Fortification of salt with iron and iodine versus fortification of salt with iodine alone for improving iron and iodine status

Abstract

Background

Iron deficiency is an important micronutrient deficiency contributing to the global burden of disease, and particularly affects children, premenopausal women, and people in low‐resource settings. Anaemia is a possible consequence of iron deficiency, although clinical and functional manifestations of anemia can occur without iron deficiency (e.g. from other nutritional deficiencies, inflammation, and parasitic infections). Direct nutritional interventions, such as large‐scale food fortification, can improve micronutrient status, especially in vulnerable populations. Given the highly successful delivery of iodine through salt iodisation, fortifying salt with iodine and iron has been proposed as a method for preventing iron deficiency anaemia. Further investigation of the effect of double‐fortified salt (i.e. with iron and iodine) on iron deficiency and related outcomes is warranted. 

Objectives

To assess the effect of double‐fortified salt (DFS) compared to iodised salt (IS) on measures of iron and iodine status in all age groups.

Search methods

We searched CENTRAL, MEDLINE, Embase, five other databases, and two trial registries up to April 2021. We also searched relevant websites, reference lists, and contacted the authors of included studies.

Selection criteria

All prospective randomised controlled trials (RCTs), including cluster‐randomised controlled trials (cRCTs), and controlled before‐after (CBA) studies, comparing DFS with IS on measures of iron and iodine status were eligible, irrespective of language or publication status. Study reports published as abstracts were also eligible.

Data collection and analysis

Three review authors applied the study selection criteria, extracted data, and assessed risk of bias. Two review authors rated the certainty of the evidence using GRADE. When necessary, we contacted study authors for additional information. We assessed RCTs, cRCTs and CBA studies using the Cochrane RoB 1 tool and Cochrane Effective Practice and Organisation of Care (EPOC) tool across the following domains: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other potential sources of bias due to similar baseline characteristics, similar baseline outcome assessments, and declarations of conflicts of interest and funding sources. We also assessed cRCTs for recruitment bias, baseline imbalance, loss of clusters, incorrect analysis, and comparability with individually randomised studies. We assigned studies an overall risk of bias judgement (low risk, high risk, or unclear). 

Main results

We included 18 studies (7 RCTs, 7 cRCTs, 4 CBA studies), involving over 8800 individuals from five countries. One study did not contribute to analyses. All studies used IS as the comparator and measured and reported outcomes at study endpoint. 

With regards to risk of bias, five RCTs had unclear risk of bias, with some concerns in random sequence generation and allocation concealment, while we assessed two RCTs to have a high risk of bias overall, whereby high risk was noted in at least one or more domain(s). Of the seven cRCTs, we assessed six at high risk of bias overall, with one or more domain(s) judged as high risk and one cRCT had an unclear risk of bias with concerns around allocation and blinding. The four CBA studies had high or unclear risk of bias for most domains.

The RCT evidence suggested that, compared to IS, DFS may slightly improve haemoglobin concentration (mean difference (MD) 0.43 g/dL, 95% confidence interval (CI) 0.23 to 0.63; 13 studies, 4564 participants; low‐certainty evidence), but DFS may reduce urinary iodine concentration compared to IS (MD −96.86 μg/L, 95% CI −164.99 to −28.73; 7 studies, 1594 participants; low‐certainty evidence), although both salts increased mean urinary iodine concentration above the cut‐off deficiency. For CBA studies, we found DFS made no difference in haemoglobin concentration (MD 0.26 g/dL, 95% CI −0.10 to 0.63; 4 studies, 1397 participants) or urinary iodine concentration (MD −17.27 µg/L, 95% CI −49.27 to 14.73; 3 studies, 1127 participants). No studies measured blood pressure.

For secondary outcomes reported in RCTs, DFS may result in little to no difference in ferritin concentration (MD −3.94 µg/L, 95% CI −20.65 to 12.77; 5 studies, 1419 participants; low‐certainty evidence) or transferrin receptor concentration (MD −4.68 mg/L, 95% CI −11.67 to 2.31; 5 studies, 1256 participants; low‐certainty evidence) compared to IS. However, DFS may reduce zinc protoporphyrin concentration (MD −27.26 µmol/mol, 95% CI −47.49 to −7.03; 3 studies, 921 participants; low‐certainty evidence) and result in a slight increase in body iron stores (MD 1.77 mg/kg, 95% CI 0.79 to 2.74; 4 studies, 847 participants; low‐certainty evidence). In terms of prevalence of anaemia, DFS may reduce the risk of anaemia by 21% (risk ratio (RR) 0.79, 95% CI 0.66 to 0.94; P = 0.007; 8 studies, 2593 participants; moderate‐certainty evidence). Likewise, DFS may reduce the risk of iron deficiency anaemia by 65% (RR 0.35, 95% CI 0.24 to 0.52; 5 studies, 1209 participants; low‐certainty evidence). 

Four studies measured salt intake at endline, although only one study reported this for both groups. Two studies reported prevalence of goitre, while one CBA study measured and reported serum iron concentration. One study reported adverse effects. No studies measured hepcidin concentration.

Authors' conclusions

Our findings suggest DFS may have a small positive impact on haemoglobin concentration and the prevalence of anaemia compared to IS, particularly when considering efficacy studies. Future research should prioritise studies that incorporate robust study designs and outcome measures (e.g. anaemia, iron status measures) to better understand the effect of DFS provision to a free‐living population (non‐research population), where there could be an added cost to purchase double‐fortified salt. Adequately measuring salt intake, both at baseline and endline, and adjusting for inflammation will be important to understanding the true effect on measures of iron status.

Plain language summary

Fortification of salt with iron and iodine compared to salt fortified with iodine only for improving iron and iodine status

Key messages

Compared to iodised salt, double‐fortified salt (salt fortified with iron and iodine) may improve some measures of iron and iodine nutrition, such as haemoglobin (i.e. the substance that gives red blood cells their colour) concentrations and body iron stores. However, it may reduce urinary iodine concentration and may make little or no difference to ferritin (i.e. iron‐storage protein) concentrations and transferrin receptor (i.e. protein that affects the uptake of iron) concentrations. It probably also reduces the prevalence of anaemia (lack of haemoglobin), and may reduce the prevalence of iron deficiency anaemia (lack of iron), compared to iodised salt. 

Well‐designed studies that assess the effects of double‐fortified salt within non‐research populations (i.e. real‐life settings), and that measure salt intake, including changes in salt consumption, are needed. 

What is iron deficiency?

Almost two billion people experience a deficiency in a vitamin or mineral (or both), with women and children in resource‐limited settings most frequently affected. Iron‐related deficiencies are among the most common deficiencies in the world and have important short‐ and long‐term health consequences. Interventions to provide iron frequently include iron supplementation, including iron tablets, powders, or syrups. However, these have known barriers, and food fortification strategies may be attractive alternatives. Salt is one of few universally consumed food vehicles. Iodised salt is fortified to provide 100% of a person's iodine requirements and is highly effective. Double‐fortified salt was developed to provide 30% of a person's daily dietary iron requirement and 100% of their iodine requirement. In some resource‐limited settings, where iron‐related deficiencies are a common problem, there has been interest in making double‐fortified salt more available to the public. This calls for further understanding of the effect of double‐fortified salt on related outcomes. 

What did we want to find out?

If double‐fortified salt is better than salt fortified with iodine alone for improving measures of iron and iodine‐related nutrition, in particular: 

‐ haemoglobin concentration;

‐ urinary iodine concentration;

‐ blood pressure;

‐ ferritin concentration;

‐ transferrin receptor concentration;

‐ prevalence of anaemia;

‐ prevalence of iron deficiency anaemia.

What did we do?

We looked for studies that provided double‐fortified salt to one group of participants and iodised salt to another. We compared their results, and rated our confidence in the evidence, based on factors such as study methods and sample size. 

What did we find?

We identified 18 studies, involving over 8800 individuals from five countries; 13 studies were conducted in India. In 13 studies the intervention lasted between 6 and 12 months; in two studies it lasted 3 months, and in single studies it lasted for 18 months, 24 months, or the duration was unclear. Nine studies were conducted in children and adolescents (5 to 17 years), four in adults (18 years and older), and five included multiple age groups. All studies compared double‐fortified salt to iodised salt. Most studies were funded by non‐profit organisations, university grants or academic institutes. In four studies, double‐fortified salt was provided by a commercial organisation, and in three studies the funding source unclear. 

Compared to iodised salt, double‐fortified salt may improve haemoglobin concentration and body iron stores slightly, and probably reduces the prevalence of anaemia by 21%. However, double‐fortified salt may also reduce urinary iodine concentration compared to iodised salt and may make little or no difference in ferritin and transferrin receptor concentration. Double‐fortified salt may reduce the prevalence of iron deficiency anaemia by 65%, compared to iodised salt, although this conclusion is uncertain because of some problems with the way the studies were conducted. Very few studies measured zinc protoporphyrin concentration, adverse effects, prevalence of goitre and salt intake. One study measured serum iron concentration. 

No studies measured blood pressure or hepcidin concentration. 

What are the limitations of the evidence? 

We have relatively low confidence in the evidence for the outcomes: haemoglobin, urinary iodine, ferritin, and transferrin receptor concentration, and prevalence of iron deficiency anaemia. Not all studies provided data about all outcomes of interest; studies delivered the intervention differently; and studies were small, both in number and size.

For the prevalence of anaemia, we are moderately confident in the evidence because studies used different ways of delivering the intervention.

Care should be taken in interpreting our findings in relation to public health policy and programmes. Most studies were conducted in monitored research settings and double‐fortified salt was provided without an added cost. We are unsure if the effect we observed would be the same in real‐life (i.e. non‐research population), where purchasing double‐fortified salt could increase the cost. More studies looking at the effect of double‐fortified salt within real‐life settings are needed to understand the true effects of double‐fortified salt with greater certainty. Given the changing guidelines for salt intake, future studies should measure salt intake to understand if double‐fortified salt should be considered to prevent anaemia at the population level and how to integrate double‐fortified salt into the supply chain.

How up to date is this evidence?

The evidence is up to date to April 2021.

Summary of findings

Summary of findings 1. Fortification of salt with iron and iodine versus fortification of salt with iodine for improving iron and iodine status.

Fortification of salt with iron and iodine versus fortification of salt with iodine for improving iron and iodine status
Patient or population: participants were of any age or sex and from any country, regardless of baseline iron and iodine status Setting: global, in any setting Intervention: double‐fortified salt  Comparison: iodised salt
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with iodised salt	Risk with double‐fortified salt
Haemoglobin concentration (g/dL) Follow‐up range: 6‐24 months	The mean haemoglobin concentration in the control group ranged from 9.2 to 13.0 g/dL.	The mean haemoglobin concentration in the intervention group ranged from 9.9 to 13.4 g/dL, and was on average 0.43 g/dL higher (0.23 higher to 0.63 higher).	‐	4597 (13 RCTs)	⊕⊕⊝⊝ Lowa,b	‐
Urinary iodine concentration (μg/L) Follow‐up range: 7.5‐18 months	The mean urinary iodine concentration in the control group ranged from 187.2 to 516.5 μg/L.	The mean urinary iodine concentration in the intervention group ranged from 136.7 to 419.8 μg/L, and was on average 96.86 μg/L lower (164.99 lower to 28.73 lower).	‐	1594 (7 RCTs)	⊕⊕⊝⊝ Lowb,c	‐
Blood pressure (systolic and diastolic) (mmHg)	‐	‐	‐	‐	‐	No studies measured or reported this outcome
Ferritin concentration (μg/L) Follow‐up range: 6‐10 months	The mean ferritin concentration in the control group ranged from 17.0 to 160.2 μg/L.	The mean ferritin concentration in the intervention group ranged from 24.0 to 83.4 μg/L, and was on average 3.94 μg/L lower (20.65 lower to 12.77 higher).	‐	1419 (5 RCTs)	⊕⊕⊝⊝ Lowb,c	‐
Transferrin receptor concentration (mg/L) Follow‐up range: 6‐10 months	The mean transferrin receptor concentration in the control group ranged from 6.8 to 35.0 mg/L.	The mean transferrin receptor concentration in the intervention group ranged from 5.8 to 11.5 mg/L, and was on average 4.68 mg/L lower (11.67 lower to 2.31 higher).	‐	1256 (5 RCTs)	⊕⊕⊝⊝ Lowb,c	‐
Prevalence of anaemia Follow‐up range: 6‐10 months	Study population		RR 0.79 (0.66 to 0.94)	2593 (8 RCTs)	⊕⊕⊕⊝ Moderated	‐
	360 per 1000	285 per 1000 (238 to 339)
Prevalence of iron deficiency anaemia Follow‐up range: 6‐10 months	Study population		RR 0.35 (0.24 to 0.52)	1209 (5 RCTs)	⊕⊕⊝⊝ Lowd,e	‐
	260 per 1000	91 per 1000 (62 to 135)
*The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; №: number;RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect 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

^aDowngraded one level for risk of bias: risk of bias was considered unclear in half the studies having judgements of 'high risk' for at least one domain. The domains with the highest risk of bias were blinding of participants, personnel outcome assessments, incomplete outcome data and selective reporting.
^bDowngraded one level for inconsistency: heterogeneity (I² value) is between 75% to 100%: considerable heterogeneity.
^cDowngraded one level for imprecision: wide CI crossing the line of no effect.
^dDowngraded one level for inconsistency: heterogeneity (I² value) is between 50% and 90%: substantial heterogeneity.
^eDowngraded one level for imprecision: does not meet optimal information size (OIS) threshold (few studies and participants).

Background

Iodine and iron are micronutrients which are important for good health. Deficiencies in these micronutrients are of public health concern given their widespread prevalence within populations (particularly in resource‐limited settings), and their association with adverse health and developmental outcomes. The fortification of staple foods is considered one of the most feasible, cost‐effective, and sustainable evidence‐based interventions to address population‐level vitamin and mineral deficiencies (WHO 2006). Salt is a good candidate for food fortification as it is one of few universally consumed food vehicles.

Description of the condition

Iron‐related deficiencies

Although iron can be found in different foods, most dietary staples consumed in low‐ and middle‐income countries (LMICs) contain only small amounts of iron or lack bioavailable iron. This is due to the concurrent consumption of iron absorption inhibitors, such as phytic acid and polyphenols (cereal grains, tea, coffee), and relatively low consumption of iron absorption enhancers, such as ascorbic acid (fruits and vegetables) (Hurrell 2021; Prentice 2017). Common consequences of insufficient dietary iron intake include anaemia, iron deficiency, and iron deficiency anaemia. These conditions can affect the formation and function of organs, including the brain, and can directly and indirectly contribute to multiple pathologies (Beard 2007; Lozoff 2007; Prado 2014).

Anaemia is a condition in which the number of red blood cells is insufficient to meet one’s physiological needs (WHO 2011a). In 2011, the global prevalence of anaemia was highest among children (42.6%), pregnant women (38.2%), and non‐pregnant women (29.0%) (WHO 2015). The prevalence of anaemia varies substantially across regions and countries; the greatest burden is observed in LMICs in Africa (62%) and Southeast Asia (54%). One of the most commonly used screening methods for anaemia in a population is measuring haemoglobin concentration. Haemoglobin concentration alone cannot be used to diagnose iron deficiency; however, employing the age‐ and sex‐specific haemoglobin concentration cut‐offs generated by the World Health Organization (WHO) can allow for the assessment of the prevalence of anaemia (WHO 2011a). Having a haemoglobin concentration that falls below the WHO cut‐offs can reduce the oxygen‐carrying capacity of the blood, and have adverse effects for individuals. Iron deficiency is thought to be the most common cause of anaemia globally, accounting for approximately 50% of cases, although anaemia can also result from other nutritional deficiencies (e.g. folate, vitamin B₁₂, vitamin A), acute and chronic inflammation, and parasitic infections (WHO 2015). In areas with a high prevalence of soil‐transmitted helminth infection (e.g. hookworm, whipworm), preventative anthelminthic treatment (i.e. deworming) is a commonly implemented public health intervention to reduce the burden of infections (WHO 2017). In such settings, deworming alone may prevent anaemia (INACG 1998).

Iron deficiency is of particular concern, as it can result in several adverse effects across the life course, including premature birth, low birthweight, and increased mortality during the perinatal period; delayed psychomotor development during infancy; and reduced work capacity in adulthood (IOM 2001). To assess iron deficiency, iron stores in the body are evaluated. The body primarily stores iron as ferritin. Small amounts of ferritin are secreted into plasma, such that the concentration of ferritin in serum is positively correlated to total body iron stores in the absence of inflammation (WHO 2011b). Transferrin receptor is another measure of iron status and is less influenced by inflammation. The concentration of transferrin receptor in serum mostly comes from developing red blood cells; thus, it reflects the intensity of erythropoiesis (the process by which red blood cells are produced) and the demand for iron. The concentration of transferrin receptor rises in iron deficiency anaemia, making it a marker of the severity of iron insufficiency when iron stores have been exhausted (WHO 2014a). Other possible measures of iron status include serum iron (indicator of iron bound to transferrin in the blood), hepcidin (regulator of iron absorption from the gut), body iron stores (ratio of transferrin receptor to ferritin, indicating body iron status), and zinc protoporphyrin (indicator of lack of iron to develop red blood cells) (WHO 2007a).

Iodine deficiency

Since iodine is found in relatively small amounts within the diet, humans must frequently consume an additional iodine source. Iodine is a key component in hormones produced by the thyroid gland, which are essential to the regulation of vital body functions. If one’s physiological requirements are not met, preventable functional and developmental abnormalities can occur due to inadequate thyroid hormone production (Zimmermann 2015). Collectively termed iodine‐deficiency disorders, adverse conditions associated with severe iodine deficiency include impaired cognitive development in children, and goitre at any age (IOM 2006). During pregnancy, maternal iodine deficiency can impair the development of the foetal brain, since normal amounts of maternal thyroid hormones are required for neuronal migration and myelination (Zimmermann 2016). Severe iodine deficiency in utero causes congenital iodine deficiency syndrome, a condition characterised by substantial intellectual disability, amongst other factors.

Iodine deficiency is considered the most common cause of preventable mental impairment worldwide (Zimmermann 2015). A worldwide assessment of the prevalence of iodine deficiency has not been determined from data collected within the last 15 years. However, among the 194 WHO countries, iodine deficiency and excessive intake are geographically dispersed; 59% are considered to have optimal iodine intake, 12% insufficient iodine intake, and 7% excess iodine intake, with the iodine intake for 22% of countries unknown (Iodine Global Network 2019). A widely implemented strategy to provide iodine has been the international iodisation of salt, and it is estimated that 86% of the population has access to iodised salt (IS) (UNICEF 2008). IS is considered commonly available in the Americas, Europe, and Southeast Asia; however, there is somewhat less coverage in Africa. The recommended indicator for measuring iodine status is urinary iodine concentration, as excess iodine is expelled from the body when one is in an iodine‐sufficient state. Urinary iodine concentration rises when one is iodine sufficient (WHO 2007b). The prevalence of goitre can also be used to reflect chronic iodine deficiency, which is the most common cause of goitre (WHO 2014b).

Description of the intervention

In settings with the greatest burden of iron deficiency, people can rarely afford iron‐rich diets or foods with highly bioavailable iron, and thus the WHO recommends the implementation of iron supplementation programmes for children, pregnant women, and non‐pregnant women (WHO 2012a; WHO 2016a; WHO 2016b). However, iron supplementation programmes (e.g. iron tablets, powders, syrups) can be hindered by factors, such as problems with access to supplies and supply availability, making food fortification strategies an attractive alternative. The iodisation of salt is a simple, cost‐effective way to improve iodine intake, and IS has been recommended by the WHO since the mid 1990s (WHO 1996). Following the relative success of IS, a salt fortified with both iodine and iron — commonly referred to as double‐fortified salt (DFS) — has been developed.

Within DFS formulations, salt is fortified to provide 100% of one’s daily dietary iodine requirement (consistent with IS) and approximately 30% of one’s daily dietary iron requirement. This typically equates to 10 mg of iron compound per day, or approximately 3 mg of elemental iron (Horton 2011). The extent to which one’s dietary iron requirement could be improved in consuming a food fortified with DFS varies depending on life stage and gender, as the food consumption of children and adults differs. Furthermore, the average daily intake amount of elemental iron sufficient to meet an individual’s daily iron needs, known as the recommended dietary allowance (RDA), varies by gender and across the life stages, as follows.

1. Adult males: 8 mg/d (37.5% of RDA for iron provided by DFS, assuming 10 g/d salt intake)

2. Women of reproductive age: 18 mg/d (16.6% of RDA for iron provided by DFS, assuming 10 g/d salt intake)

3. Pregnant women: 27 mg/d (11.1% of RDA for iron provided by DFS, assuming 10 g/d salt intake)

4. Children under three years of age: 7 mg/d (cannot assume 10 g/d salt intake) (IOM 2006)

The life stages considered as being at the greatest risk of iron deficiency include reproductive age and pregnancy for women, and younger than three years of age for children.

The different formulations of DFS that exist vary by the iron compound added (e.g. ferrous fumarate, ferrous sulphate, micronised ferric pyrophosphate); use of encapsulation technology (e.g. lipid‐coated iron particles); inclusion of additives (e.g. stabilisers, colourising agents, absorption promoters); and technology used to produce and blend the fortified salt (e.g. dry blend, fluidised‐bed agglomeration) (Shields 2021). There are challenges given the complexity of chemical interactions within foods, as well as related to maintaining the chemical integrity of the fortificants (i.e. iron and iodine compounds) themselves. It is important to note that ferrous iron compounds can react with iodine compounds, resulting in the sublimation of iodine (Diosady 2002). Additionally, different iron compounds have varying stability under normal conditions and bioavailability (Kraemer 2007). However, current forms of DFS are generally considered to be stable under ideal conditions (Baxter 2015).

How the intervention might work

The goal of food fortification strategies is to improve individuals' micronutrient status by increasing the amount of micronutrients available within the diet. By fortifying a food staple, all members of a population have the potential for exposure to the food with an increased micronutrient content, regardless of their potential to benefit. Given that a small dose of a micronutrient is typically added to a food within fortification strategies, adverse effects are infrequent; however, the risk of excessive intake when multiple strategies are employed and in areas with widespread infection and inflammation (i.e. malaria endemic areas) is important and should be monitored (Hurrell 2021). Moreover, a key consideration of using salt as a vehicle for fortification (with any nutrient), is the alignment with WHO salt‐reduction efforts. As such, DFS should not be used to justify or encourage an increase in salt intake to the public. 

The provision of DFS would be expected to improve iron status by providing more iron within the diet. Mechanistically, iron absorbed within the gastrointestinal tract would be expected to increase haemoglobin concentration and iron stores (e.g. increased ferritin levels and normal transferrin expression). Adverse effects associated with too much iron consumption can include gastrointestinal side effects (e.g. constipation, nausea, vomiting, and diarrhoea) (IOM 2006). IS would not be expected to have an effect on measures of iron status, as it does not provide iron. DFS would be expected to improve iodine status in a comparable manner to IS, as it provides more iodine within the diet. Mechanistically, iodine absorbed within the gastrointestinal tract would be taken up by the thyroid. More iodine is taken up by the thyroid in iodine‐deficient states than sufficient states, with any excess iodine being excreted in the urine (Zimmermann 2016). Excess iodine intake is generally tolerated well, although acute effects of iodine poisoning can include abdominal pain and gastrointestinal side effects (IOM 2006). Variation in the concentration or form (or both) of iron or iodine compounds, and use of additional additives, may affect any observed change in iron and iodine status from DFS. In settings where iron intake is insufficient, DFS would ideally be introduced within the food system instead of IS. Therefore, it is important to fully understand the potential effect of DFS on both iron and iodine status.

Why it is important to do this review

Food fortification strategies to improve population health are of topical interest in the field of nutrition, and there are existing Cochrane Review protocols and Cochrane Reviews investigating the effect of such strategies (Das 2019; Garcia‐Casal 2018; Hombali 2019; Santos 2019; Self 2012). Given the widespread success and feasibility of salt as a vehicle for iodine fortification, and its impact on iodine status, salt fortified with both iron and iodine could further improve iron status. In populations where fortification of other staple foods has not been found to achieve desired coverage, a cost‐benefit assessment of DFS has suggested that DFS could be a good alternative for improving iron status (Horton 2011). The extent to which DFS improves iron and iodine status within individual studies conducted longitudinally appears somewhat variable, yet its use is documented as being scaled up in several provinces in India. This has likely been encouraged by the implementation of legislation mandating the use of DFS within the midday meal programme in schools in 2011 (Government of India 2011), and many forms of DFS have become available on the Indian market. However, there is ongoing concern around excessive salt consumption (WHO 2012b), so the appropriateness of an intervention of this nature could be called into question. Understanding the potential for adverse effects will be important in ensuring the safety of any DFS‐related fortification programmes.

Although several clinical studies have been conducted, there has been no published review on the effect of salt fortified with iron and iodine on multiple measures of both iron and iodine status, as well as safety, compared to IS. Two systematic reviews with meta‐analyses on the effects of salt fortified with iron and iodine, compared to iodine alone, on haemoglobin concentration and prevalence of anaemia and iron deficiency anaemia have been published recently (Larson 2021; Ramírez‐Luzuriaga 2018). However, Ramírez‐Luzuriaga 2018, did not include outcome measures specific to iron status (e.g. ferritin and transferrin receptor concentration), iodine status (e.g. urinary iodine concentration), or safety (e.g. salt intake, blood pressure); they also did not include sources of grey literature. The review by Larson 2021 was comprehensive, although it did not include safety outcomes and some outcome measures specific to iron status (e.g. transferrin receptor, serum iron); they also included three studies with salt fortified with iron only (i.e. no iodine). Neither review employed the GRADE approach to assess the certainty and quality of the evidence for outcomes. Relative to Larson 2021, we have identified one additional study and four ongoing studies of DFS compared to IS. In undertaking the present review of salt fortified with iron and iodine, we aimed to inform these gaps in the evidence, which is particularly important given the existing scale‐up efforts around DFS and ongoing concerns about salt intake.

Objectives

To assess the effect of double‐fortified salt (DFS) compared to iodised salt (IS) on measures of iron and iodine status in all age groups.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs), with randomisation at either the individual‐ or cluster‐level (i.e. cluster‐randomised trials; cRCTs), and quasi‐randomised trials. A quasi‐randomised trial is one in which participants are allocated to different arms of the study (to receive the study intervention, or placebo, for example) using a method of allocation that is not truly random, such as by birth date or gender.

We anticipated that few RCTs would report on certain outcomes of interest (e.g. blood pressure, salt intake); hence, we decided to include studies that delivered double‐fortified salt (DFS) without employing proper randomisation techniques (studies that measured outcomes before and after the implementation of an intervention to a group that received the intervention and a control group that did not, i.e. controlled before‐after (CBA) studies).

Types of participants

Eligible participants were of any age or sex and from any country, regardless of baseline iron and iodine status.

Types of interventions

Experimental intervention

The experimental intervention was salt fortified with iron and iodine (DFS), provided to participants for use within the context of the study, irrespective of iron and iodine compound and concentration, frequency of use, or duration of use.

Comparator intervention

The comparator intervention was salt fortified with iodine only (IS), provided to participants for use within the context of the study, irrespective of iodine compound and concentration, frequency of use, or duration of use.

Types of outcome measures

We considered clinical and biochemical outcomes across all populations. All outcomes measured were determined by study authors. To be included in the review, outcomes must have been measured at the study endpoint. We placed no restrictions on duration of exposure prior to the endpoint, as we investigated this further in a subgroup analysis (see Subgroup analysis and investigation of heterogeneity).

Primary outcomes

1. Haemoglobin concentration (g/dL)

2. Urinary iodine concentration (μg/L)

3. Blood pressure (systolic and diastolic) (mmHg)

Secondary outcomes

1. Ferritin concentration (μg/L)

2. Transferrin receptor concentration (mg/L)

3. Prevalence of anaemia (haemoglobin concentrations below a cut‐off, taking into account age, altitude and smoking, when applicable)

4. Prevalence of iron deficiency anaemia (defined by the presence of anaemia with iron deficiency)

5. Serum iron concentration (μg/dL)

6. Hepcidin concentration (ng/mL)

7. Body iron stores (mg/kg)

8. Zinc protoporphyrin (μmol/mol of haemoglobin)

9. Salt intake (g/d)

10. Prevalence of adverse effects (including constipation, nausea, vomiting, heartburn or diarrhoea)

11. Prevalence of goitre

Search methods for identification of studies

We first ran searches for this review in September 2019, and ran top‐up searches in April 2021. We did not limit the searches by date or language, nor did we apply a study methods filter.

Electronic searches

We searched the following sources from inception onwards.

1. Cochrane Central Register of Controlled Trials (CENTRAL; 2021, Issue 3) in the Cochrane Library, which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialised Register (searched 29 April 2021)

2. MEDLINE Ovid (1946 to April Week 3 2021)

3. MEDLINE In‐Process & Other Non‐Indexed Citations Ovid (1946 to 28 April 2021)

4. MEDLINE EPub Ahead of Print Ovid (searched 29 April 2021)

5. Embase Ovid (1974 to 23 April 2021)

6. Web of Science Core Collection Clarivate (Science Citation Index ‐ Expanded, Social Science Citation Index, Conference Proceedings Citation Index ‐ Science, Conference Proceedings CItation Index ‐ Social Sciences & Humanities, 1970 to 28 April 2021)

7. SCOPUS Elsevier (searched 29 April 2021)

8. WHO Library Database (WHOLIS; kohahq.searo.who.int; searched 29 April 2021)

9. Epistemonikos (epistemonikos.org; searched 29 April 2021)

10. Global Index Medicus (includes regional indexes (AIM: African Index Medicus; LILACS: Latin American and Caribbean Health Sciences Literature; IMEMR: Index Medicus for the Eastern Mediterranean Region; IMSEAR: Index Medicus for South‐East Asia Region; SciELO: Scientific Electronic Library Online; and WPRIM: Western Pacific Region Index Medicus); search.bvsalud.org/ghl; searched 29 April 2021)

11. ClinicalTrials.gov (clinicaltrials.gov; searched 29 April 2021)

12. World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; who.int/ictrp/en; searched 18 September 2019; access attempted in April 2021, but site was unresponsive)

The Cochrane Information Specialist for Developmental Psychosocial and Learning Problems assisted with the development of the search strategy and running top‐up searches. The exact search strategies for each database are reported in Appendix 1. 

A search for retractions of included studies was run in MEDLINE and Embase on 22 July 2021. No corrections or retractions were found.

Searching other resources

We reviewed ancillary publications in the reference lists of included studies, and contacted the authors of included studies to identify any additional unpublished or ongoing studies. We also handsearched the following sources for other potentially eligible studies using keywords such as "double fortified salt" and "dual fortified salt".

1. Google Scholar (scholar.google.co.uk)

2. International Initiative for Impact Evaluation (www.3ieimpact.org)

3. Global Alliance for Improved Nutrition (www.gainhealth.org)

4. Nutrition International (www.nutritionintl.org)

5. Fortified Food Initiative (www.ffinetwork.org)

Two review authors first searched the sources on 26 October 2019 (J‐ABB, MK), and then subsequently on 4 May 2021 (BC, J‐ABB).

Data collection and analysis

Selection of studies

Two review authors (J‐ABB, MK) independently screened the titles and abstracts of all records retrieved from literature searches (up to September 2019) to determine which studies should be assessed. Subsequently, two review authors (BC, J‐ABB) independently screened the titles and abstracts of top‐up searches (up to April 2021) to further determine which studies should be assessed. Full‐text reports of all potentially relevant records were obtained. We resolved any disagreements by discussion or, if necessary, upon review by a third author (SHZ or ZAB). For each full‐text report, we kept a record of all eligibility decisions, with brief details on the study design, participant and intervention characteristics. Selection of studies and documentation of reasons for exclusion were managed using Covidence 2021, a web‐based platform for systematic review processes. We generated an adapted PRISMA flow diagram to show the selection process (Moher 2009).

Data extraction and management

From the reports that met the eligibility criteria (see: Criteria for considering studies for this review), three review authors (BC, J‐ABB, MK) independently abstracted the following information using a standardised data extraction spreadsheet; the data collection form was piloted by J‐ABB and MK on two included studies, before being finalised for use in this review.

1. Study characteristics (author, publication year, country, study population (age, sex, health status), study duration, study design, study year)

2. Details about the intervention and comparator (iron compound, iron concentration, iodine concentration, co‐interventions (deworming))

3. Outcome measures and associated outcome data related to iron and iodine status

See Appendix 2 for more details on extraction. 

J‐ABB and MK abstracted full‐text reports identified in the literature searches up to September 2019 and J‐ABB and BC abstracted the full‐text reports identified in the top‐up searches up to April 2021. We resolved any disagreements by discussion, or in consultation with a third review author (SHZ or ZAB), if required. In the case of multiple publications, companion documents or reports on a single study, we reviewed all available data and used the most complete data set aggregated across all sources. In cases where study data were presented in a format inconsistent with that required for the meta‐analysis (e.g. child and adult data reported together), we contacted the respective study author and requested the disaggregated data.

Main comparisons

We compared the use of DFS to the use of IS on the indicated primary and secondary outcomes of interest. We also included studies that combined the provision of DFS or IS with concurrent deworming treatment. We assessed the effect of concurrent deworming as a subgroup analysis (see Subgroup analysis and investigation of heterogeneity).

Assessment of risk of bias in included studies

We assessed risk of bias in RCTs and cRCTs using the Cochrane RoB 1 tool (Higgins 2011), and the Cochrane Effective Practice and Organisation of Care (EPOC) guidelines for assessing CBA studies (EPOC 2017). Using the criteria set out in Appendix 3, three review authors (J‐ABB, MK, BC) independently assigned the following items a rating of low, high or unclear risk of bias in each domain: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other potential sources of bias. For this last domain, ‘other potential sources of bias’, we assessed all studies for risk of bias due to similar baseline characteristics, similar baseline outcome assessments, declarations of conflicts of interest and funding sources. We also assessed cRCTs for recruitment bias, baseline imbalance, loss of clusters, incorrect analysis, and comparability with individually randomised studies. Where we suspected reporting bias, we attempted to contact study authors to ask them to provide missing outcome data. We resolved disagreements by discussion, or in consultation with a third review author (SHZ or ZAB), if required. 

Overall risk of bias

To reach an overall judgement of risk of bias for each study, two review authors (BC, J‐ABB) independently assigned a rating of one of the following.

1. Low (i.e. we judged the study to be at low risk of bias for all domains)

2. High (i.e. we judged the study to be at high risk of bias in at least one domain for this result, or we judged the study to be unclear for multiple domains in a way that substantially lowers confidence)

3. Unclear (i.e. we judged the study to be unclear in at least one domain, but not at high risk of bias for any domain)

Measures of treatment effect

Continuous data

We expressed continuous outcome data as mean differences (MDs). When two or more eligible studies were available to conduct the comparison of continuous outcomes, we reported measures with 95% confidence intervals (CIs).

Dichotomous data

We expressed dichotomous data as proportions. When two or more eligible studies were available to conduct the comparison of a dichotomous outcome, we expressed the data as risk ratios (RRs) with 95% CIs.

Unit of analysis issues

Cluster‐randomised trials

We anticipated that we would identify both cluster‐randomised and individually randomised studies for this review. If there was little heterogeneity between such studies, we combined the results from both designs within a single meta‐analysis. If there was notable heterogeneity, we attempted to reanalyse cRCTs that were not appropriately adjusted for potential clustering of participants within clusters in their analyses. We did this by incorporating an intraclass correlation coefficient (ICC) to account for the design effect. If the ICC was not reported for a study, we attempted to estimate it from the available data from each study (Ahn 2012), or we applied an ICC value that had been reported elsewhere in similar research, and conducted sensitivity analyses on higher and lower ICC values.

Studies with two or more treatment groups

If we identified a study with more than two intervention groups, we included only the relevant groups (i.e. we generated a single pair‐wise comparison). If a study had multiple relevant groups, we divided the control group by the number of groups relevant to the review (i.e. divided the events and total population of the control group) to avoid double‐counting the participants in the control group. Although this approach does not take into account possible correlations arising from including the same set of participants in multiple comparisons, it does take all relevant groups into consideration (Higgins 2021).

Dealing with missing data

In the case that data from a study were missing within a publication, we contacted the study author and attempted to obtain the missing information. We investigated attrition rates (e.g. dropouts, losses to follow‐up, withdrawals), and critically appraised issues concerning missing data and imputation methods (e.g. carrying the last observation forward). Where a study did not report a mean and standard deviation (SD) for outcomes, and we were unable to obtain the relevant information from the study authors, we calculated these values using the median, range and sample size, according to the formula described in Hozo 2005. We assessed the possible impact of imputation on the meta‐analyses by performing sensitivity analyses (see Sensitivity analysis). We reported all imputed values by outcome and study. We reported any missing data in the risk of bias tables. We analysed all data using intention‐to‐treat.

Assessment of heterogeneity

We examined clinical heterogeneity by looking at the similarity between the participants, interventions and outcomes of the identified studies. We investigated methodological heterogeneity by examining the methodological characteristics and risk of bias of the studies. We assessed statistical heterogeneity visually, by inspecting the forest plots derived from the meta‐analyses (e.g. to assess the size and direction of the treatment effect across studies on outcome of interest). We also considered the I² statistic for each meta‐analysis, which allows for the quantification of inconsistencies across studies. We used the following interpretive categories for I²: 0% to 40% heterogeneity might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; and 75% to 100% considerable heterogeneity (Higgins 2011). As we used a random‐effects model, we reported tau² as a measure of between‐study variance. If we identified moderate or substantial heterogeneity, we assessed the possible reasons by examining the characteristics of the individual studies. We advise caution around the interpretation of results where there are high levels of unexplained heterogeneity.

Assessment of reporting biases

Where we suspected reporting bias, we attempted to contact study authors and ask them to provide missing outcome data. If there were more than 10 studies reporting the same outcome of interest, we examined possible publication bias by generating funnel plots in Review Manager 5 (Review Manager 2020), and examined them visually for asymmetry. We acknowledge that asymmetric funnel plots are not necessarily caused by publication bias. Where we pooled studies in a meta‐analysis, we ordered the studies by weight, such that upon visual examination of the forest plots we could assess whether the results from smaller and larger studies were similar, or if there were any apparent differences (i.e. we checked that the effect size was similar across smaller and larger studies).

Data synthesis

We conducted meta‐analyses to provide an overall estimate of the treatment effect when more than one study examined the same intervention on an outcome of interest, provided that the studies used similar methodology, assessment methods, and participant populations. However, because of the differences between RCT and CBA study designs, we did not pool the results from RCTs and CBA studies in the meta‐analyses. We only conducted meta‐analyses for data from RCTs. We summarised data from CBA studies in a separate table (Centeno Tablante 2019). If there were conflicting data on outcomes of interest from RCTs and CBA studies, we relied on the evidence from RCTs to estimate the effect of the treatment and draw conclusions.

We summarised data for continuous outcomes (haemoglobin, ferritin, transferrin receptor, and urinary iodine concentration, blood pressure, salt intake) using generic inverse variance methods and a random‐effects model. For dichotomous outcomes (prevalence of anaemia and iron deficiency anaemia), we applied Mantel‐Haenszel methods and a random‐effects model. We conducted all analyses using Review Manager 5 (Review Manager 2020) and RevMan Web (RevMan Web 2020). If a meta‐analysis was not appropriate, we provided a narrative description of the relevant study findings. We applied all statistical analyses according to Cochrane guidelines (Deeks 2021). We assessed whether the treatment effect varies across subgroups of clinical importance (see Subgroup analysis and investigation of heterogeneity).

Subgroup analysis and investigation of heterogeneity

We anticipated that certain characteristics could introduce clinical heterogeneity. We explored heterogeneity by conducting analyses using the following subgroups.

1. Life stage (children (< 5 years old, 5 to 17 years old), adult (≥ 18 years old, separated by male and female), pregnancy (any age)). These groups were considered significant because the effect of DFS compared to IS for different life stages could differ given the amount of iron actually consumed compared to the recommended dietary allowance (RDA).

2. Iron compound used in salt (e.g. ferrous sulphate, ferrous fumarate, micronised ferric pyrophosphate). The different forms of iron used within the DFS could have affected iron outcomes because iron compounds have different bioavailabilities and environmental stability.

3. Study design (e.g. RCT, cRCT). The rationale for this analysis was that study design, particularly appropriate use of randomisation, could have affected study quality.

4. Study duration (0 to 3 months, 3 to < 6 months, 6 to < 9 months, 9 to < 12 months, 12 to < 24 months, 24 to 36 months). The study duration could have affected the magnitude of change for DFS compared to IS for outcomes of interest.

5. Anaemia status at enrolment (anaemic, non‐anaemic). Whether participants were anaemic or not at the start of the study could have affected the magnitude of change in outcomes of interest among those who receive DFS, as anaemic populations are more likely to benefit from DFS.

6. Use of concurrent deworming (deworming, no deworming). In areas with high levels of parasites, worms are a common cause of infection and could have impaired the nutritional status of the people they infected by affecting their ability to absorb ingested nutrients.

We only conducted subgroup analyses for outcomes when there were two or more studies, and when there were outcome data for at least two of the identified subgroup categorisations.

Sensitivity analysis

We performed sensitivity analyses to examine whether certain factors might have affected the effect size for outcomes of interest, by removing the following studies from the analyses.

1. Studies identified as being at high risk of bias in any of the following domains: allocation concealment; blinding of participants and study personnel; and blinding of outcome assessors

2. Studies for which we adjusted for clustering

Summary of findings and assessment of the certainty of the evidence

We provided a summary of the evidence across studies in the format of a summary of findings table, which we prepared using GRADEpro GDT software (GRADEpro GDT). In this table, we presented an estimate of the relative effect of outcomes of interest, along with the number of participants and studies contributing to the generation of the estimate for the outcome. We presented summary information on the population, setting and intervention, for the comparison between DFS and IS. We included the following outcomes, assessed at the study end point.

1. Haemoglobin concentration

2. Urinary iodine concentration

3. Blood pressure (systolic and diastolic)

4. Ferritin concentration

5. Transferrin receptor concentration

6. Prevalence of anaemia

7. Prevalence of iron deficiency anaemia

Two review authors (BC, J‐ABB) used the GRADE approach to assign the evidence for each outcome a rating of either very low, low, moderate, or high certainty (Schünemann 2013). In accordance with the GRADE recommendations, we reduced our assessment of the certainty of evidence, based on the following five factors: limitations in the design and implementation of studies, suggesting a high likelihood of bias; indirectness of evidence (e.g. intervention, control, outcomes, population); unexplained heterogeneity or inconsistency within the results (e.g. problems relating to subgroups); imprecision of the effect estimates (e.g. wide CIs); and risk of publication bias.

Results

Description of studies

See Characteristics of included studies  tables, Characteristics of excluded studies tables, Characteristics of studies awaiting classification tables, and Characteristics of ongoing studies tables.

Results of the search

The database searches produced 2796 records, while handsearching revealed an additional 636 records. Following removal of 1086 duplicates, 2346 records were screened at the title and abstract stage, which identified 91 records for full‐text review. Of these, 18 unique studies (47 reports) were eligible for inclusion in this review. One study, Banerjee 2016, presented data in a format that precluded its inclusion in analyses, but we have provided a narrative synthesis of the characteristics below. One study is awaiting classification (Jadhav 2019), and four studies are ongoing (CTRI/2019/08/020508; CTRI/2020/11/028936; NCT04404751; RIDIE‐STUDY‐ID‐58f6eeb45c050); these studies did not contribute data to the analyses. See Figure 1.

1.

Study flow diagram.

Included studies

Participants and settings

Thirteen studies were conducted in India, two in Morocco, and one each in Cote d'Ivoire, Ghana, and Sri Lanka (Table 2). Most studies were conducted in the household, with the exception of Bathla 2016, Krämer 2018 and Sivakumar 2001 (conducted within schools), and Kaur 2000 (conducted in a university hostel). A total of 8800 participants were included in this review at various life stages. Nine studies conducted double‐fortified salt (DFS) studies exclusively in school‐aged children and adolescents aged 5 to 17 years, four studies were conducted in adults aged 18 years and older (including pregnant and lactating women), and five studies included multiple age groups (Asibey‐Berko 2007; Banerjee 2016; Bathla 2017; Brahmam 2000; Vinodkumar 2007). See Characteristics of included studies tables for further details of the included studies.

1. Location of included studies.

Study	Country
Andersson 2008	India
Asibey‐Berko 2007	Ghana
Banerjee 2016	India
Bathla 2016	India
Bathla 2017	India
Brahmam 2000	India
Haas 2014	India
Jayatissa 2012	Sri Lanka
Joshi 2014	India
Kaur 2000	India
Krämer 2018	India
Rajagopalan 2000	India
Reddy 2016	India
Sivakumar 2001	India
Vinodkumar 2007	India
Wegmüller 2006	Cote d'Ivoire
Zimmermann 2003	Morocco
Zimmermann 2004	Morocco

Study design

Of the 18 studies, seven were cluster‐randomised controlled trials (cRCTs) (Banerjee 2016; Brahmam 2000; Jayatissa 2012; Krämer 2018; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007), four were controlled before‐after (CBA) studies (Bathla 2016; Bathla 2017; Joshi 2014; Kaur 2000), and the remaining seven studies were individually randomised controlled trials (RCTs). Only two studies were considered effectiveness studies (Banerjee 2016; Krämer 2018), because the intervention was deployed as part of a broader programme to a free‐living population. The remainder of the assessed were efficacy studies, as the specific intervention was delivered under ideal research conditions. Within the latter, the majority of studies were two‐arm studies that assessed DFS versus control (iodised salt; IS), with the exception of Andersson 2008, which was a three‐arm study reporting the effects from two intervention groups compared to the same control group. In our analysis, both intervention groups in this study were included as separate comparisons and the control group N was halved in pooled effect size calculations in order to avoid double‐counting participants in the control group.

Types of interventions

Composition of double‐fortified salt

The composition of DFS varied across studies, by iron compound (ferrous fumarate, ferrous sulphate, ferric pyrophosphate), and the use of encapsulation, stabilising compounds, and/or iron absorption promoters. While all studies used potassium iodate, the majority of studies (11/18) provided iron in the form of ferrous sulphate, with sodium hexametaphosphate as a stabiliser, while two studies provided micronised ferric pyrophosphate, two studies provided ferrous fumarate with encapsulation additives, and one study provided ferrous fumarate. One study did not report the composition of DFS (Kaur 2000). Andersson 2008 provided DFS through two intervention arms: one with micronised ferric pyrophosphate and another with encapsulated ferrous fumarate. DFS formulations ranged from 0.86 to 3 mg iron/1 g salt and 20 to 50 µg iodine/1 g salt.

Co‐interventions

With regards to co‐interventions, seven studies provided albendazole treatment (400 mg) (deworming) (Andersson 2008; Haas 2014; Jayatissa 2012; Kaur 2000; Rajagopalan 2000; Vinodkumar 2007; Wegmüller 2006), with Andersson 2008 also providing vitamin A (200,000 IU). Three studies provided nutrition education (Bathla 2017; Joshi 2014; Reddy 2016), with Reddy 2016 also providing additional iron‐folic acid supplementation because participants were pregnant women.

Duration of fortification

The majority of studies had an intervention duration of 6 to 12 months (13/18 studies), with two studies conducting interventions of 3 months in duration (Bathla 2016; Bathla 2017), and two studies conducting interventions of 18 and 24 months in duration respectively (Brahmam 2000; Sivakumar 2001). One study was unclear in duration (Banerjee 2016).

Administration of salt

In the majority of studies (13/18), salt was distributed monthly at the household‐level (Andersson 2008; Asibey‐Berko 2007; Banerjee 2016; Bathla 2017; Brahmam 2000; Haas 2014; Jayatissa 2012; Joshi 2014; Reddy 2016; Vinodkumar 2007; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004), although one also monitored weekly (Asibey‐Berko 2007), and one study did not specify the administration timing (Rajagopalan 2000). Only three studies distributed salts within schools (Bathla 2016; Krämer 2018; Sivakumar 2001), or a university (Kaur 2000), and three for use in the preparation of food (Kaur 2000), or midday meal programmes (Bathla 2016; Krämer 2018).

Adherence to treatment

Compliance to the intervention was infrequently measured or reported, or both, with only six studies quantitatively or qualitatively measuring and reporting compliance to DFS by 3‐day weighted food records or remaining salt in the household (Andersson 2008; Haas 2014; Jayatissa 2012; Krämer 2018; Reddy 2016; Wegmüller 2006). Reddy 2016 noted that compliance in the DFS group was 95.5%, compared to the control group at 79.6%. Andersson 2008 found 17% of households stopped using DFS (encapsulated ferrous fumarate arm) due to the adverse colour changes that occurred in cooked foods.

Funding sources

Most studies (10/18) were funded non‐commercially, through not‐for‐profit organisations, university grants or academic institutes. One study, Banerjee 2016, was funded non‐commercially, but the DFS was provided by a commercial entity. The funding for four studies was unclear (Brahmam 2000; Kaur 2000; Rajagopalan 2000; Sivakumar 2001), and three studies received DFS from a commercial entity, although the study funders were unclear (Bathla 2016; Bathla 2017; Vinodkumar 2007). 

Declarations of interest

No conflicts of interest were reported for four studies (Andersson 2008; Reddy 2016; Zimmermann 2003; Zimmermann 2004). In the study conducted by Haas 2014, one author was employed by the funder (Micronutrient Initiative) and all other authors declared no conflicts of interest. The remaining studies did not provide details as to whether or not there were any conflicts to declare (Asibey‐Berko 2007; Banerjee 2016; Bathla 2016; Bathla 2017; Brahmam 2000; Jayatissa 2012; Joshi 2014; Krämer 2018; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007; Wegmüller 2006).

Comparators

All studies provided IS to the control group.

Outcomes

The majority of outcomes were measured biochemically, through blood or urine samples (i.e. haemoglobin concentration, urinary iodine concentration, ferritin concentration, transferrin receptor concentration, prevalence of anaemia, prevalence of iron deficiency anaemia, serum iron concentration, body iron stores, and zinc protoporphyrin). Other outcomes were self‐reported (i.e. prevalence of adverse effects, salt intake) or assessed through physical examination (i.e. prevalence of goitre).

Of the 17 studies included in the meta‐analysis, all measured and reported haemoglobin concentration at study endpoint. Ten studies measured and reported urinary iodine concentration at study endpoint (Andersson 2008; Bathla 2016; Bathla 2017; Haas 2014; Jayatissa 2012; Joshi 2014; Sivakumar 2001; Vinodkumar 2007; Zimmermann 2003; Zimmermann 2004). No studies measured blood pressure (systolic or diastolic).

Eleven studies measured and reported prevalence of anaemia at study endpoint (Andersson 2008; Asibey‐Berko 2007; Bathla 2016; Bathla 2017; Haas 2014; Jayatissa 2012; Kaur 2000; Krämer 2018; Reddy 2016; Wegmüller 2006; Zimmermann 2004), while six studies measured and reported the prevalence of iron deficiency anaemia at study endpoint (Andersson 2008; Haas 2014; Joshi 2014; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004). Of the RCTs, all used WHO thresholds to define anaemia based on serum haemoglobin concentrations (Andersson 2008; Asibey‐Berko 2007; Haas 2014; Jayatissa 2012; Krämer 2018; Reddy 2016; Zimmermann 2003; Zimmermann 2004). Among the four CBA studies, three used WHO haemoglobin cut‐offs to define anaemia (Bathla 2017; Joshi 2014; Kaur 2000), while Bathla 2016 used an alternate cut‐off to enrol anaemic participants.

Six studies measured and reported serum ferritin concentration at study endpoint (Andersson 2008; Haas 2014; Jayatissa 2012; Kaur 2000; Wegmüller 2006; Zimmermann 2003) and five studies measured transferrin receptor concentration at study endpoint (Andersson 2008; Haas 2014; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004). Four studies measured and reported body iron stores (Andersson 2008; Haas 2014; Wegmüller 2006; Zimmermann 2004) and three studies measured and reported zinc protoporphyrin at study endpoint (Andersson 2008; Zimmermann 2003; Zimmermann 2004). Two studies reported prevalence of goitre (Zimmermann 2003; Zimmermann 2004). Four studies measured and reported salt intake at endline, although this was only for the DFS group; one study measured salt intake for both groups at baseline and endline (Kaur 2000). One study measured serum iron concentration (Kaur 2000). No studies measured hepcidin concentration.

Excluded studies

At full‐text stage, we deemed 23 reports to be irrelevant. We excluded the remaining 13 reports with reasons. Of these, three studies had ineligible study designs (Bhatia 2018; Diosady 2018; Horton 2011), eight studies had ineligible interventions, i.e. they used iron‐only fortified salt or iron supplements instead of DFS (Anonymous 1980; Anonymous 1982; Anonymous 1983; Dodd 1997; Jain 1987; Kumar 2020; Nadiger 1980; Working Group on Fortification 1982), and two studies had ineligible comparators (Nair 2013; Nair 2014). See Characteristics of excluded studies tables for further information.

Studies awaiting classification

One study is awaiting classification, as we are unable to access the full publication (Jadhav 2019).

Ongoing studies

We identified four ongoing studies (CTRI/2019/08/020508; CTRI/2020/11/028936; NCT04404751; RIDIE‐STUDY‐ID‐58f6eeb45c050), of which two are RCTs (CTRI/2020/11/028936; NCT04404751), one a cRCT (RIDIE‐STUDY‐ID‐58f6eeb45c050), and one uses a randomised parallel group design (CTRI/2019/08/020508). Two studies were effectiveness studies, conducted in India (RIDIE‐STUDY‐ID‐58f6eeb45c050) and Tanzania (NCT04404751), while two studies were efficacy studies conducted in India (CTRI/2019/08/020508; CTRI/2020/11/028936). Three studies were conducted exclusively in adults (CTRI/2020/11/028936; NCT04404751; RIDIE‐STUDY‐ID‐58f6eeb45c050), while one study included both children and adults (CTRI/2019/08/020508). Detailed information on the composition and administration of salts was not provided. Studies provided expected measurable outcomes including haemoglobin concentration (NCT04404751; CTRI/2019/08/020508), ferritin concentration (CTRI/2019/08/020508), body iron stores (NCT04404751), prevalence of anaemia (RIDIE‐STUDY‐ID‐58f6eeb45c050), and iron status (CTRI/2020/11/028936).

Risk of bias in included studies

We used standardised domains to assess the risk of bias of included studies (Higgins 2011). We assessed the primary outcomes for risk of bias at the study level. We considered additional domains for risk of bias among the cRCTs and CBA studies. We presented these in the risk of bias tables beneath the Characteristics of included studies tables, and give a summary in Figure 2 and Figure 3. Among the seven RCTs contributing to the meta‐analysis, five studies had unclear risk of bias (Andersson 2008; Haas 2014; Reddy 2016; Zimmermann 2003; Zimmermann 2004), with some concerns in random sequence generation and allocation concealment, as study authors provided insufficient information about the methods used to generate a random component in the sequence generation process or conceal assignment of participants. We assessed two RCTs to have a high risk of bias overall, whereby we noted high risk in at least one or more domain(s) (Asibey‐Berko 2007; Wegmüller 2006). Of the seven cRCTs, we assessed six at high risk of bias overall, with one or more domain(s) judged as high risk (Brahmam 2000; Jayatissa 2012; Krämer 2018; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007), and one at unclear risk of bias (Banerjee 2016), given concerns around allocation and blinding. The four CBA studies had high or unclear risk of bias for most domains (Bathla 2016; Bathla 2017; Joshi 2014; Kaur 2000). 

2.

3.

Allocation

Sequence generation

For random sequence generation for the recruitment of participants among the seven RCTs, we identified the risk of bias to be low for one study that used computer‐generated randomisation (Reddy 2016), and unclear for the remaining six RCTs, as insufficient details were provided or the description led to uncertainty. For the seven cRCTs, three had a low risk of bias (Banerjee 2016; Krämer 2018; Rajagopalan 2000), and we assessed the remaining four to be unclear due to insufficient description of the method used. Among the four CBA studies (Bathla 2016; Bathla 2017; Joshi 2014; Kaur 2000), we assessed all to have a high risk of bias since they were non‐random given the nature of the study design. 

Allocation concealment

We identified allocation concealment among the seven RCTs to be low risk for one study where the randomisation code was kept secret until assessment (Haas 2014), and unclear for the remaining six RCTs, as there was insufficient information about the methods used (Andersson 2008; Asibey‐Berko 2007; Reddy 2016), or the description led to uncertainty (Wegmüller 2006; Zimmermann 2003; Zimmermann 2004). For the seven cRCTs, one had a low risk of bias (Jayatissa 2012); four were unclear because insufficient details were provided (Banerjee 2016; Brahmam 2000; Sivakumar 2001), or the description led to uncertainty (Krämer 2018); and we identified two as high risk of bias because study personnel had knowledge of allocation (Rajagopalan 2000; Vinodkumar 2007). Among the CBA studies, all four had an unclear risk of bias as insufficient details were provided (Bathla 2016; Bathla 2017; Joshi 2014; Kaur 2000).  

Blinding

Blinding of participants and personnel

For blinding of participants and personnel, among the seven RCTs, we assessed six to be at low risk of bias (Andersson 2008; Asibey‐Berko 2007; Haas 2014; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004), and one had an unclear risk of bias due to insufficient details (Reddy 2016). For the seven cRCTs, three were at low risk (Jayatissa 2012; Rajagopalan 2000; Sivakumar 2001), two were unclear because insufficient details were provided (Banerjee 2016; Brahmam 2000), and two were at high risk given incomplete blinding methods (Krämer 2018; Vinodkumar 2007). Among the four CBA studies, two had an unclear risk of bias as insufficient details were provided (Bathla 2016; Bathla 2017), and two had a high risk of bias because they were not blinded (Joshi 2014; Kaur 2000). 

Blinding of outcome assessments

For the blinding of outcome assessments among the seven RCTs, we identified the risk of bias to be low for one study (Andersson 2008), with both the assessors and the households blinded to allocation. We rated the remaining six RCTs as unclear due to there being insufficient information provided to make the assessment. For the seven cRCTs, one had a low risk of bias (Jayatissa 2012), four were unclear due to insufficient information (Banerjee 2016; Krämer 2018; Rajagopalan 2000; Sivakumar 2001), and the remaining two had a high risk of bias since research personnel were not blinded. Among the four CBA studies, two studies were unclear as insufficient information was provided (Bathla 2016; Bathla 2017), and two had a high risk of bias because they were not blinded (Joshi 2014; Kaur 2000). 

Incomplete outcome data

For incomplete outcome data among the seven RCTs, we assessed the risk of bias to be low for six studies because all outcomes were reported and attrition was < 10% (Andersson 2008; Haas 2014; Reddy 2016; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004), and high for one study because of loss of outcome data and high attrition (Asibey‐Berko 2007). For the seven cRCTs, two had a low risk of bias (Banerjee 2016; Jayatissa 2012), one an unclear risk of bias as insufficient information was described (Brahmam 2000), and four had a high risk of bias because of high attrition (Krämer 2018; Sivakumar 2001) and incomplete outcome reporting (Rajagopalan 2000; Vinodkumar 2007). Among the four CBA studies, two had a low risk of bias (Bathla 2017; Kaur 2000), one was unclear because insufficient information was presented (Bathla 2016), and one had a high risk of bias, given the high attrition rate (Joshi 2014).

Selective reporting

For selective reporting among the seven RCTs, we assessed the risk of bias to be low for six studies (Andersson 2008; Haas 2014; Reddy 2016; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004), and high for one (Asibey‐Berko 2007), given selective reporting of outcomes. For the seven cRCTs, three had a low risk of bias (Jayatissa 2012; Krämer 2018; Vinodkumar 2007), two had an unclear risk of bias because it was not possible to discern from the presented data (Banerjee 2016; Rajagopalan 2000), and two had a high risk of bias because of selective reporting (Brahmam 2000; Sivakumar 2001). All CBA studies had a low risk of bias.  

Other potential sources of bias

All studies

Similarilty of baseline characteristics 

For the similarity of baseline characteristics among the seven RCTs, we assessed five to be at low risk of bias, as study arms were balanced (Andersson 2008; Haas 2014; Reddy 2016; Zimmermann 2003; Zimmermann 2004), and two to be at high risk of bias because of significant differences in baseline biomarkers between groups despite randomisation (Asibey‐Berko 2007; Wegmüller 2006). For the seven cRCTs, we assessed one at low risk of bias (Banerjee 2016), two at unclear risk of bias (Brahmam 2000; Sivakumar 2001) and four at high risk of bias for baseline characteristics (Jayatissa 2012; Krämer 2018; Rajagopalan 2000; Vinodkumar 2007). Among the CBA studies, we assessed three at low risk of bias (Bathla 2016; Bathla 2017; Kaur 2000), and one at high risk of bias (Joshi 2014) for baseline characteristics.

Similarity of baseline outcome assessment

We assessed baseline outcome assessment to be at low risk of bias for all RCTs, cRCTs and CBA studies, as measurement of the outcomes did not differ between intervention groups and the assessment of biochemical outcomes could not have been influenced by knowledge of the intervention received.

Declaration of conflicts of interest and funding sources

For declaration of conflicts of interest and funding sources among the seven RCTs, five studies reported this information and we judged them as being at low risk of bias (Andersson 2008; Haas 2014; Reddy 2016; Zimmermann 2003; Zimmermann 2004). The remaining two studies provided incomplete details on conflicts of interest and were judged as unclear risk of bias (Asibey‐Berko 2007; Wegmüller 2006). For the seven cRCTs, three studies were judged as having unclear risk of bias as they did not state their conflicts of interest but did report funding sources (Banerjee 2016; Jayatissa 2012; Krämer 2018). The remaining cRCTS, were judged as high risk of bias as they did not state either conflicts of interest or funding sources (Brahmam 2000; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007). For the four CBA studies, Kaur 2000 was judged as high risk of bias as the authors did not include information about conflicts of interest or funding. The remaining CBA studies were judged as unclear risk of bias, as they only reported sources of funding (Joshi 2014) or some indication of potential conflicts of interest or funding sources (Bathla 2016; Bathla 2017).

Cluster‐RCTs only

Recruitment bias

We assessed all cRCTs as low risk of bias, as individual recruitment after cluster randomisation did not appear to be present. 

Baseline imbalance

Reporting of the baseline comparability of clusters was at low risk of bias for one study (Banerjee 2016); unclear risk of bias in five studies, as there was insufficient information (Brahmam 2000; Jayatissa 2012; Krämer 2018; Rajagopalan 2000; Sivakumar 2001); and high risk of bias in one study where baseline cluster differences were described (Vinodkumar 2007).

Loss of clusters

Attrition at the cluster‐level was judged at low risk of bias for one study (Banerjee 2016) and unclear risk of bias for six studies, as loss of clusters was not disclosed (Brahmam 2000; Jayatissa 2012; Krämer 2018; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007). Though Krämer 2018 reported attrition over 20% for both control and treatment groups at the individual level, this was not reported at the cluster level, and therefore judged at unclear risk of bias.

Incorrect analysis

Among the seven cRCTs, two studies reported adjustment for clustering, and therefore were judged at low risk of bias (Jayatissa 2012; Krämer 2018), and only one reported the ICC (Jayatissa 2012). We assessed four studies to be high risk of bias (Brahmam 2000; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007), as adjustment for clustering was not conducted, and one study at unclear risk of bias as results are presented at the village level (Banerjee 2016).

Comparability with individually randomised studies

Contamination and herd effects did not appear to be present in the cRCTs versus RCTs. This was confirmed through subgroup and sensitivity analyses, and therefore all cRCTs were judged as low risk of bias.

Overall risk of other bias

We assessed five RCTs at low risk of other bias overall (i.e. all 'other' domains were judged as low risk of bias) (Andersson 2008; Haas 2014; Reddy 2016; Zimmermann 2003; Zimmermann 2004). We assessed one cRCT (Banerjee 2016) and two CBA studies (Bathla 2016; Bathla 2017) to be at unclear risk of other bias overall (i.e. we judged the study to be unclear in at least one 'other' domain, but not at high risk of bias for any domain). Finally, we assessed two RCTs (Asibey‐Berko 2007; Wegmüller 2006), six cRCTs (Brahmam 2000; Jayatissa 2012; Krämer 2018; Rajagopalan 2000; Sivakumar 2001; Vinodkumar 2007), and two CBA studies (Kaur 2000; Joshi 2014) to be at high risk of other bias overall (i.e. we judged the study to be at high risk of bias in at least one 'other' domain for this result, or we judged the study to be unclear for multiple 'other' domains in a way that substantially lowers confidence).

Effects of interventions

See: Table 1

Salt fortified with iron and iodine (DFS) versus salt fortified with iodine only (IS)

We adjusted all primary and secondary outcome estimates for clustering, as indicated in the Methods section and forest plots.

Primary outcomes

Haemoglobin concentration

RCT evidence

Evidence comparing DFS with IS in RCTs showed that DFS may improve haemoglobin concentration slightly (MD 0.43 g/dL, 95% CI 0.23 to 0.63; I² = 90%; 18 comparisons, 13 studies, 4564 participants; low‐certainty evidence; Analysis 1.1). 

1.1. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 1: Haemoglobin concentration

Controlled before‐after studies

Among the CBA studies, we found no clear difference in haemoglobin concentration between DFS and IS (MD 0.26 g/dL, 95% CI −0.10 to 0.63; I² = 83%; 5 comparisons, 4 studies, 1397 participants; Analysis 2.1).

2.1. Analysis.

Comparison 2: Double‐fortified salt versus iodised salt: non‐randomised controlled trials, Outcome 1: Haemoglobin concentration

Urinary iodine concentration

RCT evidence

DFS may reduce urinary iodine concentration as compared to IS (MD −96.86 µg/L, 95% CI −164.99 to −28.73; I² = 100%; 8 comparisons, 7 studies, 1594 participants; low‐certainty evidence; Analysis 1.2). However, the mean endline values for urinary iodine concentration for all IS and DFS groups were greater than the cut‐off for iodine deficiency (i.e. > 100 µg/L), and the upper urinary iodine concentration range did appear to be high. This suggests that the provision of DFS increases urinary iodine concentration sufficiently to prevent iodine deficiency. 

1.2. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 2: Urinary iodine concentration

Controlled before‐after studies

Among the CBA studies, we found that DFS made no clear difference in urinary iodine concentration compared to IS (MD −17.27 µg/L, 95% CI −49.27 to 14.73; I² = 97%; 3 comparisons, 3 studies, 1127 participants; Analysis 2.2).

2.2. Analysis.

Comparison 2: Double‐fortified salt versus iodised salt: non‐randomised controlled trials, Outcome 2: Urinary iodine concentration

Blood pressure (systolic and diastolic)

No studies measured systolic or diastolic blood pressure.

Secondary outcomes

Ferritin concentration

RCT evidence

Use of DFS may result in little to no difference in ferritin concentration compared to IS (MD −3.94 µg/L, 95% CI −20.65 to 12.77; I² = 100%; 6 comparisons, 5 studies, 1419 participants; low‐certainty evidence; Analysis 1.3).

1.3. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 3: Ferritin concentration

Controlled before‐after studies

Among the CBA studies, only one study contributed to data for ferritin concentration (Kaur 2000).

Transferrin receptor concentration

Only RCT data were available for transferrin receptor concentration. Use of DFS may result in little to no difference in transferrin receptor concentration compared to IS (MD −4.68 mg/L, 95% CI −11.67 to 2.31; I² = 100%; 6 comparisons, 5 studies, 1256 participants; low‐certainty evidence; Analysis 1.4).

1.4. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 4: Transferrin receptor concentration

Body iron stores

Only RCT data were available for body iron stores. Use of DFS may result in a slight increase in body iron stores compared to IS (MD 1.77 mg/kg, 95% CI 0.79 to 2.74; I² = 87%; 5 comparisons, 4 studies, 847 participants; low‐certainty evidence; Analysis 1.5).

1.5. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 5: Body iron stores

Zinc protoporphyrin (μmol/mol of haemoglobin)

Only RCT data were available for zinc protoporphyrin. Use of DFS may reduce zinc protoporphyrin concentration compared to IS (MD −27.26 µmol/mol, 95% CI −47.49 to −7.03; I² = 99%; 4 comparisons, 3 studies, 921 participants; low‐certainty evidence; Analysis 1.6).

1.6. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 6: Zinc protoporphyrin

Prevalence of anaemia (haemoglobin concentrations below a cut‐off, taking into account age, altitude and smoking, when applicable)

RCT evidence

Compared to IS, DFS may reduce the risk of anaemia by 21% (RR 0.79, 95% CI 0.66 to 0.94; P = 0.007; I² = 61%; 10 comparisons, 8 studies, 2593 participants; moderate‐certainty evidence; Analysis 1.7).

1.7. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 7: Prevalence of anaemia

Controlled before‐after studies

Among the CBA studies, it is uncertain if DFS improved the prevalence of anaemia (RR 0.86, 95% CI 0.73 to 1.02; I² = 72%; 4 comparisons, 4 studies, 450 participants; Analysis 2.3).

2.3. Analysis.

Comparison 2: Double‐fortified salt versus iodised salt: non‐randomised controlled trials, Outcome 3: Prevalence of anaemia

Prevalence of iron deficiency anaemia (defined by the presence of anaemia with iron deficiency)

RCT evidence

DFS may reduce iron deficiency anaemia by on average 65% as compared to IS (RR 0.35, 95% CI 0.24 to 0.52; P < 0.001; I² = 37%; 6 comparisons, 5 studies, 1209 participants; low‐certainty evidence; Analysis 1.8).

1.8. Analysis.

Comparison 1: Double‐fortified salt versus iodised salt, Outcome 8: Prevalence of iron deficiency anaemia

Controlled before‐after studies

Among the CBA studies, it is uncertain if DFS improves iron deficiency anaemia (RR 1.01, 95% CI 0.99 to 1.02; I² = 0%; 1 comparison, 1 study, 947 participants; Analysis 2.4). An illustrative forest plot is provided.

2.4. Analysis.

Comparison 2: Double‐fortified salt versus iodised salt: non‐randomised controlled trials, Outcome 4: Prevalence of iron deficiency anaemia

Serum iron concentration (μg/dL)

Data were available from only one CBA study for serum iron concentration (Kaur 2000).

Hepcidin concentration (ng/mL)

No studies measured hepcidin concentration.

Salt intake (g/d)

RCT evidence

Three studies measured salt intake at endline, but only among those in the DFS group; therefore no comparative judgements of effects can be made (Andersson 2008; Haas 2014; Wegmüller 2006).

Controlled before‐after studies

One CBA study measured salt intake in both groups at endline, and found no change in salt intake for either group from baseline (Kaur 2000). 

Prevalence of adverse effects (including constipation, nausea, vomiting, heartburn or diarrhoea)

One study contributed to data for the prevalence of adverse effects (Asibey‐Berko 2007).

Prevalence of goitre

The evidence on the presence of goitre is very uncertain. Two RCTs measured the prevalence of goitre (Zimmermann 2003; Zimmermann 2004), whereby DFS groups had a lower prevalence (38%) compared to controls (51% to 58%).

Subgroup analyses

Given the substantial heterogeneity in the primary outcome analyses on haemoglobin, we explored subgroup analyses by life stage, study design, iron compound used in salt, use of concurrent deworming, anaemia status at enrolment and study duration. Among almost all subgroup analyses, heterogeneity remained high (I² > 60%). There were insufficient studies to conduct meaningful subgroup analyses on urinary iodine concentration and the secondary outcomes.

In investigating life stage, DFS may improve haemoglobin concentration slightly in school‐aged children and adolescents (MD 0.48 g/dL, 95% CI 0.12 to 0.85; 9 comparisons, 8 studies, 3084 participants) and adults (MD 0.44 g/dL, 95% CI 0.17 to 0.71; 7 comparisons, 6 studies, 966 participants), but not in children under five years (MD 0.06 g/dL, 95% CI −0.40, 0.52; 2 comparisons, 2 studies, 514 participants); see Analysis 3.1.

3.1. Analysis.

Comparison 3: Double‐fortified salt versus iodised: subgroup analysis, haemoglobin, Outcome 1: Haemoglobin concentration by life stage

For study design, DFS slightly improved haemoglobin concentration for both RCTs and cRCTs compared to IS. However, the observed improvement was greater for RCTs (MD 0.57 g/dL, 95% CI 0.26 to 0.87; 9 comparisons, 7 studies, 1572 participants) compared to cRCTs (MD 0.26 g/dL, 95% CI 0.03 to 0.48; 9 comparisons, 6 studies, 2992 participants); see Analysis 3.2.

3.2. Analysis.

Comparison 3: Double‐fortified salt versus iodised: subgroup analysis, haemoglobin, Outcome 2: Haemoglobin concentration by study design

Furthermore, the subgroup analysis assessing the iron compound used within DFS showed greater improvement in haemoglobin concentration for ferric pyrophosphate (MD 0.79 g/dL, 95% CI 0.59 to 1.00; 3 comparisons, 3 studies, 477 participants) compared to ferrous sulphate (MD 0.25 g/dL, 95% CI 0.20 to 0.31; 10 comparisons, 7 studies, 3117 participants) or ferrous fumarate (MD 0.18 g/dL, 95% CI 0.04 to 0.31; 5 comparisons, 4 studies, 970 participants); see Analysis 3.3.

3.3. Analysis.

Comparison 3: Double‐fortified salt versus iodised: subgroup analysis, haemoglobin, Outcome 3: Haemoglobin concentration by iron compound

Haemoglobin concentration may improve slightly with DFS both among individuals who received concurrent deworming (MD 0.31 g/dL, 95% CI 0.14 to 0.47; 7 comparisons, 6 studies, 1557 participants) and those who receive no deworming (MD 0.48 g/dL, 95% CI 0.09 to 0.86; 11 comparisons, 7 studies, 3007 participants); see Analysis 3.4.

3.4. Analysis.

Comparison 3: Double‐fortified salt versus iodised: subgroup analysis, haemoglobin, Outcome 4: Haemoglobin concentration by deworming status

Studies that enrolled non‐anaemic or 'healthy' participants experienced an improvement in haemoglobin concentration with DFS (MD 0.46 g/dL, 95% CI 0.23 to 0.69; 16 comparisons, 11 studies, 4078 participants), but there was no improvement among anaemic participants (MD 0.20 g/dL, 95% CI −0.06, 0.46; 2 comparisons, 2 studies, 486 participants); see Analysis 3.5.

3.5. Analysis.

Comparison 3: Double‐fortified salt versus iodised: subgroup analysis, haemoglobin, Outcome 5: Haemoglobin concentration by anaemia status

Study durations of 6 to 12 months may demonstrate slightly greater improvements in haemoglobin concentration with the use of DFS (MD 0.48 g/dL, 95% CI 0.26 to 0.70; 13 comparisons, 11 studies, 3245 participants), as compared to study durations of 18 months to 24 months (MD 0.29 g/dL, 95% CI −0.34 to 0.92; 5 comparisons, 2 studies, 1319 participants); see Analysis 3.6.

3.6. Analysis.

Comparison 3: Double‐fortified salt versus iodised: subgroup analysis, haemoglobin, Outcome 6: Haemoglobin concentration by study duration

The test for subgroup differences was only observed for the subgroup analysis on iron compound used (P < 0.00001; Analysis 3.3).

Sensitivity analyses

When inspecting unadjusted estimates from cRCTs, as compared to RCTs, there appeared to be a negligible difference in haemoglobin in DFS groups (MD 0.31 g/dL, 95% CI 0.09 to 0.53; 9 comparisons, 6 studies, 5887 participants; Analysis 4.1) compared to adjusted estimates for clustering (MD 0.26 g/dL, 95% CI 0.03 to 0.48; 9 comparisons, 6 studies, 3025 participants; Analysis 3.2). Similarly, when removing studies with high risk of bias in any of the prespecified domains, we found that results were consistent with the main analyses for haemoglobin concentration (MD 0.51 g/dL, 95% CI 0.21 to 0.82; 10 comparisons, 8 studies, 2240 participants; Analysis 4.2). 

4.1. Analysis.

Comparison 4: Sensitivity analyses, Outcome 1: Unadjusted estimates, haemoglobin concentration

4.2. Analysis.

Comparison 4: Sensitivity analyses, Outcome 2: Removal of high risk of bias studies, haemoglobin concentration

Assessment of reporting biases

Finally, we assessed the funnel plot for haemoglobin concentration, Figure 4, which indicated asymmetry and potential publication bias.

4.

Funnel plot of comparison 1: double‐fortified salt versus iodised salt, outcome: 1.1 haemoglobin concentration [g/dL]

Discussion

 We have summarised the findings here and in Table 1.

Summary of main results

We included 18 studies in this review. Of these, 14 were randomised controlled trials (RCTs), with seven randomised at the cluster‐level (cRCTs), and four were controlled before‐after (CBA) studies. We considered 17 studies in the meta‐analyses. All studies in this review used iodised salt (IS) as a control. 

Findings from the RCT studies demonstrated that providing double‐fortified salt (DFS) appears to increase haemoglobin concentration and reduce the risk of anaemia among those who received DFS compared to those who received IS. For urinary iodine concentration, consuming DFS does appear to reduce urinary iodine concentration relative to IS, although both salts increase urinary iodine concentration overall, using the evidence from the RCTs. If considered from a ‘proof‐of‐concept’ lens, this finding suggests that the addition of iron did not compromise the delivery of iodine significantly from a programmatic perspective, and participants still achieved adequate iodine nutrition. Despite the slight improvements in body iron stores and zinc protoporphyrin, DFS did not impact ferritin or transferrin receptor concentration. DFS may reduce iron deficiency anaemia by 65%, though the evidence is of low certainty and we do not have robust measures of iron deficiency anaemia. Few studies reported on adverse effects, the prevalence of goitre, or salt intake at endline. One study measured serum iron concentration. No studies measured or reported blood pressure (systolic or diastolic) or hepcidin concentration. 

For the CBA studies, the provision of DFS did not appear to affect haemoglobin or urinary iodine concentration, and it is uncertain if it has an effect on prevalence of anaemia or iron deficiency anaemia.

We conducted subgroup analyses on haemoglobin concentration by age group (< 5 years, 5 to 17 years, 18+ years), iron compound used in salt (ferrous sulphate, ferrous fumarate, micronised ferric pyrophosphate), study design (RCT versus cRCT), anaemia status at enrolment, study duration (6 to 12 months versus 18 to 24 months) and use of concurrent deworming. We found haemoglobin concentration appeared to improve within each subgroup compared to IS, except for children under five years, participants enrolled with pre‐existing anaemia and study durations of 18 to 24 months. The test for subgroup differences revealed an effect for the iron compound, as the increase in haemoglobin concentration was greater among those who received DFS with ferric pyrophosphate.

Although we found no subgroup effect on haemoglobin concentration based on anaemia status at enrolment, it is important to note that few study authors were explicit as to whether participants were anaemic or non‐anaemic. Theoretically, iron‐deficient individuals should see the greatest benefit from DFS. In many low‐ and middle‐income countries (LMICs), various forms of malnutrition and anaemia prevalence are high, as well as inflammation and infection. Therefore, we do not believe this subgroup analysis provides an accurate picture of improvement in haemoglobin concentration with the consumption of DFS. In fact, in our review, only three studies measured or adjusted for inflammation (i.e. AGP and CRP) (Andersson 2008; Haas 2014; Zimmermann 2004). For this reason, we did not conduct a subgroup analysis on inflammatory markers, as prespecified in the protocol (Baxter 2019).

Overall completeness and applicability of evidence

With respect to the limitations of this systematic review and meta‐analysis, several factors varied among the included studies and DFS formulations, which can complicate the extrapolation and generalisability of the results. The extent to which these factors could be explored with respect to subgroup differences was limited. Most of the studies were conducted in ‘ideal’ conditions, meaning the supply of DFS was generally ensured and participants interacted regularly with study personnel. This could increase uptake of the intervention compared to a more free‐living population, as seen in the study conducted by Banerjee 2016. Notably, the study by Banerjee 2016 was large (> 30,000 participants); thus, its inclusion could have impacted estimates for haemoglobin concentration and anaemia prevalence. Notably, fortification strategies can be influenced by health equity factors, which may contribute to the utility and acceptability of DFS and have implications for DFS uptake. Furthermore, most of the studies took place in India, where salt is noted to be of high quality (e.g. less moisture, few impurities), which could mean that the stability of DFS in other settings might be affected (Horton 2011). Importantly, indicators of the pathophysiology of iron deficiency anaemia, such as hepcidin concentration, were not measured in any of the studies.

To be successful and sustainable, food fortification programmes should be embedded within existing food systems to reduce nutrition inequities and ensure that quality products are available (Osendarp 2018). In most instances, the use of DFS would be relevant to geographies with endemic iodine deficiencies, mostly found in LMICs, and where anaemia or multiple micronutrient deficiencies exist. In such settings, changing the composition of salt, without changing its taste or salt content, might be an easier, cost‐effective strategy. There may be other technical limitations, however, such as multiple suppliers of salt, which can make controlling the fortification of salt with premix difficult, particularly if the salt is being obtained from local bodies of water (Shields 2021). Furthermore, without proper preparatory and storage techniques, the iron within DFS can oxidise, and thus become less bioavailable and the visual and organoleptic properties can decrease (Hurrell 2021). To ensure quality, supply chain requirements for DFS should be considered at several time points to prevent contamination, ensure product stability, avoid unacceptable sensory changes, and oversee ongoing quality assurance and quality control (Hurrell 2021). 

In addition to its lack of targeted benefits for those with higher nutritional demands, DFS consumption could also be affected by the changing guidelines for salt intake. In 2014, the World Health Organization (WHO) recommended decreasing salt intake to 5 g/d to decrease the risk of elevated blood pressure, which is associated with heart disease and stroke (WHO 2012b). This could have implications for DFS, as levels of fortification have largely been based on the assumed salt consumption of 10 g/d. While policies for DFS and reduction of salt intake to 5 g/d could be compatible, the concentration of iron and iodine in the salt would need to increase. The consequences that this might have with regards to stability and acceptability are not clear, but it is possible that the occurrence of discolouration of the salt and certain food products could increase, as these have been reported in some studies (Andersson 2008; Wegmüller 2006; Zimmermann 2003; Zimmermann 2004). In the conducted studies, we found that neither salt intake, nor change in salt consumption from baseline, were commonly measured. The recommendations for reducing salt intake below 5 g/d must also be balanced against some of the ongoing discussion around the adverse effects of very low salt intake below 5 g/d (O'Donnell 2020).

Certainty of the evidence

We assessed the certainty of evidence comparing the effects of DFS to IS on outcomes of interest using GRADE (Table 1). We judged the primary outcomes to be of low certainty (haemoglobin and urinary iodine concentration). We downgraded the evidence by two levels due to risk of bias and inconsistency of results (i.e. high heterogeneity). We considered the certainty of the evidence for prevalence of anaemia to be moderate; we downgraded the evidence by just one level for inconsistency due to substantial heterogeneity. We downgraded other secondary outcomes, such as ferritin concentration, transferrin receptor concentration, zinc protoporphyrin, body iron stores, and prevalence of iron deficiency anaemia by two levels due to two of the following reasons: risk of bias, inconsistency of results (high heterogeneity), and/or imprecision (crossing the line of no effect and not meeting the optimal information size threshold). We acknowledge that there could have been challenges around resource constraints within the studies, which may have affected data collection, processing and analysis.

Potential biases in the review process

We employed multiple techniques to minimise potential biases. The review was independent with no authors who held prior views on the subject. Three review authors independently carried out the review process, with the same data extraction sheet and tools to assess risk of bias in the included studies. All extractions were reviewed, and inconsistencies discussed. We followed the Cochrane Developmental, Psychosocial and Learning Problems Group search strategies to be inclusive in our searches and followed the outlined review processes to reduce potential biases. To minimise publication bias in this review, we also extensively searched grey literature and study registries. While there were no language limits set for this review for the conducted searches or the identified materials, all the identified studies were written in English. Many studies included minimal information regarding the randomisation procedure, allocation concealment and blinding, and adverse effects, thus our assessments were dependent on provided information, which could have made it difficult to detect reporting biases.  

Agreements and disagreements with other studies or reviews

We determined the mean difference in haemoglobin concentration from RCTs to be comparable to other meta‐analyses evaluating the effect of iron‐fortified foods on haemoglobin concentration (Athe 2014; Gera 2012; Keats 2019; Waller 2020). With respect to previous systematic reviews and meta‐analyses of DFS (Larson 2021; Ramírez‐Luzuriaga 2018; Yadav 2019), provision of DFS was found to result in similar effects on haemoglobin concentration and the prevalence of anaemia, although these studies did not follow the Cochrane methods of systematic reviews and may have included studies we excluded. For example, these reviews did not adjust for clustering within cRCTs (Ramírez‐Luzuriaga 2018; Yadav 2019), RCTs were pooled with CBA studies (Larson 2021; Ramírez‐Luzuriaga 2018; Yadav 2019), and some reviews did not consider grey literature (Ramírez‐Luzuriaga 2018; Yadav 2019). Although our sensitivity analyses showed adjustment for clustering and removal of high risk of bias studies did not make a difference in the overall effect estimate for haemoglobin concentration, it is important to have considered these potential effects. Previous reviews also did not provide results for related outcomes such as zinc protoporphyrin, which we have provided here. Furthermore, previous reviews did not employ the GRADE approach to assess the certainty of the evidence for outcomes. 

Recently, Larson 2021 reported the mean difference in haemoglobin concentration to be 0.44 g/dL (95% CI 0.23 to 0.64; 22 studies) among those who received DFS compared to controls. This is very similar to our findings (0.43 g/dL, 95% CI 0.23 to 0.63; 13 studies). Notably, Larson 2021 included three studies that did not meet our eligibility for inclusion — Nair 2013; Nair 2014 and Working Group on Fortification 1982 — as these studies either did not have a comparator or used iron‐only fortified salt as their intervention. Further, we classified Joshi 2014 as a CBA study, as authors indicated schools were purposely chosen. We also believe that given the multiple similarities between the study setting, design, and participant characteristics described by Asibey‐Berko 2007 and Nti‐Nimako 1998, the two reflect findings from the same study (Appendix 4). More specifically, we are of the opinion that the findings presented by Nti‐Nimako 1998 are from a midline assessment of the children enrolled in the larger study reported by Asibey‐Berko 2007, and thus ineligible for inclusion in our review, which focuses on endline measures. 

With respect to anaemia prevalence, Larson 2021 reported a reduced risk of 20% (95% CI 0.70 to 0.92; 13 studies) among those who received DFS, which is very similar to the reduced risk of 21% (95% CI 0.66 to 0.94; 8 studies) that we found. Similarly, for iron deficiency anaemia prevalence, Larson 2021 reported those who received DFS had a reduced risk of 64% (95% CI 0.24 to 0.55; 5 studies) and we reported a reduced risk of 65% (95% CI 0.24 to 0.52; 5 studies). Findings differed between the two reviews for urinary iodine concentration, as Larson 2021 found no effect between intervention and control groups, whereas we found that consuming IS led to a greater improvement in urinary iodine concentration. Conversely, Larson 2021 found that DFS improved ferritin concentration, while we observed no effect. 

Authors' conclusions

Implications for practice.

As the burden of iron‐related deficiencies remains among the most important micronutrient deficiencies globally, the additional fortification of iodised salt (IS) with iron could offer a unique opportunity to leverage an almost universally consumed product to improve iron intake. In highly controlled research settings, this review shows that double‐fortified salt (DFS) may be efficacious in reducing the risk of anaemia by 21%, and that there might be positive effects on blood haemoglobin concentration (an indicator used in diagnosis of anaemia). This result was consistent across groups for those under five years of age, and regardless of concurrent deworming. However, in studies of free‐living populations (i.e. effectiveness studies), there have been few assessments of DFS. From those that have been conducted to date, there is insufficient evidence to support the effect of DFS on anaemia and measures of iron status. Adherence to fortified salt, shelf life of fortified salt and community awareness could have further implications related to the use of DFS. Additional details on the background level of anaemia in populations receiving DFS would lend to the further interpretability of the data.  

Implications for research.

Several limitations of studies within this review are attributed to risk of bias, imprecision, and inconsistency. Given these, we identify the following areas of future research.

1. The need for high‐quality randomised controlled trials (RCTs), with special attention to high internal validity within the study design, will be important. In the current review, GRADE ratings for most outcomes were found to have low‐certainty evidence. Adjustment for clustering, inflammation, and potential differences in baseline characteristics will aid in understanding the true effect of DFS among populations where the burden of iron deficiency is high.

2. RCTs that focus on effectiveness (i.e. provision of the intervention within a programmatic context), to understanding the effect of DFS within free‐living populations. There is limited evidence demonstrating the effectiveness of DFS on biochemical and functional outcomes.

3. Studies that measure and report salt intake and blood pressure, to better understand the impact of a salt fortified with iron and iodine on hypertension. Most studies to date have investigated daily salt intakes of 10 g/d, although the World Health Organization (WHO) recommends 5 g/d of salt.

4. Studies that measure and report anaemia. Because anaemia, rather than iron deficiency or mean haemoglobin, is a more common indicator of national nutrition programmes, it is important to have better evidence on this outcome.

5. Focus on specific populations under‐represented in this review, where a high burden of iron deficiency anaemia or iodine deficiency, or both, exists (e.g. women of reproductive age).

6. Explore comparisons with triple or quadruple fortified salt (i.e. the fortification of salt with iron, iodine, and other micronutrients known to contribute to the burden of anaemia), to understand if there is an added benefit on anaemia prevalence and micronutrient deficiencies.

7. Further understanding around iron overload, particularly in settings where multiple methods are employed to address iron deficiency.

History

Protocol first published: Issue 10, 2019

Appendices

Appendix 1. Search strategies

Cochrane Central Register of Controlled Trials (CENTRAL)

Searched 18 September 2019 (56 records)
Searched 29 April 2021  (9 records when limited to records added 18 September 2019 to 29 April 2021)

#1    [mh ^"Sodium Chloride"]
#2    [mh "Sodium Chloride, Dietary"]
#3    #1 or #2
#4    [mh ^Iron]
#5    [mh "Iron, Dietary"]
#6    [mh "iron compounds"]
#7    [mh "ferric compounds"]
#8    [mh "ferrous compounds"]
#9    iron:kw,ti,ab
#10    Fe:kw,ti,ab
#11    (ferric or ferrous):kw,ti,ab
#12    {or #4‐#11}
#13    [mh ^Iodine]
#14    iodine:kw,ti,ab
#15    (iodis* or iodiz*):kw,ti,ab
#16    {or #13‐#15}
#17    #3 and #12 and #16
#18    ((doubl*or dual*) Near/3 salt*):kw,ti,ab
#19    (fortif* Near/3 salt*):kw,ti,ab
#20    (iron near/5 iodine near/5 salt*):kw,ti,ab
#21    {or #17‐#20} Limited to Trials

MEDLINE Ovid

Searched 16 September 2019 (440 records)
Searched 27 April 2021 (37 records)

1 Sodium Chloride/
2 Sodium Chloride, Dietary/
3 1 or 2
4 Iron/
5 Iron, Dietary/
6 iron compounds/ or ferric compounds/ or ferrous compounds/
7 iron.tw,kf.
8 Fe.tw,kf.
9 (ferric or ferrous).tw,kf.
10 or/4‐9
11 Iodine/
12 iodine.tw,kf.
13 ((iodis$ or iodiz$) adj3 salt$).tw,kf.
14 or/11‐13
15 3 and 10 and 14
16 ((doubl$ or dual$) adj3 salt$).tw,kf.
17 (fortif$ adj3 salt$).tw,kf.
18 (iron adj5 iodine adj5 salt$).tw,kf.
19 or/15‐18 [ Final line 2019]
20     (201909* or 201910* or 201911* or 201912* or 2020* or 2021*).dt,ez,da. (1658042)
21     19 and 20  {Final line 2021]

MEDLINE In‐Process & Other Non‐Indexed Citations Ovid

Searched 16 September 2019 (244 records)
Searched 29 April 2021 (21 records)

1     (iron or Fe or ferric or ferrous).tw,kf. 
2     (iodine or iodi#ed).tw,kf. 
3     (salt or salts).tw,kf. 
4     1 and 2 and 3 
5     ((doubl$ or dual$) adj3 salt$).tw,kf. 
6     (fortif$ adj3 salt$).tw,kf. 
7     4 or 5 or 6 

MEDLINE EPub Ahead of Print Ovid

Searched 16 September 2019 (11 records)
​​​​​​Searched 29 April 2021 (10 records)

1     (iron or Fe or ferric or ferrous).tw,kf. 
2     (iodine or iodi#ed).tw,kf. 
3     (salt or salts).tw,kf. 
4     1 and 2 and 3 
5     ((doubl$ or dual$) adj3 salt$).tw,kf. 
6     (fortif$ adj3 salt$).tw,kf. 
7     4 or 5 or 6 

Embase Ovid

Searched 16 September 2019 (645 records)
Searched 27 April 2021 (133 records)

1     sodium chloride/ 
2     salt intake/ 
3     1 or 2 
4     iron/ 
5     iron intake/ 
6     iron derivative/ 
7     ferric ion/ 
8     ferrous ion/ 
9     iron.tw,kw. 
10     Fe.tw,kw. 
11     (ferric or ferrous).tw,kw. 
12     or/4‐11 
13     iodine/ 
14     iodine.tw,kw.
15     ((iodis$ or iodiz$) adj3 salt$).tw,kw. 
16     or/13‐15 
17     3 and 12 and 16 
18     ((doubl$ or dual$) adj3 salt$).tw,kw. 
19     (fortif$ adj3 salt$).tw,kw. 
20     (iron adj5 iodine adj5 salt$).tw,kw. 
21     or/17‐20 
22     exp animals/ or exp invertebrate/ or animal experiment/ or animal model/ or animal tissue/ or animal cell/ or nonhuman/
23     human/ or normal human/ or human cell/ 
24     22 and 23 
25     22 not 24 
26     21 not 25 {Final line 2019]
27     limit 26 to yr="2019 ‐Current"  {Final line 2021]

Web of Science Core Collection Clarivate (Science Citation Index‐Expanded,  Social Science Citation Index, Conference Proceedings Citation Index‐Science, Conference Proceedings CItation Index‐Social Sciences & Humanities

Searched 17 September 2019 (357 records)
Searched 29 April 2021 (132 records)

#7 #6 OR #5 OR #4    
 Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=2019‐2021    [FInal line 2021]
# 6 TS= (fortif* NEAR/3 salt*)     
Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=All years    [FInal line 2019]
# 5 TS= ((doubl* or dual*) NEXT salt*)     
Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=All years    
# 4 #3 AND #2 AND #1     
Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=All years    
# 3 TS=(salt or salts)     
Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=All years    
# 2 TS=(iodine or iodi*ed)     
Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=All years    
# 1 TS=(iron or Fe or ferric or ferrous)     
Indexes=SCI‐EXPANDED, SSCI, CPCI‐S, CPCI‐SSH Timespan=All years    
Edit    

SCOPUS Elsevier

Searched 16 September 2019 (512 records)
Searched 29 April 2021 (45 new records added since 16 September 2019 )

((TITLE‐ABS‐KEY (iron  OR  fe  OR  ferric  OR  ferrous)) AND (TITLE‐ABS‐KEY ( iodine  OR  iodis*  OR  iodiz*)) AND (TITLE‐ABS‐KEY (salt  OR  salts))) OR (TITLE‐ABS‐KEY ("doubl* fortified"  W/3  salt*)) OR ( TITLE‐ABS‐KEY ("dual* fortif*"  W/3  salt*))  

WHO Library Database (WHOLIS)

Searched 16 September 2019 (244 records)
Searched 29 April 2021 (6 new records added between 16 September 2019 to 29 April 2021)

kw,wrdl: SALT AND FORTIFICATION or kw,wrdl: SALT AND FORTIFIED or kw,wrdl: SALT AND FORTIFY or kw,wrdl: SALT AND IRON or kw,wrdl: SALT AND IODINE 
kw,wrdl: SALT AND IODISED or kw,wrdl: SALT AND IODIZED 

Epistemonikos 

Searched 17 September 2019 (63 records)
Searched 29 April 2021 (2 new records added between 18 September 2019 to 29 April 2021)

(title:((title:(salt OR salts) OR abstract:(salt OR salts)) AND (title:(iron OR ferrous OR ferric OR Fe) OR abstract:(iron OR ferrous OR ferric OR Fe)) AND (title:(iodine OR iodise* OR iodize*) OR abstract:(iodine OR iodise* OR iodize*))) OR abstract:((title:(salt OR salts) OR abstract:(salt OR salts)) AND (title:(iron OR ferrous OR ferric OR Fe) OR abstract:(iron OR ferrous OR ferric OR Fe)) AND (title:(iodine OR iodise* OR iodize*) OR abstract:(iodine OR iodise* OR iodize*)))) OR (title:((double AND fortif*) OR (dual* AND fortif*)) OR abstract:((double AND fortif*) OR (dual* AND fortif*)))

Global Index Medicus (includes regional indexes AIM, LILACS, IMEMR, IMSEAR, SciELO and WPRIM) 

Searched 17 September 2019 (40 records)
Searched 29 April 2021 (1 new record 2019 to 2021)

(tw:(SALT OR SALTS)) AND (tw:(IRON OR FERROUS OR FERRIC OR FE)) AND (tw:(IODINE OR IODISE* OR IODIZE*))

ClinicalTrials.gov

Searched 18 September 2019 (11 records)
Searched 29 April 2021 (2 new records first posted from 09/18/2019 to 04/29/2021

Interventional Studies | SALT AND IRON AND IODINE | 
Interventional Studies | DOUBLE FORTIFICATION OR DUAL FORTIFICATION | 
Interventional Studies | FORTIFIED SALT 

World Health Organization International Clinical Trials Registry Platform (WHO ICTRP)

Searched 18 September 2019 (16 records)

Access attempted 29 April 2021 but site was unresponsive

3 separate searches via the Basic search screen: 
DOUBLE FORTIFICATION OR DUAL FORTIFICATION
SALT AND IRON AND IODINE
FORTIFIED SALT 

Other Resources

Searched 26 October 2019 (622 records)
Searched 4 May 2021 (14 records)

Keywords such as "double fortified salt" and "dual fortified salt":

1. Google Scholar (scholar.google.co.uk).

2. International Initiative for Impact Evaluation (www.3ieimpact.org).

3. Global Alliance for Improved Nutrition (www.gainhealth.org).

4. Nutrition International (www.nutritionintl.org).

5. Fortified Food Initiative (www.ffinetwork.org)

Appendix 2. Data extraction 

Trial methods

1. Study design

2. Study years

3. Unit and method of allocation

4. Method of sequence generation

5. Masking of participants, personnel and outcome assessors

6. Sources of funding

7. Conflicts of interest

Participants

1. Location of the study

2. Sample size

3. Age

4. Sex

5. Socioeconomic status (as defined by trialists and where such information was available)

6. Baseline prevalence of anaemia

7. Inclusion and exclusion criteria

Intervention

1. Dose

2. Type of iron compound

3. Iron concentration

4. Iodine concentration

5. Duration of the intervention

6. Co‐intervention (i.e. deworming)

7. Compliance

Comparison group

1. No intervention

2. Placebo

3. Provision of any supplements

Outcomes

1. Primary and secondary outcomes outlined under Types of outcome measures

2. Exclusion of participants after randomisation and proportion of losses at follow‐up

3. Method of aggregation (mean, median, proportion)

4. Outcome unit

5. Adjustment for clustering

Appendix 3. Criteria for assessing risk of bias 

We adapted the following operational definitions for assessing risk of bias from the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and the Cochrane Effective Practice and Organisation of Care guidelines for controlled before‐after (CBA) studies (EPOC 2017). We assessed all studies for random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting.

Random sequence generation (selection bias due to inadequate generation of a randomised sequence)

We described the method used to generate the allocation sequence, and used the following judgements.

1. Low risk of bias: the study clearly described the method of sequence generation and employed a randomised process (e.g. coin toss, random numbers table, computer‐generated random number sequence).

2. Unclear risk of bias: the study did not clearly describe the method of sequence generation or the description led to uncertainty around the process of sequence generation.

3. High risk of bias: the method of sequence generation was non‐random or quasi‐random (e.g. date of birth, day of the week, judgement by study personnel, preference by participant).

Allocation concealment (selection bias due to inadequate concealment of allocation before assignment)

We described the method used to conceal the allocation sequence, and used the following judgements.

1. Low risk of bias: the study used a central allocation system or an independent researcher or computerised system generated a random allocation schema; the allocation was concealed using sequentially numbered, opaque, sealed envelopes or a comparable method.

2. Unclear risk of bias: insufficient information about allocation concealment methods or description led to uncertainty around the allocation concealment.

3. High risk of bias: the study used an open allocation schedule (e.g. list of random numbers) without safeguards to prevent allocation disclosure; study personnel allocated participants or had prior knowledge of participant characteristics prior to allocation.

Blinding of participants and study personnel (performance bias due to knowledge of allocation to experimental and comparator interventions by participants and personnel)

We described the methods used to blind trial participants and personnel from knowledge of which intervention participants received, and used the following judgements.

1. Low risk of bias: the blinding of participants and key study personnel was ensured, and it was unlikely that the blinding could have been broken.

2. Unclear risk of bias: insufficient information to determine whether blinding could have been broken led to uncertainty around the blinding of participants and study personnel.

3. High risk of bias: blinding was absent or incomplete; blinding of key personnel and participants was attempted, but it was likely that the blinding could have been broken.

Blinding of outcome assessors (detection bias due to knowledge of allocation interventions by outcome assessors)

We described the methods used to blind outcome assessors from which intervention participants received, and use the following judgements.

1. Low risk of bias: blinding of outcome assessors was ensured, and it was unlikely that the blinding could have been broken.

2. Unclear risk of bias: insufficient information to determine whether blinding could have been broken led to uncertainty around blinding of outcome assessors.

3. High risk of bias: blinding of outcome assessors was absent or incomplete; blinding of outcome assessors was attempted, but it was likely that the blinding could have been broken.

Completeness of data collection (attrition bias due to quantity, nature, or handling of incomplete outcome data)

We described the completeness of the outcome data collected from study participants, and used the following judgements.

1. Low risk of bias: there was no missing outcome data; attrition was unlikely to be related to the outcome; attrition was similar across study arms and thus unlikely to be related to allocation; there was appropriate imputation of any missing data.

2. Unclear risk of bias: attrition was not reported or insufficiently reported.

3. High risk of bias: there was loss of outcome data; attrition was likely to be related to the outcome due to imbalance in numbers or missing data; there was uneven attrition between groups.

Selective reporting (reporting bias due to selective reporting of outcomes)

We described whether all planned comparisons were reported, and used the following judgements.

1. Low risk of bias: the study reported all prespecified outcomes.

2. Unclear risk of bias: it was not possible to discern whether all planned outcomes were reported.

3. High risk of bias: there was evidence of selective reporting of outcomes.

Other sources of bias

For this domain, we assessed all studies for risk of bias due to similar baseline outcome measurements, similar baseline characteristics, declarations of conflicts of interest and funding sources. We also assessed cRCTs for recruitment bias, baseline imbalance, loss of clusters, incorrect analysis, and comparability with individually randomised studies. 

Similar baseline outcome measurements

We assessed whether baseline assessments were similar and reported, and used the following judgements.

1. Low risk of bias: if performance or patient outcomes were measured prior to the intervention, and no important differences were present across study groups; or if imbalanced but appropriate adjusted analysis was performed (e.g. analysis of covariance).

2. Unclear risk of bias: if randomised trials had no baseline measure of outcome.

3. High risk of bias: if important differences were present and not adjusted for in analysis.

Similar baseline characteristics

We assessed whether baseline characteristics were similar and reported, and used the following judgements.

1. Low risk of bias: if baseline characteristics of the study and control providers were reported and similar.

2. Unclear risk of bias: if it was not clear in the paper (e.g. characteristics were mentioned in text but no data were presented).

3. High risk of bias: if there was no report of characteristics in text or tables or if there were differences between control and intervention providers.

Declarations of conflicts of interest and funding sources

We described declarations of conflicts of interest and funding sources that may have made the study liable to bias, and used the following judgements.

1. Low risk of bias: both conflicts of interest and funding sources were declared.

2. Unclear risk of bias: either conflicts of interest or funding sources were declared but not both.

3. High risk of bias: neither conflicts of interest nor funding sources were declared.

Recruitment bias

We assessed studies as follows.

1. Low risk of bias: if individuals were recruited to the study before the clusters were randomised.

2. Unclear risk of bias: if there was insufficient information provided on recruitment and randomisation procedures of clusters to permit judgement of low or high risk of bias.

3. High risk of bias: if individuals were recruited to the study after the clusters were randomised or participants were aware of the allocation before clusters were randomised.

Baseline imbalance

We assessed studies as follows.

1. Low risk of bias: if baseline characteristics were reported and were similar across clusters or if study authors used stratified or pair‐matched randomisation of clusters.

2. Unclear risk of bias: if there was inconsistency in reporting of clusters or insufficient information to permit judgement of low or high risk of bias.

3. High risk of bias: if baseline characteristics were not reported or if there were differences across clusters.

Loss of clusters

We assessed studies as follows.

1. Low risk of bias: if no complete clusters were lost or omitted from the analysis.

2. Unclear risk of bias: if there was inconsistency in reporting of clusters or insufficient information to permit judgement of low or high risk of bias.

3. High risk of bias: if complete clusters were lost or omitted from the analysis.

Incorrect analysis

We assessed studies as follows.

1. Low risk of bias: if study authors appropriately accounted for clusters in the analysis or provided enough information for review authors to account for clusters in the meta‐analysis.

2. Unclear risk of bias: if the study describes adjustment for clustering in their statistical methods, though this doesn't align with results presented or there is insufficient information to permit judgement of low or high risk of bias.

3. High risk of bias: if study authors did not appropriately account for clusters in the analysis or did not provide enough information for review authors to account for clusters in the meta‐analysis.

Compatibility with individual RCTs

We assessed studies as follows.

1. Low risk of bias: if effects of the intervention were likely not altered by the unit of randomisation.

2. Unclear risk of bias: when there is insufficient information to permit judgement of low or high risk of bias.

3. High risk of bias: if effects of the intervention were likely altered by the unit of randomisation.

Overall risk of other sources of bias

We reached an overall judgement for this domain as follows.

1. Low risk of bias: if we judged the study to be at low risk of bias for all 'other' domains above.

2. Unclear risk of bias: if we judged the study to be unclear in at least one 'other domain, but not at high risk of bias for any 'other' domain.

3. High risk of bias: if we judged the study to be at high risk of bias in at least one 'other' domain, or we judged the study to be unclear for multiple 'other' domains in a way that substantially lowers confidence.

Appendix 4. Further comments on studies by Asibey‐Berko and Nti‐Nimako

We treated the report by Nti‐Nimako 1998 as a secondary reference to the Asibey‐Berko 2007 publication, given substantial similarities between the two studies as described. These similarities included: 

• enrolment criteria (exact match);

• baseline numbers (nearly identical);

• study setting (exact match);

• timing (nearly identical); and

• salt‐masking techniques (exact match).

Unfortunately, we were unable to contact either author, as one author is deceased and the contact details were not available for the other author. We contacted three individuals involved with the Asibey‐Berko 2007 study (2 collaborators, 1 funder), and they did not have further information on the studies given that they were conducted more than 20 years ago. Both collaborators agreed the characteristics of the studies were highly similar and unlikely to be independent. 

We additionally contacted the first author of a recently released DFS systematic review and meta‐analysis (Larson 2021), who agreed the details were very similar, but had decided to include the study by Nti‐Nimako 1998. 

Because the report by Nti‐Nimako 1998 is a thesis and there are extreme similarities with the publication by Asibey‐Berko 2007, we concluded the work by Nti‐Nimako 1998 likely reflects a midline assessment for the purposes of obtaining one's degree. As the study by Asibey‐Berko 2007 was assessed as high risk of bias for selective outcome reporting, we believe that the mention of a midline assessment in the publication was likely omitted. 

Data and analyses

Comparison 1. Double‐fortified salt versus iodised salt.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Haemoglobin concentration	13	4564	Mean Difference (IV, Random, 95% CI)	0.43 [0.23, 0.63]
1.2 Urinary iodine concentration	7	1594	Mean Difference (IV, Random, 95% CI)	‐96.86 [‐164.99, ‐28.73]
1.3 Ferritin concentration	5	1419	Mean Difference (IV, Random, 95% CI)	‐3.94 [‐20.65, 12.77]
1.4 Transferrin receptor concentration	5	1256	Mean Difference (IV, Random, 95% CI)	‐4.68 [‐11.67, 2.31]
1.5 Body iron stores	4	847	Mean Difference (IV, Random, 95% CI)	1.77 [0.79, 2.74]
1.6 Zinc protoporphyrin	3	921	Mean Difference (IV, Random, 95% CI)	‐27.26 [‐47.49, ‐7.03]
1.7 Prevalence of anaemia	8	2593	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.66, 0.94]
1.8 Prevalence of iron deficiency anaemia	5	1209	Risk Ratio (M‐H, Random, 95% CI)	0.35 [0.24, 0.52]

Comparison 2. Double‐fortified salt versus iodised salt: non‐randomised controlled trials.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Haemoglobin concentration	4	1397	Mean Difference (IV, Random, 95% CI)	0.26 [‐0.10, 0.63]
2.2 Urinary iodine concentration	3	1127	Mean Difference (IV, Random, 95% CI)	‐17.27 [‐49.27, 14.73]
2.3 Prevalence of anaemia	3	450	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.73, 1.02]
2.4 Prevalence of iron deficiency anaemia	1	947	Risk Ratio (M‐H, Random, 95% CI)	1.01 [0.99, 1.02]

Comparison 3. Double‐fortified salt versus iodised: subgroup analysis, haemoglobin.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Haemoglobin concentration by life stage	13	4564	Mean Difference (IV, Random, 95% CI)	0.43 [0.23, 0.63]
3.1.1 Children < 5 years	2	514	Mean Difference (IV, Random, 95% CI)	0.06 [‐0.40, 0.52]
3.1.2 School‐aged children and adolescents 5‐17 years	8	3084	Mean Difference (IV, Random, 95% CI)	0.48 [0.12, 0.85]
3.1.3 Adults 18+ years	6	966	Mean Difference (IV, Random, 95% CI)	0.44 [0.17, 0.71]
3.2 Haemoglobin concentration by study design	13	4564	Mean Difference (IV, Random, 95% CI)	0.43 [0.23, 0.63]
3.2.1 Cluster‐RCT	6	2992	Mean Difference (IV, Random, 95% CI)	0.26 [0.03, 0.48]
3.2.2 RCT	7	1572	Mean Difference (IV, Random, 95% CI)	0.57 [0.26, 0.87]
3.3 Haemoglobin concentration by iron compound	13	4564	Mean Difference (IV, Fixed, 95% CI)	0.28 [0.22, 0.33]
3.3.1 Ferrous sulphate	7	3117	Mean Difference (IV, Fixed, 95% CI)	0.25 [0.20, 0.31]
3.3.2 Ferrous fumarate	4	970	Mean Difference (IV, Fixed, 95% CI)	0.18 [0.04, 0.31]
3.3.3 Ferric pyrophosphate	3	477	Mean Difference (IV, Fixed, 95% CI)	0.79 [0.59, 1.00]
3.4 Haemoglobin concentration by deworming status	13	4564	Mean Difference (IV, Random, 95% CI)	0.43 [0.23, 0.63]
3.4.1 Dewormed	6	1557	Mean Difference (IV, Random, 95% CI)	0.31 [0.14, 0.47]
3.4.2 No deworming	7	3007	Mean Difference (IV, Random, 95% CI)	0.48 [0.09, 0.86]
3.5 Haemoglobin concentration by anaemia status	13	4564	Mean Difference (IV, Random, 95% CI)	0.43 [0.23, 0.63]
3.5.1 Non‐anaemic or unspecified	11	4078	Mean Difference (IV, Random, 95% CI)	0.46 [0.23, 0.69]
3.5.2 Anaemic	2	486	Mean Difference (IV, Random, 95% CI)	0.20 [‐0.06, 0.46]
3.6 Haemoglobin concentration by study duration	13	4564	Mean Difference (IV, Random, 95% CI)	0.43 [0.23, 0.63]
3.6.1 6‐12 months	11	3245	Mean Difference (IV, Random, 95% CI)	0.48 [0.26, 0.70]
3.6.2 18‐24 months	2	1319	Mean Difference (IV, Random, 95% CI)	0.29 [‐0.34, 0.92]

Comparison 4. Sensitivity analyses.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Unadjusted estimates, haemoglobin concentration	13	7459	Mean Difference (IV, Random, 95% CI)	0.44 [0.26, 0.63]
4.1.2 RCT	7	1572	Mean Difference (IV, Random, 95% CI)	0.57 [0.26, 0.87]
4.1.3 Cluster‐RCT unadjusted	6	5887	Mean Difference (IV, Random, 95% CI)	0.31 [0.09, 0.53]
4.2 Removal of high risk of bias studies, haemoglobin concentration	8	2240	Mean Difference (IV, Random, 95% CI)	0.51 [0.21, 0.82]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Andersson 2008.

Study characteristics
Methods	Design: double‐blind, randomised, placebo‐controlled trial Country: India Study years: December 2005 to February 2007 Duration of participation: 10 months
Participants	The study was conducted in 18 villages in the Anekal Taluk Bangakire Urban District, Karnataka State, India. Children (aged 5‐18 years) were recruited from 6 schools (from grades 1‐10). Participants were eligible if SF < 15 µg/L or TfR > 7.6 mg/L and ZnPP < 40 µmol/mol heme. Of the 934 children who were screened, 458 were enrolled and eligible for randomisation.
Interventions	Participants were randomly assigned to 1 of 3 groups to receive: control (IS: n = 151); DFS intervention arm with iodised salt with 2 mg Fe/g salt as MGFePP (n = 155); or DFS intervention arms with iodised salt with 2 mg Fe/g salt as EFF (n = 152). Children living in the same household were randomly assigned to the same group. Setting and administration: salt provided to participants for use at home. "Salt was distributed to households (n = 364) in house‐to‐house visits every second month for 10 months. Salt for 2 months of household consumption was handed over directly to the head of the household. The aim of the study was carefully explained to the participating families, and it was emphasized that the new salt should be used for all cooking and food preparation. Families were instructed to finish an opened salt bag before starting a new one. These messages were reinforced at each of the salt distributions." Intervention: Iron compound: group 1 = ferric pyrophosphate (micronised); group 2 = ferrous fumarate (encapsulated) Iron concentration: group 1 = 2 mg/g salt; group 2 = 2 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: group 2 = encapsulation ingredients (soy stearin, titanium dioxide, hydroxypropyl methylcellulose, and sodium hexametaphosphate) Co‐interventions: vitamin A and albendazole (with albendazole treatment provided at 6 and 8 months) Control: salt iodised with potassium iodate at concentration of 30 µg iodine/1 g salt at factory
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: SF concentration, ZnPP concentration, TfR concentration, body iron, salt intake, anaemia prevalence, IDA prevalence All outcomes were measured at endpoint (10 months).
Notes	Comment(s): authors declined to provide further disaggregated data. Although iodine concentration in DFS with MGFePP decreased during the study, urinary iodine concentration still increased among participants relative to baseline.  Declarations of interest: declared Funding sources: supported by "the Micronutrient Initiative, the Swiss Federal Institute of Technology Zurich, and St John’s National Academy of Health Sciences. Paul Lohmann GmbH KG (Emmerthal, Germany) provided the iron fortification compound MGFePP, and The Micronutrient Initiative donated the iron fortification form encapsulated ferrous fumarate (EFF)"
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how random sequence was generated Quote: "remaining children from the 6 schools who met the criteria for iron deficiency (n = 458) were randomly assigned to 1 of 3 groups"
Allocation concealment (selection bias)	Unclear risk	Not mentioned explicitly by authors in terms of methods to conceal knowledge of assignment, although study was noted to be double‐blinded
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "the 3 salts were randomly assigned 3 different color codes, which were kept by an investigator who was not involved in the study"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "both the study investigators and the households were blinded to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	Clear reporting on loss to follow‐up and attrition was minimal
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Low risk	Arms balanced at baseline; salt stable throughout study for IS and EFF, although iodine content did decrease for MGFePP group; declared conflicts of interest and study funders

Asibey‐Berko 2007.

Study characteristics
Methods	Design: double‐blind, randomised, placebo‐controlled trial Country: Ghana Study years: December 1997 to August 1998 Duration of participation: 8 months
Participants	The study was conducted in the Agrarian region of the Sekyere West District of Ghana. Seventeen villages were randomly selected from a cluster of villages around Mampong, and mother‐child pairs were recruited from two clusters of nine rural villages. Field workers randomly selected homes in villages and recruited mother‐child dyads. Mothers were eligible to participate if they were healthy, non‐pregnant, non‐lactating, of childbearing age (15‐45 years), cared for a child aged 1‐5 years and prepared family meals. Mothers were excluded if they were unwilling to take a supplement during the trial, were unavailable during the 8‐month trial or had Hb < 100 g/L during baseline. Children were eligible if they were aged 1‐5 years, healthy and had Hb > 100 g/L. 318 mothers‐dyads met the inclusion criteria, while 300 women and 157 children were randomised.
Interventions	Participants were randomised to 1 of 3 groups: DFS plus placebo supplement (n = 100 mothers and 52 children); IS plus iron supplement intervention (n = 100 mothers); or placebo (IS) (n = 100 mothers and 105 children). For the purposes of this review, we only considered the DFS and IS groups. Setting and administration: salt provided to participants for use at home. Women were given an excess amount of salt, 4 kg, to use for the entire household for 3 months. Each family was monitored weekly for salt use. Salt was replenished in 1 kg sachets as required. Intervention Iron compound: ferrous fumarate Iron concentration: 1 mg/g salt Iodine compound: potassium iodide Iodine concentration: 50 µg/g salt Additives to salt: none Co‐interventions: none Control: iodised salt
Outcomes	Primary outcomes: Hb concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (8 months).
Notes	Comment(s): estimates for Hb concentration taken from Horton 2011 (obtained disaggregated data from study author) Declarations of interest: not reported Funding sources: Micronutrient Initiative, Ottawa, Canada
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Study noted to be randomised, but no details as to how randomisation was provided
Allocation concealment (selection bias)	Unclear risk	Not mentioned explicitly by authors in terms of methods to conceal knowledge of assignment, although study was noted to be double‐blinded
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "randomised by assignment to red, yellow, or green‐labelled salt packages"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double‐blinding
Incomplete outcome data (attrition bias) All outcomes	High risk	Was notable loss to follow‐up of 56.1% in DFS group, 44% in IS group
Selective reporting (reporting bias)	High risk	All outcomes not reported on completely (some missing data on haemoglobin, urinary iodine concentration, depending on whether mother or child, and unable to get clarification from authors)
Other bias	High risk	Arms balanced, although did not assess salt stability within study; there was discordance between groups at baseline for characteristics, but no correction was made; no mention of conflicts of interest, although study funders reported

Banerjee 2016.

Study characteristics
Methods	Design: cluster‐randomised, placebo‐controlled trial Country: India Study years: August 2012 to February 2015 Duration of participation: unclear
Participants	The study was conducted in the Bhojpur district of Bihar. Multiple experiments were conducted with inclusion of participants of any age and health status; however, the number of participants who were eligible and subsequently randomised is unclear.
Interventions	Setting and administration: salt provided to participants for use at home. DFS was delivered to homes and replenished every 2.5 to 3 months over 2 years. Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 40 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: none Control: iodised salt
Outcomes	Primary outcomes: Hb concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (2 exact time points unclear).
Notes	Comment(s): study not included in analyses, as published data could not be extracted for outcomes of interest (i.e. endline mean/median, SD and n). We contacted authors but, due to resource constraints, they could not provide us with the required data. Declarations of interest: not reported Funding sources: "For financial support we thank the UK Department for International Development, the Initiative for Impact Evaluation, the U.S. National Institutes of Health (P01AG005842), the International Food Policy Research Institute, and the MIT Department of Economics. Our partners at Tata Chemicals Limited made the distribution of DFS possible in rural Bihar. The National Institute of Nutrition (Hyderabad) provided ongoing assistance with the verification of iron and iodine content in salt samples." (page 144)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quotes: "we stratified by block and then randomly assigned half of the villages to treatment and half to control" "within each village, we randomly select 15 households to be measurement households, for a total of 6000 households"
Allocation concealment (selection bias)	Unclear risk	No details provided by study authors
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "control households are therefore not aware of the existence of Double Fortified Salt when we start the sales experiment"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided by study authors
Incomplete outcome data (attrition bias) All outcomes	Low risk	Household attrition was roughly 4% in the control villages, 5% in sales villages, 4% among households not receiving free DFS in the free DFS villages, but only 2% for households receiving free DFS. These figures are slightly imbalanced across both experiments.
Selective reporting (reporting bias)	Unclear risk	Outcomes appear, however some ages are mentioned in particular while others are left out
Other bias	Unclear risk	Authors note having controlled for baseline characteristics, arms balanced at baseline; salt stability not reported; no mention of conflicts of interest, although study funders reported

Bathla 2016.

Study characteristics
Methods	Design: controlled before‐after trial Country: India Study years: not indicated Duration of participation: 3 months
Participants	The study was conducted in primary government schools of Ludhiana, where 140 children aged 7 to 9 years were randomly selected to participate. Of these, 120 children with mild, moderate or severe anaemia were eligible (having a hb level < 12 g/dL).
Interventions	Participants were randomised to intervention (DFS: n = 60) or control group (IS: n = 60). Setting and administration: salt was provided within meals at school (midday meal programme) for use in daily preparations.  Intervention Iron compound: ferrous sulphate Iron concentration: 1.8 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: none Control: iodised salt, 2.08 g/d/child
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (3 months).
Notes	Comment(s): none Declarations of interest: not reported Funding sources: authors acknowledge Tata Chemicals for providing DFS for research purposes, but study funding not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Study was not randomised Quote: "subjects were divided into two groups‐ Control Group and Experimental Group"
Allocation concealment (selection bias)	Unclear risk	No details provided by study authors
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No mention of blinding in publication
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No mention of blinding in publication
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No mention of loss to follow‐up and no study flow diagram; unclear if there was attrition as baseline and endline values are the same
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Unclear risk	Arms balanced at baseline; no report of salt stability; no mention of funding or conflicts of interest

Bathla 2017.

Study characteristics
Methods	Design: controlled before‐after study Country: India Study years: not indicated Duration of participation: 3 months
Participants	Anaemic, adolescent girls aged 16 to 18 years, and both adult women and men aged 18 to 35 years with any health status at enrolment were eligible to participate. A total of 60 participants were eligible for randomisation.
Interventions	Participants were randomly allocated to intervention (n = 30) or control (n = 30) groups. Setting and administration: salt was provided to participants for use at home for all meals. Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: nutrition education in intervention group only (conducted every 15 days) Control Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (3 months).
Notes	Comment(s): National Institute of Nutrition DFS used in study, so have used the iron and iodine concentration for National Institute of Nutrition DFS Declarations of interest: not reported Funding sources: authors acknowledge Tata Chemicals for providing DFS for research purposes, but study funding not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Study was not randomised, as subjects were categorised into control and experimental group
Allocation concealment (selection bias)	Unclear risk	No details provided by study authors
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No mention of blinding in publication
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No mention of blinding in publication
Incomplete outcome data (attrition bias) All outcomes	Low risk	No loss to follow‐up among participants
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Unclear risk	Arms balanced at baseline; no report of salt stability; no mention of conflicts of interest or study funding

Brahmam 2000.

Study characteristics
Methods	Design: cluster‐randomised controlled trial Country: India Study years: 1989 to 1992 Duration of participation: 24 months
Participants	This study was carried out in four randomly selected blocks of the East Godvari district in the State of Andhra Pradesh. Participants were of any age, sex or health status. The number of participants eligible and subsequently randomised is unclear.
Interventions	Setting and administration: unclear, but seems that salt was provided to participants for use at home. Intervention Iron compound: ferrous sulphate heptahydrate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: none Control: iodised salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: anaemia prevalence, prevalence of goitre All outcomes were measured at endpoint (24 months).
Notes	Comment(s): adjustment for clustering was conducted in analyses using a mean cluster size of 20, ICC: 0.035 and design effect of 1.66 Declarations of interest: not reported Funding sources: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how random sequence was generated Quote: "this study was carried out in four randomly selected blocks"
Allocation concealment (selection bias)	Unclear risk	No details provided by study authors, but also does not appear to be a blinded study
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No mention of blinding in publication
Blinding of outcome assessment (detection bias) All outcomes	High risk	Not blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No information on loss to follow‐up
Selective reporting (reporting bias)	High risk	Selective reporting of ages and outcomes, outcomes not reported on in their entirety
Other bias	High risk	Do not comment on balance between arms at baseline; salt appears to be stable throughout study; do not adjust for clusters or report ICC; no mention of conflicts of interest or study funding

Haas 2014.

Study characteristics
Methods	Design: double‐blind, randomised, placebo‐controlled trial Country: India Study years: June 2009 to August 2010 Duration of participation: 7.5 to 9 months
Participants	The study was conducted in the Panighatta Tea Estate in the Darjeeling District of India. Women were eligible if they were healthy and not severely anaemic, aged 18‐55 years old and not pregnant. Of these, 245 women were eligible for inclusion.
Interventions	Participants were randomised into intervention (DFS: n = 122) or control (IS: n = 123) groups. Loss to follow‐up resulted in the final sample of 212 women (DFS: n = 104) and (IS: n = 108) Setting and administration: salt was provided to participants for use at home. To improve compliance, DFS and IS were delivered to the tea estate in 500 g bags marked with colour coding. Two colours per type of salt were assigned by the salt manufacturer, resulting in a random assignment to participants. Households received 1 bag for every 2 household members each month. The salt was instructed to be used ad libitum in food preparation. Intervention Iron compound: ferrous fumarate Iron concentration: 1.1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 47 µg/g salt Additives to salt: encapsulation ingredients (soy stearin, titanium dioxide, hydroxypropyl methylcellulose, and sodium hexametaphosphate) Co‐interventions: deworming Control: iodised salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: SF concentration, TfR concentration, body iron, salt intake, anaemia prevalence, IDA prevalence All outcomes were measured at endpoint (7.5 to 9 months).
Notes	Comment(s): obtained estimates for iron deficiency anaemia from communication with study authors Declarations of interest: declared. JD Haas, M Rahn, S Venkatramanan, GS Marquis, MJ Wenger, LE Murray‐Kolb, and G A. Reinhart have no conflicts of interest. AS Wesley was employed by the Micronutrient Initiative. Funding sources: supported by the Micronutrient Initiative, Ottawa, Canada
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how random sequence was generated Quote: "women from each stratum were randomly assigned into either a DFS group or control group that used only iodized salt"
Allocation concealment (selection bias)	Low risk	Quote: "two colors per type of salt were assigned by the salt manufacturer, resulting in random assignment of the 245 participants into 4 color‐coded groups" (note only manufacturer knew code until first round of data analysis completed)
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "the study design was a double‐blind, randomised, controlled food‐fortification trial" Also used colour‐coding system to conceal salt
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double‐blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk	Clearly report on loss to follow‐up, losses also minimal at 13.4% (n = 18 in DFS, n = 15 in IS)
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Low risk	Arms balanced at baseline; salt stable throughout the study; declared conflicts of interest and study funding

Jayatissa 2012.

Study characteristics
Methods	Design: double‐blind, cluster‐randomised controlled trial Country: Sri Lanka Study years: not indicated Duration of participation: 9 months
Participants	The study was conducted in the Public Health Midwives areas, where there is a high burden of anaemia. Participants were eligible if they were 6 to 10 years of age, and were not suffering any chronic health problems. Of these, 704 households with a child in each were eligible and allocated randomly.
Interventions	Due to non‐response, 351 children were allocated to DFS and 350 children were allocated to control. Setting and administration: salt provided to participants for use at home. A total of 2 kg to 3 kg of salt (accounting for all household members) was distributed to mothers monthly after baseline assessment was completed. Health personnel monitored the intake of salt and any shortages during home visits. Intervention Iron compound: ferrous fumarate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 20 µg/g to 30 µg/g salt Additives to salt: encapsulation ingredients (soy stearin, titanium dioxide, hydroxypropyl methylcellulose, and sodium hexametaphosphate) Co‐interventions: deworming Control: iodised salt (20 mg to 30 mg of Iodine/kg of salt)
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: SF concentration, anaemia prevalence All outcomes were measured at endpoint (9 months).
Notes	Comment(s): adjustment for clustering was conducted in analyses, with a mean cluster size of 22, ICC: 0.04 and design effect of 1.84 Declarations of interest: not reported Funding sources: Micronutrient Initiative, India
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how random sequence was generated Quote: "A total of 704 households each with a child in the selected age group were identified and were allocated randomly to two groups"
Allocation concealment (selection bias)	Low risk	Quote: "the code given in sealed envelopes was broken after the completion of the end survey"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "All individuals involved in the trial (including parents, health workers, and research staff) were unaware of assignment of the PHM area to the intervention / control group  until  the  code given in  sealed envelopes was broken after the completion of the end survey"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "All individuals involved in the trial (parents, health workers, and research staff) were unaware of the assignment of the cluster to the intervention / control group until  the  code given in  sealed envelopes was broken after the completion of the end survey"
Incomplete outcome data (attrition bias) All outcomes	Low risk	Minimal loss to follow‐up Quote: "the drop outs rates in the DFS and Iodine group are 3.7% and 4.0% respectively"
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes (although do not comment on adverse effects, although noted collecting this data from caregivers)
Other bias	High risk	Arms not balanced at baseline; no report of salt stability; do not report whether adjusted for clustering, do state ICC used to generate sample size, did not assess stability of DFS over time; no mention of conflicts of interest, although study funders reported

Joshi 2014.

Study characteristics
Methods	Design: controlled before‐after trial, with clustering at the level of the school Country: India Study years: March 2010 to April 2011 Duration of participation: 9 months
Participants	Primary school children (aged 6 to 15 years) were enrolled from the rural schools of Waghodia, in the Vadodara district of Gujarat. At baseline, four schools were selected purposively (based on the availability of iodised salt) where 1184 children were enrolled. Of these, only 947 could complete the study due to various reasons.
Interventions	Two schools were chosen as the experimental group (DFS: n = 442) and the other two as the control group (IS: n = 505). Setting and administration: salt was provided to participants for use at home. DFS packets (1 kg) were distributed among experimental group during first week of every month Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 40 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: nutrition education in intervention group only Control: iodised salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (9 months).
Notes	Comment(s):  The authors only collected Hb concentration, which is used to assess anaemia, and did not collect ferritin or transferrin receptor protein (markers of iron deficiency), which are used in combination with anaemia to determine the prevalence of iron deficiency anaemia.  National Institute of Nutrition DFS used in study, so have used the iron and iodine concentration for National Institute of Nutrition DFS Declarations of interest: not reported Funding sources: University Grants Commission provided Research Fellowships in Science for Meritorious Students scholarship to Ms Kejal Joshi
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Study was not randomised, the four schools were "selected purposively" to form experimental and control groups
Allocation concealment (selection bias)	Unclear risk	No mention of allocation concealment
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	Not blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	20% attrition (1184 enrolled, 947 completed; have not commented on how children who did not complete study were different from those who did)
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes; authors report measuring IDA, but only collected biomarkers to be able to determine anaemia alone (i.e. haemoglobin) were measured, and measures of iron deficiency (i.e. ferritin or TfR) were not collected
Other bias	High risk	Arms were not balanced at baseline; no report of salt stability; no mention of funding or conflicts of interest

Kaur 2000.

Study characteristics
Methods	Design: controlled before‐after study Country: India Setting and administration: university hostel Study years: mid‐September 1998 to mid‐March 1999 Duration of participation: 6 months
Participants	Healthy, non‐pregnant, non‐lactating young women (aged 18 to 24 years) living in the girls' hostel at the Punjab Agricultural University, Ludhiana. Participants could be anaemic and non‐anaemic at enrolment. Residents had different mess halls that they obtained meals from, and the respective salts were added to meals at the level of the mess hall (DFS provided to Mess No. 2 and IS provided to Mess No. 1).
Interventions	150 participants were recruited to participate in the study, with 100 receiving DFS and 50 receiving IS. DFS was procured from Tamil Nadu Salt Corporation, Chennai, and IS was from Tata Salt. There was no study attrition.  Intervention Iron compound: not reported Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 23.2 µg/g salt Additives to salt: not reported Co‐interventions: dewormed with 400 mg of albendazole Control: iodised salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: SF, serum iron, anaemia prevalence, salt intake, adverse effects All outcomes were measured at endpoint (6 months).
Notes	Comment(s): none Declare conflicts of interest: not reported Funding: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Study was not randomised
Allocation concealment (selection bias)	Unclear risk	No mention of allocation concealment
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study was not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	Study was not blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition and very thorough reporting
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	High risk	Arms balanced at baseline; salt appears to be stable throughout the study; no mention of funding or conflicts of interest

Krämer 2018.

Study characteristics
Methods	Design: cluster‐randomised control trial, randomised at level of school Country: India Study years: August 2015 to August 2016 Duration of participation: 12 months
Participants	The study was conducted in the two blocks of Kako and Modanganj in the Jehanabad district, located in the state of Bihar, India. A simple random sample of 108 schools was drawn prior to the DFS intervention from a list of 228 public schools within the two selected blocks. From these 108 schools, 54 were randomly allocated to the treatment group. The remaining 53 schools did not receive any treatment and formed the control group to use the conventional iodised salt. One school dropped out. Participants from eligible schools were in the second grade (6 to 7 years old).
Interventions	A total of 1791 participants were randomised to either the intervention or control group, but the disaggregation was not provided at baseline. Setting and administration: salt delivered to public schools in Bihar, India, and used to prepare the midday meal programme. Intervention Iron compound: ferrous sulphate Iron concentration: 0.86 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: none Control: iodised salt
Outcomes	Primary outcomes: Hb concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (12 months).
Notes	Comment(s): Data about the study obtained from the authors National Institute of Nutrition DFS used in study, so have applied the iodine concentration for National Institute of Nutrition DFS Effectiveness study Adjustment for clustering was conducted in analyses using a mean cluster size of 22, ICC: 0.03 and design effect of 1.65. Declarations of interest: not reported Funding sources: Foundation Fiat Panis and the German Research Foundation (DFG)‐RTG 1666 provided financial support
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A computer‐generated list of random numbers was used for the allocation of the treatment and control groups Quote: "a simple random sample of 108 schools was drawn prior to the DFS intervention from a list of 228 public schools within the two selected blocks"
Allocation concealment (selection bias)	Unclear risk	The research team does not seem to be blinded to the intervention, and therefore potentially knew about allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "We received DFS directly from the manufacturer and delivered it to the treatment schools either every month or two, depending on consumption. Headmasters were instructed to contact the study team if they ran out of DFS before the next delivery date." Note: headmasters know which group (DFS/IS) they were in
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No mention of blinding in publication
Incomplete outcome data (attrition bias) All outcomes	High risk	21.5% attrition (1791 enrolled, 1406 completed; have not commented on how children who did not complete study were different from those who did)
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	High risk	Arms balanced at baseline; no report of salt stability; no mention of conflicts of interests, although study funders reported

Rajagopalan 2000.

Study characteristics
Methods	Design: double‐blind, cluster‐randomised controlled trial Country: India Study years: not indicated Duration of participation: 12 months
Participants	The study was conducted in the "CWS" (page 323) tea estate of Valparai in South India. Households were selected from clusters, whereby adult participants (men and women, age not specified) were eligible. Of these, 1327 were allocated to the experimental area and 1332 to the control area. Only 793 participants completed the study.
Interventions	Of those that were eligible, 1327 were allocated to the experimental area and 1332 to the control area. Only 793 participants completed the study (DFS: n = 385) and (IS: n = 408). Setting and administration: salt was provided to participants for use at home. Project workers delivered the required quantity of salt.  Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: deworming (among half of participants only) Control: iodised salt
Outcomes	Primary outcomes: Hb concentration Secondary outcomes: none All outcomes were measured at endpoint (12 months).
Notes	Comment(s): adjustment for clustering was conducted in analyses with a mean cluster size of 66, ICC: 0.04 and design effect of 3.60 Declarations of interest: authors were affiliated with Sundar Chemicals Pvt Ltd. Study was conducted jointly by Sundar Chemicals, Parry Agro Group and United Planters Association of South India. Funding sources: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "With the use of a table of random numbers, every cluster was assigned to either an experimental or a control cluster. The study was a double‐blind, randomised, placebo‐controlled trial."
Allocation concealment (selection bias)	High risk	Use of random numbers table, project leader knew allocation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "The study was a double‐blind, randomised, placebo‐controlled trial. Only the project leader knew which were the experimental and which were the control households."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double blinding (insufficient information to assess)
Incomplete outcome data (attrition bias) All outcomes	High risk	Only report endline numbers (no baseline provided) Quote: "1320 participants had blood analysis completed, with 793 participants completing all 3 rounds of haemoglobin analysis"
Selective reporting (reporting bias)	Unclear risk	Not clear about dewormed and non‐dewormed for all outcomes, even though only half the group was dewormed
Other bias	High risk	Arms not balanced at baseline; salt reported to be stable; do not report adjustment for clustering and do not state ICC; authors affiliated with salt company, no mention of study funding

Reddy 2016.

Study characteristics
Methods	Design: randomised controlled trial Country: India Study years: October 2009 to November 2010 Duration of participation: 6 months
Participants	The study was conducted in a semi‐governmental hospital in urban Vadodara, Gujarat. Pregnant women with a singleton pregnancy in their first trimester were eligible.
Interventions	Participants were randomised to intervention (DFS: n = 75) or control (IS: n = 75) groups. Of these, a total of 121 women completed the study (DFS: n = 67; IS: n = 54). Setting and administration: salt was provided to participants for use at home. DFS was packed in 1 kg packets and were supplied monthly to families. Each house received 1‐2 kg of salt/month. Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 40 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: iron and folic acid supplementation, nutrition education Control: iodised salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: anaemia prevalence All outcomes were measured at endpoint (6 months).
Notes	Comment(s): none Declarations of interest: declared Funding sources: University Grants Commission, New Delhi, awarded Research Fellowships in Science for Meritorious Students scholarship to Ms Kejal Joshi, funding for study not declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "women with singleton pregnancy were selected and were randomly divided into experimental (n=75) and control (n=75) group using a computer generated list"
Allocation concealment (selection bias)	Unclear risk	No details provided by study authors
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No mention of blinding in publication
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No mention of blinding in publication
Incomplete outcome data (attrition bias) All outcomes	Low risk	No loss to follow‐up among participants
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Low risk	Arms were balanced at baseline; salt was reported to be stable throughout the study. Declared funding and conflicts of interest

Sivakumar 2001.

Study characteristics
Methods	Design: double‐blind, cluster‐randomised controlled trial Country: India Study years: 1996 to 1998 Duration of participation: 18 months
Participants	Children were eligible if they attended any of the four randomised schools. No further information given.
Interventions	Intervention and control schools, with approximately 500 children per school enrolled. Setting and administration: salt was provided within residential schools. Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: none Control: iodised salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: none All outcomes were measured at endpoint (18 months).
Notes	Comment(s): adjustment for clustering was conducted in all analyses using a similar design of 1.66 effect as Brahmam 2000 Declarations of interest: not reported Funding sources: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Noted to be randomised, but no details as to how Quotes: "a double blind randomised study" "one boys school and one girls school were randomly allocated to experimental group where DFS was supplemented, while the rest served as control and received IS"
Allocation concealment (selection bias)	Unclear risk	No details provided by study authors
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "a double blind randomised study"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double‐blinding (insufficient information to assess)
Incomplete outcome data (attrition bias) All outcomes	High risk	No information on loss to follow‐up, although note retention issues. Authors note substantial loss to follow‐up; data have been presented inconsistently between the outcomes (boys and girls measures are presented together, as well as separately).
Selective reporting (reporting bias)	High risk	Selective reporting of ages and outcomes, outcomes not reported on in their entirety
Other bias	High risk	Arms balanced at baseline; salt reported to be stable throughout the study; do not adjust for clusters or report ICC; no mention of conflicts of interest or study funding

Vinodkumar 2007.

Study characteristics
Methods	Design: single blind, cluster‐randomised controlled trial Country: India Study years: not indicated, but authors took 2 years to complete the study due to the staggered start Duration of participation: 12 months
Participants	Families living in any of the designated seven clusters within three states: Karnataka (3 clusters: Tumkur, Uttar Kanada and Dharwad), Gujarat (2 clusters: Surat and Bharuch), and Uttar Pradesh (2 clusters: Pratapgarh and Gonda). All members of the household (with any health status) were eligible; however, measurements were taken on two to three members per household aged 10 years or older. Approximately 30 families per cluster (or 63‐90 participants) were recruited, but it is unknown how many participants were eligible.
Interventions	A total of 829 participants were randomised and completed the study (intervention group: n = 393; control group: n = 436). Both DFS and IS were obtained by SSF Chennai. Setting and administration: salt was provided to participants for use at home. The salt was packaged in 1 kg samples and supplied to families through self‐help community‐based organisations. Intervention Iron compound: ferrous sulphate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 40 µg/g salt Additives to salt: sodium hexametaphosphate Co‐interventions: deworming Control Iodine concentration: not specified
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: none All outcomes were measured at endpoint (12 months).
Notes	Comment(s): adjustment for clustering was conducted in analyses, with a mean cluster size of 63, ICC: 0.04 and design effect of 3.48 Declarations of interest: not reported Funding sources: DFS provided by Sundar Serendipity Foundation, although study funders not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "In each cluster, two adjoining villages were selected randomly, and these formed the experimental and control groups. After the random selection process, it was determined that all clusters shared a similar economic background."
Allocation concealment (selection bias)	High risk	Quote: "since the study was single blinded, the families did not know whether they were receiving iodized salt or DFS"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Single‐blinded study (participants only)
Blinding of outcome assessment (detection bias) All outcomes	High risk	Research personnel were not blinded, therefore unlikely that data collectors were blinded (although not explicitly addressed within publication)
Incomplete outcome data (attrition bias) All outcomes	High risk	Only report endline numbers (no baseline provided), therefore attrition is unknown
Selective reporting (reporting bias)	Low risk	Presentation of outcomes could be hard to follow, or include alternate data presentation, but ultimately was not incorrect
Other bias	High risk	Several members per household were recruited to the study, but it was not clear whether adjustments were made for having multiple members from one family in the analysis; arms appear balanced at baseline; salt reported to be stable; did not adjust for clustering or report ICC; no mention of conflicts of interest, and funders of study unclear

Wegmüller 2006.

Study characteristics
Methods	Design: double‐blind, randomised controlled trial Country: Cote d'Ivoire Study years: not indicated Duration of participation: 6 months
Participants	School children, aged 5 to 15 years, from 4 primary schools in a rural village in the Dabou district, with iron deficiency with or without anaemia were eligible for the intervention trial. Anaemia was defined as Hb < 120 g/L in children aged 12 years and older, and Hb < 115 g/L in children aged 5 to 12 years. Iron deficiency was indicated by serum TfR > 8.5 mg/L or SF <30 mg/L.
Interventions	Participants were randomly assigned to either the intervention or control group. Children from 2 schools at one end of the village received IS (n = 63) and the children from 2 schools at the other end of the village received DFS (n = 60). Setting and administration: salt was provided to participants for use at home. "Each participating child was given a monthly 2.5‐kg salt portion (based on a mean per capita salt intake of; 4 g/d and an mean household size of 12 persons) distributed at school to be used in their household. In a village meeting at the beginning of the study and at each of the monthly salt distribution, it was emphasized that the distributed salt should be used for all cooking and food preparation, as well as at the table." Intervention Iron compound: ferric pyrophosphate Iron concentration: 3 mg/g salt Iodine compound: potassium iodate Iodine concentration: 40 µg/g salt Additives to salt: none Co‐interventions: deworming Control Iodine concentration: not specified "Iodized salt from Namibia imported by the major Abidjan salt producer (Sagid Salt) was used for the study."
Outcomes	Primary outcomes: Hb concentration, UI concentration (although data not presented clearly) Secondary outcomes: SF concentration, TfR concentration, CRP concentration, body iron, salt intake (at baseline only), anaemia prevalence, IDA prevalence All outcomes were measured at endpoint (6 months).
Notes	Comment(s): authors declined to provide further disaggregated data Declarations of interest: not reported Funding sources: supported by ETH Zurich and the Walter Hochstrasser Foundation in Zurich, Switzerland and by Dr Paul Lohmann GmbH KG, Emmerthal, Germany
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	As described, details as to appropriateness of random sequence generation are uncertain  Quote: "randomisation at school level" "all children from 2 schools at one end of the village received the IS (n = 60); the children from 2 schools at the other end of the village received the DFS"
Allocation concealment (selection bias)	Unclear risk	Quote: "both the investigators and schools were unaware of the group assignment"  No mention of how allocation occurred
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "both the investigators and schools were unaware of the group assignment"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double‐blinding (insufficient information to assess)
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition reported by study authors
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	High risk	Arms not balanced at baseline; iron and iodine stability of salts affected with study progression; no mention of conflicts of interest, although study funders reported

Zimmermann 2003.

Study characteristics
Methods	Design: double‐blind, randomised controlled trial Country: Morocco Study years: not indicated Duration of participation: 9 months
Participants	Iodine‐deficient school children from two neighbouring primary schools in Brikcha Rural Commune, an area of endemic goiter. Boys and girls aged 6 to 15 years old were eligible for participation in the study. Of these, 377 consented and were randomised (188 in control group and 189 in intervention group).
Interventions	Participants were randomly assigned to either the intervention (DFS) or control (IS) groups. Ten children (5 in IS and 5 in DFS) dropped out before study completion, leaving 184 children in the intervention group and 183 in the control group. Setting and administration: salt was provided to participants for use at home. "Each household was provided with 2 kg salt at the beginning of each month for 9 mo to supply all household needs. The salt was dispensed directly to the head of the household from a central supply at the local health center. At baseline, the study was carefully explained to the participating families, and it was emphasized that the new salt should be used for all cooking and food preparation, as well as at the table. This message was reinforced at each of the monthly salt distributions." Intervention Iron compound: encapsulated ferrous sulphate hydrate Iron concentration: 1 mg/g salt Iodine compound: potassium iodide Iodine concentration: 25 µg/g salt Additives to salt: encapsulation ingredients (soy stearin, titanium dioxide, hydroxypropyl methylcellulose, and sodium hexametaphosphate) Co‐interventions: none Control Iodine concentration: 25 µg/g salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: SF concentration, TfR concentration, ZnPP concentration, IDA prevalence All outcomes were measured at endpoint (9 months).
Notes	Comment(s): authors declined to provide further clarification of data Declarations of interest: declared Funding sources: supported by The Nestle Foundation (Lausanne, Switzerland), The Foundation for Micronutrients in Medicine (Rapperswil, Switzerland), and The Swiss Federal Institute of Technology (Zurich, Switzerland)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how random sequence was generated Quote: "children were randomly assigned at the household level into 2 groups".
Allocation concealment (selection bias)	Unclear risk	No specific mention of allocation Quote: "both investigators and households were blind to group assignment."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "both investigators and households were blind to group assignment"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double‐blinding (insufficient information to assess)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Very low loss to follow‐up
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Low risk	Arms balanced at baseline; salt appears to be stable throughout study; declared conflicts of interest and study funders

Zimmermann 2004.

Study characteristics
Methods	Design: double‐blind, randomised controlled trial Country: Morocco Study years: not indicated Duration of participation: 10 months
Participants	Iodine‐deficient school children, with a high prevalence of anaemia from two neighbouring primary schools in the Brikcha Rural Commune, an area of endemic goiter in northern Morocco. Boys and girls aged 6 to 15 years old were eligible for participation in the study. Of these, 163 consented and completed the study.
Interventions	Participants were randomly assigned to either the intervention (DFS) or control (IS) groups (86 in control group and 77 in intervention group). Setting and administration: salt provided to participants for use at home. "Each household was provided with 2 kg salt at the beginning of each month for 10 mo to supply all household needs. The salt was dispensed directly to the head of the household from a central supply at the local health center. At baseline, the study was carefully explained to the participating families, and it was emphasized that the new salt should be used for all cooking and food preparation, as well as at the table. This message was reinforced at each of the monthly salt distributions." Intervention Iron compound: ferric pyrophosphate Iron concentration: 2 mg/g salt Iodine compound: potassium iodate Iodine concentration: 25 µg/g salt Additives to salt: none Co‐interventions: none Control Iodine concentration: 25 µg/g salt
Outcomes	Primary outcomes: Hb concentration, UI concentration Secondary outcomes: SF concentration, TfR concentration, ZnPP concentration All outcomes were measured at endpoint (10 months).
Notes	Comment(s): authors declined to provide further data clarification Declarations of interest: declared Funding sources: supported by the Thrasher Research Fund (Salt Lake City, Utah), the Foundation for Micronutrients in Medicine (Rapperswil, Switzerland), and the Swiss Federal Institute of Technology (Zürich, Switzerland). Paul Lohmann AG (Emmerthal, Germany) supplied the iron pyrophosphate compound and the iron sulfate tablets used in the study
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No mention of how random sequence was generated Quote: "children were randomly divided by household into 2 groups"
Allocation concealment (selection bias)	Unclear risk	No specific mention of allocation Quote: "both investigators and households were blind to group assignment"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "both investigators and households were blind to group assignment"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not discussed by study authors explicitly, beyond personnel being stated as blinded as part of double‐blinding (insufficient information to assess)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Minimal loss to follow‐up (loss of 3 participants in IS group, 2 in the DFS group)
Selective reporting (reporting bias)	Low risk	Study reports on all specified outcomes
Other bias	Low risk	Arms balanced at baseline; salt appears to be stable throughout study; declared conflicts of interest and study funders

CRP: C‐reactive protein; DFS: double‐fortified salt; EFF: encapsulated ferrous fumarate; g: grams; Hb: haemoglobin;ICC: intraclass correlation; IDA: iron deficiency anaemia; IS: iodised salt; kg: kilograms; mg: milligrams; MGFePP: micronised ground ferric pyrophosphate; n: number; SD: standard deviation; SF: serum ferritin; TfR: transferrin receptor; µg: micrograms; UI: urinary iodine; ZnPP: zinc protoporphyrin.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Anonymous 1980	Ineligible intervention (iron fortified salt only)
Anonymous 1982	Ineligible intervention (iron fortified salt only)
Anonymous 1983	Ineligible intervention (iron fortified salt only)
Bhatia 2018	Ineligible study design (policy brief on DFS use in India). No new DFS studies identified from the reference list
Diosady 2018	Ineligible study design (process development paper for DFS from laboratory scale to large‐scale distribution capable of reaching millions of people in India)
Dodd 1997	Ineligible intervention (Group A provided iodised salt but Group C provided an iron supplement (tablet) plus iodised salt)
Horton 2011	Ineligible study design (economic assessment that includes DFS studies). No new DFS studies identified from the reference list
Jain 1987	Ineligible intervention (iron fortified salt only)
Kumar 2020	Ineligible intervention (multiple micronutrients as the main intervention; no separate iron and iodine arm)
Nadiger 1980	Ineligible intervention (iron fortified salt only)
Nair 2013	Ineligible comparator (no control arm; intervention only)
Nair 2014	Ineligible comparator (no control arm; intervention only)
Working Group on Fortification 1982	Ineligible intervention (iron salt only)

Characteristics of studies awaiting classification [ordered by study ID]

Jadhav 2019.

Methods	Design: unclear Country: India Study years: "Between August to December 2018, data collected from 2000 households every month" Duration: 1 year ("~4000 tons of DFS every month for a whole year")
Participants	Population‐level study (authors note "intervention reached 15 million people")
Interventions	Method of allocation is unclear Setting and administration: subsidised DFS was supplied through "Public Distribution System" Intervention:  Iron compound: unclear Iron concentration: "minimum 850‐1100 mg/kg ... to provide 30% DRI of iron for women of reproductive age" Iodine compound: unclear Iodine concentration: "minimum 15 mg/kg ... to provide 100% DRI of iodine" Additives to salt: unclear Co‐interventions: unclear Control:  Iodine compound: unclear Iodine concentration: unclear Additives to salt: unclear
Outcomes	Not reported
Notes	Not reported

DFS: double‐fortified salt; DRI: dietary reference intake.

Characteristics of ongoing studies [ordered by study ID]

CTRI/2019/08/020508.

Study name	Study on iron fortified iodised salt
Methods	Design: randomised, parallel group trial  Country: India Setting and administration: a community‐based, open randomised study between iodised salt and two iron‐fortified iodised salt (DFS1 and DFS2) formulations. The study will be carried out in Anganwadis in Mehrauli Block. Randomisation is on the basis of households. Sample size: randomisation and distribution of salt will be on the basis of households. Average number of persons in the household = 5. Impact will be assessed on the basis of changes in the individual person's Hb level before and after use of salt. Available data indicate that, at the end of 12 months, persons using DFS will have a higher mean Hb of 0.5 g/dL compared to those receiving iodised salt. There are no data on increase in Hb after 18 months of DFS use. The loss to follow‐up can be up to 20% at the end of 18 months. Number of households for the study: 250 per group; 750 households for 3 groups Number of persons for the study: 5/household with 1250 per group = 3750 persons Total duration of the study: 2 years
Participants	Inclusion criteria Healthy girls and boys, men and women (aged 2‐70 years) Likely to stay in the locality for two years Willing to take salt provided by NFI and use it as the sole salt for household cooking Willing to participate in the study and provide blood or urine samples, or both, as and when required Exclusion criteria Unlikely to stay in the locality Not willing to participate in the study Individuals in the household have health problems such as hypertension and diabetes Households will receive either iodine‐fortified (iodised) salt or either of the iron‐fortified iodised salts according to the random allocation
Interventions	Intervention group 1: ferrous sulphate with sodium hexametaphosphate as stabiliser; oral route of administration; salt to be used for cooking the household meals; duration of therapy is 18 months frequency; daily dose 8 g/d/person to 10 g/d/person. This is based on available data suggesting that average per capita intake of salt is between 8 g/d to 10 g/d.  Intervention group 2: ferrous fumarate encapsulated with soy stearin; oral route of administration; salt to be used for cooking the household meals; duration of therapy 18 months frequency; daily dose of 8 g/d/person to 10 g/d/person. This is based on available data suggesting that average per capita intake of salt is between 8 g/d to 10 g/d.  Control group: iodised salt; oral route of administration; salt to be used for cooking the household meals; duration of therapy 18 months frequency; daily dose 8 g/d/person to 10 g/d/person. This is based on available data suggesting that average per capita intake of salt is between 8 g/d to 10 g/d.
Outcomes	Primary outcomes: improvement in haemoglobin status and ferritin concentration Secondary outcomes Feasibility, acceptability, continued use of rates of salt in the three groups will be computed to see if there are any differences between the three groups. Assess organoleptic properties of the three types of salt when used in different preparations commonly consumed Timing of outcome assessment Baseline At baseline, sociodemographic profiles of the households will be obtained. Information on morbidity in the preceding fortnight will be obtained. Height, weight will be measured and BMI computed. Blood pressure will be measured in all. Hb estimation will be done for all available members of the  family. All households will be supplied with 2 kg of one of these three salts on the basis of the random numbers in the first month and requested to use the salt exclusively for cooking at home. Monthly follow‐up Every month consumption of salt will be assessed and salt packets will be provided as per their need (estimated salt consumption for family of six is 2 kg/packet). Data on consumption of salt per consumption unit computed. Iodine and iron in the samples of the remaining salt from households will be tested. Information on morbidity for the last fortnight will be ascertained. Three months  Anthropometric assessment of nutritional status will be carried out for all available members of the family once in three months. Blood pressure will be measured once in three months in all adults. Biochemical estimations At baseline, and then at 6 months, 12 months and 18 months, Hb levels will be estimated in all available members of the family receiving either iodised or iron and iodine fortified salt. In a subsample of 10% of the households, urinary iodine will be measured once in six months. In 50% of the adults, efforts will be made to assess iron stores by estimating ferritin levels.
Starting date	2 September 2019
Contact information	Name: Dr Prema Ramachandran  Address: Nutrition Foundation of India  Email address: premaramachandran@gmail.com
Notes	Comment(s): none Declarations of interest: not reported Funding source: Food Safety and Standards Authority of India

CTRI/2020/11/028936.

Study name	Iron absorption from salt double fortified with iodine and an encapsulated ferric pyrophosphate compound: a stable isotope study in iron‐deficient Indian women
Methods	Design: randomised, parallel group, placebo controlled trial  Country: India Setting and administration: not reported Study years: ongoing Duration of participation (expected): 3 years
Participants	Participants Healthy, non‐pregnant, non‐lactating women (18‐40 years) SF < 25 µg/L Normal BMI (18.5 kg/m2 to25 kg/m2) Bodyweight ≤ 65 kg Sample size: n = 41
Interventions	Intervention: DFS of varying compositions: iodised salt + ferric pyrophosphate encapsulated; iodised salt + ferric pyrophosphate + CA/TSC encapsulated; iodised salt + ferric pyrophosphate + CA/NaPP encapsulated; iodised salt + ferrous fumarate encapsulated; iodised salt + ferrous fumarate non‐encapsulated Control: not reported
Outcomes	Primary outcome: "To compare encapsulated ferrous fumarate with co‐encapsulated ferric pyrophosphate and with solubilizing agents" (what measure is to be compared is not reported) Secondary outcomes "To compare in iron‐deficient Indian women (SF < 25µg/L) fractional iron absorption from double fortified with iodine and different iron compounds" "To determine the iron status and inflammatory markers" "To evaluate the sensory properties of the different double fortified salts in local foods using triangle and acceptability tests" "To determine the stability of color and iodine concentration in double fortified salt during a 6‐month storage trial"
Starting date	1 December 2020
Contact information	Name: Dr. Prashanth Thankachan  Address: Division of Nutrition, St John’s Research Institute, St Johns National Academy of Health Sciences Email address: praxmail@gmail.com
Notes	Comment(s): timing of outcome assessment is unclear Declarations of interest: not reported Funding source: Bill and Melinda Gates Foundation

NCT04404751.

Study name	Effectiveness of quadruple fortified salt in improving haemoglobin levels among anemic women of reproductive age (18‐49 years) in rural low resource setting
Methods	Design: randomised trial, with factorial allocation and triple blinding (participant, investigator, outcome assessor) Country: Tanzania Setting and administration: not reported Study years: not reported Duration of participation: not reported
Participants	Inclusion criteria Adults (aged 18‐49 years) Adult with relatively low Hb level (8 g/dL to 13 g/dL) Able to eat food cooked with salt Able to give consent
Interventions	Three intervention arms of different composition (not specified) Quadruple fortified salt: iodised salt with iron, folic acid and vitamin B12 Other salt: iodised salt with iron Control: standard iodised salt
Outcomes	Primary outcomes: haemoglobin concentration Secondary outcomes: iron store replenishment
Starting date	23 August 2019
Contact information	Not reported
Notes	Comment(s): none Declarations of interest: not reported Funding source: not reported

RIDIE‐STUDY‐ID‐58f6eeb45c050.

Study name	Impact evaluation of integrating double fortified salt (DFS) to reduce anemia in recipients of the PDS programme in UP, India
Methods	Design: controlled before‐after study Country: India Setting and administration: the programme plans to distribute DFS at a subsidised rate through the public distribution system in 10 districts, in UP. Five intervention and five control districts will be matched at the district level, to ensure they match on socioeconomic and demographic characteristics that may influence anaemia (other than DFS). Study years: not reported Duration of participation: 12 months
Participants	Participants: non‐pregnant women of reproductive age (18‐49 years) Sample size: n = 1250 individuals for baseline, 1200 for midline, 6200 for endline
Interventions	Intervention Iron compound: ferrous fumarate Iron concentration: 1 mg/g salt Iodine compound: potassium iodate Iodine concentration: 30 µg/g salt Additives to salt: semolina flour, shortening, titanium dioxide, hydroxypropyl methylcellulose, soy stearin Control Iodine concentration: 30 µg/g salt
Outcomes	Primary outcomes: haemoglobin concentration  Secondary outcomes: serum ferritin, inflammation, prevalence of anaemia
Starting date	1 January 2018
Contact information	Name: Dr. Lynnette M Neufeld
Notes	Comment(s): none Declarations of interest: none declared Funding source: Bill and Melinda Gates Foundation

BMI: body mass index; CA/NaPP: citric acid/sodium pyrophosphate;CA/TSC: citric acid/trisodium citrate; DFS: double‐fortified salt; Hb: haemoglobin; n: number; NFI: Nutrition Foundation of India; PDS: public distribution system; SF: serum ferritin; UP: Uttar Pradesh. 

Differences between protocol and review

1. For the 2021 updated searches, we attempted to access the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP), but the site was unresponsive.

2. We assessed overall risk of bias in each study and further assessed cRCTs and CBA studies using the ROB 1 tool and EPOC guidance. This was to ensure all possible biases were assessed.

3. Previous systematic reviews did not adjust for clustering in the meta‐analyses. Therefore, we also conducted an additional sensitivity analysis for unadjusted estimates from CRCTs, to compare results to adjusted estimates for haemoglobin concentration.

4. Due to limited disaggregated data availability, we revised the groupings for our subgroup analysis by life stage based on cut‐offs used by study authors (versus prespecified age bands), to ensure meaningful conclusions. The originally proposed subgroups 5 to < 10 years old and 10 to < 19 years old were merged to form a single subgroup (5 to 17 years).

5. Preplanned but unused methods are reported in additional Table 7.

2. Unused methods.

Method

Measures of treatment effect > Continuous data We did not use the SMD in our meta‐analyses on urinary iodine, serum ferritin, transferrin receptor concentration and zinc protoporphyrin, as these outcomes were measured using the same scale.

Dealing with missing data  We did not impute any SDs for outcomes using the average of the pooled baseline SD from trials reporting this information (Higgins 2019), as the SDs were either available or we were able to obtain them from the study authors, or we could recreate them using other data in the trial reports (median, range, sample size).

Assessment of reporting bias  Where we suspected reporting bias and could not obtain missing information from authors and the missing data were thought to introduce substantial bias, we conducted a sensitivity analysis to explore the impact of including or excluding such studies in relation to their respective outcomes.

Subgroup analysis and investigation of heterogeneity We did not conduct the following subgroup analyses due to limited data. Adjustment of iron deficiency anaemia estimates for inflammatory markers (adjustment for inflammation; no adjustment for inflammation). As iron biomarkers are influenced by inflammation, adjustments should be made to cut‐offs if inflammation is detected (although inflammatory marker data — e.g. CRP, AGP — are not always collected because of the associated cost).

Sensitivity analysis  We did not perform the following sensitivity analyses due to limited data. Trials reported in non‐peer‐reviewed sources or identified from grey literature. Trials for which we imputed values for outcomes of interest.

AGP: α 1 ‐acid glycoprotein; CRP: C‐reactive protein; SD: standard deviation; SMD: standardised mean difference. 

Contributions of authors

Jo‐Anna B Baxter (J‐ABB), Stanley H Zlotkin (SHZ), and Zulfiqar Bhutta (ZAB) conceived the idea for the review.

J‐ABB and Bianca Carducci (BC) co‐ordinated the review.

J‐ABB, BC and Mahdis Kamali (MK) assessed studies for inclusion in the review, collected data for the review, and assessed the risk of bias in the included studies.

J‐ABB developed and wrote the background and methods, assessed the certainty in the body of evidence, and contributed to writing the results and discussion.

BC analysed the data, assessed the certainty in the body of evidence, wrote the results, and contributed to writing the discussion.

ZAB has overall responsibility for the review.

All authors contributed to interpreting the data, and reviewed and approved the final manuscript.

Sources of support

Internal sources

• Hospital for Sick Children, Canada

  Salaries for BC, SHZ, and ZAB.

• Offord Centre for Child Studies, Canada

  Salary for MK.

External sources

• Gates Ventures, USA

  Our institution (The Hospital for Sick Children) holds funding from Gates Ventures for a range of desk‐based research activities, including this review.

• Vanier Canada Graduate Scholarship, Canada

  Dissertation research scholarship for J‐AB

• Ontario Graduate Scholarship, Canada

  Dissertation research scholarship for BC

Declarations of interest

J‐ABB: has declared that they have no conflict of interest.

BC: has declared that they have no conflict of interest.

MK: has declared that they have no conflict of interest.

SHZ: reports being involved in a study eligible for inclusion in the work; the study was funded by the Micronutrient Initiative but they had no control over the study design, methods, data analysis and reporting. SHZ also reports being a member of the Home Fortification Technical Advisory Group in an unpaid position.

ZAB: has declared that they have no conflict of interest.  

These authors contributed equally to this work.

These author contributed equally to this work.

New