Is a Polar H10 chest strap as good as an ECG?
For heart rate and RR intervals, yes, from rest to exhaustion on a bike. For DFA a1 it is close at low intensity, far less so at high intensity. One lab, 25 recreational athletes.
Schaffarczyk and colleagues recorded 25 recreational adults (14 men, 11 women) simultaneously with a Polar H10 chest strap (through the Elite HRV app) and a 12-channel ECG during 3 min of supine rest, an incremental cycling ramp to exhaustion (50 W, then +50 W every 3 min as a continuous ramp) and 3 min of rest afterwards, and analysed both signals the same way in Kubios HRV Premium 3.5.0.
Five participants were excluded from the exercise analysis and two from the post-exercise rest because more than 5% of beats were artefacts.
RR intervals and heart rate agreed almost perfectly at rest (r, concordance and ICC 1.00) and through the ramp (1.00); during exercise the tiny differences were statistically significant (p < 0.001) but RR limits of agreement were +4.3 to −2.8 ms at low intensity and +1.3 to −0.5 ms at high intensity.
DFA a1, the non-linear HRV index used to locate VT1 (about 0.75) and VT2 (about 0.5), agreed less well: concordance 0.84-0.95 at rest and 0.93 in the ramp, limits of agreement +11.6 to −9.9% at low intensity and +58.1 to −40.9% at high intensity, and the strap's values ran about 6% higher at 0.75 and 7.5% higher at 0.5, so both threshold values were crossed at a higher heart rate with the strap.
The authors recommend the H10 for linear HRV and heart rate and caution about DFA a1 at higher intensities and in the uncorrelated range. Our reading: the H10 can stand in for an ECG wherever heart rate or RR intervals are the measure, which also qualifies it as the criterion in REF-1251; for a DFA a1 threshold test it is good near VT1 and unreliable near VT2.
Protocol
Participants: n = 25 healthy adults recruited by word of mouth and online; men n = 14 (40 ± 14 y, 178.1 ± 9.0 cm, 82.2 ± 14.8 kg), women n = 11 (34 ± 10 y, 169.1 ± 4.3 cm, 67.8 ± 9.5 kg); any fitness level, no medical history, medication or recent illness; no caffeine, alcohol, tobacco or vigorous exercise for 24 h. Ethics: University of Hamburg (2021_400).
Test: mechanically braked ergometer (Ergoselect 4 SN) at 60-80 rpm; 3 min at 50 W, then +1 W every 3.6 s (50 W per 3 min) to voluntary exhaustion; breath-by-breath gas exchange (COSMED Quark CPET); 3-min supine rest before (PRE) and after (POST).
Devices: 12-channel ECG CardioPart 12 Blue (AMEDTEC), 500 Hz, lead II used as the comparator; Polar H10, 1000 Hz, Elite HRV app version 5.5.1. H10 firmware not reported.
Processing: Kubios HRV Premium 3.5.0; smoothness-priors detrending (lambda 500); Kubios "automatic" beat correction; DFA a1 window 4-16 beats; rest analysed over a 2-min window (00:30-02:30), exercise in 2-min windows recomputed every 20 s; recordings with more than 5% artefacts excluded.
Statistics: Pearson r, Lin's concordance (rc), ICC3,1, regression with SEE, Bland–Altman bias and limits of agreement (absolute for RR and heart rate, percentage for DFA a1), paired t or Wilcoxon tests, Hodges–Lehmann median differences, Cohen's d.
Outcome of the ramp: peak power 258 ± 49 W, VO2max 40.6 ± 7.6 mL/kg/min, HRmax 178 ± 14 bpm.
Funding and interests: no external funding; the authors declare no conflict of interest. No device-maker involvement is stated. (Bibliographic note, not a declared interest: Rogers is first author of the DFA a1 = 0.75 threshold study REF-873.)
Key Results
| Condition | Variable | H10 vs ECG (mean) | r / rc / ICC | Bland–Altman bias (limits of agreement) |
|---|---|---|---|---|
| Rest before | RR | 987.6 vs 987.5 ms | 1.00 / 1.00 / 1.00 | 0.2 to 0.1 ms (upper 2.6 to 1.3, lower −2.2 to −1.1 ms) |
| Rest before | Heart rate | 63.2 vs 63.2 bpm | 1.00 / 1.00 / 1.00 | — |
| Rest before | DFA a1 | 0.94 vs 0.96 | 0.95 / 0.95 / 0.95 | −2.1% (upper 14.7 to 16.4%, lower −18.9 to −20.5%) |
| Rest after | RR | 686.4 vs 686.3 ms | 1.00 / 1.00 / 1.00 | 0.2 to 0.0 ms (upper 1.6 to 0.5, lower −1.2 to −0.6 ms) |
| Rest after | DFA a1 | 1.38 vs 1.35 | 0.86 / 0.84 / 0.85 | 2.5% (upper 15.6 to 23.7%, lower −10.2 to −19.1%) |
| Ramp | RR | 485.5 vs 485.0 ms | 1.00 / 1.00 / 1.00 | low intensity 0.7 ms (+4.3 to −2.8); high intensity 0.4 ms (+1.3 to −0.5) |
| Ramp | Heart rate | 130.1 vs 130.2 bpm | 1.00 / 1.00 / 1.00 | Hodges–Lehmann −0.14 (−2.75 to 3.04); abstract: bias −0.1 to −0.2 with limits 0.3 to −0.5 (low) and 0.4 to −0.7 (high), printed in "ms" |
| Ramp | DFA a1 | 1.05 vs 1.01 | 0.93 / 0.93 / 0.93 | low intensity 0.9% (+11.6 to −9.9%); high intensity 8.6% (+58.1 to −40.9%) |
Significance: no differences at rest (p > 0.05); every variable differed during the ramp (p < 0.001), with effect sizes d 0.28-0.51.
At the threshold values (Figure 5C): the H10's DFA a1 was about 6% higher at 0.75 and 7.5% higher at 0.5, so both threshold values were crossed at a higher heart rate with the strap; the deviation varied, with some participants in very close agreement. The paper does not give the difference in bpm.
Artefacts: one woman's H10 recording exceeded 5% "artefacts" that the parallel ECG showed to be runs of atrial premature complexes; in an obese participant the ECG lost its signal while the H10 and a Movesense sensor stayed under 1% artefacts.
Numbers that do not add up (our reading): the heart-rate bias and limits of agreement are given only in the abstract and in "ms" — for heart rate they must be bpm; the Results text gives RR numbers only. The paper reports a low- and a high-intensity Bland–Altman split but never defines where it falls. The rest-condition limits are printed as ranges ("upper LoA of 14.7 to 16.4%") whose meaning is not explained.
Mechanisms
Measured: RR intervals and heart rate agreed within a few milliseconds and fractions of a beat from rest to exhaustion; DFA a1 agreement fell as intensity rose and DFA a1 approached the uncorrelated range.
Authors' interpretation: the strap's electrodes and ECG lead II see a different QRS waveform and signal strength, the two systems use different (Pan-Tompkins-like) R-peak rules, and individual factors (ventricular size, subcutaneous fat, skin) and artefact correction shift R-peak timing; because DFA a1 depends on the pattern of RR intervals over time, these small shifts matter for it and not for heart rate.They recommend checking the H10's ECG waveform before a test and rotating the strap or inverting the module for the best QRS signal (Polar's own advice).
Inference: at high heart rates the beat-to-beat variation is a few milliseconds, the same order as the timing differences between the two systems, so a fractal index of that variation is the first thing to break while the beat count stays right.
Inference: a strap that reads DFA a1 higher at a given heart rate places the 0.75 crossing later in the ramp, so an H10-derived VT1 heart rate sits a little above an ECG-derived one; the size in bpm was not reported.
Practical Application
Is a Polar H10 chest strap as good as an ECG? For heart rate and RR intervals, yes, from rest to exhaustion on a bike. For DFA a1 it is close at low intensity, far less so at high intensity. One lab, 25 recreational athletes.
Use the H10 as the field reference for heart rate (B, cycling, n = 25): heart-rate concordance 1.00 and RR limits of agreement within about 4 ms through a ramp; wrist optical readings on the bike are the weaker number (REF-1249).
DFA a1 near VT1 is usable as an estimate (at the paper's undefined "low intensity": bias 0.9%, limits +11.6 to −9.9%), but the strap read about 6% high at 0.75, which moves the VT1 heart rate slightly up; this supports SYNTH-38's HRV-threshold row and SYNTH-53's DFA a1 row as estimates, and qualifies SYNTH-38's "chest strap or ECG" as not quite interchangeable.
Do not anchor VT2 on a chest-strap DFA a1 alone: limits of agreement +58.1 to −40.9% at high intensity, and 7.5% higher readings at 0.5.
Before a DFA a1 ramp, check the signal (the authors' recommendation): view the strap's ECG trace if the app allows it and rotate the strap or flip the module for the best QRS signal; the authors also excluded recordings with more than 5% artefacts (wetting the electrodes is the base's practice, not the paper's).
Some "artefacts" may be real beats: the H10 stores RR intervals, not the ECG, so it cannot tell noise from extra beats (the authors' case showed atrial premature complexes). Frequent irregular beats flagged at rest are a reason to see a doctor, not to re-tighten the strap (Not from this paper; the base's own practice; no diagnosis from a strap).
Limitations
Small, single laboratory: 25 recreational adults, 20 in the exercise analysis after excluding recordings with more than 5% artefacts; nothing was stratified by sex, body composition or age (the authors' limitation).
One protocol and one setup: a cycling ramp indoors, the Elite HRV app and Kubios Premium with automatic correction; running, outdoor movement, sweat and other apps were not tested.
The comparator differs in lead position, so the paper cannot say which device is "right" for DFA a1, only that they differ (the authors' point).
Reporting: heart-rate limits of agreement in "ms" and only in the abstract; the intensity split is undefined; no threshold heart-rate difference in bpm.
Clean recordings only: recordings with more than 5% artefacts were dropped, so the agreement shown applies to clean signals; with more artefacts, DFA a1 in particular is likely to agree worse (inference; the authors show artefact correction producing data loss or HRV bias in single cases).
Interests: no external funding and no conflicts declared; no device-maker involvement is stated.
How this entry was checked and changed
Why this level
test validation study against a laboratory ECG, n = 25 (20 in the exercise analysis), read from the full text (open access, CC BY 4.0). B, not higher: small single-laboratory sample of recreational athletes, the low- and high-intensity split used for the limits of agreement is not defined in the Methods, and the heart-rate limits of agreement appear only in the abstract, printed in "ms"
Full text (checked 2026-10-27)
Twenty-five participants (men: n = 14, age: 40 ± 14 years; women: n = 11, age: 34 ± 10 years); five participants were excluded from exercise analysis and two from resting POST measurement due to artefacts >5% (full text)
Ramp: 50 Watts followed by a 1 watt increase every 3.6 s; Polar H10 sampling rate 1000 Hz with Elite HRV App Version 5.5.1; 12-channel ECG at 500 Hz; Kubios HRV Premium 3.5.0 (full text)
RR during the incremental test: mean difference of 0.7 to 0.4 ms with upper and lower LoA of 4.3 to −2.8 ms during low intensity and 1.3 to −0.5 ms during high intensity (full text)
Heart rate during the ramp: H10 130.1 vs ECG 130.2 bpm; Hodges-Lehmann −0.14 (−2.75 to 3.04); r, rc and ICC3,1 1.00 (full text)
DFA a1: low intensity 0.9% (LoA: 11.6 to −9.9%); high intensity 8.6% (LoA: 58.1 to −40.9%) (full text)
Index values tended to be approximately 6% and 7.5% higher at the fixed values of 0.75 and 0.5 for Polar H10 in comparison to the ECG (full text)
Pmax 258 ± 49 watts, VO2max 40.6 ± 7.6 mL/kg/min, HRmax 178 ± 14 bpm (full text)
This research received no external funding; the authors declare no conflict of interest (full text)