Changes in affect longitudinally mediate associations between emotion regulation strategy use and disordered eating

Abstract

Background:

Trait-level emotion regulation (ER) difficulties are associated with eating disorders (EDs) transdiagnostically. However, little research has examined whether within-person fluctuations in ER longitudinally predict ED behaviors in daily life or the mechanisms of ER effects. Investigating daily ER could help us better understand why people experience ED behaviors at a given time. We examined whether day-to-day changes in adaptive (e.g., cognitive reappraisal) and maladaptive (e.g., rumination) ER longitudinally predicted core ED behaviors (binge eating, purging, dieting) and whether changes in affect mediated effects.

Method:

Female participants (N = 688) ages 15–30 from the Michigan State University Twin Registry reported their adaptive and maladaptive ER use, negative affect (NA), positive affect (PA), binge eating, purging, and dieting on 49 consecutive days. Using structural equation modeling, we examined whether within-person fluctuations in ER predicted same- and next-day ED behaviors and whether changes in affect mediated longitudinal ER effects.

Results:

Greater maladaptive ER predicted increased likelihood of same-day binge eating and next-day binge eating and purging. The association between maladaptive ER and next-day binge eating and purging was mediated by increased next-day NA. In contrast, dieting was more closely related to changes in PA. Adaptive ER did not predict reduced likelihood of any ED behavior.

Conclusions:

Maladaptive ER may longitudinally increase risk for binge eating and purging by amplifying NA. Interventions focused on decreasing maladaptive ER and subsequent NA might help disrupt binge eating-purging cycles. Conversely, results add to evidence that PA fluctuations may play a unique role in maintaining restrictive behaviors.

Trait emotion regulation (ER) difficulties (i.e., difficulty modulating the intensity and/or duration of emotions; Gross, 2014), are associated with eating disorders (EDs) (Aldao et al., 2010; Brockmeyer et al., 2014; Prefit et al., 2019) and core ED symptoms such as binge eating (Kenny et al., 2017; Whiteside et al., 2007). Some evidence suggests ER difficulties prospectively predict ED symptoms (e.g., dietary restraint; McLaughlin et al., 2011) and improve following treatment (Rowsell et al., 2016), indicating such difficulties may be an important risk and maintaining factor. ER difficulties are partly driven by difficulty selecting/implementing appropriate ER strategies (Tull & Aldao, 2015). Correspondingly, EDs and related behaviors are associated with increased trait-level reliance on maladaptive ER strategies such as rumination and self-criticism (Aldao et al., 2010), with decreased use of more adaptive strategies (e.g., cognitive reappraisal) also observed in some research (Danner et al., 2012, 2014).

While there is robust evidence of an association between trait-level ER difficulties and disordered eating, much less is known regarding how day-to-day changes in ER impact disordered eating at a moment in time. ER is a dynamic process, and the specific strategies people use vary depending on both internal (e.g., fatigue, hunger) and external (e.g., social support, social norms) contexts (Colombo et al., 2020). ED behaviors are likewise discrete events that wax and wane in frequency (Smyth et al., 2009). Within-person analyses that examine fluctuations in ER from a person’s baseline can therefore provide unique insight into why an individual may experience disordered eating on one day and not another, identifying direct and timely points for intervention.

Daily, within-person analyses can also identify potential mechanisms of ER effects, helping explain why ER difficulties may lead to disordered eating. Previously, we hypothesized ER strategy use might moderate the association between negative affect (NA) and dysregulated eating, such that NA would be less strongly associated with binge eating and emotional eating when participants used more adaptive/fewer maladaptive ER strategies (Mikhail et al, 2022). Contrary to our hypothesis, we found the relationship between NA and dysregulated eating was not impacted by ER strategy use. However, maladaptive strategy use was independently associated with binge eating on the same and next day, and also predicted significantly increased next-day NA. This raised the alternative possibility that ER might impact disordered eating by changing the trajectory of an individual’s affect rather than moderating the NA-dysregulated eating association. In other words, participants who used more maladaptive strategies might experience a persistence or escalation of NA, which would in turn increase the likelihood of disordered eating. This hypothesis would be consistent with research suggesting ED behaviors, and binge eating in particular, are typically preceded by increases in NA (e.g., Berg et al., 2015; Wonderlich et al., 2022). Preliminary evidence consistent with this hypothesis was identified by Smith et al. (2021) in the specific context of rumination, whereby rumination predicted increased NA, which predicted subsequent binge eating.

In the current study, we sought to test this alternative mediation model for maladaptive ER strategy use more broadly and address two additional important knowledge gaps. First, we expanded our analyses to include purging and dieting in addition to binge eating. Nearly all the research to date on daily ER and disordered eating has focused on binge-eating phenotypes (Forester et al., 2023; Mikhail et al., 2022; Smith et al., 2021; Svaldi et al., 2019), with little known regarding how ER may impact other core disordered eating behaviors. Though ER difficulties are observed transdiagnostically across EDs (Brockmeyer et al., 2014; Prefit et al., 2019), evidence is more consistent regarding an association between NA and binge eating than NA and restrictive symptoms (Haedt-Matt & Keel, 2011; Haynos et al., 2015, 2017; Mikhail, 2021). It is possible ER strategy use and subsequent changes in affect might impact restrictive behaviors differently from binge eating. Understanding potential differences in ER effects across different symptom presentations is critical to build tailored symptom maintenance models and targeted interventions.

Second, and relatedly, we added positive affect (PA) as an additional potential mediator of ER effects. PA may play a particularly important role in the maintenance of restrictive symptoms, potentially through both physiological (e.g., release of endorphins) and psychological (e.g., positive comments from others) reinforcement mechanisms (Coniglio et al., 2019; Engel et al., 2014; Haynos et al., 2017; Selby et al., 2014). Thus, ER effects on NA might be especially impactful for binge eating (and potentially purging), with PA processes playing a stronger role in restrictive behavior.

The primary aims of the current study were therefore to examine whether 1) NA and/or PA mediate associations between ER strategy use and disordered eating, and 2) the affective mechanisms linking ER strategy use to disordered eating differ across binge eating, purging, and dieting. We believe this was the first study to examine daily ER strategy use as a longitudinal predictor of purging and dieting and the first to test affect as a mediator of ER effects on multiple forms of disordered eating. Building on prior work, we hypothesized increased NA would mediate the association between maladaptive strategy use and next-day binge eating. We expected potentially similar effects for purging given that elevated NA also tends to precede this behavior (Haedt-Matt & Keel, 2011). Our hypotheses for dieting were more tentative given the absence of prior research, but we expected we might find an association with PA processes either in addition to or in lieu of NA mediation.

Method

Participants

Analyses included a population-based sample of 688 female twins (41.3% from complete monozygotic pairs, 54.7% from complete dizygotic pairs, and 4.1% without cotwin data) ages 15–30 (mean = 21.80, SD = 3.19) from the ongoing Twin Study of Exogenous Hormone Exposure and Risk for Binge Eating within the Michigan State University Twin Registry (MSUTR; Burt & Klump, 2013, 2019; Klump & Burt, 2006). MSUTR twins are representative of Michigan, suggesting results can be generalized to non-twin populations (Burt & Klump, 2019). All participants who completed the study by September 2022 (when data were pulled for current analyses) were included.

Participants were recruited through mailings based on birth records (n = 470; 68.3%), flyers (n = 78; 11.3%), social media (n = 72; 10.5%), community events (n = 29; 4.2%), and other methods (e.g., word-of-mouth, undergraduate research pool) (n = 39; 5.7%). Because the parent study focuses on combined oral contraceptives (COC) and binge eating, eligibility criteria included: 1) member of a female same-sex twin pair (as documented on birth certificates); 2) at least one twin taking COCs (participants not taking COCs (n = 54, 17.7%) required to have regular menstruation); 3) no pregnancy/lactation in the past 6 months; and 4) no history of genetic/medical conditions known to influence hormones/appetite/weight. Race and ethnicity were assessed via questionnaire based on US Census categories and NIH reporting requirements, and participants provided parental income to indicate socioeconomic background (SES). Demographic information is included in Table 1.

Table 1.

Descriptive statistics for participant demographics and symptoms (N = 688)

Participant Characteristics	Mean (SD) or N (%)
Zygosity
Monozygotic	284 (41.3%)
Dizygotic	376 (54.7%)
Cotwin dropped/not yet completed study	28 (4.1%)
Age	21.80 (3.19); range = 15–30
Racial identity
White	619 (90.0%)
Black/African American	35 (5.1%)
Asian/Asian American	8 (1.2%)
More than one race	26 (3.8%)
Hispanic/Latinx ethnicity	40 (5.8%)
Sex on original birth certificate
Female	688 (100%)
Gender identity
Woman/female	422 (99.1%)
Transgender	2 (0.5%)
Nonbinary/genderfluid/genderqueer	2 (0.5%)
Sexual orientation
Heterosexual	322 (75.6%)
Bisexual	38 (8.9%)
Queer	1 (0.2%)
Pansexual	8 (1.9%)
Gay/lesbian	9 (2.1%)
Asexual	36 (8.5%)
Not listed	11 (2.6%)
No response	1 (0.2%)
Combined parental income
<$20,000	11 (1.6%)
$20,000-$40,000	31 (4.5%)
$40,000-$60,000	67 (9.7%)
$60,000-$100,000	208 (30.2%)
>$100,000	348 (50.6%)
Not specified/prefer not to answer	23 (3.3%)
Body mass index (BMI)	24.33 (5.05); range = 16.48–58.12
Any binge eating during the study	187 (27.2%)
Days binge eating during the study for participants who reported this behavior	5.13 (6.31); range = 1–45
Any purging during the study	58 (8.4%)
Days purging during the study for participants who reported this behavior	3.00 (3.88); range = 1–21
Any dieting during the study	272 (39.5%)
Days dieting during the study for participants who reported this behavior	17.25 (17.06); range = 1–49

Note: Sample sizes for gender identity and sexual orientation are lower than the total sample size because these items were added after some participants had already completed the study.

Procedure

Participants completed daily questionnaires after 5 p.m. and as close to bedtime as possible each day for 49 days. Items/scales directed participants to think about their experiences today to orient them to the reporting timeframe . Staff called participants 1x/week to confirm protocol adherence and answer questions. Participants received full compensation (up to $300) if they completed ≥30 daily questionnaires and had ≤4 consecutive missing questionnaires. Compensation was prorated for lower completion rates. Additional assessments were completed at the beginning (“intake”), mid-point (~day 23; “intermediate”), and end (after day 49; “final”) of data collection. Dropout was rare (0.5%), and compliance was excellent (89% of daily assessments completed on average).

Measures

Disordered Eating Behaviors

Participants reported whether they had dieted and how many times they had experienced binge eating and purging each day. Because participants rarely reported binge eating or purging >1x per day, each disordered eating variable was analyzed dichotomously (0 = behavior absent that day, 1 = behavior present ≥1x that day). To ensure accurate reports of binge eating and purging, participants were given detailed definitions prior to daily questionnaires and quizzed on their understanding at intake and intermediate assessments. Definitions were also included on the daily questionnaire. These methods have been shown to improve the validity of self-reported disordered eating behaviors (Celio et al., 2004) (see Supplemental Material for additional validation data in the current sample).

ER Strategy Use

Daily ER strategy use was assessed using a questionnaire previously validated in young adults (Mikhail & Kring, 2019; Mikhail et al., 2022). This was the same questionnaire used in our prior study of ER moderation of affect-binge eating associations (Mikhail et al., 2022). Participants were first asked to recall when they experienced the most negative emotion that day, then reported the extent to which they used ER strategies at that time from 1 (not at all) to 5 (very much). Adaptive ER strategies included situation modification (“change an aspect of the situation”), cognitive reappraisal (“think about the situation differently”), acceptance (“accept how you were feeling”), and social sharing (“talk about how you were feeling with someone else”), while maladaptive strategies included expressive suppression (“keep yourself from expressing your emotions outwardly”), self-criticism (“criticize yourself for feeling the way you did”), and rumination (“continue to focus on how you felt and why you felt that way even after the situation ended”). Mean ER strategy use on this measure shows convergent validity with other established trait-level ER measures (i.e., Difficulties in Emotion Regulation Scale; Mikhail et al., 2022). Importantly, adaptive and maladaptive ER are only moderately correlated (r = .29), indicating they represent relatively distinct constructs.

This ER measure was added after the study was already underway, and thus was only completed by a subset of participants (n = 521; 75.7%). However, as discussed below, our analyses allowed inclusion of all participants even if they did not have data on ER strategy use.

Affect

Daily PA and NA were assessed using the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988). The PANAS contains two subscales measuring PA (10 items; e.g., excited) and NA (10 items; e.g., guilty). Participants rated the extent to which they felt each emotion from 1 (very slightly or not at all) to 5 (extremely) that day. Negative and positive emotion ratings were summed separately to form NA and PA subscale scores.

Data Analyses

General Analytic Approach

All analyses were conducted in Mplus version 8.6 (Muthén & Muthén, 1998–2021) using raw data and weighted least squares mean and variance adjusted (WLSMV) estimation with theta parameterization and a probit link function. Simulation studies suggest WLSMV is superior to maximum likelihood estimation for dichotomous outcomes in structural equation modeling (Beauducel & Herzberg, 2006). WLSMV estimation in Mplus makes use of all available data, meaning data can be included even from days/participants missing some variables. Simulation studies indicate this approach yields estimates with no more bias and greater efficiency than listwise deletion (i.e., removing observations with missing data) (Asparouhov & Muthén, 2010). Similarly, all participants were included in analyses even if they did not experience binge eating, purging, or dieting during the study because their data could inform at least some model parameters. Model fit was adequate if the root mean square error of approximation (RMSEA) was <0.08 (Browne & Cudeck, 1993) and the Tucker-Lewis index (TLI) ≥ 0.95 or the standardized root mean squared residual (SRMR) ≤ 0.90 (Hu & Bentler, 1999). Following recommendations (Preacher & Hayes, 2008), significant direct and indirect effects were determined using the percentile bootstrapping method with 1000 random samples with replacement; effects were significant if the 95% confidence interval did not contain zero.

All predictors were within-person centered (i.e., the value for a given day was subtracted from a participant’s mean across the study). Effects therefore represent how changes in ER strategy use and affect from a person’s own baseline impact their probability of engaging in disordered eating behaviors on a given day. Importantly, these within-person effects are independent of between-person individual differences (e.g., body mass index) that could potentially impact mean affect and ER strategy use.

The “cluster” option was used to adjust for dependence in the data due to participants reporting their behavior on multiple days (e.g., in calculating standard errors). Observations were not statistically clustered at the family level after within-person centering variables.

Structural Equation Model Parameters

The structural equation model is shown in Figure 1. The model was estimated for binge eating, purging, and dieting independently, with the same model structure for each behavior. In this model, day T represents the current day and day T-1 represents the prior day. The model first predicts disordered eating behaviors from concurrently measured NA, PA, adaptive strategy use, and maladaptive strategy use at both day T and day T-1 (e.g., T-1 NA predicting T-1 disordered eating, T NA predicting T disordered eating). On day T-1, these effects represent concurrent associations between affect/ER strategy use and disordered eating. On day T, these effects represent concurrent associations between affect/ER strategy use and disordered eating after accounting for prior day affect/ER strategy use.

Figure 1. Structural equation model for longitudinal associations between emotion regulation, affect, and disordered eating behaviors.

T = current day; T-1 = prior day; NA = negative affect; PA = positive affect; ER = emotion regulation strategy use; DE = disordered eating behavior (binge eating, purging, or dieting). All variables measured on day T-1 are in grey shaded boxes, while variables measured on day T are in white shaded boxes. Solid lines represent associations between predictors and engagement in DE on the same day, dotted lines represent associations between predictors and engagement in DE on the next day, and dashed lines represent associations between emotion regulation and next-day affect. Covariances between the same variable at different timepoints (e.g., T-1 NA and T NA) and different variables at the same timepoint (e.g., T-1 NA and T-1 PA) were modeled but are not shown for readability.

The model also predicts disordered eating behaviors on day T from NA, PA, adaptive strategy use, maladaptive strategy use, and engagement in disordered eating behaviors on day T-1. These parameters represent the direct longitudinal effects of affect/ER strategy use on next-day disordered eating and the autoregressive effect of disordered eating. Finally, the model predicts NA and PA on day T from adaptive and maladaptive strategy use on day T-1. These paths allow for calculation of indirect effects from adaptive and maladaptive strategy use on day T-1 to disordered eating on day T through NA and PA on day T. In other words, a significant indirect effect from day T-1 ER strategy use to day T disordered eating through one of these paths would imply that ER strategy use impacts the trajectory of affect into the next day, which in turn impacts the probability of next-day disordered eating.

Model coefficients reported in tables represent unstandardized probit coefficients. To aid in interpretation, effects are presented in the text as the difference in the probability of experiencing a disordered eating behavior on a given day for high (80^th percentile) versus low (20^th percentile) values of the within-person predictor. Note that the probability of experiencing a behavior on day T is also impacted by whether a participant experienced that same behavior on day T-1. Separate changes in probabilities for longitudinal effects are therefore provided for participants who did and did not experience the behavior on day T-1.

Results

Sample Descriptives

Overall, 187 participants (27.2%) reported binge eating on a total of 960 days (mean = 5.13 days during the study, SD = 6.31, range = 1–45), 58 participants (8.4%) reported purging on 174 total days (mean = 3.00, SD = 3.88, range = 1–21), and 272 participants (39.5%) reported dieting on 4,691 total days (mean = 17.25, SD = 17.06, range = 1–49) (see Table 1). Rates of purging in the current sample were in line with prior estimates in population-based samples of young women using self-report, while rates of binge eating and dieting were somewhat higher (Luce et al., 2008).

Relationships between Affect, ER Strategy Use, and Disordered Eating Behaviors

Model Fit

Model fit was excellent for analyses of binge eating (RMSEA = .008, 95% CI [.004, .012]; TLI = .971; SRMR = .004), purging (RMSEA = .008, 95% CI [.004, .013]; TLI = .966; SRMR = .006), and dieting (RMSEA = .008, 95% CI [.004, .013]; TLI = .998; SRMR = .003).

Binge Eating

Affect.

Greater NA on day T-1 was associated with significantly greater likelihood of binge eating on day T-1 (see Table 2). Adjusting for all other variables, the probability of binge eating on days with high NA was 14.3% greater than on days with low NA. The relationship between NA and binge eating remained significant on day T when accounting for prior-day variables. However, NA on day T-1 was not longitudinally associated with binge eating on day T. PA was not significantly related to binge eating on the same or next day.

Table 2.

Direct and indirect associations between emotion regulation strategy use, affect, and binge eating

Parameter	Unstandardized Probit Estimate	SE	Bootstrapped 95% Confidence Interval
Direct associations between predictors at T-1 and BE at T-1
T-1 NA → T-1 BE	.012	.004	.004, .019
T-1 PA → T-1 BE	−.002	.003	−.008, .005
T-1 Adaptive ER → T-1 BE	.000	.025	−.050, .048
T-1 Maladaptive ER → T-1 BE	.059	.026	.007, .110
Direct associations between predictors at T-1 and BE at T
T-1 NA → T BE	−.010	.006	−.022, .002
T-1 PA → T BE	.004	.006	−.008, .016
T-1 Adaptive ER → T BE	−.006	.048	−.106, .088
T-1 Maladaptive ER → T BE	−.017	.042	−.095, .069
T-1 BE → T BE	1.091	.115	.887, 1.345
Direct associations between predictors at T and BE at T, after accounting for the effects of predictors at T-1
T NA → T BE	.016	.005	.005, .026
T PA → T BE	−.003	.005	−.013, .007
T Adaptive ER → T BE	.007	.035	−.062, .072
T Maladaptive ER → T BE	.075	.038	.002, .146
Direct associations between ER at T-1 and affect at T
T-1 Adaptive ER → T NA	.155	.078	.009, .313
T-1 Maladaptive ER → T NA	.462	.065	.337, .595
T-1 Adaptive ER → T PA	.445	.093	.274, .638
T-1 Maladaptive ER → T PA	.122	.079	−.038, .277
Indirect associations between ER at T-1 and BE at T
T-1 Adaptive ER → T NA → T BE	.002	—	.000, .006
T-1 Adaptive ER → T PA → T BE	−.001	—	−.006, .003
T-1 Adaptive ER → T-1 BE → T BE	.000	—	−.052, .054
T-1 Maladaptive ER → T NA → T BE	.007	—	.002, .013
T-1 Maladaptive ER → T PA → T BE	.000	—	−.002, .001
T-1 Maladaptive ER → T-1 BE → T BE	.064	—	.006, .123

Note. T = current day; T-1 = prior day; NA = negative affect; PA = positive affect; ER = emotion regulation strategy use; BE = binge eating; SE = standard error. Significant parameters are bolded.

Maladaptive Strategy Use.

Maladaptive strategy use on day T-1 was independently associated with greater likelihood of same-day binge eating (see Table 2). The probability of binge eating on days with high maladaptive strategy use was 14.4% greater than on days with low maladaptive strategy use.

We also found two significant indirect effects from maladaptive strategy use on day T-1 to binge eating on day T. First, as hypothesized, there was a significant indirect effect from maladaptive strategy use on day T-1 to binge eating on day T through increased day T NA. In other words, maladaptive strategy use on one day was associated with binge eating on the next in part because it predicted increased next-day NA. The probability of binge eating on day T was 1.5% higher for participants with binge eating on day T-1 and 2.2% higher for participants without binge eating on day T-1 through this path. Second, there was a significant indirect effect from maladaptive ER on day T-1 to binge eating on day T through increased likelihood of binge eating on day T-1. In other words, maladaptive strategy use was associated with a greater likelihood of binge eating on the same day, which in turn predicted a higher likelihood of binge eating the next day. The probability of binge eating was 14.5% higher on day T following high versus low use of maladaptive strategy use on day T-1 through this path (note only one value is provided because this path assumes binge eating on day T-1).

The direct effect from maladaptive strategy use on day T-1 to binge eating on day T was not significant, meaning maladaptive strategy use was not directly associated with next-day binge eating after accounting for its impacts on same-day binge eating and next-day NA.

Adaptive Strategy Use.

Adaptive strategy use was not significantly associated with binge eating through direct or indirect pathways.

Purging

Affect.

Neither NA nor PA were significantly associated with same-day purging on day T-1 (see Table 3). However, the effect size for NA was similar to that observed in the model for binge eating, and NA was significantly associated with purging on day T after accounting for prior day variables.

Table 3.

Direct and indirect associations between emotion regulation strategy use, affect, and purging

Parameter	Unstandardized Probit Estimate	SE	Bootstrapped 95% Confidence Interval
Direct associations between predictors at T-1 and purging at T-1
T-1 NA → T-1 Purge	.015	.008	.000, .031
T-1 PA → T-1 Purge	−.001	.006	−.014, .011
T-1 Adaptive ER → T-1 Purge	.011	.060	−.085, .148
T-1 Maladaptive ER → T-1 Purge	.085	.054	−.012, .204
Direct associations between predictors at T-1 and purging at T
T-1 NA → T Purge	−.009	.014	−.039, .015
T-1 PA → T Purge	−.005	.013	−.030, .022
T-1 Adaptive ER → T Purge	−.152	.085	−.327, .007
T-1 Maladaptive ER → T Purge	−.129	.109	−.376, .049
T-1 Purge → T Purge	.988	.123	.726, 1.204
Direct associations between predictors at T and purging at T, after accounting for the effects of predictors at T-1
T NA → T Purge	.024	.010	.004, .044
T PA → T Purge	−.001	.009	−.019, .017
T Adaptive ER → T Purge	.061	.084	−.081, .245
T Maladaptive ER → T Purge	.115	.071	−.016, .261
Direct associations between ER at T-1 and affect at T
T-1 Adaptive ER → T NA	.155	.078	.008, .314
T-1 Maladaptive ER → T NA	.461	.065	.335, .594
T-1 Adaptive ER → T PA	.445	.093	.274, .638
T-1 Maladaptive ER → T PA	.122	.078	−.036, .278
Indirect associations between ER at T-1 and purging at T
T-1 Adaptive ER → T NA → T Purge	.004	—	.000, .009
T-1 Adaptive ER → T PA → T Purge	.000	—	−.008, .008
T-1 Adaptive ER → T-1 Purge → T Purge	.011	—	−.087, .134
T-1 Maladaptive ER → T NA → T Purge	.011	—	.002, .020
T-1 Maladaptive ER → T PA → T Purge	.000	—	−.003, .002
T-1 Maladaptive ER → T-1 Purge → T Purge	.084	—	−.012, .180

Note. T = current day; T-1 = prior day; NA = negative affect; PA = positive affect; ER = emotion regulation strategy use; SE = standard error. Significant parameters are bolded.

Maladaptive Strategy Use.

As with NA, maladaptive strategy use was not significantly associated with same-day purging on day T-1, but the effect size was slightly larger than in the model for binge eating.

Nevertheless, as with binge eating, greater maladaptive strategy use on day T-1 significantly indirectly predicted purging on day T through its association with greater NA on day T. In other words, purging was more likely on days after participants used more maladaptive strategies than typical for them because maladaptive strategy use predicted increased next-day NA. The probability of purging on day T was 3.3% higher for participants with purging on day T-1 and 4.4% higher for participants without purging on day T-1 through this path. There were no other significant direct or indirect effects from maladaptive strategy use on day T-1 to purging on day T.

Adaptive Strategy Use.

Adaptive strategy use was not significantly associated with purging through direct or indirect pathways.

Dieting

Affect.

As hypothesized, dieting showed a different pattern of effects than binge eating and purging (see Table 4). Higher PA, rather than NA, on day T-1 was associated with dieting on day T-1. The probability of dieting on days with high PA was 5.8% greater than on days with low PA. Interestingly, however, the likelihood of dieting was lower the day after a participant experienced high PA. High versus low PA on day T-1 was associated with a 12.8% decrease in the probability of dieting on day T for participants who dieted on day T-1 and a 54.7% decrease for participants who did not diet on day T-1. In other words, dieting was most likely when PA was low the prior day but higher on the current day. This could potentially point to a role for PA instability in daily dieting.

Table 4.

Direct and indirect associations between emotion regulation strategy use, affect, and dieting

Parameter	Unstandardized Probit Estimate	SE	Bootstrapped 95% Confidence Interval
Direct associations between predictors at T-1 and dieting at T-1
T-1 NA → T-1 Diet	.002	.002	−.001, .006
T-1 PA → T-1 Diet	.005	.002	.001, .009
T-1 Adaptive ER → T-1 Diet	.000	.015	−.029, .028
T-1 Maladaptive ER → T-1 Diet	.011	.014	−.018, .037
Direct associations between predictors at T-1 and dieting at T
T-1 NA → T Diet	−.007	.009	−.024, .010
T-1 PA → T Diet	−.021	.007	−.034, −.007
T-1 Adaptive ER → T Diet	−.023	.064	−.157, .096
T-1 Maladaptive ER → T Diet	−.040	.058	−.153, .081
T-1 Diet → T Diet	4.777	.383	4.108, 5.629
Direct associations between predictors at T and dieting at T, after accounting for the effects of predictors at T-1
T NA → T Diet	.009	.008	−.007, .025
T PA → T Diet	.022	.007	.009, .036
T Adaptive ER → T Diet	.006	.061	−.114, .130
T Maladaptive ER → T Diet	.055	.056	−.060, .165
Direct associations between ER at T-1 and affect at T
T-1 Adaptive ER → T NA	.155	.078	.009, .314
T-1 Maladaptive ER → T NA	.462	.065	.337, .595
T-1 Adaptive ER → T PA	.445	.093	.273, .639
T-1 Maladaptive ER → T PA	.122	.079	−.038, .277
Indirect associations between ER at T-1 and dieting at T
T-1 Adaptive ER → T NA → T Diet	.001	—	−.001, .005
T-1 Adaptive ER → T PA → T Diet	.010	—	.004, .019
T-1 Adaptive ER → T-1 Diet → T Diet	−.001	—	−.140, .138
T-1 Maladaptive ER → T NA → T Diet	.004	—	−.003, .012
T-1 Maladaptive ER → T PA → T Diet	.003	—	−.001, .007
T-1 Maladaptive ER → T-1 Diet → T Diet	.051	—	−.086, .180

Note. T = current day; T-1 = prior day; NA = negative affect; PA = positive affect; ER = emotion regulation strategy use; SE = standard error. Significant parameters are bolded.

Maladaptive Strategy Use.

Maladaptive strategy use was not significantly associated with dieting through direct or indirect paths.

Adaptive Strategy Use.

Adaptive strategy use was not significantly directly associated with dieting. However, adaptive strategy use had somewhat complex indirect relationships with dieting through its associations with PA. Because there was a significant association between adaptive strategy use on day T-1 and increased PA on day T, adaptive strategy use on day T-1 was also indirectly associated with dieting on day T though increased day T PA. The probability of dieting on day T was 0.7% higher for participants with dieting on day T-1 and 4.4% higher for participants without dieting on day T-1 through this path. Simultaneously, adaptive strategy use on day T-1 was significantly associated with higher PA on day T-1, which (as discussed above) was associated with lower odds of dieting on day T. Though the model does not specify a directional relationship between adaptive strategy use and PA on day T-1, this raises the possibility that adaptive strategy use could also decrease odds of dieting on day T by increasing day T-1 PA.

Supplemental Analyses with Individual ER Strategies

To understand which specific ER strategies might be driving effects, we conducted supplemental analyses with each ER strategy included independently in the model (see Tables S1–S3). Rumination and criticism of emotions were independently associated with next-day binge eating and purging through increased next-day NA, while reappraisal and acceptance showed the strongest associations with increased PA.

Discussion

While trait-level ER difficulties are strongly associated with EDs and disordered eating, much less is known about how within-person fluctuations in ER may contribute to disordered eating on a day-to-day basis. This was the first study to examine prospective associations between daily ER strategy use and multiple types of disordered eating (i.e., purging and dieting in addition to binge eating), with changes in both PA and NA as potential mediators. We found that greater within-person maladaptive strategy use predicted greater probability of binge eating and purging on the next day through increases in NA. In contrast, dieting showed a more complex pattern of associations that more directly implicated PA. Findings significantly extend our understanding of when and how different types of ER strategy use may contribute to the maintenance of multiple disordered eating behaviors and highlight potential points of intervention.

Results supported our hypothesis that daily maladaptive ER strategy use might contribute to binge eating and purging by increasing NA. Findings were also consistent with research by Smith et al. (2021) showing that NA may mediate the association between rumination and binge eating. While ER is often thought of as a response to emotions rather than a precipitant, it also plays a significant role in determining subsequent affect in daily life (Brans et al., 2013). Together with our earlier work (Mikhail et al., 2022), the current study suggests maladaptive ER strategy use increases risk for binge eating and purging not because it increases the likelihood an individual will respond to NA with disordered eating, but rather because it promotes the persistence of NA itself. This highlights the importance of interventions that can decrease NA in the moment to disrupt binge-purge cycles, including interventions to reduce reliance on maladaptive ER strategies such as rumination and criticism of emotions. Interestingly, adaptive ER strategy use was associated with increased PA but did not predict decreased NA or decreased binge eating and purging. Some strategies traditionally taught in therapies like CBT (e.g., cognitive reappraisal) may therefore be less helpful for immediately diminishing NA and preventing binge eating and purging in the moment.

Dieting showed different associations with ER strategy use and affect. Rather than NA and maladaptive strategy use, PA processes appeared most influential in shaping daily dieting. Specifically, participants were most likely to diet when they showed unstable patterns of PA, with lower PA on the prior day and higher PA on the current day. This finding is consistent with past research suggesting PA is lower prior to restricting episodes in individuals with anorexia nervosa and increases following restriction (Haynos et al., 2017). Emerging theories of affective dynamics in restrictive EDs suggest restrictive behaviors may serve an ER function by temporarily boosting PA, potentially contributing to persistent restriction even in the face of negative consequences (Coniglio et al., 2019). Participants in our study may have therefore engaged in dieting in part to increase momentary positive feelings of pride or accomplishment. Because daily adaptive ER strategy use predicted increased next-day PA, we also identified an indirect effect from adaptive strategy use to next-day dieting. Rather than causing dieting, however, it might be that adaptive strategy use and dieting both represent attempts to bolster PA following dips from a person’s baseline. Nevertheless, it is notable that participants who reported dieting during the study had lower mean PA, adjusting for age, body mass index, and NA (OR = .69, p = .002; see Table S4). Dieting may therefore suppress PA in the long term. Clinically, psychoeducation regarding the longer-term ineffectiveness of restriction to increase PA and efforts to boost PA in other areas of individuals’ lives might help disrupt maladaptive patterns of dieting.

Study strengths included a large sample that reported ER strategy use, affect, and disordered eating daily over an extended period; a high daily diary compliance rate; inclusion of multiple disordered eating behaviors; and longitudinal analyses that allowed us to identify mediators of effects. Study limitations included a sample consisting primarily of White, cisgender, middle-to-upper SES women in adolescence or emerging adulthood. Future studies should examine whether ER effects on disordered eating are the same in people with different genders, SES, and racial and ethnic backgrounds.

We conducted this study in a population-based sample; possibly, associations may be stronger or different in individuals with clinical EDs. In particular, dieting may hold a somewhat different meaning for individuals without EDs than for people with a restrictive ED, and NA may play a role alongside PA for individuals with more extreme restrictive behaviors. Severe restriction and/or low weight may also impact ER processes, and future research should examine moderation of ER effects by these variables. Additionally, while binge eating and dieting were fairly common in our sample, purging was rarer. However, it is notable that we were still adequately powered to detect relatively small increases in the probability of purging following maladaptive ER and subsequent increases in NA.

The current study’s daily diary design did not allow us to examine fluctuations in ER over the course of a single day, which could help to inform more immediate ER effects on disordered eating over minutes or hours. Asking participants to reflect on their experiences at the end of the day may have also resulted in some retrospective recall bias, which could have somewhat weakened our ability to detect associations due to increased measurement error. Additional ecological momentary assessment studies are needed to test ER effects on shorter timescales. Nevertheless, our findings are notable in suggesting a person’s approach to regulating their emotions may have lingering effects into the next day, highlighting the persistent impact of ER even at slightly longer delays.

Public Significance:

Little is known about how daily changes in emotion regulation may impact disordered eating. We found that maladaptive emotion regulation (e.g., rumination) was associated with a higher likelihood of binge eating and purging on the next day because it predicted increased next-day negative affect. In contrast, dieting was more closely tied to fluctuations in positive affect. Targeting daily emotion regulation and affective processes may help disrupt cycles of disordered eating.

Data Availability Statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.