Variability of SBP and cardiovascular outcomes in hypertensive patients

Abstract

Background

Visit-to-visit variability (VVV) in blood pressure (BP) has been linked to adverse cardiovascular outcomes, but the number of measurements needed for reliable assessment is unclear. This study aimed to determine the minimum number of systolic BP (SBP) readings required to quantify VVV and predict major adverse cardiovascular events (MACE) in patients with hypertension.

Methods

We have conducted a multicenter retrospective cohort study using electronic health records of the Korea University Medical Center database. Overall, data from 4,480 patients with hypertension who had more than nine BP readings during a maximum period of 2 years were identified. This study used the coefficient of variation (CV) for analysis of variability and its correlation with MACE over a 3-year follow-up.

Results

The area under the receiver-operating characteristic curve (AUC) for MACE prediction plateaued at around 0.60 with five readings, indicating minimal gain with additional readings. The mean CV that best predicted the 3-year MACE for five SBP readings was 8.2. Patients with high SBP variability (CV > 8, based on five readings) had a hazard ratio of 1.89 (95% confidence interval, 1.48–2.41) for 3-year MACE, compared with those with low SBP variability. In multivariable analysis, high SBP variability remained an independent risk factor of MACE.

Conclusion

To conclude, a minimum of five SBP readings appear sufficient to estimate visit-to-visit variability associated with cardiovascular risk in patients with hypertension. Assessment of SBP variability using routinely collected clinic measurements may complement conventional risk evaluation, although further validation in broader hypertensive populations is warranted.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05800-1.

Introduction

Visit-to-visit variability (VVV) in various metabolic parameters have been associated with increases in adverse cardiovascular outcomes in patients with hypertension [1–4]. Variability in systolic blood pressure (SBP), in particular, is a well-known contributor to adverse outcomes such as stroke, myocardial infarction (MI) and cardiovascular death [1, 5–7]. Patients with hypertension are prone to variations in SBP owing to numerous factors such as drug compliance, hormonal influence or lifestyle factors [8].

Theoretically, the number of blood pressure (BP) measurements deployed to calculate the VVV in order to identify individuals at risk could increase indefinitely. However, as this approach is impracticable in real-life clinical settings, there is a need for an optimal number of measurements that is both feasible in clinics and able to adequately predict adverse outcomes.

A previous study has shown that for patients with diabetes mellitus (DM), five BP measurements can provide reasonable predictions for MACE, while another has suggested six measurements based on a cohort study of patients without prior cardiovascular events [9, 10]. This paper aims to evaluate the number of SBP measurements needed to reliably predict adverse cardiovascular outcomes for patients with hypertension.

Methods

Study design and population

This multicenter retrospective cohort study was conducted at three tertiary centers in South Korea (Korea University Anam Hospital, Korea University Guro Hospital and Korea University Ansan Hospital). The study used the Observational Medical Outcomes Partnership Common Data Model (OMOP-CDM) for analysis, which was provided by the Observational Health Data Sciences and Informatics Partnership to standardize hospital electronic health records (EHR) (https://github.com/OHDSI/CommonDataModel/). The International Classification of Diseases, 10th Revision (ICD-10) codes were used for classification of clinical diagnoses. In the OMOP-CDM database, a unique concept identifier (ID) is assigned to each clinical diagnosis, which then correlates with a specific ICD-10 code. The data used for the study were stored on Microsoft’s structured query language server and accessed via direct queries.

People who had been diagnosed with hypertension from January to December 2017 were included in this study. (Fig. 1) Hypertension was defined as having baseline SBP ≥ 140 mmHg or DBP ≥ 90 mmHg, the use of antihypertensive drugs, or OMOP-CDM concept ID for hypertension. The index day was defined as the date of initial BP assessment in the outpatient clinic, following the diagnosis of hypertension. People who had nine or more BP readings during a maximum period of 2 years prior to the index day were identified. VVV was calculated using the BP readings obtained from the 2-year period preceding the index day. Patients with a history of MI, stroke, heart failure, or all-cause death before the index day, or those with missing data for key laboratory values (total cholesterol [TC], high-density lipoprotein cholesterol [HDL-C], low-density lipoprotein cholesterol [LDL-C], triglyceride, glucose, HbA1c, serum creatinine, or albuminuria levels) within 1 month of the index day were excluded from the study. The study was approved by the institutional review board of each participating center (Korea University Anam Hospital IRB no., 2023AN0230). As the study utilized anonymized data using a retrospective design, the requirement for written informed consent was waived. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Fig. 1.

Study flowchart. People with hypertension who had nine or more SBP readings during a maximum period of 2 years prior to the index day were enrolled (modelling phase). The participants were followed for 3 years after the index day (monitoring phase). BP, blood pressure; CV, coefficient of variation; MACE, major adverse cardiovascular events

BP measurements and study outcomes

The patients’ BP was measured by qualified personnel using an automatic sphygmomanometer following a minimum of 5-minute rest in a seated position. They were directed to refrain from smoking, alcohol consumption, and caffeine intake for 30 min before the BP measurement. Serial SBP and DBP readings were obtained retrospectively from the EHR database at baseline and each offline visit during the modeling phase. BP-measuring personnel were trained to measure the BP three times and take the average value to be recorded in the EHR. Only a single BP reading was recorded per visit; hence, the number of BP readings corresponded to the number of outpatient clinic visits during the modeling phase. For analysis of variability, the last readings (e.g. the last 3, 4, 5, … or 9) before the index date were used in order to better represent the patient’s current health status, and also to provide better data accessibility in clinical settings.

Diabetes was defined as having a glycated hemoglobin (HbA1c) level of 6.5% or higher, using oral hypoglycemic agents, or having a specific concept ID for diabetes in the OMOP-CDM database. Dyslipidemia was defined as serum TC ≥ 240 mg/dl, LDL-C ≥ 160 mg/dl, TG ≥ 200 mg/dl, or HDL-C ≤ 40 mg/dl; the use of lipid-lowering drugs; or OMOP-CDM concept ID for dyslipidemia. The cardiovascular risk for each patient was assessed using the Systemic Coronary Risk Evaluation 2 (SCORE2) model, in conjunction with blood creatinine and albuminuria levels; the outcomes were classed as low, moderate, high, or very high. All blood samples were collected after overnight fasting.

The primary outcome was the occurrence of major adverse cardiovascular events (MACE), which comprised of cardiovascular mortality, MI, stroke, and hospitalization for heart failure. Patients were followed up for 3 years after the index day (monitoring phase; Fig. 1), and data on the dates and causes of mortality were extracted from the EHR database’s death certificates. MI was defined as a serum creatinine kinase-myocardial band level above the upper limit of normal with a rising and/or falling pattern during hospitalization in the emergency department. Stroke was defined as a brain MRI revealing acute, subacute or recent cerebral infarction, or having a matching OMOP-CDM concept ID. Hospitalization for heart failure was defined as a case of hospital admission for ≥ 3 days, including the emergency department, and a N-terminal pro-brain natriuretic peptide level of ≥ 300 pg/ml during hospitalization.

Statistical analysis

Categorical variables are shown as numeric values (percentages), while continuous variables are shown as means ± standard deviations. Continuous variables were compared using parametric (Student’s t test) and nonparametric (Mann-Whitney U) tests, and categorical variables were compared using the χ² test or Fisher’s exact test as appropriate.

Intraindividual VVV was defined as the coefficient of variation (CV), calculated as 100 × σ/µ, where σ is the standard deviation (SD) and µ is the mean of multiple SBP readings. Peak size, the difference between the maximum and mean SBP, was used to determine the presence of extreme values. Differences between the groups with various SBP readings for CV (i.e., 3 to 9) were analyzed using one-way analysis of variance. The incremental predictive value of a model using different numbers of SBP readings was assessed by the area under the curve (AUC) of the receiver-operating curves (ROC). The differences between AUCs were compared using the DeLong test [11].

Cumulative incidences were determined using the Kaplan-Meier method with censoring estimates, and the time-to-event outcomes were analyzed using the complete follow-up period. Event censoring occurred at the time of death or final follow-up. Cox proportional hazards model analysis was employed to calculate the hazard ratios (HR) and 95% confidence intervals (CI) for each outcome. The proportional hazard assumption was validated using the Schoenfeld residual test, and no significant violations were identified.

Results

Patient demographics

The characteristics of the 4,480 patients included in this study are summarized in Table 1. The mean age was 65.3 ± 12 years and 52.9% were men. Average BP at baseline were 130/75 mm Hg, 75.2% had a history of diabetes and 85.9% had dyslipidemia. Most patients (92.9%) were labelled as having low-to-moderate risk using the SCORE2 model. In this present group of patients with hypertension, the majority of patients (99.7%) were prescribed with antihypertensive agents, with the most common class of drugs being renin-angiotensin-system inhibitors (97.1%). The median number of total BP measurements was 14 times (IQR; 11–19).

Table 1.

Baseline characteristics

	Overall
	(n = 4,480)
Age (years)	65.3±12.4
Male, n (%)	2,368 (52.9)
Alcohol	1,211 (27.0)
Smoking	891 (19.9)
Baseline SBP (mmHg)	130.5 ± 14.5
Baseline DBP (mmHg)	75.0 ± 12.2
Comorbidities
Diabetes mellitus	3,371 (75.2)
Dyslipidemia	3,895 (86.9)
Cardiovascular risk using SCORE2
Low to moderate risk	4,163 (92.9)
High risk	292 (6.5)
Very high risk	25 (0.6)
Laboratory measurements
Total cholesterol, mg/dL	155.3 ± 35.9
Triglyceride, mg/dL	144.8 ± 91.2
Glucose	123.7 ± 39.8
Hemoglobin A1c	6.6 ± 1.2
Medications
Oral hypoglycemic agents	2,192 (48.9)
Insulin	1,165 (26.0)
Antihypertensive drugs	4,466 (99.7)
RAS inhibitors	4,351 (97.1)
Dihydropyridine calcium channel blockers	3.981 (88.9)
Beta-blocker	2,536 (56.6)
Diuretics	1,724 (38.5)
Statin and/or ezetimibe	3,424 (76.4)

Values are presented as numbers (percentages) or means ± standard deviation

Variability in visit-to-visit SBP readings

Table 2 illustrates the differences in variability metrics based on using different number of clinic visits. The documented maximum SBP showed a gradual increase as more SBP values were included in the calculation, from 140.3 ± 15.4 mmHg for 3 visits to 147.9 ± 15.1 mmHg for 9 visits (p < 0.001). The mean of SBP readings remained relatively constant. Both the SD (9.8 ± 6.2 for 3 visits; 11.3 ± 4.5 for 9 visits; p < 0.001) and the CV (7.5 ± 4.6 for 3 visits; 8.7 ± 3.3 for 9 visits; p < 0.001) increased incrementally with the number of visits used, suggesting that the calculated visit-to-visit variability increased as the number of measurements increased. For both SD and CV, calculation using all available readings yielded the maximal values (SD, 11.7 ± 4.1; CV, 9.0 ± 3.0).

Table 2.

Variability of SBP by the number of measurements used

	3 SBP readings	4 SBP readings	5 SBP readings	6 SBP readings	7 SBP readings	8 SBP readings	9 SBP readings	All SBP readings	P -value
Mean of SBP readings (mmHg)	130.9 ± 12.9	130.5 ± 12.1	130.3 ± 11.6	130.2 ± 11.3	130.1 ± 11.0	130.0 ± 10.8	130.0 ± 10.6	130.5 ± 14.5	<0.001
Maximum of SBP readings (mmHg)	140.3 ± 15.4	142.3 ± 15.2	143.8 ± 15.2	145.1 ± 15.1	146.2 ± 15.2	147.1 ± 15.2	147.9 ± 15.1	151.7 ± 16.1	< 0.001
SD of SBP readings	9.8 ± 6.2	10.3 ± 5.7	10.7 ± 5.3	10.9 ± 5.0	11.0 ± 4.8	11.2 ± 4.6	11.3 ± 4.5	11.7 ± 4.1	< 0.001
CV of SBP readings	7.5 ± 4.6	7.9 ± 4.2	8.2 ± 3.9	8.3 ± 3.7	8.5 ± 3.5	8.6 ± 3.4	8.7 ± 3.3	9.0 ± 3.0	< 0.001
Days between the first and last SBP readings (mean)	116.0 ± 73.8	169.4 ± 97.7	222.1 ± 116.9	274.4 ± 134.5	327.7 ± 150.9	380.2 ± 167.3	434.5 ± 185.0	689.0 ± 33.0	< 0.001
Days between the first and last SBP readings (median)	110.0 (59.5-173.5)	168.0 (96.0-234.5)	216.0 (134.0-301.0)	278.0 (180.0-371.0)	336.0 (217.0-444.5)	385.0 (261.5-518.0)	450.0 (302.0-586.0)	696.5 (666.0-717.0)	< 0.001
Days between each SBP reading (mean)	38.7 ± 24.6	42.3 ± 24.4	44.4 ± 23.4	45.7 ± 22.4	46.8 ± 21.5	47.5 ± 20.9	48.3 ± 20.6	49.8 ± 18.3	< 0.001
Days between each SBP reading (median)	28.0 (5.0-64.0)	37.5 (14.0-61.5)	34.0 (12.0-67.0)	41.0 (16.5-65.0)	38.0 (15.0-70.0)	42.2 (19.0-67.0)	42.0 (19.0-70.0)	42.0 (21.5-68.0)	< 0.001

Values are presented as means ± standard deviation or median(interquartile range). Multiple intraindividual SBP readings were used to estimate visit-to-visit variability. Differences between the groups were analyzed using one-way analysis of variance

CV = coefficient of variation; SBP = systolic blood pressure; SD = standard deviation

Number of SBP readings and clinical outcomes

As more BP measurements were used to calculate variability, the variability metric’s predictive power for MACE enhanced (Table 3; Fig. 2). The AUC for prediction of 3-year MACE using 3 SBP readings was 0.56, which increased gradually to 0.60 using 9 SBP readings. As with SD and CV, the AUC using all available SBP readings (0.64) was the highest. The hazard ratio for MACE also increased with the number of readings used. Using 3 SBP readings, the adjusted hazard ratio was 1.41, which increased to 1.63 using 9 SBP readings and 2.02 for all SBP readings.

Table 3.

Number of SBP readings used for CV and the prediction of MACE

	Coefficient of variation	AUC to predict 3-year MACE	Cutoff point of CV	Una djusted HR (95% CI)	Adjusted* HR (95% CI)	P-value for consecutive no. of SBP readings
CV of 3 SBP readings	7.5 ± 4.6	0.56	7.52	1.71 (1.34-2.18)	1.42 (1.09-1.84)	-
CV of 4 SBP readings	7.9 ± 4.2	0.56	6.84	1.72 (1.34–2.22)	1.34 (1.03–1.75)	0.875
CV of 5 SBP readings	8.2 ± 4.0	0.58	8.40	1.89 (1.48–2.41)	1.51 (1.16–1.96)	0.032
CV of 6 SBP readings	8.3 ± 3.7	0.59	7.91	1.96 (1.53–2.52)	1.55 (1.19–2.01)	0.480
CV of 7 SBP readings	8.5 ± 3.6	0.59	8.64	1.91 (1.50–2.43)	1.54 (1.19–1.99)	0.910
CV of 8 SBP readings	8.6 ± 3.4	0.59	8.33	1.88 (1.47–2.40)	1.61 (1.23–2.10)	0.378
CV of 9 SBP readings	8.7 ± 3.3	0.60	9.26	2.03 (1.60–2.59)	1.65 (1.27–2.13)	0.330
CV of all SBP readings	9.0 ± 3.0	0.64	9.27	2.81 (2.18–3.62)	2.06 (1.57–2.71)	-

Fig. 2.

Receiver-operating characteristic curves in different sets of multiple SBP readings. The prediction accuracy from four to five readings was statistically significant. *P<0.05. AUC, area under the curve; CV, coefficient of variation; MACE, major adverse cardiovascular events; SBP, systolic blood pressure

As the number of visits used increased from 3 to 9, the difference for the predictive value between two consecutive numbers of readings was only significant between 4 and 5 readings (AUC, 0.58 vs. 0.59; p = 0.032). The mean CV using 5 SBP readings was 8.2 ± 4.0, and the cutoff CV value for prediction of MACE was 8.403. The crude hazard ratio for MACE from having a CV > 8, using 5 SBP readings, was 1.89 (95% confidence interval, 1.48–2.41).

Around 45% of the total population exhibited a CV higher than 8 using 5 SBP readings (Supplemental Table S1). Patients with high CV tended to be older and more likely to be female. Baseline SBP were similar between the two groups (130.6 ± 15.1 vs. 130.4 ± 14.1, p = 0.701), as were the proportion of patients with DM or dyslipidemia.

Subsequently, using a cutoff value of 8 for CV from 5 consecutive SBP readings, 3-year MACE was significantly higher in those with elevated CV (7.8% vs. 4.4%, p < 0.001). (Table 4; Fig. 3) Among secondary outcomes, risks of MI (2.2% vs. 1.2%, p = 0.010) and heart failure hospitalization (6.5% vs. 3.3%, p < 0.001) were significantly elevated in the high-CV group. Supplemental Table 2 illustrates that after multivariable adjustment, having a CV of > 8 from 5 SBP readings yields a 50% increase in 3-year MACE risk. Age, current smoking status, high SCORE2 risk and high creatinine level were also identified as risk factors for MACE. Notably, additional sensitivity analysis using SD instead of CV yielded similar results to using CV in that there was a significant difference between using 4 band 5 readings with similar AUCs (Supplemental Table S3). The cutoff value for SD using 5 readings was 11, with comparable incidences of cardiovascular outcomes during follow-up as using a cutoff of 8 for CV (Supplemental Table S4).

Table 4.

Incidence of the primary and secondary outcomes at 3-years after the index date using CV

	Overall	CV of 5 SBP readings ≤8	CV of 5 SBP readings >8	Log-rank P
	(n=4,480)	(n=2,473)	(n=2,007)
Primary outcome
MACE	267 (6.0%)	111 (4.4%)	156 (7.8%)	<0.001
Secondary outcomes
Cardiovascular death	28 (0.6%)	16 (0.6%)	12 (0.6%)	0.863
Myocardial infarction	74 (1.7%)	30 (1.2%)	44 (2.2%)	0.010
Cerebrovascular accident	21 (0.5%)	9 (0.4%)	12 (0.6%)	0.246
Hospitalization for heart failure	213 (4.8%)	82 (3.3%)	131 (6.5%)	<0.001

Values are presented as numbers (the Kaplan-Meier estimate of cumulative incidence). CI, confidence interval; HR, hazard ratio

Fig. 3.

Three-year cumulative incidence of MACE according to high vs. low SBP variability. CV, coefficient of variation; HR, hazard ratio; MACE, major adverse cardiovascular events; SBP, systolic blood pressure

Discussion

In this cohort of patients with hypertension, we found that long-term visit-to-visit variability was a significant independent predictor of MACE. This finding aligns with a broad literature linking BP variability to adverse outcomes. Elevated visit-to-visit variability in SBP has been associated with increased risks of stroke, MI, heart failure and mortality across diverse populations [1, 12, 13]. Notably, these associations often persist even after adjusting for mean BP, suggesting that BP variability offers a risk prediction independent of mean BP [13]. A post-hoc analysis of the ALLHAT trial on patients with hypertension and one or more CV risk factors found that patients in the highest quintile of SBP variability (SD ≥ 14.4 mmHg) had 30% and 46% higher risk of coronary event and stroke, respectively, compared with those with stable BP, even after controlling for mean SBP [1]. Similarly, a recent meta-analysis reported that increase in SBP variability was associated with increased risks of stroke, MI and all-cause mortality [13]. Mechanistically, exaggerated BP fluctuations in patients with hypertension may reflect episodes of poor blood pressure control due to lack of medication adherence, or underlying pathophysiology such as arterial stiffness and autonomic dysfunction, which can exacerbate target-organ damage [14].

A practical challenge in leveraging BP variability is determining how many readings are needed to reliably define an individual’s variability. In theory, using more BP measurements should improve the precision of variability estimates, as was shown in our data. Unfortunately, indefinitely high number of measurements is not feasible in routine practice. Our analysis suggests that as few as five serial SBP readings provide a robust and clinically useful estimate of long-term variability for risk stratification, with only minimal gains in predictive power beyond the fifth reading. In fact, the discrimination of MACE plateaued after five visits, indicating a point of diminishing returns. This result is consistent with prior evidence in patients with diabetes; Kim et al. recently demonstrated that a minimum of five BP measurements was sufficient to reliably predict 3-year cardiovascular outcomes using visit-to-visit SBP variability [9]. The authors noted that predictive accuracy for MACE improved markedly between three and five readings, while little further improvement was achieved using seven or more readings. Notably, using five visits to calculate SBP coefficient of variation enabled clear risk stratification in diabetics, as patients with high variability (CV value of > 9.0) had nearly double the incidence of events compared to those with stable BP [9]. While Kim et al. only evaluated odd-number visits (3,5,7 and 9), our findings refine these findings by comparing all integer numbers between 3 and 9. Our results suggest that their decision to conclude five visits are sufficient are well grounded. Further, our findings broaden this perspective to patients with hypertension, indicating that approximately five clinic BP readings over time may be an optimal and practical threshold for capturing visit-to-visit variability in routine practice. This provides an important insight for clinicians, as it sets a feasible target for the number of measurements needed to incorporate variability into healthcare practice without overburdening either the patient or the clinician. Nonetheless, it should be noted that our study population had a high prevalence of diabetes, leading to additional subgroup analyses according to diabetes status. (Supplemental Table S5) In patients with diabetes, the predictive performance of visit-to-visit SBP variability according to the number of readings closely resembled that of the overall cohort, supporting the robustness of the five-measurement threshold in this group. In contrast, the pattern was less consistent among patients without diabetes, possibly reflecting lower event rates and weaker associations between BP variability and cardiovascular risk in lower-risk populations. These findings suggest that the proposed minimum number of measurements may be particularly applicable to hypertensive patients with elevated cardiometabolic risk such as coprevalent diabetes, and further validation in broader community-based hypertensive populations is warranted.

Despite robust evidence linking visit-to-visit variability with outcomes, current clinical guidelines have not yet formally incorporated variability metrics into hypertension management. Most guidelines still focus on achieving a set target mean BP, either by office measurement of 24-hour ambulatory BP measurement [15–17]. Currently, there are no clear recommendations on how to interpret or address high visit-to-visit variability. Our findings, together with prior studies, suggest that there is value in recognizing high BP variability as a marker of cardiovascular risk. Patients with hypertension are particularly susceptible to fluctuations in BP, and may benefit from closer monitoring, strengthened patient education or therapeutic adjustments [8]. With the widespread adoption of EHR, it is increasingly feasible to implement automated tracking of BP variability in clinical practice. EHR systems can effectively aggregate blood pressure data across visits spanning days, months, or even years to calculate variability indices for use in clinics [18]. This facilitates clinical decision support tools that could flag patients with excessive variability. Identifying such patients could prompt individualized interventions such as counselling to enhance medication adherence, switching to longer-acting antihypertensive agents or scheduling more frequent follow-ups to stabilize BP control. Ultimately, incorporating visit-to-visit BP variability into routine care – as an adjunct to mean BP – could allow for personalized risk assessment and treatment for patients with hypertension. Future research should address whether interventions aimed specifically at reducing BP variability, such as consistent medication dosing or lifestyle modifications, can translate into improved clinical outcomes.

Our study has some limitations. First, being a retrospective cohort study, the possibility of selection biases cannot be ignored. Additionally, as all measurements were obtained retrospectively, the consistency of measurement methods cannot be fully guaranteed, although experienced medical staff at each center supervised BP measurements. For instance, because only a single BP reading was retrospectively collected per encounter, it cannot be ascertained whether this was an average of three measurements taken after adequate resting. Second, in measuring the VVV of the SBP, we were unable to account for diurnal, contextual, or seasonal variations. Third, VVV measures derived from EHR are subject to influences from frequency and intervals of clinic visits. The median duration between each visit was 42.0 days, which implies that our findings are mainly applicable to visit-to-visit variability assessed at 1- or 2-month intervals over a follow-up period of 2 years. Interpretation of variability assessed from measurements obtained over different intervals may require caution. Additionally, patients with more clinic visits may reflect a population with more severe or complex conditions, which may explain why the overall study population comprises of relatively high percentage of patients with diabetes or dyslipidemia. While frequent visits or BP measurements may suggest variations in BP due to individual medical adjustments, it may also lead to higher incidences of adverse outcomes. Fourth, for definition of myocardial infarction, we have used the values and patterns of CK-MB rather than high-sensitivity cardiac troponin (hs-cTn) because hs-cTn values were not consistently available across all participating centers during the study period within the OMOP-CDM database. This may have led to underestimation of MACE due to reduced sensitivity of CK-MB compared with hs-cTn. Fifth, another limitation relates to the potential influence of competing risks, as non-cardiovascular death may have occurred before the development of cardiovascular events in some patients, especially those who are older-aged. Although conventional Kaplan-Meier and Cox proportional hazards analyses may overestimate the outcome events in such cases, as our primary endpoint was a composite outcome (MACE) rather than individual cardiovascular mortality alone, the overall association between BP variability and adverse cardiovascular events is unlikely to be substantially altered. Future studies using competing-risk models may help further refine the relationship between BP variability and cardiovascular outcomes. Finally, information on drug compliance for antihypertensive medications was not available. However, the stipulation of at least nine BP measurements over a duration exceeding 2 years for eligibility indicates a requisite level of adherence.

Conclusion

In conclusion, our analysis illustrates that feasible and practicable assessment of MACE is available via BP measurements using EHR records. Our recommended minimum number of measurements for visit-to-visit variability assessment of SBP for patients with hypertension is five. This approach may improve risk categorization and provide individualized therapy regimens for hypertension.

Abbreviations

BP

Blood pressure

EHR

Electronic health records

ICD-10

International Classification of Diseases, 10th Revision

MACE

Major adverse cardiovascular events

MI

Myocardial infarction

OMOP-CDM

Observational Medical Outcomes Partnership Common Data Model

ROC

Receiver-operating characteristic curve

SCORE2

Systemic Coronary Risk Evaluation 2

VVV

Visit-to-visit variability

Authors’ contributions

H.J. designed the study. Y.K., E.K. and H.J. collected the data. S.L. analyzed the data. S.L. wrote the first draft of the article. S.H. and C.Y. were responsible for review and editing; H.J., Y.K. and E.K. were responsible for project administration. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by Korean Cardiac Research Foundation (202303-03).

Data availability

The data used to support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The requirement for signed written consent was waived owing to the retrospective nature of the study and anonymity of the data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.