DISADVANTAGE AND DISORDERED EATING: EXAMINING PHENOTYPIC AND GENOTYPE X ENVIRONMENT ASSOCIATIONS ACROSS DEVELOPMENT By Megan Elizabeth Mikhail A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Psychology – Doctor of Philosophy 2024 ABSTRACT Background: Socioeconomic disadvantage may be a significant risk factor for disordered eating, particularly for individuals with underlying genetic risk. However, little-to-nothing is known about the impact of disadvantage on disordered eating in boys during the critical developmental risk period. Crucially, risk models developed for girls may not necessarily apply to boys, as boys show different developmental patterns of disordered eating risk (i.e., earlier activation of genetic influences during adrenarche, an early stage of puberty). This is the first study to examine phenotypic and genotype x environment (GxE) effects of disadvantage in boys. Methods: Analyses examined 3,484 male twins ages 8-17 (Mage = 12.27, SD = 2.96) from the Michigan State University Twin Registry. Disordered eating (e.g., body dissatisfaction, binge eating) was measured using the parent-report Michigan Twins Project Eating Disorder Survey. Neighborhood disadvantage was measured using a census-tract level Area Deprivation Index, and family socioeconomic status was determined from parental income and education. Adrenarche status was determined using multiple indicators, including age and Pubertal Development Scale scores. Results: GxE models suggested that genetic influences on disordered eating were activated earlier for boys experiencing familial or neighborhood disadvantage, with substantial genetic influences in early adrenarche, when genetic influences were low in more advantaged boys. Phenotypically, both neighborhood and familial disadvantage were associated with greater disordered eating for boys in late adrenarche, which could indicate a lasting impact of earlier activation of genetic influences on later risk. Conclusions: Results highlight disadvantage as a novel risk factor for disordered eating in boys, particularly those with genetic vulnerabilities. ACKNOWLEDGEMENTS This dissertation, and my graduate studies more broadly, would have been impossible without the support and guidance of a wide network of friends, colleagues, and mentors. My doctoral guidance committee, Dr. Kelly Klump, Dr. Alexandra Burt, Dr. Jason Moser, and Dr. Alexa Veenema, have gone above and beyond their formal responsibilities to help me grow as a scientist and a person, giving me invaluable advice when I have needed it and challenging me to ask novel and difficult questions and reach higher to realize my potential. I am particularly grateful to Dr. Klump, my committee chair and advisor, who has been my scientific role model, teacher, champion, guide, and friend all wrapped into one. If I grow to be half the scientist and mentor that she is, I will be extraordinarily pleased (and she will deserve ample credit). I am also deeply appreciative of the guidance Dr. Burt has given me throughout my graduate career on everything from grant writing to twin modeling. I am thankful for the support of Dr. Ann Kring, who welcomed me into her research group as a post-bac and gave me the opportunity to run my very first eating disorder study within her psychosis lab. I am also grateful to Dr. Cheryl Sisk, who served on my master’s thesis committee and has provided invaluable guidance on the intricacies of hormone effects across development, Dr. Amy Nuttall for her skillful and patient instruction in advanced statistical analyses, Dr. Blair Burnette, who has quickly become an exemplar for me with respect to conducting creative and thoughtful eating disorder research through a social justice lens, and the entire clinical science area faculty, who have supported and challenged me in just the right balance throughout graduate school. I have also benefited immensely from the guidance of my clinical supervisors, whose insights have helped me grow as a researcher as well as a clinician. I appreciate the hard work and kind support of all Klump lab and Michigan State University Twin Registry staff, whose contributions helped make this iii research possible, as well as the time and effort of study participants. I am also deeply grateful to my family, including my dad, Mike Mikhail, mom, Sheila Mikhail, “stepmothery person” Carrie Di Santo, and our beloved furry adoptees (Ralphie, Luna, and Sadie). I have been extremely fortunate to be able to rely on the support of my graduate school cohortmates, particularly Sarah Carroll, who has celebrated with me in good times, commiserated in bad, and offered excellent twin model fitting advice. Equally important have been the many friends outside of school who have served as bright points in my life and always been there for me when I have needed them, especially (but certainly not limited to) Lily Xu, Amy Blake-Iranon, Justin Bui, Naite Boham, Ian Lim, Shawn Lee, Emily Kohn, Carol Olausen, and Michael Speers. One of the great and unexpected delights of graduate school has been discovering my music community, especially Kaitlyn Riel, Michael Pierce, Hattie Danby, and everyone who has played with me in the School of Rock adult band, who have given me many of my most joyful moments over the past several years when I have felt stuck working through challenges in my research. I am also immeasurably grateful to Judy Banker, who has been a tremendous source of support and nurtured my growth in all ways, and Sarah Adler, who gave me the courage and confidence to believe I could become a clinical psychologist in the first place. Finally, I would like to thank my funding sources throughout graduate school, including the National Institute of Mental Health, National Science Foundation, and Michigan State University Graduate School. iv TABLE OF CONTENTS INTRODUCTION ............................................................................................................................. 1 METHODS ........................................................................................................................................ 6 RESULTS ........................................................................................................................................ 15 DISCUSSION .................................................................................................................................. 18 TABLES ........................................................................................................................................... 22 FIGURES ......................................................................................................................................... 27 BIBLIOGRAPHY ............................................................................................................................ 29 APPENDIX A: SUPPLMENTAL MATERIAL ............................................................................. 36 v INTRODUCTION Despite historical stereotypes that eating disorders (EDs) primarily impact individuals from relatively advantaged backgrounds (Gard & Freeman, 1996), recent research suggests risk for EDs and related symptoms may be elevated among people experiencing socioeconomic disadvantage. While relatively few studies have examined the association between disadvantage and disordered eating, increased disordered eating among disadvantaged populations has been found in both girls and adults across multiple indicators of disadvantage, including food insecurity, neighborhood disadvantage (i.e., increased neighborhood poverty and decreased community resources), and familial disadvantage (i.e., lower household income and educational attainment) (Becker et al., 2017, 2019; Coffino et al., 2020; Hazzard et al., 2021; Lydecker et al., 2019; Mikhail, Carroll, et al., 2021). Though people from disadvantaged backgrounds are underrepresented in research and treatment settings, this disparity appears to reflect reduced access to care rather than the prevalence of EDs in the general population (Gard & Freeman, 1996; Huryk et al., 2021; Sonneville & Lipson, 2018). Preliminary studies linking disadvantage to disordered eating suggest an urgent need for additional research examining disordered eating in socioeconomically disadvantaged populations, including how the etiology of disordered eating may be similar or different for people from disadvantaged backgrounds. There are several mechanisms through which disadvantage may increase disordered eating, including increased stress (DeCarlo Santiago et al., 2011; Goodman et al., 2005), reduced access to fresh foods such as fruits/vegetables and increased availability of highly palatable foods (e.g., fast food; Cooksey-Stowers et al., 2017; Dubowitz et al., 2012), and increased weight stigma among disadvantaged populations (Becker et al., 2021). The impact of these environmental risk factors may be further amplified in individuals with underlying genetic risk via 1 genotype x environment interactions (GxE). When GxE is present, the impact of latent genetic risk on a behavioral phenotype depends on the presence of environmental stressors. In some cases, genetic influences may be weaker in stressful circumstances that impede normative development (i.e., bioecological GxE; Bronfenbrenner & Ceci, 1994; Burt, 2014). Alternatively, and more commonly for internalizing phenotypes such as disordered eating (e.g., Fairweather- Schmidt & Wade, 2017; Strachan et al., 2017), stressful environmental circumstances amplify underlying genetic vulnerabilities, leading to elevated psychopathology in individuals with genetic risk (i.e., diathesis-stress GxE; Rende & Plomin, 1992). Initial research suggests the impact of disadvantage on disordered eating may be amplified for individuals with underlying genetic vulnerabilities through diathesis-stress GxE, particularly during puberty, a developmentally sensitive risk period for the emergence of EDs (e.g., Mikhail, Anaya, et al., 2021; Nagl et al., 2016). In a recent study, our group found that phenotypic ED symptoms were greater for girls experiencing familial or neighborhood disadvantage. In addition, both forms of disadvantage were associated with stronger and earlier expression of genetic influences on disordered eating (Mikhail, Carroll, et al., 2021). Though disordered eating is strongly heritable in adulthood (with ~50% of variance in disordered eating due to genetic factors), girls from more advantaged backgrounds typically show minimal genetic influences on disordered eating prior to mid-puberty (Klump et al., 2003, 2007, 2012; O’Connor et al., 2020). However, genetic influences on disordered eating were already substantial in girls from the most disadvantaged backgrounds in pre/early puberty, suggesting much earlier expression of genetic risk in disadvantaged contexts that could ultimately lead to more disordered eating (Mikhail, Carroll, et al., 2021). The considerable stress accompanying disadvantage may exacerbate genetically-based individual differences in the stress response or emotional reactivity (Gillespie et 2 al., 2009), potentiating earlier expression of genetic risk for disordered eating. It is notable that effects were largely consistent across neighborhood and familial disadvantage, which are conceptually and empirically distinct (r’s ~ .3 to .5; Hackman et al., 2012; Mikhail, Carroll, et al., 2021; Roubinov et al., 2018), suggesting that multiple forms of disadvantage (both more proximal and distal) are associated with increased ED risk in girls. Importantly, research to date has focused on the impact of disadvantage on disordered eating in girls (e.g., Mikhail, Carroll, et al., 2021) or adults (e.g., Becker et al., 2017, 2019; Hazzard et al., 2021; Lydecker et al., 2019), with no studies of disadvantage effects in boys during the critical developmental risk period. While disordered eating is less common in boys than girls, a significant number of boys and men do experience EDs and related symptoms (e.g., binge eating), with recent estimates indicating that over 10% of adolescent boys experience clinically significant disordered eating (Nagata et al., 2020). Disordered eating may be even more common among boys and men experiencing significant stress (Gadalla, 2009; Mitchell et al., 2016), potentially including those living in disadvantaged environments, and preliminary research suggests that food insecurity (Becker et al., 2017, 2019) and lower SES (Burke et al., 2022) are similarly associated with disordered eating in adult men and women. Notably, boys and men are less likely than girls and women to be diagnosed or receive treatment for EDs even when experiencing significant symptoms (Sonneville & Lipson, 2018). It is therefore critical to identify boys at increased risk for targeted prevention and intervention. Crucially, boys experience different developmental patterns of ED risk than girls, and developmentally sensitive risk models based on girls (including analyses in Mikhail, Carroll, et al. (2021) discussed above) may not necessarily apply to boys. Specifically, the developmental timing of activation of genetic influences on ED risk differs across sex. Puberty can be divided 3 into two developmental stages: adrenarche, during which adrenal androgens (e.g., androstenedione, dehydroepiandrosterone, dehydroepiandrosterone-sulphate [DHEA-S]) increase prior to pronounced outward physical changes, and gonadarche, during which increases in gonadal hormones (e.g., estradiol, testosterone) drive the development of secondary sex characteristics (e.g., breast growth, voice changes) (Auchus & Rainey, 2003). Adrenarche typically begins before gonadarche (~age 6-8) and continues through gonadarcheal development (Guran et al., 2015). Girls do not show genetic influences on disordered eating until mid- gonadarche, well after adrenarche is underway (Klump et al., 2003, 2007, 2012; O’Connor et al., 2020). However, in boys, genetic influences start to increase during the early stages of adrenarche that precede gonadarche and are fully online when gonadarche begins (Culbert et al, 2017). Genetic influences on disordered eating may be activated in males but not females during adrenarche because males display greater sensitivity to androgens following greater exposure to testosterone prenatally, leading to unique impacts of androgens on later gene expression in males (Arnold, 2009). If disadvantage impacts disordered eating in part by leading to earlier expression of genetic risk, these developmentally sensitive effects would be expected to unfold earlier in boys than girls (i.e., in adrenarche rather than gonadarche) and could reflect potentially distinct underlying molecular mechanisms (i.e., activation by androgens rather than estrogen). It is therefore crucial to examine boys independently rather than assuming that disadvantage effects during adolescence are the same in girls and boys. In this study, we examined whether boys living in more socioeconomically disadvantaged circumstances were at elevated risk for disordered eating. We examined both family SES and neighborhood disadvantage to investigate potential similarities and differences in the impact of disadvantage at different levels of proximity. Notably, prior research suggests that activation of 4 genetic influences during adrenarche/puberty may lead to lasting changes in neural organization that precede behavioral changes (Klump et al., 2018; Schulz & Sisk, 2016). If disadvantage impacts disordered eating in part through changes in gene expression that alter brain organization during adrenarche, we might expect significant GxE (i.e., elevated genetic influences on disordered eating with increasing disadvantage) in early adrenarche, but minimal phenotypic effects until late adrenarche. Conversely, we would expect smaller GxE effects (i.e., similar levels of genetic influence across disadvantage) during late adrenarche after the period of organization has ended, but greater phenotypic effects. Moderation analyses across adrenarche allowed us to examine these hypotheses regarding developmental shifts in disadvantage effects. 5 Participants METHODS Primary analyses included 3,484 boys ages 8-17 (Mage = 12.27, SD = 2.96) from same-sex twin pairs from the Michigan Twins Project (MTP), a large-scale twin registry that serves as a recruitment pool for research conducted through the Michigan State University Twin Registry (MSUTR). The MSUTR is a population-based twin registry that recruits twins through birth records in collaboration with the Michigan Department of Health and Human Services (see Burt & Klump, 2013, 2019; Klump & Burt, 2006). Response rates for the MTP are similar or better than those of other twin registries (58.9% for youth under 18) and MTP twins are demographically representative of Michigan (Burt & Klump, 2019). Approximately 14% of MTP youth live in families whose income is at or below the federal poverty level (~$26,500 for a family of four; US Department of Health and Human Services, 2021), which is similar to the overall population of Michigan (Burt & Klump, 2013). Most participants identified as white/non-Latinx (n = 2,948; 84.6%), followed by Black/non-Latinx (n = 248; 7.1%), multiracial (n = 124; 3.6%), Latinx (n = 48; 1.4%), Asian American (n = 38; 1.1%), and Native American (n = 10; 0.3%). The remaining participants (n = 68; 2.0%) identified as belonging to another race/ethnicity or did not specify their race/ethnicity. Twins varied widely in family SES (combined parental income M = $90,390, SD = $54,410, range = $0-$300,000+). Similar to our prior report examining girls from the MSUTR (Mikhail, Carroll, et al., 2021), 10.9% of participants lived in neighborhoods above the national 75th percentile for disadvantage. Additional demographic information is shown in Table S1.1. 6 Measures Zygosity Determination Zygosity was determined using a well-validated physical similarity questionnaire (Lykken et al., 1990) completed by the twins’ parents. This questionnaire is over 95% accurate in determining zygosity as verified through DNA/serologic testing (Lykken et al., 1990; Peeters et al., 1998). Disordered Eating Disordered eating was assessed using the Michigan Twins Project Eating Disorder Survey (MTP-ED; Mikhail, Carroll, et al., 2021), a nine-item parent-report questionnaire for measuring disordered eating in population-based samples. Prior research suggests parent-reported symptoms differentiate youth with and without clinical EDs (Accurso & Waller, 2021) and show similar or greater concordance with objective external measurements (e.g., BMI, clinician-reported symptoms) as adolescent-reported symptoms (Couturier et al., 2007; Steinberg et al., 2004; Swanson et al., 2014). Parent report may be particularly useful for younger boys who may have difficulty understanding disordered eating items. The MTP-ED contains questions regarding body dissatisfaction (i.e., distress regarding body shape), weight preoccupation (i.e., fear of gaining weight), and disordered eating behaviors (i.e., dieting, binge eating, purging). Each item is rated on a 3-point scale from 0 (not true) to 2 (certainly true). Detailed information on the reliability/validity of the MTP-ED in boys is included in Supplemental Material. In brief, in the current sample, the MTP-ED had acceptable internal consistency across age (ages 8-12: α = .77; ages 13-17; α = .81) and pubertal development (early adrenarche: α = .70; early gonadarche: α = .78; mid/late gonadarche: α = .80), discriminated between boys with and without a parent-reported ED (d = 1.24, p <.001), and showed expected 7 correlations with other constructs (e.g., r = .29, p <.001 with BMI; r = .25, p <.001 with internalizing symptoms). Additional validation of the MTP-ED was conducted in 299 boys ages 7-18 and their primary caregivers from a separate, ongoing study within the MSUTR. Correlations in this independent sample were large between self-reported MTP-ED and self-reported Minnesota Eating Behavior Survey 1 (MEBS; von Ranson et al., 2005) total scores (r = .66, p <.001). As is typical in the ED literature, correlations between parent- and self-reported MTP-ED were significant but small-to-moderate in magnitude (r = .26, p <.001). Disadvantage Neighborhood disadvantage was measured using a well-validated (Kind & Buckingham, 2013; Singh, 2003), census-tract level Area Deprivation Index (ADI) incorporating 17 indicators of neighborhood disadvantage (e.g., unemployment rate, median home value). The ADI has been used to examine associations between neighborhood disadvantage and numerous mental and physical health outcomes in prior work (Burt et al., 2020; Carroll et al., 2021; Kind et al., 2014; Powell et al., 2020; Suarez et al., 2022), including our previous report on disadvantage and disordered eating in girls (Mikhail, Carroll, et al., 2021). Neighborhood disadvantage, as measured by the ADI, is correlated with poorer physical (Kind et al., 2014; Powell et al., 2020) and mental (Carroll et al., 2021; Burt et al., 2020) health, as well as higher BMI (Sheets et al., 2020) and lower physical activity (Miller et al., 2020). The ADI also has excellent internal consistency (α = .95 in past research; Singh, 2003). The ADI score for each family was coded 1 The Minnesota Eating Behavior Survey (previously known as the Minnesota Eating Disorder Inventory [M-EDI]) was adapted and reproduced by special permission of Psychological Assessment Resources, 16204 North Florida Avenue, Lutz, Florida 33549, from the Eating Disorder Inventory (collectively, EDI and EDI-2) by Garner, Olmstead, Polivy, Copyright 1983 by Psychological Assessment Resources. Further reproduction is prohibited without prior permission from Psychological Assessment Resources. 8 using publicly available data from the American Community Survey for the census-tract containing the family’s address (https://www.neighborhoodatlas.medicine.wisc.edu/). Raw ADI scores were converted into percentiles relative to other families in the sample, with higher scores indicating greater neighborhood disadvantage. Family SES was measured using a latent variable factor score incorporating mother’s education level, father’s education level, and combined parental income. As with the ADI, raw scores were converted into percentiles relative to other families in the sample. To maintain consistency with Mikhail, Carroll, et al. (2021), family SES was coded such that lower scores (i.e., lower family SES) indicate greater disadvantage. Importantly, neighborhood disadvantage and family SES are only moderately correlated (r = -.47 in the current study; only 22% variance shared), indicating that they are related but distinct (Burt, 2014; Hackman et al., 2012; Roubinov et al., 2018). Adrenarche Because adrenal androgens were not directly measured, age and gonadarche were used as proxy indicators of adrenarche status based on earlier research on changes in the etiology of disordered eating across adrenarche and gonadarche (Culbert et al., 2017). Gonadarche was measured using the parent-report Pubertal Development Scale (PDS; Peterson et al., 1988), a five- item questionnaire that assesses physical markers of maturation during gonadarche. Parent-rated PDS correlates strongly with professionally rated Tanner staging and shows good psychometric properties for boys (α = .96; Koopman-Verhoeff et al., 2020). Items for boys include height changes, skin changes, body hair growth, voice deepening, and facial hair growth. Each item is rated from 1 (not yet begun) to 4 (seems completed). As in past research (Klump et al., 2003, 2012), the five items were averaged to create an overall PDS score. 9 We divided participants into early and late adrenarche groups based on research indicating that genetic influences on disordered eating begin to gradually increase during the period of adrenarche preceding gonadarche (i.e., early adrenarche) and are fully online when gonadarche begins (i.e., late adrenarche) (Culbert et al., 2017). In other words, the period of adrenarche preceding gonadarche onset is critical for activation of genetic influences on disordered eating in boys. Developmental studies indicate most boys begin adrenarche based on adrenal androgen levels by age 8 (i.e., the youngest age in our sample) (Guran et al., 2015; Ilondo et al., 1982). Therefore, we categorized all participants aged 12 or younger with a PDS score of 1 (i.e., no external indicators of gonadarche) as in early adrenarche (n = 495; 14.2%). Participants with a PDS score greater >1 (n = 2,723; 78.2%) or who were 13 or older and missing data on the PDS (n = 118; 3.4%) were categorized as in late adrenarche. We used a cutoff age of 13 as a proxy indicator of being in late adrenarche based on prior research indicating that over 95% of boys show evidence of gonadarcheal development (e.g., increase in testicular volume) by age 13 (Bundak et al., 2007). A small number of boys who were 13 but had a PDS score of 1 (n = 6; 0.2%) were also categorized as being in late adrenarche, which was a conservative decision in relation to our hypotheses (i.e., the difference between boys in early and late adrenarche would be reduced if these boys were in fact in early adrenarche). Adrenarche status for the remaining 142 participants (4.1%) could not be determined because they were under age 13 and missing data on the PDS. BMI Percentile Age- and sex-specific BMI percentiles were calculated from parent-reported height and weight using CDC growth charts (https://www.cdc.gov/healthyweight/xls/bmi-group-calculator- us-062018-508.xlsm). Parent-reported BMI shows good concordance with measured BMI in 10 youth, with parent-reported weight estimates deviating from measured weights by <5 pounds (Gordon & Mellor, 2015; Shields et al., 2011). Statistical Analyses Data Preparation MTP-ED scores were prorated if one item was missing and marked as missing if >1 item was missing. While parent-reported BMI shows good concordance with objective measures (Gordon & Mellor, 2015; Shields et al., 2011), following Mikhail, Carroll, et al. (2021), we took a conservative approach in setting extreme BMI values <0.5th percentile or >99.5th percentile to missing. MTP-ED scores were log transformed to account for positive skew and standardized. More disadvantaged youth tend to have higher BMIs (Alvarado, 2016), and higher BMIs are associated with disordered eating (Neumark-Sztainer et al., 2007). All phenotypic and GxE analyses were therefore conducted with and without BMI percentile to directly assess its impact on results. Phenotypic Analyses Multilevel models (MLMs) with a random intercept to account for nesting of twins within families were used to examine phenotypic associations between disadvantage and disordered eating. Random slopes were not estimated due to the small number of observations per group (i.e., two twins per family). Models used an identity covariance structure and maximum likelihood estimation, which makes use of all available data to produce relatively unbiased parameter estimates (Black et al., 2011). Continuous variables were z-scored. Race/ethnicity was included as a covariate because people of color are disproportionately likely to live in disadvantaged contexts due to histories of discrimination (e.g., redlining; Woods, 2012), and are also more likely to face stressors such as racism and prejudice that may increase risk for disordered eating (Mikhail & 11 Klump, 2020). Models examined adrenarche status (coded dichotomously as 0 = early adrenarche, 1 = late adrenarche) as a moderator to examine whether phenotypic associations between disadvantage and disordered eating differ across adrenarche in boys. GxE Analyses Extended univariate, double moderator twin models (van der Sluis et al., 2012) were used to examine how genetic and environmental influences on disordered eating differ across disadvantage in boys, and whether these GxE effects depend on developmental stage. The double moderator twin model is depicted in Figure S1.1. This model examines additive genetic (A; i.e., genetic influences that sum across genes), shared environmental (C; i.e., environmental factors that increase similarity between co-twins, such as attending the same school), and non-shared environmental (E; i.e., environmental factors that differentiate twins raised in the same family, such as non-overlapping friend groups) influences on disordered eating, and how these influences differ across disadvantage and adrenarche. The van der Sluis (2012) model allowed us to include twins who were discordant on adrenarche status while correcting for potential biases in significance testing resulting from the correlation between adrenarche and disordered eating. All twins were concordant on disadvantage variables, as these were measured at the family level. Because moderators are included in the means model, A, C, and E reflect the etiology of disordered eating after regressing out variance shared with the moderators. Double moderator twin models include 12 major parameters of interest: 3 initial path coefficients (a, c, e in Figure S1.1) that capture genetic/environmental influences at the lowest level of the moderators (i.e., among the least disadvantaged boys in early adrenarche), and 9 moderation coefficients that capture linear increases/decreases in the initial ACE path coefficients as a function of developmental stage (βxP, βyP, βzP in Figure S1.1), disadvantage (βxD, βyD, βzD in Figure 12 S1.1), and their interaction (βxPD, βyPD, βzPD in Figure S1.1). Quadratic moderators were not included because the data suggested only linear effects were present. This approach is consistent with our earlier study in female twins (Mikhail, Carroll, et al., 2021), and also helps to conserve power and enhance interpretability. The full model was fit first, with all path estimates and moderators freely estimated. Submodels were then fit based on the full model parameter estimates and confidence intervals to identify a best-fitting model. This approach allowed for identification of relevant submodels without conducting an excessive number of tests, as each model has numerous possible submodels. Best-fitting models were identified as those that had a non-significant difference in minus twice the log-likelihood (−2lnL) between the full and nested model, and minimized Akaike’s Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample-size adjusted BIC (SABIC). If AIC, BIC, and SABIC identified different models as best-fitting, the model that optimized two out of three fit indices was selected as best-fitting. BMI percentile was regressed out of log-transformed MTP-ED total scores, and the resulting residuals were standardized. Neighborhood disadvantage and family SES percentiles were floored at 0, then scaled from 0-1 for interpretability. Adrenarche was coded dichotomously (0 = early adrenarche, 1 = late adrenarche). Following prior recommendations for twin moderation models (Purcell, 2002), tables and figures report unstandardized path coefficient and moderation estimates. Unstandardized estimates are generally recommended because they reflect absolute differences in genetic/environmental influences across the moderators, while standardized estimates only capture differences in proportions of the total variance. However, standardized estimates are also reported where appropriate to facilitate interpretability. 13 Transparency and Openness Data, analysis code, and research materials are available from the corresponding author upon reasonable request. This study was not preregistered. 14 Sample Descriptives RESULTS A range of disordered eating symptoms was represented (MTP-ED score range = 0-15; possible range = 0-18), including more severe ED behaviors such as binge eating (8.5% of the sample). As expected, boys displayed greater disordered eating symptoms in late adrenarche than in early adrenarche (p <.001, d = .32). Disordered eating symptoms were also significantly associated with both neighborhood disadvantage (r = .10, p <.001) and family SES (r = -.11, p <.001) with a small effect size when examined using Pearson correlations. Importantly, relatively modest phenotypic associations between disadvantage and disordered eating do not preclude GxE, and in fact may reflect the presence of significant moderation (e.g., stronger associations for individuals with genetic vulnerabilities, and weaker/no association for individuals without genetic risk). Phenotypic Analyses In MLMs examining differences in associations between disadvantage and disordered eating across adrenarche, we observed expected significant main effects of adrenarche and BMI indicating greater disordered eating in boys during late adrenarche and for boys at higher BMI percentiles. We also observed significant or trend-level interactions between adrenarche status and disadvantage for both neighborhood disadvantage and family SES (see Table 1.1). For both neighborhood disadvantage and family SES, interactions indicated that the association between disadvantage and disordered eating was stronger in late adrenarche. Specifically, in the model including BMI percentile as a covariate, the association between neighborhood disadvantage and disordered eating was significant for boys in late adrenarche (β = .08, p = .001, 95% CI [.03, .13]) but not in early adrenarche (β = -.03, p = .640, 95% CI [-.14, .09]). Similarly, when controlling for 15 BMI, family SES was significantly associated with disordered eating for boys in late adrenarche (β = -.08, p = .001, 95% CI [-.12, -.03]), but not in early adrenarche (β = .05, p = .360, 95% CI [- .05, .15]). Results were similar (but with slightly larger effect sizes for boys in late adrenarche) in models not including BMI as a covariate. Findings were consistent with the hypothesis that phenotypic associations between disadvantage and disordered eating may be greatest in late adrenarche, following GxE during early adrenarche. GxE Analyses As shown in Supplemental Material (Tables S1.3-S1.4 and Figure S1.2), GxE analyses yielded very similar results with and without BMI percentile regressed out of the MTP-ED total score for family SES. However, the full GxE model of neighborhood disadvantage that did not control for BMI failed to converge, although cotwin correlations suggested a similar pattern of effects as the model that did control for BMI (see Table S1.2). Results below therefore focus on models that controlled for BMI. For both neighborhood disadvantage and family SES, genetic influences on disordered eating appeared to differ across disadvantage and adrenarche in the full model (see Figures 1.1 and 1.2). Specifically, for boys living in more advantaged contexts (low ADI or high family SES), genetic influences appeared substantially greater during late adrenarche than in early adrenarche. This pattern of results is consistent with previous findings suggesting greater genetic influences on disordered eating in late adrenarche in relatively advantaged boys (Culbert et al., 2017). However, for boys living in more disadvantaged circumstances (high ADI or low family SES), genetic influences on disordered eating appeared at least as large in early adrenarche as in late adrenarche. Differences in environmental influences across disadvantage and adrenarche appeared less pronounced than moderation of genetic effects in these models. 16 With respect to model fitting, no moderation models fit poorly for both neighborhood disadvantage and family SES, suggesting significant moderation effects (see Table 1.2). The best- fitting models for both neighborhood disadvantage and family SES retained disadvantage x adrenarche moderation of the A parameter, such that genetic influences on disordered eating were greater in late adrenarche, but only for boys living in advantaged circumstances (see Tables 1.2- 1.3 and Figures 1.1-1.2). For boys in disadvantaged neighborhoods and families, genetic influences were already substantial during early adrenarche. Consequently, the estimated proportion of variance in disordered eating due to genetic factors during early adrenarche was significantly greater for boys from more disadvantaged neighborhoods (low ADI: 19% of variance due to genes; high ADI: >95% of variance due to genes) 2 and families (high SES: 35% of variance due to genes; low SES: 67% of variance due to genes). While some moderation of C and E parameters was also retained in the best-fitting models, these effects appeared relatively modest when plotted, particularly for family SES (see Figures 1.1 and 1.2). Overall, effects were consistent with the hypothesis that disadvantage may potentiate earlier expression of genetic influences on disordered eating during early adrenarche through GxE. 2A very high estimated percentage of variance due to genetic factors could reflect non-additive genetic influences. To test this possibility, we ran an additional set of analyses that modeled non-additive genetic influences and dropped shared environmental influences (i.e., ADE models). The best-fitting ADE model fit worse on all fit indices than the best fitting ACE model, suggesting that non-additive genetic influences are not a major contributor to observed effects. 17 DISCUSSION This is the first study to examine phenotypic and GxE associations between multiple forms of disadvantage and disordered eating in boys, substantially extending our understanding of how disadvantage may impact disordered eating in youth. Both neighborhood disadvantage and lower family SES were associated with significantly greater phenotypic disordered eating symptoms in boys beginning in late adrenarche. Notably, effects remained significant even after controlling for BMI, indicating that the association between disadvantage and disordered eating in boys cannot be solely attributed to increased body weight and attendant weight stigma in disadvantaged environments. GxE analyses showed substantially stronger and earlier activation of genetic influences on disordered eating for boys living in disadvantaged environments during early adrenarche, when genetic influences were modest in more advantaged boys. This earlier activation of genetic influences could contribute to greater phenotypic ED symptoms later in development, reflecting a potentially lasting impact of disadvantage on ED risk in boys. Findings are novel in highlighting disadvantage as a significant risk factor for disordered eating in boys, perhaps especially for those with underlying genetic vulnerabilities. Prior research indicates that adrenarche is a critical period for activation of genetic influences on disordered eating in relatively advantaged boys, with genetic influences increasing gradually across early adrenarche (i.e., prior to gonadarche), then remaining constant from late adrenarche/gonadarche into adulthood (Culbert et al., 2017; Klump et al., 2012). We replicated these prior findings for boys from relatively advantaged neighborhoods and families, who showed a precipitous increase in genetic influences from early adrenarche to late adrenarche. However, for boys living in more disadvantaged circumstances, GxE analyses indicated that genetic influences on disordered eating were already substantial in early adrenarche, suggesting earlier 18 activation of genetic influences that could increase later risk. Importantly, genetic influences did not differ across disadvantage during late adrenarche, consistent with a shift in the developmental timing of expression of genetic risk in disadvantaged environments rather than a general increase in genetic influences regardless of developmental stage. Although GxE effects emerged during early adrenarche, phenotypic associations between disadvantage and disordered eating were not apparent until late adrenarche. This pattern of effects (increased genetic activation followed by later phenotypic expression) may reflect alterations to developing neurocircuitry during key hormonal/developmental periods that have enduring effects on later behavior (i.e., organizational hormone effects; Schulz & Sisk, 2016). Similar potentially organizational impacts of risk factors for EDs during puberty have been observed previously in girls and female animals (e.g., Klump et al., 2018). Both familial and neighborhood disadvantage are accompanied by considerable stress (e.g., stemming from financial instability, food insecurity, noise pollution, community violence, etc.) that could potentiate expression of genes relevant to vulnerability for disordered eating earlier than developmentally normative. Effects during adrenarche may involve interactions between rising androgen levels and the physiological stress response that could together lead to changes in gene expression and amplification of risk. Consistent with this possibility, a robust body of literature indicates that stress can alter gene expression and brain organization in neural circuits relevant to disordered eating (e.g., regions in the amygdala and prefrontal cortex involved in inhibitory control and emotion regulation; McEwen, 2013), and that androgens regulate the stress response and downstream physiological changes in males (Zuloaga et al., 2020). Relatedly, stress has been shown to alter the timing of brain development, promoting earlier maturation of emotion-related circuits that may be adaptive in the short-term, but have more deleterious long- 19 term repercussions for coping with stress and negative affect (Callaghan & Tottenham, 2016). This “stress acceleration hypothesis” is consistent with our findings of earlier activation of genetic influences in boys experiencing disadvantage. While stress is associated with increased disordered eating (Gadalla, 2009; Mitchell et al., 2016) and androgens are generally protective against disordered eating in men and boys (Culbert et al., 2014, 2020), no studies have yet examined how androgens and stress may interact to impact ED risk. Additional longitudinal research is needed to identify how the stress accompanying disadvantage may interact with androgens during development to impact gene expression and neural development in boys. Research is also needed to identify which aspects of disadvantage have the greatest impact on ED risk, and whether stressors that directly impact nutritional status (e.g., food insecurity) may have particularly pronounced effects. This study had several strengths, including a large, population-based sample, multiple measures of disadvantage, and developmentally sensitive analyses. Nevertheless, some limitations should be noted. As in our earlier study of disadvantage effects in girls (Mikhail, Carroll, et al., 2021), we relied on a parent-report measure of disordered eating. Using a consistent outcome measure across studies allows for direct comparison between the current study and Mikhail, Carroll, et al. (2021). Our disordered eating measure also demonstrated strong psychometric properties and expected associations with other key variables (e.g., BMI, puberty, internalizing) in boys. Despite this, EDs are often accompanied by considerable shame and secrecy, and parents may not be fully aware of all symptoms experienced by youth. Replication with self-reported symptoms is therefore needed. It would also be helpful to examine whether different symptom domains (e.g., binge eating versus body image concerns) relate to disadvantage differently. Interestingly, however, initial research in adults suggests disadvantage may be associated with 20 increases in all types of EDs and their symptoms, rather than only select symptoms (Becker et al., 2019; Coffino et al., 2020). Relatedly, determination of adrenarcheal development relied on indirect measures (i.e., age and outward indicators of gonadarche). Though our method of measuring adrenarche is consistent with past developmental studies of EDs in boys (i.e., Culbert et al., 2017), findings would ideally be replicated using adrenal androgen levels as a more precise, continuous measure of adrenarche. Analyses examined a population-based sample, rather than a sample enriched for disadvantage. An advantage of this approach is that the full range of disadvantage was present, allowing us to more easily detect differences between youth high and low in disadvantage. Nevertheless, effect sizes may have been larger in a sample specifically enriched for disadvantage, and future research should examine samples with larger numbers of highly disadvantaged youth. Additionally, observed associations were correlational, and causal associations between disadvantage and disordered eating cannot necessarily be inferred. Longitudinal research and research on the “active ingredients” underlying disadvantage effects is needed to continue to expand our understanding of how disadvantage may impact disordered eating for both boys and girls. 21 Variable Intercept ADI Adrenarche status ADI x adrenarche Race/ethnicity Black/African American (non-Latinx) Latinx/Hispanic Asian American Native American/ American Indian More than one race Other/unknown Table 1.1. MLMs examining associations between disadvantage and disordered eating across adrenarche status Neighborhood Disadvantage TABLES BMI Percentile Not Included as a Covariate β -.33 -.007 .38 .12 SE .06 <.001 .899 .06 .06 <.001 .050 .06 95% CI Variable -.44, -.21 Intercept -.12, .11 ADI .26, .51 Adrenarche status .0001, .24 ADI x adrenarche BMI Percentile Included as a Covariate β -.30 -.03 .36 .11 p <.001 .640 <.001 .078 SE .06 .06 .06 .06 p 95% CI -.42, -.19 -.14, .09 .24, .49 -.01, .23 .006 .10 .950 -.18, .19 .22 .007 -.13 -.008 .37 .20 .23 .50 .13 .16 .276 .976 .789 .950 .024 -.18, .62 -.45, .47 -1.12, .85 -.25, .24 .05, .69 Race/ethnicity Black/African American (non-Latinx) Latinx/Hispanic Asian American Native American/ American Indian More than one race Other/unknown BMI percentile -.04 .10 .701 -.24, .16 .03 .06 -.21 .05 .50 .29 .22 .24 .48 .13 .18 .02 .903 .806 .657 -.40, .45 -.42, .54 -1.14, .72 .716 .005 <.001 -.20, .29 .15, .84 .24, .33 Family Socioeconomic Status (SES) BMI Percentile Not Included as a Covariate β -.29 -.01 .33 -.09 SE .06 <.001 .05 .799 .06 <.001 .124 .06 p Variable Intercept SES Adrenarche status SES x adrenarche Race/ethnicity Black/African American (non-Latinx) Latinx/Hispanic Asian American .12 .09 .158 .18 .08 .19 .21 .348 .715 95% CI Variable Intercept -.40, -.19 -.12, .09 SES .21, .44 Adrenarche status SES x adrenarche -.20, .02 Race/ethnicity Black/African American (non-Latinx) Latinx/Hispanic Asian American -.19, .55 -.34, .49 -.05, .30 BMI Percentile Included as a Covariate β -.28 .05 .32 -.12 p <.001 .360 <.001 .029 SE .06 .05 .06 .06 95% CI -.39, -.17 -.05, .15 .20, .44 -.23, -.01 .04 .09 .674 -.14, .22 -.02 .08 .20 .22 .923 .709 -.41, .37 -.35, .52 22 Table 1.1 (cont’d) Native American/ American Indian More than one race Other/unknown -.39 .38 .309 -1.15, .36 -.43 .37 .237 -1.15, .28 -.17, .29 .15, .84 .25, .32 Note: MLM = multilevel model; ADI = Area Deprivation Index; adrenarche: 0 = early adrenarche, 1 = late adrenarche; BMI = body .600 .005 <.001 -.20, .26 .05, .68 .12 .18 .02 .06 .50 .28 .778 .023 .03 .36 .12 .16 Native American/ American Indian More than one race Other/unknown BMI percentile mass index. Reference group for race/ethnicity is White. Effects significant at p <.05 are bolded. 23 Table 1.2. Model fit comparisons for genotype x environment models across adrenarche status and disadvantage Model -2lnL χ2 Δ (df) p AIC BIC SABIC Neighborhood Disadvantage Full model Nested submodels No moderation Constrain all C mods Constrain all E mods Constrain all A mods Constrain C ADI and ADI x adrenarche mods Constrain C ADI and adrenarche mods Constrain C ADI and ADI x adrenarche mods, E adrenarche mod Constrain C ADI and adrenarche mods, E adrenarche mod Constrain C main effect and ADI and ADI x adrenarche mods, E adrenarche mod Constrain C main effect and ADI and adrenarche mods, E adrenarche mod 6551.806 — — 6597.807 6719.546 6646.482 6665.528 6559.144 6600.034 6572.244 6553.568 6553.760 6554.586 113.722 (9) <.001 7.338 (3) .062 48.228 (3) <.001 20.438 (3) <.001 .414 1.762 (2) .376 1.954 (2) .427 2.780 (3) 6693.529 6599.144 6640.035 6612.243 6595.568 6595.760 6594.586 6767.631 6705.004 6745.895 6718.103 6706.721 6706.913 6700.447 6723.157 6641.471 6682.361 6654.570 6640.011 6640.203 6636.913 6554.272 2.466 (3) .481 6594.271 6700.132 6636.598 6555.156 3.350 (4) .501 6593.155 6693.723 6633.366 6554.346 2.540 (4) .637 6592.346 6692.913 6632.556 Family Socioeconomic Status (SES) — 7410.740 — 7456.740 7581.386 7508.318 Full model Nested submodels No moderation Constrain all C mods Constrain all E mods Constrain all A mods Constrain E SES mod Constrain E SES and SES x adrenarche mods Constrain E SES mod, A and C adrenarche Mods Constrain E SES mod, A adrenarche and SES x adrenarche mods, C adrenarche mod Constrain E SES mod, A adrenarche mod, C SES and SES x adrenarche mods 7516.248 105.508 (9) <.001 .016 10.308 (3) 7421.048 23.190 (3) <.001 7433.930 .019 9.962 (3) 7420.702 .529 0.396 (1) 7411.136 .034 6.738 (2) 7417.478 .811 0.960 (3) 7411.700 7544.248 7461.048 7473.930 7460.703 7455.136 7459.479 7451.701 7620.119 7569.436 7582.318 7569.090 7574.363 7573.286 7560.088 7575.643 7505.899 7518.780 7505.553 7504.472 7506.572 7496.551 7419.126 8.386 (4) .078 7457.125 7560.094 7499.733 7419.976 9.236 (4) .055 7457.975 7560.943 7500.583 24 Table 1.2 (cont’d) Note: ADI = Area Deprivation Index; adrenarche = coded 0 for early adrenarche, 1 for late adrenarche; mod(s) = moderator(s); -2lnL = minus twice the log-likelihood; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; SABIC = sample size adjusted Bayesian Information Criterion; full model = model with paths and all moderators; A = additive genetic variance; C = shared environmental variance; E = nonshared environmental variance. The best-fitting model description is bolded. 25 Table 1.3. Unstandardized path and moderator estimates for full and best-fitting genotype x environment models Model a c e βxP βyP βzP βxD βyD βzD βxPD βyPD βzPD Neighborhood Disadvantage Full model Best-fitting .269 (.096, .442) .244 (.085, .404) -.042 (-.317, .234) — .472 (.360, .584) .510 (.453, .568) .510 (.280, .741) .597 (.419, .775) .253 (-.161, .668) .059 (-.077, .194) .841 (.504, 1.177) -.106 (-.689, .476) -.394 (-.577, -.211) -.843 (-1.293, -.393) .413 (-.288, 1.114) — — — .886 (.563, 1.209) -.450 (-.550, -.350) -.977 (-1.372, -.581) .591 (.331, .850) Family SES .499 (.273, .725) .587 (.499, .675) Full model .794 (.575, 1.013) .299 (-.253, .850) .417 (.295, .539) .019 (-.263, .300) .169 (-.400, .739) .244 (.100, .388) -.337 (-.695, .021) -.718 (-1.389, -.046) -.057 (-.238, .123) .276 (-.205, .758) .490 (-.266, 1.247) -.115 (-.339, .110) Best-fitting .278 (.185, .370) Note: A = additive genetic influences at the lowest levels of the moderators; c = shared environmental influences at the lowest levels of -.878 (-1.294, -.462) -.174 (-.298, -.050) -.383 (-.655, -.110) .692 (.363, 1.020) .819 (.656, .983) .433 (.130, .737) .385 (.325, .444) .316 (.087, .545) — — — the moderators; e = non-shared environmental influences at the lowest levels of the moderators; βxP, βyP, βzP = coefficients for moderation of genetic/environmental variance by adrenarche; βxD, βyD, βzD = coefficients for moderation of genetic/environmental variance by neighborhood disadvantage/family SES; βxPD, βyPD, βzPD = coefficients representing changes in the moderating effects of disadvantage across adrenarche (i.e., the disadvantage x development interaction). 95% confidence intervals of parameter estimates are included in parentheses. Effects significant at p <.05 are bolded. 26 FIGURES Figure 1.1. Additive genetic (A), shared environmental (C), and non-shared environmental (E) influences on disordered eating across adrenarche status and neighborhood disadvantage. ADI = Area Deprivation Index. 27 Figure 1.2. Additive genetic (A), shared environmental (C), and non-shared environmental (E) influences on disordered eating across adrenarche status and family socioeconomic status (SES). 28 BIBLIOGRAPHY Accurso, E. C., & Waller, G. (2021). Concordance between youth and caregiver report of eating disorder psychopathology: Development and psychometric properties of the Eating Disorder‐15 for Parents/Caregivers (ED‐15‐P). International Journal of Eating Disorders, 54(7), 1302-1306. Alvarado, S. E. (2016). Neighborhood disadvantage and obesity across childhood and adolescence: Evidence from the NLSY children and young adults cohort (1986–2010). Social Science Research, 57, 80-98. Arnold, A. P. (2009). The organizational–activational hypothesis as the foundation for a unified theory of sexual differentiation of all mammalian tissues. Hormones and Behavior, 55(5), 570-578. Auchus, R. J., & Rainey, W. E. (2004). Adrenarche–physiology, biochemistry and human disease. Clinical Endocrinology, 60(3), 288-296. Becker, C. B., Middlemas, K., Gomez, F., & Kilpela, L. S. (2021). An exploratory examination of internalized weight stigma in a sample living with food insecurity. Body Image, 37, 238-245. Becker, C. B., Middlemass, K. M., Gomez, F., & Martinez-Abrego, A. (2019). Eating disorder pathology among individuals living with food insecurity: A replication study. Clinical Psychological Science, 7(5), 1144-1158. Becker, C. B., Middlemass, K., Taylor, B., Johnson, C., & Gomez, F. (2017). Food insecurity and eating disorder pathology. International Journal of Eating Disorders, 50(9), 1031- 1040. Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88(3), 588-606. Black, A. C., Harel, O., & Betsy McCoach, D. (2011). Missing data techniques for multilevel data: Implications of model misspecification. Journal of Applied Statistics, 38(9), 1845- 1865. Bronfenbrenner, U., & Ceci, S. J. (1994). Nature-nurture reconceptualized in developmental perspective: A bioecological model. Psychological Review, 101(4), 568-586. Bundak, R., Darendeliler, F., Gunoz, H., Bas, F., Saka, N., & Neyzi, O. (2007). Analysis of puberty and pubertal growth in healthy boys. European Journal of Pediatrics, 166(6), 595-600. Burke, N. L., Hazzard, V. M., Schaefer, L. M., Simone, M., O'Flynn, J. L., & Rodgers, R. F. (2022). Socioeconomic status and eating disorder prevalence: At the intersections of 29 gender identity, sexual orientation, and race/ethnicity. Psychological Medicine, 1-11. Burt, S. A. (2014). Research review: The shared environment as a key source of variability in child and adolescent psychopathology. Journal of Child Psychology and Psychiatry, 55(4), 304-312. Burt, S. A., Clark, D. A., Pearson, A. L., Klump, K. L., & Neiderhiser, J. M. (2020). Do neighborhood social processes moderate the etiology of youth conduct problems? Psychological Medicine, 50(9), 1519-1529. Burt, S. A., & Klump, K. L. (2013). The Michigan state university twin registry (MSUTR): An update. Twin Research and Human Genetics, 16(1), 344-350. Burt, S. A., & Klump, K. L. (2019). The Michigan State University Twin Registry (MSUTR): 15 years of twin and family research. Twin Research and Human Genetics, 22(6), 741-745. Burt, S. A., Klump, K. L., Vazquez, A. Y., Shewark, E. A., & Hyde, L. W. (2021). Identifying patterns of youth resilience to neighborhood disadvantage. Research in Human Development, 18(3), 181-196. Burt, S. A., Pearson, A. L., Carroll, S., Klump, K. L., & Neiderhiser, J. M. (2020). Child antisocial behavior is more environmental in origin in disadvantaged neighborhoods: Evidence across residents’ perceptions and geographic scales in two samples. Journal of Abnormal Child Psychology, 48(2), 265-276. Callaghan, B. L., & Tottenham, N. (2016). The stress acceleration hypothesis: Effects of early-life adversity on emotion circuits and behavior. Current Opinion in Behavioral Sciences, 7, 76-81. Carroll, S. L., Clark, D. A., Hyde, L. W., Klump, K. L., & Burt, S. A. (2021). Continuity and change in the genetic and environmental etiology of youth antisocial behavior. Behavior Genetics, 1-12. Coffino, J. A., Grilo, C. M., & Udo, T. (2020). Childhood food neglect and adverse experiences associated with DSM-5 eating disorders in US national sample. Journal of Psychiatric Research, 127, 75-79. Cooksey-Stowers, K., Schwartz, M. B., & Brownell, K. D. (2017). Food swamps predict obesity rates better than food deserts in the United States. International Journal of Environmental Research and Public Health, 14(11), 1366. Couturier, J., Lock, J., Forsberg, S., Vanderheyden, D., & Yen, H. L. (2007). The addition of a parent and clinician component to the eating disorder examination for children and adolescents. International Journal of Eating Disorders, 40(5), 472-475. Culbert, K. M., Burt, S. A., & Klump, K. L. (2017). Expanding the developmental boundaries of 30 etiologic effects: The role of adrenarche in genetic influences on disordered eating in males. Journal of Abnormal Psychology, 126(5), 593-606. Culbert, K. M., Burt, S. A., Sisk, C. L., Nigg, J. T., & Klump, K. L. (2014). The effects of circulating testosterone and pubertal maturation on risk for disordered eating symptoms in adolescent males. Psychological Medicine, 44(11), 2271-2286. Culbert, K. M., Shope, M. M., Sisk, C. L., & Klump, K. L. (2020). Low testosterone is associated with dysregulated eating symptoms in young adult men. International Journal of Eating Disorders, 53(9), 1469-1479. DeCarlo Santiago, C., Wadsworth, M. E., & Stump, J. (2011). Socioeconomic status, neighborhood disadvantage, and poverty-related stress: Prospective effects on psychological syndromes among diverse low-income families. Journal of Economic Psychology, 32(2), 218-230. Dubowitz, T., Ghosh‐Dastidar, M., Eibner, C., Slaughter, M. E., Fernandes, M., Whitsel, E. A.,...Escarce, J. J. (2012). The Women’s Health Initiative: The food environment, neighborhood socioeconomic status, BMI, and blood pressure. Obesity, 20(4), 862-871. Fairweather-Schmidt, A. K., & Wade, T. D. (2017). Weight-related peer-teasing moderates genetic and environmental risk and disordered eating: Twin study. The British Journal of Psychiatry, 210(5), 350-355. Gadalla, T. M. (2009). Eating disorders in men: A community-based study. International Journal of Men’s Health, 8(1). Gard, M. C., & Freeman, C. P. (1996). The dismantling of a myth: A review of eating disorders and socioeconomic status. International Journal of Eating Disorders, 20(1), 1-12. Gillespie, C. F., Phifer, J., Bradley, B., & Ressler, K. J. (2009). Risk and resilience: Genetic and environmental influences on development of the stress response. Depression and Anxiety, 26(11), 984-992. Goodman, E., McEwen, B. S., Dolan, L. M., Schafer-Kalkhoff, T., & Adler, N. E. (2005). Social disadvantage and adolescent stress. Journal of Adolescent Health, 37(6), 484-492. Goodman, R. (1997). The Strengths and Difficulties Questionnaire: A research note. Journal of Child Psychology and Psychiatry, 38(5), 581-586. Gordon, N. P., & Mellor, R. G. (2015). Accuracy of parent-reported information for estimating prevalence of overweight and obesity in a race-ethnically diverse pediatric clinic population aged 3 to 12. BMC Pediatrics, 15(1), 1-11. Guran, T., Firat, I., Yildiz, F., Kaplan Bulut, I., Dogru, M., & Bereket, A. (2015). Reference values for serum dehydroepiandrosterone‐sulphate in healthy children and adolescents 31 with emphasis on the age of adrenarche and pubarche. Clinical Endocrinology, 82(5), 712- 718. Hackman, D. A., Betancourt, L. M., Brodsky, N. L., Hurt, H., & Farah, M. J. (2012). Neighborhood disadvantage and adolescent stress reactivity. Frontiers in Human Neuroscience, 6, 277. Hazzard, V. M., Barry, M. R., Leung, C. W., Sonneville, K. R., Wonderlich, S. A., & Crosby, R. D. (2021). Food insecurity and its associations with bulimic-spectrum eating disorders, mood disorders, and anxiety disorders in a nationally representative sample of US adults. Social Psychiatry and Psychiatric Epidemiology, 1-8. Huryk, K. M., Drury, C. R., & Loeb, K. L. (2021). Diseases of affluence? A systematic review of the literature on socioeconomic diversity in eating disorders. Eating Behaviors, 101548. Ilondo, M. M., Vanderschueren-Lodeweyckx, M., Vlietinck, R., Pizarro, M., Malvaux, P., Eggermont, E., & Eeckels, R. (1982). Plasma androgens in children and adolescents. Hormones, 16(2), 61-77. Kind, A. J., & Buckingham, W. R. (2018). Making neighborhood-disadvantage metrics accessible—the neighborhood atlas. The New England Journal of Medicine, 378(26), 2456-2458. Kind, A. J., Jencks, S., Brock, J., Yu, M., Bartels, C., Ehlenbach, W.,...Smith, M. (2014). Neighborhood socioeconomic disadvantage and 30-day rehospitalization: A retrospective cohort study. Annals of Internal Medicine, 161(11), 765-774. Klump, K. L., & Burt, S. A. (2006). The Michigan State University Twin Registry (MSUTR): Genetic, environmental and neurobiological influences on behavior across development. Twin Research and Human Genetics, 9(6), 971-977. Klump, K. L., Culbert, K. M., Slane, J. D., Burt, S. A., Sisk, C. L., & Nigg, J. T. (2012). The effects of puberty on genetic risk for disordered eating: Evidence for a sex difference. Psychological Medicine, 42(3), 627-637. Klump, K. L., Fowler, N., Mayhall, L., Sisk, C. L., Culbert, K. M., & Burt, S. A. (2018). Estrogen moderates genetic influences on binge eating during puberty: Disruption of normative processes? Journal of Abnormal Psychology, 127(5), 458–470. Klump, K. L., McGue, M., & Iacono, W. G. (2003). Differential heritability of eating attitudes and behaviors in prepubertal versus pubertal twins. International Journal of Eating Disorders, 33(3), 287-292. Klump, K. L., Perkins, P. S., Burt, S. A., McGue, M. A. T. T., & Iacono, W. G. (2007). Puberty moderates genetic influences on disordered eating. Psychological Medicine, 37(5), 627- 634. 32 Koopman-Verhoeff, M. E., Gredvig-Ardito, C., Barker, D. H., Saletin, J. M., & Carskadon, M. (2020). Classifying pubertal development using child and parent report: Comparing the Pubertal Development Scales to Tanner staging. Journal of Adolescent Health, 66, 597- 602. Lydecker, J. A., & Grilo, C. M. (2019). Food insecurity and bulimia nervosa in the United States. International Journal of Eating Disorders, 52(6), 735-739. Lykken, D. T., Bouchard, T. J., McGue, M., & Tellegen, A. (1990). The Minnesota twin family registry: Some initial findings. Acta Geneticae Medicae et Gemellologiae: Twin Research, 39(1), 35-70. McEwen, B. S. (2013). The brain on stress: Toward an integrative approach to brain, body, and behavior. Perspectives on Psychological Science, 8(6), 673-675. Mikhail, M. E., Anaya, C., Culbert, K. M., Sisk, C. L., Johnson, A., & Klump, K. L. (2021). Gonadal hormone influences on sex differences in binge eating across development. Current Psychiatry Reports, 23(11), 1-10. Mikhail, M. E., Carroll, S. L., Clark, D. A., O’Connor, S., Burt, S. A., & Klump, K. L. (2021). Context matters: Neighborhood disadvantage is associated with increased disordered eating and earlier activation of genetic influences in girls. Journal of Abnormal Psychology, 130(8), 875–885. Mikhail, M. E., & Klump, K. L. (2021). A virtual issue highlighting eating disorders in people of black/African and Indigenous heritage. International Journal of Eating Disorders, 54(3), 459-467. Miller, A., Pohlig, R. T., & Reisman, D. S. (2020). Social and physical environmental factors in daily stepping activity in those with chronic stroke. Topics in Stroke Rehabilitation, 1-9. Mitchell, K. S., Porter, B., Boyko, E. J., & Field, A. E. (2016). Longitudinal associations among posttraumatic stress disorder, disordered eating, and weight gain in military men and women. American Journal of Epidemiology, 184(1), 33-47. Mond, J., Hall, A., Bentley, C., Harrison, C., Gratwick‐Sarll, K., & Lewis, V. (2014). Eating‐ disordered behavior in adolescent boys: Eating disorder examination questionnaire norms. International Journal of Eating Disorders, 47(4), 335-341. Nagata, J. M., Ganson, K. T., & Murray, S. B. (2020). Eating disorders in adolescent boys and young men: An update. Current Opinion in Pediatrics, 32(4), 476-481. Nagl, M., Jacobi, C., Paul, M., Beesdo-Baum, K., Höfler, M., Lieb, R., & Wittchen, H. U. (2016). Prevalence, incidence, and natural course of anorexia and bulimia nervosa among adolescents and young adults. European Child & Adolescent Psychiatry, 25(8), 903-918. 33 Neumark-Sztainer, D. R., Wall, M. M., Haines, J. I., Story, M. T., Sherwood, N. E., & van den Berg, P. A. (2007). Shared risk and protective factors for overweight and disordered eating in adolescents. American Journal of Preventive Medicine, 33(5), 359-369. O’Connor, S. M., Culbert, K. M., Mayhall, L. A., Burt, S. A., & Klump, K. L. (2020). Differences in genetic and environmental influences on body weight and shape concerns across pubertal development in females. Journal of Psychiatric Research, 121, 39-46. Peeters, H., Van Gestel, S., Vlietinck, R., Derom, C., & Derom, R. (1998). Validation of a telephone zygosity questionnaire in twins of known zygosity. Behavior Genetics, 28(3), 159-163. Powell, W. R., Buckingham, W. R., Larson, J. L., Vilen, L., Yu, M., Salamat, M. S.,...Kind, A. J. (2020). Association of neighborhood-level disadvantage with Alzheimer Disease neuropathology. JAMA Network Open, 3(6), e207559-e207559. Purcell, S. (2002). Variance components models for gene–environment interaction in twin analysis. Twin Research and Human Genetics, 5(6), 554-571. Rende, R., & Plomin, R. (1992). Diathesis-stress models of psychopathology: A quantitative genetic perspective. Applied and Preventive Psychology, 1(4), 177-182. Roubinov, D. S., Hagan, M. J., Boyce, W. T., Adler, N. E., & Bush, N. R. (2018). Family socioeconomic status, cortisol, and physical health in early childhood: The role of advantageous neighborhood characteristics. Psychosomatic Medicine, 80(5), 492-501. Schulz, K. M., & Sisk, C. L. (2016). The organizing actions of adolescent gonadal steroid hormones on brain and behavioral development. Neuroscience & Biobehavioral Reviews, 70, 148-158. Sheets, L. R., Kelley, L. E. H., Scheitler-Ring, K., Petroski, G. F., Barnett, Y., Barnett, C.,... Parker, J. C. (2020). An index of geospatial disadvantage predicts both obesity and unmeasured body weight. Preventive Medicine Reports, 18, 101067. Shields, M., Gorber, S. C., Janssen, I., & Tremblay, M. S. (2011). Obesity estimates for children based on parent-reported versus direct measures. Health Reports, 22(3), 47. Singh, G. K. (2003). Area deprivation and widening inequalities in US mortality, 1969–1998. American Journal of Public Health, 93(7), 1137-1143. Smink, F. R., Van Hoeken, D., & Hoek, H. W. (2012). Epidemiology of eating disorders: Incidence, prevalence and mortality rates. Current Psychiatry Reports, 14(4), 406-414. Sonneville, K. R., & Lipson, S. K. (2018). Disparities in eating disorder diagnosis and treatment according to weight status, race/ethnicity, socioeconomic background, and sex among 34 college students. International Journal of Eating Disorders, 51(6), 518-526. Steinberg, E., Tanofsky‐Kraff, M., Cohen, M. L., Elberg, J., Freedman, R. J., Semega‐Janneh, M.,...Yanovski, J. A. (2004). Comparison of the child and parent forms of the Questionnaire on Eating and Weight Patterns in the assessment of children's eating‐ disordered behaviors. International Journal of Eating Disorders, 36(2), 183-194. Strachan, E., Duncan, G., Horn, E., & Turkheimer, E. (2017). Neighborhood deprivation and depression in adult twins: Genetics and gene x environment interaction. Psychological Medicine, 47(4), 627-638. Suarez, G. L., Burt, S. A., Gard, A. M., Burton, J., Clark, D. A., Klump, K. L., & Hyde, L. W. (2022). The impact of neighborhood disadvantage on amygdala reactivity: Pathways through neighborhood social processes. Developmental Cognitive Neuroscience, 101061. Swanson, S. A., Aloisio, K. M., Horton, N. J., Sonneville, K. R., Crosby, R. D., Eddy, K. T.,...Micali, N. (2014). Assessing eating disorder symptoms in adolescence: Is there a role for multiple informants? International Journal of Eating Disorders, 47(5), 475-482. van der Sluis, S., Posthuma, D., & Dolan, C. V. (2012). A note on false positives and power in GxE modelling of twin data. Behavior Genetics, 42(1), 170-186. Thomas, K. S., Williams, M. O., & Vanderwert, R. E. (2021). Disordered eating and internalizing symptoms in preadolescence. Brain and Behavior, 11(1), e01904. von Ranson, K. M., Klump, K. L., Iacono, W. G., & McGue, M. (2005). The Minnesota Eating Behavior Survey: A brief measure of disordered eating attitudes and behaviors. Eating Behaviors, 6(4), 373-392. Woods, L. L. (2012). The Federal Home Loan Bank Board, redlining, and the national proliferation of racial lending discrimination, 1921–1950. Journal of Urban History, 38(6), 1036-1059. Zuloaga, D. G., Heck, A. L., De Guzman, R. M., & Handa, R. J. (2020). Roles for androgens in mediating the sex differences of neuroendocrine and behavioral stress responses. Biology of sex Differences, 11(1), 1-18. 35 APPENDIX A: SUPPLMENTAL MATERIAL Additional Information Regarding the Reliability and Validity of the MTP-ED in Boys The MTP-ED was previously validated in a large, population-based sample of female twins (N = 2,922; Mikhail, Carroll, et al., 2021). For the current study, the MTP-ED was further validated in boys. The MTP-ED had acceptable internal consistency across age (ages 8-12: α = .77; ages 13-17; α = .81) and pubertal development (early adrenarche: α = .70; early gonadarche: α = .78; mid/late gonadarche: α = .80) in boys. Exploratory factor analysis (EFA) with orthogonal varimax rotation yielded a single factor with an eigenvalue above 1 (factor 1 eigenvalue = 2.87, factor 2 eigenvalue = .31), suggesting that all MTP-ED items loaded on a single factor. Confirmatory factor analysis (CFA) of the nine MTP-ED items showed adequate fit for a single latent factor model in the full sample (RMSEA = .071; CFI = .941, TLI = .922, SRMR = .036). An alternative two-factor model suggested by the EFA that placed purging and dieting on a separate factor from the other disordered eating symptoms did not have appreciably better fit (RMSEA = .071; CFI = .943, TLI = .922, SRMR = .035), and so the single factor model was preferred due to parsimony. When comparing model fit across adrenarche, a model constraining all factor loadings to equality across early and late adrenarche had adequate fit (RMSEA = .072; CFI = .921, TLI = .919, SRMR = .067) that was similar to the fit of a model that allowed factor loadings to differ for boys in early and late adrenarche (RMSEA = .075; CFI = .925, TLI = .910, SRMR = .052) (AIC = 5492.107 for the constrained model and 5510.630 for the unconstrained model; BIC = 5782.110 for the constrained model and 5740.216 for the unconstrained model). The chi-square test comparing models with and without factor loadings constrained to equality across adrenarche was significant (χ2 = 38.52, p <.001), but this was likely due to the fact that the chi-square statistic is very sensitive to sample 36 size and has a high likelihood of rejecting more parsimonious models when sample size is large (Bentler & Bonett, 1980). Altogether, we concluded that the fit of a single latent factor model was adequate to enable analyses of disordered eating as a single composite scale across adrenarche in boys. The MTP-ED showed expected correlations with age (r = .13, p < .001), pubertal status (r = .13, p < .001), BMI percentile (r = .29, p < .001), and internalizing symptoms (e.g., worry, depression; r = .25, p < .001) in boys that were similar to associations for other self-report measures of disordered eating (Mond et al., 2014; Neumark-Sztainer & Hannan, 2000; Thomas et al., 2021). The MTP-ED discriminated between boys with and without a lifetime parent-reported ED (AN, BN, or BED) on a checklist of physical and mental health conditions on the MTP intake questionnaire (M(SD) with no ED: .92 (1.92); M(SD) with lifetime ED: 3.32 (3.41); p <.001). At the time the current study was conducted, 299 boys ages 7-18 and their primary caregiver from a separate, ongoing study (Twin Study of Mood, Behavior, and Hormones in Males) had completed the MTP-ED, with parents competing the MTP-ED in relation to their child and boys completing the MTP-ED in relation to themselves. Boys in this study also completed the Minnesota Eating Behavior Survey (MEBS; von Ranson et al., 2005), an established self-report measure of ED symptoms. Correlations were large between self-reported MTP-ED and self- reported MEBS total scores (r = .66, p <.001). As is typical in the ED literature, correlations between parent- and self-reported MTP-ED were significant but small-to-moderate in magnitude (r = .26, p <.001). While parent- and youth-reported symptoms represent somewhat distinct perspectives on a youth’s disordered eating, prior research suggests parent-reported symptoms differentiate youth with and without clinical EDs (Accurso & Waller, 2021) and show similar or greater concordance with objective external measurements (e.g., BMI, clinician-reported 37 symptoms) as adolescent-reported symptoms (Couturier et al., 2007; Steinberg et al., 2004; Swanson et al., 2014). Parent report may be particularly useful for younger boys who may have difficulty understanding disordered eating items. 38 Table S1.1. Descriptive statistics for participant demographics and symptoms (N = 3,484) Tables Participant Characteristics Age Zygosity (N listed as number of pairs) Monozygotic Dizygotic Unknown zygosity Race/ethnicity White (non-Latinx) Black/African American (non-Latinx) Latinx/Hispanic Asian American Native American/American Indian More than one race Other/Unknown BMI percentile Raw BMI PDS score Categorical adrenarche status Early adrenarche Late adrenarche Combined parental income (in thousands of dollars) Mother’s education level Less than high school High school graduate Less than 4 years of college College graduate (4-6 years of college) Post-graduate education Father’s education level Less than high school High school graduate Less than 4 years of college College graduate (4-6 years of college) Post-graduate education Area Deprivation Index (ADI) percentile rank relative to all census tracts in the United States Mean (SD) or % of Sample (N) 12.27 (2.96) Range 8.05-17.99 — — — — — — — — — — 0.5–99.5 13.17–38.39 1–4 — — $0–$300+ — — — — — — — — — — 1–100 43.4% (756) 56.5% (984) 0.1% (2) 84.6% (2,948) 7.1% (248) 1.4% (48) 1.1% (38) 0.3% (10) 3.6% (124) 2.0% (68) 55.28 (30.47) 19.56 (4.16) 2.01 (.88) 14.8% (495) 85.2% (2,847) $90.39 (54.41) 2.9% (98) 15.8% (538) 33.6% (1,144) 34.9% (1,190) 12.9% (440) 4.7% (148) 23.4% (746) 28.1% (896) 31.4% (1,000) 12.4% (394) 37.34 (26.39) 39 Table S1.1 (cont’d) Symptom Measures MTP-ED total score Reported having AN, BN, or BED Reported being treated for AN, BN, or BED Internalizing symptoms Note: PDS = Pubertal Development Scale; BMI = body mass index; MTP-ED = Michigan Twins 1.47 (1.78) 0–10 0–10 .65 Possible Range 0–18 — — Cronbach’s alpha .79 — — Sample Range 0–15 — — Mean (SD) or % of Sample (N) .94 (1.95) 0.9% (31) 0.4% (14) Project Eating Disorder Survey; AN = anorexia nervosa; BN = bulimia nervosa; BED = binge- eating disorder; internalizing symptoms = score on the Emotional Symptoms subscale of the Strengths and Difficulties Questionnaire (Goodman et al., 1997). N’s may not add up to the total N for all variables due to missing values. The lower percentage of participants with reported eating disorders likely reflects the young average age of the sample, as threshold eating disorders are very rare in boys prior to mid-adolescence (Smink et al., 2012). 40 Table S1.2. MLMs examining associations between disadvantage and disordered eating, with neighborhood disadvantage and family SES included in the same model Controlling For Adrenarche BMI Not Included as a Covariate β -.30 .07 -.05 .36 SE .06 <.001 .03 .006 .03 .048 .06 <.001 p 95% CI Variable Intercept -.42, -.19 ADI .02, .13 -.10, -.0005 Family SES .23, .48 BMI Included as a Covariate p Variable Intercept ADI Family SES Adrenarche status Race/ethnicity Black/African American (non-Latinx) Latinx/Hispanic Asian American Native American/ American Indian More than one race Other/unknown .02 .09 .802 -.16, .21 .22 .02 -.17 .006 .34 .20 .23 .50 .13 .16 .267 .922 .735 .960 .036 -.17, .62 -.44, .48 -1.15, .81 -.24, .25 .02, .66 Adrenarche status Race/ethnicity Black/African American (non-Latinx) Latinx/Hispanic Asian American Native American/ American Indian More than one race Other/unknown BMI Percentile Adrenarche as a Moderator BMI Included as a Covariate p p BMI Not Included as a Covariate β -.31 -.03 -.05 .37 .12 -.005 SE .06 <.001 .663 .06 .472 .07 .07 <.001 .080 .07 .946 .07 Variable Intercept ADI Family SES Adrenarche status ADI x adrenarche SES x adrenarche Race/ethnicity Black/African American (non-Latinx) .003 .10 .977 95% CI Variable -.44, -.19 Intercept -.15, .10 ADI Family SES -.18, .08 .24, .50 Adrenarche status -.01, .25 ADI x adrenarche SES x adrenarche -.14, .13 Race/ethnicity Black/African American (non-Latinx) -.18, .19 41 β -.29 .05 -.03 .34 SE .06 <.001 .03 .048 .03 .187 .06 <.001 95% CI -.40, -.17 .0004, .10 -.08, .02 .22, .47 -.02 .10 .818 -.22, .17 .04 .07 -.23 .22 .24 .48 .864 .771 .630 -.39, .46 -.41, .55 -1.16, .70 .06 .48 .28 .631 .13 .18 .007 .02 <.001 -.19, .31 .13, .83 .24, .32 β -.31 -.01 .03 .37 .08 -.08 SE .06 <.001 .844 .06 .608 .07 .07 <.001 .262 .07 .264 .07 95% CI -.44, -.19 -.14, .11 -.10, .16 .24, .50 -.06, .21 -.22, .06 -.04 .10 .702 -.23, .16 Table S1.2 (cont’d) Latinx/Hispanic Asian American Native American/ American Indian More than one race Other/unknown .21 .03 -.19 .20 .23 .50 .304 .892 .704 -.19, .60 -.43, .49 -1.17, .79 .02 .07 -.24 .22 .24 .48 .942 .759 .617 -.41, .44 -.40, .55 -1.17, .69 Latinx/Hispanic Asian American Native American/ American Indian More than one race Other/unknown BMI Percentile -.20, .30 .14, .83 .25, .33 Note: MLM = multilevel model; ADI = Area Deprivation Index; family SES = family socioeconomic status; adrenarche = coded 0 = .697 .13 .18 .006 .02 <.001 -.25, .24 .03, .67 -.003 .35 .05 .49 .29 .984 .031 .13 .16 early adrenarche, 1 = late adrenarche; BMI = body mass index. The outcome for all models is standardized, log-transformed Michigan Twins Project Eating Disorder Survey (MTP-ED) total score. Reference group for race/ethnicity is White. Effects significant at p <.05 are bolded. 42 Table S1.3. Model fit comparisons for genotype x environment models across adrenarche status with BMI percentile not regressed out Model -2lnL χ2 Δ (df) p AIC BIC SABIC Neighborhood Disadvantage – No Convergence of Full Model Family SES — — 8046.966 Full model Nested submodels No moderation Constrain E SES mod Constrain E SES and SES x adrenarche mods Constrain E SES mod, A and C adrenarche Mods Constrain E SES mod, A adrenarche and SES x adrenarche mods, C adrenarche mod Constrain E SES mod, A adrenarche mod, C SES and SES x adrenarche mods Note: ADI = Area Deprivation Index percentile; SES = family socioeconomic status; BMI = body mass index; adrenarche = adrenarche 115.118 (9) .276 (1) 8.014 (2) .568 (3) 8265.955 8210.468 8210.786 8195.922 8162.084 8047.242 8054.980 8047.534 8190.084 8091.242 8096.979 8087.535 8221.479 8140.577 8144.072 8132.385 10.060 (4) .018 .904 8137.634 8197.994 8055.912 8092.965 8144.543 8217.611 8057.026 8095.026 8136.520 8196.881 8093.912 8.946 (4) <.001 .062 .039 status (0 = early adrenarche, 1 = late adrenarche); mod(s) = moderator(s); -2lnL = minus twice the log-likelihood; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; SABIC = sample size adjusted Bayesian Information Criterion; full model = model with paths and all moderators; A = additive genetic variance; C = shared environmental variance; E = nonshared environmental variance. Although the model examining the ADI did not converge, cotwin correlations were consistent with results from the analogous GxE model that controlled for BMI in suggesting earlier activation of genetic influences for boys in early adrenarche living in disadvantaged neighborhoods. Specifically, the difference in the cotwin correlation between MZ and DZ twins was much greater for boys in early adrenarche living in more disadvantaged neighborhoods (high ADI: MZ = .982, DZ = .115) than for boys in early adrenarche living in less disadvantaged neighborhoods (low ADI: MZ = .421, DZ = .408). 43 Table S1.4. Unstandardized path and moderator estimates for full and best-fitting genotype x environment models across adrenarche, without BMI percentile regressed out Neighborhood Disadvantage (ADI) – No Convergence of Full Model Family SES Model a c e βxP βyP βzP βxD βyD βzD βxPD βyPD βzPD Full model .797 (.566, 1.029) .392 (-.052, .836) .436 (.296, .575) .010 (-.274, .294) .110 (-.359, .580) .209 (.053, .365) -.378 (-.785, .029) -.836 (-1.397, -.276) -.060 (-.284, .164) .297 (-.225, .819) .619 (-.041, 1.279) -.117 (-.374, .141) Best-fitting .821 (.676, .966) .473 (.224, .723) -.406 (-.473, -.340) — — -.237 (-.329, -.145) -.444 (-.707, -.180) -.949 (-1.290, -.607) — .348 (.096, .601) .764 (.465, 1.064) .176 (.060, .291) Note: Outcome is standardized, log-transformed MTP-ED total score without BMI percentile regressed out. ADI = Area Deprivation Index percentile (higher values indicate greater neighborhood disadvantage); family SES = family socioeconomic status (lower values indicate greater familial disadvantage); a = additive genetic influences at the lowest levels of the moderators; c = shared environmental influences at the lowest levels of the moderators; e = non-shared environmental influences at the lowest levels of the moderators; βxP, βyP, βzP = coefficients for moderation of genetic/environmental variance by adrenarche; βxD, βyD, βzD = coefficients for moderation of genetic/environmental variance by neighborhood disadvantage/family SES; βxPD, βyPD, βzPD = coefficients representing changes in the moderating effects of disadvantage across adrenarche (i.e., the disadvantage x development interaction). 95% confidence intervals of parameter estimates are included in parentheses. Effects significant at p < .05 are bolded. 44 Figures Figure S1.1. Path diagram for the full twin moderation model. Disadvantage = Area Deprivation Index percentile (neighborhood disadvantage), or a factor score comprised of mother’s education level, father’s education level, and combined parental income (family SES); Disordered Eating = standardized, log-transformed Michigan Twins Project Eating Disorder Survey (MTP-ED) total score with or without BMI percentile regressed out; Adrenarche = adrenarche status (0 = early adrenarche, 1 = late adrenarche); A = additive genetic influences; C = shared environmental influences; E = non-shared environmental influences; P1 and P2 = adrenarche status for twin 1 and twin 2; D = disadvantage for the twin pair; μP, μD, a, c, e = intercepts; βP1 = regression coefficient representing the phenotypic association between twin 1’s adrenarche status and their own disordered eating; βP2 = regression coefficient representing the phenotypic association between 45 Figure S1.1 (cont’d) twin 2’s adrenarche status and twin 1’s disordered eating; βD = regression coefficient representing the phenotypic association between disadvantage and twin 1’s disordered eating; βP1D = regression coefficient representing moderation of the phenotypic association between disadvantage and twin 1’s disordered eating by twin 1’s adrenarche status; βP2D = regression coefficient representing moderation of the phenotypic association between disadvantage and twin 1’s disordered eating by twin 2’s adrenarche status; βxP, βyP, βzP = coefficients for moderation of genetic and environmental influences by adrenarche status; βxD, βyD, βzD = coefficients for moderation of genetic and environmental influences by disadvantage; βxPD, βyPD, βzPD = coefficients representing developmental differences in the moderating effects of disadvantage on genetic/environmental influences (i.e., the adrenarche x disadvantage interaction). 46 Figure S1.2. Additive genetic (A), shared environmental (C), and non-shared environmental (E) influences on disordered eating across adrenarche status and family socioeconomic status (SES), without body mass index regressed out. 47