Persistent Functional Impairment as an Early Indicator of Alzheimer Disease Pathology and Progression

ABSTRACT

Background

Functional impairment (FI) is a key criterion for diagnosing dementia. However, subtle functional changes may occur during preclinical and prodromal phases but may not be accurately characterized. Furthermore, research linking FI to Alzheimer disease (AD) biofluid biomarkers is limited. Here we examined cross‐sectional associations between cerebrospinal fluid (CSF) AD biomarkers and persistent versus transient FI in dementia‐free older adults, and the longitudinal association of FI with incident dementia.

Methods

Data from 1000 participants (age 72.9 ± 7.0; 45.2% female; 62.8% MCI) from the Alzheimer's Disease Neuroimaging Initiative were analyzed. CSF biomarkers included p‐tau181, Aβ42, and ptau‐181/Aβ42 ratio. Three Functional Activities Questionnaire items of “preparing a hot beverage,” “preparing a balanced meal,” and “shopping alone” were identified by factor analysis as assessing function rather than cognition directly. Persistent‐FI was operationalized as FI present at> two‐thirds of pre‐dementia visits. Comparator groups included Transient‐FI and No‐FI. Linear regression modeled the association between FI status and baseline biomarker levels, while Cox regression assessed the association between FI and incident dementia. Models adjusted for age, sex, education, APOE‐ε4 status, and MMSE.

Results

Compared to No‐FI, Persistent‐FI was associated with lower Aβ42 (Beta = −8.93; 95% CI: −13.56 to −4.03; p < 0.001), higher p‐tau181 (Beta = 10.81; 95% CI: 0.44–22.26; p = 0.041), and ptau181/Aβ42 ratio (Beta = 21.66; 95% CI: 7.02–38.31; p = 0.003). In contrast, Transient‐FI showed no significant associations. APOE‐ε4 carrier status was more prevalent in the Persistent‐FI group compared to No‐FI (p = 0.009), but not in Transient‐FI (p = 0.931). Compared to No‐FI, Persistent‐FI had a 6.66‐fold greater dementia incidence rate (95% CI: 4.98–8.91, p < 0.001), while Transient‐FI had a 1.72‐fold greater incidence rate (95% CI: 1.09–2.72, p = 0.021).

Conclusions

Findings extend the limited research on the association of FI with CSF AD biomarkers in dementia‐free populations. Operationalizing FI‐related risk by persistence enhances prognostication, identifying individuals with greater AD pathology and progression risk. This approach could enhance screening, early detection, and risk stratification, informing timely interventions before dementia onset.

Summary

• Key Points

  • ○
    Functional impairment in dementia‐free older adults is associated with cerebrospinal fluid Alzheimer disease (AD) biomarkers.
  • ○
    Persistent functional impairment, compared to transient or no impairment, is linked to greater prevalent AD pathology in advance of dementia, reflected by higher p‐tau181, lower Aβ42, and an elevated p‐tau181/Aβ42 ratio.
  • ○
    Dementia incidence rate is significantly higher in individuals with persistent functional impairment compared to those with transient or no impairment.
• Why Does This Paper Matter?

  • ○
    This study demonstrates that in a non‐dementia sample, persistent functional impairment is more strongly associated with CSF biomarkers of AD than transient functional impairment, suggesting that sustained impairment may better reflect underlying disease pathology. Distinguishing between transient and persistent symptoms reinforces the potential of functional impairment as a marker for identifying individuals at heightened risk of prevalent and incident AD. Assessing functional changes can be done remotely and at scale, aiding in screening, early detection, and clinical decision‐making.

1. Introduction

Dementia research is increasingly focused on the early identification of individuals with underlying Alzheimer disease (AD) pathology, which is critical for timely intervention and potential treatment. Loss of functional independence in performing instrumental (IADL) and basic (BADL) activities of daily living is a key criterion for diagnosis and clinical staging of dementia [1]. Yet, detecting functional impairment (FI) in early stages of the disease remains a significant challenge. IADL involve tasks critical for maintaining independent living, such as managing medications, preparing meals, shopping, handling finances, and traveling—activities that often demand complex cognitive processing. In contrast, BADL involve fundamental physical activities, including personal hygiene, eating, dressing, and toileting [2]. Recognizing the importance of functional changes in early disease stages, the National Institute on Aging‐Alzheimer's Association diagnostic criteria for mild cognitive impairment (MCI) incorporate subtle decline in functional abilities [3], reflecting potential utility as an early indicator of cognitive decline and dementia.

Longitudinal studies have shown that difficulties in performing ADL among older adults with MCI could predict progression to dementia [4, 5, 6, 7]. While several studies have explored the link between cognitive decline and FI, research examining the relationship between FI and AD biofluid biomarkers remains limited. This gap is particularly notable given the increasing importance of biofluid biomarkers in AD diagnosis and staging [1] and identifying potential candidates for use of targeted therapies in early‐stage disease.

Furthermore, most of the existing instruments for FI assessment lack the sensitivity to capture functional changes in the early stages of neurodegenerative disease. Most instruments may fall short in appropriately accounting for the natural history of functional changes, to distinguish between transient non‐pathological changes in function vs. persistent FI that could signal an underlying neurodegenerative process. Transient FI may arise from temporary disruptions unrelated to neurodegenerative disease, such as acute medical conditions (e.g., infections, pain flareups), sleep disturbances, medication side effects, or short‐term psychological stressors. In such cases, functional changes may reflect a temporary reduction in capacity or motivation rather than progressive decline associated with neurodegenerative disease. By analogy, research on neurobehavioral symptoms has shown that transient symptoms are far less effective at detecting prevalent AD neuropathology [8, 9, 10, 11, 12] and predicting incident dementia than persisting symptoms [13, 14, 15, 16], reflected in the definition of the mild behavioral impairment (MBI) syndrome, which requires symptoms to persist for at least 6 months [17]. This principle may also apply to FI, highlighting the need for more nuanced approaches to functional assessment that distinguish persistent from transitory symptoms.

To address this gap, we adopted a novel approach to FI, informed by the validated approach for MBI case ascertainment [17, 18], which accounts for the natural history of functional changes, distinguishing FI that persists over time from transient FI. In a recent study, we demonstrated that persistent FI in cognitively normal older adults is more strongly associated with incident cognitive decline and dementia than transient or no FI¹⁹. These findings [19] provided the first evidence that the principle of symptom persistence extends to FI. Building on this evidence, here in dementia‐free older adults, we investigated the cross‐sectional association of persistent and transient FI with baseline cerebrospinal fluid (CSF) AD biomarkers, and the longitudinal association with incident dementia. We hypothesized that compared to no FI or transient FI, persistent FI would show stronger associations with CSF AD biomarkers and a greater incidence of dementia. Persistent FI more strongly associating with CSF AD biomarkers would provide compelling evidence that sustained impairments in functional abilities may better reflect underlying disease pathology. This distinction reinforces the idea that, like behavioral changes, appropriately ascertained FI could serve as a valuable early marker for identifying individuals at heightened risk of AD, supporting use in screening, early detection, and clinical decision‐making.

2. Methods

2.1. Study Population

Participant data were obtained from the September 2024 release of the Alzheimer's Disease Neuroimaging Initiative (ADNI), a collaborative effort involving multiple centers across North America. Launched in 2003, ADNI aims to track participants through periods of cognitive decline and dementia by evaluating biomarkers, neuroimaging, and neuropsychological status. ADNI has established some of the following inclusion criteria: participants must be aged 55–90 years, have a Hachinski Ischemic Score of ≤ 4, and a Geriatric Depression Scale score of < 6. Participants must also demonstrate adequate visual and auditory acuity for neuropsychological testing, maintain good general health without any conditions that would prevent enrollment, and have at least a sixth‐grade education. For a complete list of inclusion and exclusion criteria by cognitive category, please refer to the ADNI protocol (https://adni.loni.usc.edu/wp‐content/themes/freshnews‐dev‐v2/documents/clinical/ADNI‐1_Protocol.pdf). The ADNI Conversion Committee reviews individual participant reports and provides a consensus diagnosis. Details of clinical diagnoses have been previously described elsewhere (https://adni.loni.usc.edu/wp‐content/uploads/2010/09/ADNI_GeneralProceduresManual.pdf). All ADNI participants provided informed consent to participate in the study, and the ethics committee approval to conduct this study was received at contributing ADNI sites. The investigators within ADNI (https://adni.loni.usc.edu/wp‐content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf) contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report.

2.2. Participant Selection

ADNI participants were included based on the following criteria: (1) age 55 or older; (2) no dementia diagnosis at baseline; (3) complete Functional Activities Questionnaire (FAQ) [20] data for all visits prior to dementia diagnosis; (4) available baseline CSF p‐tau181 and amyloid‐beta42 (Aβ42) measures; and (5) available baseline demographic covariates data required for modeling. For the longitudinal analysis, participants missing follow‐up visits were further excluded (Figure 1). The final sample included participants classified as cognitively unimpaired (CU) or with MCI. CU was defined by a Clinical Dementia Rating (CDR) global and memory box score of 0, while MCI was defined by a CDR global score of 0.5 and a memory box score ≥ 0.5. Participants with MCI may show mild functional difficulties (e.g., FAQ > 0) but must remain independent in daily activities. Diagnosis also incorporated performance on the Mini‐Mental State Examination (MMSE) and Wechsler Logical Memory II subscale, with education‐adjusted cutoffs. Detailed diagnostic procedures are available in the ADNI documentation (https://adni.loni.usc.edu/help‐faqs/adni‐documentation/).

FIGURE 1.

Flowchart demonstrating the participant selection process. CSF, cerebrospinal fluid; FAQ, Functional Activities Questionnaire; FI, functional impairment.

2.3. FI Operationalization

FI status was derived from the FAQ [20], an informant‐reported screening tool that assesses the level of dependence in performing 10 IADLs: “paying bills,” “managing taxes,” “shopping alone,” “playing games,” “preparing a hot beverage,” “preparing a balanced meal,” “tracking current events,” “traveling outside the neighborhood,” “paying attention to and understanding a TV program,” and “remembering appointments and medications.” Each item is scored from 0 (normal) to 3 (dependent) [20]. Recognizing the relationship between cognition and function, certain FAQ items show overt overlap with cognitive symptoms (e.g., “remembering appointments”), making it challenging to isolate primarily functional abilities. To reduce this overlap and identify FAQ items that predominantly assess functional ability, we performed an exploratory factor analysis (EFA) on FAQ item scores, averaged across all study visits prior to dementia onset for each participant. The ideal number of factors was identified using a scree plot, with a two‐factor solution the most appropriate for the data. The EFA identified three FAQ items—“preparing a hot beverage,” “preparing a balanced meal,” and “shopping alone”—that loaded onto a factor representing functional ability (Figure 2). Only these three items were included in subsequent analyses, with an FI composite score calculated by summing individual item scores. Given that these tasks may still engage a certain level of cognitive processing, all analyses were adjusted for MMSE scores to further minimize the potential influence of cognitive decline on the assessment of function.

FIGURE 2.

Exploratory factor analysis of the Functional Activities Questionnaire (FAQ) items. The 10 FAQ items were analyzed and loaded onto two distinct factors. Factor 1 is interpreted as representing functional abilities, while Factor 2 reflects cognitive abilities. The numbers on the arrow lines indicate factor loadings, representing the strength of the relationship between each FAQ item and the respective factor.

Persistent‐FI was defined as an FI composite score ≥ 1 at more than two‐thirds of all study visits (TTV) prior to dementia onset, following a validated approach previously operationalizing MBI [18] from conventional measures of neuropsychiatric symptoms (NPS). Comparator groups included those with transient FI not meeting persistence criteria (Transient‐FI) and those with an FI score of zero (No‐FI) in advance of dementia. This longitudinal FI algorithm captured sustained patterns of change rather than baseline‐only functional limitations. It is important to note that, under this approach, MCI participants with mild FI at baseline who showed FI at the next follow‐up visit, sufficient to meet criteria for dementia at that visit, were classified as persistent‐FI at baseline. We selected this longitudinal FI algorithm to emphasize the persistence of FI over time, given that the FAQ [20] assesses current or recent functional ability without specifying a strict timeframe, thereby ignoring the natural history of impairment. We acknowledge, however, that alternative approaches could address this limitation, including the use or development of functional measures that inherently capture symptom persistence over longer periods (e.g., 6 months), as implemented in the assessment of neuropsychiatric symptoms using the MBI Checklist [21].

2.4. Cerebrospinal Fluid Biomarkers

CSF biomarkers of interest included p‐tau181, Aβ42, and the ptau‐181/Aβ42 ratio, measured in pg/ml at baseline. Biomarker levels were measured using the Research Use Only (RUO) INNO‐BIA AlzBio3 immunoassay (Fujirebio, Belgium) and were further re‐scaled against the original baseline dataset (UPENNBIOMK) as the gold standard. A detailed overview of the re‐scaling procedures is provided elsewhere (https://adni.bitbucket.io/reference/docs/UPENNBIOMK_MASTER/ADNI_Methods_Template_Shaw%20Figurski%20Waligorska%20Trojanowski%20overview%20for%20CSF%20Ab1‐42%20tau%20and%20ptau181%20AlzBio3%20immunoassay%20datav3%20(5).pdf).

2.5. Statistical Analysis

Participant age, sex, education years, APOE‐ε4 status, and MMSE scores were compared across FI groups at baseline to identify significant differences, using t‐tests for continuous variables and Chi‐squared tests for categorical variables. Biomarker measurements for both p‐tau181 and Aβ42 were winsorized at 95% to minimize the impact of outliers without excluding them.

To examine the cross‐sectional association between FI status and baseline biomarker levels, linear regression models were fitted with FI status as the primary exposure and biomarker concentrations as the outcome. Since biomarker measurements were log‐transformed due to skewness, the estimated model coefficients were exponentiated to return to the original scale. To facilitate interpretation, the obtained coefficients were then reported as percentage differences in biomarker levels. Separate models were fitted for p‐tau181, Aβ42, and ptau‐181/Aβ42. All models adjusted for age, sex, education years, APOE‐ε4 status, and MMSE scores. Interaction terms were included to assess potential effect modification by sex, APOE‐ε4 status, and MMSE score.

To assess the longitudinal association between FI status and incident dementia, a Cox proportional hazards model was employed to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for progression to dementia across FI groups. The model adjusted for age, sex, education years, APOE‐ε4 status, and MMSE scores. Interaction effects paralleling those in the cross‐sectional model were also examined.

All statistical analyses were conducted in RStudio version Prairie Trillium (2022.02.0) (R version 4.2.2). Factor analysis was performed using the psych package, winsorization using the DescTools package, Cox models using the survival package, and forest plot using ggplot2 and survminer. The proportional hazards assumptions were assessed using the cox.zph function from the survival package.

3. Results

3.1. Sample Demographics

The final sample for the cross‐sectional analysis included 1000 participants (840 with No‐FI, 58 with Transient‐FI, and 102 with Persistent‐FI), with 372 CU participants and 628 with MCI. Over a mean follow‐up of 4.9 years (SD 3.6), 270 participants progressed to dementia, including 26 CU and 244 with MCI.

The Transient‐FI group was more educated than the No‐FI group (p = 0.018), while the Persistent‐FI group did not differ (p = 0.203). The Persistent‐FI group had a lower percentage of female participants compared to No‐FI (p < 0.001). MCI diagnosis was more prevalent in both Transient‐ (87.9%; p < 0.001) and Persistent‐FI groups (99.0%; p < 0.001) compared with no FI (56.7%). APOE‐ε4 carrier status was more prevalent in the Persistent‐FI group compared to No‐FI (53.9% vs. 39.9%; p = 0.009), while the Transient‐FI group did not differ significantly (41.4%; p = 0.931). Furthermore, Persistent‐FI participants had higher CSF p‐tau181 levels (41.7 pg/mL vs. 35.2 pg/mL; p = 0.003), lower Aβ42 levels (157 pg/mL vs. 185 pg/mL; p < 0.001), and higher ptau‐181/Aβ42 ratio (0.297 vs. 0.223; p < 0.001) on average, compared to No‐FI participants, while the Transient‐FI group did not differ significantly (Table 1).

TABLE 1.

Baseline demographic characteristics of dementia‐free participants across functional impairment (FI) groups. Between‐group differences were assessed using t‐tests for continuous variables and Chi‐squared tests for categorical variables, with No‐FI as the reference group.

	No‐FI (N = 840)	Transient‐FI (N = 58)	Persistent‐FI (N = 102)	Transient‐FI p‐value	Persistent‐FI p‐value
Age
Mean (SD)	72.8 (6.9)	73.4 (7.4)	74.0 (7.3)	0.556	0.111
Education years
Mean (SD)	16.2 (2.7)	17.0 (2.4)	15.8 (2.7)	0.018	0.203
Sex
Female	404 (48.1%)	20 (34.5%)	28 (27.5%)	0.061	< 0.001
Diagnosis
CU	364 (43.3%)	7 (12.1%)	1 (1.0%)	< 0.001	< 0.001
MCI	476 (56.7%)	51 (87.9%)	101 (99.0%)
MMSE score
Mean (SD)	28.3 (1.7)	27.9 (1.7)	27.5 (1.7)	0.066	< 0.001
APOE‐ε4 status
Carrier	335 (39.9%)	24 (41.4%)	55 (53.9%)	0.931	0.009
p‐tau181 level
Mean (SD)	35.2 (18.3)	38.2 (20.9)	41.7 (20.6)	0.301	0.003
Aβ42 level
Mean (SD)	185 (51.7)	178 (49.9)	157 (46.0)	0.271	< 0.001
p‐tau181/Aβ42
Mean (SD)	0.223 (0.164)	0.254 (0.194)	0.297 (0.176)	0.240	< 0.001

Note: Bold p‐values indicate statistical significance (p < 0.05).

Abbreviations: CU, cognitively unimpaired; FI, functional impairment; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; SD, standard deviation.

3.2. Cross‐Sectional Association Between FI Status and CSF Biomarkers of AD

Linear regression models demonstrated an association between FI status and CSF biomarker levels at baseline, after multivariable adjustments. Participants with persistent FI had 10.81% higher CSF p‐tau181 levels (95% CI: 0.44–22.26, p = 0.041), 8.93% lower Aβ42 levels (95% CI: −13.56 to −4.03, p < 0.001), and 21.66% higher ptau‐181/Aβ42 ratio (95% CI: 7.02–38.31, p = 0.003). The Transient‐FI group did not differ significantly from the No‐FI group across all biomarker levels (p‐tau181: Beta = 4.36%, 95% CI: −7.93–18.30, p = 0.504; Aβ42: Beta = −2.19%, 95% CI: −8.50–4.55, p = 0.515; ptau‐181/Aβ42: Beta = 6.70%, 95% CI: −9.39–25.66, p = 0.436) (Table 2). No significant interaction effects were identified between FI status and sex, APOE‐ε4 status, and MMSE score.

TABLE 2.

Cross‐sectional associations between cerebrospinal fluid AD biomarkers p‐tau181, Aβ42, and p‐tau181/Aβ42 ratio and functional impairment (FI) status (No‐FI, Transient‐FI, Persistent‐FI). The No‐FI group served as the reference group for FI status.

Outcome	Predictor	Beta a (%)	CI (%)	p
p‐tau181	Transient‐FI	4.36	−7.93 to 18.30	0.504
	Persistent‐FI	10.81	0.44 to 22.26	0.041
Aβ42	Transient‐FI	−2.19	−8.50 to 4.55	0.515
	Persistent‐FI	−8.93	−13.56 to −4.03	< 0.001
p‐tau181/Aβ42	Transient‐FI	6.70	−9.39 to 25.66	0.436
	Persistent‐FI	21.66	7.02 to 38.31	0.003

Note: All models were adjusted for age, sex, education, APOE‐ε4 status, and MMSE scores. Bold p‐values indicate statistical significance (p < 0.05).

Abbreviations: CI, confidence interval; FI, functional impairment.

^a Beta coefficients represent the estimated percent difference in biomarker level.

3.3. Longitudinal Association Between FI Status and Incident Dementia

The final sample for the longitudinal analysis included 972 participants, with 815 in the No‐FI group, 58 in Transient‐FI, and 99 in Persistent‐FI. Compared to the No‐FI group, the Persistent‐FI group had a 6.66‐fold greater adjusted hazard for dementia (CI: 4.98–8.91, p < 0.001), while the Transient‐FI group had a HR of 1.72 (CI: 1.09–2.72, p = 0.021) (Figure 3). No significant interaction effects were observed between FI status and sex, APOE‐ε4 status, and MMSE score.

FIGURE 3.

Forest plot of hazard ratios for incident dementia across functional impairment (FI) groups, adjusted for age, sex, education years, APOE‐ε4 status, and MMSE scores. CI, confidence interval; FI, functional impairment; MMSE, mini‐mental state examination.

4. Discussion

The present study examined the association between FI status and AD CSF biomarker levels in older adults with unimpaired cognition or MCI. Findings revealed that individuals with persistent FI exhibited significantly higher CSF p‐tau181, lower Aβ42, higher ptau‐181/Aβ42 ratio, higher percentage of APOE‐ε4 carriers, and a 6.66‐fold greater hazard for incident dementia compared to those with no FI. In contrast, participants with transient FI did not show significant differences in CSF biomarker levels or APOE‐ε4 carrier status and had only a 1.72‐fold greater dementia incidence rate. These findings suggest that, like persistent behavioral symptoms, sustained changes in functional abilities are more strongly associated with AD‐related pathology, APOE‐ε4 carriership, and greater risk of progression to dementia. These results add to the evidence base for appropriately measured symptoms representing AD proteinopathies, highlighting the importance of natural history in risk assessment and the potential of FI as an early marker of dementia.

Difficulty in performing IADL is common among individuals with MCI and is associated with a higher likelihood of progression to dementia compared to those with normal cognition [4, 5, 7, 22, 23, 24, 25]. Several positron emission tomography (PET) studies have also explored the association between IADL impairment and AD biomarkers, indicating a link between Aβ plaques and lower IADL abilities in preclinical and prodromal AD [26, 27]. Two cross‐sectional studies in dementia‐free participants demonstrated that IADL impairments were associated with greater Aβ‐PET burden in participants with MCI, relative to cognitively normal individuals [27, 28, 29]. Earlier along the cognitive continuum, a cross‐sectional study using the Financial Capacity Instrument Short Form (FCI‐SF) demonstrated that greater Aβ‐PET burden in cognitively normal participants was associated with greater difficulties in completing financial tasks [30]. Longitudinal studies have reported consistent findings. One three‐year study in participants with subjective cognitive decline (SCD) observed that those with elevated Aβ‐PET burden experienced greater decline in IADL performance [31]. Similarly, another three‐year study of dementia‐free older adults revealed a trend towards deteriorating IADL performance among Aβ+ participants compared to Aβ− [32]. Together, these PET studies indicate an association between FI and Aβ burden even in the preclinical stages of dementia.

Complementary evidence from studies of biofluid AD biomarkers further supports the link between IADL impairment and AD pathology. In early analyses of the ADNI cohort, including participants with normal cognition, MCI, or mild‐AD, lower baseline CSF Aβ42 and higher total tau (t‐tau) and p‐tau181 were associated with greater IADL impairment over time [33]. These findings were echoed in another ADNI study involving participants spanning the cognitive continuum, where IADL impairment was associated with lower CSF Aβ42 and greater t‐tau [34]. Similarly, a third ADNI study found that baseline CSF Aβ42 levels were predictive of IADL change over a three‐year period in cognitively normal and MCI individuals [35]. All three studies used the FAQ total score to assess IADL performance, thus including the overtly cognitive items. A more recent study examining driving behavior as an important component of IADL in cognitively normal older drivers found that elevated plasma Aβ40 and Aβ42 levels were associated with fewer night‐time trips and more instances of speeding, respectively. Furthermore, higher CSF t‐tau and p‐tau181 levels were associated with shorter trip durations, while a higher Aβ42/Aβ40 ratio predicted greater acceleration [36]. These converging findings from PET, CSF, and plasma biomarker studies underscore a consistent relationship between AD pathophysiology and FI, reinforcing the utility of IADL performance as an early marker of disease progression.

Our findings on the association of persistent FI with CSF AD biomarker and incident dementia in dementia‐free older adults add to the evidence base, emphasizing the importance of accounting for the natural history of FI. These results suggest that persistent FI may be a more robust indicator of dementia‐related pathology and progression risk, compared to single time‐point measures. A strength of the present study was the adoption of a novel approach to examine FI by considering its persistence across multiple study visits before the onset of dementia, consistent with the successful approach used for behavioral change [37]. Most of the existing FI assessment scales lack a well‐defined temporal framework for evaluating symptoms, especially with respect to the natural history of FI. When FI is evaluated at a single time point or within a short timeframe, it becomes challenging to distinguish between impairments arising from transient factors, such as acute illness, medication side effects, psychological stress, or environmental challenges, versus those signaling a neurodegenerative disease process. Our study distinguished between persistent and transient FI by capturing the persistence of symptoms over at least two‐thirds of annual study visits (TTV), a validated approach to identify disease markers in advance of dementia [18]. This distinction revealed that FI that persisted over time was strongly associated with higher CSF p‐tau181, lower Aβ42, and higher p‐tau181/Aβ42 ratio and greater dementia incidence rate over time. Conversely, transient FI, which did not meet the persistence criterion, showed no significant association with CSF biomarker levels or dementia incidence rate. These results align with our previous findings that persistent FI in cognitively normal older adults is more strongly associated with incident cognitive decline and dementia than transient or no FI [19]. Our classification of persistent FI was designed to identify individuals showing sustained loss of functional abilities, rather than isolated or transient difficulties. While this operationalization may not be universally adopted, we believe it offers distinct clinical value by capturing a trajectory of decline that is more predictive of progression to dementia. From a clinical standpoint, identifying persistent FI could help clinicians and families anticipate a more continuous functional decline, facilitating early planning and support, whereas transient or impersistent FI may reflect situational or reversible factors with different prognostic implications. Furthermore, compared to standard confrontational cognitive testing, evaluating functional decline provides a more practical and culturally adaptable strategy for early detection. This complementary approach is based on real‐world abilities rather than test scores. Unlike cognitive testing, which can be influenced by language, education, or cultural factors, sustained changes in daily function may offer a more universally applicable indicator of risk, based on naturalistic observations over time in comparison to one's baseline. Thus, incorporating functional changes can provide valuable prognostic information, even when cognitive scores remain within normal limits, or if scores are biased by language, education, or culture.

An additional finding of our study was the higher prevalence of APOE‐ε4 carriers in the Persistent‐FI group compared to No‐FI and Transient‐FI (53.9% vs. 39.9% and 41.4%, respectively). Given the well‐established association between APOE‐ε4 carrier status and greater AD risk relative to noncarrier [38], this finding suggests that persistent FI may reflect underlying neurodegenerative processes associated with APOE‐ε4 risk. From a practical perspective, it appears that identifying those with Persistent‐FI may also identify a group enriched for APOE‐ε4 carriership, supporting our approach to identifying risk and the utility of FI in screening.

Aside from accounting for the natural history of FI, another strength of our study was the use of exploratory factor analysis to identify FAQ items that primarily assess functional abilities rather than overt cognitive abilities. EFA in our study identified “preparing a hot beverage,” “preparing a balanced meal,” and “shopping alone” as FAQ items that were most reflective of functional ability, relatively independent of cognition. The fact that we found no significant interaction effect with MMSE score further supports the utility of this approach. In contrast, the remaining FAQ items loaded on a separate factor more closely associated with cognitive demands. While previous studies have attempted to identify FAQ items that best predicted progression from normal cognition to MCI, most of the items identified seem to primarily measure cognitive abilities rather than function. For instance, FAQ items of “remembering appointments” and “assembling tax records” best distinguished between cognitively normal and MCI individuals [39, 40]. While all FAQ items involve a certain level of cognitive processing (e.g., attention, executive function, and memory), items like “remembering appointments” or “paying attention to and understanding a TV program” are clearly proxies for memory and attention, and caregivers report on these cognitive domains when completing the questionnaire. Unsurprisingly, poorer scores on these items are associated with progression to MCI, as they overtly capture decline in cognition rather than function. On the other hand, FAQ items such as “preparing a hot beverage,” “preparing a balanced meal,” and “shopping alone” primarily reflect functional independence to the caregiver, although they still require cognitive processing and coordination. Our study aimed to focus on FAQ items that are less overtly reliant on cognitive abilities, in order to explore more specifically the role of FI as an early marker of disease progression and underlying proteinopathies. Indeed, the distinction between cognition and function is not always clear; but the factor analysis did provide a data‐driven approach to the distinction in this dataset. This distinction has not been explicitly addressed in prior studies exploring the association between IADL (as measured by FAQ) and CSF biomarkers. It is possible that the associations observed in previous studies between CSF p‐tau181 and Aβ42 and IADL impairment, as reflected by the FAQ total score, may have been confounded by the inclusion of FAQ domains more heavily reliant on cognitive function—domains that are, unsurprisingly, more strongly associated with AD biomarkers.

In summary, our findings emphasize the significance of accurately assessing FI as a potential marker for dementia risk and underlying pathology. Incorporating symptom persistence into assessment scales for FI will enhance their specificity for symptoms that are related to the neurodegenerative processes of dementia. Accurate assessment of early functional decline provides a practical and culturally adaptable approach to early detection, capturing real‐world abilities that can indicate risk even when traditional cognitive tests remain normal, making it a universally relevant marker across diverse populations.

5. Limitations

One limitation of the present study is the use of FAQ in a dementia‐free sample. Originally designed for dementia populations, FAQ may lack sensitivity to mild FI in preclinical and prodromal stages of dementia [23]. Additionally, the scope of FAQ is limited in capturing modern IADL, such as technology use [41] or operating household electrical appliances, which are covered by more contemporary scales [25, 42]. Furthermore, limiting analyses to participants with available CSF biomarker data may have introduced selection bias, as lumbar punctures are more often performed in those with cognitive decline. Consequently, our sample was enriched for MCI, which may limit generalizability. However, other demographic and clinical characteristics were comparable (Table S1), suggesting minimal impact on the observed associations. Our findings might have more significant implications if the sample were solely preclinical disease, where cognitive testing is within the normal range, and persistent FI could capture early risk and disease pathology possibly not well captured by conventional confrontational cognitive testing. However, the small number of CU individuals in our sample limited statistical power, making it difficult to draw meaningful conclusions from this subgroup. One might speculate that the higher prevalence of MCI in Transient‐ and Persistent‐FI groups, compared to No‐FI, suggests that those with greater cognitive impairment may be driving the observed effects associated with Persistent‐FI. However, all models were adjusted for MMSE scores, with no significant interactions detected with MMSE, minimizing the likelihood that differences in global cognition account for the findings. Moreover, the higher prevalence of APOE‐ε4 carriership in the Persistent‐FI group adds support to the notion that persistence confers risk. Finally, our definition of persistent FI, while based on established precedent, reflects one possible operationalization of longitudinal functional change and may differ from other approaches. Future work should evaluate whether different definitions of FI persistence yield comparable predictive utility for clinical outcomes and focus on developing functional scales that inherently assess symptoms while capturing their natural history over extended periods (e.g., at least 6 months). Our study serves as a starting point to explore these issues further.

6. Conclusion

Our findings demonstrate that assessing the persistence of FI as a risk factor enhances the detection of underlying disease pathology and improves dementia prognostication. This finding emphasizes the critical role of FI evaluation in early dementia detection and the need for continued research to refine assessment tools for detecting FI more accurately.

Author Contributions

All authors contributed to data analysis, interpretation, and manuscript preparation. M.G. and Z.I. conceived and designed the study. M.G. led data preparation and analysis and drafted the initial version of the manuscript. E.E.S. and Z.I. contributed substantially to the interpretation of findings. All authors had full access to the study data, participated in revising the manuscript, and approved the final version for submission.

Funding

Z. Ismail is supported by the Canadian Institutes of Health Research (BCA527734), Gordie Howe CARES, and the NIHR Exeter Biomedical Research Centre.

Consent

All ADNI participants provided informed consent to participate in the study, and the ethics committee approval to conduct this study was received at contributing ADNI sites.

Conflicts of Interest

Z.I. has served as an advisor/consultant to CADTH, Eisai, Lilly, Lundbeck/Otsuka, Novo Nordisk, and Roche. E.E.S. has provided consulting (unpaid) to Alnylam, Eisai, and Lilly. M. Ghahremani reports no conflicts of interest relevant to this manuscript.

Supporting information

Table S1: Baseline demographic and clinical characteristics of participants included in the analyses versus those excluded due to missing cerebrospinal fluid (CSF) biomarker data (p‐tau181, Aβ42). Between group differences were assessed using t‐tests for continuous variables and Chi‐squared tests for categorical variables. Bold p‐values indicate statistical significance (p < 0.05). CU, cognitively unimpaired; FI, functional impairment; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; SD, standard deviation.

Ghahremani M., Smith E. E., and Ismail Z., “Persistent Functional Impairment as an Early Indicator of Alzheimer Disease Pathology and Progression,” Journal of the American Geriatrics Society 74, no. 2 (2026): 447–457, 10.1111/jgs.70247.