Associations between blood-based neurodegenerative biomarkers and cognitive functioning and decline in India

Abstract

Background

Mounting evidence supports the use of blood-based neurodegenerative biomarkers as a low-cost, minimally invasive tool for studying Alzheimer's disease and other dementias, but existing data largely come from clinical samples or high-income settings. Despite emphasis in the literature on the importance of understanding the utility of neurodegenerative biomarkers in diverse populations, published analyses are limited.

Objective

To assess the utility of neurodegenerative biomarkers in India by quantifying associations between biomarkers and cognitive outcomes in a nationally representative cohort study.

Methods

We quantified associations between five neurodegenerative blood biomarkers (amyloid-β 42/40 (Aβ_42/40), total tau, phosphorylated Tau181 (pTau-181), glial fibrillary acidic protein (GFAP), neurofilament light (NfL)) and cross-sectional and longitudinal cognitive outcomes using nationally-representative data from the Longitudinal Aging Study in India–Diagnostic Assessment of Dementia (N = 4096).

Results

We observed associations between biomarkers and cross-sectional cognitive functioning (Aβ_42/40, GFAP, and NfL) and longitudinal cognitive change (pTau-181 and NfL). NfL had the strongest associations; each SD increase in log NfL was associated with 0.007 (95% CI 0.000 to 0.014) SD unit/year worse cognitive decline, equivalent to about 35% of the mean longitudinal decline in the sample. We saw little evidence of effect modification by demographic variables or APOE ε4 status.

Conclusions

Neurodegenerative biomarkers were associated with cross-sectional and longitudinal outcomes as well as mortality, though there was variation in outcome-specific findings across biomarkers. Findings generally support the use of neurodegenerative biomarkers in India. Future research in India should leverage these biomarkers to address a range of research topics, including heterogeneity in dementia phenotypes.

Keywords

Alzheimer's disease amyloid blood-based biomarkers cognition cognitive decline dementia glial fibrillary acidic protein neurodegeneration neurofilament light tau

Introduction

Though the expected tripling of Alzheimer's disease and other dementia (dementia) cases by 2050¹ will lead to large social, societal, and economic burden,^2,3 the underlying complexity and heterogeneity of the disease has handicapped progress towards identifying breakthrough treatments or highly effective prevention strategies. Limited clinical success of newly emerging biologic treatments for Alzheimer's disease^4,5 may be due in part to the fact that most individuals with dementia have more than one pathology simultaneously.^6,7 However, research pushing forward our understanding of dementia etiology has been hampered by challenges in measuring these pathologies in large-scale population-based research. The emergence of new, blood-based neurodegenerative biomarkers may enable researchers to ask and answer questions about more specific etiologic processes and begin to untangle questions around disease heterogeneity.

Though validation studies have shown close correlations between blood-based neurodegenerative biomarkers with more standard measures from PET imaging and cerebrospinal fluid (CSF)^8–11 and there is mounting evidence showing associations between these biomarkers and cognitive outcomes,^12–14 some have suggested that caution is warranted before rolling out the use of these biomarkers into all existing population-based dementia research.¹⁵ While the lower cost and lack of need for invasive procedures may help increase both racial and global diversity in research on dementia and Alzheimer's disease,¹⁵ blood-based biomarkers must have validity and reliability across groups to ensure that research in these populations or comparisons across settings are appropriate. Once the comparability of measures across populations is established, research on the underlying causes of dementia across diverse contexts and populations can yield important insights due to differences in the distribution of risk factors or differences in the economic, cultural, and social contexts, which can have large impacts on cognitive aging.^16–18

India is the most populous country in the world, and the population is rapidly aging.¹⁹ The rapid ongoing epidemiologic transition in India and known differences in the distributions of various risk factors or potential effect modifiers (e.g., air pollution)²⁰ lead to differences in the social and cultural context compared to most high-income contexts, where the majority of dementia research has been conducted.²¹ The use of blood-based neurodegenerative markers in this context therefore has the potential to help researchers uncover new insights regarding interactions between the environment and the biological and etiologic underpinnings of dementia. However, because existing research highlighting associations between blood-based neurodegenerative biomarkers and cognitive outcomes has been conducted predominantly in clinical settings^12,13,22,23 or in population-based settings in high-income contexts,^24–27 it is first necessary to evaluate these measures in the Indian setting.

The current study aims to describe the performance of blood-based biomarkers in India by assessing their association with cross-sectional and longitudinal cognitive outcomes in the Longitudinal Aging Study in India – Diagnostic Assessment of Dementia (LASI-DAD). By further investigating potential nonlinearities in associations, effect modification, and the association between biomarkers and survival in joint models for longitudinal change and survival, the current study seeks to provide a detailed understanding of how blood-based neurodegenerative markers are associated with cross-sectional and longitudinal cognitive phenotypes in India.

Methods

Data source

We used data from Waves 1 (2017–2019) & 2 (2022–2024) of the LASI-DAD study, the first nationally-representative, longitudinal study of dementia in India. Respondents 60 years of age and older were sampled from the broader Longitudinal Aging Study in India (LASI) study and represent 97.9% of the Indian population.²⁸ Of the 4096 respondents from Wave 1, 2566 completed Wave 2 interviews. The response rate among those re-contacted was 84% but the overall cumulative 5-year mortality was high (25%).²⁹

We excluded those with missing data across analyses. There were 2699 respondents from Wave 1 with data on neurodegenerative biomarkers, as the collection of blood samples began after initiation of Wave 1 data collection. The response rate among those eligible for data collection was relatively high (69%)³⁰; we would not expect bias due to the exclusion of those in the initial phases of Wave 1 data collection as these individuals would not be expected to be different from those included. We additionally excluded those with missingness on key covariates (N = 460). To maximize the available sample size, cross-sectional analyses of individual biomarkers using data from Wave 1 and longitudinal analyses using data from Waves 1 and 2 excluded those with missing data on each biomarker in turn (available sample sizes in Supplemental Material 1). In longitudinal analyses, we also excluded data from people lost to follow-up between waves (N = 128).

Outcome: Cognitive functioning

In both waves of LASI-DAD, participants completed a 1-h battery of cognitive tests adapted from the Harmonized Cognitive Assessment Protocol (HCAP),³¹ covering domains including orientation, memory, language, executive functioning, and visuospatial functioning (full list in Supplemental Material 2). Adaptations ensured adequate performance of the battery, with particular attention to issues related to language, culture, and literacy. Some minor adjustments were made to the LASI-DAD battery between waves, including shifts in scoring rules and a transition to tablet-based rather than paper-based administration for some tests (e.g., symbol cancellation).²⁸ Previous work has longitudinally co-calibrated estimates of latent cognitive functioning and specific cognitive domains using confirmatory factor analysis methods, accounting for known changes between the test batteries in Waves 1 and 2 while leveraging information from all available cognitive assessments across both time points.³² Cognitive functioning summary factors are normally distributed and scaled to have a mean of 0 and standard deviation of 1 in the full Wave 1 LASI-DAD sample. Estimates of cognitive functioning in LASI-DAD were previously shown to be invariant by language of administration (English and 12 local languages).³³ In the current study, we used continuous estimates of general cognitive functioning and four specific domains: memory, language/fluency, executive functioning, and visuospatial functioning.

Exposure: Neurodegenerative biomarkers

In Wave 1, trained phlebotomists collected 17 mL of fasting venous blood from participants, which was then shipped to a local lab for initial processing. Though participants were asked to fast, non-fasting samples were also collected. If the local lab was greater than two hours from the collection site, an automated centrifuge machine was brought to process blood samples on-site. After processing, separated plasma samples for neurodegenerative biomarker assays were then shipped via cold chain at −20 °C to the Department of Biophysics, All India Institute of Medical Sciences (AIIMS), New Delhi, where samples were stored at −80°C prior to completion of assays for amyloid-β 42 (Aβ₄₂), amyloid-β 40 (Aβ₄₀), total tau, phosphorylated tau 181 (pTau-181), glial fibrillary acidic protein (GFAP), and neurofilament light (NfL). All assays were conducted using an ultrasensitive and automated Single Molecule Array (Simoa) analyzer (Quanterix HD-X).

All biomarkers were log-transformed and standardized to have a mean of 0 and standard deviation of 1 for analyses.

Covariates

Age (continuous) and sex/gender (gender) were self-reported. Models additionally adjusted for continuous measures of body-mass index (BMI) and estimated Glomerular Filtration Rate (eGFR). Body-mass index was based on measured height and weight. eGFR was calculated using the CKD-EPI creatinine-cystatin C equation.³⁴

Inter-wave mortality

We conducted end-of-life interviews with a family member or friend who provided information on the month and year of death. If a respondent died in the same month as their Wave 1 interview (N = 7), we assumed 1/24th of a year between interview date and death. For respondents with missing month of death (N = 16), we assumed that deaths occurred halfway through the year. When year of death was missing (N = 37), we assumed deaths occurred halfway between the Wave 1 interview and the Wave 2 recontact date.

Effect modifiers

We categorized age using three age groups: 60–69, 70–74, and 80+ years. Education was self-reported and divided into categories of no school, less than primary or primary school, and middle-secondary school and higher. APOE genotype was determined via direct genotyping using TaqMan assays (Applied Biosystems, Foster City, CA), and dichotomized into a binary variable indicating the presence of an ε4 allele. Analyses using APOE genotype data additionally excluded individuals with missing genetic data (N = 193) and those with ε2/ε4 genotype (N = 27).

Statistical analysis

We used descriptive statistics to characterize the included Wave 1 sample and compare these participants to those who died between waves and those who returned and had cognition measured at Wave 2. We used density plots to visualize differences in the distributions of log biomarkers by tertile of cognitive functioning, age, APOE ε4 carrier status, and gender. We used Kaplan-Meier plots to describe inter-wave mortality.

We used linear regression models to estimate associations between neurodegenerative biomarkers and cognitive functioning using three sets of models: (1) univariate, (2) adjusted for age and gender, and (3) adjusted for age, gender, BMI, and eGFR. We estimated different models for general cognitive functioning, and specific cognitive domain outcomes. We calculated E-values to understand the potential risk that unmeasured biological factors associated with the measurement of biomarkers and cognitive outcomes may explain observed associations.³⁵ To further probe potential non-linearities, we reran models using natural cubic splines with three degrees of freedom and plotted marginal predictions for models adjusting for age, gender, BMI, and eGFR. We evaluated effect modification by estimating a further set of additional models including interaction terms between biomarkers and four potential effect modifiers: gender, age group, educational attainment, and APOE ε4 carrier status. All effect modification models also adjusted for age, gender, BMI, and eGFR. We used χ² tests for nested models to evaluate the overall statistical significance of interaction terms for categorical variables.

Given the high levels of mortality observed between waves of LASI-DAD, and evidence of strong associations between mortality and dementia-related phenotypes,²⁹ when estimating the effects of baseline risk factors or biomarkers on dementia it is helpful to consider the effects on mortality in tandem. To simultaneously model the joint effects of each neurodegenerative biomarker at Wave 1 on both mortality and longitudinal cognitive decline, we used joint models for survival and longitudinal change. In this joint modeling framework, the survival model (Cox proportional hazards model) and longitudinal model (linear mixed-effects model with random intercepts and slopes) are linked via a joint estimation process and the use of the current value of the longitudinal process as a linear predictor in the survival model. Models were estimated using the integrated nested Laplace approximation (INLA) for approximate Bayesian inference, which has been shown in simulation studies to be both reliable, computationally faster than alternatives, and capable of handling more complex models as compared to either maximum likelihood estimation or Bayesian inference with MCMC sampling.³⁶ Estimates from joint models include both the hazard ratios for the association between each biomarker and the hazard of death and estimates of the association between each biomarker and change in cognitive functioning between waves. Three sets of joint models were run, analogous to those used in cross-sectional analyses. In models adjusting for age, we also adjusted for the interaction between age and time. We excluded those who were lost to follow-up between waves (N = 128) but retained survival information on those who were recontacted with confirmed vital status, regardless of their participation in Wave 2 (full details on sample sizes in Supplemental Material 1).

We conducted a global test of the correlation between Schoenfeld residuals and time to evaluate proportionality of hazards for the survival outcome. The proportional hazards assumption was met for all biomarkers except Aβ_42/40. To further investigate time-varying hazards for Aβ_42/40 we ran follow-up survival models allowing the hazard ratio to vary over three time periods (0–2/2–3/3+ years), selected based on visual inspection of scaled Schoenfeld residuals over time.

All data processing and analysis was conducted in R version 4.4.1; R packages used for modeling include “survey” for weighted descriptive statistics and regression modeling, “survival” for survival analysis, and “INLAjoint” for joint modeling.^36–38 All descriptive and cross-sectional analyses use survey weights to account for unequal sampling probabilities.

Results

Cohort characteristics and descriptive results

The Wave 1 sample of 2239 respondents had an average age of 69.4 years. About half (51.4%) were women and slightly over half (54.3%) had no education (Table 1). About 20% (19.6%) of the sample were APOE ε4 carriers. 475 respondents died between waves (Supplemental Material 3), with 1450 returning for a Wave 2 cognitive interview; the average follow-up time between visits was 4.6 years. Compared to those who returned, those who died were older (Died = 73.1 years; Returned = 68.1 years), less likely to be women (Died = 42.6%; Returned = 53.1%), had lower Wave 1 cognitive functioning (Died = -0.38; Returned = 0.01), and were more likely to be APOE ε4 carriers (Died = 21.6%; Returned = 18.3%). Cognitive functioning was strongly associated with age, gender, and educational attainment, whereas differences by APOE ε4 carrier status were smaller (Supplemental Material 4).

Table 1.

Characteristics of the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD) sample with neurodegenerative biomarkers available in wave 1, those with data from wave 1 who died between waves, and those with data from Wave 1 and cognitive data available at both Waves 1 and 2.

	Wave 1	Died between waves	Waves 1 & 2
N	2239	475	1450
Age (y) (Wave 1), mean (SD)	69.4 (7.2)	73.1 (8.5)	68.1 (6.3)
Age (y) (Wave 2), mean (SD)	–	–	72.7 (6.3)
Women (%)	51.4 (1198)	42.6 (210)	53.1 (802)
Education (%)
No school	54.3 (1111)	57.6 (250)	53.2 (706)
Less than primary or primary school	25.6 (609)	26.5 (134)	25.9 (401)
Middle-secondary school and up	20.1 (519)	15.9 (91)	20.9 (343)
BMI (kg/m2), mean (SD)	22.1 (4.8)	21.0 (4.7)	22.4 (4.8)
eGFR (mL/min), mean (SD)	72.5 (19.9)	63.4 (20.7)	75.6 (18.7)
GCP (Wave 1), mean (SD)	−0.09 (0.90)	−0.38 (0.89)	0.01 (0.85)
GCP (Wave 2), mean (SD)	–	–	−0.06 (0.83)
APOE ε4 carrier (%)	19.6 (391)	21.6 (91)	18.3 (239)

All estimated means, standard deviations, and proportions account for unequal sampling probabilities using survey weights.

Visual assessment of the distribution of log-biomarker values by tertile of cognitive functioning, age group (60–69/70–79/80+), APOE ε4 carrier status, and gender do not suggest large differences, with a few notable exceptions (Figure 1). Both GFAP and NfL show strong patterning by tertile of cognitive functioning and age group, with higher log-biomarker values for those with lower cognition and older age. Additionally, we observe differences in the distribution of Aβ_42/40 ratio by APOE ε4 carrier status; carriers have lower ratio values compared to non-carriers, consistent with the biological effect of APOE ε4 carrier status in increasing amyloid plaques in the brain.

Figure 1.

Weighted density plots illustrating the differences in the distributions of log-transformed biomarker values by tertile of cognitive functioning (a), age group (b), APOE carrier status (c), and gender (d) in the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD) study.

Associations between neurodegenerative blood biomarkers and cognitive level

Crude associations between neurodegenerative biomarkers and cognition were attenuated after adjusting for age and gender, and age, gender, BMI, and eGFR (Figure 2). In fully adjusted models, we observed relatively consistent associations of GFAP and NfL with cognition across cognitive domains. For example, for each standard deviation higher observed level of GFAP, respondents had 0.04 (95% Confidence Interval [CI] 0.01 to 0.08) standard deviations (SD) lower general cognitive functioning and for each standard deviation higher observed level of NfL, respondents had 0.07 (95% CI 0.03 to 0.11) SD lower general cognitive functioning. These effect sizes are analogous to the effect of approximately 1 (GFAP) or 2 (NfL) years of older age on cognitive functioning in the full LASI-DAD sample. In contrast, results were null for pTau-181 (for general cognitive functioning: 0.00 [95% CI −0.03 to 0.04]) and total tau (for general cognitive functioning: 0.01 [95% CI −0.03 to 0.04]) and were more mixed for Aβ_42/40. While fully adjusted associations between Aβ_42/40 were null for memory (0.01 [95% CI −0.02 to 0.05]) and visuospatial functioning (0.00 [95% CI −0.04 to 0.03]), they were stronger for global cognition (0.04 [95% CI 0.00 to 0.07]; p = 0.04), language/fluency (0.06 [95% CI 0.02 to 0.09]; p < 0.01), and executive functioning (0.03 [95% CI 0.00 to 0.07]; p = 0.06). E-value estimates suggested that unobserved factors would have had to be associated with biomarker measurement and cognitive level with relative risks (or relative risk equivalents) of between 1.22–1.32 to fully explain cross-sectional associations with general cognitive functioning for NfL, GFAP, and Aβ_42/40.

Figure 2.

Cross-sectional associations between neurodegenerative biomarkers and continuous cognitive outcomes in the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD). Estimates come from three sets of weighted regression models (univariate, adjusted for age and gender, and adjusted for age, gender, BMI, and eGFR). Error bars show 95% confidence intervals.

Follow-up analyses using continuous splines did not indicate large deviations from linearity for most biomarkers and cognitive domains (Figure 3). However, for GFAP there was a notable threshold effect across cognitive domains, with no meaningful association between GFAP and cognition before a log-GFAP value of approximately 5, and a strong association at higher values.

Figure 3.

Spline fits to evaluate nonlinearities in the cross-sectional weighted associations between continuous cognitive outcomes and log biomarker values in the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD) study. Shaded regions represent 95% confidence intervals.

There was no strong statistical evidence of effect modification in associations by gender (Figure 4), age group, educational attainment, or APOE ε4 carrier status (Supplemental Material 3 and 4). However, summarizing broad patterns across cognitive domains, we saw suggestive evidence of stronger effects among men for GFAP and Aβ_42/40. For example, the association between Aβ_42/40 and general cognitive functioning was strong and statistically significant in men (difference per SD unit cognition: 0.06; 95% CI 0.00 to 0.11) but not women (0.01; 95% CI −0.03 to 0.06).

Figure 4.

Cross-sectional associations between neurodegenerative biomarkers and continuous cognitive outcomes stratified by gender in the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD) study. Estimates come from weighted regression models with an interaction term for gender, adjusted for age, gender, BMI, and eGFR. Error bars show 95% confidence intervals.

Associations between neurodegenerative blood biomarkers and cognitive decline

In joint models, we observed significant or marginally significant effects of Wave 1 levels of pTau-181 and NfL on cognitive decline across most cognitive domains (Figures 5 and 6). We estimated that each additional SD unit increase in log pTau-181 was associated with 0.008 (95% CI −0.002 to 0.014) SD unit/year greater rate of decline and each additional standard deviation unit increase in log NfL was associated with 0.007 (95% CI 0.000 to 0.014) SD unit/year greater rate of decline. These longitudinal effects are analogous to about 35–45% of the observed mean longitudinal decline in the sample, or the effect of 3.5–4.5 years of age at baseline. For other biomarkers, results were less consistent, though generally null or in the expected direction. The Aβ_42/40 results for language/fluency stood out as being strong, statistically significant, and in the opposite direction compared to what would be expected a priori (−0.013 [95% CI −0.022 to −0.005]). However, in follow-up sensitivity analyses using animal naming as a single indicator of semantic fluency instead of the language/fluency factor score, results were attenuated and no longer statistically significant (−0.008 [95% CI −0.022 to 0.006]) (Supplemental Material 5), indicating findings may be partially due to measurement challenges or practice effects in the language domain due to the inclusion of a number of very easy naming tasks. The largest mortality effects were observed for NfL; for each SD increase in log NfL we estimated that respondents had a 1.85 (95% CI 1.64 to 2.08) times greater hazard of death. Findings were also significant for GFAP (Hazard Ratio [HR]: 1.14; 95% CI 1.02 to 1.28), but null for pTau-181 and total tau. We observed a significant mortality effect for Aβ_42/40, but only after 3 years of follow-up (HR: 0.63; 95% CI 0.47–0.86). Effect sizes were equivalent to the effect of approximately 9 (NfL), 2 (GFAP), or 6 (Aβ_42/40) years of age.

Figure 5.

Associations between neurodegenerative biomarkers and longitudinal outcomes in the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD) study. Associations with cognitive decline (a) and survival (b) were derived from joint models incorporating both longitudinal outcomes across three models: univariate, adjusted for age and gender, and adjusted for age gender, BMI, and eGFR. Error bars show 95% credible intervals.

Figure 6.

Predicted trajectories of general cognitive functioning by biomarker quantile in the longitudinal aging study in India – diagnostic assessment of dementia (LASI-DAD) study. Marginal trajectories were derived from joint models incorporating both longitudinal outcomes, adjusted for age, gender, BMI, and eGFR.

Discussion

In a nationally representative sample in India, we observed associations between neurodegenerative blood biomarkers and both cross-sectional and longitudinal cognitive outcomes, though there were differences in the specific biomarkers related to cross-sectional cognitive levels versus longitudinal change in cognition. Specifically, we observed associations between Aβ_42/40, GFAP, and NfL with cognitive level, but associations between pTau-181 and NfL with cognitive change. Cross-sectional associations were strongest for GFAP, which was associated with inter-wave mortality as well. We also observed little evidence of effect modification by basic demographic variables or APOE ε4 status and associations were largely linear apart from GFAP, which was only associated with cognitive functioning at high biomarker levels.

Prior research in clinical samples has shown evidence of associations between blood-based biomarkers and cognitive outcomes.^23,39–41 Of the biomarkers considered here, evidence from the literature is strongest and most consistent for pTau-181,^39,41–43 whereas evidence for an association between total tau and cognitive outcomes is largely null. Findings have generally been more mixed for Aβ_42/40, NfL, and GFAP.^39,41,44 For Aβ_42/40 in particular, some of the strongest evidence has come from studies that are enriched for AD patients,^13,45 which may reflect the greater importance of these biomarkers in samples at higher risk for amyloid pathology, whereas some of the signal may be diminished in more heterogenous populations.

Findings from existing community-based studies tend to also support associations between blood-based biomarkers and cognitive outcomes, though there is some heterogeneity in the specific biomarkers and cognitive domains that were found to be statistically significantly associated with cognitive levels and cognitive decline. In the Rotterdam study, the Atherosclerosis Risk in Communities (ARIC) study, the Health ABC study, the Washington Heights-Hamilton Heights-Inwood Columbia Aging Project (WHICAP), and the Baltimore Longitudinal Study on Aging (BLSA), Aβ_42/40 was associated with either incident dementia or general cognitive decline.^{24–26,46,47} However, in the Study of Women's Health Across the Nation Michigan Cohort Aβ_42/40 was associated with cognitive decline in only 1 of 4 cognitive tests and in the Monongahela-Youghiogheny Health Aging Team (MYHAT) study Aβ_42/40 was associated with significant decline in memory and visuospatial functioning but not attention, executive functioning or language.^27,48 In contrast, in the current study, associations between Aβ_42/40 and cognitive decline were largely null, though we did observe an association between Aβ_42/40 and cognitive level.

For NfL, the Rotterdam study found significant associations with incident dementia or cognitive decline, but data from WHICAP and BLSA did not yield significant findings.^24,25,47 In MYHAT, NfL was associated only with memory decline, of the five domains assessed.²⁷ However, in the present study, NfL was positively associated with cognitive decline; the significant association in general cognition was driven largely by memory and executive functioning. Findings for GFAP and ptau-181 are also largely positive but with some variation across cohorts.^24,27,47 In the current study, we identified a significant association with cognitive decline for ptau-181 but not GFAP. Though not included in many studies, only total tau had consistent findings of null effects across community based cohorts.^25,48

Despite perceived differences between studies in findings for Aβ_42/40, pTau-180, NfL, and GFAP, most results are in expected directions; however, the generally small observed effect sizes in combination with other differences between cohorts may affect conclusions about statistical significance. Because plasma biomarkers may be differentially associated with certain cognitive domains more than other domains, differences in cognitive tests used may be a reason for inconsistent findings across studies.²⁷ Evidence of differences in the characteristics of biomarkers across the Alzheimer's disease or dementia continuum may also contribute to differences if samples differ in the distribution across the spectrum of severity or disease progression.^49,50 Additionally, given prior reporting of interactions between different neurodegenerative markers in their effects on cognitive outcomes, differences in the absolute levels of these biomarkers across settings could also lead to differences in observed effect sizes.⁴⁷ Though prior research has argued that differences in distributions of age, gender, and educational attainment across studies may impact comparisons, in the current study we did not identify significant effect modification by basic demographic factors or APOE ε4 carrier status.

Given the existing heterogeneity across community-based studies, findings from the LASI-DAD study largely fall in line with what has previously been reported. In spite of small observed effect sizes in the current study and across the prior literature (here, comparable to 1–2 years of age in cross-sectional analyses and 3.5–4 years of age in longitudinal analyses), the observed consistency of general findings across existing studies and in this setting indicate that these biomarkers perform similarly in India to what has been previously observed in high-income settings, where the majority of research using these biomarkers has been conducted to date. Relatively small effect sizes may be due to several reasons. First, the extensive literature on cognitive reserve and resilience provides compelling evidence describing how factors such as education may allow for increased ability to maintain cognitive capacity in the face of accumulating neuropathologies.⁵¹ These concepts around cognitive and brain reserve provide theoretical rationale for the lack of correspondence between cognitive functioning and neurodegeneration. However, lack of correspondence between fluid biomarkers and brain pathologies provides may also contribute; some data suggest that use of neurodegenerative blood biomarkers leads to somewhat weaker associations between biomarkers and cognitive outcomes as compared to use of neurodegenerative biomarkers in the CSF or from brain imaging.^52,53 Given that measurement of neurodegenerative biomarkers in plasma is a step further removed from measurement in the CSF or via brain imaging, the addition of measurement error likely leads to some attenuation of effects. Nonetheless, findings in the present study and in prior work suggest that despite the presence of some noise and measurement error, neurodegenerative blood biomarkers also contain important signal related to both brain pathology and cognitive functioning, promoting their use in future research. Future work could seek to combine information across biomarkers and derive novel biomarker measures to improve the signal-to-noise ratio.

The lack of consideration or incorporation of vascular pathologies in the current set of available neurodegenerative blood biomarkers could also help explain smaller observed effect sizes, given evidence on the importance of vascular pathologies in the development of dementia.⁵⁴ A growing body of evidence supports the notion that vascular pathologies interact with both Alzheimer's disease and other neurodegenerative pathologies in their effects on cognitive functioning^55–57 and suggests that future research should include vascular markers and their interactions with other biomarkers in considering associations between neurodegenerative blood biomarkers and cognitive outcomes. However, additional work in the development of novel plasma biomarkers of vascular brain pathologies is needed to facilitate the data collection required for such analyses.⁵⁸ The inclusion of co-morbid vascular pathologies is of even greater importance in the Indian context, where the prevalence of vascular disease is high and disease management is poor.⁵⁹

Findings from the current study showing associations between NfL, GFAP, and Aβ_42/40 (after 3 years of follow-up) and mortality provide further evidence that the neurodegenerative blood biomarkers capture general and consequential aspects of health status. Though observed associations with mortality may be mediated by the effects of these biomarkers on cognitive health, it is also possible that these neurodegenerative biomarkers capture aspects of general aging and health beyond their impacts on cognition. Future work quantifying associations between neurodegenerative biomarkers and other markers of general aging such as biological or epigenetic age after adjusting for cognitive functioning may be useful in elucidating the extent to which these markers capture important aspects of general heath independently of cognitive functioning. Regardless of the reason for the observed association, results highlight the importance of considering the potential impact of this differential mortality on the estimation of associations between neurodegenerative biomarkers and longitudinal changes in cognitive functioning in future population-based research.⁶⁰

Study strengths include the use of a nationally representative sample in India and a wide variety of models to probe non-linearities, effect modification, and simultaneously consider both longitudinal change in cognitive functioning and mortality. Limitations should also be considered. First, there are a number of health factors that have been shown to affect levels of neurodegenerative biomarkers,⁴⁴ which may lead to spurious associations if these health conditions are associated with cognitive outcomes. However, in current analyses we adjusted for both BMI and eGFR, which have been shown to be the most consequential factors in prior work.⁶¹ Further, E-value estimation suggested that unobserved factors would have had to have at least moderate associations (relative risk between 1.22–1.32) with biomarker measurement and cognitive level to fully explain cross-sectional associations for NfL, GFAP, and Aβ_42/40. Second, the length of available follow-up in the LASI-DAD study is currently still relatively short, with an average inter-wave interval of approximately 4.5 years; therefore, we are unable to capture associations between neurodegenerative biomarkers and long-term changes in cognitive functioning. Furthermore, the timing of the two waves of data collection sandwich the COVID-19 pandemic, which may have exacerbated the level of observed mortality. We considered and incorporated information on mortality by using joint models that simultaneously estimate the impacts of neurodegenerative biomarkers on mortality and longitudinal change while allowing for links between these two estimated associations. The extension of current analyses to future waves of LASI-DAD will further allow for analyses that could evaluate the extent to which COVID-19 impacted findings.

Findings highlight some important areas of investment for future research. Additional work to better understand the reasons underlying relatively small observed effect sizes, consider interactions between biomarkers, and incorporate information on the role of vascular pathologies in observed associations would be valuable in further elucidating the biological underpinnings of dementia in large-scale survey research, particularly in LMICs. Additionally, subsequent follow-up will allow for investigation of the longer-term links between biomarkers and cognitive outcomes.

Using nationally representative data from India, current results highlight cross-sectional associations for Aβ_42/40, GFAP, and NfL with cognitive level, as well as for pTau-181 and NfL with cognitive decline. There was little evidence of effect modification by demographic factors or APOE ε4 status. Findings are in line with existing evidence linking neurodegenerative blood biomarkers and cognitive outcomes from clinical samples or population-based research in high-income samples, supporting the use of these biomarkers in the Indian context.

Supplemental Material

sj-docx-1-alz-10.1177_13872877251361934 - Supplemental material for Associations between blood-based neurodegenerative biomarkers and cognitive functioning and decline in India

Supplemental material, sj-docx-1-alz-10.1177_13872877251361934 for Associations between blood-based neurodegenerative biomarkers and cognitive functioning and decline in India by Emma Nichols, Jinkook Lee, Alden L Gross, Masroor Anwar, Abhishek Gupta, Eileen M Crimmins, Bharat Thyagarajan and Sharmistha Dey in Journal of Alzheimer's Disease

Footnotes

Acknowledgments

We thank the field team, all LASI-DAD investigators, and LASI-DAD participants for their invaluable contributions to the project.

ORCID iDs

Emma Nichols

Sharmistha Dey

Ethical considerations

The LASI-DAD study obtained ethics approval from the Indian Council of Medical Research (2202-16741/F1) and all collaborating institutions, including the University of Southern California (UP-15-00684), All India Institute of Medical Sciences, New Delhi, Venu Geriatric Center, New Delhi, All India Institute of Medical Sciences, Bhubaneshwar, All India Institute of Medical Sciences, Rishikesh, Uttarakhand, Government Medical College, Chandigarh, Punjab, Aster MIMS Kannur, Kerala, Grants Medical College and JJ Hospital, Mumbai, Guwahati Medical College, Guwahati, All India Institute of Medical Sciences, Banaras Hindu University, Varanasi, Jawaharlal Institute of Postgraduate Medical Education and Research, Puducherry, Medical College, Kolkata, National Institute of Mental Health and Neurosciences, Bengaluru, All India Institute of Medical Sciences, Bibinagar, Hyderabad, Sher-e- Kashmir Institute of Medical Sciences, Srinagar, All India Institute of Medical Sciences, Mangalagiri, Andhra Pradesh, All India Institute of Medical Sciences, Bhopal Madhya Pradesh, and All India Institute of Medical Sciences, Raipur.

Consent to participate

All participants provided informed (written or thumbprint) consent for participation.

Author contributions

Emma Nichols: Conceptualization; Formal analysis; Methodology; Writing – original draft.

Jinkook Lee: Conceptualization; Funding acquisition; Supervision; Writing – review & editing.

Alden Gross: Conceptualization; Investigation; Writing – review & editing.

Masroor Anwar: Data curation; Investigation; Writing – review & editing.

Abhishek Gupta: Data curation; Investigation; Writing – review & editing.

Eileen Crimmins: Conceptualization; Project administration; Validation; Writing – review & editing.

Bharat Thyagarajan: Conceptualization; Project administration; Validation; Writing – review & editing.

Sharmistha Dey: Conceptualization; Funding acquisition; Project administration; Validation; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by National Institutes of Health/National Institute on Aging [grant numbers R01AG051125, U01AG064948 to J.L. and grant number RF1AG088003 to A.L.G. and E.C.].

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

LASI-DAD is a public resource for interested scientists around the world at no cost. LASI-DAD Wave 1 data is publicly available on the LASI-DAD website (g2aging. org/dad). LASI-DAD Wave 2 data will be publicly available in February 2025.

Supplemental material

Supplemental material for this article is available online.

References

Nichols

Steinmetz

Vollset

, et al. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the global burden of disease study 2019. Lancet Public Health 2022; 7: e105–e125.

Wimo

Ali

Guerchet

, et al. World Alzheimer Report 2015: The global impact of dementia: An analysis of prevalence, incidence, cost and trends, https://www.alzint.org/resource/world-alzheimer-report-2015/ (21 September 2015, accessed 10 April 2023).

Jutkowitz

Kane

Gaugler

, et al. Societal and family lifetime cost of dementia: implications for policy. J Am Geriatr Soc 2017; 65: 2169–2175.

Digma

Winer

Greicius

. Substantial doubt remains about the efficacy of anti-amyloid antibodies. J Alzheimers Dis 2024; 97: 567–572.

Walsh

Merrick

Milne

, et al. Considering challenges for the new Alzheimer’s drugs: clinical, population, and health system perspectives. Alzheimers Dement 2024; 20: 6639–6646.

Nichols

Merrick

Hay

, et al. The prevalence, correlation, and co-occurrence of neuropathology in old age: harmonisation of 12 measures across six community-based autopsy studies of dementia. Lancet Healthy Longev 2023; 4: e115–e125.

Robinson

Xie

Baer

, et al. Pathological combinations in neurodegenerative disease are heterogeneous and disease-associated. Brain J Neurol 2023; 146: 2557–2569.

Schindler

Bollinger

, et al. Validation of plasma amyloid-β 42/40 for detecting Alzheimer disease amyloid plaques. Neurology 2022; 98: e688–e699.

Mielke

Hagen

, et al. Plasma phospho-tau181 increases with Alzheimer’s disease clinical severity and is associated with tau- and amyloid-positron emission tomography. Alzheimers Dement 2018; 14: 989–997.

10.

Abdelhak

Foschi

Abu-Rumeileh

, et al. Blood GFAP as an emerging biomarker in brain and spinal cord disorders. Nat Rev Neurol 2022; 18: 158–172.

11.

Ashton

Janelidze

Al Khleifat

, et al. A multicentre validation study of the diagnostic value of plasma neurofilament light. Nat Commun 2021; 12: 3400.

12.

Tosun

Demir

Veitch

, et al. Contribution of Alzheimer’s biomarkers and risk factors to cognitive impairment and decline across the Alzheimer’s disease continuum. Alzheimers Dement 2022; 18: 1370–1382.

13.

Xiao

, et al. Plasma biomarker profiles and the correlation with cognitive function across the clinical spectrum of Alzheimer’s disease. Alzheimers Res Ther 2021; 13: 123.

14.

Meyer

Ashton

Karikari

, et al. Plasma p-tau231, p-tau181, PET biomarkers, and cognitive change in older adults. Ann Neurol 2022; 91: 548–560.

15.

Hayes-Larson

Ackley

Turney

, et al. Considerations for use of blood-based biomarkers in epidemiologic dementia research. Am J Epidemiol 2024; 193: 527–535.

16.

Adkins-Jackson

George

Besser

, et al. The structural and social determinants of Alzheimer’s disease related dementias. Alzheimers Dement 2023; 19: 3171–3185.

17.

Kobayashi

Jones

Briceño

, et al. Cross-national comparisons of later-life cognitive function using data from the harmonized cognitive assessment protocol (HCAP): considerations and recommended best practices. Alzheimers Dement 2024; 20: 2273–2281.

18.

Torres

Glymour

. Future directions for the HRS harmonized cognitive assessment protocol. Forum Health Econ Policy 2022; 25: 7–27.

19.

United Nations Department of Economic and Social Affiars, Population Division. World population prospects 2022: summary of results. New York, NY: United Nations, Department of Economic and Social Affairs, Population Division, 2022.

20.

Yadav

Arokiasamy

. Understanding epidemiological transition in India. Glob Health Action 2014; 7: 23248.

21.

Prince

. Dementia in developing countries. A consensus statement from the 10/66 dementia research group. Int J Geriatr Psychiatry 2000; 15: 14–20.

22.

Cousins

KAQ

Phillips

Das

, et al. Pathologic and cognitive correlates of plasma biomarkers in neurodegenerative disease. Alzheimers Dement 2024; 20: 3889–3905.

23.

Nallapu

Petersen

Lipton

, et al. Plasma biomarkers as predictors of progression to dementia in individuals with mild cognitive impairment. J Alzheimer’s Dis 2024; 98: 231–246.

24.

Honig

Kang

, et al. Risk of Alzheimer’s disease is associated with longitudinal changes in plasma biomarkers in the multi-ethnic Washington heights–Hamilton heights–inwood Columbia aging project (WHICAP) cohort. Alzheimers Dement 2024; 20: 1988–1999.

25.

de Wolf

Ghanbari

Licher

, et al. Plasma tau, neurofilament light chain and amyloid-β levels and risk of dementia; a population-based cohort study. Brain J Neurol 2020; 143: 1220–1232.

26.

Sullivan

Blackshear

Simino

, et al. Association of midlife plasma amyloid-β levels with cognitive impairment in late life. Neurology 2021; 97: e1123–e1131.

27.

Zhang

Ferreira

PCL

Jacobsen

, et al. Association of plasma biomarkers of Alzheimer’s disease and related disorders with cognition and cognitive decline: the MYHAT population-based study. Alzheimers Dement J Alzheimers Assoc 2024; 20: 4199–4211.

28.

Khobragade

Nichols

Meijer

, et al. Performance of the informant questionnaire on cognitive decline for the elderly (IQCODE) in a nationally representative study in India: the LASI-DAD study. Int Psychogeriatr 2022; 36: 177–187.

29.

Nichols

Dey

Khobragade

, et al. Socioeconomic and clinical predictors of all-cause mortality among older adults in India from 2017–2024 in a nationally representative, longitudinal study: The Importance of Cognitive Health and Dementia [Preprint]. 26 November 2024. DOI: https://doi.org/10.2139/ssrn.5032672.

30.

Lee

Petrosyan

Khobragade

, et al. Deep phenotyping and genomic data from a nationally representative study on dementia in India. Sci Data 2023; 10: 45.

31.

Langa

Ryan

McCammon

, et al. The health and retirement study harmonized cognitive assessment protocol project: study design and methods. Neuroepidemiology 2020; 54: 64–74.

32.

Gross

Nichols

Renteria

, et al. Risk and protective factors against cognitive decline in older adults from a nationally representative sample in India: Results from the LASI-DAD. medRxiv 2025. DOI: https://doi.org/10.1101/2025.07.07.25331027v1 [Preprint]. Posted July 10, 2025.

33.

Gross

Khobragade

Meijer

, et al. Measurement and structure of cognition in the longitudinal aging study in India-diagnostic assessment of dementia. J Am Geriatr Soc 2020; 68(Suppl 3): S11–S19.

34.

Inker

Eneanya

Coresh

, et al. New creatinine- and cystatin C–based equations to estimate GFR without race. N Engl J Med 2021; 385: 1737–1749.

35.

Haneuse

VanderWeele

Arterburn

. Using the E-value to assess the potential effect of unmeasured confounding in observational studies. JAMA 2019; 321: 602–603.

36.

Rustand

van Niekerk

Krainski

, et al. Fast and flexible inference for joint models of multivariate longitudinal and survival data using integrated nested Laplace approximations. Biostat Oxf Engl 2024; 25: 429–448.

37.

Lumley

. Analysis of complex survey samples. J Stat Softw 2004; 9: 1–19.

38.

Therneau

. A package for survival analysis in S. R Package Version 2015; 2: 2014.

39.

Simrén

Leuzy

Karikari

, et al. The diagnostic and prognostic capabilities of plasma biomarkers in Alzheimer’s disease. Alzheimers Dement 2021; 17: 1145–1156.

40.

Kivisäkk

Carlyle

Sweeney

, et al. Plasma biomarkers for diagnosis of Alzheimer’s disease and prediction of cognitive decline in individuals with mild cognitive impairment. Front Neurol 2023; 14: 1069411.

41.

Grasset

Bouteloup

Cacciamani

, et al. Associations between blood-based biomarkers and cognitive and functional trajectories among participants of the MEMENTO cohort. Neurology 2024; 102: e209307.

42.

Tsiknia

Edland

Sundermann

, et al. Sex differences in plasma p-tau181 associations with Alzheimer’s disease biomarkers, cognitive decline, and clinical progression. Mol Psychiatry 2022; 27: 4314–4322.

43.

Zabala-Findlay

Penny

Lofthouse

, et al. Utility of blood-based tau biomarkers for mild cognitive impairment and Alzheimer’s disease: systematic review and meta-analysis. Cells 2023; 12: 1184.

44.

Toledo

Vanderstichele

Figurski

, et al. Factors affecting aβ plasma levels and their utility as biomarkers in ADNI. Acta Neuropathol 2011; 122: 401–413.

45.

Chatterjee

Pedrini

Doecke

, et al. Plasma Aβ42/40 ratio, p-tau181, GFAP, and NfL across the Alzheimer’s disease continuum: a cross-sectional and longitudinal study in the AIBL cohort. Alzheimers Dement 2023; 19: 1117–1134.

46.

Yaffe

Weston

Graff-Radford

, et al. Association of plasma β-amyloid level and cognitive reserve with subsequent cognitive decline. JAMA 2011; 305: 261–266.

47.

Dark

Duggan

, et al. Alzheimer’s and neurodegenerative disease biomarkers in blood predict brain atrophy and cognitive decline. Alzheimers Res Ther 2024; 16: 94.

48.

Wang

Bakulski

Karvonen-Gutierrez

, et al. Blood-based biomarkers for Alzheimer’s disease and cognitive function from mid- to late life. Alzheimers Dement 2023; 20: 1807–1814.

49.

Guo

Zhang

Zhan

, et al. Sex differences in the association between obesity and cognitive impairment in a low-income elderly population in rural China: a population-based cross-sectional study. Front Neurol 2021; 12: 669174.

50.

Wojdała

Bellomo

Gaetani

, et al. Trajectories of CSF and plasma biomarkers across Alzheimer’s disease continuum: disease staging by NF-L, p-tau181, and GFAP. Neurobiol Dis 2023; 189: 106356.

51.

Stern

. Cognitive reserve in ageing and Alzheimer’s disease. Lancet Neurol 2012; 11: 1006–1012.

52.

Aschenbrenner

Henson

, et al. Comparison of plasma and CSF biomarkers in predicting cognitive decline. Ann Clin Transl Neurol 2022; 9: 1739–1751.

53.

Planche

Bouteloup

Pellegrin

, et al. Validity and performance of blood biomarkers for Alzheimer disease to predict dementia risk in a large clinic-based cohort. Neurology 2023; 100: e473–e484.

54.

Cortes-Canteli

Iadecola

. Alzheimer’s disease and vascular aging. J Am Coll Cardiol 2020; 75: 942–951.

55.

Rabin

Nichols

La Joie

, et al. Cerebral amyloid angiopathy interacts with neuritic amyloid plaques to promote tau and cognitive decline. Brain J Neurol 2022; 145: 2823–2833.

56.

Schneider

Wilson

Bienias

, et al. Cerebral infarctions and the likelihood of dementia from Alzheimer disease pathology. Neurology 2004; 62: 1148–1155.

57.

Schneider

Boyle

Arvanitakis

, et al. Subcortical infarcts, Alzheimer’s disease pathology, and memory function in older persons. Ann Neurol 2007; 62: 59–66.

58.

Hosoki

Hansra

Jayasena

, et al. Molecular biomarkers for vascular cognitive impairment and dementia. Nat Rev Neurol 2023; 19: 737–753.

59.

Prabhakaran

Jeemon

Roy

. Cardiovascular diseases in India. Circulation 2016; 133: 1605–1620.

60.

Davis-Plourde

Mayeda

Lodi

, et al. Joint models for estimating determinants of cognitive decline in the presence of survival bias. Epidemiol Camb Mass 2022; 33: 362–371.

61.

Pichet Binette

Janelidze

Cullen

, et al. Confounding factors of Alzheimer’s disease plasma biomarkers and their impact on clinical performance. Alzheimers Dement 2023; 19: 1403–1414.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.87 MB