How to Lower Blood Pressure Naturally: What Each Thing Is Actually Worth, in mmHg

The best-graded evidence for lowering blood pressure without a prescription is for a diet, and it does not come in a bottle. In a 2020 analysis of 22 non-drug approaches it took about 7 mmHg off systolic pressure, and the reviewers ranked it first. [1] A review of the diet on its own, 30 trials, found about 3. [5]

The exercise that topped a 2023 ranking of 270 trials was isometric training, holding a muscle under load without moving, and in those trials it mostly meant squeezing a handgrip. [2] Newer rankings mostly put it lower; two published in 2026 placed it last. [6] [7] One from 2025 put it first, but only for the bottom number, and only in people whose pressure was raised but not yet high. [12]

And the two supplements sold hardest for blood pressure did less, on their average figures, than any kind of exercise in the table.

Two words first, because everything below is measured in them. Systolic is the top number — the pressure while your heart is actually pushing. mmHg is millimetres of mercury, which is just the scale blood pressure has always been measured on; there is no American version of it. A reading of 130 over 80 means systolic 130, diastolic 80.

The whole thing on one page

Read down the middle column and the shape of the advice industry becomes obvious: the things sold in a bottle are at the bottom.

The best-graded evidence is for a diet from 1997

In 2020 a Bayesian network meta-analysis went through 60,166 articles, kept 120, and compared 22 different non-drug interventions across 14,923 people. [1]

The reviewers put the DASH diet first: 6.97 mmHg off systolic (the plausible range ran 4.50 to 9.47) and 3.54 off diastolic, on evidence they graded high quality, the top rung and rarer than it sounds; only meditation shared it. [1] First is their reading of rank and quality together, not of size. On the ranking statistic alone, tai chi came first, on two trials graded very low quality. Three approaches had bigger estimates than DASH, and aerobic exercise (6.60) and a lower-calorie diet (6.50) sat within half a point. [1]

Most of DASH’s evidence in that analysis is one trial. Five of its nine DASH entries are reports from the original 1997 feeding study, in which the food was supplied, some of it donated by food companies, and sodium was the same on every diet, about 3,000 mg a day. [11] Two more entries are reports of one 55-person trial. That is four trials and about 405 people, where the analysis counted 1,022 (our count, from the review’s own tables). [1] A 2020 review of DASH on its own, 30 trials and 5,545 people, found 3.2 mmHg, and the effect did not differ between people with and without high blood pressure. [5]

DASH stands for Dietary Approaches to Stop Hypertension. It is vegetables, fruit, wholegrains, low-fat dairy, less salt, less red meat and less sugar. It is thirty years old, it is free, and there is nothing in it to put in a bottle. It is still sold, as books: one 2011 DASH book is subtitled “Proven to Lower Blood Pressure and Cholesterol Without Medication”, which is the claim this page checks.

Also effective in the same analysis, at moderate-to-high quality: aerobic exercise, isometric training, low-sodium and high-potassium salt, breathing control, meditation, and in people carrying extra weight, a lower-calorie diet. [1] One caution on that salt: the 2025 US guideline calls potassium-based salt substitutes useful except for people with kidney disease or on drugs that make the body hold on to potassium, some water pills among them, where it advises extra monitoring. Ask before switching. [9]

The exercise result: isometric training topped one ranking, and most newer ones disagree

A 2023 network meta-analysis pooled 270 randomised trials and 15,827 people to rank exercise modes against each other. [2]

Everything worked. Aerobic training took off 4.49 systolic, dynamic resistance 4.55, intervals 4.08, and combining aerobic and resistance work 6.04. [2]

Isometric training took off 8.24. [2] Isometric means holding a position under load rather than moving through a rep, and in this review it mostly meant squeezing a handgrip: 17 of its 24 isometric groups did that, from 19 of the 270 trials. When the reviewers looked at which specific version did most, the wall sit came out top for systolic pressure, at 10.47 on four small trials, each co-written by at least one of the review’s own authors. It beat walking and aerobic intervals by more than chance would explain; against running, lifting and sprint intervals the gap could have been chance. Running came out top for the diastolic number. [2]

A note on how confident to be about that. The ranking uses a statistic called SUCRA, which is simply the probability that a treatment is among the better ones — 100% would mean it beat everything in every simulation, 50% is middling. Isometric scored 98.3%, combined training 75.7%, and the rest sat between 39% and 47%. [2] Isometric training’s lead over aerobic, interval and resistance training was bigger than chance would explain; its lead over combined training was not. It is still an ordering built from far fewer isometric trials than aerobic ones, and no trial in the analysis put isometric training against another exercise head to head. The reviewers advise caution for both reasons, and note that trials outside the analysis that did make that comparison have given conflicting results. Nor was isometric training a newcomer: a 2013 analysis had also ranked it first, on four trials. [2]

Newer rankings mostly put it lower. In one published in 2026, of 105 trials in people with raised or high blood pressure, combined aerobic-and-resistance training and interval training came top and isometric training came last of seven, level with resistance training. [6] In another, of 159 trials in adults aged 45 and over, isometric training came last of six, and every isometric trial in it was a handgrip one. [7] A third, of 31 trials that measured pressure with a monitor worn for 24 hours, put it joint last of eight for the systolic number, level with resistance training, and its effect there could not be told apart from no exercise at all; its isometric trials were mostly short handgrip programmes. [13]

One newer ranking points the other way. A 2025 analysis of 18 trials and 2,592 people whose pressure was raised but not yet high put isometric training first for the diastolic number, at 4.61 mmHg, and its authors named it and tai chi as the most effective exercises they compared. For the systolic number, tai chi, steady aerobic exercise and interval training all ranked above it. [12] Rankings like these move with the trials that go into them, which is the point: no single exercise has a settled claim to “best”.

Where the analyses disagree, we are not going to pick. The 2020 paper puts DASH first at 6.97 with high-quality evidence; the 2023 paper puts isometric exercise at 8.24. Those two numbers come from different reviews with different comparison groups, run three years apart. But the 2020 analysis did rank both, and in it DASH (6.97) and aerobic exercise (6.60) both ranked ahead of isometric training (5.77). [1]

Salt: the effect is real, and short studies hide it

133 randomised trials, 12,197 people, with sodium intake measured properly through 24-hour urine collection rather than asked about on a questionnaire. [3]

Cutting sodium took 4.26 mmHg off systolic and 2.07 off diastolic, for an average cut of 130 mmol a day: about 3,000 mg of sodium, or 1.3 teaspoons of salt (our arithmetic). [3] That is a big cut: the FDA puts the average American’s whole intake at about 3,400 mg and counts 2,300 mg as about a teaspoon of salt. [10] The effect scaled with the size of the cut. [3]

The interesting finding is about study length. In trials shorter than fifteen days, each 50 mmol cut (about 1,150 mg of sodium, half a teaspoon of salt) bought about 1.05 mmHg. In longer trials the same cut bought 2.13 — roughly double. [3] Short studies underestimate salt. Anyone quoting a two-week trial to tell you salt does not matter is quoting the half of the literature built to miss it.

The effect was also larger in older people, in people whose pressure was already higher, and in non-white populations. [3] Worth knowing that the electrolyte industry sells you the opposite instruction, a claim our electrolytes verdict did not test.

And the supplements

We have already checked the two sold hardest for blood pressure, and both of them work — just not by much.

Magnesium takes about 2 to 3 mmHg off systolic on average: 2.0 across 34 double-blind trials in 2016, and 2.81 across 38 trials in 2025, which found nothing significant in people whose blood pressure was already normal. That is a real effect and it is the most defensible thing on the magnesium shelf. It is also between about half and two-thirds of what the big salt cut above did, about what half a teaspoon of salt a day did in trials longer than two weeks, and a quarter to a third of what isometric training did in the 2023 analysis. In people already taking blood-pressure medication, the 2025 analysis found more: about 7.7 mmHg.

Beetroot juice manages about 3.55 mmHg on resting readings, in trials that ran 2 to 56 days. Reviews that pooled 24-hour readings found no clear drop. On paper that is more than magnesium. Still below the big salt cut, and it is the one on this page you have to keep buying.

“Exercise works as well as the pills” — where that came from, and what it leaves out

You have probably seen this claim. It comes from a 2019 network meta-analysis of 391 randomised trials — 197 of exercise, 194 of blood pressure medication. [4] It is a serious paper. The headline version of it is not.

Three things the paper says that the headline does not.

First: none of the 391 trials compared them directly. [4] Every exercise-versus-drug number in the paper is an indirect comparison, worked out by chaining each against its own control group. That is a legitimate method and it is not the same as a head-to-head. A few small trials outside the paper have run one. In the ones we found that set weeks or months of ordinary training (walking, running, cycling, swimming) against a pill, the exercise never came out ahead. In one, 28 older adults with high blood pressure were randomly given six months of endurance training or a thiazide water pill, and the pill took about twice as much off systolic pressure: 26.6 mmHg against 11.5. [8] Two other kinds of trial point the other way. Tai chi and qigong, the slow Chinese movement-and-breathing practices, have been set against blood-pressure drugs in small trials that the reviewers who pooled them judged mostly poor in quality; pooled, the tai chi trials came out ahead of the drugs on both the top and bottom numbers, and the qigong trials on the bottom one. [15] [16] And in trials of one exercise session against one dose, the session lowered systolic pressure more in the hours that followed, which is a different question from months of treatment. A 2021 review that gathered all of these concluded that exercise alone lowers blood pressure more than medication alone, grading that evidence limited for ordinary exercise and moderate for the Chinese practices; in its own training trials, the drug did as well or better. [14]

Second: in the overall analysis, the medications won — by 3.96 mmHg, with a plausible range of 2.91 to 5.02. [4]

Third: the similarity shows up only in one slice. Among people who actually had high blood pressure, the reviewers found no detectable difference between endurance or resistance exercise and four of the five drug classes they looked at: ACE inhibitors, angiotensin-receptor blockers, beta-blockers and diuretics (water pills). In that slice the comparisons ran both ways: the fifth class, calcium-channel blockers, did better than resistance exercise, and combined endurance-and-resistance training did better than three of the drug classes. [4] But only 56 of the 197 exercise trials enrolled hypertensive people, covering 3,508 participants — while every single medication trial did. [4] The comparison that produces the headline rests on the thinnest part of the exercise evidence. The reviewers also note the exercise trials carried a higher risk of bias, mostly from not being blinded.

So: none of this is a reason to stop taking, skip, or delay blood pressure medication. That is a conversation with a doctor and nothing on this page substitutes for it. The honest reading of that paper is that exercise is worth doing and is under-studied in the people who need it most — not that it replaces a prescription.

What this adds up to

If you want the short version:

— The diet has the best-graded evidence: about 3 to 7 mmHg depending on the review, and nothing in it comes in a bottle. [1] [5]
— Any exercise works; the newest rankings put combined aerobic-and-resistance training near the top, and the training itself is worth having for other reasons. [2] [6] [7]
— A big salt cut, about 1.3 teaspoons a day, did one and a half to two times what a magnesium capsule does; half a teaspoon, held for more than two weeks, did about as much as the capsule, and short trials undersell salt. [3]
— The two supplements did less than any exercise on their average figures, though magnesium did more in people already on blood-pressure medication.

Three levers this page has not checked are on the US guideline’s list, with its rough figures for people with high blood pressure: losing weight (about 1 mmHg off systolic for every 2.2 lb, or 1 kg, lost; 6 to 8 for a sustained loss of 5% of body weight or more), more potassium (about 6) and less alcohol (4 to 6). The guideline warns, as this page does, that its figures should not be compared with one another directly. [9]

And the part that is genuinely important: high blood pressure usually has no symptoms at all. If you do not know your number, none of this ranking is your first step — getting it measured is, and if it is high, that belongs with a clinician rather than a shopping list. How we grade evidence like this, and the simplest version of moving more.