Associations between fiber intake and Body Mass Index (BMI) among African-American women participating in a randomized weight loss and maintenance trial

Abstract

Introduction:

African-American women are at increased risk for obesity, and therefore it is important to identify dietary factors that have the potential to prevent weight gain within this population. The purpose of the current study was to examine associations between daily fiber intake and Body Mass Index (BMI) over the course of an 18-month weight loss intervention for African-American women.

Methods:

Anthropometric measures and the Block Food Frequency Questionnaire were administered at baseline, 6-month, and 18-month follow-up between 2008 and 2010. A mixed-effects linear regression model with random intercept and time slope was used to model associations between fiber consumption and BMI controlling for time trend.

Results:

Associations between fiber consumption and BMI were significantly different over time (${\widehat{\beta }}_{Fiber\ast Time}=-0.07$, $p-value=0.003$). There was no association between fiber intake and BMI at baseline; however, there was a significant inverse relation between fiber consumption and BMI at 6 months, and the association was even stronger at 18 months.

Conclusions:

Results from this study suggest that dietary fiber consumption may be particularly important within weight loss interventions tailored for African-American women.

1.1. Introduction

Obesity rates have remained substantially high over the last several decades, making it a significant public health concern.^{1, 2} African-American women have the highest prevalence of obesity compared to any other subgroup, with 56.6% classified as obese, compared to 32.8% of their European-American counterparts.¹ Obesity is associated with increased risk of several leading chronic diseases disproportionately experienced by African-American women including heart disease, type 2 diabetes, and some cancers such as breast and colon cancer.^{1, 3, 4} Thus, determining factors associated with weight management is especially important for African-American women.

Numerous epidemiological studies suggest that a diet high in fiber, especially intake of whole grains or cereal fiber, can prevent excessive weight gain.^5–9 However, these studies have been conducted in predominately European-American samples^{10, 11} and very few studies have focused exclusively on women.⁶ Results from an analysis of 12 years of dietary data from the Nurses’ Health Study revealed that women with the greatest increase in dietary fiber intake gained 1.53 kg less on average than females with the smallest increases in dietary fiber intake (p < 0.0001).⁶ These results were independent of age, body weight, body mass index (BMI), total energy, and macronutrients at baseline, and changes in age, exercise, smoking status, hormone replacement therapy status, alcohol use, and caffeine intake. In the same study, women in the highest quintile of dietary fiber intake had a 49% lower risk of major weight gain than did women in the lowest quintile (OR = 0.51; 95% CI: 0.39, 0.67; p< 0.0001). The analyses were conducted on over 74,000 women, but no subgroup analyses were reported on associations between fiber intake and weight specifically in African-American women. Given their elevated risk for obesity and related co-morbidities, there is a need for more evidence on associations between fiber intake and weight control among African-American women.

Evidence suggests African-American women have the lowest dietary fiber intake of any racial and ethnic group.^{12, 13} In a cross-sectional study examining dietary fiber intake by age, sex, and race, Storey and colleagues reported that, on average, non-Hispanic, black females consumed less fiber per day (12.7 grams/day) compared with European-American women (15.7 grams/day) and non-Hispanic, African-American males (15.0 grams/day).¹⁴ In another cross-sectional study examining the dietary quality of African-American and Hispanic families in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), African-American mothers had a median fiber intake of 8.8 grams/day compared to 14.5 grams/day in Hispanic mothers.¹³ Given that fiber intake is generally low among African-American women, it is possible that increased fiber intake may impact weight status in this high-risk group.

There is limited prospective evidence supporting the role of increased fiber intake and weight loss in African-American women.^{15, 16} Krishnan and colleagues examined the association between cereal fiber intake and risk of type 2 diabetes among 59,000 participants from the Black Women’s Health Study, and found an inverse association between cereal fiber intake and obesity. Although cereal fiber intake [IIR: 0.41 (95% CI, 0.24–0.72)] was protective for risk of type 2 diabetes, it was less protective in overweight women [IIR: 0.88 (95% CI, 0.74–1.04)].¹⁵ Further, low dietary fiber intake was also a significant predictor of weight gain in African-American women and men (n=1307) in the Coronary Artery Risk Development in Young Adults (CARDIA) study.¹⁷ Finally, in the National Heart, Lung, and Blood Institute Growth and Health Study, Affenito and colleagues found African-American girls (n = 1213) consistently consumed breakfast less often than European-American girls and consumed breakfast less frequently over the course of 9 years. Eating breakfast was significantly associated with increased fiber consumption and decreased BMI.¹⁶ No studies to date have investigated the associations between fiber intake and weight over time specifically among African-American women and no studies to date have examined this relation in the context of a weight loss trial.

Given the limited inclusion of African-American women in randomized controlled weight loss trials, there are fewer opportunities to examine the association between fiber intake and weight in this group. Successful weight loss studies that included a significant number of African-American women such as the Diabetes Prevention Program (DPP)¹⁸, the PREMIER trial¹⁹, and Look AHEAD²⁰ did not report on fiber intake by race specifically. The Obesity Reduction Black Intervention Trial (ORBIT) was a randomized controlled weight loss and weight loss maintenance trial for obese African-American women.^21–23 As part of the intervention, participants were encouraged to adopt a low-fat, high-fiber diet with increased fruit and vegetable consumption and decreased caloric intake. The purpose of the current study was to examine associations between habitual daily fiber intake and Body Mass Index (BMI) over time. The proposed study is the first to examine the association between in fiber intake and BMI during an 18 month randomized controlled trial weight loss lifestyle intervention trial for African-American women. We hypothesized that increased fiber intake would be associated with greater weight losses (lower BMI) over the course of the program. Identifying a relation between fiber intake and weight loss among this high-isk group may inform the development of future weight loss interventions.

1.2. Methods

Data from the currently study were collected between 2008 and 2010. Methods for the IRB-approved ORBIT trial have been reported previously.^21–23 Our analysis was restricted to women randomized to the weight loss intervention group given that the control participants did not receive any intervention related to diet. Women who provided informed consent completed anthropometric measures and the Block Food Frequency Questionnaire²⁴ at each time point: baseline, 6-month, and 18-month follow-up.

1.2.1. Measures

1.2.1.1. Demographics.

The sociodemographic questionnaire inquired about date of birth, education, income, employment status, and relationship status.

1.2.1.2. Anthropometrics.

Height was measured using a seca 214 portable stadiometer (seca, Hanover, MD), and weight was measured using a Tanita BWB-800 digital scale (Tanita Corporation of America, Inc., Arlington Heights, IL). Participants removed their shoes and any outer clothing for the anthropometric measurements. Height and weight were measured twice during the baseline interview, to the nearest 0.1 cm and 0.1 kg, respectively. If the two height measurements were more than 0.5 cm apart or if the two weight measurements were more than 0.2 kg apart, a third measurement was taken. The mean of the two closest measurements was used for analysis. BMI was then computed from height and weight (weight in kg/height in meters squared).

1.2.1.3. Block ‘98 Food Frequency Questionnaire (FFQ).²⁴

The Block FFQ was administered by trained interviewers and assessed frequency of dietary consumption in the past year and usual portion size for 110 different food and beverage items. Data from the Block were used to calculate nutrient intake, food group intake, and other dietary variables, including dietary fiber specifically. The food list was developed using food intake data from the National Health and Nutrition Examination Survey III (NHANES III). Reliability and validity have been established for the measure in a wide range of age, gender, income, and ethnic groups, including African American women.^24,25–29 Data was sent to Nutrition Quest for scoring (http://nutritionquest.com/).

1.2.2. Weight Loss Intervention period (baseline to 6 months)

The development and delivery of the intervention were based on Social Cognitive Theory (SCT)³⁰ and thus focused on changes in cognitions, behaviors, and social support related to weight loss. According to SCT, modeling or observational learning is a powerful contributor to behavioral change.³⁰ Individuals are more likely to model behavior that has a positive outcome, as well as the behavior of someone similar to themselves. The intervention provided ample opportunity for observational learning to occur, both via the interventionists and fellow participants. The intervention also presented multiple avenues to enhance self-efficacy, with curriculum focused on stimulus control, problem solving and other weight loss strategies, as well as the opportunity for monthly Motivational Interviewing (MI) sessions.

In addition to SCT, the intervention incorporated tenets related to the practice of culturally competent research.³¹ Culturally sensitive interventions require the recognition of the beliefs and practices of the particular social, ethnic and age group for whom the intervention is being developed, appreciation of the roles these factors play in participants’ lives, and considerate incorporation into the intervention.³¹ To create a culturally sensitive intervention we incorporated information gained from focus groups, a pilot intervention, and related scientific literature.

The weight loss intervention was conducted in a small group format and met twice weekly (60 minutes once per week and 90 minutes once per week) on the university campus. All participants were encouraged to adopt a low-fat, high-fiber diet with increased fruit and vegetable consumption and to increase their physical activity. Dietary objectives included reducing dietary fat to less than 30% of total daily calories, increasing dietary fiber to a minimum of 25 grams per day (this is equal to 14 g/1000 kal), and increasing fruit and vegetable consumption to a minimum of five servings per day.

In addition to group sessions, participants were encouraged to attend monthly individual MI sessions with a trained interventionist. These sessions were conducted face-to face or over the phone, lasting approximately 20–30 minutes. Each MI session addressed either diet or physical activity. The goal of the MI sessions was to assist individuals to work through their ambivalence about behavior change within a supportive climate where they felt comfortable expressing both the positive and negative aspects of their current behavior.

1.2.3. Maintenance Period (6 months to 18 months)

During the maintenance period, the frequency of the meetings decreased and the focus of the sessions was on structuring one’s lifestyle to support maintenance of health behaviors consistent with weight loss. During months 7–12, the group met twice weekly for 45–60 minutes, and each member received monthly MI sessions. During months 13–15, the group met once weekly for exercise class, and continued to receive MI sessions. From 16–18 months, there were no face to face group sessions, but women continued to receive MI sessions. Throughout the maintenance period, women received newsletters every other month.

1.2.4. Brief Summary of Primary Outcome Results at 6 and 18 months

Six and 18-month primary outcomes are reported elsewhere.^{22, 23} We report a brief summary here. At 6 months, women in the intervention group lost significantly more weight than women in the control group (p < 0.001). Women in the intervention group also consumed more fruit, had higher dietary quality, and were more physically active than those in the control group. At 18 months, both groups gained weight from 6 months, but the intervention group continued to have significantly lower BMIs than the control group.

1.3. Analytic Plan

As mentioned previously, the analyses were restricted to the intervention group (N=107). The main goal of the analysis was to investigate temporal associations between fiber consumption and BMI. The primary outcome was BMI reported at baseline, 6-month, and 18-month follow-ups. Descriptive statistics were calculated for participants’ demographic characteristics, height at baseline, weight, BMI, and fiber (g/1000 kcal). To examine associations between BMI and demographic and other covariates, we used bivariate mixed-effects linear models with a random intercept that controls for repeated measurements.³² Multiple mixed-effects linear regression model with random intercept and random time slope was used to model the association between BMI and fiber. In this model, we controlled for time trend. We considered controlling for physical activity and caloric intake in our model. However, the analysis of this data revealed that there was no statistically significant relation between physical activity and BMI nor caloric intake and BMI. Hence, the final model did not include physical activity nor caloric intake. Fiber intake was assessed at each time point and, hence, was modeled as a time-varying covariate. The model included the fiber by time interaction, which modeled concurrent associations between fiber and BMI and allowed for change in the association over time. To better understand the association between fiber intake and BMI, a second mixed-effects linear regression model was tested to model the association between fiber and weight loss category (more than 10% of initial weight lost at 18 months, less than 10% of initial weight lost at 18 months, and weight gain). In this model, we also controlled for time trend. The model included the weight loss category by time interaction, which modeled differential change in fiber consumption over time within each weight loss group. All analyses were conducted with SAS 9.4 (Cary, North Carolina) and R (version 3.2.2).

1.4. Results

1.4.1. Sample descriptive and bivariate regression results

Sample baseline descriptive data are presented in Table 1. Participants (n = 107) had a mean age of 46 (SD = 8.4), and a baseline BMI of 39 (SD = 5.5). About 39% of participants graduated from college or graduate school, 76% worked full time or were self-employed, and 35% were married. Mean annual income was $42,745. Mean BMI was 38.71 kg/m² at baseline (SD = 5.50), 37.70 kg/m² at 6 months (SD = 5.76), and 38.03 kg/m² at 18 months (SD = 5.92). Mean fiber (g/1000 kcals) was 8.72 (SD = 3.48) at baseline, 10.58 (SD = 3.89) at 6 months, and 10.59 (SD = 4.22) at 18 months. Associations of each of the potential predictor with BMI are presented in Table 2. Each model had subject-specific random effect that controls for repeated observations. The table parameter estimate is the regression coefficient that describes average change in BMI for one-unit increase in a covariate. For example, on average, an added consumption of 1 gram of fiber per 1000 kcal is associated with 0.24 kg/m² decrease in BMI, p < 0.0001. Time was also a significant predictor. The negative coefficients suggest a decrease in BMI over time.

Table 1.

Sample Baseline Characteristics

Variables	N	Mean(SD), %
Age (years)	107	46.43 (8.42)
Height (cm)	107	163.79 (5.74)
Weight (kg)	107	103.89 (67.40)
BMI (kg/m^2)	107	38.71(5.5)
Number of Children	107	1.70 (1.47)
Income, $K	102	42.75 (21.52)
Fiber (g/1000 kcal)	103	8.72 (3.48)
Education
HS graduate or less, GED	18	16.82%
Some college	47	43.93%
College graduate	28	26.17%
Professional degree	14	13.08%
Marital status
Single	35	32.71%
Married	37	34.58%
Widowed	4	3.74%
Separated/Divorced	31	28.98%
Employment
Full-time, self-employed	81	75.70%
Part-time	9	8.41%
Retired	3	2.80%
Unemployed	14	13.08%

Table 2.

Parameter estimates of the bivariate mixed-effect models for BMI

Variables	Estimate	Standard Error	p-value
Time	−0.22	0.08	0.0085
Age (years)	−0.003	0.06	0.97
Number of Children	0.51	0.36	0.17
Married vs. other	−0.11	1.13	0.92
Education (ref: professional degree)
HS graduate or less, GED	1.99	1.94	0.23
Some college	3.28	1.66
College graduate	1.83	1.78
Fiber (g/1000 kcal)	−0.24	0.04	<0.0001

1.4.2. Mixed-effects linear regression model of BMI and fiber association over time; Table 3

Table 3.

Parameter estimates of the mixed-effects regression model of BMI and fiber association over time

Variables	Estimate	Standard Error	p-value
Intercept	38.61	0.54	<0.0001
Time	−1.39	0.28	<0.0001
Time Square	0.39	0.08	<0.0001
Fiber (g/1000 kcal, mean centered)	−0.06	0.05	0.19
Time*Fiber	−0.07	0.02	0.003

Note. Mean BMI was 38.71 kg/m² at baseline (SD = 5.50), 37.7 kg/m² at 6 months (SD = 5.76), and 38.03 at 18 months (SD = 5.92). Mean fiber (g/1000 kcals) was 8.72 (SD = 3.48) at baseline, 10.58 (SD = 3.89) at 6 months, and 10.59 (SD = 4.22) at 18 months.

The final model included time, time squared, grams of fiber per 1000 kcals, and fiber by time interaction. Fiber was grand-mean centered around mean values of 9.92. The centering improves model estimation and interpretation of model parameters. The negative regression coefficient for linear time component and positive coefficient for quadratic term (${\widehat{\beta }}_{linear}=-1.39$ and ${\widehat{\beta }}_{square}=0.39$) suggested the diminishing effect of BMI reduction over time. BMI association with fiber consumption was found to significantly change over time (${\widehat{\beta }}_{Fiber\ast Time}=-0.07$, $p-value=0.003$). There was no statistically significant association between BMI and fiber consumption at the baseline, whereas consumption of fiber at the 6-month follow-up visit was significantly and negatively associated with BMI with an even stronger negative association at the 18-month follow-up (See Figure for details).

Figure.

Fiber association with BMI, kg/m² over time estimated from the final model

1.4.3. Mixed-effects linear regression results (Association between fiber and weight category; Tables 4 and 5)

Table 4.

Mean Fiber Intake by Weight Loss Category over Time

Weight Loss Category	N	Time	Mean	Std Dev
Lost greater than 10%	10	Baseline	10.55	3.95
		6 months	13.50	4.87
		18 months	16.82	5.23
Lost less than 10%	44	Baseline	8.70	3.13
		6 months	10.68	4.07
		18 months	10.07	4.02
Gained Weight	39	Baseline	8.63	3.68
		6 months	9.76	3.13
		18 months	9.51	2.46

Note. Weight loss category was calculated by determining those who lose greater than 10%, less than 10%, and gained weight at 18 months relative to baseline.

Table 5.

Mixed-effects linear model of fiber intake over time by weight loss categories

Variables	Estimate	Standard Error	p-value
Intercept	10.38	1.07	<.0001
Time	4.00	0.65	<.0001
Time*Time	−0.63	0.17	0.0002
Weight Category:
Gain	−1.93	1.19	0.11
No Loss	−1.56	1.17	0.18
Time*Weight Category:
Time*Gain	−1.76	0.42	<.0001
Time*No Loss	−1.68	0.41	<.0001

Note. For weight categories, weight loss is a reference group

Participants who had lost 10% or more of their initial body weight at 18 months had higher levels of fiber consumption at each time point compared to women who lost less than 10% or gained weight (see Table 4). Women who lost 10% or more of their initial body weight increased their fiber consumption over time (${\widehat{\beta }}_{linear}=4.00$, ${\widehat{\beta }}_{quad}=-0.63$) (see Table 5). The linear increase over time of fiber intake in women who lost less than 10% was almost 2 points lower compared to women who lost 10% or more, p-value<0.0001. Similar results were found in women who gained weight, ${\widehat{\beta }}_{inter}=-1.76$, p-value<0.0001.

1.5. Discussion

Obesity is a public health concern, particularly among African-American women.¹ Given the increased risk of obesity and related co-morbidities in this population, it is critical to identify factors that facilitate weight loss. Numerous epidemiological studies have found an association between increased fiber and lowered risk for excessive weight gain.^{5–9, 11} However, these samples have included predominately European-American individuals. .^{10, 11} African-American women, on average, consume significantly less dietary fiber than Latino and European-American women.^12–14 It is possible that this low level of fiber consumption impacts weight gain, weight loss, and weight maintenance in African-American women. Low fiber consumption in African Americans may partly be explained by socioeconomic status (SES).³³ Adults with lower income and levels of education consume fewer fruits, vegetables, and whole grains compared to those with higher SES.³³ Some African-American women are more likely to live in poor neighborhoods with low access to full service supermarkets and greater access to convenience stores. In neighborhoods served by smaller grocery stores, access to healthy foods including whole-grain products are specifically limited ³⁴. The current study was the first to examine the associations between habitual fiber intake and BMI during an 18 month randomized controlled lifestyle intervention trial for African-American women.

Results from this study suggest that, over the course of the 18-month weight loss and maintenance trial, there was a change in the relation between dietary fiber consumption and BMI over time. At baseline, there was no relation between dietary fiber consumption and BMI. However, at 6 months, there was a statistically significant negative association between dietary fiber consumption and BMI. This relation remained significant at 18-months and was even stronger than at 6 months. These findings are consistent with the few existing prospective studies investigating associations between fiber intake and weight in African-American women.^15–17

Our findings suggest that targeting fiber intake may be one important factor for weight loss and maintenance among African-American women. Overall, fiber intake was still relatively low at 8.72 g/1000 kcal (SD = 3.48) at baseline, 10.58 g/1000 kcal (SD = 3.89) at 6 months, and 10.59 g/1000 kcal (SD = 4.22) at 18 months. However, there was a slight increase over the course of the trial, and the mean at 18 months was less than 4 g/1000 kcal per day short of the study goal given to participants at the beginning of the trial. The ORBIT trial did include curricula on the importance of increasing fruits and vegetable consumption, but did not include substantial didactic information on fiber, specifically.^21–23 Previous weight loss and maintenance trials with African-American women included similar messages and behavioral targets around fruits and vegetables, but not explicitly fiber.^{35, 36} Although fruits and vegetables contain high levels of dietary fiber overall, there are a variety of other food groups which also contain high levels of fiber such as beans, nuts and seeds, and whole grains. Often, the main behavioral target of intervention trials is reducing daily calorie intake.^{35, 36} Although reduction of calories is essential for weight loss, given the relation between dietary fiber intake and prevention of excessive weight gain,³⁶ it may also be important to assist African-American women in building habits that include frequent consumption of fibrous foods.

It is possible that explicit targeting of dietary fiber might further enhance weight loss programs and weight maintenance. Examples of this might be incorporating a lesson on dietary fiber, its importance for weight loss and maintenance, and its benefits above and beyond weight loss (e.g., lowers risk of heart disease, stroke, hypertension, diabetes, colorectal cancer, enhances immune function⁷). Further didactic material could be presented regarding specific foods high in fiber intake. Participants could then set goals around increasing the consumption of their favorite high fiber foods. Future studies may consider incorporating these methods in an effort to enhance the initial weight loss and weight maintenance within this high risk population.

The current study has some notable limitations. First, our sample was comprised of only African-American women, and therefore findings cannot be generalized to other populations. Additionally, our analysis focused solely on the association between dietary fiber intake and BMI across 18 months, and not on associations between weight and other macro or micronutrients. Further, we did not assess SES directly and could not test the impact that SES might have had on the association between fiber and weight-related outcomes. Finally, although the FFQ is widely used for diet and nutrient intake, it has many limitations. There has been concern about not only random measurement errors in food frequency reports but also biases in reporting diet related to obesity. The FFQ as a dietary assessment tool is known to elicit high rates of under-reporting ³⁷. Reporting bias can compound this effect ^38–40.

Overall, however, our study provides compelling evidence of the importance of dietary fiber consumption among African-American women and the association between consumption and BMI across an 18-month weight loss and weight loss maintenance trial.