Participants

Imaging data were acquired as part of the ASPREE- NEURO sub-study [16], which aimed to determine the effect of low-dose aspirin on a range of MRI biomarkers (cerebral microbleeds, white matter hyperintensities) over a three-year period, in neurotypical adults aged 70-yrs and over. The ASPREE clinical trial was double-blinded, with approximately half of the participants taking aspirin, and half taking placebo. Analysis of the main study aims is in progress, and information about which participants were allocated to each group is not yet unblinded [18]. We included data from baseline (prior to study medication) to test the main hypotheses. We used the data from Year 1 and 3 (after randomisation to study medication) to test the reproducibility of the main findings (see Data Analysis section).

Healthy older adults in Australia were eligible to participate if aged 70-years and over, had no history of occlusive vascular disease, atrial fibrillation, cognitive impairment or disability; were not taking antithrombotic therapy, and did not have anaemia or high risk of major bleeding, or a diagnosis likely to cause death within five years. At study entry, each participant had a Modified Mini Mental Status Examination (3MS; [19]) score of at least 78/100, and were able to perform all six of the Katz activities of daily living (ADLS; [20]).

Participants included here (N = 573) were aged 70–88 years. Data from 21 participants contained incomplete MRI scans and were discarded. Data from five participants were excluded due to the presence of neurological conditions (history of brain cancer (n = 1), Parkinson’s disease (n = 4)). In the final sample, data from 547 participants aged 70–88 years (mean age [+/-stdev] = 73.9+/-3.5years, 89.06% right-handed) were retained, including n = 287 males (74.0 ±3.6 years), and n = 260 females (73.7 ±3.4 years). The study attrition is given in S1 Table in S1 File, and shows a 15% attrition across 3 years.

As part of a health outcomes questionnaire, ASPREE-NEURO participants were asked “How many children do you have?”. More detailed parenthood data (e.g. biological/adopted children, grandparenthood, primary/secondary caregiver etc.) were not collected. The number of children parented by males and females in our sample is shown in S1 Fig in S1 File.

Data acquisition

Data analysis

A regression was conducted to control for the effects of age, education, systolic and diastolic blood pressure, BMI, cholesterol, and the Euler characteristic (a FreeSurfer-derived measure of data quality; [28]). These covariates were selected based on previous studies reporting correlations between these factors and cortical thickness [28–30], and the expectation of a relationship between education and parity. All subsequent analyses used the Z-scores of these residuals, and were non-parametric as number of children is discrete, ordinal data, with a non-normal distribution.

The ‘dose’ effect of the number of children on cortical thickness was tested using Spearman’s correlation (two-tailed). The analysis was performed on all 70 thickness measures for 504 older adult participants, n = 269 male (74.0 ±3.5 years) and 235 female (73.8±3.5 years) participants who had one or more children. To test the hypothesis that cortical thickness would differ between parents and non-parents, for both sexes, the cortical thickness of 70 brain regions were compared between individuals with one child and individuals with no children in 80 older adult participants, including males (n = 35; 73.4 ±3.7 years) and females (n = 46; 72.8 ±3.3 years). The effect sizes were quantified using Cohen’s d. Significance was estimated using a permutation test (10,000 permutations of group assignments).

Data analysis for the cognitive tasks mirrored the analysis structure of the MRI data. All cognitive tasks (COWAT, HVLT-R, SDMT, Color Trails, Stroop and CES-D) were compared between participants who were childless (non-parents) and those who were parents to one child. The relationship between all cognitive tasks and number of children was also examined for participants with more than one child.

Note that we did not apply Bonferroni correction to our results. However, given the large number of comparisons in this study, we examined the reproducibility of the findings across timepoints within our data set to identify effects worthy of future study. Participants were scanned at three timepoints, at baseline (prior to administration of study medication), Year 1 and Year 3. We tested our hypotheses with data from Timepoint 1 (prior to aspirin exposure) and then tested the reproducibility of our MRI results by examining if they were maintained at Timepoints 2 and 3. The effect of parenthood in late life is not expected to change across the three-year measurement period; rather, any differences in the relationship between parenthood and cortical thickness across timepoints should be largely attributable to measurement error. Note, that this approach tests for false positives due to measurement error, it does not control for false negatives. We report the main results for Timepoint 1 in the Results section; the reproducibility results are reported in S2 and S3 Tables in S1 File. Whilst we acknowledge the importance of reducing the inflation of type I errors, given the novelty of the research question, it is also important to identify potential effects for future studies, so that hypotheses can be formed for specific regions in replication studies.

Motherhood and cortical thickness

We found a positive relationship between number of children parented and cortical thickness in the parahippocampal gyrus of older adult mothers. This result is consistent with findings in early stages of parenthood, where grey matter changes are seen in the parahippocampal gyrus during pregnancy [7] and the postpartum period [5]. The parahippocampal gyrus is involved in memory consolidation and cohesion [33] and mediates cortico-hippocampal communication [34], and atrophy in the parahippocampal gyrus plays a key role in age-related memory decline [34]. A positive association between number of children parented and cortical thickness in a region associated with memory is compatible with our cognitive results, where memory outcomes were improved in older adult women with more children. Parenthood appears to be associated with a small but reliable improvement in memory outcomes for older mothers. This result is consistent with evidence from early stages of parenthood, where memory is impaired during pregnancy [35], returns to pre-pregnancy levels in the postpartum period [36], and, at least in rodents, is enhanced in late parenthood [10].

Our result is consistent with results seen in preclinical models. Spatial learning and memory are improved in female rodents with reproductive experience, and this effect is stronger in mothers with more than one pregnancy [37]. This cognitive benefit persists into late-life for parous female rats, who have increased hippocampal long-term potentiation, enhanced memory capacities and fewer signs of brain aging compared with aged nulliparous females; an effect that is enhanced with multiparity [10]. Our findings, that both memory performance and parahippocampal thickness increase with the number of children parented in female humans, are consistent with these reports and suggest that parenthood is positively related to human cognition in late life. We found no difference in cognitive performance between older mothers with one child (N = 20), and older women with no children (N = 25). It is possible that these effects would be observed using a larger sample size, or that memory ability is only linked to the cumulative hormonal and environmental changes associated with mothering a number of children.

Compared with childless women, those with one child showed thinner grey matter in the left dorsolateral prefrontal cortex. The dorsolateral prefrontal cortex is involved in high-order cognitive functions including executive functioning, cognitive control and emotional regulation [38]. Neuroimaging data highlights the importance of the prefrontal cortex in parenting behaviours, as this region is activated in almost every fMRI study of human mothers’ responses to infant stimuli [32]. Increased brain activity in a mother’s prefrontal cortex in response to her own baby compared to an unfamiliar baby, is positively associated with maternal sensitivity and higher quality of mother-infant interactions [39]. Conversely, decreased fMRI activation in the prefrontal cortex is correlated with difficulties adjusting to parenthood [40]. Neural changes in early stages of motherhood are likely to have evolved to assist in the dual survival of parent and young [3]. Three recent studies have shown maternal grey matter thickness [6] and volume [5, 7] changes in the prefrontal cortex across pregnancy and the early post-partum period. In all three of these studies, structural change was associated with adaptive maternal factors (increased attachment; higher parental self-efficacy; more positive opinions about their baby). Our results indicate that parenthood-related changes in the cortical thickness of the dorsolateral prefrontal cortex in mothers are present in late life.

Our results show thinner cortical thickness of the pericalcarine sulcus and cuneus for females. These results appear in both hemispheres across both analyses (childless versus one child, and number of children). This finding is intriguing, as it is not usually a region associated with age-related decline (e.g., [41]). Interactions with children provide a rich array of sensory experiences for mothers. Interactions involving multi-sensory cues (e.g., visual, auditory, tactile) between mothers and infants throughout the postpartum period, and continuing into late-parenthood, may alter the structure and function of the maternal sensory regions over time [6]. In female rodents, neural reorganisation of the olfactory bulb, parietal lobe, and somatosensory cortex were associated with the expression of maternal behaviours during the postpartum period [42]. This result may therefore suggest that sensory changes that occur in the postpartum period that serve to enhance maternal behaviours may be still present in older age.

Fatherhood and cortical thickness

Compared with childless men, fathers showed thinner grey matter in the left anterior cingulate cortex and thicker grey matter in the right temporal pole. The effect in the temporal pole was notably weaker across timepoints, and so this effect may be a false positive result. These regions are involved in social cognition, including emotional regulation, empathy and theory of mind [43, 44]. Social cognition is a hallmark of the human parental brain, especially relevant for pre-verbal infants [45]. Theory of mind abilities, including empathy, contribute to successful social interactions and are important for human survival. Our results are consistent with previous studies in early stages of parenthood that found grey matter volume changes in the anterior cingulate cortex associated with early parenthood and fathers [8], and suggests they may persist into late life.

Parenthood and cortical thickness in late life

We found modest but reproducible relationships between parenthood and cortical thickness in late life. As expected, the results were smaller in males than in females. We argue that these late-life structural brain differences associated with parenthood are related to the environmental complexity of parenthood. While the drastic hormonal shifts of pregnancy may play an initial role in mothers, the hormonal milieu of pregnancy returns to a pre-pregnancy state following parturition, at which time many hormonal and biological changes are resolved, i.e., uterus size, lactation, etc [46]. In our sample of older adults, the hormonal exposure of pregnancy occurred three or more decades earlier. In contrast, the environmental complexity of parenthood endures throughout the entire parenting experience, possibly throughout the life-span of a parent, and compounds with increasing number of children. Indeed, it is known from preclinical models that the mere presence of offspring is sufficient to produce neural changes associated with parenting, even in the absence of pregnancy-related hormones [47]. Child-rearing provides parents of both sexes with rich sensory stimulation and increased environmental complexity, with new parents required to engage in novel tasks and challenges [48]. This is consistent with the conceptualisation of caregiving as a dynamic, learning experience in humans, and as a period of plasticity [48, 49]. As such, parenthood is a continuum of experience, and may represent a learning environment that is sustained for two or more decades of an individual’s life [49]. Parenthood may therefore contribute to a person’s cognitive reserve, by increasing the cognitive and social demands of the individual.

In the present study, cortical thickness differences between parents and non-parents were evident in both sexes. This novel finding implicates the parental environment, which exists for both sexes. The analysis of cortical thickness differences between parents and non-parents included smaller sample sizes than the multiparity analyses, and this should be taken into account when evaluating the results. The neuroanatomical changes associated with parenting multiple children were only apparent for mothers, with no significant relationship identified in fathers. This may reflect differences in care-giving responsibilities between the sexes: the cohort examined in this study most likely included participants who were predominantly in ‘traditional’ care-giving arrangements, with roles and responsibilities as primary care-giving mothers and ‘bread-winning’ fathers [50]. The social structure of this generation may have limited the cumulative impact of parenthood for the males in this cohort. Therefore, the different effects observed in females and males may reflect the likely primary and secondary caregiver status of our participants, respectively, rather than the distinction between motherhood and fatherhood. Similar studies in future generations, with a presumably larger proportion of primary care-giving fathers, may see results that instead scale with time spent in care, rather than sex of the parent.

The diverse and subtle changes in cortical grey matter thickness measured in this study point to a rich and multi-layered process of cortical plasticity that begins during pregnancy (in females) and the early post-partum period (males and females), and is evident across the lifespan. It is important to note that we have observed both positive and negative associations with cortical thickness in relation to parenthood for both males and females. These regions have also shown grey matter changes across pregnancy and the early postpartum period (Table 4). In early parenthood, these regions have shown both increases [4–6, 8, 9, 31] and decreases [7, 8] in grey matter thickness or volume associated with measures of positive parental care (e.g. attachment, self-efficacy, positive opinions of baby). Generally, decreasing grey matter thickness is understood as neurodegeneration (atrophy) in ageing brains (e.g., [51]). However, in relation to parenthood, reductions in grey matter have been explained as a ‘fine-tuning’, similar to what we see in puberty [7]. Indeed, the morphometric changes (surface area, cortical thickness) of pregnancy highly resemble those occurring during adolescence, and decreased grey matter may reflect processes of synaptic pruning and myelination [52]. The physiological processes that can alter cortical thickness include changes in the number of synapses, glial cells or neurons, morphological changes to dendritic structure and axonal myelination, and vasculature changes—including angiogenesis—that influence blood volume and perfusion. While the study design precludes identification of the neurobiological mechanisms that underpin the observations, the novelty and social importance of the findings highlight the need for further studies: in vivo, to examine the microstructural, functional and neurovascular characteristics of parenthood; and post-mortem stereological, to determine the cellular and morphological origins of the cortical thickness changes. Without studies like these, it is very difficult to determine whether increases and/or decreases in grey matter thickness are adaptive for the late life brain in relation to parenthood.

The relationships between number of children parented and cortical thickness and memory performance found for mothers are potentially non-linear. In the current study, the relationship was obtained using Spearman’s correlation, which does not assume linearity, and instead measures the strength and direction of a monotonic relationship. A non-linear relationship between brain structure and number of children parented has been shown in a large sample of middle-aged women [14]. However, this non-linearity was not replicated in their follow-up study, showing region-specific effects in the same sample which were linearly related to the number of children parented [53]. Future studies should explore the potential for non-linear associations between parenthood and brain structure in middle- and late-life.

The data used in this study were acquired as part of a larger study with broader aims. Detailed demographic information related to parenthood was not obtained from the participants, and so information about the parenting experiences is not available. Many of these unknown factors, such as birth and feeding methods, hormone levels, parental sensitivity, and involvement with child-rearing (e.g. primary or secondary caregiver, living or not living with children, adoption, and so on) may also potentially impact maternal neuroplasticity [54]. We were also unable to control for pregnancies that did not result in live birth and/or parenthood experience (miscarriage, termination, still-birth, infant loss, or adoption). Therefore, we are not able to guarantee that some of our participants didn’t experience additional exposure to pregnancy hormones. Furthermore, it is important to note that reproductive experience is related to other traits which impact brain health in human ageing, including exposure to sex steroid hormones. Some of these traits are heritable (e.g. age at menarche and menopause) and common genetic variation should be controlled for in future studies [14, 55]. Parity is also associated with socio-environmental factors, and as such, these effects may differ with different access to contraception and childcare. In the absence of these data, we cannot rule out the possibility that the observed relationships between cortical thickness and parenthood are explained by some of these uncollected variables. Future studies should control for these factors, as well as examine the potential effects of grandparenthood and other opportunities for interaction with children during later life. Further, prospective longitudinal studies are needed to draw conclusions about the longevity of these effects; whether our observed relationships are evident across the life-span and reflect effects that have stabilised following the early postpartum period, or whether these effects are specific to late-life parenthood (caring for adult children and potentially grandchildren). Future studies to investigate neurobiological changes across parenting of multiple children in humans are required, as there are currently no such studies in humans at any life-stage [48, 56]. Our findings are the first to show a relationship between number of children and grey matter thickness in the ageing human brain. These preliminary findings should be interpreted in light of their limitations, and we encourage future research to replicate these effects in a prospective longitudinal study.