < MCI < AD). There were more ApoE4 carriers in the MCI group than in the CU and AD groups and more men in the MCI and AD groups than in the CU group. Hypertension (intermediate and poor) showed a higher frequency across groups in comparison to ideal hypertension (Table 1).

Baseline Associations of Risk Factors with AD Pathology

On multiple comparison corrections, significant correlation effects were observed between baseline centiloids and tau burden (ρ = 0.459, q < 0.001), tau volume (ρ = 0.346, q < 0.001), and the NPI (ρ = 0.219, q = 0.010). Baseline tau burden was positively correlated with tau volume (ρ = 0.787, q < 0.001) and NPI (ρ = 0.209, q = 0.010). Likewise, baseline volume was positively associated with the NPI (ρ = 0.208, q = 0.010). Remaining correlations did not yield significance.

Comparisons of the variables of interest (i.e., baseline centiloids, tau burden/volume, BMI, education, NPI) between hypertension severity or ApoE4 carriership groups did not result in any significant findings after multiple comparison correction. Also, the distributions of the hypertension and ApoE4 groups were not significantly associated. Results of all tested associations can be found in Supplemental Table 1.

Tau Spatial Extent: Its Decelerating and Accelerating Factors

Model 1 yielded a significant main effect of ApoE4 status (β = 0.024; P = 0.039; 95% CI, 0.001–0.046), BMI (β = 0.002; P = 0.021; 95% CI, 0.0003–0.003), baseline tau volume (β = 0.207; P < 0.001; 95% CI, 0.107–0.308), and time in months from the baseline tau scan acquisition (β = 0.002; P < 0.001; 95% CI, 0.001–0.003) on the newly affected volume over time. More specifically, ApoE4 carriership was linked to a greater tau spatial extent. Likewise, higher BMI, tau volume, and longer time were positively associated with the newly affected volume. The remaining variables, including the control variables age and sex, were not significant (Supplemental Table 2A). The interaction analyses revealed a significant effect of baseline centiloids and tau volume (β = −0.005; P < 0.001; 95% CI, −0.008 to −0.002). Closer inspection through visualization showed that greater centiloids and tau volumes at baseline were linked to the strongest increase in the newly affected volume in comparison to either individuals with greater centiloids and lower tau volume or lower centiloids and greater tau volume at baseline. The remaining tested interaction terms were not significant. Significant fixed (Supplemental Table 2A) and interactive (Supplemental Table 3A) effects are visualized Figure 1.

FIGURE 1.

Main effects on tau spatial progression over time. Significant main effects of variables of interest concerning tau spatial extent (newly affected volume) are visualized. Continuous variables were converted into grouping variables based on common cutoffs (BMI, education) or median of baseline centiloids and tau volume across all subjects. Time refers to time differences in months from baseline tau PET scan to follow-up scans. 95% CI are provided for all group effects. BL = baseline.

The LMMs in the subcohort with at least 2 FU time points yielded the same results, except that sex turned out to be a significant fixed effect, with females presenting a greater increase (Supplemental Tables 4A and 5A).

Finally, subsequent group comparison showed that the rank totals of the yearly increase in the newly affected volume (CU, 65.55; MCI, 94.85; AD, 97.97) were significantly different (H(2,N = 162) = 17.06; P < 0.001). The mean in yearly increase per group is plotted in Supplemental Figure 1A.

Tau Level Rise: Its Decelerating and Accelerating Contributors

The tau level rise model yielded a significant effect of the AD group (β = 0.305; P = 0.006; 95% CI, 0.088–0.522), hypertension (β = −0.295; P = 0.001; 95% CI, −0.477 to −0.114), BMI (β = 0.011; P = 0.049; 95% CI, 0.00004–0.022), and time from the baseline tau scan acquisition (β = 0.011; P < 0.001; 95% CI, 0.006–0.016). The AD group presented a greater tau level rise over time in comparison to the CU group. Unexpectedly, the poor and intermediate hypertension group presented a generally lower increase in tau pathology over time in comparison to the ideal hypertension group. We subsequently tested the effects of SBP and DBP, separately as continuous variables. The results yielded no significant effect of the DBP (β = 0.001; P = 0.734; 95% CI, −0.006 to 0.009) but showed a negative effect of the SBP (β = −0.006; P = 0.005; 95% CI, −0.010 to −0.002) on tau level rise. The remaining variables, including the control variables age and sex, did not reach significance. Significant interaction effects were observed for baseline centiloids and tau burden (β = −0.003; P = 0.003; 95% CI, −0.005 to −0.001), baseline centiloids and BMI (β = 0.001; P < 0.001; 95% CI, 0.0004–0.001), and baseline tau burden and education (β = −0.035; P = 0.020; 95% CI, −0.064 to −0.006). The remaining interactions did not yield significant results. Significant fixed (Supplemental Table 2B) and interactive (Supplemental Table 3B) effects are visualized in Figure 2. Again, results could be replicated in the smaller subcohort (Supplemental Tables 4B and 5B). When using the mean tau burden in the affected volume per time point as a dependent variable, we were able to replicate the results, except that no significant effect of BMI on tau burden was found (Supplemental Table 6A).

FIGURE 2.

Main effects on tau local aggregation over time. Significant main effects of variables of interest on tau level rise (increase of tau in more affected regions) are visualized. Continuous variables were converted into grouping variables based on common cutoffs (BMI, education) or median of baseline centiloids and tau burden across all subjects. Time refers to time differences in months from baseline tau PET scan to follow-up scans. 95% CI are provided for all group effects. BL = baseline.

Finally, subsequent group comparison showed that the rank totals of the yearly z score change (CU, 68.91; MCI, 93.13; AD, 92.50) were significantly different (H(2,N = 162) = 10.58; P = 0.005). The mean in yearly z score change per group is plotted in Supplemental Figure 1B.

DISCUSSION

The current study provides novel insights into the intersection of genetics, health risk factors, and pathologic burden to the spatial progression and regional increase of tau pathology in a cohort of AD biomarker–positive individuals. Consistent with prior work (12), baseline centiloids and tau levels predicted both tau spatial extent and tau level rise, with higher initial pathology leading to greater progression. Interestingly, we observed that the baseline spatial extent of tau predicted its subsequent volumetric expansion, whereas this relation was not observable for baseline tau burden and its local increase. Potentially, the initial seeding of pathologic tau in a certain region causes a self-amplification process, which is independent of its previous seed load. However, the greater the volumetric extent of tau, the more regions are affected, and thus the more structural and functional pathways are available, which may facilitate a more rapid tau spatial extent. This may also explain why we observed a greater yearly increase in newly affected volume in the MCI and AD groups in comparison to the CU group. Predetermined factors, like ApoE4 carriership and female sex, appear to further perpetuate tau’s spatial progression rather than its local increase.

In terms of modifiable risk factors, we found that lower BMI was linked to a slower spatial progression and lower increase in tau burden, whereas greater hypertension severity was, unexpectedly, linked to a lower increase in tau burden but not to the spatial expansion of tau pathology.

The mechanistic pathways by which these tested risk factors act on either the progression or increase of tau pathology may partly be explained by impaired waste clearance (13). Indeed, a more sedentary lifestyle (14) (closely linked to obesity) has been associated with a worsening of protein clearance from the brain, likely due to a decrease in arterial pulsatility (15). Failed clearance, in turn, may lead to progressive aggregation of proteins, such as tau pathology, thereby presumably explaining the effects of BMI on tau level rise. Speculatively, the rather unexpected results of greater hypertension severity being linked to lower tau aggregation may potentially be explained by greater clearance due to higher SBP and thus to a lower increase in tau burden. This certainly requires further elucidation, while also considering confounding effects of medication intake.

Despite this line of argumentation, we observed a (trend) significant interactive effect of baseline centiloids and BMI on tau spatial extent and level rise indicating that the effect of BMI may be driven by its association with amyloid rather than tau burden. Interestingly, it has been shown that obesity in midlife is closely associated with a greater risk of developing AD (16,17). Potentially, this risk factor facilitates amyloid-β accumulation during midlife. Indeed, antecedent amyloid burden has been shown to be a key driver of tau pathology aggregation (18). It may thus be that the risk of higher BMI on tau progression is driven by its effects on amyloid aggregation in preclinical stages, which subsequently impacts tau pathology in clinical stages.

In terms of NSS and education, we only observed baseline associations with pathologic burden and an interactive effect of lower education and greater tau burden on tau level rise. This may be due to the nature of these factors as they do not directly interfere with biologic pathways, as opposed to BMI or hypertension. Hence, these factors may be more closely related to cognitive function or decline rather than the pathologic progress per se. Given that education is considered as proxy for cognitive reserve (19) (i.e., the preservation of cognition despite increased pathologic burden), it may potentially moderate the association between tau spatial extent and tau level rise and cognitive decline, which remains to be assessed.

Several limitations should be noted. We only tested predefined interactions, so other effects may have been missed. Moreover, the rather unexpected finding in terms of hypertension severity may potentially be driven by confounding effects of medication intake, which were not accounted for in the current analyses. Furthermore, blood pressure is susceptible to fluctuations but was based on a single measurement in the current study. Moreover, we assessed tau progression using a first-generation tau PET tracer, which shows greater off-target binding than second-generation tracers. Replication based on second-generation tracers, including testing different z score cutoffs necessary for the volumetric approach, may provide more refined means to study the differential role of tau spatial extent and tau level rise, in particular concerning their association with cognitive and clinical progression. Finally, due to current restrictions in data availability of the ADNI 3 medical history, we were only able to focus on 4 of the 14 risk factors that have been suggested to account for 45% world-wide dementia cases (1). Therefore, it remains unknown whether a single factor carries superior risk for tau progression than others. Refined assessment of these risk factors will be important for predicting individual cognitive decline and progression across disease stages. Evidence indicates that the spatial spread of tau pathology is more tightly linked to cognitive impairment than regional tau accumulation alone (20). Future analyses on these 2 measures of tau progression could therefore provide key insights for evaluating the effectiveness of new AD treatments.

CONCLUSION

Overall, this study provides an initial framework for examining how modifiable factors influence tau pathology, distinguishing between its spatial spread (tau spatial extent) and amplification (tau level rise). The results indicate that education, BMI, and hypertension carry differential effects in this regard and closely interact with initial pathologic burden. Promoting healthier lifestyles may help slow disease progression, potentially through reducing overall pathologic burden and supporting maintenance of overall brain health. At what time point in life this modification should take place remains to be tested. Moreover, differential consideration of tau spatial extent and tau level rise may further provide novel and more refined means to study treatment effects of currently tested drug compounds against AD.

DISCLOSURE

Thilo van Eimeren reports having received consulting and lecture fees from Lundbeck Foundation, Lilly Germany, Eisai GmbH, GT Gain Therapeutics SA, ICON plc, the Leibniz Association, and the EU-joint program for neurodegenerative disease research (JPND). Alexander Drzezga reports research support by Siemens Healthineers, Life Molecular Imaging, GE HealthCare, AVID Radiopharmaceuticals, SOFIE, Eisai, Novartis/AAA, and Ariceum Therapeutics; speaker honorary/advisory boards for Siemens Healthineers, Sanofi, GE HealthCare, Biogen, Novo Nordisk, Invicro, Novartis/AAA, Bayer Vital, and Lilly; stock in Siemens Healthineers, Lantheus Holding, Structured Therapeutics, and Lilly; patents for ¹⁸F-JK-PSMA-7 (patent no. EP3765097A1; date of patent: Jan. 20, 2021). Gerard Bischof and Merle Hoenig received funding from the Alzheimer Forschung Initiative e.V. In addition, this study was supported by the German Research Foundation (DFG; DR 445/9-1 (to Alexander Drzezga), CRC1451-C04 Project-ID 431549029 (to Alexander Drzezga and Gerard Bischof), and RTG 1960 Project ID 233886668 (to Verena Dzialas)). 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).

KEY POINTS

QUESTION: How do modifiable factors, genetic determinants, and initial pathologic burden affect the spatial progression and local amplification of tau pathology in AD?

PERTINENT FINDINGS: In this longitudinal study, we show that modifiable risk factors, such as higher BMI or level of education, appear to impact the local amplification of tau rather than its spatial extent, whereas genetic determinants, such as female sex and ApoE4 status, contribute to a greater spatial extent of tau. Yet, both measures of tau progression are highly dependent on the initial pathologic burden.

IMPLICATIONS FOR PATIENT CARE: Timely modification of risk factors may slow the progression of tau pathology. Consideration of the two aspects of tau spreading (i.e. spatial extent and local amplification) may provide more refined means to study treatment effects of currently tested drug compounds against AD.