The Influence of Birth Weight, Socio-Economic Status, and Adult Health on Brain Volumes during Ageing

The health of the ageing brain is the cumulative consequence of genetic and external factors experienced from conception. These influence childhood brain development and late-life brain deterioration. Greater brain volume in old age is associated with resilience to pathology and cognitive health [1, 2]. In contrast, brain atrophy accompanies mild cognitive impairment and Alzheimer’s dementia [3, 4]. Poverty and health are interdependent and could have direct or indirect effects on the brain. Therefore, we must know the lifelong exposure and interaction of these factors to reveal their direct or indirect effects.

Small for gestational age neonates have experienced intrauterine growth restriction (IUGR) influenced by factors exacerbated by poverty. IUGR affects infant brain development. Premature small for gestational age infants have reduced brain and grey matter (GM) volumes at 1 year of age and in adolescence [5, 6]. Birth weight is associated with cortical surface area and has lifelong effects on cognitive ability [7] lasting into late-life. In adults aged 71–74 y, their birth weight correlated with GM and brain volume [8].

Low birth weight (LBW) could directly affect brain volume or influence volume via adult health with LBW increasing the risk of adult hypertension, insulin resistance [9] and type II diabetes [10]. Childhood poverty has profound effects on brain volumes throughout life. Hanson et al. [11] found that in children aged 7–18 y, SES was correlated with amygdala and hippocampal volume. Childhood adversity and lower status paternal occupation are associated with reduced brain cortical thickness, GM volume, and hippocampal volume in mid- to late-life [12]. Lower childhood SES is associated with lower adult SES and poor adult health [13, 14]. Poor adult SES correlates with smoking prevalence [15], hypertension [16], type 2 diabetes [17], and obesity [18], all associated with brain pathology.

During late-life, a higher SES is associated with greater volume and less brain pathology [19, 20]. The bidirectional link between poor adult health and SES is well established; impaired health reduces economic activity and increases the risk of poverty, conversely, lower SES causes poor health due to reduced healthcare access, and reduced engagement with health advice [21]. Hypertension [22], obesity [23], inflammation [24], and diabetes [25] are associated with reduced brain volumes and these conditions in midlife affect the brain in later life.

Therefore, there is evidence for both direct and indirect effects of early life on the health of the ageing brain. Socio-economic status, health, and brain health are interconnected throughout life so whole life data are needed to understand their interactions and influence on the brain in late-life. We have historical data on birth weight and childhood socio-economic status, collected contemporaneously and current measures of SES, health and brain MRI. These data allow us to interrogate the direct and indirect determinants of brain health using reliable, longitudinal, and detailed measures of the life-course environment.

Using data from a community living, older, non-demented sample, we investigated whether life-course health and SES factors are associated with brain volume and structure. We used dimension reduction and structural equation modelling to test if factors could directly or indirectly affect the late-life brain. We hypothesised that birth weight, childhood SES, and adult health affect brain size and structure in late-life through direct mechanisms.

Participants were born between 1950 and 1956 in Aberdeen, Scotland. In 1962, the Aberdeen Child Development Survey (ACDS) linked childhood records, questionnaires completed by parents, and school records. Between 1999 and 2001, 7,183 members of ACDS were traced and recruited into the Aberdeen Children of the Nineteen Fifties cohort (ACONF). The ACONF participants completed a questionnaire of recollections of childhood experiences, and current health and SES. These data were linked to historical data from ACDS and the Aberdeen Maternal and Neonatal Database [26].

Between 2015 and 2017, 313 participants of the ACONF cohort were recruited into the STratifying Resilience and Depression Longitudinally (STRADL) study. 280 participants had complete datasets for analysis and were included here. Missing were MRI data (n = 25), ACONF data (n = 5), and STRADL health data (n = 3). Details are available in the online supplementary materials (for all online suppl. material, see https://doi.org/10.1159/000541918).

This study received ethical approval from the Scotland Research Ethics Committee (number 14/55/0039). Participants provided written informed consent. The ACONF research database is registered with the National Research Ethics Service.

Between 1999 and 2001, aged 43–51 y, ACONF participants completed a questionnaire of health and SES [27]. Questions on childhood socio-economic status included: father’s occupation, family and house size and structure, car use and parental tobacco use. Birth weight and pregnancy duration were obtained from the AMND. Their SES was assessed by questioning their occupation, income, family size and structure, educational achievement, marital or cohabiting status, and car availability. Health was assessed by questions on smoking and alcohol consumption and a diagnosis of diabetes or hypertension. The participants self-reported their height and weight.

Socio-economic and health indices were derived from these data. Father’s occupation was classified using the standard occupational classification of 1990 and further analysed by socio-economic group, the registrar’s general social class and by Nuffield classification [28]. Living density (persons per room) was calculated from family and house size data. Birth weight was corrected for pregnancy duration. The educational qualifications were assigned a score from 0 to 7, from no qualifications to degree level certification. Smoking history was calculated from smoking status, duration, and packs smoked per week.

Participants were imaged on a 3T Philips Achieva TX-series MRI system (Philips Healthcare, Best, Netherlands) with a 32 channel phased-array head coil (software version 5.1.7; gradients with maximum amplitude and slew rate of 80 mT/m and 100 T/m/s). Structural brain images were acquired using a 3D T1-weighted fast gradient echo with magnetisation preparation. T1-weighted images were analysed using Freesurfer version 6.0 and segmentation checked for errors. Intracranial volume (TICV), brain, GM, and hippocampal volumes were extracted.

Systolic and diastolic blood pressure were measured in duplicate by a research nurse using a sphygmomanometer (Omicron). Mean grip strength was measured with a dynamometer (Patterson Medical Jamar). Height and weight were measured using calibrated scales and BMI calculated. Participants were asked their smoking history, alcohol intake, and diagnoses of hypertension and/or diabetes. The General Health Questionnaire (GHQ) assessed self-perceived psychological and physical well-being [29].

Principal component analysis was used reduce the dimensions of the data to a single component (a factor). Factors were extracted for a participant and their father’s occupation (occupation factor and paternal occupation factor) childhood and adult SES factors and adult health factor. The first un-rotated principal component, which best describes the variance shared between tests, was used in subsequent analyses. For each factor, the contributing variables and their contribution to their general factor are detailed in the online supplementary materials.

Multiple regression was performed using brain, GM and hippocampal volume as dependent variables. Models included age at MRI and sex as covariates. For each brain structure, a “childhood” and “whole life” model were constructed. The childhood model included birth weight and childhood SES factor, and the factors found to be significantly associated with the dependent variable were added to the “whole life” model. This model included adult health factor and adult SES factor. For brain volume models, intracranial volume was added as a confounding variable. For regional volume models, brain volume was included as a confounding variable to enable detection of effects independent of whole brain atrophy. Analyses were performed using R [30]. Structural equation models were constructed using variables found, by regression, to be associated with brain volumes. The path diagram structure was design was informed by the temporal sequence of factors or their structural relationship. Non-directional relationships were connected using a correlation term. Structural equation models were constructed using AMOS version 27. The fit of each model was assessed by χ² significance (p > 0.05), the goodness, incremental, and comparative fit indices (all >0.95) and Root Mean Square Error of Approximation (RMSEA, <0.08). Models fits of the data were acceptable and indices detailed in the online supplementary materials (online suppl. Table S6).

A summary of demographic, MRI, and health data are shown in Table 1. There was no significant difference in birth weight, childhood SES factor, adult SES factor, sex, or adult health factor between those with or without imaging.

Occupations of the participant and their father were each graded using 4 methods and composite occupation measures, their factors, were extracted using principal component analysis. The participant and paternal occupation factors accounted for 78% and 82% of the variance these grades, respectively. Twenty-two variables were analysed to extract the factor of childhood SES which accounted for 24% of their variance. Eight variables were used for the adult SES factor which accounted for 35% of their variance. Fifteen variables were analysed for the adult health factor which accounted for 19% of their variance. Details of the principal component analyses performed are contained within the Supplementary Materials.

Table 2 shows of the regression analysis of childhood and adult general factors on late-life brain volume. Brain volume was unaffected by age and being male predicted a 0.2 SD (21 mL) increased volume.

Birth weight was associated with brain volume, with each SD predicting an increased brain volume of 0.1 SD equating to 22 mL brain volume per 1 kg birth weight (2% of brain volume). Increased childhood SES factor predicts greater brain volume. Birth weight and childhood SES factor effects remained significant when including adult factors in the model. Adult SES factor did not independently predict brain volume. Adult health was associated with brain volume, with each SD of adult health resulting in an 0.17 SD increase in brain volume. A structural equation model of these data is shown in Figure 1. This analysis supported the findings made with multiple regression, showing that BW correlated with adult SES factor, childhood SES factor was correlated with adult health factor and adult SES factor, and there was a strong correlation between adult SES factor and adult health factor.

Table 3 details the regression analysis of factors influencing GM and hippocampal volume. Models included brain volume as a covariate; therefore, any finding is independent of effects on brain volume. Mean GM volume was greater in men. Childhood SES factor was associated with GM volume, with an increase of 2 mL per SD childhood SES factor. No adult factors measured influenced GM volume. Hippocampal volume was only associated with brain volume. Figure 2 shows the structural equation model of determinants of GM volume and the hippocampi.

We demonstrate that birth weight, childhood SES, and adult health are associated with late-life brain volumes through direct mechanisms. In addition, a poor childhood SES is associated with a smaller proportion of GM relative to brain. Adult health was correlated with greater brain volume. Hippocampal volume was not associated with birth weight, childhood SES, adult SES, or adult health. Birth weight, SES, and adult health acted directly on late-life regional and whole brain volumes.

Birth weight was associated with late-life brain volume, independently of SES or adult health. Brain volumes are affected by their environment before birth. In utero, fetal growth is sensitive to maternal nutrition, tobacco use, and maternal stress hormones that affects birth weight [31, 32]. These factors also affect brain development [33‒35], and it is probable that IUGR is associated with altered brain growth and development. It is likely that IUGR is a measure of harmful factors that have change prenatal brain growth development and remain evident as altered brain volumes in late life.

Childhood SES had a direct association with late-life brain volumes. Until age 20 y, the brain undergoes development processes, such as neuronal growth and pruning of synapses which are sensitive to SES [36]. SES is influenced by family size and wealth and can effect nutrition, health care, education, and environment. Disruption of these factors could have permanent consequences on brain structure and health.

The direct association between birth weight and childhood SES and late-life brain volumes may operate directly by two, non-mutually exclusive, mechanisms. They can change the structure and development of the brain, causing greater initial volume and altered proportions of tissue. These changes can persist, and cause differences in late-life brain volumes. By this direct mechanism, childhood factors can increase brain volume, synaptic density and contribute to cognitive reserve.

The second mechanism is via accelerated atrophy, with early life experiences altering vulnerability to neurodegeneration. Early life factors, including nutrition, alter late-life epigenetic methylation and are associated with late-life brain volumes [12, 37]. Birth weight and childhood stress also effect adult hormonal regulation and immune system function [38, 39]. Immune system dysfunction and stress hormone exposure are associated with diseases such as AD [40, 41]. Birth weight and childhood SES may act on the brain through changes in epigenetic programming of metabolism or adult hormonal and immune system function.

We found no effect of early life factors on hippocampal volume. Evidence for the influence of childhood SES on hippocampal volume in late-life is mixed. Studies have found correlations with childhood, but not adult, hippocampal volume [42] and childhood mistreatment, not SES, affects adult hippocampal volume [43]. This cohort may be too young for hippocampal atrophy to be detectable by MRI. When modelling hippocampal volume, we included brain volume as a confounding variable and hypothesised an effect that is specific to the hippocampi. Several previous findings of hippocampal volume did not include brain volume, but rather ICV, as a confounding factor and therefore hippocampal findings may have been non-specific and actually due to brain volume changes.

Adult health was associated with late-life brain volumes, potentially by directly promoting brain atrophy or because it could share a common underlying cause of poor brain health. Hypertension, diabetes, and obesity are associated with reduced brain volume [17, 44, 45]. A major contributor to the adult health factor is smoking. Smoking is a major risk for other health factors probably causes brain atrophy [46]. Considering smoking is crucial when investigating brain health in older populations, as it may be a primary, underlying cause of poor general and brain health.

We show that early life factors have a direct effect on the late life brain. The use of a detailed longitudinal dataset enabled us to overcome the analytical challenge caused by the close association between childhood and adult health and/or SES. We found early life factors directly affect the brain and had indirect effect on the brain through adult health. Research should focus on the specific details of how and when during early life the brain is affected so interventions that benefit brain health and resilience during late life can be developed.

Using PCA analysis to create adult health factor, childhood SES factor and adult SES factor scores captured 20–40% of the variability between their constituent variables. These data reduction can miss the importance of a single variable on brain volumes. In a cohort study, it is impossible to prove causal relationships, and unmeasured factors could be involved in the relationships measured.

We used multiple measures of SES and adult health to model the wider experience of an individual during their life and using retrospective questionnaires and contemporaneous measurement, allowing longitudinal analysis. These data avoid inaccuracies from data based on potentially emotive childhood memories. We exploited this longitudinal dataset using structural equation modelling, examining causal hypotheses. By including TICV and global brain volume as confounding factors our findings are specific for the tissue of interest and not global volume changes.

During late-life, reduced GM and whole brain volume are independently linked to low birth weight and childhood hardship. Crucially, these associations are direct and not via poor adult health. Early life factors directly affect late-life brain volume. Early life poverty and IUGR may directly increase dementia risk in late-life.

This study received ethical approval from the Scotland Research Ethics Committee (No. 14/55/0039). Participants provided written informed consent. The ACONF research database is registered with the National Research Ethics Service.

No conflict of interests.

Wellcome Trust [104036/Z/14/Z, 220857/Z/20/Z, and 216767/Z/19/Z], Chief Scientist Office of the Scottish Government Health Directorates [CZD/16/6], the Scottish Funding Council [HR03006], and UKRI Award MR/W014386/1 funded the project. The funders had no role in the design, data collection, data analysis, and reporting of this study.

All authors substantially contributed to the concept or design of the project. C.J.M., T.H., A.S., and G.W. did the acquisition, analysis, and interpretation of the data. The authors either wrote the manuscript (C.J.M.) or revised it for content (T.H., A.S., G.W., H.W., A.D.M.). All authors approved the final manuscript.

Non-identifiable information from the STRADL cohort is available to researchers in the UK and to international collaborators through application to the Generation Scotland Access Committee (gro.dnaltocsnoitareneg@ssecca) and through the Edinburgh Data Vault (https://doi.org/10.7488/8f68f1ae-0329-4b73-b189-c7288ea844d7). Generation Scotland operates a managed data access process including an online application form, and proposals are reviewed by the Generation Scotland Access Committee. The data and samples collected by the STRADL study have been incorporated in the main Generation Scotland dataset and governance process. Summary information to help researchers assess the feasibility and statistical power of a proposed project is available on request by contacting gro.dnaltocsnoitareneg@secruoser. ACONF data are available for investigators by application to [email protected], where data access is managed by the ACONF Steering committee. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.