**Official Title:** Changes in sugar-sweetened beverage purchases one year after Chile’s front-of-package warning labels and marketing restrictions: a pre-post analysis **Keywords:** 1. Sugar-sweetened beverages 2. SSBs 3. Front-of-package warning labels 4. School food policies 5. Food marketing 6. Obesity prevention 7. Food policy 8. Obesity prevention policy 9. Latin America 10. Global nutrition 11. Socio-economic disparities 12. SES disparities 13. Chilean Law of Food Labelling and Advertising 14. Changes in SSB purchase 15. Chile **Funders:** - Bloomberg Philanthropies - International and Development Research Agency of Canada (IDRC) #107731-002 - Chilean National Science Council (Fondecyt #1161436) **Principal Investigator:** Lindsey Smith Taillie, University of North Carolina of Chapel Hill (UNC), USA **Sponsor**: Camila Corvalán, Institute of Nutrition and Food Technology, University of Chile **Brief Summary:** In April 2012, the Chilean Senate passed the country-wide National Law of Food Labelling and Advertising, which has 3 major components: (i) point-of-food purchase labelling to improve consumer information through front-of-package (FOP) easy-to-understand labels, including specific messaging around sugars, saturated fats, sodium, and energy, (ii) complete restrictions on marketing and advertising of unhealthy foods to children under age 14y over an array of media and hours of coverage that cover a great deal of media time, and (iii) complete restriction of sale/offer (i.e. kiosks, cafeterias, school feeding programs) of unhealthy foods in schools and nurseries. The implementation began on July 1, 2016, and over a 3-year period becomes increasingly stringent with regards to the nutritional quality of food products. The aim of this study is to assess the changes in beverages’ purchase in Chilean households after the implementation of the first phase of the Chilean Law of Food Labelling and Advertising entered in force. For this purpose, we will use purchase data from Kantar World Panel, from 2014 to December 2017. We hypothesized that compared to the counterfactual (what would have happened with beverage purchases had the regulation not be implemented based on pre-regulation trends i.e. comparing the predicted post-regulation purchases based on the pre-regulation trend with observed post-regulation purchases), post-regulation: - Overall, the total volume, sugar, and calories of regulated beverages will decline in the post regulation period. - Volume of beverages in the baseline period that were regulated in July 2016 and remained regulated in the post-regulation period will decline. - Volume of beverages in the baseline period that were regulated in July 2016 but became unregulated in the post-regulation period will increase. - Calories and sugar of regulated beverages that remained regulated will decrease. - Calories and sugar of regulated beverages that became unregulated will decrease. - Households with a higher level of household head education will show larger declines in regulated beverage purchases than lower-educated households. - Households with children <14y will show larger declines in regulated beverage purchases than households without children. - Volume, calories, and sugar of unregulated beverages will increase. - Calories and sugars of total beverage purchases will decline. - Volume of total beverage purchases will not change **Study Detailed Description:** We will use a longitudinal data set of household food purchases from 2014-2017 obtained from Kantar World Panel Chile. The data are based on weekly purchases of fast-moving consumer goods by households from cities with more than 20,000 inhabitants, representing 74% of the urban population. The total sample is approximately 2,000 households. Interviewers visited households weekly to collect data on food purchases using a handheld bar code scanner. First, information on purchases was collected either by scanning products' bar codes on the packages or by using a codebook to assign bar codes tor bulk products or other products without bar codes. Second, households were instructed to keep all receipts so that interviewers could match purchases each week and determine the store where they were purchased (specifically for frequently consumed products). Finally, interviewers checked household pantries, and households stored empty product packages in a bin between interviews to ensure that products were not counted twice. Information collected for each beverage purchase included volume or weight, bar code expenditure, price per unit, retail channel, brand, package size, and date of purchase [1]. The data on each beverage purchase were linked at the bar code level to a nutrition facts panel gathered from photographs that a team of Chilean nutrition research assistants collected directly at stores (79.8% of products) [2-7], Mintel Latin America (19.9%), or Mintel North America (0.2%) or imputed with a systematic match based on similar products using package description, brand, and manufacturer (less than 0.1% of each beverage category) [1]. We categorized households into four socioeconomic status (SES) categories using the information on household head schooling provided by Kantar World-Panel Chile. In addition, in the absence of an experimental design and a control group to compare with as trends in income, economic activity, and other market factors could influence the price and quantity of beverage purchases [8], we will control for several macroeconomic measures at the region-month level. These may include unemployment rate, population (in thousands), construction permits (meters2) as a predictor of beverage purchases [9], the Supermarket Sales Index, and the Regional Economic Index, using official figures from the Chilean National Institute of Statistics [10]. ****Statistical analysis**** First, we will conduct analyses to check the consistency of the data, including box plots of mean volume, calories, and sugar of beverage purchases (unadjusted) overall and by subcategory in each month to ensure that trends look roughly the same across each of the 3 main outcomes (volume, calories, and sugar). Second, we will conduct an outliers analyses. Within each beverage subcategory, for volume, calories, and sugar, we will examine purchases in the 99th percentile and above to understand if there are beverage purchases that seem implausible based on typical daily per-capita intakes of fluids. We will not consider outliers at the opposite end of the distribution (near zero), as these could simply reflect households who chose not to purchase packaged beverages (i.e., they only drink tap water or beverages from restaurants). We will consider excluding outliers based on implausible values, and conduct sensitivity analyses to understand whether excluding these outliers changes results. Next, we will conduct descriptive analyses, including changes in unadjusted trends in the percent consumers of regulated and unregulated beverages, overall and by beverage type, across all months. We will also examine unadjusted changes in volume, sugar, and calories over time. The unit of analyses for these and subsequent analyses will be volume, calories, or sugar per capita per month. Third, we will implement a model building procedure to determine which model to use for the adjusted comparison of observed post-regulation beverage purchases compared to counterfactual post-regulation purchases (i.e., predicted beverage purchases in the post-regulation period predicted from pre-regulation trends). We will use fixed effects models of changes in volume of overall regulated and unregulated beverages as our main model for model building, in order to provide comparability with previous studies of regulated beverages. The overall models and model-building procedure will use the Colchero 2016 [11] and Colchero 2017 [12] papers as a reference, because these are also pre-post analyses of beverage purchases after a regulation was instituted. We will consider January 1, 2015 to June 30, 2016 to be the pre-regulation period and July 1, 2016- December 31, 2017 to be the post-regulation period. If the distribution of the outcomes of interest is not normal, we will log-transform the outcomes for the analyses and then apply Duan-smearing factors to back-transform outcomes for presentation in the paper. Our first model consideration will be whether there is a high proportion of non-consumers (<10%). If the probability of non-consuming within a month is <10%, we will use inverse probability weights. Otherwise, we will use two-part models to account for the correlation between probability of consuming a beverage and the amount consumed. Secondly, we will explore methods for specifying the regulation period and controlling for seasonality and time trends. We will consider two approaches: 1) using a pre-post regulation dummy variable with an interaction with a time count (i.e., months, quarters, semesters), which was used in the Colchero 2016 paper [11]; and 2) using a pre-post dummy with binary variables for each month to adjust for seasonality, which was used in the 2017 Colchero paper [12]. The former approach is preferred due to its ability to allow us to model non-linear trends and will be selected so long as there is enough sample size within each month. Third, we will consider different approaches for examining differences in response to the regulation by SES and households including children. We will consider using a stratified model (the approach used by Colchero 2016 and 2017 [11, 12]) if the sample size is enough. Alternately, we will consider including interactions for household SES (either specified as two dummy variables for high vs. low, or a four-level variable for low, low-middle, high-middle, and high, depending on sample size) and whether households have children with <14 years. In each model, we will consider including the following covariates as potential confounders based on previous literature, including the Colchero papers [11, 12]: household size and gender/age composition, educational attainment, employment status of household head and the main food preparer, and region-month level contextual factors, including unemployment rate, construction permits, population size, supermarket sales index, and regional economic index. We will not include variables which have high correlations with other co-variates of interest (for example, educational attainment may be highly correlated with household SES; in this case we will select one of these variables as the main measure of a households social, economic, and educational status). We will not include prices in these models because prices are a possible mediator of time-based changes in purchases (that is, companies could have changed the prices of products in response to the regulations -though unlikely-, which could change consumer purchases). Sensitivity analyses will include: • Using a longer pre-regulation period (from January 2014). In this case, we will also need to include a dummy variable for the pre/post tax period (the sugar sweetened beverages (SSB) tax rate changes implemented in October 2014). • A second set of analyses will attempt to deal with the fact that nutrition facts panel data was collected in early 2016 (6 months prior to implementation) and early 2017 (6 months after implementation). This means that for 2016, we do not know when products were reformulated (which may cause them to shift from being regulated to unregulated). To examine this issue, we will consider the following: o In descriptive analyses, exclude the period immediately around the regulation implementation (i.e. 6 months before and 6 months after), so that the relevant comparisons are for January 1, 2015-December 31, 2015 to January 1, 2017- December 31, 2017. o If possible, test different methods of linking the household beverage purchase data to the nutrition facts panel data for the period of July 1, 2016 to December 1, 2016. For example, we will experiment with allowing the percentage of products that link to the post-regulation nutrition facts panel data to increase slowly over time. For example, in July 2016, we would select a random subsample of 15% of beverage purchases to link to the post-period, and 85% go to pre-period; we will increase the % going to post-period by 15% each month) **OUTCOME MEASURES** **Outcome 1** Primary Outcome Measure: Title: Changes in the purchase of regulated beverages post-regulation Time Frame: Jan 1, 2015- December 31, 2017 Description: - Changes in volume, calories, and sugars of regulated beverages post-regulation - Changes in the purchase of regulated beverages post-regulation, by household head education level - Changes in the purchase of regulated beverages post-regulation, by households with and without children **Outcome 2** Primary Outcome Measure: Title: Changes in the purchase of unregulated beverages post-regulation Time Frame: Jan 1, 2015- December 31, 2017 Description: - Changes in volume, calories, and sugars of unregulated beverages post-regulation - Changes in the purchase of unregulated beverages post-regulation, by household head education level - Changes in the purchase of unregulated beverages post-regulation, by households with and without children **Outcome 3** Primary Outcome Measure: Title: Changes in the purchase of total beverages post-regulation Time Frame: Jan 1, 2015- December 31, 2017 Description: - Changes in volume, calories, and sugars of total beverages post-regulation - Changes in the purchase of total beverages by post-regulation, by household head education level - Changes in the purchase of total beverages post-regulation, by households with and without children **Outcome 4** Primary Outcome Measure: Title: Changes in volume and nutritional content of total, regulated and unregulated beverages post-regulation, by beverages category Time Frame: Jan 1, 2015- December 31, 2017 Description: - Changes in volume, calories, and sugars of total, regulated and unregulated beverages post-regulation, by beverage categories. - Beverage categories may include: Regulated carbonated soda, Regulated non-carbonated fruit drinks, Regulated flavored milks, Regulated other beverages, Unregulated carbonated soda, Unregulated non-carbonated fruit drinks, Unregulated milks, Unregulated waters, Unregulated other beverages. Citations 1. Caro JC, Corvalán C, Reyes M, Silva A, Popkin B, Taillie LS. Chile's 2014 sugar-sweetened beverage tax and changes in prices and purchases of sugar-sweetened beverages: An observational study in an urban environment. PLoS Med. 2018;15(7):e1002597. PMID: 29969444. doi: 10.1371/journal.pmed.1002597. 2. Kanter R, Reyes M, Corvalán C. Photographic Methods for Measuring Packaged Food and Beverage Products in Supermarkets. Curr Dev Nutr. 2017;1(10):e001016. PMID: 29955678. doi: 10.3945/cdn.117.001016. 3. Sanchez M, Oliva D, Kanter R, Reyes M, Corvalan C (2017) Photo taking protocol: for the monitoring of labels from packaged food products in supermarkets, small stores and elsewhere. Center for the Prevention of Obesity and Chronic Diseases (CEPOC). Institute of Nutrition and Food Technology (INTA)-University of Chile 4. Swinburn B, Sacks G, Vandevijvere S, Kumanyika S, Lobstein T, Neal B, Barquera S, Friel S, Hawkes C, Kelly B, L'abbé M, Lee A, Ma J, Macmullan J, Mohan S, Monteiro C, Rayner M, Sanders D, Snowdon W, Walker C; INFORMAS. INFORMAS (International Network for Food and Obesity/non-communicable diseases Research, Monitoring and Action Support): overview and key principles. Obes Rev. 2013;14 Suppl 1:1-12. PMID: 24074206. doi: 10.1111/obr.12087. 5. Neal B, Sacks G, Swinburn B, Vandevijvere S, Dunford E, Snowdon W, Webster J, Barquera S, Friel S, Hawkes C, Kelly B, Kumanyika S, L'Abbé M, Lee A, Lobstein T, Ma J, Macmullan J, Mohan S, Monteiro C, Rayner M, Sanders D, Walker C; INFORMAS. Monitoring the levels of important nutrients in the food supply. Obes Rev. 2013;14 Suppl 1:49-58. PMID: 24074210. doi: 10.1111/obr.12075 6. Ng SW, Dunford E. Complexities and opportunities in monitoring and evaluating US and global changes by the food industry. Obes Rev. 2013 Nov;14 Suppl 2:29-41. PMID: 24103006. doi: 10.1111/obr.12095. 7. Dunford E, Webster J, Metzler AB, Czernichow S, Ni Mhurchu C, Wolmarans P, Snowdon W, L'Abbe M, Li N, Maulik PK, Barquera S, Schoj V, Allemandi L, Samman N, de Menezes EW, Hassell T, Ortiz J, Salazar de Ariza J, Rahman AR, de Núñez L, Garcia MR, van Rossum C, Westenbrink S, Thiam LM, MacGregor G, Neal B; Food Monitoring Group. International collaborative project to compare and monitor the nutritional composition of processed foods. Eur J Prev Cardiol. 2012;19(6):1326-32. PMID: 21971487. doi: 10.1177/1741826711425777. 8. awkes C, Alderman H, Chaloupka F, Harris J, Kumanyika S, Smed S, Story M, Swinburn B, Willett W. Principles behind evaluations of national food and beverage taxes and other regulatory efforts. Obes Rev. 2017 Nov;18(11):1374-1375. PMID: 28925079. doi: 10.1111/obr.12594. 9. Stock JH, WatsonMW. New indexes of coincident and leading economic indicators. NBER macroeconomics annual 1989;4: 351±394. 10. Instituto Nacional de Estadística de Chile [National Institute of Statistics]. Estadisticas sociales, demograficas y economicas [Social, demographic and economic statistics]. 2017. 11. Colchero A, Rivera JA, Popkin BM, Ng SW. Beverage purchases from stores in Mexico under the excise tax on sugar sweetened beverages: observational study. BMJ. 2016;352(h6704). PMID: 26738745 12. Colchero MA, Rivera-Dommarco J, Popkin BN, Shu Wen. In Mexico, Evidence of Sustained Consumer Response Two Years after Implementing a Sugar-Sweetened Beverage Tax. Health Affairs. 2017;Published online before print. PMID: 28228484. doi: 10.1377/hlthaff.2016.1231