What GLP-1s Do to Your Wearable Metrics

Last updated: September 3, 2026

If you started semaglutide or tirzepatide and your resting heart rate climbed while your HRV fell, that is the expected pattern, not a warning sign of overtraining. Across 12 randomised trials the average heart-rate rise is +3.47 bpm, and the FDA label for Wegovy puts the mean increase at 1 to 4 bpm. A 12-week study of 66 wearable users found RHR +3 bpm and HRV -6 ms. The part almost nobody explains is the mechanism: a 2024 study showed GLP-1 raises heart rate by acting directly on the pacemaker cells of the sinus node, not through the nervous system. Your recovery score does not know that, and reads the same numbers as stress.

Curious what else moves your resting heart rate? The resting heart rate factor explorer ranks the evidence-backed causes.

This is not medical advice and nothing here is a reason to change how you take a prescribed medication. The Wegovy label instructs prescribers to monitor heart rate and asks patients to report palpitations or a racing heart at rest. If that is you, that conversation belongs with your prescriber, not with a blog post. What follows is what the trials measured, so you can read your own data with less alarm and more accuracy.

I went looking for this because the explainers I could find all described the heart-rate rise as sympathetic activation, and the primary literature says something different and considerably more interesting.

What actually happens to your numbers

Metric	Direction	Best evidence
Resting heart rate	Up, typically a few bpm	Mean 1 to 4 bpm (Wegovy FDA label); +3.47 bpm pooled across 12 RCTs; +3 bpm in a 12-week wearable cohort
HRV	Down, but the evidence is split	-6 ms in the wearable cohort; one crossover RCT found SDNN and RMSSD both fell, another RCT found no change at all
Sleep apnea (AHI)	Sharply down	-27.4 and -30.4 events/hour in the two SURMOUNT-OSA trials, a 55% to 63% reduction
Self-reported sleep quality	Up, in people with sleep apnea	PROMIS sleep impairment improved ~7.5 points, Epworth Sleepiness -1.33 to -2.18
Body weight	Down	90% of wearable-cohort users lost 5% or more in 12 weeks, mean loss 10%
Physical activity	Up on average, with a dosing-day dip	+30 min/week on average, but lowest step counts on dosing days
VO2 max (wearable-estimated)	Probably up, partly for arithmetic reasons	No trial has measured it. See the section below

The heart-rate rise is real, and it varies a lot by drug

A 2026 network meta-analysis pooled 12 placebo-controlled RCTs in non-diabetic adults with overweight or obesity. The per-drug increases versus placebo are not interchangeable.

Drug	Mean heart-rate increase vs placebo
Tirzepatide (Mounjaro, Zepbound)	+2.05 bpm
Liraglutide (Saxenda, Victoza)	+2.37 bpm
Semaglutide, injectable (Ozempic, Wegovy)	+3.35 bpm
Retatrutide	+3.46 bpm
Semaglutide, oral (Rybelsus)	+4.50 bpm
Orforglipron	+7.30 bpm
Pooled across all GLP-1 RAs	+3.47 bpm

At the extremes of the network analysis, tirzepatide 5 mg raised heart rate by just 0.52 bpm while orforglipron 36 mg raised it by 9.29 bpm. So "GLP-1s raise your heart rate" is true but coarse. The drug and the dose matter more than the class.

The averages also hide the spread, and the FDA label is unusually explicit about this. Looking at each patient's maximum change from baseline at any visit, 26% of Wegovy patients recorded a rise of 20 bpm or more at some point, against 16% on placebo. For a 10 to 19 bpm rise it was 41% versus 34%. So roughly one in four people will see a genuinely large jump at least once, and the placebo arm shows that a meaningful share of that is ordinary day-to-day variation rather than the drug.

If your resting heart rate has moved and you are not on any medication, the causes are different and worth reading separately: we cover why your resting heart rate is suddenly higher with the full ranked list.

The mechanism is not what your recovery score assumes

This is the part worth the whole article.

Almost every explainer, including some manufacturer-adjacent write-ups, describes the GLP-1 heart-rate rise as sympathetic nervous system activation. A 2024 study in Cardiovascular Research tested that directly and found it is wrong, or at least badly incomplete.

Working in pigs, in living animals, in isolated perfused hearts and in isolated sinus node preparations, the researchers found that GLP-1 raises heart rate by a direct action on the pacemaker cells of the sinus node itself. The evidence that it is not autonomic is unusually thorough. The effect persisted through:

• Beta- and alpha-adrenergic blockade, so it is not catecholamine-mediated.

• Cervical vagotomy and autonomic ganglion blockade, so it is not coming from the brain or the autonomic nerves at all.

• Ivabradine, which rules out the HCN "funny current" pathway.

And it was abolished by exendin 9-39, a GLP-1 receptor-specific blocker, confirming it runs through the GLP-1 receptor. The receptor turned out to be expressed in the HCN4-positive pacemaker cells of the sinus node, with the effect running through calcium signalling rather than adrenaline.

The caveat, stated plainly: this was done in pigs. Pig sinoatrial physiology is closer to human than rodent physiology is, but human confirmation has not been published. It is the best available mechanistic account, not a settled human fact.

Why it matters for anyone wearing a tracker: your recovery score is built on the assumption that a higher resting heart rate and a lower HRV mean autonomic stress. That inference is usually reasonable. On a GLP-1 it may not be, because the drug appears to act on the pacemaker directly rather than shifting your autonomic balance. Same numbers, different cause, and the algorithm cannot tell the difference.

The HRV picture is genuinely unsettled

It would be neat to say HRV drops on these drugs. The honest answer is that the evidence disagrees with itself.

The strongest study design available is a double-blind, placebo-controlled crossover RCT of liraglutide in overweight adults with newly diagnosed type 2 diabetes and stable coronary artery disease. It found both SDNN and RMSSD decreased versus placebo, and that the effect persisted despite significant weight loss, pointing to a direct drug effect rather than a weight-mediated one.

Against that, a randomised open-label study using 24-hour Holter monitoring compared liraglutide with glimepiride in 62 adults with type 2 diabetes and found no significant difference in SDNN, RMSSD, HF power or LF power. Liraglutide raised heart rate and daytime heart-rate variation, but did not change the HRV measures.

Two randomised studies, same drug, opposite HRV conclusions. The comparators differed, the populations differed, and the durations differed. The 12-week wearable cohort found -6 ms, which supports the first, but it is observational.

The defensible summary: the heart-rate rise is consistent across the literature, and the HRV change is not. If your HRV has fallen, that is compatible with the evidence. If it has not, that is also compatible. Neither tells you the drug is or is not working. For the non-drug causes of an HRV drop, we have twelve common causes ranked.

What this does to your recovery score

Put the two findings together and the practical consequence is straightforward. Recovery and readiness scores weight resting heart rate and HRV heavily. A GLP-1 pushes RHR up and may push HRV down. So for the first weeks to months on the drug your recovery score will likely read worse, and the drop is not telling you what it normally tells you.

One nuance from the wearable cohort is worth having: users who were more physically active saw smaller heart-rate increases. That is observational rather than controlled, so it cannot support a causal claim, but it is the only moderator anyone has measured.

The right move is not to ignore the score but to re-baseline it. Give it eight to twelve weeks, then use your new normal as the reference and watch deviations from that instead of comparing against pre-drug numbers indefinitely. Recovery scores have shakier validation than most people assume even without a drug involved, which we cover in whether you can trust your recovery score.

This is also the moment a food log stops being optional. On a GLP-1 your intake, your body weight and your metrics are all moving at once, and untangling which change drove which is exactly the problem correlation is for. Kygo logs meals in 20 to 30 seconds by voice, photo or plain description, pulls your Oura, WHOOP, Garmin, Fitbit and Apple Health data alongside it, and runs the correlation over 12 to 36 hour windows. Get it on iOS or Android.

Sleep is where the news is straightforwardly good

The two SURMOUNT-OSA trials tested tirzepatide in adults with moderate-to-severe obstructive sleep apnea and obesity over 52 weeks, one trial in people not using CPAP and one in people on stable CPAP.

Apnea-hypopnea index fell by 27.4 events per hour in the first trial and 30.4 in the second, reductions of 55% and 62.8% against roughly 5% to 6% on placebo. Up to 51.5% of participants on the highest dose met the criteria for disease resolution.

The follow-up paper on patient-reported outcomes matters more for how you will actually feel, because AHI is a number and tiredness is an experience. At 52 weeks, PROMIS Sleep-related Impairment improved by about 7.5 points, PROMIS Sleep Disturbance by about 5, and the Epworth Sleepiness Scale by 1.33 to 2.18 points depending on whether participants were sleepy at baseline. Functional outcomes and general quality-of-life measures improved too.

The scope limit is important and routinely ignored: this was people with diagnosed obstructive sleep apnea and obesity. It is strong evidence that treating obesity improves apnea and how rested people feel. It is not evidence that a GLP-1 improves sleep in someone without sleep apnea, and no trial has tested that.

Your VO2 max will probably rise, and part of that is arithmetic

Here is a claim you will see confidently stated and should not believe: that GLP-1s improve VO2 max. No cardiopulmonary exercise testing trial of a GLP-1 appears to have been published at all. There is no measured VO2 max finding to cite.

What can be said is mechanical. Wearable-estimated VO2 max is expressed per kilogram of body weight, so if you lose 10% to 22% of your body mass without a proportional fall in cardiac output, the denominator shrinks and the number rises. Some of that is real improvement in relative fitness, which matters for how running feels. Some is the arithmetic of the normalisation. The split between the two has not been published, and anyone giving you a percentage is guessing.

The closest genuine functional evidence is from heart-failure trials, where six-minute walk distance improved by about 20.3 metres in one and 14.3 metres in another. Real gains, in a specific and unwell population, on a measure that is not VO2 max. For how far to trust the number your watch shows, see the most accurate VO2 max wearable.

The dosing-day step pattern nobody warns you about

In a larger ancillary cohort of 538 wearable users, step counts were lowest on dosing days and in the one to seven days afterwards.

This is obvious in hindsight and invisible in the moment. If you inject on a Sunday and your activity sits at its weekly low every Sunday through Tuesday, both your activity streak and your energy-expenditure estimate absorb it. Overall activity in that cohort was up by about 30 minutes a week, so the direction across the period was positive. The weekly rhythm underneath it is the part worth knowing.

If you test one thing in your own data, test this. It needs no equipment, and if the pattern holds for you it tells you which day to schedule hard sessions.

How to read your data on a GLP-1

Expect the first eight to twelve weeks to look worse, and do not act on it. Resting heart rate up and recovery down is the documented pattern, and on the mechanistic evidence it is not the autonomic stress your app thinks it is.

Re-baseline rather than comparing to your old self. Your pre-drug numbers stop being the right reference once a drug is acting directly on your sinus node. Deviations from your new normal still carry information. The absolute comparison does not.

Separate the drug from the weight loss. Both are happening at once and they move metrics in different directions. Weight loss tends to improve apnea, sleep quality and per-kg VO2 max. The drug pushes heart rate up. A single score blends them into one number and tells you nothing about which is which.

Watch the dosing-day rhythm in your step count and activity minutes before concluding anything about your training.

Take the FDA label seriously on the one point where it is directive. It asks patients to report palpitations or a racing heart at rest, and roughly one in four users records a 20 bpm-plus jump at some visit. A tracker is a reasonable early warning here, and an unreasonable substitute for telling your prescriber.

Common questions

Does Ozempic raise your resting heart rate?

Yes, modestly on average. The FDA label for Wegovy (semaglutide 2.4 mg) reports mean increases of 1 to 4 bpm, and a 2026 meta-analysis of 12 randomised trials found +3.35 bpm for injectable semaglutide and +3.47 bpm pooled across all GLP-1 receptor agonists. Individual variation is wide: about 26% of Wegovy patients recorded a rise of 20 bpm or more at some visit, versus 16% on placebo.

Why did my HRV drop after starting a GLP-1?

An HRV drop is consistent with what has been observed, but the evidence is genuinely split. A 12-week study of wearable users found HRV fell by 6 ms. One placebo-controlled crossover RCT found SDNN and RMSSD both decreased; a separate randomised study found no significant change in any HRV measure. The heart-rate rise is consistent across studies. The HRV change is not settled.

Is the heart rate increase from GLP-1s caused by stress?

Apparently not. A 2024 study in Cardiovascular Research found GLP-1 raises heart rate by acting directly on the pacemaker cells of the sinus node. The effect survived beta- and alpha-blockade, cervical vagotomy, autonomic ganglion blockade and ivabradine, and was blocked only by a GLP-1 receptor antagonist. The work was done in pigs and has not been confirmed in humans, but it suggests the rise is a direct cardiac effect rather than sympathetic activation.

Will my recovery score go down on Ozempic or Mounjaro?

Probably, at least at first. Recovery and readiness scores weight resting heart rate and HRV heavily, and a GLP-1 pushes RHR up and may lower HRV. Because the mechanism appears to be a direct pacemaker effect rather than autonomic stress, the lower score may not carry its usual meaning. Re-baselining after eight to twelve weeks is more useful than comparing against pre-drug numbers.

Do GLP-1s help sleep apnea?

Substantially, in people who have it. In the two SURMOUNT-OSA trials, tirzepatide reduced the apnea-hypopnea index by 27.4 and 30.4 events per hour over 52 weeks, reductions of 55% and 62.8%, and up to 51.5% of participants on the highest dose met criteria for disease resolution. Patient-reported sleep quality and daytime sleepiness also improved. This was in people with diagnosed obstructive sleep apnea and obesity, and does not generalise to people without apnea.

Does a GLP-1 improve your VO2 max?

No trial has measured it. There is no published cardiopulmonary exercise testing study of a GLP-1, so any specific VO2 max figure you see is inference rather than measurement. Wearable-estimated VO2 max is calculated per kilogram of body weight, so substantial weight loss raises the reported number partly through the arithmetic. The closest real functional evidence is six-minute walk distance improving by roughly 14 to 20 metres in heart-failure trials.

Which GLP-1 raises heart rate the least?

In the 2026 network meta-analysis, tirzepatide had the smallest pooled increase at +2.05 bpm, with tirzepatide 5 mg the lowest single estimate at +0.52 bpm. Orforglipron had the largest at +7.30 bpm pooled. That is not a reason to choose or switch a medication, which is a decision for your prescriber, but it explains why two people on different drugs see very different changes.

Should I stop taking my GLP-1 because my wearable numbers look bad?

No, and a wearable should not drive that decision on its own. The FDA label asks patients to report palpitations or a racing heart at rest, so a persistent or large change is worth raising with your prescriber. But a lower recovery score in the first weeks is the documented, expected pattern.

The bottom line

On a GLP-1, your resting heart rate goes up by a few bpm on average, your HRV may fall, and your recovery score will probably read worse for a while. The single most useful thing to know is that the heart-rate rise appears to be a direct effect on your heart's pacemaker rather than autonomic stress, which means your recovery score is describing the change accurately and interpreting it wrongly.

Meanwhile the things that genuinely improve, sleep apnea above all, are the ones your wearable is least likely to put in a headline number.

Re-baseline after a couple of months, separate the drug's effects from the weight loss, watch the dosing-day dip in your activity, and take a persistent large heart-rate jump to your prescriber rather than to your app.

Want to see which of these changes are the drug and which are what you are eating? Kygo connects your food log to your wearable data and surfaces the correlations rather than leaving you to guess. Get it on iOS or Android.

---

Key sources

Wegovy FDA prescribing information for the 1 to 4 bpm mean increase and the maximum-change distribution; Zhang et al. 2026, European Journal of Medical Research, systematic review and network meta-analysis of 12 RCTs, for the per-drug heart-rate figures; Lubberding et al. 2024, Cardiovascular Research, for the direct sinus-node mechanism (porcine model); Grosicki et al. 2025, American Journal of Physiology - Heart and Circulatory Physiology, the 12-week wearable cohort of 66 GLP-1 users and 66 matched controls, note that several authors are WHOOP performance science researchers; Kumarathurai et al. 2017, Diabetes Care, the placebo-controlled crossover HRV RCT; Nystrom et al. 2019, Endocrinology Diabetes and Metabolism, the contradicting HRV RCT; SURMOUNT-OSA (Malhotra et al. 2024), NEJM, for the apnea-hypopnea results; Kanu et al. 2025, Sleep Medicine, for the patient-reported sleep outcomes; and STEP-HFpEF (Kosiborod et al. 2023) for six-minute walk distance.

---

A note on what this post leaves out

One number that circulates widely is deliberately not used here. A 2016 Holter study reported liraglutide raising 24-hour mean heart rate by 13.2 bpm. That figure is real but it is a 24-hour mean rather than a resting heart rate, from a small open-label study with no placebo arm, and it is several times larger than every controlled trial and the FDA label. Quoting it as a typical expectation would be alarming and wrong.

We also do not give a VO2 max percentage, because no cardiopulmonary exercise testing trial of a GLP-1 has been published and every figure in circulation is inferred from weight loss rather than measured.

And we do not describe the heart-rate rise as sympathetic activation, which is how most explainers frame it. The primary mechanistic study contradicts that directly, and it is the main reason this post exists.

---

Disclaimer: Kygo is a personal data aggregation and insights platform designed for informational purposes only. The information provided by Kygo, including correlations, patterns, and trends identified in your data, does not constitute medical advice, diagnosis, or treatment. Nothing in this article is a recommendation to start, stop, or change any medication. Always consult a licensed healthcare provider with any questions regarding medical conditions or prescribed treatment.

If you have worn a tracker through your first months on a GLP-1, I would like to know how many bpm your resting heart rate moved and how long it took to settle.