Psychosocial Predictors of Weight Loss by Race and Sex

Abstract

Objective

This paper examined psychosocial predictors of weight loss among race and sex subgroups.

Design

Analyses included overweight and obese participants from the PREMIER study, a previously published randomized trial which examined the effects of two multicomponent lifestyle interventions on blood pressure among pre-hypertensive and stage 1 hypertensive adults. Both intervention conditions received behavioral recommendations for weight loss and group sessions. Weight and psychosocial measures of self-efficacy and social support for diet and exercise were assessed at baseline and 6 months.

Results

There were 157 African American (AA) women, 46 AA men, 203 non-AA women, and 182 non-AA men with an average age of 50 years and average body mass index of 34 at baseline. Multiple predictor regression models were performed individually by race and sex subgroup. Among AA women, increases in diet self-efficacy were associated with weight loss. Among AA men, increases in diet related social support and self-efficacy along with increases in family support to exercise were associated with weight loss (all ps <0.05). Among Non-AA women, increases in support to exercise and exercise related self-efficacy were associated with weight loss, and among Non-AA men, only increases in diet self-efficacy were associated with weight loss (all ps<0.05).

Conclusions

These results emphasize the need for targeted interventions based on race and sex to optimize the impact of lifestyle based weight loss programs.

Introduction

Obesity is associated with significant medical expenditures, increased disease burden, and high risk of morbidity and mortality.(1-5) Approximately 36% of all adults in the United States are considered obese, however the rates of obesity among African American adults is disproportionately higher at 50%.(6) Although obesity rates for women in general have not been increasing in the last decade, the rates for African American females have been increasing as have the rates among men in general.(6) Prevention and treatment of obesity within sex and race subgroups deserves further examination.

Clinical trials have demonstrated that lifestyle approaches to weight loss achieve up to 10% weight loss in a 6 month period.(7,8) There have been numerous efforts to develop culturally tailored weight loss programs, yet reviews of clinical trials indicate that African American participants tend to lose less weight than non-African American counterparts.(9-16) Moreover, African American women tend to lose less weight than other subgroups.(15) Unfortunately, the reasons for these sub-group disparities are largely unknown.

Traditional lifestyle approaches used in weight loss trials include coaching or counseling along with recommendations for caloric restriction, increased physical activity and use of behavioral strategies such as goal setting, self-monitoring and problem solving. These approaches frequently incorporate social cognitive principles into education materials and coaching sessions. There is both theoretical support and clinical evidence suggesting that psychosocial variables such as social support and self-efficacy mediate behavior change associated with weight management.(17,18) Although social support is associated with health behavior change, it has been theorized that social support acts through its influence on cognition (e.g. self-efficacy). Social cognitive theory suggests that those with higher levels of self-efficacy, or confidence in completing a task, are more likely than their lower efficacious peers to take on a challenge and persist at the task.(19) In a social cognitive framework, social support plays an important role as a source of efficacy information. Many clinical trials utilizing lifestyle approaches have been too small in sample size or lack sufficient diversity to investigate race and sex differences among psychosocial predictors. Developing a better understanding of sex-race differences in psychosocial predictors could provide insight regarding targeted strategies to promote lifestyle changes for weight control.

The PREMIER trial was a multi-site randomized clinical trial that compared the effects of two multi-component interventions to a usual care control.(20,21) The trial demonstrated improvements in lifestyle changes (e.g., diet and exercise) and decreases in systolic blood pressure among African American (AA) men and women, as well as non-AA men and women.(22) Although the trial was focused on blood pressure control, behavioral weight loss was a key component of the intervention. The trial did not include a full complement of psychosocial predictors to examine any particular social cognitive theory in its entirety, but it did include the commonly identified predictors of weight loss, specifically social support and self-efficacy for both diet and exercise. This paper examines the association among race, sex, putative psychosocial predictors (e.g. social support and self-efficacy) and weight loss among PREMIER trial participants. Specifically, we examine changes in social-support and self-efficacy among subgroups and determine if self-efficacy mediates the relationship among social support and six month weight change.

Methods

Details of the PREMIER trial have been previously reported.(20,21,23) PREMIER participants were men and women age 25 years and older with pre-hypertension or stage 1 hypertension. Individuals on anti-hypertension medication, insulin or oral hypoglycemic drugs were excluded. Household members of another PREMIER participant or of a PREMIER staff member were ineligible for the study. The trial was conducted at four clinical centers (Baltimore, MD; Baton Rouge, LA; Durham, NC; and Portland, OR, USA), a coordinating center (Kaiser Permanente Center for Health Research), and the National Heart Lung, and Blood Institute project office. The study was approved by institutional review boards at each clinical center and all participants provided informed consent.

Participants were randomized into one of two treatment conditions (established condition, established plus DASH diet condition) or the advice only comparison condition. The advice only group received print materials and advice regarding limiting alcohol, limiting sodium, increasing physical activity, and losing weight. Materials were given at baseline and again after 6 month data collection. Participants in the two active treatment conditions shared common lifestyle recommendations of: 1) weight loss of at least 4.5 kg for those overweight and obese, 2) 180 mins/wk of moderate to vigorous physical activity, 3) reducing sodium to ≤ 2400 mg/day, 4) daily maximum of two alcoholic drinks for men and one for women, and 5) a caloric goal based on baseline weight. In addition to these recommendations those in the established plus DASH diet condition were encouraged to follow the DASH diet guidelines.(24) Participants in the two active treatment groups were offered 18 lifestyle behavioral sessions (14 group, 4 individual sessions) in the first 6 months of the intervention. During months 7-18, participants were offered monthly group sessions and 3 additional individual sessions. All sessions were theoretically-based on behavioral and social cognitive principles and focused on helping participants achieve lifestyle goals. Improving social support and increasing self-efficacy were two specific strategies presented in group and individual sessions. Social support was fostered through group session interactions and by providing strategies to increase social support from others such as family and friends.

Measures

Standardized measures of height and weight were taken by blinded study staff at baseline and weight was reassessed at 6 month follow-up. Weight was taken in light indoor clothing without shoes on a calibrated scale, and height was measured using a wall mounted stadiometer. Participants completed domain specific self-report questionnaires regarding social support and self-efficacy at baseline and 6 month follow-up. Social support for diet was measured using a modified 14-item Social Support and Eating Habits Survey, and participants used a 5-point Likert scale ranging from 1(none) to 5(very often) to indicate the amount of support from both family and friends.(25) Seven of the items measure encouragement (e.g. “reminded me to eat fruits and vegetables”). The encouragement items were summed independently for family and friends with higher encouragement scores indicating more support. The Cronbach's alpha for the current study ranged from 0.84 to 0.87 for support. Exercise social support was measured using the 11-item Social Support and Exercise Questionnaire using the same Likert scoring as the diet social support measure.(25) Scores on the 11 items were summed, with higher scores indicating more social support for exercise. Cronbach's alpha for the current study ranged from 0.91 to 0.92.

Diet self-efficacy was measured using a modified 27-item Eating Habits Confidence Questionnaire.(26) Participants indicated their confidence in performing specific diet related behaviors (e.g. “cook smaller portions so there are no leftovers”) using a 5-point Likert scale ranging from 1(I know I cannot) to 5(I know I can). Scores were summed and greater diet self-efficacy scores correspond with higher levels of self-efficacy. The Cronbach's alpha for the current study ranged from 0.90-0.91. Exercise self-efficacy was measured using the 12-item Exercise Confidence Survey. Participants indicated their confidence in performing a range of exercise related tasks (e.g. “exercise even though you are feeling depressed”) using a 5-point Likert scale ranging from 1(I know I cannot) to 5(I know I can).(26) Scores were summed and greater scores indicate higher levels of self-efficacy. The Cronbach's alpha for the current study ranged from 0.86 to 0.90.

Data Analysis

Analyses included obese and overweight participants with psychosocial predictor variables from all three treatment groups. We examined 6 month changes in the psychosocial variables and weight by subgroup. Correlations between social support and self-efficacy were examined separately for diet and exercise by subgroup. Next, changes in weight were regressed on changes in psychosocial variables by race and sex subgroups. Independent variables were centered on the median, and binary predictors were coded as -0.5 and 0.5.(27) Each predictor was tested individually to optimize statistical power and the analyses were adjusted for baseline weight. We also examined a multiple predictor models of weight change individually for race and sex subgroup. These models included all of the psychosocial predictors, age, baseline weight, and treatment condition.

Results

These analyses examined 588 PREMIER participants who were overweight or obese with an average age of 50 years (SD = 8.7) and average BMI of 33.7 kg/m² (SD = 5.5) at baseline. There were 157 African American (AA) women, 46 AA men, 203 non-AA women, and 182 non-AA men (see Table 1). Each race by sex subgroup had significant 6 month weight loss (see Table 2). Weight loss at 6 months was greater (p ≤ 0.05) for non-AA women (-5.0±5.6 kg), and non-AA men (-4.6±6.0 kg) compared to AA women (-2.4±3.9 kg). Each subgroup had significant 6 month decreases in diet and exercise self-efficacy. Although there were a few weak, statistically significant, associations between social support and self-efficacy (see Table 3), we did not find a consistent association among changes in social support and changes in self-efficacy within behavioral domains (i.e. diet and exercise) and participant subgroups.

Table 1.

Baseline Characteristics by Race and Sex.

	AA women	AA men	Non-AA women	Non-AA men
	n=157	n=46	n=203	n=182
	M(SD)	M(SD)	M(SD)	M(SD)
Age	48.7(9.0)	48.3(9.1)	50.7(8.3)1	50.7(8.8)1
Weight, kg	94.4(16.8)	107.5(20.4)1	92.1(17.4)2	102.4(17.4)1,3
BMI	34.6(5.5)	34.1(6.2)	34.0(5.6)	32.5(5.1)1,3
Diet
Family Support	14.6(6.8)	14.8(7.8)	12.0(5.5)1,2	14.8(6.7)3
Friends Support	12.3(5.9)	12.4(6.6)	10.8(4.9)1	9.4(3.7)1,2,3
Self-efficacy	119.9(13.3)	119.5(14.9)	120.3(10.8)	117.9(12.9)3
Exercise
Family Support	20.2(10.0)	19.9(9.6)	18.8(9.0)	23.0(9.3)1,3
Friends Support	20.1(10.0)	18.3(9.0)	17.5(8.4)1	14.8(6.7)1,2,3
Self-efficacy	49.6(7.8)	52.5(7.1)1	49.0(7.9)2	50.1(7.0)2

Note:

¹different from AA women, p < 0.05;

²different from AA men, p < 0.05;

³different from Non-AA women, p < 0.05;

Table 2.

Changes at Six Months for Weight and Psychosocial Variables by Race and Sex.

Change at 6 Month	AA women	AA men	Non-AA women	Non-AA men
	n=157	n=46	n=203	n=182
	M(SD)	M(SD)	M(SD)	M(SD)
Weight, kg	-2.4(3.9)*	-4.0(5.7)*	-5.0(5.6)*1	-4.6(6.0)*1
Diet
Family Support	1.1(6.3)*	1.8(8.0)	1.4(5.2)*	1.5(6.3)*
Friends Support	1.0(6.0)*	0.8(6.4)	1.6(5.4)*	1.9(4.6)*
Self-efficacy	-5.4(15.3)*	-2.8(15.8)	-4.6(12.9)*	-6.5(16.1)*
Exercise
Family Support	-0.4(8.6)	1.3(8.9)	0.9(7.8)	0.4(8.0)
Friends Support	-0.2(9.4)	-0.8(7.5)	1.5(8.2)*	1.4(6.6)*
Self-efficacy	-5.2(10.6)*	-6.5(8.3)*	-6.1(10.1)*	-5.4(9.8)*

Note: Change values were calculated so a negative number represents a reduction from baseline value.

^*significant change from baseline, p ≤ 0.05;

¹different from AA women, p < 0.05. No other between group differences were identified.

Table 3.

Correlation between Changes in Social Support and Changes in Self-Efficacy by Race and Sex.

	AA women	AA men	Non-AA women	Non-AA men
	n=157	n=46	n=203	n=182
	r(p)	r(p)	r(p)	r(p)
Diet
Family Support	.04(.59)	.09(.56)	.15(.03)	.07(.38)
Friends Support	.09(.25)	.02(.85)	.09(.19)	.03(.67)
Exercise
Family Support	.13(.10)	.02(.87)	.14(.04)	.14(.05)
Friends Support	.16(.05)	.03(.86)	.14(.05)	.09(.22)

Note: Correlations reflect association among changes in domain specific social support with changes in corresponding domain specific self-efficacy.

The single predictor models were adjusted for age, baseline weight, and treatment group and were conducted individually for each of the race by sex subgroups. A seen in Table 4, increases in diet self-efficacy were associated with weight loss in the single predictor models for all subgroups. Increased exercise self-efficacy was a significant single predictor of weight loss for all groups except AA women. The multiple predictor models were conducted individually for each of the race by sex subgroups, and were adjusted for the aforementioned variables along with social support and self-efficacy measures for both diet and exercise. The multiple predictor models explained 20-58% of the variance in weight change, and there was heterogeneity across subgroups regarding which putative predictors were associated with weight loss. Among AA women, increase in diet self-efficacy was associated with weight loss. Among AA men, increased diet support from family and friends in addition to increased diet self-efficacy were associated with weight loss. Among non-AA women, increase in friends' support to exercise and exercise self-efficacy were significant predictors of weight loss. For non-AA men, only increased diet self-efficacy was associated with weight change.

Table 4.

Association among Changes in Psychosocial Predictors and Weight Change at 6 Months by Race and Sex.

	AA women		AA men		Non-AA women		Non-AA men
	Single Predictor	Multiple Predictor	Single Predictor	Multiple Predictor	Single Predictor	Multiple Predictor	Single Predictor	Multiple Predictor
	β (p)	β (p)	β (p)	β (p)	β (p)	β (p)	β (p)	β (p)
Diet
Family support	-0.07(.39)	-0.05(.61)	-0.08(.63)	0.41(.01)	-0.11(.09)	0.03(.72)	0.01(.92)	0.08(.28)
Friends support	0.03(.72)	0.12(.18)	-0.39(.01)	-0.52(<.01)	-0.14(.04)	-0.07(.39)	-0.08(.20)	-0.10(.21)
Self-efficacy	-0.19(.01)	-0.21(.02)	-0.40(.01)	-0.31(.02)	-0.18(.01)	-0.10(.14)	-0.27(<.01)	-0.28(<0.01)
Exercise
Family support	-0.13(.09)	-0.12(.16)	-0.28(.07)	-0.46(<.01)	-0.18(.01)	-0.05(.48)	-0.04(.59)	-0.04(.56)
Friends support	-0.10(.20)	-0.06(.45)	-0.29(.05)	-0.03(.86)	-0.27(<.01)	-0.20(<.01)	-0.06(.39)	-0.03(.68)
Self-efficacy	-0.10(.19)	0.03(.72)	-0.32(.04)	-0.22(.13)	-0.21(<.01)	-0.14(.04)	-0.15(.02)	-0.01(.93)
Model
N	-	155	-	46	-	203	-	182
R2	-	0.19	-	0.55	-	0.25	-	0.34
p	-	<0.001	-	<0.001	-	<0.001	-	<0.001

Note: Change values were calculated so a negative number represents a reduction from baseline value. Standardized coefficients are reported. Single predictor models were adjusted for age, baseline weight, and treatment condition. Multiple predictor models were adjusted for age, baseline weight, treatment condition and all variables listed.

Discussion

The multiple predictor models explained between 20 and 58% of the variance in weight change and highlighted differences between subgroups. These are some of the first analyses to examine differences in psychosocial predictors of weight loss among race and sex subgroups within the same study. Increased self-efficacy for both diet and exercise was associated with weight loss among non-AA women and increased self-efficacy for diet was associated with weight loss for the other three subgroups. The association between self-efficacy and weight loss was significant even after controlling for treatment condition suggesting that this personal cognitive factor played an important role independent of the formal behavioral change program. Increased social support for diet was also a significant predictor of weight loss among AA men, which is notable as AA men were the only subgroup without a significant increase in diet social support. This suggests that although significant increases in diet related social support were not prevalent in this subgroup, increased support for diet changes were important. Future efforts should include a more targeted approach to helping this group develop stronger support for dietary changes.

Although a number of studies have reported the importance of social support for weight loss programs targeting AA women, we did not find significant association between changes in social support and weight loss in this subgroup.(9-14,14) In fact, increased diet self-efficacy was the only psychosocial predictor that was associated with weight loss for AA women. In this subgroup there were significant increases in diet related social support, but these increases were not identified as a significant predictor of weight loss. Social support for weight loss among AA women may be complicated by social norms as attractiveness is not as strongly linked to being thin among AA women compared to non-AA women.(28) Another explanation is that diet support was provided due to difficulty in achieving the weight loss goal; this is congruent with social support literature suggesting that increased support is received during times of need.(29) Similarly, non-AA men and women reported increases in diet related social support, however increases in social support were not associated with weight loss in either the single or multiple predictor models. It may be necessary to develop large social support reserves for the support to have a direct effect on weight loss.

The treatment conditions were designed to increase both social support and self-efficacy, yet there were declines in both diet and exercise self-efficacy. Decreases in self-efficacy are not unique to this study. It has been suggested that those entering physical activity interventions may be overly confident at baseline and underestimate the barriers and challenges associated with regular physical activity. (30,31) Similar phenomenon have been reported in weight loss studies.(30,31) It has been hypothesized that after the start of the study a participant develops a more realistic understanding of the related challenges and lower self-efficacy scores represent a more informed reporting on the construct. In addition, lower self-efficacy at the end of a study may occur as the participant considers the challenges of maintaining activity level and dietary changes without the support of the intervention program. Despite the average decreases in self-efficacy reported in this study, increases in self-efficacy were associated with weight loss. Specific efforts targeting those with high baseline self-efficacy may help ensure those who underestimated barriers to behavior change develop the necessary self-confidence to facilitate long term health behavior change.

Social support is theorized to be an important source of self-efficacy information, yet we did not find an association between changes in social support and changes in self-efficacy. Perhaps the changes in social support were not sufficient to increase self-efficacy levels. Alternatively these psychosocial predictors were operating independently within the sub-groups. Further efforts to examine the association among social support and self-efficacy in sub-groups should ensure the measures were determined valid within the target population. It has been noted that few efforts are made to ensure that psychosocial measures are validated among different populations.(32)

As noted above, one limitation of the study was the self-reported psychosocial measures. However, the measures used in this study showed acceptable reliability and have been used in previous behavior interventions.(17,18) Additionally, these analyses examined change scores focusing in intra-individual differences. Generalizability of the results are limited by the relatively high income status of the participants with 55.3% reporting household incomes of at least sixty thousand dollars per year (data not shown). Even though a larger proportion of AA (3%) compare to non-AA (6.1%) reported annual income below thirty thousand dollars, the limited sample size precludes further analyses adjusting for income status. The trial was not optimally powered for these secondary analyses; however, the diversity in this multi-center trial provided a unique opportunity to examine sub-group differences within the same study. Strengths of the study also include the randomized design with longitudinal data allowing us to establish temporal sequencing of the predictors and outcomes.

In conclusion, our results demonstrated significant sub-group differences in psychosocial factors associated with weight loss. These results may provide important insight for healthcare and wellness systems that attempt to provide a single weight loss program to meet the needs of a diverse membership. Unfortunately, it is not clear if these subgroup differences were due to the intervention being differentially received or if different subgroups rely on different change mechanisms. Increased efforts should target those with high self-efficacy at baseline to ensure any underestimation of barriers are sufficiently addressed to ensure successful behavior change. Moreover, statistically significant increases in psychosocial factors may not have clinical significance that translates into weight loss. Ensuring sub-groups receive additional resources and assistance in building social support reserves could further facilitate individual weight loss efforts. The translational application of these findings suggests that targeting lifestyle interventions based on race and sex would improve weight loss outcomes.

What is already known

• There are race-sex subgroup differences in weight loss success.

• Social support and self-efficacy are common psychosocial predictors of behavior change.

What this study adds

• Changes in psychosocial factors varied by race-sex subgroups.

• Statistically significant increases in psychosocial factors may not have clinical significance that translates into weight loss.

• There are race-sex subgroup differences in psychosocial predictors of weight loss success.