Cuffless 24-hour ambulatory blood pressure monitoring-guided antihypertensive titration in real-world clinical practice: a prospective non-randomized study

Abstract

Background

Twenty-four-hour ambulatory blood pressure monitoring (ABPM) is the reference standard for diagnosing hypertension and assessing treatment response, but routine use remains limited by conventional cuff-based devices. We compared 6-month ambulatory blood pressure (BP) between cuffless 24-hour ABPM-guided titration and usual care in outpatient hypertension management.

Methods

We conducted a single-center, open-label, non-randomized prospective study of adults with uncontrolled hypertension between March and November 2025. Participants were assigned by the treating physician, without randomization, to CART BP pro-guided titration (CART BP arm, n = 67) or usual care (Control arm, n = 179). The primary endpoint was 24-hour mean systolic BP at 6 months, analyzed using analysis of covariance with 10 prespecified covariates. Sensitivity analyses addressed measured confounding, missing outcomes, and visit frequency.

Results

Among 162 participants with a valid 6-month ABPM recording, the adjusted between-group difference in 24-hour systolic BP was −7.35 mmHg (95% confidence interval [CI], −12.03 to −2.67 mmHg; P < 0.01), favoring the CART BP arm. Sensitivity analyses for measured confounding, missing outcomes, and visit frequency yielded similar estimates. In exploratory phenotype-specific comparisons, the controlled-BP phenotype was more prevalent in the CART BP arm (65.8% vs. 36.3%; P < 0.01), and masked hypertension was less prevalent (5.3% vs. 20.2%; P = 0.04). Adjusted office systolic BP differed by −3.14 mmHg (95% CI, −9.25 to 2.96; P = 0.31). Adverse events were more frequent per participant (20.0% vs. 8.1%; P = 0.02); exposure-adjusted incidence rate ratios were 1.78 (95% CI, 0.87 to 3.65) per participant-visit and 2.75 (95% CI, 1.35 to 5.61) per person-year.

Conclusions

In this non-randomized study, a care pathway incorporating cuffless 24-hour ABPM was associated with lower 24-hour ambulatory systolic BP at 6 months compared with usual care, whereas office systolic BP did not differ significantly between arms. Given the non-randomized design, these findings are hypothesis-generating and require confirmation in a randomized trial.

Trial Registration

ClinicalTrials.gov Identifier: NCT07454148

BACKGROUND

Hypertension is a major modifiable cardiovascular risk factor; approximately 30% of Korean adults have hypertension, and gaps in treatment and control remain [1,2]. Despite guideline emphasis on out-of-office blood pressure (BP) measurement, intermittent office BP measurement remains dominant in clinical practice. Office BP measurements cannot identify clinically important phenomena revealed by out-of-office monitoring, including white-coat and masked hypertension and early-morning BP surges [3,4].

Twenty-four-hour ambulatory blood pressure monitoring (ABPM) is the reference standard for diagnosing hypertension and stratifying cardiovascular risk [5,6]. Current European and US guidelines recommend out-of-office BP measurement to confirm hypertension and support treatment monitoring [7,8]. Twenty-four-hour and nighttime BP provide prognostic information beyond office BP [9,10,11,12]. However, sleep disruption from nocturnal cuff inflation and arm discomfort from repeated cuff constriction limit the routine use of conventional ABPM.

Cuffless BP technologies, including photoplethysmography-based devices, have emerged as potential tools for out-of-office monitoring, although validation and implementation concerns remain [13,14]. CART BP pro (Sky Labs Inc., Seongnam, Republic of Korea) is a ring-type cuffless BP device approved by the Korean Ministry of Food and Drug Safety. In a 33-participant feasibility study, the ring showed agreement with simultaneous cuff-based ABPM measurements [15]. However, real-world BP outcomes after cuffless 24-hour ABPM-guided antihypertensive titration have not been prospectively evaluated. We therefore compared adjusted 24-hour systolic BP at 6 months between CART BP pro-guided titration and usual care in adults with uncontrolled hypertension. Because assignment reflected physician and participant preference rather than randomization, the study was designed to estimate the magnitude of the between-arm difference and assess its consistency across sensitivity analyses, not to establish causality.

METHODS

Study design and ethics

This was a single-center, open-label, prospective non-randomized study (ClinicalTrials.gov NCT07454148) of consecutively enrolled adults with uncontrolled hypertension managed at Korea University Anam Hospital (Seoul, Republic of Korea). Participants were enrolled between 25 March 2025 and 5 November 2025. The study protocol was approved by the Institutional Review Board of Korea University Anam Hospital (2025AN0115). Written informed consent was obtained from each participant. The study was conducted in accordance with the Declaration of Helsinki.

Participants

Eligible participants were adults aged ≥ 19 years with uncontrolled hypertension, defined as office systolic BP ≥ 140 mmHg or office diastolic BP ≥ 90 mmHg at the index visit, who provided written informed consent. Exclusion criteria were a history of intolerance to olmesartan, amlodipine, or hydrochlorothiazide; suspected secondary hypertension; atrial fibrillation, which compromises cuffless waveform-based BP estimation; pregnancy or lactation; and any other serious medical condition that could limit study completion. Recruitment procedures were identical in both arms. The treating physician assigned each participant to a group on the basis of clinical judgment and the participant’s willingness to wear the cuffless ring, without random allocation.

CART BP pro-guided titration intervention

Participants in the CART BP arm were scheduled to undergo a 24-hour CART BP pro recording on the non-dominant hand during the titration phase of follow-up. The protocol specified a single titration-phase recording. In the final population, participants in the CART BP arm underwent a median of 2 recordings (interquartile range 1 to 2; range 0 to 3) over the 6 months, compared with a median of 1 (interquartile range 1 to 1) in the Control arm. A titration-phase recording was obtained in 54 of 65 participants in the CART BP arm, a median of 26 days (interquartile range 21 to 29) after the index visit; 23 Control-arm participants also received a titration-phase recording, a median of 76 days (interquartile range 57 to 85) after the index visit, and these participants are addressed in an as-treated analysis. The device provided summary 24-hour systolic and diastolic BP values, daytime and nocturnal mean values, BP variability, and dipping pattern. The intervention was therefore an information intervention: the device did not deliver treatment, and no device-specific BP target or titration algorithm was applied. Antihypertensive class additions, dose intensifications, and class switches were left to the discretion of the treating physician, who titrated toward the BP targets recommended in current guidelines [7,8]. Participants in the Control arm received usual care based on office BP measurements without scheduled use of CART BP pro.

Office BP measurement

Office BP was measured in routine practice by using a validated electronic sphygmomanometer (WatchBP Office AFIB; Microlife, Taipei, Taiwan). Participants rested for at least 5 minutes before measurement and were seated with the arm supported at heart level. If the initial reading was elevated, the measurement was repeated after at least 10 minutes.

Endpoints

The protocol specified that all participants in both arms undergo a 24-hour CART BP pro recording at the 6-month visit for outcome assessment. This recording was performed a median of 181 days (interquartile range 172 to 189) after the index visit in the CART BP arm and 188 days (interquartile range 176 to 196) in the Control arm. Ambulatory BP was not systematically measured at baseline; a day-0 recording was available in only 14 of 238 participants (10 in the CART BP arm and 4 in the Control arm). Accordingly, the primary endpoint was the 6-month ambulatory BP level rather than within-participant change. The primary endpoint was 24-hour mean systolic BP at 6 months measured by CART BP pro. Daytime and nighttime means were calculated over 07:00–21:59 and 22:00–06:59, respectively. Follow-up visits were counted from the day after the index visit to the last recorded study visit, and person-time accrued over the same interval; participants with no follow-up visit were excluded from exposure-adjusted analyses, which therefore included 236 of 238 participants. Each participant contributed at most one event per adverse-event category. Secondary endpoints included 24-hour mean diastolic BP, daytime and nighttime mean systolic and diastolic BP, the proportion of participants with a 24-hour mean BP below the guideline-based ambulatory threshold of 130/80 mmHg and below 2 exploratory cut points of 135/85 and 140/90 mmHg, office systolic and diastolic BP at the 6-month visit, and the 4 BP phenotypes at 6 months (controlled BP, white-coat hypertension, masked hypertension, and sustained hypertension) classified using office BP (≥ 140 mmHg threshold) and 24-hour ambulatory mean systolic BP (≥ 130 mmHg threshold). Safety endpoints included a composite of adverse events (dizziness or symptomatic hypotension, peripheral edema, headache, and palpitation) and office systolic BP < 100 mmHg during follow-up. Device-related events were identified by retrospective review of free-text adverse-event records for skin irritation, digital ischemia, local pressure injury, and paresthesia of the wearing hand.

Statistical analysis

Continuous variables were summarized as mean ± standard deviation and categorical variables as count (percentage). Between-group differences in baseline characteristics were tested using Welch’s t-test for continuous variables and Fisher’s exact test for categorical variables. The overall 4-category BP phenotype distribution at 6 months was compared using a 4 × 2 Pearson χ² test. Exploratory category-specific comparisons used Fisher’s exact test for all 4 phenotypes because of sparse cells. The primary analysis used analysis of covariance for 24-hour mean systolic BP at 6 months among participants with a valid 6-month ambulatory BP recording, with adjustment for 10 covariates: baseline systolic BP, baseline diastolic BP, age, sex, body mass index, estimated glomerular filtration rate (eGFR) calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 2021 equation [16], baseline antihypertensive class count, baseline use of a renin-angiotensin system (RAS) inhibitor, diabetes mellitus, and new-onset hypertension. The group coefficient was reported with its 95% confidence interval (CI) and 2-sided P-value.

Because treatment was not randomized, baseline balance was summarized using standardized mean differences (SMDs), with an absolute value below 0.10 indicating adequate balance. A propensity score for CART BP assignment was estimated by logistic regression using the same 10 covariates. Propensity-score overlap weighting was the principal propensity-score sensitivity analysis because it targets the population with clinical equipoise and, with a logistic propensity score, has the exact mean-balance property for modeled covariates [17,18]; the resulting estimate pertains to the overlap population. Additional analyses used stabilized inverse probability of treatment weighting and 1:1 nearest-neighbor matching without replacement within a caliper of 0.2 standard deviations of the logit propensity score. Residual imbalance was reported for each method. Weighted models used heteroscedasticity-consistent standard errors.

Missing 6-month ambulatory BP was addressed using multiple imputation by chained equations (20 imputations) in the final population under a missing-at-random assumption, inverse probability of censoring weighting derived from a logistic model for outcome ascertainment, and a tipping-point analysis in which imputed values were shifted across a range of ± 15 mmHg. Regression to the mean was examined by restricting the analysis to participants in whom the mean of all office readings recorded on the index day remained ≥ 140 mmHg systolic or ≥ 90 mmHg diastolic. To examine whether more frequent clinical contact rather than device information accounted for the difference, the primary model was refitted with the number of all follow-up visits added as a covariate and stratified at the median of 2 follow-up visits in the primary-analysis cohort; because visit frequency may lie on the causal pathway between assignment and BP, these were sensitivity analyses. An as-treated analysis addressed Control-arm participants who received a titration-phase cuffless recording. Adverse events were additionally expressed as rates per participant-visit and per person-year and compared using Poisson regression with the corresponding logarithmic offset; rates were reported descriptively when one arm contained no events.

No a priori sample-size calculation was performed because this prospective cohort comprised consecutively enrolled participants. The study was not designed to achieve prespecified precision; the primary estimate had a 95% CI half-width of 4.68 mmHg. All analyses were performed using R version 4.4 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Cohort flow

A total of 246 participants were enrolled (CART BP, n = 67; Control, n = 179). Eight participants were lost to follow-up. Among the remaining 238 participants (CART BP, n = 65; Control, n = 173), 162 (CART BP, n = 38; Control, n = 124) had a valid 6-month CART BP pro recording and were included in the primary analysis (Fig. 1). Six-month outcome ascertainment was 68.1% overall and was lower in the CART BP arm than in the Control arm (58.5% vs. 71.7%). In a logistic model for successful ascertainment adjusted for the 10 primary covariates, with the CART BP arm as the referent, the odds of ascertainment were higher in the Control arm (adjusted odds ratio [aOR], 2.08; 95% CI, 1.12 to 3.89; P = 0.02). The between-arm difference in ascertainment primarily reflected noncompletion of the scheduled 6-month device recording. Reasons for noncompletion were abstracted retrospectively from the study log and were available for all 76 participants without a valid recording. CART BP pro was not ordered by the attending physician in 48 participants, at similar rates in the 2 arms (21.5% and 19.7%). Of these 48 participants, 41 (85%) attended the 6-month visit, and their office systolic BP did not differ from that of participants who completed the measurement (135.9 vs. 137.1 mmHg, P = 0.65). CART BP pro was performed outside the protocol window in 6 participants, 16 participants did not complete it for patient-side reasons, and 6 did not complete it for clinical reasons. One Control-arm participant declined because of the visit required to return the device.

Fig. 1. Participant flow through the CART BP prospective study. A total of 246 participants were enrolled (CART BP, n = 67; Control, n = 179). Eight were lost to follow-up. Among the remaining 238 participants (CART BP, n = 65; Control, n = 173), 162 (CART BP, n = 38; Control, n = 124) had a valid 6-month CART BP pro recording.

ABPM, ambulatory blood pressure monitoring.

A day-0 cuffless recording was available in 14 participants (10 in the CART BP arm and 4 in the Control arm), of whom 9 also had a 6-month recording. Among these 9 participants, mean 24-hour systolic BP changed from 147.3 to 126.8 mmHg in the CART BP arm (n = 6) and from 141.6 to 131.1 mmHg in the Control arm (n = 3). Because this sample was small and not systematically selected, no between-arm estimate or hypothesis test was performed.

Baseline characteristics

Baseline characteristics are presented in Table 1. The arms differed substantially at assignment. Participants in the CART BP arm had higher baseline office systolic BP (148.8 ± 12.6 vs. 144.5 ± 14.6 mmHg; SMD 0.32) and a higher prevalence of new-onset hypertension (13.4% vs. 2.2%; SMD 0.42), whereas participants in the Control arm were older (SMD −0.25), had higher baseline use of a RAS inhibitor (77.7% vs. 56.7%; SMD −0.45), lower eGFR (84.2 ± 21.9 vs. 91.8 ± 14.0 mL/min/1.73 m²; SMD 0.41), and a higher prevalence of CKD (eGFR < 60 mL/min/1.73 m²; 12.3% vs. 3.0%; SMD −0.35). Seventeen of the 22 baseline characteristics had an absolute SMD above 0.10, including 13 above 0.20, indicating substantial measured baseline imbalance and motivating adjusted analyses.

Table 1. Baseline characteristics.

Characteristic		CART BP (n = 67)	Control (n = 179)	P-value	SMD
Demographics
	Age (yr)	63.9 ± 13.7	67.3 ± 13.1	0.08	−0.25
	Female sex	27 (40.3)	67 (37.4)	0.77	0.06
	Body mass index (kg/m2)	26.1 ± 4.2	26.8 ± 4.9	0.26	−0.16
	New-onset hypertension	9 (13.4)	4 (2.2)	< 0.01	0.42
Office BP at index visit (mmHg)
	Systolic BP	148.8 ± 12.6	144.5 ± 14.6	0.02	0.32
	Diastolic BP	87.5 ± 13.6	84.7 ± 13.7	0.15	0.21
Comorbidities
	Diabetes mellitus	19 (28.4)	55 (30.7)	0.76	−0.05
	Dyslipidemia	49 (73.1)	149 (83.2)	0.10	−0.25
	CKD	2 (3.0)	22 (12.3)	0.03	−0.35
	eGFR (mL/min/1.73 m2)	91.8 ± 14.0	84.2 ± 21.9	< 0.01	0.41
	Prior heart failure	3 (4.5)	24 (13.4)	0.06	−0.32
Laboratory variables (mg/dL)
	Serum creatinine	0.83 ± 0.19	0.99 ± 0.89	0.02	−0.26
	Total cholesterol	154.3 ± 35.8	149.5 ± 40.4	0.37	0.13
	LDL cholesterol	84.0 ± 30.7	76.1 ± 33.8	0.10	0.24
	HDL cholesterol	51.5 ± 12.2	51.1 ± 11.8	0.83	0.03
	Triglycerides	125.1 ± 50.4	135.7 ± 88.0	0.26	−0.15
	Fasting glucose	119.1 ± 42.1	114.5 ± 30.1	0.43	0.13
Baseline antihypertensive use
	No. of classes	1.6 ± 1.3	1.9 ± 1.0	0.08	−0.26
	RAS inhibitor	38 (56.7)	139 (77.7)	< 0.01	−0.45
	Beta-blocker	17 (25.4)	43 (24.0)	0.87	0.03
	Dihydropyridine CCB	37 (55.2)	119 (66.5)	0.14	−0.23
	Diuretic	17 (25.4)	43 (24.0)	0.87	0.03

Data are presented as mean ± standard deviation or number (%). P-values were reported to 2 decimal places. SMDs are shown for all baseline characteristics; an absolute value below 0.10 indicates negligible imbalance.

SMD, standardized mean difference; BP, blood pressure; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; LDL, low-density lipoprotein; HDL, high-density lipoprotein; RAS, renin-angiotensin system; CCB, calcium channel blocker.

Primary endpoint at 6 months

In the 10-covariate analysis of covariance model, the adjusted between-group difference in 24-hour mean systolic BP at 6 months was −7.35 mmHg (95% CI, −12.03 to −2.67 mmHg; P < 0.01), favoring the CART BP arm (Table 2). The 95% CI half-width was 4.68 mmHg. The unadjusted 6-month mean values were 124.1 ± 9.9 mmHg in the CART BP arm and 131.2 ± 12.5 mmHg in the Control arm. Exploratory subgroup analyses were not adjusted for multiplicity; point estimates were negative in all estimable subgroups, with no evidence of arm-by-subgroup interaction (P for interaction, 0.30–0.85; Supplementary Table 1).

Table 2. Outcomes at 6 months.

Endpoint		CART BP	Control	Effect (95% CI)	P-value
Primary endpoint at 6 mon: 24-hr mean systolic BP (mmHg, adjusted Δ)
	Primary 10-covariate ANCOVA	124.1 ± 9.9 (n = 38)	131.2 ± 12.5 (n = 124)	−7.35 (−12.03 to −2.67)	< 0.01
	Overlap-weighted ANCOVA	-	-	−7.68 (−11.91 to −3.44)	< 0.001
Secondary 24-hr ambulatory BP endpoints (mmHg, adjusted Δ)
	24-hr mean diastolic BP	78.9 ± 8.4 (n = 38)	79.2 ± 12.5 (n = 124)	−3.03 (−6.79 to 0.74)	0.11
	Daytime mean systolic BP	126.6 ± 10.3 (n = 38)	134.1 ± 12.2 (n = 124)	−8.08 (−12.72 to −3.43)	< 0.01
	Nocturnal mean systolic BP	121.2 ± 10.5 (n = 38)	127.3 ± 14.2 (n = 123)	−6.15 (−11.32 to −0.99)	0.02
	Daytime mean diastolic BP	81.0 ± 9.3 (n = 38)	81.5 ± 12.8 (n = 124)	−3.58 (−7.34 to 0.18)	0.06
	Nocturnal mean diastolic BP	76.4 ± 8.3 (n = 38)	76.1 ± 12.9 (n = 123)	−2.13 (−6.25 to 1.99)	0.31
	Peak hour at 16:00, systolic BP	123.1 ± 14.2 (n = 35)	134.2 ± 13.8 (n = 118)	−12.17 (−17.82 to −6.52)	< 0.01
Office systolic BP at the 6-mon visit (mmHg, adjusted Δ)
	Office systolic BP, 10-covariate ANCOVA	134.1 ± 12.9 (n = 38)	136.2 ± 16.5 (n = 124)	−3.14 (−9.25 to 2.96)	0.31
BP control thresholds at 6 mon (CART BP, n = 38; Control, n = 124; aOR, 10-covariate logistic regression)
	24-hr ABPM < 140/90 mmHg	35 (92.1)	90 (72.6)	aOR 4.74 (1.25 to 17.97)	0.02
	24-hr ABPM < 135/85 mmHg	29 (76.3)	65 (52.4)	aOR 3.36 (1.34 to 8.39)	< 0.01
	24-hr ABPM < 130/80 mmHg	17 (44.7)	42 (33.9)	aOR 2.13 (0.90 to 5.01)	0.09
Medication titration over 6 months (CART BP, n = 65; Control, n = 173)
	Δ No. of antihypertensive classes	+0.85 ± 1.04	+0.37 ± 0.81	+0.48	< 0.01
	Class addition, RAS inhibitor	22 (33.8)	27 (15.6)	RR 2.17 (1.33 to 3.52)	< 0.01
	Class addition, dihydropyridine CCB	21 (32.3)	32 (18.5)	RR 1.75 (1.09 to 2.80)	0.03
	Class addition, beta-blocker	6 (9.2)	8 (4.6)	RR 2.00 (0.72 to 5.53)	0.21
	Class addition, diuretic	13 (20.0)	26 (15.0)	RR 1.33 (0.73 to 2.43)	0.43
Adverse events (n = 238; CART BP, n = 65; Control, n = 173)
	Composite	13 (20.0)	14 (8.1)	RR 2.47 (1.23 to 4.97)	0.02
	Composite, per 100 participant-visits	5.53	3.11	IRR 1.78 (0.87 to 3.65)	0.12
	Composite, per 100 person-years	28.4	10.3	IRR 2.75 (1.35 to 5.61)	< 0.01
	Peripheral edema	4 (6.2)	0 (0.0)	Not estimable	< 0.01
	Headache	4 (6.2)	2 (1.2)	RR 5.32 (1.00 to 28.4)	0.05
	Dizziness or symptomatic hypotension	5 (7.7)	11 (6.4)	RR 1.21 (0.44 to 3.35)	0.77
	Palpitation	0 (0.0)	2 (1.2)	Not estimable	1.00
	No. of device-related events	0	0	-	-
Safety endpoint and healthcare utilization
	Office systolic BP < 100 mmHg	2 (3.1)	9 (5.2)	RR 0.59 (0.13 to 2.67)	0.73
	Cardiology visits with same-day BP measurement	2.6 ± 0.9	1.6 ± 1.1	IRR 1.58 (1.31 to 1.91)	< 0.01
	All follow-up visits	3.6 ± 2.3	2.4 ± 1.9	-	-

Data are presented as mean ± standard deviation or number (%). Adjusted Δ values are from analyses of covariance and aORs are from logistic regression, both using the 10 covariates detailed in the Methods. For adverse events, rows reporting number (%) show participant-level RRs; the composite exposure-adjusted rows show IRRs from Poisson regression with logarithmic offsets for the number of follow-up visits or person-time. Per-participant analyses included 65 CART BP and 173 Control participants; exposure-adjusted analyses included 64 and 172 participants, respectively.

CI, confidence interval; ANCOVA, analysis of covariance; BP, blood pressure; ABPM, ambulatory blood pressure monitoring; aOR, adjusted odds ratio; RAS, renin-angiotensin system; RR, relative risk; CCB, calcium channel blocker; IRR, incidence rate ratio.

Sensitivity analyses of the primary endpoint

Before weighting, all 10 propensity-score covariates had an absolute SMD above 0.10, with the largest for eGFR (0.46). Overlap weighting achieved exact mean balance for the modeled covariates, whereas stabilized inverse probability of treatment weighting left a maximum absolute SMD of 0.18 and 1:1 matching, which retained 32 of 38 possible pairs, left a maximum of 0.25 (Supplementary Table 2). In the overlap-weighted model, the adjusted difference was −7.68 mmHg (95% CI, −11.91 to −3.44; P < 0.001); this estimate pertains to the overlap population. The corresponding estimates were −8.63 mmHg (95% CI, −13.09 to −4.17) with inverse probability of treatment weighting and −8.71 mmHg (95% CI, −14.68 to −2.74) with matching.

With multiple imputation of the 76 participants without a 6-month recording, the difference in the final population of 238 participants was −6.70 mmHg (95% CI, −11.18 to −2.22; P < 0.01); inverse probability of censoring weighting gave −7.58 mmHg (95% CI, −12.10 to −3.06). In the tipping-point analysis, the between-arm difference ceased to be statistically significant when noncompleters in the CART BP arm were assigned a systolic BP 10 mmHg above their arm’s observed mean or when noncompleters in the Control arm were assigned a systolic BP 12 mmHg below their arm’s observed mean. These thresholds indicate how far the unmeasured participants would have to differ from those measured in the same arm before the result changed; they do not exclude attrition bias.

Restriction to participants whose office BP averaged over all readings recorded on the index day remained ≥ 140/90 mmHg yielded an adjusted difference of −9.36 mmHg (95% CI, −14.71 to −4.01); regression to the mean nevertheless cannot be excluded. The as-treated estimate was −7.50 mmHg (95% CI, −11.99 to −3.02). Estimates are summarized in Supplementary Table 3.

Distribution of effects across the 24-hour cycle

The hourly adjusted between-group difference (Fig. 2A) was larger during daytime hours. The largest hourly difference occurred at 16:00 (adjusted Δ, −12.17 mmHg; 95% CI, −17.82 to −6.52; P < 0.01), with estimates ranging from −11.58 to −9.93 mmHg at 12:00, 13:00, and 17:00. The adjusted between-group difference in daytime mean systolic BP was −8.08 mmHg (95% CI, −12.72 to −3.43; P < 0.01), and the nocturnal between-group difference was −6.15 mmHg (95% CI, −11.32 to −0.99; P = 0.02). The morning surge (mean systolic BP at 06:00–08:00 minus the pre-awake nadir at 04:00–06:00) was 4.4 ± 8.8 mmHg in the CART BP arm and 5.0 ± 7.7 mmHg in the Control arm (P = 0.74). The adjusted between-group difference in 24-hour mean diastolic BP was −3.03 mmHg (95% CI, −6.79 to 0.74; P = 0.11; Fig. 2B). At the exploratory 135/85 mmHg cut point, 76.3% of participants in the CART BP arm and 52.4% in the Control arm met the threshold (aOR, 3.36; 95% CI, 1.34 to 8.39; P < 0.01; Table 2).

Fig. 2. Twenty-four-hour ambulatory BP effect at 6 months. (A) Adjusted hourly between-group difference (CART BP minus Control) in 24-hour systolic BP across the 24-hour cycle. Each point represents the per-hour analysis of covariance estimate from the 10-covariate model, and vertical bars represent 95% CIs. The night-time window (22:00–06:00) is shaded. (B) Forest plot summarizing 6 endpoints with their 95% CIs and P-values.

BP, blood pressure; CI, confidence interval; SBP, systolic blood pressure; DBP, diastolic blood pressure.

Office and ambulatory BP outcomes

Office BP at the 6-month visit declined from 148.8 ± 12.6 to 134.1 ± 12.9 mmHg in the CART BP arm and from 144.5 ± 14.6 to 136.2 ± 16.5 mmHg in the Control arm (Fig. 3A). After adjustment for the same 10 covariates, the between-arm difference in office systolic BP at 6 months was −3.14 mmHg (95% CI, −9.25 to 2.96 mmHg; P = 0.31). The between-arm difference observed for ambulatory BP was not evident in office BP.

Fig. 3. Office and 24-hour ambulatory BP outcomes at 6 months. (A) Group mean office systolic BP at baseline and 6 months by arm; lines connect group means and do not represent participant-level paired trajectories. Error bars denote standard errors. (B) Distribution of the 4 BP phenotypes at 6 months: controlled BP (office < 140 mmHg and ambulatory < 130 mmHg), white-coat hypertension (office ≥ 140 mmHg and ambulatory < 130 mmHg), masked hypertension (office < 140 mmHg and ambulatory ≥ 130 mmHg), and sustained hypertension (office ≥ 140 mmHg and ambulatory ≥ 130 mmHg). (C) Distribution of the participant-level office-minus-ambulatory systolic BP difference at 6 months by arm.

BP, blood pressure.

At 6 months, the overall 4-category BP phenotype distribution differed between arms (P = 0.006). In exploratory phenotype-specific comparisons, controlled BP was more prevalent in the CART BP arm (65.8% vs. 36.3%; P < 0.01), sustained hypertension was less frequent numerically (13.2% vs. 28.2%; P = 0.08), and masked hypertension was less prevalent (5.3% vs. 20.2%; P = 0.04; Fig. 3B). Because ambulatory BP was not systematically measured at baseline, these between-arm phenotype differences are cross-sectional and do not establish phenotype conversion or an intervention effect. The participant-level office-minus-ambulatory systolic BP difference at 6 months was modestly greater in the CART BP arm (Fig. 3C), and the same limitation applies to its interpretation.

Medication titration and healthcare utilization

Antihypertensive titration intensity was greater in the CART BP arm. The mean 6-month change in antihypertensive class count was +0.85 ± 1.04 vs. +0.37 ± 0.81 (P < 0.01). Class additions were more frequent for RAS inhibitors (33.8% vs. 15.6%; relative risk [RR], 2.17; 95% CI, 1.33 to 3.52; P < 0.01) and dihydropyridine calcium channel blockers (CCBs, 32.3% vs. 18.5%; RR, 1.75; 95% CI, 1.09 to 2.80; P = 0.03; Table 2). There was no evidence of between-arm differences in beta-blocker (9.2% vs. 4.6%; P = 0.21) or diuretic additions (20.0% vs. 15.0%; P = 0.43). Cardiology visits with a same-day BP measurement were more frequent in the CART BP arm (incidence rate ratio [IRR], 1.58; 95% CI, 1.31 to 1.91; P < 0.01).

Visit frequency and co-intervention

Participants in the CART BP arm attended more follow-up visits than those in the Control arm (3.6 ± 2.3 vs. 2.4 ± 1.9) (Supplementary Table 4). With visit count added to the primary model, the adjusted difference was −8.02 mmHg (95% CI, −12.91 to −3.13; P < 0.01). Stratified estimates, using the median of 2 follow-up visits in the primary-analysis cohort as the cut point, were −7.37 mmHg (95% CI, −15.75 to 1.00) among participants at or below the median and −6.47 mmHg (95% CI, −13.26 to 0.33) among those above it (Supplementary Table 3). Because visit count was measured after assignment and may lie on the causal pathway, these analyses may be affected by post-treatment adjustment and do not isolate device information from attention or other co-interventions.

Safety

Adverse events were more frequent in the CART BP arm per participant (20.0% vs. 8.1%; RR, 2.47; 95% CI, 1.23 to 4.97; P = 0.02); the most frequent component was dizziness or symptomatic hypotension (7.7% vs. 6.4%). Because participants in the CART BP arm attended more visits, events were also expressed per unit of exposure among the 236 participants with at least one follow-up visit. The composite rate was 5.53 vs. 3.11 events per 100 participant-visits (IRR, 1.78; 95% CI, 0.87 to 3.65; P = 0.12) and 28.4 vs. 10.3 events per 100 person-years (IRR, 2.75; 95% CI, 1.35 to 5.61; P < 0.01). The per-visit estimate was attenuated and imprecise, whereas the person-time estimate remained elevated; these exposure scales do not establish whether the difference reflects surveillance opportunity, the underlying event rate, or both. Peripheral edema occurred only in the CART BP arm, so a rate ratio was not estimable. For dizziness or symptomatic hypotension, there was no evidence of a between-arm difference on the per-participant or per-visit scale (7.7% vs. 6.4%; per-visit IRR, 0.89; 95% CI, 0.32 to 2.44). Exposure-adjusted rates for the composite and selected events are shown in Supplementary Table 5. Office systolic BP below 100 mmHg occurred in 2 of 65 participants (3.1%) in the CART BP arm and 9 of 173 (5.2%) in the Control arm (RR, 0.59; 95% CI, 0.13 to 2.67); this estimate was imprecise.

One participant in the CART BP arm reported transient digital paresthesia, which the investigators judged unrelated to device wear. No participant had skin irritation, digital ischemia, or local pressure injury; no event was judged device-related, and no device-related discontinuation occurred.

DISCUSSION

To our knowledge, this is the first prospective study to evaluate a cuffless 24-hour ABPM platform for guiding antihypertensive titration in routine clinical practice. We report 3 principal findings. First, in this cohort of 246 adults, the primary analysis among 162 participants with a valid 6-month ABPM recording found an adjusted between-arm difference of −7.35 mmHg in 24-hour mean systolic BP; similar estimates were obtained in propensity-score, missing-data, and visit-frequency sensitivity analyses. Second, between-arm differences were observed during both daytime and nighttime hours, and the CART BP arm had a higher prevalence of controlled BP (65.8% vs. 36.3%; P < 0.01) and a lower prevalence of masked hypertension (5.3% vs. 20.2%; P = 0.04). Third, the adjusted between-arm difference in office systolic BP was −3.14 mmHg (P = 0.31), smaller than that observed for ambulatory systolic BP.

Previous trials of self-monitoring and telemonitoring have reported office systolic BP reductions that varied by intervention and follow-up. The Home BP Telemonitoring and Pharmacist Management trial reported adjusted between-group differences of −10.7 mmHg at 6 months and −6.6 mmHg at 18 months [19], and TASMINH4 reported 12-month differences versus usual care of −3.5 mmHg with self-monitoring and −4.7 mmHg with self-monitoring plus telemonitoring [20]. Both trials used office BP as the endpoint and did not characterize ambulatory BP. In the present study, mean office systolic BP at 6 months was 134.1 mmHg in the CART BP arm and 136.2 mmHg in the Control arm, with an adjusted between-arm difference of −3.14 mmHg. The adjusted difference was larger for ambulatory than for office systolic BP, suggesting that a care pathway incorporating cuffless monitoring may provide information not captured by office measurement; this comparison does not establish a causal effect or comparative measurement sensitivity.

Hourly differences were larger during daytime hours. The antihypertensive effect can vary across the dosing interval [21], and daytime BP varies with physical activity and autonomic influences [22]. Because medication timing, physical activity, and sleep timing were not recorded, this pattern should be considered exploratory. Ambulatory monitoring provides a complementary assessment of the 24-hour BP response to treatment [23]. Masked and sustained hypertension are associated with adverse prognosis in cohort studies [3,24,25], and the 24-hour ambulatory BP profile is prognostically informative [26]; however, the lower prevalence of masked hypertension in the CART BP arm at 6 months is cross-sectional and should not be interpreted as phenotype conversion or cardiovascular risk reduction.

Medication intensification may have contributed to the observed between-arm difference (change in antihypertensive class count, +0.85 vs. +0.37). Home BP monitoring has been associated with less therapeutic inertia [27], although inertia can persist after ambulatory monitoring in routine practice [28]. Because mediation was not assessed, the contribution of prescribing changes cannot be separated from adherence, visit frequency, or other co-interventions. The added classes were predominantly RAS inhibitors and dihydropyridine CCBs, both recommended as first-line therapies [7,8,29]. Korean observational studies have evaluated these drug classes in routine practice [30,31].

This study has important limitations that constrain the inferences that can be drawn from it.

First, treatment was assigned by physician and participant preference rather than at random, and the arms differed materially at baseline, with 13 of 22 baseline characteristics showing an absolute SMD above 0.20. Participants in the CART BP arm were younger; lower wearable-device use among older adults has been reported [32], but other unmeasured factors may have influenced arm assignment. Propensity-score overlap weighting achieved exact mean balance for the modeled covariates and changed the estimate little, but unmeasured confounding cannot be excluded. Because baseline office systolic BP was higher in the CART BP arm, regression to the mean is a further concern; restriction to participants with confirmed elevated baseline office BP did not attenuate the estimate, but this does not exclude it.

Second, ambulatory BP was not systematically measured at baseline. The primary endpoint is therefore an ambulatory BP level at 6 months rather than a within-participant change, and the cohort may include participants with white-coat hypertension whose ambulatory BP was never elevated. If such participants were unevenly distributed between the arms, which is plausible given the higher office BP at entry in the CART BP arm, they could contribute to both the observed difference and the reclassification of BP phenotypes. The phenotype and masked-hypertension findings should therefore be interpreted as cross-sectional descriptions at 6 months and as hypothesis-generating rather than as evidence of phenotype conversion.

Third, measurement was asymmetric between the arms. Participants in the CART BP arm wore the device during titration and again at 6 months, whereas most participants in the Control arm wore it only at the 6-month visit, so differences in familiarity and wearing adherence could have influenced the recordings. Although the outcome was derived by the same automated algorithm in both arms and titration targeted guideline BP thresholds rather than any device-specific target, an element of measurement in the direction of the intervention cannot be excluded. Relatedly, cardiology visits with a same-day BP measurement were 1.58-fold more frequent in the CART BP arm and all follow-up visits were also more frequent (3.6 vs. 2.4), so the information provided by the device cannot be fully separated from the effect of more frequent clinical contact; adjustment for and stratification by visit frequency did not attenuate the estimate, but a non-randomized design cannot disentangle these components definitively. Adherence to antihypertensive therapy was not measured, and differential adherence remains an alternative explanation for the difference in ambulatory BP.

Fourth, a 6-month ambulatory recording was obtained in 68.1% of participants, and ascertainment was lower in the CART BP arm. Multiple imputation and inverse probability of censoring weighting produced similar estimates, but both rely on modeling assumptions and selection bias cannot be excluded. Within the CART BP arm, 41.5% did not complete the repeat 6-month measurement. These participants were numerically older and more often diabetic, although neither difference reached statistical significance. The most frequent reason in both arms was that the attending physician did not order CART BP pro, which was left to the discretion of the treating physician. Most of these participants attended the 6-month visit with office BP no lower than in those who completed the measurement, which argues against a clinical decision that the test was not needed and suggests instead that ambulatory monitoring is not yet routine in this setting. The excess noncompletion in the CART BP arm came mainly from measurements performed outside the protocol window, indicating delay rather than refusal, and only one participant declined because of the device-return visit. Uptake therefore depends more on whether ambulatory monitoring is ordered than on whether patients will wear the device.

Finally, this was a single-center study, which limits generalizability to community primary care. No a priori sample-size calculation was performed; the resulting 95% CI half-width of 4.68 mmHg for the primary endpoint reflects the available cohort.

CONCLUSIONS

A care pathway incorporating cuffless 24-hour ABPM-guided antihypertensive titration was associated with a 7.35-mmHg lower 24-hour systolic BP and a higher prevalence of controlled BP at 6 months than usual care, with similar estimates in propensity-score, missing-data, and visit-frequency sensitivity analyses. Because the design was non-randomized and ambulatory BP was not systematically measured at baseline, these findings should be regarded as hypothesis-generating. They nevertheless provide estimates of magnitude and implementation, including repeat-monitoring completion, treatment intensification, and safety surveillance, that inform the design and interpretation of the ongoing randomized CART-BP trial (ClinicalTrials.gov NCT07465549), which is designed to test whether cuffless 24-hour ABPM-guided titration improves ambulatory BP control when treatment is randomly allocated and ambulatory BP is measured at both baseline and follow-up.

Abbreviations

ABPM

ambulatory blood pressure monitoring

ANCOVA

analysis of covariance

aOR

adjusted odds ratio

BP

blood pressure

CCB

calcium channel blocker

CI

confidence interval

CKD

chronic kidney disease

CKD-EPI

Chronic Kidney Disease Epidemiology Collaboration

DBP

diastolic blood pressure

eGFR

estimated glomerular filtration rate

HDL

high-density lipoprotein

IPTW

inverse probability of treatment weighting

IRR

incidence rate ratio

LDL

low-density lipoprotein

RAS

renin-angiotensin system

RR

relative risk

SBP

systolic blood pressure

SMD

standardized mean difference

SUPPLEMENTARY MATERIALS

Supplementary Table 1

Exploratory subgroup analyses of the primary endpoint

Supplementary Table 2

Baseline balance and PS diagnostics in the primary-analysis cohort

Supplementary Table 3

Sensitivity analyses of the primary endpoint

Supplementary Table 4

Cuffless recording dose, timing, and clinical contact

Supplementary Table 5

Exposure-adjusted rates for the adverse-event composite and selected events