Cutting salt reliably lowers blood pressure. Almost everything past that is less certain.
- Reducing sodium lowers blood pressure — the best-established link in this whole area, and larger in people who already have high blood pressure than in those who do not.
- Whether that prevents heart attacks, strokes and death is far less certain. The one strong trial on hard outcomes raised potassium at the same time it cut sodium, so its benefit cannot be credited to salt reduction alone.
- Potassium is the coupled lever — worth raising from food, but on current evidence it lowers blood pressure only in people who already have hypertension.
- The scare that very low sodium is dangerous has never faced a decisive test — it is unresolved, not disproven, and it is the largest open question here.
- The people who gain most are those with high blood pressure and high cardiovascular risk. A lean, normotensive person gains little in absolute terms and has no reason to chase a gram target.
The honest bottom line: cut sodium if your blood pressure or cardiovascular risk is high, raise potassium from fruit, vegetables and pulses, and treat every published gram-target as the edge of the evidence rather than a proven cliff.
Cutting sodium lowers blood pressure, and that step is as settled as this field gets
The surrogate is the strong part of the story. Pooling 34 randomised trials (3,230 participants, reductions maintained four or more weeks), a modest cut in salt lowered systolic blood pressure by 4.18 mmHg (95% CI 3.18 to 5.18) and diastolic by 2.06 mmHg (He et al., 2013). WHO’s own 36-trial pool put the resting systolic effect at 3.39 mmHg (2.46 to 4.31) and graded it HIGH certainty (World Health Organization, 2012). The two pools differ in size because they are different trial sets, not the same analysis restated.
«A modest reduction in salt intake for four or more weeks causes significant and, from a population viewpoint, important falls in blood pressure in both hypertensive and normotensive individuals, irrespective of sex and ethnic group.» (He et al., 2013)
Sodium is not even the only dietary blood-pressure lever: the whole DASH pattern lowers systolic pressure by 5.2 mmHg (95% CI 3.4 to 7.0) in a 20-trial meta-analysis, and its between-arm sodium difference does not predict that fall (P=0.67), so its effect is carried by the pattern, not by its incidental salt contrast (Siervo et al., 2014). The two effects act through the same channel and must not be added together.
The blood-pressure fall is biggest in people who already have high blood pressure
The effect is real in everyone but not equal across strata. In He’s pool the systolic fall was 5.39 mmHg (4.15 to 6.62) in hypertensive participants versus 2.42 mmHg (1.29 to 3.56) in normotensive ones (He et al., 2013); WHO’s figures are 4.06 mmHg (2.96 to 5.15) and 1.38 mmHg (0.02 to 2.74) — a roughly three-fold gap either way (World Health Organization, 2012).
The cleaner reading of that gap is that the real modifier is continuous baseline blood pressure, not a hypertensive/normotensive dichotomy. Huang’s larger pool found the dichotomy non-significant (P=0.08) while the graded baseline-pressure trend was significant (P=0.01), with an effect present even below 120 mmHg systolic (Huang et al., 2020). Older age and non-white ethnicity are further modifiers; baseline sodium intake is not. So the sodium lever pays across the pressure range and pays most where pressure is already high — a genuine effect-modification finding, not merely arithmetic.
Whether less sodium prevents heart attacks and strokes is far less certain
Blood pressure is a surrogate. It matters only because lower pressure is known to cut hard events: the randomised evidence (344,716 people, 48 trials) is unambiguous on that second step —
«a 5 mm Hg reduction of systolic blood pressure reduced the risk of major cardiovascular events by about 10%, irrespective of previous diagnoses of cardiovascular disease, and even at normal or high-normal blood pressure values» (Blood Pressure Lowering Treatment Trialists Collaboration, 2021)
— with stroke down 13%, heart failure 13%, ischaemic heart disease 8% per 5 mmHg. But that trial lowered pressure with drugs. Transferring the same magnitude to a sodium-induced fall is an assumption, not the same evidence.
On sodium’s own hard-outcome evidence, WHO rated stroke, cardiovascular disease and mortality VERY LOW certainty — cohort data only (e.g. stroke RR 1.24, 1.08 to 1.43 per higher intake) — against HIGH for blood pressure (World Health Organization, 2012). It bridged the gap by transfer rather than by direct evidence:
«the evidence of an effect of sodium on blood pressure was also considered moderate-quality evidence that reduced sodium is beneficial for reducing risk of cardiovascular disease, stroke and coronary heart disease» (World Health Organization, 2012)
WHO said as much itself: the blood-pressure evidence «was of high quality, whereas the evidence regarding sodium intake and all-cause mortality, cardiovascular disease, stroke and coronary heart disease was of lower quality» (World Health Organization, 2012). GRADE’s rule is that a surrogate is downgraded «by one, or even two, levels» for indirectness (Schünemann et al., n.d.); the certainty on the outcome a person actually cares about is therefore well below the certainty on the marker.
The strongest hard-outcome trial cut sodium and raised potassium at once
The one hard-outcome RCT here is SSaSS: 20,995 high-risk rural-Chinese adults (72.6% with prior stroke; mean pressure 154/89; mean age 65) randomised to a 75% sodium chloride / 25% potassium chloride salt substitute versus regular salt for 4.74 years. It cut stroke (rate ratio 0.86, 0.77 to 0.96), major cardiovascular events (0.87, 0.80 to 0.94) and all-cause death (0.88, 0.82 to 0.95) — absolute reductions of 4.5, 7.2 and 5.3 events per 1,000 person-years (Neal et al., 2021).
The catch is the exposure. The switch raised potassium more than it cut sodium (urinary potassium +20.6 mmol; sodium -15.2 mmol; systolic pressure -3.34 mmHg), so it moves two levers at once and its benefit cannot be attributed to sodium reduction alone (Neal et al., 2021). SSaSS is strong evidence that a salt substitute helps a high-risk group; it is not clean evidence that sodium reduction per se prevents events, and it says nothing about which sodium-to-potassium ratio is best, because it used one formulation.
Two further reads matter. The event benefit slightly outran what a 3.34 mmHg fall predicts (RR ~0.91 expected from BPLTTC, 0.86 observed), consistent with a blood-pressure-independent potassium effect or with the measured pressure fall understating the sustained one. And the large absolute benefit rides on an extreme baseline risk; the ~12-14% relative reductions may transport to lower-risk people, but the absolute gain shrinks with them -> Baseline Risk and the Relative-Absolute Split.
Potassium is the coupled lever, but it only lowers blood pressure in people with hypertension
The companion WHO review (22 trials, 1,606 participants; 11 cohorts, 127,038 people) sets the potassium half:
«Increased potassium intake reduced systolic blood pressure by 3.49 (95% confidence interval 1.82 to 5.15) mm Hg and diastolic blood pressure by 1.96 (0.86 to 3.06) mm Hg in adults, an effect seen in people with hypertension but not in those without hypertension.» (Aburto et al., 2013)
That asymmetry is the decision-relevant point: raising potassium lowered systolic pressure by 5.32 mmHg (3.43 to 7.20) in hypertensive participants but by 0.09 mmHg (-0.77 to 0.95) — essentially zero — in normotensive ones (Aburto et al., 2013). So cut sodium and raise potassium are not symmetric advice: for a normotensive person the sodium lever still moves pressure while the potassium lever, on current evidence, does not.
That normotensive null is not fully settled, and Aburto flags the reason: «However, the studies in people without hypertension were of relatively short duration and did not consider the effect that increased potassium intake may have over time on the prevention of elevated blood pressure» (Aburto et al., 2013) — a short-duration limit, so read it as no effect shown over weeks, not no effect over a lifetime. Where potassium does lower pressure the dose-response is flat-topped rather than graded: systolic pressure fell «7.16 (1.91 to 12.41) mm Hg when the higher potassium intake was 90-120 mmol/day, without any dose response» (Aburto et al., 2013) — the benefit concentrates at moderate repletion and does not keep climbing with intake.
As with sodium, stroke is the one hard cohort outcome that reaches significance (RR 0.76, 0.66 to 0.89, LOW certainty); cardiovascular disease, coronary disease and mortality stay null and underpowered.
Potassium raising is safe for lipids, catecholamines and renal function (all null, HIGH certainty) in people with normal potassium handling — and food is a sufficient vehicle; supplements are not required (Aburto et al., 2013). Fruit, vegetables and pulses are the named sources, and processing strips potassium.
The sodium-to-potassium ratio is asserted beneficial near 1:1 but never graded against a hard outcome — that figure traces to a 2003 expert consultation, not to this evidence (Aburto et al., 2013). SSaSS moved the joint lever from asserted to partly-evidenced; the optimal ratio remains an open, ungraded question on both sides.
The claim that very low sodium is dangerous has never faced a decisive test
Some cohorts show higher event rates at low sodium intake — a J-shaped curve read by some as a reason not to cut salt far. This is exactly the pattern the wiki treats with suspicion: a protective-looking or harmful-looking lower arm in observational data is often an artifact of reverse causation, sick-quitter bias or confounding by frailty, and must survive a referent-correction or a genetic check before it is believed -> The U-Shaped Association Artifact.
The largest direct test of that lower arm is PURE. Mente pooled 133,118 people across 49 countries with more than 10,000 hard events (median 4.2 years) and found a J-shaped curve on death-plus-cardiovascular-events, its nadir at 4-5 g/day and risk raised at both ends: below 3 g/day the risk rose in hypertensive (HR 1.34, 1.23 to 1.47) and normotensive participants alike (HR 1.26, 1.10 to 1.45), while above 7 g/day the harm appeared only in hypertensives (HR 1.23, 1.11 to 1.37) and was null in everyone else (HR 0.90, 0.76 to 1.08) (Mente et al., 2016). Because that high-intake harm concentrates in hypertensives — who consuming above 7 g/day make up only ~11% of the studied population — Mente reads the data as a case for targeting salt reduction at high-intake hypertensives rather than pushing the whole population toward a low-sodium target (Mente et al., 2016).
That targeting conclusion is genuinely contested, not a verbal dispute, and a filed tension turns on it -> Should Sodium Reduction Be Population-Wide or Targeted. Two counters keep PURE from settling the question. First, its exposure is a single fasting spot urine run through the Kawasaki formula, and a one-day recovery biomarker is least accurate at the extremes — exactly where the J-arms sit — so unequal measurement error can bend a flat relationship into a U with no confounder at all; PURE ran only weak artifact checks (excluding early events, prevalent disease and medicated participants) and neither a referent-correction nor a Mendelian-randomization analysis of the low arm (Mente et al., 2016). Second, the targeting case leans on a premise that blood-pressure lowering helps only above 140 mmHg systolic, which the larger BPLTTC evidence contradicts — benefit reaches down to normal pressure and no subgroup is harmed -> Blood Pressure Lowering and Cardiovascular Events.
The two literatures also partly measure different things: the population-wide case rests on a randomised reduction against a surrogate (blood pressure), PURE on an estimated habitual level against directly observed hard outcomes, and the disputed low arm sits largely outside the reduction trials’ range — so the shape clash is smaller than it looks while the policy conflict is exactly as large. Where the poles agree is the useful part: for a hypertensive with high intake, both say reduce. The live disagreement is over the normotensive at moderate intake (3-6 g/day), and there the honest state is insufficient evidence on the low arm, not a resolved answer.
For sodium that referent-correction or genetic check has not been run. WHO named the J-curve as the reason it commissioned its review, then never returned to it, and excluded a priori the very strata (heart failure, type 1 diabetes) where a low-intake harm is most plausible (World Health Organization, 2012). He and Huang argue the low arm away as measurement error and reverse causation — the right mechanisms — but only as a critique, from salt-reduction advocates, with no bias-corrected or Mendelian-randomization analysis of the low-intake arm (He et al., 2013). An independent modelling consortium simply encoded the disagreement as uncertainty:
«To reflect the uncertainty in existing evidence on optimal level of intake for sodium, 1—5 g per day was considered as the uncertainty range for the optimal level of sodium where less than 2.3 g per day is the intake level of sodium associated with the lowest level of blood pressure in randomised controlled trials and 4—5 g per day is the level of sodium intake associated with the lowest risk of cardiovascular disease in observational studies.» (Afshin et al., 2019)
The blood-pressure-minimising intake (trials, <2.3 g) and the event-minimising intake (observational, 4-5 g) point to different optima — precisely what a real J-curve or a confounded one would produce. Named gap: the wiki holds no Mendelian-randomization or bias-corrected analysis that adjudicates the low-sodium arm. It is unadjudicated, not settled either way.
There is no proven floor: every realistic reduction still buys some blood-pressure fall
Across the studied range the dose-response is monotone or not-estimable — no plateau or knee has been located where a person would actually eat. Larger reductions bought larger falls in He’s meta-regression (roughly 5.8 mmHg systolic per 100 mmol/24h cut), though the author flags it as exploratory and ecological (He et al., 2013). Huang’s larger pool shows the raw slope is null and the dose-response appears only after adjusting for trial duration (about 1.10 mmHg per 50 mmol) — the effect is recovered by the model, not visible in the raw contrast, but the direction survives down to baseline pressures below 120 mmHg (Huang et al., 2020).
The measurement caveat cuts one way here. Short trials underestimate the effect:
«Our review identifies an approximate doubling of the effect of sodium reduction on blood pressure in studies of longer than two weeks’ duration versus shorter studies… Very short term studies of sodium reduction are not a sound basis for drawing conclusions.» (Huang et al., 2020)
Because a large share of the pooled trials ran two weeks or less, the pooled surrogate effect is a lower bound on the sustained one. Practically: measurement error can hide a plateau but cannot manufacture one, so over-shooting a hidden plateau merely fails to help — the operative default is that every realistic reduction still pays (the dose-response shape reasoning).
The people who gain most are those at high blood pressure and high cardiovascular risk
Stratify on baseline risk first, because it does most of the work with no subgroup claim needed. A relative effect (the ~10% per 5 mmHg on events, the ~3-5 mmHg per intervention on pressure) applied to a high absolute risk is a large absolute benefit; the same relative effect on a low risk is small -> Baseline Risk and the Relative-Absolute Split. This is why SSaSS earns a strong recommendation for its own high-risk stratum and not a universal one, and why a lean, normotensive, low-stroke-risk person gets little absolute benefit from the same switch.
- Baseline blood pressure (route a + b). Higher pressure means both a larger relative blood-pressure fall and a higher baseline risk — the two compound. This is the best-supported stratifier.
- Salt-sensitivity (a named gap). The term does not appear in the held guidance; the hypertensive/normotensive split is the closest proxy, and no source isolates a salt-sensitive responder subgroup. Personalising beyond baseline pressure is not yet evidence-supported.
- The potassium contraindication (route c). Raising potassium — including via a salt substitute — is a harm in impaired potassium excretion. SSaSS found no excess hyperkalaemia (RR 1.04, 0.80 to 1.37) but excluded known kidney disease, potassium-sparing diuretics and potassium supplements and did no biochemical screening (Neal et al., 2021). So the reassurance holds only where renal handling is normal; chronic kidney disease remains a real contraindication (Aburto et al., 2013).
Guideline targets mark where the data thin, not a proven cliff-edge
WHO recommends reducing sodium to below 2 g/day (5 g/day salt) as a strong recommendation (World Health Organization, 2012). Treat that number first as the edge of the evidence rather than a curve feature: no trial directly assessed the threshold against a hard outcome, and the claim rests on high-certainty blood-pressure evidence transferred down one level plus a handful of indirect cohort comparisons. Which objective set the exact figure is not separable from the evidence here, but the achievability load is explicit — typical Western intake is 9-10 g/day salt against a 5 g/day target, so the recommendation asks for roughly a halving, population-wide (European Society of Cardiology, 2021) -> Which Objective Moved This Recommendation.
Two guidance families reporting a near-identical ~3:1 hypertensive-to-normotensive gradient (WHO and ESC) look like independent corroboration but are not: both trace to the same He/MacGregor research lineage, so the agreement is shared-source, not a second witness. Read guideline convergence here as one programme reaching two committees.
One thing sodium has that most nutrients lack: it is auditable. Urinary sodium is one of only four recovery biomarkers —
«In addition to urinary nitrogen, urinary potassium, urinary sodium, and energy intake estimated using the doubly labeled water technique are recovery biomarkers.» (Willett, 2012)
— so it yields absolute intake, and a person can in principle be measured against the 2 g/day number in a way a food-frequency-questionnaire nutrient cannot. (A single 24-hour urine still captures only one day, so calibration is fixed but day-to-day variation is not.)
The bottom line
- If your blood pressure or overall cardiovascular risk is high, cut sodium. This is where the proven blood-pressure fall (largest in hypertension) meets a high baseline risk, so the absolute benefit is real.
- Raise potassium from food — fruit, vegetables, pulses — not supplements. It adds a blood-pressure benefit if you are hypertensive and reinforces the sodium effect.
- Consider a potassium-enriched salt substitute if you are older or high-risk and have no kidney disease and take no potassium-sparing drugs. It is the one intervention here with a hard- outcome trial behind it — but that trial changed sodium and potassium together.
- Do not chase a precise gram target if you are lean, normotensive and low-risk. Your absolute gain is small; the large levers are elsewhere.
- If you have chronic kidney disease or take potassium-sparing drugs, do not add a potassium salt substitute without medical supervision — that is the one place the coupled lever turns harmful.
- Hold the low-sodium harm claim as unresolved, in both directions — very low sodium is dangerous is neither established nor disproven, because the low-intake arm has never faced a confounder-immune test.
Read this as appraisal, not a prescription
- The loop is open. Nothing here has been checked against a realized outcome for any individual; the wiki grades whether claims are sound and faithful to their sources, never whether following them makes a specific person better off.
- This appraises; it does not prescribe. Selecting, dosing, screening for contraindications and managing drug interactions require this person’s labs, medications and history, which are out of scope. The potassium contraindication above is a flag to raise with a clinician, not a green light.
- These are population-level estimates. Stratify on baseline risk (well-supported) and on baseline blood pressure; do not personalise further — beyond baseline pressure, the effect- modification evidence (salt-sensitive responders) does not yet exist.
- One axis only. This weighs health outcomes. Salt substitutes, food-processing choices and population salt policy also carry cost, palatability and equity considerations; these are real, and this page names them without pricing them or netting them against the health finding.
Evidence box
Question ’For an adult deciding what to do about dietary sodium and the coupled potassium: effect of sodium reduction, of raising potassium, and of the Na/K ratio on each patient-important outcome (BP surrogate; CV events + mortality endpoints); dose-response shape across the studied range; variation by stratum (baseline BP, salt-sensitivity).’ Evidence included 12 sources — 5 gold, 5 high Overall certainty Medium (see Rating Certainty of Evidence) Source-selection note All sources are gold or high tier. Last updated 2026-08-28 · Independently reviewed: No · Full edit history