The WHO-commissioned SR+MA (Aburto 2013, BMJ) behind WHO’s first potassium guideline — the potassium analog of the WHO sodium evidence, and the electrolyte partner of Sodium Intake and Blood Pressure. 22 RCTs (1606 participants) on BP/lipids/renal/catecholamines + 11 cohorts (127 038) on hard outcomes. The one-sentence result: raising potassium lowers BP in people with hypertension but not in those without, is safe for renal function in people with normal potassium handling, and tracks a lower stroke risk — while the harder CV endpoints stay null and underpowered. (Aburto et al., 2013)
The effect on blood pressure — the HIGH-certainty core
«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)
GRADE high for resting SBP (21 RCTs, n=1892). The un-trimmed pooled estimate (SBP -5.93) carried I2=96%; removing one outlier study brought it to the reported -3.49 at I2=65%, and baseline hypertension status explained most of the remaining heterogeneity — the effect is not a weak average, it is a strong effect in one stratum diluted by a null in another:
«Increased potassium intake significantly reduced systolic and diastolic blood pressure in the 16 studies targeting people with hypertension (systolic blood pressure 5.32 (3.43 to 7.20) mm Hg; diastolic 3.10 (1.66 to 4.53) mm Hg) but not the three studies targeting people without hypertension (systolic 0.09 (−0.77 to 0.95) mm Hg; diastolic (0.56 (−0.42 to 1.55) mm Hg).» (Aburto et al., 2013)
This is a route-b effect-modification finding with an unusually clean signal (transportability’s five routes): the normotensive SBP point estimate is 0.09 mmHg — essentially zero, not merely smaller. Aburto’s own hedge is that the three normotensive trials were short and did not test whether potassium prevents a future BP rise, so not effective in normotensives is established for lowering an already-normal BP, not for long-run prevention.
The hard outcome — stroke reaches significance, the rest do not
The same split the sodium evidence shows: stroke is the one hard cohort endpoint that reaches significance; CVD, CHD and mortality do not.
| Outcome | Effect (95% CI) | Design | GRADE |
|---|---|---|---|
| Resting SBP | 3.49 mmHg lower (1.82 to 5.15) | 21 RCTs, n=1892 | HIGH |
| Incident stroke | RR 0.76 (0.66 to 0.89) | 9 cohort comparisons | LOW (direct) |
| Cardiovascular disease | RR 0.88 (0.70 to 1.11) — NS | 4 cohorts, n=29 067 | VERY LOW |
| Coronary heart disease | RR 0.97 (0.77 to 1.24) — NS | 3 cohorts, n=31 162 | VERY LOW |
| All-cause mortality | RR 1.08 (0.91 to 1.29) — NS | 1 cohort, n=1766 | VERY LOW |
«The meta-analysis of nine cohort studies with nine comparisons detected a protective effect of higher potassium intake on risk of incident stroke (risk ratio 0.76, 0.66 to 0.89).» (Aburto et al., 2013)
The headline blends a direct magnitude with a surrogate-path certainty — a Surrogate Outcomes tell. The conclusion reads «a 24% lower risk of stroke (moderate quality evidence)». But 24% is the direct cohort estimate (RR 0.76), which Table 2 grades LOW (cohort floor, not downgraded); the moderate rating belongs to a different row — the BP-surrogate-transfer path (BP high, downgraded one level for indirectness). So a single sentence pairs the direct magnitude with the surrogate path’s grade. The surrogate transfer is explicit, and it is the same GRADE machinery the sodium review used — literally so, since the boilerplate still names sodium:
«Recognising the limitations of any biomarker, we considered data on change in systolic blood pressure as indirect evidence regarding the effect of sodium intake on risk of cardiovascular disease, stroke, and coronary heart disease.» (Aburto et al., 2013)
The word sodium in a potassium paper’s GRADE paragraph is a copy-carryover from the sodium review — concrete evidence the two WHO reviews share one methodological template (bears on the independence question in the Synthesis). (inferred from Aburto et al., 2013)
Dose — a plateau near 90-120 mmol/day, no monotone gradient
«high quality evidence showed that a higher potassium intake of 90-120 mmol/day reduced blood pressure and was associated with a lower risk of incident stroke. Intake above 120 mmol/day did not seem to have any additional benefit. However, the evidence… did not identify a precise optimal level.» (Aburto et al., 2013)
The largest BP fall was at 90-120 mmol/day (SBP -7.16, 1.91 to 12.41), «without any dose response» and not statistically distinguishable from other intake bands. Aburto’s difference-subgroups (achieved intervention-minus-control potassium: <30 / 30-60 / >60 mmol/day -> SBP -4.89 / -1.97 / -3.01) also show no clear gradient. This is a plateau/no-gradient reading on both axes — but read it with the telos dose-response caution: a flat inter-band contrast built on wide, overlapping intervals is weak evidence of a true plateau, so the defensible statement is a minimum effective region around 90 mmol/day with no demonstrated benefit above 120, WHO’s threshold (>90 mmol/day) at the low knee and the 120 where the data thin, not a demonstrated ceiling. No clean contrast with the sodium curve is available — and stating one would compare different axes: the sodium page’s monotone reading is across magnitude of reduction achieved (He), not achieved intake bands, and even there it is monotone only after multivariable adjustment (Huang’s raw slope is null) and author-flagged exploratory. So potassium-flat-by-intake and sodium-rising-by-reduction are not opposed head-to-head; both are consistent with the same decision default (raise toward ~90-120, reduce sodium, neither curve shows a knee that changes the action; the dose-response shape reasoning).
Safety — null on lipids/catecholamines/renal (HIGH), with one contraindication
Aburto’s novel contribution: the first systematic search for potassium’s adverse effects. All null, all GRADE high — no effect on total cholesterol (-0.12 mmol/L, NS), LDL, HDL, triglycerides, plasma adrenaline/noradrenaline, or serum creatinine. The safety story has a stratum boundary (route-c contraindication):
«People with impaired urinary potassium excretion can be at risk of hyperkalaemia… The risk is confined to those patients, most of whom are under medical supervision, and they were not included in this review. In people without renal impairment, the body is able to efficiently adapt and excrete excess potassium… Intervention trials including potassium consumption as high as 400 mmol/day from food for several weeks… reported no adverse effects.» (Aburto et al., 2013)
So the benefit is scoped to normal renal handling; in CKD or on potassium-sparing drugs the same exposure is a harm. No toxicity from food potassium has been reported (isolated acute cases from supplements only) — which is why Aburto set no upper limit for dietary intake. This is a genuine exposure-stratum interaction of the kind the fabric ranks as decision-relevant, not a blanket safe.
Food is a sufficient vehicle — supplements are not required:
«the vehicle of increased potassium intake can be food and does not have to be a supplement… a beneficial effect of higher potassium intake from food on risk of stroke.» (Aburto et al., 2013)
Fresh fruit, vegetables and pulses are the named sources; food processing strips potassium, and modern intakes (<70-80 mmol/day in many countries) sit far below both the guideline and ancestral (>200 mmol/day) levels. -> Plant Foods
The exposure measure is a recovery biomarker — like sodium, unlike most of the corpus
RCTs were required to estimate intake from 24-hour urinary potassium excretion (conversion factor 1.3). Urinary potassium is one of the four recovery biomarkers Van Dam & Hunter name (with urinary sodium, urinary nitrogen, doubly-labelled water) — so potassium, like sodium and unlike FFQ-measured exposures, yields absolute intake and is auditable against a guideline cutpoint. The cohorts, however, mostly used dietary recall (two were dropped in sensitivity analysis for single-24h-recall exposure, stroke RR held at 0.79). Full taxonomy and why this matters: Measurement Error in Dietary Assessment. (inferred from Aburto et al., 2013)
Synthesis — the electrolyte pair is one lever, not two independent witnesses
Nucleus link. This page orbits Sodium Intake and Blood Pressure (nucleus of the
cardiometabolic-exposures cluster): sodium and potassium are the two arms of one BP/stroke story, and
the Na:K ratio is the joint target neither guideline scores alone.
(inferred from Aburto et al., 2013)
The parallel is structural — and it is NOT independent corroboration
Both electrolytes: (1) move BP most in hypertensives; (2) reach stroke as the one significant hard
cohort outcome while CVD/CHD/mortality stay null; (3) claim their hard-outcome benefit through the
same BP-surrogate transfer (shared WHO GRADE template — see the «sodium» carryover above); (4) are
measured by recovery biomarker; and (5) come from the same CASH/WASH advocacy lineage (Cappuccio
and Elliott here are unpaid CASH/WASH members; He/MacGregor on the sodium side; D’Elia/Strazzullo/
Cappuccio authored the reused stroke cohort meta). So the corpus’s sodium and potassium findings are
two arms of one WHO/CASH programme, not cross-group convergence — the same laundering trap the sodium
page caught for the WHO/ESC/He agreement, one level up. No [E-independent] is claimable between the
two electrolytes’ BP-stroke stories.
(inferred from Aburto et al., 2013)
But the two exposures are NOT mirror images — a genuine asymmetry
Parameter table (op-weave 2a) — potassium-raising vs sodium-reduction on SBP, by BP stratum. These are different exposures, so no cell is a same quantity match; the object of comparison is the stratum pattern, individually sourced:
| Parameter | Potassium raised (Aburto) | Sodium reduced (held on sodium page) | Same quantity? |
|---|---|---|---|
| SBP effect, hypertensive | -5.32 mmHg (3.43-7.20), 16 RCTs | -5.39 (He) / -4.06 (WHO) | NO — opposite exposures; magnitudes comparable |
| SBP effect, normotensive | +0.09 mmHg (−0.77 to 0.95) — null | -2.42 (He, sig) / -1.38 (WHO, sig) | NO — and the pattern differs: potassium null, sodium NOT |
| The one significant hard outcome | stroke RR 0.76 | stroke (WHO cohort RR ~1.24 per higher intake) | parallel structure, opposite direction of intake |
The asymmetry is the finding: sodium reduction lowers BP in both strata (larger in hypertensives); potassium raising lowers BP only in hypertensives (normotensive effect is a flat zero). So cut sodium and raise potassium are not symmetric advice — for a normotensive person the sodium lever still moves BP while the potassium lever (on current evidence) does not, though Aburto flags the normotensive potassium trials as too short to rule out a prevention effect. (inferred from Aburto et al., 2013)
The Na:K ratio is the joint target — asserted beneficial, not graded
«if a person consumes 90 mmol/day or more potassium and the WHO recommended sodium intake of less than 2 g/day, his or her intake would have a molar ratio of sodium to potassium of approximately one to one, a ratio considered beneficial for health.» (Aburto et al., 2013)
Potassium is more effective at higher sodium intake (greatest SBP fall, -6.91, when sodium >4 g/day), but still works at 2-4 g/day — so the two levers reinforce and studies changing both show more benefit than either alone (Aburto’s refs 94/95). Caveat: the 1:1-beneficial claim is sourced to a 2003 WHO/FAO consultation, not graded in this review, and the joint-intervention evidence is two studies, not a meta-analysis — an asserted target, not an evidenced one. This is what the sodium page flagged as explicitly out of scope for WHO 2012; the potassium half now exists, but the ratio itself remains un-adjudicated on either page. -> Surrogate Outcomes (a target is legitimate only if its transmission to a hard outcome is itself evidenced).
The joint Na:K lever now has a hard-outcome RCT — SSaSS [2026-08-04, Neal SSaSS]
The AWAITS this page lodged (a salt-substitute RCT against a hard outcome) is cashed: SSaSS (Neal 2021, NEJM) randomised 20,995 high-risk rural-Chinese adults (prior stroke or age >=60 + high BP) to a 75% NaCl / 25% KCl salt substitute vs regular salt and found lower stroke (RR 0.86, 0.77-0.96), major CV events (RR 0.87, 0.80-0.94) and all-cause death (RR 0.88, 0.82-0.95) over 4.74 y — absolute reductions of ~4.5-7.2 events per 1000 person-years. (Neal et al., 2021)
This moves the joint Na:K target from asserted to partly-evidenced — but only the lever, not the ratio. SSaSS achieved urinary potassium +20.6 mmol and sodium −15.2 mmol together, with SBP −3.34 mmHg. So it is the first hard-outcome RCT of raising K and cutting Na at once, and it points the joint direction the 1:1-ratio claim asserted. But the confound runs in this page’s favour too: just as the sodium page cannot credit the benefit to sodium alone, this page cannot credit it to potassium alone — the two exposures move together and SSaSS used a single formulation, so it says nothing about which ratio is optimal. The Na:K ratio as a graded, dose-optimised target stays open; what is now evidenced is that the combined intervention reduces hard outcomes in a high-risk stratum. (inferred from Neal et al., 2021)
Hyperkalaemia — SSaSS refines the route-c contraindication with RCT data. This page held the CKD contraindication from Aburto’s mechanism (impaired excretion -> hyperkalaemia risk). SSaSS supplies the RCT safety datum: no excess serious hyperkalaemia (3.35 vs 3.30 per 1000 py; RR 1.04, 0.80-1.37; P=0.76) and no excess sudden death — in a population screened to exclude known kidney disease and K-sparing/K-supplement use, with no biochemical prescreening and no serial serum-K measurement. So the contraindication is confirmed, not lifted: safe where renal handling is normal (as Aburto predicted), untested where it is not. (Neal et al., 2021)
The surrogate transfer upgrades through BPLTTC (type-F composite)
Aburto’s hard-outcome warrant is a bare assertion that «blood pressure is recognised as a reliable biomarker». That assertion is bounded and upgraded by Blood Pressure Lowering and Cardiovascular Events: BPLTTC’s 344 716-participant RCT evidence establishes that a 5 mmHg SBP reduction cuts major CV events ~10% (stroke 13%), even in primary prevention. Potassium lowers SBP ~5.3 mmHg in hypertensives — so the composite chain potassium -> BP (RCT) -> events (RCT) is a stronger warrant than Aburto’s proxy claim alone. But the arithmetic does not fully close: BPLTTC’s stroke -13% per 5 mmHg predicts only ~RR 0.87 from a 5.3 mmHg fall, whereas the direct cohort shows RR 0.76 (~twice the BP-mediated effect). The two are not statistically inconsistent (0.87 sits inside the cohort CI 0.66-0.89), but the exposures are unmatched — a modest RCT supplement dose vs a highest-vs-lowest cohort intake contrast — and the excess of the cohort effect over the BP-channel prediction is exactly what a BP-independent effect OR residual confounding would produce. So the chain corroborates direction, not magnitude. It is strongest where potassium moves BP (hypertensives); it does not license a hard-outcome benefit in normotensives, where the BP effect is null. (inferred from Aburto et al., 2013)
What is absent — the gaps that drive the next source
- No absolute risk, no NNT for stroke — the RR 0.76 cannot be ranked against another exposure in the fabric (telos layer 1). Absolute benefit scales with baseline stroke risk (route-a); a high-baseline- risk person gains more. -> Baseline Risk and the Relative-Absolute Split
- Hard-outcome RCTs absent. CVD/CHD/mortality rest on few underpowered cohorts (CVD 4, CHD 3,
mortality 1) — «detecting an effect is difficult even through meta-analysis».
G (needs aggregation): the direct potassium->events effect is a magnitude the current evidence base cannot pin. - The Na:K ratio is partly-adjudicated
[updated 2026-08-04]— the joint lever now has a hard-outcome RCT (SSaSS, above: salt substitute cut stroke/CV events/death in high-risk adults), so raising K + cutting Na together reduces hard outcomes is evidenced. What stays un-adjudicated is the ratio itself — SSaSS used one formulation, so no optimal Na:K ratio and no ratio dose-response is established; the 1:1-beneficial figure is still the ungraded 2003 consultation.AWAITSa graded meta-analysis of the sodium:potassium ratio (or the broader salt-substitute SR, e.g. Jafarnejad 2020 / Greer 2020) against a hard outcome to pin the optimal ratio. - Children: 3 trials, non-significant BP fall, high risk of bias — an evidenced gap, not a null.
- No LMIC hard-outcome data; and Aburto did not examine the mechanism by which potassium lowers BP (explicitly deferred).
(inferred from Aburto et al., 2013)
Limits
- Single primary source (Aburto/WHO 2013); gold-tier SR+MA design, but the hard-outcome arm is cohort evidence at LOW/VERY-LOW certainty and the synthesis above is single-lineage.
- Competing interest, disclosed and material to the interpretive claims. Cappuccio and Elliott are unpaid CASH/WASH members; Aburto was WHO staff; funding included the Kidney Evaluation Association Japan and the governments of Japan and Korea. The RCT-pooled BP effects and null safety findings are hard data taken at face value; the pro-intake framing (the benefits most people in most countries extrapolation, the 1:1-ratio target) is advocacy-shaped and caveated at point of use — symmetric standards, the same discipline applied to the sodium lineage.
- Some small-study bias possible (funnel plots not clear); overall SBP I2=96% before stratifying.