Puberty as a Critical Risk Period for Eating Disorders: A Review of Human and Animal Studies

Abstract

Puberty is one of the most frequently discussed risk periods for the development of eating disorders. Prevailing theories propose environmentally mediated sources of risk arising from the psychosocial effects (e.g., increased body dissatisfaction, decreased self-esteem) of pubertal development in girls. However, recent research highlights the potential role of ovarian hormones in phenotypic and genetic risk for eating disorders during puberty. The goal of this paper is to review data from human and animal studies in support of puberty as a critical risk period for eating disorders and evaluate the evidence for hormonal contributions. Data are consistent in suggesting that both pubertal status and pubertal timing significantly impact risk for most eating disorders in girls, such that advanced pubertal development and early pubertal timing are associated with increased rates of eating disorders and their symptoms in both cross-sectional and longitudinal research. Findings in boys have been much less consistent and suggest a smaller role for puberty in risk for eating disorders in boys. Twin and animal studies indicate that at least part of the female-specific risk is due to genetic factors associated with estrogen activation at puberty. In conclusion, data thus far support a role for puberty in risk for eating disorders and highlight the need for additional human and animal studies of hormonal and genetic risk for eating disorders during puberty.

Historically, puberty has been one of the most frequently discussed risk periods for the development of eating disorders (i.e., anorexia nervosa (AN), bulimia nervosa (BN)) and their symptoms (e.g., body dissatisfaction, weight concerns, dieting, binge eating). Traditional theories of risk focused on the psychosocial effects (e.g., increased body dissatisfaction) of the physical changes associated with puberty (e.g., increased adiposity) and their potential consequences for eating disorder development (Fornari & Dancyger 2003). However, human and animal data are converging to suggest that biological and genetic factors also significantly contribute to eating disorder risk during the pubertal transition. The overarching goal of this review is to evaluate the evidence in support of biological and genetic risk and synthesize these data with existing neuroendocrinological models of puberty’s effects on behavior. Perhaps not surprisingly, most theories focus on gonadal hormones and their influence on sex-differentiated behaviors during and after puberty. This review will be no exception.

The review begins with an overview of the types of eating disorders and symptoms that have been examined in puberty studies. Data in support of puberty as a critical risk period for these phenotypes will then be reviewed, with a particular emphasis on differentiating pubertal status (i.e., pubertal stage at a given point in time) from pubertal timing (i.e., onset of puberty relative to peers, including early, on-time, and late onset) effects. Evidence from human and animal studies suggesting that biological and/or genetic factors contribute to puberty’s effects will then be evaluated and integrated into existing behavioral neuroendocrinological models of sex-differentiated behaviors during puberty.

Eating Disorder Definitions

To date, studies of puberty have primarily focused on AN and BN and their component symptoms. DSM-IV criteria for each disorder (American Psychiatric Association, 2000) are presented in Table 1 . Notably, two subtypes of AN exist, the restricting subtype which is characterized by strict dieting and/or excessive exercise only, and the binge/purge subtype which is characterized by strict dieting and/or excessive exercise as well as binge eating, purging, or binge eating and purging. Thus, despite common misconceptions, AN frequently includes binge eating and purging behaviors commonly thought to be exclusive to BN. Likewise, BN includes behaviors and cognitions that are commonly attributed to AN, including strict dieting and extreme concerns about shape/weight that significantly impact the individual’s self-esteem (see Table 1). To avoid overlap between the two disorders, DSMIV stipulates that individuals with BN cannot be underweight (i.e., their weight must be > 85% of ideal), otherwise they would receive a diagnosis of AN, binge/purge subtype.

Table 1.

Definitions of DSM IV Diagnoses and Eating Disorder Symptoms that have been Investigated in Studies of Puberty.

Diagnoses/Symptoms	Definition
Diagnoses:
Anorexia nervosa (AN)	All of the following are present: Low body weight (i.e., < 85% of ideal for age and height). Fear of becoming fat or of gaining weight. Distortion in the way in which body weight and shape are perceived (i.e., thinks part(s) of body too fat, undue influence of body weight/shape on self-esteem, denial of seriousness of low body weight). Primary amenorrhea or (in post-menarcheal females) loss of menses for ≥3 consecutive months.
Bulimia nervosa (BN)	All of the following are present: Binge eating (i.e., the consumption of a large amount of food in short period of time (i.e., 2 hours) with a loss of control over the binge episode). Use of inappropriate compensatory behaviors including purging (i.e., self-induced vomiting, or abuse of laxatives, diuretics, or enemas) and non-purging (e.g., excessive exercise, fasting) behaviors. Binge eating and inappropriate compensatory behaviors occur, on average, ≥2 times per week for 3 months. Undue influence of body weight and/or shape on self-esteem. The BN symptoms do not occur exclusively during periods of AN.
Eating Disorder Not Otherwise Specified (EDNOS)	Clinically significant syndromes that do not meet criteria for either AN or BN – examples include: A woman who would meet criteria for BN but who engages in binge eating/purging episodes only 1 time per week for 3 months. A woman who would meet criteria for AN except that her body weight is 90% of ideal. Binge eating disorder, i.e., a woman who binge eats but does not regularly engage in compensatory behaviors.
Symptoms:
Body dissatisfaction, weight/shape concerns	Dissatisfaction with the size/shape of one’s body or body parts, and/or preoccupation with weight and a desire to lose weight
Dieting/Weight management	Behavioral attempts to restrict food intake and/or engage in other behaviors to lose weight (e.g., exercise).
Dietary restraint	A cognitive intent to diet (i.e., desire to lose weight, plans to restrict food intake) as well as actual attempts to lose weight through dieting, avoidance of high fat foods, and/or fasting.
Binge eating	See definition above under BN
Purging behaviors	See definition above under BN
Drive for muscularity	Both an intent/desire to increase muscle mass as well as actual attempts to increase muscle mass (e.g., weight lifting, protein shakes).

Both AN and BN are relatively rare disorders, with AN occurring in only .5% of the population and BN occurring in 1-3% (American Psychiatric Association, 2000). However, subthreshold cases of the disorders (see Table 1) are much more common with estimates as high as 10% (Fairburn and Bohn, 2005). These and other symptom presentations (i.e., those with binge eating disorder who binge eat but do not purge) are categorized as eating disorders not otherwise specified (EDNOS) within DSMIV (American Psychiatric Association, 2000). The fact that EDNOS diagnoses are the most frequently encountered eating disorders in clinical (Thomas et al., 2009) and community (Machado et al., 2007) samples has led the DSM 5 Workgroup to consider broadening the criteria for AN and BN to capture these more commonly encountered cases (Walsh, 2009).

The higher prevalence of EDNOS diagnoses also has led to a push for dimensional rather than categorical models of eating disorders and other forms of psychopathology. There is a growing recognition that categorical diagnoses may not accurately “carve nature at its joints” and that dimensional models (i.e., models that characterize patients based on quantitatively measured dimensions (e.g., degree of weight loss; frequency of binge/purge behaviors) rather than yes/no diagnoses) may get us closer to identifying the etiology of psychiatric disorders (Cuthbert, 2005; Krueger and Markon, 2011). Fortunately, because of the rarity of AN and BN, the eating disorders field has focused on dimensions of eating disorders for some time by examining symptoms of eating disorders (e.g., the tendency to binge eat, concerns about weight/shape, dissatisfaction with the size/shape of body parts) that are known to predate the onset of the full disorders (Jacobi et al., 2004) and cut across diagnostic boundaries of AN and BN. This has been particularly true of studies examining puberty that must account for the fact that full syndromal AN and BN are very rare before puberty (see review below). Major dimensions of eating disorders that have been examined in studies of puberty are described in Table 1.

One final comment on eating disorder definitions is in order. Eating disorders are the most sex-differentiated disorders in our psychiatric nomenclature, as the female to male ratio ranges from 4:1 (Hudson et al., 2007) all the way up to10:1 (American Psychiatric Association, 2000). Sex differences extend to the symptoms in Table 1 and have contributed to views that gonadal hormones may contribute to risk for eating disorders during puberty. This point will be further discussed later in the review, but suffice it to say that studies examining puberty’s influence on eating disorder risk have focused more on female than male subjects.

Associations between Pubertal Status, Pubertal Timing, and Eating Disorders

Data over the last four decades have confirmed that puberty is a significant risk period for the development of eating disorders and eating disorder symptoms in girls. These studies have examined the frequency of eating disorders and their symptoms in terms of pubertal status (e.g., pre-puberty versus post-puberty) as well as pubertal timing (i.e., early, on-time, versus late maturers) ¹. Most studies have examined these processes cross-sectionally, examining how pubertal status and/or timing at one point in time predict eating disorders and their symptoms. Although cross-sectional studies are important for understanding proximal correlates of the disorder, longitudinal/prospective studies are needed to ensure that pubertal status/timing are risk factors for eating disorders even after all girls “catch up” and have completed their pubertal development. Persistence of pubertal status and timing effects can be examined via: 1) longitudinal/prospective studies that examine whether pubertal status/timing effects at time 1 predict later risk at time 2, particularly when time 2 occurs in post-puberty; and/or 2) retrospective studies that assess whether recalled markers of pubertal development (e.g., menarche) occurred earlier or later than peers and/or population norms. The later studies are obviously less rigorous, as they rely on retrospective reports that are prone to memory biases (Coughlin, 1989). Nonetheless, rare disorders like eating disorders can be difficult to investigate in prospective studies due to the small number of expected cases, and so most studies of clinical eating disorders rely on retrospective reports.

Notably, it remains an open question whether persistence of status effects reflect processes that are independent of pubertal timing, i.e., since all girls eventually achieve advanced pubertal development, longitudinal effects of pubertal status may simply reflect the timing of these processes (i.e., early or late) at the time of initial assessment. These points highlight the fact that although persistence of puberty’s effects across time provides strong support for pubertal processes as risk factors, they cannot disentangle pubertal status from pubertal timing as the critical mechanisms underlying the effects.

Despite these caveats, overall, findings from cross-sectional and longitudinal studies are remarkably consistent in suggesting that pubertal status and timing effects persist across time and are associated with a significantly increased risk of developing eating disorders and their component symptoms in girls, although findings are much less clear in boys.

Girls

Eating Disorder Diagnoses

Two types of studies have examined the effects of puberty on risk for AN, BN, and EDNOS. The first indirectly examined the influence of puberty by exploring rates of eating disorders in children versus adolescents. These studies consistently show a significantly increased prevalence of all forms of eating disorders in individuals in mid-late adolescence as compared to pre-adolescence (Bulik, 2002). Indeed, pre-pubertal onset of eating disorders is rare (Bulik, 2002), and mean ages of onset for the disorders (i.e., 15–19 years for AN and BN) (American Psychiatric Association, 2000; Favaro et al., 2009; Favaro et al., 2003) reflect a post-pubertal bias in onset.

Nonetheless, several psychosocial and biological events take place between childhood and adolescence, leaving a specific role for puberty unclear. The second type of study directly examined rates of eating disorders by pubertal status and timing. More of these investigations have examined pubertal timing than pubertal status, likely due to the difficulty of capturing new onset cases during pubertal development. Studies thus far consistently suggest significant associations between pubertal status/timing and both BN and EDNOS (see Table 2). Girls who are at more advanced stages of pubertal development have increased rates of both disorders, even after controlling for age. Likewise, with few exceptions, a higher percentage of early maturers are among those diagnosed with BN as compared to women without eating disorders. This pattern of results is observed in combined samples of AN and BN women as well (see Table 2).Importantly, although two prospective studies failed to find an effect of pubertal status (Killen et al., 1994) or timing (Stice et al., 2001) on risk for BN, findings from retrospective studies clearly show that the effects of early pubertal timing on BN persist across late adolescence (Biederman et al., 2007; Graber et al., 1997; Kaltiala-Heino et al., 2003; Kaltiala-Heino et al., 2001; Ruuska et al., 2003) into young (Corcos et al., 2000; Fairburn et al., 1997; Romans et al., 2001) and middle (Romans et al., 2001) adulthood.

Table 2.

Summary of Studies Examining the Effects of Advanced Pubertal Status and Early Pubertal Timing on Eating Disorders and Their Symptoms.

	Advanced Pubertal Status		Early Pubertal Timing
Study	Yes	No	Yes	No
	FEMALES
Eating Disorder Diagnoses
AN
Crisp (1970)			X
Fairburn et al. (1999)				X
Favaro et al. (2003)				X
Nicholls and Viner (2009)			Xa
Nielsen (1985)				X
Mangweth-Matzek et al. (2007)				X
Ruuska et al. (2003)				X
Stice et al. (2001)				X
BN
Corcos et al. (2000)			X
Fairburn et al. (1997)			X
Fairburn et al. (1999)			X
Favaro et al. (2003)				X
Kaltiala-Heino et al. (1999)		X
Kaltiala-Heino et al. (2001)			X
Kaltiala-Heino et al. (2003)			X
Killen et al. (1992)	X
Killen et al. (1994)		X
Mangweth-Matzek et al. (2007)				X
Mousa et al. (2010)	X
Ruuska et al. (2003)			X
Stice et al. (2001)				X
AN and BN Combined
Biederman et al. (2007)	X
Favaro et al. (2003)	X
Graber et al. (1997)	X
Romans et al. (2001)	X
EDNOS
Mousa et al. (2010)	X
Eating Disorder Symptoms
Global Measures
Attie and Brooks-Gunn (1989)	X
Baker et al. (2012)			X
Baker et al. (in press)			X
Berger et al. (2009)			X
Culbert et al. (2009)	X
Culbert et al. (submitted)	X
Graber et al. (1994)			X
Hayward et al. (1997)	X		X
Johnson-Sabine et al. (1988)				X
Kaluski et al. (2008)			X
Keel et al. (1997)	X
Klump et al. (2003)	X
Klump et al. (2007b)	X
Leon et al. (1993)				X
Leon et al. (1995)				X
Levine et al. (1994)		X
The McKnight Investigators (2003)		X
Wichstrom (1995)			X
Body Dissatisfaction and Weight/Shape Concerns:
Ackard and Peterson (2001)			X
Barker and Galambos (2003)		X
Berger et al. (2009)	X
Blyth et al. (1985)			X
Brooks-Gunn et al. (1989)			X
Cattarin and Thompson (1994)		X
Cauffman and Steinberg (1996)	X
Cotrufo et al. (2007)	X
Culbert et al. (submitted)	X
de Castro and Goldstein (1995)	X
Fabian and Thompson (1989)	X
Folk et al. (1993)		X
Fonesca and Matos (2011)			X
Hermes and Keel (2003)	X
Jaszyna-Gasior et al. (2009)				X
Keski-Rahkonen et al. (2005)	X
Koff and Rierdan (1993)	X
Levine et al. (1994)		X
McCabe et al. (2001)	X
McCabe and Ricciardelli (2004)	X
McHale et al. (2001)	X
Michaud et al. (2006)	X
Muris et al. (2005)		X
Niven et al. (2007)	X
Ohring et al. (2002)	X
Oinonen and Bird (2012)	X
O'Dea and Abraham (1999a)	X
O'Dea and Abraham (1999b)	X
Petersen et al. (1994)			X
Rodriguez-Tome et al. (1993)		X
Siegel et al. (1999)	X
Stormer and Thompson (1996)				X
Striegel-Moore et al. (2001)			X
Suka et al. (2005)	X
Thompson and Chad (2000)	X
Tremblay and Lariviere (2009)	X
Zehr et al. (2007)			X
Dieting and Weight Management
Abraham and O'Dea (2001)	X
Baker et al. (in press)	X
Cotrufo et al. (2007)	X
Field et al. (1999b)	X
Gralen et al. (1990)	X
Johansson and Ritzen (2005)			X
Levine et al. (1994)	X
McCabe et al. (2002)	X
McCabe and Ricciardelli (2004)	X
Harden et al. (2012)			X
Shisslak et al. (1998)		X
Tremblay & Lariviere (2009)	X
Dietary Restraint
de Castro & Goldstein (1995)	X
Patel et al. (2011)	X
Zehr et al. (2007)			X
Binge Eating
Cattarin & Thompson (1994)		X
Combs et al. (2011)
Cotrufo et al. (2007)	X
Field et al. (1999b)	X
Gralen et al. (1990)	X
Koff & Rierdan (1993)		X
McCabe & Ricciardelli (2004)				X
Reichborn-Kjennerud et al. (2004) (Reichborn- Kjennerud et al., 2004)			X
Striegel-Moore et al. (2001)				X
Tenconi et al. (2006)			X
Thompson and Chad (2000)	X
Purging
Combs et al. (2011)	X
Field et al. (1999b)	X
Field et al. (1999a)	X
	MALES
Eating Disorder Diagnoses
AN
Nielsen (1985)			X
BN
Kaltialo-Heino (1999)		X
Kaltialo-Heino (2003)			X
Eating Disorder Symptoms
Global Measures
Baker et al. (in press)	X
Cortufo et al. (2007)	X
Culbert et al. (submitted)		X
Domine et al. (2009)			X
Keel et al. (1997)		X
Leon et al. (1995)				X
Reichborn-Kjennurud et al. (2004)				X
Body Dissatisfaction and Weight/Shape Concerns
Barker & Galambos (2003)		X
Cortufo et al. (2007)	X
Folk et al. (1993)	X
Keski-Rahkonen et al. (2005)		X
Michaud et al. (2006)			X
McCabe et al. (2001)	X
McHale et al. (2001)	X
Siegel et al. (1999)			X
Suka et al. (2005)	X
Zehr et al. (2007)			X
Dieting and Weight Management
Baker et al. (in press)		X
Cortufo et al. (2007)	X
McCabe et al. (2002)	X
McCabe & Ricciardelli (2004)			X
Suka et al. (2005)	X
Dietary Restraint
Zehr et al. (2007)			X
Binge Eating
Cortufo et al. (2007)	X
Pearson et al. (2010)		X
Reichborn-Kjennerud et al. (2004)				X
Zehr et al. (2007)			X
Purging
Field et al. (1999b)	X
Pearson et al. (2010)		X
Drive for Muscularity
McCabe & Ricciardelli (2004)	X
Ricciardelli and McCabe (2003)		X
Smolak and Stein (2006)		X

Note. The “X” indicates whether the study did (yes) or did not (no) find a significant effect of advanced pubertal status or early pubertal timing on eating disorders/symptoms. Some studies included males and females in the same sample but did not examine them separately in analyses. In all of these cases, sex was a significant predictor of eating problems, with females being at significantly increased risk for eating pathology relative to males. Consequently, the results of these studies are included in table entries for females but not males.

^aEffect was trend-level (p = .06) and became non-significant when included in models that controlled for sex and socioeconomic status.

Results for AN only have been more mixed. Although some studies find early maturers are at increased risk for AN (see Table 2), even when assessed in adulthood (Crisp, 1970; Nicholls and Viner, 2009), many failed to find an effect, or found a higher prevalence of late maturers in AN samples (Root et al., 2011). No studies have examined differential rates of AN across pubertal status. The general lack of support for pubertal effects on AN may be due to the relatively small number of studies conducted and/or the difficulty of studying early-onset AN that has a much lower prevalence (i.e., .5%) than BN (i.e., 1-3%) or EDNOS (i.e., ~10%) (American Psychiatric Association, 2000). In addition, because extreme dieting frequently begins months/years before AN onset and can forestall pubertal development (Lantzouni et al., 2002; Roze et al., 2007), disentangling the effects of the disorder on puberty versus the effects of puberty on the disorder can be difficult. This tends to be less of an issue for BN and EDNOS where the dieting is less extreme and other core symptoms (e.g., binge eating, purging) tend not to disrupt pubertal processes.

Eating Disorder Symptoms

Studies have investigated component symptoms of eating disorders (e.g., body dissatisfaction, dieting; see Table 1) as well as global measures that sum across items tapping body dissatisfaction/weight concerns, binge eating, and purging/non-purging behaviors. These global measures assess the overall likelihood of developing eating disorders given that these symptoms are known risk factors for AN and BN (Jacobi et al., 2004). Nevertheless, given that global measures frequently do not include low body weight as an indicator, these measures likely tap BN or EDNOS pathology more than AN per se.

Most studies have found significant effects of both pubertal status and timing on global as well as specific eating disorder symptoms (see Table 2). With few exceptions, studies tend to find higher levels of symptoms in girls at more advanced stages of pubertal development, particularly when comparing pre-early puberty to mid-late puberty (e.g., (Culbert et al., 2009; Klump et al., 2007a). Pubertal timing results largely replicate these results where the highest rates of eating disorder symptoms are found in early maturing as compared to on-time and late-maturing girls (see Table 2). Findings have been particularly robust for body dissatisfaction and weight/shape concerns where studies are numerous and negative findings have been relatively rare (for interesting exceptions, see (Brooks-Gunn and Warren, 1988) and (Slap et al., 1994) showing positive associations between breast development and body satisfaction).

Although most studies of both pubertal timing and status have been cross-sectional in nature, some prospective and retrospective studies have examined persistence across time. With some exceptions (Leon et al., 1993; Leon et al., 1995; McCabe and Ricciardelli, 2004), pubertal timing effects have been observed in late (Baker et al., 2012; Graber et al., 1994; Harden et al., 2012; Hayward et al., 1997; Ohring et al., 2002; Oinonen and Bird, 2012) as well as young (Ackard and Peterson, 2001; Graber et al., 1994; Keski-Rahkonen et al., 2005; Ohring et al., 2002; Petersen et al., 1994; Reichborn-Kjennerud et al., 2004; Striegel-Moore et al., 2001; Tenconi et al., 2006; Zehr et al., 2007) and middle (Johansson and Ritzen, 2005) adulthood. Several studies were prospective and found that timing effects assessed during puberty persisted one to 30 years after initial assessment. Prospective studies of pubertal status have been fewer in number (Attie and Brooks-Gunn, 1989; Cattarin and Thompson, 1994; Field et al., 1999; The McKnight Investigators, 2003; Keel et al., 1997; McCabe and Ricciardelli, 2004), but many found that advanced pubertal development at the time of initial assessment predicted eating disorder symptoms 8 months to 3 years later (Field et al., 1999; Keel et al., 1997; McCabe and Ricciardelli, 2004).

Boys

Far fewer studies have examined the effects of puberty on eating disorders in males (see Table 2) and to date, findings are quite mixed. Similar to results in girls, 22 studies found that early maturing boys and/or those at advanced stages of puberty had higher rates of AN, BN, and eating disorder symptoms (see Table 2). Results have been most consistent for body dissatisfaction and weight/shape concerns where several studies found significant effects of pubertal status and timing.

However, 13 studies failed to find significant effects of pubertal status/timing on BN or disordered eating symptoms (see Table 2), four studies found that late maturers had higher rates of body dissatisfaction (McCabe and Ricciardelli, 2004; Michaud et al., 2006; Petersen et al., 1994; Siegel et al., 1999), and five studies found that advanced pubertal development was associated with improved body image (McCabe et al., 2001; Suka et al., 2005), enhanced self-concept (O'Dea and Abraham, 1999a, 1999b), and/or increased perceptions of physical attractiveness and/or decreased social pressures around eating (Muris et al., 2005).

Results from prospective and retrospective studies examining persistence across time were also equivocal. While four studies found early maturers to be at increased risk for eating disorder symptoms into late adolescence/young adulthood (Domine et al., 2009; McCabe and Ricciardelli, 2004; Nielsen, 1985; Zehr et al., 2007), three studies failed to find an effect of early pubertal timing on eating disorder risk during these later time periods (Keski-Rahkonen et al., 2005; Leon et al., 1995; Reichborn-Kjennerud et al., 2004). Only one prospective study of pubertal status found persistence of effects (Baker et al., 2012) while others found no significant associations between advanced pubertal development at time 1 and later risk for eating disorders at time 2 (Keel et al., 1997; Ricciardelli and McCabe, 2003).

Summary

Overall, results are relatively consistent in girls showing significant effects of both pubertal status and timing on most eating disorder phenotypes. The strongest effects have been observed for BN and its symptoms, including body dissatisfaction, weight/shape concerns, binge eating, and purging. Findings for AN have been much less consistent, although studies are fewer in number and the trajectory of the disorder makes it more difficult to detect pubertal effects. Nonetheless, the consistency of results for BN and eating disorder symptoms is striking given study differences in the eating constructs measured (e.g., weight concerns versus binge eating) and pubertal measures (i.e., self-reports of secondary sex characteristics, physician ratings of development, age at menarche). These data provide strong support for a role for puberty in the etiology of these clinical phenomena in humans.

Findings for boys have been much more mixed. Several studies show no effect of pubertal status/timing on eating disorders and their symptoms, and persistence of associations across time has not been established. The lone exception might be body dissatisfaction and weight/shape concerns where the majority of studies show positive effects of both status and timing. However, only two studies examined persistence across time, and their findings were discrepant (i.e., Zehr et al. (2007) found persistence while Keski-Rahkonen et al. (2005) did not). Moreover, advanced pubertal development has been associated with positive as well as negative body image in boys (see above). Overall, the bulk of the data suggest that puberty is likely to play a much larger role in eating disorder risk in girls than boys and that sex-specific processes during puberty may differentially contribute to eating disorders in females versus males.

Mechanisms underlying Puberty’s Effects on Eating Disorder Risk

Prevailing theories of risk have posited that significant pubertal status and timing effects in girls are due to the physical changes of puberty and their effects on body satisfaction, self-esteem, and mood (Bulik, 2002; Fornari and Dancyger, 2003). These theories emphasize the role of body dissatisfaction in causing increased dieting that then leads to AN and BN. Early maturers are thought to be at particular risk given that they experience these physical changes earlier than their peers and may therefore experience even more body dissatisfaction than their developmentally on-time counterparts. Girls are believed to be at increased risk relative to boys given that pubertal physical changes in girls (i.e., increased adiposity) move them away from beauty ideals, while physical changes in boys (i.e., increased muscle mass) move them closer to their beauty ideals (Bulik, 2002; Fornari and Dancyger, 2003).

Although data testing the full psychosocial model are relatively sparse, some studies have supported its major tenants, at least in regard to the association between pubertal status, body dissatisfaction, and dieting (Attie and Brooks-Gunn, 1989; Bulik, 2002; Gralen et al., 1990). However, no studies have examined whether these processes ultimately lead to the development of AN or BN, and researchers are beginning to consider biological/genetic influences as well. The pubertal activation of ovarian hormones and their substantial role in sex-differentiated behaviors have made this system a prime candidate in the hunt for biological factors underlying pubertal risk. Emerging data from both human and animal studies are providing support for these hypotheses.

Human Studies

Twin Studies of Adolescence

Initial hints that biological and/or genetic factors might be important for pubertal risk in girls came from twin studies examining developmental differences in genetic influences on eating disorder symptoms during adolescence. These studies examined global measures (assessing body dissatisfaction, weight concerns, binge eating, and compensatory behaviors) as well as specific eating disorder symptoms (e.g., body dissatisfaction, weight concerns) and found significant increases in genetic influences across adolescence, where there was essentially no genetic influence (i.e., 0% heritability) on eating disorder symptoms in pre-adolescent female twins (age 11), but significant genetic effects (≥50% heritability) in twins during late adolescence (age 17) (Klump et al., 2010a). Two follow-up studies confirmed that developmental changes emerged between pre-adolescence and middle adolescence (Klump et al., 2007a; Klump et al., 2010a), as increases in genetic effects (from 0% to ~50%) were only observed between early and middle adolescence (i.e., between ages 11 and 14), while the magnitude (~50% heritability) and type of genetic factors (Klump et al., 2007a) remained constant from mid-adolescence (i.e., age 14) into middle adulthood (up to age 41). Developmental differences were observed in both cross-sectional (Klump et al., 2010a; Klump et al., 2000) and longitudinal (Klump et al., 2007a) studies, suggesting that the effects were robust and reflected within-twin pair shifts in etiologic influences across adolescence. The pattern of results remained unchanged after controlling for body mass index (BMI) and mood symptoms (e.g., anxiety) (Klump et al., 2007a; Klump et al., 2000), suggesting that the developmental differences were specific to eating disorder symptoms rather than secondary changes due to developmental differences in correlated phenotypes. Finally, the lack of changes in magnitude or type of genetic influences after mid-adolescence further highlighted processes specific to the early adolescent period as key to genetic (and likely phenotypic) risk for eating disorders.

Twin Studies of Puberty

Given that puberty typically begins between pre-adolescence and middle adolescence in girls, puberty was an obvious candidate for studies examining mechanisms underlying age differences in genetic risk. Twin studies took up the cause and began investigating whether shifts in genetic effects occur across puberty rather than across adolescent ages per se. Instead of dividing twins by age, these studies examined differences in genetic risk in girls who varied in their pubertal status and/or timing, as assessed via secondary sex characteristics (e.g., breast development, hair growth, onset of menses).

Most studies of pubertal status used a 2-group analysis that compared twins in pre-early puberty to those in mid-post puberty (Culbert et al., 2009; Klump et al., 2003), although one study examined changes across the full range from pre-puberty to post-puberty (Klump et al., 2007b). Regardless of the grouping method used, substantial differences in genetic effects across pubertal status were observed, such that genes accounted for 0% of the variance in eating disorder symptoms (e.g., overall levels, body dissatisfaction, weight concerns) in pre-early puberty, but ~50% of the variance during mid-puberty and beyond. Increases in genetic effects were found to be linear across puberty (Klump et al., 2007b), although changes were complete by the onset of menses (Culbert et al., 2009; Rowe et al., 2002; Silberg and Bulik, 2005). All studies of puberty were cross-sectional, not longitudinal; however, pubertal differences in genetic effects persisted even after controlling for age and BMI (Culbert et al., 2009; Klump et al., 2003), and there were no changes in the degree of genetic influence from post-puberty through late adolescence (i.e., heritability was ~50% from post-puberty on) (Klump et al., 2003). These data strongly suggest that puberty accounts for age differences observed previously and that there may be genetic links between processes occurring during puberty and eating disorder symptoms in girls.

Two recent twin studies of pubertal timing confirmed that this is the case. Baker et al. (in press) and Harden et al. (2012) directly examined these genetic links by investigating shared genetic risk factors between eating disorder symptoms (e.g., overall eating disorder symptoms, dieting) and age at menarche or early pubertal timing, respectively. Associations between these menarche variables and eating disorder symptoms were due to a shared set of genetic risk factors contributing to both phenotypes. Importantly, after controlling for these shared genetic risk factors, phenotypic associations between early pubertal timing and dieting were eliminated, suggesting that genetic risk factors (rather than environmental influences) entirely account for the co-occurrence of early menarche and eating disorder behaviors in girls (Harden et al., 2012).

Overall, data from twin studies provide strong support for puberty as a period of significant genetic risk for eating disorders. Genetic influences appear to become “activated” during the pubertal period and may increase phenotypic risk for the disorders in girls. These results are leading to a re-thinking of traditional theories of pubertal risk focusing on environmental mechanisms (see above) as well as renewed interest in ovarian hormones (most notably, estrogen) as potential risk factors that are activated during puberty and may drive increases in, and genetic risk for, eating disorder symptoms during the critical pubertal period.

Twin Studies of Estrogen

Researchers cite several lines of evidence in support of a focus on estrogen (Klump et al., 2010b). For example, estradiol release during puberty closely mimics pubertal activation of genetic effects on eating disorder symptoms. Although estrogen is present at low levels in girls prior to puberty, the activation of the GnRH pulse generator during pre-puberty leads to linear increases in ovarian secretion of estradiol across puberty (Wilson et al., 1998). This pattern of release is similar to the increases in genetic influences on disordered eating across puberty (Klump et al., 2007b) and is in contrast to the activation of the other primary ovarian hormone, progesterone, which increases only after first ovulation. Given twin study data showing increases in genetic effects around mid-puberty (Culbert et al., 2009; Klump et al., 2003) and before menarche, progesterone is not viewed as a major player in phenotypic and genetic risk for disordered eating during the pubertal period (Klump et al., 2010b).

Estrogen also drives many of the pubertal changes in secondary sex characteristics (e.g., breast development) (Wilson et al., 1998) that have been shown to moderate genetic influences on eating disorder symptoms. Indeed, developmental twin studies showing increased genetic effects (see above) relied on assessments of these secondary sex characteristics for assessing pubertal development. The fact that estrogen causes the physical changes that have been found to moderate genetic influences on DE has led to suggestions that estrogen might be one of the mechanisms underlying puberty’s effects on genetic risk (Klump et al., 2010b).

Finally, researchers have highlighted the fact that one of the primary functions of estrogen and other steroid hormones is to regulate gene transcription, and thereby protein synthesis, within the central nervous system (Ostlund et al., 2003; Wilson et al., 1998). Some of the genes regulated by estrogen have been found to be significantly associated with eating disorders, particularly those in the serotonin system (Hildebrandt et al., 2010; Ostlund et al., 2003) and those associated with brain-derived neurotrophic factor (BDNF) (Klump & Culbert, 2007). Researchers have proposed (Klump and Culbert, 2007; Klump et al., 2010b) that increases in estrogen at puberty may contribute to genetic risk via the hormones’ effects on the production of these important neurotransmitters/neurotrophins, their receptors, and/or their signal transduction mechanisms. Moreover, variants of ovarian hormone receptor genes (e.g., estrogen receptor beta) (Klump and Gobrogge, 2005) may increase risk for binge eating by causing different patterns of gene regulation by ovarian hormones, leading to distinct cellular and behavioral responses. In these scenarios, individual differences in the production of ovarian hormones are not of consequence – the key individual difference variable is the presence or absence of susceptibility alleles that are regulated by ovarian hormones. These individual differences would become evident after ovarian hormone activation (i.e., after puberty).

To date, only one pilot study has investigated this possibility, and it did so indirectly using a twin study design. Klump et al. (2010) compared the magnitude of genetic effects on disordered eating in twins with high (i.e., above the median value) versus low (i.e., below the median value) estradiol levels during puberty. Results suggested significant differences in genetic influences on overall eating disorder symptoms by estradiol levels, with no genetic influence in twins with low estradiol levels but significant genetic effects in twins with high estradiol levels. Findings remained unchanged after controlling for age and BMI, suggesting direct effects of estrogen on genetic risk for eating disorders during puberty in girls. Most importantly, results remained unchanged after controlling for changes in secondary sex characteristics, suggesting that differences in estradiol account for differences in genetic effects across puberty in previous studies.

Clearly, additional research is needed to replicate these results, particularly given the small sample size (N = 99 twin pairs) for this pilot twin project. However, corroborating data come from twin studies of males examining the effects of puberty on genetic risk for disordered eating (Klump et al., 2012). Boys do not experience increases in estrogen during puberty, as their pubertal development is driven primarily by increases in testosterone. If pubertal increases are present in boys, then factors other than (or in addition to) estrogen may drive increases in phenotypic and genetic effects in girls. In contrast to findings in girls, results showed no changes in genetic effects on overall levels of disordered eating in male twins across pre-early puberty, mid-late puberty, or young adulthood. The heritability remained constant at ~50% of the variance in all groups (Klump et al., 2012). These findings strongly suggest that pubertal increases in genetic influences are specific to girls and may be related to the estrogen effects described above.

Summary

Findings from twin studies provide support for a role for genetic factors in pubertal risk for eating disorders in girls. Dramatic increases in genetic effects have been observed across adolescence that appear to be due to estrogen activation during puberty. These changes are not present in boys, providing further, indirect support for pubertal hormone influences on risk for eating disorders in girls. Finally, emerging data confirm that associations between early pubertal timing and key disordered eating symptoms are due to genetic (rather than environmental) risk factors.

Although these studies do not rule-out psychosocial explanations for pubertal risk, they call into question the predominance of these risk factors in etiologic models of risk. Harden et al. (2012) showed rather convincingly that phenotypic associations between early pubertal timing and key eating disorder symptoms (e.g., dieting) are eliminated when within-family, genetic risk factors are controlled. Likewise, twin studies of puberty that control for age and BMI, and the twin study of estrogen that controlled for age, BMI, and the physical changes of puberty, suggest direct effects of puberty and estrogen on genetic and phenotypic risk.

Animal Studies

Nonetheless, it is difficult to control for all potentially relevant psychosocial risk factors in human studies, as the level of experimental control is necessarily limited. Animal studies can be helpful in these cases for teasing apart biological/genetic from psychosocial/environmental risk and identifying causal mechanisms. To date, only one study has examined the effects of puberty on eating disorder risk in animals by investigating whether pubertal increases in eating disorder symptoms observed in girls are present in female animals as well. Klump et al. (2011) used the binge eating resistant/binge eating prone rat model of binge eating (Boggiano et al., 2007) to determine if binge eating proneness (i.e., the tendency to consistently consume large amounts of palatable food (i.e., food that is high in fat and/or sugar) in a short period of time) emerges during puberty in female rats. The focus on binge eating was predicated on the fact that binge eating is a key eating disorder phenotype that is disrupted in both AN and BN and is easier to model in animals than the cognitive features of the disorders (Klump et al., 2011b).

Klump et al. (2011) found that differences in palatable food intake between binge eating prone and binge eating resistant rats (i.e., rats who consistently consume smaller amounts of palatable food in a short period of time) in adulthood did not emerge until puberty. In pre-puberty, there were no significant differences in PF consumption between the two groups of rats, but significant differences emerged during mid-late puberty and persisted into adulthood. Importantly, the two groups did not differ in their chow intake or body weight at any stage of development, suggesting that the pubertal effects were specific to palatable food intake and binge eating phenotypes. Findings were replicated across two independent samples of rats.

Notably, these findings are unlikely to be due to environmental or psychosocial influences, as female rats do not experience key psychosocial/psychological risk factors (i.e., increased body dissatisfaction) that are present in humans. However, data showing that these effects are due to ovarian hormones (and estrogen in particular) are lacking, as this observational study did not manipulate ovarian hormones. Nonetheless, indirect data from animal studies of food intake support a role for ovarian hormones in these effects. These studies date back more than 40 years and convincingly show that estrogen inhibits food intake in adult female rats and other species, but these causal effects do not emerge until puberty (Asarian and Geary, 2006; Swithers et al., 2008; Wade, 1976). Before puberty, exogenously administered estradiol does not have any effect on food intake in female rats (Asarian and Geary, 2006). The same pubertal pattern has been observed for food intake patterns that are typical of binge eating episodes (e.g., preference for sweet foods); female rats’ preference for sweet solutions (relative to males) are mediated by estrogen in adulthood, but these sex differences do not emerge until puberty (i.e., male and female rats do not differ in preference for sweet solutions prior to puberty) (Wade, 1976).

Summary, Integrative Theories and Future Directions

In summary, data from humans are clear in showing that both advanced pubertal status and early pubertal timing significantly increase risk for eating disorders (particularly BN and EDNOS) and their symptoms in girls. Findings further suggest that these effects are at least partially genetically mediated via the effects of estrogen during puberty. Animal studies are fewer in number, but results thus far support a role for biological factors (and possibly estrogen) in pubertal risk for binge eating phenotypes. Although additional animal and human studies are clearly needed (see below), findings thus far are sufficient to begin integrating current data with existing hormonal theories of pubertal risk for behavioral phenotypes.

Integrative Theories

An ideal framework for conceptualizing pubertal and hormonal risk for eating disorders is the activational and organizational theory of hormone action on sex-differentiated behavior. This theory posits two types of gonadal hormone effects on behavior: activational (i.e., transient effects that depend on the presence/absence of hormone) versus organizational (i.e., permanent effects that persist beyond the period of exposure to hormone and program later activational responses to hormone) influences. To date, no studies have directly compared the influence of these two hormone effects for eating disorders, although data support the possible importance of both for eating disorder risk during puberty. For example, estrogen exhibits robust activational effects on eating disorder symptoms in adulthood, where it is inversely correlated with food intake (Buffenstein et al., 1995), binge eating (Edler et al., 2007; Klump et al., 2008; Klump et al., 2013), and weight concerns (Racine et al., 2012) in women, and food intake (Asarian and Geary, 2006), binge eating (Babbs et al., 2011; Yu et al., 2008), and preference for sweet foods (Wade, 1976) in adult female animals. These data would suggest that increases in phenotypic and genetic risk for eating disorders by pubertal status may be due to increased levels of circulating estrogen starting early in puberty and continuing through adulthood.

However, if estrogen’s effects were purely activational in nature, we would expect significant phenotypic associations between estradiol levels and eating disorder symptoms in girls during puberty, as levels of eating disorder symptoms should correlate with circulating levels of estradiol. This association has not been observed, as correlations between estradiol levels and eating disorder symptoms (e.g., body dissatisfaction, binge eating) during puberty have been small and non-significant (Nottelmann et al., 1986; Slap et al., 1994; Warren and Brooks-Gunn, 1989). This is in contrast to significant and moderate associations between disordered eating symptoms and estradiol levels in adult women (Edler et al., 2007; Klump et al., 2008; Klump et al, 2013; Racine et al., 2012).

These differential phenotypic associations across development fit nicely with the organizational hypothesis of gonadal hormone action during puberty which predicts: 1) hormones organize behavior during puberty through permanent changes in brain structure/function; and 2) these changes organize the brain to respond to circulating levels of hormones in adulthood which activate and/or influence the expression of behavior (Sisk and Zehr, 2005). Indeed, when hormone organizational effects are present, little-to-no phenotypic association between hormones and behavior is observed before/during puberty, whereas phenotypic associations are present in adulthood (Sisk and Zehr, 2005). Another key component of this theory is that organizational effects cause permanent changes in behavior that persist beyond the initial period of organization and are relatively immutable to hormone manipulations in adulthood. The pubertal timing effects described above fit with this aspect of the theory, as several studies have found that early maturers report elevated levels of eating disorder symptoms well into adulthood. New animal data provide additional support for this possibility, as the binge prone phenotypes that emerge during puberty in rats are immutable to ovarian hormone manipulations in adulthood, i.e., they persist in adulthood even after the removal of the primary source of estrogen via ovariectomy (Klump et al., 2011a). Thus, they are permanent phenotypes that are organized at puberty and continue into adulthood. Finally, in terms of general food intake, estradiol treatment fails to alter chow intake in female animals that are ovariectomized prior to puberty, whereas predictable changes in food intake are observed with exogenous estradiol treatment in animals ovariectomized in adulthood (Asarian and Geary, 2006; Wade, 1976). Thus, at least for food intake, the presence of ovarian hormones during puberty is required for the activational effects of estrogen on food intake in adulthood. Pubertal ovarian hormones appear to organize the brain in order to respond to the activating effects of estrogen on food intake (and possibly eating disorder symptoms) in adulthood.

Future Directions

Although still in its infancy, research on biological and genetic influences on pubertal risk for eating disorders has made great strides in confirming the likely presence and importance of these effects, as well as highlighting several areas in need of future research. On the phenotypic side, it will be important for future studies to disentangle the long-term effects of pubertal status from pubertal timing. As noted previously, it is currently unknown whether the persistence of status effects reflects an actual effect of pubertal status at the time of initial assessment or the relative timing of that development in relation to peers. Studies could disambiguate these processes by comparing the longitudinal effects of initially assessed pubertal status (i.e., pre, mid-, late-puberty) to the effects of pubertal timing (i.e., early versus late timing, defined using within-sample distributions of pubertal status at the time of initial assessment and/or population norms). Such data could clarify potential differences in the proximal versus the long-term effects of pubertal status and timing on eating disorder risk.

Much more data are needed on pubertal status/pubertal timing effects on AN and its symptoms. This area can be most fruitfully tackled using a combination of animal and human studies that can maximize inferential power as well as ecological validity. It would be helpful to examine symptoms that are specific to AN (i.e., extreme weight loss and dieting) rather than those common to both AN and BN (i.e., body dissatisfaction, weight concerns) to determine if puberty’s effects extend to the AN spectrum as well. Investigators are currently focusing on the activity-based anorexia (ABA) animal model (Aoki et al., 2012) and examining whether the phenotype emerges during puberty and the neural substrates underlying pubertal risk. This line of work is likely to produce the type of AN-specific data that is currently lacking in the field.

Likewise, more studies are needed on the effects of puberty on binge eating disorder (BED) and EDNOS. BED will be a new disorder in DSM 5, and EDNOS diagnoses are of significant interest due to their increased prevalence in the population (relative to AN and BN). To some extent, the dimensional approach of most studies ameliorates concerns about the lack of data, as key symptoms contributing to these diagnoses (e.g., binge eating, dieting) show the pubertal phenotypic and genetic effects described herein. Nonetheless, given the later age of onset of BED and a need to more clearly demarcate similarities/differences between this disorder and other well-established eating disorders, future phenotypic and genetic studies should examine puberty’s effects on BED/EDNOS and their early precursors (e.g., loss of control over eating) to determine the role of puberty in their development.

With regard to genetic and hormonal mechanisms, more twin studies of estrogen are clearly needed to replicate initial pilot data and confirm a role for estrogen in pubertal emergence of phenotypic and genetic risk for eating disorders. Longitudinal studies would go a long way in this regard, as these studies could determine whether within-person changes in phenotypic and genetic risk across puberty are due to estrogen activation. Much like studies of puberty, these studies should more clearly differentiate status versus timing effects to determine if estrogen levels alone, or early versus late exposure to estrogen, is important. Studies thus far suggest that both types of effects might be operating, as Klump et al. (2010) found that higher estrogen levels during puberty (i.e., status effects) significantly increased genetic effects on disordered eating, regardless of age (i.e., the impact of low/high estrogen levels did not depend on the twin’s age). By contrast, Harden et al. (2012) and Baker et al. (in press) found that early pubertal timing shared genetic risk factors with eating disorder symptoms, suggesting that timing of puberty (and possibly, timing of estrogen activation) might influence genetic risk for eating disorders. More twin studies directly examining status versus timing effects of estrogen activation are clearly needed to confirm these impressions.

It would be helpful for these studies to also examine other phenotypes that are comorbid with eating disorders and have been linked to pubertal processes (e.g., depression, anxiety, substance use). Some developmental twin studies show that changes in genetic risk for disordered eating are independent of mood, but most studies have not examined comorbid psychiatric traits. Additional research into the specificity of puberty/hormone effects for eating disorders will help clarify etiologic models of, and pathways for, eating disorders and inform dimensional models of psychopathology. Ideally, these investigations would identify similarities as well as differences in hormone effects across disorders to more clearly delineate processes that predict differential disorder expression (e.g., common genetic/hormonal mechanisms for a range of pathology that then interact with disorder-specific environmental risk factors (e.g., pressures for thinness) (see Kendler et al., 2010 for data suggesting the presence of common genetic versus disorder-specific environmental risk factors).

Clearly, additional studies are needed to directly examine the significance of activational versus organizational influences of estrogen on eating disorder phenotypes during puberty. Animal studies would be very useful in this regard, as they could examine risk for eating disorder phenotypes in female animals differentially exposed to ovarian hormone manipulations across development. For example, if estrogen influences risk for eating disorders via its organizational properties, one would expect decreases in eating disorder phenotypes in pre-pubertally ovariectomized animals that cannot be restored with hormone treatment in adulthood. These animals would be permanently altered to be low risk phenotypes in adulthood, regardless of ovarian hormone exposure in later life.

However, human studies of developmental periods other than puberty could also increase understanding of organizational versus activational influences. If estrogen has only activational influences on eating disorder risk, we would expect changes in phenotypic and genetic risk for eating disorders across the menstrual cycle and after menopause. In terms of the menstrual cycle, we should see robust changes in eating disorder symptoms and genetic risk across menstrual cycle phases that are due to differential hormone activation across the cycle. In terms of menopause, we should see significant changes in phenotypic and genetic risk for eating disorders after menopause when the ovarian hormone system is de-activated in women. By contrast, if estrogen’s influence is primarily organizational, we would expect much less of a change in phenotypic and genetic risk for eating disorders and their symptoms across the menstrual cycle and in later life. In this case, the “turning on” of genes during puberty would cause permanent changes in gene action and/or brain structure/function that make the eating disorder phenotype relatively immutable to changes in hormone concentrations across the adult reproductive cycle. Indirect data exist to support both possibilities, as phenotypic changes in symptoms (albeit, somewhat modest) are observed across the menstrual cycle (Edler et al., 2007; Klump et al., 2008; Klump et al., 2013; Racine et al., 2012), and significant eating disorder symptoms persist in women over the age of 50 (Gagne et al., 2012). However, no studies have directly compared genetic risk across menstrual cycle phase or pre- versus post- menopause. Data from such investigations would provide critical insights into the relative importance of activational versus organizational influences of ovarian hormones on risk for eating disorders.

Studies of the prenatal/perinatal period are also needed, as recent studies show that the effects of estrogen during puberty may depend upon earlier organizational influences of testosterone during prenatal/perinatal development. These theories posit that it is the absence of prenatal testosterone, and the presence of pubertal estrogen, that organizes sex-specific behavior in females (Bell and Zucker, 1971; Gentry and Wade, 1976). Intriguing new data suggest that these processes may be critical for eating disorders, as sex differences in eating disorder risk during puberty (i.e., higher rates in girls relative to boys) may be due to the protective effects of prenatal testosterone in boys and the “risky” effects of ovarian hormones in girls (Culbert et al., submitted).

Additional studies are also needed to identify the specific neurobiological systems that may be activated and/or organized by estrogen during puberty. Emerging data show strong phenotypic and genetic effects of estrogen on the pattern and timing of changes in brain structure and function during puberty (Ahmed et al., 2008; Nunez et al., 2002; Peper et al., 2009a; Peper et al., 2009b; Primus and Kellogg, 1991; Spear, 2000; Zehr et al., 2006). Yet, none of these studies have linked these changes to eating disorders or their symptoms. Several of the neurobiological systems that are regulated by estrogen and have been implicated in the etiology of eating disorders would be ideal targets for such investigations (Hildebrandt et al., 2010; Ostlund et al., 2003; Young, 2010). For example, estrogen regulates gene transcription within the serotonin system (Hildebrandt et al., 2010; Ostlund et al., 2003) including the serotonin 2a receptor and serotonin transporter, which have been associated with AN and BN in neuroimaging and genetic research (Hildebrandt et al., 2010; Kaye, 2008; Klump and Culbert, 2007; Ostlund et al., 2003). Estrogen also regulates the production of BDNF (Schafman and MacLusky, 2008; Sohrabji et al., 1995; Solum and Handa, 2002; Toran-Allerand, 1996) which shows consistent genetic associations with eating disorders (Klump and Culbert, 2007; Rask-Andersen et al., 2010). Estrogen plays a highly significant role in the regulation of brain reward systems (Becker, 2009; Craft, 2008), including dopamine and opoiods, consistently linked to preferences for the palatable and sweet foods ingested during binges in humans and animals (Berridge, 2009). Notably, many of these systems experience significant changes/maturation during the pubertal period (Becker, 2009; Friemel et al., 2010; Iughetti et al., 2011) that may be linked to estrogen (Becker, 2009) and may contribute to increases in pubertal risk. Human and animal studies examining whether estrogen regulation of these systems accounts for increased risk are needed to confirm a role for estrogen in pubertal increases in eating disorders.

Finally, studies that integrate across multiple levels of risk factors are needed to address traditional and emerging theories of eating disorder development during puberty. Similar to other behavioral phenotypes, eating disorders are unlikely to be caused by one type of risk factor, particularly since the same factor can impact risk via both biological and environmental processes. As an example, in the Harden et al. (2012) twin study discussed above, subjectively rated pubertal timing (i.e., whether the twin thought her development was earlier than her peers) was environmentally associated with increased risk for dieting in girls. This was in stark contrast to the findings for objectively measured pubertal timing (i.e., comparison of the twin’s actual pubertal development to population norms) that was associated with increased dieting through genetic factors. The possibility that pubertal timing may be differentially associated with eating disorder risk depending upon the measurement/perspective underscores the strong need for multi-method, multi-level studies that examine biological and psychosocial risk factors and their interplay in the progression of eating disorder symptoms across development. These types of studies are likely to have the strongest impact on etiologic models as well as the eventual development of prevention programs that target the specific aspects of pubertal development (e.g., perception of timing) most amenable to change and psychosocial intervention.

Research Highlights.

• Human and animal studies of pubertal risk for eating disorders are reviewed.

• Advanced pubertal status and early pubertal timing increase risk in girls.

• Puberty is less consistently linked to eating disorder risk in boys.

• Female-specific risk may be due to estrogen activation of genetic influences.