In a large observational dataset collected from consumer wearables, nights with self-reported marijuana use were associated with small but consistent changes in nocturnal cardiac autonomic markers. Using linear mixed models at both within- and between-person levels, the study found that, within individuals, use nights had a lower nocturnal resting heart rate (RHR) by −0.12 bpm (99.5% CI: −0.14 to −0.10) compared with nights without reported use. Concurrently, heart rate variability (HRV), a marker of cardiac autonomic modulation, was reduced on use nights by −0.47 ms (99.5% CI: −0.53 to −0.42).
These associations extend laboratory observations of acute cannabis effects to a real-world setting using wearable-derived physiological signals. The reductions in HRV on use nights indicate altered autonomic balance during sleep when marijuana was reportedly used.
The analysis showed that reported marijuana use was associated with modestly longer sleep duration but objectively poorer sleep continuity. Specifically, sleep was longer by +7.60 minutes on use nights (99.5% CI: 7.37 to 7.83). Despite the increase in duration, measures of sleep fragmentation were higher: disturbances increased by +0.47 units (99.5% CI: 0.45 to 0.48) and wake time after sleep onset increased by +1.79 minutes (99.5% CI: 1.70 to 1.89).
Taken together, these findings suggest that although users slept slightly longer on nights they reported marijuana use, sleep was more disrupted, with more awakenings and greater wake time. The wearable-derived metrics thus portray a trade-off of increased total sleep time with reduced sleep quality on nights of marijuana consumption.
The study reported alterations in wearable-estimated sleep architecture on days with reported marijuana use. The source states that sleep architecture differed on use nights, consistent with other observed changes in duration and fragmentation, but does not provide granular numerical detail for each architecture stage in the abstract. Therefore, specific stage-by-stage values (for example, percent time in N1, N2, N3, or REM) were not reported in the source document excerpt.
Using binomial mixed models across 72 weeks of participant tracking, the authors observed a decline in the probability of daily reported marijuana use over time. The estimated change in daily use probability over 72 weeks was a decrease of −11.1% (99.5% CI: −11.5 to −10.8%).
The decline occurred while participants were wearing the consumer device and reporting use, raising the possibility, noted by the authors, that longitudinal self-monitoring may influence consumption behavior. The source does not report causal analyses or confirm that monitoring caused the decline; it only reports the observed temporal trend.
The dataset comprised 16,965 consumer-wearable users contributing a total of 5,226,739 person-days with self-reported marijuana use data. Daily self-reported use was logged alongside continuous wearable recordings that provided nocturnal RHR, HRV, and sleep metrics. Median marijuana use frequency among logged days was 25.1% (interquartile range, 7.1–62.6%).
Analyses relied on linear mixed-effects models to relate day-level reported use to nocturnal cardiac autonomic function (RHR, heart rate variability) and sleep outcomes (duration, fragmentation, and architecture) at both the within-person and between-person levels. Binomial mixed models estimated the probability of reported use over the longitudinal tracking period (72 weeks).
Key limitations and declarations from the source include that participant data are not publicly available due to WHOOP, Inc. intellectual property constraints. Deidentified data and a data dictionary may be available on reasonable request subject to WHOOP review and a signed data use agreement.
All listed authors are employees of WHOOP, Inc., the manufacturer of the wearable device used in this study, and received salary support and stock options; WHOOP provided data infrastructure. The source states that marketing and non-research groups at WHOOP had no role in study approval, data interpretation, manuscript preparation, or publication decisions.
The study protocol was reviewed and approved by an independent institutional review board (Salus Institutional Review Board; approval number 251121), and participants consented to use of deidentified data for research with the option to revoke consent.
In this large real-world cohort using consumer wearables, self-reported marijuana use was associated with measurable differences in nocturnal autonomic physiology—lower RHR and reduced HRV—and with longer but more fragmented sleep. The observed decline in reported use over extended tracking suggests that wearable-based longitudinal monitoring could be associated with changes in consumption patterns, though causality was not established in the report. The authors propose that consumer wearables offer scalable opportunities to study substance use and its physiological correlates outside laboratory settings.
Note: The source abstract reports these associations and aggregated effect estimates. Detailed sleep architecture stage values and causal inferences were not presented in the abstract and therefore are not described here.