Effect of calories-only vs physical activity calorie expenditure labeling on lunch calories purchased in worksite cafeterias

The overall design of the PACE Study has been previously published in detail [9]. In brief, we partnered with Blue Cross and Blue Shield of North Carolina (BCBSNC) to examine the effects of PACE labels compared to calorie-only labels in three worksite campus cafeterias. Employees who regularly ate lunch in the campus cafeterias were invited to participate in the study. During a pre-intervention year, we collected baseline data from participants in all three cafeterias. At the beginning of the second year of the study, one cafeteria had its food items prominently labeled with PACE labels and two cafeterias had their food items labeled with calorie-only information. The reasons for combining two cafeterias were (1) to maintain enough cohort participants over the duration of the study, and (2) to better balance the characteristics of employees.

The primary outcome was calories purchased during lunchtime as measured every 3 months over a 2-year period by individuals enrolled in two cohorts. This outcome was assessed at baseline (prior to labeling) and during the 1-year labeling intervention.

We advertised the study using paper and electronic flyers. Study coordinators actively recruited participants in the worksite cafeterias by setting up an informational table for employees to visit and to learn more about or sign up for the study. Enrollment of participants continued on a rolling basis throughout the baseline year to help compensate for attrition. To be eligible, a participant needed to (1) be a BCBSNC employee or contractor, and (2) eat lunch or be willing to eat lunch in the BCBSNC cafeteria at least 3 times per 5-day work week.

We collected detailed information on the lunch purchases made by the participants using a specially designed photo capture system. The system consisted of a touchscreen monitor and camera positioned above a shelf that allowed participants to take a picture of their entire meal contents. Photographs were saved with the participants’ study identification number, initials, date and site location. Study coordinators were on-site during all data collection periods to record details that may not have been evident from the photograph (e.g., dressings, condiments, soup or drink contents).

Participants submitted lunch photos over 2-week periods on a quarterly basis during the 24-month study period. Study staff analyzed all lunch photos and entered calorie information into a detailed database. For any items requiring estimation of portion sizes (e.g., self-serve items or self-built salads), Study coordinators used a food atlas along with their on-site notes to estimate portion sizes [9].

Following the year of baseline data collection, one cafeteria received PACE labels which showed the calories in the food as well as an image of someone walking and the estimated number of miles needed to “burn off” the calories [9]. We determined PACE label values as previously described [9]. The two other cafeterias received calorie-only labels. The labels, which were bright green, bright blue, or bright yellow and measured 3 X 4 in., were prominently displayed above or beside each food item. For food prepared to order at the grill and deli, lists of commonly purchased items were posted with the PACE or calorie label [9]. For salad bars, we posted lists of common items as well as representative salads showing the sum of calories or PACE equivalent from all ingredients included. Beverage cooler doors were labeled with lists of all their beverages.

We based sample size on the outcome of change in purchase calories before versus after the labeling intervention. For sample size calculations, we assumed standard deviation of purchased lunch meals is 350 cal and expected that the PACE labeling intervention would reduce purchase calories by 100 cal based on our preliminary study [8]. There were four measures of purchased calories (one for each quarter) prior to the intervention and four after the intervention. We analyzed purchased calorie data from all time points using a longitudinal generalized linear mixed model (GLMM) that included study groups, time, and study group by time interaction as fixed effects, each participant as random effect and a covariance structure that provided the best fit for the model to obtain estimates of slopes before and after introducing the calorie label interventions and compared between groups. We compared baseline sample characteristics of the two study groups using t-test and chi-square tests. Study groups differed significantly for age, sex, race, and occupation categories. These variables were included as fixed covariates along with literacy and numeracy scores (considered a priori) in a similar GLMM to compare slopes before and after the interventions between groups. In addition, a small number of participants changed worksite cafeterias prior to the intervention, so we analyzed participants using the same GLMMs based on the cafeteria in which they took > 50% of their lunch photos during the intervention year.

A total of 416 individuals initially consented to be in the study. Due to relocations, layoffs, and other life circumstances, some employees did not participate for the full 2 years. A total of 371 participants contributed lunchtime photographic lunch tray data to at least one of 8 time points (Fig 1). The majority of the 371 participants were female (78.4%), and the sample was racially diverse with 46% white and 44% Black (Table 1). The mean body mass index in the cohort at baseline was 32.0 kg/m² and did not differ significantly between groups. The demographics of the cohort reflected those of the entire employee population. Characteristics that differed between the participants across worksites included age, sex, race and occupational roles.

During the baseline year, 371 of the enrolled participants submitted a total of 4721 lunch photographs. During the intervention year, 246 of the participants submitted 3237 photographs. The averages of meal calories over each data collection period in intervention vs. comparison groups are shown in Table 2. The averages were lower in the calories-only group at all eight time points.

In unadjusted models comparing average meal calories after vs before labeling, participants exposed to PACE labels purchased 40.4 fewer calories (P = 0.002), and participants exposed to calorie-only labels purchased 38.2 fewer calories (P = 0.0002). The small difference of 2 fewer calories purchased among participants exposed to PACE labeling vs calorie-only labeling was not significant (P = 0.90). Models adjusting for age, sex, race, occupational role, numeracy level, and health literacy level or those models that had the participants assigned based on > 50% of their lunch photos during the intervention year did not change estimates appreciably.