Table 5.

 Results of linear regression analyses for mean energy content (kcal) of lunchtime fast food purchases in 2009, by customers’ use of information on calorie labelling

	Model 1*			Model 2*
	Mean (95% CI) energy content/purchase (kcal)	P value		Mean (95% CI) energy content/purchase (kcal)	P value
Estimated marginal means
Did not use information	862.0 (849.0 to 875.2)	—		859.4 (851.8 to 867.0)	—
Used information	765.9 (743.0 to 788.9)	—		781.6 (767.3 to 795.9)	—
Parameter estimates
Did not use information	Reference	—		Reference	—
Used information	−96.2 (−71.8 to −120,7)	<0.001		−77.8 (−62.3 to −93.4)	<0.001
Sex (women–men)	−95.2 (−113.6 to −76.7)	<0.001		−46.9 (−57.3 to −36.5)	0.08
Age (years):
18–24	98.4 (63.3 to 133.5)	<0.001		44.6 (27.9 to 61.3)	<0.001
25–34	112.7 (89.5 to 125.9)			27.0 (13.4 to 40.5)
35–44	97.1 (71.6 to 122.6)			26.5 (14.1 to 39.0)
≥45	Reference			Reference
Poverty level†:
Of customer residence	63.7 (−13.3 to 140.6)	0.10		40.1 (6.2 to 74.1)	0.02
Of store location	10.4 (−63.2 to 84.1)	0.78		26.1 (−16.3 to 68.6)	0.23
Description of purchase:
No of food items	—	—		148.6 (130.9 to 166.4)	<0.001
Beverage (0=No; 1=Yes)	—	—		25.2 (8.7 to 41.7)	0.003
Cost (inflation adjusted)	—	—		102.8 (96.4 to 109.2)	<0.001

Models include city residents only (poverty level for customer residence could not be determined for non-city residents).

*Model 1 adjusted for restaurant chain, sex, and neighbourhood poverty. Model 2 further controlled for the number of food items purchased, beverage purchased, and cost. Both models included a variable for chain, to adjust for change in customer volume across the two years.

†Poverty is a continuous variable defined as the percentage of households in the store’s zip code that were below twice the national poverty level.