Validation study on cuffless ambulatory blood pressure monitoring method using bio-impedance, impedance cardiography and electrocardiography.
Continuous, non-invasive blood pressure (BP) monitoring is essential for understanding cardiovascular dynamics in real-world settings. This study evaluated the reliability and validity of a cuffless method for ambulatory BP monitoring using pulse wave travel times derived from bio-impedance, impedance cardiography, and electrocardiography. We examined whether this method can accurately estimate systolic and diastolic BP cross varying conditions.
Forty-one young, healthy participants completed a protocol involving postural changes, physical activity, and mental stress tasks. Pulse arrival time (PAT) and pulse transit time (PTT) were extracted, and person-specific models estimated BP using (1) all data of the person ('double-dipping') - used in most previous studies -, (2) half of a person's data (calibration), or (3) person-invariant parameters (calibration-free) which would enhance practical utility. Cuff-based BP served as the reference. Test-retest reliability, and construct and convergent validity were assessed.
Test-retest reliability was high for PAT- and PTT-based BP for all approaches (ICC > 0.7), comparable to cuff-based measurements. Construct validity was generally poor: BP changes during physical activity were detected, but responses to mental stressors were not. Only double-dipping showed acceptable convergent validity, likely reflecting overfitting. In the calibration approach, convergence with cuff-based BP was moderate (MAE = 10.1-12.9 mmHg), with 59.2-68.5% of estimates within 10 mmHg of reference values. The strongest deterioration of convergentvalidity occurred with person-invariant models.
Although cuffless pulse wave travel time-based BP estimators showed strong reliability, validity was insufficient for clinical or stress research use, even with person-specific calibration. Future work should explore extended calibrationperiods before testing in naturalistic settings.
Forty-one young, healthy participants completed a protocol involving postural changes, physical activity, and mental stress tasks. Pulse arrival time (PAT) and pulse transit time (PTT) were extracted, and person-specific models estimated BP using (1) all data of the person ('double-dipping') - used in most previous studies -, (2) half of a person's data (calibration), or (3) person-invariant parameters (calibration-free) which would enhance practical utility. Cuff-based BP served as the reference. Test-retest reliability, and construct and convergent validity were assessed.
Test-retest reliability was high for PAT- and PTT-based BP for all approaches (ICC > 0.7), comparable to cuff-based measurements. Construct validity was generally poor: BP changes during physical activity were detected, but responses to mental stressors were not. Only double-dipping showed acceptable convergent validity, likely reflecting overfitting. In the calibration approach, convergence with cuff-based BP was moderate (MAE = 10.1-12.9 mmHg), with 59.2-68.5% of estimates within 10 mmHg of reference values. The strongest deterioration of convergentvalidity occurred with person-invariant models.
Although cuffless pulse wave travel time-based BP estimators showed strong reliability, validity was insufficient for clinical or stress research use, even with person-specific calibration. Future work should explore extended calibrationperiods before testing in naturalistic settings.