Age‐related differences in functional network segregation in the context of sex and reproductive stage

Abstract Age is accompanied by differences in the organization of functional brain networks, which impact behavior in adulthood. Functional networks become less segregated and more integrated with age. However, sex differences in network segregation declines with age are not well‐understood. Further, network segregation in the context of female reproductive stage is relatively understudied, though unmasking such relationships would be informative for elucidating biological mechanisms that contribute to sex‐specific differences in aging. In the current work, we used data from the Cambridge Centre for Ageing and Neuroscience (Cam‐CAN) repository to evaluate differences in resting‐state network segregation as a product of sex and reproductive stage. Reproductive stage was categorized using the Stages of Reproductive Aging Workshop (STRAW+10) criteria. Replicating prior work, we investigated the following functional networks: auditory, cerebellar‐basal ganglia, cingulo‐opercular task control, default mode, dorsal attention, fronto‐parietal task control, salience, sensory somatomotor mouth, sensory somatomotor hand, ventral attention, and visual. First, our results mirror findings from previous work indicating that network segregation is lower with increasing age. Second, when analyzing associations between network segregation and age within each sex separately, we find qualitative differences between females and males. Finally, we report significant effects of reproductive stage on network segregation, though these findings are likely driven by age. Broadly, our results suggest that impacts of sex may be important to evaluate when investigating network segregation differences across adulthood, though further work is needed to determine the unique role of menopause and sex hormones on the organization of functional brain networks within aging females.

tion in the context of female reproductive stage is relatively understudied, though unmasking such relationships would be informative for elucidating biological mechanisms that contribute to sex-specific differences in aging. In the current work, we used data from the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) repository to evaluate differences in resting-state network segregation as a product of sex and reproductive stage. Reproductive stage was categorized using the Stages of Reproductive Aging Workshop (STRAW+10) criteria. Replicating prior work, we investigated the following functional networks: auditory, cerebellar-basal ganglia, cingulo-opercular task control, default mode, dorsal attention, fronto-parietal task control, salience, sensory somatomotor mouth, sensory somatomotor hand, ventral attention, and visual. First, our results mirror findings from previous work indicating that network segregation is lower with increasing age. Second, when analyzing associations between network segregation and age within each sex separately, we find qualitative differences between females and males. Finally, we report significant effects of reproductive stage on network segregation, though these findings are likely driven by age. Broadly, our results suggest that impacts of sex may be important to evaluate when investigating network segregation differences across adulthood, though further work is needed to determine the unique role of menopause and sex hormones on the organization of functional brain networks within aging females.

K E Y W O R D S
aging, functional connectivity, menopause, network segregation, sex differences 1 | INTRODUCTION With older age comes normative functional differences in both cognitive and motor domains (Harada et al., 2013;Leal & Yassa, 2014;Stöckel et al., 2017). These age-related behavioral differences are linked to structural differences in brain volume (Bernard & Seidler, 2013;Raz & Rodrigue, 2006), as well as differences in the large-scale organization of brain networks (Chan et al., 2014;Damoiseaux, 2017;King et al., 2018). Thus, understanding the factors that contribute to these brain-behavior relationships is important for advancing care and improving quality of life for the aging population.
Many aging investigations focus on task-based functional activation (Mirelman et al., 2017;Qin & Basak, 2020), though connectivity in the absence of a task is also informative for assessing differences in brain organization over the course of the adult lifespan (Ferreira & Busatto, 2013). The organization of functional brain networks is partially defined by network segregation, which represents greater within-network connectivity strength relative to between-network strength. Network segregation is thought to benefit specialized information processing and efficiency (Bullmore & Sporns, 2012;Wig, 2017) and is often evaluated in comparison to network integration, or dedifferentiation, which corresponds to greater connectivity between networks.
These measures of brain network organization are impacted by age. Young adults demonstrate multiple segregated functional networks with unique behavioral contributions (Power et al., 2011). However, network segregation is typically reduced in advanced age, resulting in increased integration/dedifferentiation of functional brain networks (Chan et al., 2014;Geerligs et al., 2015;Goh, 2011;Setton et al., 2022). Importantly, reduced network segregation, decreased modularity, and dedifferentiation are associated with worsened cognitive and motor performance (King et al., 2018;Kong et al., 2020). In fact, some work shows that network segregation mediates relationships between neurotransmitter systems and behavior in later life (Cassady et al., 2019). As such, the literature suggests that differences in the organization of functional brain networks across adulthood may be a key contributor to age-related behavioral declines.
Notably, aging females are more affected by behavioral and brain differences, compared to males. For example, females demonstrate faster declines in global cognition and greater deficits in balance than males (Levine et al., 2021;Wolfson et al., 1994). Relatedly, females incur higher risk for age-related diseases, such as Alzheimer's disease (Alzheimer's Association, 2021;Gao et al., 1998). These sex-specific impacts of age may be related to biological characteristics, such as sex hormone changes with menopause. Menopause is characterized by the cessation of ovarian function, which initiates a decrease in estrogen and progesterone levels. Estrogen and progesterone have been shown to benefit cognition and brain health (Duka et al., 2000;Hara et al., 2015;Jacobs & D'Esposito, 2011;Singh & Su, 2013), and brain circuitry has been associated with hormonal fluctuations across the menstrual cycle (Jacobs et al., 2017;Pritschet et al., 2020). Thus, the loss of neuroprotective hormones with menopause may contribute to disproportionate aging impacts on older females. However, research on network segregation with respect to sex differences and reproductive aging is lacking. The influence of menopause on the severity of functional declines in aging females is important to factor in when interrogating the origins of sex-specific differences in aging. Such insight would offer important new avenues through which age-related declines may be more effectively addressed. Given the increased incidence and severity of age-related diseases (e.g., Alzheimer's disease) in females, this knowledge may also promote efforts in the early detection and treatment of disease progression. To address these gaps, we investigated differences in resting-state network segregation between females and males, as well as between reproductive and postmenopausal females, across several functional brain networks. We also looked at associations between network segregation and age across sexes, replicating past research (Cassady et al., 2019;Chan et al., 2014), and within females and males separately.
In the interest of evaluating both cortical and subcortical network segregation, we included 10 cortical networks, as defined by Power et al. (2011), and one subcortical network, following Hausman et al. (2020). For the subcortical network, we included striatal seeds originally from Di Martino et al. (2008) and lobular cerebellar seeds that were created using the SUIT atlas (Diedrichsen, 2006;Diedrichsen et al., 2009), in lieu of using the undefined subcortical nodes/all-encompassing subcortical network from Power et al. (2011).
This approach allows us to investigate a more distinct and functionally-specific subcortical network in the context of aging (i.e., a cerebellar-basal ganglia network). In addition, we have previously shown differences in cerebellar-basal ganglia connectivity between young and older adults (Hausman et al., 2020), further motivating the inclusion of this particular network in the current study. Striatal seeds were localized to the left hemisphere while cerebellar seeds were placed in the right hemisphere, given the known lateralization of cerebellar networks with cortical regions. These subcortical regions have reported age differences in connectivity, wherein older adults primarily show reduced resting-state connectivity relative to young adults (Hausman et al., 2020), and are implicated in both motor and cognitive function (Bernard et al., 2017;Helie et al., 2013;King et al., 2019;Stoodley et al., 2012;Stoodley & Schmahmann, 2009). As such, subcortical networks may show differences in brain organization with advanced age, potentially contributing to age-related functional declines. To follow-up on prior work and further explore subcortical structures in the context of aging, we included this cerebellar-basal ganglia network as an additional point of comparison in our analyses.
The current investigation was designed to answer several overarching questions. Question 1: Can we replicate prior findings showing reduced segregation in cortical networks with increasing age? Given that several studies have shown lower functional network segregation in advanced age (Chan et al., 2014;Geerligs et al., 2015;Goh, 2011;Setton et al., 2022), we predicted that similar age-segregation associations would be present in the current sample. Question 2: Do we see the same age-segregation relationships with subcortical structures?
Considering the role of the cerebellum and basal ganglia in behaviors associated with age-related declines (Bernard et al., 2017;Helie et al., 2013;King et al., 2019;Stoodley et al., 2012;Stoodley & Schmahmann, 2009), we anticipated that reduced network segregation with increased age would also emerge within this subcortical network. Question 3: Do patterns of network segregation declines with age differ between females and males? As females generally experience heavier burdens with older age (Alzheimer's Association, 2021; Gao et al., 1998;Levine et al., 2021;Wolfson et al., 1994), we predicted that reduced network segregation would be more pronounced in females, relative to males. Question 4: Does reproductive stage play a role in potential sex differences with age-segregation relationships?
With the benefits of sex hormones in mind (Duka et al., 2000;Hara et al., 2015;Jacobs & D'Esposito, 2011;Singh & Su, 2013), we expected to see greater differences in network segregation between female reproductive stages, as related to hormone loss with menopause, compared to age-matched male controls. These questions are revisited when discussing the present findings.

| Study sample
Data was accessed through the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) repository (Shafto et al., 2014;Taylor et al., 2017). Data for this repository was gathered from a large sample of healthy adults, ranging from 18 to 88 years of age. We used raw structural and resting-state magnetic resonance imaging (MRI) data, along with demographic variables including sex, age, and menstrual cycle characteristics. We initially acquired data for 652 participants; however, 54 of those participants were excluded for being left-handed or for lacking handedness data. Handedness was assessed using the Edinburgh Handedness Inventory (Oldfield, 1971). This exclusionary criterion was applied to avoid the potential influence of brain organization differences between left-handed and right-handed individuals (Levy & Reid, 1978;Li et al., 2014). Additional individuals were excluded due to MRI data discrepancies, such as missing restingstate scans (n = 4), significant motion artifacts that could not be corrected during image preprocessing (n = 1), and lack of full restingstate volumes (n = 3). As a result, our initial sample consisted of 590 right-handed participants (297 females).

| Reproductive stage groupings
Our approach for categorizing females into reproductive, late perimenopausal, early postmenopausal, and late postmenopausal groups was replicated from previous work (Ballard et al., 2022). Here, a brief overview is provided. We used the Stages of Reproductive Aging Workshop Criteria (STRAW+10) to assign females to each reproductive stage group (Harlow et al., 2012). To distinguish between reproductive and late perimenopausal females, we used the reported length of menstrual cycles in days and number of days since last menstrual period. Females with 0-59 days for both variables were classified as reproductive, and females with 60-365 days were put in the late perimenopause group. Further, those within 1 year of their final menstrual period were also included in the late perimenopause group. To separate postmenopausal females into early and late groups, we used the number of years since final menstrual period. Females with 2-8 years since their final menstrual period were categorized as early postmenopausal, while those with 9+ years since their final menstrual period were assigned to the late postmenopause group.
Females lacking data for menstrual cycle characteristics were categorized by age cut-offs (n = 24): 18-39 for reproductive, 40-49 for late perimenopausal, 55-70 for early postmenopausal, and 71 or older for late postmenopausal. Females ages 50-54, lacking menstrual cycle data, were excluded from final analyses (n = 5) due to variability in reproductive stage for females in this age range (Kato et al., 1998;Morabia & Costanza, 1998;Palmer et al., 2003). To minimize external influences on hormone levels and examine impacts of natural menopause, we excluded females with an intrauterine device (IUD) (n = 12), possible use of continuous birth control (n = 2), and history of hysterectomy (n = 1). Notably, we only excluded females who indicated a hysterectomy that were less than 71 years of age, given that those over the age of 71 with a hysterectomy (n = 7) are likely in a comparable hormonal state to naturally menopausal females of a similar age. The resulting groups from this staging approach were corroborated with subjective responses from females regarding the occurrence of menopause. For further details on our grouping approach, please refer to Ballard et al. (2022).

| Age-matching
To help account for the intrinsic impact of age on reproductive stage, we formed age-matched male control groups to be used as an indirect reference for female groups. Each male was matched to a female using age, resulting in 1:1 age-matching, along with two variables of quality assurance where necessary: number of outlier scans and maximum motion. Females and males did not significantly differ in either quality assurance variable (ps ≥ 0.09). When presented with multiple males of the same age, we chose the male with the number of outlier scans most similar to that of the female in question. If males of the same age also contained identical counts for outlier scans, we chose the male whose maximum motion value was closest to that of the female. In cases where there were more females than males for a particular age, the same approach using number of outlier scans and maximum motion was used to choose female matches. Un-matched males and females were excluded from analyses (n = 156, 70 females); thus, our final sample consisted of 414 participants (207 females, ages 18-87, mean age 56.39 ± 18.80).
Our age-matching method helps account for the natural linkage between age and menopause by facilitating sex comparisons between groups of equal age makeups and sample sizes. In fact, there is notable age overlap between the resulting female groups, even slightly between reproductive (ages 18-55) and late postmenopausal (ages 54-87) females. Characteristics of age-matched groups are reported in Table 1. A graphical representation of this data is also available in Ballard et al. (2022), as groups were identical to those used in this prior work.

| Imaging analyses
A full overview of the study parameters and sample demographics for the Cam-CAN repository can be found in Taylor et al. (2017) and Shafto et al. (2014). For our analyses, we used raw T1 MPRAGE structural scans and raw resting-state EPI scans. The following parameters were used to collect resting-state data: 8 min and 30 s of acquisition using a 3T Siemens TimTrio, 3 Â 3 Â 4.4 mm voxel size, and repetition time (TR) of 1.97 s.
Image preprocessing and analyses were performed using the CONN toolbox, version 19b (Whitfield-Gabrieli & Nieto-Castanon, 2012). We used the default preprocessing pipeline, which consists of realignment and unwarping with motion correction, centering to (0, 0, and 0) coordinates, slice-timing correction, outlier detection using a 95th percentile threshold and the Artifact Rejection Toolbox (ART), segmentation of grey matter, white matter, and cerebrospinal fluid, normalization to MNI space, and spatial smoothing with a 5 mm full width at half-maximum (FWHM) Gaussian kernel. A band-pass filter of 0.008-0.09 Hz was applied to denoise data. The threshold for global-signal z-values was set at 3, while the motion correction threshold was set at 0.5 mm. After being de-spiked during denoising to adhere to the global mean, six-axis motion data and frame-wise outliers were included as first-level covariates.
Next, we looked at group differences between female reproductive stages, as well as age-matched male controls, to investigate relationships with age and effects of sex and reproductive stage on network segregation. Replicating the approach of Chan et al. (2014) and Cassady et al. (2019), network segregation values were determined using Equation (1) below.
Pearson's correlation coefficients were first transformed via Fisher's rto-z conversion (Zar, 1996). In Equation (1)  Note: Sample size and age, in years, for each female reproductive group and relative male control group, after apply age-matching exclusions. The numbers presented here correspond to both the female reproductive group and relative male control group, as 1:1 age-matching resulted in equal age makeups and sample sizes between sexes.
corresponding to 20 standard deviations above the original network mean, was removed. To fairly screen for true outliers in the cerebellar-basal ganglia network, this extreme outlier was removed before performing subsequent exclusions. Values above or below three standard deviations from the adjusted network mean, after removing the extreme outlier, were also excluded as outliers for the cerebellar-basal ganglia network. All statistical analyses were conducted in R programming software.

| Age correlations
Detailed results for network segregation and age correlations are reported in Table 2. When considering age and mean network segregation (computed across all networks), there is a significant correlation (p < .001) such that increased age is associated with lower overall network segregation across participants ( Figure 1;  (Figure 1; Table 2).
Correlations in females only reveal that older females exhibit lower segregation in almost all of the same networks as the whole sample: cingulo-opercular task control, default mode, dorsal attention, fronto-parietal task control, salience, and visual ( Figure 2; Table 3).
Further, females show a significant negative relationship between age and mean network segregation (p < .001). In contrast to the whole sample results, females do not demonstrate a significant correlation between age and segregation of the sensory somatomotor hand network (p = .124).
In age-matched male controls, we see broadly similar associations between age and network segregation, though some differences emerge as well (Figure 2; Table 3). Mirroring females, male controls demonstrate lower segregation with increased age across all networks as well as within the cingulo-opercular task control, fronto-parietal task control, salience, and visual networks. However, unique to males, a negative association between age and network segregation is also present for the sensory somatomotor hand network, though this correlation does not survive multiple comparisons correction (p = .077).
Interestingly, males do not demonstrate significant correlations between age and network segregation for the default mode and dorsal attention networks; thus, those relationships are unique to females. Overall, it seems that though some similarities exist between females and males, a few sex-specific differences in functional network segregation declines with age are present in the current sample.
To follow-up on these results and test potential interactions between sex and age, we then performed a series of ANCOVAs (Table 4). Each ANCOVA included sex and age, as well as the interaction between these variables, as factors with respect to network segregation. In brief, we found a main effect of age for the mean across all networks and for several individual networks after multiple comparisons corrections: cingulo-opercular task control, default mode, fronto-parietal task control, salience, and visual, which parallels the whole-sample results. A significant effect of age was also observed for the ventral attention network, though this effect did not survive multiple comparisons correction (p = .072). Importantly, main effects of sex and interactions between sex and age were not revealed in any case with this particular series of analyses, standing in contrast to our prior within-sex correlation results.

| Reproductive stage comparisons
An overview of reproductive stage and sex effects on network segregation is provided in here. In addition, interactions between reproductive stage and sex were not significant for any networks. As such, given the lack of sex effects or interactions between stage and sex, this series of analyses suggests that reproductive stage within females does not differentially impact functional network segregation beyond the impacts of age more generally. The consistent significance of main effects for reproductive stage across both sexes may be more directly attributed to age alone ( Figure 3). However, exploratory analyses with the reproductive and early postmenopause groups indicate that sex-specific differences in network segregation, with respect to reproductive stage, may be present during the mid-life transition to menopause (see Supplementary Table 2). This may suggest that the menopausal transition is particularly important for network dynamics, but in later life once hormones reach a more stable low state, sex differences are no longer present. However, we would note that this is exploratory and speculative.

| DISCUSSION
In the current study, we investigated age-related differences in functional network segregation in the context of sex and female reproductive stage. Following previous work (Cassady et al., 2019;Chan et al., 2014), we examined 10 cortical networks (Power et al., 2011) and also included an additional subcortical network (Di Martino et al., 2008;Diedrichsen, 2006;Diedrichsen et al., 2009;Hausman et al., 2020). We first replicated past work showing lower network segregation with higher age across participants, as well as within females and males separately, answering Question 1 and consistent with our predictions. Contrary to what we predicted for Question 2 however, we did not find significant age-segregation relationships with the subcortical network. Next, when evaluating relationships between age and network segregation within each sex separately, we found distinct patterns between females and males, along with some

| Resting-state network segregation and aging
When looking at associations between mean network segregation and age across all participants, we find a significant correlation in the negative direction, indicating lower overall network segregation with increased age. This parallels the overarching theme in the literature, wherein older age is associated with lower network segrega- age-related differences in network segregation, as observed here and in prior work, may contribute to functional declines in the aging population. This shift from network segregation to integration, or dedifferentiation, may represent a compensatory mechanism for the natural depreciation of brain function with age, in turn, impacting functional performance.
Relatedly, when investigating age-segregation relationships within individual networks (using the whole sample), we found that segregation was lower with higher age in 7 out of 11 functional networks after correcting for multiple comparisons with an FDR approach: cingulo-opercular task control, default mode, dorsal attention, frontoparietal task control, salience, sensory somatomotor hand, and visual. We did not observe significant correlations with age in 4 of the 11 networks after multiple comparisons correction: auditory, cerebellar-basal ganglia, sensory somatomotor mouth, and ventral attention. This indicates that, contrary to our expectations, subcortical network segregation, at least in the cerebellar-basal ganglia network, is not correlated with age. Notably, we investigated this network as a whole for the current investigation, though in our prior work, we did show some degree of functional dedifferentiation within the cerebellar-basal ganglia network in older adults. That is, motor nodes became less strongly associated with one another as did nodes associated with cognitive networks and structures (Hausman et al., 2020).
Here, in relation to other cortical networks, we did not see any agesegregation associations for this specific subcortical network. As such, we suggest that within network dynamics change and show evidence for dedifferentiation subcortically, but this is distinct from broader global dynamics with cortical networks.
In addition, this demonstrates that networks responsible for attentional filtering (ventral attention) (Vossel et al., 2014)

| Sex-specific differences in network segregation
When breaking down age and network segregation associations by sex, we find both similarities and differences in females and males.
T A B L E 3 Pearson's correlations between age and network segregation within each sex separately Both sexes exhibit lower segregation with older age in cinguloopercular task control, fronto-parietal task control, salience, and visual networks, though females also demonstrate negative correlations with age in the default mode and dorsal attention networks while these relationships were not seen in males. Thus, females may endure greater consequences with age in respect to the organization and efficiency of these particular functional networks, which may contribute to the disproportionate impact of normative behavioral declines and age-related disease on aging females, compared to males. Interestingly, the default mode network is strongly implicated in Alzheimer's disease pathology, which is more prevalent and severe in females (Alzheimer's Association, 2021; Greicius et al., 2004;Jones et al., 2011). Suppression of default mode network activity is associated with better performance on cognitive tasks (Anticevic et al., 2012). Consequently, the lack of default mode network segregation with age, specifically in females, may also reflect an inability to successfully inhibit default mode function during task-positive processing, in turn, contributing to age-related behavioral deficits.
On the other hand, males demonstrate unique age-related declines in segregation of the sensory somatomotor hand network, whereas females do not, though this correlation was only trending towards significance after correcting for multiple comparisons. Interestingly, in our prior work examining resting-state connectivity differences in cerebellar-whole brain networks between female reproductive stages and age-matched males, we found that male control groups exhibited greater differences in cerebellar-somatosensory connectivity compared to female reproductive groups (Ballard et al., 2022). Specifically, late postmenopausal male controls showed lower connectivity between cerebellar regions associated with cognition (Crus I/II) and regions of the somatosensory cortex, relative to both reproductive male controls and female counterparts. Though this prior work used the same sample from the current investigation, our results here align with these previous findings and illustrate that connectivity differences in somatosensory regions, specific to males, stand when using an alternative analysis approach. In sum, our findings highlight that sex-specific differences are important to consider when exploring relationships between age and the organization of functional brain networks.
When testing the statistical effects of sex and interactions between age and sex on network segregation in our ANCOVA series, we did not observe results that were indicative of sex-specific differences. Rather, we observed several instances of a main effect of age on network segregation, akin to our whole-sample results, though this age-segregation relationship was not significantly modified by sex for any network variable. Thus, these subsequent findings challenge the conclusions drawn from our qualitative comparisons of the correlations in males and females alone and suggest that variability in network segregation declines with age may not be statistically or directly associated with biological sex.
In our pre-versus post-menopause ANOVAs, reproductive stage influenced segregation for several networks while no effects of sex were observed. These findings may point to age effects, though impacts of sex hormone changes with menopause may also be at play and would be useful for better categorizing reproductive stage. Notably, in an exploratory analysis evaluating network segregation differences between reproductive and early postmenopausal groups, effects of sex and interactions between stage and sex begin to emerge in a few networks, though these effects are no longer statistically significant after correcting for multiple comparisons (Supplementary Table 2).
While these results are not conclusive, this indicates that the transition to menopause and initial declines in sex hormones may be important to evaluate in the context of sex differences in aging outcomes, however after the menopausal transition there are fewer sex differences. As such, it may be the case that the dynamics of hormonal change are important in midlife, though we would caution that this is highly speculative at this point. We would also note that, without access to direct hormone data we cannot accurately tease out these potential influences on the present results. Further work assessing direct effects of sex hormones is needed to fully understand the impact of menopause on the brain in aging females.
To this point, the current body of work on network segregation in older age does not include sex-specific analyses or comparisons between female reproductive stages (Cassady et al., 2019;Chan et al., 2014;Geerligs et al., 2015;Grady et al., 2016;King et al., 2018;Spreng et al., 2016). In fact, none of this work reports any analyses on network segregation differences between females and males. However, females endure more severe functional declines with age (Levine et al., 2021;Wolfson et al., 1994); thus, functional network segregation differences may contribute, at least in part, to the imbalance in aging trajectories between females and males. Results from the present work suggest that the effects of sex may be important and should be further interrogated when examining the organization of resting-state networks with respect to aging, and more work including the potential influence of sex steroid hormones is needed in the context of female reproductive aging. Future investigations should include sex-specific analyses and evaluate the effects of hormone changes with menopause when probing brain differences in older adulthood.

| Exploratory analyses
An ongoing topic of discussion in the neuroimaging literature is the use of strict thresholds for motion correction in older adults, who typically have greater motion in the scanner. Some work suggests that a conservative threshold of 0.5 mm is sufficient to control for motion artifacts in older populations (Adams et al., 2019;Cassady et al., 2019;Cassady et al., 2021;Reagh et al., 2018), while other work recommends a more stringent threshold in general (Ciric et al., 2018;Power et al., 2014).
However, recent work has also adopted liberal thresholds for motion correction in older adults (e.g., 0.9 mm) Riphagen et al., 2020), challenging these standards. Thus, the current stance of the literature on this particular parameter is unclear and more work is needed to fully understand the implications of different motion correction thresholds, with particular respect to motion in older populations.
In the current study, we adopted a conservative threshold of 0.5 mm for motion correction, given that our older subjects, perhaps surprisingly, moved less than the young adults (mean motion-age correlation: p < .001, r = À.44; maximum motion-age correlation: p < .001, r = À.25). Prior work (e.g., Power et al., 2012;Van Dijk et al., 2012) suggests that increased motion may result in lower measures of connectivity and influence group difference analyses. Here, motion was higher in the young participants, suggesting that the decreased segregation measures cannot be explained solely by higher possible motion in the older individuals in our sample. In addition, the averages for mean motion (0.095 mm) and maximum motion (0.489 mm) across our cohort of 414 subjects were relatively low. However, in the interest of thoroughly exploring the effects of motion correction thresholds on the outcomes tested here, we include a supplementary analysis that replicates the present analyses using a stringent threshold of 0.2 mm. These findings are presented within the Supplementary results. In brief, we find that many of the significant effects from the initial analysis with 0.5 mm were reduced or completely lost after applying the 0.2 mm threshold; thus, these modified results are largely different from what previous work reports with respect to age-segregation relationships (Cassady et al., 2019;Chan et al., 2014).
While this may indicate that a more stringent threshold tapers the risk of observing false positives, it could also suggest that the quality of data may be sacrificed as the information that remains after filtering with a more stringent approach may not be entirely representative of the true variability across subjects. Rather than offering premature speculations on these contrasting results, we present the new analysis in a supplement in order to contribute to ongoing conversations on the issue of motion artifacts, particularly in older adults.

| Limitations
Though this investigation contributes to current advances in aging research, there are limitations worth noting. First, we lacked access to direct hormone data or data regarding consecutive cycle lengths for our reproductive stage categorizations. As a result, reproductive stage was characterized using self-report menstrual information, and females undergoing hormone therapy or taking hormonal contraceptives may have been included in our sample. As the effects of menopause on functional network segregation may be more explicitly linked to hormone fluctuations, as opposed to broad reproductive stage differences, the lack of hormone data has limited our investigation. Moreover, given the lack of consecutive cycle data, our reproductive group may inherently include females in early perimenopause.
Second, we did not evaluate behavioral performance. Therefore, interpretation of functional relevance is purely theoretical. Third, though a network-level approach is informative for assessing differences in large-scale brain organization, our network segregation measure lacks the nuance of a purely ROI-based analysis that may yield added insight on a finer scale. Relatedly, the size (i.e., number of ROIs) per network varies and is, thereby, unequal across the functional networks included here, which may have impacted our comparisons.
Finally, given that menopause is a product of aging, we cannot discount impacts of age on the current findings. However, female reproductive groups overlap in age and age-matched male controls help to limit age impacts.

| CONCLUSION
The current study, using data from the CamCAN repository, offers new insight into sex-specific differences in the aging brain. Here, we evaluated the influence of sex and female reproductive stage on agerelated associations with functional network segregation. We provide preliminary evidence for distinct patterns of functional network segregation between females and males, through a qualitative lens, along with potential effects of reproductive stage, indicating that these biological factors may contribute to some degree to the differing aging trajectories between sexes. However, subsequent work is needed to determine the particular role of sex hormone fluctuations with menopause on brain differences within aging females. Such work is necessary to support findings from the present investigation and provide potential avenues through which age-related declines may be alleviated. Further, given sex differences in non-normative aging, elucidating relationships between menopause and the aging brain may also offer treatment alternatives for age-related diseases, such as Alzheimer's disease and other dementias.