The neural determinants of age-related changes in fluid intelligence: a pre-registered, longitudinal analysis in UK Biobank

A General Cognitive Ability Factor for the UK Biobank

UK Biobank participants do not have a high-quality measure of intelligence or polygenic scores (PGSs) of intelligence to simultaneously examine the genetic and neural underpinnings of intelligence. We created a standardized measure of general intelligence (g factor) relative to the UK population and estimated its quality. After running a GWAS of g on UK Biobank participants with a g factor of good quality and without neuroimaging data (N = 187,288), we derived a g PGS for UK Biobank participants with neuroimaging data. For individuals with at least one cognitive test, the g factor from eight cognitive tests (N = 501,650) explained 29% of the variance in cognitive test performance. The PGS for British individuals with neuroimaging data (N = 27,174) explained 7.6% of the variance in g. We provided high-quality g factor estimates for most UK Biobank participants and g factor PGSs for UK Biobank participants with neuroimaging data.

Analysis of associations between dietary patterns, genetic disposition, and cognitive function in data from UK Biobank

PURPOSE

Research suggests that diet influences cognitive function and the risk for neurodegenerative disease. The present study aimed to determine whether a recently developed diet score, based on recommendations for dietary priorities for cardio metabolic health, was associated with fluid intelligence, and whether these associations were modified by individual genetic disposition.

METHODS

This research has been conducted using the UK Biobank Resource. Analyses were performed using self-report data on diet and the results for the verbal-numerical reasoning test of fluid intelligence of 104,895 individuals (46% male: mean age at recruitment 57.1 years (range 40-70)). For each participant, a diet score and a polygenic score (PGS) were constructed, which evaluated predefined cut-offs for the intake of fruit, vegetables, fish, processed meat, unprocessed meat, whole grain, and refined grain, and ranged from 0 (unfavorable) to 7 (favorable). To investigate whether the diet score was associated with fluid intelligence, and whether the association was modified by PGS, linear regression analyses were performed.

RESULTS

The average diet score was 3.9 (SD 1.4). After adjustment for selected confounders, a positive association was found between baseline fluid intelligence and PGS (P < 0.001). No association was found between baseline fluid intelligence and diet score (P = 0.601), even after stratification for PGS, or in participants with longitudinal data available (n = 9,482).

CONCLUSION

In this middle-aged cohort, no evidence was found for an association between the investigated diet score and either baseline or longitudinal fluid intelligence. However, as in previous reports, fluid intelligence was strongly associated with a PGS for general cognitive function.

Multifactorial structure of cognitive assessment tests in the UK Biobank: A combined exploratory factor and structural equation modeling analyses

Introduction

The UK Biobank cognitive assessment data has been a significant resource for researchers looking to investigate predictors and modifiers of cognitive abilities and associated health outcomes in the general population. Given the diverse nature of this data, researchers use different approaches - from the use of a single test to composing the general intelligence score, g, across the tests. We argue that both approaches are suboptimal - one being too specific and the other one too general - and suggest a novel multifactorial solution to represent cognitive abilities.

Methods

Using a combined Exploratory Factor (EFA) and Exploratory Structural Equation Modeling Analyses (ESEM) we developed a three-factor model to characterize an underlying structure of nine cognitive tests selected from the UK Biobank using a Cattell-Horn-Carroll framework. We first estimated a series of probable factor solutions using the maximum likelihood method of extraction. The best solution for the EFA-defined factor structure was then tested using the ESEM approach with the aim of confirming or disconfirming the decisions made.

Results

We determined that a three-factor model fits the UK Biobank cognitive assessment data best. Two of the three factors can be assigned to fluid reasoning (Gf) with a clear distinction between visuospatial reasoning and verbal-analytical reasoning. The third factor was identified as a processing speed (Gs) factor.

Discussion

This study characterizes cognitive assessment data in the UK Biobank and delivers an alternative view on its underlying structure, suggesting that the three factor model provides a more granular solution than g that can further be applied to study different facets of cognitive functioning in relation to health outcomes and to further progress examination of its biological underpinnings.

The collective effects of genetic variants and complex traits

Traditional approaches in studying the genetics of complex traits have focused on identifying specific genetic variants. However, the collective effects of variants have remained largely unexplored. Here, we evaluated whether traits could be influenced by the collective effects of variants across the entire protein coding-region of the genome or the entire genome. We studied the UK Biobank exome sequencing data of 167,246 individuals as well as the genome-wide SNP array data of 408,868 individuals. We calculated for each individual four different measures of genetic variation such as heterozygosity and number of variants and two different measures of the overall deleteriousness of all variants, and performed correlations with 17 representative traits that have been studied previously. Linear regression analysis was performed with adjustment for age, sex, and genetic principal components. The results showed a high correlation among the six different measures and an inverse association of two well-correlated traits (educational attainment and height) with the total number of all variants as well as the overall deleteriousness of all variants. We have also categorized the genes based on whether they are expressed in the brain and found that the association with educational attainment only held for the brain-expressed genes. No other traits examined showed a significant correlation with the brain-expressed genes. The study demonstrates that common traits could be studied by analyzing the overall genetic variation and suggests that educational attainment is inversely related to genetic variation.

A multicohort geometric deep learning study of age dependent cortical and subcortical morphologic interactions for fluid intelligence prediction

The relationship of human brain structure to cognitive function is complex, and how this relationship differs between childhood and adulthood is poorly understood. One strong hypothesis suggests the cognitive function of Fluid Intelligence (Gf) is dependent on prefrontal cortex and parietal cortex. In this work, we developed a novel graph convolutional neural networks (gCNNs) for the analysis of localized anatomic shape and prediction of Gf. Morphologic information of the cortical ribbons and subcortical structures was extracted from T1-weighted MRIs within two independent cohorts, the Adolescent Brain Cognitive Development Study (ABCD; age: 9.93 ± 0.62 years) of children and the Human Connectome Project (HCP; age: 28.81 ± 3.70 years). Prediction combining cortical and subcortical surfaces together yielded the highest accuracy of Gf for both ABCD (R = 0.314) and HCP datasets (R = 0.454), outperforming the state-of-the-art prediction of Gf from any other brain measures in the literature. Across both datasets, the morphology of the amygdala, hippocampus, and nucleus accumbens, along with temporal, parietal and cingulate cortex consistently drove the prediction of Gf, suggesting a significant reframing of the relationship between brain morphology and Gf to include systems involved with reward/aversion processing, judgment and decision-making, motivation, and emotion.

Using large, publicly available data sets to study adolescent development: opportunities and challenges

Adolescence is a period of rapid change, with cognitive, mental wellbeing, environmental biological factors interacting to shape lifelong outcomes. Large, longitudinal phenotypically rich data sets available for reuse (secondary data) have revolutionized the way we study adolescence, allowing the field to examine these unfolding processes across hundreds or even thousands of individuals. Here, we outline the opportunities and challenges associated with such secondary data sets, provide an overview of particularly valuable resources available to the field, and recommend best practices to improve the rigor and transparency of analyses conducted on large, secondary data sets.

Longitudinal changes in the neural oscillatory dynamics underlying abstract reasoning in children and adolescents

Fluid reasoning is the ability to problem solve in the absence of prior knowledge and is commonly conceptualized as “non-verbal” intelligence. Importantly, fluid reasoning abilities rapidly develop throughout childhood and adolescence. Although numerous studies have characterized the neural underpinnings of fluid reasoning in adults, there is a paucity of research detailing the developmental trajectory of this neural processing. Herein, we examine longitudinal changes in the neural oscillatory dynamics underlying fluid intelligence in a sample of typically developing youths. A total of 34 participants age 10 to 16 years-old completed an abstract reasoning task during magnetoencephalography (MEG) on two occasions set one year apart. We found robust longitudinal optimization in theta, beta, and gamma oscillatory activity across years of the study across a distributed network commonly implicated in fluid reasoning abilities. More specifically, activity tended to decrease longitudinally in additional, compensatory areas such as the right lateral prefrontal cortex and increase in areas commonly utilized in mature adult samples (e.g., left frontal and parietal cortices). Importantly, shifts in neural activity were associated with improvements in task performance from one year to the next. Overall, the data suggest a longitudinal shift in performance that is accompanied by a reconfiguration of the functional oscillatory dynamics serving fluid reasoning during this important period of development.

How to remove or control confounds in predictive models, with applications to brain biomarkers

BACKGROUND

With increasing data sizes and more easily available computational methods, neurosciences rely more and more on predictive modeling with machine learning, e.g., to extract disease biomarkers. Yet, a successful prediction may capture a confounding effect correlated with the outcome instead of brain features specific to the outcome of interest. For instance, because patients tend to move more in the scanner than controls, imaging biomarkers of a disease condition may mostly reflect head motion, leading to inefficient use of resources and wrong interpretation of the biomarkers.

RESULTS

Here we study how to adapt statistical methods that control for confounds to predictive modeling settings. We review how to train predictors that are not driven by such spurious effects. We also show how to measure the unbiased predictive accuracy of these biomarkers, based on a confounded dataset. For this purpose, cross-validation must be modified to account for the nuisance effect. To guide understanding and practical recommendations, we apply various strategies to assess predictive models in the presence of confounds on simulated data and population brain imaging settings. Theoretical and empirical studies show that deconfounding should not be applied to the train and test data jointly: modeling the effect of confounds, on the training data only, should instead be decoupled from removing confounds.

CONCLUSIONS

Cross-validation that isolates nuisance effects gives an additional piece of information: confound-free prediction accuracy.

Neuroanatomical Correlates Underlying the Association Between Maternal Interleukin 6 Concentration During Pregnancy and Offspring Fluid Reasoning Performance in Early Childhood

BACKGROUND

Maternal inflammation during pregnancy can alter offspring brain development and influence risk for disorders commonly accompanied by deficits in cognitive functioning. We therefore examined associations between maternal interleukin 6 (IL-6) concentrations during pregnancy and offspring cognitive ability and concurrent magnetic resonance imaging-based measures of brain anatomy in early childhood. We further examined newborn brain anatomy in secondary analyses to consider whether effects are evident soon after birth and to increase capacity to differentiate effects of pre- versus postnatal exposures.

METHODS

IL-6 concentrations were quantified in early (12.6 ± 2.8 weeks), mid (20.4 ± 1.5 weeks), and late (30.3 ± 1.3 weeks) pregnancy. Offspring nonverbal fluid intelligence (Gf) was assessed at 5.2 ± 0.6 years using a spatial reasoning task (Wechsler Preschool and Primary Scale of Intelligence-Matrix) (n = 49). T1-weighted magnetic resonance imaging scans were acquired at birth (n = 89, postmenstrual age = 42.9 ± 2.0 weeks) and in early childhood (n = 42, scan age = 5.1 ± 1.0 years). Regional cortical volumes were examined for a joint association between maternal IL-6 and offspring Gf performance.

RESULTS

Average maternal IL-6 concentration during pregnancy was inversely associated with offspring Gf performance after adjusting for socioeconomic status and the quality of the caregiving and learning environment (R² = 13%; p = .02). Early-childhood pars triangularis volume was jointly associated with maternal IL-6 and childhood Gf (p_corrected < .001). An association also was observed between maternal IL-6 and newborn pars triangularis volume (R² = 6%; p = .02).

CONCLUSIONS

These findings suggest that the origins of variation in child cognitive ability can, in part, trace back to maternal conditions during the intrauterine period of life and support the role of inflammation as an important component of this putative biological pathway.

Population modeling with machine learning can enhance measures of mental health

BACKGROUND

Biological aging is revealed by physical measures, e.g., DNA probes or brain scans. In contrast, individual differences in mental function are explained by psychological constructs, e.g., intelligence or neuroticism. These constructs are typically assessed by tailored neuropsychological tests that build on expert judgement and require careful interpretation. Could machine learning on large samples from the general population be used to build proxy measures of these constructs that do not require human intervention?

RESULTS

Here, we built proxy measures by applying machine learning on multimodal MR images and rich sociodemographic information from the largest biomedical cohort to date: the UK Biobank. Objective model comparisons revealed that all proxies captured the target constructs and were as useful, and sometimes more useful, than the original measures for characterizing real-world health behavior (sleep, exercise, tobacco, alcohol consumption). We observed this complementarity of proxy measures and original measures at capturing multiple health-related constructs when modeling from, both, brain signals and sociodemographic data.

CONCLUSION

Population modeling with machine learning can derive measures of mental health from heterogeneous inputs including brain signals and questionnaire data. This may complement or even substitute for psychometric assessments in clinical populations.

Analysis of Diffusion Tensor Imaging Data From the UK Biobank Confirms Dosage Effect of 15q11.2 Copy Number Variation on White Matter and Shows Association With Cognition

BACKGROUND

Copy number variations at the 15q11.2 BP1-BP2 locus are present in 0.5%-1.0% of the population, and the deletion is associated with several neurodevelopmental disorders. Previously, we showed a reciprocal effect of 15q11.2 copy number variation on fractional anisotropy, with widespread increases in deletion carriers. We aim to expand these findings using a larger sample of participants (N = 29,166) and higher resolution imaging and by examining the implications for cognitive performance.

METHODS

Diffusion tensor imaging measures from participants with no neurological or psychiatric diagnoses were obtained from the UK Biobank database. We compared 15q11.2 BP1-BP2 deletion (n = 102) and duplication (n = 113) carriers to a large cohort of control individuals with no neuropsychiatric copy number variants (n = 28,951). Additionally, we assessed how changes in white matter mediated the association between carrier status and cognitive performance.

RESULTS

Deletion carriers showed increases in fractional anisotropy in the internal capsule and cingulum and decreases in the posterior thalamic radiation compared with both duplication carriers and control subjects (who had intermediate values). Compared with control subjects, deletion carriers had lower scores across cognitive tasks, which were partly influenced by white matter. Reduced fractional anisotropy in the posterior thalamic radiation partially contributed to worse cognitive performance in deletion carriers.

CONCLUSIONS

These results, together with our previous findings, provide convergent evidence for an effect of 15q11.2 BP1-BP2 on white matter microstructure, this being more pronounced in deletion carriers. Additionally, changes in white matter were found to partially mediate cognitive ability in deletion carriers, providing a link between white matter changes in 15q11.2 BP1-BP2 carriers and cognitive function.

Bridging Brain and Cognition: A Multilayer Network Analysis of Brain Structural Covariance and General Intelligence in a Developmental Sample of Struggling Learners

Network analytic methods that are ubiquitous in other areas, such as systems neuroscience, have recently been used to test network theories in psychology, including intelligence research. The network or mutualism theory of intelligence proposes that the statistical associations among cognitive abilities (e.g., specific abilities such as vocabulary or memory) stem from causal relations among them throughout development. In this study, we used network models (specifically LASSO) of cognitive abilities and brain structural covariance (grey and white matter) to simultaneously model brain-behavior relationships essential for general intelligence in a large (behavioral, N = 805; cortical volume, N = 246; fractional anisotropy, N = 165) developmental (ages 5-18) cohort of struggling learners (CALM). We found that mostly positive, small partial correlations pervade our cognitive, neural, and multilayer networks. Moreover, using community detection (Walktrap algorithm) and calculating node centrality (absolute strength and bridge strength), we found convergent evidence that subsets of both cognitive and neural nodes play an intermediary role 'between' brain and behavior. We discuss implications and possible avenues for future studies.

Crystallized and fluid intelligence are predicted by microstructure of specific white-matter tracts

Studies of the neural basis of intelligence have focused on comparing brain imaging variables with global scales instead of the cognitive domains integrating these scales or quotients. Here, the relation between mean tract-based fractional anisotropy (mTBFA) and intelligence indices was explored. Deterministic tractography was performed using a regions of interest approach for 10 white-matter fascicles along which the mTBFA was calculated. The study sample included 83 healthy individuals from the second wave of the Cuban Human Brain Mapping Project, whose WAIS-III intelligence quotients and indices were obtained. Inspired by the "Watershed model" of intelligence, we employed a regularized hierarchical Multiple Indicator, Multiple Causes model (MIMIC), to assess the association of mTBFA with intelligence scores, as mediated by latent variables summarizing the indices. Regularized MIMIC, used due to the limited sample size, selected relevant mTBFA by means of an elastic net penalty and achieved good fits to the data. Two latent variables were necessary to describe the indices: Fluid intelligence (Perceptual Organization and Processing Speed indices) and Crystallized Intelligence (Verbal Comprehension and Working Memory indices). Regularized MIMIC revealed effects of the forceps minor tract on crystallized intelligence and of the superior longitudinal fasciculus on fluid intelligence. The model also detected the significant effect of age on both latent variables.

A Hierarchical Watershed Model of Fluid Intelligence in Childhood and Adolescence

Fluid intelligence is the capacity to solve novel problems in the absence of task-specific knowledge and is highly predictive of outcomes like educational attainment and psychopathology. Here, we modeled the neurocognitive architecture of fluid intelligence in two cohorts: the Centre for Attention, Leaning and Memory sample (CALM) (N = 551, aged 5-17 years) and the Enhanced Nathan Kline Institute-Rockland Sample (NKI-RS) (N = 335, aged 6-17 years). We used multivariate structural equation modeling to test a preregistered watershed model of fluid intelligence. This model predicts that white matter contributes to intermediate cognitive phenotypes, like working memory and processing speed, which, in turn, contribute to fluid intelligence. We found that this model performed well for both samples and explained large amounts of variance in fluid intelligence (R2CALM = 51.2%, R2NKI-RS = 78.3%). The relationship between cognitive abilities and white matter differed with age, showing a dip in strength around ages 7-12 years. This age effect may reflect a reorganization of the neurocognitive architecture around pre- and early puberty. Overall, these findings highlight that intelligence is part of a complex hierarchical system of partially independent effects.

Aging-related changes in fluid intelligence, muscle and adipose mass, and sex-specific immunologic mediation: A longitudinal UK Biobank study

BACKGROUND

Obesity in midlife and early late-life is associated with worse normal cognitive aging. Dual-energy X-ray absorptiometry (DEXA) suggests that visceral adipose mass (VAM) plays a predominant role, whereas non-visceral adipose mass (NVAM) and lean muscle mass (LMM) have shown conflicting relationships. It is unknown how longitudinal, cognitive changes in age-sensitive domains like fluid intelligence (FI) correspond to VAM, NVAM, and LMM in women and men. Furthermore, changes over time in blood leukocyte sub-populations may partially or fully account for sex-specific associations.

METHODS

Data on 4431 late middle-aged, cognitively unimpaired adults (mean = 64.5 y) was obtained from the UK Biobank prospective cohort across 22 centers. FI scores, blood leukocyte counts, and covariates (age, social class, education) were measured at three 2-year intervals over 6 years. DEXA collection overlapped with these intervals. Sex-stratified growth curves, structural equations, and Preacher-Hayes mediation were used to estimate direct and indirect effects. β-weights were standardized.

RESULTS

More LMM predicted gains in FI scores among women (β = 0.130, p < .001) and men (β = 0.089, p < .001). Conversely, more VAM and NVAM independently predicted FI decline equally among sexes (e.g., NVAM: women: β = -0.082, p < .001; men: β = -0.076, p < .001). Among women, FI associations were fully mediated by higher eosinophil counts via VAM (λ = 30.8%, p = .028) and lower lymphocyte counts via LMM (λ = 69.2%, p = .021). Among men, FI associations were partially mediated by lower basophils counts via LMM (λ = 4.5%, p = .042) and higher counts via VAM (λ = 50%, p = .037).

CONCLUSION

The proportion of LMM and VAM equally influenced male FI changes over 6 years, whereas higher LMM among women appeared to more strongly influence. FI changes. Leukocyte counts strongly mediated VAM- and LMM-related FI changes in a sex-specific manner, but not for NVAM. For clinical translation, exercise studies in older adults may benefit from assessing sex-specific values of DEXA-based tissue mass, FI, and leukocyte sub-populations to gauge potential cognitive benefits of less VAM and more LMM.

Structural brain imaging correlates of general intelligence in UK Biobank

The associations between indices of brain structure and measured intelligence are unclear. This is partly because the evidence to-date comes from mostly small and heterogeneous studies. Here, we report brain structure-intelligence associations on a large sample from the UK Biobank study. The overall N = 29,004, with N = 18,426 participants providing both brain MRI and at least one cognitive test, and a complete four-test battery with MRI data available in a minimum N = 7201, depending upon the MRI measure. Participants' age range was 44-81 years (M = 63.13, SD = 7.48). A general factor of intelligence (g) was derived from four varied cognitive tests, accounting for one third of the variance in the cognitive test scores. The association between (age- and sex- corrected) total brain volume and a latent factor of general intelligence is r = 0.276, 95% C.I. = [0.252, 0.300]. A model that incorporated multiple global measures of grey and white matter macro- and microstructure accounted for more than double the g variance in older participants compared to those in middle-age (13.6% and 5. 4%, respectively). There were no sex differences in the magnitude of associations between g and total brain volume or other global aspects of brain structure. The largest brain regional correlates of g were volumes of the insula, frontal, anterior/superior and medial temporal, posterior and paracingulate, lateral occipital cortices, thalamic volume, and the white matter microstructure of thalamic and association fibres, and of the forceps minor. Many of these regions exhibited unique contributions to intelligence, and showed highly stable out of sample prediction.

Age and cognitive decline in the UK Biobank

OBJECTIVES

Age-related cognitive decline is a well-known phenomenon after age 65 but little is known about earlier changes and prior studies are based on relatively small samples. We investigated the impact of age on cognitive decline in the largest population sample to date including young to old adults.

METHOD

Between 100,352 and 468,534 participants aged 38-73 years from UK Biobank completed at least one of seven self-administered cognitive functioning tests: prospective memory (PM), pairs matching (Pairs), fluid intelligence (FI), reaction time (RT), symbol digit substitution, trail making A and B. Up to 26,005 participants completed at least one of two follow-up assessments of PM, Pairs, FI and RT. Multivariable regression models examined the association between age (<45[reference], 45-49, 50-54, 55-59, 60-64, 65+) and cognition scores at baseline. Mixed models estimated the impact of age on cognitive decline over follow-up (~5.1 years).

RESULTS

FI was higher between ages 50 and 64 and lower at 65+ compared to <45 at baseline. Performance on all other baseline tests was lower with older age: with increasing age category, difference in test scores ranged from 2.5 to 7.8%(P<0.0001). Compared to <45 at baseline, RT and Pairs performance declined faster across all older age cohorts (3.0 and 1.2% change, respectively, with increasing age category, P<0.0001). Cross-sectional results yielded 8 to 12-fold higher differences in RT and Pairs with age compared to longitudinal results.

CONCLUSIONS

Our findings suggest that declines in cognitive abilities <65 are small. The cross-sectional differences in cognition scores for middle to older adult years may be due in part to age cohort effects.