Replacing salt with low‐sodium salt substitutes (LSSS) for cardiovascular health in adults, children and pregnant women

Abstract

Background

Elevated blood pressure, or hypertension, is the leading cause of preventable deaths globally. Diets high in sodium (predominantly sodium chloride) and low in potassium contribute to elevated blood pressure. The WHO recommends decreasing mean population sodium intake through effective and safe strategies to reduce hypertension and its associated disease burden. Incorporating low‐sodium salt substitutes (LSSS) into population strategies has increasingly been recognised as a possible sodium reduction strategy, particularly in populations where a substantial proportion of overall sodium intake comes from discretionary salt. The LSSS contain lower concentrations of sodium through its displacement with potassium predominantly, or other minerals. Potassium‐containing LSSS can potentially simultaneously decrease sodium intake and increase potassium intake. Benefits of LSSS include their potential blood pressure‐lowering effect and relatively low cost. However, there are concerns about potential adverse effects of LSSS, such as hyperkalaemia, particularly in people at risk, for example, those with chronic kidney disease (CKD) or taking medications that impair potassium excretion.

Objectives

To assess the effects and safety of replacing salt with LSSS to reduce sodium intake on cardiovascular health in adults, pregnant women and children.

Search methods

We searched MEDLINE (PubMed), Embase (Ovid), Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science Core Collection (Clarivate Analytics), Cumulative Index to Nursing and Allied Health Literature (CINAHL, EBSCOhost), ClinicalTrials.gov and WHO International Clinical Trials Registry Platform (ICTRP) up to 18 August 2021, and screened reference lists of included trials and relevant systematic reviews. No language or publication restrictions were applied.

Selection criteria

We included randomised controlled trials (RCTs) and prospective analytical cohort studies in participants of any age in the general population, from any setting in any country. This included participants with non‐communicable diseases and those taking medications that impair potassium excretion. Studies had to compare any type and method of implementation of LSSS with the use of regular salt, or no active intervention, at an individual, household or community level, for any duration.

Data collection and analysis

Two review authors independently screened titles, abstracts and full‐text articles to determine eligibility; and extracted data, assessed risk of bias (RoB) using the Cochrane RoB tool, and assessed the certainty of the evidence using GRADE. We stratified analyses by adults, children (≤ 18 years) and pregnant women. Primary effectiveness outcomes were change in diastolic and systolic blood pressure (DBP and SBP), hypertension and blood pressure control; cardiovascular events and cardiovascular mortality were additionally assessed as primary effectiveness outcomes in adults. Primary safety outcomes were change in blood potassium, hyperkalaemia and hypokalaemia.

Main results

We included 26 RCTs, 16 randomising individual participants and 10 randomising clusters (families, households or villages). A total of 34,961 adult participants and 92 children were randomised to either LSSS or regular salt, with the smallest trial including 10 and the largest including 20,995 participants. No studies in pregnant women were identified. Studies included only participants with hypertension (11/26), normal blood pressure (1/26), pre‐hypertension (1/26), or participants with and without hypertension (11/26). This was unknown in the remaining studies. The largest study included only participants with an elevated risk of stroke at baseline. Seven studies included adult participants possibly at risk of hyperkalaemia. All 26 trials specifically excluded participants in whom an increased potassium intake is known to be potentially harmful. The majority of trials were conducted in rural or suburban settings, with more than half (14/26) conducted in low‐ and middle‐income countries. The proportion of sodium chloride replacement in the LSSS interventions varied from approximately 3% to 77%. The majority of trials (23/26) investigated LSSS where potassium‐containing salts were used to substitute sodium. In most trials, LSSS implementation was discretionary (22/26). Trial duration ranged from two months to nearly five years. We assessed the overall risk of bias as high in six trials and unclear in 12 trials. LSSS compared to regular salt in adults: LSSS compared to regular salt probably reduce DBP on average (mean difference (MD) ‐2.43 mmHg, 95% confidence interval (CI) ‐3.50 to ‐1.36; 20,830 participants, 19 RCTs, moderate‐certainty evidence) and SBP (MD ‐4.76 mmHg, 95% CI ‐6.01 to ‐3.50; 21,414 participants, 20 RCTs, moderate‐certainty evidence) slightly. On average, LSSS probably reduce non‐fatal stroke (absolute effect (AE) 20 fewer/100,000 person‐years, 95% CI ‐40 to 2; 21,250 participants, 3 RCTs, moderate‐certainty evidence), non‐fatal acute coronary syndrome (AE 150 fewer/100,000 person‐years, 95% CI ‐250 to ‐30; 20,995 participants, 1 RCT, moderate‐certainty evidence) and cardiovascular mortality (AE 180 fewer/100,000 person‐years, 95% CI ‐310 to 0; 23,200 participants, 3 RCTs, moderate‐certainty evidence) slightly, and probably increase blood potassium slightly (MD 0.12 mmol/L, 95% CI 0.07 to 0.18; 784 participants, 6 RCTs, moderate‐certainty evidence), compared to regular salt. LSSS may result in little to no difference, on average, in hypertension (AE 17 fewer/1000, 95% CI ‐58 to 17; 2566 participants, 1 RCT, low‐certainty evidence) and hyperkalaemia (AE 4 more/100,000, 95% CI ‐47 to 121; 22,849 participants, 5 RCTs, moderate‐certainty evidence) compared to regular salt. The evidence is very uncertain about the effects of LSSS on blood pressure control, various cardiovascular events, stroke mortality, hypokalaemia, and other adverse events (very‐low certainty evidence). LSSS compared to regular salt in children: The evidence is very uncertain about the effects of LSSS on DBP and SBP in children. We found no evidence about the effects of LSSS on hypertension, blood pressure control, blood potassium, hyperkalaemia and hypokalaemia in children.

Authors' conclusions

When compared to regular salt, LSSS probably reduce blood pressure, non‐fatal cardiovascular events and cardiovascular mortality slightly in adults. However, LSSS also probably increase blood potassium slightly in adults. These small effects may be important when LSSS interventions are implemented at the population level. Evidence is limited for adults without elevated blood pressure, and there is a lack of evidence in pregnant women and people in whom an increased potassium intake is known to be potentially harmful, limiting conclusions on the safety of LSSS in the general population. We also cannot draw firm conclusions about effects of non‐discretionary LSSS implementations. The evidence is very uncertain about the effects of LSSS on blood pressure in children.

Summary of findings

Background

Description of the condition

High blood pressure is the leading cause of preventable deaths worldwide, contributing to more than 10 million deaths and 211 million disability‐adjusted life years annually, mainly due to acute coronary syndrome (formerly called ischaemic heart disease) and stroke (Forouzanfar 2017).

Hypertension is typically defined by a diastolic blood pressure (DBP) ≥ 90 mmHg and a systolic blood pressure (SBP) ≥ 140 mmHg, although recent guidelines define stage 1 hypertension as a SBP ranging from 130 to 139 mmHg and a DBP ranging from 80 to 89 mmHg to reflect current blood‐pressure lowering targets (Arnett 2019).  After 50 years of age, SBP increases disproportionately to DBP in many individuals due to factors such as reduced arterial stiffness (Franklin 2011; Lee 2010), with an elevated SBP being a prominent risk factor for cardiovascular events in older people (Staessen 2000). In 2015, an estimated 874 million adults had a SBP of 140 mmHg or higher (Forouzanfar 2017).

High sodium together with insufficient potassium intake contribute to hypertension, thereby increasing the risk of cardiovascular disease and stroke. Current global estimates of sodium intake are 3950 mg (172 mmol) per person per day (Powles 2013), which equates to nearly ten grams of salt (sodium chloride) per person per day. For sodium, current World Health Organization (WHO) guidelines strongly recommend reducing intake in adults to < 2 g/day sodium (equating to about 5 g salt per day) and a downward adjusted intake in children (WHO 2012a). Global estimates of potassium intake for all ages, education levels, residences and sexes in 2018 are 2.3 grams per person per day (Global Dietary Database 2022), which equates to an intake of 59 mmol per person per day. For potassium, WHO conditionally recommends an intake in adults of at least 90 mmol/day (3150 mg/day) and a downward adjusted intake in children (WHO 2012b).  Although antihypertensive drug therapy is an effective method for controlling blood pressure, poor adherence to antihypertensive therapy substantially increases the near‐ and long‐term risk of stroke among patients with hypertension (Herttua 2013), and access to health care such as blood‐pressure lowering medication is not universally available.

Hypertension is also a major contributor to the development and progression of chronic kidney disease (CKD). Adequate blood pressure control has been shown to be effective in slowing the progression of CKD to end stage renal disease (ESRD). In addition, adequate treatment of diabetes and cardiovascular risk factors such as dyslipidemia, are also linked to lower rates of progression to ESRD, and associated with significant reductions in cardiovascular morbidity and mortality (Couser 2011).

Description of the intervention

The WHO target of a 30% relative reduction in mean population salt/sodium intake by 2025 requires effective and safe strategies to reduce population intake. One of several existing salt reduction strategies is using salt products with lower concentrations of sodium ‐ usually replaced by potassium or other minerals, or both. These low‐sodium salt substitutes (LSSS) vary widely in their formulations and are available in high‐income as well as low‐ and middle‐income countries. In many LSSS, a proportion of sodium chloride (NaCl) is replaced with potassium chloride (KCl), which shares many properties with NaCl but also has unwanted relatively offensive side tastes (bitter, acrid, and metallic). A recent narrative review (Cepanec 2017) described the many various formulations of KCl‐based LSSS, which include the use of numerous taste‐improving agents (TIAs) and formulation concepts. Authors concluded that “within the great number of various compositions of KCl‐based salt substitutes, presumably the most effective ones are based on well‐balanced mixtures of KCl and NaCl, maintaining a sodium reduction range from −25% to −50% (relative to NaCl), which always include certain percentages of one or more TIAs. A typical formulation of a KCl‐based salt substitute with 50% in sodium reduction is 50% NaCl + 30‐45% KCl + 5‐20% taste‐improving agents.” Incorporating salt substitutes into population strategies to reduce sodium intake has increasingly been recognised by health authorities and public health organisations (Greer 2020), especially in countries where the majority of sodium intake comes from the discretionary use of salt by households.

How the intervention might work

The dose‐response relationship between reduced dietary sodium and blood pressure change was examined in a recent systematic review (133 studies with 12,197 participants). Authors showed that in diverse populations, lower sodium intakes resulted in blood pressure reductions, with greater reductions in sodium intake producing greater reductions in BP (Huang 2020). Additionally, older and non‐white populations (for SBP), as well as those with higher baseline blood pressure (for SBP and DBP) achieved greater blood pressure reductions from the same amount of sodium reduction (Huang 2020). Reductions in blood pressure, such as a reduction of 5% in SBP, translate to important reductions (10%) in the risk of major cardiovascular events (e.g. fatal or non‐fatal stroke or myocardial infarction), as demonstrated by a recent meta‐analysis of individual participant‐level data from 48 trials of blood‐pressure lowering medication (Rahimi 2021). Observational studies have demonstrated that stroke risk is inversely associated with dietary potassium intake (Vinceti 2016). In addition, data from randomised clinical trials have shown that potassium supplements have a blood‐pressure lowering effect in people with hypertension, particularly those with a high sodium intake (Filippini 2017). As described in the Description of the condition section, global estimates of potassium intake are lower than what is currently recommended by the WHO. The low dietary intake of potassium, in addition to high dietary sodium intake, contributes to hypertension. Therefore, interventions or strategies promoting the use of a potassium‐enriched LSSS could aid in reducing sodium intake, while concurrently increasing potassium intake, at the population level.

Reduction in sodium intakes ‐ either through reduction of dietary salt intake, salt substitution, or a combination of these ‐ may also be a practical choice for patients with hypertension who are resistant to antihypertensive medications or who experience side effects from medications. It may also play an important role as an adjunctive therapy in the management of hypertensive individuals by potentially lowering the doses of antihypertensive medication required. In cases where the behavioural changes required to reduce dietary salt intake are very difficult or unfeasible, salt substitutes may offer convenience and practicality. Therefore, salt substitution as a cost‐effective strategy could result in reductions in health‐care costs associated with non‐communicable diseases at a population level. However, it should be noted that if foods with high levels of non‐discretionary sodium chloride are regularly consumed, the discretionary use of LSSS may not result in a sufficient reduction in sodium intake to be beneficial.

LSSS may offer a potential solution for the food industry to develop lower sodium food products without compromising on taste or safety, particularly in countries where non‐discretionary sodium intake contributes significantly to the overall population intake of sodium. However, because KCl costs more than NaCl, significant consumer demand, industry‐targeted subsidies or taxes on high sodium content foods will likely be required before the food industry will absorb the costs of product reformulation. Therefore, the application of LSSS strategies at population level to reduce sodium consumption is dependent on several factors, including its main uses within a population, as well as its effects on food taste and cost (Greer 2020).

The greatest risk with potassium‐based LSSS is the potential for adverse effects resulting from hyperkalaemia, particularly the increased risk of arrhythmias and sudden cardiac death. The risk of adverse events is greater at higher levels of serum potassium. There is no absolute threshold at which these adverse events occur, however a serum potassium level of ≥ 6.0 mmol/L is commonly considered to be a clinically significant threshold above which the most serious manifestations of hyperkalaemia occur (Ahee 2000; Hollander‐Rodriguez 2006). Multiple factors influence the occurrence of these adverse events, such as the underlying cause of hyperkalaemia and the rate at which serum potassium increases. High intakes of dietary potassium have not been linked to adverse effects in healthy adults and children with normal kidney function. However, the effects of high dietary potassium intakes on the risk of adverse effects are a key concern among people with impaired potassium excretion, such as those with chronic kidney disease or taking medications that impair potassium excretion (Greer 2020; Kovesdy 2018).

A reduction of dietary sodium intake through the population‐level implementation of LSSS may also result in hyponatraemia in people with impaired renal function (Sahay 2014), including older people and people treated with thiazide diuretics (Upadhyay 2009).

Why it is important to do this review

If the best available evidence on replacing salt with LSSS shows adequate effectiveness and safety for important outcomes, it could be recommended as a population‐level intervention for reducing cardiovascular disease risk. However, concerns exist about potential adverse effects of LSSS, such as hyperkalaemia, particularly in those at risk, such as people with chronic kidney disease or on medications that impair potassium excretion.

The WHO is currently developing a guideline on the use of LSSS in adults and children. This review was commissioned by the WHO Nutrition Guidance Expert Advisory Group (NUGAG) Subgroup on Diet and Health in order to inform and contribute to the development of a WHO recommendation on the use of LSSS for this guideline. The results of this review, including Grading of Recommendations, Assessment, Development and Evaluations (GRADE) assessments, were discussed and reviewed by the WHO NUGAG Subgroup on Diet and Health as part of their guideline development process.

Objectives

To assess the effects and safety of replacing salt with LSSS to reduce sodium intake on cardiovascular health in adults, pregnant women and children.

Methods

Criteria for considering studies for this review

Types of studies

The Populations, Intervention, Comparison, and Outcomes (PICO) were agreed by the WHO NUGAG Subgroup on Diet and Health, who ranked the outcomes and also agreed on subgroups and study designs to be included. As per these agreements and our prospective registration on the international prospective register of systematic reviews (PROSPERO 2020 CRD42020180162; available at https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=180162), we included individually randomised controlled trials (RCTs) and cluster‐randomised controlled trials (cluster‐RCTs) with true randomisation methods, regardless of the unit of allocation. We also planned to include prospective analytical cohort studies, where LSSS intake/exposure was assessed at baseline and related to any of the prespecified outcomes at a later time point using empirical data. We excluded RCTs with a cross‐over trial design if data for the first phase per group were unavailable, due to the possible period and carry‐over effects that would arise with the eligible dietary interventions/exposures and outcomes not being easily reversible, as required for a valid cross‐over design (Younge 2015). We additionally excluded cluster‐RCTs with fewer than two intervention and two control clusters.

Types of participants

We included studies in the general population, from any setting in any country, including participants with the following condition(s) and/or risk factors: hypertension, cardiovascular disease (CVD), diabetes mellitus, renal impairment and those taking medications that impair potassium excretion.

In accordance with the WHO NUGAG PICO conceptualisations and agreements,  the following three comparisons were planned if data allowed:

LSSS versus regular salt or no active intervention in adults (aged 18 years and older)LSSS versus regular salt or no active intervention in children aged 2 to < 18 yearsLSSS versus regular salt or no active intervention in pregnant women

Types of interventions

We included studies that assessed the health effects associated with the use of LSSS at an individual, household or community level. LSSS interventions/exposures of any type or duration were included, provided they aimed to replace the dietary intake of any amount of sodium with another mineral or compound. Studies investigating either discretionary (i.e. salt on table or added during cooking) or non‐discretionary use of LSSS (i.e. included during food manufacturing), or both, were included.

Eligible comparators/controls included the use of regular salt (NaCl) or no active intervention to reduce salt intake. Studies where the control group received only basic information on sodium reduction at baseline, were included. Studies with multi‐component interventions were included if effects of LSSS could be isolated from the multifactorial design.

We excluded studies with multi‐component interventions if the additional intervention components were not aimed primarily at promoting LSSS use by participants or communities, but were instead focussed more broadly on reducing sodium intake (e.g. changing lifestyle and dietary behaviour of which LSSS use is only one component) or aimed at improving health in general (e.g. counselling for exercise or smoking cessation), such that LSSS effects could not be isolated.

Types of outcome measures

We did not exclude studies on the basis of outcomes measured. However, we did exclude studies measuring only sensory or organoleptic outcomes (e.g. taste of or preference for LSSS).

Primary outcomes

Table 3, Table 4 and Table 5 detail the prespecified primary outcomes for each comparison, with outcome ranking by the WHO NUGAG Subgroup on Diet and Health indicated as follows: critical^c, importantⁱ and not important ⁿⁱ.  The following primary outcomes were regarded as safety outcomes related to the intake of LSSS with potassium: change in blood potassium, hyperkalaemia and hypokalaemia.

Secondary outcomes

Table 3, Table 4 and Table 5 detail the prespecified secondary outcomes for each comparison, with outcome ranking by the WHO NUGAG Subgroup on Diet and Health indicated as follows: critical^c, importantⁱ and not important ⁿⁱ.  The following secondary outcomes were regarded as safety outcomes related to the intake of LSSS with potassium: adverse events, renal function and hyponatraemia.

Search methods for identification of studies

The search strategy was developed, peer‐reviewed and implemented by Cochrane information specialists in consultation with the review team. We used a comprehensive search strategy aiming to identify all eligible studies regardless of language, publication type or publication status. Publication date restrictions were not imposed, except for conference abstracts identified through Embase which covered only those published in the past two years. With this, we specifically aimed to find recent proceedings of studies that may not yet have been published as full articles at the time of the search. We used filters for trials (Lefebvre 2022), cohort studies (Li 2019) and adverse effects (Golder 2006; Golder 2012) to inform our search strategy.

Electronic searches

We aimed to identify RCTs and prospective analytical cohort studies through systematic searches of the following bibliographic databases:

MEDLINE (PubMed, from 1946 to 18 August 2021)Embase (Ovid, from 1947 to 18 August 2021)Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library (Issue 8 of 12, 2021)Web of Science Core Collection with Indexes SCI‐Expanded, SSCi, CPCI‐S (Clarivate Analytics, from 1970 to 18 August 2021)Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCOhost, from 1937 to 18 August 2021)

We also conducted a search of ClinicalTrials.gov (www.ClinicalTrials.gov) and the WHO International Clinical Trials Registry Platform (ICTRP) for ongoing and unpublished trials (https://trialsearch.who.int/). The date of the last searches here was also 18 August 2021. Search strategies per database/registry searched are detailed in Appendix 1.

Searching other resources

To identify any additional eligible records, two reviewers also screened the reference lists of three recent systematic reviews evaluating the effects of LSSS use (Hernandez 2019; Jafarnejad 2020; Jin 2020), as well as the bibliographies of all studies included in this review.

Data collection and analysis

Selection of studies

After de‐duplication of search records, titles and abstracts were screened independently by two reviewers using Covidence (Covidence). Full‐text articles for all records identified as potentially eligible for inclusion were then screened by two reviewers independently to determine final eligibility. Records where we could not obtain the full text or more details of the study in order to determine eligibility, were classified as ‘Studies awaiting classification’.  We resolved any disagreements between reviewers at any stage of the eligibility assessment process through discussion and consultation with a third reviewer, where necessary.

Data extraction and management

Two reviewers independently extracted data onto forms designed and piloted for the review, and we resolved any disagreements during the data extraction and management process through discussion and consultation with a third reviewer, where necessary. Where necessary, translations of records in non‐English were obtained. We extracted data on the following:

Study details, including author details, conflict of interest declaration, funding source, settingMethods, including design, aim, dates, limitations as reported by authors, sample size calculation, participants, including eligibility criteria; method of recruitment; number of clusters per trial arm and how authors accounted for the effect of clustering; participant flow details such as number assessed for eligibility, number randomised; baseline characteristic such as demographic and lifestyle characteristics, health status and intake of sodium and potassium; and any differences in these characteristics by trial armInterventions/exposures, including description, delivery/use, addition of fortificants, duration, co‐interventions, integrity of deliveryComparators, including description, delivery, duration, co‐interventions, integrity of deliveryOutcomes, including numeric data relevant to all primary and secondary outcomes according to the following time point ranges, when available: baseline to 3 months, > 3 to 12 months and > 12 months, except for cardiovascular events, all‐cause mortality, cardiovascular mortality and adverse events, for which data were extracted for the duration of the study. When outcome data were reported at more than one point, we extracted data from the latest point available. For studies that did not use the International System of Units (SI) to report outcomes, we converted values to SI units, where possible. For trials, we extracted change data (change in the outcome from baseline to outcome assessment) with relevant data on variance for intervention and control groups (along with numbers of participants at the time point). Where change data were not available, we extracted end‐values at the time point, along with the variance and numbers of participants for each group, or mean differences (MDs) and measures of variance per group. Where outcome data were only reported per subgroup of the total sample of study participants (e.g. participants with hypertension and participants with normal blood pressure), we extracted these data and calculated the combined mean and standard deviation (SD) for the total sample according to the guidance by (Higgins 2020a), where possible. We preferentially extracted and used supine over standing blood pressure measurements, 24‐hour measurements over measurements done at a single time point, and ambulatory measurements over those conducted in a clinic setting. For cohort studies, we planned to extract the most adjusted odds ratio, risk ratio, mean change or mean end values per group, when comparing the most exposed group of participants with the least exposed group, and the most adjusted regression outputs when LSSS intake was assessed at baseline and related to an outcome measure later.

Assessment of risk of bias in included studies

We assessed the risk of bias in RCTs and cluster‐RCTs using the Cochrane tool for assessment of risk of bias (Higgins 2017). Two reviewers conducted these assessments independently for each included study.  We resolved disagreements by discussion or through consultation with a third reviewer. We assessed the risk of bias for RCTs according to the following domains:

Random sequence generation (selection bias)Allocation concealment (selection bias)Blinding of participants and personnel (performance bias)Blinding of outcome assessment (detection bias)Incomplete outcome data (attrition bias)Selective outcome reporting (reporting bias)Other bias

We also assessed the risk of bias for cluster‐RCTs according to the following domains (Higgins 2017):

Recruitment bias (selection bias)Comparability with RCTsBaseline imbalance (selection bias)Loss of clusters (attrition bias)Incorrect analysis

For cohort studies, we planned to use the following domains to assess risk of bias (Naude 2018):

Were adequate outcome data available?Was there matching of less‐exposed and more‐exposed participants for prognostic factors associated with outcome, or were relevant statistical adjustments done?Did the exposures between groups differ in components other than only LSSS exposure?Could we be confident in the assessment of outcomes?Could we be confident in the assessment of exposure?Could we be confident in the assessment of presence or absence of prognostic factors?Was selection of less‐exposed and more‐exposed groups from the same population?

Overall risk of bias assessment

As this review addressed mainly objective outcomes (e.g. blood pressure measurements, laboratory‐determined electrolyte values), we did not regard blinding to be of key importance for informing judgements on overall bias. Consequently, we judged overall risk of bias for each included study using two key domains for RCTs and four key domains for cluster‐RCTs, as follows:

RCTs: allocation concealment (selection bias) and incomplete outcome data (attrition bias), andCluster‐RCTs: baseline imbalance (selection bias), recruitment bias (selection bias), incomplete outcome data (attrition bias) and loss of clusters (attrition bias).

We assessed the overall risk of bias of each included study as follows:

low risk (low risk of bias for all key domains);high risk (high risk of bias for one or more key domains); orunclear risk (unclear risk of bias for one or more key domains).

For cohort studies, we planned to consider domains relevant to confounding to inform judgements of the overall risk of bias.

Measures of treatment effect

For dichotomous outcomes, we presented proportions; for two‐group comparisons where numbers of events and participants were provided, we presented results as risk ratios (RRs) with 95% confidence intervals (CIs). Where event rates were reported per person‐years followed in separate groups, we calculated incidence rate ratios (IRRs) with 95% CIs to enable meta‐analysis of these studies with studies reporting rate ratios for the same outcomes. Rate ratios were calculated by dividing the rate in the intervention group by the rate in the control group. The 95% confidence interval (95% CI) of these rate ratios were calculated by taking the antilogarithm of the natural log of the rate ratio (log(IRR)), plus or minus 1.96 times the standard error of the log(IRR) (Boston University School of Public Health 2018). Briefly, the standard error was calculated as the square root of the sum of the inverse of events in the intervention and control group.

Where hazard ratios (HRs) were reported for incident hypertension in the stepped‐wedge trial (Bernabe‐Ortiz 2014), we presented these results with 95% CIs. Due to the different way of analysing these data and the unique design of this trial, these measures were not combined with other data reporting on hypertension.

For continuous outcomes, we used the mean difference (MD) with 95% CIs if outcomes were measured in the same way between trials. Where continuous data were reported using different units across included studies, we planned to calculate and present the standardised mean difference (SMD).

Unit of analysis issues

Studies with more than two intervention groups

For the single study with more than two intervention groups (Pan 2017), we combined event outcome data reported separately for both intervention groups (LSSS < 50% KCl and LSSS ≥ 50% KCl) in our meta‐analyses. These intervention groups were combined using the methods set out in the Cochrane Handbook (Higgins 2020a). Another study randomised participants to receive LSSS or continue with their usual practice, after which intervention participants were again randomised to receive LSSS with or without price subsidy (Li 2016). As the LSSS intervention was the same in both these arms, we only extracted and used data for the overall LSSS group (both with and without subsidy) and the usual practice group.

Cluster‐RCTs

Four included cluster‐RCTs did not report sufficient information on adjustment for clustering in the statistical analysis or results section of either the full text (Hu 2018; Li 2014; Zhou 2013), or conference abstract (Zhang 2015).  We calculated the effective sample sizes for these trials by calculating the design effect (DE), which is 1 + (c ‐ 1) x ICC, where c is the average cluster size.  Our calculations were based on an estimated intra‐cluster correlation coefficient (ICC) of 0.04, reported by a study conducted in similar trial settings in Northern China (Neal 2017). For continuous data (e.g. DBP, SBP), we adjusted for the sample size only; while for dichotomous outcomes (e.g. cardiovascular events) we divided both the sample size and the number of people who experienced the event by the design effect. Where cluster‐RCTs reporting rates did not account for the effect of clustering in its analyses, we adjusted for clustering by inflating the standard errors by multiplying the standard error of the log(IRR) by the square root of the DE (Higgins 2011). All the estimates from cluster‐RCTs were combined with those from RCTs that had individual group assignment in our meta‐analyses (Higgins 2020b).

Dealing with missing data

We contacted study authors to request any missing or unreported data, such as group means, SDs, details of attrition, or details of the type of analysis conducted (e.g. intention‐to‐treat).

In cases where there were missing data due to attrition, we used the data available to conduct available case (modified intention‐to‐treat) meta‐analyses. We assessed the extent and impact of missing data and attrition for each included study during the Risk of bias assessment.

Assessment of heterogeneity

For each meta‐analysis, we examined the forest plots visually to determine whether heterogeneity of the size and direction of treatment effect was present between studies. We used the I² statistic, Tau², and the Chi² test to estimate the level of heterogeneity among the studies in each analysis. We defined substantial heterogeneity as Tau² > 0, and either I² > 50% or a low P value (< 0.10) in the Chi² test. Where substantial heterogeneity was found, we noted this in the text and explored it by conducting prespecified subgroup analyses to account for potential sources of clinical heterogeneity (see section: Subgroup analysis and investigation of heterogeneity). We also considered other potential sources of heterogeneity, for example, differences in the nature of the interventions delivered. In addition, we explored methodological sources of heterogeneity by examining studies with different levels of risk of bias in a sensitivity analysis (see section: Sensitivity analysis). We used caution in the interpretation of results with high levels of unexplained heterogeneity. We did not perform a meta‐analysis if the I² statistic was 90% or higher (considerable heterogeneity) (Deeks 2020).

Assessment of reporting biases

Where more than 10 included studies addressed a primary outcome, we used funnel plots to assess the possibility of small‐study effects and, in the case of asymmetry, intended to consider various explanations such as publication bias, poor study design and the effect of study size (Sterne 2017).

Data synthesis

All syntheses were conducted using Review Manager Web 2021 (RevMan Web 2021). We used a random‐effects meta‐analysis to combine data across more than one study, as we anticipated that there may be natural heterogeneity between studies, attributable to the different study settings, intervention strategies, or both. If a study only reported an MD and variance per group for an outcome, we first calculated MDs and 95% CI for the other studies reporting on that outcome, and then combined MDs and 95% CIs from all studies in a meta‐analysis using generic inverse variance (GIV). If a study reported rate ratios or events per person‐years, from which rate ratios could be calculated (see section: Measures of treatment effect), we also combined these rate ratios and 95% CIs in a meta‐analysis using GIV. Where studies reported event outcomes as rate ratios and risk ratios, these were combined in a meta‐analysis using GIV by using rate ratios as approximations for risk ratios.

We sought to only generate pooled estimates where data from separate studies were similar enough to be combined (see section: Assessment of heterogeneity). Data not suitable for pooling (defined as considerable heterogeneity, I² ≥ 90%) in meta‐analyses were presented in forest plots without the pooled estimate, or in tables, as appropriate. Data from peer‐reviewed publications and conference abstracts were eligible for inclusion in meta‐analysis. Data from conference abstracts were identified in forest plots using footnotes. If needed, we also planned to conduct a narrative synthesis, by adopting a systematic approach to presentation, guided by the reporting guideline, Synthesis Without Meta‐analysis (SWiM) in systematic reviews (Campbell 2020).

Subgroup analysis and investigation of heterogeneity

We performed subgroup analyses where data allowed using a test for interaction (i.e. heterogeneity across subgroups rather than across studies), calculating summary effect sizes for each subgroup in a univariate analysis for prespecified subgroups provided by WHO NUGAG, as follows.

All comparisons:

Study duration: short‐term (≤ 3 months) versus medium‐term (> 3 to 12 months) versus long‐term (> 12 months)Gender: male versus female versus mixed versus unknownEthnicity: African versus Asian versus European versus mixed versus conducted in one setting (e.g. Europe), but ethnicity unspecifiedBlood pressure status (as defined by study authors): hypertensive versus normotensive versus hypotensive versus mixed versus unknownBaseline potassium intake: lower (urinary 24‐hour [24‐h] potassium excretion < 59 mmol [2.3 g] per day) versus higher (urinary 24‐h potassium excretion ≥ 59 mmol [2.3 g] per day) versus unknown or not reported as 24‐h excretion; based on global potassium intake estimates (Global Dietary Database 2022)Baseline sodium intake: lower (urinary 24‐h sodium excretion < 172 mmol [3.95 g sodium or 9.88 g sodium chloride] per day) versus higher (urinary 24‐h sodium excretion ≥ 172 mmol [3.95 g sodium or 9.88 g sodium chloride] per day) versus unknown or not reported as 24‐h excretion; based on global sodium intake estimates (Powles 2013)Iodine status (as defined by study authors): within normal ranges versus insufficient or deficient versus mixed versus unknownType of LSSS: based on proportion of potassium chloride: ≥ 30% KCl versus < 30% KCl versus unknown versus non‐potassium containing LSSS (based on the description of a 'typical' formulation of potassium‐based salt substitutes with a 50% sodium chloride reduction in Cepanec 2017)LSSS implementation: discretionary only (through added LSSS in cooking and at table) versus non‐discretionary only (through consumption of manufactured products) versus discretionary and non‐discretionarySalt as fortification vehicle: using salt as fortification vehicle versus not or unknown

All comparisons (only safety outcomes):

Possible risk of hyperkalaemia versus not at risk or unclear risk of hyperkalaemia (according to the criteria and assessment in Table 6), regardless of heterogeneity, only in the primary analyses of the following safety outcomes: change in blood potassium, hyperkalaemia, hypokalaemia and adverse events. The WHO NUGAG made the decision to limit this subgrouping to the safety outcomes since there are no clinical justifications to expect differences in the effects of LSSS on the effectiveness outcomes in populations possibly at risk of hyperkalaemia.

Additionally, for adults:

Age: adults younger than 65 years versus 65 years and older versus mixed ages versus unknown agesBody mass index (BMI): underweight (< 18.5 kg/m²) versus normal weight (18.5 to 24.9 kg/m²) versus overweight (25 to 29.9 kg/m²) versus obese (≥ 30 kg/m²) for non‐Asian adults, or underweight (< 18.5 kg/m²) versus normal weight (18.5 to 22.9 kg/m²) versus overweight (23 to 24.9 kg/m²) versus obese (≥ 25 kg/m²) for Asian adults (WHO 2000)

Additionally, for children:

Age at start of study: 2 to 5 years versus 6 to 12 years versus 13 to 18 years versus mixed versus unknown

We also planned the following additional subgroup analyses, but available data did not allow these:

Term of pregnancy at start of study: first trimester versus second trimester versus third trimester versus mixed versus unknown (in comparison of pregnant women);Conditions and risk factors: renal impairment versus other NCDs versus medication use that impair potassium excretion versus mixed versus unknown (in comparison of pregnant women).

Sensitivity analysis

We conducted sensitivity analyses for primary outcomes if we had three or more studies per meta‐analysis, assessing the impact of:

Risk of bias: removing studies with a high risk of overall bias (see section: Assessment of risk of bias in included studies);Study design: removing cluster‐RCTs

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach to judge the certainty of the evidence as it relates to the studies contributing data to the meta‐analyses for the main outcomes, using GRADEprofiler (GRADEpro) software (GRADEpro GDT). The GRADE approach assesses certainty as high, moderate, low, or very low according to five criteria, namely, risk of bias, inconsistency of results, indirectness, imprecision and publication bias (Schünemann 2020).

For the following outcomes, we presented assessments in a GRADE Evidence Profile for Comparisons 1 and 3 (i.e. per type of population included in this review): DBP, SBP, hypertension, blood pressure control, cardiovascular events, cardiovascular mortality, blood potassium, hyperkalaemia, hypokalaemia and adverse events (other). We could not compile GRADE Evidence Profiles for the comparison in pregnant women as no eligible studies reported on outcomes in pregnant women. The effects of interventions on the outcomes included in the GRADE Evidence Profiles were interpreted according to magnitude of effect and certainty of the evidence, using GRADE guidance on informative statements to combine size and certainty of an effect (Santesso 2020).

We used the approaches described below for each domain to guide our ratings and included explanations as footnotes in the GRADE Evidence Profiles.

Risk of bias

We considered downgrading if the majority (> 50%) of the weighted outcome data in a meta‐analysis were from studies at a high or unclear overall risk of bias.

Inconsistency

We considered downgrading due to either unexplained considerable (defined as I² ≥ 90%) or substantial heterogeneity (defined as I² between 50% to < 90%). We explored heterogeneity using prespecified subgroup analysis, and examined sensitivity analyses of study design and quality (overall risk of bias).

Indirectness

We considered downgrading based on population characteristics such as age, ethnicity, blood pressure status; or due to intervention characteristics (e.g. purpose of intervention, LSSS type), comparator, direct comparison and outcome.

Imprecision

Number of events or participants

We considered downgrading based on an insufficient number of events (i.e. 300 events) for dichotomous outcomes or sample size not meeting the optimal information size (OIS) (i.e. 400 people providing outcome measures) for continuous outcomes.

Minimally contextualised approach to GRADE ratings

In line with recent GRADE guidance (Hultcrantz 2017; Zeng 2021), we selected a minimally contextualised approach that required us to specify thresholds for minimally important differences for key outcomes. The upper and lower limits of the 95% CIs were assessed in the same way to determine if they included the possibility of a small, trivial or no effect and an important benefit or harm (Zeng 2021).

Applying this approach to rating the certainty of evidence using GRADE in relation to thresholds (other than no effect) ideally requires the use of absolute numbers (Zeng 2021). To further support decision‐making by the WHO NUGAG Subgroup on Diet and Health about a population‐level intervention, we generated estimated population impacts for effect estimates and variation (95% CIs) for key clinical effectiveness outcomes when LSSS use was compared to regular salt use (Verbeek 2021). This was applied to the following key clinical effectiveness outcomes in adults: change in DBP, change in SBP, cardiovascular events: non‐fatal stroke, cardiovascular events: non‐fatal acute coronary syndrome, cardiovascular mortality and stroke mortality. We used a simplified model to estimate absolute numbers from relative cardiovascular measures, as well as the absolute numbers of stroke deaths prevented or caused by changes in blood pressure, as a surrogate outcome (Verbeek 2021). More detail on this simplified modelling approach can be found in Appendix 2.

Results

Description of studies

For detailed information, see Characteristics of included studies; Characteristics of excluded studies; Characteristics of ongoing studies; Characteristics of studies awaiting classification.

Results of the search

The study selection flowchart is available in Figure 1. We screened the titles and abstracts of 6511 de‐duplicated records identified through searching electronic databases, as well as 14 records identified through handsearching of three relevant systematic reviews. We assessed the full texts of 161 records against our eligibility criteria, of which four were in Chinese and one in Portuguese; we obtained language translation assistance for assessment of these. We included 26 studies reported in 74 full‐text records (Included studies), of which one did not provide data that could be used in the quantitative syntheses (meta‐analyses) (Arzilli 1986). Eight studies were identified as ongoing. We placed three studies under awaiting classification because we were unable to obtain further study details or data from the study authors in order to assess their eligibility for inclusion. We excluded a total of 75 full‐text records, of which 42 were duplicates (Excluded studies).

1 PRISMA flow diagram (Figure not reproduced here; see the original article.)

Included studies

Study designs

We included 26 eligible RCTs and did not identify any eligible prospective analytical cohort studies. The details of the included studies are summarised in Characteristics of included studies. Of the included trials, 16 were individually randomised trials (Allaert 2013; Allaert 2017; Arzilli 1986; CSSS Collaborative Group 2007; Geleijnse 1994; Gilleran 1996; Kawasaki 1998; Mu 2003; Omvik 1995; Pan 2017; Pereira 2005; Sarkkinen 2011; Suppa 1988; Yu 2021; Zhao 2014; Zhou 2009) and 10 were cluster‐RCTs (Bernabe‐Ortiz 2014; Chang 2006; Hu 2018; Li 2014; Li 2016; Mu 2009; Neal 2021; Toft 2020; Zhang 2015; Zhou 2013), including one stepped‐wedge cluster‐RCT (Bernabe‐Ortiz 2014). One RCT reported a cross‐over design (Allaert 2013) for which we only used first‐phase data. Twenty‐five of the eligible trials were published in peer‐reviewed journals; one cluster‐RCT was published in three separate conference abstracts and one trial as an abstract in a journal supplement. Full‐text publications of these could not be sourced after numerous attempts to contact authors.

Sample sizes and follow‐up

Nine of the 16 RCTs randomised ≤ 100 participants while seven were larger, randomising > 100 participants up to a maximum of 608 participants (CSSS Collaborative Group 2007; Mu 2003; Pan 2017; Suppa 1988; Yu 2021; Zhao 2014; Zhou 2009). Fewer than half of these RCTs (n = 7) reported sample size calculations, based on expected changes of between 3 and 10 mmHg in SBP (Allaert 2013; Allaert 2017; CSSS Collaborative Group 2007; Geleijnse 1994; Yu 2021; Zhao 2014; Zhou 2009); of these, two trials additionally based sample size calculations on expected changes of between 1.7 and 4 mmHg in DBP (CSSS Collaborative Group 2007; Geleijnse 1994).

Of the cluster‐RCTs, five randomised families or households (Hu 2018; Li 2014; Mu 2009; Toft 2020; Zhou 2013), ranging between 89 and 325 households, including 309 to 659 individual participants. Three cluster‐RCTs randomised villages; one conducted in Peru (N = 6 clusters; 2376 participants; Bernabe‐Ortiz 2014) and two conducted in China (N = 120 clusters; 2566 participants; Li 2016 and N = 600; 20,995 participants; Neal 2021). The cluster‐RCT by Zhang 2015 randomised nursing homes (N = 30); another randomised kitchens within a retirement home (N = 5; 2764 participants) (Chang 2006). Of the ten cluster‐RCTs, only four reported appropriate sample size calculations (including an ICC) based on expected changes in blood pressure, 24‐h sodium excretion, relative reduction in stroke and sodium intake reduction (Bernabe‐Ortiz 2014; Li 2016; Neal 2021; Toft 2020, respectively); two cluster‐RCTs (Chang 2006; Mu 2009) did not report a sample size calculation but did adjust for the effect of clustering in their analyses. One cluster‐RCT (Hu 2018) reported a sample size calculation based on an expected change in SBP, but did not report incorporating an ICC in this calculation.

Eleven (Allaert 2013; Allaert 2017; Arzilli 1986; CSSS Collaborative Group 2007; Gilleran 1996; Kawasaki 1998; Mu 2003; Omvik 1995; Sarkkinen 2011; Suppa 1988; Zhou 2009) of the 16 RCTs included a run‐in period, ranging between five days and six weeks. Ten RCTs tested an active LSSS intervention for a period of up to three months (Allaert 2013; Allaert 2017; Arzilli 1986; Geleijnse 1994; Kawasaki 1998; Pereira 2005; Sarkkinen 2011; Suppa 1988; Yu 2021; Zhao 2014), and five for between three and 12 months (CSSS Collaborative Group 2007; Gilleran 1996; Omvik 1995; Pan 2017; Zhou 2009). One RCT implemented a LSSS intervention for longer than 12 months (Mu 2003). For the 10 cluster‐RCTs, the duration of the LSSS intervention was two months in one trial (Li 2014), four months in another (Toft 2020), and ranged between one and five years in the other eight cluster‐RCTs.

Settings

Four of the 16 RCTs were conducted in northern China or Tibet; most were done in rural or suburban households (CSSS Collaborative Group 2007; Mu 2003; Zhao 2014; Zhou 2009). The remaining individually randomised trials from Brazil (Pereira 2005), France (Allaert 2013; Allaert 2017), Finland (Sarkkinen 2011), India (Yu 2021); Italy (Suppa 1988), Japan (Kawasaki 1998), Netherlands (Geleijnse 1994), Norway (Omvik 1995), Taiwan (Pan 2017) and the UK (Gilleran 1996) were conducted at household level, except for one trial in a European hospital setting (Arzilli 1986). Seven cluster‐RCTs were conducted in northern China or Tibet; most were done in rural or suburban households or communities (Hu 2018; Li 2014; Li 2016; Mu 2009; Neal 2021; Zhou 2013), with one cluster‐RCT having been conducted in nursing homes (Zhang 2015). One included cluster‐RCT from Taiwan (Chang 2006) was also conducted in a nursing home setting. In Peru, one stepped‐wedge cluster‐RCT was conducted in rural villages and households (Bernabe‐Ortiz 2014); a cluster‐RCT in Southwestern Denmark was conducted in families (Toft 2020).

Participants

We did not find any eligible studies in pregnant women. Trial participants were adults with a mean age ranging from 20 to 75.21 years and children with a mean age ranging from 8.4 to 9.5 years. Most of the included trials (15/26) were conducted in populations living in Asian countries. Eleven studies specifically included only participants with hypertension (Allaert 2013; Arzilli 1986; Geleijnse 1994; Gilleran 1996; Mu 2003; Omvik 1995; Pereira 2005; Sarkkinen 2011; Suppa 1988; Yu 2021; Zhao 2014), 11 included participants with and without hypertension (Bernabe‐Ortiz 2014; Chang 2006; CSSS Collaborative Group 2007; Hu 2018; Kawasaki 1998; Li 2014; Mu 2009; Neal 2021; Pan 2017; Zhou 2009; Zhou 2013), one each included only participants with normal blood pressure (Toft 2020) and participants who were pre‐hypertensive (Allaert 2017), and blood pressure status at baseline in the remaining studies were unknown (Li 2016; Zhang 2015). The largest trial (Neal 2021) included participants with an elevated risk of stroke and approximately 70% of participants in intervention and control groups had a history of stroke at baseline. Fifteen studies reported outcome data separately in participants with hypertension (Allaert 2013; Arzilli 1986; Geleijnse 1994; Gilleran 1996; Hu 2018; Kawasaki 1998; Mu 2003; Mu 2009; Omvik 1995; Pereira 2005; Sarkkinen 2011; Suppa 1988; Yu 2021; Zhao 2014; Zhou 2009). Twelve of these studies specifically included only participants with hypertension. Three of these studies included participants with hypertension and their family members (Hu 2018; Mu 2009; Zhou 2009) and one study included clinically healthy middle‐aged and elderly volunteers (Kawasaki 1998); these studies additionally reported outcome data in participants with normal blood pressure separately.

One cluster‐RCT compared LSSS to regular salt in 92 children (numbers analysed) (Toft 2020) by randomising families to LSSS or regular salt in bread. Seven studies included participants possibly at risk of hyperkalaemia (Chang 2006; Geleijnse 1994; Hu 2018; Neal 2021; Yu 2021; Zhao 2014; Zhou 2013), four studies included participants at unclear risk of hyperkalaemia (Arzilli 1986; Li 2016; Mu 2003; Zhang 2015) and the remaining trials included participants considered not to be at risk of hyperkalaemia. The criteria and assessments applied to classify the hyperkalaemia risk of participants per included study are summarised in Table 6. All 26 included trials excluded participants in whom an increased intake of potassium is known to be potentially harmful, for example, people with chronic kidney disease, type 1 or 2 diabetes mellitus, impaired renal function, or those using potassium‐sparing medications.

Interventions

In 23 of the 26 studies, combinations of potassium and/or magnesium and/or calcium salts were used as sodium substitutes in the LSSS interventions, with two studies assessing a LSSS intervention consisting of NaCl combined with 3% chitosan (Allaert 2013; Allaert 2017) and the remaining study assessing a LSSS intervention ‘naturally low in sodium’ in bread (Toft 2020). Product characteristics of the latter, obtained through author correspondence (Toft 2020), showed that the compound contained trace amounts of potassium (approximately 0.1 to 0.2%). Two RCTs (Arzilli 1986; Mu 2003) and one cluster‐RCT (Mu 2009) assessed LSSS with an unknown KCl content. Four cluster‐RCTs (Chang 2006; Li 2014; Neal 2021; Zhang 2015) and six RCTs (Geleijnse 1994; Gilleran 1996; Pan 2017; Pereira 2005; Yu 2021; Zhou 2009) assessed the effects of a LSSS intervention containing ≥ 30% KCl, while the remainder of the trials used LSSS interventions containing < 30% KCl. One RCT included two LSSS intervention arms, both including ≥ 30% KCl (Pan 2017).

Most trials (22/26) administered the LSSS intervention as a discretionary intervention (at the individual, household, institution or salt supply chain level). Of these, most trials replaced the supply of regular salt with LSSS within each household, to be used at the table and during food preparation (Bernabe‐Ortiz 2014; CSSS Collaborative Group 2007; Gilleran 1996; Hu 2018; Li 2014; Li 2016; Mu 2003; Mu 2009; Neal 2021; Omvik 1995; Pan 2017; Pereira 2005; Yu 2021; Zhao 2014; Zhou 2009; Zhou 2013). Two trials were conducted in nursing homes where LSSS was used during food preparation in the intervention kitchens of one trial (Chang 2006), while the specific implementation of the LSSS was unclear in the other trial (Zhang 2015). LSSS was administered as ‘added salt’ in four trials (Allaert 2013; Allaert 2017; Arzilli 1986; Suppa 1988).

A cluster‐RCT from Peru replaced the supply of regular salt in each village with LSSS in the salt supply chain, including households, food vendors, bakeries, community kitchens and restaurants. A social marketing/education strategy promoting LSSS in each village was aimed at women who were responsible for household food preparation (Bernabe‐Ortiz 2014). Another cluster‐RCT from northern China provided LSSS via the local food supply chain. LSSS was available for purchase at local village shops at either a subsidised price (same as regular salt) in half of the intervention villages, or at a regular price (approximately double that of regular salt). A community‐based health education programme to promote the use of LSSS was implemented via public announcement systems, bulletin boards, and specially developed promotional materials (Li 2016).

Three RCTs incorporated LSSS into prepared test foods, such as processed main dishes, bread, cheese, luncheon meats, soups or smoked sausage (Geleijnse 1994; Sarkkinen 2011), or seasonings containing LSSS, such as miso and soy sauce (Kawasaki 1998). These trials also provided trial participants with LSSS as salt for household food preparation and for use as table salt. A fourth cluster‐RCT used bread as the exclusive method of LSSS implementation by incrementally replacing normal salt with LSSS in bread over a period of five to six weeks, with participants followed up for four months in total (Toft 2020).

Trial participants in four studies were instructed not to change their dietary habits during the study period (Allaert 2017; Geleijnse 1994; Hu 2018; Kawasaki 1998), whereas participants from two trials were advised to either reduce their salt intake (Omvik 1995), or avoid salt‐rich foods (Sarkkinen 2011). Two trials reported co‐interventions such as lifestyle advice about eating less fat and sugar and doing more physical exercise (Allaert 2013), or a hypocaloric diet with increased physical exercise (Pereira 2005).

Outcome measures

Outcomes were regarded as clinical effectiveness outcomes, or safety outcomes related to the intake of LSSS with potassium (as guided by the WHO NUGAG Subgroup on Diet and Health).

Clinical effectiveness outcomes for comparisons in both adults and children were change in DBP, change in SBP, hypertension, blood pressure control, cardiovascular events, cardiovascular mortality, all‐cause mortality, antihypertensive medication use, change in fasting blood glucose, change in blood triglycerides, change in total blood cholesterol, and change in 24‐h urinary sodium and potassium excretion. In addition, clinical effectiveness outcomes in adults only were diabetes mellitus diagnosis and change in BMI; and in children only were growth changes, bone densitometry and bone health.

Safety outcomes for comparisons in both adults and children were change in blood potassium, hyperkalaemia, hypokalaemia, adverse events, renal function and hyponatraemia.

Primary outcomes

Only one RCT (Pan 2017) and one cluster‐RCT (Chang 2006) did not report on changes in DBP and SBP. However, we were also unable to use relevant outcome data from four trials, due to blood pressure data being reported in a figure (CSSS Collaborative Group 2007 (DBP only); Kawasaki 1998 (DBP and SBP)) or study authors providing insufficient information on the number of participants per treatment group (Arzilli 1986; Mu 2009). Three RCTs reported two types of DBP and SBP outcome data, i.e. ambulatory and clinic BP measurements (Allaert 2017; Omvik 1995; Pereira 2005). In order to minimise potential heterogeneity between studies, we included only the clinic BP measurements from these trials in our meta‐analyses.

Two cluster‐RCTs reported on the outcome hypertension; defined as SBP ≥ 140 mmHg, DBP ≥ 90 mmHg, or the use of blood‐pressure lowering therapy in the last two weeks (Li 2016) or SBP ≥ 140 mmHg, DBP ≥ 90 mmHg, a self‐reported physician diagnosis or current treatment for hypertension (Bernabe‐Ortiz 2014). Two RCTs reported on blood pressure control, defined as achieving SBP ≤ 140 mmHg and DBP ≤ 90 mmHg in both trials (Allaert 2013; Zhao 2014).

For the outcome cardiovascular events, we extracted data related to events such as stroke (Gilleran 1996; Hu 2018; Neal 2021; Pan 2017; Zhao 2014), myocardial infarction (Hu 2018) or acute coronary syndrome (Neal 2021), coronary heart disease or heart failure (Li 2016), hypotension (Li 2016), angina (Allaert 2017; Omvik 1995), bradycardia (Suppa 1988), as well as composite outcomes such as cardiovascular events (CSSS Collaborative Group 2007; Zhou 2009) or cardiovascular symptoms (Sarkkinen 2011). Three trials reported insufficient data on the number of events per group (Li 2016; Suppa 1988) as described in Table 7. Cardiovascular mortality was reported by one RCT (Zhao 2014) and three cluster‐RCTs (Chang 2006; Neal 2021; Zhou 2013); stroke mortality was reported by two cluster‐RCTs (Neal 2021; Zhou 2013).

Six trials reported on hyperkalaemia events; two trials (CSSS Collaborative Group 2007; Mu 2009) did not explicitly define criteria for hyperkalaemia, and one assessed self‐reported hyperkalaemia without defined criteria (Li 2016). Yu 2021 and Zhang 2015 defined the outcome as a serum potassium level more than 6.5 mmol/L and 5.5 mmol/L, respectively. Neal 2021 reported on definite, probable, possible and unlikely hyperkalaemia ‐ only data on definite and probable events, defined as elevated serum potassium > 5.5 mmol/L and typical electrocardiogram (ECG) changes documented in medical notes, were extracted; events that were possible (self‐reported serum potassium > 5.5 mmol/L or ECG changes but no supporting documentation to verify) and unlikely (clinical history and documentation suggest minimal indication for the diagnosis) were excluded from our review. Insufficient information was available from one trial (Li 2016) as described in Table 7. Only one RCT reported on hypokalaemia events (Pereira 2005), though criteria for the condition were not explicitly defined.

Seven trials reported changes in serum (Allaert 2017; Geleijnse 1994; Kawasaki 1998; Pereira 2005; Zhang 2015) or plasma concentrations of potassium (Omvik 1995; Sarkkinen 2011) between intervention and control groups. Data could not be extracted from one study (Sarkkinen 2011), as described in Table 7.

Secondary outcomes

Four RCTs (CSSS Collaborative Group 2007; Pan 2017; Zhao 2014; Zhou 2013), and three cluster‐RCTs (Chang 2006; Neal 2021; Zhang 2015) reported on all‐cause mortality.

Three trials reported on the occurrence of serious adverse events (not defined) during the study period (Bernabe‐Ortiz 2014; CSSS Collaborative Group 2007; Pan 2017). Other adverse event outcomes extracted included gastrointestinal symptoms (stomach ache or abdominal distension) (Sarkkinen 2011; Zhao 2014), hypercalcemia and renal calculi (Mu 2009), appendicitis, nephritis, nephrosis (Hu 2018), influenza (Allaert 2017), respiratory symptoms (Sarkkinen 2011) and dorsalgia (Allaert 2017). Suppa 1988 included self‐reported adverse events including asthenia, bradycardia, drowsiness, insomnia, decreased libido and depression, but did not report the number of events per group. None of the included trials reported adverse events such as nausea or vomiting.

Five trials reported on changes in antihypertensive medication use (Bernabe‐Ortiz 2014; Hu 2018; Li 2016; Zhao 2014; Zhou 2013). Four trials reported the number of participants using antihypertensive medications: two trials reported on this outcome in participants with hypertension (Hu 2018 included participants with hypertension and their family members, but only reported on the former; Zhao 2014 included only participants with hypertension) and one trial each in participants with unknown hypertensive status (Li 2016) and participants using hypotensive medication at baseline (Zhou 2013). Bernabe‐Ortiz 2014 assessed participant‐reported changes in medication use for hypertension and type 2 diabetes mellitus combined.

Three trials reported changes in serum creatinine (Omvik 1995; Pereira 2005; Zhang 2015). One cluster‐RCT (Li 2016) reported the mean urinary albumin‐to‐creatinine ratios of participants in the intervention and control groups, as well as the proportion of participants with albuminuria (including micro‐ and macro‐albuminuria) in both groups.

Four studies reported changes in BMI (Li 2016; Pereira 2005; Sarkkinen 2011; Toft 2020), while two reported changes in fasting blood glucose concentrations (Toft 2020; Zhou 2013). Five studies reported changes in blood triglycerides (Gilleran 1996; Kawasaki 1998; Pereira 2005; Toft 2020; Zhou 2009) and total blood cholesterol (Geleijnse 1994; Gilleran 1996; Kawasaki 1998; Toft 2020; Zhou 2009), respectively.

A total of 12 studies collected 24‐h urine samples and reported on 24‐h urinary sodium excretion (Bernabe‐Ortiz 2014; Geleijnse 1994; Gilleran 1996; Kawasaki 1998; Li 2016; Neal 2021; Omvik 1995; Sarkkinen 2011; Suppa 1988; Toft 2020; Yu 2021; Zhou 2009); the same studies reported on 24‐h urinary potassium excretion from 24‐h urine samples.

None of the included studies reported on the outcomes of diabetes mellitus diagnosis or hyponatremia.

Funding sources and conflicts of interest

Nine of the 26 studies did not disclose their funding source(s). The remaining studies were funded as follows:

nine public/non‐commercial funding only, including government bodies and research institutions (Bernabe‐Ortiz 2014; CSSS Collaborative Group 2007; Li 2014; Li 2016; Mu 2009; Pan 2017; Toft 2020; Zhao 2014; Zhou 2013);four public/non‐commercial funding plus LSSS provided for the trial by LSSS manufacturer (Geleijnse 1994; Neal 2021; Omvik 1995; Yu 2021);two commercial funding by LSSS manufacturers plus LSSS provided for the trial (Chang 2006; Sarkkinen 2011);one commercial funding for the study from food industry research fund plus LSSS provided for the trial by LSSS manufacturer (Hu 2018); andone LSSS provided for the trial by the LSSS manufacturer (Kawasaki 1998).

The authors of 13 included studies did not report on potential conflicts of interest (COI), whereas those from 13 studies did. Of these 13 studies, authors of eleven declared that they had no potential COI (Bernabe‐Ortiz 2014; Chang 2006; CSSS Collaborative Group 2007; Mu 2009; Pan 2017; Pereira 2005; Toft 2020; Yu 2021; Zhao 2014; Zhou 2009; Zhou 2013), whereas the author of one study declared a potential conflict of interest as the chair of the Australian Division of World Action on Salt and Health (Li 2016), and some members of the author team of a large cluster‐RCT declared potential conflicts of interest, while the remaining members declared no conflict (Neal 2021).

Excluded studies

We contacted nine corresponding authors for further information to assist with study inclusion. We excluded 32 studies (33 full‐text records) due to the following reasons: 

Wrong study design (single‐arm trial): 5
Wrong study design (commentary/letter): 3
Wrong study design (case report/study): 2
Wrong study design (case series): 2
Wrong study design (non‐randomised trial): 2
Wrong study design (quasi‐randomised trial): 2
Wrong study design (cross‐over with first phase data not available): 1
Wrong type of intervention (multifactorial): 4
Wrong type of intervention (dietary): 2
Wrong type of intervention (LSSS administered as supplement): 1
Wrong type of intervention (salt restriction education): 1
Wrong comparator: 4
Wrong outcome (sensory/organoleptic): 2
Wrong outcome (sodium concentration of homemade food): 1

The Characteristics of excluded studies section illustrates these 32 studies with reasons for exclusion. We also excluded 42 duplicates at the full‐text screening stage. The remaining references where we could not reach the authors or information provided was not sufficient to make a clear judgement (n = 3) were included as Studies awaiting classification. The eight ongoing studies are detailed in Characteristics of ongoing studies.

Risk of bias in included studies

The Characteristics of included studies provides details of the judgements for each risk of bias domain per study. Figure 2 presents a summary of the risk of bias judgements for each included study and Figure 3 the summary of the judgements per risk of bias domain.

2 Summary of the risk of bias judgements for each included study (Figure not reproduced here; see the original article.)

3 Summary of the judgements per risk of bias domain (Figure not reproduced here; see the original article.)

Allocation

Random sequence generation

Fifteen trials described adequate methods of random sequence generation and were at low risk of selection bias (Bernabe‐Ortiz 2014; Chang 2006; CSSS Collaborative Group 2007; Geleijnse 1994; Hu 2018; Li 2014; Li 2016; Mu 2009; Neal 2021; Pan 2017; Toft 2020; Yu 2021; Zhao 2014; Zhou 2009; Zhou 2013). Eleven trials did not report how the random sequence had been generated and were at unclear risk of selection bias.

Allocation concealment

Eleven trials described methods of allocation concealment judged to be at low risk of selection bias (Allaert 2017; Bernabe‐Ortiz 2014; CSSS Collaborative Group 2007; Hu 2018; Li 2016; Neal 2021; Pan 2017; Toft 2020; Yu 2021; Zhao 2014; Zhou 2009). The remaining fifteen did not report sufficient information on allocation concealment, and were at unclear risk of selection bias.

Blinding

Blinding of participants and personnel (performance bias)

Performance bias was assessed as unlikely in fifteen trials, while eleven trials had an unclear risk of bias, mainly due to insufficient information on blinding of either participants or personnel (Allaert 2013; Arzilli 1986; Bernabe‐Ortiz 2014; Chang 2006; Kawasaki 1998; Li 2014; Li 2016; Mu 2003; Mu 2009; Suppa 1988; Zhang 2015).

Blinding of outcome assessment (detection bias)

Nineteen trials were assessed to have a low risk of detection bias, while seven had an unclear risk mainly due to insufficient information on blinding of outcome assessors (Li 2016; Mu 2003; Mu 2009; Pereira 2005; Suppa 1988; Zhang 2015; Zhou 2013). Some of these trials reported the measurement of blood pressure with non‐automatic devices, increasing the likelihood of detection bias (Mu 2003; Pereira 2005; Suppa 1988).

Incomplete outcome data

Four trials were at a high risk of bias due to high overall or differential attrition (≥ 10%) (Geleijnse 1994; Gilleran 1996; Hu 2018; Mu 2003). Fourteen studies were at low risk because they reported low overall or differential attrition (Allaert 2013; Allaert 2017; Chang 2006; CSSS Collaborative Group 2007; Kawasaki 1998; Li 2014; Neal 2021; Omvik 1995; Sarkkinen 2011; Suppa 1988; Toft 2020; Yu 2021; Zhou 2009; Zhou 2013). Two studies reported high attrition; however intention‐to‐treat analyses (ITT) were conducted using multiple imputation in one study (Zhao 2014) and the last‐observation‐carried‐forward method in the other study (Pan 2017); therefore, these were judged as being at low risk. High attrition was reported at some time points in the stepped‐wedge cluster‐RCT; however, it was unclear whether any data were imputed and therefore this study had an unclear risk (Bernabe‐Ortiz 2014). Five additional studies were at unclear risk of bias since insufficient information on attrition was provided.

Selective reporting

Two trials were assessed as being at high risk of bias.  Substudies of both trials reported outcomes not prespecified in the study protocol (CSSS Collaborative Group 2007; Li 2016). Two trials at low risk of bias reported outcomes prespecified in the study protocol; the remaining studies had an unclear risk due to inadequate reporting of all prespecified outcomes, or the unavailability of the study protocol.

Other potential sources of bias

One trial was assessed as being high risk for misclassification bias due to limited information for adjudication of clinical outcome events, as well as self‐reported potential hyperkalaemia events (Neal 2021). Two trials were assessed as being at unclear risk. In the stepped‐wedge trial, a reduction in blood pressure was observed in some clusters (villages) before the intervention (Bernabe‐Ortiz 2014) while, in another cluster‐RCT, there was a considerable risk of contamination in the intervention group due to the unlimited availability of condiments and spices such as soy sauce and monosodium glutamate (Chang 2006). No potential sources of bias were identified in the remaining studies.

Additional domains assessed for cluster‐RCTs

Recruitment bias

One trial had a high risk for recruitment bias since a number of participants were recruited after the clusters (kitchens) were randomised (Chang 2006).  Six trials reported recruitment of participants before randomisation of clusters and were therefore considered at low risk (Bernabe‐Ortiz 2014; Hu 2018; Li 2014; Li 2016; Neal 2021; Toft 2020). Three trials did not provide sufficient information regarding the timing of recruitment.

Comparability with individually randomised trials (RCTs)

Six trials at low risk of bias reported effect estimates comparable to those reported by similar RCTs (Bernabe‐Ortiz 2014; Chang 2006; Hu 2018; Li 2014; Zhang 2015; Zhou 2013), whereas it was not possible to compare the effect estimates in four trials.

Baseline imbalance

Three trials reported no differences in baseline characteristics of the participants in the intervention and control groups (Hu 2018; Li 2014; Neal 2021), while two adjusted for baseline differences and were therefore also considered to be at low risk (Bernabe‐Ortiz 2014; Zhou 2013). Five studies that reported insufficient information regarding baseline characteristics were at unclear risk of bias.

Loss of clusters

One trial reported an analysis without outcome data from almost one‐third of their included clusters, thus was considered at high risk of attrition bias (Zhang 2015). Seven trials had a low risk (Bernabe‐Ortiz 2014; Chang 2006 ; Hu 2018; Li 2014; Li 2016; Neal 2021; Toft 2020) while two did not provide sufficient information on the number of clusters (families) lost to follow‐up (Mu 2009), or the reasons for loss to follow‐up (Zhou 2013).

Incorrect analysis

Three trials did not report adjustment for clustering in their analysis, thus were considered at high risk of bias (Hu 2018; Li 2014; Zhou 2013). Six trials were at low risk; five reported statistical adjustment for clustering (Chang 2006; Li 2016; Mu 2009; Neal 2021; Toft 2020) and one also accounted for time trends in their analysis (Bernabe‐Ortiz 2014). One trial that reported insufficient information regarding statistical adjustment for clusters was considered at unclear risk.

Overall risk of bias

Three RCTs and three cluster‐RCTs had a high overall risk of bias due to attrition (Geleijnse 1994; Gilleran 1996; Hu 2018; Mu 2003; Zhang 2015) or recruitment bias (Chang 2006). Six RCTs had a low overall risk (Allaert 2017; CSSS Collaborative Group 2007; Pan 2017; Yu 2021; Zhao 2014; Zhou 2009) and two cluster‐RCTs (Li 2014; Neal 2021) had a low overall risk, while the remaining seven RCTs and five cluster‐RCTs had an unclear risk mainly due to uncertainty about selection, attrition and recruitment bias (Allaert 2013; Arzilli 1986; Bernabe‐Ortiz 2014; Kawasaki 1998; Li 2016; Mu 2009; Omvik 1995; Pereira 2005; Sarkkinen 2011; Suppa 1988; Toft 2020; Zhou 2013).

Effects of interventions

See: Table 1; Table 2

See: Table 1; Table 2.

Comparison 1. Low‐sodium salt substitutes versus regular salt or no active intervention in adults

Table 1 presents the effects of LSSS compared to regular salt or no active intervention in adult participants on changes in DBP and SBP; as well as the number of participants per group with hypertension, blood pressure control, cardiovascular events (various events and non‐fatal stroke); the rate ratio of participants in the intervention group with non‐fatal acute coronary syndrome (ACS), cardiovascular mortality and stroke mortality; changes in blood potassium; and the number of participants per group with hyperkalaemia, hypokalaemia and other adverse events.

A total of 26 RCTs, 16 randomising individual participants and 10 randomising clusters, reporting on 34,961 adult participants, were included in this comparison. Key details about studies in this comparison, including study design, setting and overall risk of bias; characteristics of the intervention, comparator, population and outcomes; method of synthesis, and time points of measurement are included in the Overview of Synthesis and Included Studies (OSIS) table (Table 8).

Primary outcomes

Change in diastolic blood pressure (DBP, mmHg)

GRADE assessment suggests that LSSS probably reduce DBP slightly, on average, compared to regular salt in adults (moderate‐certainty evidence, downgraded once for inconsistency).

Average reductions in DBP ranged from 0.6 mmHg to 11.33 mmHg with LSSS and from a reduction of 7 mmHg to an increase of 2.6 mmHg with regular salt in the 19 trials that reported this outcome. Two trials did not have average changes per group available: Li 2016 reported end values and no baseline measures; Neal 2021 reported only mean differences between groups. The meta‐analysis showed small, important effects on DBP on average between LSSS and regular salt groups (MD ‐2.43 mmHg, 95% confidence interval (CI) ‐3.50 to ‐1.36, I² = 88%, 20,830 participants, 19 RCTs, moderate‐certainty evidence, Analysis 1.1). Follow‐up ranged from four weeks to 60 months.

This small yet important effect was confirmed by sensitivity analyses, including only trials with 'low' or 'unclear' overall risk of bias (Analysis 1.12) and including only trials randomising participants at the individual level, i.e. excluding cluster‐RCTs (Analysis 1.13). This direction of effect was also seen in a large stepped‐wedge cluster trial following 2376 participants over 30 months (unclear overall risk of bias), though the magnitude of the effect was considerably diminished (Analysis 1.14).

The estimated population impact, as described in Appendix 2, indicated that the effect of the primary meta‐analysis (Analysis 1.1) corresponded to an estimated 60 (ranging from 35 to 83) stroke deaths prevented per 100,000 persons, aged 50 years and older, per year.

Subgroup analyses were undertaken for this outcome due to the presence of substantial heterogeneity. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping by study duration (Analysis 1.2) suggests there may be no important differences in average effects between subgroups. Subgrouping participants by age (Analysis 1.3), gender (Analysis 1.4), ethnicity (Analysis 1.5), BMI (Analysis 1.6), blood pressure status (Analysis 1.7) or baseline 24‐h urinary sodium (Analysis 1.10) or potassium (Analysis 1.11) excretion further suggested there may be no important clinical differences in average effects between subgroups. Subgrouping by intervention characteristics also suggests there may be no important differences in average effects between the manner of LSSS implementation (Analysis 1.8) or the type of LSSS (Analysis 1.9).

Five trials reported this outcome in an unusable format (e.g. reported only in a figure from which we could not extract exact values, or did not report standard deviations and participant numbers along with mean change), and usable data were not provided when requested from authors (Table 7). The funnel plot (Figure 4) shows that most trials had similar effect sizes despite varying inter‐trial standard errors, thereby limiting the ability to assess asymmetry. The mean difference in the fixed‐effect model, an analytical approach that gives less weight to small studies, for the primary analysis (‐2.27 mmHg, 95% CI ‐2.56 to ‐1.98) was similar to the mean difference when using the random‐effects model, which gives more weight to smaller studies, for the primary analysis (‐2.43 mmHg, 95% CI ‐3.50 to ‐1.36, Analysis 1.1). This suggests that any small‐study effects have little impact on the intervention effect estimate.

4 Funnel plot for change in DBP (Analysis 1.1) in comparison 1 (Figure not reproduced here; see the original article.)

Change in systolic blood pressure (SBP, mmHg)

GRADE assessment suggests that LSSS probably reduce SBP slightly, on average, compared to regular salt in adults (moderate‐certainty evidence, downgraded once for inconsistency).

Average reductions in SBP ranged from 1.5 mmHg to 15.25 mmHg with LSSS and from a reduction of 6.8 mmHg to an increase of 4 mmHg with regular salt in the 20 trials that reported this outcome. Three trials did not have average changes per group available: Li 2016 reported end values and no baseline measures; CSSS Collaborative Group 2007 and Neal 2021 reported only mean differences between groups. The meta‐analysis showed small, important effects on SBP on average between LSSS and regular salt groups (MD ‐4.76 mmHg, 95% CI ‐6.01 to ‐3.50, I² = 78%, 21,414 participants, 20 RCTs, moderate‐certainty evidence, Analysis 1.15). Follow‐up ranged from four weeks to 60 months.

This small yet important effect was confirmed by sensitivity analyses, including only trials with 'low' or 'unclear' overall risk of bias (Analysis 1.26) and including only trials randomising participants at the individual level, i.e. excluding cluster‐RCTs (Analysis 1.27). This direction of effect was also seen in the stepped‐wedge cluster trial that followed 2376 participants over 30 months (unclear overall risk of bias), though the magnitude of the effect was considerably diminished (Analysis 1.28).

The estimated population impact, as described in Appendix 2, indicated that the effect of the primary meta‐analysis (Analysis 1.15) corresponded to an estimated 53 (ranging from 40 to 65) stroke deaths prevented per 100,000 persons, aged 50 years and older, per year.

Subgroup analyses were undertaken for this outcome due to the presence of substantial heterogeneity. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping by study duration (Analysis 1.16), and subgrouping participants by age (Analysis 1.17), gender (Analysis 1.18), ethnicity (Analysis 1.19), BMI (Analysis 1.20), blood pressure status (Analysis 1.21) or baseline 24‐h urinary sodium (Analysis 1.24) or potassium (Analysis 1.25) excretion suggests there may be no important clinical differences in average effects between subgroups. Subgrouping by intervention characteristics also suggested there may be no important differences in average effects between the manner of LSSS implementation (Analysis 1.22) or the type of LSSS (Analysis 1.23).

Four trials reported this outcome in an unusable format (e.g. reported only between‐group P values or did not report standard deviations and participant numbers along with mean change), and usable data were not provided when requested from authors (Table 7). The funnel plot (Figure 5) shows that most trials had similar effect sizes and standard errors, with some outliers, thereby limiting the ability to assess asymmetry. The mean difference in the fixed‐effect model for the primary analysis (‐3.92 mmHg, 95% CI ‐4.37 to ‐3.47) was similar to the mean difference when using the random‐effects model for the primary analysis (‐4.76 mmHg, 95% CI ‐6.01 to ‐3.50, Analysis 1.15). This suggests that any small‐study effects have little impact on the intervention effect estimate.

5 Funnel plot for change in SBP (Analysis 1.14) in comparison 1 (Figure not reproduced here; see the original article.)

Hypertension

GRADE assessment suggests that, on average, LSSS may result in little to no difference in hypertension in the adult population, when compared to regular salt (low‐certainty evidence, downgraded once for risk of bias and once for imprecision).

One study following participants for 18 months reported on this outcome (risk ratio (RR) 0.97, 95% CI 0.90 to 1.03, 2566 participants, 1 RCT, low‐certainty evidence, Analysis 1.29), with 725 participants in the LSSS group and 738 participants in the regular salt group having prevalent hypertension at the end of the study. The absolute effect for hypertension was 17 fewer per 1000 (95% CI 58 fewer to 17 more). The stepped‐wedge cluster trial (unclear overall risk of bias) reported on incident hypertension in 1914 participants represented by 2712.3 person‐years at risk in the LSSS group and 1961.1 person‐years at risk in the regular salt group, and found a reduction in hypertension with LSSS compared to regular salt (Analysis 1.30).

Blood pressure control

GRADE assessment suggests that the evidence is very uncertain about the effect of LSSS on blood pressure control in adults, when compared to regular salt (very low‐certainty evidence, downgraded once for risk of bias, once for indirectness and once for imprecision).

Two small studies reported on this outcome (RR 2.12, 95% CI 1.32 to 3.41, I² = 0%, 253 participants, 2 RCTs, very low‐certainty evidence, Analysis 1.31), with the number of participants in the LSSS group achieving blood pressure control ranging from 16 to 19 and participants in the regular salt group achieving blood pressure control ranging from seven to 10. The absolute effect for blood pressure control was 143 more per 1000 (95% CI 41 more to 308 more). The two trials reporting this outcome had follow‐up at eight weeks and three months.

Cardiovascular events: various

GRADE assessment suggests that the evidence is very uncertain about the effect of LSSS interventions on various other cardiovascular events when compared to regular salt in the adult population (very low‐certainty evidence, downgraded once for indirectness and twice for imprecision).

It should be noted that a very small number of participants presented with various other cardiovascular events in both groups for the five trials reporting this outcome. Event numbers ranged from zero to eight with LSSS and from zero to five with regular salt. The meta‐analysis of the RR was 1.22 (95% CI 0.49 to 3.04, I² = 0%, 982 participants, 5 RCTs, very low‐certainty evidence, Analysis 1.32) when comparing LSSS and regular salt. The absolute effect for various other cardiovascular events was 357 more per 100,000 (95% CI 828 fewer to 3310 more). Two trials reporting this outcome had follow‐up at ≤ 3 months, while three followed participants for > 3 to 12 months.

Two trials reported this outcome in an unusable format (i.e. numbers of events per group and numbers of participants per group not reported), and usable data were not provided when requested from authors (Table 7).

Cardiovascular events: non‐fatal stroke

GRADE assessment for this outcome suggests that, on average, LSSS probably reduce non‐fatal stroke events slightly in adults, when compared to regular salt (moderate‐certainty evidence, downgraded once for indirectness).

The number of participants with non‐fatal stroke in two small trials (Gilleran 1996; Pan 2017) reporting on this outcome at ≤ 3 months and > 3 to 12 months, respectively, ranged from zero to four with LSSS and one participant each with regular salt. The third, a large cluster‐RCT (Neal 2021) that followed participants for a mean of 4.75 years, reported rates of events: 22.36 events per 1000 person‐years in the LSSS group and 24.86 events per 1000 person‐years in the regular salt group (rate ratio 0.90, 95% CI 0.80 to 1.01). The meta‐analysis combining these data as risk ratios resulted in an RR of 0.90 (95% CI 0.80 to 1.01, I² = 0%, 21,250 participants, 3 RCTs, moderate‐certainty evidence, Analysis 1.33) when comparing LSSS with regular salt. This result translates to an absolute effect for non‐fatal stroke of 20 fewer per 100,000 (95% CI 40 fewer to 2 more). Since this pooled effect was driven by a large secondary prevention trial with a preponderance of participants with previous stroke, we downgraded once for indirectness.

The estimated population impact, as described in Appendix 2, indicated that the effect of the primary meta‐analysis (Analysis 1.33) corresponded to an estimated 10 non‐fatal strokes prevented (ranging from 21 prevented to 1 caused) per 100,000 persons per year.

Sensitivity analyses, including only trials with 'low' or 'unclear' overall risk of bias (Analysis 1.34) and including only trials randomising participants at the individual level, i.e. excluding cluster‐RCTs (Analysis 1.35), did not reflect this benefit with LSSS; instead showing highly imprecise effects of little to no effect, or harm.

One trial reported this outcome in an unusable format (i.e. numbers of participants per group not reported), and usable data were not provided when requested from authors (Table 7).

Cardiovascular events: non‐fatal acute coronary syndrome

GRADE assessment suggests that LSSS probably reduce non‐fatal ACS events slightly, on average, when compared to regular salt in adults (moderate‐certainty evidence, downgraded once for indirectness).

A single large cluster‐RCT contributed data to this outcome at a mean follow‐up of 4.75 years, reporting rates of 3.79 events per 1000 person‐years in the LSSS group and 5.12 events per 1000 person‐years in the regular salt group. The rate ratio was 0.70 (95% CI 0.52 to 0.94, 20,995 participants, 1 RCT, moderate‐certainty evidence, Analysis 1.36) and the absolute effect for non‐fatal acute coronary syndrome was 150 fewer per 100,000 person‐years (95% CI 250 fewer to 30 fewer), when comparing LSSS with regular salt in this large secondary prevention trial in which most of the participants had a history of previous stroke. This setting limited generalisability to the general adult population, and the evidence was consequently downgraded once for indirectness.

The estimated population impact, as described in Appendix 2, indicated that the effect of the primary analysis (Analysis 1.36) corresponded to an estimated 50 (ranging from 10 to 80) non‐fatal ACS events prevented per 100,000 persons per year.

Cardiovascular mortality

GRADE assessment of this outcome suggests that LSSS probably reduce cardiovascular mortality slightly, on average, in adults when compared to regular salt (moderate‐certainty evidence, downgraded once for indirectness).

The number of cardiovascular mortality events per 1000 person‐years in the three trials reporting on this outcome ranged from 4.53 to 22.94 in the LSSS groups and 7.81 to 26.30 in the regular salt groups. The meta‐analysis comparing LSSS with regular salt resulted in a rate ratio of 0.77 (95% CI 0.60 to 1.00, I² = 35%, 23,200 participants, 3 RCTs, moderate‐certainty evidence, Analysis 1.37). The absolute effect for cardiovascular mortality was 180 fewer per 100,000 person‐years (95% CI 310 fewer to 0 fewer). We downgraded this finding once for indirectness since the pooled effect was driven by the secondary prevention trial including a large proportion of participants with previous stroke (Neal 2021). Two trials reporting on this outcome had a mean follow‐up of between 2.6 and 4.75 years (Chang 2006; Neal 2021) while a third (Zhou 2013) reported on this outcome following three years of active intervention and ten years follow‐up.

The estimated population impact, as described in Appendix 2, indicated that the effect of the primary meta‐analysis (Analysis 1.37) corresponded to an estimated 53 cardiovascular deaths prevented (ranging from 92 prevented to none prevented or caused) per 100,000 persons per year.

A sensitivity analysis including only trials with 'low' or 'unclear' overall risk of bias (Analysis 1.38) confirmed this effect.

Stroke mortality

GRADE assessment suggests that the evidence is very uncertain about the effect of LSSS on stroke mortality in adults, when compared to regular salt (very low‐certainty evidence, downgraded once for indirectness and twice for imprecision).

The number of stroke mortality events per 1000 person‐years in the two trials reporting on this outcome ranged from 2.01 to 6.78 in the LSSS groups and 5.85 to 8.79 in the regular salt groups. The meta‐analysis comparing LSSS with regular salt resulted in a rate ratio 0.64 (95% CI 0.33 to 1.25, I² = 45%, 21,423 participants, 2 RCTs, very low‐certainty evidence, Analysis 1.39). The absolute effect for stroke mortality was 145 fewer per 100,000 person‐years (95% CI 270 fewer to 100 more). The pooled effect was driven to a considerable extent by a large secondary prevention trial including a large proportion of participants with previous stroke, resulting in limited generalisability to the general adult population. This trial (Neal 2021) had a mean follow‐up of 4.75 years, while the second trial reporting on this outcome (Zhou 2013) had a follow‐up of three years of active intervention and ten years thereafter.

The estimated population impact, as described in Appendix 2, indicated that the effect of the primary meta‐analysis (Analysis 1.39) corresponded to an estimated 28 stroke deaths prevented (ranging from 53 prevented to 20 caused) per 100,000 persons per year.

Change in blood potassium (mmol/L)

GRADE assessment suggests that, on average, LSSS probably increase blood potassium slightly compared to regular salt in the adult population (moderate‐certainty evidence, downgraded once for risk of bias).

Average changes in blood potassium ranged from a reduction of 0.2 mmol/L to an increase of 0.38 mmol/L with LSSS and from a reduction of 0.2 mmol/L to an increase of 0.3 mmol/L with regular salt in the six trials that reported this outcome. The meta‐analysis showed small, important effects on blood potassium on average between LSSS and regular salt groups (MD 0.12, 95% CI 0.07 to 0.18, I² = 0%, 784 participants, 6 RCTs, moderate‐certainty evidence, Analysis 1.40).

This small yet important effect was confirmed by sensitivity analyses, including only trials with 'low' or 'unclear' overall risk of bias (Analysis 1.42) and including only trials randomising participants at the individual level, i.e. excluding cluster‐RCTs (Analysis 1.43). The trials reporting on this outcome reported results at 56 days, five weeks, 12 weeks and between one and 1.5 years each, while two trials reported results at approximately six months.

Subgroup analyses were undertaken for this outcome to explore whether there were differences in effects in subgroups based on hyperkalaemia risk. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping participants by risk of hyperkalaemia as per Table 6 suggests there may be no important clinical differences in average effects between participants not at risk, at unclear risk and those at possible risk of hyperkalaemia (Analysis 1.41).

One trial reported this outcome in an unusable format (i.e. reported change and significance of change in the control group only), and usable data were not provided when requested from authors (Table 7).

Hyperkalaemia

GRADE assessment suggests that, on average, LSSS likely result in little to no difference in hyperkalaemia in the adult population when compared to regular salt (moderate‐certainty evidence, downgraded once for risk of bias).

It should be noted, however, that a very small number of participants presented with hyperkalaemia in both groups across the trials that reported this outcome. The number of participants with hyperkalaemia in the five trials reporting this outcome ranged from zero to 11 with LSSS and from zero to nine with regular salt. The meta‐analysis of the RR was 1.04 (95% CI 0.46 to 2.38, I² = 0%, 22,849 participants, 5 RCTs, moderate‐certainty evidence, Analysis 1.44) when comparing LSSS to regular salt. The absolute effect for hyperkalaemia was 4 more per 100,000 (95% CI 47 fewer to 121 more).

A sensitivity analysis including only trials with 'low' or 'unclear' overall risk of bias (Analysis 1.46) confirmed this effect, though this result was highly imprecise. A sensitivity analysis including only trials randomising participants at the individual level, i.e. excluding cluster‐RCTs (Analysis 1.47) was not informative due to zero events in both trial arms. These five trials reported results after three months, 12 months, one to 1.5 years, 2 years and a mean of 4.75 years follow‐up.

Subgroup analyses were undertaken for this outcome to explore whether there were differences in effects in subgroups based on hyperkalaemia risk. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping participants by risk of hyperkalaemia as per Table 6 suggests there may be no important clinical differences in average effects between participants not at risk, at unclear risk and those at possible risk, of hyperkalaemia (Analysis 1.45).

Hypokalaemia

GRADE assessment of this outcome suggests that the evidence is very uncertain about the effects of LSSS on hypokalaemia when compared to regular salt in the adult population (very low‐certainty evidence, downgraded once for risk of bias and twice for indirectness).

A single, small trial (Pereira 2005) reported no hypokalaemia events in either trial arm comparing LSSS and regular salt in young participants with hypertension requiring potassium supplementation due to the use of potassium‐depleting diuretics (RR and 95% CI not estimable, 22 participants, 1 RCT, very low‐certainty evidence, Analysis 1.48). This study reported outcomes at 12 weeks.

Secondary outcomes

For comparison 1, no studies reported on diabetes mellitus diagnosis or hyponatraemia.

All‐cause mortality

The number of all‐cause mortality events in two trials (CSSS Collaborative Group 2007; Pan 2017) reporting on this outcome at > 3 to 12 months ranged from three to four with LSSS and one to four with regular salt. Three additional trials (Chang 2006; Neal 2021; Zhou 2013) reported rates of events and rate ratios; these ranged from 11.08 to 93.45 events per 1000 person‐years in the LSSS groups and 13.66 to 101.29 events per 1000 person‐years in the regular salt groups and corresponded to rate ratios (95% CIs) of 0.92 (0.77 to 1.10), 0.88 (0.82 to 0.95) and 0.81 (0.46 to 1.42), respectively. The meta‐analysis combining these data as risk ratios resulted in an RR of 0.89 (95% CI 0.83 to 0.95, I² = 0%, 24,005 participants, 5 RCTs, Analysis 1.49) when comparing LSSS with regular salt.

Adverse events (other)

GRADE assessment suggests that the evidence is very uncertain about the effect of LSSS on other adverse events when compared to regular salt in adults (very low‐certainty evidence, downgraded once for risk of bias, once for inconsistency and once for imprecision).

The number of participants with other adverse events in the eight trials reporting this outcome ranged from zero to 17 with LSSS and from zero to seven with regular salt. The events reported were highly diverse and not suitable for pooling in a meta‐analysis (2109 participants, 8 RCTs, very low‐certainty evidence, Analysis 1.50). Four trials reporting on other adverse events reported these at ≤ 3 months, three reported on this outcome at > 3 to 12 months and one trial reported other adverse events at > 12 months.

Subgroup analyses were undertaken for this outcome to explore whether there were differences in effects in subgroups based on hyperkalaemia risk. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping participants by risk of hyperkalaemia as per Table 6 suggests there may be no important clinical differences in average effects between participants not at risk, and those at possible risk, of hyperkalaemia (Analysis 1.51).

One trial reported this outcome in an unusable format (i.e. numbers of events per group not reported), and usable data were not provided when requested from authors (Table 7).

Antihypertensive medication use

The number of participants using antihypertensive medication across the four trials reporting this outcome ranged from 34 to 246 with LSSS and from 52 to 267 with regular salt. The meta‐analysis of the RR of the number of participants using antihypertensive medication was 0.80 (95% CI 0.67 to 0.95, I² = 53%, 3301 participants, 4 RCTs, Analysis 1.52) when comparing LSSS and regular salt groups. The one individually randomised trial that reported this outcome had follow‐up at three months; the three cluster trials reporting this outcome had follow‐up at 12, 18 and 36 months. The stepped‐wedge cluster trial (unclear overall risk of bias) reported no changes in medication use, with 10.5% at baseline and 10.1% (P = 0.73) at the end of the study three years later; this was for hypertension and type 2 diabetes mellitus medication use combined.

Subgroup analyses were undertaken for this outcome due to the presence of substantial heterogeneity. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping by duration of study (Analysis 1.53) as well as age (Analysis 1.54), gender (Analysis 1.55), BMI (Analysis 1.56) or hypertensive status (Analysis 1.57) of participants at baseline suggests there may be no important clinical differences in average effects between subgroups. Subgrouping by ethnicity, type and implementation of LSSS, 24‐h sodium excretion at baseline and 24‐h potassium excretion at baseline could not be used to explore clinical differences, as all studies were categorised into the same subgroup.

Change in BMI (kg/m2)

Average change in BMI ranged from a reduction of 1.6 kg/m² to no change with LSSS and from a reduction of 1.0 kg/m² to no change with regular salt in the four trials that reported this outcome. We did not pool these results into an overall effect estimate due to considerable heterogeneity (I² = 96%, Tau² = 0.80, Chi² = 70.15, df = 3 (P < 0.00001)), but rather presented the individual effect sizes between LSSS and regular salt per study (2060 participants, 4 RCTs, Analysis 1.58). Trials reporting on this outcome followed participants for eight weeks, 12 weeks, four months and 18 months.

Change in serum creatinine (µmol/L)

Average change in serum creatinine ranged from a reduction of 0.8 µmol/L to an increase of 2 µmol/L with LSSS and from a reduction of 1.0 µmol/L to an increase of 3.54 µmol/L with regular salt in the three trials that reported this outcome. The mean difference in serum creatinine, on average, between LSSS and regular salt groups was 2.56 µmol/L (95% CI ‐0.59 to 5.71, I² = 0%, 616 participants, 3 RCTs, Analysis 1.59). Trials reporting on this outcome followed participants for 12 weeks, six months and between one and 1.5 years.

Subgroup analyses were undertaken for this outcome to explore whether there are differences in effects in subgroups based on hyperkalaemia risk. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping participants by risk of hyperkalaemia as per Table 6 suggests there may be no important clinical differences in average effects between participants not at risk, and those at possible risk, of hyperkalaemia (Analysis 1.60).

Microalbuminuria

Two studies reported on this outcome (RR 0.67, 95% CI 0.53 to 0.84, I² = 0%, 2382 participants, 2 RCTs, Analysis 1.61), with the number of participants in the LSSS group with microalbuminuria ranging from 45 to 64, and participants in the regular salt group with microalbuminuria ranging from 58 to 84 across trials. The two trials reporting this outcome had follow‐up at 18 months and three years.

Subgrouping participants by risk of hyperkalaemia could not be conducted as participants in both trials were at unclear risk of hyperkalaemia as per Table 6.

Macroalbuminuria

One trial reported on this outcome at 18 months (RR 0.48, 95% CI 0.16 to 1.39, 1903 participants, 1 RCT, Analysis 1.62), with five participants in the LSSS group experiencing macroalbuminuria events and 10 participants in the regular salt group experiencing macroalbuminuria events.

Change in urinary albumin‐to‐creatinine ratio (uACR)

One trial reported end values for this outcome at 18 months (MD ‐1.68, 95% CI ‐2.87 to ‐0.49, 1903 participants, 1 RCT, Analysis 1.63).

Change in fasting blood glucose (mmol/L)

Two trials reported on this outcome for this comparison. Average increases in fasting blood glucose ranged from 0.1 mmol/L to 0.22 mmol/L with LSSS, while average changes ranged from no change to an increase of 0.10 mmol/L in fasting blood glucose with regular salt, across the two trials. We did not pool these results into an overall effect estimate due to considerable heterogeneity (I² = 94%, Tau² = 0.56, Chi² = 17.06, df = 1 (P < 0.0001)), but rather presented the individual effect sizes between LSSS and regular salt per study (338 participants, 2 RCTs, Analysis 1.64). The two trials reporting on the outcome followed up participants for four and six months each.

Change in blood triglycerides (mmol/L)

Five trials reported on change in blood triglycerides for this comparison. Average changes in blood triglycerides ranged from a reduction of 0.7 mmol/L to an increase of 0.15 mmol/L with LSSS and from a reduction of 0.01 mmol/L to an increase of 0.9 mmol/L in blood triglycerides with regular salt across the five trials. The change in blood triglycerides, on average, between LSSS and regular salt groups was ‐0.11 mmol/L (95% CI ‐0.91 to 0.69, I² = 81%, 420 participants, 5 RCTs, Analysis 1.65). Trials reporting on the outcome followed up participants for five weeks, 12 weeks, four months, six months and nine months each.

Subgroup analyses were undertaken for this outcome due to the presence of substantial heterogeneity. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses.  Subgrouping by study duration (Analysis 1.66), age (Analysis 1.67), ethnicity (Analysis 1.68), BMI (Analysis 1.69), blood pressure status (Analysis 1.70), type of LSSS (Analysis 1.72), baseline sodium excretion (Analysis 1.73) and baseline potassium excretion (Analysis 1.74) suggests there may be no important differences in average effects between these subgroups. Subgrouping by the method of implementation of LSSS yielded a few studies in each subgroup, resulting in an analysis that was not powered to detect any differences in effect (Analysis 1.71). Subgrouping by gender could not be used to explore clinical differences as all studies were categorised into the same subgroup.

Change in total blood cholesterol (mmol/L)

Six trials reported on change in total blood cholesterol for this comparison. One trial each followed up participants for five weeks, 12 weeks and four months; two trials reporting on this outcome had follow‐up of approximately six months and one trial followed up participants for nine months. Across the trials, average changes in total blood cholesterol ranged from a reduction of 0.7 mmol/L to an increase of 0.16 mmol/L with LSSS and from a reduction of 0.63 mmol/L to an increase of 0.24 mmol/L with regular salt. When LSSS were compared to regular salt, the mean difference in total cholesterol change on average was ‐0.31 mmol/L (95% CI ‐0.74 to 0.12, I² = 85%, 509 participants, 6 RCTs, Analysis 1.75).

Subgroup analyses were undertaken for this outcome due to the presence of substantial heterogeneity. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping by study duration (Analysis 1.76), age (Analysis 1.77), ethnicity (Analysis 1.78), BMI (Analysis 1.79), blood pressure status (Analysis 1.80), implementation of LSSS (Analysis 1.81), type of LSSS (Analysis 1.82), baseline sodium excretion (Analysis 1.83) and baseline potassium excretion (Analysis 1.84) suggests there may be no important differences in average effects between these subgroups. Subgrouping by gender could not be used to explore clinical differences as all studies were categorised into the same subgroup.

Change in 24‐h urinary sodium excretion (mmol/24‐h)

Eleven trials reported on change in 24‐h urinary sodium excretion for this comparison. Average changes in this outcome ranged from a reduction of 75.5 mmol (1730 mg) sodium/24‐h to an increase of 20.2 mmol (460 mg) sodium/24‐h with LSSS and from a reduction of 31 mmol (710 mg) sodium/24‐h to an increase of 11 mmol (250 mg) sodium/24‐h with regular salt across the trials.  We did not pool these results into an overall effect estimate due to considerable heterogeneity (I² = 91%, Tau² = 595.13, Chi² = 107.72, df = 10 (P < 0.00001)), but rather presented the individual effect sizes between LSSS and regular salt per study (3885 participants, 11 RCTs, Analysis 1.85). Three trials reporting on this outcome followed up participants for four, five, and eight weeks; five trials followed up participants for three, four, nine, 18 and 60 months each; three trials reported on the outcome at approximately six months.

The stepped‐wedge cluster trial (unclear overall risk of bias) reporting on 605 participants reported little to no difference between a LSSS and regular salt for this outcome (Analysis 1.86).

Change in 24‐h urinary potassium excretion (mmol/24‐h)

Eleven trials reported on change in 24‐h urinary potassium excretion for this comparison. Average changes in this outcome ranged from a reduction of 4.4 mmol (170 mg) potassium/24‐h to an increase of 18.5 mmol (720 mg) potassium/24‐h with LSSS and from a reduction of 16 mmol (630 mg) potassium/24‐h to an increase of 4.6 mmol (180 mg) potassium/24‐h with regular salt across the trials. The meta‐analysis showed a difference, on average, favouring LSSS when compared to regular salt in the effect on 24‐h urinary potassium excretion between LSSS and regular salt groups (MD 11.44 mmol (450 mg) potassium/24‐h, 95% CI 7.62 to 15.26 mmol/24‐h [298 to 597 mg/24‐h], I² = 82%, 3885 participants, 11 RCTs, Analysis 1.87). Three trials reporting on the outcome followed up participants for four, five, and eight weeks, five trials followed up participants for three, four, nine, 18 and 60 months each; three trials reported on the outcome at approximately six months.

The stepped‐wedge cluster trial (unclear overall risk of bias) reporting on 605 participants found a similar direction of effect, though it was far smaller (Analysis 1.98).

Subgroup analyses were undertaken for this outcome due to the presence of substantial heterogeneity. In line with Cochrane guidance (Deeks 2020) detailing the limitations of subgroup analyses, caution was taken in the interpretation of findings from these subgroup analyses. Subgrouping by study duration (Analysis 1.88), age (Analysis 1.89), gender (Analysis 1.90), ethnicity (Analysis 1.91), BMI (Analysis 1.92), type of LSSS (Analysis 1.95), baseline sodium excretion (Analysis 1.96) and baseline potassium excretion (Analysis 1.97) suggests there may be no important differences in average effects between these subgroups.

Subgrouping by blood pressure status suggested some differences between subgroups (Analysis 1.93), but this was driven mainly by a large subgroup of participants of mixed hypertensive status from one study (Neal 2021). Therefore, the observed difference in effect was not considered to be attributable to hypertensive status. Subgrouping by the method of implementation of LSSS yielded a few studies in two of the three subgroups, resulting in an analysis that was not powered to detect any differences in effect (Analysis 1.94).

Comparison 2. Low‐sodium salt substitutes versus regular salt or no active intervention in children

Table 2 presents the effects of LSSS compared to regular salt in children on changes in DBP and SBP.

A single RCT randomising families as clusters and reporting on 92 children was included in this comparison. Key details about the study in this comparison, including study design, setting and overall risk of bias; characteristics of the intervention, comparator, population and outcomes; method of synthesis, and time points of measurement are included in the Overview of Synthesis and Included Studies (OSIS) table (Table 9).

Primary outcomes

For comparison 2, no studies reported on hypertension, blood pressure control, change in blood potassium, hyperkalaemia or hypokalaemia.

Change in diastolic blood pressure (DBP, mmHg)

GRADE assessment suggests that the evidence is very uncertain about the effect of LSSS on changes in DBP, when compared to regular salt, in children (very low‐certainty evidence, downgraded once risk of bias, once for indirectness and once for imprecision).

The average change in DBP was a reduction of 2.1 mmHg in the group that ate bread containing LSSS and a reduction of 5.87 mmHg in the group that ate bread containing regular salt for the single cluster‐RCT that reported this outcome at four months follow‐up. The mean difference, when comparing these groups, was 1.28 mmHg (95% CI ‐1.56 to 4.12, 92 participants, 1 RCT, very low‐certainty evidence, Analysis 2.1).

Change in systolic blood pressure (SBP, mmHg)

GRADE assessment suggests that the evidence is very uncertain about the effect of LSSS on changes in SBP, when compared to regular salt, in children (very low‐certainty evidence, downgraded once risk of bias, once for indirectness and once for imprecision).

The average change in SBP was a reduction of 5.1 mmHg in the group that ate bread containing LSSS and a reduction of 6.05 mmHg in the group that ate bread containing regular salt for the single cluster‐RCT that reported this outcome at four months follow‐up. The mean difference when comparing these groups was 0.12 mmHg (95% CI ‐4.41 to 4.64, 92 participants, 1 RCT, very low‐certainty evidence, Analysis 2.2).

Secondary outcomes

For comparison 2, no studies reported on adverse events (other), cardiovascular events, antihypertensive medication use, all‐cause mortality, cardiovascular mortality, bone densitometry measures, renal function, bone health, hyponatraemia, changes in fasting blood glucose, changes in blood triglycerides or changes in total blood cholesterol.

Growth changes (e.g. z‐scores for height‐ or length‐for‐age (HAZ or LAZ), weight‐for‐height (WHZ), weight‐for‐age (WAZ), BMI‐for‐age)

In the trial reporting on BMI changes in children, the unadjusted average reductions in BMI were 1.62 kg/m² with bread containing LSSS and 1.50 kg/m² with bread containing regular salt. The mean difference in BMI, on average, was 0.94 kg/m² (95% CI 0.85 to 1.03, 92 participants, 1 RCT, Analysis 2.3) when comparing these groups at four months.

Change in 24‐h urinary sodium excretion (mmol/24‐h)

The average change in 24‐h urinary sodium excretion was an increase of 11.4 mmol (262 mg) sodium/24‐h in the group that ate bread containing LSSS and a reduction of 3.2 mmol (74 mg) sodium/24‐h in the group that ate bread containing regular salt for the single cluster‐RCT that reported this outcome at four months follow‐up. The mean difference when comparing these groups was 14.60 mmol (336 mg) sodium/24‐h (95% CI ‐11.22 to 40.42 mmol/24‐h [‐258 to 929 mg/24‐h], 92 participants, 1 RCT, Analysis 2.4).

Change in 24‐h urinary potassium excretion (mmol/24‐h)

The average change in 24‐h urinary potassium excretion was a reduction of 1.6 mmol (64 mg) potassium/24‐h in the group that ate bread containing LSSS and a reduction of 5.7 mmol (223 mg) potassium/24‐h in the group that ate bread containing regular salt for the single cluster‐RCT that reported this outcome at four months follow‐up. The mean difference when comparing these groups was 4.10 mmol (160 mg) potassium/24‐h (95% CI ‐5.13 to 13.33 mmol/24‐h [‐201 to 521 mg/24‐h], 92 participants, 1 RCT, Analysis 2.5).

Comparison 3. Low‐sodium salt substitutes versus regular salt or no active intervention in pregnant women

No eligible studies in pregnant women were found.

Discussion

Summary of main results

This review examined the effects and safety of LSSS compared to regular salt or no active intervention on blood pressure and cardiovascular health in adults, children and pregnant women. We included 16 RCTs and ten cluster‐RCTs (n = 26) conducted in adults from a range of different settings, including nursing homes, hospitals, rural and suburban households and communities, as well as rural villages. Importantly, these 26 trials included various clinical subpopulations, with nearly two‐thirds of trials conducted in people with existing hypertension. We also included one cluster‐RCT including healthy children. The proportion of sodium chloride replacement in the LSSS interventions varied from approximately 3% to 77% with 24 trials replacing the sodium chloride with some potassium chloride (more details in the Characteristics of included studies section). We did not find any eligible studies in pregnant women. We also did not find any eligible prospective analytical cohort studies.

Low‐sodium salt substitutes versus regular salt or no active intervention in adults

Adult participants in groups allocated to LSSS had lowered DBP and SBP (range of reductions: 0.6 to 11.33 mmHg and 1.5 to 15.25 mmHg, respectively) on average, while those allocated to regular salt had smaller reductions and more variable results for both DBP (range of change: 7 mmHg reduction to 2.6 mmHg increase) and SBP (range of change: 6.8 mmHg reduction to 4 mmHg increase) on average.

In adult participants, meta‐analysis showed that LSSS probably reduce DBP slightly at up to 60 months. The 95% CIs of the pooled mean differences did not include clinically meaningful benefit (ranging from 1.36 to 3.50 mmHg lower), as we considered changes in DBP of greater than 5 mmHg to be clinically meaningful due to a 'significant' reduction in stroke risk of approximately 60% in high risk individuals (Thomopoulos 2017). However, small mean reductions for an entire population are more beneficial than very large reductions in only those at high risk (Verbeek 2021). Since the review focussed on the population‐level substitution of regular salt with LSSS, we applied a population perspective and derived a simplified population impact estimate (as described in Appendix 2) of the reduction in DBP observed in our meta‐analysis. This estimate suggested that the observed reduction in DBP corresponded to an estimated 60 (ranging from 35 to 83) stroke deaths prevented per 100,000 persons aged 50 years and older, per year. The observed small, important mean difference in DBP between LSSS and regular salt was confirmed by sensitivity analyses. Substantial heterogeneity was, however, detected in the pooled analysis for this outcome: subgrouping by various characteristics of included studies, participants and interventions suggests there may be no important clinical differences in average effects between the various subgroups.

In adult participants, meta‐analysis showed that LSSS probably reduce SBP slightly at up to 60 months. The 95% CIs of the pooled mean differences did not include clinically meaningful benefit (ranging from 3.50 to 6.01 mmHg lower), as we considered changes of at least 10 mmHg in SBP to be clinically meaningful. This cut‐off was informed by a systematic review with meta‐regression, quantifying the effects of blood pressure reduction on cardiovascular outcomes and death from large‐scale blood‐pressure lowering trials, which indicated that relative risk reductions are proportional to the magnitude of blood‐pressure reduction, with every 10 mmHg reduction in SBP significantly reducing the risk of major cardiovascular disease events (Ettehad 2016).  In addition, the overview and meta‐analysis by Thomopoulos 2017, investigating the effects of blood‐pressure lowering treatment and stratifying participants by total cardiovascular risk, reported a 'significant' stroke reduction of 60% with a 10 mmHg reduction in SBP in individuals at high risk. Our simplified population impact estimate suggested that the observed reduction in SBP with LSSS compared to regular salt corresponds to an estimated 53 (ranging from 40 to 65) stroke deaths prevented per 100,000 persons aged 50 years and older, per year. The observed small, important mean difference in SBP between LSSS and regular salt was broadly confirmed by sensitivity analyses. We explored the substantial heterogeneity detected in the pooled analysis for SBP using subgroup analyses, which suggest there may be no important clinical differences in average effects between the various subgroups.

For the presence of hypertension at 18 months, one trial showed little to no difference between the effects of LSSS and regular salt in adult participants. A stepped‐wedge cluster trial measuring incident hypertension indicated a more pronounced difference, with the hazard ratio for this outcome favouring LSSS.

We do not know whether there is a difference in the number of adult participants per group who achieve blood pressure control as the 95% CI limits of the pooled effect were consistent with the possibility for unimportant and important benefit, and most of the information for this outcome came from a study at unclear overall risk of bias. Furthermore, this study had limited generalisability as it investigated the effect of LSSS comprising 97% sodium chloride (regular salt) and therefore did not represent the composition of the majority of LSSS formulations on the market, most of which contain 65% or less sodium chloride.

We also do not know whether there is a difference in the number of adult participants per group who experience various cardiovascular events. Only 18 of these events were reported in total, including angina, serious cardiovascular events and cardiovascular symptoms, resulting in very imprecise 95% CI limits around the pooled effect. In addition, the pooled effect was driven by a large study in individuals at high risk of future vascular disease, thereby limiting the generalisability of the findings.

In adult participants, meta‐analysis showed that LSSS probably reduce non‐fatal stroke slightly at up to 60 months. The 95% CIs of the pooled risk and rate ratios included unimportant benefit and unimportant harm or no effect (ranging from a RR of 0.80 to 1.01), as we considered relative measures of less than 0.75 or greater than 1.25 to be important or 'appreciable' (Guyatt 2011). The simplified population impact estimate we derived (as described in Appendix 2) suggested that the observed relative risk when LSSS was compared to regular salt corresponds to an estimated 10 non‐fatal strokes prevented (ranging from 21 prevented to 1 caused) per 100,000 persons per year. The observed benefit with LSSS was not reflected in sensitivity analyses; with these instead showing highly imprecise effects of little to no effect, or harm.

Meta‐analysis in adult participants showed that LSSS probably reduce non‐fatal ACS slightly at up to 60 months. The 95% CIs of this effect included important and unimportant benefit (ranging from a rate ratio of 0.52 to 0.94), as we considered relative measures of less than 0.75 or greater than 1.25 to be important. The simplified population impact estimate we derived suggested that the observed relative risk when LSSS was compared to regular salt corresponds to an estimated 50 non‐fatal ACS events prevented (ranging from 10 to 80 prevented) per 100,000 persons per year.

In adult participants, meta‐analysis showed that LSSS probably reduce cardiovascular mortality slightly at up to 60 months. The 95% CIs of the pooled rate ratios included important benefit and no effect (ranging from a rate ratio of 0.60 to 1.00), as we considered relative measures of less than 0.75 or greater than 1.25 to be important.  The simplified population impact estimate we derived suggested that the observed relative risk when LSSS was compared to regular salt corresponds to an estimated 53 cardiovascular deaths prevented (ranging from 92 prevented to none caused or prevented) per 100,000 persons per year. The observed relative effect between LSSS and regular salt was confirmed by sensitivity analysis excluding trials at high risk of overall bias.

We do not know whether there is a difference in the number of stroke deaths per group in adults as the 95% CI limits of the pooled effect were consistent with the possibility for important harm and important benefit. Furthermore, the generalisability of the results to the general population was limited as the pooled effect was driven by a large secondary prevention trial in which 73% of included participants had previously had a stroke.

In adult participants, meta‐analysis showed that LSSS probably increase blood potassium slightly at up to one and a half years. The 95% CIs of the pooled mean differences did not include clinically meaningful changes (ranging from 0.07 to 0.18 mmol/L higher), as we considered changes in blood potassium of greater than 1.0 mmol/L to be clinically meaningful. This is based on the variations around the 'normal' blood potassium levels of 3.6 to 5.0 mmol/L (Cohn 2000), which can cause moderate hyperkalaemia (defined as 6.0 to 6.9 mmol/L; Hollander‐Rodriguez 2006 and Ahee 2000, respectively).

For the presence of hyperkalaemia at up to 60 months, some trials reported no events while others reported hyperkalaemic events in both the LSSS and regular salt groups. In adult participants, meta‐analyses showed that LSSS likely result in little to no difference in hyperkalaemia. The 95% CIs of the pooled risk ratios included important benefit and important harm (ranging from a RR of 0.46 to 2.38), as we considered relative measures of less than 0.75 or greater than 1.25 to be important. Only five trials reported on this outcome, though all information included in the meta‐analysis came from two trials in participants judged to be at possible and unclear risk of hyperkalaemia.

We do not know whether there is a difference in the number of adult participants per group who experience hypokalaemia events as one small trial reporting on this outcome reported zero events in both groups. In addition, this trial was at unclear overall risk of bias, and included only younger participants with hypertension treated with potassium‐sparing diuretics. Consequently, as the rationale for the administration of LSSS was potassium supplementation, we considered the generalisability of the evidence to be limited.

We also do not know whether there is a difference in the number of adult participants per group who experience other adverse events. Most of the information for this outcome was from studies at high or unclear overall risk of bias, events were very sparsely reported (39 in total), and outcomes were too diverse to pool.

Low‐sodium salt substitutes versus regular salt or no active intervention in children

Children (mean age 9.5 (SD 4.2) years) allocated to bread containing LSSS had lowered DBP and SBP (reduction of 2.1 mmHg and 5.1 mmHg, respectively) on average. Children (mean age 8.4 (SD 3.5) years) allocated to bread containing regular salt had larger reductions for both DBP and SBP (reduction of 5.87 mmHg and 6.05 mmHg, respectively) on average.

In all participants aged 18 or younger, results from a single included trial showed that the evidence is very uncertain about the effect of LSSS compared to regular salt on change in DBP as well as SBP at four months. The trial contributing to these outcomes was at unclear overall risk of bias, and reported effects with wide 95% CIs including both reductions and increases in DBP and SBP. Furthermore, the intervention was delivered in bread only, thereby limiting generalisability to discretionary use settings.

Overall completeness and applicability of evidence

Our review made use of a comprehensive search strategy with no language or date restrictions to identify all RCTs and prospective analytical cohort studies assessing the effect of LSSS on cardiovascular health in adults, children and pregnant women in the general population. We searched multiple sources of information for all studies and handsearched three relevant systematic reviews to identify additional studies. We also contacted study authors in cases where we required additional data or information.

We found only one trial, included under Comparison 1 for the effect of LSSS on adults, which additionally reported certain outcomes in children. The sparse evidence in children may be due to the relatively low prevalence of elevated blood pressure in children, with two large systematic reviews conducted in low‐ and middle‐income settings reporting pooled prevalence of 5.5% and 9.8% in children and adolescents from Africa and China, respectively (Noubiap 2017; Wang 2019). A large systematic review and meta‐regression conducted in 122,000 adolescents further indicated that this prevalence was disproportionately affecting adolescents in low‐ and middle‐income countries (De Moraes 2014). Consequently, maximal population‐level effects would not be achieved through targeting children, but rather adults; a group in which the global prevalence of hypertension (defined as a SBP ≥ 140 mmHg and DBP ≥ 90 mmHg) was approximately 32.5% for adults aged 30 years and older in 2019 (NCD Risk Factor Collaboration 2021).

We found no studies assessing the effect of LSSS in pregnant women. While chronic hypertension is a known significant risk factor for pre‐eclampsia (Bilano 2014), mean arterial pressure (MAP) in the first and second trimester has been suggested as a better predictor of pre‐eclampsia than SBP and DBP (Cnossen 2008). This study additionally reported that high MAP before pregnancy can be used as a predictor of pre‐eclampsia (Cnossen 2008) suggesting that the timing of the management of blood pressure is important, and that interventions to lower blood pressure prior to pregnancy might have the most success in avoiding complications for the mother and infant. In addition, pre‐eclampsia is a complex disease involving multiple organ systems (Palei 2013) and risk factors (Bilano 2014), and there is a paucity of evidence to support lifestyle interventions, such as reducing dietary sodium intake, for preventing pre‐eclampsia (Thangaratinam 2011).

Though our review included a reasonable distribution of studies from low‐ and middle‐income (n = 15) and high‐income countries (n = 12), the majority of trials (n = 15) were conducted in Asian populations. Furthermore, only one included study was conducted in South America; no eligible studies conducted in Africa, Oceania or North America were found. As different populations may use different quantities of discretionary salt, as a proportion of total salt intake, this may have an impact on the degree to which substitution with discretionary LSSS will alter sodium and potassium intakes. Subgroup analyses suggest there may be no important clinical differences in average effects on blood pressure between subgroups by ethnicity, although these analyses were limited. This suggested that the mix of ethnicities included in the review may not systematically bias our pooled estimates to Asian populations. It is more difficult to judge whether evidence from other countries and regions not represented in the review may have changed our pooled estimates. Such potential systematic differences could likely be categorised as biological and behavioural; such categories might plausibly include differences in baseline prevalence, and extent, of hypertension and differences in adherence to LSSS, respectively. Our review included populations with diverse baseline risks of hypertension and diverse baseline 24‐h urinary excretion of sodium and potassium, but subgroup analyses of these factors suggest there may be no important clinical differences in average effects on blood pressure.

Importantly, the findings of subgroup analyses should always be interpreted with caution as these can often be misleading (Deeks 2020). The likelihood of false negative and positive results increase rapidly when numerous subgroup analyses are undertaken and statistical power to find significant differences between subgroups is often lacking (Cuijpers 2021; Deeks 2020). In our review in particular, subgroup analyses were often limited by very few studies or participants contributing information to certain subgroups. As a result, findings from the subgroup analyses in our review may not all be sufficiently robust and should be interpreted with caution and with consideration of the described limitations.

Most of the trials included in our review assessed the effects of LSSS in participants with elevated blood pressure at enrolment. Due to limited data, we could not adequately examine effect modification for the relationship between LSSS use and outcomes by hypertension status.

All trials included in the review specifically excluded participants in whom it is known that an increased intake of potassium could cause harm, for example, people with CKD, type 1 or 2 diabetes mellitus, impaired renal function or those using potassium‐sparing medications. This limits the generalisability of our findings regarding the effects and safety of LSSS to these subpopulations, as well as to settings where a considerable proportion of the population may have undiagnosed conditions rendering increased potassium intake as potentially harmful.

Furthermore, the majority of included trials investigated the implementation of LSSS as a discretionary intervention. This limits the generalisability of our findings to non‐discretionary applications of LSSS, particularly use in condiments, and in manufactured food products or foods sold in restaurants, markets, cafeterias and street vendors.

We did not find evidence for a number of prespecified outcomes in the review. Across comparisons, no studies reported on diabetes mellitus (DM) diagnosis or hyponatraemia. The absence of evidence on hyponatraemia represents a gap in the evidence related to the safety of LSSS. While global sodium intakes are approximately double the WHO recommendation at present, thereby lowering the overall likelihood of hyponatraemia events, the individual risk of this event remains ‐ particularly in older people and those using thiazide diuretics (Filippatos 2017; Upadhyay 2009).

In Comparison 2, a single study reporting on the effects of LSSS in children did not report on hypertension, blood pressure control, hyper‐ or hypokalaemia, changes in blood potassium or adverse events. The paucity of evidence for these outcomes in children is likely due to the low general prevalence of hypertension as well as conditions and risk factors related to blood potassium imbalances.

Quality of the evidence

The interpretation of many of the trials included in the review is constrained by small sample sizes and considerable loss to follow‐up. Lack of baseline exposure and dietary status, as well as adherence to the allocated intervention, have been identified as factors that undermine the translation of dietary clinical trials into practice (Mirmiran 2021); these were generally sparsely and diversely reported by the included trials in the review. Nine trials included in the review were judged as having low risk of bias overall; 12 were at unclear overall risk of bias.

The pooled estimates of the effect of LSSS interventions compared to regular salt on hypertension, blood pressure control, and stroke mortality were downgraded for imprecision in line with the minimally contextualised approach we used. Imprecision was also identified for the outcomes: various cardiovascular events and other adverse events. These were composite outcomes and studies reporting on them were not designed or powered to detect differences between LSSS and regular salt groups.

The evidence for blood pressure control was considered indirect because it is questionable whether the intervention for the study contributing the most data was sufficiently lower in sodium compared to regular salt (only 3% replacement).

The evidence for hypokalaemia was downgraded for indirectness since the participants included in the single study reporting on this outcome were not sufficiently generalisable, being younger hypertensive adults on potassium‐depleting diuretics.

Indirectness was also identified for all outcomes relating to cardiovascular events and mortality. This was due to most, or all, of the information for these outcomes coming from studies including participants at high risk of cardiovascular disease, or individuals who had already experienced a cardiovascular event; thereby limiting the generalisability of the findings to the general population.

Furthermore, the certainty in pooled estimates for changes in blood pressure was affected by unexplained substantial heterogeneity. This may be due to clinical heterogeneity related to the various ways in which blood pressure measurements are collected in practice; various studies have shown variability between single measures and the mean of consecutive measurements (Burkard 2018), between blood pressure measured in a clinical setting and those obtained from daytime ambulatory measurements (Banegas 2017), and between measurements obtained from aneroid (inflatable cuff) and electronic sphygmomanometers (Shahbabu 2016).

Potential biases in the review process

The review may be affected by non‐reporting bias. For the unknowns that we are aware of ('known unknowns'), i.e. particular results from a trial not reported in a usable format, we contacted trial authors to request the data in a usable format. In cases where we did not obtain these data and they were consequently excluded, this is a limitation, as we cannot be certain how the inclusion of these data would have affected the pooled estimates. Despite this, we do observe agreements in the pooled mean differences in blood pressure changes reported in other similar reviews on this topic; though the estimated reductions are typically more conservative in our review. As a result, we think it is unlikely that these missing data would have meaningfully changed our pooled estimates. A number of studies (Li 2014; Li 2016; Mu 2009; Zhou 2013) reported allocating entire villages or households to LSSS or regular salt without explicit exclusion of children, though only one study reported separate data for the effect of the intervention in children. It is possible that data from family members aged 18 years or younger who were included as part of a randomised village or household may have changed our pooled estimates for children, though we anticipate that these data would be from a small subset of participants included in these trials. We are unsure whether these missing data would have meaningfully changed our pooled estimates of effect in children.

It is more difficult for us to judge the effect of 'unknown unknowns', i.e. entire eligible studies not detected by our comprehensive search strategy. We acknowledge the possibility that small, unpublished studies may not have been identified and included in the review as the interpretation of funnel plot asymmetry could not definitively rule this out. This was due to similar effect sizes despite varying inter‐trial standard errors for DBP and similar effect sizes as well as standard errors between trials for SBP.

We did not exclude any studies based on the duration of intervention, the formulation of LSSS, or participant characteristics. We did, however, exclude studies with multifactorial designs where the effect of LSSS could not be isolated, though it may have been relevant to the review question. The reason for excluding studies with such multi‐component interventions was that any observed changes in outcomes of interest could not be attributed to LSSS alone.

Agreements and disagreements with other studies or reviews

Pooled mean differences in blood pressure in our review are in line with previous systematic reviews on the effects of LSSS use in adults. These recent systematic reviews reported reductions in DBP and SBP ranging from 2.00 to 4.04 mmHg and 7.81 to 8.87 mmHg, respectively (Hernandez 2019; Jafarnejad 2020; Jin 2020).

However, it is important to take note of differences between these reviews and our review. One of the reviews included only studies conducted in Chinese study participants (Jin 2020), while another restricted studies to participants with stage 2 hypertension (Jafarnejad 2020). Both of these can be considered limited in their ability to generalise to guidelines for the general population: the review by Jin 2020, through its restriction to an ethno‐geographic group with, according to a recent publication from the China Hypertension Survey, a high prevalence of hypertension (Wang 2018); the review by Jafarnejad 2020, for including only studies in participants with progressive disease (Giles 2009). The inclusion criteria for these reviews consequently resulted in enriched populations in respect of hypertension, which may have influenced treatment effect through an increased number of participants with resistant hypertension (Yaxley 2015). Jafarnejad 2020 reported stratified analyses by several effect modifiers, demonstrating that LSSS use resulted in reductions in SBP and DBP in hypertensive adults of all ages, though reductions in SBP and DBP were slightly more pronounced in hypertensive adults younger than 65 years. Though these results were numerically very similar in the subgroup analyses conducted as part of our review, they were not fully reflected since we found greater reductions in SBP and smaller reductions in DBP in younger participants compared to older participants in the general population. As discussed previously, these differences may be due to differences in the included populations. A large review investigating the effect of antihypertensive medication (therefore including participants with overt hypertension) reported larger reductions in DBP in younger participants when compared to older and very old participants (Guang Wang 2005).

Only one review evaluated the certainty of evidence and presented low‐certainty evidence for reductions of 3.96 mmHg and 7.81 mmHg, in DBP and SBP respectively, at any length of follow‐up (Hernandez 2019). Low‐certainty evidence was also presented for the effect on triglycerides. For other outcomes, Hernandez 2019 concluded, on the basis of moderate‐certainty evidence, that LSSS use probably has little or no effect on mortality, while its effects on detected hypertension, total blood cholesterol, glucose, as well as urinary sodium and potassium excretion were very uncertain. The pooled effect estimates of these outcomes were similar between the review by Hernandez 2019 and our review, despite slight differences in included studies and the exact definitions of outcomes.

Authors' conclusions

Implications for practiceIn an adult population, the small pooled mean differences in DBP and SBP favoured LSSS when compared to regular salt. Though these were statistically significant, they were not considered to be clinically important at the individual level (moderate‐certainty evidence). However, small mean reductions from a population‐level intervention can be more beneficial than larger reductions in at‐risk patients (Verbeek 2021). When taking this perspective and considering the general population for which the WHO NUGAG Subgroup on Diet and Health is formulating the guidance, we considered the observed reductions in DBP and SBP when comparing the use of LSSS to regular salt to be small but important from a population perspective. While maximum follow‐up for these outcomes was at 60 months, the majority of trials reporting on blood pressure outcomes followed participants for six months or less.We observed a slight reduction in the risk ratio for non‐fatal stroke, non‐fatal ACS and cardiovascular mortality, favouring LSSS when compared to regular salt, in our review. Maximal follow‐up for these outcomes was at 60 months. The reduction in non‐fatal stroke and cardiovascular mortality was not considered to be clinically important at the individual level (both moderate‐certainty evidence). The reduction in non‐fatal ACS was considered to be clinically important at the individual level (moderate‐certainty evidence), though variation around the point estimate showed that both clinically important and unimportant benefits are possible. However, these benefits were all considered to be small but important from a population perspective.Importantly, many of the trials included in our review restricted participants to those with elevated blood pressure or hypertension at enrolment. Therefore, the use of this evidence to inform population‐based public health guidance requires careful consideration as offered by evidence‐informed approaches to guideline development (e.g. the GRADE approach (Zhang 2019)). Our systematic review failed to show that LSSS meaningfully reduces prevalent hypertension when compared to regular salt, with little or no difference in this outcome between the two groups at maximal follow‐up of 18 months; though a large stepped‐wedge cluster trial showed a more pronounced effect on incident hypertension over 30 months. Evidence on blood pressure control, various cardiovascular events, and stroke mortality was limited and we could not draw any conclusions about these outcomes.Furthermore, our systematic review found no meaningful increase in hyperkalaemia with LSSS when compared to regular salt, with little or no difference in effect for this important safety outcome at maximal follow‐up of five years. While global estimates of potassium intakes (Global Dietary Database 2022) are currently less than levels conditionally recommended by the WHO (WHO 2012b), thereby potentially delaying the onset of hyperkalaemia events, we consider five years likely to be sufficient follow‐up time to detect an event which usually develops over the course of weeks (Hollander‐Rodriguez 2006). It should be noted, however, that the evidence on hyperkalaemia presented in this review has several limitations. Very few studies reported on this important safety outcome, and studies that did report on hyperkalaemia also used variable criteria to define the condition (see Included studies). Most studies included in our review also had strict inclusion and exclusion criteria with regard to hyperkalaemia risk factors. All included trials specifically excluded participants in whom it is known that an increased intake of potassium could cause harm (e.g. people with chronic kidney disease). Only seven trials included participants judged to be at possible risk of hyperkalaemia, and four trials included participants at unclear risk of hyperkalaemia due mostly to limited reporting of the criteria assessed (Table 6). Though only five trials reported on this outcome, two including participants judged to be at possible risk and one including participants at unclear risk, all information in the meta‐analysis came from participants at possible or unclear risk. Therefore, caution should be taken in directly applying the results of our review to the general population, which would likely consist of some people in whom it is known that an increased intake of potassium could cause harm, as well as proportions of people with risk factors for hyperkalaemia, both diagnosed and undiagnosed.A small pooled mean difference in blood potassium between LSSS and regular salt indicated a small increase in this outcome for the LSSS group; this was not considered to be clinically important (moderate‐certainty evidence) given its impact on the upper end of a 'normal' blood potassium level would not reach levels indicative of moderate hyperkalaemia. Again, these findings should be interpreted with caution given the limited number of studies reporting on this outcome, and the strict exclusion by all studies of participants in whom it is known that an increased intake of potassium could cause harm. Evidence on hypokalaemia events and various adverse effects was limited and we could not draw any conclusions about these safety outcomes.Nearly all the trials in our review investigated the effects of LSSS implemented as a discretionary intervention. This restricts the generalisability of our findings to discretionary LSSS implementation, and we are unable to draw firm conclusions about non‐discretionary LSSS implementations (for example, where LSSS is added to manufactured foods). Furthermore, the contribution of discretionary salt to total sodium intake varies considerably across countries and settings. This variation is an important consideration when decisions are made about the implementation of LSSS since the absolute intakes of LSSS may vary across settings, thereby creating the possibility of variable impacts on both effectiveness and safety.It should also be noted that the majority of trials included in our review (24/27) investigated the effects of potassium‐containing LSSS, therefore, we are unable to draw firm conclusions about the effects and safety of LSSS that do not displace sodium with potassium. Variable impacts on effectiveness and safety might also be expected as a result of the potassium content of the LSSS used. In our review, potassium content in potassium‐containing products ranged from 10.1% to 50%. The selection of potassium‐containing LSSS based on potassium content is another important implementation consideration.  There was sparse evidence for the effect and safety of LSSS in children, with no studies assessing the discretionary use of LSSS in this population group. Given the limited evidence, we could not draw any conclusions about the effect of LSSS on DBP and SBP in children. We could not draw any conclusions about the safety of the use of LSSS in children given that no studies reported on safety outcomes. Given that no studies were found in pregnant women, we also could not draw any conclusions about the effect and safety of LSSS in pregnant women.Importantly, multi‐component and multi‐sector strategies are usually used to reduce sodium intake, blood pressure and cardiovascular disease risk in populations, and the use of LSSS to reduce sodium intake would be regarded as one of the approaches within these overall strategies.

Implications for researchGiven the sparse evidence available for the safety and effect of LSSS in children, and the lack of evidence in pregnant women, studies assessing the benefits and potential harms of LSSS in these groups are needed as a priority. In addition, the lack of information on hyponatraemia events in our review indicates that studies assessing this outcome in the context of LSSS use are needed, particularly since older people and those using certain classes of medication used to treat hypertension are disproportionately at risk. Given the limitations of the evidence relating to hyperkalaemia, robust studies with explicitly defined measures of this outcome are needed to better understand the safety implications of widespread LSSS use (discretionary and non‐discretionary). In addition, highly monitored trials including participants who may be at risk of hyperkalaemia, or evidence from prospective cohort studies including participants that are representative of the general population, would provide evidence that is more generalisable to widespread population‐level LSSS implementation. Similarly, robust studies assessing a participant population that is representative of the general population in terms of blood pressure status are required to better understand the effectiveness and safety of LSSS in people with normal blood pressure.Important evidence is still needed to answer the question of whether LSSS use sustainably decreases overall sodium intake, or whether it results in dietary compensation through behavioural modifications; such as, for example, an increased intake of products that are sources of non‐discretionary salt. Studies using reliable measures of dietary sodium and potassium intake, such as 24‐h urinary excretion are needed to assess the extent to which LSSS use reduces sodium intake, and increases potassium intake (when potassium‐containing LSSS are used) over the longer term. Other measures of dietary intake, such as dietary recalls, food frequency questionnaires and spot urine tests, have recently been shown to exhibit poor accuracy in individuals at high cardiovascular risk (Tsirimiagkou 2022). Evidence on the use of LSSS in manufactured and processed foods, particularly sauces and condiments, which represent a large proportion of dietary sodium intake in some regions of the world is also required to understand the extent to which population‐level exclusive replacement of discretionary salt would change global sodium intakes. Finally, studies assessing sodium intakes in populations and quantifying the level and extent of iodisation in LSSS are needed to ensure proactive recalibration of iodisation levels in salt in the context of increased LSSS use. Robust studies examining the effectiveness of multi‐component, multi‐sectoral strategies that include LSSS could further inform decision‐making to reduce sodium intake and cardiovascular disease risk.The quality and utility of future research on these questions would be optimised by prospective registration of studies, as well as publication of protocols and detailed data analysis plans of future studies. The use of appropriate reporting guidelines (e.g. CONSORT) would also support in the production of high‐quality evidence of maximum utility.Robust evidence linked to the resource implications of LSSS use is needed to better inform considerations related to population‐level implementation.

Acknowledgements

AcknowledgementsThe World Health Organization (WHO) provided funding to Stellenbosch University towards the cost of carrying out this systematic review. AB, MV, AS and CN are partly supported by the Research, Evidence and Development Initiative (READ‐It). READ‐It (project number 300342‐104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government’s official policies. This Cochrane Review is associated with the Research, Evidence and Development Initiative (READ‐It). READ‐It (project number 300342‐104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government’s official policies.We thank the following people:Vittoria Lutje, Information Specialist for Cochrane Infectious Diseases Group, for the Embase searches.Robin Featherstone, Information Specialist for Cochrane, for guidance in terms of the search strategy.Ms Nina Robertson, for assistance with screening and data extraction.Ms Yuan Chi, Information Specialist for Cochrane Campbell Global Ageing Partnership for Chinese translation, screening and data extraction of papers in Chinese.Dr Marty Chaplin, Senior Research Associate, Liverpool School of Tropical Medicine, UK and Statistical Editor, Cochrane Infectious Diseases for guidance related to the stepped‐wedge cluster‐randomised controlled trial.Prof Alfred Musekiwa, Associate Professor in Biostatistics, University of Pretoria, South Africa for providing guidance on meta‐analytic approach for rate ratios and conversion of incidence rates to rate ratios.Prof Razeen Davids, Head of the Division of Nephrology, Department of Medicine, Faculty of Medicine and Health Sciences, Stellenbosch University and Tygerberg Hospital, South Africa, for providing guidance on criteria for the assessment of hyperkalaemia risk and the time point ranges relevant to the review outcomes.Members of the WHO Nutrition Guidance Expert Advisory Group (NUGAG) Subgroup on Diet and Health for inputs on this review in line with the WHO guideline development process.Cochrane Public Health supported the authors in the development of this review. The following people conducted the editorial process for this review:Sign‐off Editor (final editorial decision): Julia L Finkelstein, Associate Professor, Division of Nutritional Sciences, and Deputy Director, Affiliate Cochrane Center for Nutrition, Cornell UniversityManaging Editors (selected peer reviewers, collated peer‐reviewer comments, provided editorial guidance to authors, edited the article): Colleen Ovelman and Joey Kwong, Cochrane Central Editorial ServiceEditorial Assistant (conducted editorial policy checks and supported editorial team): Lisa Wydrzynski, Cochrane Central Editorial ServiceCopy Editor (copy‐editing and production): Anne Lethaby, c/o Cochrane Production ServicePeer‐reviewers (provided comments and recommended an editorial decision): Rachael McLean, Department of Preventive and Social Medicine, Dunedin School of Medicine, University of Otago, New Zealand (clinical review); Jessica Rigutto‐Farebrother, Human Nutrition Laboratory, Institute of Food, Nutrition and Health, ETH Zürich, Switzerland (clinical review); Robert Walton, Cochrane UK (summary versions review); Rachel Richardson, Cochrane Evidence Production & Methods Directorate (methods review); Jo Platt, Cochrane GNOC (search review). One additional peer reviewer provided clinical peer review but chose not to be publicly acknowledged.

Competing interests

Declarations of interestAB: partly supported by the Research, Evidence and Development Initiative (READ‐It). READ‐It (project number 300342‐104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government's official policies; partial support paid to my institution for a scoping review on total fat intake and health outcomes other than measures of unhealthy weight gain (2020); a systematic review on low sodium salt substitutes and cardiovascular health (2020‐2021); rapid scoping reviews on coconut and palm oil intake and cardiovascular health (2021); a scoping review on the health effects of tropical oil consumption (2022). MV: partly supported by the Research, Evidence and Development Initiative (READ‐It). READ‐It (project number 300342‐104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government's official policies; partial support paid to my institution for a scoping review on total fat intake and health outcomes other than measures of unhealthy weight gain (2020); a systematic review on low sodium salt substitutes and cardiovascular health (2020‐2021); rapid scoping reviews on coconut and palm oil intake and cardiovascular health (2021); a scoping review on health effects of tropical oil consumption (2022). AS: partly supported by the Research, Evidence and Development Initiative (READ‐It). READ‐It (project number 300342‐104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government's official policies.CN: partly supported by the Research, Evidence and Development Initiative (READ‐It). READ‐It (project number 300342‐104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government's official policies; partial support paid to my institution for a scoping review on total fat intake and health outcomes other than measures of unhealthy weight gain; a systematic review on low sodium salt substitutes and cardiovascular health; rapid scoping reviews on coconut and palm oil intake and cardiovascular health; a scoping review on the health effects of tropical oil consumption. *CN is Co‐director of Cochrane Nutrition, and AB and MV are members of the Cochrane Nutrition local coordination team. These authors had no involvement in the editorial process for this review.

References

1. AllaertFA. Double-blind, randomized, crossover, controlled clinical trial of NaCl + Chitosan 3% versus NaCl on mild or moderate high blood pressure during the diet and lifestyle improvement period before possible prescription of an antihypertensive treatment. International Angiology2013;32(1):94-101. 23435397 PubMed
2. AllaertFA. Double blind, randomized, controlled clinical trial of NaCl + Chitosan 3% vs. NaCl on the decrease of blood pressure induced by a low salt diet in healthy prehypertensive volunteers. Archives of Cardiovascular Diseases Supplements2019;11(Suppl.3):e372. [DOI: 10.1016/j.acvdsp.2019.05.098] doi:10.1016/j.acvdsp.2019.05.098
3. AllaertFA. Effect of NaCl + Chitosan 3% vs. NaCl on high blood pressure parameters of healthy volunteers with prehypertension. Minerva Cardioangiologica2017;65(6):563-76. [DOI: 10.23736/S0026-4725.17.04451-6]28862406 doi:10.23736/S0026-4725.17.04451-6 PubMed
4. ArzilliF, TaddeiS, GraziadeiL, BichisaoE, GiovannettiR, SalvettiA. Potassium-rich and sodium-poor salt reduces blood pressure in hospitalized patients. Journal of Hypertension1986;4(5 Suppl 1):S347-50. 3553479 PubMed
5. Bernabe-OrtizA, Diez-CansecoF, GilmanRH, CárdenasMK, SackstederKA, MirandaJJ. Launching a salt substitute to reduce blood pressure at the population level: a cluster randomized stepped wedge trial in Peru. Trials2014;15:93. [DOI: 10.1186/1745-6215-15-93]PMC397655124667035 doi:10.1186/1745-6215-15-93 PubMed
6. Bernabe-OrtizA, Sal y RosasVG, Ponce-LuceroV, CárdenasMK, Carrillo-LarcoRM, Diez-CansecoF, et al. Effect of salt substitution on community-wide blood pressure and hypertension incidence. Nature Medicine2020;26:374-8. [DOI: 10.1038/s41591-020-0754-2]PMC761308332066973 doi:10.1038/s41591-020-0754-2 PubMed
7. NCT01960972. Launching a salt substitute to reduce blood pressure at the population level in Peru. clinicaltrials.gov/ct2/show/NCT01960972 (first received 11 October 2013).
8. Chang H-Y, Hu Y-W, Yue C-S, Wen Y-W, Yeh W-T, Hsu L-S, et al. Effect of potassium-enriched salt on cardiovascular mortality and medical expenses of elderly men. American Journal of Clinical Nutrition2006;83(6):1289-96. [DOI: 10.1093/ajcn/83.6.1289]16762939 doi:10.1093/ajcn/83.6.1289 PubMed
9. Chang H-Y, Pan W-H. Reply to CK Chow. American Journal of Clinical Nutrition2006;84(6):1553-4. [DOI: 10.1093/ajcn/84.6.1553]17158444 doi:10.1093/ajcn/84.6.1553 PubMed
10. ChowCK. Does potassium-enriched salt or sodium reduction reduce cardiovascular mortality and medical expenses?American Journal of Clinical Nutrition2006;84(6):1552-3. [DOI: 10.1093/ajcn/84.6.1552]17158444 doi:10.1093/ajcn/84.6.1552 PubMed
11. China Salt Substitute Study Collaborative Group. Salt substitution: a low-cost strategy for blood pressure control among rural Chinese. A randomized, controlled trial. Journal of Hypertension2007;25(10):2011-8. [DOI: 10.1097/HJH.0b013e3282b9714b]17885542 doi:10.1097/HJH.0b013e3282b9714b PubMed
12. HeFJ, MacGregorGA. Can a low-sodium, high-potassium salt substitute reduce blood pressure in rural Chinese people?Nature Reviews Cardiology2008;5(4):186-7. [DOI: 10.1038/ncpcardio1122]18227813 doi:10.1038/ncpcardio1122 PubMed
13. HuJ, JiangX, LiN, YuX, PerkovicV, ChenB, et al. Effects of salt substitute on pulse wave analysis among individuals at high cardiovascular risk in rural China: a randomized controlled trial. Hypertension Research2009;32(4):282-8. [DOI: 10.1038/hr.2009.7]19262499 doi:10.1038/hr.2009.7 PubMed
14. JiangX, HuJ, LiN, WuY. Effects of a reduced sodium increased potassium salt substitute on periphral and central blood pressure among high risk individuals in rural China: a randomised controlled trial. Journal of Hypertension2009;27 Suppl 4:S276.
15. LiN, PrescottJ, WuY, BarziF, YuX, ZhaoL, et al, China Salt Substitute Study Collaborative Group. The effects of a reduced-sodium, high-potassium salt substitute on food taste and acceptability in rural northern China. British Journal of Nutrition2009;101(7):1088-93. [DOI: 10.1017/S0007114508042360]18710605 doi:10.1017/S0007114508042360 PubMed
16. LiNY, WuY, BarziF, YuX, ZhaoL, NealBC. A low sodium, high potassium salt substitute substantially lowers blood pressure levels among high-risk individuals in rural northern China – the China Salt Substitute Study. Journal of Hypertension. Supplement: Official Journal of the International Society of Hypertension2006;24(4):S150-1.
17. NCT00145756. China Salt Substitute Study [China Salt Substitute Study - a randomised trial to determine the long-term effects of a low sodium, high potassium salt substitute on blood pressure among high-risk individuals in Northern China]. clinicaltrials.gov/ct2/show/NCT00145756 (first received 5 September 2005).
18. WangH, YangY, ShiJ, LiH, ZhouB, FuL. A randomized double controlled trial: the effect of salt substitution on blood pressure for hypertension patients among rural people in north of China. Zhongguo Shiyong Neike Za Zhi [Chinese Journal of Practical Internal Medicine]2010;30(6):552-3.
19. GeleijnseJM, WittemanJC, BakAA, Den BreeijenJH, GrobbeeDE. Long-term moderate sodium restriction does not adversely affect the serum HDL/total cholesterol ratio. Journal of Human Hypertension1995;9(12):975-9. 8746642 PubMed
20. GeleijnseJM, WittemanJC, BakAA, Den BreeijenJH, GrobbeeDE. Reducing blood pressure by use of a low sodium high potassium, high magnesium mineral salt in older persons with mild to moderate hypertension [Verlaging van de bloeddruk door gebruik van een mineraalzout met een verlaagd natriumgehalte en een verhoogd kalium- en magnesiumgehalte bij ouderen met milde tot matige hypertensie]. Nederlands Tijdschrift voor Geneeskunde1995;139:512-8.
21. GeleijnseJM, WittemanJC, BakAA, Den BreeijenJH, GrobbeeDE. Reduction in blood pressure with a low sodium, high potassium, high magnesium salt in older subjects with mild to moderate hypertension. BMJ (Clinical research ed.)1994;309(6952):436-40. [DOI: 10.1136/bmj.309.6952.436.]PMC25409677920126 doi:10.1136/bmj.309.6952.436. PubMed
22. GilleranG, O'LearyM, BartlettWA, VinallH, JonesAF, DodsonPM. Effects of dietary sodium substitution with potassium and magnesium in hypertensive type II diabetics: a randomised blind controlled parallel study. Journal of Human Hypertension1996;10(8):517-21. 8895035 PubMed
23. HuJ, ZhaoL, ThompsonB, ZhangY, WuY. Effects of salt substitute on home blood pressure differs according to age and degree of blood pressure in hypertensive patients and their families. Clinical and Experimental Hypertension2018;40(7):664-72. [DOI: 10.1080/10641963.2018.1425415]29400576 doi:10.1080/10641963.2018.1425415 PubMed
24. Hu J-H, Zhao L-C, LiX, Wu Y-F. Effects of salt substitution on blood pressure using home measurements in essential hypertensive patients: a double-blinded, randomized controlled trial. Chinese Journal of Hypertension2014;22(1):42-6.
25. VisserME. Low sodium salt Substitutes (LSSS) systematic review: request for outcome data [personal communication]. Email to: J Hu 10 July 2020.
26. KawasakiT, ItohK, KawasakiM. Reduction in blood pressure with a sodium-reduced, potassium- and magnesium-enriched mineral salt in subjects with mild essential hypertension. Hypertension Research1998;21(4):235-43. [DOI: 10.1291/hypres.21.235]9877516 doi:10.1291/hypres.21.235 PubMed
27. LiZ, FengX, WuT, YanL, ElliottP, WuY. Randomised trial on effect of involving media reporters in salt reduction programme to increase media reports and the public's knowledge, belief and behaviors on salt and health: Changzhi reporters trial. PLoS One2021;16(7):e0252989. [DOI: 10.1371/journal.pone.0252989]PMC829410134283844 doi:10.1371/journal.pone.0252989 PubMed
28. LiZ, YangH, WuY, ElliottP, DuS, X Feng, et al. Relationship of the dietary sodium substitution with low sodium and high potassium salt with the blood pressure among middle aged and elderly people: a randomized controlled study [低钠富钾替代盐与中老年人群血压关系的 随机对照研究]. Chinese Journal of Geriatrics2014;33(4):365-7.
29. ChuH, ZhangJ, FettersMD, NiuW, LiH, LiN, et al. A mixed methods process evaluation of a clustered-randomized controlled trial to determine the effects of community-based dietary sodium reduction in rural China. Frontiers in Medicine2021;8:646576. [DOI: 10.3389/fmed.2021.646576]PMC819279934124088 doi:10.3389/fmed.2021.646576 PubMed
30. JardineM, LiN, NinomiyaT, FengX, ZhangJ, ShiJ, et al. Dietary sodium reduction reduced albuminuria in 1,903 rural Chinese: a cluster randomised trial. Nephrology2014;19 Suppl 4:28.
31. JardineM, LiN, NinomiyaT, FengX, ZhangJ, ShiJ, et al. Dietary sodium reduction reduces albuminuria: a cluster randomized trial. Journal of Renal Nutrition2019;29(4):276-84. [DOI: 10.1053/j.jrn.2018.10.009]30591358 doi:10.1053/j.jrn.2018.10.009 PubMed
32. LiN, YanL, NiuW, YaoC, FengX, ZhangJ, et al. China Rural Health Initiative - Sodium Reduction Study: the effects of a community-based sodium reduction program on 24hr urinary sodium and blood pressure in rural China. Circulation2013;128(24):2707.
33. LiN, YanLL, NiuW, LabartheD, FengX, ShiJ. A large-scale cluster randomized trial to determine the effects of community-based dietary sodium reduction - the China Rural Health Initiative Sodium Reduction Study. American Heart Journal2013;166(5):815-22. [DOI: 10.1016/j.ahj.2013.07.009]PMC391968424176436 doi:10.1016/j.ahj.2013.07.009 PubMed
34. LiN, YanLL, NiuW, YaoC, FengX, ZhangJ, et al. The effects of a community-based sodium reduction program in rural China – a cluster-randomized trial. PLOS One2016;11(12):e0166620. [DOI: 10.1371/journal.pone.0166620]PMC514783427935977 doi:10.1371/journal.pone.0166620 PubMed
35. NCT01259700. High cardiovascular risk management and salt reduction in rural villages in China [China Rural Health Initiative: high cardiovascular risk management and salt reduction in rural villages in China - two parallel large cluster randomized controlled trials]. clinicaltrials.gov/ct2/show/NCT01259700 (first received 14 December 2010).
36. WangX, LiX, VaartjesI, NealB, BotsML, HoesAW, et al. Does education level affect the efficacy of a community based salt reduction program? - a post-hoc analysis of the China Rural Health Initiative Sodium Reduction Study (CRHI-SRS). BMC Public Health2016;16(1):759. [DOI: 10.1186/s12889-016-3454-6]PMC498243427515930 doi:10.1186/s12889-016-3454-6 PubMed
37. YanLL, LiN, NiuW, ChuH, ZhangJ, HaoZ, et al. China Rural Health Initiative - Sodium Reduction Study: a cluster randomized trial on a community-based dietary sodium reduction program. Annals of Nutrition and Metabolism2013;63 Suppl 1:48.
38. Mu J-J, Liu Z-Q, YangJ, Liang Y-M, Zhy D-J, Wang Y-X, et al. Long term observation in effects of potassium and calcium supplementation on arterial blood pressure and sodium metabolism in adolescents with higher blood pressure. Zhonghua Yu Fang Yi Xue Za Zhi [Chinese Journal of Preventive Medicine]2003;37(2):90-2. 12839657 PubMed
39. MuJ, LiuZ, LiuF, XuX, LiangY, ZhuD. Family-based randomized trial to detect effects on blood pressure of a salt substitute containing potassium and calcium in hypertensive adolescents. American Journal of Hypertension2009;22(9):943-7. [DOI: 10.1038/ajh.2009.136]19661927 doi:10.1038/ajh.2009.136 PubMed
40. HuangL, TianM, LiN, ElliottP, YanL, LabartheD, et al. Interim effects of salt substitution on urinary electrolytes and blood pressure in the China Salt Substitute and Stroke Study (SSaSS). Journal of Hypertension2018;36(Suppl.3):e279. [DOI: 10.1097/01.hjh.0000549138.27560.0d]31986290 doi:10.1097/01.hjh.0000549138.27560.0d PubMed
41. HuangL, TianM, YuJ, LiQ, LiuY, YinX, et al. Interim effects of salt substitution on urinary electrolytes and blood pressure in the China Salt Substitute and Stroke Study (SSaSS). American Heart Journal2020;221:136-45. [DOI: 10.1016/j.ahj.2019.12.020]31986290 doi:10.1016/j.ahj.2019.12.020 PubMed
42. HuangL, WoodwardM, StepienS, TianM, YinX, HaoZ, et al. Spot urine samples compared with 24-h urine samples for estimating changes in urinary sodium and potassium excretion in the China Salt Substitute and Stroke Study. International Journal of Epidemiology2018;47(6):1811-20. [DOI: 10.1093/ije/dyy206]30307504 doi:10.1093/ije/dyy206 PubMed
43. IngelfingerJR. Can salt substitution save at-risk persons from stroke?New England Journal of Medicine2021;385(12):1137-8. [DOI: 10.1056/NEJMe2112857]34459568 doi:10.1056/NEJMe2112857 PubMed
44. NCT02092090. China Salt Substitute and Stroke Study (SSaSS) [A large-scale cluster randomized trial to determine the effects of sodium reduction on stroke: the China Salt Substitute and Stroke Study (SSaSS)]. clinicaltrials.gov/ct2/show/NCT02092090 (first received 19 March 2014).
45. NealB, TianM, LiN, ElliottP, YanLL, LabartheDR, et al. Rationale, design, and baseline characteristics of the Salt Substitute and Stroke Study (SSaSS)-a large-scale cluster randomized controlled trial. American Heart Journal2017;188:109-17. [DOI: 10.1016/j.ahj.2017.02.033]28577665 doi:10.1016/j.ahj.2017.02.033 PubMed
46. NealB, WuY, FengX, ZhangR, ZhangY, ShiJ, et al. Effect of salt substitution on cardiovascular events and death. New England Journal of Medicine2021;385(12):1067-77. [DOI: 10.1056/NEJMoa2105675]34459569 doi:10.1056/NEJMoa2105675 PubMed
47. ZhangT, ChangX, LiuW, LiX, WangF, HuangL, et al. Comparison of sodium, potassium, calcium, magnesium, zinc, copper and iron concentrations of elements in 24-h urine and spot urine in hypertensive patients with healthy renal function. Journal of Trace Elements in Medicine and Biology2017;44:104-8. [DOI: 10.1016/j.jtemb.2017.06.006]28965563 doi:10.1016/j.jtemb.2017.06.006 PubMed
48. ZhaoY, LiuWL, LiuS, LiXX, YinT, LiuXY, et al. Estimating 24-h urinary sodium excretion from casual spot urine specimen among hypertensive patients in Northwest China: the Salt Substitute and Stroke Study. Public Health Nutrition2021;24(4):604-10. [DOI: 10.1017/S1368980019005019]PMC1157482232345383 doi:10.1017/S1368980019005019 PubMed
49. OmvikP, MykingOL. Unchanged central hemodynamics after six months of moderate sodium restriction with or without potassium supplement in essential hypertension. Blood Pressure1995;4(1):32-41. [DOI: 10.3109/08037059509077565]7735495 doi:10.3109/08037059509077565 PubMed
50. LaiYH, ChenJR, JengJS, ChenCM, BaiCH, LinRT, et al. The effect of intervention with potassium and magnesium-enriched salt on neurological performance in stroke patients. Cerebrovascular Diseases2014;38(Suppl.1):98. [DOI: 10.1159/000367674] doi:10.1159/000367674
51. NCT02910427. The effect of intervention with potassium and/or magnesium-enriched salt on neurological performance of stroke patients. https://clinicaltrials.gov/ct2/show/NCT02910427 (first received 22 September 2016).
52. Pan W-H, Lai Y-H, Yeh W-T, Chen J-R, Jeng J-S, Bai C-H, et al. Intake of potassium- and magnesium-enriched salt improves functional outcome after stroke: a randomized, multicenter, double-blind controlled trial. American Journal of Clinical Nutrition2017;106(5):1267-73. [DOI: 10.3945/ajcn.116.148536]28877896 doi:10.3945/ajcn.116.148536 PubMed
53. PereiraMA, GalvãoR, ZanellaMT. Effects of potassium supplementation by salt on arterial blood pressure and insulin resistance in hypertensive obese patients on diuretic therapy [Efeitos da suplementação de potássio via sal de cozinha sobre a pressão arterial e a resistência à insulina em pacientes obesos hipertensos em uso de diuréticos]. Revista de Nutrição [Brazilian Journal of Nutrition]2005;18(1):5-17. [DOI: 10.1590/S1415-52732005000100001] doi:10.1590/S1415-52732005000100001
54. ISRCTN01739816. The effect of replacing regular salt with SmartSalt® mineral salt on blood pressure in middle-aged subjects with high blood pressure or with mild hypertension: a randomised controlled two-arm human study. doi.org/10.1186/ISRCTN01739816 (first received 11 November 2009).
55. SarkkinenES, KastarinenMJ, NiskanenTH, KarjalainenPH, VenäläinenTM, UdaniJK, et al. Feasibility and antihypertensive effect of replacing regular salt with mineral salt - rich in magnesium and potassium - in subjects with mildly elevated blood pressure . Nutrition Journal2011;10(1):88. [DOI: 10.1186/1475-2891-10-88]PMC317515121888642 doi:10.1186/1475-2891-10-88 PubMed
56. SuppaG, PollaviniG, AlbertiD, SavonittoS. Effects of a low-sodium high-potassium salt in hypertensive patients treated with metoprolol: a multicentre study. Journal of Hypertension1988;6(10):787-90. 3058796 PubMed
57. BjoernsboKS, RiisNL, AndreasenAH, PetersenJ, LassenAD, TrolleE, et al. Salt reduction intervention in families investigating metabolic, behavioral and health effects of targeted intake reductions: study protocol for a four months three-armed, randomized, controlled “real-life” trial. International Journal of Environmental Research and Public Health2019;16(19):3532. [DOI: 10.3390/ijerph16193532]PMC680145631546576 doi:10.3390/ijerph16193532 PubMed
58. RiisNL, BjoernsboKS, ToftU, TrolleE, HyldigG, HartleyIE, et al. Impact of salt reduction interventions on salt taste sensitivity and liking, a cluster randomized controlled trial. Food Quality and Preference2021;87:104059. [DOI: 10.1016/j.foodqual.2020.104059] doi:10.1016/j.foodqual.2020.104059
59. ToftU, RiisNL, LassenAD, TrolleE, AndreasenAH, FrederiksenAK, et al. The effects of two intervention strategies to reduce the intake of salt and the sodium-to-potassium ratio on cardiovascular risk factors. A 4-month randomised controlled study among healthy families. Nutrients2020;12(5):1467. [DOI: 10.3390/nu12051467]PMC728465232438659 doi:10.3390/nu12051467 PubMed
60. VisserME. Request for information on the composition of Viva salt [personal communication]. Email to: A Woitrin 22 September 2021.
61. NCT03909659. The effects on blood pressure of salt substitute among adults with hypertension in India [A randomized-controlled trial to determine the effects on blood pressure of a reduced-sodium added-potassium salt substitute among adults with hypertension in India]. https://clinicaltrials.gov/ct2/show/NCT03909659 (first received 10 April 2019).
62. ThoutSR, YuJ, TianM, HuffmanMD, ArnottC, LiQ, et al. Rationale, design, and baseline characteristics of the Salt Substitute in India Study (SSiIS): the protocol for a double-blinded, randomized-controlled trial. Journal of Clinical Hypertension2020;22(8):1504-12. [DOI: 10.1111/jch.13947]PMC802975332710677 doi:10.1111/jch.13947 PubMed
63. VisariaA, ShahaniJ, ShahM, ModakA, ChilakapatiR. Maximizing the potential of the Salt Substitute in India Study. Journal of Clinical Hypertension2021;23(1):197-8. [DOI: 10.1111/jch.14109]PMC802972333222393 doi:10.1111/jch.14109 PubMed
64. YuJ, ThoutSR, LiQ, TianM, MarklundM, ArnottC, et al. Effects of a reduced-sodium added-potassium salt substitute on blood pressure in rural Indian hypertensive patients: a randomized, double-blind, controlled trial. American Journal of Clinical Nutrition2021;114(1):185-93. [DOI: 10.1093/ajcn/nqab054.]33782684 doi:10.1093/ajcn/nqab054. PubMed
65. ZhangH, LiDF, LiWW, WangQ, BrunnerHR, ZhangBQ, et al. Safety of long-term enriched potassium salt consumption and effect on blood pressure in Chinese. Journal of Hypertension2015;33 e-Suppl 1:e148-9.
66. ZhangH, WangQ, BrunnerHR, LiW, ZhangB, DongY, et al. Effects of long-term enriched potassium salt consumption on premature death in Chinese living in nursing houses - preliminary analysis. Journal of Hypertension2018;36 e-Suppl 3:e286-7.
67. ZhangH, WangQ, GuoY, LiD, ZhangB, DongY, et al. Effect of long-term enriched potassium salt consumption on all causes mortality in Chinese living in nursing houses - a preliminary analysis. Journal of Hypertension2016;34 e-Suppl 1:e52-3.
68. NCT01429246. China Salt Substitute Study in Tibet (CSSS-Tibet) [China Salt Substitute Study in Tibet: efficacy of salt substitute in reducing blood pressure in hypertensive adults]. clinicaltrials.gov/ct2/show/NCT01429246 (first received 7 September 2011).
69. ZhaoX, KeL, LiS, LiN, YanLL, BaS, et al. Effects on blood pressure of a low-sodium, high-potassium salt substitute among Tibetan Chinese: a randomized controlled trial. Circulation2010;122(2):e353.
70. ZhaoX, YinX, LiX, YanLL, LamCT, LiS, et al. Using a low-sodium, high-potassium salt substitute to reduce blood pressure among Tibetans with high blood pressure: a patient-blinded randomized controlled trial. PLOS One2014;9(10):e110131. [DOI: 10.1371/journal.pone.0110131]PMC420628925338053 doi:10.1371/journal.pone.0110131 PubMed
71. Yang G-H, ZhouX, Ji W-J, Liu J-X, SunJ, ShiR, et al. Effects of a low salt diet on isolated systolic hypertension: a community-based population study. Medicine2018;97(14):e0342. [DOI: 10.1097/MD.0000000000010342]PMC590226929620663 doi:10.1097/MD.0000000000010342 PubMed
72. YangGH, ZhouX, Ji W-J, Liu J-X, SunJ, ShiR, et al. Effects of low salt diet on isolated systolic hypertension: a community-based population study. Journal of Hypertension2018;36 e-Suppl 3:e298. 10.1097/MD.0000000000010342PMC590226929620663 doi:10.1097/MD.0000000000010342 PubMed
73. ZhouX, Liu J-X, ShiR, YangN, Song D-L, PangW, et al. Compound ion salt, a novel low-sodium salt substitute: from animal study to community-based population trial. American Journal of Hypertension2009;22(9):934-42. [DOI: 10.1038/ajh.2009.135]19661926 doi:10.1038/ajh.2009.135 PubMed
74. SunH, MaB, WuX, WangH, ZhouB. Long-term effect of salt substitute on all-cause and cardiovascular disease mortality: an exploratory follow-up of a randomized controlled trial. Frontiers in Cardiovascular Medicine2021;8:645902. [DOI: 10.3389/fcvm.2021.645902]PMC816510334079827 doi:10.3389/fcvm.2021.645902 PubMed
75. ZhouB, Wang H-L, Wang W-L, Wu X-M, Fu L-Y, Shi J-P. Long-term effects of salt substitution on blood pressure in a rural north Chinese population. Journal of Human Hypertension2013;27(7):427-33. [DOI: 10.1038/jhh.2012.63.]23254595 doi:10.1038/jhh.2012.63. PubMed
76. ZhouB, WebsterJ, Fu L-Y, Wang H-L, Wu X-M, Wang W-L, et al. Intake of low sodium salt substitute for 3 years attenuates the increase in blood pressure in a rural population of North China - a randomized controlled trial. International Journal of Cardiology2016;215:377-82. [DOI: 10.1016/j.ijcard.2016.04.073]27128565 doi:10.1016/j.ijcard.2016.04.073 PubMed
77. AllaertF, MeleroC. CO-42: Observational study of the effect of substituting NaCl with NaCl+ chitosan 3% (Symbiosal®) in the diet of elderly subjects on their blood pressure values [Étude observationnelle de l‘effet d‘une substitution du NaCl par du NaCl + chitosan 3% (Symbiosal®) dans l‘alimentation des personnes âgées sur leurs paramètres tensionnels]. Annales de Cardiologie et d'Angéiologie2015;64(Supplement 1):S19.
78. AllaertFA. Observational study of the effect of substituting NaCl with NaCl+Chitosan 3% (Symbiosal) in the diet of elderly subjects on their blood pressure values. Minerva Cardioangiologica2015;63(6):515-23. 26657533 PubMed
79. BaekSH, AhnJW, LeeHR, ChoSH, KimJH. Anti-hypertensive effect of a solar salt diet in elderly hypertensive patients: a preliminary randomized, double-blind clinical trial. Korean Journal of Health Promotion2015;15(3):98-107.
80. BarrosCL. Substitution of Regular Salt with Light Salt: Impact on Blood Pressure of Hypertensive Patients (Impacto da Substituição do Sal Convencional Pelo Sal Light sobre a Pressão Arterial de Hipertensos [Master's thesis]. Goiânia (Brazil): Universidade Federal de Goiás Faculdade de Nutrição Programa de Pós-Graduação em Nutrição e Saúde, 2013.
81. BarrosCL, SousaAL, ChinemBM, RodriguesRB, JardimTS, CarneiroSB, et al. Impact of light salt substitution for regular salt on blood pressure of hypertensive patients. Arquivos Brasileiros de Cardiologia 2015;104(2):128-35. 10.5935/abc.20140174PMC437565625409877 doi:10.5935/abc.20140174 PubMed
82. ChiCTR1800016804. Home cook intervention to reduce salt intake: in China - a cluster randomized controlled trial. www.chictr.org.cn/showprojen.aspx?proj=25400 (first received 25 June 2018).
83. ChiCTR1800018119. A community-based comprehensive intervention to reduce salt intake in the Chinese population: a cluster randomized controlled trial. www.chictr.org.cn/com/25/showproj.aspx?proj=30631 (first received 31 August 2018).
84. ChiCTR1800019727. Effects of salt substitution on blood pressure using home blood pressure, ambulatory blood pressure and clinical blood pressure in essential hypertensive patients: a randomized controlled trial. www.chictr.org.cn/showprojen.aspx?proj=31036 (first received 25 November 2018).
85. ChiCTR2000033349. Zhejiang province pilot of intervention evaluation project of salt reduction and prevention and control of hypertension: a cluster randomized controlled trial. www.chictr.org.cn/hvshowproject.aspx?id=36685 (first received 29 May 2020).
86. DentA, WalmsleyD, DhandapaniS. Hyperkalaemia is a risk with low sodium salt in vulnerable patients. British Medical Journal 2011;343:2. 10.1136/bmj.d451421771833 doi:10.1136/bmj.d4514 PubMed
87. DoorenbosCJ, VermeijCG. Lesson of the week: danger of salt substitutes that contain potassium in patients with renal failure. British Medical Journal (International Edition) 2003;326(7379):35-6. 10.1136/bmj.326.7379.35PMC112492612511461 doi:10.1136/bmj.326.7379.35 PubMed
88. FarrandC, MacGregorG, CampbellNR, WebsterJ. Potential use of salt substitutes to reduce blood pressure. Journal of Clinical Hypertension2019;21(3):350-4. 10.1111/jch.13482PMC803035130690859 doi:10.1111/jch.13482 PubMed
89. HeFJ, TanMN, SongJ, MacGregorGA. Salt substitution to lower population blood pressure. Nature Medicine2020;26(3):313-4. 10.1038/s41591-020-0784-932132716 doi:10.1038/s41591-020-0784-9 PubMed
90. HuestonWJ. Use of salt substitutes in the treatment of diuretic-induced hypokalemia. Journal of Family Practice1989;29(6):623-6. 2687427 PubMed
91. ItohK, KawasakiT. Clinical and nutritional study of low sodium, high potassium and high magnesium salt on sensory test and electrolyte balance. Japanese Journal of Nutrition1997;55:263-72.
92. JandaJ, VeleminskyM, SulakovaT, ProchazkaB, EliasekJ, StranskyP, et al. Effect of the DASH-diet and salt Kardisal on blood pressure in adolescents with prehypertension (cooperative multicentre interventional study). Neuroendocrinology Letters2017;38(8):544-8. 29504732 PubMed
93. JefferyRW, PiriePL, ElmerPJ, Bjornson-BensonWM, MullenbachVA, KurthCL, et al. Low-sodium, high-potassium diet: feasibility and acceptability in a normotensive population. American Journal of Public Health1984;74(5):492-4. 10.2105/ajph.74.5.492PMC16516276711725 doi:10.2105/ajph.74.5.492 PubMed
94. JPRN-UMIN000012560. Effect of using low sodium seasonings (miso and soy sauce) or monitoring salt concentration of dishes prepared at home for 12 weeks on the reduction of sodium intake among healthy middle-aged Japanese individuals: a randomized, controlled, crossover trial. upload.umin.ac.jp/cgi-open-bin/icdr_e/ctr_his_list.cgi?recptno=R000013246 (first received 12 December 2013).
95. KarppanenH, TanskanenA, TuomilehtoJ, PuskaP, VuoriJ, JanttiV, et al. Safety and effects of potassium-containing and magnesium-containing low sodium-salt mixtures . Journal of Cardiovascular Pharmacology1984;6:S236-43. 10.1097/00005344-198400061-000376204148 doi:10.1097/00005344-198400061-00037 PubMed
96. KatzA, RosenthalT, MaozC, PelegE, ZeidensteinR, LeviY. Effect of a mineral salt diet on 24-h blood pressure monitoring in elderly hypertensive patients. Journal of Human Hypertension1999;13(11):777-80. 10.1038/sj.jhh.100083710578223 doi:10.1038/sj.jhh.1000837 PubMed
97. LambertK, ConleyM, DumontR, MontgomeryR, NobleS, NotarasS, et al. Letter to the editor on "Potential use of salt substitutes to reduce blood pressure". Journal of Clinical Hypertension2019;21(10):1609-10. 10.1111/jch.13685PMC803044731448881 doi:10.1111/jch.13685 PubMed
98. LittleP, KellyJ, BarnettJ, DorwardM, MargettsB, WarmD. Randomised controlled factorial trial of dietary advice for patients with a single high blood pressure reading in primary care. BMJ2004;328(7447):1054. 10.1136/bmj.38037.435972.EEPMC40384915082472 doi:10.1136/bmj.38037.435972.EE PubMed
99. VisserME. Request for study data [personal communication]. Email to: P Little 10 September 2021.
100. MalekiA, SoltanianAR, ZeraatiF, SheikhV, PoorolajalJ. The flavor and acceptability of six different potassium-enriched (sodium reduced) iodized salts: a single-blind, randomized, crossover design. Clinical Hypertension2016;22(18):eCollection 2016. [DOI: 10.1186/s40885-016-0054-9]PMC517898528031983 doi:10.1186/s40885-016-0054-9 PubMed
101. MaruyaS, TakachiR, KandaM, NakadateM, IshiharaJ. Short-term effects of salt restriction via home dishes do not persist in the long term: a randomized control study. Nutrients2020;12(10):3034. [DOI: 10.3390/nu12103034]PMC760070733022957 doi:10.3390/nu12103034 PubMed
102. MatlouSM, IslesCG, HiggsA, MilneFJ, MurrayGD, SchultzE, et al. Potassium supplementation in blacks with mild to moderate essential hypertension. Journal of Hypertension1986;4(1):61-4. 10.1097/00004872-198602000-000103514747 doi:10.1097/00004872-198602000-00010 PubMed
103. MuL, LiC, LiuT, XieW, LiG, WangM, et al. A pilot study on efficacy and safety of a new salt substitute with very low sodium among hypertension patients on regular treatment. Medicine 2020;99(8):e19263. 10.1097/MD.0000000000019263PMC703469932080135 doi:10.1097/MD.0000000000019263 PubMed
104. NakanoM, EguchiK, SatoT, OnoguchiA, HoshideS, KarioK. Effect of intensive salt-restriction education on clinic, home, and ambulatory blood pressure levels in treated hypertensive patients during a 3-month education period. Journal of Clinical Hypertension2016;18(5):385-92. 10.1111/jch.12770PMC803148926732187 doi:10.1111/jch.12770 PubMed
105. NCT02105727. Changing population salt consumption in Lithgow, Australia [A before after comparison of the effectiveness of a community-based salt reduction program done in Lithgow, Australia]. clinicaltrials.gov/ct2/show/NCT02105727 (first received 7 April 2014).
106. PietinenP, RuotsalainenP, TanskanenA, PuskaP. Sodium intake reduction in volunteer families by using a salt substitute and nutrition counselling. Annals of Nutrition & Metabolism1981;25(6):371-80. 10.1159/0001765187332315 doi:10.1159/000176518 PubMed
107. RobareJF, MilasNC, BaylesCM, WilliamsK, NewmanAB, LovalekarMT, et al. The key to life nutrition program: results from a community-based dietary sodium reduction trial. Public Health Nutrition2010;13(5):606-14. 10.1017/S1368980009991583PMC488633119781124 doi:10.1017/S1368980009991583 PubMed
108. SalvettiA, BichisaoE, CaiazzaA, BartolomeiG, CagianelliMA, FederighiG, et al. The combination of a low-Na/high-K salt with metoprolol in the treatment of mild-moderate hypertension. A multicenter study. American Journal of Hypertension1988;1(3 Pt 3):201s-5s. 10.1093/ajh/1.3.201s3046626 doi:10.1093/ajh/1.3.201s PubMed
109. SciarroneSE, BeilinLJ, RouseIL, RogersPB. A factorial study of salt restriction and a low-fat/high-fibre diet in hypertensive subjects. Journal of Hypertension1992;10(3):287-98. 10.1097/00004872-199203000-000131315827 doi:10.1097/00004872-199203000-00013 PubMed
110. ZoccaliC, CummingAM, HutchesonMJ, BarnettP, SemplePF. Effects of potassium on sodium balance, renin, noradrenaline and arterial pressure. Journal of Hypertension1985;3(1):67-72. 10.1097/00004872-198502000-000113889148 doi:10.1097/00004872-198502000-00011 PubMed
111. JonesD. Dietary sodium and potassium intake: an event-based randomized clinical trial. Journal of Hypertension 2019;37(Supplement 1):e252.
112. NeutelJ. Replacing regular salt with sodium-reduced, potassium- and magnesium-enriched mineral salt may offer non-pharmacological approach to lowering blood pressure. American Journal of Hypertension1996;9:94A.
113. VoloshynaI, VizirVA, VoloshynMA, KrivenkoVI, PonomarenkoVI. Effect of potassium/magnesium enriched salt on functional capacity and quality of life in symptomatic patients with chronic heart failure. European Journal of Heart Failure2017;19(Supplement 1):53.
114. ACTRN12619000352101. Salt ALTernatives Study (SALTS): a smartphone app and dietary alternative salt to lower blood pressure for adults with high blood pressure. trialsearch.who.int/?TrialID=ACTRN12619000352101 (first received 06 March 2019).
115. ChiCTR-TRC-08000296. China Salt trial. www.chictr.org.cn/showproj.aspx?proj=9234 (first received 30 December 2008).
116. ChiCTR-TRC-09000538. The study on pathogeny of hypertension and the salt intervention on diet in the population of Zhangwu County. www.chictr.org.cn/showproj.aspx?proj=8997 (first received 11 September 2009).
117. ChiCTR2000029017. Study on the application of low-sodium formula salt in hypertensive patients with diabetes. www.chictr.org.cn/showproj.aspx?proj=47080 (first received 01 November 2020).
118. IRCT2016103130572N1. The effects of a salt substitute on an Iranian population. trialsearch.who.int/?TrialID=IRCT2016103130572N1 (first received 08 December 2016).
119. NCT02016404. The effect of a low sodium-high potassium salt on blood pressure in Vietnamese adults. clinicaltrials.gov/ct2/show/NCT02016404?term=NCT+02016404&draw=2&rank=1 (first received 20 December 2013).
120. NCT02021435. Tibet salt reduction study. clinicaltrials.gov/ct2/show/NCT02021435?term=NCT02021435&draw=2&rank=1 (first received 27 December 2013).
121. JinA, LiuK, LabartheDR, FengX, ZhangR, WangH, et al. Impact of salt substitute and stepwise reduction of salt supply on blood pressure in residents in senior residential facilities: design and rationale of the DECIDE-Salt trial. American Heart Journal2020;226:198-205. [DOI: 10.1016/j.ahj.2020.05.013]32615357 doi:10.1016/j.ahj.2020.05.013 PubMed
122. NCT03290716. Diet, exercise and cardiovascular health - effect of salt substitute and stepwise salt supply control in reducing blood pressure in the elderly in nursing homes in China. clinicaltrials.gov/ct2/show/NCT03290716?term=NCT03290716&draw=2&rank=1 (first received 25 September 2017).
123. AheeP, CroweAV. The management of hyperkalaemia in the emergency department. Journal of Accident & Emergency Medicine2000;17(3):188-91. [DOI: 10.1136/emj.17.3.188]PMC172536610819381 doi:10.1136/emj.17.3.188 PubMed
124. ArnettDK, BlumenthalRS, AlbertMA, BurokerAB, GoldbergerZD, HahnEJ, et al. 2019 ACC/AHA guideline on the primary prevention of cardiovascular disease: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Journal of the American College of Cardiology2019;74(10):e177-232. [DOI: 10.1016/j.jacc.2019.03.010]PMC768556530894318 doi:10.1016/j.jacc.2019.03.010 PubMed
125. BanegasJR, RuilopeLM, De la SierraA, VinyolesE, GorostidiM, De la CruzJJ, et al. Clinic versus daytime ambulatory blood pressure difference in hypertensive patients: the impact of age and clinic blood pressure. Hypertension2017;69(2):211-9. [DOI: 10.1161/HYPERTENSIONAHA.116.08567]28028191 doi:10.1161/HYPERTENSIONAHA.116.08567 PubMed
126. BilanoVL, OtaE, GanchimegT, MoriR, SouzaJP. Risk factors of pre-eclampsia/eclampsia and its adverse outcomes in low- and middle-income countries: a WHO secondary analysis. PLOS One2014;9(3):e91198. [DOI: 10.1371/journal.pone.0091198]PMC396237624657964 doi:10.1371/journal.pone.0091198 PubMed
127. Boston University School of Public Health. Comparing frequencies: rate ratios. sphweb.bumc.bu.edu/otlt/MPH-Modules/PH717-QuantCore/PH717_ComparingFrequencies/PH717_ComparingFrequencies9.html (accessed 8 October 2021).
128. BurkardT, MayrM, WinterhalderC, LeonardiL, EcksteinJ, VischerAS. Reliability of single office blood pressure measurements. Heart2018;104(14):1173-9. [DOI: 10.1136/heartjnl-2017-312523]29530989 doi:10.1136/heartjnl-2017-312523 PubMed
129. CampbellM, McKenzieJE, SowdenA, KatikireddiSV, BrennanSE, EllisS, et al. Synthesis without meta-analysis (SWiM) in systematic reviews: reporting guideline. BMJ2020;368:l6890. [DOI: https://doi.org/10.1136/bmj.l6890]10.1136/bmj.l6890PMC719026631948937 doi:10.1136/bmj.l6890 PubMed
130. CepanecK, VugrinecS, CvetkovićT, RanilovićJ. Potassium chloride-based salt substitutes: a critical review with a focus on the patent literature. Comprehensive Reviews in Food Science and Food Safety2017;16(5):881-94. [DOI: https://doi.org/10.1111/1541-4337.12291]10.1111/1541-4337.1229133371617 doi:10.1111/1541-4337.12291 PubMed
131. CnossenJS, VollebregtKC, De VriezeN, Ter RietG, MolBW, FranxA, et al. Accuracy of mean arterial pressure and blood pressure measurements in predicting pre-eclampsia: systematic review and meta-analysis. BMJ2008;336(7653):1117-20. [DOI: 10.1136/bmj.39540.522049.BE]PMC238662718480117 doi:10.1136/bmj.39540.522049.BE PubMed
132. CohnJN, KoweyPR, WheltonPK, PrisantLM. New guidelines for potassium replacement in clinical practice: a contemporary review by the National Council on Potassium in Clinical Practice. Archives of Internal Medicine2000;160(16):2429-36. [DOI: 10.1001/archinte.160.16.2429]10979053 doi:10.1001/archinte.160.16.2429 PubMed
133. CouserWG, RemuzziG, MendisS, TonelliM. The contribution of chronic kidney disease to the global burden of major noncommunicable diseases. Kidney International2011;80(12):1258-70. [DOI: 10.1038/ki.2011.368]21993585 doi:10.1038/ki.2011.368 PubMed
134. Covidence. Version accessed prior to 4 November 2021. Melbourne, Australia: Veritas Health Innovation, 2017. Available at covidence.org.
135. CuijpersP, GriffinJW, FurukawaTA. The lack of statistical power of subgroup analyses in meta-analyses: a cautionary note. Epidemiology and Psychiatric Sciences2021;30:e78. [DOI: 10.1017/S2045796021000664]PMC867983234852862 doi:10.1017/S2045796021000664 PubMed
136. DeeksJJ, HigginsJP, AltmanDG. Chapter 10: Analysing data and undertaking meta-analyses. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 (updated February 2022). Cochrane, 2022. Available from training.cochrane.org/handbook .
137. De MoraesAC, LacerdaMB, MorenoLA, HortaBL, CarvalhoHB. Prevalence of high blood pressure in 122,053 adolescents: a systematic review and meta-regression. Medicine2014;93(27):e232. [DOI: 10.1097/MD.0000000000000232]PMC460280525501086 doi:10.1097/MD.0000000000000232 PubMed
138. EttehadD, EmdinCA, KiranA, AndersonSG, CallenderT, EmbersonJ, et al. Blood pressure lowering for prevention of cardiovascular disease and death: a systematic review and meta-analysis. Lancet2016;387(10022):957-67. [DOI: 10.1016/S0140-6736(15)01225-8]26724178 doi:10.1016/S0140-6736(15)01225-8 PubMed
139. FilippatosTD, MakriA, ElisafMS, LiamisG. Hyponatremia in the elderly: challenges and solutions. Clinical Interventions in Aging2017;12:1957-65. [DOI: 10.2147/CIA.S138535]PMC569419829180859 doi:10.2147/CIA.S138535 PubMed
140. FilippiniT, VioliF, D'AmicoR, VincetiM. The effect of potassium supplementation on blood pressure in hypertensive subjects: a systematic review and meta-analysis. International Journal of Cardiology2017;230:127-35. [DOI: 10.1016/j.ijcard.2016.12.048.]28024910 doi:10.1016/j.ijcard.2016.12.048. PubMed
141. ForouzanfarMH, LiuP, RothGA, NgM, BiryukovS, MarczakL, et al. Global burden of hypertension and systolic blood pressure of at least 110 to 115 mm Hg, 1990-2015. JAMA2017;317(2):165-82. [DOI: 10.1001/jama.2016.19043]28097354 doi:10.1001/jama.2016.19043 PubMed
142. FranklinSS, LevyD. Aging, blood pressure, and heart failure: what are the connections?Hypertension2011;58(5):760-2. [DOI: 10.1161/HYPERTENSIONAHA.111.179119]PMC471126021947472 doi:10.1161/HYPERTENSIONAHA.111.179119 PubMed
143. GilesTD, MatersonBJ, CohnJN, KostisJB. Definition and classification of hypertension: an update. Journal of Clinical Hypertension2009;11(11):611-4. [DOI: 10.1111/j.1751-7176.2009.00179.x.]PMC867328619878368 doi:10.1111/j.1751-7176.2009.00179.x. PubMed
144. GeraldJ, Dorothy R Friedman School of Nutrition Science and Policy at Tufts University. Global Dietary Database 2018 Final Estimates. www.globaldietarydatabase.org/data-download (accessed 13 March 2022).
145. GolderS, McIntoshHM, DuffyS, GlanvilleJ, Centre for Reviews and Dissemination and UK Cochrane Centre Search Filters Design Group. Developing efficient search strategies to identify reports of adverse effects in MEDLINE and EMBASE. Health Information and Libraries Journal2006;23(1):3-12. [DOI: 10.1111/j.1471-1842.2006.00634.x.]16466494 doi:10.1111/j.1471-1842.2006.00634.x. PubMed
146. GolderS, LokeYK. The performance of adverse effects search filters in MEDLINE and EMBASE. Health Information and Libraries Journal2012;29(2):141-51. [DOI: 10.1111/j.1471-1842.2012.00980.x]22630362 doi:10.1111/j.1471-1842.2012.00980.x PubMed
147. GRADEpro GDT. Version accessed prior to 4 November 2021. Hamilton (ON): McMaster University (developed by Evidence Prime), 2021. Available at gradepro.org.
148. GreerRC, MarklundM, AndersonCA, CobbLK, DalcinAT, HenryM, et al . Potassium-enriched salt substitutes as a means to lower blood pressure: benefits and risks. Hypertension2020;75(2):266-74. [DOI: 10.1161/HYPERTENSIONAHA.119.13241]31838902 doi:10.1161/HYPERTENSIONAHA.119.13241 PubMed
149. Guang WangJ, StaessenJA, FranklinSS, FagardR, GueyffierF. Systolic and diastolic blood pressure lowering as determinants of cardiovascular outcome. Hypertension2005;45(5):907-13. [DOI: 10.1161/01.HYP.0000165020.14745.79.]15837826 doi:10.1161/01.HYP.0000165020.14745.79. PubMed
150. GuyattGH, OxmanAD, KunzR, BrozekJ, Alonso-CoelloP, RindD, et al. GRADE guidelines 6. Rating the quality of evidence - imprecision. Journal of Clinical Epidemiology2011;64(12):1283-93. [DOI: 10.1016/j.jclinepi.2011.01.012]21839614 doi:10.1016/j.jclinepi.2011.01.012 PubMed
151. HernandezAV, EmondsEE, ChenBA, Zavala-LoayzaAJ, ThotaP, PasupuletiV, et al. Effect of low-sodium salt substitutes on blood pressure, detected hypertension, stroke and mortality. Heart2019;105(12):953-60. [DOI: 10.1136/heartjnl-2018-314036.]30661034 doi:10.1136/heartjnl-2018-314036. PubMed
152. HerttuaK, TabákAG, MartikainenP, VahteraJ, KivimäkiM. Adherence to antihypertensive therapy prior to the first presentation of stroke in hypertensive adults: population-based study. European Heart Journal2013;34(38):2933-9. [DOI: 10.1093/eurheartj/eht219]PMC379139323861328 doi:10.1093/eurheartj/eht219 PubMed
153. HigginsJP, GreenS, editor(s) . Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.1 (updated March 2011). The Cochrane Collaboration, 2011. Available from training.cochrane.org/handbook/archive/v5.1/ .
154. HigginsJP, AltmanDG, SterneJA. Chapter 8: Assessing risk of bias in included studies. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.2 (updated February 2021). Cochrane, 2021. Available from training.cochrane.org/handbook .
155. HigginsJP, LiT, DeeksJJ. Chapter 6: Choosing effect measures and computing estimates of effect. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.2 (updated February 2021). Cochrane, 2021. Available from training.cochrane.org/handbook .
156. HigginsJP, EldridgeS, LiT. Chapter 23: Including variants on randomized trials. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.2 (updated February 2021). Cochrane, 2021. Available from training.cochrane.org/handbook .
157. Hollander-RodriguezJC, Calvert JrJF. Hyperkalemia. American Family Physician2006;73(2):283-90. [PMID: https://pubmed.ncbi.nlm.nih.gov/16445274/]16445274 PubMed
158. HuJH, ZhaoLC, LiX, WuYF. Effects of salt substitution on blood pressure using home measurements in essential hypertensive patients: a double-blinded, randomized controlled trial. Chinese Journal of Hypertension2014;22(1):42-6.
159. HuangL, TrieuK, YoshimuraS, NealB, WoodwardM, CampbellNR, et al. Effect of dose and duration of reduction in dietary sodium on blood pressure levels: systematic review and meta-analysis of randomised trials. BMJ2020;368:m315. [DOI: https://doi.org/10.1136/bmj.m315]10.1136/bmj.m315PMC719003932094151 doi:10.1136/bmj.m315 PubMed
160. HultcrantzM, RindD, AklEA, TreweekS, MustafaRA, IorioA, et al. The GRADE Working Group clarifies the construct of certainty of evidence. Journal of Clinical Epidemiology2017;87:4-13. [DOI: 10.1016/j.jclinepi.2017.05.006.]PMC654266428529184 doi:10.1016/j.jclinepi.2017.05.006. PubMed
161. Institute for Health Metrics and Evaluation . The Global Burden of Disease Study 2016: Global prevalence, incidence, and YLDs for 2016, percentage change of YLD counts, and percentage change of age-standardised YLD rates between 2006 and 2016 for all causes and nine impairments. ghdx.healthdata.org/record/ihme-data/gbd-2016-incidence-prevalence-and-ylds-1990-2016 (accessed 4 November 2021).
162. JafarnejadS, MirzaeiH, ClarkCC, TaghizadehM, EbrahimzadehA. The hypotensive effect of salt substitutes in stage 2 hypertension: a systematic review and meta-analysis. BMC Cardiovascular Disorders2020;20(1):98. [DOI: 10.1186/s12872-020-01347-x.]PMC704742032106813 doi:10.1186/s12872-020-01347-x. PubMed
163. JinA, XieW, WuY. Effect of salt reduction interventions in lowering blood pressure in Chinese populations: a systematic review and meta-analysis of randomised controlled trials. BMJ Open2020;10(2):e032941. [DOI: 10.1136/bmjopen-2019-032941.]PMC704485832071177 doi:10.1136/bmjopen-2019-032941. PubMed
164. KovesdyCP, MatsushitaK, SangY, BrunskillNJ, CarreroJJ, ChodickG, et al. Serum potassium and adverse outcomes across the range of kidney function: a CKD Prognosis Consortium meta-analysis. European Heart Journal2018;39(17):1535-42. [DOI: 10.1093/eurheartj/ehy100.]PMC593024929554312 doi:10.1093/eurheartj/ehy100. PubMed
165. Lee H-Y, Oh B-H. Aging and arterial stiffness. Circulation Journal2010;74(11):2257-62. [DOI: 10.1253/circj.cj-10-0910]20962429 doi:10.1253/circj.cj-10-0910 PubMed
166. LefebvreC, GlanvilleJ, BriscoeS, FeatherstoneR, LittlewoodA, MarshallC, et al. Technical Supplement to Chapter 4: Searching for and selecting studies. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 (updated February 2022). Cochrane, 2022. Available from training.cochrane.org/handbook.
167. LeveyAS, LevinA, KellumJA. Definition and classification of kidney diseases. American Journal of Kidney Diseases2013;61(5):686-8. [DOI: https://doi.org/10.1053/j.ajkd.2013.03.003]10.1053/j.ajkd.2013.03.00323582249 doi:10.1053/j.ajkd.2013.03.003 PubMed
168. LewingtonS, ClarkeR, QizilbashN, PetoR, CollinsR, Prospective Studies Collaboration. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet2002;360(9349):1903-13. [DOI: 10.1016/s0140-6736(02)11911-8]12493255 doi:10.1016/s0140-6736(02)11911-8 PubMed
169. LiL, SmithHE, AtunR, Tudor CarL. Search strategies to identify observational studies in MEDLINE and Embase. Cochrane Database of Systematic Reviews2019;3(3):MR000041. [DOI: 10.1002/14651858.MR000041.pub2]PMC810356630860595 doi:10.1002/14651858.MR000041.pub2 PubMed
170. LiuZQ, MuJJ, LiuWM, LiangYM, YangDY, ZhuDJ, et al. Reduction of blood pressure with calcium and potassium supplementation in children: a randomized double-blind placebo controlled trial of two years. Chinese Circulation Journal1996;11:534-6.10.1038/sj.jhh.100185415759021 doi:10.1038/sj.jhh.1001854 PubMed
171. MirmiranP, BahadoranZ, GaeiniZ. Common limitations and challenges of dietary clinical trials for translation into clinical practices. International Journal of Endocrinology and Metabolism2021;19(3):e108170. [DOI: 10.5812/ijem.108170]PMC845365134567133 doi:10.5812/ijem.108170 PubMed
172. NaudeCE, VisserME, NguyenKA, DurãoS, SchooneesA. Effects of total fat intake on bodyweight in children. Cochrane Database of Systematic Reviews2018;2(2):CD012960. [DOI: 10.1002/14651858.CD012960.]PMC649133329446437 doi:10.1002/14651858.CD012960. PubMed
173. NCD Risk Factor Collaboration (NCD-RisC) . Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants. Lancet2021;398(10304):957-80. [DOI: 10.1016/S0140-6736(21)01330-1]PMC844693834450083 doi:10.1016/S0140-6736(21)01330-1 PubMed
174. NealB, TianM, LiN, ElliottP, YanLL, LabartheDR, et al. Rationale, design, and baseline characteristics of the Salt Substitute and Stroke Study (SSaSS) - a large-scale cluster randomized controlled trial. American Heart Journal2017;188:109-17. [DOI: 10.1016/j.ahj.2017.02.033.]28577665 doi:10.1016/j.ahj.2017.02.033. PubMed
175. NoubiapJJ, EssoumaM, BignaJJ, JingiAM, AmindeLN, NansseuJR. Prevalence of elevated blood pressure in children and adolescents in Africa: a systematic review and meta-analysis. Lancet Public Health2017;2(8):E375-86. [DOI: 10.1016/S2468-2667(17)30123-8]29253478 doi:10.1016/S2468-2667(17)30123-8 PubMed
176. PaleiAC, SpradleyFT, WarringtonJP, GeorgeEM, GrangerJP. Pathophysiology of hypertension in pre-eclampsia: a lesson in integrative physiology. Acta Physiologica2013;208(3):224-33. [DOI: 10.1111/apha.12106]PMC368701223590594 doi:10.1111/apha.12106 PubMed
177. PengYG, LiW, WenXX, LiY, HuJH, ZhaoLC. Effects of salt substitutes on blood pressure: a meta-analysis of randomized controlled trials. American Journal of Clinical Nutrition2014;100(6):1148-54. [DOI: 10.3945/ajcn.114.089235]25411279 doi:10.3945/ajcn.114.089235 PubMed
178. PowlesJ, FahimiS, MichaR, KhatibzadehS, ShiP, EzzatiM, et al. Global, regional and national sodium intakes in 1990 and 2010: a systematic analysis of 24 h urinary sodium excretion and dietary surveys worldwide. BMJ Open2013;3(12):e003733. [DOI: 10.1136/bmjopen-2013-003733.]PMC388459024366578 doi:10.1136/bmjopen-2013-003733. PubMed
179. RahimiK, BidelZ, NazarzadehM, CoplandE, CanoyD, RamakrishnanR, et al, Blood Pressure Lowering Treatment Trialists' Collaboration. Pharmacological blood pressure lowering for primary and secondary prevention of cardiovascular disease across different levels of blood pressure: an individual participant-level data meta-analysis. Lancet2021;397(10285):1625-36. [DOI: 10.1016/S0140-6736(21)00590-0.]PMC810246733933205 doi:10.1016/S0140-6736(21)00590-0. PubMed
180. Review Manager Web (RevMan Web). Version 3.12.1. The Cochrane Collaboration, 2021. Available at revman.cochrane.org.
181. SahayM, SahayR. Hyponatremia: a practical approach. Indian Journal of Endocrinology and Metabolism2014;18(6):760-71. [DOI: 10.4103/2230-8210.141320]PMC419297925364669 doi:10.4103/2230-8210.141320 PubMed
182. SantessoN, GlentonC, DahmP, GarnerP, AklEA, AlperB, et al. GRADE guidelines 26: informative statements to communicate the findings of systematic reviews of interventions. Journal of Clinical Epidemiology2020;119:126-35. [DOI: 10.1016/j.jclinepi.2019.10.014]31711912 doi:10.1016/j.jclinepi.2019.10.014 PubMed
183. SchünemannHJ, HigginsJP, VistGE, GlasziouP, AklEA, SkoetzN, et al. Chapter 14: Completing ‘Summary of findings’ tables and grading the certainty of the evidence. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.2 (updated February 2021). Cochrane, 2021. Available from training.cochrane.org/handbook .
184. ShahbabuB, DasguptaA, SarkarK, SahooSK. Which is more accurate in measuring the blood pressure? A digital or an aneroid sphygmomanometer. Journal of Clinical and Diagnostic Research2016;10(3):LC11-4. [DOI: 10.7860/JCDR/2016/14351.7458]PMC484328827134902 doi:10.7860/JCDR/2016/14351.7458 PubMed
185. StaessenJA, GasowskiJ, WangJG, ThijsL, Den HondE, BoisselJP, et al. Risks of untreated and treated isolated systolic hypertension in the elderly: meta-analysis of outcome trials. Lancet2000;355(9207):865-72. [DOI: 10.1016/s0140-6736(99)07330-4]10752701 doi:10.1016/s0140-6736(99)07330-4 PubMed
186. SterneJA, EggerM, MoherD, BoutronI. Chapter 10: Addressing reporting biases. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.2 (updated February 2021). Cochrane, 2021. Available from training.cochrane.org/handbook.
187. ThangaratinamS, LangenveldJ, MolBW, KhanKS. Prediction and primary prevention of pre-eclampsia. Best Practice & Research: Clinical Obstetrics & Gynaecology2011;25(4):419-33. [DOI: 10.1016/j.bpobgyn.2011.02.008]21454131 doi:10.1016/j.bpobgyn.2011.02.008 PubMed
188. ThomopoulosC, ParatiG, ZanchettiA. Effects of blood-pressure-lowering treatment on outcome incidence. 12. Effects in individuals with high-normal and normal blood pressure: overview and meta-analyses of randomized trials. Journal of Hypertension2017;35(11):2150-60. [DOI: 10.1097/HJH.0000000000001547]28858984 doi:10.1097/HJH.0000000000001547 PubMed
189. TsirimiagkouC, KaratziK, BasdekiED, ArgyrisAA, PapaioannouTG, YannakouliaM. Dietary sodium estimation methods: accuracy and limitations of old and new methods in individuals at high cardiovascular risk. Public Health Nutrition2022;25(4):866-78. [DOI: 10.1017/S1368980021004390]PMC999176434693901 doi:10.1017/S1368980021004390 PubMed
190. UpadhyayA, JaberBL, MadiasNE. Epidemiology of hyponatremia. Seminars in Nephrology2009;29(3):227-38. [DOI: 10.1016/j.semnephrol.2009.03.004]19523571 doi:10.1016/j.semnephrol.2009.03.004 PubMed
191. VerbeekJ, HovingJ, BoschmanJ, Chong L-Y, Livingstone-BanksJ, BeroL. Systematic reviews should consider effects from both the population and the individual perspective. American Journal of Public Health2021;111(5):820-5. [DOI: 10.2105/AJPH.2020.306147.]PMC803400033826374 doi:10.2105/AJPH.2020.306147. PubMed
192. VincetiM, FilippiniT, CrippaA, De SesmaisonsA, WiseLA, OrsiniN. Meta-analysis of potassium intake and the risk of stroke. Journal of the American Heart Association2016;5(10):e004210. [DOI: 10.1161/JAHA.116.004210]PMC512151627792643 doi:10.1161/JAHA.116.004210 PubMed
193. WangZ, ChenZ, ZhangL, WangX, HaoG, ZhangZ, et al. Status of hypertension in China: results from the China Hypertension Survey, 2012–2015. Circulation2018;137:2344-56. [DOI: 10.1161/CIRCULATIONAHA.117.032380]29449338 doi:10.1161/CIRCULATIONAHA.117.032380 PubMed
194. WangL, SongL, LiuB, ZhangL, WuM, CaoZ, et al. Trends and status of the prevalence of elevated blood pressure in children and adolescents in China: a systematic review and meta-analysis. Current Hypertension Reports2019;21(11):88. [DOI: 10.1007/s11906-019-0992-1]31599364 doi:10.1007/s11906-019-0992-1 PubMed
195. World Health Organization. The Asia-Pacific perspective: redefining obesity and its treatment; February 2000. apps.who.int/iris/handle/10665/206936 (accessed 4 November 2021).
196. World Health Organization. Guideline: Sodium intake for adults and children; December 2012. www.who.int/publications/i/item/9789241504836 (accessed 4 November 2021).
197. World Health Organization. Guideline: Potassium intake for adults and children; December 2012. www.who.int/publications/i/item/9789241504829 (accessed 4 November 2021).
198. World Health Organization. Global health estimates 2019: deaths by cause, age, sex, by country and by region, 2000-2019. www.who.int/data/gho/data/themes/mortality-and-global-health-estimates (accessed 4 November 2021).
199. YanLL, FangW, DelongE, NealB, PetersonED, HuangY, et al. Population impact of a high cardiovascular risk management program delivered by village doctors in rural China: design and rationale of a large, cluster-randomized controlled trial. BMC Public Health2014;14:345. [DOI: 10.1186/1471-2458-14-345]PMC398643824721435 doi:10.1186/1471-2458-14-345 PubMed
200. YaxleyJP, ThambarSV. Resistant hypertension: an approach to management in primary care. Journal of Family Medicine and Primary Care2015;4(2):193-9. [DOI: 10.4103/2249-4863.154630]PMC440870025949966 doi:10.4103/2249-4863.154630 PubMed
201. YoungeJO, Kouwenhoven-PasmooijTA, Freak-PoliR, Roos-HesselinkJW, HuninkMM. Randomized study designs for lifestyle interventions: a tutorial. International Journal of Epidemiology2015;44(6):2006-19. [DOI: 10.1093/ije/dyv183.]26377509 doi:10.1093/ije/dyv183. PubMed
202. ZengL, Brignardello-PetersenR, GuyattG. When applying GRADE, how do we decide the target of certainty of evidence rating?Evidence-Based Mental Health2021;24:121-3. [DOI: 10.1136/ebmental-2020-300170]PMC831109734127510 doi:10.1136/ebmental-2020-300170 PubMed
203. ZhangY, AklEA, SchünemannHJ. Using systematic reviews in guideline development: the GRADE approach. Research Synthesis Methods2019;10(3):312-29. [DOI: 10.1002/jrsm.1313]30006970 doi:10.1002/jrsm.1313 PubMed