The effect of calorie menu boards on calorie ordering and purchasing
All seven studies compared calorie ordering and purchasing in two conditions: calorie label versus no calorie label. Two studies reported that calorie menu labels reduced the calories purchased [6, 21], one reported significant reductions in calories purchased at some chains (but not others), three reported no effect on calories purchased [17, 18, 20] and one reported a slight increase in calories purchased .
Among the observational studies, Elbel et al. found that in New York City, purchasing behavior of children and adolescents did not differ before and after calorie labels were implemented on menu boards, with patrons purchasing a mean of 643 calories before labeling and 652 calories (p = 0.82) after restaurants introduced menu labels . The authors also observed a non-significant change in purchasing behavior over the same time period among children and adolescents in Newark, NJ, where calorie labels were not introduced (611 vs. 673 calories, p = 0.37). A companion study of adults also showed a non-significant difference in New York City . Adults purchased a regression-adjusted mean of 825 calories (95% CI: 779-870) before calorie labeling and 846 calories (95% CI: 758-889) after calorie labeling. There was also a non-significant trend among adults in Newark, NJ with 823 calories (95% CI: 802-890) in the pre-labeling time period and 826 calories (95% CI: 746-906) in the post-labeling time period .
Also in New York City, Dumanovsky et al. collected survey and purchase data before calorie labeling implementation in 2007, and nine months after implementation in 2009. They collected data from the 11 largest fast food chains, and found no change in mean calories purchased overall between study periods in 2007 and 2009 (828 vs. 846 calories, P = 0.22). When examining data for each chain individually, they found a reduction in mean calories purchased for three chains (McDonald's 829 vs. 785 calories, P = 0.02; Au Bon Pain 555 vs. 475 calories, P < 0.001; KFC 927 vs. 868 calories, P < 0.01), no significant difference for 7 chains (Burger King, Wendy's, Popeye's, Domino's, Pizza Hut, Papa John's, Taco Bell), and an increase for one chain (Subway 749 vs. 882, P < 0.001). The study did not include a control population.
Though it was a small change, Finkelstein et al. did observe a small, statistically significant (but we do not think clinically significant) increase in calories purchased per transaction after calorie labels were added to menus in King County, WA . Patrons purchased 5.7 (p < 0.05) more calories after calorie labels were introduced on menu boards inside restaurants, and 2.9 (p < 0.05) more calories after calorie labels were introduced on drive-thru menu boards. In the control county, they did not observe a significant trend. Moreover, a difference-in-difference regression analysis found that calories per transaction were not reduced after the legislation .
In a study of entrée purchasing in a college dining hall, Chu et al. reported a significant but modest decrease in calories per entrée sold during the two weeks that calorie labels were posted on menu boards (treatment). They calculated average calories per sale using sales data furnished by the cafeteria. In the two weeks before posting calorie information (pretreatment), the average energy content was 646.5 calories per entrée. This average dropped 12.4 calories per entrée sold on the first day of calorie posting (p = 0.007) and remained lower throughout the treatment period. Though statistically significant, an average reduction of 12.4 calories may not be clinically significant.
In contrast to studies utilizing only purchasing behavior, the two experimental studies conducted in laboratory settings allowed researchers to measure both calories ordered and calories consumed (discussed below)[20, 21]. Harnack et al. found no significant difference in calories ordered among four menu labeling conditions manipulating availability of calorie labels and value pricing (calorie labels + value pricing 874, calorie labels without value pricing 842, no calorie labels + no value pricing 882, and no calorie labels + value pricing (control) 828 calories, p = 0.62).
Roberto et al. tested three types of menus: one had no calorie labels (no label), one had calorie labels (calorie), and one had calorie labels and a statement that the recommended daily caloric intake was 2000 calories (calorie + information). They found that menu type had a statistically significant effect on calorie ordering (p = 0.04). Significant differences were found between the no label and calorie labeled menus (no label 2189, calorie 1862 calories, p = 0.03), and also a significant difference between the no label menus and the calorie + information menus (1860 calories, p = 0.03). The difference between the calorie menus and calorie + information menus was not statistically significant (p = 0.99). It is not clear why the difference in calories ordered between the groups appears to be more clinically significant than those noted in other studies . However, the average number of calories ordered was also high compared to previous studies, which may account for some of this difference.
The effect of calorie menu labels on calorie consumption
As noted above, two studies measured calories consumed in addition to calories ordered or calories purchased [20, 21]. The distinction is an important one since consumers might theoretically respond to calorie posting on menus by changing the amount they eat rather than the amount they order. Harnack et al. found, however, that participants overall did not differ significantly in the number of calories they consumed by menu type (no label 739, calorie labels 805, no value pricing 761 calories, p = 0.25). Subgroup analysis did demonstrate a difference in calories consumed. Men in groups with menus listing calorie information and those without value pricing consumed more calories than those with control menus (p = 0.01).
Roberto et al. also found no significant difference between calorie consumption when they examined consumption by menu type overall (no label 1459 vs. calorie label 1335 vs. calorie + information 1256, p = 0.12). However, when they combined the two calorie label menus and compared them to the no label menu, they did find those in the labeled condition consumed fewer calories than those in the no label condition (label 1286 vs. no label 1466, p = 0.04). The credibility of this result is questionable considering the exploratory circumstances in which it was found. The average number of calories consumed was very high for a single meal.