Cannabis use linked to more broken sleep on wrist monitors in a 35-year study, but not to self-rated sleep

Plenty of people who use cannabis at night say it helps them sleep. A study that tracked cannabis use for 35 years, and then put sleep monitors on participants’ wrists, found something different. The more cannabis people had used over their lives, the more broken their sleep was in their early 60s. Yet when the same people rated their own sleep, cannabis users did not score worse than anyone else.

The study, published on October 9 in JAMA Network Open, cannot say that cannabis caused the difference. People who sleep badly may be more likely to use it. What it does is put measured sleep and felt sleep side by side in the same people, and the two did not agree.

Why cannabis and sleep needed a long study

The paper starts from a gap. Cannabis is widely used and often promoted as a sleep aid, but evidence of its long-term implications for sleep is limited.

What evidence exists is mixed. A review of the field in 2017 put it this way: Delta-9 tetrahydrocannabinol (THC) may decrease sleep latency but could impair sleep quality long-term. THC is the main intoxicating compound in cannabis, and sleep latency is the time it takes to fall asleep. The reviewers added that research on cannabis and sleep is in its infancy and has yielded mixed results.

The largest recent summary looked at recreational use. Its authors carried out a systematic review and meta-analysis of 120 studies identified from six databases, excluding studies on medicinal cannabis or clinical samples. Among 102 observational studies identified, current recreational cannabis use was associated with poorer sleep quality, both short and long sleep duration, more insomnia symptoms, and a later chronotype, compared to non-use. Chronotype means a person’s natural timing of sleep, so a later chronotype is a tendency to sleep and wake late. Observational studies watch what people do without assigning them to anything.

Experiments told a different story. No associations between cannabis and sleep quality and duration were observed across 19 experimental studies.

What was missing was a long record of cannabis use set against sleep that was measured as well as reported.

How the study measured cannabis use and sleep

The new work comes from Northwestern University, with Thanh-Huyen Vu as first author and Mercedes Carnethon as senior author. This cohort study analyzed data from the Coronary Artery Risk Development in Young Adults (CARDIA) study. It is, in the university’s description, a long-running, community-based study that began in 1985-86 and follows Black and white adults around the country.

Participants were asked about cannabis again and again. The scientists followed participants from young adulthood into middle age for nearly 35 years, repeatedly tracking their cannabis use over time.

From those answers the team built two measures. The first was lifetime exposure, calculated in cannabis-years, with 1 cannabis-year equating to 365 days of use. Someone who used cannabis every day for a year would have one cannabis-year. So would someone who used it once a week for seven years. The second measure was recent use, defined as past 30-day use vs former use and never use.

Sleep was measured at the examination 35 years in, held between 2020 and 2022, when participants were about 61 on average. Objective sleep was assessed using data from 7-day wrist actigraphy. Actigraphy means recording movement with a device worn like a watch and estimating from that movement when the wearer is asleep. Alongside it, subjective sleep was assessed with the Pittsburgh Sleep Quality Index (PSQI), a standard questionnaire in which people rate their own sleep.

A total of 1266 participants (891 with objective and 1128 with subjective sleep data) were analyzed. Of the 891 with monitor data, 63.3% were women, 41.0% were Black and 59.0% were White.

The monitor data were turned into four measures. Sleep percentage, the share of the night spent asleep, counted as low if it was under 85%. Fragmentation, an index of restlessness during sleep, counted as high in the most restless quarter of participants. Wakefulness after sleep onset (WASO), the time spent awake after first falling asleep, counted as prolonged at 30 minutes or more. And sleep duration was classed as short under six hours, optimal from six to eight, and long above eight.

What the wrist monitors showed about sleep

Lifetime use tracked with worse measured sleep. After adjustment for other factors, each additional cannabis-year was associated with higher odds of low sleep percentage. The increase was 10%, with a range from 6% to 14%. The odds of high fragmentation were 4% higher for each cannabis-year, with a range from 1% to 8%. The odds of prolonged wakefulness were 21% higher, with a range from 10% to 34%.

Odds are a way of expressing how likely something is: the chance of it happening set against the chance of it not happening.

The university summarized the pattern in plainer terms. People with higher cumulative cannabis use spent a lower percentage of the night asleep, experienced more fragmented sleep and spent more time awake after initially falling asleep.

Recent use showed the same pattern, more strongly. Compared with former and never users, those with recent use had higher odds of low sleep percentage, high fragmentation, prolonged WASO, and long (vs optimal) sleep duration. Set against people who had never used cannabis, recent users had 3.23 times the odds of prolonged wakefulness and 4.60 times the odds of sleeping more than eight hours.

Those two estimates are loose. The range around the first runs from 1.44 to 7.27, and around the second from 1.81 to 11.70. Ranges that wide usually mean a group was small.

Long sleep among users has been reported before. The 2025 analysis of recreational use found current use tied to both short and long sleep duration.

The associations held after the team accounted for other influences on sleep. The university lists age, sex and race, along with cigarette smoking, alcohol use, physical activity, depression, chronic health conditions and use of other substances.

How people rated their own sleep

On the questionnaire, there was nothing to see. Cannabis use was not associated with subjective sleep quality.

Vu singled this out. “The wearable device detected more nighttime disruption among participants with long-term cannabis exposure, but the participants themselves generally did not report worse sleep,” Vu said.

The researchers’ reading is that the disruption goes unnoticed. “Our findings suggest cannabis may actually be associated with poorer sleep quality, even if people don’t perceive those changes,” Vu said. Carnethon went further. “Our data do not support the belief that cannabis is helpful as a sleep aid,” Carnethon said.

There is another way to read the gap. The differences the monitor picked up may be real but too small to matter to how a person functions. The study does not settle which reading is right.

The same mismatch has turned up before, pointing the other way. A 2022 review looked at studies in which cannabis was given for sleep. While subjective improvements in sleep quality were often observed, diagnostic testing showed no improvements in sleep architecture. Sleep architecture is the pattern of sleep stages through the night. A 2025 analysis of sleep-laboratory studies, by some of the same authors, called for research to address the discrepancies between subjective sleep improvements and objective sleep metrics.

So across several kinds of study, how people feel about their sleep on cannabis and what instruments record have parted ways.

What trials of cannabis for sleep have found

Trials, in which cannabis products are assigned at random, are the stronger test of whether it helps. Most concern medical products given to people with illnesses, and their results are modest.

The largest review covered randomized trials of medical cannabis for impaired sleep. Thirty-nine trials (5100 patients) were eligible for review, and most trials (33 of 39) enrolled patients living with chronic cancer or noncancer chronic pain. They were short. The median follow-up was 35 days, median meaning the middle value. The reviewers concluded that medical cannabis and cannabinoids may improve impaired sleep among people living with chronic pain, but the magnitude of benefit is likely small. Side effects, dizziness above all, were more common with cannabis than with placebo.

A review of dosing was cooler. Sleep improvements were seen in 7 out of 19 randomized studies and in 7 out of 12 uncontrolled trials. Its conclusion was that high-quality evidence to support cannabis use for sleep remains limited.

In the sleep laboratory the picture is flatter still. Eighteen studies were identified, with nine suitable for meta-analysis. Pooled, they showed that cannabis administration does not consistently alter sleep duration, latency, wake time, efficiency, or sleep staging.

One result was consistent, and it concerned stopping. Withdrawal from active cannabis use was consistently associated with sleep disturbances, including reduced total sleeping times and prolonged sleep onset latency.

That bears on why regular users may feel cannabis helps. If sleep worsens for a while when a regular user stops, a night with cannabis will feel better than a night without it, whatever the effect over years. The new study did not test this.

What the cannabis study cannot show

Which came first. Sleep was measured during one week, when participants were about 61. The study cannot show whether broken sleep followed years of cannabis use or whether people who slept poorly were drawn to it. The 2025 analysis of recreational use flagged the same problem, noting that the association between cannabis and sleep may be bi-directional. The university’s statement says the study was not designed to determine whether cannabis directly causes sleep problems.

Other differences. The analysis adjusted for many influences on sleep. People who use cannabis for decades may still differ from those who do not in ways the adjustments did not capture.

What a wrist monitor sees. Actigraphy records movement, not brain activity. It estimates when a person is asleep and cannot tell one stage of sleep from another.

Products and doses. The researchers did not distinguish between modes of cannabis consumption (e.g., smoking versus consuming edibles). Over 35 years, the university notes, patterns of cannabis use likely changed over time.

Loose estimates for recent use. The wide ranges around the recent-use estimates mean those figures could be much smaller or much larger than reported.

Who took part. Participants were Black and White adults who had stayed in a study for 35 years and joined its sleep project. The results describe people in their early 60s and may not apply to younger users.

Recreational use. The paper’s conclusion refers to long-term recreational cannabis exposure and midlife sleep health. It does not address cannabis-based medicines prescribed for a condition. Do not stop or change a prescribed medicine without talking to your doctor.

Whether persistent trouble sleeping in a particular person points to an underlying sleep disorder is assessed by their doctor.

After 35 years of records, the study’s plainest result is that how people who used cannabis rated their sleep and how a monitor recorded it did not match.