Product Reformulation in Response to the 2016 FDA Updated Nutrition Facts Label Authors: C. Ross Hatton, Joshua Petimar, Yutong Zhang, Caroline E. Collis, Tao Hou, Alexandra Ross, Michele Polacsek, Anna H. Grummon, Eric B. Rimm, Jason P. Block Published: 2026 ABSTRACT This cohort study examines whether changes in the nutritional profiles of packaged foods in the US are associated with the adoption of the revised Nutrition Facts label. Introduction Poor diet is a major risk factor for several leading causes of death in the US, including heart disease, diabetes, and certain cancers.1,2 Multiple factors drive the poor nutritional quality of US diets, including the consumption of packaged foods, which constitute most of the calories consumed by adults in the US and are often dense in nutrients the intake of which individuals are recommended to limit, such as added sugar, saturated fat, and sodium.3,4,5 Improving the nutritional quality of packaged foods could therefore meaningfully improve US diet quality.3 Nutrition labels are placed on packaged foods and communicate information about a product’s nutrient content to help consumers make informed food choices.6 In 1993, the US Food and Drug Administration (FDA) mandated the inclusion of the Nutrition Facts label on packaged foods sold in the US to provide consumers with standardized nutrition information.7 These labels, placed on the back or side of the package, require manufacturers to list the serving size, the per-serving content of several nutrients, and the percentage of the recommended daily value that each nutrient comprises. In an effort to better communicate nutrient content and improve the healthfulness of consumers’ food purchases, the FDA revised the Nutrition Facts label in 2016 to emphasize calories (ie, larger and bolder font), remove some nutrients (eg, calories from fat), include new nutrients (eg, added sugars), and update serving sizes to reflect typical amounts consumed.8 Depending on their annual total sales, companies were required to implement the revised Nutrition Facts label by July 2018 (>$10 million) or July 2019 (<$10 million), although these dates were later extended to January 2020 and January 2021, respectively. However, companies could adopt the label any time after the regulation went into effect in July 2016. Given the revised label’s emphasis on calories and added sugar and its changes to serving size, it is possible that this policy was associated with a reformulation of products by the food industry to be healthier, but this potential change has not previously been evaluated. Past evidence suggests that front-of-package labels implemented outside the US have at times encouraged the food industry to reformulate products to reduce those nutrients individuals are recommended to limit.6 For example, evaluations of mandatory warning labels in Chile, Mexico, and Peru, which identify products high in calories, sugar, saturated fat, or sodium, found that food manufacturers often reformulated products to reduce these nutrients enough to avoid the warning labels.9,10,11 In the US, calorie labeling for prepared foods was associated with the introduction of lower-calorie items in packaged and restaurant foods.12,13 Whether the revised Nutrition Facts label was associated with similar changes in the US food supply remains unclear. To address this gap, this study leveraged the staggered adoption of the revised Nutrition Facts label to evaluate whether companies reformulated their products on adoption of the revised label. Companies could respond to label adoption by (1) making their products less calorie or nutrient dense; (2) reducing the serving size, which would make products appear to have fewer calories or nutrients per serving on the label; or (3) introducing healthier products or discontinuing less-healthy products. These responses could also differ across food groups (eg, yogurt vs cereal). Accordingly, the objective of this study was to examine, by food group, whether adoption of the revised Nutrition Facts label was associated with changes in nutrient content or serving size among packaged foods. Methods This cohort study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. The Harvard Pilgrim Health Care Institutional Review Board approved the study. Sample and Data Sources We obtained a dataset of products sold between January 2016 and December 2022 in a large supermarket chain with 185 stores located in the northeastern US. The dataset came from Guiding Stars, an internal shelf-tag labeling system that records product nutrition information for all items sold across the chain. The dataset included universal product codes (UPCs), product descriptions, serving sizes, and nutrient information (eg, calories, added sugar, sodium) for 51 053 products, which we categorized into 32 mutually exclusive food groups using a previously developed classification system.14 We restricted the study period to end in December 2019 because supply chain disruptions, changes in consumer behavior, and lack of label enforcement by the FDA during the COVID-19 pandemic may have influenced labeling and reformulation decisions.15 Additional information on the Guiding Stars dataset and how it was prepared for analyses is included in eAppendix 1 and eFigure 1 in Supplement 1. Exposure To identify the timing of Nutrition Facts label adoption, we merged 2 commercially available databases that record historical label information for packaged foods sold in the US: Label Insight and Syndigo.16,17 Manufacturers and retailers representing more than 80% of the top-selling brands in the US periodically report product data to these databases, including when products displayed the old or revised Nutrition Facts label (details on how Label Insight and Syndigo identify labels are provided in eAppendix 2 in Supplement 1).18 Between June 2016 and December 2019, 199 292 packaged foods were recorded in these datasets, of which 185 008 were recorded solely in Label Insight, 3576 were recorded solely in Syndigo, and 10 708 were recorded in both datasets. Among products recorded in both datasets, the median (IQR) difference in label adoption dates between the 2 datasets was 8 (3-15) months. Companies only periodically report data to Label Insight and Syndigo. As such, products may not have been recorded the month the revised label was adopted, or they may not be recorded at all after label adoption. For example, 67% of products (n = 133 666) are only observed once in these datasets. Identifying the date of label adoption at the product level would lead to extensive missing data regarding the timing of label adoption and would remove any products observed only before or after label adoption. This could also bias results to the null, as the estimated exposure date would often be after label adoption. Using the first date of adoption across a brand’s products is more reliable, and it is reasonable. Implementing new label adoption across all products within a brand avoids the unnecessary additional expenses needed to maintain multiple supply chains for different labels on different products, and enforcement of the new label requirements occurs at the brand level. Moreover, this behavior has been seen in other policy contexts, such as Vermont’s Genetically Modified Organism label, where brands applied the label across all products nationally, even though they were only required to do so in Vermont.19 We therefore identified the earliest date a new label appeared on any product within a brand and used that implementation date across all products within that brand. To create the final analytic dataset of implementation dates and nutrients, we matched these dates to products in the Guiding Stars dataset using a combination of UPCs and brand names. Exclusion Criteria First, we excluded products from food groups that were unlikely to have a Nutrition Facts label or be affected by the 2016 revision (eg, fruit, milk, culinary ingredients, and items with zero or few calories; a full list is available in eAppendix 3 in Supplement 1). We also excluded products sold by weight (eg, bulk nuts and cheese sliced at the delicatessen counter), which typically do not have a Nutrition Facts label. Second, we excluded products from eggs and egg dishes and processed soy and meat alternatives, which comprised too few products to be evaluated. Third, we excluded products that had missing data or poor identification of label adoption dates; among this group, we excluded products that could not be matched from Guiding Stars to Label Insight or Syndigo using either UPCs or brand names. We also removed products from brands with fewer than 5 products to improve precision in our estimated label dates. Brands with fewer products have fewer opportunities to be observed by Syndigo or Label Insight and could be subject to missing or lagged information on the timing of label adoption. Lastly, we removed products from brands that had no new label date recorded, which could have been due to companies ending their reporting of data or Label Insight or Syndigo failing to enter a label date. This produced a final sample that included products across 15 food groups: bread; candy; cereal; cheese; cold or frozen dessert; condiments, sauces, and salad dressings; convenience food (cold); nuts; pizza; processed meat; soup; sugar-sweetened drinks; sweet bread, cake, or cookies; sweet or salty snacks; and yogurt. Outcomes We evaluated 3 primary outcomes: changes in calories per serving, grams of total sugar per serving, and serving size, as these values are presented on the Nutrition Facts label, and food manufacturers may seek to alter the components most visible to consumers. We did not evaluate added sugars per serving, because this information was not present on the original Nutrition Facts label, and we did not have preintervention data on this measure. However, most reductions in added sugar would likely be captured by changes in total sugar. Secondary outcomes included calorie density (calories per 100 g or 100 mL) and sugar density (grams of total sugar per 100 g or 100 mL), as these outcomes are unaffected by changes in serving size. In exploratory analyses, we examined the densities of other nutrients not directly targeted by changes to the Nutrition Facts label: grams of carbohydrates, protein, and saturated fat and milligrams of sodium. Statistical Analysis Although required to adopt the revised Nutrition Facts label by January 2020 or January 2021, depending on their size, companies could do so earlier. We leveraged variation in the timing of adoption to conduct a staggered difference-in-differences analysis comparing early adopters with late adopters. We used the Callaway and Sant’Anna20 estimator, which accommodates heterogeneity in treatment timing and effect, to estimate the average treatment effect on the treated (ATT) of label adoption. This analysis assumes that trends in the outcome in the treated (early adopters of the revised label) and control (never or late adopters) groups would have continued if not for the intervention. While this counterfactual is inherently unobservable, inspecting pretrends for differences can increase confidence that outcome trends would have continued in parallel. We evaluated pretrends without including any adjustment variables and found no evidence of substantially different pretrends for all outcomes (eTable 1 in Supplement 1). We therefore proceeded without covariate adjustment. We used sales weights in all analyses to ensure our findings represented the products customers typically buy; we constructed sales weights by estimating each product’s portion of all sales made during the entire study period. We used these static sales weights, rather than time-varying weights, to avoid allowing changes in sales over time to affect the outcomes. We restricted these analyses to products available at the start of the study period (ie, in the months before the FDA rule was finalized). We conducted 3 exploratory analyses. First, we tested whether products newly introduced during the study period but before their brand adopted the revised Nutrition Facts label had different nutrient profiles than those introduced during the study period but after their brand adopted the label. Second, we tested whether products discontinued before and after adoption of the revised Nutrition Facts label differed in their nutrient profiles. More information about the construction of these samples is provided in eAppendix 4 in Supplement 1. We used linear regression with time fixed effects and SEs clustered at the brand level to test the association between label adoption and the 5 outcomes tested in the primary analyses. Third, we repeated our primary analyses for nutrients not directly targeted by the revised Nutrition Facts label (carbohydrates, protein, saturated fat, and sodium). We conducted 4 sensitivity analyses. First, we assessed whether our primary results were sensitive to the inclusion of low-calorie products by excluding products with a mean of fewer than 25 calories during the study period. Second, we tested moving the label dates 3 months earlier in case our brand-level dates retained some lag. Third, we limited the sample to brands with 2 or more products for which Label Insights or Syndigo reported the same adoption date for the new Nutrition Facts label. Fourth, we repeated our primary analyses without sales weights to test whether higher-selling products drove our findings. Data were analyzed between July 2025 and March 2026 using Stata version 18 (StataCorp). P values are 2 sided, with statistical significance set at P < .05. Results Product Characteristics We identified label adoption dates for 24 816 products in 15 food groups across 1647 brands that had complete nutrition information and were sold between January 2016 and December 2019. There were 1433 (87%) brands that adopted the revised Nutrition Facts label before January 2020 (eFigure 1 in Supplement 1). The final sample included 12 097 products that were available both before and after their brands adopted the updated label, 11 667 of which were from brands that adopted the label during the study period (Table). Table. Characteristics of Products Before and After Adoption of the Revised Nutrition Facts LabelaCharacteristicMean (SD)Reformulated productsNew productsDiscontinued productsPreadoption (n = 11 667)Postadoption (n = 11 667)Never adopted (n = 430)bPreadoption (n = 4570)Postadoption (n = 4806)Preadoption (n = 3510)Postadoption (n = 5852)Calories per serving159 (130)153 (118)126 (94)156 (94)162 (112)150 (93)160 (101)Calories per 100 g or 100 mL297 (168)303 (169)260 (158)313 (165)321 (173)302 (164)330 (166)Total sugar per serving, g10 (11)9 (11)8 (11)9 (9)8 (9)10 (11)9 (10)Total sugar per 100 g or 100 mL, g18 (20)16 (20)15 (20)18 (21)16 (20)20 (23)19 (21)Serving size, g or mL85 (96)80 (93)75 (86)78 (83)81 (90)78 (85)75 (84) a Products were sold at a large northeastern supermarket retailer (n = 185 stores) between January 2016 and December 2019. b Never adoption refers to products in brands that did not adopt the revised label during the study period (ie, as of December 2019). Characteristics of Products Before and After Adoption of the Revised Nutrition Facts Labela Products were sold at a large northeastern supermarket retailer (n = 185 stores) between January 2016 and December 2019. Never adoption refers to products in brands that did not adopt the revised label during the study period (ie, as of December 2019). Changes in Calories per Serving, Total Sugar per Serving, and Serving Size After label adoption, calories per serving, total sugar per serving, and serving size did not change for most food groups (Figure 1). Bread (ATT, 2.0 [95% CI, 0.1-4.0] kcal), candy (ATT, −6.4 [95% CI, −10.6 to −2.2] kcal), cereal (ATT, 11.6 [95% CI, 1.3-21.9] kcal), nuts (ATT, −2.9 [95% CI, −5.7 to 0.0] kcal), and sugar-sweetened drinks (ATT, 4.0 [95% CI, 0.4-7.6] kcal) were the only food groups with statistically significant changes in calories per serving; these changes were small and in variable directions. Statistically significant changes in total sugar per serving after label adoption were found for candy (ATT, −0.6 [95% CI, −1.0 to −0.2] g) and sweet bread, cake, or cookies (ATT, 0.4 [95% CI, 0.1-0.7] g). Variable and generally small changes in serving size were also found after label adoption for bread (ATT, 0.6 [95% CI, 0.0-1.1] g), candy (ATT, −1.3 [95% CI, −2.2 to −0.4] g), cereal (ATT, 2.6 [95% CI, 0.6-4.5] g), soup (ATT, 1.2 [95% CI, 0.1-2.4] g), and sugar-sweetened drinks (ATT, 9.0 [95% CI, 0.0-17.9] mL). Dot Plots of Product Changes in Primary Outcomes Associated With Adoption of the Revised Nutrition Facts Label Products (n = 12 097 packaged foods) were sold at a large northeastern supermarket retailer (n = 185 stores) between January 2016 and December 2019. Average treatment effects on the treated (ATTs) were estimated using a staggered difference-in-differences analysis comparing changes in outcomes among early adopters of the revised Nutrition Facts label vs late and never adopters. Changes in Calorie and Total Sugar Densities Calorie and total sugar densities did not change substantively in any food group, although small, statistically significant changes were found in some groups (Figure 2). The largest change in calorie density was a decline of 8.3 (95% CI, −16.4 to −0.1) calories per 100 g among nut products. Statistically significant changes in total sugar density in variable directions were found for pizza (ATT, 0.4 [95% CI, 0.1-0.7] g per 100 g); soup (ATT, −0.1 [95% CI, −0.3 to 0.0] g per 100 g); sweet bread, cake, or cookies (ATT, 0.3 [95% CI, 0.1-0.5] g per 100 g); and sweet or salty snacks (ATT, 0.4 [95% CI, 0.2-0.7] g per 100 g). Dot Plots of Product Changes in Secondary Outcomes Associated With Adoption of the Revised Nutrition Facts Label Products (n = 12 097 packaged foods) were sold at a large northeastern supermarket retailer (n = 185 stores) between January 2016 and December 2019. Average treatment effects on the treated (ATT) were estimated using a staggered difference-in-differences analysis comparing changes in outcomes among early adopters of the revised Nutrition Facts Label vs late and never adopters. Exploratory and Sensitivity Analyses Consistent with our primary findings, there was limited evidence of differences between products introduced before (n = 4570) and after (n = 4806) or discontinued before (n = 3510) and after (n = 5852) Nutrition Facts label adoption (eTables 2 and 3 in Supplement 1). We also found no evidence of product reformulation across carbohydrates, protein, saturated fat, and sodium (eTable 4 in the Supplement). In sensitivity analyses, removing low-calorie items, moving adoption dates earlier, limiting the sample to brands for which multiple products identified the same date of adoption of the new Nutrition Facts label, and removing the sales weights did not substantially alter our primary findings (eTable 5 in Supplement 1). Discussion This cohort study of the staggered adoption of the 2016 revised Nutrition Facts label found limited evidence that food manufacturers reformulated packaged foods after revised Nutrition Facts label adoption. Manufacturers also did not appear to introduce healthier items or discontinue less healthy items after label adoption, unlike what has been found for menu calorie labeling in the US.12,13 Together, these findings suggest that an association between the 2016 Nutrition Facts label revision and the nutritional quality of US packaged food purchases is unlikely. The 2016 Nutrition Facts label revision could improve nutrition in 3 ways. First, consumers could use the labels to select healthier products. Second, it could prompt reformulation of packaged foods. Third, it could prompt industry to shrink package and serving sizes to appear healthier to consumers. Arguably the largest changes made to the label were the bolding of calorie information and the inclusion of added sugar. Given prior evidence that consumers may struggle to use the Nutrition Facts label, due in part to its location on the back or side of packaged foods and the complexity of the information it communicates,21 these changes may be unlikely to alter consumer behavior directly. Rather, product reformulation by the food industry may be the most promising mechanism by which these revisions could improve US consumers’ food purchases. However, we found no evidence that companies reformulated products, introduced healthier products, or discontinued less-healthy products after label adoption. We similarly found no evidence that the food industry meaningfully changed serving sizes to lower calories or total sugar per serving without actually changing products’ nutrient densities. This suggests the 2016 revision ultimately did little to change the healthfulness of packaged food purchases in the US. In contrast, the capacity to prompt product reformulation has been demonstrated with other labeling systems, and policymakers should consider similar approaches to improve nutrition in the US. For example, menu calorie labeling in the US and international front-of-package warning labels have led to product reformulation across many targeted nutrients and food groups.9,10,11,13,22,23 Key distinctions of these systems include their visibility (ie, front vs side or back of package) and their presentation of nutrient content (eg, high-in or traffic light labels) that make it easier for consumers to understand nutritional quality.6 Labeling systems that are more prominent and facilitate consumer understanding may be more likely to prompt product reformulation, as retaining or attracting customers is a powerful financial incentive for the food industry. In 2025, the FDA proposed a front-of-package label that would convey nutrient content by indicating whether a product is low, medium, or high in saturated fat, sodium, or added sugar.24 Because such labels explain nutrient content, they may be more likely to both alter consumer behavior and instigate product reformulation.6 However, no country has implemented labels identical to the FDA’s proposal; therefore, their effectiveness is uncertain. Policymakers may also wish to explore other nonlabeling policies that could prompt product reformulation, such as voluntary food industry guidelines and taxation of nutrients individuals should limit. For example, the UK’s voluntary sodium reduction targets, which encouraged the food industry to voluntarily reduce the sodium content of packaged foods, led to modest reductions in sodium content.25 The FDA implemented a similar phased sodium reduction initiative in 2021,26 but associated changes in product reformulation have not been formally evaluated. Similarly, sugar-sweetened beverage taxes, particularly when sugar based instead of volume based, can prompt product reformulation to reduce added sugar content.27,28 Limitations This study had several limitations. First, Label Insight and Syndigo may not have documented the revised Nutrition Facts label for individual products at the time of adoption, leading to a lag time. However, we assumed that brands adopted the label across all products at once given the logistic and financial burdens of implementing different labels across different products. This brand-level analysis increased confidence that the correct label date was identified, particularly for brands with many products. Sensitivity analyses assuming earlier adoption of the new label did not alter the study’s findings. Second, the study sample included products sold at a large supermarket retailer in the northeastern US, which does not represent all products available in the US food system. However, the chain sells many top-selling, nationwide brands that are found in many grocery stores. Third, we were unable to evaluate the association between the revised Nutrition Facts label and added sugar, as added sugar was not included on labels (and thus not included in nutrient databases) prior to the revised label adoption. While total sugar may serve as a proxy for added sugar, it could capture changes in naturally occurring sources of sugar. Fourth, we used sales data to identify newly introduced and discontinued products, but these dates may reflect the supermarket’s decision to introduce or withdraw products, instead of decisions by brands to either introduce or discontinue products. Finally, we evaluated early adoption of the revised Nutrition Facts label, which could be subject to selection bias. However, we found similar pretrends among early vs late adopters. In addition, because the staggered analysis compared products that adopted the revised label at a particular time with those that had not yet done so at that time point, factors shaping adoption timing would affect both the treated and control products. For example, if less-healthy products adopted the revised label later, then the treated group would become less healthy over time, as would the control group (as increasingly unhealthy products remained available over time). Conclusions This study found no substantive evidence that adoption of the 2016 revised Nutrition Facts label was associated with changes in the nutrient profile of packaged foods. Policymakers seeking to improve the nutritional quality of consumer food purchases and reduce related chronic disease should prioritize evidence-based interventions (eg, nutrient warning labels) that have the potential to both influence consumers to make healthier food choices and prompt the food industry to improve the nutritional quality of the food supply.