Salt and blood pressure: what the trials show

Thirty-four randomised trials, 3,230 adults, four weeks or longer. A reduction of 4.4g of salt a day lowered systolic blood pressure by 4.18 mmHg, 95% confidence interval 3.18 to 5.18, and diastolic pressure by 2.06 mmHg [1]. That is the number behind the salt advice, and it is an average taken across people whose readings barely moved and people whose readings moved a good deal.

The site's own advice is a quarter teaspoon of salt in water, twice a day, for the first month. That rests on a mechanism, not a trial, and the trials were run in a different population. This is what they measured.

The review the guidance rests on

He, Li and MacGregor pooled 34 randomised trials lasting at least four weeks, with a modest change in salt rather than a drastic one. Mean 24-hour urinary sodium fell by 75 mmol, the equivalent of 4.4g of salt a day, measured in urine rather than estimated from a food diary [1].

Split by baseline pressure, the systolic fall was 5.39 mmHg in people with hypertension and 2.42 mmHg in people without it. Age, ethnic group, blood pressure status and the size of the sodium change together explained 68 per cent of the variation between trials, and each 100 mmol of sodium removed was associated with a 5.8 mmHg adjusted fall [1].

The hormonal response was small and measurable: plasma renin activity rose 0.26 ng/mL/h, aldosterone 73.20 pmol/L and noradrenaline 187 pmol/L. Total cholesterol moved 0.05 mmol/L, on a confidence interval from 0.02 below to 0.11 above, which is to say it did not move [1].

A feeding trial, where somebody else controls the food

DASH-Sodium randomised 412 adults in the United States to a control diet or the DASH diet, then fed each group high, intermediate and low sodium for 30 days at a time, in random order [2]. Nobody had to remember what they ate.

Going from high to intermediate sodium lowered systolic pressure by 2.1 mmHg on the control diet and 1.3 mmHg on DASH. Intermediate to low bought a further 4.6 mmHg on the control diet and 1.7 mmHg on DASH. Control diet at high sodium against DASH at low sodium was 7.1 mmHg lower in people without hypertension and 11.5 mmHg lower in people with it [2].

The size of the sodium effect depends on where you start and on the rest of the plate. On a diet already built from fruit, vegetables and low-fat dairy, the last step down was worth 1.7 mmHg. DASH is not this diet.

Where the effect is about one millimetre

The second Cochrane review asks a blunter question: any low-sodium diet against any high-sodium diet. The 2020 update draws on 195 references [3].

In white participants with normal pressure, systolic pressure fell 1.14 mmHg (0.63 to 1.65 lower) across 95 trials and 5,982 participants, and diastolic pressure did not move. White participants with hypertension fell 5.71 mmHg (4.74 to 6.67 lower) across 88 trials. Black participants with normal pressure fell 4.02 mmHg, and with hypertension 6.64 mmHg. Asian participants with normal pressure fell 1.50 mmHg on a confidence interval spanning zero, 3.09 lower to 0.10 higher, and with hypertension 7.75 mmHg [3].

The two reviews disagree because they ask different questions. He's includes only trials lasting four weeks or more with a modest change, and those moved sodium by 75 mmol a day. Graudal's sets no duration floor, and its white-participant trials moved sodium by 138 mmol a day, from 203 down to 65 [3]. Bigger, faster changes produce a bigger short-term fall and a bigger hormonal counter-response, and that difference in design accounts for much of the gap between 4.18 and 1.14 mmHg.

The one trial with heart attacks and strokes in the outcome

SSaSS randomised 600 villages in rural China, 20,995 people in all, mean age 65.4, with 72.6 per cent having had a stroke and 88.4 per cent with hypertension. Households received a substitute of 75 per cent sodium chloride and 25 per cent potassium chloride, or carried on with ordinary salt, and were followed for a mean of 4.74 years [4]. Stroke occurred at 29.14 against 33.65 events per 1,000 person-years, a rate ratio of 0.86. Major cardiovascular events ran at 49.09 against 56.29, and death from any cause was lower with the substitute as well [4].

That trial changed two things at once, sodium down and potassium up, in communities where mean sodium intake exceeded 5g a day. It is the strongest evidence in this article and the least transferable. A reader in Hampshire is not a 65-year-old in a rural Chinese village, and NICE is explicit that salt substitutes containing potassium chloride are unsuitable for older people, people with diabetes, pregnant women, people with kidney disease and people taking some blood pressure drugs, including ACE inhibitors and angiotensin II receptor blockers [7].

The population data, which cuts both ways

PURE followed 95,767 adults aged 35 to 70 across 369 communities in 18 countries. Mean systolic pressure rose 2.86 mmHg for each additional gram of sodium, but only in communities in the highest third of intake. Cardiovascular outcomes in 82,544 people over a median 8.1 years were associated with sodium only where mean intake was above 5g a day [5].

Single morning urine, community averages, no randomisation. That is the weakest design in this article and it cannot show cause. It does explain why "less is always better" is as unsupported as "salt does not matter".

The question this site actually has to answer

Every trial above was run in people eating an ordinary Western diet, some with hypertension and some without. Nobody has randomised people on a very-low-carbohydrate diet to two sodium intakes. That trial does not exist, so the advice to salt food in the first month is a judgement call rather than a finding.

What does exist is the mechanism. Insulin reduces sodium excretion at the kidney. In six healthy men, with blood glucose held at the fasting level, an insulin infusion raising plasma insulin to between 98 and 193 µU/mL cut urinary sodium from 401 to 213 µeq/min, with no change in aldosterone [6]. Six men in a laboratory in 1975 is a long way from a salt cellar in a kitchen, and the gap between them is where the week-one advice lives.

The measured part is narrower and more useful. A clinic reading of 150/95 puts you in the group where the trials found a systolic fall of roughly 5 mmHg. A reading of 118/76 puts you in the group where the fall was 1.14 mmHg, which a single cuff reading cannot tell apart from no change.

What 4.4g looks like, and how you would know

Salt is about 40 per cent sodium by mass, so 4.4g of salt is 1,760mg of sodium, and the NHS maximum for an adult is 6g of salt a day, around one level teaspoon, counting what is already in the food [8]. The trial's average reduction is roughly three-quarters of a teaspoon, every day, for a month or longer.

A change of that size is not something a single home reading can resolve. NICE asks for a clinic reading to be confirmed by ambulatory or home monitoring before a diagnosis is made, and its stage 1 threshold is a clinic reading of 140/90 to 159/99 with a daytime average of 135/85 to 149/94 [7]. That is the number to compare against, not the one taken after you have walked up the stairs.

The 4.18 mmHg in the pooled analysis is real, and so is the 1.14 mmHg in the subgroup of white participants with normal pressure. SACN's 6g a day for adults came out of that literature and has not been revised since [9].