Migraine and HRV: What the Evidence Shows
Editorial: ONDA Life (Yakiv Bilenko) · Updated November 18, 2026

Migraine is classified by its clinical features, not by an HRV score [S8]. A meta-analysis found no significant pooled HRV differences between people with migraine and controls [S1]. Standardised autonomic reflex tests have shown group-level differences, but they are not the same as resting smartwatch HRV [S2]. Wearable forecasts remain early research, with no external validation in the systematic review [S3]. HRV does not diagnose migraine, identify its subtype or reliably predict an individual attack [S1, S3, S8].
Key points
- The main HRV meta-analysis did not find a consistent pooled difference between migraine and control groups [S1].
- Some HRV outcomes had very high between-study heterogeneity, so they do not define one typical migraine pattern [S1].
- Autonomic reflex tests during deep breathing, standing and other challenges are not interchangeable with resting smartwatch HRV [S2].
- The systematic review of wearable forecasts found no external validation and only rare reporting of calibration [S3].
- A wrist-wearable prediction study excluded HRV from its model because daytime ambulatory HRV was considered insufficiently reliable [S4].
- HRV does not diagnose migraine, identify its subtype or reliably predict an individual attack [S1, S3, S8].
What is known about HRV in migraine?
Heart rate variability (HRV) is the beat-to-beat variation in heart rate [S2]. The main meta-analysis found no significant pooled HRV differences between people with migraine and control groups [S1]. The authors noted that differences in recording periods may have affected the results [S1]. Between-study heterogeneity was very high for some HRV measures, so those pooled estimates should not be treated as one typical migraine pattern [S1]. Current evidence does not establish a consistent resting-HRV signature for migraine [S1].
Are resting HRV and autonomic reflex tests the same?
No. Resting HRV describes beat-to-beat variation during a recording, while autonomic reflex tests examine how the cardiovascular system responds to a specific challenge [S2]. These tests include deep breathing, standing, a sustained handgrip and the Valsalva manoeuvre [S2]. A deep-breathing test measures the heart's response to the breathing challenge, rather than a routine resting smartwatch score [S2].
A meta-analysis found lower results on several standardised autonomic tests in migraine groups during headache-free periods [S2]. This is evidence of group-level differences in autonomic responses, not proof that resting HRV is low in every person with migraine [S1] [S2]. The authors also noted important risk-of-bias concerns and limits in the timing of the measurements [S2].
Can a wearable predict an attack?
A 2026 systematic review examined wearable-derived and ambulatory signals used to forecast migraine or headache [S3]. No study in that review reported external validation, and calibration was rarely reported [S3]. The authors concluded that these models remain early proof-of-concept research, not clinically validated alert systems [S3]. They also warned that a model may appear to perform well because some people have more headache days than others, rather than because it can forecast an attack within one person [S3].
That distinction matters. Recognising a person who often has headaches is not the same as predicting when that person will have the next attack [S3]. The review also found no evidence that acting on wearable forecasts improves clinical outcomes [S3].
What do newer wearable studies actually measure?
A wrist-biosensor study measured electrodermal activity, heart rate, skin temperature and activity-related signals while testing migraine prediction models [S4]. Although the device stream included HRV, the researchers excluded HRV from the prediction features because daytime ambulatory HRV was considered insufficiently reliable [S4]. Personalised models performed better than generalised models, but the authors called for larger prospective validation and real-time testing before routine clinical use [S4].
A separate nocturnal pilot combined pulse-rate variability (PRV), electrodermal activity, respiratory rate and sleep measures [S5]. PRV is derived from a pulse signal and is not identical to ECG-derived HRV [S5]. Group averages did not significantly distinguish nights before headache days from headache-free nights, and individual model performance varied considerably [S5]. These studies test combinations of signals; they do not show that HRV alone can predict migraine [S4] [S5].
What do camera-based studies add?
A camera-only feasibility preprint reported a group-level rise in heart rate before attacks, but other signals did not remain different after statistical correction [S6]. Individual prediction was weak in this small study [S6]. Because the report is a preprint and has not been peer reviewed, it should be treated as an early research lead, not evidence of a usable phone-camera migraine alert [S6].
What does the evidence show?
Context-dependent — resting HRV. The available meta-analysis did not find a consistent pooled difference in resting HRV between migraine and control groups, and some measures had very high heterogeneity [S1].
Context-dependent — autonomic reflex tests. Standardised tests found group-level differences during headache-free periods, but the included studies had risk-of-bias and timing limitations [S2].
Emerging — wearable prediction. Wearable and camera-based models show proof of concept, but they lack the validation needed for dependable individual forecasts [S3] [S4] [S5] [S6].
Unknown — clinical benefit. It is not established whether acting on a wearable migraine forecast improves health outcomes [S3].
What does it not tell you?
HRV does not diagnose migraine, identify its subtype or reliably predict an individual attack [S1] [S3] [S8]. A normal reading does not rule out migraine, and an unusual reading does not prove that a migraine is starting [S1] [S3]. PRV from a wrist sensor, ECG-derived HRV and standard autonomic reflex tests are different measurements and should not be treated as interchangeable [S2] [S4] [S5].
Seek urgent medical care for a sudden 'worst-ever' headache, or a headache with weakness, confusion, vision loss or fever [S7].
How does ONDA handle these signals?
ONDA builds a personal baseline from HRV, resting heart rate and breathing rate stored in Apple Health [S9]. ONDA is not a medical device. It does not detect, predict or treat migraine: it describes your own signals, not a migraine screen or an attack forecast. See How ONDA measures and interprets your signals for the app's method [S9].
Educational information, not a diagnosis or medical treatment.
Evidence at a glance
| Claim | Evidence | Limitation |
|---|---|---|
| The main meta-analysis found no significant pooled HRV differences between women or men with migraine and controls. [S1] | Context-dependent | The meta-analysis was part of a broader hormone review; it does not establish that every migraine population or every individual has the same HRV pattern. |
| Differences in recording periods may help explain the lack of a consistent pooled HRV result. [S1] | Context-dependent | The review included varied recording methods and durations; the statement is a methodological caveat, not a recommended protocol. |
| Standardised autonomic reflex tests show group-level differences between migraine patients and healthy controls during headache-free periods. [S2] | Context-dependent | The synthesis covered seven studies and selected standardised tests; it does not describe resting smartwatch HRV or diagnose migraine. |
| The deep-breathing reflex test assesses cardiovagal function by measuring the heart's response during a breathing challenge. [S2] | Context-dependent | A supervised challenge test is not interchangeable with an ordinary resting HRV reading. |
| The autonomic reflex-test evidence has important study-quality and timing limitations. [S2] | Context-dependent | Risk of bias and the timing of a single interictal measurement limit generalisation to the full migraine cycle. |
| The systematic review of wearable prediction models found no external validation and very limited calibration reporting. [S3] | Emerging | The review covered varied wearable and ambulatory signals; it was not an HRV-only review. |
| Prediction-model performance may partly reflect differences between people rather than reliable forecasting within one person. [S3] | Emerging | This is a concern across the reviewed prediction models, not proof about every individual model. |
| A wrist-wearable migraine prediction study excluded HRV from its model features because daytime ambulatory HRV was considered insufficiently reliable. [S4] | Emerging | One small, device-specific study; its multimodal prediction results are not an HRV validation. |
| A nocturnal wearable pilot found no significant group-level difference in average autonomic and sleep measures on nights before headache days versus headache-free days. [S5] | Emerging | Small pilot using personalised models and pulse-derived PRV; results are exploratory and do not establish an individual alert. |
| A camera-only preprint found a group-level heart-rate change before attacks, but other signals did not remain different after correction. [S6] | Emerging | Not peer reviewed; very small feasibility sample; no ECG validation and weak individual prediction. |
| Sudden onset, neurological deficit and systemic symptoms such as fever are warning signs that can point to a secondary headache. [S7] | Context-dependent | A warning-sign review does not diagnose the cause of a headache; urgent clinical assessment is needed for red flags. |
| The International Classification of Headache Disorders defines migraine using clinical attack features rather than an HRV threshold. [S8] | Guideline / expert consensus | A diagnostic classification; it does not evaluate wearable HRV or migraine prediction models. |
| ONDA builds a personal baseline from HRV, resting heart rate and breathing rate stored in Apple Health. [S9] | Established | Describes app behaviour only; it is not evidence for a health benefit or a migraine-related capability. |
Sources
- [S1] Beech et al. (2023). Sex and stress hormone dysregulation as clinical manifestations of hypothalamic function in migraine disorder: A meta-analysis. European Journal of Neuroscience. DOI 10.1111/ejn.16087 · PMID 37452646 · Systematic search and random-effects meta-analysis; recording periods differed across studies; open-access publishing facilitated by La Trobe University; authors declare no conflicts of interest; Crossref: no corrections or updates (checked 2026-11-18).
- [S2] Pavelić et al. (2024). Reconceptualizing autonomic function testing in migraine: a systematic review and meta-analysis. The Journal of Headache and Pain. DOI 10.1186/s10194-024-01758-7 · PMID 38600467 · Systematic review/meta-analysis of standardised autonomic reflex tests, not resting wearable HRV; no funding and no competing interests declared; Crossref: no corrections or updates (checked 2026-11-18).
- [S3] Díaz-Aldana et al. (2026). Clinical performance and applicability of prediction models using wearable-derived physiologic signals to anticipate migraine or headache attacks: A systematic review. Headache. DOI 10.1111/head.70265 · PMID 43000936 · Systematic review of wearable/ambulatory prediction models; search through 18 May 2026; funding and COI were not confirmed in the accessible record; Crossref: no corrections or updates (checked 2026-11-18).
- [S4] Andruškevičius et al. (2026). Migraine attack prediction using wearable biosensor data. Cephalalgia. DOI 10.1177/03331024261492220 · PMID 42789370 · Prospective observational wrist-biosensor study; HRV was excluded from prediction features because of daytime ambulatory reliability concerns; authors report no financial support and no potential conflicts; Crossmark says Document is current; Crossref: no corrections or updates (checked 2026-11-18).
- [S5] Tomalin et al. (2026). Wearable Nocturnal Autonomic and Sleep Biomarkers for Predicting Next-Day Headache and Identifying Nociplastic Pain in Patients with Migraine. Journal of Clinical Medicine. DOI 10.3390/jcm15103802 · PMID 42194763 · Peer-reviewed pilot observational study using nocturnal PRV, EDA, respiratory and sleep metrics; supported by NIH/NINDS and NCATS awards; authors disclosed equity, consultancy, speaking and honoraria relationships; Crossref: no corrections or updates (checked 2026-11-18).
- [S6] Van Gils (2026). Contactless Camera-Based Home Measurement of Pre-Attack Physiology in Migraine: A Longitudinal Feasibility Study. medRxiv. DOI 10.64898/2026.09.13.26362930 · Camera-only longitudinal feasibility preprint; not peer reviewed; a group-level heart-rate change did not translate into useful individual prediction; funding and COI not confirmed; Crossref: no corrections or updates (checked 2026-11-18).
- [S7] Do et al. (2019). Red and orange flags for secondary headaches in clinical practice: SNNOOP10 list. Neurology. DOI 10.1212/WNL.0000000000006697 · PMID 30587518 · Clinical review of warning signs for secondary headache; funding and COI details were not confirmed in the accessible record; Crossref: no corrections or updates (checked 2026-11-18).
- [S8] Headache Classification Committee of the International Headache Society (2018). The International Classification of Headache Disorders, 3rd edition. Cephalalgia. DOI 10.1177/0333102417738202 · PMID 29368949 · International clinical classification for headache disorders; diagnostic criteria are symptom-based and do not use HRV; Crossref: no corrections or updates (checked 2026-11-18).
- [S9] ONDA — product documentation: How ONDA works. How ONDA works. · ONDA product documentation; cited only for the app's own baseline and medical-positioning statements.
Related
- ScienceChronic Pain and HRV: What the Evidence Shows
- ScienceHRV and Heart Rate During Sleep: Why the Night Is the Best Window
- ScienceMenstrual Cycle and HRV: Why Your Baseline Moves
- ScienceRMSSD — What This HRV Metric Reflects, and What It Doesn't
- ScienceHow ONDA Measures and Interprets Your Body's Signals
- GlossaryHeart Rate Variability
How ONDA Science pages are made: every number comes from one checked list of facts, every claim is mapped to its sources and graded by strength of evidence, and sources need a DOI or PMID (manufacturer documentation is used only for device facts).