Abstract
Background Attention-Deficit/Hyperactivity Disorder (ADHD) persists in older age and is postulated to be a risk factor for cognitive impairment and Alzheimer’s Disease (AD). However, this notion relies exclusively on epidemiological associations, and no previous study has linked ADHD with a decline in cognitive performance in older adults or with AD progression. Therefore, this study aimed to determine whether genetic liability for ADHD, as measured by a well-validated ADHD polygenic risk score (ADHD-PRS), is associated with longitudinal cognitive decline and the development of AD pathophysiology in cognitively unimpaired (CU) older adults.
Methods We calculated a weighted ADHD-PRS in 212 CU individuals without a clinical diagnosis of ADHD (55-90 years) using whole-genome information. These individuals had baseline amyloid-β (Aβ) positron emission tomography, as well as longitudinal cerebrospinal fluid (CSF) phosphorylated tau at threonine 181, structural magnetic resonance imaging, and cognitive assessments for up to 6 years. Linear mixed-effects models were used to test the association of ADHD-PRS with cognition and AD biomarkers.
Outcomes Higher ADHD-PRS was associated with greater cognitive decline over 6 years. The combined effect between high ADHD-PRS and brain Aβ deposition on cognitive deterioration was more significant than each individually. Additionally, higher ADHD-PRS was associated with increased CSF p-tau181 levels and frontoparietal atrophy in CU Aβ-positive individuals.
Interpretation Our results suggest that genetic liability for ADHD is associated with cognitive deterioration and the development of AD pathophysiology in the CU elderly. These findings indicate that ADHD-PRS might inform the risk of developing cognitive decline in this population.
Funding National Institute of Health and Brain & Behavioral Research Foundation.
Introduction
Attention-deficit/hyperactivity disorder (ADHD) is characterized by impairing and pervasive symptoms of inattention, hyperactivity-impulsivity, or both (1). Although first conceptualized as a neurodevelopmental disorder of childhood, current evidence supports ADHD as a waxing and waning condition with fluctuating symptoms and impairments throughout the lifespan (2), including in older adulthood (3). According to a recent meta-analysis, the prevalence of ADHD in older adults (>50 years) is approximately 2.18% (4). Furthermore, as the geriatric population grows, the absolute number of patients aged 50 years and older fulfilling the criteria for diagnosis of ADHD will likely increase (3). Therefore, understanding the disorder’s association with prevalent age-related diseases is a pressing concern.
The association between ADHD and age-related cognitive impairment is of particular interest. Throughout the lifespan, cognitive deficits across various neurocognitive domains have been extensively described in ADHD (1). Additionally, cognitive function in older adults with ADHD may closely resemble early manifestations of neurodegenerative conditions (5). Recent population-based large epidemiological studies suggested that ADHD is associated with a higher risk for mild cognitive impairment (MCI) and Alzheimer’s disease (AD) (6–8). However, these findings rely primarily on electronic health records and consequently can present biased estimates of ADHD and dementia prevalence. Since the differential diagnosis between undiagnosed ADHD and early dementia can be challenging (5), it is crucial to clarify whether ADHD is a risk factor for MCI and AD dementia or misdiagnosis due to symptom overlap (5). Additionally, it is unclear whether ADHD is associated with progressive cognitive decline in older age, above and beyond the cognitive deficits originating in childhood (1).
An obstacle to investigating age-related cognitive decline in ADHD is the absence of large-scale studies following patients with childhood-diagnosed ADHD into older age. An alternative approach is to consider the association of well-established dimensional biomarkers of ADHD and AD in samples not selected for ADHD. The ADHD polygenic risk score (ADHD-PRS) represents the combined genetic liability for the disorder and is highly associated with ADHD diagnosis and related traits in independent clinical and population samples (9). In this study, we explored the association of ADHD-PRS with cognitive impairment in older age. More specifically, we tested the following hypotheses: (1) ADHD-PRS is associated with progressive cognitive decline in cognitively unimpaired (CU) older people; (2) the association between ADHD-PRS and cognitive decline varies according to baseline Aβ burden; (3) ADHD-PRS is associated with brain tau pathology and degeneration.
Methods
Participants
We used data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a longitudinal multicenter study designed to develop clinical, imaging, genetic, and biochemical biomarkers for the early detection and tracking of AD (http://adni.loni.usc.edu; for more information, see previous reports (10)). ADNI’s inclusion criteria relevant for this study are age between 55 and 90, absence of major depression or bipolar disorder (DSM-IV criteria) within the past one year, no history of schizophrenia (DSM-IV criteria), and a Geriatric Depression Scale (GDS) score less than 6. All data were downloaded from the ADNI data repository in December 2021. Institutional Review Boards of all involved sites approved the ADNI study, and all research participants or their authorized representatives provided written informed consent.
We included CU participants with baseline medical data, baseline Aβ [18F]florbetapir positron emission tomography (PET), whole-genome information, and a minimum of two clinical assessments with neuropsychological testing. At baseline, all CU participants included in the analysis had a Clinical Dementia Rating (CDR) score of 0 and a Mini-Mental State Exam (MMSE) score of 24 to 30. After the baseline assessment, subjects returned at six months, one year, and annually after that (mean number of observations with complete data = 3.6, with a minimum of 2 and a maximum of 7 observations per individual). Analyses evaluating the longitudinal trajectory of (CSF) tau phosphorylated at threonine 181 (p-tau181) and gray matter (GM) volume were performed in a subsample based on data availability (Figure S1).
Polygenic risk score
Data were available across three genotyping platforms: (1) the Human610-Quad platform, (2) the HumanOmniExpress, and (3) Omni 2.5M platform. Imputation and merging of the different platforms were performed as previously described (11). ADHD-PRS was calculated using the additive model, which is the weighted sum of risk alleles for ADHD according to the most recent genome-wide association study (GWAS) (12). The calculation was performed using the PRSice software v2.2 (13). Independent single nucleotide polymorphisms (SNPs) were classified based on a 250-kb window and 0.1 r2 linkage disequilibrium criteria. After applying the quality control filters, 212,846 variants were retained for PRS analysis. Only genes located outside the MHC region (chr6: 26-33Mb) were included. Nine ADHD-PRSs were calculated using subsets of SNPs selected according to the following GWAS p-value thresholds: 1, 0.5, 0.4, 0.3, 0.2, 0.1, 0.05, 0.005, and 0.0005. Main analyses were performed with the threshold of 1, assuming all genetic markers contributed to ADHD diagnosis. ADHD-PRSs were transformed into z-scores for better visualization. To investigate populational structure, principal components analysis was conducted using PLINK 1.9 (14). We retained seven principal components to account for any ancestry differences in genetic structure that could bias the results, as previously done for ADNI datasets (11).
Cognitive function
Each participant from ADNI was submitted to a broad clinical and neuropsychological assessment in selected visits. Our primary outcome was cognitive function as measured by the Preclinical Alzheimer’s Cognitive Composite (PACC) adapted in the ADNI study (15). The PACC was developed to detect the first signs of cognitive decline in CU subjects with biomarker evidence of AD pathology. The PACC adapted for ADNI was obtained by summing the following four standardized z-scores: Alzheimer Disease Assessment Scale - Cognitive Subscale Delayed Word Recall, Logical Memory Delayed Recall, MMSE, and Trail-Making Test B Time to Completion (15). Higher scores in PACC indicate better cognitive performance. To explore specific aspects of cognitive function, we used the ADNI composite score for executive function (ADNI-EF) (16) and the ADNI composite score for memory function (ADNI-Mem) (17). The ADNI-EF includes the performance on WAIS-R Digit Symbol Substitution, Digit Span Backwards (Trails A and B), Category Fluency, and Clock Drawing (16). The ADNI-Mem includes the Rey Auditory Verbal Learning Test, AD Assessment Schedule – Cognition, MMSE, and Logical Memory (17). For both, higher scores indicate better performance.
MRI and PET
Magnetic resonance imaging (MRI) and [18F]florbetapir Aβ-PET were acquired following ADNI protocols and pre-processed as previously described (18). Briefly, MRIs were segmented into probabilistic GM maps using the SPM12 segmentation tool. Each GM probability map was then non-linearly registered (with modulation) to a stereotaxic space using DARTEL and smoothed with a Gaussian kernel of full-width half maximum of 8 mm to generate GM density voxel-based morphometry (VBM) images. We visually inspected all images to ensure proper alignment to the ADNI template. [18F]Florbetapir standardized uptake value ratio (SUVR) images used the whole cerebellum as the reference region and were generated from a weighted average of the mean uptake from the cortical GM of frontal, anterior and posterior cingulate, lateral parietal, and temporal regions (19). Individuals with [18F]florbetapir SUVR higher than 1.11 were classified as Aβ-positive, a widely validated cutoff for this population (19).
CSF p-tau181
CSF p-tau181 was measured using fully automated Elecsys immunoassays (Roche Diagnostics). Measurements outside the analytical range (< 8 pg/mL or > 120 pg/mL) were handled by setting them to the lower or upper detection limit. One individual presenting CSF p-tau181 concentrations three standard deviations above the mean was considered an outlier and excluded from the analyses.
Statistical analysis
Statistical analyses were performed using Stata version 14.0 (StataCorp, College Station, Texas USA) and voxel-wise statistics using MATLAB software version 9.2 (http://www.mathworks.com) with VoxelStats package (20). We used random field theory (RFT) to correct brain imaging results for multiple comparisons. The association between ADHD-PRS and baseline demographic characteristics was explored using linear or logistic regressions. A two-sided p-value lower than 0.05 was considered statistically significant.
Linear mixed-effects models were used to assess the main effects of ADHD-PRS on cognitive function (Model 1, Supplementary Material). To test the hypothesis that ADHD-PRS is associated with progressive cognitive decline, we examined the interaction between ADHD-PRS and time (Model 2, Supplementary Material). To test whether the association between ADHD-PRS and cognitive decline varies according to baseline Aβ burden, we added an interaction between ADHD-PRS, time, and Aβ-PET (Model 3, Supplementary Material). Cohen’s f2 allows the estimation of the effect size within the context of mixed-effects linear models and was obtained as previously described (21). Cohen’s f2 ≥ .02, f2 ≥ .15, and f2 ≥ .35 represent small, medium, and large effect sizes, respectively (21).
In exploratory analyses, we tested the association between ADHD-PRS and longitudinal changes in CSF p-tau181 or VBM. For that, we evaluated the interaction between ADHD-PRS and time, also with linear mixed-effects models (Models 4 and 5, respectively, Supplementary Material). Following a significant triple interaction between ADHD-PRS, time, and Aβ-PET, analyses with CSF p-tau181 and brain atrophy were conducted stratifying individuals according to their Aβ status. All models were adjusted for sex, age at baseline, apolipoprotein E ε4 (APOE ε4) carriership status (carriers vs. non-carriers), years of education, and ancestry (using the first seven principal components, as previously performed for ADNI datasets (11)). Furthermore, as recommended for proper adjustment in genetic studies, we adjusted for the interaction between the independent variables and each covariate (22). Time was defined as years from baseline for each participant. Mixed models were fit including subject-specific random slopes and intercepts to cluster the multiple assessments per individual.
We conducted the primary analyses including all SNPs (ADHD-PRSs with a threshold of 1). Sensitivity analyses with additional thresholds were performed and can be found in the Supplementary Material. Additionally, we performed sensitivity analyses to explore the role of potential confounders previously associated with ADHD and AD dementia, such as vascular risk factors (VRFs) (1,23) and depression (1,23). VRF burden was assessed using a composite score, and a score equal to or higher than two was defined as elevated (24) (Supplementary Material). Depression symptoms were assessed using baseline GDS scores.
Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript.
Results
A total of 212 participants had genetic data and Aβ-PET measures at baseline (Figure S1). From those, 196 and 193 participants had CSF p-tau181 and MRI data, respectively (Figure S1). The mean (SD) age of the sample was 73.1 (5.96) years, 116 (54.7%) were women, and all self-reported being White (210 as Not Hispanic/Latino). The mean and median observation time was 3.96 and 4.05 years, respectively (SD = 1.5, interquartile range = 3.01 – 5.31, maximum follow-up of 6 years). Table 1 shows participants’ characteristics at baseline. ADHD-PRSs were normally distributed (Figure S2) and were not associated with age, sex, or APOE ε4 carrier status (Table S1). Higher ADHD-PRS was associated with decreased years of education (Table S1). Additionally, ADHD-PRS was not associated with length of follow-up, suggesting no attrition bias (Table S1).
ADHD-PRS is associated with a persistent executive function deficit from baseline
We observed a significant main effect of ADHD-PRS on ADNI-EF (β = -.09, 95% CI = -.18 to -.008, p-value = .03), suggesting that higher ADHD-PRS was related to a persistent executive function deficit in all time points including baseline (Figure 1A). No significant effect was observed for ADHD-PRS in the model with PACC global cognitive composite as the outcome variable (β = -.12, 95% CI = -.44 to 19, p-value = .45; Figure 1B) or in the model containing ADNI-Mem as the outcome variable (β = .01, 95% CI = -.05 to .08, p-value = .63; Figure 1C).
ADHD-PRS associates with longitudinal cognitive decline
We observed a significant interaction between ADHD-PRS and time on PACC, demonstrating that higher ADHD-PRS was associated with a higher decline in general cognitive performance over 6 years (ADHD-PRS x time; β = -.10, 95% CI = -.16 to -.03 p-value = .003; Figure 2A) with a Cohen’s f2 of .21, indicating a medium effect size. As a comparison, the Cohen’s f2 for baseline [18F]florbetapir Aβ-PET, a well-established predictor of cognitive decline in AD, was .48, indicating a large effect size. Similar findings were observed for ADNI-Mem (ADHD-PRS x time; β = -.01, 95% CI = -.02 to -.002, p-value = .01; Figure 2B), indicating that higher ADHD-PRS was related to a progressive decline in memory function. No longitudinal effects were observed for ADNI-EF (ADHD-PRS x time; β = -.003, 95% CI = -.01 to .01, p-value = .62; Figure 2C).
ADHD-PRS and Aβ show an interaction in longitudinal cognitive decline
We observed a significant interaction between ADHD-PRS and baseline Aβ-PET on worsening cognitive performance, indicating that the association between higher ADHD-PRS and decreased PACC scores over time was present in Aβ-positive but not in Aβ-negative individuals (ADHD-PRS x time x baseline Aβ-PET; β = -.17, 95% CI = -.31 to -.02, p-value = .01; Figure 3A). Similar findings were obtained for ADNI-Mem (ADHD-PRS x time x baseline Aβ-PET; β = -.02, 95% CI = -.05 to -.0001, p-value = .04; Figure 3B), but not ADNI-EF (ADHD-PRS x time x baseline Aβ-PET; β = .001, 95% CI = -.03 to .03, p-value = .92; Figure 3C).
ADHD-PRS associates with forthcoming tau pathology and brain atrophy in Aβ-positive individuals
Higher ADHD-PRS was highly associated with increased CSF p-tau181 over time in Aβ-positive individuals (ADHD-PRS x time; β = .05, 95% CI = .01 to .08, p-value = .003, Figure 4B). On the other hand, no significant association was observed for Aβ-negative individuals (ADHD-PRS x time; β = -.003, 95% CI = -.02 to .01, p-value = .70, Figure 4A). Similarly, ADHD-PRS was associated with longitudinal neurodegeneration as measured by reduction of GM density in the superior frontal gyrus and supramarginal gyrus (Figure 5) in Aβ-positive individuals.
Sensitivity analyses
The aforementioned findings were replicated using most ADHD-PRS thresholds, supporting the robustness of our results (Table S2). In addition, similar findings were obtained by adjusting for possible confounders such as VRFs and depression symptoms (Table S3 and Table S4).
Discussion
This study aimed to determine whether ADHD-PRS was associated with longitudinal cognitive impairment in CU older adults and whether this association was related to the core markers of AD pathology. For the first time, we described that higher ADHD-PRS was associated with progressive longitudinal cognitive decline, particularly in memory function. Furthermore, cognitive decline was mostly observed in Aβ-positive individuals, suggesting that individuals carrying a genetic liability for ADHD are characterized by cognitive susceptibility to the presence of Aβ pathology. Finally, in Aβ-positive individuals, higher ADHD-PRS was associated with longitudinal increases in CSF p-tau181 and brain atrophy in frontal and parietal brain regions. These findings suggest that the genetic liability to ADHD increases the susceptibility to cognitive decline, tau pathology, and neurodegeneration in the presence of Aβ pathology in the brain of older individuals.
Our results corroborate previous studies showing deficits in executive function among older adults with ADHD (3). We observed that higher genetic liability for ADHD was associated with executive function deficits, which remained relatively constant over time. These results are unsurprising and consistent with previous literature showing an association between higher ADHD-PRS and decreased executive function during childhood (9). Importantly, our study provides unique evidence that higher ADHD-PRS was associated with progressive cognitive decline, predominantly in the memory domain. Since the prototypical clinical phenotype of AD is progressive amnestic symptoms (23), our results support previous epidemiological findings demonstrating that ADHD might be a risk factor for cognitive decline, potentially leading to MCI and dementia syndromes due to AD (6–8).
We found that future cognitive decline in older adults is associated with the presence of both high ADHD-PRS and brain Aβ pathology. Specifically, ADHD-PRS potentiated the effects of Aβ on longitudinal clinical and pathophysiological progressions in our older population. It is postulated that abnormal Aβ deposition triggers a cascade of events leading to AD progression (23). Although brain Aβ load is associated with AD-related cognitive decline (25), Aβ pathology alone seems not to be not sufficient to cause it (25). For example, it is well established that around 30% of CU individuals older than 55 years of age present brain Aβ pathology and that a large portion of these individuals remains cognitively intact during their lives (26). This supports that the association between Aβ accumulation and cognitive deterioration depends on patients’ intrinsic resilience and susceptibility mechanisms. Together, the results above indicate that the genetic liability to ADHD plays a role in increasing the susceptibility to the harmful effects of Aβ in the human brain.
A widely held view in the AD field posits that Aβ triggers the spread of tau pathology, leading to neurodegeneration and cognitive impairment (25). In line with this hypothesis, our study showed that the association between ADHD-PRS and cognitive decline was accompanied by a longitudinal increase in CSF levels of p-tau181, a well-validated marker of brain tau pathology (27). Our findings suggest that ADHD-PRS is related to both tau deposition and cognitive decline in Aβ-positive individuals, highlighting the genetic liability for ADHD as a relevant factor influencing AD progression in the presence of Aβ pathology.
We showed that higher ADHD-PRS was associated with brain atrophy in frontal and parietal brain regions in Aβ-positive individuals. Specifically, decreased GM density over time was observed in the superior frontal gyrus and supramarginal gyrus. Previous AD studies have shown that brain atrophy closely correlates with tau deposition and cognitive deficits, being a well-established marker of disease progression (28). In CU populations, AD-related brain atrophy has been reported predominantly in the medial temporal cortex. In contrast, reduced cortical thickness in regions such as the superior frontal gyrus and supramarginal gyrus is present in later stages of AD (28). Interestingly, atrophy in parietal and frontal cortices has been demonstrated in middle-aged (29) and older adults (30) with ADHD. These results suggest that while in our study ADHD-PRS-related atrophy in the superior and supramarginal gyrus was Aβ pathology-dependent, it recapitulated regions showing atrophy in ADHD rather than early AD. This supports the notion that the combination of underlying ADHD-related vulnerability with Aβ pathology is a factor associated with cognitive dysfunction in older adults.
This study should be viewed in light of some limitations. First, our sample did not have a detailed clinical assessment for ADHD diagnosis. Since ADHD is a relatively recent diagnostic category, older adults are unlikely to have received the diagnosis as children (1). Thus, ADHD-PRS in our results may identify either asymptomatic older adults with genetic susceptibility to ADHD and/or patients with undiagnosed ADHD. In addition, the fact that neuropsychiatric conditions such as depression, bipolar disorder, or schizophrenia were not included in the ADNI cohort further limits the external validity of our results. The sample included in this study likely represents a less affected population when compared to the general ADHD population treated in clinical centers, where psychiatric comorbidities are the rule rather than the exception (1). The population used to generate ADHD-PRS (12) and our study population were composed almost exclusively of white participants. Therefore, future studies in ADHD and AD should focus on enrolling more diverse populations. Finally, no adjustments were made to control type I error, and consequently, the analyses should be considered exploratory.
Conclusion
To conclude, our results suggest that ADHD-PRS can be used to inform the risk of cognitive decline in Aβ-positive CU older adults. Since ADHD-PRS was associated with cognitive decline in our entire population, ADHD-PRS may also be used to predict cognitive deterioration in the absence of AD biomarkers. Importantly, ADHD-PRS was associated with longitudinal CSF p-tau hyperphosphorylation and brain atrophy in frontoparietal but not temporal regions in Aβ-positive individuals, suggesting that an ADHD-related brain susceptibility to the harmful effects of Aβ plays a role in the early development of AD in genetically vulnerable patients.
Data Availability
All data produced in the present study are available upon reasonable request to the authors.
Disclosures
LAR has received grant or research support from, served as a consultant to, and served on the speakers’ bureau of Aché, Bial, Medice, Novartis/Sandoz, Pfizer/Upjohn, and Shire/Takeda in the last three years. The ADHD and Juvenile Bipolar Disorder Outpatient Programs chaired by LAR have received unrestricted educational and research support from the following pharmaceutical companies in the last three years: Novartis/Sandoz and Shire/Takeda. LAR has received authorship royalties from Oxford Press and ArtMed. AC has acted as a consultant for Knight Therapeutics in the last three years.
Acknowledgments
DTL is supported by a CNPq postdoctoral fellowship (#154116/2018-1), and supported by a NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation (#29486). JPF-S receives financial support from CAPES (#88887.627297/2021-00). CT receives funding from Faculty of Medicine McGill and IPN McGill. PCLF is supported by the Alzheimer’s Association (#AARFD-22-923814). WSB is supported by CAPES (#88887.372371/2019-00 and #88887.596742/2020-00). TKK is funded by the Swedish Research Council’s career establishment fellowship (#2021-03244), the Alzheimer’s Association Research Fellowship (#850325), the BrightFocus Foundation (#A2020812F), the International Society for Neurochemistry’s Career Development Grant, the Swedish Alzheimer Foundation (Alzheimerfonden; #AF-930627), the Swedish Brain Foundation (Hjärnfonden; #FO2020-0240), the Swedish Dementia Foundation (Demensförbundet), the Swedish Parkinson Foundation (Parkinsonfonden), Gamla Tjänarinnor Foundation, the Aina (Ann) Wallströms and Mary-Ann Sjöbloms Foundation, the Agneta Prytz-Folkes & Gösta Folkes Foundation (#2020-00124), the Gun and Bertil Stohnes Foundation, and the Anna Lisa and Brother Björnsson’s Foundation. ERZ receives financial support from CNPq (#435642/2018-9 and #312410/2018-2), Instituto Serrapilheira (#Serra-1912-31365), Brazilian National Institute of Science and Technology in Excitotoxicity and Neuroprotection (#465671/2014-4), FAPERGS/MS/CNPq/SESRS–PPSUS (#30786.434.24734.231120170), ARD/FAPERGS (#54392.632.30451.05032021), and Alzheimer’s Association (#AARGD-21-850670). TAP is supported by the NIH (#R01AG075336 and #R01AG073267) and the Alzheimer’s Association (#AACSF-20-648075).
Footnotes
↵† Data used to prepare this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. A complete listing of ADNI investigators can be found at http://adni.Loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf