Taxation of unprocessed sugar or sugar‐added foods for reducing their consumption and preventing obesity or other adverse health outcomes

Abstract Background Global prevalence of overweight and obesity are alarming. For tackling this public health problem, preventive public health and policy actions are urgently needed. Some countries implemented food taxes in the past and some were subsequently abolished. Some countries, such as Norway, Hungary, Denmark, Bermuda, Dominica, St. Vincent and the Grenadines, and the Navajo Nation (USA), specifically implemented taxes on unprocessed sugar and sugar‐added foods. These taxes on unprocessed sugar and sugar‐added foods are fiscal policy interventions, implemented to decrease their consumption and in turn reduce adverse health‐related, economic and social effects associated with these food products. Objectives To assess the effects of taxation of unprocessed sugar or sugar‐added foods in the general population on the consumption of unprocessed sugar or sugar‐added foods, the prevalence and incidence of overweight and obesity, and the prevalence and incidence of other diet‐related health outcomes. Search methods We searched CENTRAL, Cochrane Database of Systematic Reviews, MEDLINE, Embase and 15 other databases and trials registers on 12 September 2019. We handsearched the reference list of all records of included studies, searched websites of international organisations and institutions, and contacted review advisory group members to identify planned, ongoing or unpublished studies. Selection criteria We included studies with the following populations: children (0 to 17 years) and adults (18 years or older) from any country and setting. Exclusion applied to studies with specific subgroups, such as people with any disease who were overweight or obese as a side‐effect of the disease. The review included studies with taxes on or artificial increases of selling prices for unprocessed sugar or food products that contain added sugar (e.g. sweets, ice cream, confectionery, and bakery products), or both, as intervention, regardless of the taxation level or price increase. In line with Cochrane Effective Practice and Organisation of Care (EPOC) criteria, we included randomised controlled trials (RCTs), cluster‐randomised controlled trials (cRCTs), non‐randomised controlled trials (nRCTs), controlled before‐after (CBA) studies, and interrupted time series (ITS) studies. We included controlled studies with more than one intervention or control site and ITS studies with a clearly defined intervention time and at least three data points before and three after the intervention. Our primary outcomes were consumption of unprocessed sugar or sugar‐added foods, energy intake, overweight, and obesity. Our secondary outcomes were substitution and diet, expenditure, demand, and other health outcomes. Data collection and analysis Two review authors independently screened all eligible records for inclusion, assessed the risk of bias, and performed data extraction.Two review authors independently assessed the certainty of the evidence using the GRADE approach. Main results We retrieved a total of 24,454 records. After deduplicating records, 18,767 records remained for title and abstract screening. Of 11 potentially relevant studies, we included one ITS study with 40,210 household‐level observations from the Hungarian Household Budget and Living Conditions Survey. The baseline ranged from January 2008 to August 2011, the intervention was implemented on September 2011, and follow‐up was until December 2012 (16 months). The intervention was a tax ‐ the so‐called 'Hungarian public health product tax' ‐ on sugar‐added foods, including selected foods exceeding a specific sugar threshold value. The intervention includes co‐interventions: the taxation of sugar‐sweetened beverages (SSBs) and of foods high in salt or caffeine. The study provides evidence on the effect of taxing foods exceeding a specific sugar threshold value on the consumption of sugar‐added foods. After implementation of the Hungarian public health product tax, the mean consumption of taxed sugar‐added foods (measured in units of kg) decreased by 4.0% (standardised mean difference (SMD) −0.040, 95% confidence interval (CI) −0.07 to −0.01; very low‐certainty evidence). The study was at low risk of bias in terms of performance bias, detection bias and reporting bias, with the shape of effect pre‐specified and the intervention unlikely to have any effect on data collection. The study was at unclear risk of attrition bias and at high risk in terms of other bias and the independence of the intervention. We rated the certainty of the evidence as very low for the primary and secondary outcomes. The Hungarian public health product tax included a tax on sugar‐added foods but did not include a tax on unprocessed sugar. We did not find eligible studies reporting on the taxation of unprocessed sugar. No studies reported on the primary outcomes of consumption of unprocessed sugar, energy intake, overweight, and obesity. No studies reported on the secondary outcomes of substitution and diet, demand, and other health outcomes. No studies reported on differential effects across population subgroups. We could not perform meta‐analyses or pool study results. Authors' conclusions There was very limited evidence and the certainty of the evidence was very low. Despite the reported reduction in consumption of taxed sugar‐added foods, we are uncertain whether taxing unprocessed sugar or sugar‐added foods has an effect on reducing their consumption and preventing obesity or other adverse health outcomes. Further robustly conducted studies are required to draw concrete conclusions on the effectiveness of taxing unprocessed sugar or sugar‐added foods for reducing their consumption and preventing obesity or other adverse health outcomes.


T A B L E O F C O N T E N T S
This review may be of interest to government public health agencies, policy decision-makers, food retailers, and food industries. This review and subsequent updates of this review may change policy and a ect a government's motivation to create a tax on unprocessed sugar and foods with added sugar. It also may motivate food industries to reformulate their products to contain lower levels of added sugar.

What question does this review aim to answer?
We wanted to know if taxation of unprocessed sugar and foods with added sugar (other than sugar-sweetened beverages (SSBs)) reduced their consumption, changed people's energy intake, and reduced overweight and obesity. We also wanted to know if taxation changed people's diet and spending, and had an e ect on other diet-related health problems.
Which studies were included in the review?
We searched for ongoing or published studies up to October 2019. Of a total of 24,454 records retrieved, we identified one 'interrupted time series' (ITS) study meeting our eligibility criteria to assess the impact of a tax on sugar-added foods (but not unprocessed sugar). The study used data from the Hungarian Household Budget and Living Conditions Survey, with observations from 40,210 households. Evidence from the study included a 'baseline' (the situation before taxation), ranging from January 2008 to August 2011. The Hungarian public health product tax was implemented on September 2011. The duration of the follow-up period (measuring the e ects of taxation) was 16 months. The study was funded by the Scottish Institute for Research in Economics (SIRE) Early Career Engagement Grant.
What does the evidence from the review reveal?
The included study provided very limited evidence that taxing foods with added sugar reduced their consumption by 4%. We are very uncertain about this evidence because the study did not use the strongest methods, looked at other kinds of taxation as well as taxing foods with added sugar, and may not have correctly classified food types. We are uncertain whether taxing foods with added sugar has an e ect on reducing their consumption. The included study did not investigate the e ects of taxing unprocessed sugar.

What should happen next?
Further research is needed to assess the e ectiveness of taxing unprocessed sugar or foods with added sugar for reducing their consumption and preventing obesity or other adverse health outcomes. Studies should take place in countries that have implemented these taxes and should look at cost-e ectiveness as well as the health benefits of taxing unprocessed sugar or foods with added sugar as a public health policy for preventing overweight, obesity or other adverse health outcomes. Countries that have implemented these taxes are Bermuda, Dominica, Hungary, India, Norway, the Navajo Nation (USA), and St. Vincent and Grenadines.

Summary of findings for the main comparison. 'Summary of findings' table for primary outcomes: Taxation of sugar-added foods compared to no taxation for reducing consumption of sugar-added foods
Taxation of sugar-added foods compared to no taxation for reducing consumption of sugar-added foods Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
a Non-randomised study (downgraded to low), downgraded one additional level (to very low) for risk of bias due to simultaneous intervention of other taxes and likely misclassification of food products as to whether taxed or untaxed. Certainty is also a ected by indirectness because the Hungarian tax is related to specific sugar contents in the particular food categories that were taxed, and the study measured purchased quantities and not consumption. Thus, it is not a direct representation of the e ect of a complete tax on sugar or sugar-added foods. b No study measured e ects of taxing unprocessed sugar or sugar-added foods on the consumption of unprocessed sugar, energy intake from unprocessed sugar or sugar-added foods, total energy intake, overweight or obesity. There was no direct substitution effect. The mean consumption of untaxed sugar-added foods even decreased after the implementation of the tax by 1.3% (SMD −0.013, 95% CI −0.05 to 0.02). The effect is based on very low-certainty evidence.

Substitution: difference in mean consumption (purchased quantities) of untaxed sugar-added foods compared with untaxed sugar-added foods
Assessed with percentage change Follow-up: 16 months The mean consumption of taxed sugar-added foods differed from the mean consumption of untaxed sugar-added foods after the implementation of the intervention by 2.8% (SMD −0.028, 95% CI −0.07 to 0.02). The effect is based on very low-certainty evidence.

Very low a
Mean expenditure on taxed sugar-added foods Assessed with percentage change Follow-up: 16 months There was an effect of the intervention on the mean expenditure of taxed sugar-added foods. Data show that the mean expenditure decreased after the implementation of the intervention slightly by 0.6% (SMD −0.006, 95% CI −0.03 to 0.02). The effect is based on very low-certainty evidence.

Very low a
Mean expenditure on untaxed sugar-added foods Assessed with percentage change Follow-up: 16 months The mean expenditure on untaxed sugar-added foods increased after the implementation of the intervention by 3.0% (SMD 0.03, 95% CI −0.01 to 0.07). The effect is based on very low-certainty evidence.

Very low a
Difference in mean expenditure on taxed sugar-added foods compared with untaxed sugar-added foods

Assessed with percentage change
The mean expenditure on taxed sugar-added foods differed from the mean expenditure on untaxed sugar-added foods after the implementation of the intervention by 3.7% (SMD −0.037, 95% CI −0.08 to 0.01). The effect is based on very low-certainty evidence.

Description of the condition Epidemiological background
Preventive action comprising both policies and interventions are urgently needed to curb the obesity and overweight epidemics and their detrimental health impacts (WHO 2000). The World Health Organization (WHO) Commission on the Social Determinants of Health called for intersectoral action to address the social determinants of health to improve population health and health equity, including fiscal interventions such as taxes (CSDH 2008

Social inequalities
As the WHO Commission on Social Determinants of Health and similar subsequent reports noted (CSDH 2008;Marmot 2012), the unequal distribution of overweight and obesity within and between countries also poses a serious challenge for achieving health equity nationally and globally. Within a country, overweight and obesity are usually (but not always) distributed along a social gradient. Regarding socioeconomic status, for example, higher prevalences of overweight and obesity are generally observed in people with a lower socioeconomic status. However, in some low-income countries, such as Cameroon and many Pacific Island countries and territories, people with a higher socioeconomic status are relatively more likely to be overweight or obese. In some low-and middle-income countries (e.g. China), the relationship of socioeconomic status with overweight and obesity, respectively, is unclear (Dinsa 2012;McLaren 2007;Ogden 2015;Wang 2012). Furthermore, it is possible that disadvantaged population groups may be at greater risk of obesity-related harms, even without experiencing greater exposure levels (Diderichsen 2019).

Nutritional transitions
Across the globe, major dietary shi s are occurring, resulting in nutritional transitions. Nutritional transitions -reflecting changes in diet, physical activity and health -are major contributors to overweight and obesity becoming increasingly prevalent globally. In the last four decades, the daily food consumption rose by a global average of about 400 calories. However, the main sources of calorie intake greatly di er between low-and middle-income countries and high-income countries.

Definition of 'unprocessed sugar' and 'sugar-added foods'
Consumption of unprocessed sugars and sugar-added foods contribute substantially to overweight and obesity (WHO 2018). We define 'unprocessed sugar' for the purpose of this review -on the basis of the definitions of 'sugars' and 'free sugars' given belowas monosaccharides (such as glucose, fructose, and galactose), disaccharides (such as lactose, maltose, and sucrose) and higher saccharides (such as cellulose).
Traditionally, the term 'sugars' describes mono-and disaccharides (FAO/WHO 1998). Monosaccharides include fructose, galactose, and glucose. Disaccharides include lactose, maltose, sucrose, and trehalose. Some sweeteners, such as corn syrups, also consist of higher saccharides. In 2002, the Joint WHO and Food and Agriculture Organization of the United Nations (FAO) Expert Consultation introduced the term 'free sugars' (Amine 2002). In 2015, the definition of the term was elaborated for the WHO guideline on sugar intake for adults and children. 'Free sugars' are defined as mono-and disaccharides (such as lactose, maltose, and sucrose) that are added to foods (WHO 2015a).
We define 'sugar-added foods' for the purpose of this review -on the basis of the following definitions -as non-liquid food products (i.e. this review does not include drinks, including SSBs) that contain artificially-added sugar in various quantities, where sugar refers to monosaccharides, disaccharides, and higher saccharides (as defined above).
Based on the definition of the US Department of Agriculture (USDA), added sugars are either pure sugars or natural products with high sugar content (e.g. honey) that are added to food during processing or preparation. In the preparation of a food product, sugars can be processed in any way, e.g. baked or cooked. Added sugar mainly appears in cakes, cookies, desserts, pies, and candy. "Specifically, added sugars include white sugar, brown sugar, raw sugar, corn syrup, corn-syrup solids, high-fructose corn syrup, maple syrup, pancake syrup, fructose sweetener, liquid fructose, honey, molasses, anhydrous dextrose, and crystal dextrose. Added sugars do not include naturally occurring sugars such as lactose in milk or fructose in fruits" (USDA/HHS 2000).

E ects of sugar consumption on health, society and economy
Overweight and obesity are risk factors for several diseases.
Overweight and obesity are defined as an excess of adipose tissue in one's body caused by an imbalance of energy intake and energy expenditure resulting from diverse genetic, environmental, cultural, behavioral, social and/or economic factors (Kopelman 2007;WHO 2015b). Increased energy intake is the result of overconsumption of foods and especially consumption of surplus quantities of high-caloric foods. Unprocessed sugar and sugaradded foods are a main source of excessive calorie intake (Bowman 2004;Popkin 2003). Thus, a sugar-rich diet, especially when combined with physical inactivity, may cause overweight and obesity, which, in turn, increases the risk of high blood pressure (e.g. hypertension), dyslipidaemia, peripheral insulin resistance, inflammation, and dental caries (Kopelman 2007; Moynihan 2014; WHO 2015b). These adverse e ects of overweight and obesity may lead to substantial health loss across many bodily systems, including disorders of the cardiovascular (e.g. ischaemic heart disease), gastrointestinal (e.g. bowel cancer), musculoskeletal (e.g. osteoarthritis), endocrine (e.g. type 2 diabetes mellitus), and respiratory (e.g. obstructive sleep apnoea) systems (Aronne 2002). In addition to its contribution to specific diseases, obesity may also reduce psychological well-being at the individual level and adversely a ect societies and economies at the population level by, for example, reducing economic productivity and increasing demands on healthcare resources (Colditz 1999;Wardle 2005).
Overweight and obesity in childhood and adolescence are associated with increased risks of overweight and obesity in adulthood (Power 1997). Thus, early development of overweight and obesity has substantial and long-lasting consequences for a person's physical and mental health (Must 1999;WHO 2016).
Overweight and obesity are the most o en cited e ects of a sugarrich diet. However, the e ects of a sugar-rich diet are far-reaching. For instance, in the USA, dental caries is one of the most prominent childhood diseases with a minimum of one filing or caries lesion among 77.1% of children aged 0 to 17 years (Touger-Decker 2003). Worldwide, one in 10 people is a ected by diabetes (Basu 2013; James 2018).
Overweight and obesity incur direct costs (e.g. disease-related preventive, treatment and diagnosis service costs) and indirect costs (e.g. disease-related costs of lost productivity), both in the health sector and in other sectors, including labour and economic development (Van Nuys 2014;Wolf 1998

Food-related fiscal policies
Food-related fiscal policies generally aim to either lower prices (e.g. subsidisation) or increase prices (e.g. taxation) for specific food groups. We evaluated the e ects of taxes on unprocessed sugar and sugar-added foods (as defined above). The Organization for Economic Co-operation and Development (OECD) defines taxes as "compulsory unrequited payments to general government" (OECD 2014).

Typologies of taxes
There are two di erent types of indirect taxes with subcategories on sugar-related products as shown in Figure 1: (1) import (or export) taxes (or fees) on unprocessed sugar or sugar-added foods, and (2) within-border (local, regional, national, and international) taxes (Fletcher 2010;Meessen 2007;Mytton 2012). From the perspective of the WHO Commission on Social Determinants of Health and its recommendations for actions (CSDH 2008), food-related fiscal policies can be classified as an intersectoral socioeconomic intervention on the social determinants of health to improve health equity (Pega 2017a).

Description of types of food-related taxes
Indirect taxes are paid by the consumer, collected by the seller or intermediary, and forwarded to government. Sales taxes -as one form of indirect taxes -are paid by the consumer at the moment of purchase of the taxed goods and services. Sales taxes are common tax interventions to reduce the consumption of a specific good, such as unprocessed sugar or sugar-added foods (Brownell 2009). Value Added Tax (VAT) is the most popular tax across the globe and the principal type of indirect taxes. The term 'VAT' is used as a synonym for 'goods and services tax'. The underlying principle of the VAT system includes "the application to goods and services of a general tax on consumption exactly proportional to the price of the goods and services" (Schenk 2015). VAT is more commonly applied to di erent food categories than are targeted food taxes (Mytton 2007). The level of a sales tax may di er according to the type of tobe-taxed product and service. Sales taxes and VAT are added to the price of an item without consideration of the item's volume. Thus, goods of a larger size, that in most cases are comparably cheaper than the same goods of smaller sizes, result in a lower impact of the tax in goods with larger package sizes. An excise tax is an inland tax on the (production for) sale and the goods produced for sale. Custom duties (or 'border taxes') are taxes applied to imported products. The Cook Islands and Fiji, for example, implemented such custom duties on SSBs to increase the cost of these drinks as a means to fight the obesity epidemic (Snowdon 2013). Governments similarly also use import sales taxes, these being taxes on goods imported from countries that are not a contracting party of the importing country (Cnossen 1993). All taxes may encourage a reformulation of the taxed item to lower its price and thus, decrease the content of the taxed ingredient in a processed food product. For an overview on the di erent tax typologies, as described above, see Figure 1.

Aims and rationales of food-related taxes
Fiscal policies such as excise taxes on food have been proposed, developed and implemented, generally with the goal of curbing overweight and obesity, but sometimes also to increase governmental revenue (Kim 2006). Taxes raise revenue for government, and these revenues may or may not be earmarked (also referred to as hypothecated) for public health programmes. These types of food taxation policies include taxes on salt, fats, SSBs, and unprocessed sugar or sugar-added foods (other than beverages) more generally.
The underlying policy and economic rationale for implementing food taxation policies, including those on unprocessed sugar and sugar-added foods, is a government's motivation to create or increase a financial charge for a specific (unhealthy) food in order to increase consumer prices and usually also to raise public revenue. This price increase may then lead to a decrease in demand, which in turn may reduce the intake of the taxed (unhealthy) food product by reducing its consumption in the population (Ecorys 2014).
The implementation of food taxes may lead to changes in food composition, in an e ort to minimise taxes paid. In other words, in response to the implementation of a tax on unprocessed sugar or sugar-added foods, food industries may reformulate their products (Brownell 2009). This may lead to products with lower added sugar content, with potential benefits to human health. However, on the other hand, this reformulation of the product may make it even unhealthier, e.g. by adding other ingredients, such as fat, with potential detrimental health e ects.

Focus of this review
This review will focus specifically on the taxation of unprocessed sugar and foods that contain added sugar (e.g. sweets, ice cream, confectionery, and bakery products) regardless of the taxation level. In tandem, we are also conducting systematic reviews of the e ectiveness of taxes of fats (Lhachimi 2016), and SSBs (Heise 2016), for improving human health.

How the intervention might work
See Figure 2 for a logic model describing the causal pathways through which taxation of unprocessed sugar and sugar-added foods may work to reduce overweight, obesity and other health outcomes.

Figure 2. Study's initial logic model with causal pathways
The typical aim of prevention and treatment of overweight and obesity is weight reduction. This can be achieved by decreasing energy intake through changes in dietary behaviours (e.g. reduce consumption of foods high in added sugar and fats), drug treatment, a surgical intervention, or increased energy expenditure through increased physical activity, or a combination of some or all of these (Wadden 2002). Taxation of food might be an e ective mechanism in reducing overweight and obesity prevalence.
In general, food taxes are o en hypothesised to lead to reduced consumption of unhealthy foods (Mytton 2012). However, the decrease from food taxes in the percentage share of unprocessed sugar and sugar-added food consumption in the total energy intake is likely to have one of two e ects on health-related behaviour: either it may lead to a reduction in total daily energy intake, or the consumption of unprocessed sugar and sugar-added foods may be substituted by foods that are also relatively high in calories (e.g. high fat content) or by other unhealthy products, such as cigarettes and salts (Briggs 2013). While the former may lead to weight reduction, the latter may result in (1) weight gain, (2) a zero e ect, or (3) weight reduction (Ecorys 2014). In any case, any e ects of food taxation on public health and consumption patterns either take some time to become detectable (Fletcher 2010;Meessen 2007) or only show short-term e ects (Wansink 2014).
According to economic theory, the taxation of unprocessed sugar or sugar-added foods is expected to cause an increase in price, which in turn will lead to a decrease in demand, sales, and consumption (Mytton 2012). Across cultures, a higher product price is also associated with a higher product reputation and quality (Dawar 1994). Thus, as a response to a tax implementation, the consumption of taxed items may rise. With regard to withincountry inequalities, as the price of a product determines the level of a ordability, low-income groups are usually more strongly a ected by taxation policies than higher-income groups (Eyles 2012; Maniadakis 2013). If low-income populations have higher prevalences of overweight, obesity, type 2 diabetes, dental caries and other sugar-related diseases and conditions than middleand high-income populations, then unprocessed sugar or sugaradded foods taxation policies may disproportionately reduce consumption of unprocessed sugar and sugar-added foods among the low-income population, and thus improve health equity in the population. Furthermore, with regard to between-country inequalities, these tax interventions may reduce overweight, obesity, type 2 diabetes, dental caries and other sugar-related diseases and conditions di erently across countries of di erent income levels. For example, it is theoretically plausible that such taxes are more e ective in reducing sugar-related diseases and conditions in low-income countries than in middle-and highincome countries. Thus, taxes on unprocessed sugar and sugaradded foods have the potential to also improve between-country health equity (Eyles 2012; Lorenc 2012; Maniadakis 2013).
In several countries, food taxes were implemented in the past and some were subsequently abolished. How the taxation of unprocessed sugar and sugar-added foods might a ect outcomes is described in a logic model with causal pathways ( Figure 2). The taxation of unprocessed sugar and sugar-added foods -introduced by local, regional, national, or multinational governments -is hypothesised to result in price changes (e.g. increased prices of chocolate, ice cream, and bakery products; Epstein 2012; Jensen 2013; Maniadakis 2013), which in turn may lead to altered expenditure patterns for food. Financial resources -also dependent on expenditure on food -and contextual and individual factors (e.g. income), determine the demand for food products. These market components impact consumer purchases and consumption choices for di erent food categories, including unprocessed sugar and sugar-added foods (Briggs 2013;Sharma 2014). This may result in a lower intake of the taxed food products (unprocessed sugar and sugar-added foods) and in a substitution of these by other (food) products (Fowler 2015; Yang 2010). As a consequence, food tax-induced changes in consumption patterns should result directly in changes to intake of unprocessed sugar and sugar-added foods (Epstein 2012; Maniadakis 2013). A decrease in the intake of unprocessed sugar and sugar-added foodsas one hypothesised consequence of taxing these foods -can reduce overweight, obesity, and other health outcomes, both directly and indirectly. To exemplify the direct path from the intake of unprocessed sugar and sugar-added foods to other health outcomes, a decrease in the intake of unprocessed sugar and sugaradded foods has the potential to reduce the risk of dental caries (Moynihan 2014; WHO 2015a). The indirect path from the intake of unprocessed sugar and sugar-added foods to overweight and obesity operates through energy intake. For example, a decreased energy intake as a consequence of decreased intake of unprocessed sugar and sugar-added foods lowers the risk of being overweight and obese, respectively (Kim 2006;Malik 2013). Moreover, food tax-induced changes in consumption patterns may directly result in changes in nutrient intake (Epstein 2012; Maniadakis 2013). The direct path from intake of other nutrients (e.g. fat or dietary minerals) as a consequence of substitution e ects has the potential to directly increase, decrease or not a ect the risk of other health outcomes (e.g. fatty liver). The indirect path from intake of other nutrients to overweight, obesity and other health outcomes goes through energy intake. To illustrate this, a higher intake of other nutrients (e.g. saturated or unsaturated fat) as a substitution e ect of decreased intake of unprocessed sugar and sugar-added foods a ects energy intake (increase, decrease or zero e ect) and is therefore associated with the risk of overweight, obesity and other health outcomes (Marriott 2010). Decreased risks of overweight and obesity, in turn, can reduce the risk of developing other dietrelated diseases and conditions (e.g. chronic diseases such as type 2 diabetes, cardiovascular diseases, and dental caries; Guh 2009).
Contextual and individual factors (see Figure 2) influence the process from the input to the outcomes, alter e ect sizes and help us to understand the causal relationships (Qi 2012). Alternative interventions may be possible comparators but also potential co-interventions (i.e. complementary interventions to reduce the consumption of unprocessed sugar and sugar-added foods, such as bans on marketing, which are designed to enhance intervention e ectiveness). Therefore, the e ect of taxation may be modified by other interventions by governments, communities and the food or other (e.g. agricultural) industry to reduce consumption of unprocessed sugar and sugar-added foods (

Why it is important to do this review
There is increasing public health interest in taxing unprocessed sugar and sugar-added foods as an intervention, sometimes spurred by the recent implementation of food taxes in several countries, such as Hungary and Norway. However, the implementation of a tax on unprocessed sugar and sugar-added foods is only one of many policy options for reducing consumption of these foods (Hawkes 2015). WHO recommends a daily sugar consumption of less than 10% of the total energy intake. Thus, the recommended maximum level in adults is approximately 50 g. Keeping the daily sugar intake on a level below 5% (approximately 25 g) of the recommended total energy intake might have even greater health benefits (WHO 2015a). In view of the excess consumption of sugar and the worldwide increase in overweight and obesity prevalence, governments must urgently act. Taxes for unprocessed sugar and sugar-added foods are interventions that may help to fulfil the policy aim of reducing the prevalence of overweight and obesity and the associated burden of disease, and the associated costs to the health and other sectors. . This is the first systematic review to investigate the e ects of taxes of unprocessed sugar and non-liquid sugar-added foods. Evidence is required regarding the e ectiveness of taxing unprocessed sugar and sugar-added foods so that policy makers can make evidencebased decisions.

Consumption of unprocessed sugar and sugar-added foods
This research is part of a series of three systematic reviews of di erent types of food taxation carried out by the same author group using a similar methodological approach. For reasons of comparability, the methodological content is similar across the three reviews. These reviews focus on the e ects of governmental taxation to increase the prices of: (1) unprocessed sugar or sugaradded foods (this review), (2) processed or packaged food with high content of saturated fat (Lhachimi 2016), and (3) SSBs (Heise 2016).

O B J E C T I V E S
To assess the e ects of taxation of unprocessed sugar or sugaradded foods in the general population on the: • consumption of unprocessed sugar or sugar-added foods; • prevalence and incidence of overweight and obesity; and • prevalence and incidence of other diet-related health outcomes.

Types of studies
Our pre-published review protocol guided this review (New Reference). We included various study designs and adopted an approach previously used in at least two other Cochrane Reviews in order to summarise 'best available evidence' (Gruen 2004;Turley 2013). This approach clearly separates studies into two broad categories: (1) studies meeting rigorous Cochrane E ective Practice and Organisation of Care (EPOC) criteria (EPOC 2012;EPOC 2015), and (2) supporting studies -those not meeting EPOC criteria with greater risk of bias as well as lower external generalisability.
First, for the synthesis of main results, in line with EPOC criteria we included: • randomised controlled trials (RCTs); • cluster-randomised controlled trials (cRCTs); • non-randomised controlled trials (nRCTs); • controlled before and a er (CBA) studies; and • interrupted time series (ITS) studies.
As recommended by EPOC, we included controlled studies with more than one intervention or control site and ITS studies with a clearly defined intervention time and at least three data points before and three a er the intervention (EPOC 2012).
There were no restrictions by publication date and language, but we only included studies focusing on humans (CPH 2011). We had no restriction on study duration and participants. Closed field experiments suggest that consumer behaviour adaptations, expressed in terms of sales of unprocessed sugar or sugar-added foods, become apparent within a short time frame, such as one month (Block 2010). Implementation of taxes on sugar or sugaradded foods at a national level might feature a longer time lag between intervention and outcomes, especially for health outcomes. However, in one study the e icacy of food taxes with respect to purchases was apparent a er one year (Popkin 2016).
In general, field experiments on food taxes recruit small numbers of participants. Nevertheless, they were considered as a valuable source to identify important outcome pathways and e ects on food patterns relevant to the taxation of unprocessed sugar or sugaradded foods (Epstein 2012).
We excluded simulation studies due to their potential limitations from their underpinning assumptions (e.g. lack of potential supplyside changes, static models to predict weight loss), and other methodological restrictions (e.g. the use of a combination of heterogeneous data sources; Lin 2011; Shemilt 2015).

Supporting studies
We considered as supporting studies: • studies that use an RCT, cRCT, nRCT, CBA or ITS design but do not fulfil the EPOC criteria (hence, are not included in the main results as outlined above); • prospective cohort studies; • retrospective/non-concurrent cohort studies; • repeated cross-sectional studies; and • uncontrolled before-a er (UBA) studies.
However, we did not find any eligible supporting studies.
In future updates, we will stick to our initially planned methodology on supporting studies. We originally planned not to include 'supporting studies' in the statistical synthesis of the primary included studies (i.e. alongside those meeting EPOC criteria), but aimed to narratively synthesise them in addition to the main findings. We planned to extract the same type of data from supporting studies as we did for the included studies and planned to document these in a separate 'Characteristics of supporting studies' table. We planned to carry out 'Risk of bias' assessments on these studies and to undertake quality assessments using the GRADE approach, then to present the findings from these supporting studies separately, as supplemental information in the results section and in a separate 'Summary of findings' table. We planned to make observations as to similarities and di erences of findings between the included studies and the supporting studies in the 'Discussion' section, to help summarise the breadth, quality and findings of the totality of research on the e ects of these interventions.

Cochrane Database of Systematic Reviews
In future updates, supporting studies may either support or challenge results in the main findings and highlight uncertainty and potential research gaps. We will consider known limitations of UBA studies, cohort studies, and repeated cross-sectional studies, especially confounding or lack of control for underlying time trends, when we assess these studies' eligibility for inclusion. If UBA studies, cohort studies, and repeated cross-sectional studies are likely to be biased and do not use analytic strategies (e.g. stratification) or other designs (e.g. regression discontinuity (Craig 2017), fixed e ects regression (Gunasekara 2014) or marginal structural models (Pega 2016)), to control for confounders and time trends, we will exclude these studies from the 'supporting studies' analysis.

Types of participants
We included studies of children (0 to 17 years) and adults (18 years and over) from any country and setting.
We excluded studies investigating the e ects of taxing unprocessed sugar or sugar-added foods focusing on specific subgroups, particularly: • people receiving a pharmaceutical intervention; • people undergoing a surgical intervention; • pregnant women; • elite athletes; • people with any disease who are overweight or obese as a sidee ect of the disease or of a clinical treatment they receive for the disease, such as those with thyroiditis and depression; and • people with any chronic illness(es); at baseline and at the post-intervention phase due to higher or lower health risks compared to the general population. The rationale is that tax policies may a ect these subgroups di erently from the general population since di erent causal mechanisms may be operating.

Types of interventions
This review included studies of the taxation of unprocessed sugar or sugar-added foods, defined as: • a tax of goods; • enacted by and/or paid to local, regional, or national governments or international organisations; • of any value or level of taxation; • added to sales prices of foods with unprocessed sugar or sugaradded foods, or both (as defined above), and • implemented for any duration.
Interventions were defined as public policies (i.e. in the form of a tax) of local, regional, national, and multinational governments or field experiments that imitate taxation e ects for research purposes in clearly defined environments (e.g. cafeterias, supermarkets and vending machines). A tax was eligible for inclusion if it operated or was payable, or both, at the local, regional, national or international level. We included any comparator intervention (e.g. no intervention, educational interventions, bans, media campaigns, and subsidies on healthy food). We also included studies that compared an eligible tax with another eligible tax that is of a lower value. We have pursued the same strategy in previous Cochrane Reviews on other financial interventions (Pega 2013; Pega 2015; Pega 2017b). We excluded virtual and hypothetical interventions imitating a taxation on unprocessed sugar or sugar-added foods if participants' purchase decisions are not binding so that they do not all result in a real purchase or if the money is virtual or not belonging to the study participant.

Types of outcome measures
Our outcome selection and grouping was guided by preliminary evidence already discussed in the Background and on the basis of the logic model (Figure 2), and incorporated feedback and recommendations from the review advisory board members (email and online survey; Table 2). All pre-selected outcomes achieved 'critical' or 'important' ratings on average, following the GRADE approach. For primary outcomes we favoured outcomes of critical importance in line with our review scope and Objectives (Table 3). Detailed information on advisory group involvement is provided in the section Searching other resources under the subheading 'Advisory group'. Primary outcomes include intermediate outcomes directly a ected by tax-induced changes in prices for unprocessed sugar or sugar-added foods. As a result, consumption of unprocessed sugar or sugar-added foods may directly alter the primary health outcomes included in this review, including overweight and obesity. Secondary outcomes focused on food patterns (substitution and diet), expenditure, and other prioritised health outcomes directly or indirectly influenced by the taxation of unprocessed sugar or sugar-added foods. We included demand as a proxy for the consumption of unprocessed sugar or sugar-added foods.

Primary outcomes
We included changes from baseline (pre-intervention) to postintervention of the following primary outcomes.

Consumption of unprocessed sugar or sugar-added foods
• consumption of unprocessed sugar or sugar-added foods (e.g. frequency, amount)

Energy intake
• energy intake from unprocessed sugar or sugar-added foods only • total energy intake Overweight and obesity • incidence of overweight • incidence of obesity • prevalence of overweight • prevalence of obesity Of these outcomes eligible for inclusion in this review, we found evidence on, and were therefore able to include, the outcome of consumption of sugar-added foods (purchased quantities) as a primary outcome.

Secondary outcomes
We considered changes from baseline to post-intervention of the following secondary outcomes.

Substitution and diet
• composition of diet (expressed as food groups or ingredients, e.g. any consumption of any items in the food groups of fats, sugars, salts, and alternative low-caloric sweeteners), including the consumption of untaxed sugar and sugar-added foods • di erence in mean consumption of taxed sugar-added foods compared with untaxed sugar-added foods

Expenditure
• total expenditure on food • total expenditure on unprocessed sugar or sugar-added foods • expenditure on untaxed sugar-added foods • di erence in mean expenditure on taxed sugar-added foods compared with untaxed sugar-added foods

Demand
• total sales of unprocessed sugar or sugar-added foods

Other health outcomes
• health-related quality of life (e.g. Short Form 36 (SF-36) and Health-Related Quality of Life (HRQOL-14)) • mortality • any other health outcomes (e.g. dental caries, type 2 diabetes, cardiovascular diseases, etc.) Of the considered secondary outcomes, we only found evidence on, and were therefore able to include in this review, the outcome of expenditure on sugar-added foods as a secondary outcome.

Search methods for identification of studies Electronic searches
We searched the following 12 databases: • We applied a search strategy with additional keywords for possible comparators (e.g. 'subsidy') and we did not use filters for study types, in order to maximise the sensitivity of the literature search (Lefebvre 2011, chapter 6.4.4). We used the strategy presented in Appendix 1 to search MEDLINE. We modified this strategy as presented in Appendix 2 to search other electronic databases for records written in any language and published since and until the dates mentioned above. We did not search African Index Medicus (AIM) -a valuable resource for low-and middle-income country literature -in our review, as a sensitive pre-search with intervention keywords (e.g. tax, taxation, etc.) resulted in zero hits.
We performed one initial search and four search updates in electronic databases. •

Grey literature databases
We searched the following six grey literature databases with search strategies as presented in Appendix 3. We performed an initial search in all grey literature databases starting at 27 April 2016 and applied the same search time frames for updates as described for the electronic databases.
We searched the following two databases for completed or ongoing studies with keywords relevant to the intervention (e.g. taxation, pricing): Cochrane Database of Systematic Reviews

Internet search engines
We screened the first 30 hits in Google Scholar via Google on 11 August 2016 and 14 October 2019. The search strategy is presented in Appendix 4.

Targeted internet searching of key organisational and institutional websites
We searched websites of major organisations and institutions in the initial search in 2016 and on 11 October 2019, specifically: • World Obesity Federation (www.worldobesity.org); • OECD (

Searching other resources
We handsearched the reference lists of all included studies.

Advisory group
We established a review advisory group of experts in the field of food taxation and health to comment and provide advice and suggestions to improve the systematic review and its manuscript at the protocol stage. Following the GRADE approach, the advisory group members participated in an online survey and ranked preselected outcomes according to their relative importance on a 9point Likert scale (categories 1 to 3: of limited importance; 4 to 6: important; 7 to 9: critical; GRADE 2013). The review advisory group consisted of policy makers, researchers and academics.
We provided the members of the review advisory group with detailed background information on this review. At the protocol stage, the review advisory group members were asked to provide feedback specifically on the focus and relevance of this review's research question, selected outcomes, study design, search strategy, database selection, and ongoing or unpublished studies (Green 2011, chapter 2. 3.4.3). In the review stage, prior to final submission, we contacted review advisory board members for relevant ongoing and unpublished studies. We received feedback via email and the online survey. All members of the advisory group and results from the online survey are found in Table 2 and Table 3.

Selection of studies
An information specialist (CF) and an additional author (TLH) conducted the electronic database searches, searches within grey literature databases and internet search engines. One review author (MP) handsearched the reference lists of included studies. We performed targeted internet searching of key organisational and institutional websites, using a standardised template to document the search (MP, THL, SKL, UG, GG, FP, IS, SVK).
We conducted screening in six stages. If a reference, an abstract or a full-text report was in a language other than English, German or French, we translated it using internet-based translation tools or by asking native speakers. First, at least two review authors (MP, TLH, SKL, UG, GG, FP, IS or SVK prior to 2018; MP, TLH or SKL in 2018 and 2019) independently screened studies' titles and abstracts (when available). MP, THL and SKL provided a detailed screening guideline for all review authors and used Covidence for screening titles and abstracts (MP, TLH, SKL, UG, GG, FP, IS, SVK). If an abstract was not provided by the database it originated from, and the title appeared to be potentially relevant, we progressed the record to full-text review within Covidence. Second, we resolved disagreement by consensus and in consultation with a third review author (SKL, TLH or MP) and eliminated all records that did not fit the inclusion criteria (see Criteria for considering studies for this review). Third, we retrieved full texts of potentially relevant studies for assessment. Fourth, two review authors (MP and TLH) independently screened the full texts. Fi h, both review authors created a list of studies that they considered to fulfil the inclusion criteria. Sixth, the review authors compared their lists and in cases of disagreement, a third review author (SKL) was decisive. Based on these six steps, we included studies in the review. At each stage, we recorded the records retrieved and excluded. For key records of which we screened the full texts, we recorded reasons for exclusion. We present a PRISMA flowchart in Figure   We stored all records obtained by the electronic searches in a reference management so ware (Endnote 2012). We recorded and managed the results of the abstract and full-text screening. MP, SVK, MHB, TLH and SKL further discussed the reasons for exclusion at full-text assessment and we stored results in an Excel spreadsheet. We used a modified data extraction and assessment template from Cochrane Public Health (CPH; CPH 2011), for the complex intervention addressed in this review. We extracted e ect estimates for study populations based on PROGRESS categories (place of residence, race/ethnicity/culture/language, occupation, gender/ sex, religion, education, socioeconomic status, social capital) to evaluate impacts on equity. We Cochrane Database of Systematic Reviews process, MP, TLH, SVK, UG, FP, and SKL piloted and adapted the data extraction form to ensure standardised extraction (Higgins 2011a, chapter 7.6.3). MP, SVK, and FP (prior to 2018), and MHB (in 2019) extracted general information (publication type, country of study, funding source of study, potential conflict of interest), study eligibility (type of study, participants, type of intervention, duration of intervention, and type of outcome measures), study details (study aim, methods, results, intervention group, confounders, and confounder-adjusted and unadjusted outcomes), indicators of changes in food prices (price of unprocessed sugar or sugaradded foods, price of other food categories), and other relevant information. We also extracted contextual factors (e.g. political system, co-interventions, reason for implementation, reason for particular tax level, intended beneficiaries, implementation costs, country and region-specific level of gross domestic product (GDP), food security (availability, access, and use), and process evaluation criteria (e.g. satisfaction of participants, adherence) that facilitate or hinder the implementation of taxation on unprocessed sugar or sugar-added foods (Anderson 2011a; Campbell 2018). If studies did not provide information on these criteria but referred to another study, we extracted information from these other sources. In the Characteristics of included studies we described methods, participants, interventions, outcomes and further notes. We did not extract qualitative data.
MP entered, stored and managed extracted data in Review Manager 5 and MHB double-checked the data entered (Review Manager 2014).

Assessment of risk of bias in included studies
Two review authors (MP, FP and SVK prior to 2018; MP and MHB in 2018 and 2019) independently evaluated the risk of bias of every included study. In case of disagreement, they discussed discrepancies with a third review author (TLH or SKL) and resolved them by consensus. Based on the template provided by CPH, we assessed the risk of bias using the criteria for judging risk of bias in Cochrane's 'Risk of bias' assessment tool (Higgins 2011), and Cochrane EPOC's guidance (EPOC 2015). Both tools examine the following biases: selection, performance, detection, attrition, reporting, and others. The EPOC 'Risk of bias' tool for ITS examines three further risks of bias: was the intervention independent of other changes, was the shape of the intervention e ect pre-specified and was the intervention unlikely to a ect data collection? For studies included in the main quantitative evidence synthesis (i.e. RCTs, cRCTs, nRCTs, CBA and ITS studies), we planned to assess the risk of bias using the 'Risk of bias' criteria for EPOC reviews, based on the Cochrane tool for assessing risk of bias (Higgins 2011b, Table 8.5.a).
We planned to assess the study quality and risk of bias of 'supporting studies' (i.e. studies that do not meet EPOC criteria, cohort studies, repeated cross-sectional studies, UBA studies) with the Quality Assessment Tool for Quantitative Studies, developed by the E ective Public Health Practice Project (EPHPP) (EPHPP 2007).
To judge the risk of bias according to Cochrane's 'Risk of bias' assessment tool, we used the following three ratings: 'low', 'high', and 'unclear' (adequate information is unavailable or there is uncertainty about the risk of bias; Higgins 2011b, chapter 8.6). For studies not meeting EPOC criteria, we planned to judge the risk of bias according to the Quality Assessment Tool for Quantitative Studies, using following three categories: 'strong', 'moderate', and 'weak' (EPHPP 2007). We provide 'Risk of bias' tables for all included studies in the Results section.

Measures of treatment e ect
In the data synthesis, we aimed to quantitatively pool the results of di erent studies using meta-analysis. However, since we were not able to perform meta-analyses, we have not presented a pooled e ect estimate in the systematic review.
The included studies reported treatment e ects of a tax on sugar-added foods for consumption (purchased quantities) and expenditure, with the outcomes measured using continuous data, and the treatment e ect measures being a standardised mean di erence (SMD) with standard error (SE). We calculated the standard deviations (SDs): SD = SE* √n where SE= standard error, and n = number of household-level observations. Then, we calculated the 95% confidence interval (CI) of the e ect estimate: μ = M ± T(s M ) where M = sample mean, T = T statistic determined by confidence level (here: 95%) and s M = standard error = √(s 2 /n).
We did not find evidence on the e ects of the treatment (i.e. tax intervention) on any dichotomous outcomes, and thus we have not reported any relative or absolute measures of treatment e ect for dichotomous outcomes (e.g. odds ratios (ORs), risk ratios (RRs) or risk di erences (RDs)).

Unit of analysis issues
As per Cochrane guidelines, we planned to collect data on allocation and to analyse the level at which allocation occurred for the same outcome (Deeks 2011, chapter 9.3.1). We considered data from cross-over trials (e.g. by incorporating the study data similar to a parallel-group trial) and studies with multiple observations (e.g. by defining di erent periods of follow-up) for potential analyses (Deeks 2011, chapter 9.3.4;Higgins 2011c, chapter 16.4.5).
We planned to request individual-level data from the corresponding author of the study if control for clustering was missing or insu icient and if individual-level data were not presented in the study. We planned to reduce the size of each trial to its 'e ective sample size' in order to correct intervention e ects in cluster-randomised trials. The e ective sample size of an intervention group is the original sample size divided by the 'design e ect'. We planned to calculate the design e ect with the formula 1 + (M -1) ICC, where M is the average cluster size and ICC is the intracluster correlation coe icient (Higgins 2011c, chapter 16.3.4).
For dichotomous data, we planned to divide the total number of participants and the number of participants who experienced the event by the same design e ect. For continuous data, we planned to reduce only the sample size, with means and standard deviations to remain unchanged (Higgins 2011c, chapter 16.3.4).
We included only one study, and thus, it was not feasible to perform analyses on the level at which allocation occurred (e.g. for multiple interventions).

Dealing with missing data
We planned to request all missing information and data from principal study authors via email. In the study from Biró 2015, data are based on a household-level survey not conducted by the principal study author. The number of distinct households and participants was not available in the study from Biró 2015. Therefore, and for further methodological issues, we contacted the study author via email. We received responses via email, but the number of distinct households and participants could not be clarified and thus, in agreement with the study author, we refer to household-level observations.

Assessment of heterogeneity
We were not able to perform meta-analyses for the outcomes because only one study was included in this review. There were not enough studies included to assess heterogeneity across studies regarding potential sources of heterogeneity, such as study population, intervention area/setting, intervention characteristics (tax definition, basis for calculating taxation, level of taxation), implementation level, comparisons, co-interventions, and outcomes.

Assessment of reporting biases
Reporting bias, including publication bias, time lag bias, multiple (duplicate) publication bias, location bias, citation bias, language bias, and outcome reporting bias occurs when the dissemination of research results depends on their magnitude or direction, or both (Sterne 2011). To assess the presence of reporting bias, we planned to produce funnel plots if we found 10 or more studies of the same outcome. Had we found 10 or more studies to include, then we would have tested for asymmetry in funnel plots (small study e ects) by investigating whether the relationship between a measure of study size and the estimated intervention e ect is asymmetrical (Sterne 2011). However, since the review included fewer than 10 studies of the same outcome, we did not assess reporting bias.

Data synthesis
As already described, we could not perform meta-analyses with one included study. We narratively summarised the study results. We structured the summary by the outcome categories of this review. Within these categories, we planned to make further separation according to the intervention setting and the study design or study quality (Ryan 2016). However, this was not feasible due to the inclusion of only a single study. In addition to reporting findings as text and tables, we considered both harvest plots and e ect direction plots to summarise data not suitable for meta-analyses. Harvest plots are graphical summaries of data represented by multiple shaded or non-shaded bars with varying heights, and can be utilised to indicate e ect directions across included studies with non-standardised e ect estimates of outcomes (e.g. anthropometric measures). Similarly, e ect direction plots can be used to visualise information on e ect directions, with more focus on direct comparisons across studies (Ogilvie 2008; Thomson 2013). However, as we have only included one study in this review, we did not represent data by harvest plots or e ect direction plots.
For reports of multiple follow-ups for the same outcome (e.g. six months during the intervention, one year during the intervention, and six months a er the end of the intervention), we planned to prioritise the longest follow-up during the intervention (e.g. one year during the intervention, in the example given). However, in the included study, data of only one follow-up were available.
We planned to map results of the data synthesis against our initial logic model, to refine the theory of change and assess the credibility of the assumed causal pathways (Anderson 2011a; Thomson 2013). Due to limited results, we have illustrated the mapped and unmapped causes, e ects and outcomes within the included study in an adapted logic model with causal pathways in Figure 4.

Figure 4. Study's adapted logic model with causal pathways Subgroup analysis and investigation of heterogeneity
Due to the inclusion of only one study in this review, it was not feasible to conduct meta-analyses or produce harvest plots for primary outcomes with subgroups of interest.
We planned to run subgroup analyses for primary outcomes with regard to: • high-income countries versus middle-and low-income countries; • high-income groups versus middle-and low-income groups; • high-educated groups versus low-educated groups; • di erent tax values of unprocessed sugar or sugar-added foods; • single tax on unprocessed sugar or sugar-added foods versus multiple taxes on unprocessed sugar or sugar-added foods; • tax on unprocessed sugar or sugar-added foods alone versus tax on unprocessed sugar or sugar-added foods accompanied by other fat taxes or interventions (e.g. bans, minimum pricing, media campaigns, or subsidies on healthy foods); • di erent types of taxation: • indirect taxes levied within national borders (e.g. excise tax, sales tax, value added tax (VAT)); and • import (or export) taxes including custom duties and import sales taxes; • children versus adults; • BMI subgroups; • indigenous populations; • chronically ill people with overweight and obesity as sidee ects.
We planned to perform subgroup analyses with data on PROGRESS categories (e.g. age, gender, education, and ethnicity; Anderson 2011b).
Finally, we planned to investigate the statistical significance of di erences in the treatment e ect between subgroups using t-tests and Chi tests (Deeks 2011, chapter 9.6.2).

Sensitivity analysis
We planned to perform sensitivity analyses to determine the robustness of our results by conducting meta-analyses and harvest plots for the studies included in our review: • with respect to source of funding; • with studies considered as 'low risk of bias' compared to studies considered as 'high risk of bias'; • with published versus unpublished studies; • with respect to the intervention duration; • with respect to follow-up time; • with objective measures versus subjective measures; • with respect to study design; • with respect to cut-o points of the measures of treatment e ect;

Cochrane Database of Systematic Reviews
• with respect to imputation of data.
We planned not to include studies in sensitivity analyses if the studies had a high or unclear risk of bias with respect to incomplete outcome data or baseline di erences. For cRCTs with adequate data provided, we planned to perform intracluster correlation value sensitivity analysis. We planned to report findings of sensitivity analyses as a summary table (Deeks 2011, chapter 9.7).
However, since we did not perform a meta-analysis, none of the planned sensitivity analyses were feasible.

Summary of findings
We have provided two 'Summary of findings' tables, one for primary outcomes and one for secondary outcomes (Schünemann 2011, chapter 11.5). As suggested by an external referee, we reported the following pre-selected outcomes from included studies: consumption of unprocessed sugar and sugar-added foods (purchased quantities; primary outcome) and expenditure on unprocessed sugar and sugar-added foods (secondary outcome).

GRADE
For each outcome, two review authors (MP, MHB) assessed the certainty of the evidence for the domains 'risk of bias', 'inconsistency', 'indirectness', 'imprecision', and 'publication bias'. Summary of findings for the main comparison includes information on the primary outcomes; the summary of findings on reported secondary outcomes is included in Summary of findings 2. Both the tables include anticipated absolute e ects, the number of participants, the number of studies included, the certainty of evidence based on the GRADE guidelines, and additional comments. We used the computer so ware GRADEpro GDT to prepare the 'Summary of findings' table. As we included only one study, we have provided GRADE considerations narratively, in accordance with Cochrane MECIR standards for the inclusion of one single study (Higgins 2019). Altogether, we retrieved 23,281 records from the 12 electronic databases, 802 records through grey literature databases, 205 records through databases for completed or ongoing studies, 60 records through internet search engine searches, 38 records through organisational and institutional websites and 68 records through handsearching. A er removing duplicates of the records retrieved from di erent sources, a total of 18,767 records remained, for which we screened titles and abstracts, using the computer program Covidence. In the process of title and abstract screening, we excluded 18,756 records, resulting in 11 records that we considered potentially eligible for inclusion. Of these 11 records, we screened full texts, excluding 10 studies (see Excluded studies), resulting in one study (Biró 2015), fulfilling the inclusion criteria of this review. Thus, we included the Biró 2015 study in our review.

Included studies
According to our eligibility criteria, we included one study with a total of 44,608 household-level observations over five survey waves (Biró 2015). The study reported 40,210 household-level observations for the primary and secondary outcomes of interest. Information on the study's methods, participants, interventions, outcomes and sources of funding are given in the Characteristics of included studies tables.

Biró 2015 was funded by the Scottish Institute for Research in Economics (SIRE) Early Career Engagement Grant. The Hungarian
Central Statistical O ice provided access to the data. The study author states that the views expressed in the study do not in any way represent the views of the Hungarian Central Statistical O ice.

Interrupted time series (ITS)
The included study is an ITS study, meeting EPOC criteria for study inclusion in a review (EPOC 2012). ITS studies are non-randomised study designs. In the included study, large-scale panel data are taken from the Hungarian Household Budget and Living Conditions Survey from the years 2008 to 2012 (five waves). Data were collected on a monthly basis. Beginning 1 September 2011, Hungarians had to pay a content-based tax on specific food categories high in sugar, salt, and ca eine. Details on the taxed products and the taxation level are provided in Table 4  Cochrane Database of Systematic Reviews being. The study assessed the e ectiveness of the Hungarian public health product tax on the consumption and expenditure of sugar-added foods (taxed sweets, untaxed sweets and di erences between taxed and untaxed sweets) and other food categories high in salt and ca eine. We considered the e ect of the tax on sugaradded foods (as part of the Hungarian public health product tax) on consumption and expenditure of these foods specifically. The quantity of consumed taxed and untaxed sugar-added foods is measured in kg. The quantity of expenditure is nominal. Results on the intervention e ects are based on 42,100 observations. We did not find any study that provided evidence on the health, consumption, substitution, and expenditure e ects of the taxation of unprocessed sugar.

Participants
Biró 2015 describes the study population as follows: "The annual sample covers around 10 thousand households, 26 thousand individuals. This gives overall 44,608 household level observations throughout the 5 survey waves. The survey is a 4-years rotational survey: each household remains in the survey for 4 years" (p.110).
The survey was set up in 2005. Thus, data from 2008 belong to the first four years of data collection and therefore, data from 2008 contain information from di erent households and individuals than data from 2009 to 2012.

Taxation of unprocessed sugar
We did not identify any studies on the taxation of unprocessed sugar.

Taxation of sugar-added foods
Biró 2015 addressed the Hungarian public health product tax that came into e ect in September 2011 and was modified in January 2012. We have provided detailed information on the taxed products, the sugar threshold levels and the tax rate of the Hungarian public health product tax in Table 4. The Hungarian public health product tax includes the taxation of foods with a specific sugar content. However, the intervention also includes the taxation of SSBs and foods high in sugar, and ca eine. As the government implemented this intervention at the national level, the intervention was universal, covering the whole population.

Primary outcome
The included Biró 2015 study reported on one primary outcome: the consumption of sugar-added foods (purchased quantities). The intervention was not applied to the total of all sugar-added foods, but to the following specific categories only: • pre-packaged products with added sugar (total sugar content more than 25 g per 100 g) • chocolates (sugar content more than 40 g per 100 g and cocoa content less than 40 g per 100 g) • sugar-sweetened cocoa powder (sugar content more than 40 g per 100 g and cocoa content less than 40 g per 100 g) • jam (sugar content more than 35 g per 100 g).
Untaxed sugar-added foods primarily include fresh confectionary, fresh bakery products, and sugar-added foods from the categories above with lower sugar levels.
Biró 2015 analysed all of these taxed sugar-added foods under the category of 'taxed sweets', whereas it categorised the sugar-added foods that were not captured by the tax as 'untaxed sweets'. In our review, we have therefore included the mean consumption of taxed sugar-added foods as primary outcome.
Biró 2015 measured this outcome at the household level. They derived measures from a monthly consumption diary. These were self-reported measures. They assessed the outcome continuously on the basis of consumption diaries, starting 44 months before the implementation of the tax until 16 months a er the implementation of the intervention.

Secondary outcome
Biró 2015 reported five secondary outcomes related to substitution and expenditure on sugar-added foods.
• Substitution. Consumption of untaxed sugar-added foods and the di erence in the consumption of taxed sugar-added foods as compared to untaxed sugar-added foods are considered as a direct measurement of substitution and an indicator of the strength of substitution, respectively: • mean consumption of untaxed sugar-added foods; and • di erence in the mean consumption of taxed sugar-added foods, compared with untaxed sugar-added foods. • Expenditure: • mean expenditure on taxed sugar-added foods; • mean expenditure on untaxed sugar-added foods; and • di erence in the mean expenditure on taxed sugar-added foods, compared with untaxed sugar-added foods.
Biró 2015 measured these outcomes at the household level; measures were from a monthly consumption diary and they were self-reported. They assessed these outcomes continuously on the basis of consumption diaries, starting 44 months before the implementation of the tax until 16 months a er the implementation of the intervention.

Excluded studies
We screened full texts from a total of 11 potentially relevant studies. Ten of these studies did not fulfil our a priori-defined eligibility criteria for this review. In the Characteristics of excluded studies table, we describe the reasons for their exclusion from this review. In all excluded studies the intervention was ineligible. Such ineligible interventions were, for example, the taxation of food categories according to a pre-defined caloric content that consists of di erent high caloric ingredients, such as fat, sugar, and carbohydrates (Batis 2016; Mauricio 2019; Taillie 2017). Taxation of energy-dense and high-caloric foods does not contain a minimum threshold value on the content of sugar per 100 g or per kg: this type of intervention and its e ect is therefore not attributable to the taxation of unprocessed sugar or sugar-added foods.
We excluded the study design 'simulation study' and studies with ineligible study outcomes from our review. Bridgman 2007, for example, was consequently excluded from this review for both reasons.

Ongoing studies
We did not identify any ongoing studies.

Risk of bias in included studies
For the included ITS design, we have presented details for the risk of bias in Figure 5. Further details are provided in the tables of the section Characteristics of included studies.

Figure 5. Risk of bias summary for ITS study
In summary, we judged Biró 2015 at low risk of bias in five domains, unclear risk of bias in one domain and high risk of bias in two domains. Given the limited number of outcomes, we conducted the 'Risk of bias' assessment at the level of the study as a whole.

Allocation
In ITS designs, generation of allocation sequence and adequate concealment of allocation sequence is not applicable.

Blinding participants and personnel (performance bias)
Biró 2015 used data from the Hungarian Household Budget and Living Conditions Survey. All participants in this household survey received the intervention, since the Hungarian public health product tax was implemented at the national level. The participants and the personnel in the Hungarian Household Budget and Living Conditions Survey, however, did not know that consumer data would subsequently be used to estimate the health and expenditure e ects of the Hungarian public health product tax. Although neither participants nor personnel were blinded, we rated the risk for performance bias as low, because the participants could not know that their information would later be used to measure the e ect of the Hungarian public health product tax, especially since data collection started six years prior to the implementation of the Hungarian public health product tax. Thus, we believe that knowledge about the intervention did not a ect participants' and personnel's reporting behaviour and therefore, it may not have influenced the outcome.

Blinding outcome assessment (detection bias)
The outcomes assessed, expenditure and purchased quantities of sugar-added foods, were self-reported by the survey participants in the same month of di erent survey waves. Expenditure and purchased quantities of sugar-added foods were self-reported subjective measures. Self-reporting of outcomes is likely to bias the results in unknown ways. However, all participants received Cochrane Database of Systematic Reviews the intervention and the participants did not know that their data would later be used to assess the e ect of the Hungarian public health product tax on expenditure and purchased quantities of sugar-added foods. The survey was not linked to sugar consumption at all. Therefore, we rated the risk of detection bias for blinding of outcome assessors as low, as participants' knowledge on the intervention was unlikely to have biased the results.

Incomplete outcome data
Biró 2015 did not report information on incomplete outcome data. As data were used from the Hungarian Household Budget and Living Conditions Survey administered by the Hungarian Central Statistical O ice, we searched the statistical o ice's website and publications for information on the rates of study nonparticipation, item non-response, loss to follow-up and methods used for handling missing data. We did not find any such information and therefore rated attrition bias as unclear.

Selective reporting
ITS (observational) studies of household-level data do not generally provide study protocols or trial registrations. The included ITS study (Biró 2015) did report all expected outcomes and presented complete data and additional analyses in a supplement. We did not find evidence of selective reporting and therefore rated the risk of reporting bias as low.

Freedom from other risks of bias
Biró 2015 derived data from the Hungarian Household Budget and Living Conditions Survey. As Biró 2015 reported, while this survey did collect data on expenditure on di erent foods and on quantities of di erent foods purchased, it did not collect data on the exact quantity of sugar in these di erent foods, for example sugar content in a jam purchased. Therefore, it is very likely that untaxed foods were misclassified as taxed foods, and vice versa. This may have biased the results of the outcomes, leading probably to an underestimate of the e ect. Therefore, we judged the risk of other bias (here: misclassification of the outcome) as high.

Interventions independent
On 1 September 2011, the Hungarian government implemented the public health product tax, a content-based tax on specific food categories high in sugar (including SSBs), salt, and ca eine, which was revised on 1 January 2012 (including tax increases). Therefore, the taxation of sugar-added foods was accompanied by the taxation of SSBs, salt and ca eine. Table 4 provides an overview of the implementation of the Hungarian public health product taxes on 1 September 2011 and 1 January 2012 (Biró 2015; Martos 2017), and how this intervention implementation corresponds with the di erent waves of data collection used in the included study (Biró 2015). The taxation of sugar-added foods was not fully independent of other changes, because the other co-interventions implemented in parallel (i.e. taxation of SSBs and products high in salt and ca eine) may have influenced our outcomes, in unknown ways. We rated the risk of bias from co-interventions as high.

Shape of e ect pre-specified
Although the time of analysis was not the same time as the point of intervention, we judged the risk of bias to be low, as the change in the e ects occurred in a plausible timeframe and manner. Biró 2015 included time as a variable in the analysis with a time-specific indicator of taxation.

Intervention had no e ect on data collection
Biró 2015 used data from the Hungarian Household Budget and Living Conditions Survey, which was conducted fully independently of the implementation of the Hungarian public health product tax. The same methods of data collection were applied pre-and post-intervention and we consequently rated the risk of bias from the intervention a ecting data collection as low.

E ects of interventions
See: Summary of findings for the main comparison 'Summary of findings' table for primary outcomes: Taxation of sugar-added foods compared to no taxation for reducing consumption of sugaradded foods; Summary of findings 2 'Summary of findings' table for secondary outcomes: Taxation of sugar-added foods compared to no taxation for reducing expenditure on and assessing substitution of sugar-added foods Summary of findings for the main comparison presents an overview of the e ects of the taxation of sugar-added foods for the primary outcome. Summary of findings 2 presents an overview of the e ects of the taxation of sugar-added foods for reported secondary outcomes.

Consumption of unprocessed sugar or sugar-added foods
Biró 2015 did not measure consumption of unprocessed sugar. However, the study provided evidence on the e ects of taxing foods exceeding a specific sugar threshold value on the consumption of sugar-added foods.

Consumption of taxed sugar-added foods
Biró 2015 provided evidence on the e ect of taxing foods exceeding a specific sugar threshold value on the consumption of sugar-added foods. A er implementation of the Hungarian public health product tax, the mean consumption of taxed sugar-added foods (measured in units of kg) decreased by 4.0%, corresponding to a reduction of 40 g per kilo (SMD −0.040, 95% CI −0.07 to −0.01; P < 0.05; SE 0.02; SD 3.41; 40,210 households; very low-certainty evidence).

Energy intake
The included study did not measure energy intake through unprocessed sugar or sugar-added foods, or total energy intake.

Overweight and obesity
The included study did not measure the incidence or prevalence of overweight or obesity.

Consumption of untaxed sugar-added foods
The mean consumption of untaxed sugar-added foods (measured in units of kg) decreased a er implementation of the tax intervention by 1.3%, corresponding to a reduction of 13 g per kg (SMD −0.013, 95% CI −0.05 to 0.02; P > 0.10; SE 0.02; SD 3.41; 40,210 households; very low-certainty evidence).
The mean consumption of taxed sugar-added foods (measured in units of kg) did not di er meaningfully from the mean consumption of untaxed sugar-added foods a er the implementation of the intervention (SMD −0.028, 95% CI −0.07 to 0.02; P > 0.10; SE 0.02; SD 4.61; 40,210 households; very low-certainty evidence).

Expenditure
Biró 2015 did not measure total expenditure on food or total expenditure on sugar. However, the study provided evidence on the e ects of taxing foods exceeding a specific sugar threshold value on the mean expenditure on sugar-added foods.

Expenditure on taxed sugar-added foods
There was no e ect of the intervention on the mean expenditure on taxed sugar-added foods (measured in units of Hungarian Forint (HUF)), although data show that the mean expenditure decreased by 0.6% (SMD −0.006, 95% CI −0.03 to 0.02; P > 0.10; SE 0.01, SD 2.81; 40.210 households; very low-certainty evidence).

Expenditure on untaxed sugar-added foods
The mean expenditure on untaxed sugar-added foods (measured in units of HUF) increased a er the implementation of the Hungarian public health product tax by 3.0% (SMD 0.03, 95% CI −0.01 to 0.07; P < 0.10; SE 0.02; SD 3.61; 40.210 households; very low-certainty evidence).

Di erence in the expenditure on taxed sugar-added foods compared with untaxed sugar-added foods
The mean expenditure on taxed sugar-added foods (measured in units of HUF) di ers from the mean expenditure on untaxed sugaradded foods by −3.7% (SMD −0.037, 95% CI −0.08 to 0.01; P < 0.10; SE 0.02; SD 4.41; 40.210 households; very low-certainty evidence).

Demand
The included study did not measure the total sales of unprocessed sugar or sugar-added foods.

Other health outcomes
The included study did not measure health-related quality of life (e.g. Short Form 36 (SF-36), Health-Related Quality of Life (HRQOL-14)), mortality, or any other health outcomes (e.g. dental caries, type 2 diabetes, cardiovascular diseases, etc.).

Summary of main results
From a total of 24,454 records, one study (Biró 2015), met the a priori-defined eligibility criteria for inclusion in our systematic review. We identified evidence on the e ects of taxing sugaradded foods regarding their consumption (primary outcome), substitution and expenditure (secondary outcomes). However, we found no studies that looked at the e ects of taxing sugar-added foods on other consumption-related outcomes such as energy intake, or on other expenditure-related outcomes, such as total sales of sugar-added foods. Moreover, we found no studies that looked at the e ects of taxing sugar-added foods on health-related outcomes, such as overweight, obesity and other health outcomes. Findings from our systematic review show that there is a substantial lack of evidence on the e ects of taxing unprocessed sugar as we did not identify any study investigating this kind of intervention and its e ects.
From the results of this systematic review, as derived from one included study, we do not know whether the taxation of sugaradded foods is e ective for reducing their consumption. The results from the primary study indicated a small reduction in consumption, but the certainty of the evidence is very low. The e ect on the mean consumption on untaxed sugar-added foods was small and inverse, that is, not reflecting a substitution e ect. However, regarding the very low certainty of the evidence, we do not know whether the taxation of sugar-added foods in fact results in a substitution e ect or not. Also, the e ect of taxing sugar-added foods on the di erence in the consumption of taxed sugar-added foods as compared to untaxed sugar-added foods is considered to be small. However, we are uncertain if taxing sugar-added foods has an e ect on the di erence in the consumption of taxed sugaradded foods as compared to untaxed sugar-added foods. We do not know whether the taxation of sugar-added foods is e ective for reducing expenditure on taxed sugar-added foods. Although our single included study showed an e ect of taxing sugar-added foods on the expenditure on untaxed sugar-added foods and the e ect on the di erence in the expenditure on taxed sugar-added foods as compared to untaxed sugar-added foods, the certainty of the evidence is very low. Therefore, we are uncertain whether taxing sugar-added foods increases expenditure of untaxed sugar-added foods and a ects the di erence in the expenditure of taxed sugaradded foods as compared to untaxed sugar-added foods.
We could not pool any study results or combine intervention groups in a meta-analysis. On the individual level, the clinical significance of the results is minimal. We are uncertain about the e ectiveness of taxing sugar-added foods, but if there is evidence for small e ects in future updates, taxing sugar-added foods may be meaningful on the population level and important for public health policy actions. Our results are derived from one single study with very low-certainty evidence and we have to be cautious with the generalisability of the results, as the results apply to the Hungarian setting and it is unclear whether similar and comparable results could be achieved with the same interventions in other European countries or across the globe. For all results, the certainty of evidence is very low, and therefore, we have to be cautious with interpretations as it is not known whether taxing sugaradded foods is e ective to decrease their consumption or improve health outcomes. Our findings demonstrate the necessity of further research to investigate the e ectiveness of taxing unprocessed sugar and sugar-added foods on consumption, expenditure, and health-related outcomes. As demonstrated in Figure 4, a large part of the assumed pathways remained empty and we were not able to follow one pathway to the end. In summary, there is insu icient evidence to assess whether the taxation of unprocessed sugar and sugar-added foods other than SSBs is e ective to reduce their consumption, demand and expenditure, to improve health outcomes and to cause a dietary shi in terms of substitution and total energy intake.

Overall completeness and applicability of evidence
In this review, the current body of evidence is insu icient to adequately address the review's objectives. Existing evidence is derived from one study (Biró 2015) and thus, the evidence is limited with respect to comparability (i.e. countries: limited to Hungary) and generalisability of treatment e ects. There is a substantial lack of evidence on the e ects of taxing unprocessed sugar as we did not identify any study investigating this kind of intervention and its e ects. Evidence is completely lacking on the e ects of taxing sugar-added foods on energy intake, total sales of sugaradded foods, and health-related outcomes such as overweight and obesity. Evidence on the e ects of taxing sugar-added foods on their consumption and expenditure is addressed by the included study. However, evidence needs to be improved, particularly as the included results might be biased by co-interventions and misclassification of products into taxed and untaxed food categories. Furthermore, the Hungarian public health product tax did not cover the taxation of all sugar-added foods, but only selected foods with high sugar content. As a result of the cointerventions, we have no evidence on the e ects of sugar-added foods implemented as a standalone intervention. For example, the observed e ects on the consumption and substitution of and expenditure on sugar-added foods may result from the interaction of taxation of foods high in salt, sugar and/or ca eine together (for details of the co-interventions see Table 4), whereas the net e ect of the taxation of foods high in sugar alone -on their consumption and expenditure -could be lower and on the substitution e ect potentially higher. Households' consumption and expenditure are based on a monthly consumption diary. Accurately measuring consumption and expenditure is challenging. Diary data were found to be substantially biased by measurement error in recall food expenditure (Brzozowski 2017). The study from Schmidt 2014 suggests that individuals report more transactions in the consumption diary within the first days of the diary period as compared to reports in later diary periods. Similarly, the taxed and untaxed products may have been misclassified. These misclassification biases may have a ected the results, with underestimation of the e ect size likely. For the reasons outlined, evidence from this study may have limited applicability.

Quality of the evidence
For the taxation of sugar-added foods, we assessed the certainty of evidence of consumption, substitution and expenditure outcomes as very low. Therefore, the real e ect of taxing sugar-added foods may di er substantially from the estimated e ects and thus, our confidence in the e ect estimates is very low. Further studies are likely to change the e ect estimates of all outcomes included in this review.
We downgraded the maximum of two levels for risk of bias because the study design is an observational study (downgraded once, to 'low', for all outcomes) with a simultaneous intervention of other taxes and likely misclassification of food products as to whether taxed or untaxed (further downgraded to 'very low' for all outcomes). Although it is only possible to downgrade two levels, we could also have downgraded two levels for indirectness because the Hungarian public health product tax is related to specific sugar contents in the particular food categories that were taxed, and the study measured purchased quantities and not consumption; thus, it is not a direct representation of the e ect of a complete tax on all sugar-added foods. The downgrade in this category would apply to all outcomes. We did not downgrade for imprecision. We considered the study to be precise because it is large enough and confidence intervals did not include conflicting values.

Potential biases in the review process
We rated the risk of potential bias in the review process as low. We have strong confidence that we identified all eligible studies for inclusion in this review. We applied a very broad search strategy for three systematic reviews conducted in tandem, including the taxation of SSBs (Heise 2016), and saturated fat (Lhachimi 2016). The large-scale search was conducted in 12 electronic databases, six grey literature databases, internet search engines, and key organisational and institutional websites, and was supplemented by handsearching of reference lists to ensure that we identified all potentially relevant records. An Information Specialist (CF), established the search strategy, which a second review author (TLH) partially adapted, and a further Information Specialist reviewed the search strategy on behalf of Cochrane Public Health (Review Milestone 1). Our Information Specialist (CF) or another review author (TLH) conducted all database searches. We also asked our review advisory group members for relevant published and unpublished records. A minimum of two review authors independently screened titles, abstracts and full texts, extracted data and assessed quality (GRADE). We included a broad range of study designs to maximise the completeness of the evidence. In the section Di erences between protocol and review, we describe changes between the protocol and review that may have introduced bias. However, in this review, we made no major changes and thus, we prevented reporting bias.

Agreements and disagreements with other studies or reviews
We are not aware of any previously conducted systematic reviews on the e ects of taxing unprocessed sugar or sugar-added foods on our a priori-defined primary and secondary outcomes. However, systematic reviews on general food taxes and subsidies suggest a positive e ect on consumption, body weight and disease incidence, with greater e ects in higher tax rates (e.g. Afshin 2017; Alagiyawanna 2015; Maniadakis 2013; Niebylski 2015; Thow 2011; Thow 2014). Niebylski 2015 suggests a minimum tax of 10% to 15% on di erent foods and beverages for their e ectiveness on consumption and public health. Maniadakis 2013 concludes that e ects of food taxes on total caloric intakes might be much smaller. A review of simulation studies suggests that taxes on SSBs and saturated fat and subsidies on healthy foods result in a reduced calorie intake and a decreased consumption of the taxed food (Eyles 2012). However, the evidence base in the mentioned reviews is of low quality with high heterogeneity with respect to tax rates, taxed items, and study designs of included studies. As the interventions in existing reviews and the methodological approaches are completely di erent to our review, we cannot compare our findings with those of the reviews mentioned above.

Implications for practice
Due to very limited and very low-certainty evidence, we cannot derive concrete conclusions on the e ectiveness of taxing sugaradded foods for reducing their consumption and preventing obesity or other adverse health outcomes. Despite the reported reduction of the consumption of sugar-added foods in the primary study, we are uncertain whether taxing sugar-added foods has an e ect on reducing their consumption and preventing obesity or other adverse health outcomes as the evidence is Cochrane Database of Systematic Reviews very low certainty. We did not find any studies that looked at the e ectiveness of taxing unprocessed sugar for reducing its consumption and preventing obesity or other adverse health outcomes. We did not find any studies that looked at the e ects of taxing sugar-added foods for consumption of unprocessed sugar, energy intake, overweight, and obesity or any health-related outcomes that would be of great interest to derive implications for practice.

Implications for research
Further studies supporting greater certainty of the evidence are required to assess the e ectiveness of taxing unprocessed sugar or sugar-added foods for reducing their consumption and preventing obesity or other adverse health outcomes. To our knowledge, taxes on unprocessed sugar or sugar-added foods are currently implemented in the following seven countries: Bermuda ( . Most of these taxes were implemented relatively recently and therefore, there is great potential for further studies on the e ects of taxing unprocessed sugar or sugar-added foods.
Further research is particularly needed in these countries to assess the e ects of taxes on unprocessed sugar or sugar-added foods. Specifically countries that tax unprocessed sugar are of high interest for this review as the taxation of unprocessed sugar a ects all other products with sugar as an ingredient in processed foods. Therefore, further studies on the e ects of taxing unprocessed sugar, and possibly simultaneously sugar-added foods, should focus on the taxation e ects in Norway and on St. Vincent and the Grenadines. All future studies should also consider health e ects as relevant outcome domains.

Thomson 2013
Thomson H. Improving utility of evidence synthesis for healthy public policy: the Three Rs (Relevance, Rigor, and Readability [and Resources]). American Journal of Public Health 2013;103:e17-23.

Thow 2010
Thow AM, Jan S, Leeder S, Swinburn B. The e ect of fiscal policy on diet, obesity and chronic disease: a systematic review.

Thow 2014
Thow AM, Downs S, Jan S. A systematic review of the e ectiveness of food taxes and subsidies to improve diets: understanding the recent evidence. Nutrition Reviews 2014;72 (9):551-65.

Vareiro 2009
Vareiro D, Recruitment: the study's primary report does not offer information on participants' recruitment.
Equity considerations: quote: "The estimated effects were driven by households belonging to the lowest income quartile, who are more responsive to increases in price." Interventions Intervention: the Hungarian public health product tax was implemented on the national level in Hungary on 1 September 2011 and revised on 1 January 2012. The intervention includes taxes on sugar-added foods: • pre-packaged products with added sugar, containing sugar contents > Cochrane Database of Systematic Reviews • energy intake: not measured • overweight and obesity: not measured Secondary outcomes: • expenditure: mean expenditure on taxed sugar-added foods (measured in units of HUF, mean expenditure on untaxed sugar-added foods (measured in units of HUF), and SMD in the expenditure on taxed sugar-added foods vs untaxed sugar-added foods, assessed continuously throughout the 5 study waves with routinely collected data from consumption diaries • substitution and diet: not measured • demand: not measured • other health outcomes: not measured

Notes
Sources of funding: the study was funded by the Scottish Institute for Research in Economics (SIRE) Early Career Engagement Grant

Bias Authors' judgement Support for judgement
Blinding of participants and personnel (performance bias) All outcomes Low risk Comment: although participants and personnel were not blinded and all participants received the intervention, participants and personnel of the Hungarian Household Budget and Living Conditions Survey did not know that data were collected to measure the effect of the Hungarian public health product tax.
Blinding of outcome assessment (detection bias) All outcomes Low risk Comment: outcome assessment is based on objective self-reports Incomplete outcome data (attrition bias) All outcomes Unclear risk Comment: information on how and whether incomplete outcome data were addressed is not reported.
Selective reporting (reporting bias) Low risk Comment: supplement gives an overview of the complete data.
Other bias High risk Comment: the categorisation of foods into taxed and untaxed food categories may be wrong in parts as relevant information for building correct categories was missing.
Intervention independent High risk Comment: the intervention on the taxation of sugar-added foods was accompanied by the taxation of SSBs and foods high in salt and caffeine. Outcomes might be affected by the other interventions.
Shape of effect pre-specified Low risk Quote from report: "The variable time captures a linear trend measured in months. Vector includes a set of household characteristics. The time-specific indicator of taxation is Tt=I(T >=September 2011)." (p. 111) Intervention had no effect on data collection Intervention not eligible: no baseline directly before implementation of the tax intervention conditions 3-5. The definition of 'healthier' and 'less-healthier' remains unclear as, "researchers defined items that met at least two of the three standards in their entirety as healthier, and other items as less healthy" (p. 50). Outcomes on sugar contain both foods and beverages together.

Hanks 2013
Intervention not eligible: less healthy foods were taxed, healthier foods were subsidised. Taxation without information on sugar levels. Healthy foods contain sugar-added foods.

Hanks 2014
Intervention not eligible: less healthy foods were taxed, healthier foods were subsidised. Taxation without information on sugar levels. Healthy foods contain sugar-added foods.

Martos 2016
Study design not eligible: cross-sectional study without pre-intervention baseline. Intervention not eligible: two independent sections after implementation of tax.

Mauricio 2019
Intervention not eligible: no intervention on unprocessed sugar or sugar-added foods, intervention addresses foods with energy density ≥ 250 kcal/100 g

Taillie 2017
Intervention not eligible: untaxed food contains salty snacks, cereals and sugar-added foods. Intervention is applied to high caloric foods.

Unar-Munguia 2019
Intervention not eligible: NAFTA as intervention. No continuity of taxation: simultaneous and overlapping price increases and decreases.   meal replacement beverages, infant formula, concentrates used in combination with other ingredients to create a beverage   Comments (1): I imagine some evidence will be presented as simply a change in BMI or other markers of obesity rather than a change in incidence or prevalence of obesity (Cristina Cleghorn) 3.1. Do you think the same outcomes are appropriate for both reviews (SSB; sugar or sugar-added foods)?

External sources
• Cochrane Review Support Programme 2019, UK.
Pfinder successfully applied to the Cochrane Review Support Programme 2019. The award is granted upon completion of the project by 1st September 2020.

D I F F E R E N C E S B E T W E E N P R O T O C O L A N D R E V I E W Background
We updated the background to outline the current state of research.

Types of secondary outcomes
We further specified and included the following secondary outcomes.
• Substitution: • consumption of untaxed sugar-added foods; • di erence in mean consumption of taxed sugar-added foods compared with untaxed sugar-added foods. Cochrane Database of Systematic Reviews • Expenditure: • expenditure on untaxed sugar-added foods; • di erence in mean consumption of taxed sugar-added foods compared with untaxed sugar-added foods.

Types of interventions
We further specified our eligibility criteria on types of interventions. We excluded virtual and hypothetical interventions imitating a taxation on unprocessed sugar or sugar-added foods if participants' purchase decisions were not binding so that they did not all result in a real purchase or if the money was virtual or not belonging to the study participant.

Searches
We did not search the website from The Obesity Society (www.obesity.org), as it was not possible to enter the search terms. The database from the website The Directory of Open Access Repositories -OpenDOAR via OpenDOAR was not accessible a er the search on 12 December 2016. The free text search on the website Trials Register of Promoting Health Interventions (TRoPHI) via EPPI-Centre was not accessible a er the search on 11 August 2016.

Data synthesis
We did not undertake harvest plots for narrative synthesis.