Participants were recruited from the population of Clemson University and surrounding areas for two different studies [2, 12]. The first study evaluated bite counter records in a free-living population against self-reported intake (FREE-LIVING) . The second study compared bite counter records with actual energy intake recorded in a cafeteria setting (CAFETERIA) . Both studies were approved by the Clemson University Institutional Review Board.
Detailed information on the original study was previously published . Seventy-seven participants (age = 32.5 ± 12.4 y; BMI = 26.7 ± 5.9 kg/m2; 39 females, 38 males) with no history of eating disorders were told that the purpose of the study was to investigate if a new device could estimate the amount of food eaten during meals. History of eating disorder was determined by asking each participant the question: “Do you have a history of eating disorders (e.g., anorexia, bulimia)?” The participants self-reported yes or no. Each participant wore the bite counter for 14 days to obtain bite data for a total of 2975 meals (an average of 2.76 meals per person per day).
Subjects were asked to wear the bite counter continuously unless they were engaging in activities that could damage it, such as taking a shower. The bite counter recorded meal duration in seconds and the bite counts per second per meal. Participant intake was not guided and they freely consumed meals and snacks as part of their daily routine. The wrist-worn bite counter was returned to the study site for download and recording of the data.
Additional study details have been previously published . Participants with self-reported eating disorders were excluded from the study. Participants ate in a cafeteria on campus at Clemson University seated at a single, four-person customized table. The table was equipped with scales underneath place settings for monitoring weight changes, wrist-worn sensors for detecting bites, and cameras mounted in the ceiling that recorded videos of each meal. The participants were made aware of each recording device. The participants were allowed to select from a wide variety of meals (approximately 380 foods and beverages) in a cafeteria setting, which they self-selected and consumed while wearing the bite counter. A few of these items were available for nearly every session; these included all beverages, ice cream, pepperoni pizza, cheese pizza, hamburgers/cheeseburgers, shoestring French fries, chicken sandwiches, sandwich-bar sandwiches, and salads. The rest of the items, with some recurring once or twice, largely varied day to day. The participants were instructed to eat as much as they liked. A record was kept of all of the food items available for each day and time of the study.
Participants ate in groups of up to four and were allowed to schedule sessions with friends if they wished. Of the original sample of 280 participants, 44 ate with someone they knew. Four participants could not always be recruited for a single session: 136 ate in groups of four, 93 in groups of three, 46 in pairs, and 5 participants ate alone. As there was only one instrumented table in the cafeteria, participants always ate with their assigned cohort.
Energy intake (kcal) in the CAFTERIA participants was determined using a validated visual method [13, 14]. Food items selected by each individual participant were first identified from video footage. The selected portion of each food item was defined as a percentage of the reference serving size of the food item. The percentage of the selected portion of each food item consumed by each participant was then visually estimated by three raters. Energy intake (kcal) for each food item was calculated by multiplying the calorie content of the selected food item by the percentage of the reference portion selected and the percentage of the selected portion consumed which accounted for plate waste.
Of the original sample of 280 participants, 66 were absent from the final analysis due to data recording errors and outlier analysis for caloric intake and bite count. A reference database of 214 participants (age = 30.0 ± 12.1 y; BMI = 25.4 ± 5.6 kg/m2; 114 female, 100 male) containing simultaneous measures of total bite counts, total energy consumed per meal, meal duration, age, gender, and body mass index (BMI) was analyzed for this study.
All statistical analysis was performed in the statistical software SPSS (IBM, Armonk, NY 2012). The BMI-bite count/min plots were developed in Microsoft Excel (Seattle, WA 2011).
In the FREE-LIVING study, the number of meals was substantial (2975 meals) and varied across participants. Secondary data analysis was performed. In order to collapse data while ensuring integrity of results, analysis was performed by aggregating bite count rate information for each participant as average bite counts (bite count/min) over 14 days. Linear regression was conducted to test whether bite count rate (bite count/min) increased as a function of body mass index (BMI) or body weight. Linear regression was also performed to identify relationships between body mass index (BMI) and eating duration for both FREE-LIVING and CAFETERIA in Microsoft Excel (Seattle, WA, 2011).
In the CAFETERIA data, average bite count rates per individual were computed by averaging total bite counts over meal duration (bite count/min). In order to determine whether longer meal durations were related to higher energy intake, linear regression analyses examing relationships between meal duration and energy intake were performed. This analysis was only conducted in the CAFETERIA data because only the CAFETERIA study had objectively measured energy intake.