Participants included patients with FXS at the Cincinnati Fragile X Research and Treatment Center seen for a clinical research visit between 2014 and 2024 as part of multiple past and ongoing studies of FXS (age: M = 21.8 years, range = 4.7 - 72.1 years; N = 123 males; N = 52 females). Participants were included if they had at least one visit during which they completed the Stanford-Binet, Fifth Edition and a blood draw. A subset of participants (N = 28) provided data at multiple independent visits. Repeated visits occurred either because (1) participants were enrolled in the FORWARD FXS natural history study which entails longitudinal follow-up visits, or (2) participants were enrolled in multiple studies (e.g., a participant completed the Stanford-Binet and a blood draw in 2020 as part of Study 1, then returned in 2022 to complete the Stanford-Binet and a blood draw as part of Study 2). There were 175 unique participants providing data across 206 unique visits. Participants providing data at multiple visits were not significantly different from those providing data only at one visit in terms of age, sex, race, ethnicity, FMRP, full-scale IQ, nor total COI scores (all p >.05).

All participants had a confirmed diagnosis of FXS, defined as having the full FMR1 mutation (>200 CGG repeats), confirmed via past testing results available in a participant’s medical record or via Southern Blot and/or PCR conducted in collaboration with the Molecular Diagnostic Laboratory at Rush University. Full demographic details are reported in Table 1 .

All studies were approved by the Cincinnati Children’s Hospital Medical Center Institutional Review Board. Parents, caregivers, or other legally authorized representatives provided written consent for participants younger than 18 years of age or for participants who were unable to provide consent due to limited informed decision-making capacity stemming from their intellectual disability. Participants provided their written consent when possible and otherwise provided verbal assent when possible.

Participants’ self-reported addresses at the time of clinical testing were extracted from our hospital’s electronic health record (EHR). Addresses were geocoded using the publicly available US Census Geocoder (https://geocoding.geo.census.gov/geocoder/ ) to extract 2010 census tract Federal Information Processing Standards (FIPS) codes. “Non-exact” and “tied” address matches were manually resolved and verified (e.g., “1234 Appletree Lane” may match to two US Census addresses: “1234 Apple Tree Lane” and “1234 Appletree Lane”). Remaining non-matches (~1.9% of addresses) occurred due to the home being recently constructed, demolished, or located in an extremely rural area. For these addresses, the nearest neighboring home address was used as input to the Census Geocoder as an approximation of the participant’s home address, which enabled systematic and accurate extraction of the census tract for the original address. This form of address interpolation is well-established for projects that rely only on coarse, tract-level matching as in the present study (30).

Census tracts for participant addresses were linked to the Child Opportunity Index 3.0 (COI) (25). The COI is a publicly available dataset providing standardized information about neighborhood characteristics related to health outcomes (i.e., social determinants of health). The COI 3.0 includes an overall composite score (“total neighborhood opportunity”), 3 domains (social and economic, health and environment, education), and 14 subdomains based on 44 component indicators ( Table 2 ). Component indicators are weighted based off the strength of their correlation with key socioeconomic and health outcomes. COI outcomes are rank-ordered percentiles ranging from 1 (least amount of opportunity) to 100 (most amount of opportunity). Because our data encompasses patients living across the United States, we used the nationally-ranked COI.

Linear mixed effects models were used for all analyses. Dependent outcomes were either IQ (FSIQ, VIQ, NVIQ) or adaptive behavior performance (Adaptive Behavior Composite and the Communication, Daily Living Skills, and Socialization subdomains). Independent variables included peripheral FMRP, chronological age (a covariate of no interest), and the COI ranking. Separate models were run for COI rankings of total neighborhood opportunity, the 3 COI domains, and the 14 COI subdomains. Models were run separately for males and females due to known sex differences in FMRP expression and previously reported sex differences in the association between environmental factors and clinical outcomes in FXS (15, 19). Participant/subject was the random effect in all models to account for participants who were tested at multiple visits. The variance accounted for by each fixed effect in our figures reflects the coefficient of determination (semi-partial R squared) calculated using methods from Nakagawa and Schielzeth (31) using the r2glmm R package (32). Unless otherwise noted in the text, the direction of all associations was positive (e.g., more positive social-environmental exposures related to greater IQ). Additional tertiary models (associations between FMRP and IQ, FMRP and adaptive behaviors, FMRP and COI) are presented in the Supplementary Materials document. Due to the hypothesis-generating and exploratory nature of this study as well as the limited work in this area, we did not correct for multiple comparisons.

All statistical analyses were conducted using R version 4.3.1 (33). Linear mixed effects models were constructed using the lme4 R package (34). All data was visualized using the ggplot2 R package (35).

Participants came from 23 different states and 90 different counties across the United States, reflecting the national makeup of patients seen in our clinic.

Overall neighborhood opportunity as measured by the COI widely varied among the full sample, but skewed towards higher opportunity areas (M = 66; SD = 26; Range: 3-99). The distribution of neighborhood opportunities was similar across the social and economic (M = 64; SD = 27; Range: 10-99), health and environment (M = 63; SD = 24; Range: 9-100), and education (M = 68; SD = 26; Range: 1-100) COI domains ( Figure 1 ). IQ and adaptive behavior sample distributions are depicted in Figure 2 . Males and females did not differ on the COI (p’s >.89).

Building on previous ID research (15, 37, 38), we show that several social-environmental factors as measured by the COI (e.g., social resources, access to healthcare, neighborhood walkability) are positively related to IQ in males, but not females, with FXS. For example, these factors explained about 3 to 5% of the unique variance in full-scale IQ among males, similar to the 3-5% of the variance in full-scale IQ accounted for by FMRP in this same sample. A largely distinct set of economic and educational factors were associated with adaptive behaviors; again, these associations were only present in our male participants.

These sex-specific findings echo two previous studies in FXS showing that environmental factors are more strongly related to IQ and adaptive behaviors in males compared to females with FXS (15, 19). Although females with FXS also present with clinical concerns such as anxiety and difficulties with conversational pragmatics (39), cognitive and adaptive skills are typically less severely impacted in females with FXS due to their retaining a second unaffected allele and random X-chromosome inactivation (3, 4). These sex differences in cognitive and adaptive outcomes are illustrated for our sample in Figure 2 .

We and others have hypothesized that sex-specific environment-behavior associations in FXS are driven by an inherently greater need for environmental accommodations and behavioral supports for more clinically affected males (16, 19). This pattern is reflected in FXS studies finding discrepancies in service utilization between males and females (40), greater median healthcare expenditures in males (41) and more frequent specialty Fragile X clinic attendance in patients with greater symptom severity (42). Although we have not demonstrated causality, this suggests that IQ and adaptive behaviors may be especially malleable in males with FXS when supported by environmental resources and interventions (e.g., early childhood intervention programs, adequate insurance coverage). Longitudinal research will help provide further evidence for this hypothesis and clarify the most salient social-environmental factors to target.

Statistically, this sex-specific pattern also may reflect the relatively greater proportion of variance in clinical outcomes accounted for by FMRP in females compared to males, leaving less unique variance that may be accounted for by social-environmental actors in females. About 38% of the males in our sample had no detectable peripheral FMRP (FMRP = 0), limiting the magnitude of observed associations between FMRP and clinical outcomes in males (i.e., leaving more variance to be accounted for by the COI). We believe analyses of females with FXS and ID (i.e., IQ < 70) and more males with partial FMRP expression will clarify if these sex-specific associations are underpinned by differences in clinical severity, differences in FMRP expression, or both. The relatively small number of females with IQ < 70 in the present study (N = 15) precludes our ability to conduct an adequately powered analysis, but future large-scale studies of FXS should consider such an approach to clarify this point. Statistically, sex-specific associations also may reflect the larger number of male participants subsequently driving increased statistical power compared to the smaller group of females for whom the risk of Type II errors is increased.

The SDOH most consistently associated with nearly all outcomes was social-economic resources. The COI uses three indicators in the calculation of its social resources subdomain: mobility-enhancing friendship networks (i.e., mixed low- and high-income friendships), multi-parent families, and proximity of non-profit organizations within a given census tract. The first two metrics are weighted the most heavily when calculating COI social resources rankings (weighted 0.90 out of 1.0) and may explain why this association is so robust across outcomes. We hypothesize this pattern is largely due to the positive effects of increased caregiver support and reduced parental stress in families embedded within mobility-enhancing social networks and multi-caregiver households. Mobility-enhancing friendships enable transmission of knowledge and sharing of resources, provide emotional support, and reduce parental stress, leading to a range of positive health and developmental outcomes (43, 44). Similarly, parental stress is lower in married parents relative to single parents, likely due to a combination of enhanced social supports (45) and increased household income among married parents (46). Taken together, both forms of caregiver support can reduce family stress.

The Family Stress Model provides a theoretical framework for the impact of this stress on downstream child development (47). Briefly, family stress, through its direct effects on caregiver-child interactions and the home environment, has detrimental effects on cognitive development, including in patients with FXS (16). This effect has been implicated in correlational studies of FXS: greater parental stress is associated with lower fluid and crystallized intelligence (17) and a fewer number of parent-child interactions (20) critical for child development. Follow-up studies of social cohesion will help further specify the mechanisms through which social supports drive greater child IQ and adaptive skills in families of people with FXS. We are especially interested in studies soliciting the lived experiences and input of family members on unique aspects of social support in FXS, such as the availability of respite care, that reduce family stress and support the development of their children.

Most neighborhood economic factors, including economic equity (e.g., reduced income disparity), employment opportunities, economic resources (e.g., household income, public assistance), and wealth (i.e., capital assets), were associated with stronger adaptive behaviors in males with FXS. Consistent with our hypotheses, these factors were most strongly associated with adaptive communication skills. This is consistent with research in the general population demonstrating that neighborhood-level opportunities are most strongly associated with language outcomes (13). These findings also are in keeping with research in other IDs, including 22q11.2 deletion syndrome (26) and Down syndrome (27), which have found stronger associations between family socioeconomic status (e.g., proxy measures like parental education) and VIQ compared to NVIQ. Although we did not observe the same pattern for VIQ compared to NVIQ, this may reflect the increased susceptibility of adaptive communication skills, which emphasize the application of pragmatic verbal skills across settings (e.g., for reading, writing, telling stories, receiving and giving verbal instructions, etc.) (48), to environmental differences.

We hypothesize that these economic factors so strongly relate to communication skills due to the mediating roles of parental responsivity and home enrichment. Several studies of the general population demonstrate that economic stability enables greater parental responsiveness (49, 50). Responsive parent-child interactions, in turn, are longitudinally associated with greater lexical diversity (21) and expressive and receptive language in FXS (18) and the general population (51). This is likely a bidirectional and cyclical relationship, as children with stronger language elicit more responsive parental interactions (52). Importantly, parental responsiveness is a modifiable factor in parents of youth with IDs (for review, see (60)), underscoring the utility of a multipronged intervention approach targeting both distal economic factors and proximal family factors to improve adaptive communication skills in FXS.

Economic resources also likely enable greater home enrichment that supports language development in FXS, as demonstrated in the general population (53). Consistent with this, home enrichment is associated with adaptive communication skills in FXS (19). Home enrichment encompasses materials and activities that stimulate intellectual growth, such as literacy materials, cognitively-stimulating games and toys, and novel experiences (e.g., trips to parks, museums, zoos, playgrounds, etc.) (54, 55). Enrichment can be manipulated in animal laboratory environments (56), making it particularly beneficial for the FXS field which has a strong set of animal models and suggesting the neurobiological mechanisms through which enrichment supports development in FXS can be directly interrogated. Like parental responsiveness, home enrichment is malleable and may serve as a key intervention target when working with families affected by FXS (53, 57).

Several limitations of our study highlight directions for future research. First, we quantified peripheral FMRP as our measure of FMRP expression for both males and females with FXS. X-inactivation ratios provide one other method of characterizing molecular heterogeneity in female patients and may help account for additional phenotypic variance in females with FXS. Second, our findings are drawn from cross-sectional data, limiting insights into causality. Longitudinal work enabled by FXS natural history studies (i.e., FORWARD and FORWARD-MARCH) will help clarify causal pathways through which social-environmental factors impact development in FXS. Third, and relatedly, although the broad age range of our sample allowed us to study more individuals and maximize statistical power, this also limits the utility of some COI metrics such as the education subdomains (i.e., adults with FXS who no longer benefit from strong public education opportunities). Our use of concurrent addresses also deemphasizes the accumulative effect that limited access to social-environmental opportunities can have across the lifespan, beginning prenatally (58). However, participants in our study who moved between their visits typically remained in neighborhoods with similar levels of opportunity, suggesting that concurrent neighborhood resources may be similar to resources to which participants had access across their life. Retrospective studies of early childhood social-environmental exposures will clarify this accumulative effect. Fourth, consistent with national diagnostic and service access trends in FXS (42, 59), our sample was predominantly comprised of white, non-Hispanic patients. The neighborhoods for our sample also skewed towards higher opportunity areas ( Figure 1 ). We have recently begun targeted efforts (e.g., in-home research visits) to minimize barriers to research participation, but this remains an ongoing need for the ID field more broadly. It is not known to what extent these findings generalize to a more racially and ethnically diverse sample or a sample more diverse in terms of neighborhood opportunity. Fifth, we have not yet incorporated other explanations of phenotypic variance into our study, including family genetics (e.g., as accounted for by a proxy like parental IQ) and epigenetic factors. These remain important areas for future study to comprehensively address factors driving individual differences in FXS. Sixth, we substituted ABIQ for FSIQ for the small number of participants who only completed the routing version of the SB-5. While we believe this is justified given findings demonstrating relatively strong ABIQ-FSIQ agreement when routing subtest scatter is minimal as in the present sample (28), this may introduce additional heterogeneity into our sample. Last, our interpretation of the COI assumes that a neighborhood’s level of resources (e.g., adequate insurance coverage) translates to an individual family (e.g., that family has adequate insurance coverage). The COI may minimize the role of individual experiences, and follow-up studies probing individual families’ social-environmental exposures will be necessary to translate these findings into actionable changes.