Cognitive reserve as a predictor of cognitive decline, but not age of diagnosis in patients with possible young-onset Alzheimer's disease: An underexplored population

Abstract

Background

The cognitive reserve (CR) theory aims to explain the disparity often observed between brain damage and its clinical manifestation.

Objective

To explore the CR theory in young-onset Alzheimer's disease (YOAD), a population not previously investigated in this context. The goal is to assess whether, similar to late-onset Alzheimer's disease (LOAD), a high CR delays diagnosis but may accelerate cognitive decline.

Methods

This is a retrospective study including 72 patients (ages: 46–64) who were diagnosed with possible YOAD. They were followed up for three years, using the Mini-Mental State Examination.

Results

Unlike the findings with LOAD, age of diagnosis of the YOAD did not correlate significantly with CR variables, years of education or family size. However, years of education predicted greater cognitive decline in the first year, and women showed increased deterioration. Family size showed inconsistent associations, highlighting its limitations.

Conclusions

In contrast to studies among LOAD, there was no significant correlation between age of diagnosis and CR among YOAD, suggesting that other mechanisms might be more influential than CR parameters in younger individuals. However, similar to LOAD, YOAD patients with higher education experienced faster disease progression, implying that diagnoses are frequently made when brain pathology is already severe. These findings reinforce the growing perspective that YOAD and LOAD may constitute distinct forms of AD, each with unique clinical and pathological features. They also underscore the urgent need for early detection tools and cognitive interventions to ease the challenges faced by these young patients.

Introduction

Dementia represents a significant decline in cognitive function beyond what is typical for one's biological age. While dementia primarily affects older individuals, with prevalence increasing with age, there is another category known as early or young-onset Alzheimer's disease (YOAD), occurring before the age of 65. Initially termed “Presenile Dementia”, this condition was first described by Aloïs Alzheimer, who documented features of dementia in a 51-year-old patient. ¹ YOAD prevalence is estimated to be approximately 119 per 100,000, with ages ranging from 30 to 64. ²

There are significant differences between YOAD and late-onset dementia (LOAD), leading some researchers and clinicians to argue that these differences warrant distinct categorization of the disease.^3,4 Notable variances between these types include genetic characteristics, with certain genetic markers being more prevalent in YOAD.^5,6 Autosomal dominant familial genetic cases with known mutations, such as in APP, PSEN1, and PSEN2, are also much more common in YOAD. Specifically, individuals with YOAD have a higher frequency of two apolipoprotein E (APOE) ε4 alleles. APOE is the primary susceptibility gene for AD, ⁷ and the presence of an APOE ε4 allele accounts for approximately 9.12% of the heritability of YOAD. ⁵ Additionally, individuals with YOAD tend to experience a faster rate of decline and earlier mortality, compared to those with LOAD. ³ Clinical, cognitive and neuropsychological manifestations also differ: patients with YOAD frequently present language impairments, visuospatial difficulties, and executive function impairments, along with significant praxis impairment. By contrast, LOAD more frequently manifests memory deficits, with patients showing worse performance in visual memory and orientation.^4,5 Although YOAD accounts for a smaller proportion of dementia cases, its impact is disproportionately severe. It affects individuals in their most active and productive years—disrupting employment, placing financial strain on families, and intensifying emotional stress, often while dependent children are still at home. ⁸ Emerging research also suggests possible gender differences in disease progression, with some studies indicating that women may experience faster cognitive decline than men.^9,10 Delayed diagnoses and a lack of tailored services further deepen the burden. Recognizing and responding to these distinct challenges is critical for developing more effective, equitable care pathways. ¹¹

Thus far, no cure or preventive treatment has been discovered for AD and other forms of dementia, though there have been some developments in this area.^12–14 Current approaches focus on symptomatic management, such as enhancing cholinergic transmission by using cholinesterase inhibitors (ChE-I) such as donepezil ¹⁵ and rivastigmine, ¹⁶ along with memantine, an N-methyl-D-aspartate receptor antagonist. ¹⁷ In LOAD, ChE-I provides moderate cognitive improvements, as evidenced by a meta-analysis showing mean Mini-Mental State Examination (MMSE) score enhancements after 3, 6, and 12 months of treatment. ¹⁸ However, the optimal duration of ChE-I therapy remains uncertain due to limited observation periods. ¹⁹ Conversely, for YOAD, treatment typically also involves ChE-I and memantine, but unfortunately, these medications do not modify the progression of the disease.^20,21

In individuals diagnosed with AD and other brain diseases (e.g., traumatic brain injury ²² ; Parkinson's disease), ²³ research has discovered intriguing discrepancies: postmortem examinations often reveal notable brain pathologies in patients who did not exhibit corresponding cognitive impairments during their lifetime. This discordance has prompted exploration into theories such as cognitive reserve (CR) and brain reserve (BR), to reconcile the difference between observed brain changes and clinical symptoms.^24,25 According to the CR theory, certain life experiences, intelligence, and inherent traits collectively contribute to resilience against cognitive decline, by establishing a reserve of brain resources capable of compensating for age- or brain disease-related changes.^26,27 A higher CR may delay or mitigate the clinical progression of the disease, potentially postponing the diagnosis of the disease.^28,29 CR variables encompass factors such as years of education, occupation, socioeconomic status (SES), leisure activities, and IQ level.^22,24,30 The neural mechanisms underlying this protective effect have been extensively investigated, with evidence suggesting both structural and functional compensatory processes.^31,32

Studies have found that education is a pertinent variable for assessing CR, particularly in advanced dementia stages. ^30,33,34 Occupational complexity and leisure activities have also demonstrated significant contributions to CR, with differential effects on various cognitive domains.^22,35 A less common variable for assessing SES is family size (more children indicating lower SES),^36,37 although research has also incorporated additional indicators, such as neighborhood deprivation, income levels, and social mobility, all showing significant associations with dementia risk and cognitive decline.^38,39

The theoretical model of CR represents a complex construct that poses significant measurement challenges.^40,41 No single standardized method exists for quantifying CR, resulting in considerable heterogeneity across studies in terms of both proxy variables and assessment methodologies.^42,43 While years of education remains the most frequently utilized proxy, researchers increasingly recognize its limitations as a standalone measure.^44,45 This has prompted the development of composite indices, incorporating multiple CR components to better capture the construct's multidimensional nature.^22,46,47

To the best of our knowledge, CR among patients with YOAD has not been explored. Given both the shared characteristics and distinct differences between various forms of AD, this study aims to investigate whether a higher level of CR in YOAD may delay or mitigate the clinical manifestations of dementia as reported in LOAD. Another goal is to identify demographic variables (e.g., gender, country of origin) and CR measures that influence the progression of cognitive decline in YOAD patients receiving symptomatic medications. Our hypothesis is that like patients with LOAD, YOAD individuals with a higher CR would also experience a delayed onset of clinical symptoms, while at the same time, once symptoms manifest, the rate of decline would be accelerated compared to that of individuals with lower CR.^25,48

Methods

Participants and procedures

This research was consistent with the relevant ethical guidelines and received approval from the institutional IRB. This retrospective study explores a clinical database from patients diagnosed with AD who underwent extended monitoring at the Hadassah Mount Scopus Neuro-Geriatric and Memory Clinic, Israel. The clinic's extensive database covered 2040 patients with dementia receiving care between 1996 and 2022. Inclusion criteria for this study were limited to patients diagnosed with possible or probable AD under the age of 65. From this cohort, 72 patients met these criteria (average age 59.77 (SD = 4.08, range = 46–64 years).

Diagnosis was conducted by neurologists following the standard criteria appropriate for the time period of assessment. Given the study's initiation in 1990, NINCDS-ADRDA criteria ⁴⁹ were initially employed, with diagnostic approaches evolving to incorporate updated guidelines such as the AA-NIA criteria ⁵⁰ as they became established in clinical practice. Diagnosis relayed on significant decline in cognitive abilities and daily function, as reported either by the patient or a caregiver, and as evident in cognitive assessments. ⁵¹ At the diagnostic stage, neurologists routinely considered caregiver reports of functional and executive impairments, and when appropriate, referred patients for neuropsychological evaluation to support differential diagnosis between YOAD and mild cognitive impairment or other types of dementia. However, in follow-up visits, MMSE scores were most often the only cognitive measure consistently recorded in clinical files. To rule out other causes of dementia, patients underwent brain imaging (MRI or CT scans) and comprehensive laboratory assessments, including B12 and thyroid function tests. Biological and genetic markers—such as plasma or cerebrospinal fluid biomarkers, were gradually incorporated into the diagnostic process, as they became clinically validated. While these were available for some patients, they were not uniformly applied across the cohort, in order to maintain methodological consistency throughout the retrospective timeline. Clinical follow-ups occurred every 6–9 months, during which cognitive deterioration was monitored and recorded, utilizing the MMSE.

Measures

Demographic and clinical data were extracted from hospital records. Demographic data included age, gender, education, marital status, ethnicity, country of origin and family history of memory diseases. The epidemiological characteristics of AD include the year of diagnosis and the drug treatment (e.g., donepezil, rivastigmine, galantamine, memantine). CR variables included years of education as the primary proxy and family size (number of siblings) as a potential proxy. MMSE is a 30-point screening tool that assesses multiple cognitive domains. ⁵² It was administered at each follow-up by a neurologist. For consistency, MMSE scores were recorded for all patients at fixed intervals: six months, one year, two years, and three years (after three years, the number of the sample became too small). While some patients had assessments at various time points, scores were aligned with these set time points, allowing for a margin of up to three months before or after each interval. Cognitive decline was assessed via the MMSE difference, measuring the difference between the initial assessment and subsequent time points (six months, one year, two years, and three years).

Statistical analyses

Data was entered into a Microsoft Excel file (Microsoft, Redmond, WA, USA), then transferred to a statistical analysis program (SPSS 26.0, Chicago, IL, USA). Additionally, R software (version 4.5.0) was used for further analysis. We conducted LASSO regression analyses to predict the age of diagnosis and cognitive decline. Given that CR has been studied previously in LOAD but not in YOAD, we used our prior study, ⁵³ which examined a sample of LOAD (n = 642), to compare whether the observed behaviors were consistent or divergent from YOAD, using an independent t-test and a mixed variance analysis across groups and measurement points.

Results

Description of demographic characteristics and cognitive decline-comparison between YOAD and LOAD

The cohort of YOAD comprised 72 patients, including 39 females (55%). The average age was 59.77 years (SD = 4.08, range = 46–64 years) (see Table 1). Over half of the participants (54%, n = 32) were treated with donepezil, a third (32%, n = 23) received Exelon, and a smaller proportion (12%, n = 9) were later treated with memantine.

Table 1.

Demographic and cognitive variables.

		YOAD Sample n = 72	LOAD Sample n = 642	p (Between YOAD and LOAD)
Variables		Mean (range)	Mean (range)
Age at diagnosis		59.77 (46–64)	77.01 (65–97)	0.001
Gender (men)		33 (45.8%)	281 (43.8%)	N.S.
Country of origin	Israel	32 (47.1%)	217 (34.6%)	N.S.
	North Africa	11 (16.2%)	74 (11.8%)
	Europe	5 (7.4%)	165 (26.3%)
	Former USSR	5 (7.4%)	49 (7.8%)
	Iran	5 (7.4%)	33 (5%)
Marital status	Married	69 (95.8%)	419 (65.8%)	0.001
Family history of memory diseases		44 (61.1%)	398 (62%)	N.S.
Education		12.4 (0–20)	11.82 (0–28)	N.S.
Family size (Number of children)		3.96 (1–13)	3.6 (0–16)	N.S.
MMSE: first follow-up		21.04 (4–29)	21.51 (3–30)	N.S.
MMSE: after half year		20.16 (5–28)	21.06 (0–30)	N.S.
MMSE: after one year		20.10 (5–27)	20.35 (0–30)	N.S.
MMSE: after two years		16.69 (0–27)	18.46 (0–30)	N.S.
MMSE: after three years		14.25 (0–27)	17.17 (0–29)	N.S.

Table 2.

Cr variables and demographic variable predicting cognitive decline.

Timepoint	Education (β)	Family Size (β)	Gender (β)	Ethnicity (β)	MSE	Lambda (λ)
Half-year	0.09*	0.79	−1.46**	−2.83**	70.97	0.33
One-year	0.22*	n.s.	−0.85	n.s.	41.23	0.78
Two-year	0.00	n.s.	n.s.	n.s.	42.02	1.76
Three-year	n.s.	−0.30	−4.17*	3.44**	81.42	0.51

*p < 0.05, **p < 0.01, n.s.: not significant.

LASSO regression analyses to predict age of diagnosis and cognitive decline

To examine predictors of age at diagnosis and cognitive decline over time, we conducted LASSO regression analyses using CR measures (years of education, family size) and demographic factors (gender, country of origin, marital status, and ethnicity). The outcomes included age at diagnosis and changes in MMSE scores at six months, one year, two years, and three years post-diagnosis (see Table 2).

Age at diagnosis

The LASSO model for age at diagnosis retained gender as the sole predictor following regularization (λ = 0.6998), with a small negative coefficient (β = −0.14). No other CR measures or demographic variables were selected, indicating limited predictive value for these factors in determining age at diagnosis within this sample. The model yielded a mean squared error (MSE) of 13.89.

To assess the potential clinical significance of the gender effect, we conducted a Two One-Sided Test (TOST) for equivalence using bounds of ±2 years. The results were statistically significant: the lower bound test yielded a t-value of 2.291 (p = 0.013), and the upper bound test yielded a t-value of −6.113 (p < 0.001). These results suggest that the observed gender difference in age at diagnosis is statistically significant but falls within the predefined equivalence bounds, indicating it is not large enough to be considered clinically meaningful.

Cognitive decline

LASSO regression analyses were conducted to identify predictors of cognitive decline, measured by changes in MMSE scores over time, with a primary focus on the role of education as a CR proxy. Results revealed that education was a key predictor in the early stages following diagnosis, although its predictive value diminished over time:

Half-year MMSE difference

Education (β = 0.087), family size (β = 0.79), gender (β = −1.46), and ethnicity (β = −2.83) were retained (λ = 0.3261). Notably, the positive coefficient for education suggests that individuals with higher educational attainment experienced greater cognitive decline in the short term. The model's MSE was 70.97.

One-year MMSE difference

Education (β = 0.22) and gender (β = −0.85) remained as predictors (λ = 0.7783). Education continued to show a positive association with decline, further supporting its role in shaping early post-diagnosis trajectories. Model MSE was 41.23.

Two-year MMSE difference

Education was the only variable retained, though its coefficient was shrunk to zero (λ = 1.76), indicating no significant predictive effect at this time point. The MSE was 42.02.

Three-year MMSE difference

Education was not retained. Instead, family size (β = −0.30), gender (β = −4.17), and ethnicity (β = 3.44) were selected (λ = 0.5153). At this later stage, larger family size and female gender were associated with less decline, while certain ethnic groups showed greater decline. Model MSE was 81.42.

Comparison between YOAD and LOAD

To investigate the differences between YOAD and LOAD, we conducted an independent samples t-test to compare demographic and clinical characteristics between the groups, using a previous sample. ⁵³ The analysis revealed significant differences only in marital status (see Table 1). To examine cognitive deterioration, we performed a mixed variance analysis across groups and measurement points. However, the interaction between time and group was not significant (p = 0.45), indicating that the cognitive decline did not differ between the YOAD and LOAD groups.

Discussion

This study explored the role of CR in individuals with YOAD, a group often underrepresented in CR research. Specifically, we examined the impact of education as the primary proxy for CR, family size as a potential proxy, and other demographic factors such as gender and ethnicity on both age of diagnosis and cognitive decline over time. While CR variables did not predict age at diagnosis, higher educational attainment was associated with faster cognitive deterioration one-year post-diagnosis. Larger family size was linked to slower decline at six months and three years, while women showed greater cognitive deterioration at six months, one year, and three years.

By contrast with LOAD findings, where CR has been associated with delaying the age of diagnosis,^28,53 our regression analysis did not reveal significant associations between age of diagnosis and CR variables in YOAD. Although clinical and demographic parameters between the LOAD and YOAD groups did not differ significantly, the correlation between age and education varied between groups, suggesting a non-linear association across the age range. This indicates that the two populations are qualitatively different, supporting the view that YOAD may represent a distinct form of disease.^3,4

This discrepancy may be influenced by factors specific to YOAD, such as the criteria of age and genetic characteristics (e.g., APOE ε4). ⁵⁴ Additionally, a lesser likelihood of suspecting AD in younger individuals may impact diagnosing timing. YOAD patients often seek medical evaluation in settings not typically focused on geriatric conditions, potentially delaying diagnosis until severe brain pathology has developed.

We conducted a LASSO regression analysis to determine whether CR variables predicted cognitive decline over time. Our primary hypothesis, which posited that higher CR, as indicated by greater education and smaller family size, would be associated with faster cognitive deterioration, was largely supported by the results. Education emerged as the most robust and consistent predictor of cognitive decline, particularly in the first-year post-diagnosis, with individuals having more years of education showing greater cognitive deterioration. This finding aligns with previous research, reinforcing education as a reliable proxy for CR.

The effect of education on cognitive decline diminished over time, likely due to the rapid progression of cognitive impairment. Initially, MMSE scores were 21.04 at the first follow-up and 20.16 after six months. However, by the second year, scores had decreased to 16.69 and further declined to 14.25 by the third year. As cognitive impairment progressed, the influence of education likely diminished, possibly due to a floor effect, where further deterioration became less influenced by baseline factors such as education.

Larger family size was associated with slower cognitive decline at six months and three years. We tested family size as a CR proxy, despite its limited use in CR theory, due to its potential reflection of SES in our population^36,37 and the broader approach to include additional variables beyond education to capture CR's multidimensional nature.^22,46,47 However, the inconsistent results highlight the limitations of this proxy.

Although concurring with prior research,^53,55,56 it is important to note that findings of accelerated decline among higher education have not been widely replicated across studies, thus underscoring its significance. Lövdén et al. ⁵⁷ argued that studies examining the link between CR and cognitive decline in dementia are limited and yield mixed results. They argued that while there appears to be a trend toward faster decline in more highly educated individuals and no evidence for the opposite effect, some studies suggest null effects, indicating that education may not have a protective influence.

However, our findings that women with YOAD experience a more rapid cognitive decline than men, are well-supported by existing research. Studies have shown a higher prevalence of LOAD among women,^58,59 with women exhibiting accelerated cognitive deterioration post-diagnosis. ⁶⁰ This disparity is linked to risk factors that disproportionately affect women, such as the APOE ε4 and earlier menopause, both of which negatively impact CR and contribute to faster disease progression. ⁶⁰ Specifically, regarding YOAD, a systematic review ⁶¹ exploring prevalence differences between genders did not yield conclusive results, due to limited prevalence data and wide confidence intervals. In addition to these biological explanations, social and cultural factors may also contribute to the observed gender differences. Women tend to assume greater caregiving roles, which can demand additional emotional and physical stress, potentially impacting their cognitive resilience. Furthermore, factors such as the interaction of gender with race/ethnicity and societal influences may shape cognitive decline, as they interact with biological and educational determinants of reserve. ⁶² These sociocultural aspects highlight the importance of considering both gendered experiences and biological markers when examining the cognitive trajectories of individuals with YOAD. Notably, there is a scarcity of research specifically examining CR and gender differences in AD. Our results are consistent with current evidence, underscoring the need for more targeted investigations in this critical area.

It can be observed that, contrary to what is generally accepted in the literature and in our previous study on LOAD,^28,53 where higher CR is associated with a later age of diagnosis, age in YOAD is not predicted by CR variables. However, similar to LOAD, higher CR in YOAD is associated with accelerated deterioration. This pattern may indicate a more acute disease progression in YOAD, given that in the first year, both the MMSE averages are relatively the same, as well as the percentages of patients who deteriorated, but later on the MMSE scores are worse in YOAD, as well as the percentages of those who deteriorate. This finding highlights that CR may only influence early-stage disease progression in YOAD, with possible substantial brain pathology already present, even when initial differences are apparent. The greater cognitive decline observed in patients with higher CR may reflect a “compressed” disease trajectory, consistent with the CR hypothesis. ⁶³ However, this pattern could also be influenced by greater sensitivity to change among caregivers or clinicians, leading to more accurate recognition and reporting of decline in this group.

It is important to note that the patients in this sample were receiving medication. Consequently, a majority of participants either showed initial stability or improvement, whereas this proportion decreased steadily over subsequent testing points up to three years, indicating a progressive trend toward cognitive decline. Although determining this pattern decisively required a comparison between treated patients and an untreated group, it may seem that the initial benefit observed within the first six months may be attributed to the effect of medical treatment. This finding adds to existing knowledge by addressing previous challenges in delineating the impact of treatment duration. ¹⁹ Moreover, it provides insights into the short-term impact of medication, contributing to ongoing discussions about the efficacy of treatment in modifying the course of YOAD.^20,21

One notable strength of this study was its focus on a less frequently studied cohort: individuals diagnosed with YOAD. Moreover, the sample was drawn from a comprehensive clinical database of patients monitored and treated over several years by the same neurologists, ensuring consistent treatment approaches. This allowed for a relatively homogeneous treatment methodology. Importantly, the database tracked the progression and treatment of YOAD in real-life settings, rather than within the controlled environment of clinical trials.

The present study, conducted within a memory clinic, provides valuable insights due to its ecological validity. However, several limitations should be noted. Firstly, the relatively small sample size may limit the generalizability of the findings, although this is consistent with the lower prevalence of YOAD compared to LOAD in older age groups. Hence the difference between the sample sizes between the current sample and the previous sample of LOAD, ⁵³ even though they were collected in the same years. Future research with larger sample sizes would be valuable in further validating these findings. Additionally, the study would have been enriched by the inclusion of other variables reflecting cognitive decline beyond the MMSE, and additional CR measures like leisure activities or more direct SES indicators, but such data were not collected. Furthermore, the analysis of brain pathology and genetic information was beyond the scope of this study, which limits our understanding of disease progression. It is important to acknowledge that CR represents a complex theoretical construct, and its measurement is heavily influenced by the specific proxy variables employed, a challenge widely recognized in the literature.^40,41 Lastly, it would have been beneficial to compare our findings with a group of individuals who did not receive drug treatment. However, ethical considerations pose a challenge in withholding medical treatment from individuals diagnosed with a degenerative disease like AD.

In conclusion, this study revealed no significant correlation between age of diagnosis and CR in YOAD, by contrast with findings in LOAD, suggesting that mechanisms other than CR may play a more critical role in younger individuals. However, consistent with LOAD, higher education were associated with faster disease progression, indicating that diagnoses often occur when brain pathology is already severe. These findings underscore the need for further research into YOAD, particularly in developing early detection tools to better support young patients and their families.