The previous version of this page did something unusual and good: it gave the reader a formula, a threshold, and spreadsheet instructions. Collect five standardised systolic readings, compute the standard deviation, divide by the mean, compare to SD 6.72 mmHg or CV 9.05%.
Then it stopped. It said no intervention is formally established to target variability, and ended.
That is three steps of a four-step procedure. Compute, compare, interpret — and then nothing. This rebuild supplies the fourth step, and it also says how firmly each of the first three stands, because one of them is softer than two decimal places make it look.
Two things to establish before the arithmetic:
What the number is worth. Per one standard deviation of systolic variability: myocardial infarction or coronary heart disease HR 1.19, cardiovascular death HR 1.21, stroke HR 1.18, heart failure HR 1.14.
What the number connects to. A meta-analysis of 389 trials found that calcium-channel blockers and diuretics reduce visit-to-visit variability, while beta-blockers, ACE inhibitors and ARBs increase it — and that in one major trial, adjusting for variability completely explained the outcome difference between two drug regimens.
That second finding is the missing fourth step. It is also from 2010, it has not changed any guideline, and the article will be explicit about both.
A note on the tables. Every row states not just the finding but how much weight it carries — the number of studies, the heterogeneity, or the age of the evidence and whether guidelines moved. In a subject where nobody measures the thing being discussed, the strength of each claim is part of the claim.
Average Versus Variability: Two Different Questions
| Measure | What it describes | Who tracks it | Evidence strength |
|---|---|---|---|
| Average (mean) BP | Typical arterial pressure | Every clinic, every visit | The basis of all diagnosis, dosing and targets |
| Visit-to-visit variability (VVV) | Spread across weeks or months | Research settings — almost never calculated clinically | Strongest of the variability measures; 49 studies in the 2025 meta-analysis |
| Day-to-day variability | Spread across consecutive home days | Some specialists | Less studied than VVV; the meta-analysis above is VVV |
| 24-hour variability | Swings within a single day | Specialist referral only | Requires ambulatory monitoring |
The two hypothetical patients in the old version make the point better than any definition, so they are worth keeping. Both average 135/85 across six visits in a year.
Patient 1: 134, 136, 135, 133, 137, 135. Patient 2: 118, 152, 126, 148, 121, 145.
Identical averages. A clinic records the same number for both and treats them identically. Patient 2’s systolic range is 34 mmHg wide.
What Variability Actually Predicts
| Outcome | Hazard ratio per 1 SD of systolic variability | 95% CI |
|---|---|---|
| Myocardial infarction or coronary heart disease | 1.19 | 1.09–1.30 |
| Cardiovascular mortality | 1.21 | 1.13–1.30 |
| Stroke (fatal and non-fatal) | 1.18 | 1.10–1.26 |
| Heart failure | 1.14 | 1.08–1.21 |
| Composite cardiovascular disease | 1.11 | 1.07–1.14 |
| Atrial fibrillation | 1.09 | 1.00–1.19 — lower bound touches 1 |
From a 2025 systematic review and dose-response meta-analysis in the European Journal of Preventive Cardiology: 49 studies, with individual study sizes ranging from 457 to 8,063,922 participants.
Diastolic variability was studied in 19 of those studies and showed a similar pattern — composite cardiovascular disease HR 1.27 (1.07–1.51), stroke 1.15 (1.02–1.28), heart failure 1.20 (1.09–1.32). Atrial fibrillation at 1.13 (0.88–1.46) was not significant.
So the old version’s claim that variability predicts events independently of the average holds up, and now has figures attached. Note the modesty of them: these are hazard ratios in the 1.1 to 1.2 range, not the two- and three-fold figures that appear elsewhere in cardiology. Variability is a real signal of moderate size.
How Firm Is the 6.72 Threshold?
This is the part the old version could not have known without reading the paper closely, and it changes how to use the number.
| Feature of the threshold | Detail |
|---|---|
| SD threshold | 6.72 mmHg — above this, roughly 10% higher cardiovascular risk |
| CV threshold | 9.05% — same association |
| The reassuring floor | Below SD 4 mmHg or CV 5.65%, hazard ratios were below 1 |
| Shape of the relationship | Positive and non-linear (quadratic) for both SD and CV |
| Per 20 units of SD | HR 1.85 (1.82–1.87) |
| Heterogeneity across studies | I² = 99.03% |
| After removing 12 outlier studies | I² fell to 17.97%, pooled HR 1.13 (1.11–1.15) |
That heterogeneity figure deserves a sentence of its own. An I² of 99% means the included studies disagreed with each other almost completely. The authors addressed it, and removing twelve outliers brought it down to a reasonable 17.97% with a stable pooled hazard ratio of 1.13 — which is the figure to trust. But it means 6.72 is a central estimate from a very noisy literature, not a cliff edge.
The practical consequence: compute your SD, but treat it as a band rather than a line. Below 4 is genuinely reassuring. Above 9 or 10 is worth raising. Between 4 and 9 you are in the region where the evidence itself is uncertain, and a figure of 6.8 versus 6.6 means nothing.
The non-linear shape matters too: risk rises faster at higher variability than at lower, which is why the gap between an SD of 5 and an SD of 7 is worth less than the gap between 12 and 14.
Calculating Your Own
The old version’s method was sound. Here it is with the zones added.
| Step | Action | Why |
|---|---|---|
| 1 | Collect at least 5 systolic readings, days or weeks apart | Fewer gives an unstable estimate |
| 2 | Standardise arm, cuff, time of day, 5 minutes seated | Uncontrolled conditions manufacture false variability |
| 3 | Use systolic values only | The systolic evidence is substantially stronger |
| 4 | Calculate the standard deviation — =STDEV(range) | This is your SD |
| 5 | Divide by the average, multiply by 100 — =STDEV(range)/AVERAGE(range)*100 | This is your CV% |
| 6 | Read the zone below | Not a single cut point |
| Your result | What it means |
|---|---|
| SD below 4 (CV below 5.65%) | Hazard ratios were below 1 in the meta-analysis. Genuinely reassuring |
| SD 4 to 6.72 | Inside the band where the evidence is noisiest. No action implied |
| SD above 6.72 (CV above 9.05%) | The threshold associated with roughly 10% higher cardiovascular risk |
| Systolic swings over 30 mmHg across properly standardised readings | Worth a specific conversation with a doctor |
| Any reading above 180/120 | Not a variability question. Prompt evaluation |
Step 2 is the one that invalidates most home attempts, and it is worth more than the arithmetic. Different arm, different chair, coffee beforehand, no rest period — each of those adds spread that has nothing to do with your arteries. If your readings were not standardised, you have measured your technique, not your variability. The protocol matters so much that it has its own article: How to Measure Blood Pressure at Home Accurately, which covers the error budget in mmHg — an unsupported arm alone adds about 6.5.
On the number of readings: five is a working minimum, and a 2026 preprint on the minimum number of measurements supports it. A preprint is not peer-reviewed evidence, so treat five as pragmatic rather than established. Ten to twelve is better if your monitor stores them.
The Missing Fourth Step: Drug Class
Here is what the old version gestured at — “some drug classes control variability better than others” — with the classes named.
| Drug class | Effect on visit-to-visit variability | Associated stroke prevention |
|---|---|---|
| Calcium-channel blockers | Reduce it the most — described as almost a unique property of the class | Among the best, independent of mean systolic BP |
| Diuretics | Reduce it | Among the best, independent of mean systolic BP |
| Beta-blockers | Increase it, dose-dependently | Least effective for stroke prevention |
| ACE inhibitors and ARBs | Increase variation in systolic BP | — |
From a meta-analysis of 389 trials published in The Lancet in 2010, with the mechanism examined further in a companion Lancet Neurology paper.
The finding that makes this more than a curiosity comes from ASCOT, a large hypertension trial comparing an amlodipine-based regimen with an atenolol-based one. Adjusting for blood pressure variability completely explained the difference in stroke and coronary outcomes between the two. Not partially — the variability difference accounted for it.
Now the three caveats, because they are the reason this has not become a treatment target.
The authors themselves said the results do not prove a causal link, and they did not propose abandoning mean blood pressure. Commentators at the time said guidelines should not yet change.
They have not. This evidence is from 2010 and sixteen years later no guideline sets a variability target or selects drugs on that basis. The old version was correct that no intervention is formally established.
And the sources disagree on diuretics — one account of the Lancet meta-analysis specifies loop diuretics, another non-loop. If this matters to your situation, it is a question for a clinician rather than a detail to act on.
So what the table is actually for: if you have computed a high variability and you are on a beta-blocker, that is a legitimate thing to raise. It is a question about whether your regimen suits you, not a reason to change anything yourself. Never adjust or stop an antihypertensive on the basis of a number you calculated at home.
The Complication the Evidence Contains
Honesty requires one finding that cuts against this article’s own framing.
| In four transient ischaemic attack cohorts, each over 2,000 patients | Stroke risk |
|---|---|
| Patients with the most systolic variation across seven clinic visits | about 6 times more likely to have a stroke |
| Patients with the highest systolic BP across those visits | about 15 times more likely |
Read the second row. In that analysis, maximum blood pressure was a stronger predictor than variability — considerably stronger.
That does not undo the case for variability, which predicted outcomes independently of the mean across 49 studies. But it does mean variability is an addition to the standard picture rather than a replacement for it, and the page’s title overstates slightly: your doctor is tracking the thing that matters most, and not tracking something that matters additionally.
It also suggests a cheap practical move the old version did not make. If you are collecting five standardised readings anyway, note your highest one as well as your spread. It costs nothing and, on that evidence, may carry more information than either the average or the SD.
Why Nobody Is Tracking This
Four reasons, and they are not mysterious.
Nobody has a target. Without a threshold a guideline endorses, there is no number to treat toward, and clinicians reasonably decline to act on a figure with no agreed action attached.
The measurement burden falls on the patient. Visit-to-visit variability needs five or more standardised readings across weeks, which no clinic visit produces. Only home monitoring can generate it, and most home monitoring is not standardised enough to be meaningful.
The arithmetic is not in any workflow. Clinic software records blood pressures; it does not compute standard deviations across them. The calculation is trivial and simply nobody’s job.
And the evidence, while real, is of moderate size and noisy. Hazard ratios of 1.1 to 1.2 with an I² of 99% before outlier removal is a genuine signal in a messy literature — enough to be interesting, not enough to force a change in practice.
What People Actually Run Into
The commonest outcome is manufactured variability. Someone takes five readings in different chairs at different times of day, computes an SD of 11, and concludes they have a problem. What they have is five different measurement conditions. The standardisation step is not a formality; it is most of the measurement, and getting it wrong produces exactly the result you were worried about.
The second is the doctor who is not interested, and it is worth understanding why rather than pushing. A clinician who declines to act on a home-computed standard deviation is not being dismissive — there is no guideline target, no endorsed threshold, and no approved intervention. The productive version of the conversation is narrower: not “my variability is high, what do we do,” but “my readings swing 30 points and I am on a beta-blocker — is this regimen right for me?”
The third is medication timing, which the old version identified correctly and is the most actionable contributor. Taking a dose at wildly different times, or taking a short-acting drug that wears off before the next one, produces real variability — not a measurement artefact. That is fixable, and it is fixable without changing what you take.
And the fourth is over-reading a single high day. Variability is a property of a series. One reading of 158 in a set of six does not mean your pressure is unstable; it means one reading was 158, which may itself be worth noting under the maximum-BP finding above.
Common Mistakes
| The mistake | Why it happens | What to do instead |
|---|---|---|
| Not standardising conditions | Five readings feels like five readings | Same arm, same cuff, same time, 5 minutes seated. Otherwise you measured your technique |
| Treating 6.72 as a cliff | It has two decimal places | I² was 99% before outlier removal. Below 4 is reassuring, above 9–10 is worth raising, between is noise |
| Using fewer than five readings | Faster | Five is the working minimum, and that rests on a preprint. Ten to twelve is better |
| Using diastolic values | There are two numbers | The systolic evidence is substantially stronger |
| Expecting your doctor to act on it | You did the work | There is no guideline target. Ask about drug class and timing instead |
| Changing or stopping medication yourself | The drug-class table is suggestive | The 2010 evidence has not changed any guideline. This is a question, not an instruction |
| Ignoring your highest reading | The average and the SD feel like the answer | Maximum BP was ~15× for stroke in one cohort set, against ~6× for variability |
| Reading the title literally | It is this article’s own headline | Your doctor tracks the thing that matters most. This matters additionally |
When Should You See a Doctor?
Seek prompt evaluation for any reading above 180/120, and emergency care if that reading comes with chest pain, breathlessness, vision change, weakness, numbness or difficulty speaking — see Heart Attack Warning Signs and Stroke Warning Signs: The FAST Method Explained.
Book an appointment, and bring your readings, if: your systolic values swing more than 30 mmHg across properly standardised measurements; you are taking a beta-blocker and your variability is high; your medication timing is inconsistent or your dose seems to wear off before the next one; or your average is at target but your readings are visibly unstable.
The two questions worth asking, both narrow: does my drug class suit a pattern like this, and is my dosing schedule contributing? Both are answerable. “What is my variability target” is not, because there isn’t one.
Frequently Asked Questions
What is blood pressure variability? How much your readings swing from one measurement to the next, as distinct from how high they average. Visit-to-visit variability — spread across weeks or months — has the strongest evidence, and it predicts cardiovascular outcomes independently of the average across 49 studies.
Is high blood pressure variability dangerous? It is associated with modestly higher risk: hazard ratios around 1.11 to 1.21 per standard deviation depending on the outcome. Those are real but moderate figures, from a literature with high heterogeneity. It is a signal worth knowing, not an emergency.
How do I measure my own? Five or more systolic readings under identical conditions, days or weeks apart. Compute the standard deviation, then divide by the mean and multiply by 100 for the coefficient of variation. The standardisation matters more than the arithmetic.
What counts as high? SD above 6.72 mmHg or CV above 9.05% was the threshold associated with roughly 10% higher risk. Below SD 4 or CV 5.65%, hazard ratios were under 1. Treat these as a band, not a line — the underlying studies disagreed substantially.
Can blood pressure variability be reduced? No intervention is formally established to target it. The best available evidence is from 2010: calcium-channel blockers and diuretics reduce visit-to-visit variability, while beta-blockers, ACE inhibitors and ARBs increase it. That has not changed any guideline, so it is a question for your doctor rather than a basis for action.
Why doesn’t my doctor measure this? Because there is no guideline target, no clinic visit generates enough standardised readings, no software computes it, and the evidence — though genuine — is of moderate size and noisy. None of those are unreasonable.
Is variability more important than average blood pressure? No, and this page’s title slightly overstates it. In one set of transient ischaemic attack cohorts, the highest systolic reading was associated with about a fifteen-fold stroke risk against about six-fold for the most variability. Variability adds to the standard picture; it does not replace it.
The Bottom Line: One Calculation, Then One Narrow Question
The useful version of this subject is short.
Do the arithmetic properly or not at all. Five or more systolic readings, same arm, same cuff, same time of day, five minutes seated. Standard deviation, then CV. Unstandardised readings produce a number that describes your chair.
Read it as a zone. Below SD 4 is reassuring. Above 9 or 10 is worth raising. In between, the literature is too noisy for a decimal point to mean anything.
Note your highest reading too. It costs nothing, and on the available evidence it may carry more weight than your spread.
Then ask one narrow question, because the broad one has no answer:
“My systolic readings swing about [X] points across standardised measurements. Does my current drug class suit that pattern, and could my dosing schedule be contributing?”
That question is answerable. It points at the only body of evidence connecting variability to an action — calcium-channel blockers and diuretics reducing it, beta-blockers and renin-angiotensin drugs increasing it — while acknowledging that this evidence is from 2010 and has moved no guideline.
And if the answer is that nothing should change, that is a legitimate answer. Variability is currently a thing worth knowing and not yet a thing worth treating. An article that tells you to compute a number owes you that sentence as much as it owes you the formula.
This article is for general informational purposes and is not a substitute for professional medical advice. Blood pressure variability is not an established treatment target. Never start, stop or adjust any antihypertensive medication on the basis of an article or a calculation you performed at home.
References
- Visit-to-visit blood pressure variability and cardiovascular outcomes: a systematic review and dose-response meta-analysis (49 studies). European Journal of Preventive Cardiology, 33(7):1127
- Same meta-analysis, PubMed record
- Visit-to-visit blood pressure variability and outcome: a traditional and a dose–response meta-analysis. European Journal of Preventive Cardiology
- Webb AJS, Fischer U, Mehta Z, Rothwell PM (2010). Effects of antihypertensive-drug class on interindividual variation in blood pressure and risk of stroke: a systematic review and meta-analysis (389 trials). The Lancet
- Rothwell PM, Howard SC, Dolan E, et al. (2010). Effects of β blockers and calcium-channel blockers on within-individual variability in blood pressure and risk of stroke. The Lancet Neurology
- CardioBrief — Landslide of Lancet Papers Suggest Importance of Blood Pressure Variability (reporting the 2010 papers and the TIA cohort figures)
- Medscape — New Analyses: BP Variability Key in Determining CV Outcomes
- Blood pressure variability and cardiovascular disease: systematic review and meta-analysis. PMC
- [The minimum number of blood pressure measurements needed (2026 preprint,
