HRV (heart rate variability) is the variation in time between successive heartbeats — a small but measurable signal of autonomic nervous system balance. Higher HRV reflects greater parasympathetic (rest-and-digest) tone; lower HRV reflects sympathetic activation, fatigue, or autonomic strain.

In migraine research, HRV has emerged as one of the more consistently useful wearable-derived signals — multiple cohort studies find that HRV drops in the 24 hours before some migraine attacks, likely reflecting the autonomic shifts of the prodromal window.

How HRV is measured

Apple Watch and similar consumer wearables measure HRV as SDNN — the standard deviation of beat-to-beat intervals over some window (typically 5-minute sleeping intervals). Units are milliseconds.

Normal SDNN values are highly individual but typically fall between 20–100 ms for adults. Athletes often run higher; older adults often run lower. Day-to-day variability of 10–20 percent is normal even when nothing’s wrong.

The migraine-HRV finding

Several studies converge on the same pattern:

  • Kapustynska 2024 — XGBoost on Empatica E4 data found nocturnal HRV among the top predictive features for next-day migraine.
  • Empatica/Gottesman 2025 — personalised wearable model including HRV reached AUROC 0.68 for next-day migraine in episodic patients.
  • Earlier HRV-migraine literature — multiple smaller studies show lower HRV during prodromal and ictal phases vs interictal.

The signal is modest in size — adding HRV doesn’t transform a 50% prediction into a 95% prediction. It contributes a few AUC points alongside sleep, weather, and cycle. But it’s consistently positive across studies, which is more than can be said for many migraine predictors.

Personal baseline is the key

Absolute HRV values vary so much between individuals that fixed thresholds don’t work. A 40 ms SDNN is normal for one person and concerning for another.

The pattern that matters is within-person change:

  • Today’s HRV vs your 30-day rolling baseline — drops below ~80 percent are flagged in Hermly’s rule engine.
  • Day-before vs day-of — some research shows the lag pattern (yesterday’s HRV drop predicting today’s attack) matters as much as same-day.

Hermly computes the 30-day baseline from your own HealthKit data — typically takes ~14 days of consistent wearing to stabilise.

Workout-day correction

One specific subtlety: vigorous exercise produces a measurable HRV dip for 24+ hours afterward. This is healthy autonomic response, not autonomic strain.

Without context, the model would misread a post-workout HRV drop as a migraine warning sign. Hermly’s scoring logic explicitly excludes workout days from low-HRV trigger detection so the model doesn’t over-flag athletes.

What “improving HRV” does and doesn’t do

HRV responds to many lifestyle factors:

  • Sleep quality and duration — better sleep raises HRV over weeks.
  • Aerobic exercise conditioning — sustained over months, raises HRV.
  • Stress reduction — meditation, breathing practice, social connection.
  • Alcohol reduction — alcohol acutely lowers HRV; abstinence raises it.

The honest framing: these interventions help migraine through their direct mechanisms — sleep and stress reduction are themselves preventive. HRV is a downstream marker of these improvements, not a separate lever.

“HRV biofeedback” and “HRV training” products typically don’t have meaningful migraine-specific evidence. Improving sleep and exercise patterns has evidence; targeting the HRV number itself does not.

What this isn’t

Not a recommendation that everyone with migraine needs an Apple Watch — the marginal AUC lift, while real, doesn’t justify the cost on its own. If you already have one, Hermly uses HRV alongside the other signals. If you don’t, the iPhone-only path still works.