A systematic review of economic evaluations of population-based sodium reduction interventions

Objective To summarise evidence describing the cost-effectiveness of population-based interventions targeting sodium reduction. Methods A systematic search of published and grey literature databases and websites was conducted using specified key words. Characteristics of identified economic evaluations were recorded, and included studies were appraised for reporting quality using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist. Results Twenty studies met the study inclusion criteria and received a full paper review. Fourteen studies were identified as full economic evaluations in that they included both costs and benefits associated with an intervention measured against a comparator. Most studies were modelling exercises based on scenarios for achieving salt reduction and assumed effects on health outcomes. All 14 studies concluded that their specified intervention(s) targeting reductions in population sodium consumption were cost-effective, and in the majority of cases, were cost saving. Just over half the studies (8/14) were assessed as being of ‘excellent’ reporting quality, five studies fell into the ‘very good’ quality category and one into the ‘good’ category. All of the identified evaluations were based on modelling, whereby inputs for all the key parameters including the effect size were either drawn from published datasets, existing literature or based on expert advice. Conclusion Despite a clear increase in evaluations of salt reduction programs in recent years, this review identified relatively few economic evaluations of population salt reduction interventions. None of the studies were based on actual implementation of intervention(s) and the associated collection of new empirical data. The studies universally showed that population-based salt reduction strategies are likely to be cost effective or cost saving. However, given the reliance on modelling, there is a need for the effectiveness of new interventions to be evaluated in the field using strong study designs and parallel economic evaluations.


Results
Twenty studies met the study inclusion criteria and received a full paper review. Fourteen studies were identified as full economic evaluations in that they included both costs and benefits associated with an intervention measured against a comparator. Most studies were modelling exercises based on scenarios for achieving salt reduction and assumed effects on health outcomes. All 14 studies concluded that their specified intervention(s) targeting reductions in population sodium consumption were cost-effective, and in the majority of cases, were cost saving. Just over half the studies (8/14) were assessed as being of 'excellent' reporting quality, five studies fell into the 'very good' quality category and one into the 'good' category. All of the identified evaluations were based on modelling, whereby inputs for all the key parameters including the effect size were either drawn from published datasets, existing literature or based on expert advice. PLOS  reduction initiative in the United Kingdom of Great Britain and the Northern Ireland demonstrated a significant reduction in average intakes from 9.5 grams per day in 2000 to 8.1 grams per day in 2011 [24]. The initiatives consisted of an awareness campaign shown on TV through a series of adverts along with a series of partnerships with institutions running programs such as peer education and social cooking classes. The strategy also involved working with the food industry to encourage product reformulation.
Interventions that reduce sodium intake have been shown to be one of the most cost-effective measures to improve public health worldwide [25]. These interventions generally target whole populations and seek to reduce exposure to dietary sodium [26]. It is estimated that a 15% reduction in population-wide sodium consumption would avert up to 8Á5 million deaths in 23 high-burden countries over 10 years [6].
Whilst numerous studies explore the effectiveness of sodium reduction interventions on salt intake through urine collections, effect on blood pressure or cardiovascular disease outcomes [14,19,27], decision makers are also interested in which interventions deliver value-formoney in the context of limited health care resources. Economic evaluations are extremely valuable in decision making as they enable the best course of action to be identified based on the evidence available by systematically analyzing the costs and benefits associated with an intervention and assessing its value for money [28]. Whilst there is broad agreement that sodium reduction strategies are cost-effective, there are many different evaluation approaches and perspectives used, and the completed evaluations vary in quality.
The objective of this paper was to conduct a systematic review of the literature to identify economic evaluation studies of interventions targeting sodium reduction and summarise evidence about their cost-effectiveness.

Important definitions
A table of important definitions has been compiled below in Table 1, defining important terms used throughout this paper.

Search strategy
Databases searched. Literature on economic evaluations of sodium reduction interventions published between 1980 and 2015 were identified from a search of journal databases, grey literature and other articles identified by experts in the field. During January 2015, the published literature was searched using the following search engines which comprise the main health databases: Pubmed, Embase, EBSCO Host, OVID and Google Scholar. This review explores the existing literature on both economic evaluations of sodium reduction interventions actually implemented in the field and involving new empirical data collection as well as desk-based modelled simulation studies.
A search of grey literature was also undertaken using the same search terms in order to find information that may only have been published in government reports or discussion papers. The search was undertaken using Google, Open Grey, the World Health Organization database and website and the World Bank website. The reference lists of extracted articles were also searched for any additional studies.
Search terms. Each database was searched using the following key words: "Economic Evaluat Ã ", "Cost Effect Ã ", "Cost Benefit", "Cost Utility", "Cost Analyses" and "Intervention Ã ". Each search term was combined with the key words "Sodium OR Salt" and "Reduc Ã ".
Study inclusion criteria. To be included, a study had to comply with all of the following criteria: 1. Be an intervention or simulation study that targeted the reduction of sodium intake at a population level (i.e. targeting populations rather than individuals). Both prospective and retrospective studies were included.
2. Presented the findings of full economic evaluations which explore both costs and benefits in relation to a comparator. A full economic evaluation was defined as the comparative analysis of alternative courses of action in terms of both costs (resource use) and consequences (outcomes, effects) [30]. Full economic evaluations include studies utilising CBA, CEA or CUA. Partial economic analyses, which focused solely on costs and resource used, or which did not entail a comparator, were excluded.

Reported in English.
The systematic review was conducted by SH following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [32]. The results were identified by title, then screened by abstract, followed by a full text assessment for eligibility.
Analyses. Key characteristics of the economic evaluation of each of the identified sodium reduction studies were extracted into a spreadsheet including the economic evaluation study design, year and country of study, setting, sample size, time horizon, study perspective, study comparator, intervention(s) analyzed, the methods or models used to conduct the economic evaluation, costs included, the primary outcome measure and the main results and conclusions of the study.
The reporting quality of the identified studies was measured against the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist for assessing economic  [29].
Cost-effectiveness analysis (CEA) An evaluation in which the effects of an intervention (and its comparators) are measured in identical units of outcome (e.g. mortality, myocardial infarctions) and alternative interventions are compared in terms of 'cost per unit of effect' [30].
Cost-utility analysis (CUA) When alternative interventions produce different levels of effect in terms of both quantity and quality of life (or different effects), the effects may be expressed in utilities. Utilities are measures which comprise both length of life and subjective levels of well-being. The best known utility measure is the quality-adjusted life year, or QALY. Alternative interventions are compared in terms of cost per unit of utility gained (e.g. cost per QALY) [30].
Cost-benefit analysis (CBA) When both resource inputs and effects of alternative interventions are expressed in monetary units, so that they compare directly and across programs within the healthcare system, or with programs outside health care (e.g. healthcare intervention vs. criminal justice intervention) [30].
Incremental cost-effectiveness ratio (ICER) Entails determination of the incremental cost of an additional unit of health benefit thereby enabling different interventions to be ranked in terms of their economic credentials. The ICER is calculated by difference in cost between two possible interventions, divided by the difference in their effect [30].
evaluations [33,34]. The 24 item checklist is designed to improve reporting of economic evaluations. Each of the included articles were assessed for reporting quality independently by two reviewers (SH, MM) against the criteria to calculate a score out of 24 (or the number of applicable items). Each item on the checklist was assigned one point, but half points were awarded where the article partially filled the criteria (e.g. provided no explanation for choice of discount rate or choice of model). The two reviewers (SH and MM) discussed any differences in criteria ratings in order to reach consensus. A percentage score for each study was then calculated.
In the absence of a broadly accepted method for reporting quality appraisal, categories were set based on methods from other literature [35][36][37]-a study was deemed to be of excellent reporting quality if it scored 85% or higher, 70-<85% very good quality, 55-<70% good quality and studies scoring below 55% were classified as poor quality.

Search results
From the initial search, a total of 3647 potentially relevant publications were identified. Some 924 duplicates were removed. Of the remaining 2723 titles, 2639 were found to be not relevant based on the title key words. A review of the abstracts of the remaining 84 articles identified a total of 25 potentially relevant studies. After a partial review of the full article, a further 10 papers were excluded as they did not meet the selection criteria (seven did not present findings of an economic evaluation, two were not available in English, and one was a conference abstract). In addition to the fifteen articles identified from the database search, an additional five articles were identified through either the grey literature search or referral from persons working in the field. So in total, 20 studies were identified that met the study inclusion criteria and were subjected to a full paper review [22,[38][39][40][41][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56].
The majority of studies (11/14) have been published in the past five years (2010 or later), with the other three studies being in the decade 2000-2009. This is a reflection of both the relative newness of attention being focused on sodium reduction interventions as a measure to curb hypertension, and of the infancy of the exposure of such interventions to economic evaluation.
Target settings and populations. The identified articles contained economic evaluations of interventions from a wide range of countries. Four were from low and middle income countries or regions [43,44,52,55], and 10 were from high income countries as classified by the World Bank [57]. The former group included studies from Vietnam, Syria and the Middle East (four countries in one study) and South-East Asia & Sub Saharan Africa. The latter group comprised four studies from the USA, England, Australia and Argentina, and Norway. The target group for the majority of studies was a national population; however one study targeted the population of a single city (Buenos Aries), whilst three were regional studies targeting multiple countries.
Seven of the 14 studies evaluated interventions which targeted the whole population of either a specific country [46,52], multiple countries [44,53,54], several regions [47,55], or a city [51]. Mason et al. [44] evaluated the intervention separately for the population of four Eastern Mediterranean countries (Palestine, Syria, Tunisia and Turkey), whilst Webb [53] modelled results separately for 187 different countries worldwide, Murray et al [47] for 14 epidemiological sub regions and Ortegon et al [55] for the populations of sub-Saharan Africa and South East Asia. Rubinstein 2009 focused on the city population of Buenos Aires [51]. Cobiac et al [46] and Ha et al [52] evaluated intervention for the populations of Australia and Vietnam respectively.
Of the 14 studies, the interventions were targeted at adults of varying age ranges. Four targeted young to middle age adults (35 or 40 years and over) [42,[48][49][50], although the latter study was confined to adults (35-85 years) who had never experienced a CVD event. Two studies lowered the youngest age to between late adolescence or early adult years (between 16-25 years and upwards) [43,45]. There were no studies focusing exclusively on children.
Study perspective. The economic perspective of a study is important in determining the costs and benefits included. Six of the studies [42,44,45,47,49,52] purport to include a societal perspective, which means that all costs and benefits are included irrespective of who incurs them. The remaining studies reported from a health sector perspective [46,48,50,54,55], or government perspective [51,53]. The perspective taken by Wilcox et al. [43] was not specified.
Interventions and comparator. A range of interventions aiming to reduce sodium consumption were identified. These consisted of activities aiming to influence both the supply and demand side of the food system. Supply side interventions aimed to alter the available food by providing access to lower sodium options. The main example is product reformulation (both voluntary and mandatory) to reduce the salt content of food. Demand side interventions aimed to influence demand by changing people's behavior so that they select lower sodium  options. Examples included health promotion campaigns, labeling of salt content on packaged food and taxes on salty food products. All interventions analysed were compared to either the status quo (current practice) or a null comparator. The latter, based on WHO-CHOICE methodology [58], entails an assumption of no interventions being in place, meaning that the intervention is compared to a situation of no costs and no interventions.
Over half of the studies (8/14) evaluated multiple sodium reduction interventions [43][44][45][46][47]49,52,55], while the remaining six studies evaluated only one sodium reduction strategy [42,48,[50][51][52][53][54]. Seven of the studies [42][43][44][45][46]49,53] were focused exclusively on sodium reduction strategies, whilst in the other seven studies [47,48,[50][51][52]54,55], a broader focus on the reduction of cardiovascular disease meant that the salt reduction interventions were evaluated along with a range of other non-salt initiatives. As an illustration of the latter, Murray et al 2013 [47] considered three salt interventions (health education through mass media, legislation and voluntary agreements on food labelling and salt content) amongst a total of 17 population and individual strategies to lower systolic blood pressure and cholesterol. Likewise, Cobiac et al. 2012 [48] included the mandatory reduction of salt in the manufacture of breads, margarines and cereals as part of a broader study of nine interventions exploring the best value for money in the primary prevention of cardiovascular disease. Of these eight studies, three evaluated two salt interventions [49,52,55], three had three interventions [43,44,46] and two had four interventions [45,47].
In the case of two of the six studies evaluating one salt reduction intervention only [42,53], the intervention was a multi-component intervention targeting sodium reduction, comprising product reformulation with a health promotion/education component. The other four were single component interventions [48,50,51,54].
The majority of the studies included a product reformulation strategy designed to reduce the sodium content of processed foods. Five studies [45][46][47]53,55] evaluated the cost-effectiveness of both voluntary and mandatory (regulatory) measures to restrict the salt content of processed foods, whilst four [43,44,48,54] included mandatory reformulation and five [42,[49][50][51][52] included voluntary programs targeting the food industry. The other salt reduction initiative common to seven of the papers was a health promotion/education program [42][43][44][45]47,52,53]. Whilst the English study by Collins et al. [45] specified a particular health program (Change4-Life), and Ha et al. [52] specified health education via a mass media campaign, generally little detail was provided on the nature of the program. Only two studies evaluated tax legislation for salt reduction. Selmer et al. [42] included taxes on salty foods and subsidies on products with less salt as components within a multipronged salt reduction program. In contrast to point-of-sale tax measures, Smith-Spangler et al. [49] evaluated the impact of a tax imposed on sodium used in food production.
All of the studies assumed that the full effect of the intervention would be maintained over time. This assumption seems reasonable for interventions such as product reformulation and tax legislation however this may not always be realistic for health promotion and education programs. Three studies [42,46,53] assumed that effects would appear gradually from the onset of the intervention and increase to full effect. Only study [48] explicitly mentions that the effect of the interventions is only assumed if the delivery of interventions is ongoing.
Time horizons. Economic evaluations should specify time horizons, both for the provision of the intervention itself and for tracking the associated costs/cost offsets and consequences. The evaluated duration of intervention delivery should ideally reflect how the intervention would be applied in real life. There were a range of study time lines in the identified studies; the studies generally do not justify their choice of time frame for tracking costs and benefits. Half of the studies (7/14) had a 10 year study time line in which the costs and consequences of the interventions were evaluated [43][44][45][52][53][54]59]. One study [42] had a 25 year timeline and one estimated annual costs and benefits [50]. The remaining five studies measured results over 100 years or the lifetime of the target group [46][47][48][49]55]. In the small number of studies which actually specified the intervention duration, it ranged from five to 25 years.
Study designs and models employed. All of the 14 studies entailed a cost-utility analysis where the incremental cost-effectiveness ratios were reported as a ratio of costs against a measure of utility. Four studies [42][43][44][45] reported cost per life years gained, and two [49,54], costs per quality-adjusted life years (QALY) gained. All of the other studies measured costs per disability-adjusted life year (DALYs) saved.
The studies employed various forms of simulation modelling to examine the impact of the specified intervention on population health. Four studies based their analytic model on the WHO-CHOICE methodology [47,51,52,55]-reductions in population attributable risks of cardiovascular events resulting from an intervention were calculated, and then translated into changes in population health using the standard multi-state model, Pop Mod. Pop Mod estimates the lifetime health gains for each age and sex cohort of the given population (divided into different health states) both with and without the intervention. Three studies [43][44][45] used country specific versions of IMPACT, a comprehensive, validated coronary heart disease (CHD) model to estimate the reduction in CHD mortality stemming from an intervention. Other studies [42,48,49] developed purpose-built Markov models which assume that each participant is always in one of a finite number of discrete health states and events are represented as transitions from one state to another. Discount rates. The majority of studies (11/14) applied a 3% discount rate to costs and benefits, whilst Collins et al [45] and Barton et al [54] used a 3.5% rate and Selmer et al [42] a 5% rate. The choice of discount rate was expected to vary between settings and location but most of the studies did not justify their choice of rate level.
Resource use costing. Items included in the cost measurement varied depending on the study purpose and perspective and the nature and number of intervention(s) being evaluated. Some studies such as Wilcox et al [43] and Collins et al [45], assumed a broad, societal approach to costing, in order to facilitate the inclusion of costs to all sectors, including the food industry (for example, the costs of product reformulation and relabeling). Others adopted a narrower focus and confined themselves, for instance, to costs to government [51,53] or the health care sector [54].
Quality assessment of the studies. The reporting quality of the 14 studies was assessed against 24 checkpoints and allocated a score of 1 for each point that was met in full (symbolized as p ), a score of 0.5 for each point that was partially met (symbolized as 6 ¼) and a score of 0 for each point that was not met (symbolized as X) ( Table 4). The majority of studies (8/14) [44,46,[48][49][50][51][52]55] were found to be of excellent reporting quality (scoring 85% or higher), with the remaining five to be of 'very good' quality (scoring 70-85%) [42,43,45,47,54] and one to be of 'good' quality (scoring 55-70%) [53].
The two criteria which were least well addressed were the time horizon and model choice. Whilst the time horizon for each study was generally specified, most studies omitted to provide reasons for choice. Likewise, very few studies provided justification for their choice of economic model. Other key areas where studies lost quality points related to study perspective (sometimes it was not explicitly stated or related to the costs included) and health outcomes (where their relevance was not made clear). It should be noted that the assessment of reporting quality is not indicative of the quality of the actual study results.

Cost-effectiveness results
All of the fourteen studies concluded that their specified interventions targeting reductions in sodium consumption were cost-effective, and in the majority of cases, were cost saving (in other words, they resulted in health gains at a lower cost, measured against the comparator) ( Table 5). For example, Barton et al. [54] concluded that any sodium reduction initiative that achieved even a modest population-wide reduction in any major cardiovascular risk factor would produce a net cost saving to United Kingdom's National Health Service.
Many studies examined the cost-effectiveness of a combination of interventions making it sometimes difficult to ascertain the effectiveness of a single intervention. In the seven studies [47,48,[50][51][52]54,55] which evaluated both salt reduction strategies and other non-salt strategies to reduce poor cardiovascular outcomes, the strategies focused on salt reduction generally represented the best 'value-for-money' given their low-cost and population wide impacts. For example, Ha et al. [52] evaluated 12 population and individual level interventions to prevent cardiovascular disease in Vietnam and found a mass media campaign to reduce salt intake as the most cost-effective. Likewise, Cobiac et al 2012 [48] showed that mandating the more moderate use of salt in breads, margarines and cereals was easily the most cost-effective strategy for primary prevention of cardiovascular disease in Australia.
Whilst the results are generally not comparable between studies due to the heterogeneous nature of the methods used, the studies that evaluated multiple salt interventions indicate that some initiatives are consistently more cost-effective than others. Mandatory product reformulation was found to be substantially more cost-effective than the food industry sector undertaking voluntary reformulation [45][46][47][48]. The 2010 study by Cobiac et al. [46] found that making recommended limits for salt in bread, margarine and cereal products mandatory would potentially avert 18% of the disease burden arising from excessive salt consumption which was 20 times greater than the health gains achieved with the voluntary approach.
There was also evidence from two studies that multiple interventions working together are likely to be more cost-effective than any single intervention (e.g. [43,44]. Mason et al. [44] found that in all four Eastern Mediterranean countries targeted a comprehensive strategy of health education, food labelling and mandatory product reformulation would produce the greatest benefit in terms of life years gained and cost savings. shown to be more cost-effective in avoiding poor cardiovascular disease outcomes than other non-salt strategies. This finding was in line with results from previous studies [60][61][62].
Whilst most of the 14 studies were from high income countries, there were several studies in middle or low income countries. The majority of the studies have been published in the past ten years. Whilst growing attention is being given to effective interventions to reduce salt consumption, very few interventions to date have been subjected to economic evaluation. Also, the interventions which have been evaluated in terms of their economic credentials are narrow in terms of their content; most related to product reformulation, relabeling, or health promotion programs, with only a couple targeting tax legislation.
A recent systematic review of salt reduction initiatives around the world identified interventions in different categories: food reformulation, consumer education, front of pack labelling and interventions in public institution settings and taxation [23]. This review found that economic evaluations have been completed for all sodium reduction intervention categories except for the 'interventions in public institution settings' category. There was also one study identified [49] relating to sodium taxation. This indicates a gap in existing literature and a need for economic evaluations of these different interventions.
None of the identified studies were based on actual implementation and evaluation of interventions. Instead the interventions were simulated using economic modelling and intervention effectiveness data were drawn from external sources or the academic literature. None of the papers made explicit mention of procedures for checking their models. Five of the studies used an existing validated model for their analyses. Three of the studies [43][44][45] reported using the existing validated IMPACT CHD model to compare their results, whilst two of the studies [47,51] used PopMod to model their analyses. Future evidence would be strengthened by the actual implementation of intervention trials within real-life settings. Despite this reliance on modelling and associated assumptions, the studies evaluated are important as model-based health economic evaluations are today widely accepted as policy-making tools that can inform resource allocation decisions.
A key strength of this review is the systematic and comprehensive method of data collection. A comprehensive search strategy was employed encompassing both peer reviewed and grey literature. The quality assessment of the economic evaluations undertaken as part of this review adds strength to the conclusions since all studies were found to be of good, very good or excellent reporting quality. The results in this review are limited to those published in English representing a potential limitation. Another limitation is that studies identified were not based on actual implementation of intervention(s) and the associated collection of new empirical data. Given the result of the studies are based on modelling and assumed costs and effectiveness, researcher bias may have influenced these findings. All studies identifies were based on modelling where inputs were drawn from published datasets, existing literature or expert advice. As the results from the study rely heavily on modelling, there is a need for the effectiveness of new interventions to be evaluated in the field using strong study designs and parallel economic evaluations.

Conclusions
Reducing the consumption of salt is now seen as a key priority in many strategies targeting the prevention of cardiovascular disease but relatively few interventions designed to lower salt intake have been rigorously evaluated. Even fewer have been examined in terms of their economic credentials. Nevertheless, the economic evaluations identified in this field suggest that salt lowering strategies are potentially cost effective and offer better value-for-money than many other non-salt strategies. In addition to simulation modelling studies, there is an urgent need for the effectiveness of salt interventions to be actually evaluated in the field using strong study designs, and economic evaluations conducted in parallel.