How Accurate Is One Blood Pressure Reading?

Blood pressure is not a fixed property of a person, like height. It is a continuously regulated variable that changes beat to beat, minute to minute, and across the day. The most direct demonstration of how much this matters came from a secondary analysis of a Veterans Affairs hypertension trial, in which 444 patients produced 111,181 systolic readings over 18 months by three methods: standardised research measurements, routine clinic measurements, and home measurements transmitted electronically.^[1]

The three methods did not agree. At baseline, 28% of the same patients were classified as controlled by their clinic reading, 47% by home readings, and 68% by standardised research measurement. Short-term variability was large and — importantly — similar across all three methods, with a mean within-patient coefficient of variation of 10%.^[1] That is the key point: the noise is not an artefact of sloppy clinic practice. It is in the signal.

The authors then asked the question that matters clinically: from one clinic systolic reading, at what value can you be 80% sure the patient's blood pressure is in or out of control? The answer was that no single clinic systolic reading between 120 and 157 mm Hg could classify a patient with that certainty.^[1] Averaging fixed most of it; the benefit accrued up to about five or six readings.

This does not mean a reading of 180 should be ignored, and it does not mean blood pressure is unknowable. It means the precision of one reading is much lower than the two-decimal confidence with which it is usually reported and reacted to.

Two randomised crossover trials from Johns Hopkins quantified how much a reading moves for reasons that have nothing to do with a person's arteries.

In the ARMS trial, 133 adults had triplicate readings with the arm in three positions, with a fourth desk-supported set used to subtract intrinsic variability. Resting the hand in the lap raised systolic by 3.9 mm Hg (95% CI 2.5–5.2) and diastolic by 4.0 (3.1–5.0). Leaving the arm unsupported at the side raised systolic by 6.5 mm Hg (5.1–7.9) and diastolic by 4.4 (3.4–5.4).^[2] Both are everyday positions in a busy clinic.

The CuffSZ trial randomised 195 adults to readings with an appropriately sized cuff and with a regular cuff regardless of arm size. For people who needed a small cuff, the regular cuff read 3.6 mm Hg low (−5.6 to −1.7). For those needing a large cuff it read 4.8 mm Hg high (3.0–6.6). For those needing an extra-large cuff it read 19.5 mm Hg high (16.1–22.9).^[3] A single "one-size" cuff on a rack is therefore enough, by itself, to move a person across a diagnostic threshold.

Hypertension Canada's 2025 guideline notes that non-standardised office readings run on average 5–10 mm Hg higher than standardised ones.^[4] These errors stack, and they stack in the same direction more often than not.

Blood pressure is normally lowest during sleep, rises on waking, and varies through the daytime with activity, posture and stress. The conventional description of a normal nocturnal fall is a night-to-day systolic ratio of ≥0.80 to <0.90 — roughly a 10–20% dip.^[5]

The pattern carries information. In an international cohort of 7,458 people from Denmark, Belgium, Japan, Sweden, Uruguay and China, followed a median of 9.6 years, night-time blood pressure predicted total, cardiovascular and non-cardiovascular mortality even after adjustment for daytime blood pressure. Adjusted the other way round, daytime pressure lost much of its independent signal.^[5] The authors cautioned that participants whose night-time pressure exceeded their daytime pressure (ratio ≥1, "reverse dippers") were older and died at older ages, and the investigators judged that the excess mortality there probably reflects reverse causality — illness disturbing sleep and the rhythm, rather than the rhythm causing the illness.

The morning rise has its own literature. In 519 older hypertensive patients followed an average of 41 months, those in the top decile of morning surge (≥55 mm Hg rise from the lowest sleeping pressure to the first two hours after waking) had a stroke incidence of 19% versus 7.3% in everyone else, and a relative risk of 2.7 after matching for age and 24-hour pressure.^[6] That was a selected elderly hypertensive population, not a general one.

The paper most often cited for this — a 2018 New England Journal of Medicine analysis of the Spanish Ambulatory Blood Pressure Registry — was retracted in 2020 over data problems.^[7] The registry was then re-analysed by an independent group at Oxford and University College London and republished in The Lancet in 2023, and that paper, not the retracted one, is the evidence to use. Some of its conclusions differ from the withdrawn version, which is exactly why the distinction matters.

In 59,124 primary-care patients followed a median of 9.7 years, 7,174 (12.1%) died, 2,361 (4.0%) from cardiovascular causes. Per one standard deviation, 24-hour systolic pressure was more strongly associated with death from any cause (HR 1.41, 95% CI 1.36–1.47) than clinic systolic pressure was (HR 1.18, 1.13–1.23). When each was adjusted for the other, the asymmetry became stark: 24-hour pressure held its association (HR 1.43, 1.37–1.49) while clinic pressure essentially collapsed (HR 1.04, 1.00–1.09).^[8]

The most striking single figure is the informativeness comparison. Taking clinic systolic pressure as the 100% reference, night-time systolic pressure was 591% as informative about all-cause death and 604% as informative about cardiovascular death.^[8] The number you can only get while asleep is the one that carries most of the signal.

White-coat hypertension means office pressure is elevated while out-of-office pressure is normal; masked hypertension means office pressure is normal while out-of-office pressure is elevated. Hypertension Canada's 2025 guideline puts white-coat hypertension at 15–30% of people with elevated office readings, and masked hypertension at 10–15%.^[4]

In the re-analysed Spanish registry, relative to pressure in the normal range, raised all-cause mortality was seen for masked hypertension (HR 1.24, 1.12–1.37) and sustained hypertension (HR 1.24, 1.15–1.32) but not for white-coat hypertension; the cardiovascular mortality pattern was the same — masked 1.37 (1.15–1.63), sustained 1.38 (1.22–1.55), white-coat not elevated.^[8]

But do not overstate that. A meta-analysis of 27 observational studies comparing 25,786 people with untreated white-coat hypertension or treated white-coat effect against 38,487 normotensive people, followed 3 to 19 years, found that untreated white-coat hypertension was associated with more cardiovascular events (HR 1.36, 1.03–2.00), higher all-cause mortality (1.33, 1.07–1.67) and higher cardiovascular mortality (2.09, 1.23–4.48), while treated white-coat effect was not associated with any of them.^[9] A separate meta-analysis found untreated white-coat hypertension carried a relative risk of 1.33 (1.10–1.62) for cardiovascular outcomes and, strikingly, 2.85 (2.32–3.49) for progressing to sustained hypertension.^[10]

The honest synthesis: white-coat hypertension is lower-risk than sustained hypertension but is not a clean bill of health — it is a reason for follow-up rather than treatment. Masked hypertension is the mirror image and the more dangerous one, because nothing in the clinic flags it. A normal reading in the office does not rule out a problem, and a high reading in the office does not establish one.

Almost no consumer coverage makes this point. The same underlying blood pressure corresponds to different numbers depending on where and when it is measured, because averages taken over a whole day — and especially over the night — are lower than readings taken awake in a clinic.

Using the corresponding-values table from the American College of Cardiology / American Heart Association guideline: a clinic 130/80 corresponds to a home average of 130/80, a daytime ambulatory average of 130/80, a 24-hour ambulatory average of 125/75, and a night-time average of 110/65. A clinic 140/90 corresponds to a home average of 135/85, a daytime average of 135/85, a 24-hour average of 130/80, and a night-time average of 120/70.^[11]

Canada's 2025 guideline simplified this: hypertension is defined as ≥130/80 mm Hg on a validated device under optimal conditions, and out-of-office confirmation — home monitoring or daytime ambulatory monitoring — uses the same ≥130/80 mean.^[4] The 2025 American guideline keeps the same categories for office readings: normal below 120/80, elevated 120–129 systolic with diastolic under 80, stage 1 at 130–139 or 80–89, stage 2 at 140/90 or above.^[12]

The practical consequence: a home average of 128 is not "almost hypertensive by clinic standards" — it is being judged against a home threshold. Comparing a home number to a clinic cut-point, or a 24-hour average to either, is a category error.

Guidelines converge. The US Preventive Services Task Force gives hypertension screening in adults an A grade and specifically advises obtaining blood pressure measurements outside the clinical setting for diagnostic confirmation before treatment begins.^[13] Hypertension Canada recommends out-of-office assessment to confirm the diagnosis, and where that is not feasible, repeating standardised office measurement at a separate visit — noting that this is less specific.^[4]

• A validated automated oscillometric device with an upper-arm cuff. Hypertension Canada notes that globally only about 10% of devices have evidence of validation; in Canada roughly 90% of devices sold in pharmacies are validated versus about 45% sold by online retailers.^[4] Hypertension Canada publishes a list of recommended devices, and independent validation lists exist for the same purpose.
• Correct cuff size for the arm circumference, cuff on a bare arm, lower edge about 3 cm above the elbow crease.^[4]
• Seated, back supported, feet flat, legs uncrossed, arm supported on a desk at heart level — not in the lap, not hanging at the side.^[2]
• Five minutes of quiet rest, then three readings a minute apart, averaged. Standardised automated office measurement taken this way closely approximates daytime ambulatory and home readings.^[4]
• Home monitoring: in duplicate, twice daily, for a week, and the average is what counts.^[4]
• The 2025 American guideline is explicit that cuffless devices, including smartwatches, should not be relied on for accurate blood pressure measurement.^[12]
• Ambulatory monitoring, which samples every 20–30 minutes across the day and night, is the only method that captures the night-time pressure that carries most of the prognostic signal.^[4][8]

Blood pressure belongs alongside other cardiovascular biomarkers in a cardiovascular risk picture, not in isolation. Vascular health also involves the endothelium, explored in our review of the endothelial glycocalyx.