Beyond fitness tracking: The use of consumer-grade wearable data from normal volunteers in cardiovascular and lipidomics research

Figures

• 

  Fig 1. Stratification of volunteers based on wearable-derived activity metrics. (A) Heatmap showing activity profiles of study subjects over a 24-hour period, grouped by cluster (red = AM, green = PM, blue = MidDay). (B) Average activity profiles of the AM, PM, and MidDay clusters, respectively. (C) Boxplots showing the age distribution of each cluster. (D) Distribution of sleep and wake times for each cluster. The code to generate this figure can be found in S2 Data. Asterisks denote significance of Tukey’s range test between cluster pairs. * = p < 0.05; ** = p < 0.01; *** = p < 0.001.

• 

  Fig 2. Associations between wearable data (DailySteps and RestingHR) and CVMD risk markers. The forest plot shows the effect and significance of wearable metrics as predictors for clinical risk markers. For steps, the OR is for each additional 1,000 steps. For RestingHR, the OR is for each additional bpm. Details of models used in the logistic regressions and the thresholds used to define the clinical features are provided in the Materials and methods section. p-Values and ORs for DailySteps are for interactions between gender and steps, with the female gender being the reference level. The code to generate this figure can be found in S2 Data. bpm, beats per minute; BMI, body mass index; CVMD, cardiovascular and metabolic disorder; DailySteps, average daily steps; DBP, diastolic blood pressure; FBG, fasting blood glucose; HDL, high-density lipoprotein; LDL, low-density lipoprotein; OR, odds ratio; RestingHR, wearable-derived RHR; RHR, resting heart rate; SBP, systolic blood pressure; TG, triglyceride; TotalChol, total cholesterol; WC, waist circumference.

• 

  Fig 3. Relationship between wearable-derived physical activity and cardiac parameters. Distribution of (A) LVM, (B) LVEDV, (C) RVEDV, and (D) AoF values across the 4 activity categories (Cat I–Cat IV). The code to generate this figure can be found in S2 Data. Asterisks denote significance of activity category as a GLM predictor with reference to Cat I. * = p < 0.05; ** = p < 0.01. AoF, aortic forward flow; BSA, body surface area; GLM, generalized linear model; LVEDV, left ventricular end-diastolic volume; LVM, left ventricular mass; RVEDV, right ventricular end-diastolic volume.

• 

  Fig 4. Wearable-derived activity and sphingolipid abundance. Heatmap showing the abundance of sphingolipids that are significantly associated with DailySteps. Columns represent volunteers ordered by increasing DailySteps. For comparison, values of FBG and DailySteps are shown. All values are z-score normalized by row. The code to generate this figure can be found in S2 Data. Cer, ceramide; DailySteps, average daily steps; FBG, fasting blood glucose; GlcCer, glucosylceramide; SM, sphingomyelin.