Scientific validation of Sapiens' heart rate recovery analytics

When we tell someone their nervous system spent Tuesday night in physiological stress rather than recovery, that statement is only as trustworthy as the signal underneath it.

At Sapiens we infer stress and recovery from heart rate variability (HRV) — the tiny, beat-to-beat fluctuations in the timing of the heartbeat that reflect how the autonomic nervous system is regulating the body. HRV is a genuinely powerful window into recovery, but it is also fragile: it lives in differences of a few milliseconds between heartbeats, so the way you capture those beats decides whether the conclusion is real or noise.

This article explains the science behind that choice. It covers what HRV measures and why it indexes recovery, why the measurement method (electrocardiogram versus the optical sensor in a typical wrist wearable) is the single biggest determinant of data quality, and what the independent, peer-reviewed evidence says about the tool we use — Firstbeat Life, which records a single-lead ECG. Every claim below is referenced to a primary study, listed in full at the end.

What heart rate variability actually measures

A healthy heart is not a metronome. The interval between consecutive beats — the R-R interval — constantly lengthens and shortens as the two branches of the autonomic nervous system push and pull on the sinoatrial node. The parasympathetic (vagal) branch slows the heart and dominates during rest and recovery; the sympathetic branch speeds it up during stress, effort and arousal. Because vagal activity changes the heart's rhythm on a beat-to-beat timescale, the variability between beats is a usable proxy for autonomic balance, and specifically for the parasympathetic activity that characterises recovery.^{[1, 2]}

Not every HRV number means the same thing. The metric most tied to vagal (recovery) activity is RMSSD — the root mean square of successive differences between R-R intervals — which is why short-term, parasympathetically driven indices like RMSSD are the workhorse of recovery monitoring rather than, say, raw heart rate alone.^{[2, 3]} This also sets a hard requirement for the hardware: if a metric is built from millisecond differences between adjacent beats, the device has to time each beat to within a few milliseconds. That requirement is exactly where measurement methods diverge.

Why the measurement method matters more than the brand

There are two common ways to capture the heartbeat. An electrocardiogram (ECG) measures the heart's electrical activity directly and marks each beat at the sharp R-wave, the same event a clinical Holter monitor uses. This is the reference standard — the "gold standard" — against which all other HRV methods are compared.^[1] Most wrist and ring wearables instead use photoplethysmography (PPG): an optical sensor that infers the pulse from light reflected by blood flow at the skin. PPG estimates a related but different quantity — pulse rate variability (PRV) — from the arrival of the pulse wave at the periphery rather than the electrical beat at the heart.

The distinction is not academic. A systematic review comparing PPG-derived PRV against ECG concluded that agreement is acceptable at rest, but that physical activity and several mental stressors degrade the agreement "often to an unacceptable extent."^[4] A second systematic review of wearable HRV devices reached the same conclusion: correlation with ECG ranged from very good to excellent at rest but "declined progressively as exercise level increased."^[5] An independent study of consumer and research-grade optical sensors quantified one driver of this: absolute heart-rate error during activity was, on average, about 30% higher than at rest, and accuracy differed meaningfully between devices and between activity types.^[6] (Notably, that same study did not find a statistically significant accuracy gap across skin tones in its sample — a reminder that the robust, well-replicated problem with wrist optical sensors is movement and the indirect nature of the pulse signal, not a single simple cause.)

The practical implication is that the most important decision is not which brand of wearable to buy, but whether the data come from ECG or from peripheral optics. For a measurement whose entire value rests on millisecond precision during everyday life — including walking, commuting and exercise — an ECG signal is the conservative, defensible choice. A narrative review of devices suitable for continuous 24-hour HRV monitoring likewise singled out chest ECG sensors as the most accurate option, outperforming wrist-based alternatives.^[7]

The tool we use: Firstbeat Life

Firstbeat Life records a single-lead ECG from two electrodes on the chest, capturing R-R intervals continuously for several days at a time. The chest-ECG form factor is what lets it sidestep the motion and peripheral-signal limitations described above: it timestamps the electrical beat directly, the same way a clinical monitor does, rather than reconstructing it from a pulse at the wrist. (Device sampling and battery specifications quoted in product materials are manufacturer figures; the claims that matter for trust are the independent ones below.)

The strongest evidence that the underlying ECG signal is clinically usable comes from a domain that demands genuine signal fidelity: arrhythmia detection. In an independent diagnostic-accuracy study, the Firstbeat single-lead chest ECG was recorded alongside a three-lead Holter monitor (the clinical reference) in patients, and an independent arrhythmia algorithm screened the recordings for atrial fibrillation. On a per-patient basis it reached 100% sensitivity and 94.9% specificity (97.2% accuracy); on a per-recording-time basis, 99.6% sensitivity and 98.0% specificity.^[8] Detecting atrial fibrillation requires correctly resolving the timing of individual beats — so this result is, in effect, third-party evidence that the device's raw beat-to-beat signal is accurate enough for medical-grade rhythm analysis, not just wellness estimates.

Does it track real stress and recovery biology?

Accurate beats are necessary but not sufficient; the derived stress and recovery readouts also have to correspond to real physiology. Three independent lines of evidence support that they do.

Against a stress hormone. In a study of 197 participants spanning young and middle-aged healthy adults and patients with cardiometabolic risk factors, researchers induced stress with the Trier Social Stress Test and compared an HRV-based stress index against salivary cortisol, the standard biochemical marker of the stress response. The HRV-based index was the strongest predictor of the cortisol response — a closer correlate of the hormonal stress reaction than heart rate itself — and the relationship held across age and health groups.^[9] In other words, the HRV stress signal moves with the body's actual endocrine stress axis.

Against a known nightly perturbation. Recovery is supposed to happen during sleep, and HRV should register anything that disrupts it. A controlled laboratory dose-response study showed that evening alcohol — even at moderate doses — produces a dose-dependent suppression of vagal (high-frequency) HRV and a shift toward sympathetic dominance across the night.^[10] This is independent confirmation that nocturnal HRV is a sensitive readout of real recovery quality, which is precisely what a multi-night recording is designed to capture.

Against sleep itself. Kadangi autonominis tonusas kinta priklausomai nuo miego stadijų, ŠSD kintamumas (ŠSDK) gali padėti nustatyti miego stadijas. Palyginus su laboratorine polisomnografija sveikiems suaugusiesiems, „Firstbeat“ metodas nustatė budrumą 0,95 jautrumu ir 0,93 bendru tikslumu, o lėtojo (gilaus) miego fazę – 0,72 jautrumu ir 0,91 specifiškumu. Pagrindiniai jo trūkumai buvo įvardyti atvirai ir konkrečiai: jis nepakankamai įvertino REM miego trukmę (vidutiniškai ~18 minučių) ir šiek tiek pervertino budrumo laiką.^[11] Atkūrimo stebėsenai, kai pagrindinis klausimas yra „ar miegas buvo atkuriamasis ir kokio jis buvo lygio“, šio rodiklio visiškai pakanka, o prietaisas aiškiai nurodo, kur jo galimybės yra ribotos.

Kodėl matuojame kelias dienas namuose

Vienkartinis trumpas matavimas yra tik akimirksnio vaizdas sistemos, kuri kinta kas valandą. Nuolatinė stebėsena visą dieną ir kelias dienas iš eilės suteikia stabilesnius ir lengviau atkartojamus ŠSDK duomenis nei trumpi patikrinimai, o ambulatoriniai EKG jutikliai puikiai tinka šiems duomenims fiksuoti įprastoje aplinkoje, o ne laboratorijoje.^[7] Tai yra „Sapiens“ protokolo pagrindas: mes registruojame duomenis kelias dienas iš eilės ir sąmoningai įtraukiame bent vieną savaitgalio naktį. Palyginę atsistatymą darbo dienomis su mažo krūvio naktimi, galime atskirti ūmią, gyvenimo būdo nulemtą įtampą (atsistatymas pagerėja sumažėjus spaudimui) nuo lėtinio, susikaupusio nuovargio (atsistatymas išlieka prastas net ir poilsio dieną) – tai kliniškai reikšmingas skirtumas, kurio neįmanoma nustatyti iš vienos nakties duomenų.

Atvirai apie ribotumus

Tikslumas taip pat reiškia aiškų suvokimą, kuo šis metodas nėra. „Firstbeat Life“ yra mokslinių tyrimų ir sveikatingumo lygio EKG įrašymo įrenginys, o ne diagnostikos priemonė; aukščiau pateiktas prieširdžių virpėjimo rezultatas parodo signalo kokybę, tačiau tai nereiškia, kad prietaisas gali pakeisti klinikinę kardiologiją. HRV pagrįstas miego stadijų nustatymas yra tikslus vertinant būdravimą ir gilų miegą, tačiau mažiau tikslus nustatant REM miego fazę.^[11] Be to, pats HRV yra jautrus tokiems veiksniams kaip amžius, kvėpavimas, vaistai ir ligos, todėl jį būtina interpretuoti atsižvelgiant į kontekstą ir individualiai, lyginant žmogų su jo paties baziniais rodikliais, o ne su populiacijos vidurkiu.^[3] Dėl šių išlygų mes teikiame pirmenybę kelių dienų individualizuotiems matavimams, o ne vienkartiniams įverčiams.

Apibendrinimas

Atsistatymo analizė yra tik tiek patikima, kiek patikimi duomenys apie širdies ritmą, kuriais ji remiasi. Nepriklausomi įrodymai vieningi: EKG yra HRV matavimo etalonas, optiniai riešo jutikliai praranda tikslumą būtent tada, kai gyvenimas tampa aktyvus, o krūtinės srities vieno kanalo EKG, toks kaip „Firstbeat Life“, fiksuoja signalą pakankamai tiksliai, kad atitiktų medicininio lygio aritmijos testus, o iš jo gauti streso ir atsistatymo rodikliai atspindi kortizolio pokyčius, alkoholio sukeltus miego sutrikimus ir polisomnografijos būdu nustatytą miego kokybę. Būtent dėl šio derinio – pagrįsto matavimo metodo ir rezultatų, kuriuos su realiais biologiniais duomenimis patvirtino komercinio suinteresuotumo neturintys mokslininkai – mes jį ir naudojame.

Literatūra

1. Europos kardiologų draugijos ir Šiaurės Amerikos širdies ritmo stimuliavimo ir elektrofiziologijos draugijos darbo grupė. Širdies ritmo kintamumas: matavimo standartai, fiziologinė interpretacija ir klinikinis pritaikymas. Circulation. 1996;93(5):1043–1065. https://doi.org/10.1161/01.CIR.93.5.1043
2. Shaffer F, Ginsberg JP. Širdies ritmo kintamumo rodiklių ir normų apžvalga. Frontiers in Public Health. 2017;5:258. https://doi.org/10.3389/fpubh.2017.00258
3. Laborde S, Mosley E, Thayer JF. Širdies ritmo kintamumas ir širdies vaginis tonusas psichofiziologiniuose tyrimuose – rekomendacijos eksperimentų planavimui, duomenų analizei ir ataskaitų rengimui. Frontiers in Psychology. 2017;8:213. https://doi.org/10.3389/fpsyg.2017.00213
4. Schäfer A, Vagedes J. Kiek tiksliai pulso ritmo kintamumas atspindi širdies ritmo kintamumą? Fotopletizmografijos technologijos ir elektrokardiogramos palyginamųjų tyrimų apžvalga. International Journal of Cardiology. 2013;166(1):15–29. https://doi.org/10.1016/j.ijcard.2012.03.119
5. Georgiou K, Larentzakis AV, Khamis NN, Alsuhaibani GI, Alaska YA, Giallafos EJ. Ar nešiojamieji prietaisai gali tiksliai išmatuoti širdies ritmo kintamumą? Sisteminė apžvalga. Folia Medica. 2018;60(1):7–20. https://doi.org/10.2478/folmed-2018-0012
6. Bent B, Goldstein BA, Kibbe WA, Dunn JP. Nešiojamųjų optinių širdies ritmo jutiklių netikslumo šaltinių tyrimas. npj Digital Medicine. 2020;3:18. https://doi.org/10.1038/s41746-020-0226-6
7. Hinde K, White G, Armstrong N. Nešiojamieji prietaisai, tinkami dvidešimt keturių valandų širdies ritmo kintamumo stebėsenai karinėse populiacijose. Sensors. 2021;21(4):1061. https://doi.org/10.3390/s21041061
8. Santala OE, Halonen J, Martikainen S ir kt. Nuolatinis mHealth pleistro stebėjimas algoritmams pagrįstam prieširdžių virpėjimo aptikimui: galimybių ir diagnostinio tikslumo tyrimas. JMIR Cardio. 2022;6(1):e31230. https://doi.org/10.2196/31230
9. Seipäjärvi SM, Tuomola A, Juurakko J ir kt. Psichosocialinio streso matavimas širdies ritmo kintamumu pagrįstais metodais skirtingose sveikatos ir amžiaus grupėse. Physiological Measurement. 2022;43(5). https://doi.org/10.1088/1361-6579/ac6b7c
10. de Zambotti M, Forouzanfar M, Javitz H ir kt. Vakarinio alkoholio vartojimo poveikis naktinei autonominei ir širdies bei kraujagyslių funkcijai suaugusiems vyrams ir moterims: dozės ir atsako laboratorinis tyrimas. Sleep. 2021;44(1):zsaa135. https://doi.org/10.1093/sleep/zsaa135
11. Kuula L, Pesonen AK. Širdies ritmo kintamumas ir „Firstbeat“ metodas miego stadijoms nustatyti sveikiems jauniems suaugusiesiems: galimybių studija. JMIR mHealth and uHealth. 2021;9(2):e24704. https://doi.org/10.2196/24704