Poor Reliability of Wrist Blood Pressure

Self-Measurement at Home
A Population-Based Study

Edoardo Casiglia

Valérie Tikhonoff

Federica Albertini

, and

Paolo Palatini

Originally published22 Aug

2016https://doi.org/10.1161/HYPERTENSIONAHA.116.07961Hypertension. 2016;68:896–903

 Other version(s) of this article

Abstract
The reliability of blood pressure measurement with wrist devices, which has not
previously been assessed under real-life circumstances in general population, is
dependent on correct positioning of the wrist device at heart level. We determined
whether an error was present when blood pressure was self-measured at the wrist in
721 unselected subjects from the general population. After training, blood pressure was
measured in the office and self-measured at home with an upper-arm device (the UA-
767 Plus) and a wrist device (the UB-542, not provided with a position sensor). The
upper-arm−wrist blood pressure difference detected in the office was used as the
reference measurement. The discrepancy between office and home differences was the
home measurement error. In the office, systolic blood pressure was 2.5% lower at wrist
than at arm (P=0.002), whereas at home, systolic and diastolic blood pressures were
higher at wrist than at arm (+5.6% and +5.4%, respectively; P<0.0001 for both); 621
subjects had home measurement error of at least ±5 mm Hg and 455 of at least ±10
mm Hg (bad measurers). In multivariable linear regression, a lower cognitive pattern
independently determined both the systolic and the diastolic home measurement error
and a longer forearm the systolic error only. This was confirmed by logistic regression
having bad measurers as dependent variable. The use of wrist devices for home self-
measurement, therefore, leads to frequent detection of falsely elevated blood pressure
values likely because of a poor memory and rendition of the instructions, leading to the
wrong position of the wrist.
Introduction
Blood pressure (BP) independently predicts cardiovascular risk in affluent countries. 1–
5 However, several studies have shown that awareness of hypertension and BP control

is still suboptimal,1and thus effective strategies should be developed to improve BP

control and adherence to therapy. One of the methods used to achieve these goals is
represented by self-measurement of BP at home by automatic devices. Many studies
have shown that BP self-measurement at home allows a better BP control in
hypertension and has a greater prognostic value than office BP.6,7In addition, self BP
measurement is more appealing to the patient than the customary procedure in the
physician’s surgery often causing long waits.6

Upper-arm automatic devices gave fresh impetus to home self-measurement because

of their user friendliness,8 the only conditions to be respected being adequate cuff
dimensions and correct positioning of the cuff on the patient’s arm. 9–14

For reasons of market penetration and following a general tendency to miniaturization,

many wrist devices recently appeared on the market15 having the advantage of being
smaller and easier to fit than upper-arm monitors,16 and today wrist devices are used by
a large portion of people who measure their BP at home.17However, in spite of technical
improvement, their reliability in real-life conditions is not unanimously accepted.18–
20 Accuracy of BP measurement at wrist largely depends on the difference in height

between the wrist and the heart because of the confounding effect of the hydrostatic
pressure caused by the limb blood column.21–23 Wrists kept at a higher level in
comparison with the heart lead to a false lower, and wrists at a lower level lead to a
false higher BP values. Very rarely do patients receive appropriate training from family
doctors or other healthcare personnel. Instructions attached to commercial packages
require a certain degree of personal discernment, i.e., a good cognitive pattern. Only a
few models have a position sensor to verify that the wrist is placed properly at heart
level,24 but even for these devices, no study has evaluated their reliability in real-life
conditions.

The aim of the present study was to ascertain whether wrist home BP is performed
reliably or an error is present in measurement, and to identify the determinants of this
error if any.

To verify the reliability of wrist BP self-measurement, we determined at a population

level the upper-arm−wrist BP difference in the office under a doctor’s supervision and
then verified whether this difference was maintained during home self-measurement.

Methods
Study Population and General Protocol
The analysis involved in the frame of the GOLDEN study (Growing Old With Less
Disease Enhancing Neurofunctions), 721 unselected subjects were recruited from an
Italian general population (response rate to call 73%), aged 49.3±15.4 (range, 18.0–
89.8) years, living in an area of ≈107 km2 and sharing homogeneous lifestyle.25 At
screening, all underwent anthropometrics and a questionnaire.26 Education was defined
as years of schooling based on the highest educational qualification achieved. Height
(in m) and weight (in kg) were recorded without shoes with the subjects wearing light
indoor clothing.27 Biceps skinfold thickness28 was measured (in mm) with a plicometer
(Holtain Ltd, Crymych, United Kingdom) applying a caliper pressure of 10 g mm−2. At the
screening, no subjects had atrial fibrillation that could decrease the accuracy of
automatic BP measurement, so that no one was excluded because of this reason.

Devices
The AND device Model UA-767 Plus was used for upper-arm BP measurement and the
AND device Model UB-542 for wrist measurement (A&D Company, Tokyo, Japan). The
UA-767 Plus was validated for BP measurement by the British Hypertension Society
(A/A grading),29 and the UB-542 was validated according to the European Society of
Hypertension International Protocol revision 2010.30 For the upper-arm measurement, a
standard cuff was used for arm circumferences of up to 32 cm and a large cuff for
circumferences of >32 cm. For the wrist measurement, the wrist circumference had to
be within the 13.5- to 21.5-cm range as recommended by the manufacturer. No subject
had a wrist circumference outside this range.

BP Measurement and Training

During the initial visit, after a 5-minute rest in the sitting posture, all subjects had their
BP measured at upper-arm and wrist by a physician in triplicate to minimize alert
reaction. The last 2 readings were averaged and considered as upper-arm office BP
and wrist office BP. Subjects then received an individual training about the use of the 2
devices, teaching theory and answering questions, if any. For upper-arm BP, they were
trained to undress their nondominant arm, to rest in the sitting position for 5 minutes, to
apply the cuff taking care of the position on the artery, to keep the elbow on a desk and
the forearm horizontal, and to proceed with automatic BP measurement without moving
from their seat. Then, they had to remove the cuff, to apply the wrist device on the same
arm, to keep the elbow on the desk with the forearm bent to place the wrist at heart
level (Figure 1A), and to measure wrist BP, once more without moving from their seat.
The waiting times were the same for all measurements. Then, the subject had to self-
measure upper-arm and wrist BP under a physician’s supervision.
 Download figure
 Download PowerPoint

Figure 1. Correct (A) and incorrect (B–D) forearm positions in wrist blood pressure
measurement. Dashed line indicates heart level. Position B (wrist higher than heart
level) leads to falsely lower values. Positions C and D (forearm in horizontal position or
vertical close to the body) lead to falsely higher values.

It was decided to measure first the upper-arm and then the wrist BP both at home and
in the office to avoid confusion, as a random protocol seemed to be too difficult to
pursue in uncontrolled home conditions.

Home Self-Measurement
Subjects were asked to self-measure the upper-arm and then the wrist BP at home
every morning and evening at the same time of the day for 7 consecutive days by
means of the 2 devices cited above, following the instructions received during training.
After getting the BP reports back, the arm and wrist BP values self-measured in the 7
days were averaged separately for systolic and diastolic BP and considered as upper-
arm home BP and wrist home BP in data analysis.
Presumptive Error in Home Self-Measurement
The difference between BP recorded at the upper arm and BP measured at the wrist
was calculated (separately for systolic and diastolic) both in the office and at home. At
home, self-measured BP on each morning and evening were averaged, obtaining for
each day a systolic and a diastolic mean value. The differences between upper-arm
BP−wrist BP day 1, upper-arm BP−wrist BP day 2, and so on, for each day were
calculated for systolic and diastolic BP and averaged to obtain the mean home upper-
arm−wrist BP difference for systolic and diastolic, respectively. The upper-arm−wrist BP
difference obtained in the office under doctor’s supervision was considered as the
reference value. To evaluate how much the upper-arm−wrist BP difference measured
by the study participant at home deviated from the office reference value, we calculated
the discrepancy between the office and home differences, separately for systolic and
diastolic BP, as (office upper-arm−office wrist BP difference) −(home upper-arm−home
wrist difference), and defined this discrepancy as home measurement error. According
to most authorities considering as tolerable a mean difference between 2
measurements up to 5 mm Hg when automatic devices are used,31 subjects showing a
home measurement error greater than ±5 mm Hg were labeled as bad measurers and
subjects with home measurement error within ±5 mm Hg as good measurers. As this
limit might be considered too restrictive, the same operation was then performed using
±10 instead of ±5 mm Hg as boundaries. Both the continuous variable home
measurement error (in mm Hg) and the categorical variable bad measurer (0=no,
1=yes) were separately used in data analysis.

Neuropsychological Assessment
At the initial visit, cognitive assessment was performed by a Mini-Mental State
Examination32 and by a comprehensive neuropsychological battery of validated paper-
and-pencil tests standardized for Italian people33 administered in a single session lasting
≈2 hours. Details are described Cognitive Pattern Assessment section of online-only
Data Supplement and in Table S1 in the online-only Data Supplement.

Ethics
The investigation conformed to the Declaration of Helsinki and institutional guidelines,
and was approved by the Ethics Committees of the University of Padua, of the
University Hospital of Verona, and of the Local Health Units No. 4 and No. 20 of the
Veneto Region (Italy). Each subject gave and signed informed consent.

Statistics

Sample size
As to our knowledge, no experience is available about arm and wrist BP self-measured
at home at a population level in epidemiological setting, it was assumed a priori a
plausible difference between arm and wrist BP at home around 10±10 mm Hg for
systolic and 5±10 for diastolic BP. Power analysis showed that 148 subjects per group
in equality for the 2 proportions test were sufficient to show effects with a power of 0.90
and a test level of 0.10 for β error and of 0.20 for α error. Therefore, the cohort of 721
subjects recruited for the present study seemed to be adequate even after stratifying by
upper-arm and wrist and by office and home measurements.

General and Descriptive Statistics

Linearity assumption was ascertained for each variable by the residuals method and
normality assumption by the Kolmogorov–Smirnov 1-sample test. Continuous variables
were expressed as mean and SD and compared between groups with ANOVA and the
Bonferroni post hoc test. In a first model, crude values were considered to describe BP
values. Then the analysis was adjusted for age, sex, highest educational level achieved,
hypertension, forearm length, upper-arm circumference, and the cognitive tests cited in
Cognitive Pattern Assessment section of online-only Data Supplement. Categorical
variables were expressed as percent rate and compared with the χ2 test. To show the
discrepancies between upper-arm and wrist BP measurements in the 2 settings, the
Bland–Altman approach was used.

Multiple Regression Analysis

The hypothesis was advanced that the home measurement error could depend on an
erroneous position of the forearm at home with the wrist at a lower level than the heart,
because of imperfect comprehension, memorization, or practical execution of
instructions. To test this hypothesis, home systolic and diastolic errors as defined above
were separately used as dependent variables in regression analyses adjusted for the
confounders listed above, having forearm length and cognitive assessment variables as
putative independent determinants.

The item bad measurer was used as dependent variable in logistic regression adjusted
for the confounders listed above. Coefficients were shown with 95% confidence
intervals. The null hypothesis was rejected for P<0.05.

Results
Descriptive Analysis
The characteristics of the cohort are summarized in Table 1. In Table S2, data are
stratified by arterial hypertension and in Table S3 by normalization of arterial
hypertension. In Tables S4, subjects are stratified according to the trait of bad
measurer.

Table 1. Characteristics of the Cohort, Also Showing

Stratification by Sex

Whole
Men Women
Items Cohort PValue
(n=310) (n=411)
(n=721)
 Table 1. Characteristics of the Cohort, Also Showing
Stratification by Sex

Whole
Men Women
Items Cohort PValue
(n=310) (n=411)
(n=721)

Age, y 49.3±15.4 49.9±15.6 48.8±15.3 0.3

Forearm
23.8±3.0 24.8±3.2 23.0±2.6 0.0001
length, cm

Upper-arm
36.9±16.0 38.4±17.0 35.9±15.1 0.04
length, cm

Upper-arm
circumference, 28.9±3.2 29.3±3.0 28.5±3.1 0.0005
cm

Office upper-
arm SBP, 131.2±19.0 135.9±17.2 127.6±19.5 0.0001
mm Hg

Office upper-
arm DBP, 82.5±10.8 85.0±10.5 80.7±10.7 0.0001
mm Hg

Office wrist
127.9±18.7 134.3±17.4 123.1±18.1 0.0001
SBP, mm Hg

Office wrist 81.9±12.0 85.5±12.0 79.1±11.2 0.0001

 Table 1. Characteristics of the Cohort, Also Showing
Stratification by Sex

Whole
Men Women
Items Cohort PValue
(n=310) (n=411)
(n=721)

DBP, mm Hg

Heart rate,
65.7±10.6 63.3±11.1 67.5±9.8 0.0001
bpm

Serum LDL-C,
126.6±30.2 130.9±30.3 123.4±29.9 0.001
mg/dL

Serum
triglycerides, 105.1±69.8 124.6±87.5 90.3±47.6 0.0001
mg/dL

Serum HDL-C,
59.4±16.7 52.7±13.7 64.6±17.0 0.0001
mg/dL

Smoking (0: 106

50 (16.1%) 56 (13.6%) 0.3
no; 1: yes) (14.7%)

Alcohol intake 325 204 121

0.0001
(0: no; 1: yes) (45.1%) (65.8%) (29.4%)

Diabetes
mellitus (0: no; 20 (2.8%) 14 (4.5%) 6 (1.5%) 0.02
1: yes)
 Table 1. Characteristics of the Cohort, Also Showing
Stratification by Sex

Whole
Men Women
Items Cohort PValue
(n=310) (n=411)
(n=721)

Hypertension 353 184 169

0.0001
(0: no; 1: yes) (49.0%) (59.4%) (41.1%)

COPD (0: no;

12 (1.7%) 6 (1.9%) 6 (1.5%) 0.7
1: yes)

Systolic and diastolic blood pressures (SBP and DBP) are those measured at the
screening, before the individual training. COPD indicates chronic obstructive pulmonary
disease; HDL-C, high-density-lipoprotein cholesterol; and LDL-C, low-density-lipoprotein
cholesterol.

All subjects had Mini-Mental State Examination >25 and were classified as having no
patent clinical cognitive impairment. As shown in Figure 2, office systolic BP was lower
at the wrist than at the arm, whereas office diastolic BP showed a negligible difference
(−2.5%; P=0.002 for systolic; −0.7% for diastolic, nonsignificant). In contrast, at home
both systolic and diastolic BPs were higher at the wrist than at the arm
(+5.6%, P<0.0001 for systolic; +5.4%, P<0.0001 for diastolic). The corresponding
values adjusted for the confounders listed above are shown in Figure S1.
 Download figure
 Download PowerPoint

Figure 2. Upper-arm and wrist blood pressures in the office and at home. P<0.0001, *vs
upper-arm systolic home blood pressure (BP), †vs wrist systolic home BP, ‡vs upper-
arm diastolic home BP, §vs wrist diastolic home BP.

These discrepancies were also present within the subgroup of people with
hypertension, both in crude and adjusted analysis (Figures S2 and S3). The upper-
arm−wrist BP differences in the office and at home are presented in Figure 3. Bland–
Altman plots for the office and home systolic BP measurements are shown in Figure 4.
A greater dispersion around the mean was present at office (SD, 13.0 mm Hg) than at
home (8.1 mm Hg). The average systolic measurement error was 9.6±15.1 mm Hg and
the diastolic measurement error was 4.6±10.0 mm Hg.
 Download figure
 Download PowerPoint

Figure 3. Differences between arm and wrist blood pressure (BP) values in the office
and at home. P<0.0001, *vs upper-arm systolic home blood pressure (BP), †vs wrist
systolic home BP, ‡vs upper-arm diastolic home BP, §vs wrist diastolic home BP.

 Download figure
 Download PowerPoint

Figure 4. Bland–Altman plot of upper-arm−wrist blood pressure differences measured

in the office and at home. Lines depict mean and 95% limits of agreement.
According to the ±5 mm Hg cutoff, 100 subjects (13.9%) turned out to be good
measurers (home measurement error within ±5 mm Hg) and 621 (86.1%) to be bad
measurers (home measurement error outside the ±5-mm Hg interval); among the latter,
508 (81.8% of the bad measurers and 70.4% of the entire cohort) had higher wrist than
upper-arm home BP and 113 had lower wrist than upper-arm home BP (18.2% of the
bad measurers and 15.7% of the entire cohort). The same analysis for ±10-mm Hg
home measurement error is shown in the Results for ±10 mm Hg Home Measurement
Error section of online-only Data Supplement.

Among the good measurers (within the ±5-mm Hg interval), the correlation between
office and home measurement was 0.854 (P<0.0001) for upper-arm systolic BP and
was 0.868 (P<0.0001) for wrist systolic BP (z statistics, −0.62; P=0.6 between the 2).
Within the bad measurers with home measurement error <−5 mm Hg (presumably those
who kept the wrist at a lower level than the heart), the correlation between office and
home measurement was 0.655 (P<0.0001 versus good measurers) for upper-arm
systolic BP and 0.591 (P<0.0001 versus good measurers) for wrist systolic BP.

Multivariable Regression Analysis

In multivariable linear regression (Table 2), lower praxic abilities were determinants of
both systolic and diastolic errors. Lower memory with interference was also a
determinant of the systolic error, and clock drawing test was a determinant of diastolic
error. Forearm length was a direct determinant of the systolic error only (Table 2).
Arterial hypertension was a direct confounder for both errors, whereas the systolic error
was also directly confounded by older age, greater arm circumference, higher systolic or
pulse pressure. In sensitivity analysis, taking selectively into consideration the 508 bad
measures presumably keeping the wrist lower than the heart level, the association
between forearm length and the systolic error was greater (coefficient, 7.22; SE, 1.46;
confidence interval, 1.20–8.31; P=0.0005), without any significant changes in the rest of
the model. The multivariable linear analysis stratified by sex is shown in Tables S5 and
S6: in men, lower memory was a determinant of both systolic error and diastolic error
and lower praxic abilities and clock drawing test of the diastolic error only. Among the
women, lower memory, lower praxic abilities, and longer forearm were determinant of
the systolic error and clock drawing test of the diastolic error.

Table 2. Whole Cohort

Independent Dependent Variable: Home Systolic Dependent Variable: Home Diastolic

Covariables Error Error
 95% CI of 95% CI of
Coefficient Coefficient
the PValue the PValue
(SE) (SE)
Coefficient Coefficient

Determinants

Praxic abilities −7.67 −15.02 to −5.94 −10.84 to

0.04 0.02
(score) (3.75) −0.32 (2.50) −1.05

−2.20 −3.81 to −3.94 −5.72 to

MI 10 (score) 0.007 0.0001
(0.82) −0.59 (0.91) −2.15

Forearm 0.56 to −0.69 −1.76 to

5.94 (2.74) 0.03 0.2
length, cm 11.32 (0.55) 0.38

Forearm 1.20 to −2.64 to

7.22 (1.46) 0.0005 0.95 (1.83) 0.6
length, * cm 8.31 4.53

Clock drawing −2.25 −4.92 to −5.94 −10.84 to

0.1 0.02
test (score) (1.36) 0.42 (2.50) −1.05

Confounders

0.04 to −0.01 to
Age, y 0.10 (0.03) 0.001 0.03 (0.02) 0.1
0.16 0.07

Sex (0:
−1.28 0.19 to −0.94 −1.91 to
women; 1: 0.1 0.1
(0.75) 2.74 (0.50) 0.04
men)
 Table 2. Whole Cohort

Dependent Variable: Home Systolic Dependent Variable: Home Diastolic

Error Error

Independent
Covariables
95% CI of 95% CI of
Coefficient Coefficient
the PValue the PValue
(SE) (SE)
Coefficient Coefficient

Arm
13.35 7.41 to −0.02 −3.97 to
circumference, 0.0001 1.0
(3.03) 19.28 (2.02) 3.94
cm

Hypertension 0.36 to 0.06 to

2.02 (0.85) 0.02 1.17 (0.56) 0.04
(1: yes) 3.68 2.28

Systolic blood
1.45 to −0.60 to
pressure, 9.81 (4.27) 0.02 4.89 (2.84) 0.1
18.18 10.46
mm Hg

Pulse
0.14 to −0.48 to
pressure, 3.23 (1.58) 0.04 1.57 (1.05) 0.1
6.33 3.63
mm Hg

Abstraction −0.80 −2.58 to −1.05 −2.24 to

0.4 0.1
(score) (0.91) 0.97 (0.60) 0.13

Digit span −2.99 to −1.74 to

0.72 (1.89) 0.7 0.73 (1.26) 0.6
(score) 4.42 3.20

IPM (score) −0.48 −2.27 to 0.6 0.11 (0.61) −1.08 to 0.9

 Table 2. Whole Cohort

Dependent Variable: Home Systolic Dependent Variable: Home Diastolic

Error Error

Independent
Covariables
95% CI of 95% CI of
Coefficient Coefficient
the PValue the PValue
(SE) (SE)
Coefficient Coefficient

(0.91) 1.31 1.31

−0.56 to −1.20 to
Education, y 1.78 (1.19) 0.1 0.36 (0.80) 0.7
4.12 1.92

Multiple linear regression. Systolic and diastolic errors as dependent variable,

respectively. Age, sex, hypertension, arm circumference, forearm length, systolic blood
pressure, educational level, memory with interference at 10 s test (MI 10), immediate
prose memory (IPM), praxic abilities, clock drawing, abstraction, and digit span tests as
independent variables. Pulse pressure was used in alternative to systolic and diastolic
blood pressure. CI indicates confidence interval.

Selectively in the 508 subjects keeping the wrist higher than the heart level.
*

In multivariable logistic regression, lower praxic abilities (odds ratio, 0.167; P=0.0005),
memory with interference (odds ratio, 0.921; P=0.007), and abstraction (odds ratio,
0.870; P=0.02) were negatively associated, and forearm length (odds ratio,
1.066; P=0.03) was positively associated with ±5 bad measurers (complete analysis is
available in Table S7).

Discussion
The main finding of our study is that the relationship between BP measured at the
upper-arm and at the wrist varied according to whether BP measurements were made
in the office under a doctor’s supervision or at home in a real-life situation. When BP
was taken in the office, the values measured at the wrist were, as shown by others, 18–
25 slightly lower than those measured at the upper-arm. In contrast, when BP was self-

measured at home by the study participants, higher BP values were obtained at the
wrist than at the arm. In the large majority of the participants classified as bad
measurers, the home measurement error was because of a disproportionately high wrist
BP.
The discrepancy between the upper-arm−wrist BP difference obtained in the 2 settings
is probably because of an error in home self-measurement despite appropriate training.
It is likely that many subjects, when left free to measure their BP at home, did not follow
the instructions received during the training session because of a deficit in memory or in
executive functions, a limitation that persisted after adjustment for age and was not
prevented by years of schooling.

Inability to follow the instructions for poor memorization and carelessness were likely to
affect wrist BP rather than upper-arm BP measurement because of the important effect
of an incorrect forearm position on wrist BP. Accurate measurement of BP at the wrist
requires that the heart and the wrist are kept at the same level to avoid the effects of
hydrostatic pressure. If the forearm is kept horizontal on the supporting desk (Figure
1C), leaning or even vertical along the subject’s side (Figure 1D), the hydraulic pressure
caused by the upper limb blood column mass is added to the hemodynamic pressure
and their sum is recorded by the wrist device. Based on the difference between density
of human blood and mercury, in the present study, the magnitude of the home
measurement error would translate into an average level discrepancy between the heart
and the wrist of 10±11 cm (confidence interval, 9.1–11.5). According to this
extrapolation, in our experience, the range of the error was from 10 over to 65 cm under
the heart level. Obviously, part of the upper-arm−wrist BP difference may be because of
random BP variability between 1 measurement and the other or to unreliable upper-arm
BP measurements, and thus the differences in level reported above can only be
considered as indicative. This interpretation was confirmed by significant effect of
forearm length in multivariable linear analyses for the systolic home measurement error.
When the wrist is kept at a lower level than the elbow (Figure 1D), a longer forearm
magnifies the hydrostatic effect of the wrong arm position,34 an effect that is notoriously
more pronounced for systolic than for diastolic BP.35–37

The reason for the wrong position of forearm in a high number of subjects was probably
because of a worse cognitive pattern, as shown by the inverse association of the
measurement error with memory, praxic abilities, visuospatial, and executive functions,
as shown for instance by the clock drawing test. It is presumable that subjects having
worse cognition were those more prone to make a mistake in wrist self-measurement.
This interpretation is corroborated by previous observations that wrist devices provided
with a position sensor, helping subjects to keep the wrist at heart level, usually give
lower values than those without a sensor.24 In the present study, it was decided to use
wrist monitors without a sensor because these are the devices mostly used in the real
world.

The home measurement error was also associated with higher BP levels or the
diagnosis of hypertension. A higher BP is likely to affect the magnitude of the
measurement error. In addition, in keeping with previous study,38 the hypertensive
participants had a much worse cognitive pattern, as shown in Table S2. However, the
measurement error was also present among the normotensives.
Clinical Relevance
An upper-arm−wrist difference greater than ±5 or ±10 mm Hg was reached at home by
the majority of the participants, presumably for an improper use of wrist BP monitors.
This is alarming as it is estimated24 that wrist devices for self-measurement have gained
30% to 50% of the market share of the BP measuring devices sold in affluent countries,
and most people use them without any preliminary training from healthcare personnel.

Limitations
The main limitation is that we had no gold standard for establishing the reliability of wrist
self-measurement at home. We could not use office wrist BP as the reference because,
as shown by our results, home BP is generally lower than office BP and is devoid of the
alarm reaction. For this reason, we used the difference between upper-arm and wrist
office BP as a reference presuming that, although at a different BP level, the upper-
arm−wrist BP difference would not vary in the 2 settings. A further limitation is that we
assumed that the home measurement error was mainly because of misuse of the wrist
rather than the arm device because of the well known problems inherent to the use of
wrist devices. However, the relationship of the forearm length with the home
measurement error lends support to our assumption. It was also decided to measure
first upper-arm and then wrist BP both at home and in the office to avoid confusion, as it
appeared that a random protocol was too difficult to pursue by subjects free to act in
uncontrolled home conditions. This is not a major problem, because at home, wrist BP
was higher than upper-arm BP. Finally, in our study, a remarkable white-coat effect was
found, probably because of the epidemiological setting in the office; however, a similar
white-coat effect was to be expected with upper-arm and wrist measurement.39

Strengths of the study are general population setting, the large sample size, taking into
consideration both morning and evening measurements as home BP, and the use of the
same 2 devices in each subject for the office and home measurements, thereby
avoiding the confounding effect of different BP monitors.

Perspectives
At a population level, the use of wrist devices for self BP measurement often leads to
detection of falsely elevated BP values. This is likely to be because of poor
comprehension and/or to incorrect application of instructions leading to incorrect
position of the arm during self-measurement, a problem that for hydraulic reasons can
be magnified by longer forearms. This represents a public health problem because of
the overestimation of people with hypertension, with increase in costs for the community
and deterioration of quality of life for the patient. A worse cognitive pattern is a key
factor in this chain of events. It is thus prudent to discourage the use of wrist devices in
patients in whom cognitive deterioration is likely to be present.

Sources of Funding
This study was funded by the Italian Ministry of Health (RF-2009-1469148, GOLDEN
study [Growing Old With Less Disease Enhancing Neurofunctions]).
 Disclosures
None.

Footnotes
The online-only Data Supplement is available with this article
at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.116.079
61/-/DC1.

Correspondence to Edoardo Casiglia, Department of Medicine, University of Padua, Via

Giustiniani, 2, Padua, Italy. E-mail edoardo.casiglia@unipd.it

References

 1.Casiglia E, Tikhonoff V, Mazza A, Pessina AC. Systolic and pulse hypertension.Aging

Health. 2005; 1:85–94.CrossrefGoogle Scholar

 2.Casiglia E, Mazza A, Tikhonoff V, Pavei A, Privato G, Schenal N, Pessina AC. Weak effect of
hypertension and other classic risk factors in the elderly who have already paid their toll.J Hum
Hypertens. 2002; 16:21–31. doi: 10.1038/sj.jhh.1001288.CrossrefMedlineGoogle Scholar

 3.Tikhonoff V, Kuznetsova T, Stolarz K, Bianchi G, Casiglia E, Kawecka-Jaszcz K, Nikitin Y,

Tizzone L, Wang JG, Staessen JA. beta-Adducin polymorphisms, blood pressure, and sodium
excretion in three European populations.Am J Hypertens. 2003; 16:840–
846.CrossrefMedlineGoogle Scholar

 4.Casiglia E, Tikhonoff V, Mazza A, Piccoli A, Pessina AC. Pulse pressure and coronary
mortality in elderly men and women from general population.J Hum Hypertens. 2002; 16:611–
620. doi: 10.1038/sj.jhh.1001461.CrossrefMedlineGoogle Scholar

 5.Casiglia E, Palatini P. Cardiovascular risk factors in the elderly.J Hum

Hypertens. 1998; 12:575–581.CrossrefMedlineGoogle Scholar

 6.Little P, Barnett J, Barnsley L, Marjoram J, Fitzgerald-Barron A, Mant D. Comparison of

acceptability of and preferences for different methods of measuring blood pressure in primary
care.BMJ. 2002; 325:258–259.CrossrefMedlineGoogle Scholar
 7.Boggia J, Thijs L, Hansen TW, et al.; International Database on Ambulatory blood pressure in
relation to Cardiovascular Outcomes Investigators. Ambulatory blood pressure monitoring in
9357 subjects from 11 populations highlights missed opportunities for cardiovascular prevention
in women.Hypertension. 2011; 57:397–405. doi:
10.1161/HYPERTENSIONAHA.110.156828.LinkGoogle Scholar

 8.McManus RJ, Wood S, Bray EP, Glasziou P, Hayen A, Heneghan C, Mant J, Padfield P,
Potter JF, Hobbs FD. Self-monitoring in hypertension: a web-based survey of primary care
physicians.J Hum Hypertens. 2014; 28:123–127. doi:
10.1038/jhh.2013.54.CrossrefMedlineGoogle Scholar

 9.Manning DM, Kuchirka C, Kaminski J. Miscuffing: inappropriate blood pressure cuff

application.Circulation. 1983; 68:763–766.CrossrefMedlineGoogle Scholar

 10.Mourad JJ, Lopez-Sublet M, Aoun-Bahous S, Villeneuve F, Jaboureck O, Dourmap-Collas C,

Denolle T, Fourcade J, Baguet JP. Impact of miscuffing during home blood pressure
measurement on the prevalence of masked hypertension.Am J Hypertens. 2013; 26:1205–
1209. doi: 10.1093/ajh/hpt084.CrossrefMedlineGoogle Scholar

 11.Mengden T, Asmar R, Kandra A, Di Giovanni R, Brudi P, Parati G. Use of automated blood

pressure measurements in clinical trials and registration studies: data from the VALTOP
Study.Blood Press Monit. 2010; 15:188–194. doi:
10.1097/MBP.0b013e328339d516.CrossrefMedlineGoogle Scholar

 12.Niyonsenga T, Vanasse A, Courteau J, Cloutier L. Impact of terminal digit preference by

family physicians and sphygmomanometer calibration errors on blood pressure value:
implication for hypertension screening.J Clin Hypertens (Greenwich). 2008; 10:341–
347.CrossrefMedlineGoogle Scholar

 13.Nietert PJ, Wessell AM, Feifer C, Ornstein SM. Effect of terminal digit preference on blood
pressure measurement and treatment in primary care.Am J Hypertens. 2006; 19:147–152. doi:
10.1016/j.amjhyper.2005.08.016.CrossrefMedlineGoogle Scholar
 14.Graves JW, Bailey KR, Grossardt BR, Gullerud RE, Meverden RA, Grill DE, Sheps SG. The
impact of observer and patient factors on the occurrence of digit preference for zero in blood
pressure measurement in a hypertension specialty clinic: evidence for the need of continued
observation.Am J Hypertens. 2006; 19:567–572. doi:
10.1016/j.amjhyper.2005.04.004.CrossrefMedlineGoogle Scholar

 15.Shirasaki O, Terada H, Niwano K, Nakanishi T, Kanai M, Miyawaki Y, Souma T, Tanaka T,

Kusunoki T. The Japan Home-health Apparatus Industrial Association: investigation of home-
use electronic sphygmomanometers.Blood Press Monit. 2001; 6:303–
307.CrossrefMedlineGoogle Scholar

 16.Westhoff TH, Schmidt S, Meissner R, Zidek W, van der Giet M. The impact of pulse pressure
on the accuracy of wrist blood pressure measurement.J Hum Hypertens. 2009; 23:391–395.
doi: 10.1038/jhh.2008.150.CrossrefMedlineGoogle Scholar

 17.Vaïsse B, Mourad JJ, Girerd X, Hanon O, Halimi JM, Pannier B; Comité Français de lutte
Contre l’hypertension Artérielle. Flash Survey 2012: the use of self-measurement in France and
its evolution since 2010.Ann Cardiol Angeiol (Paris). 2013; 62:200–203. doi:
10.1016/j.ancard.2013.03.003.MedlineGoogle Scholar

 18.Palatini P, Longo D, Toffanin G, Bertolo O, Zaetta V, Pessina AC. Wrist blood pressure
overestimates blood pressure measured at the upper arm.Blood Press Monit. 2004; 9:77–
81.CrossrefMedlineGoogle Scholar

 19.Uen S, Fimmers R, Brieger M, Nickenig G, Mengden T. Reproducibility of wrist home blood

pressure measurement with position sensor and automatic data storage.BMC Cardiovasc
Disord. 2009; 9:20. doi: 10.1186/1471-2261-9-20.CrossrefMedlineGoogle Scholar

 20.Kikuya M, Chonan K, Imai Y, Goto E, Ishii M; Research Group to Assess the Validity of
Automated Blood Pressure Measurement Devices in Japan. Accuracy and reliability of wrist-cuff
devices for self-measurement of blood pressure.J Hypertens. 2002; 20:629–
638.CrossrefMedlineGoogle Scholar
 21.Khoshdel AR, Carney S, Gillies A. The impact of arm position and pulse pressure on the
validation of a wrist-cuff blood pressure measurement device in a high risk population.Int J Gen
Med. 2010; 3:119–125.CrossrefMedlineGoogle Scholar

 22.Mourad A, Gillies A, Carney S. Inaccuracy of wrist-cuff oscillometric blood pressure devices:

an arm position artefact?Blood Press Monit. 2005; 10:67–71.CrossrefMedlineGoogle Scholar

 23.Sato H, Koshimizu H, Yamashita S, Ogura T. Blood pressure monitor with a position sensor
for wrist placement to eliminate hydrostatic pressure effect on blood pressure
measurement.Conf Proc IEEE Eng Med Biol Soc. 2013; 2013:1835–1838. doi:
10.1109/EMBC.2013.6609880.MedlineGoogle Scholar

 24.Guggiari C, Büla C, Iglesias K, Waeber B. Measurement with an automated oscillometric

wrist device with position sensor leads to lower values than measurements obtained with an
automated oscillometric arm device from the same manufacturer in elderly persons.Blood
Press Monit. 2014; 19:32–37. doi: 10.1097/MBP.0000000000000013.CrossrefMedlineGoogle
Scholar

 25.Tikhonoff V, Casiglia E, Guidotti F, Giordano N, Martini B, Mazza A, Spinella P, Palatini

P. Body fat and the cognitive pattern: a population-based study.Obesity (Silver
Spring). 2015; 23:1502–1510. doi: 10.1002/oby.21114.CrossrefMedlineGoogle Scholar

 26.Rose GA, Blackburn H, Gillum RF, Prineas RJ. Survey questionnaires., Rose GA, Blackburn
H, Gillum RF, Prineas RJ. In: Cardiovascular Survey Methods. Geneva,
Switzerland: WHO;1982:64–77.Google Scholar

 27.Wormser D, Di Angelantonio E, Kaptoge S, et al.; The Emerging Risk Factors

Collaboration. Adult height and the risk of cause-specific death and vascular morbidity in 1
million people: individual participant meta-analysis.Int J Epidemiol. 2012; 41:1419–
1433.CrossrefMedlineGoogle Scholar

 28.Haymsfield SB, Smith J. Muscle mass: reliable indicator of protein-energy malnutrition

severity and outcome.Am J Clin Nutr. 1982; 32:1192–1199.CrossrefGoogle Scholar
 29.Kobalava ZD, Kotovskaia LV, Rusakova OS, Babaeva LA. Validation of the UA-767 Plus
device for self-measurment of blood pressure.Clin Pharmac Ther. 2003; 12:2.Google Scholar

 30.Saladini F, Benetti E, Fania C, Palatini P. Validation of the A&D BP UB-542 wrist device for
home blood pressure measurement according to the European Society of Hypertension
International Protocol revision 2010.Blood Press Monit. 2013; 18:219–222. doi:
10.1097/MBP.0b013e3283624aa2.CrossrefMedlineGoogle Scholar

 31.Blood Pressure Measurement Devices. Medicines and Healthcare Products Regulatory

Agency, 2013, 9 pages.Google Scholar

 32.Magni E, Binetti G, Bianchetti A, Rozzini R, Trabucchi M. Mini-Mental State Examination: a

normative study in Italian elderly population.Eur J Neurol. 1996; 3:198–202. doi:
10.1111/j.1468-1331.1996.tb00423.x.CrossrefMedlineGoogle Scholar

 33.Casiglia E, Giordano N, Tikhonoff V, Boschetti G, Mazza A, Caffi S, Guidotti F, Bisiacchi

P. Cognitive Functions across the GNB3 C825T Polymorphism in an Elderly Italian
Population.Neurol Res Int. 2013; 2013:597034. doi:
10.1155/2013/597034.CrossrefMedlineGoogle Scholar

 34.Thomas SS, Nathan V, Zong C, Soundarapandian K, Shi X, Jafari R. BioWatch: a non-

invasive wrist-based blood pressure monitor that incorporates training techniques for posture
and subject variability.IEEE J Biomed Health Inform. 2015; 94:1–10. doi:
10.1109/JBHI.2015.2458779.Google Scholar

 35.Gavish B, Gavish L. Blood pressure variation in response to changing arm cuff height cannot
be explained solely by the hydrostatic effect.J Hypertens. 2011; 29:2099–2104. doi:
10.1097/HJH.0b013e32834ae315.CrossrefMedlineGoogle Scholar

 36.Gavish B, Gavish L. Simple determination of the systolic-diastolic pressure relationship from

blood pressure readings taken at different arm heights.Blood Press Monit. 2013; 18:144–150.
doi: 10.1097/MBP.0b013e328361c8fd.CrossrefMedlineGoogle Scholar
 37.Deutsch C, Krüger R, Saito K, Yamashita S, Sawanoi Y, Beime B, Bramlage P. Comparison
of the Omron RS6 wrist blood pressure monitor with the positioning sensor on or off with a
standard mercury sphygmomanometer.Blood Press Monit. 2014; 19:306–
213.CrossrefMedlineGoogle Scholar

 38.Semplicini A, Amodio P, Leonetti G, Cuspidi C, Umiltà C, Schiff S, Scheltens P, Barkhof F,

Emanueli C, Cagnin A, Pizzolato G, Macchini L, Realdi A, Royter V, Bornstein NM, Madeddu
P. Diagnostic tools for the study of vascular cognitive dysfunction in hypertension and
antihypertensive drug research.Pharmacol Ther. 2006; 109:274–283.CrossrefMedlineGoogle
Scholar

 39.Fujita H, Matsuoka S, Awazu M. White-coat and reverse white-coat effects correlate with 24-
h pulse pressure and systolic blood pressure variability in children and young adults.Pediatr
Cardiol. 2016; 37:345–352. doi: 10.1007/s00246-015-1283-5.CrossrefMedlineGoogle Scholar