Kivorent News

Sunday, November 22, 2026

Health techLooking back

February 2026: Belgian study tested heart rate on 10 wearables in heat and cold

Researchers at KU Leuven and Thomas More University tested optical heart rate on 10 wearables in hot, cold and neutral rooms, in a study published February 17, 2026. Temperature barely mattered; stop-start walking did.

Pulse, covering wearables and health sensors
Published , 3 min read

3xerror gap between two Fitbit trackers in the same test

This is a look back at an event from . We wrote it on November 21, 2026, from the sources listed at the end.

What to know

  • The study, in JMIR Formative Research, compared nine watches and one ring against a Zephyr BioHarness chest strap in 45 healthy adults.
  • Hot (36 °C) and cold (10 °C) rooms did not significantly change accuracy; intermittent walking raised errors for several devices.
  • The Fitbit Charge 6 and Google Pixel Watch 2 agreed best with the reference; the Fitbit Inspire 3 had the largest median error, 14.3 beats per minute.
Median heart rate error across the full protocol (beats per minute)
Fitbit Charge 64.5 bpm
Pixel Watch 24.9 bpm
Apple Watch SE5 bpm
Garmin Vivoactive 55.1 bpm
Garmin Vivosmart 57 bpm
Polar Ignite 39.5 bpm
Fitbit Inspire 314.3 bpm

Source: Gielen et al., JMIR Formative Research, Feb 17, 2026 (Table 1, median absolute error)

The most useful result in this study is the one that didn't happen. Researchers put 10 wearables through a hot, humid room, a cold room and a neutral one, expecting the weather to scramble the light-based sensors. It didn't, by any statistically significant margin. What tripped the devices up was something far more ordinary: walking for 30 seconds, stopping, and walking again.

The paper, from KU Leuven's M3-BIORES group and Thomas More University of Applied Sciences in Belgium, appeared in JMIR Formative Research on February 17, 2026. We didn't cover it then. It's worth a look now because it tested devices people still wear, under conditions closer to real life than a seated lab test.

What they measured

Forty-five healthy adults, aged 21 to 68, wore two devices at a time, one per wrist, with a Zephyr BioHarness 3.0 chest strap as the reference. The strap reads heart rate from the heart's electrical signal; the wrist devices shine light into the skin and infer the pulse from changes in blood volume, the method we explained in how optical heart rate sensors work. Each device got 10 sessions.

Each session ran about 22 minutes per climate: rest, a mental stress task, recovery, steady treadmill walking at 6 km/h, stop-start walking and standing. The rooms were 23 °C, 36 °C at 70% humidity, and 10 °C.

The lineup: Fitbit Charge 6 and Inspire 3, Garmin Vivosmart 5 and Vivoactive 5, Apple Watch SE, Google Pixel Watch 2, Polar Ignite 3 and Pacer, Xiaomi Watch 2 and the Oura Ring Gen 3.

Same brand, very different numbers

The Charge 6 and Pixel Watch 2 agreed best with the chest strap, according to the authors. The Charge 6's median error across the whole protocol was 4.5 beats per minute; the Inspire 3, also a Fitbit, came in at 14.3. That's roughly three times the error (14.3 ÷ 4.5 ≈ 3.2) from one company's two trackers, and the Inspire's average bias was 14.4 beats low, so it mostly under-read.

The Polar watches, the Inspire 3 and the Oura ring sat in a weaker group, with median errors of 9 to 14 beats per minute and errors of 11% to 16%. Apple Watch SE and Garmin Vivoactive 5 landed close to the leaders on average error, but the SE's limits of agreement ran from about 21 beats under to 23 over, wide enough that single readings could be well off.

For anyone reading the brand logo as a guarantee, that Fitbit gap is the lesson. Sensor hardware, fit and the algorithm that cleans up motion noise differ from model to model, and the paper's numbers apply to the firmware it tested.

Where it stops

The activity blocks lasted four to six minutes, the climates were moderate rather than extreme, and the testers were healthy adults. Several models have since been replaced (Pixel Watch 2, Oura Gen 3), so the results describe those devices, not their successors. Our guide to reading these papers covers why limits of agreement matter more than a single ranking.

The authors' advice was to read wearable heart rate in context for clinical use. The stop-start finding is the part that travels: a watch that's fine on a steady walk can lag on the kind of movement most days are made of. That matters most for features built on long-run heart rate, such as resting heart rate trends, where small steady biases add up.

This article is general information about how technology works. It is not medical advice. Talk to a clinician about your own health.

Sources

WearablesHeart health techSmart rings

Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.