Alzheimer's disease neuropathology and longitudinal change in subjective cognitive complaints

Abstract

INTRODUCTION

Subjective cognitive complaints often precede declines in objective measures of cognitive performance. Associations of incipient Alzheimer's disease (AD) neuropathology with subjective cognitive complaints may be detectable earlier than associations with neuropsychological testing among cognitively normal individuals.

METHODS

We examined the independent associations of positron emission tomography measures of amyloid beta and tau pathologies with longitudinal subjective complaints and memory among 91 cognitively normal Baltimore Longitudinal Study of Aging participants using linear mixed effects models. Subjective complaints and memory performance were assessed with the Cognitive Failures Questionnaire and the California Verbal Learning Test, respectively.

RESULTS

Greater parahippocampal tau, independent of amyloid, was associated with higher subjective complaints (estimate = 0.25, standard error [SE] = 0.1, P =$0.015$), while greater entorhinal tau corresponded to an attenuated increase in complaints over time (estimate = –0.06, SE = 0.03, P =$0.047$). Hippocampal tau was associated with steeper memory decline (estimate = –0.03, SE = 0.01, P =$0.040$).

CONCLUSION

Subjective cognitive complaints may be a reflection of early cerebral tau pathology in cognitively normal individuals.

Highlights

• Greater parahippocampal tau was linked with higher subjective cognitive complaints.

• Entorhinal tau was associated with slower increases in cognitive complaints over time.

• Subjective complaints may reflect early amyloid and tau in cognitively normal adults.

1. INTRODUCTION

Subjective cognitive complaints are individuals’ perceived and self‐reported memory complaints or memory complaints reported by a close relative or friend who has observed changes in the individual's cognition. ¹ The 2018 National Institute on Aging–Alzheimer's Association's Alzheimer's disease (AD) research framework ² as well as the Alzheimer's Association Workgroup's revised criteria for diagnosis and staging of AD ³ identify clinical stage 2 as the time when an individual is experiencing subtle cognitive change, which can be measured by subjective cognitive complaints, but does not show cognitive impairment on objective testing. Some studies have shown that participants reporting higher subjective cognitive complaints had worse scores on neuropsychological testing ⁴ , ⁵ and faster rates of decline in neuropsychological testing. ⁶ Participants with subjective cognitive complaints are at a higher risk for progressing to dementia versus those without these complaints. ⁷ , ⁸ , ⁹ However, other studies did not find an association between subjective cognitive complaints and objective cognitive assessments. ¹⁰ , ¹¹

The revised criteria for AD suggest that for individuals in clinical stage 2, defined by subtle cognitive decline including subjective decline, the typical expected biological stage is B, which is characterized by the presence of amyloid beta (Aβ) plaques and medial temporal tau. ³ This is supported by research showing that individuals with subjective cognitive complaints have higher levels of amyloid and tau in the brain. ⁸ Detecting these neuropathological changes in early stages is essential, as studies have shown that the onset of these changes precedes the onset of clinical AD symptoms by up to 15 years for cerebrospinal fluid (CSF) tau and 20 years for CSF Aβ₄₂. ¹² , ¹³ Furthermore, a study using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) found that the median time from estimated amyloid onset to clinical AD symptoms, as determined with a Clinical Dementia Rating (CDR) ≥ 1, to be ≈ 14 years. ¹⁴

Prior research has offered insight into the association between neuropathology and subjective cognitive decline. Amyloid positivity has been associated with higher levels of cognitive complaints even after accounting for depression and anxiety. ¹⁵ , ¹⁶ In addition, the severity of subjective cognitive complaints has been associated with higher amyloid burden, ¹⁷ especially in the parietal and frontal cortices ¹⁸ and among apolipoprotein E ε4 carriers. ¹⁹ Compared to amyloid, tau pathology is a stronger predictor of cognitive decline and progression to dementia. ²⁰ , ²¹ , ²² However, there is limited research on the relationship between subjective cognitive complaints and tau pathology. One previous study found that subjective cognitive complaints were associated with tau aggregation, particularly in the frontal and parietal regions, ²³ while other studies did not find associations with tau pathology. ²⁴ , ²⁵

RESEARCH IN CONTEXT

1. Systematic review: We conducted a literature review using the PubMed database to investigate the relationship of subjective cognitive complaints with amyloid beta plaques and neurofibrillary tau tangles. Although longitudinal research on tau pathology and subjective complaints is limited, existing research offered valuable insights in formulating the study's scientific questions. All relevant literature is cited in the references.

2. Interpretation: The findings support the hypothesis that subjective cognitive complaints may reflect elevated cerebral tau pathology in cognitively normal individuals. This contributes to a growing knowledge of the relationship between Alzheimer's disease (AD) neuropathology and cognition. Subjective cognitive questionnaires may be useful as an initial screen prior to further cognitive and clinical testing among cognitively normal individuals.

3. Future directions: This work highlights the importance of further research into the longitudinal relationship between AD neuropathology and subjective cognitive complaints, particularly in larger cohorts.

Our study focuses on the following three research questions: First, what is the cross‐sectional association between AD neuropathology and subjective cognitive complaints in cognitively normal individuals? Second, is AD neuropathology a predictor of subjective cognitive decline over time? Third, is AD neuropathology a stronger predictor of subjective cognitive complaints compared to neuropsychological assessment–based measures of memory among cognitively normal individuals? To address these questions, we assessed the relationship of subjective cognitive complaints as measured by the Cognitive Failures Questionnaire (CFQ) with global amyloid positivity and tau positron emission tomography (PET) levels in regions that are known to exhibit early tau pathology. We then compared these associations of pathology to verbal episodic memory as assessed by the California Verbal Learning Test (CVLT).

2. METHODS

2.1. Participants

We used data from cognitively normal participant visits in the neuroimaging sub‐study of the Baltimore Longitudinal Study of Aging (BLSA) for participants who had a ¹⁸F‐flortaucipir (FTP) PET scan to measure phosphorylated tau burden, a ¹¹C‐Pittsburgh compound B (PiB) PET scan to measure fibrillar Aβ burden, completed the CFQ, ²⁶ and received the CVLT. ²⁷ Participant selection for this analysis is detailed in the supporting information. Cognitively normal status was based on either (1) a CDR ²⁸ of 0 and ≤ 3 errors on the Blessed Information, Memory, and Concentration (BIMC) test ²⁹ and therefore, the participant did not meet criteria for a consensus conference; or (2) the participant was determined to be cognitively normal based on thorough review of clinical and neuropsychological data at consensus conference. In longitudinal analyses, we included only CFQ and CVLT assessments corresponding to visits at which participants were cognitively normal. Participants below the age of 60 were evaluated every 4 years, 60‐ to 79‐year‐old participants every 2 years, and participants ≥ 80 annually.

Research protocols were conducted in accordance with the United States federal policy for the protection of human research subjects (45 CFR 46), approved by local institutional review boards (IRBs) and the National Institutes of Health, and all participants gave written informed consent at each visit. The BLSA PET study was governed by the IRB of the Johns Hopkins Medical Institutions (protocol numbers NA_00051793, IRB00047185) and the BLSA study was overseen by the IRBs of the National Institute of Environmental Health Sciences and the National Institutes of Health Intramural Research Program.

2.2. Cognitive measures

The CFQ evaluates subjective cognitive complaints using 25 questions about the frequency of common cognitive issues within the past 2 weeks. We calculated total score as the sum of scores divided by four times the total number of questions answered to account for missing responses. Higher scores represent greater subjective cognitive complaints. Using the loadings by Rast et al., ³⁰ we also calculated the three CFQ factors reflecting forgetfulness, distractibility, and false triggering as a weighted mean of the individual CFQ items (see supporting information).

Verbal episodic memory was assessed by the CVLT using a single form to ensure consistency across participant visits. We quantified immediate recall as the number of correctly remembered items across the five learning trials. A memory composite score was calculated by averaging the z scores of the immediate and long delay free recall.

The Center for Epidemiological Studies–Depression (CES‐D) ³¹ scale was used to quantify subjective depressive symptoms.

2.3. PET imaging

To assess phosphorylated tau, PET scans were acquired over 30 minutes on a Siemens high resolution research tomograph (HRRT) scanner starting 75 minutes after an intravenous bolus injection of ≈ 370 MBq of FTP. FTP PET analysis workflow is detailed in Ziontz et al. ³² The partial volume corrected ³³ 80 to 100 minute image was used to compute standardized uptake value ratio (SUVR) images with inferior cerebellar gray matter as the reference region. We computed the average bilateral SUVR in the entorhinal cortex (EC), hippocampus, parahippocampal gyrus, and the inferior temporal gyrus (ITG). We investigated tau burden as a continuous rather than a dichotomous variable to be able to examine its subtle effects. One participant with outlying regional tau SUVRs was excluded from analyses (entorhinal SUVR of 2.0 and ITG SUVR of 2.7).

To assess fibrillar Aβ, PET scans were obtained over 70 minutes on either a General Electric (GE) Advance or a Siemens HRRT scanner immediately after an intravenous bolus injection of ≈ 555 MBq of PiB. PiB PET analysis is detailed in Bilgel et al. ³⁴ HRRT scans were smoothed with a 3 mm full width half‐maximum Gaussian to match their spatial resolution to that of the GE advance scans. Distribution volume ratio (DVR) images were computed using a spatially constrained simplified reference tissue model with cerebellar gray matter as the reference region. ³⁵ Mean cortical Aβ burden was calculated as the average of the DVR in cingulate, frontal, parietal (including precuneus), lateral temporal, and lateral occipital cortical regions, excluding the sensorimotor strip. Leveraging longitudinal PiB PET data available on both scanners for 79 BLSA participants, we estimated the parameters of a linear model mapping mean cortical DVR between the GE Advance and HRRT scanners and applied this mapping to all HRRT values to harmonize them with the GE Advance values. Individuals were categorized as PiB −/+ based on a mean cortical DVR threshold of 1.06, which was derived from a Gaussian mixture model fitted to harmonized values at baseline.

2.4. Statistical analysis

We fitted linear mixed effects models to assess the associations between longitudinal cognitive outcomes and PET biomarkers at index visit using a separate model for CFQ and memory and per each of the four tau PET regions of interest, yielding a total of eight models. Inclusion of both amyloid status (PiB −/+) and regional tau SUVR terms allowed for the examination of their independent cross‐sectional associations with the outcome (CFQ or memory) and the interactions of each of these terms with time from PET allowed for the examination of their associations with the rate of change in the CFQ and memory outcomes. We additionally included an amyloid and tau interaction to capture the synergistic cross‐sectional effect of these pathologies. Age at the time of PET, age², CES‐D, sex, and years of education were included as covariates. For memory models, we additionally adjusted for CVLT practice effects by including a binary term reflecting whether it was the participant's first CVLT administration. We included a random intercept and slope for time per person. The construction of our linear mixed effects models is detailed in (Tables S1–S4 in supporting information). We confirmed our cross‐sectional findings using linear regression models fitted using index visits only (Tables S5 and S6 in supporting information), fitted linear mixed effects models investigating each of the three CFQ factors and the CVLT measures that make up the memory composite score as outcomes (Tables S7–S11 in supporting information), and assessed whether using a continuous index of global amyloid burden rather than binary amyloid status as a covariate affects the main results (Tables S12 and S13 in supporting information).

Prior to fitting the models, we centered age at the time of the PET by subtracting the median age of the sample. Amyloid group was coded as −0.5 for PiB– and +0.5 for PiB+. Each cognitive outcome and continuous PET variable was standardized by subtracting its mean and dividing by its standard deviation. The means and standard deviations for this standardization were computed using a larger cross‐sectional BLSA data set of cognitively normal individuals between the ages of 70 and 80 using the visit closest to age 75 per participant. As a result, the regression coefficients reported for the amyloid and tau variables are standardized coefficients.

To assess if the associations of amyloid and tau neuropathology with subjective cognitive complaints differ from their associations with verbal episodic memory, we tested the equality of the standardized regression coefficients between the CFQ and memory models using seemingly unrelated regression equations (SURE). ³⁶ This analysis was limited to examining cross‐sectional effects using data at index visit and we negated CFQ to make its direction consistent with that of memory.

Linear mixed effects analyses were conducted in R version 4.3.3 ³⁷ using lmerTest. ³⁸ We checked the fit of all statistical models using performance ³⁹ to ensure they were not susceptible to outliers and that there were no convergence issues. The SURE analysis was conducted using systemfit and car. ⁴⁰ , ⁴¹

2.5. Data availability

Code is provided in an open repository (https://gitlab.com/bilgelm/PET‐and‐CFQ). BLSA data are available upon request from https://www.blsa.nih.gov after approval by the BLSA Data Sharing Proposal Review Committee.

3. RESULTS

The sample consisted of 91 cognitively normal participants. Table 1 presents participant characteristics at index visit, which is defined by the requirement that they have a PiB PET within 2 years of their baseline FTP PET and a CFQ and a CVLT assessment within 6 months of their baseline FTP PET. Pair plots of continuous variables are presented in Figures S1 and S2 in supporting information. A total of cognitively normal 241 visits were included for CFQ and 350 for CVLT (Figures 1 and S3 in supporting information).

TABLE 1.

Participant characteristics at index visit.

Characteristic	N = 91
Age (year)	77 (69, 83)
Female	52 (57%)
White	70 (77%)
Education	18 (16, 18)
APOE ε4+	30 (34%)
PiB+	24 (26%)
Entorhinal tau SUVR	1.06 (0.96, 1.12)
Hippocampal tau SUVR	1.29 (1.18, 1.44)
Parahippocampal tau SUVR	1.01 (0.94, 1.08)
ITG tau SUVR	1.22 (1.16, 1.28)
CES‐D	4 (1, 8)
CFQ score (sum of scores / 4 × # answered)	0.28 (0.19, 0.34)
CFQ z score	−0.13 (−0.88, 0.38)
CVLT long delay free recall	11 (9, 13)
CVLT immediate recall	52 (41, 61)
Memory z score	0.19 (−0.74, 0.77)
Number of CFQ visits	3 (1, 3)
Number of CVLT visits	4 (3, 5)
CFQ follow‐up duration (year)	3.45 (0.00, 4.58)
CVLT follow‐up duration (year)	5.06 (4.10, 6.71)

Note: For continuous and categorical variables, we report the median and interquartile range and the N and percentage, respectively.

Abbreviations: APOE, apolipoprotein E; CES‐D, Center for Epidemiological Studies–Depression; CFQ, Cognitive Failures Questionnaire; CVLT, California Verbal Learning Test; ITG, inferior temporal gyrus; PiB, Pittsburgh compound B; SUVR, standardized uptake value ratio.

FIGURE 1.

Longitudinal Cognitive Failures Questionnaire (CFQ) (left) and memory (right) scores versus age. Lower CFQ z scores indicate lower subjective cognitive complaints and higher memory z scores indicate better performance.

3.1. Subjective cognitive complaints

In adjusted linear mixed effects models, we did not find any cross‐sectional associations between amyloid positivity (independent of tau) and CFQ. Amyloid positivity was related to steeper longitudinal increases in CFQ, but this effect was statistically significant only in the entorhinal tau model (estimate = 0.11, standard error [SE] = 0.05, P = $0.041$). In contrast, higher parahippocampal tau SUVR, independent of amyloid status, was associated with higher CFQ at index visit (estimate = 0.25, SE = 0.1, P = $0.015$; Table 2). Higher entorhinal tau SUVR at index visit was associated with an attenuated increase in CFQ over time (estimate = −0.06, SE = 0.03, P = $0.047$; Figure 2). Similarly, we found a statistically significant association between higher entorhinal tau SUVR, independent of amyloid, and attenuated increase over time in the CFQ false triggering factor, whereas the cross‐sectional association with parahippocampal tau SUVR was observed only for the CFQ distractibility score (Tables S8 and S9).

TABLE 2.

Longitudinal CFQ linear mixed effects model results.

	Entorhinal	Hippocampus	Parahippocampal	ITG
Tau PET SUVR	0.13 (SE = 0.11)	0.05 (SE = 0.10)	0.25 (SE = 0.10)	0.07 (SE = 0.12)
	P = 0.255	P = 0.640	P = 0.015*	P = 0.582
Tau PET SUVR × amyloid group	−0.12 (SE = 0.23)	−0.18 (SE = 0.19)	0.13 (SE = 0.20)	−0.03 (SE = 0.23)
	P = 0.615	P = 0.353	P = 0.504	P = 0.897
Tau PET SUVR × time	−0.06 (SE = 0.03)	−0.04 (SE = 0.03)	−0.03 (SE = 0.03)	−0.00 (SE = 0.04)
	P = 0.047*	P = 0.125	P = 0.316	P = 0.974

Note: Estimated coefficients along with their SEs and P values for the tau PET SUVR (i.e., cross‐sectional effect of tau), tau PET SUVR by amyloid group interaction (i.e., the synergistic cross‐sectional effect of tau and amyloid), and the tau PET SUVR by time interaction (i.e., longitudinal effect of tau) terms in linear mixed effects models where longitudinal CFQ Z score is the outcome. Each column corresponds to a separate model. * Indicates P < 0.05.

Abbreviations: CFQ, Cognitive Failures Questionnaire; ITG, inferior temporal gyrus; PET, positron emission tomography; SE, standard error; SUVR, standardized uptake value ratio.

FIGURE 2.

Longitudinal Cognitive Failures Questionnaire (CFQ) trajectories by entorhinal tau standardized uptake value ratio (SUVR) z score based on linear mixed effects model estimates. PET, positron emission tomography; SD, standard deviation.

3.2. Verbal episodic memory

In adjusted linear mixed effects, we did not find any cross‐sectional or longitudinal associations between the amyloid positivity main effect and memory. The cross‐sectional associations of entorhinal and parahippocampal tau SUVR with memory were exacerbated in the presence of amyloid positivity (entorhinal tau × amyloid interaction: estimate = −0.53, SE = 0.25, P = $0.041$; parahippocampal tau × amyloid interaction: estimate = −0.61, SE = 0.22, P = $0.007$). Greater hippocampal tau SUVR was associated with lower concurrent memory scores (estimate = −0.27, SE = 0.11, P = $0.019$) and a steeper decline in memory over time (estimate = −0.03, SE = 0.01, P = $0.040$; Table 3). We found similar cross‐sectional and longitudinal associations between hippocampal tau SUVR and CVLT free recall long delay. Greater parahippocampal tau SUVR was also associated with lower concurrent memory scores (estimate = −0.36, SE = 0.11, P = $0.002$) and individual CVLT components (Tables S10 and S11).

TABLE 3.

Longitudinal memory linear mixed effects model results.

	Entorhinal	Hippocampus	Parahippocampal	ITG
Tau PET SUVR	−0.24 (SE = 0.13)	−0.27 (SE = 0.11)	−0.36 (SE = 0.11)	−0.17 (SE = 0.13)
	P = 0.061	P = 0.019*	P = 0.002**	P = 0.198
Tau PET SUVR × amyloid group	−0.53 (SE = 0.25)	−0.40 (SE = 0.21)	−0.61 (SE = 0.22)	−0.10 (SE = 0.25)
	P = 0.041*	P = 0.062	P = 0.007**	P = 0.697
Tau PET SUVR × Time	−0.00 (SE = 0.02)	−0.03 (SE = 0.01)	−0.00 (SE = 0.01)	−0.00 (SE = 0.02)
	P = 0.830	P = 0.040*	P = 0.776	P = 0.922

Note: Estimated coefficients along with their SEs and P values for the tau PET SUVR (i.e., cross‐sectional effect of tau), tau PET SUVR by amyloid group interaction (i.e., the synergistic cross‐sectional effect of tau and amyloid), and the tau PET SUVR by time interaction (i.e., longitudinal effect of tau) terms in linear mixed effects models where longitudinal memory Z score is the outcome. Each column corresponds to a separate model. * P < 0.05, ** P < 0.01.

Abbreviations: ITG, inferior temporal gyrus; PET, positron emission tomography; SE, standard error; SUVR, standardized uptake value ratio.

3.3. Comparison of CFQ and memory models

To examine whether AD neuropathology is a stronger predictor of subjective cognitive complaints or neuropsychological assessment‐based measures of memory among cognitively normal individuals, we assessed if the cross‐sectional associations of neuropathology with subjective cognitive complaints differed from the associations of neuropathology with verbal episodic memory using SURE. Differences between the standardized regression coefficients estimated in the CFQ and memory models were not statistically significant for amyloid group or regional tau SUVR (Table S14 in supporting information).

4. DISCUSSION

We investigated the independent associations between PET measurements of two hallmark neuropathologies of AD and subjective cognitive complaints among cognitively normal older individuals, adjusting for age, sex, education, and depressive symptoms. Cross‐sectionally, greater tau burden in the parahippocampal gyrus was associated with greater subjective cognitive complaints as assessed using the CFQ. Greater entorhinal tau SUVR was associated with an attenuated increase over a median follow‐up duration of 3.4 years. In contrast, greater hippocampal SUVR was associated with a steeper decline in memory performance over a median follow‐up duration of 5.1 years. When we compared models for CFQ and verbal memory directly, we did not find statistically significant differences between the associations with AD neuropathology of subjective cognitive complaints and memory scores.

The parahippocampal gyrus is part of the medial temporal lobe, a crucial region for memory processing. In agreement with our findings, several studies have also reported associations between tau in the medial temporal lobe and worse cognitive performance and cognitive decline among cognitively unimpaired individuals. ⁴² , ⁴³ An ADNI study similarly found that self‐reported complaints measured by the ECog (measurement of everyday cognition) questionnaire were associated with parahippocampal tau burden as well as parietal, frontal, and global tau, although they found the strongest associations in the frontal lobe, ²³ whereas our study primarily analyzed regions in the temporal lobe. However, this ADNI study included those with mild cognitive impairment (MCI) whereas we restricted our analysis to include only cognitively unimpaired individuals.

The association between greater entorhinal tau and attenuated increases in CFQ over time may be explained by the possibility that subjective cognitive complaints exhibit steep increases early, prior to the elevation in entorhinal tau burden, with the rate of increase in subjective cognitive complaints tapering off later as entorhinal tau burden increases (Figure 3). This possible explanation accounts for both the cross‐sectional associations we observed, with higher tau burden being associated with greater subjective cognitive complaints and the association of higher tau burden with attenuated increases in subjective cognitive complaints. Another possible explanation for this association is that as an individual experiences advancing pathology and presumably poorer objective memory, they may experience a lack of insight into their cognitive deficits, resulting in a greater denial of symptoms. Larger studies examining both subjective cognitive complaints and tau burden longitudinally will be needed to test these hypotheses.

FIGURE 3.

A possible explanation of the observed cross‐sectional and longitudinal associations between tau positron emission tomography (PET) standardized uptake value ratio (SUVR) and Cognitive Failures Questionnaire (CFQ). The trajectories shown in (A) for tau (blue) and CFQ (red) yield a positive cross‐sectional association between tau and CFQ as shown in (B) and a negative association between tau and rate of change in CFQ as shown in (C).

Compared to the total CFQ score, the CFQ false triggering factor, but not forgetfulness or distractibility, had a statistically significant longitudinal association of similar size with entorhinal tau, indicating that our longitudinal finding may be driven by this factor. ³⁰

Our findings differ from previous reports in several ways. First, the Harvard Aging Brain Study found that entorhinal tau PET was associated cross‐sectionally with subjective cognitive decline as measured by the memory functioning questionnaire, everyday cognition battery, and a 7‐item questionnaire. ⁴⁴ While our cross‐sectional results are consistent with this finding, our results did not reach statistical significance. In addition, we did not find a relationship between amyloid positivity and longitudinal change in memory, in contrast to longitudinal studies that have consistently indicated this association, ⁴⁵ , ⁴⁶ including from the BLSA. ⁴⁷ This is likely attributable to our inclusion of participants who remained cognitively normal throughout the longitudinal follow‐up period in these analyses and our limited sample size. Because we restricted the sample to cognitively normal participants, the effect of amyloid and tau on verbal memory decline was likely subtle, therefore requiring a larger sample size to detect a statistically significant association. We note, however, that the effect we estimated was in the expected direction. Additionally, the present study adjusted for regional tau pathology in the amyloid models, whereas the previous analysis in the BLSA did not. Second, unlike a previous BLSA analysis that found that elevated baseline entorhinal tau was linked to steeper memory decline prior to tau PET, ³² we did not find an association between entorhinal tau and rate of memory decline. This difference may be attributed to the extent of CVLT follow‐up in the present study compared to the Ziontz et al. study, as we limited the inclusion of CVLT scores to within 5 years of their index tau PET visit. In our models, amyloid and tau pathologies were assumed to have additive effects on the outcomes we investigated, but research suggests that their effects might be synergistic, with individuals having both pathologies exhibiting steeper cognitive decline. However, we were unable to investigate this synergistic effect due to our limited sample size.

We did not adjust our results for multiple comparisons across the four a priori selected brain regions we investigated for tau pathology. At a per hypothesis type I error rate of $\alpha =0.05$, our family‐wise error rate (FWER) is bounded above by $1-{(1-0.05)}^4\approx 18.5\%$, which is the FWER under the assumption of independence across hypotheses. Given that these four brain regions exhibit moderate to high correlations in their tau SUVRs (Figure S2), the true FWER of our analyses is much lower than this upper bound. The associations with CFQ would not survive multiple comparison correction ensuring FWER ≤ 0.05 (using the Bonferroni, Holm, Hochberg, or Hommel procedure), but the cross‐sectional associations of parahippocampal tau and its interaction with amyloid group on memory would remain statistically significant. Given that these FWER‐controlling procedures are conservative when hypotheses are correlated (as in our case), multiple comparison correction would be better suited in larger samples. Our results should be confirmed in samples with sufficient statistical power that enables the implementation of a multiple comparison correction framework.

This study has limitations. Our study accounted for certain factors that are associated with subjective cognitive complaints, such as severe psychiatric or neurological disorders, which are exclusion criteria for the BLSA, and depressive symptoms, which we adjusted for using CES‐D as a covariate. However, there may be other conditions that affect subjective cognitive complaints that we did not account for, such as personality, medication use, substance use, and cultural background. For instance, one study found that CFQ was more associated with personality domains, such as conscientiousness and neuroticism, rather than objective cognitive performance. ¹⁰ In addition, our PET measures, particularly in the hippocampus, may be confounded by spill‐over of non‐specific binding signal in the choroid plexus. Finally, the BLSA cohort is a highly educated, healthy sample largely comprising White participants. Therefore, it is uncertain how these findings will generalize to a more socioeconomically diverse cohort.

Our study also has important strengths. This study extensively characterized participants through our available longitudinal CFQ measures. We made a direct comparison between subjective and objective measures by investigating associations of tau pathology with CFQ and CVLT. We also created a proportion variable to include those who had missing items on the questionnaire to maximize our sample size. In addition, we used subjective cognitive complaints as a continuous variable, unlike other studies that used subjective cognitive measures to identify participants as part of a subjective cognitive impairment group and then compared to healthy, MCI, and AD groups.

In conclusion, these findings add to our growing knowledge of the relationship between AD pathology and cognition. A previous BLSA study found that subjective cognitive complaints also measured by the CFQ were predictive of declining cognitive performance, particularly in verbal memory. ⁶ Other studies have found that those with subjective memory complaints combined with baseline AD neuropathology may be more likely to develop dementia and cognitive decline subsequently. ⁴⁸ , ⁴⁹ Therefore, our findings support the utility of including measures of subjective cognitive assessments as early indicators for AD. Early amyloid and tau accumulation may have cognitive effects that could be reflected in subjective cognitive questionnaires such as the CFQ. Future studies investigating the relationship between AD neuropathology and subjective cognitive complaints, particularly among larger and more diverse cohorts, can help clinicians detect those at risk for developing AD at the earliest opportunity and develop targeted interventions.

CONFLICT OF INTEREST STATEMENT

The authors of this study declare that there are no conflicts of interest related to this work. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human subjects involved in this study provided informed consent prior to participation.

Supporting information

Supporting Information

Supporting Information

Bannerjee A, Elkins W, Resnick SM, Bilgel M. Alzheimer's disease neuropathology and longitudinal change in subjective cognitive complaints. Alzheimer's Dement. 2025;17:e70176. 10.1002/dad2.70176