Abstract
Background and Purpose
Apolipoprotein E (APOE) ε4 is associated with Alzheimer’s disease risk and amyloid pathology, but whether cerebrospinal fluid (CSF) Alzheimer’s disease biomarkers show statistical indirect effects in its association with longitudinal clinical worsening in mild cognitive impairment (MCI) remains unclear.
Methods
We analyzed 397 participants with MCI from the Alzheimer’s Disease Neuroimaging Initiative. The primary analysis used 36-month change in Clinical Dementia Rating–Sum of Boxes (CDR-SB) in the full cohort. APOE ε4 carrier status was the predictor, baseline CSF amyloid beta 1–42 (Aβ42) and phosphorylated tau 181 (p-tau181) were serial mediators, and age, sex, and education were covariates. Secondary analyses were stratified by amyloid status, and 12- and 24-month analyses and APOE ε4 allele-count models were exploratory and sensitivity analyses, respectively. Indirect effects were evaluated using 5,000 percentile bootstrap resamples.
Results
In the full cohort, the total effect of APOE ε4 on 36-month CDR-SB change was significant (estimate, 1.047; 95% bootstrap confidence interval [CI], 0.552 to 1.540), whereas the residual direct effect was not (0.097; 95% CI, −0.447 to 0.636). The Aβ42-specific indirect effect was significant (0.751; 95% CI, 0.474 to 1.035), while the p-tau181-specific (0.068; 95% CI, −0.007 to 0.167) and sequential Aβ42-to-p-tau181 indirect effects (0.131; 95% CI, −0.002 to 0.309) were not. Secondary and allele-count sensitivity analyses were directionally consistent.
Conclusions
Lower CSF Aβ42 showed a significant statistical indirect association between APOE ε4 carrier status and 36-month clinical worsening in MCI. These observational findings do not establish a causal mediating mechanism.
Keywords: Alzheimer Disease, Apolipoproteins E, Amyloid Beta-Peptides, Tau Proteins, Mild Cognitive Impairment
INTRODUCTION
Alzheimer’s disease (AD) is the most common cause of dementia. AD substantially impairs global functioning and severely compromises patients’ quality of life. Detecting factors that predispose individuals with mild cognitive impairment (MCI) to progressive cognitive change is crucial for promptly initiating interventions aimed at slowing disease progression.
Apolipoprotein E (APOE) polymorphic alleles are major determinants of genetic risk for AD.1 There are three common APOE alleles in humans: ε2, ε3, and ε4. The risk of developing AD varies according to the combination of APOE alleles.2 The APOE ε4 allele increases AD risk, whereas ε2 is generally protective.3 APOE ε4 is also associated with an increased risk of progression from MCI to AD dementia.4
AD pathologic changes can be detected using fluid and neuroimaging biomarkers that support diagnosis, differential diagnosis, disease-course tracking, and treatment evaluation.5 In particular, amyloid beta (Aβ) and pathologic tau can be assessed using cerebrospinal fluid (CSF) biomarkers throughout the disease course.6 CSF amyloid beta 1–42 (Aβ42) is reduced in AD in association with cortical amyloid deposition, whereas elevated phosphorylated tau reflects AD-related tau pathology.7 Aβ plaques and neurofibrillary tangles composed of hyperphosphorylated tau are core pathological features of AD.8
Low CSF Aβ42 levels have long been associated with symptomatic AD.9 Amyloid positivity is therefore relevant not only to diagnosis but also to clinical prognosis, and is associated with clinical deterioration and progression to dementia in MCI.10 Although recent criteria increasingly emphasize biomarker ratios and newer plasma phosphorylated tau (p-tau) markers, single CSF Aβ42 and p-tau181 measurements from established Alzheimer’s Disease Neuroimaging Initiative (ADNI) platforms remain useful proxies for AD pathophysiology.
APOE ε4 is strongly associated with amyloid accumulation, while amyloid and tau pathologies are associated with cognitive decline. However, whether baseline CSF Aβ42 and p-tau181 show statistical indirect effects in the association between APOE ε4 and subsequent clinical worsening in MCI remains uncertain. The primary objective of this study was therefore to evaluate total, direct, and biomarker-specific indirect effects of APOE ε4 carrier status on 36-month Clinical Dementia Rating–Sum of Boxes (CDR-SB) change in the full MCI cohort.
Secondary analyses examined the same pathways within Aβ-positive and Aβ-negative subgroups. Because amyloid status was defined using the same CSF Aβ42 measure entered as the first mediator, these stratified analyses were considered secondary and interpreted in light of restricted mediator variance. Analyses at 12 and 24 months were exploratory, and APOE ε4 allele count was evaluated as a sensitivity exposure.
METHODS
Data source
Data used in this study were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). ADNI was launched in 2003 as a public-private partnership led by Principal Investigator Michael W. Weiner, MD. Its goals include validating biomarkers for clinical trials, improving generalizability by increasing cohort diversity, and providing data on the diagnosis and progression of Alzheimer’s disease to the scientific community. This analysis used de-identified data from participants enrolled in the ADNI-GO and ADNI-2 phases, accessed from the Image and Data Archive in February 2024.
Ethics statement
The Institutional Review Board (IRB) of Kangwon National University Hospital determined that this study was exempt from review (IRB No. KNUH-2023-01-010). The study was conducted in accordance with the Declaration of Helsinki. All ADNI study procedures were approved by the IRBs of the participating institutions, and written informed consent was obtained from all participants or their authorized representatives.
Study population
Participants with amnestic MCI who had completed baseline CSF assessments were selected. MCI was diagnosed according to Petersen et al.11; participants had objective memory impairment without dementia. All participants had a Mini-Mental State Examination (MMSE) score of at least 24,12 a global Clinical Dementia Rating (CDR) score of 0.5,13 a CDR memory score of at least 0.5, and impaired delayed recall of Story A from the Wechsler Memory Scale–Revised according to education-adjusted criteria.14,15 Sex was recorded as male or female in ADNI demographic data and included as a covariate. None of the participants were taking anti-dementia medication at baseline.
The analytic dataset comprised participants with baseline CSF Aβ42 and p-tau181, APOE genotype, and complete CDR-SB data at 12, 24, and 36 months. A common complete-case cohort was retained across the primary and exploratory time-point analyses. The final cohort included 397 participants.
CSF biomarker measurements
CSF Aβ42 and p-tau181 were collected by lumbar puncture using a standardized ADNI protocol and measured with the Innogenetics INNO-BIA AlzBio3 immunoassay and multiplex xMAP Luminex platform.16 Amyloid positivity was defined as CSF Aβ42 <192 pg/mL. This threshold yielded 252 Aβ-positive and 145 Aβ-negative participants.
APOE genotyping
APOE genotyping was performed using blood samples. In the primary analysis, participants with at least one ε4 allele were classified as APOE ε4 carriers, and those without ε4 were classified as non-carriers. APOE ε4 allele count, coded as 0, 1, or 2 and modeled continuously, was used in sensitivity analyses.
Clinical outcome
Longitudinal clinical change was assessed using CDR-SB. ΔCDR-SB was calculated as follow-up CDR-SB minus baseline CDR-SB. The primary outcome was ΔCDR-SB at 36 months, representing the maximum observed follow-up. ΔCDR-SB at 12 and 24 months was evaluated in exploratory analyses.
Statistical analysis
Continuous variables are presented as mean ± standard deviation and categorical variables as number (percentage). Aβ-positive and Aβ-negative groups were compared using independent-samples t-tests for continuous variables and χ2 tests for categorical variables.
A serial multiple-mediation model corresponding to Hayes’ PROCESS Model 6 was used (Fig. 1).17 APOE ε4 carrier status was the independent variable (X), baseline CSF Aβ42 was the first mediator (M1), baseline CSF p-tau181 was the second mediator (M2), and ΔCDR-SB was the outcome (Y). The path coefficients were defined as a1 for X→M1, a2 for X→M2, d21 for M1→M2, b1 for M1→Y, b2 for M2→Y, c for the total effect, and c′ for the direct effect. The specific indirect effects were a1×b1 through Aβ42, a2×b2 through p-tau181, and a1×d21×b2 through the sequential Aβ42-to-p-tau181 pathway. The total indirect effect was the sum of the three specific indirect effects.
Fig. 1. Serial multiple-mediation model. Diagram of the primary serial multiple-mediation model. APOE ε4 carrier status was modeled as the independent variable (X), baseline CSF Aβ42 as the first mediator (M1), baseline CSF p-tau181 as the second mediator (M2), and 36-month change in CDR-SB (ΔCDR-SB) as the outcome (Y). The specific indirect effects were a1×b1 through Aβ42, a2×b2 through p-tau181, and a1×d21×b2 through the sequential Aβ42-to-p-tau181 pathway. Age, sex, and education were included as covariates in all endogenous-variable equations. The total effect (c) was estimated in a separate model without the mediators; c′ denotes the residual direct effect.

APOE: apolipoprotein E gene, CSF: cerebrospinal fluid, Aβ42: amyloid beta 1–42, p-tau181: phosphorylated tau 181, CDR-SB: Clinical Dementia Rating–Sum of Boxes.
The primary model was fitted in the full MCI cohort using 36-month ΔCDR-SB. Secondary 36-month models were fitted separately in Aβ-positive and Aβ-negative subgroups; because the stratification variable and first mediator were both based on baseline CSF Aβ42, these subgroup results were interpreted cautiously because of restricted mediator variance. Exploratory models evaluated 12- and 24-month outcomes, and sensitivity models replaced carrier status with APOE ε4 allele count.
Age, sex, and education were included as covariates in all three endogenous-variable equations. Unstandardized effects were estimated. Total, direct, total indirect, and specific indirect effects were evaluated with 5,000 percentile bootstrap resamples. An effect was considered statistically significant when its 95% percentile bootstrap confidence interval (CI) excluded zero. Indirect effects were evaluated directly and were not conditioned on a statistically significant total or direct effect. The directions of the total, direct, and specific indirect effects were examined for inconsistent mediation or statistical suppression.
All analyses were performed using custom R scripts implementing the serial mediation framework in R version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria). Bootstrap CIs were calculated with the boot package (version 1.3-30), using a fixed random seed of 1234. All results, tables, and figures were regenerated in a unified analytic pipeline.
Generative AI assisted with drafting and checking R code; all final analyses and numerical outputs were independently executed and verified by the authors.
RESULTS
The study included 397 participants with MCI, including 252 Aβ-positive and 145 Aβ-negative participants. Mean age was 71.9±7.4 years, 163 (41.1%) participants were female, and 192 (48.4%) were APOE ε4 carriers. Compared with the Aβ-negative group, the Aβ-positive group was older and had higher baseline CDR-SB, lower MMSE, lower CSF Aβ42, higher p-tau181, more frequent APOE ε4 carriage, and greater CDR-SB worsening during follow-up (Table 1).
Table 1. Baseline characteristics of the study population.
| Characteristic | MCI (n=397) | Aβ-positive MCI (n=252) | Aβ-negative MCI (n=145) | p-value | |
|---|---|---|---|---|---|
| Age (yr) | 71.9±7.4 | 73.0±7.0 | 70.1±7.8 | <0.001 | |
| Female | 163 (41.1) | 100 (39.7) | 63 (43.4) | 0.530 | |
| Education (yr) | 16.1±2.8 | 16.1±2.8 | 16.3±2.7 | 0.551 | |
| APOE ε4 allele count | 0.59±0.68 | 0.81±0.70 | 0.21±0.41 | <0.001 | |
| 0 alleles | 205 (51.6) | 90 (35.7) | 115 (79.3) | ||
| 1 allele | 149 (37.5) | 119 (47.2) | 30 (20.7) | ||
| 2 alleles | 43 (10.8) | 43 (17.1) | 0 (0.0) | ||
| APOE ε4 carrier | 192 (48.4) | 162 (64.3) | 30 (20.7) | <0.001 | |
| Baseline CDR-SB | 1.4±0.9 | 1.5±0.9 | 1.2±0.7 | <0.001 | |
| Baseline MMSE | 27.8±1.8 | 27.4±1.9 | 28.3±1.5 | <0.001 | |
| CSF Aβ42 (pg/mL) | 173.7±54.0 | 138.1±26.0 | 235.4±28.2 | <0.001 | |
| CSF p-tau181 (pg/mL) | 38.5±22.2 | 46.2±23.1 | 25.1±12.1 | <0.001 | |
| Progression to dementia | 116 (29.2) | 104 (41.3) | 12 (8.3) | <0.001 | |
| ΔCDR-SB at 12 mon | 0.23±1.00 | 0.39±1.03 | −0.06±0.88 | <0.001 | |
| ΔCDR-SB at 24 mon | 0.71±1.57 | 1.12±1.71 | −0.01±0.95 | <0.001 | |
| ΔCDR-SB at 36 mon | 1.25±2.55 | 1.97±2.82 | −0.00±1.22 | <0.001 | |
Values are presented as mean ± standard deviation or number (%).
Group comparisons used independent-samples t-tests for continuous variables and χ2 tests for categorical variables.
MCI: mild cognitive impairment, Aβ: amyloid beta, APOE: apolipoprotein E gene, CDR-SB: Clinical Dementia Rating–Sum of Boxes, MMSE: Mini-Mental State Examination, CSF: cerebrospinal fluid, Aβ42: amyloid beta 1–42, p-tau181: phosphorylated tau 181.
In the primary full-cohort 36-month model, the total effect of APOE ε4 carrier status on ΔCDR-SB was statistically significant (estimate, 1.047; 95% bootstrap CI, 0.552 to 1.540), while the residual direct effect was not (estimate, 0.097; 95% CI, −0.447 to 0.636). The total indirect effect was significant (estimate, 0.950; 95% CI, 0.691 to 1.229). Among the specific indirect pathways, the Aβ42-specific indirect effect was significant (estimate, 0.751; 95% CI, 0.474 to 1.035), whereas the p-tau181-specific indirect effect (estimate, 0.068; 95% CI, −0.007 to 0.167) and sequential indirect effect (estimate, 0.131; 95% CI, −0.002 to 0.309) were not significant (Table 2, Fig. 2).
Table 2. Total, direct, and indirect effects of APOE ε4 carrier status on 36-month CDR-SB change in the full MCI cohort.
| Effect | Estimate | Bootstrap SE | 95% percentile bootstrap CI |
|---|---|---|---|
| Total effect (c) | 1.047 | 0.257 | 0.552, 1.540 |
| Direct effect (c′) | 0.097 | 0.279 | −0.447, 0.636 |
| Total indirect effect | 0.950 | 0.140 | 0.691, 1.229 |
| Indirect effect via CSF Aβ42 (a1×b1) | 0.751 | 0.146 | 0.474, 1.035 |
| Indirect effect via CSF p-tau181 (a2×b2) | 0.068 | 0.046 | −0.007, 0.167 |
| Sequential indirect effect via CSF Aβ42 and p-tau181 (a1×d21×b2) | 0.131 | 0.078 | −0.002, 0.309 |
Effects are unstandardized estimates from the primary full-cohort model (n=397), adjusted for age, sex, and education. CIs were obtained from 5,000 percentile bootstrap resamples. An effect was considered statistically significant when the 95% bootstrap CI excluded zero; statistically significant rows are shown in bold. The total indirect effect is the sum of the three specific indirect effects.
APOE: apolipoprotein E gene, CDR-SB: Clinical Dementia Rating–Sum of Boxes, MCI: mild cognitive impairment, SE: standard error, CI: confidence interval, CSF: cerebrospinal fluid, Aβ42: amyloid beta 1–42, p-tau181: phosphorylated tau 181.
Fig. 2. Total, direct, and indirect effects of APOE ε4 on 36-month clinical change. Forest plots show unstandardized effect estimates and 95% percentile bootstrap CIs. (A) presents the total, direct, total indirect, Aβ42-specific indirect, p-tau181-specific indirect, and sequential indirect effects in the full MCI cohort. (B) compares the Aβ42-specific indirect effect in the full cohort and the Aβ-positive and Aβ-negative subgroups. CIs excluding zero indicate statistical significance. Subgroup analyses were secondary and should be interpreted in light of restricted Aβ42 variance.

APOE: apolipoprotein E gene, CI: confidence interval, Aβ42: amyloid beta 1–42, p-tau181: phosphorylated tau 181, MCI: mild cognitive impairment, Aβ: amyloid beta.
In the secondary Aβ-positive subgroup analysis, the total and direct effects at 36 months were not significant. The total indirect effect (estimate, 0.403; 95% CI, 0.116 to 0.715) and Aβ42-specific indirect effect (estimate, 0.309; 95% CI, 0.021 to 0.613) were significant, while the p-tau181-specific and sequential indirect effects were not. In the Aβ-negative subgroup, no total, direct, total indirect, or specific indirect effect was significant (Table 3).
Table 3. Secondary 36-month mediation analyses stratified by amyloid status.
| Effect | Aβ-positive | Aβ-negative | ||
|---|---|---|---|---|
| Estimate | 95% CI | Estimate | 95% CI | |
| Total effect (c) | 0.479 | −0.277, 1.255 | −0.111 | −0.496, 0.285 |
| Direct effect (c′) | 0.076 | −0.682, 0.859 | −0.115 | −0.492, 0.279 |
| Total indirect effect | 0.403 | 0.116, 0.715 | 0.004 | −0.066, 0.077 |
| Indirect effect via CSF Aβ42 (a1×b1) | 0.309 | 0.021, 0.613 | −0.003 | −0.066, 0.056 |
| Indirect effect via CSF p-tau181 (a2×b2) | 0.057 | −0.042, 0.173 | 0.008 | −0.031, 0.067 |
| Sequential indirect effect via CSF Aβ42 and p-tau181 (a1×d21×b2) | 0.037 | −0.012, 0.124 | 0.000 | −0.007, 0.004 |
Values are unstandardized estimates from secondary subgroup models adjusted for age, sex, and education. CIs were obtained from 5,000 percentile bootstrap resamples. An effect was considered statistically significant when the 95% bootstrap CI excluded zero; statistically significant cells are shown in bold. Because amyloid status was defined using the same baseline CSF Aβ42 measure entered as the first mediator, these results should be interpreted in light of restricted mediator variance.
Aβ: amyloid beta, CI: confidence interval, CSF: cerebrospinal fluid, Aβ42: amyloid beta 1–42, p-tau181: phosphorylated tau 181.
Exploratory full-cohort analyses at 12 and 24 months also showed significant total and Aβ42-specific indirect effects, whereas p-tau181-specific and sequential indirect effects were not significant (Supplementary Table 1). Sensitivity analysis using APOE ε4 allele count yielded conclusions consistent with the primary carrier-status analysis at 36 months, including a significant total effect (estimate, 0.917; 95% CI, 0.558 to 1.272) and Aβ42-specific indirect effect (estimate, 0.554; 95% CI, 0.332 to 0.774) (Supplementary Table 2).
DISCUSSION
In the primary full-cohort analysis, APOE ε4 carrier status showed a significant total association with 36-month clinical worsening, and a significant statistical indirect effect was observed through lower baseline CSF Aβ42. No statistically significant residual direct effect, p-tau181-specific indirect effect, or sequential Aβ42-to-p-tau181 indirect effect was detected. Secondary Aβ-stratified analyses were directionally consistent, while allele-count sensitivity analysis supported the robustness of the primary finding.
The significant indirect effect should be interpreted as a statistical pathway rather than proof of a causal biological mechanism. Contemporary mediation methods do not require a statistically significant total effect before evaluating indirect effects17; nevertheless, our primary analysis also identified a significant total effect. In this model, the total, direct, and specific indirect effects were directionally concordant, providing no evidence of inconsistent mediation or statistical suppression. The absence of a statistically significant residual direct effect does not establish that the direct effect is exactly zero and should not be described as complete or full mediation.
The Aβ42-specific indirect effect is biologically plausible. APOE ε4 is a major genetic risk factor associated with cerebral amyloid accumulation through effects on Aβ aggregation and clearance.1,18 Lower CSF Aβ42 reflects greater cerebral sequestration of soluble Aβ into amyloid plaques.7,9 Previous neuropathologic and biomarker studies have similarly suggested that AD pathology, particularly amyloid deposition, contributes to the association between APOE ε4 and clinical outcomes (Table 4).19,22,24 Our full-cohort analysis extends this literature by demonstrating a robust statistical indirect association through baseline CSF Aβ42 over 36 months in MCI.
Table 4. Summary of previous studies on APOE ε4, AD pathology, and cognitive outcomes.
| Study | Participants | Main finding |
|---|---|---|
| Bennett et al., 200519 | 44 AD, 50 CN; single cohort | At least one APOE ε4 allele was associated with lower cognitive function proximate to death; the association was related to amyloid deposition and subsequent tangle formation. |
| Risacher et al., 201318 | 209 E-MCI, 123 CN; ADNI | APOE ε4 status was associated with amyloid accumulation and lower CSF Aβ. |
| Buerger et al., 200520 | 102 MCI, 29 CN | CSF p-tau231P levels were higher in APOE ε4 carriers than in non-carriers. |
| Young et al., 202321 | 220 Aβ-positive CU/MCI; ADNI | Amyloid mediated the association between APOE ε4 and tau. |
| Paradela et al., 202322 | 648 CN | The association between APOE ε4 and cognition was partially mediated by AD pathology. |
| Xia et al., 202323 | NACC autopsy cohort | Most of the effect of APOE ε4 alleles on cognition occurred through AD neuropathologic change. |
| Mortimer et al., 200924 | 689 Nun Study participants | The effect of APOE ε4 on dementia was largely accounted for by more severe AD neuropathology. |
| Bennett et al., 200425 | 44 AD, 53 CN | The effect of amyloid deposition on clinical disease was mediated by neurofibrillary tangles. |
| Ma et al., 202126 | 370 CN, 623 MCI; ADNI | In MCI, tau-related pathology and neurodegeneration mediated the effect of Aβ on cognition. |
| Liu et al., 201627 | 1,718 AD/MCI/CN | APOE ε4 was associated with reduced CSF Aβ42, increased cerebral Aβ deposition, tau-related biomarkers, neurodegeneration, and cognitive decline. |
| Dage et al., 201628 | 378 CN, 161 MCI | Plasma tau was associated with cortical thickness and memory performance. |
APOE: apolipoprotein E gene, AD: Alzheimer’s disease, CN: cognitively normal, E-MCI: early mild cognitive impairment, ADNI: Alzheimer’s Disease Neuroimaging Initiative, CSF: cerebrospinal fluid, Aβ: amyloid beta, MCI: mild cognitive impairment, p-tau: phosphorylated tau, CU: cognitively unimpaired, NACC: National Alzheimer’s Coordinating Center.
Baseline CSF p-tau181 did not show a significant specific indirect effect, and the sequential Aβ42-to-p-tau181 pathway was also not significant. This finding should not be interpreted as evidence that tau pathology is unimportant to cognitive decline. A single baseline CSF p-tau181 measurement reflects soluble tau phosphorylation but may not fully capture longitudinal tau accumulation or the regional burden of neurofibrillary tangles that is more closely related to domain-specific cognitive change.8 MCI is clinically and pathologically heterogeneous,11,15 and CDR-SB is a global clinical measure that may be less sensitive to subtle tau-related changes than domain-specific neuropsychological composites. In addition, a sequential indirect effect is estimated as a product of multiple path coefficients and can be less precisely estimated than a single-mediator pathway.
The Aβ-positive and Aβ-negative analyses were retained as secondary comparator analyses. The Aβ42-specific indirect effect in the Aβ-positive subgroup was directionally consistent with the primary result, whereas no indirect effect was detected in the Aβ-negative subgroup. However, amyloid status was defined using the same baseline CSF Aβ42 variable used as the first mediator, which restricts mediator variance within each subgroup. Accordingly, subgroup significance should not be viewed as independent causal confirmation of the full-cohort pathway, and the full-cohort model provides the principal evidence.
This study has several strengths, including a well-characterized multicenter longitudinal cohort, a clearly defined 36-month primary endpoint representing the maximum observed follow-up, bootstrap-based reporting of total, direct, and specific indirect effects, and sensitivity analyses using alternative APOE ε4 coding. Several limitations should also be acknowledged. First, the observational design and possible unmeasured confounding preclude causal interpretation. Second, biomarkers were measured only at baseline, and longitudinal biomarker trajectories were unavailable. Third, the complete-case design may introduce selection bias. Fourth, the Aβ-stratified analyses were affected by restricted Aβ42 variance. Fifth, amyloid PET, tau PET, and imaging markers of neurodegeneration were not included. Sixth, MCI subtypes were not differentiated. Finally, the selected, highly educated ADNI cohort may limit generalizability to more diverse clinical populations.
In conclusion, lower baseline CSF Aβ42 showed a significant statistical indirect association between APOE ε4 carrier status and 36-month clinical worsening in MCI. Baseline p-tau181 and the sequential Aβ42-to-p-tau181 pathway did not show significant specific indirect effects. These findings support an amyloid-related statistical pathway but do not establish causal mediation. Longitudinal multimodal biomarker studies are needed to clarify how APOE ε4-associated genetic risk translates into clinical progression.
Footnotes
Funding: Data collection and sharing for the Alzheimer’s Disease Neuroimaging Initiative (ADNI) were 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. ADNI has also received support 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, including contributions from 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 authors received no specific funding for this analysis.
Conflict of Interest: The authors have no financial conflicts of interest.
Generative AI Disclosure: OpenAI ChatGPT (OpenAI, GPT-5.6 Sol; accessed August 30, 2026) and Claude (Anthropic, Claude Fable 5; accessed August 30, 2026) were used for language editing, manuscript restructuring, and drafting and checking R code. All AI-assisted outputs were reviewed and verified by the authors, who take full responsibility for the final manuscript.
- Conceptualization: Jang JW.
- Data curation: Heo M.
- Formal analysis: Heo M.
- Methodology: Lee S, Jang JW.
- Supervision: Kim Y, Kim S, Lee SY, Jang JW.
- Writing - original draft: Lee S, Heo M.
- Writing - review & editing: Lee S, Heo M.
SUPPLEMENTARY MATERIALS
Exploratory mediation analyses at 12 and 24 months using APOE ε4 carrier status
Sensitivity analyses using APOE ε4 allele count at 36 months
References
- 1.Liu CC, Kanekiyo T, Xu H, Bu G. Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol. 2013;9:106–118. doi: 10.1038/nrneurol.2012.263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Raulin AC, Doss SV, Trottier ZA, Ikezu TC, Bu G, Liu CC. ApoE in Alzheimer’s disease: pathophysiology and therapeutic strategies. Mol Neurodegener. 2022;17:72. doi: 10.1186/s13024-022-00574-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Corder EH, Saunders AM, Strittmatter WJ, Schmechel DE, Gaskell PC, Small GW, et al. Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer’s disease in late onset families. Science. 1993;261:921–923. doi: 10.1126/science.8346443. [DOI] [PubMed] [Google Scholar]
- 4.Elias-Sonnenschein LS, Viechtbauer W, Ramakers IHGB, Verhey FRJ, Visser PJ. Predictive value of APOE-ε4 allele for progression from MCI to AD-type dementia: a meta-analysis. J Neurol Neurosurg Psychiatry. 2011;82:1149–1156. doi: 10.1136/jnnp.2010.231555. [DOI] [PubMed] [Google Scholar]
- 5.Hampel H, Broich K, Hoessler Y, Pantel J. Biological markers for early detection and pharmacological treatment of Alzheimer’s disease. Dialogues Clin Neurosci. 2009;11:141–157. doi: 10.31887/DCNS.2009.11.2/hhampel. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Olsson B, Lautner R, Andreasson U, Öhrfelt A, Portelius E, Bjerke M, et al. CSF and blood biomarkers for the diagnosis of Alzheimer’s disease: a systematic review and meta-analysis. Lancet Neurol. 2016;15:673–684. doi: 10.1016/S1474-4422(16)00070-3. [DOI] [PubMed] [Google Scholar]
- 7.Seppälä TT, Nerg O, Koivisto AM, Rummukainen J, Puli L, Zetterberg H, et al. CSF biomarkers for Alzheimer disease correlate with cortical brain biopsy findings. Neurology. 2012;78:1568–1575. doi: 10.1212/WNL.0b013e3182563bd0. [DOI] [PubMed] [Google Scholar]
- 8.Long JM, Holtzman DM. Alzheimer disease: an update on pathobiology and treatment strategies. Cell. 2019;179:312–339. doi: 10.1016/j.cell.2019.09.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Motter R, Vigo-Pelfrey C, Kholodenko D, Barbour R, Johnson-Wood K, Galasko D, et al. Reduction of beta-amyloid peptide42 in the cerebrospinal fluid of patients with Alzheimer’s disease. Ann Neurol. 1995;38:643–648. doi: 10.1002/ana.410380413. [DOI] [PubMed] [Google Scholar]
- 10.Doraiswamy PM, Sperling RA, Coleman RE, Johnson KA, Reiman EM, Davis MD, et al. Amyloid-β assessed by florbetapir F 18 PET and 18-month cognitive decline: a multicenter study. Neurology. 2012;79:1636–1644. doi: 10.1212/WNL.0b013e3182661f74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tangalos EG, Kokmen E. Mild cognitive impairment: clinical characterization and outcome. Arch Neurol. 1999;56:303–308. doi: 10.1001/archneur.56.3.303. [DOI] [PubMed] [Google Scholar]
- 12.Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–198. doi: 10.1016/0022-3956(75)90026-6. [DOI] [PubMed] [Google Scholar]
- 13.Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology. 1993;43:2412–2414. doi: 10.1212/wnl.43.11.2412-a. [DOI] [PubMed] [Google Scholar]
- 14.Wechsler D. WMS-R: Wechsler Memory Scale--Revised: Manual. San Antonio: Psychological Corporation; 1987. [Google Scholar]
- 15.Bondi MW, Edmonds EC, Jak AJ, Clark LR, Delano-Wood L, McDonald CR, et al. Neuropsychological criteria for mild cognitive impairment improves diagnostic precision, biomarker associations, and progression rates. J Alzheimers Dis. 2014;42:275–289. doi: 10.3233/JAD-140276. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Shaw LM, Vanderstichele H, Knapik-Czajka M, Clark CM, Aisen PS, Petersen RC, et al. Cerebrospinal fluid biomarker signature in Alzheimer’s Disease Neuroimaging Initiative subjects. Ann Neurol. 2009;65:403–413. doi: 10.1002/ana.21610. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hayes AF. Beyond Baron and Kenny: statistical mediation analysis in the new millennium. Commun Monogr. 2009;76:408–420. [Google Scholar]
- 18.Risacher SL, Kim S, Shen L, Nho K, Foroud T, Green RC, et al. The role of apolipoprotein E (APOE) genotype in early mild cognitive impairment (E-MCI) Front Aging Neurosci. 2013;5:11. doi: 10.3389/fnagi.2013.00011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bennett DA, Schneider JA, Wilson RS, Bienias JL, Berry-Kravis E, Arnold SE. Amyloid mediates the association of apolipoprotein E e4 allele to cognitive function in older people. J Neurol Neurosurg Psychiatry. 2005;76:1194–1199. doi: 10.1136/jnnp.2004.054445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Buerger K, Teipel SJ, Zinkowski R, Sunderland T, Andreasen N, Blennow K, et al. Increased levels of CSF phosphorylated tau in apolipoprotein E epsilon4 carriers with mild cognitive impairment. Neurosci Lett. 2005;391:48–50. doi: 10.1016/j.neulet.2005.08.030. [DOI] [PubMed] [Google Scholar]
- 21.Young CB, Johns E, Kennedy G, Belloy ME, Insel PS, Greicius MD, et al. APOE effects on regional tau in preclinical Alzheimer’s disease. Mol Neurodegener. 2023;18:1. doi: 10.1186/s13024-022-00590-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Paradela RS, Justo AFO, Paes VR, Leite REP, Pasqualucci CA, Grinberg LT, et al. Association between APOE-ε4 allele and cognitive function is mediated by Alzheimer’s disease pathology: a population-based autopsy study in an admixed sample. Acta Neuropathol Commun. 2023;11:205. doi: 10.1186/s40478-023-01681-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Xia F, Culhane JE, Chan KCG, Keene CD, Kukull WA. Simultaneous mediation effect estimation of APOE through multiple neuropathological pathways on cognitive outcomes. Alzheimers Dement. 2023;19(S8):e062099 [Google Scholar]
- 24.Mortimer JA, Snowdon DA, Markesbery WR. The effect of APOE-ε4 on dementia is mediated by Alzheimer neuropathology. Alzheimer Dis Assoc Disord. 2009;23:152–157. doi: 10.1097/wad.0b013e318190a855. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bennett DA, Schneider JA, Wilson RS, Bienias JL, Arnold SE. Neurofibrillary tangles mediate the association of amyloid load with clinical Alzheimer disease and level of cognitive function. Arch Neurol. 2004;61:378–384. doi: 10.1001/archneur.61.3.378. [DOI] [PubMed] [Google Scholar]
- 26.Ma LZ, Hu H, Wang ZT, Ou YN, Dong Q, Tan L, et al. P-tau and neurodegeneration mediate the effect of β-amyloid on cognition in non-demented elders. Alzheimers Res Ther. 2021;13:200. doi: 10.1186/s13195-021-00943-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Liu Y, Tan L, Wang HF, Liu Y, Hao XK, Tan CC, et al. Multiple effect of APOE genotype on clinical and neuroimaging biomarkers across Alzheimer’s disease spectrum. Mol Neurobiol. 2016;53:4539–4547. doi: 10.1007/s12035-015-9388-7. [DOI] [PubMed] [Google Scholar]
- 28.Dage JL, Wennberg AMV, Airey DC, Hagen CE, Knopman DS, Machulda MM, et al. Levels of tau protein in plasma are associated with neurodegeneration and cognitive function in a population-based elderly cohort. Alzheimers Dement. 2016;12:1226–1234. doi: 10.1016/j.jalz.2016.06.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Exploratory mediation analyses at 12 and 24 months using APOE ε4 carrier status
Sensitivity analyses using APOE ε4 allele count at 36 months
