Long-term accuracy and stability of blood pressure measurements from a smartwatch: Prospective validation study.

Background: Regular monitoring of blood pressure (BP) provides early detection of hypertension. Automated cuff-based oscillometric devices are commonly used for simplicity and reduced observer subjectivity. However, these devices can be uncomfortable, prompting the development of non-invasive, cuffless BP monitoring in smartwatches. Current validation standards, including ISO 81060-2:2018 and IEEE 1708TM-2019, specify accuracy criteria, yet most studies focus on short-term performance and do not address their long-term stability across the calibration period. Objective: This prospective single-arm study aimed to assess the long-term accuracy and stability of BP measurements obtained from a smartwatch throughout its recommended calibration interval. Methods: Thirty-seven participants completed a 28-day protocol, consisting of an initial calibration on day 0 followed by 27 consecutive days of paired measurements of systolic BP (SBP) and diastolic BP (DBP). BP was simultaneously measured using a Samsung Galaxy Watch 5 and a validated reference sphygmomanometer Omron M4. Accuracy was assessed against ISO and IEEE standards, and the recalibration test consistent with European Society of Hypertension (ESH) guidelines was performed using the last three measurements before scheduled recalibration. Results: The smartwatch met the accuracy criteria based on ISO and IEEE standards in all but one measure, with mean absolute differences, mean differences (MD), and standard deviations for SBP and DBP within acceptable limits. Bland-Altman analysis revealed negligible MD for SBP (-0.34 mmHg) and DBP (0.62 mmHg), with minimal drift over the calibration period (-0.19 mmHg for SBP and 1.02 mmHg for DBP). However, when reference BP was 10 mmHg away from the calibration point, MD was 3.4 mmHg for SBP and 5.1 mmHg for DBP. Conclusions: The smartwatch demonstrated acceptable long-term stability and accuracy for BP monitoring and trend tracking. However, accuracy declined as values diverged from the calibration point. Cuff-based confirmation is advised when BP fluctuates substantially or when diagnostic or therapeutic decisions are planned. Trial Registration: ClinicalTrials.gov NCT06098092.