| | | | HSRe,k | Calories (kcal)f,l | Sodium (mg)f,l | Saturated fat (g)f,l | Trans fat (g)g,k | Total sugars (g)h,k | Free sugars (g)f,l |
|---|
Number of products included | n | 14,767 | 14,767 | 14,230b,i | 14,230b,i | 14,230b,i | 14,767 | 14,491j |
Year1 | | X2 | 124.74 | 43.42 | 117.42 | 7.57 | 25.30 | 1.10 | 6.91 |
| | p-value | < 0.001 | < 0.001 | < 0.001 | 0.006 | < 0.001 | 0.29 | 0.009 |
2017 vs. 2013 | Exp(β)e, f, g, h | 1.095 | 0.990 | 0.945 | 1.028 | 0.963 | | 0.986 |
95% CI | 1.078, 1.112 | 0.987. 0.993 | 0.935, 0.955 | 1.008, 1.049 | 0.949, 0.977 | | 0.976, 0.997 |
P-value | < 0.001 | < 0.001 | < 0.001 | 0.006 | < 0.001 | | 0.009 |
FCR score1,2 | X2 | 72.09 | 3.15 | 0.35 | 6.91 | 3.29 | 42.44 | 56.28 |
p-value | < 0.001 | 0.08 | 0.56 | 0.009 | 0.07 | < 0.001 | < 0.001 |
Exp(β)e, f, g, h | 0.892 | | | 0.980 | | 1.545 | 1.109 |
95% CI | 0.866, 0.918 | | | 0.949, 1.013 | | 1.359, 1.756 | 1.081, 1.139 |
P-value | < 0.001 | | | 0.23 | | < 0.001 | < 0.001 |
Year*FCR score1,2 | | X2 | 0.42 | 10.20 | 2.87 | 11.88 | 27.16 | 3.24 | 3.58 |
| | p-value | 0.52 | 0.001 | 0.09 | 0.001 | < 0.001 | 0.07 | 0.06 |
2017 vs. 2013 | Exp(β)5–8 | | 0.994 | | 0.961 | 0.949 | | |
95% CI | | 0.990, 0.998 | | 0.940, 0.983 | 0.931, 0.968 | | |
P-value | | 0.001 | | 0.001 | < 0.001 | | |
- aThe generalized estimating equation analyses tested whether higher FCR tool scores (entered as a standardized continuous variable) were associated with greater increases in HSRs (i.e., healthier) or decreases in calories, sodium, saturated fat, trans fat, total sugars or free sugars per 100 g (or mL) between 2013 and 2017. Coefficients for the “year” variable measured whether HSRs, calories or nutrient levels (per 100 g or mL) tended to be higher or lower in 2017 compared to 2013. The “FCR score” variable coefficients measured whether there was an association between FCR scores and HSRs, calories or nutrient levels in products (per 100 g or mL). The “year*FCR score” variable examined whether the change in HSRs or calorie/nutrient density over time differed between products offered by companies with higher versus lower FCR tool scores. An association between higher FCR scores and greater improvements in nutritional quality between 2013 and 2017 would be suggested if all three of the following results were observed and statistically significant (p < 0.05): 1) calorie or nutrient amounts were lower in 2017 than 2013 (and HSRs were higher in 2017 than 2013); 2) FCR scores were negatively associated with calorie/nutrient amounts (and positively associated with HSRs); and 3) the “year*FCR score” interaction term variable was negative for calorie/nutrient outcome models (i.e., exp.(β) < 1) and positive for the HSR outcome model (exp(β) > 1; indicating a greater change in nutritional quality in higher-scoring companies’ products between 2013 and 2017). bThe FCR tool quantifies the strength of voluntary recent actions and commitments reported by food companies to reduce energy/portion sizes, sodium, saturated fat, trans fat and (total, added or free) sugars in their products; scores were calculated out of 100%. Companies only or primarily offering beverages were not evaluated for sodium, saturated fat or trans fat. HSRs of products were examined in relation to FCR tool total scores. Details about the FCR scoring tool methodology and a complete breakdown of companies’ scores is provided elsewhere [26]. cFood categories are defined in Health Canada’s Table of Reference Amounts for Food [39]. dBoldface values indicate statistically significant tests of model effects and pairwise comparisons; p-values < 0.05 were considered significant. For all contrasts, the reference category is listed second. Pairwise contrasts are not shown for variables that did not have a significant overall effect. eBased on a model with an inverse gaussian distribution and an identity link. fBased on a model with a Tweedie distribution and a logarithm link. gBased on a model with a normal distribution and an identity link. hBased on a model with a Tweedie distribution and an identity link. iProducts offered by beverage companies did not receive FCR tool scores for sodium, saturated fat or sugars, resulting in the exclusion of 537 products and a smaller sample size for these models. jModel excluded 276 products for which free sugars could not be estimated in the absence of an ingredients list. kExp(β) is the expected change in the mean of the dependent variable for each 1-unit change in a covariate. lA 1-unit change in a covariate multiplies the mean value of the dependent variable by exp.(β)
