A Controlled Evaluation of Staging Dietary Patterns to Reduce the Risk of Diabetes in African-American Women

Participants in the EWLW were recruited individually through a social service agency that served as the program’s sponsor and through advertisements in neighborhood newspapers that targeted African-American audiences. African-American women ages 25–55 years and living in the neighborhoods were eligible for the study if they did not have diabetes, were not pregnant, and were >20% over ideal body weight, as determined by self-report (BMI >27). Eligible subjects were randomly assigned to a treatment or control group. Baseline data were collected before the intervention, at posttest intervals (immediately after the 3-month program), and at a 3-month follow-up assessment. The total study time period was 6 months. There was no further intervention during the posttreatment period for the treatment group. Participants in the control group did not receive any intervention during the treatment or follow-up phases, but were given a self-help workbook that reflected the content of the program and were offered a half-day workshop on healthy, low-fat eating after their follow-up assessment.

The EWLW was developed as a result of the collaborative partnership of health professionals affiliated with Washington University and peer educators from the Wellness Initiative of the sponsoring agency. Peer educators, who were African-American women from the target community with no background in nutrition or education, were recruited by the lead agency to deliver the intervention. The peer educators were trained by a team consisting of dietitians, social workers, and health educators over a 4-month period (∼3 half-days per week). A more detailed description of the training and peer-led, community approach is described elsewhere (14–16).

The manual-based program consisted of six group sessions (approximately six to eight participants per group) and six individual sessions with a peer educator, integrated over the 3-month intervention phase. Participants met weekly with the peer educator. Each individual session focused on a dietary pattern that represented a way to reduce fat in the diet, including “avoid fat as seasoning,” “substitution” of specially manufactured foods for higher fat counterparts, “modify meat” or removing fat and skin from meat, “avoid fried foods,” and “replacement” or replacing high-fat foods with fruits, vegetables, grains, and bread. During the individual sessions, the peer educator assessed each participant’s stage or readiness to change each of the five dietary patterns, and then tailored the session content to that stage.

The content of the six group sessions focused on specific skill areas that included the following: 1) “rate your plate” (participants learned how to assess the fat in their diet and target areas for change, 2) label reading (emphasizing portion size and the total fat and saturated fat content of food), 3) comparison shopping (emphasizing skills to purchase low-fat foods on a budget), 4) recipe modification (keeping culturally rich recipes in the diet while reducing fat content), 5) eating out (making healthy food choices in fast food and other restaurants), and 6) coping with high-risk situations.

To assess program integrity, sessions were randomly audiotaped and scored by independent raters using detailed session checklists. Results of the process evaluation indicated that the peer educators delivered 91.42% of the content across 12 sessions and that the overall accuracy of information delivered was 88.52% (averaged across the three peer educators). The EWLW process evaluation methods and results are discussed in more detail elsewhere (16).

Evaluation of the effectiveness of the intervention was performed on data from before and after treatment and from follow-up assessments on the following behavioral and physical outcome variables. In addition, structured interview questions were asked to obtain information regarding demographics and medical history. Demographic information included variables such as age, marital status, number of children, educational status, work status, and monthly income.

Participants’ knowledge was assessed by 15 items originally developed by Kristal and colleagues (17), and then modified for the present study.

This assessment tool was developed for this program and consisted of 10 items (Cronbach’s α = 0.74) that tested the respondent’s ability to interpret fat and calorie content in foods through reading sample food labels.

Attitudes were assessed using a revised 10-item scale (Cronbach’s α = 0.62) (17). Items measured attitudes regarding the importance of meat, models (what friends do), attitudes about high-fat meals, and attitudes toward eating fiber-rich foods, such as fruits and vegetables, on a four-point Likert scale (“strongly agree” to “strongly disagree”). Higher scores indicated healthier attitudes toward low-fat diets.

Eating patterns were assessed using the Eating Patterns Questionnaire, a 1994 revised version of the Fat and Fiber-Related Diet Habits Questionnaire (18). In the revised questionnaire, 34 items on a four-point scale (“always” to “never”) related to food patterns addressed in the program were assessed (total scale Cronbach’s α = 0.83). Five dietary patterns were measured: “avoid fat as seasoning” (α = 0.50), “substitution” (α = 0.61), “modify meat” (α = 0.67), “avoid fried foods” (α = 0.71), and “replacement” (α = 0.61). Higher scores on the Eating Patterns Questionnaire indicated lower fat dietary habits.

The Staging of Eating Patterns Assessment determined participants’ general readiness to perform each of the aforementioned five dietary patterns. Each pattern was assessed by one item; subjects responded to a five-point scale that indicated their degree of readiness to make or maintain changes based on Prochaska’s five stages of change. Validation of this assessment measure has been previously demonstrated (19). To present the findings in a parsimonious manner, stages were combined to create two variables: pre-action (including precontemplation, contemplation, and preparation) and action (including action and maintenance).

Participants’ daily intake of fat and energy was measured by the Food Frequency Questionnaire (FFQ). The FFQ has become a well-accepted method for quantifying usual nutrient intake because, in part, it minimizes the high intra-individual, day-to-day variability in nutrient intake without relying on multiple day assessments of actual foods consumed. The FFQ used in the present study was developed for the Women’s Health Trial-Feasibility Study in Minority Populations (20).

Each subject’s height and weight were assessed while she was wearing indoor clothing without shoes. Body weight was measured using a Health-O-Meter physician beam scale. Scales were calibrated quarterly using standard weights. BMI was calculated by taking the subject’s weight divided by her height squared.

The design of this study was a two (treatment versus control) by three (pretest, posttest, and follow-up) factorial design. For outcome measures that were either interval or ratio level measures, differences between the two conditions (pre- versus posttest and pretest versus follow-up) were tested using the ANCOVA procedure, with baseline values as the covariate. The t statistics and associated two-tail significance levels were based on the post hoc comparisons of each outcome measure between treatment and control conditions. For the staging of dietary patterns (action versus pre-action) variables, ANCOVA via logistic regression was used to determine significant differences between conditions.

The sample consisted of 294 African-American women who completed the pre- and posttest and follow-up assessments. Table 1 compares the sociodemographic characteristics of the participants in the treatment and control groups at baseline assessment. Comparisons between the participants in the two conditions indicated that there were no significant differences between the treatment and control groups in any of the participant characteristics. The retention rate for study participants was 73.7%. Comparisons were made to determine if there were any differences between the women who completed pre- and posttest and follow-up assessments (n = 294) and those who dropped out of the study (n = 104)—that is, completed pretests, but not posttests and follow-up tests. Results showed that those who completed the study were older than the dropouts (mean age 40.7 vs. 37.3 years, respectively; t (396) = 3.75, P < 0.001). Participation rates among those that completed the study showed that 68.6% attended at least 10 of 12 possible sessions (mean = 9.4 sessions).

Overall, participants in the treatment group reported a greater readiness to change their dietary patterns than those in the control group at the posttest assessment. Table 2 shows the percent of participants who reported being in the action stages (versus the pre-action stages) to perform each low-fat dietary pattern. χ² tests of the treatment effect, controlling for pretest scores via logistic regression, were significant for all dietary patterns. These significant differences were maintained at follow-up assessment.

As shown in Table 3, between-group comparisons of the knowledge of fat in diet at the posttest assessment revealed that there was a significant difference between the treatment and control groups (F[2,289] = 58.38, P < 0.0001). The treatment group had significantly higher scores than the control group, after adjusting for the baseline scores. This difference remained significant at the 3-month follow-up assessment (F[2,290] = 76.26, P < 0.0001). Skill-based knowledge as measured by the Knowledge of Label Reading Questionnaire also showed significant differences between groups at the posttest assessment (F[2,290] = 141.71, P < 0.0001) and remained significant at the follow-up assessment (F[2,291] = 133.49, P < 0.0001). Comparisons of the participants’ attitudes about diet and health revealed that there were no significant differences between the treatment and control groups at the postintervention or 3-month follow-up assessments.

The impact of the intervention indicated significant differences in the participants’ actual dietary behaviors (Table 3). Assessment of the participants’ dietary behaviors showed that the treatment group reported significantly more low-fat dietary patterns (total score) compared with the control group at the posttest (F[2,286] = 97.71, P < 0.0001) and follow-up assessments (F[2,290] = 57.38, P < 0.0001). Analyses of the separate dietary patterns were consistent with the results of the total scale, except for one pattern: there was no significant difference between the treatment and control groups in “replacement” at the posttreatment and follow-up assessments.

As shown in Table 3, dietary fat intake measured by the FFQ at the posttest assessment revealed a significant difference between the treatment and control groups. At posttest, the intervention was effective in reducing fat intake, as measured by the percent of calories from total fat (F[2,290] = 33.96, P < 0.0001). The post hoc t tests indicated that the treatment group reported significantly less fat intake than the control group, and that the differences were maintained at the follow-up assessment (F[2,291] = 29.52, P < 0.0001). The actual percent of calories from fat for the treatment group was reduced from 35.9% at pretest to 32.1% at posttest and 32.3% at follow-up, versus the same values for the control group (36, 35.6, and 34.5%, respectively).

Fat intake, as measured by the percent of calories from saturated fat, was also significantly reduced in the treatment group from pre- to posttest assessment (F[2,290] = 30.85, P < 0.0001). The post hoc t tests showed that the treatment group reported significantly less fat intake than the control group and that these differences were maintained at follow-up (F[2,291] = 25.59, P < 0.0001). The total daily energy intake of the treatment group was significantly lower than that of the control group from the pre- to posttest assessment (F[2,290] = 46.96, P < 0.0001) and at follow-up (F[2,291] = 30.75, P < 0.0001). Despite significant reductions in fat intake, results indicated that no significant group differences were detected in the weight and BMI of the participants.

This study addressed the following question: To what extent does a peer-led program that tailors content to participants’ readiness to make dietary changes reduce fat intake and increase low-fat dietary patterns, and maintain these changes over a follow-up phase? Other intermediate outcomes of interest included increasing skill-based knowledge (e.g., label reading) and knowledge and attitudes about dietary fat. The EWLW was evaluated using an experimental design with 3-month intervals between pre- and posttest assessments and posttest and follow-up assessments.

A critical finding in this study was that greater reductions were seen in fat intake of the EWLW group when compared with the control group and that the reductions were maintained at the 3-month follow-up assessment. At the posttest assessment, women in the EWLW condition had reduced their fat intake by 3.8% vs. 0.4% for the control group. At follow-up, fat intake remained significantly lower in the EWLW than in the control group (32.3 vs. 34.5%), which brought the former group closer to the public health goal of ≤30% of calories from fat, as recommended by the Healthy People 2000 guidelines (21). Within-group comparisons of the control group showed a slight decrease in fat intake from pretest to follow-up assessment. One possible reason for the decrease in fat intake in the control group is the effect of testing. Because subjects completed three 45-min food frequency questionnaires over 6 months, it is possible that by the follow-up assessment, the control group subjects had increased their awareness of fat and overall food intake simply by completing the questionnaires, and that this influenced their eating habits. This is particularly plausible given that the women in the control condition were motivated to change their eating habits, as evidenced by the fact that they responded to our recruitment efforts and may have sought out other strategies to modify their eating patterns outside of our program.

A second finding was the significant increases in low-fat dietary patterns among the EWLW group compared with the control group at posttest and follow-up assessments. It is likely that changes in dietary patterns led to the reduction of total fat intake. This was true across all dietary patterns except for “replacement” (replacing high-fat foods with fruits, vegetables, grains, and bread). Several explanations for why participants did not make significant changes in this dietary pattern are possible: 1) content on “replacement” was delivered in the 11th session (out of 12 sessions), and 38.7% did not receive the content on “replacement” because of attrition; 2) the program did not emphasize increasing intake of fiber (i.e., fruits and vegetables), but rather focused on reducing total fat; and 3) this pattern may involve a greater change in cuisine in that the replacement food looks and tastes different than the high-fat food. The data on increasing participants’ readiness to change dietary patterns is consistent with the above-mentioned findings on dietary patterns. A greater percentage of participants in the treatment group moved from pre-action to action stages across all dietary patterns, and maintained their significant gains at the follow-up assessment.

Intermediate outcomes, such as knowledge of fat in foods and reading and interpreting food labels, were increased and maintained for the participants in the treatment group when compared with the control group. Attitudinal change was not influenced by the intervention. This may be in part have been because of the low-to-fair reliability of the total attitudinal scale (α = 0.62). Another explanation for this result may be that attitudes about food and dietary preferences are difficult to change because they are embedded in family tradition and ethnic and cultural practices.

There was no significant weight loss among participants at the follow-up assessment. The reasons for the lack of weight loss are unclear. This finding may be attributable to measurement problems associated with dietary assessment. Specifically, it is possible that the women in both conditions, either through social desirability or difficulty in recalling intake over the previous month, underreported their intake. However, there was consistency in underreporting across pre- and posttest and follow-up assessments and across conditions, which lends some support to the suggestion that the differences between groups may have been real.

It is important to note that the objective of the EWLW was not to lower caloric intake but to lower fat intake and change dietary patterns, a recognized strategy for reducing the risk of diabetes (5,6). This focus, which proved effective, was culturally appropriate, as African-American women are not generally as concerned about weight loss as their Caucasian counterparts (22–24). Also, changing dietary patterns may be a first step toward other diabetes risk-reduction strategies, as successful attainment of more immediate goals holds relevance as a motivating factor for more long-term prevention goals (e.g., maintaining weight loss).

Finally, these findings do not negate the importance of weight reduction in diabetes prevention, but rather suggest what strategies might work for certain outcomes in this population. Reduction in fat intake alone does not necessarily result in a reduction of weight. Future programs that emphasize increases in fiber (fruits and vegetables) and physical activity, in addition to decreases in fat intake, may be more effective in decreasing weight and BMI than the EWLW.

In conclusion, the gathered data indicated that a stage-based intervention conducted by trained peer leaders in the community is effective in changing dietary patterns and reducing fat intake among low-income African-American women. The EWLW program uses strategies that could be expanded to include a greater emphasis on physical activity and increased fiber intake. A peer-led approach with the collaboration of a community organization that is located in the target neighborhoods holds promise for reducing the risk of diabetes among lower-income African-American women.

This work was supported by grant no. RO1-DK-48143 from the National Institutes of Diabetes and Digestive and Kidney Diseases, and the Office of Research on Minority Health of the National Institutes of Health to the George Warren Brown School of Social Work of Washington University.

We would like to thank the following individuals for their contribution to this project: the staff of the Wellness Initiative of Grace Hill Neighborhood Services; Hope Krebill, BSN, RN; and Alan Kristal, PhD.