|Home | About | Journals | Submit | Contact Us | Français|
Consumption of energy-containing beverages may lead to weight gain, yet research investigating this issue is limited. An easily-administered beverage intake assessment tool could facilitate research on this topic. The purpose of this cross-sectional investigation was to determine the validity and reliability of a self-administered beverage intake questionnaire (BEVQ), which estimates mean daily intake of beverages consumed (g, kcals) across 19 beverage categories. Participants (n=105; aged 39±2 yrs) underwent assessments of height, weight, body mass index, and dietary intake using 4-day food intake records (FIR) from June, 2008-June, 2009. The BEVQ was completed at two additional visits (BEVQ1, BEVQ2). Urine samples were collected to objectively determine total fluid intake and encourage accurate self-reporting. Validity was assessed by comparing BEVQ1 with FIR results; reliability was assessed by comparing BEVQ1 and BEVQ2. Analyses included descriptive statistics, bivariate correlations, paired samples t-tests, and independent samples t-tests. Self-reported water and total beverage intake (g) were not different between the BEVQ1 and FIR (mean difference: 129±77g [P=0.096] and 61±106g [P=0.567], respectively). Total beverage and sugar-sweetened beverage (SSB) energy intake were significantly different, although mean differences were small (63 and 44 kcal, respectively). Daily consumption (g) of water (r=0.53), total beverages (r=0.46), and SSB (r=0.49) determined by the BEVQ1 were correlated with reported intake determined by the FIR, as was energy from total beverages (r=0.61) and SSB (r=0.59) (all P<0.001). Reliability was demonstrated, with correlations (P<0.001) detected between BEVQ1 and BEVQ2 results. The BEVQ is a valid, reliable, and rapid self-administered dietary assessment tool.
Obesity has become an epidemic in the United States (1), with more than 66% of adults overweight (Body Mass Index [BMI] 25–29.9 kg/m2) or obese (BMI ≥ 30 kg/m2) (2). Despite efforts to identify strategies which effectively promote weight management, the prevalence of obesity has not declined (3). According to laboratory-based feedings studies (4), energy-containing beverages are less satiating than solid foods. Specifically, intake of solid food is not spontaneously reduced when energy-containing beverages are consumed (5, 6), regardless of nutrient composition (e.g., low fat milk, soda, or juice) (7, 8). Thus, consumption of energy-containing beverages may increase energy intake and lead to weight gain (4).
Interventions targeting energy-containing beverage consumption could lead to weight loss for overweight and obese individuals. Self-reported energy intake declines when sugar-sweetened beverage (SSB) intake is replaced with water (10); premeal water consumption reduces subsequent meal energy intake and facilitates weight loss over time (9,11). Furthermore, a sugared beverage tax is being enforced to discourage SSB consumption in several US states (12). A recent Scientific Statement from the American Heart Association highlighted the contribution of SSB to total added sugar intake, and recommended added sugar intake guidelines (13). However, the need for direct evidence linking beverage consumption patterns with weight outcomes has been suggested (14).
Food diaries and recalls are commonly used to assess dietary intake; however these methods are resource-intensive, time-consuming, burdensome for participants, provide only recent intake data (i.e. not habitual intake patterns), and are not always feasible in large-scale studies (15–17). There is currently no rapid (< 5 minutes) method for determining habitual beverage intake in adults, including quantities and energy contribution. A brief, self-administered, valid and reliable beverage intake assessment tool could enhance nutrition research targeting beverage intake patterns.
The purpose of this investigation is to test the validity and reliability of a newly developed self-administered beverage intake questionnaire (BEVQ) as compared to a “gold standard” of measuring dietary intake–food intake records (FIR), which have been used in numerous validation studies (18–23). Although their limitations are recognized (24), FIR are suitable for comparison to questionnaires to establish validity, and have the least correlated errors of the dietary intake methods available (17).
Healthy adults (n=105) aged ≥21 years were recruited for this cross-sectional investigation from a local university community between June 2008-June 2009. The Virginia Tech Institutional Review Board approved the study protocol. Participants provided written informed consent prior to enrollment, however they were not aware of the specific purpose of the study; they were informed that the study was evaluating a new food intake questionnaire.
Participation entailed three laboratory visits within a two-week period; visits were completed in one of two randomly assigned visit sequences. The three study visits included the completion of two BEVQ and one four-day FIR, as follows: Sequence 1: (visit 1) BEVQ1, (visit 2) FIR, (visit 3) BEVQ2; Sequence 2: (visit 1) FIR, (visit 2) BEVQ1, (visit 3) BEVQ2. Completing the FIR before the BEVQ could heighten participant’s awareness of their food and beverage intake, and falsely increase correlations between the FIR and BEVQ (17). Randomizing visit sequence provided a means to determine if randomization sequence influenced results. All visits were conducted between 12 pm – 5 pm to avoid the differences in urinary specific gravity (SG) measurements that may occur throughout the day.
For all participants, visit 1 included the following procedures: height, measured in meters without shoes using a wall mounted stadiometer; body weight, measured in light clothing without shoes, to the nearest 0.2 kg using a physician’s balance scale (Seca; Hanover, Maryland); and BMI, calculated as weight (kg)/height (m2). Participants provided information on demographic characteristics and health status (e.g., age, race/ethnicity, medical history, medications). Sequence 1 participants then completed a BEVQ (BEVQ1) and provided a urine sample to determine SG; sequence 2 participants received instructions for completing a four-day FIR, including the use of two-dimensional food models to assist with portion size determination. Urinary SG was determined using a handheld refractometer (ATAGO 4410 Digital Urine Specific Gravity Refractometer, Bellevue, Washington). The urine sample provided an objective indicator of total fluid intake, and also served to encourage the accuracy of participant’s self-reported dietary intake (17). Food records were kept either from Sunday through Wednesday or Wednesday through Saturday, in order to capture both weekend and weekday dietary habits; FIR were reviewed for completeness upon return, and analyzed using nutritional analysis software (Nutrition Data Systems for Research [NDS-R] 2007, University of Minnesota, Minneapolis, MN).
At visit 2, sequence 1 participants were provided with instructions for completing the FIR identical to that for the initial visit of sequence 2 participants; sequence 2 participants completed a BEVQ (BEVQ1), provided a urine sample and returned the FIR. At visit 3, sequence 1 participants completed a BEVQ (BEVQ2), provided a urine sample, and returned the FIR; sequence 2 participants completed a BEVQ (BEVQ2) and provided a urine sample. Participants were compensated $10 upon completion of all three study visits.
The BEVQ was developed to estimate mean daily intake of water, SSB and total beverages (grams [g], kcals) across 19 beverage categories plus one open-ended section for “other” beverages not listed (Figure 1). This tool is a quantitative food frequency questionnaire; the frequency of food items consumed and amounts consumed are assessed (24). Beverage categories were grouped by energy and macronutrient content using published food composition tables (25) and nutritional analysis software (NDS-R 2007, University of Minnesota, Minneapolis, MN). Common beverage portion sizes (e.g., 12 fl oz can of soft drinks, 20 fl oz bottles of juice/water/soft drinks), and common cup sizes (e.g., juice glasses [(4–6 fl oz] and cups [8 fl oz]) were utilized to assess amounts consumed. Due to the desire to develop a brief, single-page BEVQ, the most commonly consumed beverage units were included. To score the BEVQ, frequency (“How often”) is converted to the unit of times per day, then multiplied by the amount consumed (“How much each time”) to provide average daily beverage consumption in fl oz. Energy and grams (per fl oz) for each beverage category were determined using food composition tables (25). Total energy and grams of each beverage were determined by multiplying the number of fl oz per day by the energy and grams per fl oz of each category. To quantify total SSB consumption, beverage categories containing added sugars were summed (sweetened juice beverages/drinks, regular soft drinks, sweet tea, sweetened coffee, energy drinks, mixed alcoholic drinks, meal replacement beverages). During pilot testing of the BEVQ, average administration time was determined to be ~3.5 minutes (range: 2 min 12 sec − 4 min 26 sec).
Statistical analyses were performed using SPSS statistical analysis software (v. 12.0 for Windows, 2003, SPSS Inc, Chicago, IL). Descriptive statistics (mean±standard error of the mean [SEM]; frequencies) are reported for demographic characteristics and average total consumption of beverages and beverage categories (g, kcal). Paired sample t-tests were used to compare the energy intake (kcal) and the g consumed of specific beverages across dietary assessment tools. To assess validity, the BEVQ1 responses were compared to FIR responses, and to assess test-retest reliability, BEVQ1 responses were compared to BEVQ2. Independent samples t-tests were used to assess potential differences in the randomization sequence. Associations among variables (beverage intake variables, SG) were assessed using correlational analyses (Spearman’s r). The alpha level was set a priori at P≤0.05.
One hundred and five individuals (45 males; 60 females) completed all study visits. Participants were primarily Caucasian (85% of sample), with remaining participants self-identified as Asian (8%), African American (4%), or “other” (4%). Mean age of participants was 39±2 yrs (range: 21–93 yrs), which was distributed across the adult age range as follows: 21–39 yrs, 60%; 40–59 yrs, 23%; ≥ 60 yrs, 17%. Body mass index was widely distributed (mean=25.6±0.6 kg/m2; range 16.2–62.5 kg/m2), although participants were primarily of “normal” BMI status (BMI <18.5 kg/m2, 2%; 18.5–24.9 kg/m2, 53%; 25.0–29.9 kg/m2, 30%; >30.0 kg/m2, 15%). Of 72 participants who provided information on their educational level, most reported being college-educated (n=67).
Results from the validity and test-retest reliability assessment of the BEVQ are presented in Table 1. Of the 21 beverage variables assessed (grams and energy for 19 individual beverage categories, plus SSB and total beverages), responses on the two assessment tools (BEVQ1, FIR) were significantly correlated (all P<0.001) with two exceptions: sweetened coffee and mixed alcoholic drinks. Responses between the BEVQ1 and FIR were not different for intake (g) of water, juice drinks, vegetable juice, milk (all types), soft drinks (regular and diet), light beer, liquor, mixed alcoholic drinks, wine, and total beverage intake. Differences in beverage energy content between assessment tools were < 35 kcal across all categories, although this difference was significant for 100% fruit juice, sweet tea, sweetened coffee, beer, meal replacement and energy drinks. Significant mean differences were detected in total beverage and sweetened beverage energy intake determined using the two tools, although this difference was minimal (63 and 44 kcal, respectively). These two variables were, however, each significantly correlated between the tools. Reliability was acceptable (r=0.45−0.87; all p<0.001), as food frequency questionnaires considered reliable typically report correlations ranging 0.5–0.7 (17, 26). Significant correlations were detected between all variables, although the correlation for energy drinks was lower than that for other beverage categories. No significant differences were found between BEVQ responses based on the two study sessions (BEVQ1, BEVQ2), or between the two visit sequences (data not shown). Urinary SG measurements were not significantly different across visits (1.0146±0.0008 vs. 1.0146±0.0008 UG; mean difference: −0.000019±0.007 UG). As would be expected for a possible biomarker of total fluid intake, SG was negatively correlated with grams of total daily beverage consumption (BEVQ) at time one and time two (r=−0.202 and r=−0.238; p<0.05, respectively). SG was also negatively correlated with BEVQ water intake (g) at time one (r=−0.236, p<0.05) and time two (r=−0.319, p<0.01). Thus, the BEVQ appears to be a valid, reliable, and easily-administered questionnaire for assessing beverage intake in adults.
Beverage consumption is timely topic in the weight management field (12, 14) and particularly for SSB, there are broad public health implications (13). This tool may be useful for researchers and clinicians interested in assessing habitual beverage consumption patterns, particularly in large-scale investigations where lengthier, resource-intensive dietary intake assessment techniques are not feasible. Among dietetic practitioners, this tool could be utilized as rapid method to assess beverage consumption as part of a Nutrition Assessment in the Nutrition Care Process, and potentially in Nutrition Monitoring and Evaluation.
The present findings are consistent with others using more extensive dietary intake assessment methods, reporting a mean beverage energy intake of 458 kcal per day (27). Water is the most consumed beverage in the US, followed by coffee, soft drinks, whole milk, fruit juices, and alcohol (27). The present findings are consistent with this pattern, with the exception of whole milk. In the general population, the majority of beverage energy (~50%) comes from SSB, such as regular soft drinks, fruit drinks, sweet tea, and energy drinks (27, 28). In this sample, SSB contribute ~40% of total beverage energy. Furthermore, Block (29) reported that energy containing soft drinks are the greatest contributor to total daily energy intake (i.e., all food and beverages) at 7.1%, while beer was also among the top contributors (2.6% of total energy). In this sample, soft drinks were the fifth highest contributor of energy from beverages, preceded by fat-free milk (greatest contributor of energy), fruit juice, reduced-fat milk, and sweet tea. These differences may be attributed to the demographics of our sample, as age, weight status, educational level and socioeconomic status may influence beverage consumption (30).
After completing this initial evaluation of the BEVQ, several limitations were determined. Questions from participants during completion of the BEVQ suggested some refining may be needed, for example the BEVQ does not include a category for hot cocoa and participants were uncertain how to report sports drink intake. Beverage category descriptions may also need modification, for example, “coffee with cream and/or sugar” may be misinterpreted as coffee with cream. Participants were uncertain as to whether milk in cereal and coffee should be included in their responses. To address this issue, future versions will include additional respondent instructions such as to only report consumption of liquids when consumed as beverages. It is possible that the BEVQ underestimates certain beverage categories due to the upper limits on quantities (60 fl oz per day), for example water intake. However, estimated BEVQ mean daily water intake is similar to that reported by National Health and Nutrition Examination Surveys (NHANES)(31) and the present findings did not indicate a ceiling effect. A final limitation is the use of a self-reported FIR for validation, as underreporting errors are common (24). However, FIR are recommended for validation of food-frequency questionnaires due to a reduced likelihood of correlated errors (17), when direct measurement of food intake is not feasible. Future work will determine if reducing the length of the tool is possible without impacting results, if the tool is sensitive to changes in beverage intake, and if the tool may be used in low-literacy populations. Due to the primarily Caucasian composition of this sample, future studies including larger numbers of minorities are warranted to determine if the BEVQ is a valid tool across ethnic/racial groups.
An easily-administered, valid, and reliable beverage intake questionnaire may be desirable for practitioners, as well as for researchers assessing habitual beverage intake and possible influence on weight and health status. This tool may also be useful for large-scale studies, and for interventions targeting changes in beverage intake, particularly in light of data indicating that increasing water consumption and reducing energy-containing beverage consumption facilitates weight loss (9).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Author Contributions:Experimental Design: B. Davy, P. Estabrooks, V. Hedrick
Collection of Data: V. Hedrick, D. Comber
Analysis of Data/Data Interpretation: all authors
Writing of Manuscript: all authors