Heart rate variability: what your HRV reveals about stress and recovery

You wake up, reach for your phone and there it is: your HRV from last night. Lower than yesterday. Next to it a coloured dot that does not bode well.

And straight away your mind starts racing. Am I ill? Too stressed? Was that glass of wine one too many? Is something wrong with my heart?

Many people know this moment. A number you never ordered suddenly sets the mood for the day.

In this article we look behind this number together. How it is calculated, how accurately your device measures, what pushes it down and what the studies say about it. And just as clearly: what it cannot do.

What HRV is: why a healthy heart is not a metronome

Place two fingers on your wrist. The pulse feels even, like a steady beat. That is exactly what it is not.

Heart rate variability, HRV for short, describes how much the time intervals between two heartbeats fluctuate. At a pulse of 60, there is on average one second between two beats. But only on average.

A made-up example: one interval lasts 940 milliseconds, the next 1,030, then 980, then 1,060. The display still says pulse 60. The intervals underneath are dancing all the same.

Shaffer and Ginsberg put this into one memorable sentence in their review of HRV measures: a healthy heart is not a metronome.

Your breath rocks along with it

At rest, your breathing can generate most of this fluctuation. When you breathe in, the intervals get slightly shorter and the heart beats faster. When you breathe out, they get longer. The technical term is respiratory sinus arrhythmia. The name sounds like a disorder and describes a healthy process.

According to Shaffer and Ginsberg, this breath-coupled, parasympathetically mediated fluctuation is the main source of variability in short resting measurements, especially with slow, even breathing.

According to current understanding, this fast component is carried mainly by the vagus nerve, the great nerve of recovery. How the vagus as a whole relates to stress regulation is covered in detail in The vagus nerve and stress regulation in burnout. One sentence is enough here: the vagus can brake the heart from one beat to the next and release it again.

Picture a good dancer. She keeps the beat, and still she responds to every small change in the music. Someone who dances rigidly to a metronome looks stiff. The heart is similar: the small deviations are a sign of flexibility.

Only normal beats are counted

A distinction that often gets lost: HRV is not the same as an irregular heartbeat. The calculation uses the intervals between normal beats, technically called NN intervals, from normal to normal. Extra beats or an arrhythmia such as atrial fibrillation do not belong in this calculation. They can distort the number, and then it no longer says anything about your regulation.

This is measurement methodology, not a study finding. But it has a practical consequence: if your pulse feels bumpy, that is not a sign of a high HRV but a reason for an ECG.

And now you know why the fluctuation between your heartbeats is not a flaw, but exactly what is being measured.

What HRV reflects: vagus, sympathetic nervous system, RMSSD and SDNN

The display shows a number with ms after it. It almost never says how it was calculated. Yet that is exactly what determines what it means.

Your autonomic nervous system has two big arms. The sympathetic nervous system switches to performance: pulse up, blood to the muscles, attention outwards. The parasympathetic nervous system, with the vagus as its most important nerve, switches to recovery, digestion and regeneration. Both work at the same time, like two hands on the same rope, not like a light switch.

Alongside this fast nervous system there is a second, slower stress system: the hormonal axis with cortisol. How it behaves under prolonged strain is described in Cortisol and the HPA axis in burnout. HRV mainly looks at the fast system.

RMSSD and SDNN: two lenses on the same heart

The two measures you will come across most often do not measure the same thing. RMSSD looks at the differences between directly consecutive intervals. According to Shaffer and Ginsberg, it is the most important time-domain measure for estimating the vagally mediated component. Laborde, Mosley and Thayer describe RMSSD as closely linked to high-frequency HRV and as relatively independent of respiratory influences, unlike the high-frequency measures.

SDNN, by contrast, captures all cyclical fluctuations within the measurement period. The longer you measure, the more slow waves are included. Over 24 hours, SDNN is considered the gold standard for cardiac risk assessment according to Shaffer and Ginsberg. And they stress: 24-hour values, short-term values of around five minutes and ultra-short-term values are not interchangeable.

There are also frequency measures. The Task Force assigned the high-frequency fluctuation, HF, to the range of 0.15 to 0.4 hertz, roughly the pace of normal breathing, and the low-frequency fluctuation, LF, to the range of 0.04 to 0.15 hertz.

And now you know why two numbers with ms after them can mean two quite different things.

What studies show about HRV as a marker, and what they do not

At some point you read the sentence: a low HRV is linked to higher mortality. And then the watch lies on the bedside table like a verdict.

Let us look closely at what these studies measure. Everything in this section describes associations in groups. None of these studies shows that your personal number determines your personal risk.

Heart and mortality

The pattern also appears without an acute heart attack. In the Framingham Study, the first two hours of an ambulatory ECG were analysed in 736 older people with an average age of 72. Over four years, lower HRV measures were associated with higher all-cause mortality. A log-transformed LF power that was one standard deviation lower was linked to a 1.70-fold higher risk (Tsuji and colleagues).

An association also appears in people without known cardiovascular disease. The meta-analysis led by Hillebrand, with eight studies and 21,988 participants, compared the lowest with the highest HRV level. The relative risk of a first cardiovascular event was 1.35 for SDNN and 1.45 for the LF component, which the team summarises as a 32 to 45 percent increased risk. The lower figure of 1.32, however, belongs to the HF component, and that was not significant. These figures cannot be compared directly with those of the heart attack study, because measures, cut-offs and groups differ.

Mental health, work stress and burnout

Similar patterns appear with psychological strain, mostly with small to moderate effects.

In a meta-analysis led by Koch with 21 studies, 2,250 people with major depression without medication and 1,982 control participants, all HRV measures were lower, for RMSSD with an effect size of Hedges' g = -0.462. Kemp and colleagues found that the severity of depression correlated negatively with HRV (r = -0.354). HRV decreased under tricyclic antidepressants, while it did not change significantly under serotonin reuptake inhibitors, mirtazapine and nefazodone, even though those treated responded to therapy.

That is one reason never to read an HRV value without a medication list. It is expressly not a reason to switch, reduce or stop an antidepressant. Stopping abruptly can carry its own risks, and any change belongs with the doctor who prescribed the medication.

For anxiety disorders, a meta-analysis led by Chalmers with 36 studies showed lower HRV, in the time domain with g = -0.45. For obsessive-compulsive disorder there was no significant difference in high-frequency HRV, although only 40 patients were included. A literature review led by Kim with 37 publications describes that HRV values changed under stress in most studies, most often with a lower parasympathetic component.

For work stress, the systematic review led by de Looff found an association with lower HRV. In a meta-analysis led by Kane, RMSSD and SDNN in doctors were lower during periods of strain than during recovery periods, with standardised mean differences of -0.63 and -1.05. However, it rests on seven small studies whose quality the team itself rates as moderate at best.

And now you know why the sentence about mortality risk holds true at group level and still tells you nothing about your own number from last night.

Measuring HRV: chest strap, watch and ring, and why there is no table of norms

You wear a ring, your friend wears a watch. At breakfast you compare your values. Hers is clearly higher. And already your night feels a little worse.

Before you conclude anything from that, it is worth asking three questions: What was used to measure, when was it measured, and what are you actually comparing it with?

Electrical or optical

There are two measuring principles. ECG and chest strap measure electrically, directly at the heart, beat by beat. Watch and ring measure optically: a light sensor picks up the pulse wave in the tissue. Strictly speaking, this is pulse rate variability. The occupational medicine guideline led by Sammito calls for keeping the two strictly separate and sees acceptable agreement only in young, healthy people at rest.

A review by Schäfer and Vagedes arrives at a similar picture. At rest, optical measurement is sufficiently accurate, even if it can somewhat overestimate short-term variability. During physical activity and some mental stressors, agreement suffers, often to an unacceptable degree.

The chest strap is a different story. Gilgen-Ammann and colleagues had ten healthy people wear a Holter ECG and a chest strap at the same time. The strap delivered 99.6 percent usable beat intervals overall and 99.4 percent at high intensity, the Holter ECG 94.6 percent and 89.8 percent at high intensity. Ten people and one device: a clue, not a comprehensive device test.

What validation studies against the ECG show

How wide the range can be was shown by an even smaller study led by Stone with five young healthy people and seven commercial devices and apps. An app that measures the pulse via the phone camera was off by 112.36 percent for RMSSD. The most accurate methods were at 4.10 and 6.84 percent.

Why watch and ring show different numbers

Now comes a point that is rarely explained, and it answers the question from the breakfast table.

And then the calculation method. According to this study, this smartwatch series calculates HRV as SDNN, every two to four hours. The rings, the wristband and the sports watches from the night-time study report RMSSD. Two calculation methods, two different numbers. It is as if you were comparing kilometres with miles, except that there is no fixed conversion factor.

Incidentally, the error figures from the two studies cannot be set against each other. Both used an ECG chest strap as reference, but one measured five minutes in the morning, the other whole nights, and the devices calculate differently. How accurately trackers detect sleep stages and where their limits lie is covered in Sleep trackers: what they measure and how accurate they are.

Why night and morning are more comparable than daytime

During the day a lot interferes: movement, coffee, conversations, thoughts, stairs. Herzig and colleagues studied 15 healthy young men over three nights in a sleep laboratory. In deep sleep, HF power was well reproducible within and between nights, whereas LF and LF/HF were poorly reproducible in all sleep stages. The team describes deep sleep as a stable state that, unlike wakefulness, is not influenced by internal and external factors.

How much the time of day plays a role is shown by the largest data collection to date. Natarajan and colleagues analysed the pulse data of around eight million users of a wearable manufacturer and found a strong daily fluctuation in HRV. Three of the four authors worked for this manufacturer, which belongs to the context.

If you do not measure during sleep, you will find in Laborde, Mosley and Thayer the recommendation of some experts to measure in the morning right after waking, lying down, because external factors interfere least then. If you want to capture HRV as a relatively stable trait, their recommendations are to combine at least two measurements. In elite sport, Plews and colleagues describe that suitable averages over several days are more informative than a single value, and that in top athletes both a rise and a fall can go along with unfavourable adaptation. How such trends can be read in training is covered in Recognising overtraining.

Why there is no reliable table of norms by age

And now the question almost everyone asks: Is my value good? Online you will find tables by age and sex, often without a source and without information on recording length, device or calculation method. I am deliberately not showing you one here. Not out of secrecy, but because the data do not yield a table you could sensibly measure yourself against.

Nunan and colleagues analysed 44 studies with 21,438 healthy adults. The values were lower than the Task Force norms, and between individuals there were differences of up to 260,000 percent, mainly in the spectral measures. In a twin study led by Golosheykin with 1,060 young adults, the estimated heritable share was 47 to 64 percent, with no notable influence of the shared environment.

Then there is age. Umetani and colleagues measured 24-hour HRV in 260 healthy people between 10 and 99 years of age. RMSSD fell to 47 percent of the starting value by the sixth decade of life and then remained stable. Under the age of 30, values were lower in women; beyond 50, the difference disappeared. In the large wearable analysis led by Natarajan, RMSSD showed no difference between the sexes.

What follows from this is unspectacular and still the most important sentence of this chapter: your own baseline from several measurements, ideally at the same time of day, with the same device, in the same situation, and the trend over days instead of today's value. There is no evidence-based fixed number of weeks after which a baseline is established.

And now you know why the question is not whether your value is good, but whether and when it changes for you.

What influences your HRV: from age to beta blockers

The number dropped noticeably this morning, and you do not feel bad at all. What happened?

Most of the time there is a very ordinary answer. Here are the influences for which there are solid data, and a few that are beyond your control.

Age, genes and fitness

You have already met age and genes above. They explain a large part of the differences between people, and none of this is a personal failure. The methods guideline led by Carter distinguishes non-modifiable factors such as age and sex from modifiable ones such as exercise and physical fitness. In the large wearable analysis led by Natarajan, more daily physical activity was associated with more favourable HRV measures in a dose-dependent way across age and sex. The authors themselves stress that this needs prospective studies for confirmation.

Alcohol

The guideline led by Sammito puts this into context: acute alcohol consumption usually lowers HRV, and chronic alcohol misuse leads to lower HRV. What alcohol can set in motion in the gut, even in small amounts, is covered in Alcohol and the gut.

Sleep

A meta-analysis led by Zhang pooled eleven randomised studies with 549 participants on sleep deprivation. RMSSD fell significantly; SDNN showed only a non-significant decrease. LF and LF/HF rose, and HF showed a downward trend without significance. Effect sizes are not given in the abstract, so there is no number here. How sleep and prolonged strain can reinforce each other is covered in Burnout and sleep disorders. If you snore loudly and are tired during the day despite long nights, it is also worth taking a look at Sleep apnoea.

Infections

An infection can change HRV, sometimes before you feel ill. In an observational study led by Hirten, employees of a health system wore a watch that recorded SDNN. The daily fluctuation of SDNN differed between people with and without COVID-19, and significant changes in HRV appeared even before the positive swab.

Natarajan, Su and Heneghan analysed wearable data from 2,745 people with a confirmed infection. Breathing rate and pulse typically rose, HRV fell. From the body signals, illness on a given day could be predicted with an area under the curve of 0.77. All three authors were funded by the manufacturer. This is an early signal at group level and not a test.

Menstrual cycle

If your HRV drops in the days before your period, that is a known pattern. A meta-analysis led by Schmalenberger with 37 studies and 1,004 naturally cycling women found a significant decrease in cardiac vagal activity from the follicular to the luteal phase (d = -0.39). From menstruation to the premenstrual phase the decrease was stronger (d = -1.17), as it was from the mid to late follicular phase to the premenstrual phase (d = -1.32). These two large values, however, are based on only five and eight studies with wide confidence intervals. In practice this means: preferably compare your values within the same cycle phase.

Medication

This is where it gets important, and where there is a trap. In a small randomised, double-blind crossover study led by Niemelä, 18 men with stable coronary heart disease received two different beta blockers and placebo in sequence, two weeks each. Depending on the drug, RMSSD rose by 70 and by 62 percent, SDNN by 20 and by 16 percent.

So a higher value due to a medication does not automatically mean a more relaxed nervous system. If you take beta blockers, you cannot compare your values with times without the medication. Conversely, in the meta-analysis by Kemp, HRV fell under tricyclic antidepressants. The guideline led by Sammito names beta blockers, acetylcholinesterase inhibitors, antiarrhythmics and psychotropic drugs as examples of drug classes that are relevant to HRV.

Stress and smoking

In most studies in the review led by Kim, stress was most often associated with a lower parasympathetic component. According to the guideline led by Sammito, smoking can lower HRV, and it does so in a dose-dependent way.

In clinical practice I observe that many people only understand their value once alcohol, a late heavy meal or a beginning infection become visible in their own trend. This is experience from consultations, not a study finding, and I did not research any study on late meals for this article.

And now you know why a low morning value can often tell you more about your previous evening than about your heart.

What was able to raise HRV in studies

You have understood what the number measures. Now you want to know whether you can move it. The short answer: yes, within limits, and most likely via the system behind it.

Endurance training

This is one of the best-studied levers. A meta-analysis led by Sandercock pooled 13 studies with 322 people. Exercise training was associated with an increase in HF power, with an effect size of d = 0.48. In older people there was a trend towards a weaker response.

An effect was also measurable later in life in studies. A meta-analysis led by Raffin with twelve studies in healthy people aged 60 and over found an increase in 24-hour SDNN in the controlled studies, with g = 0.721. The more often people trained, the larger the effect on short-term SDNN. The authors themselves point to a certain risk of bias in the studies.

And HRV-guided training? A meta-analysis led by Manresa-Rocamora compared it with predefined training. For vagally mediated HRV, the guided approach was superior (SMD = 0.50); for maximal oxygen uptake the difference was not significant (SMD = 0.20). If there is an advantage for fitness and performance at all, it is small at group level.

Why calm, long endurance sessions can be interesting for metabolism is covered in Understanding Zone 2 training. Fitness and HRV are related but different markers; what maximal oxygen uptake can say about life expectancy is covered in VO2max and life expectancy. And how HRV can be read as a recovery signal in everyday training is covered in Recovery after exercise.

Slow breathing

The second well-supported lever costs nothing. A team led by Laborde pooled 223 studies on voluntary slow breathing. Vagally mediated HRV rose during the exercise, immediately afterwards and after programmes lasting several weeks. The team describes slow breathing as a simple, low-cost technique with few expected side effects, for prevention and as a complementary measure.

HRV biofeedback

In HRV biofeedback you breathe slowly while watching your heart data in real time, usually as a curve. The aim is to find the pace at which the fluctuation becomes largest.

For self-reported stress and anxiety, an older meta-analysis led by Goessl with 24 studies and a total of 484 participants arrived at g = 0.83 compared with control conditions, measured with questionnaires.

Sleep and less alcohol as indirect levers

Two levers are already contained in the previous chapter. If sleep deprivation can push RMSSD down and alcohol lowered night-time HRV in a dose-dependent way in studies, then sufficient sleep and alcohol-free evenings are the most obvious way to avoid dragging your own measurement down artificially. This is an inference from the observations, not a separate intervention study.

And once more, because it is so easily overlooked: a rise due to a medication such as a beta blocker is not the same as a rise due to training or breathing. The number goes up in both cases, but the system behind it is a different one.

I did not review studies on supplements, cold exposure or sauna as HRV levers for this article. That is why they do not appear here as a recommendation.

And now you know why the best levers for HRV are unspectacular, and why that is exactly where their strength lies.

The limits: not a diagnostic tool, and when measuring itself becomes stress

Some people look at their watch at three in the morning to check whether recovery has set in. And then they lie awake because they checked.

What HRV cannot do

The guidelines are remarkably unanimous here. The occupational medicine guideline led by Sammito sees no basis for general health statements from cut-off values. The methods guideline led by Carter sees some value for estimating cardiovascular risk, but no indication of a sympathovagal balance. And a joint position statement by a working group of the European Society of Cardiology and the European Heart Rhythm Association, led by Sassi, states that newer analysis methods have broadened the technical understanding of the signal, but that their success in developing new clinical tools has remained rather limited. New prospective studies are needed.

This also applies to the daily readiness, recovery and stress scores of many apps. These are often manufacturer indices, like the recovery index from the alcohol study above. An independent validation of such scores was not part of the research for this article, so I classify them as common practice without a strong study base. That does not mean they are worthless. It means they deserve no more weight than the question of how you feel.

And the biggest gap remains causality. All the data on mortality, cardiovascular risk, depression and burnout are observational. Whether a low HRV is a cause, a consequence or merely a companion cannot be read from them.

When self-tracking can amplify anxiety and compulsive checking

In some people, self-tracking can amplify exactly what it is meant to calm: worry, constant rechecking that can feel like a compulsion, the feeling of having to pass an exam every morning. Sleep medicine has a term for this: orthosomnia. In a case series, Baron and colleagues described people in whom the presumed link between tracker data and daytime tiredness can turn into a perfectionist quest for the ideal sleep; more on this in Sleep trackers: what they measure and how accurate they are.

In clinical practice I also observe that people who tend to worry anyway often look particularly closely at their values. And anxiety disorders, as described above, are themselves associated with lower HRV. This could create a loop in which worry pushes the number down and the number feeds the worry. This is a supposition from practice, not a study finding.

If you notice that looking at the number determines your morning, taking a break from displaying the values may be worth considering. This is not a study recommendation, but a question you are allowed to ask yourself. And if brooding, checking and anxiety extend beyond the measurements, that is a matter for medical or psychotherapeutic support, not for another device. If you are exhausted right now and want to get back into exercise, Training despite burnout explains why the nervous system needs rest first.

Through the KPNI lenses, and three levers for the coming weeks

In Clinical Psychoneuroimmunology we look at the same number through four lenses. Through the lens of the nervous system, HRV mainly shows how flexibly the vagus guides the heart. Through the lens of the immune system, a drop can point to a beginning infection, as the wearable studies show at group level. Through the lens of metabolism, alcohol and fitness can be reflected in night-time HRV. And through the lens of the hormonal system, the menstrual cycle can shift vagal activity, while the cortisol axis works in parallel as a second, slower stress system.

None of these lenses turns HRV into a diagnosis. Together they make it a useful starting point for a conversation about sleep, alcohol, exercise, strain and medication.

If you would like to put your values into context with someone who looks at sleep, strain, medication, lab values and symptoms together: below this article you will find the option to book an appointment. More articles on exhaustion and stress regulation can be found in the Burnout topic area.

And now you know why your HRV says the most when you take it quietly: as a trend, under the same conditions, alongside how you feel and never in its place.

Frequently asked questions about heart rate variability

Where HRV connects with other topics

HRV never stands alone. It is linked to the vagus, to sleep, to training, to mood and to what is in the glass in the evening. Here is where to go next.