Substitute foods are more likely than their traditional food counterparts to display front-of-package references

Our study drew on a larger survey of FOP references found on foods and beverages sold in three large grocery stores in Toronto conducted between July 2010 and August 2011 (Sacco et al. 2013; Sumanac et al. 2013; Christoforou et al. 2017). A single store was selected from each of the top three food retailers in Canada (Loblaws, Metro, and Sobeys), representing 71% of the total Canadian retail market share (Canadian Grocer 2009). Although these retailers operate multiple store banners, this sample comprised conventional stores in recognition of their dominance both in market share and retail square footage (Canadian Grocer 2009). Data collectors systematically recorded all descriptive text on the front of packaged foods, including product identifiers (i.e., brand and product name, variety, and product size), nutrient content claims, quantitative statements, generic and product-specific health claims, third-party- (e.g., The Heart and Stroke Foundation of Canada’s Health Check) and manufacturer- (e.g., Kraft’s Sensible Solutions) developed symbols and summary systems, as well as any other descriptive or implied references emphasizing the presence or absence of a specific nutrient. References to “natural”, “pure”, or “real” (henceforth all referred to as “natural”), and “organic” references were also noted.

Fresh produce, meat, poultry and fish, and dried herbs and spices were not included in the data collection because they were unlikely to bear FOP labelling. Products found in the pharmacy and infant food sections were also excluded because these products are designed for specific population subgroups and special dietary usage. After removing duplicate products, 20520 unique packaged items were captured in the database. Products were considered unique if they differed from similar products on the basis of any product identifier (e.g., different product names, package sizes, etc.).

For the purposes of this study, substitute products were defined as foods that (i) are designed to function as alternates or substitute foods for traditional foods in the diet, (ii) have been created through the use of novel or unusual ingredients and (or) formulation processes (Carayannis et al. 2003; Moskowitz et al. 2006), and (iii) are more highly processed than their traditional counterparts (Monteiro 2009; Moubarac et al. 2017). All products were classified by food category (n = 52) using The Bureau of Nutritional Sciences’ food grouping developed by Health Canada (Bureau of Nutritional Sciences, Food Directorate 2004). This categorization was expanded to 79 groupings to more fully capture major differences in composition, variations in modes of preparation, and form of presentation (e.g., frozen fruit versus dried fruit). All foods were coded for level of processing using the NOVA framework developed by Monteiro et al. (2013), which classifies foods into four groups along a continuum based on the nature, extent, and purpose of industrial food processing (Moubarac et al. 2017). Food categories were then screened to identify foods that comprised substitute–traditional pairs, defined as foods that had interchangeable use within the diet but inherently different formulations such that the substitute member had innovative ingredients and (or) formulation processes, resulting in a higher degree of product processing when compared with its traditional counterpart. Following these criteria, products that differed by a single adulteration (e.g., plain yogurt and yogurt with inulin or probiotics, or plain milk and lactose free milk) were not included in the analysis (Table S1).

Each product was coded for the presence of any form of nutrition reference (e.g., summary systems and symbols, quantifying statements, nutrient content, and health claims). To investigate the nature of the nutritional promotion, these references were further classified as “negative” or “positive”. Negative references were defined as those that conveyed the reduction or absence of a nutrient for which there is public health messaging to limit exposure (e.g., “low sodium”, “trans fat free”, and “no added sugar”) (Trans Fat Task Force 2006; Sodium Working Group 2010; Health Canada 2014). Positive nutrition references were those that highlighted the presence or addition of a nutrient deemed beneficial (e.g., “good source of calcium”, and “high in fibre”). FOP nutrition references were also categorized as being regulated (e.g., nutrient content and disease risk reduction claims) or unregulated, following previous work that suggests that the use of unregulated references may signal lower concentrations of a particular nutrient than is found in a product displaying regulated references (Sacco et al. 2013; Metcalfe and Elliott 2015). Regulated nutrition references were identified based on the prescribed wording and permitted wording variations outlined in the Canadian Food and Drug Regulations (Government of Canada 2003) and Canadian Food Inspection Agency’s Guide to Food Labelling and Advertising (CFIA 2014). All other FOP nutrition references, including quantifying statements (e.g., “x grams of protein”) and those that appeared to more deliberately avoid compositional requirements through wording manipulation (e.g., regulated “plus energy” versus unregulated “energy+”) were classified as unregulated. To more explicitly compare the nature of the nutritional messaging found within substitute and traditional pairs, all FOP nutrition references on substitute foods were additionally dichotomized as highlighting a compositional attribute that was either inherent to its traditional counterpart or one that reported a formulation that was unattainable by the traditional food. The total number of references on the FOP, including nutrition, “organic”, and “natural” references, on each product was also coded for each product. In addition, the number of references on substitute and traditional foods that relayed information about nutritional attributes inherent to the traditional counterpart was recorded.

Multilevel logistic regression modeling using PROC GLIMMIX was conducted in SAS statistical software package (version 9.4, SAS Institute, Cary, North Carolina) to assess the relationship between the presence and nature of FOP references on substitute versus traditional products, accounting for the fact that products are nested within their individual substitute–traditional pairs. The extent of FOP labelling (i.e., total number of the nutrition, “organic”, and “natural” references) on substitute and traditional pairs was compared by fitting a negative binomial distribution, appropriate for count outcome variables of this nature (e.g., values of 0,1,2,3…) (Gardner et al. 1995), to the PROC GLIMMIX model. The model was repeated to compare the number of nutrition references on substitute and traditional foods that referred to compositional characteristics inherent to the traditional foods.

Recognizing that the direction of one of these pairs (fruit/vegetable drinks and juices) diverged from the pattern observed in the others, a detailed examination of the nature of FOP nutrition references on these products was undertaken (Table S2). Multilevel logistic regression modeling was repeated excluding this pair as a robustness test to confirm that the effect of this divergent pair was to attenuate the findings of the analyses described above (Tables S3 and S4).