155   The impact of excessive salt intake on human health

Abstract

Intake of salt is a biological imperative, inextricably woven into physiological systems, human societies and global culture. However, excessive salt intake is associated with high blood pressure. As this effect likely drives cardiovascular morbidity and mortality, excessive salt intake is estimated to cause ~5 million deaths per annum worldwide. Animal research has identified various mechanisms by which high salt intake drives disease in the kidney, brain, vasculature and immune system. The potential for therapeutic interventions in many of these pathways has yet to be tested. Salt-reduction interventions lower blood pressure but for most individuals, ‘hidden’ salt in processed foods disconnects salt intake from discretionary control. This problem is compounded by growing inequalities in food systems, which form another hurdle to sustaining individual dietary control of salt intake. The most effective salt-reduction interventions have been implemented at the population level and comprise multi-component approaches, involving government, education and the food industry. 

Introduction 

Excessive intake of salt (NaCl) — defined as an intake that is surplus to physiological requirements — can raise blood pressure.1 Such excessive salt intake 2 is well within the range of typical dietary intake in most modern societies. The mean daily global intake of ~10–15 g NaCl (170–260 mmol)2-4 far exceeds both physiological requirements and the World Health Organization target of <5g NaCl (85 mmol) per day5. Excessive salt intake is associated with high blood pressure in diverse human populations6. The crucial question is whether salt exerts a causal influence on patient-centred health outcomes such as myocardial infarction, stroke and death. This is a controversial, fiercely debated7 area. Some authors argue that there are compelling data to support a causal link between dietary salt intake and cardiovascular disease8,9. Others argue that an observational association between low estimated sodium intake and higher mortality shows that salt reduction may be neutral or even harmful in many individuals10. The extent to which an individual is susceptible to the health effects of dietary salt depends on their salt intake and their salt-sensitivity; that is, the gradient between salt intake and risk of disease.Figure 2 | Salt consumption and the effects of interventions on blood pressure and cardiovascular risk. a | The World Health Organization (WHO)5 and US National Academy of Sciences (NAS)48 recommend an upper limit of daily salt intake for all adults of 5 g NaCl (85 mmol) or 5.85 g NaCl (100 mmol), respectively. Kidney Disease: Improving Global Outcomes (KDIGO)201 and the European Society of Hypertension (ESC)202 have made similar recommendations for patients with hypertension. However, the daily salt intake of many populations worldwide exceed these recommendations with a global average of ~10–15 g NaCl per day (170–260 mmol) 2,3,10,83,187,189,203. In remote isolated populations such as the Amazonian Yanomami, daily salt intake is thought to be similar to that in Paleolithic man at ~1g NaCl (20 mmol)83,203. Following multi-modal public health interventions, mean daily population salt intake reduced by ~40% in Finland187 and ~15% in the UK184. b | Salt-reduction interventions have been shown to lower blood pressure with a linear dose-response relationship in randomized controlled trials (RCTs)1,30,58. In these trials, a mean reduction in sodium intake of 80–130 mmol (5–8 g NaCl) induced a ~4 mmHg decrease in systolic blood pressure1,30. The effect size was smaller for interventions lasting longer than 6 months. c | The NAS meta-analysis of selected RCTs showed that sodium reduction can reduce the risk of cardiovascular disease (relative risk reduction of ~25 %)48. In observational studies, estimated sodium intake has exhibited either a linear35,52 or J-shaped21,37,38,44 association with cardiovascular disease risk. The J-shaped association can likely be explained by methodological artefacts owing to use of spot urine samples, confounding and reverse causality, rather than a genuine increase in cardiovascular risk at low sodium intakes. Both salt intake and salt-sensitivity of blood pressure are associated with increased risk of cardiovascular disease (CVD) and death11,12 Diseases including heart failure and CKD also alter individual sensitivity to the health effects of salt (Box 3). The biological mechanisms whereby salt perturbs normal physiology13-18, whether salt causes CVD7 and effective interventions9,19 to improve health outcomes have been reviewed separately elsewhere. Here, we aim to integrate these disparate areas into a unifying thesis regarding the impact of salt on human health. Given the difficulties in performing definitive long-term RCTs with patient-centred endpoints, we argue that it is necessary to form an opinion based on the total evidence from multiple sources (Figure 2 | Salt consumption and the effects of interventions on blood pressure and cardiovascular risk. a | The World Health Organization (WHO)5 and US National Academy of Sciences (NAS)48 recommend an upper limit of daily salt intake for all adults of 5 g NaCl (85 mmol) or 5.85 g NaCl (100 mmol), respectively. Kidney Disease: Improving Global Outcomes (KDIGO)201 and the European Society of Hypertension (ESC)202 have made similar recommendations for patients with hypertension. However, the daily salt intake of many populations worldwide exceed these recommendations with a global average of ~10–15 g NaCl per day (170–260 mmol) 2,3,10,83,187,189,203. In remote isolated populations such as the Amazonian Yanomami, daily salt intake is thought to be similar to that in Paleolithic man at ~1g NaCl (20 mmol)83,203. Following multi-modal public health interventions, mean daily population salt intake reduced by ~40% in Finland187 and ~15% in the UK184. b | Salt-reduction interventions have been shown to lower blood pressure with a linear dose-response relationship in randomized controlled trials (RCTs)1,30,58. In these trials, a mean reduction in sodium intake of 80–130 mmol (5–8 g NaCl) induced a ~4 mmHg decrease in systolic blood pressure1,30. The effect size was smaller for interventions lasting longer than 6 months. c | The NAS meta-analysis of selected RCTs showed that sodium reduction can reduce the risk of cardiovascular disease (relative risk reduction of ~25 %)48. In observational studies, estimated sodium intake has exhibited either a linear35,52 or J-shaped21,37,38,44 association with cardiovascular disease risk. The J-shaped association can likely be explained by methodological artefacts owing to use of spot urine samples, confounding and reverse causality, rather than a genuine increase in cardiovascular risk at low sodium intakes. ). The effects of salt reduction on blood pressure have been tested in individual-level RCTs; the effects of potassium-rich salt-substitutes on cardiovascular disease have been tested in cluster-level RCTs. In observational studies, statistical adjustment for confounders or Mendelian randomization are used to infer a causal role for salt in cardiovascular disease. Animal models have been used to define the likely molecular mechanisms of salt-induced cardiovascular effects. We begin by reviewing the evidence that salt intake affects patient-centred health outcomes. We then highlight the factors that determine individual susceptibility to dietary salt and discuss how best to intervene to reduce salt intake and improve health at the population and individual levels. 

Salt and blood pressure

Observational data 

The mean global sodium intake is ~170 mmol (4 g Na; 10 g NaCl) per day2. Above this threshold, salt intake is positively associated with blood pressure. This observation has been replicated in large, multinational studies in the general population, notably INTERSALT6, PURE20 and the UK Biobank cohort21. The relationship between salt intake and blood pressure is steeper in patients with hypertension than in those who are normotensive22. In addition, a robust negative association exists between potassium intake and blood pressure (Box 1). Mendelian randomisation analysis has been used to infer a causal role for dietary sodium in setting blood pressure. In the UK Biobank cohort, genetic variants that were associated with urine Na/K ratio exerted a probable causal influence on blood pressure23,24. Similarly, in East Asian populations, genetic variants that associated with sodium intake were also associated with blood pressure 25,26. 

Randomized controlled trial data

 Prior to the development of antihypertensive therapies, the low-salt rice diet was used to successfully treat malignant hypertension, but was not tested in RCTs27.

Subsequent trials in Chimpanzees showed that blood pressure varies in response to interventions in dietary salt28, an observation that has now been replicated in hundreds of human RCTs. Two large meta-analyses that quantified the ‘dose-response’ relationship between sodium intake and blood pressure in adults gave similar results. The first by Huang et al., which included 133 studies with >12,000 participants, included trials of any duration excluding those in pregnant women and patients with chronic kidney disease (CKD) 1. The analysis showed that in studies that exceeded 14 days in duration, systolic blood pressure fell by ~2.1 mmHg (95% CI 0.9–3.4) per 50 mM reduction in Na. The effect size was roughly half as large in studies of shorter duration. The mean reduction in sodium intake in the intervention group was 131 mmol per day (range of change in sodium intake -336–8 mmol) and the mean reduction in systolic blood pressure was 4.3 mmHg (95% confidence interval 3.6–4.9 mmHg) 1. This effect size is broadly comparable to that achieved with antihypertensive medications: a meta-analysis published in preprint form that included 51 RCTs reported that antihypertensive drugs induce a mean blood pressure decrease of ~5 mmHg compared to placebo29. The second meta-analysis by Filippini et al. included 85 trials of >4 weeks’ duration and a total of >10,000 participants30. A linear relationship between urinary sodium excretion and blood pressure was observed over the range 20–330 mmol per day. The gradient for systolic blood pressure reduction was ~2.8 mmHg (2.3–3.3 mmHg) per 50 mM reduction in Na and the median reduction in Na intake in the intervention group was ~80 mmol per day. Most published RCTs are fairly small and of short duration. In the meta-analysis by Huang et al., only 5 studies including 3,252 participants had interventions that lasted more than 6 months1. However, longer-term population-level salt-substitution interventions have achieved sustained reductions in salt intake and blood pressure, as we discuss in detail below.

Salt and cardiovascular disease

Observational data

Large observational studies differ in their conclusions regarding the relationship between sodium intake and CVD (supplemental Table 1). A meta-analysis of highquality prospective cohort studies showed a dose-dependent association between salt intake and CVD, with a ~12% increase in stroke risk for every 100 mmol per day increase in salt consumption34. Similarly, a meta-analysis of six prospective cohort studies in over 10,000 healthy, predominantly White adults demonstrated a broadly linear association between sodium intake – estimated from at least two 24h urine collections – and risk of a composite cardiovascular outcome (stroke, MI, coronary revascularisation)35. However, some large cohort studies, notably the PURE study, have shown a ‘J-shaped curve’ with an inverse association between salt intake, CVD and all-cause mortality when estimated salt intake is less than ~170 mmol (~4 g Na; 10 g NaCl) per day36-40. Although a well-publicised debate exists regarding the explanation for the ‘Jshaped curve’, consensus is growing that much of the effect can be explained by methodological artefacts inherent in the estimation of sodium intake using spot urine samples (Box 2) or epidemiological analyses (confounding and reverse causation)19. The World Hypertension League have been particularly vocal in their criticism of the PURE study, primarily for its reliance on spot urine samples8,41 Even if the ‘J-shaped curve’ is predominantly an artefact, low sodium intake could potentially be harmful in certain individuals (Box 3). Some have argued that sodium restriction could activate the renin-angiotensin-aldosterone system (RAAS) and sympathetic nervous systems with deleterious consequences, but no empirical support for this hypothesis exists42. Controversy also exists regarding the upper end of sodium excretion. Most large cohort studies have reported a positive association between estimated sodium intake and CVD for intakes >4–6 g per day (170–260 mmol; 10–15 g NaCl) 22,37,40,43. However, three analyses of data from the UK Biobank cohort did not replicate this observation 21,24,44. Whether this discrepancy arises from differences in the study populations or is a methodological artefact is unclear (supplemental Table 1). The most parsimonious explanation is that use of spot urine samples, which underestimate sodium intake at higher levels, flattens any association between salt intake and CVD in this range. Consistent with this effect being a methodological artefact, an analysis of UK Biobank participant data found no significant association between estimated sodium intake and CVD in the primary analysis, but a significant positive association in a model that did not adjust for BMI.

Randomized controlled trial data

Several individually-randomized controlled trials have attempted to determine the effects of salt reduction on CVD and mortality. However, none has had sufficient sample size and duration to give a definitive answer and meta-analyses of these RCTs have reached different conclusions. A Cochrane review concluded that only weak evidence suggests that dietary salt reduction reduces the incidence of CVD45,46, whereas a contemporaneous analysis by different researchers concluded that dietary salt reduction (~34 mmol (2g) NaCl per day) causes a meaningful reduction in CVD risk (RR 0.80, 95% CI 0.64–0.99)47. A US National Academy of Sciences (NAS) report concluded that moderate evidence suggests that dietary sodium reduction (from ~150 to 100 mmol (~3.5 to 2.3 g) NaCl per day) prevents CVD (RR 0.74, 95% CI 0.58–0.93)48, whereas an associated systematic review concluded that confidence in this observation was low49. These discrepancies arise because meta-analyses are highly sensitive to the study inclusion criteria. This issue was exemplified by a meta-analysis in which the researchers reversed their initial opinion that salt restriction reduces all-cause and cardiovascular mortality50, and in light of additional sensitivity analyses, ultimately concluded that no strong evidence of any effect exists51. The conclusion of the NAS report was largely based on a meta-analysis of trials lasting for >12 months and excluding interventions based on salt substitutes; only three trials met these criteria: TOHP-1, TOHP-2 and TONE52,53. These trials are potentially informative; they are methodologically sound with a low risk of bias, they were conducted in populations with moderate CVD risk (middle-aged prehypertensive adults in TOHP and elderly patients with hypertension in TONE) and they used serial 24h urine collections to accurately define the effect of the low sodium intervention. They also studied the effect of sodium reduction within a clinically relevant range (decrease of ~40 mmol from a baseline of 150–180 mmol per day). However, confidence in the findings of the NAS meta-analysis is limited by the low number of study participants (3,087, of whom 282 experienced a cardiovascular event). Furthermore, as the included trials were conducted in the late 80s and early 90s, when definitions of cardiovascular events and the approach to CVD prevention differed substantially from the current standards, whether the results are generalizable to a contemporary population is uncertain. Despite the controversy, two broad points of consensus exist. The first is that the uncertainty could be resolved by better quality data, that is, large RCTs with long enough follow-up to assess robust cardiovascular outcomes. The second is that the barriers to performing such trials are so substantial that they will be rarely, if ever, surmounted. The number of participants required to test the effect of sodium reduction on CVD in an individually randomized controlled trial was estimated to be between 17,000 and 37,000, depending on baseline cardiovascular risk54. The estimated cost of such a trial was USD$400–900 million54. Cluster-randomized trials enable greater cost-efficiency at the expense of reduced statistical power and precision. This approach has been used to test the effect of potassium-rich salt substitutes in high-risk populations31,32,55. The largest such study to date, the SSaSS RCT, included ~21,000 individuals with a history of stroke or aged >60 years with hypertension in rural China31. In this study, the intervention, substitution of table salt with 75% NaCl/25% KCl, was randomly assigned to entire villages. This substitution was associated with reductions in the incidence of stroke (~14%), major cardiovascular events (~13%) and all-cause mortality (~12%), without any detectable increase in hyperkalaemia. These effects were broadly consistent with the observed drop in systolic blood pressure of ~3.3 mmHg. The pragmatic design of SSaSS renders it subject to criticisms of internal and external validity. The control group did not receive a placebo intervention and the unusual table salt in the treatment group might have served as a daily reminder to participants that they were at high cardiovascular risk, encouraging other healthy behaviours (a Hawthorne effect). Moreover, the beneficial effects of sodium reduction could not be differentiated from those of potassium supplementation. This concept is broadly relevant, particularly when renal potassium secretion is compromised (Box 1). As patients with kidney disease or who were taking potassium-sparing diuretics were excluded from SSaSS, whether the substitution intervention is safe in these populations is unknown. Nevertheless, in our view, SSaSS provides high confidence that salt reduction reduces cardiovascular risk. These cluster-randomized controlled trials31,32,55 prove that intervention at a population level is feasible and cost-effective in populations with discretionary control of dietary salt (that is, when salt is mainly added to food in the home). Beyond this conclusion, one can either take the view that judgement should be reserved until the ideal RCT data are available or can argue – as we do – that the existing evidence from multiple sources should be used to estimate the likely consequences of sodium reduction on CVD in a population. To this end, investigators have used modelling to predict the effect of sodium reduction on CVD from the effect on blood pressure and/or make inferences from observational data.

Modelling data

In RCTs of pharmacological interventions to lower blood pressure, a 5 mmHg reduction in blood pressure confers a ~10% relative risk reduction in major adverse cardiovascular events and a ~6% relative risk reduction in all-cause mortality56,57. Thus, the blood pressure reductions that can be achieved using dietary sodium restriction would be expected to confer a clinically meaningful reduction in the risk of stroke, heart failure and death. Models constructed using this reasoning predict that sodium reduction strategies could prevent millions of premature deaths worldwide. The global excess of deaths attributed to excessive sodium intake is estimated at 1– 5 million per year3,58 and sodium reduction strategies could save over 5 million disability-adjusted life-years per year59,60. These models rely primarily on blood pressure as a surrogate or intermediate end point on the pathway from salt intake to CVD. In general, surrogate end points must be used with caution61; however, blood pressure is recognised as a validated intermediate end point by the US FDA and the EMA because of its incontrovertible role in the causation of stroke, myocardial infarction and heart failure56,62. Indeed, given the difficulties in performing a RCT with robust cardiovascular event end points, use of blood pressure as a validated intermediate end point has been essential in obtaining data on the effect of salt reduction on cardiovascular health. However, the available models are predicated on some partially unverified assumptions. First, that dietary sodium reduction does not activate deleterious pathways that would oppose any beneficial effects of blood pressure lowering. Second, that changes in blood pressure are sustained in the long term. Furthermore, how effective sodium reduction strategies are at the individual and population levels is unknown.

Salt and other diseases

High sodium intake is associated with stomach cancer63, obesity, metabolic syndrome64, autoimmunity65,66, kidney stones and osteoporosis9. These associations are supported by plausible mechanisms in cell and animal models and by epidemiological data. Salt intake is likely associated with kidney stones and osteoporosis through a common mechanism. In animal models and humans, urinary excretion of sodium and calcium are strongly correlated67-69. The majority of renal calcium reabsorption occurs passively in the proximal tubules70, where it is coupled to sodium and water reabsorption through the effects of water reabsorption on the translumenal calcium gradient and by solvent drag69. Salt loading also reduces the expression of claudin 2, an important calcium channel, in the rat proximal tubule71. Therefore sodium loading / volume expansion will reduce proximal tubular fluxes and decrease net calcium reabsorption; sodium restriction / volume depletion will do the opposite. This a priori reasoning has been used for many years to support a role for sodium restriction in prevention of kidney stones and is supported by one small RCT72. For stomach cancer and autoimmune disease, epidemiological and animal data support intriguing mechanistic hypotheses. For example, it has been suggested that high salt intake increases the risk of stomach cancer because it promotes infection with Helicobacter pylori73,74 and that salt intake promotes autoimmune disease by perturbing TH17-lymphocyte function65. Nevertheless, in the absence of RCTs, we must be guarded about ascribing a meaningful causative role for high dietary salt intake in these diseases. 

Determinants of salt intake

The determinants of salt intake are myriad and include cultural, socioeconomic and biological factors. The reasons why salt intake is consistently high are unknown. One possibility is that the inclusion of salt within processed foods75, which are a dominant food source in many countries, negates discretionary control of salt intake. Even ostensibly natural cuts of meat include hidden salt76 due to the common industry practice of injecting saline during processing in a practice known as ‘plumping’ For most people, these practices create an environment in which sustaining a low salt intake is a difficult and resource-intensive choice. However, salt intake is also high in regions where the majority of salt intake is under discretionary control, for example, in many areas of China75. The uniformly high salt intake in most populations worldwide raises the question of whether a biological rationale exists for habitual salt consumption far in excess of physiological need. Understanding the mechanism might be important to identify and support individuals who are more sensitive to the adverse effects of high-salt intake and to inform effective salt-reduction strategies at the population level. 

Genetic variants

Although salt is an environmental risk factor, its effects might be primarily determined by other modifying exposures, including genetic susceptibilities. Human data support this hypothesis. In the UK Biobank cohort, genetic variants were associated with estimated sodium intake at genome-wide significance. These variants were implicated in the regulation of dietary preferences, behaviour, learning, cognition, thermoregulation and weight loss and were associated with genes expressed in brain, adipose and vascular tissue23. Similar observations have been made in other populations. A meta-analysis of GWAS in European populations (n = 6,500) identified seven variants that were associated with urinary sodium excretion77, including variants in the CARPTP and ZSWIM5 genes, which regulate eating, appetite and neural development. These findings fit well with current understanding of salt biology because, as we discuss below, ‘salt preference’ – a behavioural phenotype – is an important determinant of sodium intake (and hence sodium excretion). 

Salt appetite

Salt appetite and salt preference describe different parameters in the afferent arc of salt homeostasis. Salt appetite is the innate motivation to find and eat salt evoked by a physiological need relating to salt deficiency. The primordial emotion arose during the transition from aquatic to terrestrial life with selection pressure imposed by the scarcity of salt across much of the land mass. The sodium content of plants is typically low and animals with predominantly plant-based diets are vulnerable to salt deficiency even if calorific intake is adequate. The evolution of behavioural pathways to ensure adequate salt intake is a protective adaptation to safeguard extracellular fluid volume status. Thus, salt appetite is strongly evident in herbivores; for example, sheep will avidly eat mineral sodium salts to precisely correct deficiency78, elephants quarry and eat mineral salt as part of their foraging behaviour79 and the provision of salt-licks has long been used as a strategy for domestication. For carnivores, the salt content of food is always adequate and they do not exhibit an innate salt appetite, even when sodium depletion is induced experimentally80. The earliest hominins preceding Homo erectus were omnivorous, eating meat from terrestrial and aquatic animals81. Their estimated salt intake of ~1 g (20 mmol; 400 mg Na) per day82 is similar to, or even slightly higher than, that of present-day isolated populations, such as the Amazonian Yanomami83. Whether this salt intake – low by contemporary standards imposed sufficient selection pressure to retain evolved salt appetite is uncertain. On the one hand, chimpanzees retain salt-seeking behaviours when environmental sodium is low84 and salt-craving is reported in humans with salt-wasting disorders such as Gitelman Syndrome85 and adrenal insufficiency86, including a classical case report of a child who consumed salt in mineral form87. On the other hands, salt craving is not a consistent feature of saltwasting disorders and controlled experiments that induce salt-depletion in healthy people do not provide evidence of an innate salt appetite

Salt preference

In contrast to their lack of salt appetite, people display a strong salt preference; that is, a desire to eat salty food in the absence of physiological need. Despite an aversive response to mineral salt in isolation, equivalent molar concentrations of salt within the food matrix are highly palatable89. The evolution of hedonistic instinct serves two purposes: first, it provides positive reinforcement to a physiological imperative; second it signals satiation, closing the negative feedback loop. In mice, for example, satiation of experimentally-induced salt appetite induces rapid transcription of hypothalamic genes important for gratification and reward. Direct blockade of limbic dopaminergic reward pathways in mice eliminates salt appetite in the absence of satiety90. Thus, while it is probable that salt appetite in humans is at most a vestigial instinct, hedonic preference remains. Coupled with ready accessibility to dietary salt, this preference might provide the key to understanding continuing high levels of salt intake as well as poor long-term adherence to lower salt intake. People mostly experience reduced salt alternatives as unappealing and weeks91 or months92 are required to induce the hedonic shift that restores a positive sensory experience at a lower salt threshold, whereupon foodstuffs that were previously palatable are regarded as unpleasantly salty93. Salt preference is plastic and influenced by the context of habitual exposure, which makes it difficult to gauge tractability as a means to reduce intake. For example, whether a sustained reduction in salt intake is necessary to induce a preference for foods containing lower sodium levels is not known. Moreover, if this goal is attained, the durability of sensory resetting in the face of fluctuating salt intake is unknown. Nevertheless, the ramifications for clinical management of patients are intriguing. Recognition of salt taste is impaired in people hospitalised with heart failure94 and recovery of taste may be a prognostic biomarker of improved outcomes95. Patients with CKD also have impaired salt-taste acuity and increased salt-preference96,97, possibly due to structural abnormalities in the taste buds98. 

The salt taste response

 Of the five basic tastes, sweet and umami are attractive and promote appetite, whereas bitter and sour are innately aversive. Salt taste is the only modality with a biphasic response ranging from appetitive (<500 mM) to powerfully aversive (>500 mM). In evolutionary terms, this response might serve to ensure adequate salt intake while protecting against the emetic and potentially fatal99 effects of concentrated salt ingestion100. The transduction of appetitive salt taste occurs through at least two distinct pathways. The major pathway is via the amiloride-sensitive sodium channel (ENaC) on the apical membrane of type 1 taste receptor cells in fungiform papillae. Sodium entry depolarises the cell, initiating action potentials in the chorda tympani101,102. Studies using conditional knockout strategies have confirmed that ENaC is the dominant salt-taste pathway in mice103,104. The molecular nature of the amilorideinsensitive pathway(s), located to the circumvallate and foliate taste buds, has not been resolved but a role of transient receptor potential cation channel subfamily V member 1 (TRPV1) has been suggested105. However, this channel is non-selective and appears broadly tuned to cations and osmolarity102 and TRPV1-knockout mice exhibit amiloride-insensitive sodium taste106. Intriguingly, when sodium is tasted the accompanying anion modulates the information encoded into rodent chorda tympani action potentials102. This mechanism might also apply to humans where large anion size increases the latency and reduces the intensity of the salt-taste response, such that monosodium glutamate is experienced as less intensely salty than the equivalent molar amount of sodium chloride107. This interaction provides insight into why molecular entities are often perceived differently within the food matrix compared to when experienced in isolation108. Thus, modifying the chemical microenvironment within complex structures could potentially maintain the perception of salt taste while achieving a reduction in salt content. Single nucleotide polymorphisms (SNPs) in ENaC subunits and TRPV1 are associated with salt-taste perception109 but whether these SNPs influence sodium intake is not clear110,111. The available association studies are small and caution must be applied when interpreting the results: ENaC and TRPV1 are also expressed in the gut and kidney so genetic variation in these channels will affect salt homeostasis at multiple points112. In the mouse brain, in vivo calcium imaging has mapped the organization of the primary gustatory cortex, locating the neuronal activity engaged by salt taste (100 mM NaCl) to a distinct area, spatially segregated from those engaged by other tastants113. Taste, appetite, and the motivation of behaviour is integrated between the lamina terminalis in the forebrain (an array of nuclei in the median preoptic area, subfornical organ (SFO), and the organum vasculosum laminae terminalis (OVLT)) and hindbrain (area postrema and nucleus tractus solitaris (NTS)). Projections from some of these areas into the nucleus accumbens integrate salt-appetite with the dopaminergic mesolimbic ‘reward’ system to engage salt-seeking behaviours. Pharmacological blockade of these central reward pathways disengages saltseeking behaviour in sodium-depleted mice90. It is possible that such gratification pathways contribute to salt preference and high salt intake in humans114 but this has not been extensively examined.

The renin-angiotensin-aldosterone system (RAAS) has a major role in driving the behavioural responses to experimentally-induced sodium appetite. Angiotensin II (ANGII) and aldosterone are synthesized de novo within the brain115. In rats, brain aldosterone levels increase with salt-restriction and are supressed by high salt intake, but whether this response reflects altered central synthesis is not clear as the majority of aldosterone in brain tissue comes from the circulation116. The SFO and OVLT have no blood–brain barrier and neurones express ANGII receptors117 that receive information about peripheral salt and water homeostasis by detecting circulating ANGII. Distinct groups of type 1 angiotensin II receptor-positive SFO neurones drive thirst and salt appetite and are mutually suppressive via activation of interconnecting GABAergic neurons 118. However, in rodents aldosterone-sensitive neurones in the NTS are necessary and sufficient to engage salt-appetite119. Neuronal activity is increased by salt-deficiency and reduced upon satiety106 and activation engenders a motivational state specific for sodium rather than thirst or more generalised hunger120. This population of neurones express both the mineralocorticoid receptor and the glucocorticoid-metabolising hormone 11bhydroxysteroid dehydrogenase 2 (11b-HDS2) and are therefore classical aldosterone target cells121. Conditional deletion of 11b-HSD2 in the NTS promotes abnormal mineralocorticoid receptor activation and in sodium-replete mice induces a preference for saline solutions and, when given free choice, a large increase in salt intake122. Such salt preference in the absence of physiological need is akin to the typical human condition. Notably, ANGII signalling in the SFO and aldosterone signalling in the NTS are synergistic, such that maximum, rapid salt intake is only engaged when both are elevated119. This mechanism ensures that salt intake is appropriately promoted in the context of hypovolaemia (which results in high ANGII and high aldosterone levels, but less so when aldosterone is elevated in the context of hyperkalaemia. 

Determinants of salt-sensitivity

 Salt-sensitivity usually refers to the gradient of the relationship between salt intake and blood pressure. However, a more patient-centred concept is the salt-sensitivity of CVD; that is, the gradient of the relationship of salt intake with cardiovascular risk.

In a substantial proportion of published studies, individuals are classified as having blood pressure that is either salt-sensitive (the minority) or salt-resistant (the majority). Such binary categorisation is useful in a research context, for example because it enables demonstration of an independent association between saltsensitivity per se and cardiovascular morbidity and mortality11,12, but is generally unhelpful in real-world settings for at least four reasons. First, the categorisation is an artificial construct: in humans, salt-sensitivity of blood pressure is a continuously distributed and reproducible trait123-126. Second, the thresholds used to define saltsensitivity are arbitrary; there is no consensus definition. Third, the protocols that are used to define salt-sensitivity in a research setting are not practical in the clinic127,128. Fourth, and most importantly, binary categorisation in a societal context could perpetuate the idea that habitually high salt intake has no adverse health consequences for the majority of the population and therefore pose a barrier to health improvement. Populations that exhibit increased salt-sensitivity can be identified from RCT data. Sodium restriction induces greater decreases in blood pressure in populations that are older, non-white or hypertensive1,30. Evidence from animal models suggests that kidney impairment and diabetes mellitus also induce salt-sensitivity, but no wellcontrolled studies have been conducted in humans to confirm or quantify this effect. In sum, the sensitivity of an individual to the damaging effects of dietary salt is neither categorical nor static and life events, including normal ageing, will move a person along this spectrum such that the adverse response is exaggerated. 

Guyton’s hypothesis 

The body has many cardiovascular control systems, and yet for over 50 years, mechanistic explanations for salt-sensitivity have been dominated by ‘The Guyton Hypothesis’, in which the proximate cause of salt-sensitivity is a failure to appropriately regulate renal sodium disposal129. As blood pressure increases, so too does renal artery perfusion pressure which in turn stimulates sodium excretion. This relationship – termed ‘pressure natriuresis’ – was originally described in the isolated, perfused canine kidney130,131 and is the central tenet of a hypothesis to explain the maintenance salt homeostasis by a simple negative-feedback loop (

Constructing a computational model of circulatory function, Guyton and colleagues found that the pressure-natriuresis feedback loop operated with infinite gain – i.e. that any deviation from ‘set-point’ blood pressure would be corrected exactly by a corresponding change in renal sodium excretion and extracellular fluid volume132. They argued that salt-sensitivity in blood pressure therefore must reflect some defect in pressure-natriuresis129,132,133. Certainly, animals with salt-sensitive blood pressure can be shown also to have a blunted pressurenatriuresis relationship (Figure 3B) 130,131,134, but whether this ‘blunting’ is necessary and sufficient to render blood pressure salt-sensitive in the long term is uncertain and extremely challenging to investigate. A small number of studies in conscious dogs have addressed this, using servo-controllers to dissociate renal perfusion pressure from systemic arterial blood pressure: these experiments showed that blocking pressure-natriuresis causes sustained hypertension over days (e.g. in the context of noradrenaline infusion135; Figure 3C).

Conceptually, a defect in pressure natriuresis could be intrinsic to the kidney, reflecting hyperactive renal sodium transporters or a defect in the paracrine signalling that controls pressure natriuresis136. However the kidney does not operate in isolation and neuroendocrine systems exert powerful modulatory effects on the intrinsic pressure-natriuresis relationship. The RAAS is particularly influential, dynamically adjusting sodium output to match input and thereby minimizing the impact of salt intake on long-term blood pressure. When salt intake is high, the RAAS is suppressed and the pressure-natriuresis system is free to drive commensurately high sodium excretion. When sodium intake is low, the RAAS is activated and ANGII drives renal sodium retention and vasoconstriction, maintaining a blood pressure that would have otherwise fallen. Thus, dynamic modulation of the RAAS with sodium intake is a key factor in salt-resistance of blood pressure. Conversely, when the RAAS is tonically suppressed137,138 (e.g. with an ACE inhibitor), blood pressure falls; when the RAAS is tonically stimulated137,139,140 (e.g. in primary hyperaldosteronism), blood pressure rises. In both cases, blood pressure becomes salt-sensitive because the RAAS is unable to dynamically match renal sodium output to input

Figure 3D). Guyton and his colleagues pioneered the use of computational approaches to understand complex physiological systems and introduced new concepts that were profoundly influential in clinical practice and research. However, whether their hypothesis explains salt-sensitive hypertension is controversial and the approach has been criticised for an over-reliance on theoretical modelling rather than empirical data141,142, for conflating dependent and independent variables (a criticism also levelled at Guyton’s venous return curve143) and for tautologous reasoning144,145. Moreover, predictions made using the model do not consistently replicate experimental outcomes146. In our view, a defect in regulating renal sodium excretion is neither necessary nor sufficient for salt-sensitive hypertension. That such a defect is not necessary is evident from studies in which salt-induced hypertension occurs independently of changes in total body sodium. For example, in African American volunteers, saltloading induced a positive sodium balance (i.e. an intake exceeding output) and increases body weight and extracellular volume to the same extent in all participants, regardless of whether they were classified as ‘salt-sensitive’ or ‘saltresistant’ 147. That such a defect is not sufficient is apparent from the many patients who have sodium retention without hypertension (for example, those with nephrotic syndrome, cardiac failure or portal hypertension) and the observation that infusion of sufficient levels of ANGII to blunt the excretion of a sodium load did not increase blood pressure in healthy volunteers148. Experiments in animals have illuminated fundamental mechanisms of saltsensitive hypertension but high-salt models often employ exaggerated, superphysiological salt intake with exposure times restricted to days or weeks, rather than months or years. Although this acute approach pushes homeostatic systems to their limits, such modelling likely provides a poor representation of the chronic, moderate elevations in salt intake observed in human populations. Moreover, manipulation of individual dietary components is easy in animal models but unusual for human diets in which interventions designed to change sodium intake almost always change other dietary parameters (for example, potassium or chloride) that have independent effects on cardiovascular physiology. 

The contemporary view

 Current evidence indicates that the physiological response to excess salt is complex and involves changes in renal, vascular, neural, metabolic and immune functions .

Figure 3). Criticisms of the Guytonian model notwithstanding, renal salt handling and vascular tone are central determinants of salt-sensitive blood pressure changes. Numerous strands of evidence suggest that the kidneys exert a dominant influence on salt-sensitivity in many settings. In experimental models, salt-sensitivity can be induced by impairing natriuresis through nephrectomy151 or by infusing ANGII139 or mineralocorticoids152. Differential activation of natriuretic peptide signalling can also contribute to salt-sensitive hypertension: atrial natriuretic peptide levels were lower in salt-sensitive than in salt-resistant subjects in one small study in hypertensive adults153. Rodent studies have shown that cross-strain transplantation of kidneys from salt-sensitive donors leads to the development of salt-sensitivity in previously salt-resistant recipients154 The mechanism of salt-sensitivity sometimes involves volume expansion; a meticulous histochemical analysis found that salt Fig 3 Manuscript number NRNEPH-19-123 Bailey Perspective 09|12|21 a c b d Arterial pressure Salt intake ACE inhibitor AngII Control Aterial pressure ENa (urinary salt excretion) Salt-resistant Salt-sensitive Servo-control (renal perfusion pressure clamped) Salt intake ↑Aterial pressure ↑Total body sodium ↑Extracellular volume ↑Renal sodium excretion Pressure natriuresis Mean systemic arterial pressure (mmHg) Servocontrol Noradrenaline infusion 90 100 110 120 130 140 Urinary sodium excretion (ENa mmol per day) 40 60 –2 –1 0 1 2 3 4 5 6 7 8 80 100 120 Days No servo-control (renal perfusion pressure matches systemic arterial pressure) 22 loading caused an increase in the sodium and water contents of multiple body tissues in salt-sensitive but not salt-resistant rats155. Increased vascular tone is also a common feature of most salt-induced hypertension models. People with salt-sensitivity exhibit a blunted drop in systemic vascular resistance after a salt load147,156. No convincing support exists for Guyton’s hypothesis that this effect is the result of whole-body autoregulation of blood flow. However, there is evidence that high sodium intake induces direct effects on vascular smooth muscle. Using both pharmacological and genetic perturbations in mice with salt-induced hypertension, Iwamoto et al. found that high sodium intake stimulated production of endogenous cardiac glycosides (ouabains), leading to an accumulation of sodium within vascular smooth muscle cells. This sodium was then extruded through the sodium-calcium exchanger NCX1, provoking calciumdependent vasoconstriction and a rise in blood pressure 157. Mice carrying a selective PPARg mutation in vascular smooth muscle cells exhibit global impairment of vasodilation and salt-sensitive hypertension158. It would be interesting to investigate the relative contribution of the systemic and renal vascular beds to saltsensitive hypertension in this model, for example using cross-transplantation studies. The central and autonomic nervous systems have an important role in salt sensitivity through their effects on vascular tone. In C57BL/6 mice, a widely-used strain that is often considered to be salt-resistant159, high salt intake raises blood pressure not via renal sodium retention, but by increased sympathetic activity and an augmented vascular response to catecholamines160. Salt-induced changes in blood pressure and sympathetic activity are abolished in mice lacking the critical central sodium sensor, NaX161. The immune system is important in establishing salt-sensitive hypertension and in modifying hypertension-induced organ damage. In rats, salt-loading induces VEGFC secretion by macrophages and blocking this system enhances salt-induced hypertension162. VEGF-C stimulates lymphangiogensis; it is hypothesised that this provides an interstitial sodium ‘buffer’, preventing sodium loads from inducing intravascular volume expansion and raising blood pressure162. Whether the proximate stimulus in this pathway is isotonic or hypertonic tissue sodium accumulation is debated155. A high-salt diet also induces T helper 17 (TH17) cells, inciting a pro-inflammatory response163,164. TH17 cells secrete IL-17, which sustains 23 ANGII-dependent hypertension by stimulating inflammatory signalling pathways in vascular smooth muscle 165 and by stimulating the expression of sodium transporters in the renal tubule166. Conversely, salt-wasting tubulopathies are associated with reduced TH17 cell activation, which may explain the increased prevalence of mucosal infections and allergy in this patient population167. In addition, excess dietary salt modifies the gut microbiome in mice and humans, inducing proinflammatory changes that perturb diurnal rhythms in blood pressure and drive hypertension168-170. 

Genetic associations 

As understanding of the physiology of salt-sensitivity has evolved beyond a simplistic, renal-vascular model to one involving diverse body systems, a parallel shift has occurred in knowledge of the genetic associations of salt-sensitivity. Initially, genetic variants that were associated with salt-sensitivity were identified through studies of rare Mendelian disorders and candidate gene approaches , which implicated genes controlling renal sodium handling171 and vascular tone127. However, these approaches are subject to confirmation bias and do not provide information about the relative importance of genetic variants in the general population. Unbiased studies of gene-sodium interactions in East Asian populations have identified a small number of genetic variants that are associated with salt-sensitivity of blood pressure with genome-wide significance25,172,173. The full functional consequences of these variants are not known, but nearby genes participate in diverse physiological processes including adrenergic and ANGII signalling (MKNK1), T-lymphocyte differentiation (BCL11B), inflammation (IRAK1BP1) and insulin signalling (PHIP). 

Salt-reduction strategies

 Expert guidelines advocate dietary salt restriction in a range of populations owing to the proven beneficial effects on blood pressure and projected beneficial effects on CVD (supplemental Table 2). However, these recommendations are not universally accepted. In a review published in 2020, an international group of experts stated that insufficient evidence exists to support the widespread recommendation to limit 24 sodium intake to <2.3 g per day (~100 mmol; 5.85 g NaCl)174. They recommended salt restriction only in populations with a mean sodium intake >5 g per day (~220 mmol; 13 g NaCl). Others have argued that because sodium intake in many populations worldwide is within a range that is associated with long life-expectancy and low rates of CVD (130–200 mmol per day; 8–12 g NaCl), attempts to reduce sodium intake are neither desirable nor possible10. We do not follow the logic of these arguments, which place an undue emphasis on low-quality observational data and ignore the many other environmental and genetic determinants of health. One argument against population-level intervention is that if only 20–40% of healthy individuals have salt-sensitive blood pressure, the majority of people will not benefit from reductions in salt intake. At least five counterarguments, in our view, support population-level intervention. First, salt reduction strategies have been effective in whole populations but are relatively untested when applied to individuals in the longer term. Second, accessible methods of identifying at-risk, ‘salt-sensitive’ individuals are lacking. Even if salt intake could be effectively reduced only in individuals with hypertension, this approach would exclude the 40–50% of these individuals in whom hypertension is undiagnosed (estimated to number in the tens of million in the USA alone)175-177. Third, any risk-stratification is likely to become increasingly complex in an ageing population with a high prevalence of multimorbidity; identifying approaches to account for several interacting health conditions will be challenging178. Fourth, most societies have already fully adopted the principle that it is acceptable to intervene in many individuals to improve the lives of a small subset. This principle provides a foundation for public health interventions as well as for the prescription of most cardiovascular medicines, including statins179. Finally, intervention at the population level provides an opportunity to redress health inequalities. Maintaining a healthy diet requires knowledge, time and money180. In a survey of ~4,700 adults in the USA, adherence to a low-salt Dietary Approaches to Stop Hypertension (DASH) diet correlated with higher income and educational attainment181. Intervention at a population level would remove these socioeconomic barriers to dietary salt reduction. 

Population-level interventions 

 At least 96 nations have implemented population-level salt-reduction strategies182. However, salt intake still exceeds the recommended thresholds in many populations (Figure 2). The success of salt-reduction strategies in Finland and the UK demonstrate that wholesale reduction in salt intake is possible. In these countries, multi-modal strategies involving third-sector lobbying groups, government and industry led to sustained reductions in population salt intake of ~40% and ~15%, respectively183-187. These reductions are likely to have contributed to a parallel improvement in cardiovascular outcomes188. In the UK, robust monitoring of salt intake and labelling of food salt content enabled the use of a target-based approach that resulted in a ~30% reduction in the salt content of processed foods during the first phase of the program184. Despite further targeted reductions, government failures such as tasking the food industry to self-regulate without robust, external monitoring of salt content in foods, stalled the decline in population salt intake189. Systematic reviews of salt-reduction strategies report that multi-component interventions involving legislation, mandatory reformulation of processed foods and food labelling are the most effective approaches to reduce salt intake190,191. A large meta-analysis that included >2 million observations in 11 countries, confirmed that food labelling legislation results in a reduction in salt content192. Given that other dietary constituents including potassium and fructose193 also modulate blood pressure and cardiovascular risk, the most effective interventions would likely rely on holistic dietary changes at multiple levels within society, including policy, industry, healthcare and education. Novel population-level interventions involving the provision of salt substitutes31,32, salt-restriction spoons194 and education have also led to substantial reductions in salt intake in clinical trials. 

Individual-level interventions 

Individual-level interventions have achieved sizeable reductions in salt intake, averaging around 100–130 mmol (6.0–7.5 g NaCl) in the short term (Figure 2) 1,58. However, even in the context of clinical trials, achieving a sodium intake of <5g NaCl (85 mmol) per day as recommended by the WHO5 has proven to be extraordinarily difficult. In a meta-analysis of RCTs in patients with heart failure, the mean achieved sodium intake in groups with a targeted intake of <1.5 g (65 mmol) per day ranged from 1.9 to 4.6 g (~80–200 mmol) per day195. Furthermore, reductions in salt intake 26 might not be sustained in the longer term. The Huang et al. and Filippini et al. metaanalyses of sodium-reduction RCTs discussed above found that interventions lasting for longer than 12 weeks or 6 months, respectively, induced smaller reductions in blood pressure than did shorter interventions1,30. In small RCTs in Dutch patients with CKD, patient-centred interventions aimed at reducing sodium intake reduced sodium excretion at 3 months but these differences were not sustained at 6 months196,197. These findings are perhaps not surprising given the powerful cultural and biological forces driving salt intake. They suggest that the biggest obstacle to improved health outcomes is not a lack of understanding about how and why salt causes disease but rather a lack of evidence of how to intervene most effectively at the individual level. Many factors shape the optimal salt-reduction strategy for individuals. To our knowledge, attempts to evaluate the effect of family-based interventions and feedback using spot urine samples or genetic risk scores to identify individuals most likely to benefit from salt restriction have been limited to small, uncontrolled studies. Individuals vary in the extent to which they prefer to use dietary rather than pharmacological means to reduce total body sodium. Although short-term studies indicate that diuretics and dietary sodium restriction exert broadly equivalent shortterm effects on body water and blood pressure198, whether they confer similar longterm benefits on disease outcomes is unknown. The efficiency of salt-reduction strategies could potentially be increased by leveraging knowledge of the biology of salt preference. For example, amiloride effectively blocks salt taste when applied topically to the lingual epithelium and in mice, targeted knockout of ENaC in taste receptors reduces salt preference and salt intake104. In humans, one small trial suggested that therapeutic doses of amiloride may limit salt intake without inducing any conscious perception of change in taste199. It would be productive to explore this potential therapeutic benefit in larger clinical trials. 

Conclusions

Overwhelming evidence indicates that excessive intake of dietary salt raises blood pressure. In addition, evidence from multiple sources suggests that excessive salt 27 intake causes CVD, particularly in high-risk populations. This hypothesis is yet to be confirmed unequivocally in a RCT of sodium restriction, but the cost of such a trial is likely to be prohibitive. Powerful socioeconomic, cultural and biological drivers explain the high dietary salt intake that is observed in many countries worldwide. The extent to which individuals are susceptible to salt-induced disease outcomes depends on salt preference and salt-sensitivity of blood pressure, which are determined by genetic, demographic and environmental factors. The high burden of ‘hidden salt’ in processed foods can make it difficult for individuals to control salt intake, but does not render them completely powerless. Exercising individual discretion is a crucial part of any salt reduction strategy and individuals should be given the tools and knowledge to facilitate this approach. Multi-component interventions have achieved sustained reductions in salt intake in large populations, whereas individual-level interventions have achieved sizeable reductions in salt intake over days to weeks in clinical trials. The effectiveness of individual salt-reduction strategies outside of clinical trials is uncertain as is whether such strategies can sustain reductions in blood pressure in the long term (months to years). Reformulation of foods to reduce salt content is a successful strategy to lower population salt intake but requires government intervention to ensure fair commercial competition. Strategies legislating for reformulation, encouraging reformulation by robust monitoring of reduction targets and even taxation of salt used in food manufacturing have been used to reduce population salt intake with some success180, but require governments to look beyond short-term political impact to a long-term view of public health.

Author contributions

 All authors contributed to researching the data, discussing the content, writing the text and reviewing or editing the manuscript before submission. 

Acknowledgements

 28 The authors have received research funding from The British Heart Foundation (PG/16/98/32568), The Chief Scientist’s Office (SCAF/19/02), Diabetes UK (17/0005685), Kidney Research UK (RP02/2019; IN001/2017), the Medical Research Council (MR/S01053X) and The Wellcome Trust (209562/Z/17/Z).