Two policy readings of the sodium -> cardiovascular evidence that give opposite decisions for ~90% of the population. The held pole (WHO 2012, the He/MacGregor/Cochrane lineage) recommends a population-wide reduction toward <2 g/day sodium. The PURE pole (Mente 2016) recommends reduction only for hypertensives who also eat a lot of salt, on the strength of a directly-observed J-curve on hard outcomes. The held evidence sits on Sodium Intake and Blood Pressure; this page is the joined issue. It does not resolve — the deciding check exists on neither side — and the value is in seeing that the two literatures partly measure different things over different ranges, so the raw clash overstates the real disagreement while the policy conclusions genuinely conflict.
Both positions in their own terms
- Population-wide (WHO 2012 + He/MacGregor + Cochrane lineage). Sodium -> BP is a high-certainty, roughly linear dose-response with no plateau in the studied range; WHO transfers that surrogate evidence one certainty level down to hard outcomes (Annex 8) and issues a strong recommendation to reduce to <2 g/day for everyone. The observed low-sodium-harm signal is read as artifact — He 2013 attributes it to measurement error, confounding, and reverse causality, and Huang 2020 to «biased estimation of sodium intake» -> Sodium Intake and Blood Pressure.
- Targeted (PURE / Mente 2016). In a pooled analysis of 133,118 people from 49 countries (three cohorts; ~10,000 events; median 4.2 y), the sodium -> hard-outcome curve is J-shaped: risk lowest at 4-5 g/day and higher at both <3 g/day and >7 g/day, with the high-intake harm confined to hypertensives. Mente concludes sodium reduction should be aimed only at hypertensives with high intake — only ~10% of the studied population had both hypertension and >6 g/day intake — and argues against a population-wide approach except in very-high-intake regions. (Mente et al., 2016)
Parameter table — is this the same quantity? (op-weave 2a)
Built before any prose comparing the poles. The columns are matched claims about the sodium -> outcome relationship; the last column is the joined-issue test.
| Parameter | Population-wide pole | PURE / Mente pole | Same quantity? |
|---|---|---|---|
| Primary outcome | BP (surrogate, HIGH certainty), transferred to CVD one level down | death + MI + stroke + HF (hard, DIRECTLY observed) | NO — surrogate-transfer vs direct hard outcome |
| Exposure measured | achieved reduction in RCTs (24h urine / randomised contrast) | estimated habitual level from a single fasting spot urine + Kawasaki formula | NO — a randomised reduction is not an estimated habitual intake |
| Intake range covered | RCTs mostly reduce ~9-10 -> ~5-6 g/day salt; low arm barely reached | the harm signal lives at <3 g/day sodium, largely OUTSIDE the RCT range | NO — the disputed low arm is where the RCTs have little data |
| Shape claimed | linear/monotone over studied range, no plateau shown | J (nadir 4-5 g/day), both arms elevated | the shapes describe different ranges — not directly comparable |
| Design / confounding | RCT for BP (randomised); cohort for hard outcomes (VERY LOW) | cohort throughout — confounding + reverse causation live | both lean on observational data for the hard-outcome claim |
| Policy conclusion | reduce everyone to <2 g/day (strong) | reduce only hypertensives with high intake | the genuine, joined conflict |
Is the issue joined? — the not-joined checks
- (i) Same observable, different words? No. The poles predict opposite signs for the sodium -> CVD association at <3 g/day (population-wide: lower is better, transferred from BP; PURE: lower is worse, directly observed). A real contradiction, not a verbal one.
- (ii) Different scope / horizon / unit? Partly, and this is the payoff. The poles are not measuring the same exposure (a randomised reduction vs a spot-urine-estimated habitual intake), the population-wide pole’s hard-outcome claim is transferred from a surrogate while PURE’s is direct, and the disputed low arm sits outside the RCT-studied range. So the two literatures partly talk past each other — but they are matched on the one thing a policy needs: should the whole population reduce sodium?, where they give opposite answers. Joined on the decision.
- (iii) Cherry-picking part of a multi-position author? Checked against WHO’s full position. WHO does not assert monotone benefit on hard outcomes with confidence — it rates that evidence VERY LOW, names the J-shaped hypothesis as the reason for its review, then never adjudicates it, and excludes a priori the strata where a lower-arm harm is most plausible -> Sodium Intake and Blood Pressure. So the population-wide pole’s hard-outcome leg rests on surrogate-transfer plus the He/Huang advocacy dismissal of the J, not on a confident direct finding. Represented at that width, the issue is still joined on policy.
Verdict: JOINED on the decision, but asymmetric in evidence type and non-identical in exposure. The disagreement is real and consequential, and it is smaller than it looks on the shape question (the poles describe different ranges) while exactly as large as it looks on the policy question.
The hidden insight — one dispute is actually three stacked ones
The sodium-policy clash is not a single disagreement about a number; it is three, and none is settled:
- Exposure measurement. A randomised sodium reduction (population-wide pole) is not the same exposure as a habitual intake estimated from one spot urine (PURE). Spot-urine + Kawasaki recovers an absolute intake but from a single day, and it is error-prone precisely at the extremes — the He/Huang critique’s «biased estimation of sodium intake» — which is where PURE’s J-arms live. Measurement error that is unequal across the range can bend a flat relationship into a U with no bias and no confounder -> Measurement Error in Dietary Assessment, The U-Shaped Association Artifact.
- Evidence type. Surrogate-BP-RCT-transferred-to-outcomes (population-wide pole) vs direct-hard-outcome-observational (PURE). The first is randomised but on a surrogate; the second is on the real endpoint but confounding-prone. Neither is a randomised hard-outcome trial of sodium level -> Surrogate Outcomes.
- Intake range. The RCTs barely reach <3 g/day, so they cannot see the low arm PURE reports; PURE observes it but cannot randomise it. The literatures are strongest in non-overlapping ranges.
The decisive check exists on neither side. A confounder-immune adjudication of the low arm on hard outcomes — a referent-correction or a genetic/Mendelian-randomization instrument, or a large long-term RCT of sodium level (not just reduction) — is what would settle it, and the corpus holds none. Mente himself concedes that observational analysis cannot definitively prove causality and calls for such RCTs.
A cross-source note on the targeting rationale — not an adjudication
Part of Mente’s case for targeting rests on a premise about blood-pressure lowering: citing pre-2016 meta-analyses and HOPE-3, he argues BP-lowering benefit is observed only «in those with a systolic BP of >140 mm Hg». The wiki holds a later, larger source that reaches the opposite conclusion on that specific premise: BPLTTC 2021 (IPD-MA, 344,716 participants) finds a 5 mmHg SBP reduction cuts major CV events by ~10% «irrespective of previous diagnoses of cardiovascular disease, and even at normal or high-normal blood pressure values», with «in no subgroup was the HR above 1·0» -> Blood Pressure Lowering and Cardiovascular Events. This weakens the targeting rationale’s BP-lowering premise (evidence-base / lag, telos divergence class 2/4) — but it does not settle the tension, because PURE’s low-arm signal is a direct sodium -> CVD observation adjusted for BP and still present, which is a separate claim from does lowering BP help below 140. The shape dispute (points 1-3 above) stands; only one supporting argument for Pole B is undercut. (Blood Pressure Lowering Treatment Trialists Collaboration, 2021; inferred from Mente et al., 2016)
Where this leaves the decision
For a hypertensive with high intake, both poles agree: reduce sodium (large BP effect in this stratum; the high-intake harm signal is concentrated here). For a normotensive at moderate intake (3-6 g/day), the poles diverge and the evidence is genuinely unsettled — the population-wide pole offers a surrogate-transferred benefit, PURE offers a direct-but-unadjudicated low-arm harm signal, and neither has the confounder-immune check. The honest fabric state is insufficient evidence on the low arm, not a resolved recommendation -> Sodium Intake and Blood Pressure.