Evening Workouts Went With Worse Sleep Only Close to Bed
Four million nights of wearable data point to one threshold. A meta-analysis of experiments found no average harm from evening exercise.
Evidence rating for the proposition “Insufficient sleep duration is associated with worse health outcomes”: Established
Only when it is late and strenuous. In 14,689 physically active adults (4,084,354 person-nights), exercise ending 4 or more hours before sleep onset was not associated with changes in sleep, whatever the strain. Later and harder sessions went with delayed sleep onset, shorter sleep, lower sleep quality, higher nocturnal resting heart rate and lower heart rate variability. A meta-analysis of 23 experiments found evening exercise did not harm sleep on average. Vigorous exercise ending within an hour of bed might.
Does evening exercise hurt sleep on average?
Not in experiments. A meta-analysis screened 11,717 references and included 23 studies. Each compared sleep after a single evening exercise session with a no-exercise control in healthy adults.
Evening exercise increased REM latency by 7.7 minutes and slow-wave sleep by 1.3 percentage points. It decreased stage 1 sleep by 0.9 points. The authors concluded that the studies do not support the idea that evening exercise harms sleep. If anything, they point the other way.
One exception appears. Sleep-onset latency, total sleep time and sleep efficiency might be worse after vigorous exercise ending within an hour of bedtime. Every significant moderator effect disappeared when one study was removed.
1 study
- In 23 studies of healthy adults, evening exercise increased REM latency by 7.7 minutes (p = 0.032) and slow-wave sleep by 1.3 percentage points (p = 0.041), and decreased stage 1 sleep by 0.9 points (p = 0.001) against no exercise. Sleep-onset latency, total sleep time and sleep efficiency might be impaired after vigorous exercise ending within one hour of bed.
Does timing and strain change the answer?
In a large real-world dataset, yes. The analysis used 14,689 physically active people who wore a biometric device for a year. That covered 4,084,354 person-nights.
Later exercise timing and higher strain were associated with delayed sleep onset and shorter sleep. They were also associated with lower sleep quality, higher nocturnal resting heart rate and lower nocturnal heart rate variability.
Exercise ending 4 or more hours before sleep onset was not associated with changes in sleep. That held regardless of strain. The authors suggest evening exercise involving high strain may disrupt sleep and nocturnal autonomic function.
1 study
- In 14,689 physically active people over 4,084,354 person-nights, later exercise timing and higher strain were associated with delayed sleep onset, shorter sleep, lower sleep quality, higher nocturnal resting heart rate and lower nocturnal heart rate variability. Exercise ending 4 or more hours before sleep onset showed no association with sleep changes, regardless of strain.
Why do the two studies point in slightly different directions?
They ask different questions. The meta-analysis pooled experiments of one evening session against a no-exercise control. The wearable study tracked what active people did over a year, with timing and strain as they occurred.
The experiments average over many timings and intensities. The dataset splits by timing and strain. Both flag the closest-to-bed, hardest sessions. The experiments name the last hour. The dataset places the line at four hours.
The two thresholds differ. Neither is a universal law. The dataset is observational, so strain and timing were not assigned.
Do the night heart rate signals on a wearable reflect this?
They moved in the same direction as the sleep outcomes. Late, strenuous sessions went with a higher nocturnal resting heart rate and lower nocturnal heart rate variability. Those are the signals most sleep and recovery scores are built from.
The abstract describes no test of any brand's score. The overlap is a reasonable connection and not a tested result. The wider causes of score drops are covered in why recovery scores drop.
What does this not settle?
The four-hour figure marks where the association vanished in one dataset. It is not a guarantee that a session ending at three hours and 59 minutes costs sleep.
Participants were physically active adults already wearing a device. Nothing here covers sedentary adults, people new to exercise, older adults specifically, or differences by chronotype or fitness. The experiments tested single sessions in healthy adults.
The wearable dataset was observational and drawn from physically active people already using a biometric device, and the meta-analysis covered single sessions in healthy adults. Neither shows how repeated late sessions affect people with sleep disorders, sedentary adults or older adults.
Common questions
Does exercising in the evening ruin sleep?
Not on average in experiments. A meta-analysis of 23 studies found evening exercise did not harm sleep and slightly increased slow-wave sleep. Vigorous exercise ending within an hour of bedtime might impair sleep onset, total sleep time and efficiency.
How long before bed does exercise stop affecting sleep?
In one dataset of 14,689 people, exercise ending 4 or more hours before sleep onset was not associated with changes in sleep, regardless of strain. That is an observational threshold, not a rule.
Does late exercise change nighttime heart rate?
In that dataset, later timing and higher strain were associated with higher nocturnal resting heart rate and lower nocturnal heart rate variability, along with delayed sleep onset and shorter sleep.
Does the evidence apply to people who are not already active?
Not established. The large dataset covered physically active people wearing a biometric device, and the meta-analysis covered single sessions in healthy adults.
Do these findings apply to a wearable's sleep score?
Untested. The study measured signals that sleep and recovery scores draw on. Its abstract describes no test of any brand's score.