Background and Rationale
Exercise, nutrition, sleep, and mental well-being are described as the four foundational pillars of longevity 1,2. While there are now consumer grade trackers available to track exercise, sleep and nutrition, mental health still remains in the realm of self-reported subjective metrics. The reverse of a mind at ease is a stressed one, and mental stress can manifest in a variety of physiological effects both acutely and chronically via its effect on the autonomic nervous system 3,4.
Stress can enhance or impair productivity, depending on its timing and persistence 5. While small bursts of stress can be stimulating and propel the body to an enhanced state of performance, chronic and unpredictable stress can depress body functions, immune systems and cause mental health problems, including sleep disturbances. Exercise has been positioned as a reliable stress management technique, however, the type, quantity and timing of the activity do have a bearing on the outcome 6. Recommendations and benefits of morning exercise are usually put forward on the basis of the principle of circadian alignment, which posits that physical activity early in the day has greater benefits for sleep and recovery. However, a large section of individuals are only able to exercise later in the day due to work-life commitments. Present work habits of a large population of adults span team collaborations across multiple time zones, which also puts work pressures on people in different geographies asymmetrically, leading to specific impact on healthspan metrics that include their daily activity, sleep and stress levels 7.
In this associational study, we followed up on the cohort of Ring AIR users described earlier to examine: a) daily stress scores and sleep quality of users with different levels of activity; b) whether there are distinct geographic signatures to stress recorded; and c) whether the timing of cardio (aerobic) activity is associated with any of these trends.
Methods
The study is a follow-up of the same cohort investigated in our
previous white paper on daily activity and sleep times. Briefly, Ultrahuman Ring AIR users in the Europe (EU), Gulf Cooperation Council Region (GCC), India (IND), and New York (NY) metropolitan areas were tracked between May 6 and May 31, 2024, who met the following criteria:
1) users aged 20-40 years;
2) who had logged cardiovascular workout sessions (indoor/outdoor walking, cycling and running) on the Ultrahuman app;
3) wore the ring during sleep on the same day; and
4) slept between 8 p.m. and 11.59 p.m. local time the same day. Users were categorised into before - 1p.m. (AM) and after - 1p.m. (PM) cohorts.
The Ultrahuman Stress Rhythm Score feature provides a dynamic measure of heart rate variation aligned with circadian rhythms, taking into account performance-related eu-stressors and anxiety-related dis-stressors throughout the day. The score is set to 100 at midnight daily, and depending on the elevation of heart rate associated with non-fitness events, points are lost. Hence, a poor stress rhythm score is indicative of high stress experienced by the user. In addition, stress occurring earlier in the day when resilience is higher impacts the score less than when the stressor occurs at night, which can adversely impact sleep quality and recovery. Similarly, exercise or physical activity, which can promote wellness, does not cause a decrease in stress scores when done anytime during the day.
We focused on the males for this analysis with the final combined population of 2345 users across four target regions: EU (AM = 482, PM = 630), GCC (AM = 73, PM = 44), IND (AM = 557, PM = 309), and NY (AM = 131, PM = 119).
We stratified the users by their average daily step count into three groups:
1) Individuals logging less than 5000 steps daily (<5k);
2) individuals logging between 5,000-10,000 steps daily (5-10k); and
3) individuals logging more than 10,000 steps a day (10k+).
We evaluated the daily average stress rhythm score (SR) and sleep heart rate variability (HRV) for the same day across each user for those who logged cardio workouts in the morning or evening, i.e., the AM and PM cohorts. We did not find enough users in the <5K group in the GCC area to allow for statistical tests, and hence these individuals were excluded from the current analysis.
Data analyses and aggregation are covered under terms of use of the Ultrahuman Ring AIR and application. Since this is a retrospective observational analysis, data analysts had no contact with any users and worked with de-identified data. Data was statistically analysed in-house using Python-MATLAB modules, and the Mann-Whitney U test was employed for comparison within and between groups.
Result
Volume of daily activity is not associated with the directionality of stress score.
The cohort was chosen based on the regularity of their cardiovascular workout routines, and were consistently found to have superior sleep quality. However, the step count logged over the day is a combination of exercise and daily activity. Hence, we decided to refine the cohort based on their average daily step count. <5K sub-group mostly represents individuals for whom activity was primarily recorded in the workout zone and subsequent activity was sporadic. The 5-10K and 10K+ sub-groups were generally active throughout the day. Within-group comparison revealed that increasing step count did not directionally correlate with higher or lower SR uniformly and had regional dependence. Therefore, in the EU, India and NY, individuals with higher step counts reported poorer stress scores in the AM cohort, but this was not statistically significant. The PM cardio workout cohort displayed a slight reduction in SR but was variable across regions and activity levels
(Table 1).
Evening workouts create a short-term stress-relieving effect but impact sleep HRV subsequently.
Ring AIR users with low daily activity (<5K group) across India and NY displayed a 5-point reduction in SR in people who exercised during the evening (Figure 1a). This decrease in SR with evening exercise was not apparent to EU Ring AIR users. For the moderately active group (5-10K), we noted a statistically significant decrease in the SR for users who logged evening workouts (Figure 1b), as compared to those who worked out in the morning in the GCC countries. Interestingly, we see the converse for users in NY, with a significantly higher SR for evening exercisers. In the highly active users (10K+) we found a marginally decreased SR for PM cohorts from India and NY, increased SR for the GCC region, and no difference in EU-resident users. None of the differences were statistically significant.
Figure 1: Graphical representation of median stress score for the AM and PM groups in the four target regions. Statistical analysis comparing AM to PM groups using Mann-Whitney U tests. ** denotes p<0.01, n.s. denotes no significant difference, N (EU) = 1112 , N (GCC) = 117, N (IND) = 866, and N (NY) = 250. N = combined AM and PM users across activity levels.
Although we found region-specific changes in SR, upon examining HRV during the following night’s sleep across users, we found a frequent decrease in this metric for users who logged evening/afternoon exercise as compared to users who did their workouts in the morning. The difference seemed to appear in higher activity groups, but with a strong regional bias. We found a statistically significant decrease in sleep HRV for the 5-10K groups of GCC and Indian users, whereas for EU and NY-residing individuals, this trend seems to be towards better sleep HRV. The differences between AM vs. PM cohorts for the 5-10K group were significantly elevated in the NY region. For the groups with the highest activity levels (10k+ user base), there was a general lowering of sleep HRV across all regions except NY when workouts were done in the evening (Figure 2c).
Figure 2: Graphical representation of median sleep HRV for the AM and PM groups in the four target regions. Statistical analysis comparing AM to PM groups using Mann-Whitney U tests. *** denotes p<0.001, ** denotes p<0.01, * denotes p<0.05, n.s. denotes no significant difference, N (EU) = 1112 , N (GCC) = 117, N (IND) = 866, and N (NY) = 250. N = combined AM and PM users across activity levels.
Table 1. Median values (Q1-Q3) for stress score and sleep HRV, segregated by activity level in the four target regions during the study period.
Conclusions, Limitations and Future Directions
The trifecta of sleep, activity, and stress are intimately intertwined and interact together to determine a person’s wellness. This follow-up analysis was motivated by the sub-clusters that we had found in the cardio-workout cohort of Ring AIR users whom we had studied previously. We looked to understand how much an algorithmic SR score can relate to a primary biomarker like sleep HRV to provide a sense of how heart rate variations throughout the day, coupled with a specific type of exercise, would relate to sleep HRV and hence recovery. At another level, there are conflicting reports on the benefit of aerobic exercise done in the morning versus evening, but the impact is usually studied at a specific time 8,9. The SR score allowed us to measure heart rate spikes before or following exercise and measure the proximal effect of the activity. For morning workout practitioners, this meant that while the increased HR in the workout was not included in calculating the SR score, they had more events in the day that could have adversely impacted their final SR score. For evening workout users, the cumulative stressors for the day would have decreased the SR even before their workout session; hence, there were similar probabilities of poor SR at the end of the day for both cohorts. The sleep HRV, on the other hand, allowed us to measure a more distal and unconscious readout of sleep heart rate, and by filtering for activity volume, we were able to pull out regional and sub-group effects.
In Asian-resident users, we generally found comparatively lesser SR metrics in general as compared to European and North American resident users. This may be due to them having to extend their work hours later in the day to interact globally. NYC users, though having higher SR numbers generally, reported the lowest median sleep HRVs, indicating an overall poorer sleep quality.
Although there was no difference in the SR based on when the user worked out, there was a negative impact on sleep HRV in users who worked out in the evening. This means that though the increased HR associated with the “good” stress of a cardio workout done in the evening could diminish before sleep, the general arousal and secreted endorphins may prevent the body from attaining a more restful state during sleep.
Morning exercise individuals had time to allow the endorphins, and other chemicals released due to their workout to taper and were primed to wind down in the evening and night 10, 11, 12. They may have had work- and social-related stressors in the evening that caused their SR score to drop, but this was processed rapidly to allow the body adequate rest and recovery.
This is an associational study that requires it to be interpreted in the context of measuring non-specific SR readout and HRV metrics, which can be influenced by multiple external and internal factors for each user. The grouping of the Ring AIR users was also done based on resident geography and likely captures an ethnically and age-wise diverse group, which can lead to wider variations. In addition, only a specific set of workout modes were included in this analysis; hence, the trends are not generalizable to functional weight training, yoga, and other forms of exercise. Finally, we have also carried out preliminary statistical analyses that can be further refined to provide greater predictive accuracy.
In summary, here we provide a glimpse into the complex interplay of sleep, activity, and stress during the day for wearable users across geographies. A deeper analysis is merited across all types of activities and their impact on sleep quality and recovery in the future.
Reach out to support@ultrahuman.com for commercial queries and science@ultrahuman.com for scientific queries.
