There were no significant differences at baseline in characteristics of the subjects assigned to the 2 intervention groups (). Subjects ranged in age from 22 to 60 y, and 76% were employed full-time. At 12 mo, 71 subjects (73%) remained in the trial; these subjects are referred to as the study completers (). The reasons given most frequently for withdrawal were inability to attend the counseling sessions because of schedule conflicts and relocation out of the area. There were no significant differences in baseline characteristics, including body weight, between the participants who completed the trial and those who withdrew.
Baseline characteristics in the 2 intervention groups of all randomly assigned subjects and for subjects who completed the trial1
Changes in body weight and body composition
After 1 y, study completers in both intervention groups (n = 71) had lost a significant amount of weight compared with their baseline values (P < 0.0001; ). Over the course of the trial, however, study completers who were advised to reduce their dietary fat intake and increase their intake of water-rich foods (RF+FV group) had a significantly different pattern of weight loss than did subjects who were advised to reduce their fat intake alone (RF group) (P = 0.002 for group × time interaction; ). During the first 6 mo of the trial (phase 1), completers in the RF+FV group lost more body weight (8.9 ± 0.8 kg, or 19.6 ± 1.8 lb) than did those in the RF group (6.7 ± 0.7 kg, or 14.7 ± 1.5 lb; P = 0.034), a difference of 33%. During the second 6 mo of the trial (phase 2), the pattern of weight change did not differ significantly between the groups (P = 0.056). In phase 2, when subject contact was less frequent, weight was well maintained; subjects in both groups regained an average of 0.7 ± 0.4 kg (1.5 ± 0.9 lb). At the end of the trial, mean weight loss was 7.9 ± 0.9 kg (17.4 ± 1.9 lb) in the RF+FV group and 6.4 ± 0.9 kg (14.1 ± 1.9 lb) in the RF group.
Body-composition measures in study completers over the course of 1 y1
FIGURE 2 Mean (± SEM) change in body weight for study completers in the reduced-fat (RF) and reduced-fat plus increased fruit and vegetable (RF+FV) intervention groups over time. Random coefficients analysis was used to model the longitudinal response (more ...)
The analysis of all randomly assigned subjects (n = 97) yielded conclusions similar to the analysis of study completers: over the course of the year, subjects in the RF+FV group lost significantly more body weight than did subjects in the RF group (P = 0.021). When all available data were included, the estimate of weight loss at the end of the trial was smaller in both groups (6.4 ± 0.8 kg in the RF+FV group and 4.9 ± 0.8 kg in the RF group). As in the analysis of completers, during phase 1 of the trial, there was a greater decline in body weight in the RF+FV group than in the RF group (P = 0.003), and during phase 2, the pattern of weight change did not differ significantly between the groups (P = 0.074).
Over the course of the study, study completers in both groups showed significant decreases in BMI, percentage body fat, and waist circumference from baseline values (P ≤ 0.0001; ). Subjects in the RF+FV group had a significantly different pattern in the change in BMI than did subjects in the RF group (P < 0.0001 for group × time interaction). During phase 1, completers in the RF+FV group had a greater decrease in BMI than did those in the RF group (−3.3 ± 0.3 versus −2.5 ± 0.3; P = 0.0001), whereas during phase 2, changes in BMI did not differ significantly between the groups (−2.9 ± 0.4 in the RF+FV group and −2.4 ± 0.4 in the RF group; P = 0.514). After 1 y of intervention, 49% of subjects in the RF+FV group were no longer classified as obese (BMI ≥ 30) compared with 28% of subjects in the RF group (P = 0.071). Measured body fat decreased by an average of 7% for subjects in both groups. Apart from differences in weight loss and BMI, there were no significant differences between the groups in body-composition measures.
Reported diet composition, hunger and satiety, and physical activity
Examination of diet records indicated that study completers in both groups made dietary changes that resulted in a significant decrease in the energy density of their diets (P < 0.0001; ). Participants in the RF+FV group, however, reduced the energy density of their diets during the intervention to a greater extent (1.23 ± 0.02 kcal/g) than did those in the RF group (1.46 ± 0.02 kcal/g; P =0.019). Diet records showed that subjects in both groups significantly reduced their fat intake from 34.5 ± 0.6% of energy at baseline to 28.4 ± 0.4% during the intervention (P ≤ 0.0001). Fat intake did not differ significantly between the groups (P = 0.50; ); this was expected because both groups received the same advice for dietary fat reduction. Similarly, reported intakes of carbohydrates and protein did not differ significantly between the groups during the intervention; the groups consumed a mean of 54.4 ± 0.5% of energy from carbohydrate and 18.0 ± 0.2% of energy from protein. The groups differed, however, in reported fruit and vegetable intake; this was expected because the dietary advice concerning fruit and vegetables differed between the groups. Subjects in the RF+FV group reported a greater intake of fruit and vegetables (P = 0.037; ) than did those in the RF group; this was true regardless of whether fried and dried fruit and vegetables were included in the analysis. Subjects in the RF+FV group also consumed significantly more dietary fiber (g/1000 kcal; P = 0.001).
FIGURE 3 Mean (± SEM) food energy density (kcal/g) for study completers in the reduced-fat (RF) and reduced-fat plus increased fruit and vegetable (RF+FV) intervention groups over time. Differences in dietary energy density between the groups were determined (more ...)
Dietary measures in study completers over the course of 1 y1
Diet records during the intervention showed a reduction in energy intake from baseline values of ≈500 kcal/d (P <0.0001; ), which did not differ significantly between the groups. The difference in weight loss between the groups corresponded with an estimated difference in energy intake of ≈100 kcal/d, whereas the SE of the difference in energy intake between the groups was 113 kcal. Thus, given the variability in reported energy intake, we were not able to detect a difference in energy intake between the groups that would account for the difference in weight loss.
There was a significant difference between the groups in the weight of food consumed during the trial. Subjects in the RF+FV group reported consuming significantly more food than did subjects in the RF group (P = 0.025; ). Specifically, the subjects who were encouraged to eat satisfying portions of low-energy-density, water-rich foods (RF+FV group) consumed a mean of 225 g (25%) more food daily.
Daily ratings of hunger, as measured by visual analogue scales at the end of the day, did not differ significantly from baseline for subjects in the RF group. Over the course of the study, however, hunger ratings reported by subjects in the RF+FV group were significantly lower than their ratings at baseline (P = 0.030). This resulted in significantly different ratings for hunger between the groups during the intervention (RF+FV group: 46.7 ± 2.2; RF group: 53.5 ± 2.2; P = 0.003). Ratings for fullness did not differ significantly between the groups. Analyses to assess the reliability of the daily measures showed that daily hunger ratings were significantly correlated with hunger ratings collected before and after meals; these results will be reported in a separate publication.
Physical activity, as measured by step counts, increased significantly from baseline for subjects in both groups (P < 0.0001; ). Step counts did not differ significantly between the groups during the trial (P = 0.50); this was expected because the advice given for physical activity was the same for both groups. Subjects in both groups increased their daily step counts by an additional 2427 ± 100 steps/d to a total of 8735 ± 70 steps/d over the course of the trial. This exceeded the goal of an additional 2000 steps/d and increased the average daily step count by 38% from baseline, an increase that was maintained over the course of the year.
Overall scores for satisfaction were significantly higher for the intervention diets (3.6 ± 0.1) than for the baseline diets (3.2 ± 0.1) for subjects in both groups (maximum score 5.0; P < 0.0001). There were no significant differences between the 2 dietary approaches in any of the ratings of diet satisfaction. Subjects in both groups rated their intervention diets as providing greater health benefits, having fewer negative aspects, working better within the family context, and leading to less preoccupation with food than did their baseline diets (P < 0.0001). Ratings of perceived cost effectiveness, preparation time, and convenience were lower for both diet strategies than for the diets consumed at baseline (all P < 0.05), but there were no significant differences between the groups.
Although there were significant changes from baseline on scores from the Eating Inventory, the Beck Depression Inventory II, and the Eating Habits Checklist, there were no significant differences between the groups. Analysis of the Eating Inventory showed that for all subjects there was a significant increase in dietary restraint scores at 6 and 12 mo (mean score 14.7 ± 0.3; P < 0.0001), whereas scores for disinhibition and hunger decreased at these time points (mean scores: 7.3 ± 0.3 and 4.4 ± 0.3, respectively; P < 0.0001). There were significant improvements in depression scores for subjects in both groups (mean score: 4.6 ± 0.4; P < 0.0001), as well as a significant decrease in scores for degree of binge-eating problems at both 6 and 12 mo (mean score: 7.7 ± 0.4; P < 0.0001).
Laboratory outcomes and blood pressure
Subjects showed beneficial changes from baseline in several physiologic values, even though baseline values were within normal ranges. Significant changes from baseline were seen in both groups for insulin, and at some time points for all cholesterol measures except LDL cholesterol (). Improvements in physiologic measures were significantly greater in the RF+FV group than in the RF group at various time points during the study for insulin, non-HDL cholesterol, and triacylglycerols (all P < 0.044; ). For LDL cholesterol and HDL cholesterol, no significant differences were found between the groups.
Insulin, lipid, and blood pressure measures in study completers over the course of 1 y1
There were increases in both groups over 1 y in serum α-carotene, lycopene, and β-cryptoxanthin concentrations (all P < 0.03). The RF+FV group showed a significantly greater increase during the trial in α-carotene than did the RF group (54% increase versus 37% increase; P = 0.023). For lutein and zeaxanthin concentrations, there was a significant change between groups over time; subjects in the RF+FV group had greater concentrations of lutein and zeaxanthin than did subjects in the RF group at 6 mo only (P = 0.047).
For all subjects, there was a significant decrease from baseline in systolic and diastolic blood pressure (; P < 0.001); however, there were no significant differences between the groups in the blood pressure measures over time (both P > 0.85).
Factors predicting weight loss
To investigate which factors were most predictive of weight loss, stepwise regression analyses were performed by using data from dietary measures (energy intake, energy density of food, total fat intake, combined fruit and vegetable intake, and total weight of food), physical activity, and scores from the Eating Inventory. The primary predictor of weight loss at month 1 was the energy density of food (R2 = 0.16) along with step counts (R2 = 0.05; both P < 0.040). Together, these factors accounted for 21% of the variability in weight loss. Energy density was the only significant predictor of weight loss at month 2 (R2 = 0.07; P = 0.022) and also at month 3 (R2 = 0.17; P = 0.0004). The strength of the relation (regression coefficient) between food energy density and weight loss increased significantly from month 1 to month 3 (P = 0.025).
At months 6, 9, and 12, fruit and vegetable intake became the primary predictor of weight loss from baseline (R2 = 0.26, 0.16, and 0.20, respectively; all P < 0.0006). The strength of the relation (regression coefficient) between fruit and vegetable intake and weight loss did not vary significantly from month 6 to month 12 (P = 0.92). Secondarily, the increase in the Eating Inventory dietary restraint score and the decrease in the Eating Inventory hunger score predicted weight loss during the second phase of the study (all P < 0.029). These scores predicted an additional 20% of the variability in weight loss at month 6 and an additional 11% at month 12.