Beverage vs. Solid Fruits and Vegetables: Effects on Energy Intake and Body Weight

Authors


(mattes@purdue.edu)

Abstract

Beverage consumption has been implicated in weight gain, but questions remain about the veracity of the association, whether the relationship is causal and what property of beverages is responsible. It was hypothesized that food form is the most salient attribute. Thus, a randomized controlled trial of food form was conducted. Energy-matched beverage or solid forms of fruits and vegetables were provided to 34, lean or overweight/obese adults for two 8-week periods with a 3-week washout interspersed. Dietary compensation was incomplete (beverage 53% solid 78%) and body weight increased after the beverage (1.95 ± 0.33 kg) (77% fat mass) and solid (1.36 ± 0.30 kg) (85% fat mass) treatments (both P < 0.0005). Differences between food forms were not significant. The lean group had the highest dietary compensation (119%) and no significant weight change (0.84 ± 0.53 kg) after consuming the solid fruits and vegetables whereas the overweight/obese group had lower compensation and significant weight gain during the solid arm (46%, 1.77 ± 0.32 kg, P < 0.0001). In contrast, incomplete dietary compensation and weight gain occurred in both the lean (43%, 1.61 ± 0.44 kg, P = 0.003) and overweight/obese (61%, 2.22 ± 0.47 kg, P < 0.0005) groups during the beverage arm. Secondary analyses revealed the obese group gained more weight than the lean and overweight groups during the beverage intervention (P = 0.024). These data demonstrate energy consumed as beverages may be especially problematic for weight gain. They also indicate that advice to increase fruit and vegetable consumption should emphasize total energy intake because the additional energy contributed may promote weight gain, especially among overweight and obese individuals.

Introduction

Recent reviews have concluded that energy-yielding beverages are (1,2) or are not (3) related to higher body weight and most call for further investigation. Some disagreement arises from methodological issues (4) such as differential inclusion of studies to target specific populations (e.g., children (3) vs. adults (1,2)) or inclusion of studies that are purported to address the role of beverages or food form when the trial actually tested questions related to energy source (5,6) or approaches to achieve behavior change (7,8,9,10). Concern has been focused on sweetened beverages because they are the largest source of beverage energy in the diet (11), but findings that beverages containing different energy sources elicit weak dietary compensation indicate the medium (i.e., beverages), not the energy source, is the likely basis of the association between drinking and energy balance (12). Only one small study has directly tested the effects of food form on energy intake and body weight (13). Those data showed an increase in body weight with beverages only compared with baseline, but not between beverage and solid food forms. Here it was hypothesized there would be increased energy consumption and a gain of body weight during the beverage study arm compared with baseline and the solid study arm; we further hypothesized there would be no change or a reduction of body weight during the solid arm.

Fruits and vegetables were used for the energy-matched solid and beverage loads because both forms are commercially available in an array of options to meet different dietary preferences. In addition, both forms are commonly consumed in meal and snack settings, thereby reducing potential confounds related to customary use patterns (13). Finally, increased consumption of these items is recommended in the Dietary Guidelines for Americans (14) based on evidence that their consumption reduces the risk of selected chronic diseases.

As reviewed previously (15), increasing fruits and vegetables in the diet may lead to increased (16,17), decreased (18,19,20), or no change (21,22,23) in body weight. Consistent with findings in the general population (24), the present study participants were low fruit and vegetable consumers. The quantity of these foods they were required to consume was in accord with the Dietary Guidelines. Thus, the current intervention also provides insights on the effects of including recommended amounts of fruits and vegetables in the diet on weight status.

Finally, purposeful recruitment of individuals in divergent BMI categories allowed comparisons of responses to each food form in lean and overweight/obese individuals. The hypothesis was that beverage energy elicits lower dietary compensation and increased weight gain in overweight/obese individuals compared to lean counterparts.

Methods and Procedures

Subjects

Participants were a convenience sample from West Lafayette, Indiana. Their characteristics included: 18–38 years of age (mean 23 ± 1), BMI of 18.4–23.0 kg/m2 (mean 20.9 ± 0.3) (lean group) or 27.4–33.5 kg/m2 (mean 29.9 ± 0.4) (overweight/obese group), weight stable (within 5% of usual body weight over the last 3 months), agreement to maintain habitual activity patterns throughout study duration, low fruit/vegetable consumer (mean 3.1 ± 0.2 servings/day (including potatoes)), not taking medications known to influence appetite, nondiabetic (mean fasting blood glucose 81.2 ± 1.2 mg/dl), nonrestrained eater (mean score 5.7 ± 0.5 on restraint scale (25)), and willingness to consume the study foods. The sample size calculation, conducted with data from a preliminary group of participants, indicated 27 participants were needed to detect a weight change difference of 1.55 kg with a standard deviation of 2.74 (α = 0.05, power = 0.80). Body weight was the primary outcome in this study and the only parameter used to calculate power. Participants were enrolled between July 2005 and March 2008. The Purdue University institutional review board approved this protocol. Each participant signed a consent form and received monetary compensation for participation.

Thirty-one participants completed the protocol. An additional three participants completed half of the study. All 34 participants were included in analyses by using group means in place of missing values (Figure 1).

Figure 1.

Participant flow diagram.

Experimental design

This 21-week, randomized, cross-over design included 1 baseline week for each study arm (beverage or solid) followed by an 8-week intervention with a 3-week washout period between treatments. During each intervention period, participants were required to consume, by random assignment, their dietary load (either beverage or solid) every day. No other dietary or activity guidance was provided. Group assignments were made (after securing consent) by J.A.H. using an internet-based random number generator. By chance, a higher proportion of participants received the beverage intervention first, but no order effects were observed. Participants were informed that the trial was designed to assess the effects of fruit and vegetable consumption on antioxidant status.

Food load

The dietary load was based on 20% of each individual's estimated energy requirement (Harris Benedict equation with an activity factor of 1.55) and rounded to the nearest 50 kcal (between 400–550 kcal/day). The solid dietary energy load was comprised of 10% vegetables. This amount of raw carrots, broccoli, and cauliflower was ∼1.3–2.8 servings of vegetables/day based on the USDA National Nutrient Database for Standard Reference (26). 90% of the dietary energy load was fruit (40% as fresh fruit and 60% as dried fruit). Total servings of vegetables and fruits were 6–8 per day. The beverage load included commercially available juices that corresponded to fruits consumed during the solid study arm. Wheat dextrin was added to the beverages to match the soluble fiber intake during the solid arm. All fresh fruits and vegetables were washed, portioned, and provided ready to eat. Participants picked up their foods at least weekly.

Body weight and composition

Fasting body weight was measured (± 0.01 kg) at baseline and end of each study arm, with the participants wearing a bathing suit or similar clothing. Body composition was measured by air displacement plethysmography (Bod Pod) at baseline and end of each study arm with participants in a similar fasted and hydrated state.

Dietary intake and pattern

Dietary intake data were collected by 24-h diet records on 3 days (one weekend day) at each baseline and final intervention week. Energy and macronutrient data were analyzed using the University of Minnesota Nutrition Data System for Research 2005 and 2006. Daily dietary pattern data were analyzed by counting the total number of eating events (energy-containing food or beverage consumed within 15 min blocks) and those comprised of: intervention foods alone, no intervention foods, or intervention foods with other self-selected items for each study arm. Dietary compensation values were calculated as:

100 − (((Difference in total energy intake during intervention vs. baseline)/(Intervention energy load)) × 100)

Activity energy expenditure

Activity energy expenditure was estimated using tri-axial accelerometers (Stayhealthy, Monrovia, CA). These data were collected on one weekday and one weekend day during each baseline and final intervention week.

Resting energy expenditure (REE) and thermic effect of feeding (TEF)

REE and the TEF were measured by indirect calorimetry (Medical Graphics, St Paul, MN) at the end of each study arm to compare acute and chronic responses to the beverage and solid loads. Participants arrived to morning appointments rested and 12-h fasted. Immediately upon arrival, they were placed in a supine position on a bed in a semi-darkened room. After a 15-min rest period, they were fitted with a face mask and gas exchange data were collected for 60 min. The final 30 min of data were used to determine the REE. After completion of REE measurement, participants consumed a standard 400 kcal load (beverage or solid) within 20 min. This load was 200 kcal of apple juice and 200 kcal of grape juice during the beverage arm of the study or 200 kcal of fresh and dried apples (∼80 kcal as fresh apple) and 200 kcal of fresh red grapes and raisins (∼80 kcal as grapes) during the solid arm of the study. Gas exchange data were collected for an additional 5 h after meal ingestion (TEF).

Compliance

Group compliance to the intervention was documented with ten 12-h fasted blood draws to measure plasma ascorbic acid and carotenoids. Fasting blood samples were collected in 3 ml EDTA tubes at baseline, 3 randomized weeks during the intervention, and the final intervention week for each arm of the protocol. Samples were protected from light to prevent photo-oxidation of carotenoids. All samples were kept cool (4 °C) during centrifugation and were stored at −80 °F until analysis. Plasma alone was stored, in the dark, for analysis of carotenoids. Plasma for analysis of ascorbic acid was stabilized with an equal volume of chilled 10% meta-phosphoric acid to preserve reduced ascorbate. The precipitate was separated by centrifugation and the 50/50 plasma/10% meta-phosphoric acid was stored. All extraction and HPLC solvents were HPLC grade.

To extract the ascorbic acid, 200 µl of plasma were treated with 10% meta-phosphoric acid and combined with 90 µl 0.5 mol/l Trizma buffer, pH 9.0, plus 10 µl DL-dithiothreitol to buffer the sample. The mixture was vortexed, allowed to stand for 5 min and the reaction was quenched with 100 µl 0.2 mol/l H2SO4 and revortexed. Samples were micro-centrifuged for 2 min and the supernatant was filtered (4 mm Teflon syringe filter 0.45 µm) before injecting into the HPLC for analysis.

The prepared plasma samples were loaded into a binary pump system with an automatic sampler (ESA, Chelmsford, MA) and 30 µl was run through a Luna 3u C18(2) 1—A, 150 × 4.6 mm 3 micron column with 50 mmol/l sodium phosphate, pH 3, at 0.8 ml/min. Ascorbic acid was detected with a CoulArray electrochemical detector between 200–300 mV. A standard was run daily and ascorbic acid was identified between 4–6 min. Ascorbic acid concentration was determined from a daily standard curve.

Carotenoids were extracted under yellow light to minimize photo-oxidative reactions. 200 µl of plasma was vortexed with 1 ml methanol. Immediately after, 1 ml acetone was added and vortexed to aid the phase separation and extraction. Next, 3 ml of petroleum ether with 0.01% butylated hydroxytoluene was added and the mixture was vortexed for ∼30 s and centrifuged for 1 min at 1,000g. The ether layer was collected and 3 ml petroleum ether with 0.01% butylated hydroxytoluene was added, vortex-mixed, centrifuged, and collected two additional times. Samples were combined, dried immediately under nitrogen gas, and resolubilized with 100 µl ethyl acetate followed by 300 µl methanol.

Resolubilized samples were loaded into a Hewlett-Packard model 1090A HPLC system with a 79880A diode array detector. A YMC C30 carotenoid column (2.0 × 100 mm) and a guard column of identical packing material (Waters, Milford, MA) were used for separation. Carotenoids were resolved using a gradient elution at 0.37 ml/min with different proportions of a solvent with 98% methanol and 2% 1 mol/l ammonium acetate then a solvent containing ethyl. Detection and identification of carotenoids was completed with diode array data between 250–600 nm. Carotenoid concentrations in serum were calculated based on multi-level calibration curves constructed with authentic external standards for lutein, zeaxanthin, β-cryptoxanthin, β-carotene and lycopene. α-Carotene concentration was based on β-carotene response due to lack of authentic standards. Similarly, quantities of cis-lycopene were estimated based on response of trans-lycopene.

Statistics

Changes (end of each treatment period minus the corresponding baseline) were analyzed using a general linear mixed model that enabled the examination of the effects of treatment (beverage vs. solid), BMI (normal vs. overweight/obese), the interaction of treatment and BMI, and order. Normality was checked using the Shapiro-Wilk test and homogeneity of variance was verified with Levene's test. For some analyses, the overweight/obese group was disaggregated and Bonferroni corrections were used to compare the resulting three BMI groups (lean, overweight, obese). One-sample t-tests, paired t-tests, one-way and repeated measures ANOVA, Pearson correlations, and χ2 tests were used when indicated. Statistical significance was defined as P < 0.05, two-tailed and data are reported as means ± s.e. Statistical testing was conducted with the Statistical Package for the Social Sciences, version 17.0.

Results

Body weight

Overall, participants gained 1.60 ± 0.24 kg during the study (t = 6.73, df = 33, P < 0.0001). Gains were observed during both the beverage (1.95 ± 0.33 kg) and solid (1.36 ± 0.30 kg) study arms (both P < 0.0001) (Figure 2). Although gains were 43% greater during beverage consumption, no statistically significant effect of treatment was noted (P = 0.19) (Table 1), nor were there significant effects of order (P = 0.13), BMI (P = 0.08), or the interaction of treatment and BMI (P = 0.72). However, the data indicate all obese participants gained weight during both study arms (Figure 3). Further analysis with overweight and obese individuals in distinct groups revealed a BMI effect (F = 4.55; df = 2,31; P < 0.02) with overall weight gains of 1.22 ± 0.31, 1.61 ± 0.32, and 3.07 ± 0.53 kg for lean, overweight and obese, respectively. Bonferroni adjusted P values were 0.02 and 0.07 for obese vs. lean and vs. overweight, respectively. Lean participants (n = 15) gained weight during the beverage (1.61 ± 0.44 kg, P = 0.003), but not during the solid intervention (0.84 ± 0.53 kg, P = 0.133). The overweight/obese participants (n = 19) gained weight during the beverage (2.22 ± 0.47 kg) and solid (1.77 ± 0.32 kg) arms (both P < 0.0005) with the obese participants (n = 5) experiencing greater weight gain (4.04 ± 1.12 kg) compared with the lean (n = 15) and overweight (n = 14) participants (1.56 ± 0.40 kg) during the beverage arm (F(2,31) = 4.220, P = 0.024). A greater proportion of overweight/obese participants gained weight during the solid arm compared with lean participants (95% vs. 67%) (χ2 = 4.55, df = 1, P < 0.03) (Figure 3), but no group difference was evident for the beverage arm. A strong relationship was observed between baseline BMI and weight change for the full sample of overweight/obese participants (r = 0.58, P = 0.009) and remained evident after omitting one obese individual with marked weight gain (r = 0.468, P = 0.05) (Figure 4). Initial BMI was not associated with weight gain during the solid arm for any group.

Figure 2.

Body weight change was measured for human participants in a randomized-crossover trial (8 weeks for each study arm). The beverage and solid fruits and vegetables were 20% of each participant's estimated energy needs. The lean and overweight/obese groups were purposefully recruited and were compared to test our primary hypothesis. The overweight and obese groups were separated out in secondary analysis. *P < 0.0005 vs. baseline. P = 0.003 vs. baseline. P = 0.002 vs. baseline. §P = 0.022 vs. baseline, P = 0.024 vs. lean and overweight.

Table 1.  Beverage vs. solid body weight change, body fat change, energy intake change, and energy compensation
inline image
Figure 3.

Body weight change was measured for 34 human participants in a randomized-crossover trial (8 weeks for each study arm). The beverage and solid fruits and vegetables were 20% of each participant's estimated energy needs. The dotted line indicates equal weight change during both study arms.

Figure 4.

Body weight change was measured for 19 human participants in a randomized-crossover trial (8 weeks for each). The beverage and solid fruits and vegetables were 20% of each participant's estimated energy needs (up to 550 kcal). Graphed here is the weight change during the beverage study arm for the overweight and obese participants. One extreme outlier was removed (BMI 32.5 kg/m2, kg change beverage = 7.9 kg). The 18 analyzed points indicate a trend for a moderate relationship between starting BMI and weight change during the beverage study arm. r = 0.468, P = 0.05, n = 18.

Body composition

Relative to the preceding baseline assessment, participants gained 1.50 ± 0.39 kg of fat mass during the beverage arm (P < 0.0005, n = 33) (∼77% weight gain) and 1.16 ± 0.33 kg of fat mass during the solid arm (P = 0.002, n = 33) (∼85% weight gain) (Table 1). Total fat gains were 0.65 ± 0.51, 1.37 ± 0.53, and 3.61 ± 0.88 kg for lean, overweight and obese (F = 4.23; df = 2,31; P < 0.03), respectively, with a Bonferroni adjusted P value of 0.02 for obese vs. lean. 95% of the overweight/obese subjects gained fat mass vs. 53% of the lean subjects during the solid arm (χ2 = 7.99, df = 1, P < 0.005). There were no group effects during the beverage arm. There were no statistically significant changes of lean body mass.

Dietary intake

Changes in total energy intake are displayed in Figure 5 and Table 1. Carbohydrate consumption increased by 73.7 g/d (t = 5.15, df = 33, P < 0.001) while protein consumption decreased by 10.2 g/d (t = 2.13. df = 33, P < 0.05) and fat consumption did not significantly change (decrease of 9.0 g/d). Beverages were consumed alone or with other foods 49% or 51% of the time, respectively. Solid fruits and vegetables were consumed alone and with other foods 56% and 44% of the time. The food use pattern was not different between treatments. Total eating events increased from 4.43 ± 0.28 to 5.54 ± 0.36 during the solid arm (t = −3.085, P = 0.004, n = 32), but after correction for multiple comparisons, the increase from 4.50 ± 0.22 to 5.16 ± 0.30 during the beverage arm was not significant. There were no BMI effects.

Figure 5.

Energy consumption change was measured for 34 human participants in a randomized-crossover trial (8 weeks for each study arm). The beverage and solid fruits and vegetables were 20% of each participant's estimated energy needs (minimum of 400 kcal, maximum of 550 kcal). The lean (n = 15) and overweight/obese (n = 19) groups were purposefully recruited and were compared to test our primary hypothesis. The overweight (n = 14) and obese (n = 5) groups were separated out in secondary analysis. *P = 0.03 vs. baseline. P = 0.046 vs. baseline.

Dietary compensation

Dietary compensation scores for the full sample during the beverage and solid arms were 53 ± 21% and 78 ± 24%, respectively. The lean group fully compensated for the solid load (119 ± 32%); indeed they exhibited slight reverse compensation. Partial compensation was observed for the overweight/obese group during the solid arm (46 ± 34%). Neither group compensated completely for the beverage load: 43 ± 32% (lean) and 61 ± 27% (overweight/obese). Overweight participants partially compensated for the beverage (73 ± 36%) and solid (74 ± 39%) arms. The obese participants compensated weakly for the beverage intervention (28 ± 23%) and increased energy intake in addition to the required fruits and vegetables during the solid arm (−32 ± 56% compensation). Dietary compensation was negatively correlated with weight change during the beverage (r = −0.39, P < 0.04) and solid (r = −0.34, P < 0.06) arms.

Energy expenditure

No statistically significant effects of order, treatment, BMI, or the interaction of treatment with BMI were found for measured REE or TEF. Accelerometer data indicated a significant treatment effect (F = 8.11; df = 1.33; P < 0.008), where energy expenditure decreased relative to baseline by 120 ± 227 kcal/min during the beverage arm and increased by 18 ± 212 kcal/min during the solid arm.

Palatability and ease ratings

Participants indicated it was easier to include the juice load (3.00 ± 0.36 units) into their usual diet compared with the solid load (5.12 ± 0.32 units) (F = 18.13: df = 1.33; P < 0.0002). Overall, participants liked the juice (7.38 ± 0.21 units) more than the fresh and dried fruit (6.74 ± 0.21 units) (F = 6.04: df = 1.33; P < 0.002).

Compliance

Plasma ascorbic acid increased during the beverage portion of the study by 15 ± 5 µU/l (P = 0.005, n = 34); 76.5% of participants had a positive change. The change during the solid arm was 6 ± 6 µU/l and not significant; 44.1% of participants had a positive change. Total carotenoids increased by 605 ± 104 nmol/l during the beverage arm (z = 4.42, P < 0.0005, n = 34) and 257 ± 171 nmol/l during the solid arm (z = 2.40, P = 0.016, n = 34).

Discussion

There are conflicting data regarding the role of energy-yielding beverages in promoting positive energy balance and elevated body fat (1,2,3,8,27,28). Secondary analyses suggest their ingestion may be particularly problematic for weight management among the overweight and obese (7,28,29). This body of evidence is largely based upon epidemiological studies that do not allow determination of causation; assessments of appetitive responses to beverages compared with other foods (30) which provide only weak presumptive evidence for effects on food choice; clinical trials that have not directly tested beverage-specific effects (5,31) and meta-analyses (1,2,3) that have attempted to combine these limited data sets, but to-date, have failed to yield more definitive answers. Resolution of the issue is best achieved by randomized controlled trials (4). This was the approach of the present trial.

Central to interpretation of findings from trials with free-living individuals, particularly when null hypotheses cannot be rejected, is evidence that participants complied with the dietary intervention. The increases of plasma carotenoids during both study arms and plasma ascorbic acid during the beverage arm indicate participant's increased total fruit and vegetable consumption during both arms (32).

The primary hypothesis tested was that prescribed inclusion of energy-yielding beverages in the diet would lead to increased energy intake and weight gain, especially among overweight/obese individuals and this would not occur with inclusion of solid food forms matched on energy and macronutrient content. Consistent with this hypothesis, beverage consumption promoted weight gain in 87% of the lean and 84% of the overweight participants. 77% of the noted weight gain during the beverage arm was body fat indicating the weight change was not simply due to positive fluid balance. The observed dietary compensation score during the beverage arm was only 53% and weight change was significantly correlated with dietary compensation. Thus, the dietary data strongly support the observed effects of beverages on body weight change.

A direct association was observed between baseline BMI and weight gain during the beverage intervention in the overweight/obese sub-group. Further, secondary analyses suggested the obese exhibited significantly greater weight gain with beverages compared with the overweight or lean participants. Individuals with higher BMI are at particular risk for weight gain with energy-yielding beverage consumption (7,28,29). Only the overweight/obese participants gained a significant amount of weight during the solid food arm and a higher proportion of these individuals gained weight compared with the lean participants with this treatment.

Measured REE and TEF did not differ between treatments or BMI groups. However, the accelerometer readings which summate an estimate of REE and measured activity revealed an unexpected decrease of activity energy expenditure during the beverage arm (accelerometer readings minus REE). However, the validity of this change, which was not corroborated by reports of changes in activity by participants, is uncertain.

While measured weight gain was 43% greater during the beverage arm, it did not differ significantly from the solid arm. This may be due to the larger than expected variance in the overweight/obese group resulting in lower than predicted statistical power (Table 1). This may also be attributable to methodological factors. First, the solid intervention entailed provision of some dried fruits to enhance dietary compliance. Inclusion of these energy dense solid foods as 60% of fruit energy may have helped the participants consume the food items provided, diminishing treatment differences that may have occurred if only fresh fruits were provided. Second, there are no recommendations to meet fruit and vegetable intake goals by drinking them in beverage form. They were used here only as vehicles to explore effects of beverages on energy balance and were supplemented with fiber to match the composition of the solid foods. Given evidence that fiber contributes to satiety and weight management (33), this minimized a potentially important property difference between beverages and solid foods and likely reduced the response discrepancy between food forms. Potentially offsetting these possible biases was the requirement that solid foods be consumed in raw form to optimize their satiety effects. This could have exaggerated treatment differences if this reduced compliance on this arm. However, participants were selected who indicated these foods were acceptable in raw form and the increase in plasma carotenoids indicate they were consumed. Third, although the variety of provided items did not faithfully match the Dietary Guidelines for Americans (14), the use of fruits and vegetables allowed assessment of their influence on energy balance and body weight. The literature on this topic contains reports of weight loss (18,19,20), stability (21,22,23), and gain (16,17). A previous review (15) suggested that variability of outcomes may stem from different emphases on total energy or fat intake in addition to recommendations to increase fruits and vegetables. This may account for some (e.g., 20,34,35,36,37), but not all (e.g., 16,17,38) interventions. It is possible that weight stability may be observed more often in trials where the goal of including fruits and vegetable in the diet is to enhance antioxidant or nutrient intake vs. when the aim is to dilute energy density or enhance satiety as a weight management approach (e.g., 21,22,39,40). Compliance is a potential complication for interpreting prior study outcomes. Some trials provided the fruits and vegetables (18,23,38) and were closer to tests of efficacy whereas others only recommended greater intake and were tests of effectiveness (16,17,19,20,21,22,36,39,40). With provision of fruits and vegetables to maximize protocol compliance and biochemical verification of their consumption, this study indicates the inclusion of fruits and vegetables in the diet without guidance on total energy intake may result in increased energy intake and weight gain. These results and limitations raise important issues for the design of future research on the health effects of fruit and vegetable intake. They indicate the need to carefully consider population characteristics (e.g., BMI status and customary intake patterns), food properties (e.g., physical form and sensory appeal) and methodological approaches (e.g., intervention intent and duration, compliance verification, power) to obtain clear outcomes.

The data from this randomized controlled trial strongly suggest energy-containing beverages promote positive energy balance and weight gain, especially for overweight and obese individuals. That said, these beverages can be important sources of water and nutrients, so may have an important place in the diet. Strategies for their healthful inclusion in the diet warrant further consideration. Similarly, in low fruit and vegetable consuming overweight/obese individuals, adding solid fruits and vegetables to the diet may provide a variety of health benefits, but advice to offset the energy they contribute to the diet must be emphasized.

Acknowledgment

We are grateful for the tenacious participants who completed this protocol. We are indebted to our Purdue undergraduate assistants who helped in all parts of data collection: Amy L. Conklin, Mindy Crawford, Tarrah K. DeClemente, Brianne N. Haupert, Jeanine Hiett, and Jennifer L. Vyduna. Plebotomy for compliance assessment was conducted by Robin L. Rhine and Judy Cassini George. This study was funded by NIH grant# R01-DK063185.

Disclosure

The authors declared no conflict of interest.

Ancillary