By linking wearable data with pregnancy hormones, researchers show how everyday devices could one day help detect risks like miscarriage or preterm birth long before symptoms appear.
Study: Association between wearable sensor signals and expected hormonal changes in pregnancy. Image credit: Taras Grebinets/Shutterstock.com
A recent study in eBioMedicine investigated the effectiveness of wearable sensors in tracking physiological and behavioral changes during pregnancy and assessing their association with pregnancy-related hormonal fluctuations.
Rising Trends in Adverse Pregnancy Outcomes
Pregnancy complications, including miscarriage and preterm birth, remain serious threats to both maternal and fetal health. In the U.S., adverse pregnancy outcomes are increasing, and disparities in maternal mortality persist between White and non-White women. These trends highlight the urgent need for better ways to monitor and address pregnancy risks.
Early detection of complications could enable timely clinical interventions and reduce adverse outcomes. Researchers are exploring digital health technologies and artificial intelligence to meet this need to track maternal and fetal health. Yet, these approaches are still in the early stages and require validation in larger, more diverse populations.
Pregnancy hormones, such as human chorionic gonadotropin (hCG), progesterone (P), oestrogens like oestrone (E1), oestradiol (E2), and oestriol (E3), relaxin, prolactin, human placental lactogen, oxytocin, and cortisol (C), are essential for supporting and maintaining a healthy, full-term pregnancy. Their levels naturally fluctuate at specific stages throughout gestation.
Many pregnant women experience difficulties in accessing healthcare, which could be attributed to geographic limitations and socioeconomic disparities. In addition, many working mothers, despite not having financial difficulties, also experience challenges in attending regular prenatal appointments due to lack of paid leave, employment demands, and childcare responsibilities. Many of these difficulties could be addressed through remote monitoring and mobile health technology.
Previous research has confirmed that remote monitoring and mobile health have enhanced patient empowerment and home pregnancy care, effectively reducing pregnancy stress. However, further research is needed better to understand the impact of remote monitoring on health outcomes.
About the study
The current observational study obtained data collected in PowerMom, a bilingual digital research platform that allows participants to remotely self-assess and report survey data. It also enables real-world data collection by integrating digital health technologies, such as wearable sensors. Wearable sensor data were analyzed from three months before pregnancy to six months after the baby's delivery.
Participants who were pregnant or within eight weeks postpartum, 16 years or older, living in the US or its territories, and had access to devices supporting the PowerMom application were recruited. The selected participants shared their wearable sensor data collected in the PowerMom mobile research platform using the MyDataHelps™ application. Furthermore, they provided demographic information and completed a “Health and History” survey detailing their health status.
During the study period, participants completed a biweekly “Health and Well-being” survey and a bimonthly “Pregnancy Support” survey. Any participants with incomplete or missing data, as well as fraudulent accounts, were excluded.
Since resting heart rate (RHR) changes are associated with expected pregnancy-related hormone changes, multivariable linear regression was used to model the expected RHR curve in live birth pregnancies using the expected hormone curves.
Study findings
Out of 5612 participants, whose average age was 31.1 years, enrolled in the PowerMom study, only 697 agreed to share data. Only 108, 108, and 56 participants provided sufficient RHR, number of daily steps (STEPS), and total minutes of daily sleep (SLEEP) data, respectively, during their pregnancy. For the main analysis, 99 live birth pregnancies were included for RHR and steps, and 54 for sleep, while the adverse outcome group was very small (9 for RHR and steps, 2 for sleep).
On average, participants with RHR wearable data were older than the overall cohort. The average duration of pregnancy for enrolled participants was 276.2 days, and for RHR participants, it was 275.2 days.
No statistically significant differences were observed in terms of access to maternity care between enrolled and RHR participants. Notably, no significant difference was observed in the RHR, STEPS, and SLEEP subsets among participants. Most participants who shared wearable data had singleton pregnancies, except for two candidates. Approximately 93% of the cohort had live birth pregnancies, while 6.9% indicated pregnancy loss, either miscarriage or stillbirth.
In the first trimester of the pregnancy, the average RHR decreased compared to pre-pregnancy levels in week 1, followed by an increase until week 5. After week 5, the decrease reached a minimum around week 9; subsequently, RHR increased during the second and third trimesters.
RHR peaks approximately 8–9 weeks before delivery and then steadily declines up to delivery. After the baby's delivery, RHR decreases below pre-pregnancy levels and stabilizes about 6 months post-delivery.
The number of steps decreased throughout the pregnancy compared to pre-pregnancy levels. Although the number of daily steps improved post-delivery, it remained below pre-pregnancy levels even 6 months postpartum.
An increase in total sleep minutes of almost 40 minutes compared to pre-pregnancy levels was observed after week 5, peaking between weeks 8 and 9. However, sleep duration decreased in the second and third trimesters and dropped sharply after delivery (about 50 minutes below baseline), before partially recovering. It remained about 20 minutes below the pre-pregnancy levels even after 6 months postpartum.
RHR curves were positively correlated with E1, E2, E3, P, and C, while negatively correlated with hCG. In regression modelling, however, only E1 and cortisol showed positive coefficients, while E2, E3, P, and hCG showed negative associations, indicating that regression analyses can reveal more nuanced hormone effects than simple correlations. SLEEP presented mild negative correlations with E1, E2, E3, P, and C, and was positively correlated with hCG. In contrast, STEPS curves exhibited a mild negative correlation with E1, E2, E3, P, and C.
In line with previous research, RHR was highly correlated with hormone changes. Multivariable linear regression modeling revealed a high coefficient of determination for the RHR curve, indicating a close relationship between pregnancy hormones and the RHR signal. During the second and third trimesters, the increase in RHR was likely induced by the hormones E1, E2, E3, P, and C. Since the length of pregnancy varied, this model failed to capture the decrease in RHR around the 31st and 32nd week of the pregnancy.
A robust correlation was observed between RHR fluctuations and pregnancy-induced hormonal changes. For pregnancies ending in miscarriage or stillbirth, RHR patterns differ from live birth pregnancies. Still, these findings are preliminary because of the very small sample size (n=9) and should be considered an exploratory feasibility assessment rather than a definitive conclusion.
Conclusions
The current study highlighted the potential of wearable sensors in monitoring maternal health throughout the pregnancy and postpartum. However, the findings of this study must be validated in the future using a larger cohort of diverse ethnicities. These studies must consider additional parameters influencing RHR, such as age and body mass index (BMI).
Download your PDF copy now!
Uterine corpus malignancies with pregnancy-like features lead to poor survival outcomes