Combined Oral Contraceptive Use and Risk for Binge Eating in Women: Potential Gene x Hormone Interactions

Abstract

Extant animal and human data suggest endogenous ovarian hormones increase risk for binge eating in females, possibly via gene x hormone interactions and hormonally induced increases in genetic influences. Approximately 85% of women will take combined oral contraceptives (COCs) that mimic the riskiest hormonal milieu for binge eating (i.e., post-ovulation when both estrogen and progesterone are present). The purpose of this narrative review is to synthesize findings of binge eating risk in COC users. Few studies have been conducted, but results suggest that COCs may increase risk for binge eating and related phenotypes (e.g., craving for sweets), particularly in genetically vulnerable women. Larger, more systematic human and animal studies of COCs and binge eating are needed. The goal of this work should be to advance personalized medicine by identifying the extent of COC risk as well as the role of gene x hormone interactions in susceptibility.

The most prevalent forms of eating disorders are those characterized by binge eating (i.e., the consumption of a large amount of food, in a short period of time, with a loss of control over eating), including bulimia nervosa (BN), binge eating disorder (BED), and other specified feeding and eating disorders (OSFEDs) (American Psychiatric Association, 2013). These binge-related eating disorders (hereto referred to as BE syndromes or BE-S) are associated with some of the highest mortality rates of all psychiatric disorders (American Psychiatric Association, 2013; Arcelus et al., 2011; Crow et al., 2009; Crow et al., 2014) and often have chronic courses characterized by substantial psychiatric and medical morbidity (American Psychiatric Association, 2013; Keel and Brown, 2010; Kessler et al., 2013). Their decidedly bleak psychosocial outcomes (e.g., Keel et al., 2000) further attest to their public health significance and the urgent need to understand their development.

Given this clinical significance, experts have called for transdiagnostic studies that increase understanding of the full array of BE-S (Sysko et al., 2011; Walsh and Sysko, 2009). Thus, in this narrative review, we focus on binge eating as the core maladaptive behavior that is common to all BE-S. This approach directly addresses recent initiatives (e.g., the NIMH’s Research Domain Criteria; Sanislow et al., 2010) that focus on transdiagnostic dimensions of observable behavior contributing to mental disorders. This focus on observable behaviors (rather than the complex constellation of behavior, cognition, and affect) provides a more powerful approach for identifying etiological processes that form the core of full syndromes (Sanislow et al., 2010) and it allows for translational research in non-human species. In addition to its significance for understanding BE-S, binge eating is an important target for women’s health research because it disproportionally affects females¹ relative to males (2:1 up to 6:1) (Klump, Culbert, and Sisk, 2017) and contributes to significant psychiatric comorbidity, medical complications, and risk for weight gain (American Psychiatric Association, 2013; Mitchell and Crow, 2010; Wassenaar et al., 2019).

Critically, this research is also needed to inform public health policy. Animal and human data show endogenous ovarian hormones (i.e., estrogen and progesterone) substantially increase risk for binge eating potentially via hormonally induced increases in genetic risk (Asarian and Geary, 2013; Harden et al., 2014; Klump, Culbert, and Sisk, 2017). Approximately 85% of women will take contraceptives (Bensyl et al., 2022; Chadwick et al., 2012; Hall and Trussell, 2012; Mosher and Jones, 2010) that mimic the riskiest hormonal milieus for binge eating (i.e., combined oral contraceptives (COCs) containing both synthetic estrogens (ethinyl estradiol) and progestins). Although the effects of the synthetic COC hormones within the central nervous system (CNS) are less studied, initial evidence suggests that they have regulatory effects (Sing & Su, 2013; Spona, Leible, & Bieglmayer, 1980; Tofoletto et al., 2014) that are similar to, or in some cases more deleterious (e.g., decrease neuroprotection – see Sing & Su, 2013), than those of estrogen and progesterone. Unfortunately, these COCs are some of the most commonly used hormonal contraceptives (Bensyl et al., 2022; Mosher and Jones, 2010), and women typically begin taking these medications during the highest period of risk for BE-S (i.e., late adolescence/early adulthood) (American Psychiatric Association , 2013).

Given the above, the goal of this narrative review is to describe and synthesize studies examining the impact of COCs on risk for binge eating. We begin by reviewing evidence that endogenous estrogen and progesterone are associated with increased risk for binge eating in women. We then discuss the ways in which COCs may mimic and enhance this risk and review the (scant) data that are available thus far. We end by discussing potential mechanisms of effects, including individual differences in genetic risk that may explain why some, but not all, women develop binge eating in the presence of risky endogenous and exogenous hormone levels. We also outline directions for future research with an emphasis on large-scale, longitudinal, and translational studies in humans and animals that can substantially move the field forward in this important area of women’s health research.

Importantly, our goal in writing this review is NOT to dissuade women from using COCs. Our goal is to highlight potential adverse consequences of one type of hormonal contraceptive (COCs) for one type of woman (e.g., those with genetic risk) and spearhead personalized medicine approaches to this critical area of women’s health.

Endogenous Hormone Effects on Binge Eating

Menstrual cycle studies show a substantial role for ovarian hormones in phenotypic risk for binge eating (Edler et al., 2007; Lester et al., 2003; Klump et al., 2013b; Klump et al., 2014). In regularly menstruating women, cyclic variations in these hormones (see Figure 1) occur every month in preparation for possible conception. Because hormonal changes are dictated by the reproductive system, changes in ovarian hormones that precede changes in binge eating likely reflect the impact of the hormones on binge eating rather than the reverse.

Figure 1. Changes in Estradiol and Progesterone Levels across the Menstrual Cycle.

Klump, K.L., Keel, P.K., Racine, S.E., Burt, S.A., Neale, M., Sisk, C.L., Boker, S., & Hu, J.Y. (2013). The interactive effects of estrogen and progesterone on changes in emotional eating across the menstrual cycle. Journal of Abnormal Psychology, 122(1), 131-137. Copyright © 2012 by the American Psychological Association. Reproduced and adapted with permission.

Results have consistently shown post-ovulatory peaks in food intake (e.g., Buffenstein, 1995), binge eating (Edler et al., 2007; Klump et al., 2014; Lester et al., 2003), and emotional eating (i.e., the tendency to overeat/binge eat in response to negative emotions) (Klump et al., 2013b) in women with BE-S and from the community. Studies confirmed that changes in estrogen and progesterone drive these post-ovulatory increases (Klump et al., 2013b; Klump et al., 2014). Indeed, interactions between estrogen and progesterone (i.e., higher levels of each hormone) were the strongest predictors of increases in binge eating and emotional eating across the cycle (Klump et al., 2013b; Klump et al., 2014). These hormone effects explained post-ovulatory peaks in binge eating, as post-ovulation is characterized by the presence of both estrogen and progesterone (see Figure 1). In sharp contrast, the pre-ovulatory phase (with higher estradiol and low progesterone levels) is a very low risk period for binge eating (Edler et al., 2007; Klump et al., 2013b; Klump et al., 2014). Importantly, in these studies, hormone effects could not be accounted for by differences in body mass index (BMI) (Edler et al., 2007; Klump et al., 2013b; Klump et al., 2013a; Klump et al., 2014; Lester et al., 2003), negative affect (Edler et al., 2007; Klump et al., 2013b; Klump et al., 2013a; Klump et al., 2014; Lester et al., 2003), dietary restraint (Klump et al., 2013a; Klump et al., 2014), or impulsivity (Racine et al., 2013). In addition, they could not be accounted for by the surges or precipitous drops in ovarian hormones that occur before, during, and after ovulation, as these dramatic changes in hormones did not significantly predict levels of emotional eating (Fowler et al., 2019). Finally, results were remarkably consistent across participant groups (population-based samples and clinical cases) and the spectrum of binge eating phenotypes (i.e., binge frequencies, emotional eating) (Klump et al., 2013b; Klump et al., 2014), suggesting that hormone/binge eating associations are present and detectable across the full range of dysregulated eating.

Collectively, these human studies extend 40+ years of animal studies showing significant and causal effects of estrogen and progesterone on the control of food intake (Asarian and Geary, 2013). Using direct manipulations of hormones via ovariectomies and hormone administration, animal studies show increases in food intake when both estradiol and progesterone levels were high, as progesterone antagonizes the anorexic effects of estrogen on food intake (Asarian and Geary, 2013). By contrast, food intake decreases when estradiol levels are high and progesterone levels are low, as estrogen’s anorexic effects run uninhibited in this hormonal environment (Asarian and Geary, 2013). Animal studies of binge eating rather per se are fewer in number, but initial data show increased rates of binge eating (i.e., intermittent consumption of highly palatable food) in female rats ovariectomized in adulthood (Klump et al., 2011; Micioni Di Bonaventura et al., 2017) that are reversed with exogenous estradiol treatment (Micioni Di Bonaventura et al., 2017; Yu et al., 2008).

COC Effects on Binge Eating

Significant associations between endogenous hormones and binge eating in humans and animals leads to questions about the effects of exogenous hormones and COCs. There are many different types of COCs, but the most commonly prescribed are the monophasic COCs (Bensyl et al., 2022; Hall and Trussell, 2012; Mosher and Jones, 2010). Monophasic COCs provide stable doses of synthetic estrogen (i.e., ethinyl estradiol) and progesterone (i.e., progestins) in 28-day packs that include 21-24 days of active pills and 4-7 days of inactive/placebo pills (Nelson, 2008). Biphasic and triphasic COCs also provide stable doses of ethinyl estradiol but the progestin dose increases once (biphasic) or twice (triphasic) during each pill pack (Nelson, 2008).

Although hormone doses and type of progestin vary by COC brand, the function of the active pills is to create an environment that mimics the post-ovulatory phase of the menstrual cycle (i.e., the presence estrogen and progesterone) (Nelson, 2008). This is accomplished via the provision of the synthetic hormones that bind to receptors and create a negative feedback loop that inhibits the release of hormones from the brain (gonadotropin-releasing hormone) and pituitary (follicle stimulation hormone, luteinizing hormone) that trigger ovarian estrogens and ovulation (Nelson, 2008). Thus, although the synthetic COC hormones attenuate endogenous estrogen and progesterone release (Dickson and Eisenfeld, 1981; Fleischman, Navarette, & Fessler, 2010; Hyder et al., 1999; Nelson, 2008), the presence of both hormones is similar to the post-ovulatory phase of the menstrual cycle. This mimicking of post-ovulation could increase risk for binge eating in COC users, as estrogen and progestin are present for much longer periods of time in COC users (21-24 days of active pills) than non-users (12-14 days after ovulation). Binge eating also could be higher in COC users because they never experience the protective, pre-ovulatory milieu (higher estradiol, lower progesterone) experienced by non-users.

Unfortunately, to date, very few studies have examined these hypotheses for binge eating, although some studies have examined appetitive processes (e.g., sweet food preferences, food cravings) that are associated with and predict the development of binge eating (Chao et al., 2016; Dalton and Finlayson, 2014). We review below the limited data that are available in humans, focusing on between-person (comparing COC users versus non-users) and within-person (comparing changes in binge eating within a woman when exposed versus not exposed to COCs) studies of risk. Notably, we discuss animal studies in the Conclusions and Future Directions section, as the even smaller number of animal studies underscores the strong need for translational research.

Between-Person Studies of COC Users versus Non-Users

By comparing COC users to non-users, between-person studies can estimate population-level risk that is needed to design public health and personalized medicine approaches to reproductive care. However, it is difficult to collect representative samples, and matching COC users to non-users on all relevant variables (e.g., age, demographic characteristics, personality traits) is challenging. Without careful matching, group differences in binge eating could be due to confounding factors or “third variables” rather than COC use.

To date, there are no population-based studies of between-person risk for binge eating. A large, school-based study of depression found higher rates of “eating problems” in adolescent COCs users versus non-users, but the types of eating problems examined were not described (De Wit et al., 2020). Nonetheless, a number of between-person studies have examined binge-related phenotypes in convenience samples (e.g., college women). These studies have limitations in terms of representativeness and generalizability, but they provide initial indications of potential between-woman differences in risk by COC status.

Consistent with the hypotheses described earlier, COC users report greater overall food intake (Tucci et al., 2010; Wallace et al., 1987) and greater consumption of foods high in sugar and/or fat (Eck et al., 1997) as compared to non-COC users. Findings from an fMRI study suggest that these group differences may be due to COC-induced neural changes (Arnoni-Bauer et al., 2017). Increased activation to food cues was observed in homeostatic and brain reward regions in normally cycling women (N = 20) during their luteal as compared to their follicular phase. By contrast, sustained activation in these neural regions was observed during all of the fMRI scans in women taking their active monophasic COC pills (N = 12). This activation was indistinguishable from that observed in the normally cycling women during their luteal phase and was significantly higher than the activation during the follicular phase. These data suggest that increased activation to food cues occurs for a much longer period of time (i.e., for 21 active pill days) in the COC users as compared the non-users (i.e., ~12 days).

Unfortunately, this study did not examine food intake, but findings from Tucci et al. (2010) suggest results may extend to actual food consumption. COC users’ (N = 29) intake of sweet and savory foods was indistinguishable from the luteal phase intake and much higher than the follicular phase intake of non-COC users (N = 26) (see Table 3 - Tucci et al., 2010). This study was underpowered to detect significant interactions between COC use and phase (see p. 391 - Tucci et al., 2010), but findings indicate that COC users may increase their food intake across all active pill days, while non-COC users increase their intake only during the luteal phase.

Taken together, data suggest that COC use might be associated with increased “eating problems”, food intake, preference for and consumption of binge-type food, and neural activation to food cues. However, other analyses, even within the same study (e.g., see Eck et al., 1997) have shown no or more modest COC user/non-user differences for binge-related variables (Bancroft and Rennie, 1993; Eck et al., 1997; Lund and Jacobsen, 1990; Wallace et al., 1987; Wiksten-Almströmer et al., 2008). Discrepant results may be due to substantial methodological limitations, including small sample sizes and reliance on convenience samples (Arnoni-Bauer et al., 2017; Eck et al., 1997, Tucci et al., 2010). There also is a lack of careful matching of users to non-users (Eck et al., 1997; Lund and Jacobsen, 1990; Wallace et al., 1987). Indeed, in most cases, basic demographic information such as race and ethnicity and socioeconomic status of participants is not described, making it difficult to determine whether the samples are well-matched or representative of the population at large. The use of very different measures of binge-related constructs across all studies also makes it difficult to compare/contrast findings. Given these methodological constraints, conclusions await larger, more tightly controlled and population-based studies that carefully assess binge eating in carefully matched samples of users and non-users.

Within-Person Studies of COC Exposure versus No Exposure

Within-person studies circumvent many of these limitations by having women serve as their own controls and examining within-woman changes in binge eating when exposed ( e.g., during active pill phases) versus not exposed (e.g., during inactive/placebo pills) to COCs. These within-person comparisons do not require matching COC users to non-users and can provide some of the strongest evidence for COC effects.

Within-person studies can be naturalistic or intervention studies. Naturalistic studies examine women who are already taking COCs and compare rates of binge eating between different “phases” of the pill pack (e.g., active hormone pills versus inactive/placebo pills). An advantage of these designs is the ability to not only compare rates of binge eating across pill type, but also determine if the length of the inactive/placebo pill phase (e.g., 4 days versus 7 days) and synthetic hormone “washout” period is differentially associated with binge eating. By contrast, intervention studies randomly assign participants to receive COCs or placebo/no treatment and compare rates of binge eating across baseline (no COC exposure) and treatment (COC or placebo exposure) conditions. Although there are ethical challenges in assigning women to a hormonal treatment that is hypothesized to increase risk for binge eating and binge-related disorders, intervention studies are unique in that they can also examine between-person effects by comparing rates of binge eating in women assigned to COCs versus placebo.

Naturalistic Studies

No naturalistic, within-person studies have examined changes in binge eating across active and inactive pills. However, studies have examined binge-eating related phenotypes and found increased hunger (McVay et al., 2011), cravings for sweet food (Bancroft and Rennie, 1993), general food cravings (McVay et al., 2011), and food intake (McVay et al., 2011; Eck et al., 1997) in COC users during active versus inactive pills. These data provide initial evidence of COC effects on several binge eating constructs.

Nevertheless, other comparisons yielded more modest or variable changes (e.g., increased binge-related behaviors in the week prior to menses/inactive pills, in addition to the week of inactive pills) (Bancroft and Rennie, 1993; Eck et al., 1997; McVay et al., 2011). Reasons for discrepant results may again be due to substantial methodological limitations. Samples tend to be small and include women taking different types of COCs (e.g., Eck et al., 1997). Some studies failed to examine any inactive pill phases (Tucci et al., 2010), while others relied on retrospective reports of past pill phases that could unduly influence results (Bancroft and Rennie, 1993; McVay et al., 2011). For example, reports of increased binge-related behaviors in the week prior to inactive pills/menses may be due to women’s pre-conceived notions that mood and eating worsen during “premenstrual” phases. Most studies failed to report demographic information about race and ethnicity and socioeconomic status. Although there is no matching of users to non-users in within-person studies, it is still critical to recruit representative samples whose data can be generalized to the population at large and to report demographic information in order to know to whom results apply. Given these limitations, and the lack of data on binge eating per se, conclusions await larger and representative studies controlling for COC type and obtaining real-time assessments of active and inactive pill phases.

Intervention Studies

An early case report described dramatic decreases in binge eating in a woman with BE-S after discontinuing COCs (Moskovitz and Lingao, 1979) while another study found that reports of overeating as a result of COCs were five times greater than reports of no or beneficial COC changes (Condon et al., 1995). However, neither study was an intervention trial randomizing women to treatment conditions.

Only two intervention studies have investigated binge eating and COCs. Plessow et al. (2019) conducted a randomized control trial of the effects of COCs (N = 11), transdermal estradiol (via a PATCH) with cyclic progesterone (N = 20), and no hormone treatment (N = 21) on uncontrolled eating (defined as “overconsumption of food accompanied by a perceived loss of control”) in female endurance athletes who were experiencing oligo-amenorrhea and suppressed estradiol levels. The authors hypothesized that uncontrolled eating would be significantly lower in the PATCH group relative to the other groups, due to their elevated and constant levels of estradiol (100 mcg of 17β-transdermal estradiol) across all study days and only intermittent exposure (12 days/month only) to micronized progesterone (dose not provided). By contrast, levels of uncontrolled eating would be expected to be higher in the COC (30 mcg ethinyl estradiol, 0.15 mg desogestrel) and no treatment groups, as these groups would not be exposed to the protective hormonal environment of higher and constant estradiol levels with only intermittent exposure to progestin. After 12 months of treatment, the athletes assigned to the PATCH group did show significant decreases in uncontrolled eating as compared to the COC (but not the no hormone treatment) group. As the authors noted, the athletes assigned to COCs were the only group to show substantial increases (COC group change score = 2.78 ± 2.72; PATCH group change score = −6.48 ± 2.40; no hormone group change score = 0.81 ± 2.92) in uncontrolled eating across the study period. Although the COC changes were not significantly different from those observed in the no hormone group, the effect size (d = .70) approached Cohen’s benchmark for a large effect size (i.e., Cohen’s d = .80). The failure to detect significant differences is therefore likely due to small sample sizes (N = 11 COC participants; N = 21 no hormone participants) rather than a lack of COC effects. Overall, although this study was not without limitations (e.g., small samples sizes, higher attrition (48%) in the COC group), results suggested that hormonal contraceptives with longer (and higher) estrogen and shorter progestin exposure may decrease binge eating, and that COCs might be associated with increases in binge eating through extended exposure to both estrogen and progestin.

Findings from a second intervention study were more difficult to interpret. Naessén et al. (2007) reported no significant changes in binge eating or other phenotypes (e.g., hunger, craving for sweets, concentrations of cholecystokinin (CCK) - a gut-brain satiety peptide) in women with BN after 3 months of treatment with an anti-androgen COC. However, changes in binge eating from pre-to-post treatment were not shown in the paper, making it difficult to determine if data were trending toward increases or decreases in binge eating. This is particularly problematic given the study’s very small sample sizes (i.e., N = 21 women with BN) and the likelihood that even medium-effect size changes would be non-significant in this sample. Nonetheless, significant increases in hunger and cravings for sweets were observed from pre-treatment to follow-up in the healthy control group of women who also took the COC (N = 17 women) (Naessén et al., 2007). These women experienced significant decreases in CCK concentrations, a neuropeptide that increases satiety and is significantly reduced in women who binge eat (e.g., Hannon-Engel, 2012; Keel et al., 2007). Significant decreases in 24-hour CCK concentrations were also observed in two small-scale studies of healthy control women (Ns = 9-10 women) assessed at baseline (no COC use) and after 2-5 months of COC use (Hirschberg et al., 1996; Karlsson et al., 1992). These decreases in CCK are noteworthy given that higher levels of estrogen (characteristic of the pre-ovulatory phase) increase the activity of CCK and promote satiety (Asarian & Geary, 2013); it is possible that the lower levels of estrogen in COCs, coupled with the antagonizing effects of progestins, result in lower CCK concentrations that lead to decreased satiety and increased craving and consumption of sweet foods that are commonly consumed during a binge.

In addition to intervention studies of binge eating, there is a fairly large literature examining changes in body weight (a binge-related phenotype) in COC treatment trials. Perceptions of increased weight are repeatedly reported by COC users (e.g., Hall et al., 2013; O’Connell et al., 2007), and weight gain is one of the most cited reasons for discontinuing COCs (Condon et al., 1995; Hirschberg, 2012; Metz et al., 2022; Hall et al., 2013). Nonetheless, intervention studies show no significant changes in objectively measured weight from pre-to-post treatment in the majority of participants (e.g., Brown et al., 2016; Gallo et al., 2014; Metz et al., 2022; O’Connell et al., 2007; Park and Kim, 2016; Piccoli et al., 2008; Rosenberg, 1998). However, many studies find individual differences in weight changes, with a significant subsample of women experiencing weight gains (e.g., 28% of users in Rosenberg, 1998). These findings suggest that some women may be more sensitive to COCs and their effects on weight. Studies carefully tracking daily weight changes show that the picture may be even more complicated. Rosenberg (1998) observed a mean weight gain of .5 pounds during the early active pill days (that occur immediately after inactive pills) and the same mean weight loss at the end of the pill cycle. These findings suggest that minor fluctuations across pill cycles could lead to perceived weight gain in COC users (Rosenberg, 1998), even if overall weight does not change.

Overall, intervention studies of body weight are much more sophisticated and advanced than those of binge eating and suggest that there is a subsample of women for whom COC use leads to weight gain. Studies of binge eating are fewer in number and suffer from a number of methodological limitations (e.g., small sample sizes, lack of careful demographic reporting) (Hirschberg et al., 1996; Karlsson et al., 1992;Naessén et al., 2007; Plessow et al., 2019) that make it difficult to draw firm conclusions. Findings thus far suggest that COCs might increase risk for binge eating, possibly through modifications to CCK and satiety signals, but many more intervention studies are needed before conclusions can be drawn.

Summary

The most striking features of the literature so far are the lack of data on binge eating per se and variability in findings. Several studies suggest that COCs might increase risk for binge eating and binge-related phenotypes, but studies also find no significant or more modest COC effects. Variability in results could be due to the vast methodological limitations described above; indeed, small sample sizes, lack of control for type of COC or pill type, lack of information on demographic characteristics of participants, and other limitations constrain the strength of the conclusions that can be drawn.

However, differences in findings could also reflect meaningful variations that tell us something about the nature of COC effects on binge eating. Discrepant results may signal the presence of moderation of COC-binge eating associations by other variables that make some women more vulnerable to the effects of COCs than others. When such moderation is present, findings are typically modest/non-significant or more variable, as samples would naturally vary in the extent to which individuals who are most at risk are included. A potential example in this regard is body weight; although in aggregate, findings suggest no significant COC effects, most studies identify a subsample of COC users who experience significant weight gains. These differences in vulnerability may reflect differential sensitivity to the effects of COCs that could contribute to some, but not all, women developing binge eating. Differences in vulnerability are clearly present for endogenous hormones, as not every woman binge eats during the luteal phase of the cycle. Data suggest that gene x hormone interactions may account for this differential susceptibility to endogenous hormones and potentially also COCs.

Gene x Hormone Interactions: Differential Susceptibility to Hormones

Gene x hormone interactions occur when a person’s response to a risk factor (e.g., ovarian hormones) varies by their level of genetic risk. Individuals with genetic predispositions for the behavior will be most sensitive to the environmental risk and most vulnerable to developing the behavior (Ottman, 1996). Individuals who are not at genetic risk are unlikely to develop the behavior, even in the presence of environmental risk.

In the case of ovarian hormones, it may be that women who are at genetic risk for binge eating are most sensitive to ovarian hormones and most likely to develop binge eating. Indirect data suggest this is the case - women who are most likely to be at genetic risk for binge eating/eating disorders (i.e., those with clinical disorders and/or more severe levels of pathology) appear to be most sensitive to COC side effects and mood changes. For example, women with a history of eating disorders, depression, or anxiety disorders are significantly more likely to experience adverse mood during COC use than women without these histories (Bengtsdotter et al., 2018). Women with elevated eating disorder symptoms are also more likely to discontinue OCs (Hall et al., 2013), and women reporting a history of OC side effects (a potential marker of hormone sensitivity) report significantly higher levels of body weight/shape concerns than those without OC side effects (Bird and Oinonen, 2011). The action of ovarian hormones in the CNS adds further support for potential gene x hormone interactions. Ovarian hormones are steroid hormones that regulate gene transcription and protein synthesis in several key systems for BE-S (Barth et al., 2015), including the dopamine (Becker, 2009; Barth et al., 2015) and serotonin system (Ostlund et al., 2003). Although progesterone’s effects on gene transcription have been studied less frequently (Arbo et al., 2014; Rivera et al., 2009), it may serve to modify or interact with the regulatory effects of estrogen (Bethea and Reddy, 2010; Jayaraman and Pike, 2014; Rivera and Bethea, 2013), particularly since estrogen x progesterone interactions predict binge eating in women (see above).

The exact way in the genomic effects of hormones may translate into gene x hormone interactions remains unknown. Theoretically, interactions between hormones and risk variants could result in differential production of neurotransmitters, neuromodulators, their receptors, or their signal transduction mechanisms. Variants of estrogen and/or progesterone receptors may also result in different patterns of gene regulation by the hormones, leading to distinct cellular and behavioral responses. In all of these scenarios, two key individual difference variables impact binge eating risk - the presence/absence of risk alleles in hormonally responsive genes and the presence of a risky hormonal environment. Increased binge eating would be evident in individuals who carry risk variants for binge eating and who are exposed to estrogen and progesterone/progestins either during post-ovulation or during the COC use. Indeed, women with genetic risk for binge eating would be more likely to experience the post-ovulation or COC hormonal environment as “risky.”

One approach for directly examining these gene x hormone hypotheses would be to investigate whether rates of binge eating vary by the presence/absence of risk variants and endogenous hormones and COC use. Unfortunately, as with many forms of psychopathology, replicated risk genes for binge eating have yet to be identified (Breithaupt et al., 2018). A powerful alternative approach uses twin analyses to examine COC effects on the aggregate of all genetic influences on binge eating rather than on only one gene or set of measured genes. These analyses examine whether the heritability of binge eating (i.e., the extent to which individual differences in binge eating are due to genetic differences between people) varies by hormone levels. If higher heritability is observed in the presence of risky hormonal environments (i.e., higher heritability with higher estradiol/progesterone levels and in COC users), then gene x hormone interactions are likely present, as the genetic variance is increasing only in the presence of the risky environment.

To date, no studies have used this approach for examining COCs, but our research group examined endogenous hormones in a series of between-person analyses (i.e., differences in heritability between women with different trait hormone levels) and within-person analyses (i.e., differences in heritability within women across changing hormone levels) (Klump et al., 2015; Klump et al., 2016). Using a population-based, representative sample of normally cycling identical (monozygotic (MZ)) and fraternal (dizygotic (DZ)) female twins (N = 571), the heritability of emotional eating varied significantly between twins with different trait-levels of hormones; heritability was 37% in twins with lower levels of progesterone, whereas the heritability was 61% in twins with higher levels of progesterone (Klump et al., 2016). Differences in twins with higher/lower estradiol levels were non-significant, although the study was underpowered to examine whether combinations of higher estradiol and higher progesterone resulted in the highest heritability of binge eating. Taken together, data suggested that the degree of genetic influence on emotional eating varies in systematic ways between women with different ovarian hormone levels, and that these differences may contribute to differential susceptibility and expression of binge eating across women.

Nonetheless, these between-person twin analyses suffer from the same limitations as the between-person findings described above - differences in genetic effects could be due to third variables or confounding factors between twins with higher versus lower hormone levels. Fortunately, daily binge eating scores were also available across a full menstrual cycle in these twins. Thus, in addition to between-person analyses, we examined within-person changes in genetic effects when twins moved from low-risk, pre-ovulatory phases of the menstrual cycle (marked by rising estradiol and low progesterone levels) to the high-risk, post-ovulatory phases (characterized by higher levels of both hormones) (Klump et al., 2015). Findings revealed dramatic changes in genetic effects; the heritability of emotional eating increased from ~0% during pre-ovulation to 47% during post-ovulation. Because analyses were conducted in the same sample of twins across time, changes in genetic effects reflected within-woman differences associated with each menstrual cycle phase (e.g., changes in hormones) rather than confounding variables or differences between women.

Collectively, these findings provide initial support for the presence of gene x endogenous hormone interactions for binge eating, whereby the degree of genetic influence on binge eating varies between and within women by endogenous hormone levels. Nonetheless, there are no independent replications of these findings, and smaller sample sizes and limitations of quantitative methods (see Klump et al., 2015) made it difficult to examine estrogen x progesterone interactions. There also are no twin studies of gene x COC effects. Based on the hypotheses and data described above, we would expect that between women, the heritability of binge eating would be higher in COC users than non-users, and within-women, the heritability of binge eating would be highest during the active versus the inactive pill phase. A large-scale twin study of these hypotheses is on-going in our laboratory and will increase understanding of individual differences in phenotypic and genetic risk for binge eating.

Conclusions and Future Directions

Research of the effects COCs on binge eating in women is in its nascent stage. Animal and human data suggest that endogenous estrogen and progesterone contribute to phenotypic risk for binge eating. Although lower estradiol levels contribute to binge eating, the riskiest environment appears to be a combination of estradiol and progesterone that characterize the mid-luteal phase of the menstrual cycle. The hormone concentrations of COCs mimic this risky time period, leading to concerns that these medications may increase risk for binge eating. Between- and within-person studies are plagued by methodological limitations, but findings suggest that binge eating may be greater in COC users relative to non-users, and that binge eating may increase within a woman when exposed to the synthetic hormones in COCs. However, data also show substantial individual differences in COC responses that may reflect gene x hormone interactions that increase susceptibility in some, but not all, women.

There are numerous areas for future research. Studies that directly assess binge eating, in addition to binge-related phenotypes, are needed to confirm that COC effects are present across the full spectrum of binge eating. Population-based, between-person studies in carefully matched COC user/non-user samples are needed to estimate community-level risk. These studies would also be useful for ruling out the possibility that COC use could actually be beneficial for decreasing binge eating; studies in premenstrual dysphoric disorder (PMDD) have shown that longer-term COC use may actually decrease, rather than increase, negative symptoms by providing stable hormone levels and decreased hormonal surges (Schmidt et al., 1991; Schmidt et al., 2017). Although previous research suggests that more pronounced changes in endogenous hormones are not significantly associated with binge eating (see Fowler et al., 2019), very few studies have been conducted, and the longer-term effects (beyond 1-2 cycles) of hormone surges or COC use on binge eating risk are unknown.

Detailed, within-person studies that conduct daily assessments of women across active and inactive pills are needed to document changes in binge eating with COC exposure. These studies should take advantage of the differences in placebo pill phase length (e.g., 4 versus 7 days) across different COCs to examine possible hormone wash-out effects and differential risk. Intervention studies of COC use on binge eating are more difficult to conduct, given ethical concerns noted earlier, but it would be extremely useful and beneficial from a personal/public health perspective for existing intervention studies of COCs to add measures of binge eating and related phenotypes to their protocols. It will be critical for all studies to include large, representative samples to ensure adequate power to detect significant effects (even with differential sensitivity to COCs) and that results apply to all women in the community.

Studies of differential susceptibility to both endogenous and exogenous hormones are also needed. Large-scale twin studies that examine differential heritability of binge eating across COC users and non-users, and across active/inactive pills, would help identify potential gene x COC mechanisms of effects. Genome-wide association studies are currently underway that will identify genes contributing to binge eating (Eating Disorders Work Group, Psychiatric Genetics Consortium, accessed March 2022. Studies that collect binge eating, genomic, ovarian hormone, and COC data would substantially advance this area of research by examining whether any identified risk genes are particularly sensitive to ovarian hormones and COCs and contribute to individual susceptibility in binge eating.

It is critically important to conduct animal studies of COCs. Studies in humans are unable to fully document causal effects, as findings are always limited by third variables, confounding factors, and issues of ethics and statistical power. There is a surprising lack of animal studies examining synthetic hormone effects on binge eating. One early study suggested that higher doses of ethinyl estradiol may suppress sweet preferences in female rats (Dippel et al., 1983), but findings for progestin were unexpected and opposite to those of endogenous progesterone in past work (i.e., increased sweet preferences, via interactions with estrogen). Clearly, additional studies are needed, particularly those using individual difference animal models of binge eating (e.g., Boggiano et al., 2007). Individual difference models identify animals that are particularly prone to binge eating and align more closely with individual differences in humans (where some women are more, and less, vulnerable to binge eating). Some of these animal models show strong effects of endogenous hormones on binge eating that are similar to those in humans (e.g., Klump et al., 2011). These models could test whether binge-eating prone animals are more sensitive to the effects of COCs than controls and directly probe genetic and neurobiological systems that may underlie differential sensitivity. Given data thus far, it is likely that systems involved in reward pathways, homeostasis, and satiety may be particularly relevant (Arnoni-Bauer et al., 2017; Hirschberg et al., 1996; Karlsson et al., 1992; Naessén et al., 2007).

Finally, it will be important to examine the effects of different types of hormonal contraceptives on binge eating risk. Within COCs, it will be critical to compare risk between monophasic, biphasic, and triphasic formulations. Although no studies to date have examined differences in binge eating across these different formulations, it is possible that the biphasic and triphasic pills pose the greatest risk for binge eating, given their increasing progestin levels across the pill pack. COCs containing progestins with significant androgenic potency may also show differential effects on binge eating risk (see discussion in Naessén et al., 2007), particularly given data suggesting links between polycystic ovary syndrome (PCOS) and binge eating (e.g., Jeanes et al., 2017). Non-COC forms of contraception (e.g., medroxyprogesterone acetate, intrauterine devices (IUDs)) are becoming increasingly popular. Initial data show inconsistent results in terms of risk for binge eating (e.g., Silva and Qadir., 2018), but studies are even fewer in number and study designs are highly variable. Finally, more studies are needed to identify birth control options that pose minimal binge eating risk (e.g., non-hormonal forms of IUDs) or are beneficial for decreasing risk (e.g., estrogen PATCH – see Plessow et al., 2019). Indeed, we owe it to women to identify both “risky” and “protective” reproductive options. Identifying and quantifying these effects will substantially advance personalized medicine approaches to contraceptive use and allow women to make informed decisions about their reproductive health.