Real-world performance of a ring-type cuffless blood pressure device compared with ambulatory blood pressure monitoring

Abstract

Background

Cuffless blood pressure (BP) monitoring devices are convenient alternatives to conventional ambulatory blood pressure monitoring (ABPM). We evaluated the real-world agreement between a ring-type cuffless BP device (continuous ambulatory ring-type blood pressure monitor, CART BP pro) and ABPM and identified the clinical and physiological factors associated with measurement discrepancies.

Methods

This real-world observational study included 154 adults (mean age, 62.1 ± 14.2 years; 41.6% men) who underwent simultaneous 24-hour ABPM and CART BP pro after exclusion of predefined clinical conditions and insufficient recordings. CART BP pro measurements were calibrated at baseline to ABPM in the contralateral arm. BP values from the 2 devices were compared for 24-hour, daytime, and nighttime periods, as well as for the classification of nocturnal hypertension and dipping status. Predictors of close systolic agreement (|24-hour systolic blood pressure [SBP] difference| ≤ 5 mmHg) were identified using multivariable logistic regression.

Results

CART BP pro-derived period summaries were lower than ABPM for systolic BP (–7.1 mmHg for 24-hour SBP) and showed a modest positive bias for diastolic BP. Only 33.1% of measurements met the ≤ 5 mmHg accuracy criterion. Daytime SBP showed a similar bias (–7.8 mmHg), whereas nighttime SBP differences were smaller (–1.7 mmHg). By contrast, diastolic BP was consistently higher with the CART BP pro, particularly at night (+4.8 mmHg). Older age, male sex, lower body mass index, and higher heart rate independently predicted a closer systolic agreement. The agreement for nocturnal hypertension was moderate (κ = 0.516), with an overall concordance of 76.0%. However, dipping concordance was limited, with the CART BP pro showing blunted nocturnal dipping compared with ABPM (mean, 3.7% vs. 7.8%).

Conclusions

This exploratory, real-world study found that CART BP pro-derived period summaries showed lower SBP values and smaller nocturnal BP declines than ABPM. Because period-level aggregates were compared rather than simultaneous point-to-point measurements, these findings should not be interpreted as a formal accuracy validation of the device. Further studies are needed to establish device-specific diagnostic thresholds and prospectively validate the clinical interpretation of CART BP pro-derived period summaries for hypertension diagnosis and circadian BP phenotyping.

BACKGROUND

Ambulatory blood pressure monitoring (ABPM) is widely considered a reference standard for the diagnosis and management of hypertension [1,2,3]. By capturing 24-hour blood pressure (BP) profiles, ABPM provides essential information on day–night variability, nocturnal hypertension, and the identification of masked and white-coat hypertension [4,5,6]. These BP phenotypes have independent prognostic significance beyond office BP measurements and are strongly associated with cardiovascular outcomes [7]. Despite its clinical value, the conventional cuff-based ABPM has limitations. Repeated cuff inflation can cause discomfort, disrupt daily activities, and disturb sleep, all of which may reduce patient adherence and the quality of measurement. Concerns have also been raised regarding the reliability and accuracy of BP measurements, especially at night [8,9]. Recent advances in wearable technology have enabled the development of cuffless BP monitoring devices to address these limitations [10,11]. The ring-type cuffless BP device (continuous ambulatory ring-type BP monitor ring-type BP device, CART BP pro; Sky Labs Inc., Seoul, Republic of Korea), provides automated BP measurements without intermittent cuff inflation and offers high wearability [12,13]. The CART BP pro has undergone regulatory validation and received approval from the Korean Ministry of Food and Drug Safety for clinical use, supporting its implementation in routine ambulatory BP monitoring practices in Korea. However, systematic analyses of the measurement differences between CART BP pro and conventional ABPM in clinical practice remain limited, and their ability to assess nocturnal BP patterns remains unclear.

Therefore, in this prospective real-world study, we performed simultaneous monitoring with the CART BP pro using conventional 24-hour ABPM. We aimed to evaluate inter-device agreement across 24-hour, daytime, and nighttime periods; identify clinical factors associated with close agreement; examine BP range-dependent bias; and assess concordance for nocturnal hypertension and dipping phenotypes.

METHODS

Study population

This real-world observational study was conducted at 2 centers within the same university system in Korea: a tertiary referral hospital and a secondary care hospital. From August 2025 to December 2025, consecutive adults undergoing simultaneous 24-hour ABPM and CART BP pro were screened for inclusion in cardiology outpatient and inpatient clinics. Eligible participants were adults with either elevated office BP indicating hypertension or a prior diagnosis of hypertension who were receiving antihypertensive medication. Of the 168 individuals initially enrolled, 14 were excluded because of predefined clinical conditions (ankylosing spondylitis, Takayasu arteritis, and congenital aortic stenosis; n = 3) or insufficient ambulatory BP recordings (fewer than 20 daytime or 7 nighttime readings; n = 11). The final analysis included 154 participants (Fig. 1).

Fig. 1. Study population.

Flowchart of participant enrollment and exclusions leading to the final analysis population.

ABP, ambulatory blood pressure; ABPM, ambulatory blood pressure monitor; CART, continuous ambulatory ring-type.

The study protocol complied with the Declaration of Helsinki and was reviewed and approved by the Institutional Review Board (IRB) of each center, including Hanyang University Guri Hospital (IRB No. 2025-08-023). Written informed consent was obtained from all the participants.

Data collection

ABPM was performed on the upper arm using a validated oscillometric device (TM-2441; A&D Co., Ltd., Tokyo, Japan). BP was measured automatically every 30 minutes at night, following the standard clinical protocol [3]. The daytime measurement interval was set at 15 minutes at one center and 20 minutes at the other. All participants received standardized verbal instructions on arm immobilization during cuff inflation at the time of device application, and monitoring was initiated under clinical supervision in the outpatient or inpatient setting. Simultaneously, participants wore the CART BP pro on the finger of the contralateral hand, which recorded BP every 20 minutes over 24 hours. Before monitoring, the device was initialized and calibrated using 2 consecutive BP readings obtained simultaneously from the ABPM device applied to the contralateral upper arm, in accordance with the manufacturer's calibration protocol. Both devices were worn continuously during daily activities and sleep. Daytime and nighttime periods were determined based on the participants’ reported waking and sleeping times. An ABPM dataset was considered valid if it included at least 20 valid daytime recordings and 7 valid nighttime recordings, excluding unmeasured or extreme values (systolic blood pressure [SBP] ≥ 350 or < 60 mmHg; diastolic blood pressure [DBP] ≥ 250 or < 30 mmHg) [14]. The number and proportion of valid readings were calculated separately for the ABPM and CART BP pro for 24-hour, daytime, and nighttime periods. Office BP and heart rate were measured before device application using an automated sphygmomanometer (HEM-907; Omron Healthcare Co., Ltd., Kyoto, Japan) after at least 5 minutes of seated rest.

Definitions

The mean 24-hour, daytime, and nighttime values for SBP, DBP, and heart rate were calculated separately for ABPM and CART BP pro using all valid readings from each period. In addition, the time-weighted mean BP values were calculated to account for unequal intervals between consecutive measurements during monitoring. The primary analytical endpoint of the study was the absolute difference in 24-hour SBP between the ABPM and CART BP pro (|24-hour SBP difference|). Close agreement was predefined as an absolute difference ≤ 5 mmHg. SBP was selected as the primary endpoint because systolic BP has the strongest prognostic relevance in ambulatory BP assessment and is the principal target of hypertension management.

Nocturnal hypertension was defined as nighttime SBP ≥ 120 mmHg, consistent with guideline-recommended ambulatory BP criteria for nocturnal hypertension (≥ 120/70 mmHg) [1,2,3].

Dipping percentage was calculated as dipping percentage = [(daytime SBP − nighttime SBP)/daytime SBP] × 100. Based on dipping percentage, participants were categorized into 4 phenotypes: riser (< 0%), non-dipper (0% to < 10%), dipper (10% to < 20%), and extreme dipper (≥ 20%). Close agreement in 24-hour SBP between ABPM and CART BP pro was defined as an absolute mean difference ≤ 5 mmHg. Office SBP categories for the stratified analyses were defined as < 110, 110–119, 120–129, 130–139, and ≥ 140 mmHg.

Statistical analysis

Continuous variables are reported as means ± standard deviation (SDs) and categorical variables as counts and percentages. Between-group comparisons were performed using Student’s t-test or one-way analysis of variance for continuous variables and χ² or Fisher’s exact tests for categorical variables, as appropriate. Agreement between ABPM and CART BP pro was assessed by calculating mean differences (CART BP pro − ABPM) and the SD of these differences. Bland–Altman plots were used to visualize the bias and limits of agreement for 24-hour, daytime, and nighttime SBP and DBP. The proportion of participants meeting predefined accuracy thresholds (absolute mean difference ≤ 5 mmHg and SD of difference ≤ 8 mmHg) was determined. BP variability was characterized by SD, coefficient of variation (CV), and variability independent of the mean (VIM) for both devices across all measurement periods. Pearson correlation coefficients between ABPM and CART BP pro values were calculated and illustrated with scatter plots. Proportional bias was assessed by regressing the difference (CART BP pro − ABPM) on the mean of the 2 devices for each Bland–Altman panel. The slope and its statistical significance were reported. The agreement in the nocturnal hypertension classification was evaluated using Cohen’s kappa coefficient. McNemar’s test was used to assess the directional imbalance in discordant classifications. Hypertension classification agreement and reclassification between the 2 devices were assessed using standard ABPM diagnostic thresholds. Agreement was quantified using Cohen’s kappa coefficient, while discordant classifications (false positives and false negatives) were evaluated using McNemar’s test. Logistic regression analysis identified factors associated with close agreement (absolute 24-hour SBP difference ≤ 5 mmHg). Significant variables in the univariate analysis were included in the multivariable models. The multivariable model was extended to include ABPM SBP SD and CART BP pro SBP SD as additional candidate predictors. Odds ratios (ORs) and 95% confidence intervals (CIs) were determined. All statistical analyses were conducted using Python 3.11.5 (Python Software Foundation, Wilmington, DE, USA) with NumPy 1.25.2, pandas 2.2.3, SciPy 1.11.2, statsmodels 0.14.3, scikit-learn 1.3.0, matplotlib 3.7.2, and openpyxl 3.1.2.

RESULTS

Baseline characteristics

The baseline characteristics are shown in Table 1. The mean age was 62.1 ± 14.2 years, and 41.6% were men; hypertension was present in 75.3% and 59.7% were taking antihypertensive medication. ABPM provided 67.7 ± 7.5 valid readings over 24 hours (validity proportion, 95.8% ± 7.1%), whereas CART BP pro yielded 61.1 ± 7.1 readings (validity proportion 86.2% ± 9.6%). CART BP pro recorded a higher number of nighttime measurements than ABPM (21.0 ± 4.6 vs. 14.3 ± 3.4), with similarly high nighttime validity proportions in both methods (~99%) (Table 1).

Table 1. Baseline characteristics.

Variable		Values (N = 154)
Age (yr)		62.14 ± 14.24
Male		64 (41.6)
BMI (kg/m2)		24.92 ± 3.25
Previous history
	Hypertension	116 (75.3)
	Diabetes	24 (15.6)
	Dyslipidemia	77 (50.0)
	Smoking, current	31 (20.1)
	Social alcohol drinking	36 (23.4)
Anti-hypertensive medication		92 (59.7)
	ARB	80 (51.9)
	ACEi	0 (0.0)
	BB	41 (26.6)
	Thiazide	14 (9.1)
	CCB	62 (40.3)
Number of valid ABPM reading (counts)
	24-hr	67.72 ± 7.50
	Daytime	53.47 ± 7.81
	Nighttime	14.25 ± 3.38
Proportion of valid ABPM reading (%)
	24-hr	95.82 ± 7.10
	Daytime	94.89 ± 7.78
	Nighttime	98.78 ± 6.14
Number of valid CART BP pro reading (counts)
	24-hr	61.06 ± 7.13
	Daytime	40.06 ± 8.38
	Nighttime	20.99 ± 4.60
Proportion of valid CART BP pro reading (%)
	24-hr	86.18 ± 9.64
	Daytime	80.50 ± 13.89
	Nighttime	99.14 ± 3.19

Data are presented as mean ± standard deviation or number (%).

BMI, body mass index; ARB, angiotensin II receptor blocker; ACEi, angiotensin-converting enzyme inhibitor; BB, beta-blocker; CCB, calcium channel blocker; ABPM, ambulatory blood pressure monitoring; CART BP pro, continuous ambulatory ring-type blood pressure monitor.

Agreement between CART BP pro and ABPM for 24-hour, daytime, nighttime BP, and heart rate

Table 2 summarizes the overall BP agreement, and the Bland–Altman plots are shown in Fig. 2. Compared with ABPM, the CART BP pro demonstrated a negative bias for SBP and a positive bias for DBP.

Table 2. BP value of 24-hour ambulatory BP monitoring and CART BP pro.

Variable	ABPM (mmHg)	CART BP pro (mmHg)	Mean difference (CART BP pro − ABPM) (mmHg)	SD difference (CART BP pro − ABPM) (mmHg)	|Mean difference| ≤ 5 mmHg	|SD difference| ≤ 8 mmHg
24-hr SBP	132.12 ± 15.37	124.98 ± 14.17	−7.14	10.56	51 (33.1)	77 (50.0)
24-hr DBP	77.72 ± 9.74	79.91 ± 11.71	2.20	6.95	75 (48.7)	108 (70.1)
24-hr HR	69.58 ± 9.98	67.64 ± 9.58	−1.94	2.49	137 (89.0)	151 (98.1)
Daytime SBP	134.51 ± 16.24	126.70 ± 14.32	−7.81	10.91	46 (29.9)	76 (49.4)
Daytime DBP	79.24 ± 10.46	81.48 ± 12.11	2.24	7.15	81 (52.6)	110 (71.4)
Daytime HR	71.76 ± 10.26	70.92 ± 9.95	−0.84	2.71	140 (90.9)	152 (98.7)
Nighttime SBP	123.60 ± 17.12	121.88 ± 15.19	−1.72	12.18	52 (33.8)	77 (50.0)
Nighttime DBP	72.31 ± 10.03	77.11 ± 11.94	4.80	8.11	67 (43.5)	93 (60.4)
Nighttime HR	61.70 ± 10.25	61.70 ± 9.53	0.01	2.06	149 (96.8)	153 (99.4)
Time-weighteda 24-hr SBP	131.26 ± 15.27	125.46 ± 14.09	−5.79	10.29	54 (35.1)	81 (52.6)
Time-weighteda 24-hr DBP	77.14 ± 9.54	80.35 ± 11.73	3.20	6.78	78 (50.6)	103 (66.9)
Time-weighteda daytime SBP	132.78 ± 16.51	126.41 ± 14.35	−6.37	11.44	53 (34.4)	75 (48.7)
Time-weighteda daytime DBP	77.92 ± 10.58	81.10 ± 12.17	3.19	7.60	68 (44.2)	99 (64.3)
Time-weighteda nighttime SBP	123.41 ± 17.30	121.78 ± 15.20	−1.63	12.30	51 (33.1)	70 (45.5)
Time-weighteda nighttime DBP	72.11 ± 10.19	77.02 ± 11.95	4.90	8.29	66 (42.9)	90 (58.4)

Data are presented as mean ± SD or number (%).

BP, blood pressure; CART BP pro, continuous ambulatory ring-type blood pressure monitor; ABPM, ambulatory blood pressure monitoring; SD, standard deviation; SBP, systolic blood pressure; DBP, diastolic blood pressure; HR, heart rate.

^aTime-weighted BP values were calculated according to the duration between consecutive measurements during ambulatory monitoring.

Fig. 2. Bland–Altman plot.

Bland–Altman plots showing the agreement between CART BP pro and ABPM for (A) 24-hour SBP, (B) 24-hour DBP, (C) daytime SBP, (D) daytime DBP, (E) nighttime SBP, and (F) nighttime DBP. The solid red line represents the mean difference (CART BP pro − ABPM), and the dashed lines indicate the 95% limits of agreement (mean difference ± 1.96 standard deviation).

CART BP pro, continuous ambulatory ring-type blood pressure monitor; ABPM, ambulatory blood pressure monitoring; SBP, systolic blood pressure; DBP, diastolic blood pressure.

For 24-hour SBP, CART BP pro values were on average 7.14 mmHg lower than ABPM (SD, 10.56 mmHg), with only 33.1% of participants meeting the ≤ 5 mmHg accuracy criterion. Daytime SBP showed a similar underestimation (mean difference, −7.81 mmHg; SD, 10.91 mmHg), with 29.9% within ≤ 5 mmHg. Nighttime SBP bias was smaller (−1.72 mmHg), whereas variability remained high (SD, 12.18 mmHg). For DBP, CART BP pro consistently overestimated values, with mean differences of +2.20 mmHg (24-hour), +2.24 mmHg (daytime), and +4.80 mmHg (nighttime). The agreement for DBP was modest, with 48.7% (24-hour) and 43.5% (nighttime) within ≤ 5 mmHg. The heart rates showed excellent agreement between the devices across all periods (Table 2). The mean differences were small (−1.94 bpm for 24-hour; −0.84 bpm for daytime; 0.01 bpm for nighttime), and the proportion within 5 bpm was 89.0%, 90.9%, and 96.8%, respectively. Additional analyses using time-weighted BP values, which account for unequal intervals between consecutive measurements, showed similar patterns of systolic and diastolic overestimation. Pearson correlation coefficients between ABPM and CART BP pro were moderate-to-good across all measurement periods (24-hour SBP r = 0.748, daytime SBP r = 0.752, nighttime SBP r = 0.722, 24-hour DBP r = 0.805; all P < 0.001; Supplementary Fig. 1). The hour-by-hour mean BP profile across clock time (Supplementary Fig. 2) illustrates these temporal patterns. The systolic gap between ABPM and CART BP pro was attenuated during nighttime hours, while CART BP pro showed relatively higher diastolic values than ABPM particularly at night, consistent with the period-level agreement data reported in Table 2. Assessment of proportional bias revealed significant proportional bias for daytime SBP (slope = −0.143, P = 0.020), nighttime SBP (slope = −0.139, P = 0.034), and all DBP periods (24-hour slope = +0.203, P < 0.001; daytime slope = +0.162, P = 0.002; nighttime slope = +0.200, P = 0.002), indicating that systolic underestimation was greater at lower mean BP values and diastolic overestimation was greater at higher mean BP values. Proportional bias for 24-hour SBP did not reach significance (slope = −0.093, P = 0.134) (Supplementary Table 1).

CART BP pro substantially underestimated BP variability across all measurement periods compared with ABPM (Supplementary Table 2). For 24-hour SBP, the SD was 8.68 ± 2.62 mmHg for CART BP pro versus 19.33 ± 5.21 mmHg for ABPM (CV: 7.00% vs. 14.73%), corresponding to an approximately 55% reduction in variability. This attenuation remained consistent across daytime and nighttime periods for both SBP and DBP (all P < 0.001), with similar findings observed using VIM.

Clinical characteristics according to 24-hour SBP agreement

Participants were grouped by the absolute difference between 24-hour SBP measured by ABPM and CART BP pro (≤ 5 mmHg vs. > 5 mmHg) (Table 3) to identify factors linked to measurement agreement. Fifty-one participants (33.1%) met the criteria for close agreement. Participants with close agreement were significantly older (66.04 ± 11.41 vs. 60.20 ± 15.13 years, P = 0.009) and had a lower body mass index (BMI) (24.14 ± 2.82 vs. 25.31 ± 3.39 kg/m², P = 0.025) than those with larger discrepancies. The close agreement group also included a higher proportion of men, although the difference was not significant (52.9% vs. 35.9%, P = 0.056). No significant differences in prevalence of hypertension, diabetes, dyslipidemia, smoking status, alcohol consumption, or use of antihypertensive medications were observed between the groups. The office BP and heart rate were also similar. The number and proportion of valid ABPM and CART BP pro readings over 24 hours, including daytime and nighttime periods, did not differ significantly. Thus, the differences in agreement were unlikely to be attributable to measurement completeness or recording quality.

Table 3. Baseline characteristics according to primary endpoint: absolute difference between 24-hour ABPM SBP and CART BP pro-SBP.

Variable		|24-hr SBP mean difference| ≤ 5 mmHg (n = 51)	|24-hr SBP mean difference| > 5 mmHg (n = 103)	P-value
Age (yr)		66.04 ± 11.41	60.20 ± 15.13	0.009
Male		27 (52.9)	37 (35.9)	0.056
BMI (kg/m2)		24.14 ± 2.82	25.31 ± 3.39	0.025
Previous history
	Hypertension	37 (72.5)	79 (76.7)	0.692
	Diabetes	6 (11.8)	18 (17.5)	0.480
	Dyslipidemia	26 (51.0)	51 (49.5)	1.000
	Smoking, current	12 (23.5)	19 (18.4)	0.523
	Social alcohol drinking	14 (27.5)	22 (21.4)	0.423
Anti-hypertensive medication		31 (60.8)	61 (59.2)	1.000
	ARB	28 (54.9)	52 (50.5)	0.613
	ACEi	0 (0.0)	0 (0.0)	1.000
	BB	12 (23.5)	29 (28.2)	0.568
	Thiazide	5 (9.8)	9 (8.7)	1.000
	CCB	21 (41.2)	41 (39.8)	1.000
Office SBP (mmHg)		137.87 ± 21.13	133.20 ± 18.91	0.185
Office DBP (mmHg)		84.57 ± 13.33	82.47 ± 12.96	0.356
Office HR (beats/min)		71.42 ± 13.76	71.46 ± 13.61	0.987
24-hr ABPM SBP (mmHg)		127.12 ± 13.88	134.59 ± 15.53	0.003
24-hr ABPM DBP (mmHg)		75.75 ± 8.99	78.69 ± 10.00	0.067
24-hr ABPM HR (beats/min)		69.73 ± 8.90	69.50 ± 10.52	0.889
Daytime ABPM SBP (mmHg)		129.51 ± 14.60	136.98 ± 16.50	0.005
Daytime ABPM DBP (mmHg)		77.14 ± 9.61	80.28 ± 10.75	0.069
Daytime ABPM HR (beats/min)		71.78 ± 9.39	71.75 ± 10.71	0.986
Nighttime ABPM SBP (mmHg)		119.42 ± 16.51	125.66 ± 17.12	0.032
Nighttime ABPM DBP (mmHg)		71.17 ± 10.09	72.87 ± 9.99	0.324
Nighttime ABPM HR (beats/min)		62.43 ± 8.36	61.33 ± 11.08	0.494
24-hr CART BP pro SBP (mmHg)		126.97 ± 13.84	123.99 ± 14.30	0.218
24-hr CART BP pro DBP (mmHg)		81.19 ± 11.74	79.28 ± 11.71	0.346
24-hr CART BP pro HR (beats/min)		67.87 ± 8.35	67.52 ± 10.17	0.821
Daytime CART BP pro SBP (mmHg)		128.49 ± 13.78	125.81 ± 14.56	0.268
Daytime CART BP pro DBP (mmHg)		82.48 ± 12.11	80.99 ± 12.14	0.474
Daytime CART BP pro HR (beats/min)		70.75 ± 8.90	71.00 ± 10.48	0.874
Nighttime CART BP pro SBP (mmHg)		124.23 ± 15.35	120.72 ± 15.04	0.181
Nighttime CART BP pro DBP (mmHg)		78.84 ± 12.08	76.25 ± 11.83	0.211
Nighttime CART BP pro HR (beats/min)		62.61 ± 8.34	61.25 ± 10.07	0.378
Number of valid ABPM reading (counts)
	24-hr	67.29 ± 8.43	67.93 ± 7.03	0.642
	Daytime	52.55 ± 8.86	53.93 ± 7.23	0.337
	Nighttime	14.75 ± 3.89	14.00 ± 3.10	0.236
Proportion of valid ABPM reading (%)
	24-hr	95.79 ± 8.76	95.83 ± 6.16	0.977
	Daytime	94.87 ± 9.52	94.89 ± 6.80	0.990
	Nighttime	98.34 ± 7.66	99.00 ± 5.25	0.582
Number of valid CART BP pro reading (counts)
	24-hr	61.86 ± 6.66	60.66 ± 7.35	0.311
	Daytime	40.10 ± 7.56	40.05 ± 8.80	0.971
	Nighttime	21.76 ± 4.82	20.61 ± 4.46	0.156
Proportion of valid CART BP pro reading (%)
	24-hr	87.73 ± 8.98	85.42 ± 9.90	0.150
	Daytime	82.81 ± 12.74	79.36 ± 14.35	0.133
	Nighttime	98.85 ± 4.10	99.28 ± 2.65	0.498

Data are presented as mean ± standard deviation or number (%).

ABPM, ambulatory blood pressure monitoring; SBP, systolic blood pressure; CART BP pro, continuous ambulatory ring-type blood pressure monitor; BMI, body mass index; ARB, angiotensin receptor blocker; ACEi, angiotensin-converting enzyme inhibitor; BB, beta-blocker; CCB, calcium channel blocker; DBP, diastolic blood pressure; HR, heart rate.

Predictors of close agreement in 24-hour SBP

Variables associated with close agreement in univariate comparisons were further evaluated using logistic regression analysis (Table 4). In univariate models, older age, male sex, and lower BMI were significantly associated with achieving an absolute 24-hour SBP difference ≤ 5 mmHg. In multivariable analysis (model A), older age (OR, 1.04 per year; 95% CI, 1.01–1.07; P = 0.011), male sex (OR, 2.57; 95% CI, 1.21–5.44; P = 0.014), and lower BMI (OR, 0.88 per kg/m²; 95% CI, 0.78–1.00; P = 0.048) remained independently associated with close SBP agreement. In addition, a higher 24-hour ABPM heart rate (≥ 68.5 beats/min) was independently associated with improved agreement (OR, 2.19; 95% CI, 1.02–4.70; P = 0.044). However, heart rate was modeled as a continuous variable, it was not significantly associated with close agreement (OR, 1.02 per bpm; 95% CI, 0.98–1.06; P = 0.273). When ABPM SBP SD and CART BP pro SBP SD were additionally incorporated (model B) in multivariable analysis, higher ABPM SBP SD was independently associated with lower odds of close agreement (OR, 0.89 per mmHg; 95% CI, 0.82–0.97; P = 0.010), suggesting that device agreement is reduced in patients with greater BP variability. In contrast, CART BP pro SBP SD was not independently associated with agreement (P = 0.932), further suggesting that variability derived from CART BP pro may not adequately reflect true BP fluctuations.

Table 4. Logistic regression analysis for “|24-hour SBP mean difference| ≤ 5 mmHg” group.

Predictor	Univariate OR (95% CI)	Univariate P-value	Model A: multivariate OR (95% CI)	Multivariate P-value	Model B: multivariate OR (95% CI)	Multivariate P-value
Age (yr)	1.03 (1.01–1.06)	0.019	1.04 (1.01–1.07)	0.011	1.04 (1.01–1.07)	0.010
Male (vs. female)	2.01 (1.02–3.97)	0.045	2.57 (1.21–5.44)	0.014	2.13 (0.98–4.67)	0.058
BMI (kg/m2)	0.89 (0.80–0.99)	0.038	0.88 (0.78–1.00)	0.048	0.88 (0.78–1.01)	0.060
24-hr ABPM HRa ≥ median (68.5 beats/min)	1.70 (0.86–3.36)	0.125	2.19 (1.02–4.70)	0.044	2.45 (1.11–5.40)	0.026
ABPM SBP SD (per mmHg)	0.90 (0.83–0.97)	0.004			0.89 (0.82–0.97)	0.010
CART BP pro SBP SD (per mmHg)	0.94 (0.82–1.07)	0.353			1.01 (0.87–1.17)	0.932

Model A: original model including age, sex, BMI, and HR (≥ vs. < 68.5 bpm). Model B: extended model additionally including ABPM SBP SD and CART BP pro SBP SD as variability predictors.

SBP, systolic blood pressure; OR, odds ratio; CI, confidence interval; BMI, body mass index; ABPM, ambulatory blood pressure monitoring; HR, heart rate; SD, standard deviation; CART BP pro, continuous ambulatory ring-type blood pressure monitor.

^aHR was dichotomized at the population median (68.5 bpm) as a pre-specified neutral split for exploratory purposes, without data-driven cutoff optimization. When analyzed as a continuous variable, HR did not reach significance (OR, 1.02 per bpm; 95% CI, 0.98–1.06; P = 0.273); the dichotomized result should be interpreted as exploratory.

BP range-dependent bias according to office SBP

Stratification by office SBP category showed that systolic bias varied across BP ranges (Supplementary Table 3). CART BP pro underestimated 24-hour SBP more in participants with lower office SBP (< 130 mmHg), with mean differences of −10 to −12 mmHg. This bias decreased with higher office SBP, reaching −3.16 mmHg among participants with office SBP ≥ 140 mmHg. For DBP, CART BP pro overestimated the values more in the higher office SBP categories, especially during nighttime measurements. Thus, the measurement bias of the cuffless device may depend on the baseline BP level.

Diagnostic agreement and misclassification of hypertension between ABPM and CART BP pro

Using the 24-hour ABPM threshold of ≥ 130/80 mmHg, 89 participants (57.8%) were classified as hypertensive by ABPM versus 87 (56.5%) by CART BP pro, with an overall agreement of 77.9% (κ = 0.549). Of the 89 participants classified as hypertensive by ABPM, 18 (20.2%) were classified as normotensive by CART BP pro (false negatives). Using the daytime threshold of ≥ 135/85 mmHg, this underclassification was more pronounced: 24 of 76 participants classified as hypertensive by ABPM (31.6%) were misclassified as normotensive by CART BP pro, with an overall agreement of 73.4% (κ = 0.467). McNemar's test showed no significant directional imbalance for either threshold (24-hour: P = 0.864; daytime: P = 0.349) (Supplementary Table 4).

Agreement in nocturnal hypertension and dipping phenotypes

The agreement between the ABPM and CART BP pro for nocturnal hypertension was moderate (Fig. 3). A concordant classification was observed in 76.0% of the participants (33.1% normotension and 42.9% nocturnal hypertension), whereas 24.0% were discordant. Cohen’s kappa coefficient was 0.516, indicating a moderate agreement. McNemar’s test showed no significant directional imbalance in the discordant classifications (P = 0.324), suggesting that CART BP pro did not systematically over- or under-classify nocturnal hypertension. However, the agreement between the dipping patterns was limited (Fig. 4). CART BP pro generally assigned participants to less pronounced dipping categories, with a few classified as extreme dippers. Quantitatively, the mean dipping percentage was lower with CART BP pro than with ABPM (3.74% ± 5.94% vs. 7.75% ± 10.18%), and the night-to-day SBP ratio was correspondingly higher (0.96 ± 0.06 vs. 0.92 ± 0.10) (Supplementary Table 5). The CART BP pro blunted estimates of physiological nocturnal dipping, which may limit its utility in the detailed assessment of dipping phenotypes.

Fig. 3. Agreement between ABPM- and CART BP pro-based classification of nocturnal hypertension.

Bar plot showing concordant and discordant classification of nocturnal hypertension (nighttime systolic blood pressure ≥ 120 mmHg) between ABPM and CART BP pro. The overall agreement was 76.0%, with a Cohen’s kappa coefficient of 0.516, indicating moderate agreement. McNemar’s test showed no significant directional imbalance in the discordant classification (P = 0.324).

ABPM, ambulatory blood pressure monitoring; CART BP pro, continuous ambulatory ring-type blood pressure monitor.

Fig. 4. Agreement between ABPM- and CART BP pro-based classification of dipping.

(A) Heatmap showing concordance of dipping phenotypes (riser, non-dipper, dipper, and extreme-dipper) between ABPM and CART BP pro. (B) Comparison of quantitative dipping indices, including dipping percentage and night/day SBP ratio, presented as mean ± standard deviation for ABPM and CART BP pro.

ABPM, ambulatory blood pressure monitoring; CART BP pro, continuous ambulatory ring-type blood pressure monitor; SBP, systolic blood pressure.

DISCUSSION

This prospective real-world study compared a ring-type cuffless BP device with conventional 24-hour ABPM and identified several clinically important findings. First, CART BP pro showed a consistent negative bias in systolic BP, with only one-third of participants achieving ≤ 5 mmHg agreement. This agreement varied across patient subgroups, with older age, male sex, lower BMI, and higher heart rate associated with greater SBP concordance. Second, CART BP pro showed substantially lower BP variability, approximately half of that measured by ABPM, and higher ABPM-derived BP variability independently predicted worse agreement. Third, CART BP pro missed 20.2% of 24-hour hypertension cases and 31.6% of daytime hypertension cases when standard ABPM thresholds were applied. Fourth, CART BP pro showed blunted nocturnal dipping relative to ABPM, with limited dipping phenotype concordance. Taken together, these findings indicate that CART BP pro is not yet suitable as a direct substitute for ABPM in hypertension diagnosis or circadian BP phenotyping under current performance characteristics.

The systematic underestimation of SBP observed with the CART BP pro reflects a structured negative bias relative to ABPM, rather than random measurement variability. Because the CART BP pro relies on an initial calibration of cuff-based readings, inter-device differences observed over 24 hours may also reflect physiological variability that is not fully captured by a single baseline calibration. Stratified analyses showed greater systolic underestimation in the lower-office SBP categories, suggesting a possible BP-range-dependent pattern. Proportional bias analysis showed that systolic underestimation was disproportionately greater at lower mean BP values (daytime SBP slope = −0.143, P = 0.020; nighttime SBP slope = −0.139, P = 0.034), while diastolic overestimation was more pronounced at higher mean BP values (24-hour DBP slope = +0.203, P < 0.001). This range-dependent variation may still be physiologically plausible in pulse-wave–based or pulse transit time (PTT)-derived BP estimation systems, where the relationship between waveform features and intra-arterial pressure is nonlinear and affected by vascular tone and arterial stiffness [10,15,16]. By contrast, CART BP pro generally overestimated the diastolic BP relative to ABPM. This pattern has also been observed in oscillometric and cuffless BP validation studies that compared noninvasive measurements with intra-arterial reference standards [17,18]. Oscillometric and pulse-based devices estimate the mean arterial pressure directly and algorithmically derive systolic and diastolic values. Because systolic and diastolic values are calculated from empirically derived ratios or model-based transformations of the oscillation envelope or pulse waveform, small deviations in waveform morphology or calibration slope can affect the diastolic estimation differently. Diastolic BP also corresponds to the lower-amplitude portion of the arterial waveform, making it more susceptible to signal processing variability and waveform smoothing. Consequently, a modest positive diastolic bias may reflect pressure-dependent algorithmic characteristics rather than isolated device errors. This framework also explains why higher heart rate independently predicted closer systolic agreement. The pulse-based algorithms perform more reliably with shorter, stable cardiac cycles [15,16]. Unlike many prior cuffless validation studies that reported small mean errors and concluded overall equivalence [19,20], our analysis identified specific patient characteristics and physiological contexts in which agreement systematically degrades. In addition to physiological factors related to waveform characteristics, patient-level clinical characteristics were independently associated with device agreement. Older age, male sex, and lower BMI predicted closer systolic agreement in multivariable analysis, suggesting that body composition and demographic factors modulate the fidelity of photoplethysmography-based BP estimation—likely through their effects on peripheral vascular tone and arterial wall properties. A supplementary analysis restricted to readily available clinical parameters (age group, sex, BMI, and office SBP) yielded a c-statistic of 0.77, identifying age ≥ 70 years, male sex, and office SBP ≥ 140 mmHg as the strongest predictors of close agreement, and providing clinicians with a practical pre-test framework for interpreting CART BP pro values in real-world settings (Supplementary Table 6).

The approximately 55% attenuation of BP variability, consistently reproduced across SD, CV, and VIM suggests that CART BP pro may attenuate true BP fluctuations rather than fully capturing them, a pattern consistent with calibration-anchored signal tracking. This finding has direct clinical relevance. Higher ABPM SBP SD independently predicted worse device agreement (model B: OR, 0.89; P = 0.010), indicating that the device performs less reliably in patients with the greater hemodynamic variability, particularly in those for whom accurate assessment of BP variability is most clinically important.

A physiology-based interpretation may also explain the reduced nocturnal BP decline observed with the CART BP pro. Nocturnal BP reduction is mainly driven by changes in autonomic balance, reduced sympathetic tone, peripheral vasodilation, and altered vascular compliance during sleep [21,22]. These hemodynamic changes affect pulse wave velocity and waveform morphology, which are key inputs for pulse-derived or PTT-based BP estimation algorithms [15]. During sleep, reduced vascular tone and increased arterial compliance may change the relationship between pulse timing and intra-arterial pressure. Because pulse-derived algorithms assume a stable relationship between waveform characteristics and BP, these circadian changes may reduce the sensitivity of the device in detecting the full extent of nocturnal BP decline. In addition, the blunted nocturnal dipping may reflect the limited generalizability of a single baseline calibration to nocturnal hemodynamic conditions.

Calibration drift over a 24-hour period may also contribute to the reduced detection of dipping patterns [10]. Since CART BP pro relies on the initial calibration of cuff-based measurements, any divergence between central arterial pressure and peripheral waveform characteristics during circadian transitions can increase the estimation error [23]. While nighttime bradycardia and fewer motion artifacts may improve signal quality, lower heart rates and altered stroke dynamics may reduce the sensitivity of the algorithm to subtle BP changes [16]. Although categorical nocturnal hypertension was moderately identified, the extent of physiological dipping was consistently underestimated. Thus, cuffless ring-type monitoring can capture general nocturnal BP levels but may be less effective for a detailed assessment of circadian BP variability without recalibration or algorithmic refinement. Positional hydrostatic pressure differences between the contralateral arm and finger sites during lateral decubitus sleep may have further contributed to the attenuated nighttime SBP bias and elevated nighttime variability observed in this study.

Based on the findings of this study, 3 specific clinical recommendations can be made. First, CART BP pro values should not be interpreted using standard ABPM diagnostic thresholds without device-specific adjustment. As this was associated with a 20–32% false-negative rate for hypertension diagnosis. Second, CART BP pro should not currently be used as the sole basis for circadian BP phenotyping or dipping status assessment, given the consistent blunted nocturnal dipping and poor dipping phenotype concordance. Third, CART BP pro may have a role in longitudinal tracking of BP trends within individuals, given its moderate-to-good inter-individual correlation with ABPM (r = 0.72–0.81), but this requires prospective validation with device-specific reference ranges before clinical implementation. These recommendations should be regarded as hypothesis-generating and require prospective validation using study designs tailored to each intended clinical application, including device-specific threshold validation, circadian BP phenotype assessment, and longitudinal within-person tracking.

This study has some limitations. First, this observational study was conducted at 2 centers within a single healthcare system, which may limit the generalizability of our findings. The 24-hour ABPM validity rate of 95.8% is higher than the 80–90% range typically reported in community-based studies, likely reflecting the supervised research protocol and hospital-based setting of this study, agreement metrics may therefore not be fully generalizable to unsupervised routine clinical monitoring. Second, CART BP pro measurements were calibrated only at baseline, with no recalibration during the 24-hour period, potentially affecting agreement, particularly at night. A further limitation is the absence of time-matched, point-to-point comparisons between the 2 devices. ABPM cuff inflation transiently alerts the patient and prompts brief movement cessation, whereas the CART BP pro acquires measurements without patient awareness and may capture values during physical activity. The aggregated period-mean comparisons reported in this study are therefore influenced by these behavioral and temporal confounders, and the observed daytime SBP bias may partly reflect this measurement-context difference rather than intrinsic device inaccuracy alone. Future studies using temporally matched measurements under controlled behavioral conditions would be needed to isolate true device-level discrepancies. Third, the contralateral limb placement of the 2 devices may have introduced hydrostatic pressure artifacts during sleep. In the lateral decubitus position, the vertical height difference between the upper arm and the contralateral finger can generate positional BP discrepancies of up to 10–20 mmHg, which may partly explain the attenuated nighttime SBP bias (−1.72 mmHg) and the pronounced nighttime DBP overestimation (+4.80 mmHg) observed in this study. Sleep position data were not collected, and future studies should incorporate posture monitoring to decouple hydrostatic effects from true device discrepancies. Fourth, the study population consisted predominantly of adults with suspected or established hypertension (mean age 62 years, 75% hypertensive), which may limit the generalizability of the findings to younger or normotensive populations, as well as to individuals with markedly elevated BP, in whom device performance may differ. Although the CART BP pro was calibrated to the ABPM arm's BP at baseline, thereby incorporating any inter-arm difference into the calibration reference, the possibility of time-varying inter-arm BP differences during the 24-hour monitoring period cannot be fully excluded. Among the 49 participants (31.8%) with available simultaneous bilateral upper-arm BP measurements, no significant systematic inter-arm asymmetry was observed (mean directional difference −2.1 ± 10.1 mmHg, P = 0.161), and the magnitude of inter-arm difference was not correlated with inter-device bias (r = −0.071, P = 0.627). Furthermore, both ABPM and CART BP pro are non-invasive indirect measurements of peripheral arterial pressure. Although ABPM is widely accepted as the reference standard for ambulatory BP assessment, the absence of intra-arterial reference measurements means that potential measurement error in ABPM itself cannot be entirely excluded, and central arterial pressure was not directly assessed in this study. Finally, the proprietary nature of the CART BP pro algorithm restricts a detailed analysis of waveform processing and calibration dynamics.

CONCLUSIONS

This exploratory, real-world study found that CART BP pro-derived period summaries showed lower systolic BP values and smaller nocturnal BP declines than ABPM. Because the analysis compared period-level aggregates obtained under different sampling conditions rather than simultaneous point-to-point measurements, these findings characterize the real-world agreement of summary measures and should not be interpreted as a new formal validation of individual device readings. Device-specific diagnostic thresholds and prospective, outcome-based validation studies with point-to-point comparison are required before CART BP pro can be considered for hypertension diagnosis or circadian BP phenotyping.

Abbreviations

ABPM

ambulatory blood pressure monitoring

BMI

body mass index

BP

blood pressure

CART BP pro

continuous ambulatory ring-type blood pressure monitor

CI

confidence interval

CV

coefficient of variation

DBP

diastolic blood pressure

IRB

Institutional Review Board

OR

odds ratio

PTT

pulse transit time

SBP

systolic blood pressure

SD

standard deviation

VIM

variability independent of the mean

SUPPLEMENTARY MATERIALS

Supplementary Table 1

Proportional bias analysis

Supplementary Table 2

Comparison of blood pressure variability indices between ABP monitoring and CART BP pro

Supplementary Table 3

BP value of 24-hour ABP monitoring and CART BP pro according to office BP

Supplementary Table 4

HTN classification agreement and reclassification between ABPM and CART BP pro using standard ABPM diagnostic threshold

Supplementary Table 5

Dipping indices

Supplementary Table 6

Logistic regression analysis for “|24-hour SBP mean difference| ≤ 5 mmHg” group

Supplementary Fig. 1

Scatter plots of ABPM vs. CART BP pro BP values.

Supplementary Fig. 2

Hour-by-hour mean BP profiles of ABPM and CART BP pro across 24-hour clock time.