Impact of an Internet-Based Lifestyle Intervention on Behavioral and Psychosocial Factors during Postpartum Weight Loss

Abstract

Objective:

This cluster randomized trial tested whether an effective internet-based weight loss intervention for low-income, postpartum women resulted in greater improvements in targeted social cognitive theory (SCT) constructs and psychosocial outcomes.

Methods:

Fit Moms/Mamás Activas was a 12-month cluster randomized controlled trial of 370 postpartum women at 12 WIC (Women, Infants, Children Nutrition Program) clinics who were randomly assigned to intervention or standard WIC. SCT constructs (self-regulation behaviors, cognitive restraint, self-efficacy, social support, body image) and psychosocial outcomes (depressive symptoms and stress) were measured at study entry and after 6 and 12 months.

Results:

The intervention versus standard WIC resulted in greater 12-month increases in weight control behaviors (3.28 points [95% CI, 1.06 to 5.50]), self-monitoring of weight and eating (2.28 points [1.17 to 3.40]), and cognitive restraint (1.93 points [0.78 to 3.08]); ps ≤ 0.004. The intervention effect was only modestly (η_p² = 0.02) mediated by improvements in these factors. No significant group x time effect was observed for disinhibition, self-efficacy, social support, body image, depressive symptoms, and stress.

Conclusions:

Among low-income postpartum women, an effective internet-based weight loss program resulted in improved weight control behaviors and cognitive restraint, but did not affect other SCT targets and psychosocial outcomes.

Trial Registration:

clinicaltrials.gov Identifier: NCT01408147

Introduction

The postpartum period is a time of increased risk of weight retention and gain (1,2), unhealthy eating and activity behaviors, stress, fatigue, depressive symptoms, and worsening body image (3–5). These factors may be particularly exacerbated among low-income Hispanic women (6–9) and those facing food insecurity (10–12).

While few effective postpartum weight loss interventions exist, an internet-based weight loss intervention promoted significant weight loss and reductions in waist circumference among low-income postpartum women in the Supplemental Nutrition Program for Women, Infants, and Children (WIC) (13). The intervention was rooted in social cognitive theory (SCT) and was designed to promote healthy eating, activity, and weight loss by increasing self-regulation of behaviors (self-monitoring, cognitive restraint), as well as self-efficacy, social support, and body image (14). Despite promoting significant weight loss, the intervention did not significantly affect self-reported diet and accelerometer-based measures of physical activity (13). However, the effect of the intervention on targeted SCT constructs and weight-related psychosocial outcomes, including depressive symptoms and perceived stress (15), remained unknown.

The purpose of this study was to examine whether an internet-based postpartum lifestyle intervention in WIC (vs. Standard WIC) resulted in greater improvements in targeted SCT constructs and psychosocial outcomes. A priori hypotheses were that the intervention relative to Standard WIC would improve practice of weight control behaviors, cognitive restraint, self-efficacy, social support, body image, and would result in fewer depressive symptoms and less stress.

Methods

Design

Fit Moms/Mamás Activas was a cluster randomized clinical trial. The trial protocol and primary outcomes have been published previously (13,14,16,17).

Participants

As previously reported (13,14), participants were recruited between July 2011 and May 2015 across 12 WIC clinics in counties along California’s central coast: Santa Barbara (n = 6), San Luis Obispo (n = 4), and Ventura (n = 2). The twelve clinics were randomized to two conditions: Standard care or Intervention. Participant eligibility was based on self-report and included being 6 to 12 months postpartum and having a body mass index (BMI) of 25 or more or a BMI of 22 to 24.9 but exceeding pre-pregnancy weight by 4.5 kg or more. Participants were between the ages of 18 to 40 years, spoke English or Spanish, were non-smoking, owned a cell phone, and had a fifth-grade education or higher.

Interventions

Standard WIC.

The standard care group received all aspects of the standard WIC program (18) plus a brief orientation to the study and newsletters every 2 months with basic information about healthy eating, activity, and wellness during the postpartum period.

SCT-based internet intervention.

The intervention group received all aspects of the standard WIC program plus a 12-month primarily internet-based weight loss program adapted from prior programs (19,20). The program was found to promote significant weight loss (13). Rooted in SCT, the intervention targeted eating, activity, and several related behavioral strategies (cognitive restraint, self-monitoring, stimulus control, problem solving) and psychosocial factors (self-efficacy, social support, body image) (13,14). The program provided a website with guidance and resources, automated feedback, weekly online lessons, a web diary, a weight and physical activity tracker, instructional and inspirational videos, and a message board. Weekly motivational text-messages were sent, and monthly group meetings at WIC were provided to promote or problem-solve website access. Intervention topics included selecting appropriate calorie goals, grocery shopping, label reading, restaurant eating, beginning exercise, self-monitoring, stimulus control, problem solving, social assertion, goal setting, body image, cognitive strategies for avoiding negative thinking and cognitive errors, overcoming stress and other barriers, relapse prevention training, strategies of successful weight losers, and numerous other topics (21,22).

Assessments

Research assistants were masked to randomization and conducted assessments at study entry, 6 months, and 12 months.

SCT Constructs.

Weight control practices were measured using the 26-item Eating Behavior Inventory (EBI; 23) that assesses the extent to which participants adhere to behavioral weight control strategies. Total score on the EBI was assessed, as well as separate scores for three of its subscales: Control of Eating (e.g., “I eat and just can’t seem to stop”), General Attention to Weight and Eating Patterns (e.g., “I weigh myself daily”, “I carefully watch the quantity of food which I eat”), and Stimulus Control Techniques (e.g., “I eat at only one place in my home”). Cognitive restraint and disinhibition were measured using subscales of the Eating Inventory (TFEQ; 24), a 51-item self-report instrument assessing dietary restraint, disinhibition, and hunger. The restraint factor assesses the degree of conscious control one exerts over eating behaviors, and the disinhibition factor measures susceptibility to loss of control over eating. Self-efficacy was defined as confidence in ability to lose weight, eat healthy, and exercise after pregnancy and was assessed using the self-efficacy subscale of the Kendall et al. attitudes questionnaire (25). Social support was assessed using a validated six-item scale (26), which focuses on instrumental (e.g., “help with caring for the baby”) and emotional (e.g., “tell you that you are doing a good job as a mother”) support for postpartum women. Body image was assessed using the validated 14-item Body Shape Questionnaire (27), which measures concerns about body shape and feelings of fatness over the four weeks prior to taking the questionnaire.

Psychosocial Outcomes.

Levels of depressive symptoms were assessed by the well-validated 10-item Edinburgh postnatal depression scale (28). Perceived stress over the past month was assessed using the validated 10-item Perceived Stress Scale (29).

Statistical Analyses

Generalized linear mixed-effect model (GLMM) was used for all analyses and included random effects for clinic, county, participant nested in clinic, and clinic nested in randomized group. GLMM was first used to test whether those who attended versus did not attend the final study (12-month) visit differed on sociodemographic variables (age, lactation, ethnicity, weeks postpartum at study entry, and BMI). Next, to assess whether the intervention vs. standard WIC was effective in improving SCT targets (weight control practices, restraint, disinhibition, self-efficacy, social support, body image) and psychosocial outcomes (depressive symptoms, stress), the GLMM also included participant-level covariates (baseline variable of interest, age, lactation, ethnicity, weeks postpartum at study entry, and BMI) that were identified a priori due to their association with postpartum weight loss in other studies (30). A group x time interaction term (fixed effect) tested whether the change in these outcomes over time differed significantly. Partial F tests using the Kenward-Roger approximation for degrees of freedom were used to simultaneously test all main effects and interactions. (Note that differences in degrees of freedom across analyses likely reflect variation in cluster sizes and patterns of missingness, and potentially other reasons (31)). If the group x time interaction was significant (p < 0.050), the equality of mean changes in the two groups at each intermediate time point was tested. Group differences in outcome parameter estimates and the CIs were estimated from the mixed models that adjusted for covariates. Our null hypotheses were that there would be no change in the SCT targets and psychosocial outcomes.

Other analyses explored whether the effect of treatment group on 12-month weight loss was mediated by SCT constructs. If a statistically significant treatment group difference with a SCT mediator was observed, subsequent GLMM examined whether the significant effect of treatment group on 12-month weight loss was changed after inclusion of the potential mediator in the model.

To evaluate the robustness of findings, sensitivity analyses using the same GLMM framework were conducted using imputed values for missing data. For the weight control measure, 15.68% and 16.49% of data were missing at the 6- and 12-month time point, respectively. For the restraint measure, 7.03% and 12.43% of the data were missing at the 6- and 12-month time point, respectively. Imputed values were assumed missing at random and calculated using fully conditional specification (32) and using the default number of 10 cycles. Inferences were obtained by comparing the results over the imputed datasets with the results from the original dataset (that had no imputations). The imputations and all analyses were done using IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, N.Y., USA).

Results

Participant flow into the study and participant demographic characteristics have been published previously (14). Of the 12 WIC clinics approached for participation, one intervention clinic withdrew, leaving a target and obtained sample size of 11 clinics and 374 participants (N = 34 participants per clinic). Four participants did not complete the full battery of psychosocial measures, leaving a final sample size of 370 for this study. Participants had a mean age of 28.07 years and were predominantly Hispanic (81.62%), a mean 7.82 kg above pre-pregnancy weight, and a mean 5.28 months postpartum. The two study groups did not significantly differ on any baseline measures (14). Participant retention in completing the full behavioral questionnaire was 92.97% at 6 months and 87.57% at 12 months. The demographic characteristics did not significantly differ between participants who attended vs. did not attend the 12-month visit.

Intervention Effects on SCT Targets and Psychosocial Outcomes

As summarized in Table 1, a significant interaction between group and time was observed for weight control practices, as measured by the EBI total score; there was a greater increase in weight control practices in the intervention group than standard WIC (F(2,969) = 14.83, p < 0.001; Figure 1). Weight control practice scores increased by 3.28 points more (95% CI, 1.06 to 5.50) in the intervention group compared with standard WIC over 12 months (p = 0.004). Examining patterns over time, the intervention was associated with an increase of weight control practices from baseline to 6 months (F(1,309) = 60.56, p < 0.001) but not from 6 to 12 months (F(1,280) = 1.18, p = 0.279).

Table 1.

Twelve-month changes (Mean ± SE) in SCT targets among intervention and standard WIC participants.

Outcome	Standard care (n = 196)	Intervention (n = 174)	Group	Time	Group × Time
Weight control (total score)			F = 8.39, P = 0.004	F = 43.78, P < 0.001	F = 14.83, P < 0.001
Baseline	74.12 ± 1.88	74.23 ± 1.90
6 months	75.51 ± 1.89	81.37 ± 1.93
12 months	76.64 ± 1.89	80.50 ± 1.93
Weight control (general attention to weight and eating patterns subscale)			F = 16.25, P < 0.001	F = 226.48, P < 0.001	F = 27.09, P < 0.001
Baseline	12.47 ± 0.95	12.57 ± 0.96
6 months	15.93 ± 0.95	20.16 ± 0.97
12 months	15.93 ± 0.95	18.45 ± 0.97
Weight control (stimulus control techniques subscale)			F = 2.30, p = 0.129	F = 10.27, P < 0.001	F = 2.76, p = 0.064
Baseline	12.42 ± 0.72	12.34 ± 0.73
6 months	12.59 ± 0.73	13.36 ± 0.74
12 months	12.99 ± 0.73	14.00 ± 0.74
Weight control (control of eating subscale)			F = 0.01, p = 0.920	F = 51.93, P < 0.001	F = 0.10, p = 0.947
Outcome	Standard care (n = 196)	Intervention (n= 174)	Group	Time	Group × Time
Baseline	28.33 ± 0.77	28.46 ± 0.78
6 months	26.03 ± 0.77	26.07 ± 0.79
12 months	26.30 ± 0.77	26.25 ± 0.79
Restraint			F = 10.87, p = 0.001	F = 14.92, P < 0.001	F = 17.88, P < 0.001
Baseline	10.84 ± 0.97	10.87 ± 0.99
6 months	10.51 ± 0.98	13.56 ± 0.99
12 months	10.91 ± 0.98	13.63 ± 0.99
Disinhibition			F = 0.73, p = 0.395	F = 4.35, p = 0.013	F = 1.77, p = 0.171
Baseline	5.38 ± 0.51	5.42 ± 0.52
6 months	5.13 ± 0.51	4.88 ± 0.52
12 months	5.28 ± 0.52	4.75 ± 0.52
Self-efficacy			F = 3.00, p = 0.083	F = 13.95, P < 0.001	F = 3.00, P = 0.050
Baseline	31.10 ± 0.51	31.14 ± 0.55
6 months	27.98 ± 0.55	30.27 ± 0.59
12 months	28.59 ± 0.55	29.18 ± 0.59
Social support			F = 1.34, p = 0.248	F = 8.38, P < 0.001	F = 1.74, P = 0.176
Baseline	38.28 ± 0.62	38.19 ± 0.66
6 months	37.39 ± 0.67	35.31 ± 0.71
12 months	36.06 ± 0.67	35.95 ± 0.72
Outcome	Standard care (n = 196)	Intervention (n= 174)	Group	Time	Group × Time
Body image			F = 0.72, p = 0.397	F = 54.92, P < 0.001	F = 1.47, p = 0.230
Baseline	51.65 ± 3.28	51.90 ± 3.33
6 months	44.22 ± 3.30	42.33 ± 3.35
12 months	43.83 ± 3.31	40.66 ± 3.36

Figure 1.

Effects of the intervention vs. standard WIC on weight control behaviors (i.e., total score on the Eating Behavior Inventory) over time. The intervention group improved in weight control behaviors to a greater extent than the standard care group.

Examining EBI subscales, a significant interaction between group and time was also observed for the EBI subscale reflecting General Attention to Weight and Eating Patterns, with greater increases in the intervention relative to standard WIC [F(2,969) = 27.09, p < 0.001]. While scores on this subscale increased in both groups, Attention to Weight and Eating Patterns increased by 2.28 points more (95% CI, 1.17 to 3.40) in the intervention group compared with standard WIC over 12 months (p < 0.001). Examining patterns over time, intervention-related increases in Attention to Weight and Eating Patterns occurred from baseline to 6 months (F(1,309) = 326.33, p < 0.001) and also from 6 to 12 months (F(1,280) = 12.29, p = 0.001). No significant treatment interaction effects over time were observed for the other EBI subscales: Stimulus Control Techniques or Control of Eating.

A significant interaction between group and time was also observed for cognitive restraint, with greater increases in restraint in the intervention group than standard WIC [F(2,1016) = 17.88, p < 0.001; Figure 2]. Restraint increased by 1.93 points more (95% CI, 0.78 to 3.08) in the intervention group compared with standard WIC over 12 months (p = 0.001). Examining patterns over time, standard WIC’s level of restraint remained generally stable, whereas increases in restraint were seen in the intervention from baseline to 6 months (F(1,326) = 31.92, p < 0.001) and were stable thereafter (F(1,305) = 0.011, p = 0.917). No significant treatment interaction effects over time were observed for disinhibition, self-efficacy, social support, or body image. In examining the effect of intervention group on psychosocial outcomes, no significant effects were observed on depressive symptoms or perceived stress (Table 2). Sensitivity analyses using imputing values for missing data indicated that missingness exerted little to no appreciable effect on the results (Table S1).

Figure 2.

Effects of the intervention vs. standard WIC on restraint over time. The intervention, but not standard care, group increased in restraint over the duration of the study.

Table 2.

Twelve-month changes (Mean ± SE) in psychosocial outcomes among intervention and standard WIC participants.

Outcome	Standard care(n = 196)	Intervention (n= 174)	Group	Time	Group × Time
Depressive symptoms			F = 3.57, P = 0.059	F = 4.74, P = 0.009	F = 1.03, P = 0.357
Baseline	5.37 ± 0.28	5.27 ± 0.29
6 months	6.50 ± 0.30	5.70 ± 0.31
12 months	6.12 ± 0.30	5.50 ± 0.32
Perceived stress			F = 0.28, P = 0.598	F = 6.25, p = 0.002	F = 0.54, P = 0.586
Baseline	15.88 ± 0.98	15.90 ± 0.99
6 months	17.11 ± 0.99	16.84 ± 1.01
12 months	16.68 ± 0.99	16.06 ± 1.01

Intervention Moderators and Mediators

Similar to our previous report (14), none of the demographic variables significantly moderated the intervention’s effects on 12-month postpartum weight loss. We explored whether the effect of the intervention on weight loss was mediated in part by observed changes in weight control practices and cognitive restraint. The effect of treatment group on weight loss was only modestly reduced with inclusion of EBI total score in the model examining the group effect on weight over time (partial eta square for group reduced from 0.06 to 0.05, ps < 0.001). Similarly, the effect of group on weight loss was modestly reduced with inclusion of restraint (partial eta square for group reduced from 0.06 to 0.04, ps < 0.001).

Discussion

An effective SCT-based internet weight loss intervention for low-income postpartum mothers was associated with increased cognitive restraint, defined as conscious control over food intake and counting calories, and an improvement of mother’s attention to weight and eating patterns, including daily self-weighing and self-monitoring of food intake. These factors contributed to the success of this primarily online postpartum weight loss intervention and should continue to be included in future postpartum weight loss interventions in WIC.

Prior research outside of the postpartum period has demonstrated the positive impact of dietary restraint for successful weight loss (33) and maintenance (34). In a study of low-income Latina women, greater dietary restraint was also associated with more healthful food choices (35). Similarly, frequent self-monitoring and self-weighing has been correlated with successful weight loss and increased weight loss success during SCT-based interventions outside of pregnancy (36,37).

During the postpartum period, however, women face many barriers to adhering to standard SCT-based weight loss programs. Increased fatigue, unexpected schedule changes, and increased childcare responsibilities may make it difficult for women to successfully monitor and manage food intake and practice other weight control strategies, and these barriers may be exacerbated among low-income women who lack resources (38–40). Here, the remote delivery of the intervention was designed to lessen the burden of weekly, face-to-face meetings that typify standard interventions. Cognitive restraint, self-weighing, and self-monitoring of food and calorie intake were encouraged and reinforced during weekly “check-ins” that were automated online and during monthly in-person meetings. In this context, the women in this study were able to increase practice of these critical weight control skills during the postpartum period. Indeed, our findings suggest that interventions designed for low-income postpartum women should include components to increase dietary restraint and attention to weight and eating patterns.

Self-efficacy is considered one of the most important determinants of behavior in SCT (41) and thus was another intervention target. However, it did not change relative to standard WIC. A previous observational weight loss intervention of postpartum women in WIC found that greater improvements in self-efficacy were related to greater weight loss (42). However, similar to the current study, dietary restraint emerged as the most significant determinant of postpartum weight loss, and weight loss self-efficacy, by comparison, did not contribute appreciably to the final model (42). Another possible reason self-efficacy did not improve for the intervention group relative to standard WIC may be the online format of the intervention. Although the intervention included frequent text messages and online lessons with motivational and encouraging words that were designed to improve self-efficacy for being physically active and eating a low-calorie diet, the impact of these messages might have been diluted in the context of a primarily online intervention. Further, the self-efficacy measure used here was validated, but only among rural pregnant women and not in Hispanic populations. Other studies have shown positive (42,43) or no (39) significant relationships between self-efficacy and weight changes in women. In this study, in both groups, self-efficacy declined over time. Future research aimed to improve the efficacy of current interventions is needed to determine effective methods to prevent the decline in self-efficacy observed during the postpartum period.

The SCT intervention also aimed to foster a socially supportive environment. However, the intervention had no significant effect on improving perceived social support. Importantly, the social support measure used here only assessed support from family and “especially from husband/partner” and was a global measure of ability to “count on” family/partner, to listen, help with household tasks, and encourage. This measure did not assess perceived support specifically for healthy eating and activity, which were targeted in the intervention. Also, because the intervention included only a single lesson that focused on improving assertion and communication skills with family and friends, it is likely this lesson was not intensive enough to change behaviors to elicit or improve perceived social support from partners and family. Of note, our intervention included methods to promote interpersonal and institutional support, including monthly group meetings with other WIC moms and support during regular meetings with WIC staff. Although not measured, participants could have experienced increased perceived support from these sources. Future research should examine the effects of more robust social support intervention components using a rich array of social support measures in the context of postpartum weight loss in WIC.

The intervention similarly did not result in improved body image, depressive symptoms, or stress. These components were either very minimally targeted by the intervention (e.g., one lesson addressed body image) or not directly targeted at all (stress, depressive symptoms). Typically, effective weight loss programs outside of the postpartum period result in improvements in these parameters. However, during the postpartum period, prior research in low-income postpartum women similarly found that changes in body image did not correlate with changes in weight (39). It is possible that the 2 kg weight loss produced by our study’s intervention might not have been enough to improve these parameters relative to standard WIC. Alternatively, limitations in the sensitivity of the measures and validity in low-income, Hispanic participant populations could further contribute to the null findings. Over time, both groups experienced worsening of body image, stress, and depressive symptoms, which is consistent with other postpartum research suggesting that negative body image, stress, and depressive symptoms pose significant barriers to achieving postpartum weight loss (44). Thus, future research is needed to bolster intervention components to address these domains.

This study had several strengths and some limitations. This study is the first to have large sample size, high retention (≥87%), and longitudinal follow-up in postpartum women using a comprehensive battery of psychosocial measures. The sample included predominantly Hispanic women who face food insecurity and in whom prospective, psychosocial assessment research had been lacking. The study used a standard care control group, which also had been lacking in other studies. Validated measures were used with published reliability and validity coefficients > 0.80, with the exception of the Eating Behavior Inventory, which had an odd-even split-half reliability of 0.62 and total score test-retest reliability of 0.74 (24–30). The study is limited by its assessments being based on self-report and subject to social desirability, which was not measured in the study. While the measures have been validated, they have had only limited prior validation in primarily Hispanic and low-income women. Also, while multiple dimensions of SCT exist (45), not all could be tested in this study (e.g., outcome expectancies, environmental factors) due to the high participant burden for completion of questionnaires. Lastly, findings from our study of low-income women in WIC in California may not generalize to other WIC or postpartum populations.

In sum, among low-income postpartum women, an internet-based weight loss program in addition to WIC and compared with the WIC program alone resulted in a significant increase in cognitive restraint and practice of weight control behaviors over 12 months. Interventions designed to promote postpartum weight loss should aim to increase cognitive restraint and mother’s attention to weight and eating patterns, including self-weighing and self-monitoring of food intake. Future postpartum research should continue to include psychosocial measures to allow a more comprehensive understanding of the relative contributions of social and behavioral factors to health outcomes in postpartum women and foster development of more effective interventions to support postpartum health.

Study Importance Questions.

What is already known about this subject?

• Weight loss interventions can improve practice of weight control behaviors and psychosocial functioning.

• The postpartum period is a time when many women experience weight retention, weight gain, and a worsening of psychosocial functioning.

• An internet-based postpartum lifestyle intervention for low-income women in the Women, Infants, and Children (WIC) nutrition program promoted significant weight loss and reduced postpartum weight retention.

What does this study add?

• An internet-based postpartum program that promoted significant weight loss also resulted in increased maternal attention to weight and eating patterns, daily self-monitoring of weight and eating, and cognitive restraint.

• Greater improvements in attention to weight and eating patterns, daily self-monitoring of weight and eating, and cognitive restraint were modestly related to greater postpartum weight loss.

• The intervention did not significantly affect disinhibition, self-efficacy, social support, body image, depression, or stress.

How might your results change the direction of research or the focus of clinical practice?

• Postpartum weight loss interventions should consider targeting cognitive restraint and attention to weight and eating patterns, including self-weighing and monitoring of food intake.

• Psychosocial factors should be taken into account when designing postpartum lifestyle interventions both to better understand these contributions to postpartum health and to develop more effective interventions.

• Future intervention research is needed to bolster intervention components to address worsening of self-efficacy, body image, stress, and depressive symptoms during the postpartum period.