Reliability of Office, Home, and Ambulatory Blood Pressure Measurements and Correlation with Left Ventricular Mass

Abstract

Background.

Determining the reliability and predictive validity of office blood pressure (OBP), ambulatory BP (ABP), and home BP (HBP) can inform which is best for diagnosing hypertension and estimating cardiovascular disease risk.

Objectives:

To assess the reliability of OBP, HBP and ABP, and evaluate their associations with left ventricular mass index (LVMI) in untreated individuals.

Methods:

The Improving the Detection of Hypertension (IDH) study, a community-based observational study, enrolled 408 participants who had OBP assessed at three visits, and completed three weeks of HBP, two 24-hour ABP recordings and a 2D echocardiogram. Mean±SD age was 41.2±13.1 years, 59.5% were women, 25.5% African American, and 64.0% Hispanic.

Results.

The reliability of one week of HBP, three office visits with mercury sphygmomanometry, and 24-hr ABP were 0.938, 0.894, and 0.846 for systolic and 0.918, 0.847, and 0.843 for diastolic BP, respectively. The correlations among OBP, HBP, and ABP, corrected for regression dilution bias, were 0.74–0.89. After multivariable adjustment including OBP and 24-hr ABP, 10 mm Hg higher systolic and diastolic HBP was associated with 5.07 (standard error [SE]: 1.48) and 3.92 (SE: 2.14) g/m² higher LVMI, respectively. After adjustment for HBP, neither systolic or diastolic OBP nor ABP were associated with LVMI.

Conclusion:

OBP, HBP and ABP assess somewhat distinct parameters. Compared with OBP (3 visits) or 24-hour ABP, systolic and diastolic HBP (1 week) were more reliable and more strongly associated with LVMI. These data suggest one week of HBP monitoring may be the best approach for diagnosing hypertension.

CONDENSED ABSTRACT:

We analyzed the reliability of office, home and ambulatory blood pressures, and their associations with left ventricular mass index (LVMI). The 1-week mean of home BP was more reliable than office BP (averaged over 3 visits) or mean awake, asleep, or 24-hr ambulatory BP. We show that office, home and ambulatory BP would not be equivalent, even if each could be assessed with perfect reliability. Home BP – not office or ambulatory BP – was independently associated with LVMI after adjustment for age, sex, race/ethnicity, BMI, and the other BP measures. Home BP may be the best approach for measuring BP.

Tweet:

In a rigorous research study, home blood pressure monitoring was more reliable and had a stronger association with sub-clinical cardiovascular disease compared with office measurements and ambulatory blood pressure monitoring

Introduction

The accurate measurement of blood pressure (BP) is essential for the diagnosis and management of hypertension.(1) While the measurement of blood pressure (BP) in an office setting has been the cornerstone of clinical assessment,(1–3) several meta analyses/systematic reviews have shown that out-of-office BP, measured with either ambulatory BP monitoring (ABPM) or home BP monitoring (HBPM), has a stronger association with the risk for cardiovascular disease (CVD) events and target organ damage.(4–7) This stronger association may result from out-of-office BP being less susceptible to a white-coat effect and/or better reflecting an individual’s BP during everyday life,(8) but it may also result from out-of-office BP being more reliable than office BP, due to the larger number of readings obtained; ceteris paribus, the more readings that are being averaged to create a summary BP measure, the greater the reliability of this average will be.(9,10)

If office measurements, ABPM and HBPM were all estimating the same parameter and were not affected by the measurement conditions and setting in which they are performed, then the correlation between any pair of these approaches would be perfect after correction for random measurement error.(11,12) However, each type of BP measurement is taken under different conditions and in different settings, which have been shown to affect BP.(13,14) ABPM and HBPM are both assessed in the person’s everyday environment rather than a medical office setting, and are therefore thought to be more ecologically valid than office BP. Meanwhile, HBPM and office BP are both assessed in the seated position while ABPM measurements are taken throughout the day, at rest and during/after activity, regardless of body position. Also, ABPM assesses BP inside and outside of the home throughout the day, typically for a single 24-hour period, whereas HBPM assesses BP only at home, usually on two occasions per day but for several days.(15) Although it is not possible to perform perfectly reliable assessments of office BP, ABPM or HBPM in the real world, information about i) the reliability of each method, and ii) what their correlations with one another and with target end-organ damage would be if each method were perfectly reliable is essential for identifying the best approach for BP measurement, and targeting the development of strategies to improve that approach for real world use.

The goals of the current study were to assess, 1) the reliability of BP measured in the office, and by ABPM and HBPM, 2) the correlations among these three types of BP measurement and the extent to which they assess the same parameter, and 3) the associations of BP assessed by each measurement approach with left ventricular mass index (LVMI).

Methods

Sample Population

The study sample was comprised of participants from the Improving the Detection of Hypertension (IDH) study, a community-based sample of 408 women and men, recruited between 2011 and 2013, in New York City. Participants were ≥18 years of age and were free of known CVD. Individuals were ineligible if they had a systolic/diastolic BP (SBP/DBP) ≥ 160/105 mm Hg, based on the mean of their 2^nd and 3^rd readings obtained during a screening visit, had evidence of secondary hypertension, were pregnant, or were taking antihypertensive medication or other medications that affect BP. The protocol for the IDH study was approved by Columbia University’s Institutional Review Board and all participants provided written informed consent. The Supplemental Appendix provides further details on study design, sample population, exclusion criteria, and the methods used to select the sample for the current analyses.

Study Procedures

Participants attended 5 study visits over a 4-week period (Figure 1). Demographic information was obtained at Visit 1 and office BP was measured during Visits 1, 3, and 4. HBPM was conducted during the 3 weeks between Visits 2 and 4, and 24-hour ABPM was performed following Visits 1 and 4. During Visit 5, participants completed a detailed medical history, blood collection, anthropometric measurements, and an echocardiogram. Of the 408 enrolled participants, 400 completed all 5 study visits and were included in the current analyses.

Figure 1. Design of the Improving the Detection of Hypertension study.

This shows the study visits and procedures completed by participants. Each participant had their blood pressure measured 9 times at office visits 1, 3 and 4 including three measurements using a mercury sphygmomanometer, the BpTRU automated oscillometric device and an Omron oscillometric device. Participants also completed three weeks of home blood pressure monitoring between visits 2 and 4 and 24-hour ambulatory blood pressure monitoring two times, between visits 1 and 2 and between visits 4 and 5. Actigraphy was assessed during both 24-hour ABPM recordings, primarily to accurately determine the start and end of each participant’s sleep period. During visit 5, participants had a cardiovascular evaluation which included a 2D echocardiogram. BP – blood pressure

Office BP

Upper arm office BP measurements were obtained with a person in the room (i.e., attended) using three separate devices: a mercury sphygmomanometer (Baum, Copiague, NY) and stethoscope, a BpTRU clinic-grade oscillometric device (Model BPM-200; (16,17)), and an Omron oscillometric device designed for HBPM (Model HEM-790IT [HEM-7080-ITZ2](18) or HEM-791IT [HEM-7222-ITZ])(19). At Visits 1, 3, and 4, a research technician obtained 3 BP readings with each device, using the appropriate-sized cuff, on the non-dominant arm following a standardized protocol. The order was randomized such that participants had their BP measured with each device first, second, and third at alternate visits. If a participant did not have a valid office BP or it was inadvertently not taken, it was treated as missing. Of the 400 participants who completed the 5 study visits, 391/392 had 3 SBP/DBP readings from each office BP device at all 3 visits. The adjustment for order effects is described in the Supplemental Appendix.

Ambulatory BP

During Visits 1 and 4, participants were fitted for ABPM (Spacelabs Model 90207, Snoqualmie, WA)(20) with an appropriate-sized cuff on their non-dominant arm. BP measurements were taken at 30-minute intervals throughout the 24-hour period. Mean awake and asleep BP were calculated for each recording, based on sleep and awake periods determined from activity and light data obtained from a wrist-worn research grade actigraphy device (ActiWatch 2, Philips Respironics, Murrysville, PA, OR).(21–23) A complete ABPM assessment was defined as having ≥10 awake readings and ≥5 asleep readings with at least 70% of all attempted BP readings being valid. There were 364 participants (91%) with two complete ABPM recordings.

Home BP

During Visit 2, participants were trained in the use of the same HBPM Omron device used at Visits 1, 3 and 4. Participants received a 5-minute training, delivered by a research coordinator, on how to conduct HBPM as well as a one-page handout containing HBPM instructions. Participants were instructed to take measurements from their non-dominant arm while seated with back supported and feet flat on the floor after resting for 5 minutes. Participants were also instructed to rest their arm on a table or desk so that the cuff was at heart level. The Omron HBPM device stores all BP measurements electronically. For the 21 days between Visits 2 and 4, participants were asked to measure their BP at home, twice in the morning (one minute apart) shortly after awakening and twice in the evening (one minute apart) before going to bed. Participants brought the device to Visit 4 and BP readings were downloaded by research staff. The monitoring period was divided into 3 separate 7-day periods and BP readings were averaged for each period. Overall, 383 (96%) participants had complete HBPM data defined as having at least 12 successful BP readings, out of the targeted 28 readings, as recommended in guidelines and scientific statements,(1,24) for each 7-day period.

Additional Data Collection

At Visit 5, height and weight were measured according to a standard protocol. Two-dimensional echocardiograms were performed by a trained sonographer and analyzed by a cardiologist who made left ventricular (LV) measurements according to American Society of Echocardiography (ASE) recommendations.(25) LV mass was calculated using the ASE formula. LVMI (g/m²) was calculated as LV mass divided by body surface area derived by the DuBois method.(26)

Statistical Analyses

Participant characteristics were summarized using means and percentages. Data from the 3 visits with office BP measurements were used to estimate the reliability (i.e., intraclass correlation coefficient) for a single visit of: a) the first office BP reading; b) the average of the first 2 readings; c) the average of all 3 readings, and d) the average of the 2^nd and 3^rd readings separately, for each office device. Applying the Spearman-Brown formula (27), the reliability estimate for a single visit was used to calculate the reliability of the mean BP over 2 visits and separately, over 3 visits. The reliability of the mean BP for one week of HBPM, and for the mean awake, asleep and 24-hour BP by ABPM was also calculated. For each BP measure, we also estimated the mean absolute deviation (MAD); i.e., the average difference between the observed and true BP.

We calculated Pearson correlations among the office, home, and ambulatory SBP and DBP measurements. Given that BP measurements are not perfectly reliable, the correlation of one with another, or with a third variable such as LVMI, will be attenuated (i.e., closer to zero) than what it would be if the measurements were perfectly reliable; this phenomenon is referred to as “attenuation due to unreliability” or “regression dilution bias”. To address this, structural equation modeling (SEM) was used to model the unreliability and estimate what the correlations among the different BP measures would be if each could be assessed with perfect reliability – that is, what the correlations are among the unobservable “true” values of office, home and ambulatory BP; these are reported as “correlations corrected for regression dilution bias”. Next, we calculated Pearson correlation coefficients of office, home, and ambulatory BP measures with LVMI, and SEM was again used to correct these correlations for regression dilution bias. Linear regression was used to estimate the association of office, home, and ambulatory BP with LVMI. In addition to an unadjusted model, two levels of adjustment were performed. First, we adjusted for age, sex, black race, Hispanic ethnicity, body mass index, and diabetes status. Second, we further adjusted for office, home, and ambulatory SBP or DBP. We used SEM to re-estimate these associations with LVMI, correcting for regression dilution bias. Separate analyses were performed for the mercury, BpTRU and the Omron office BP measurements. Mplus (version 8.1)(28) was used to estimate the SEMs. SAS (version 9.4, Cary NC) was used to perform all other analyses.

Role of the funding source

Program project grant P01-HL47540 (PI: JE Schwartz) from the National Heart, Lung, and Blood Institute (NHLBI) covered all costs associated with finalizing the protocol, participant recruitment, collection of data, and data management for the IDH study. NHLBI had no role in the design or execution of the IDH study. The authors are solely responsible for its execution, as well as the analysis and content of this manuscript.

Results

For the 400 participants who completed all 5 study visits, the mean age was 41.2 (SD, 13.1) years, 59.5% were women, 25.5% were African American, and 64.0% were Hispanic (Table 1). The mean office SBP across the three visits ranged from 116.0 to 117.2 mm Hg, and DBP from 75.6 to 76.5 mm Hg, depending on the device used. Across 21 days of HBPM measurements, the mean SBP/DBP was 115.6/76.7 mm Hg. Across the two 24-hour ABPM periods, the mean awake, asleep, and 24-hour SBP/DBP were 124.5/77.9, 109.1/63.8, and 120.1/73.9 mm Hg, respectively. Although peripheral to this paper’s focus, it is noteworthy that as in other studies of younger, untreated individuals,(29) the mean awake SBP was 8–9 mm Hg higher than office and home SBP.

Table 1.

Characteristics of Improving the Detection of Hypertension Study Participants (N = 400)

Age, yrs	41.2 ± 13.1
Female	59.5
Black	25.5
Hispanic	64.0
Education, yrs	14.5 ± 3.8
Smoking status
Never	82.0
Past smoker	9.5
Current smoker	8.5
Alcohol consumption
Nondrinker	22.3
Light/moderate drinker	72.5
Heavy drinker	5.3
Body mass index, kg/m2	27.6 ± 5.2
Diabetes	4.8
Left ventricular mass index, g/m2	79.3 ± 16.1
SBP, mm Hg
Office SBP, mercury	117.2 ± 13.8
Office SBP, oscillometric, BpTRU	116.0 ± 14.0
Office SBP, oscillometric, Omron	116.6 ± 14.2
Home SBP	115.6 ± 13.8
Awake ambulatory SBP	124.5 ± 11.9
Asleep ambulatory SBP	109.1 ± 11.7
24-h ambulatory SBP	120.1 ± 11.5
DBP, mm Hg
Office DBP, mercury	76.5 ± 8.6
Office DBP, oscillometric, BpTRU	75.6 ± 9.4
Office DBP, oscillometric, Omron	76.0 ± 9.2
Home DBP	76.7 ± 9.1
Awake ambulatory DBP	77.9 ± 7.7
Asleep ambulatory DBP	63.8 ± 8.3
24-h ambulatory DBP	73.9 ± 7.5
Hypertenstion status: ACC/AHA
Office, mercury	35.5
Office, oscillometric, BpTRU*	31.2
Office, oscillometric, Omron†	33.2
Home	36.3
Awake ambulatory	41.9
Asleep ambulatory	48.4
24-h ambulatory	45.1
Hypertension status: JNC7
Office, mercury	9.6
Office, oscillometric, BpTRU	10.2
Office, oscillometric, Omron	8.9
Home	9.4
Awake ambulatory	24.1
Asleep ambulatory	25.8
24-h ambulatory	26.1

Values are mean ± SD or %. Alcohol use is defined as: nondrinker (no alcohol consumption), light to moderate drinker (1–14 and 1–7 alcoholic beverages per week for men and women, respectively), or heavy drinker (> 14 and > 7 alcoholic beverages per week for men and women, respectively) The ACC/AHA definitions of hypertension are SBP/DBP ≥130/80 mm Hg for office, home and daytime blood pressure, ≥125/75 mm Hg for 24-h blood pressure and ≥110/65 mm Hg for asleep blood pressure. The JNC7 definitions of hypertension are office SBP/DBP blood pressure ≥140/90 mm Hg, ≥135/85 mm Hg for home and daytime blood pressure, ≥130/80 mm Hg for 24-h blood pressure and ≥120/70 mm Hg for asleep blood pressure. Of the 400 participants who completed the study, the number with complete office blood pressure data (9 readings at each of 3 visits) was 391 and 392 for SBP and DBP, respectively. The numbers with complete home, awake ambulatory, asleep ambulatory and 24-h ambulatory BP data were 383, 377, 364 and 364, respectively.

^*BpTRU: VSM MedTech Ltd., Vancouver, British Columbia, Canada.

^†Omron Healthcare Inc., Lake Forest, Illinois

ACC = American College of Cardiology, AHA = American Heart Association, DBP = diastolic blood pressure, JNC = Joint National Committee on Prevention,. Detection, Evaluation, and Treatment of High Blood Pressure, SBP = systolic blood pressure.

Reliability of office, home and ambulatory BP measurements

The reliability of SBP and DBP was higher for one week of HBPM when compared with office measurements or one 24-hour ABPM (Table 2). The reliability of office-measured BP was greater when based on readings obtained at different visits versus the equivalent number of readings obtained during a single visit. For example, the reliability of office SBP was 0.883 for the average of 1 mercury reading taken at 3 separate visits versus 0.738 for 3 readings taken at a single visit. Furthermore, the reliability of office BP measurements averaged over three visits was greater than that of awake, asleep, or 24-hour BP from a single 24-hour ABPM recording. The average all 3 readings from a visit was consistently more reliable than the average of the 2^nd and 3^rd readings.

Table 2.

Reliability and Estimated Mean Absolute Deviation of Mean Office, Home, and Ambulatory Systolic and Diastolic Blood Pressures.

Method	Readings per visit	Reliability			Estimated Mean Absolute Deviation, mm Hg
		1 Visit	2 Visits	3 Visits	1 Visit	2 Visits	3 Visits
Systolic blood pressure
Mercury	1	0.716	0.835	0.883	6.53	4.62	3.77
	2	0.731	0.845	0.891	6.32	4.47	3.65
	2nd & 3rd	0.724	0.840	0.887	6.43	4.55	3.71
	3	0.738	0.849	0.894	6.20	4.38	3.58
BpTRU, oscillometry	1	0.657	0.793	0.852	7.91	5.59	4.56
	2	0.716	0.835	0.883	6.73	4.76	3.89
	2nd & 3rd	0.715	0.834	0.883	6.54	4.63	3.78
	3	0.738	0.849	0.894	6.29	4.45	3.63
Omron, oscillometry	1	0.673	0.805	0.861	7.54	5.33	4.36
	2	0.721	0.838	0.886	6.65	4.70	3.84
	2nd & 3rd	0.741	0.851	0.896	6.32	4.47	3.65
	3	0.746	0.855	0.898	6.26	4.42	3.61
1 week of HBPM		0.938			2.81
Mean awake ABPM		0.834			4.04
Mean asleep ABPM		0.800			4.40
Mean 24-hr ABPM		0.846			3.73
Diastolic blood pressure
Mercury	1	0.594	0.746	0.815	5.10	3.61	2.95
	2	0.637	0.778	0.840	4.74	3.35	2.74
	2nd & 3rd	0.643	0.782	0.844	4.75	3.36	2.74
	3	0.648	0.787	0.847	4.63	3.28	2.68
BpTRU, oscillometry	1	0.580	0.734	0.806	6.16	4.36	3.56
	2	0.637	0.778	0.840	5.23	3.69	3.02
	2nd & 3rd	0.654	0.791	0.850	4.96	3.50	2.86
	3	0.677	0.808	0.863	4.80	3.39	2.77
Omron, oscillometry	1	0.570	0.726	0.799	5.97	4.22	3.45
	2	0.641	0.781	0.843	5.09	3.60	2.94
	2nd & 3rd	0.658	0.794	0.852	4.88	3.45	2.82
	3	0.668	0.801	0.858	4.79	3.39	2.76
1 week of HBPM		0.918			2.13
Mean awake ABPM		0.806			2.84
Mean asleep ABPM		0.797			3.14
Mean 24-hr ABPM		0.843			2.47

Reliability: For the column Labeled “1 Visit”, this is the intraclass correlation coefficient for the 3 visits for the first reading, the mean of the first 2 readings, the mean of the second and third readings, or the mean of all 3 readings obtained at each visit. For the next 2 columns (“2 Visits” and “3 Visits”), the Spearman-Brown formula was used to estimate the reliability of the average of readings obtained during 2 Visits or during 3 Visits using the intraclass correlation coefficient for a single visit. For HBPM, the intraclass correlation coefficient for the 3 1-week means is reported; for ABPM, the intraclass correlation coefficients for the awake, asleep, and 24-h means from the 2 recordings is reported. Estimated mean absolute deviation from the “true” value : Assuming that repeat measurements for each person are normally distributed around his/her “true” value for that method of measurement, this is calculated as (2/π)^0.5 times the pooled within-person standard deviation of the measurements. Of the 400 participants who completed the study, the number of participants used in the analysis of office BP (Mercury, BpTRU and Omron) was 391 and 392 for SBP and DBP, respectively. The numbers used in the analyses of home, awake ambulatory, asleep ambulatory, and 24-h ambulatory BP were 383, 377, 364, and 364, respectively.

ABPM = ambulatory blood pressure measurement; HBPM = home blood pressure measurement, other abbreviations as in Table 1.

Correlations among office, home and ambulatory BP measurements

The correlations among office mercury BP, home BP, and awake, asleep, and 24-hour ambulatory BP corrected for regression dilution bias ranged from 0.74 to 0.89 (Figure 2). The correlations for BpTRU and Omron office measurements with home BP and ambulatory BP are presented in Supplemental Figures 1 and 2. The squared correlations indicate that any two types of BP measure share at most 80%, and as little as 55%, of their variance in common, and therefore 20% or more is not shared. Uncorrected correlations among office BP, HBPM and ABPM were substantially lower (r = 0.47 to 0.80) than the corrected correlations (Supplemental Table 1).

Figure 2. Correlations among office mercury, home and ambulatory systolic and diastolic blood pressure measurements, corrected for regression dilution bias, N=342.

Correlations among office mercury, home and 24-hour mean ambulatory systolic and diastolic blood pressure measurements, corrected for regression dilution bias, are shown as curved bi-directional arrows connecting pairs of blood pressures (green circles). For example, the correlation between participants’ systolic home blood pressure values and their mean awake ambulatory systolic blood pressure values is 0.89 after correction for regression dilution bias. Each green circle represents the “true” values of the specified type of blood pressure measurement; i.e., the latent (unobserved) variable that is being approximated by each repeat assessments of that blood pressure. The small black lines emanating from each green circle represent the multiple assessments and show the correlations of these observed blood pressure measures with the “true” values of that blood pressure. For example, the correlation of participants’ mean systolic home blood pressure for week 1 with their true systolic home blood pressure is 0.97. The squared value of this correlation (r²) equals the reliability of that specific measurement. The bottom portion of the figure shows the correlations among the three different methods of assessing office blood pressure, with each observed measurement (e.g., Visit 1) based on the average of 3 readings. BP – blood pressure, ABP – ambulatory blood pressure

Associations of office, home and ambulatory BP with LVMI

In analyses corrected and not corrected for regression dilution, the correlations of home SBP and DBP with LVMI were higher than for office, or awake, asleep, or 24-hour ambulatory measurements (Table 3). SBP and DBP measured in the office by mercury, by HBPM, or ABPM were each associated with LVMI in unadjusted models and, with the exception of mercury office DBP, remained so after adjustment for age, sex, black race, Hispanic ethnicity, BMI, and diabetes status (Table 4, Supplemental Table 2). In models that simultaneously included mercury office SBP, home SBP, and ambulatory SBP, home SBP was associated with LVMI, while the other measurements were not. Similarly, home DBP, but not mercury office or ambulatory DBP, was associated with LVMI in models including each of these variables. Neither office mercury nor ambulatory awake, asleep, or 24-hour DBP were associated with LVMI after adjustment for home DBP. Results using BpTRU and Omron office readings instead of mercury office BP readings are provided in Supplemental Tables 3–6.

Table 3.

Correlations of mean office, home and ambulatory systolic and diastolic blood pressures with left ventricular mass index, before and after correction for regression dilution bias, N=340.

Method		Uncorrected	Correcteda
Systolic BP
Mercury, auscultatory	1 reading, 1 visit	0.336***	0.389***
	3 readings, 3 visits	0.365***
BpTRU, oscillometry	1 reading, 1 visit	0.263***	0.320***
	3 readings, 3 visits	0.303***
Omron, oscillometry	1 reading, 1 visit	0.369***	0.434*
	3 readings, 3 visits	0.413**
Home BP	1 week	0.486	0.501
Ambulatory BP	Mean awake	0.388***	0.426**
	Mean asleep	0.338***	0.378***
	Mean 24-hour	0.394***	0.430**
Diastolic BP
Mercury, auscultatory	1 reading, 1 visit	0.216**	0.267
	3 readings, 3 visits	0.246*
BpTRU, oscillometry	1 reading, 1 visit	0.202**	0.254
	3 readings, 3 visits	0.239*
Omron, oscillometry	1 reading, 1 visit	0.193***	0.266
	3 readings, 3 visits	0.244*
Home BP	1 week	0.315	0.328
Ambulatory BP	Mean awake	0.231**	0.255*
	Mean asleep	0.245	0.278
	Mean 24-hour	0.258	0.280

BP – Blood pressure

^aThe estimates in this column are correction for regression dilution bias.

^*p≤0.05

^**p≤0.01

^***p≤0.001 for difference between this correlation and the correlation for Home BP with LVMI.

Among the uncorrected correlations, the correlation of LVMI with a single systolic BpTRU reading taken at a single visit (0.263) was lower than the correlations of LVMI with mean awake systolic BP (0.388, p≤0.001) and mean 24-hour systolic BP (0.394, p≤0.001) and the correlation of LVMI with the mean of 9 systolic BpTRU readings taken over 3 visits (0.303) was lower than the correlations of LVMI with mean awake systolic BP (0.388, p≤0.05) and mean 24-hour systolic BP (0.394, p≤0.05). The uncorrected correlation of LVMI with mean asleep systolic BP (0.338) was lower than the correlations of LVMI with mean awake systolic BP (0.388, p≤0.05) and mean 24-hour systolic BP (0.394, p≤0.01). The uncorrected correlation of LVMI with mean awake diastolic BP (0.231) was lower than the correlation of LVMI with mean 24-hour diastolic BP (0.394, p≤0.01). No other comparisons between uncorrected correlations of LVMI with two different ABP measures or of LVMI with one ABP measure and an office BP measure were statistically significant. Comparisons among the correlations of LVMI with different clinic BP measures are not reported because all clinic BPs were taken at the same visit.

Among the corrected correlations, the correlation of LVMI with BpTRU systolic BP (0.320) was lower than the correlations of LVMI with mean awake systolic BP (0.426, p≤0.05) and mean 24-hour systolic BP (0.430, p≤0.01). The correlation of LVMI with mean asleep systolic BP (0.378) was lower than the correlation of LVMI with mean 24-hour systolic BP (0.430, p≤0.05) and the correlation of LVMI with mean awake diastolic BP (0.255) was lower than the correlation of LVMI with mean 24-hour distolic BP (0.280, p≤0.05). No other comparisons between corrected correlations of LVMI with two different ABP measures or of LVMI with one ABP measure and an office BP measure were statistically significant.

Table 4.

Associations of mean mercury office, home and ambulatory systolic and diastolic blood pressures with left ventricular mass index, corrected for regression dilution bias, N=340.

Systolic blood pressure	Model 1	Model 2	Model 3a	Model 3b	Model 3c
Office, Mercury	4.55 (0.62) P<0.001	3.00 (0.77) P<0.001	−1.49 (1.57) P=0.34	−1.60 (1.45) P=0.27	−1.59 (1.52) P=0.30
Home	5.68 (0.54) P<0.001	4.05 (0.64) P<0.001	5.22 (1.47) P<0.001	5.16 (1.39) P<0.001	5.07 (1.48) P=0.001
Awake	5.98 (0.73) P<0.001	3.78 (0.81) P<0.001	−0.10 (1.86) P=0.96
Asleep	5.42 (0.77) P<0.001	3.54 (0.84) P<0.001		0.07 (1.39) P=0.96
24-hour	6.20 (0.75) P<0.001	4.00 (0.83) P<0.001			0.21 (1.80) P=0.91
Diastolic blood pressure
Office, Mercury	5.40 (1.15) P<0.001	2.32 (1.27) P=0.07	−2.82 (3.13) P=0.37	−3.40 (2.44) P=0.16	−3.83 (2.85) P=0.18
Home	5.88 (0.93) P<0.001	3.53 (0.96) P<0.001	4.91 (2.24) P=0.03	3.40 (2.05) P=0.10	3.92 (2.14) P=0.07
Awake	5.59 (1.22) P<0.001	3.17 (1.17) P=0.007	0.79 (3.57) P=0.85
Asleep	5.67 (1.14) P<0.001	4.26 (1.17) P<0.001		3.64 (2.05) P=0.08
24-hour	6.20 (1.21) P<0.001	3.79 (1.20) P=0.002			3.05 (3.07) P=0.32

Numbers in the table represent the estimated difference in left ventricular mass index (g/m²) associated with a 10 mm Hg difference in systolic (top panel) or diastolic (bottom panel) blood pressure with the standard error in parentheses.

Model 1 is unadjusted and regresses left ventricular mass index on each blood pressure measure separately.

Model 2 regresses left ventricular mass index on each blood pressure measure separately, adjusting for age, sex, black race, Hispanic ethnicity, body mass index, and diabetes status. Models 3a, 3b, and 3c regress left ventricular mass index on office, home and ambulatory blood pressure simultaneously, adjusting for the covariates in Model 2.

Model 3a includes awake blood pressure from ambulatory blood pressure monitoring.

Model 3b includes asleep blood pressure from ambulatory blood pressure monitoring.

Model 3c includes 24-hour blood pressure from ambulatory blood pressure monitoring.

Discussion

In the current study, the mean SBP from one week of HBPM was associated with LVMI after adjustment for office-measured SBP and awake, asleep or 24-hour ambulatory SBP. Mean DBP from one week of HBPM was associated with LVMI after adjustment for office-measured DBP and awake DBP on ABPM. In contrast, SBP and DBP measured in the office or by ABPM were not associated with LVMI after adjustment for home SBP or DBP. Also, the reliability of SBP and DBP was higher for one week of HBPM than for nine office measurements collected over three visits, or for mean awake, asleep, or 24-hour ambulatory BP (Central Illustration). These findings suggest that when diagnosing hypertension, HBPM may be preferred to office-and ambulatory BP assessments.

Central Illustration. Systolic Home BP is more reliable and more strongly correlated with left ventricular mass than either office BP or ambulatory BP.

Participants had their office BP (OBP) assessed at 3 visits (3 readings/visit) and completed 3 weeks of home BP monitoring (HBPM) and two 24-hour ambulatory BP monitoring (ABPM) recordings. We estimated the reliability (reproducibility) of mean BP of one week of HBPM, one 24-hour ambulatory BP recording ABPM and 3 office visits; the reliability of HBPM was greater than that of both 24-hr ABPM and OBP. We also estimated the correlation of each with left ventricular mass index (LVMI), with and without correction for regression dilution bias. HBPM was more highly correlated with LVMI than either 24-hr ABPM or OBP, both before (not shown) and after correction for regression dilution bias.

If office BP, HBPM, and ABPM were all measuring the same parameter, then the correlation, after correction for regression dilution bias, between any two would be perfect. Furthermore, they would have been equally strongly associated with LVMI after correction for regression dilution bias. This is not what we found. No two BP measures shared more than 80% of their variance and it was HBPM that was most strongly associated with LVMI.

The US Preventive Services Task Force and the UK NICE recommend ABPM as the preferred method for the diagnosis of hypertension based on data showing a strong association with CVD events and all-cause mortality and cost-effectiveness analyses. (6,13,30,31) Both consider HBPM an acceptable alternative out-of-office BP assessment, in part because ABPM is inaccessible to most primary care providers and rarely conducted in the US,(32) and a systematic review published in 2017 found few data to support the superiority of one approach over the other with regard to their association with future CVD events.(7) In the current study, HBPM was superior to office BP and ABPM with regard to both reliability and the association with LVMI, a marker of sub-clinical CVD.

There are several possible reasons for why HBPM was superior to office BP and ABPM for predicting LVMI. Office BP measurements are obtained in an environment that has low ecological validity, where individuals spend very little time. ABPM measurements are not taken in a standardized manner and, except perhaps for sleep, do not measure resting BP. Further, ABPM readings are only taken over a single 24-hour period. HBPM, in contrast, is consistently performed at home, at rest, while seated. Further, by collecting readings over 7 days, HBPM may average out the day-to-day variability in BP. Thus, HBPM may provide the best estimate of resting or basal BP in the natural environment, which may be what is most strongly associated with LVMI, and thus CVD risk. Importantly, the analyses that correct for regression dilution bias remove any effect of number of BP readings obtained, and demonstrate that the superiority of HBPM in this study is not attributable to differences in the number of readings.

The vast majority of studies have found that out-of-office BP is superior to office BP for predicting target organ damage or major adverse events.(4–7). Given the present results, the observed differences in reliability suggest that the stronger association with CVD events for out-of-office versus office BP may be due, at least in part, to out-of-office BP being more reliable. Reanalysis of prior studies, correcting for regression dilution bias, would provide a comparison of alternative approaches after removing the effects of differential reliability. In the current study comparing all three approaches, HBPM had the strongest association with LVMI, and was the only approach associated with LVMI after adjustment for the other BP measurement approaches. If further research confirms that HBPM is the measure that is most strongly associated with adverse outcomes, then efforts to improve the assessment of office BP (e.g., with more readings per visit or more than 3 visits), or ABPM (by having a longer recording period or multiple recordings) might be misdirected.

The current study provides empirical data supporting guideline recommendations to average multiple office BP readings taken at more than one visit for clinical decision making.(2) Specifically, the reliability of office SBP and DBP was substantially higher when based on two or three visits versus a single visit, and the correlation of office-measured SBP with LVMI was stronger when based on three BP readings at three visits versus a single measurement at one visit. Furthermore, when measured over three visits, SBP measured in the office setting was more reliable than awake, asleep or 24-hour SBP from a single 24-hr ABPM recording.(33) Most prior studies comparing the associations of office BP and ABPM with CVD events have relied on office measurements obtained during a single visit. Perhaps the finding that the association with CVD events was stronger for ABPM than office BP would have been different if prior studies had assessed office BP over multiple visits or corrected their analyses for regression dilution bias.(9)

The Controlling High Blood Pressure measure of the National Committee for Quality Assurance 2019 Healthcare Effectiveness Data and Information Set (HEDIS) was recently updated to include BP readings taken from remote patient monitoring devices including HBPM.(34) Furthermore, new procedure codes for HBPM have been approved for use starting in 2020.(35) According to NHANES 2011– 2014, 17% of US adults self-reported performing HBPM at least monthly.(36) In contrast, ABPM has been reported to not be widely available in the US.(37) Given its greater reliability, predictive value and availability, HBPM has the potential to become the primary approach for out-of-office BP monitoring.(15) Further, one week of HBPM may be more cost effective and less burdensome than having a patient attend 3 office visits over several weeks. This should be investigated in future studies. There are several barriers to the widespread implementation of HBPM in the US.(24,37) Patient-level barriers include concerns by patients about the requirement for a rigid daily schedule of BP measurement over a long period of time, and costs of purchasing the HBPM device.(38,39) Provider-level barriers include concerns about the possible inaccuracy of HBPM devices, low adherence to self-measured BP monitoring schedules by patients including those who are older and/or have low health literacy, burden on practice resources, and lack of reimbursement for HBPM devices.(37) Healthcare system-level barriers include insufficient infrastucture and support. A recent policy statement for the American Heart Association and American Medical Association recommended additional investment in building and supporting infrastructure for HBPM in the US.(24)

Strengths of the current study include the repeat assessments of each approach for assessing BP: standardized measurement of BP during three office visits, during three weeks by HBPM, and during two 24-hour ABPM recordings. This allowed us to determine the reliability of each method.(10) Another strength of the current study was the ability to compare HBPM and office BP using both the identical oscillometric device (Omron) and a different oscillometric device (BpTRU). Additional strengths include the use of a mercury device for office BP measurement, and the racial/ethnic diversity of the study population. The statistical approach provided estimates of observed associations, based on available research quality (yet imperfect) BP assessments, as well as estimates corrected for regression dilution bias of what these associations would be if each type of BP could be assessed with perfect reliability. The results of the current study should be interpreted in the context of known and potential limitations. The IDH study enrolled adults not taking antihypertensive medication, participants were relatively young, and only a small percentage of participants had diabetes. Its focus was on evaluating the performance of alternative methods of assessing BP in the context of diagnosing, or ruling out, hypertension in the general practice setting. While the mean office BP levels were relatively low, 30 to 50% of the sample met ACC/AHA criteria for hypertension, depending on the method of assessment; 25% met the higher JNC7 criteria based on their ABPM. Therefore, the present results may be especially relevant for individuals without high office BP in whom out-of-office monitoring is recommended to exclude masked hypertension.(2) Nevertheless, the generalizability of these results to individuals with severely elevated BP (≥160/105 mm Hg), those taking antihypertensive medication, older adults, and populations with major comorbidities is unclear. This was a cross-sectional study, and we were unable to determine which BP measurement approach best predicts future CVD events. However, LVMI was assessed following a standardized protocol and provided the ability to assess associations with subclinical CVD. BP was measured in the non-dominant arm only. Therefore, it is unknown whether similar results would be obtained if BP was measured using the arm with the higher BP.

Conclusions

Our analyses demonstrate that office BP, HBPM, and ABPM do not measure the same underlying parameter, and that one week of HBPM has a stronger association with LVMI than awake, asleep or 24-hour ABPM or 9 office BP readings taken over 3 visits. Furthermore, HBPM was the only BP measurement associated with LVMI when all three types of BP were simultaneously included in a regression model. This was true for analyses performed with and without correction for regression dilution bias. Also, one week of HBPM provided more reliable estimates of both SBP and DBP than either office BP or 24-hour ABPM. These data support the use of HBPM over ABPM and office BP for the diagnosis of hypertension and its associated CVD risk.

Clinical Perspectives.

Competency in Patient Care:

One week of home blood pressure monitoring (HBPM) is more reliable than repeated office or ambulatory monitoring (ABPM) and correlates more closely with left ventricular mass.

Translational Outlook:

Further efforts are needed to define and educate patients in the optimum methodology and frequency of HBPM to guide management and improve long-term management.

ABBREVIATIONS:

ABPM

ambulatory blood pressure monitoring

ACC/AHA

American College of Cardiology / American Heart Association

BMI

body mass index

CVD

cardiovascular disease

DBP

diastolic blood pressure

ESC/ESH

European Society of Cardiology / European Society of Hypertension

HBPM

home blood pressure monitoring

IDH

Improving the Detection of Hypertension (study)

MAD

mean absolute deviation

LVMI

left ventricular mass index

SBP

systolic blood pressure

SEM

structural equation modeling