Samsung Galaxy Watch Accuracy: The Only DXA-Checked Sensor

A 2022 study ran 28 participants through 24 hours of free-living monitoring, comparing a Samsung Gear Sport smartwatch, PPG-based hardware and firmware that predates the Galaxy Watch branding but uses the same underlying sensing approach, against a medical-grade chest ECG for both heart rate and HRV parameters. During sleep, heart rate and time-domain HRV correlated strongly with the ECG reference, with low mean bias. During waking hours, only average heart rate and average NN interval held up; every other HRV parameter tested, including RMSSD and SDNN, showed weaker correlation and higher error variance.

That is the same nightly-average-versus-daytime split found on other brands' HRV validations elsewhere on this site. A separate, more recent preliminary study specifically tested the Galaxy Watch 5, but its results were not accessible for independent verification during research for this page and are not cited here as a result.

A 2022 evaluation ran a Galaxy Watch Active2 and a Fitbit Charge 4 through five resting or sedentary phases, two low-intensity phases, and three high-intensity exercise phases in 23 healthy adults, referenced against a medically certified Holter ECG. The Galaxy Watch Active2 overestimated heart rate by 3.84 bpm on average, the opposite direction of bias from the Fitbit tested alongside it, which underestimated by 1.66 bpm. Both devices kept mean absolute percentage error under 10 percent overall.

The pattern ran opposite to what might be expected: the Galaxy Watch had higher error during seated rest and sedentary activity (9.4 to 11.6 bpm mean absolute error vs the Fitbit's 6.0 to 8.1 bpm) but performed better than the Fitbit during and after higher-intensity movement, such as squats (6.1 vs 15.7 bpm) and stair climbing (6.9 vs 8.8 bpm). Neither device reached the authors' accuracy bar during seated rest or keyboard typing specifically.

A 2023 polysomnography study put a Galaxy Watch 3 against overnight PSG in 32 healthy adults. Sleep detection was strong: 0.954 sensitivity for identifying sleep, though specificity for identifying wake was considerably weaker at 0.524, the same asymmetry found across most of this device category. Total sleep time was overestimated by an average of 9.5 minutes, with moderate agreement (intraclass correlation 0.640).

Sleep staging was the harder problem. Per-stage sensitivity was 0.695 for light sleep, 0.612 for deep sleep, and 0.598 for REM sleep, and the study describes REM duration specifically as not reliably estimated by the watch. Overall four-stage classification accuracy was 0.651, and the authors note this is comparable to what other consumer wearables achieve on the same test, not a distinctly weaker result for Samsung.

Nothing. No independent study of the composite Energy Score itself, as opposed to the sleep, resting-heart-rate, and HRV inputs it's built from, turned up in this search.

Those individual inputs carry their own separate evidence, reviewed elsewhere on this page, but a composite score is not automatically as accurate as the sum of its validated parts. Whether Samsung's specific weighting of sleep consistency, bed and wake time, and sleeping heart rate and HRV tracks readiness the way it's framed to is untested on the current published record.

A 2022 study running 200 patients (100 in AFib, 100 in sinus rhythm) through a 12-lead ECG plus three smartwatches found the Galaxy Watch Active 3's automated AFib diagnosis (88 percent sensitivity, 81 percent specificity) comparable to an Apple Watch Series 5 (87 percent and 86 percent) and ahead of a Withings device in the same cohort. A separate 2022 study of 163 patients found similar numbers for a Galaxy Watch 3 specifically (97 percent sensitivity, 91 percent specificity), in the same range as an Apple Watch 6 tested alongside it.

The first study also found something the automated-accuracy numbers alone don't show: when cardiologists manually reviewed the raw tracings rather than relying on the algorithm, accuracy for the Galaxy Watch actually dropped (86 percent sensitivity, 76 percent specificity), worse than either the Apple Watch or the Withings device under manual review, because 15 percent of Samsung's tracings were rated uninterpretable by a human expert against 2 to 4 percent for the other two devices. The algorithm and the raw signal it works from are not the same thing to evaluate, and this is the one study in this search that separated them.

A 2025 study of 108 physically active adults (56 female, 52 male) compared a Galaxy Watch5 and a clinical standing analyzer against DXA as the reference standard. The two measures it reports come out very differently, and the gap between them is the finding.

For body fat percentage the watch performed well, at r = 0.93 and Lin's concordance of 0.91 against DXA, with 14.3 percent mean absolute percentage error. It was most accurate in female participants specifically, where concordance held at 0.91 with error falling to 9.19 percent and statistical equivalence supported. Notably the watch beat the clinical analyzer on error for this measure, which recorded 21.1 percent.

For skeletal muscle mass the picture inverts. Correlation stayed high at r = 0.92, but agreement collapsed to a concordance of 0.45 with 20.3 percent error. Correlation and agreement are not the same property: the watch tracks the direction of differences between people while getting the absolute figure wrong. Bland-Altman analysis also found proportional bias, with accuracy degrading in people with higher body fat.

In a controlled study of 18 healthy young adults breathing reduced-oxygen gas mixtures, the Samsung Galaxy Watch 5 stayed within the 4 percent root-mean-square deviation limit that medical pulse oximeters must meet under ISO 80601-2-61, measured against a Masimo Radical-7 clinical oximeter, both at normal oxygen levels and with saturation below 90 percent. The Apple Watch 8 and Withings ScanWatch tested alongside it also met the limit, and the Apple Watch 8 showed the lowest bias of the three (Sensors, 2023).

That is one small study in young, healthy participants on one model, which is why the rating is Emerging rather than higher. It says nothing yet about other Galaxy Watch generations, older users, or people with lung or circulatory conditions.

The general pulse-oximetry literature documents real accuracy problems this feature would inherit regardless of brand, including skin-pigmentation bias and reduced accuracy from motion artifact or poor peripheral circulation. Whether Samsung's implementation handles those conditions well is unknown rather than good or bad on the current published record.

No independent validation of Samsung's VO2 max estimate specifically turned up in a dedicated search.