For how we live.
Wrist trackers infer light, deep and REM sleep from movement and pulse. Two validation studies find they track total sleep passably but each misjudges the stage breakdown its own way.
Since 2024 the FDA has cleared over-the-counter CGMs for people who don't use insulin. The first independent accuracy check in adults without diabetes finds the sensors track the trend but miss the exact number.
In a 64-patient randomised trial, home cardiac rehab monitored by a smart ring improved fitness. But both arms wore the ring; what separated them was a weekly call from a clinician.
Rings and watches sold online promise needle-free glucose readings. The FDA says do not use any of them, and even research-grade non-invasive sensors miss the mark by about a fifth.
Wrist and body sensors detect roughly 9 in 10 tonic-clonic seizures. A 192-patient validation against video-EEG found the harder focal seizures still slip past, or bury the alarm in false alerts.
Contactless photoplethysmography turns an ordinary camera into a heart-rate monitor, and pooled studies show it tracks pulse well. Oxygen, respiration and the stress scores layered on top rest on far thinner evidence.
Smart rings and watches measure peripheral skin temperature and report drift from your own baseline, not a thermometer reading. That makes them decent trend sensors and poor fever detectors.
Automatic fall alerts are mostly tested on staged laboratory falls, with real-world validation in older people rare. A 2026 meta-analysis of fall prediction found high specificity but only moderate sensitivity.
Green-light pulse sensors raise a real concern about melanin, but independent tests find the skin-tone effect small next to device and movement errors — and the validation studies barely include darker skin.
In 152 adults tested against two nights of polysomnography, the Galaxy Watch reached an AUROC of 0.94 for moderate-to-severe sleep apnea — with 66.7% specificity at the setting it ships with.
European hypertension specialists found only two wrist-cuff blood pressure watches with validation studies done to established protocols. A sleep group ran a Galaxy Watch against two nights of polysomnography.
Head-to-head against an ECG chest strap, popular optical wrist devices differed substantially, and one graded test found a smartwatch running a couple of beats low with a 28-bpm spread in its limits of agreement.
A National Sleep Foundation position statement and a new national survey land on the same problem: a third of US adults track their sleep, and the numbers they trust most are the ones devices measure worst.
Twenty months of under-mattress sensor data from 95,819 people found that chronic irregularity was common, persistent, and tracked with 46% higher odds of hypertension.
In a free-living comparison, three devices counted steps within about 18% of a research monitor but one overestimated active-calorie burn by 139%. A lab study found step counts collapse when walking is slow.
A meta-analysis of 45 cohorts found risk rising steadily with every 10 beats per minute. A study of 92,457 wearable users found one person's normal can differ from another's by 70 beats.
An umbrella review of 42 reviews found only two wearable measures — heart rate and atrial fibrillation — with comparably strong support. Getting any device into a trial is harder than buying it.
A 2025 meta-analysis of 57 studies found the mortality curve bending at about 7,000 steps a day. The 10,000 figure traces to the trade name of a Japanese step counter sold in 1965.
In a small pilot of older veterans, those walking fewer than about 4,100 steps a day before an operation were more likely to be frail. Most weren't walking enough to recover afterward.
A low-probability alert still sends most recipients to a doctor. Researchers are now naming the flip side of continuous self-monitoring — a device-driven health anxiety they call technohypochondria.
Two 2026 meta-analyses put a number on it: roughly a thousand extra steps daily in older adults and in heart patients. The catch is duration — the gains faded once the intervention stopped.
In 230 adults followed a year, lower heart-rate variability and less moderate-to-vigorous activity at baseline were independently linked to developing hypertension. It is a risk signal, in one unreplicated cohort.
An audit of 17 manufacturers' privacy policies found transparency the weakest area, with wide variation between brands. The data these devices generate largely falls outside HIPAA and GDPR's clinical protections.
Smart-ring studies report flagging COVID nearly three days before symptoms and IBD flares weeks ahead. But two-thirds of that research carries a high risk of bias, and clinical usefulness is largely unproven.
Wrist temperature rises after ovulation, not before it. That makes the watch's retrospective estimate biologically sound and useless as a way to predict the fertile window.
Pooled against the sleep lab, finger-worn devices call sleep versus wake at 87% but fall to the mid-60s on light sleep. An Oura meta-analysis found no significant gap on total sleep time.
A network meta-analysis put pooled accuracy for the fertile window at 0.88, best in the three days around ovulation. A wrist-temperature study hit ovulation within three days about 78% of the time.
Against a laboratory gas-exchange test, the Series 10 underestimated VO2 max by about 6 mL/kg/min on average, with wide limits of agreement. It reads trends better than absolute numbers.
It is a wellness feature, not a cleared pulse oximeter, and it inherits pulse oximetry's known skin-tone bias. A patent fight also stripped it from US watches for about 19 months.
Pooled across dozens of trials, activity trackers add roughly 1,800 steps a day and about 1 kg of weight loss. Effects on blood pressure, cholesterol and blood sugar are small and often not significant.
A February scientific statement declines to recommend the category for clinical decisions. Six weeks earlier, a journal retracted a validation study of a popular consumer model.
A meta-analysis of continuous glucose monitors in people without diabetes found a real effect on blood sugar — but only in those with prediabetes. In healthy users, the evidence for a metabolic benefit was not there.
A small qualitative study finds device dependency, blurred motivation and users who no longer trust their own bodies. It lands as health services keep integrating the same devices.
The FDA cleared it to sort one rhythm from one other. It cannot detect a heart attack, and about one reading in eight comes back inconclusive.
Among watch wearers told their pulse looked irregular, a third had atrial fibrillation confirmed on a follow-up ECG patch. Most who were notified did not.
A 250-person randomised trial paired a wearable tracker with just-in-time energy-management messages. Post-exertional malaise fell in both arms, and the difference between them was trivial.
A UK Biobank analysis finds high step counts do not offset short sleep. A US cohort finds one-minute bursts matter. And in older adults, most step-count differences vanish after adjusting for age and sex.
A review found wearables gave valid maximal-oxygen-uptake estimates in most studies of untrained and recreational exercisers. A separate meta-analysis pooled a 0.83 correlation, with high variability.
FDA-cleared in September 2025, the feature notifies fewer than half of people with hypertension in Apple's own validation. Hypertension specialists warn the bigger risk is false reassurance.
Against an ECG over 536 nights, Oura's heart-rate-variability error was under 8% and Whoop's acceptable, while Garmin and Polar lagged. The recovery scores layered on top stay proprietary and largely unvalidated.
The FDA's revised general wellness policy lets non-invasive devices infer blood pressure, glucose and other physiologic parameters. The regulated line is no longer the measurement — it is the sentence next to it.
A mirror that scores physiological age in 30 seconds. A watch that watches your gait for dementia. The gap between what these products measure and what they imply is the story of the show.
A year of wearable data from 76 subsistence farmers shows adaptation happening in real time. The women in the study had less room to slow down, and their strain readings show it.
A smartwatch bioimpedance sensor tracked body-fat percentage against DXA with about 14% average error — but muscle-mass agreement was weak, and even sitting versus standing shifted the reading.
A meta-analysis of 26 studies put pooled sensitivity at 95% and specificity at 97%. Samsung and Withings devices performed comparably to Apple — but the numbers come from people already suspected of arrhythmia.
Across 111 runners completing up to four trials each, two consumer devices tracked the direction of fluid loss reasonably well but missed the amount by 25% to 33% on average.
One proposes a rate-of-change benchmark from 153 non-diabetic wearers. The other put ten young adults through food, exercise and a laboratory stress test to see what actually moves the line.
PpgAge predicts chronological age from consumer wearable photoplethysmography and tracks disease. The same optical method has documented accuracy problems across skin tones.
Twenty untrained women cycled at four fixed workloads wearing four popular low-cost bands, checked against indirect calorimetry. Every device drifted further from the truth as effort rose.
Google's PH-LLM beat sampled human experts on multiple-choice tests but only matched them on real cases. A new reporting checklist published the same month explains why such claims are hard to compare.
A meta-analysis of pedometer, tracker and smartwatch interventions lasting 12 to 48 weeks found no significant change in systolic or diastolic pressure — nor in weight, glucose or HbA1c.
FDA-cleared in September 2024, the feature caught 89% of severe obstructive sleep apnea in Apple's validation but 43% of moderate cases. It is a nudge toward testing, not a test.
One lab test measured 4.7% sensitivity for falls from a wheelchair. Crash Detection has flooded some 911 centres with automated calls from ski slopes and roller coasters.