Content uploaded by Alexandre C B Ramos
Author content
All content in this area was uploaded by Alexandre C B Ramos on Apr 24, 2018
Content may be subject to copyright.
Relationship Between Upper Arm Muscle
Index and Upper Arm Dimensions in Blood
Pressure Measurement in Symmetrical Upper
Arms: Statistical and Classification
and Regression Tree Analysis
Letícia Helena Januário
1(&)
, Alexandre Carlos Brandão Ramos
2
,
Paôla de Oliveira Souza
3
, Rafael Duarte Coelho Santos
4
,
Helen Cristiny T. Couto Ribeiro
1
, JoséMaria Parente de Oliveira
3
,
and Hevilla Nobre Cezar
2
1
Campus Centro-oeste, Universidade Federal
de SãoJoão del Rei, Divinópolis, MG, Brazil
{leticiahj,helen.cristiny}@ufsj.edu.br
2
Instituto de Matemática e Computação,
Universidade Federal de Itajubá, Itajubá, MG, Brazil
ramos@unifei.edu.br, hevilla@gmail.com
3
Instituto Tecnológico de Aeronáutica, São Josédos Campos, SP, Brazil
paola.cefet@gmail.com, parente@ita.br
4
Instituto Nacional Pesquisas Espaciais, São Josédos Campos, SP, Brazil
rafael.santoa@inpe.br
Abstract. Objective: to identify the influence of upper arm muscle index
(AMI) and upper arm dimensions on the measurement of blood pressure (BP).
Methodology: 489 university students were evaluated in Divinópolis, Brazil and
data were collected on anthropometric and BP measurements and elaborated
multiple linear regression and regression tree models using data mining tech-
niques. Results: length arm (AL) and arm circumference (AC) showed positive
correlation with systolic blood pressure (SBP) and diastolic blood pressure
(DBP). The AMI presented positive correlation with SBP and negative corre-
lation with DBP. The regression tree showed interactions between BP and AL,
AC and AMI. Conclusion: BP values in the upper right upper arm were higher
than in the left upper arm in population of healthy young adults. AL and AC
were predictors of overestimation of indirect measurement of SBP and
DBP. AMI overestimates SBP and underestimates DBP. There were interactions
between arm dimensions and BP.
Keywords: Blood pressure measurement Blood pressure CART
Body composition
©Springer International Publishing AG, part of Springer Nature 2018
Á. Rocha et al. (Eds.): WorldCIST'18 2018, AISC 746, pp. 1178–1187, 2018.
https://doi.org/10.1007/978-3-319-77712-2_113
1 Introduction
Arterial hypertension (AH) is one of the most important diseases from an epidemio-
logical and economic perspective. It is a health problem that is directly associated with
cardiovascular diseases that constitute the major cause of human morbidity and mor-
tality in the world [1,2]. AH is characterized by a persistent increase in blood pressure
(BP) which is usually asymptomatic [3]. If there are no symptoms, it is through the
systematic measurement of these values that it is possible to identify changes, thus
making the introduction of therapeutic measures also possible. The BP measurement is
usually performed by the indirect method which is simpler, non-invasive and less
costly. This method consists, in brief, of the compression and decompression of the
artery, correlating the pulsation with the unit in millimeter of mercury (mmHg) of the
manometer. In this sense, BP measurement is extremely relevant and must surround
itself with techniques and procedures that enable the highest possible precision.
Analogue BP measurements with a mercury sphygmomanometer may be the most
accurate. However, there is growing evidence that the use of uncalibrated equipment,
inappropriate cuff size, and integer preference in reading the values make the method
less precise and even inconsistent. As a result, one can get diagnostic errors that can
lead individuals to unnecessary drug therapy or lack of essential treatment. There is
considerable reduction in the risk of these errors through the use of automatic electronic
equipment properly validated for BP measurement [4]. On the other hand, many studies
have questioned the possible variation, as well as the lack of exposure of algorithms
used in electronic equipment [5–7].
In the standardization of the BP measurement technique, the selection of cuffs,
proportional to upper arm circumference (AC), is recommended. However, in the same
AC there are different proportions of tissues with different densities such as muscle, fat,
skin and attachments, nerves, blood vessels and bone. The difference in tissue density
characterizes different degrees of hardness which means greater or less resistance to
compression of the arm. Therefore, the pressure required to be applied to the arm body
structures for brachial artery occlusion may lead to the highest BP result in the case of a
higher proportion of structures with higher density, such as muscles, or the lower BP in
the case of higher proportion of structures with lower density as fat.
Thus, studying the variation of the proportion of muscles and fat in the arm may be
relevant to increase the accuracy of the BP measurement. The possible variation of
pressure values as a function of the different percentages of fat and upper arm muscles
was not found in the specific literature.
This paper presents a study to identify the influence of the muscle index of the
upper arm on the BP measurement in symmetrical upper arms.
2 Methodology
The study was conducted at the Federal University of SãoJoão del Rei (UFSJ) in
Divinópolis, Brazil. Data were collected from a population of 489 healthy young adult
university students aged 18 to 29 years using anthropometric and blood pressure
measurements and questionnaires. The option for young adults is justified by the
Statistical and Classification and Regression Tree Analysis 1179
possibility of finding several arm circumferences and comparing BP values within the
normal range. To identify the proportion of arm muscle, the muscle index of the arm
was used, which was calculated according to Frisancho [9]. The AMI is the result of the
estimation of the muscular circumference (MC) in relation to the AC value, which
results from the sum of muscle, fat, blood vessels and bone. The AMI value is cal-
culated as follows:
AMI ¼MC
AC ð1Þ
Participant data that informed the use of medications and/or the presence of dis-
eases that could influence blood pressure values were excluded from the study. The
equipment used was duly validated and calibrated for anthropometric measures of
height weight, perimeters, circumferences and skinfolds according to NAHNES [8]
(anthropometric tape lufkin, scientific adipometer Harpenden Skinfold CaliperI). The
anthropometric measurements were performed in triplicate and the mean was used. The
variables derived from the anthropometric measurements were estimated according to
Frisancho [9].
BP was measured on both upper arms simultaneously with properly calibrated
electronic devices and a cuff suitable for upper arm circumference. Three measure-
ments were taken and the average of the last two measurements was used. The tech-
nique used to measure BP is in accordance with international guidelines [10,11].
The collected and estimated data were submitted to statistical analysis and data
mining. Multiple linear regression and classification and regression trees (CART)
models were performed using the tool R.
3 Results and Discussions
Blood pressure values were normal and the mean number of participants (n=489) was
109.26 mmHg for systolic blood pressure (SBP) and 64.72 mmHg for diastolic blood
pressure (DBP). The mean pressure in the right upper arm was SBP: 109.91 mmHg and
DBP: 64.01 mmHg. In the left upper arm, mean SBP was 108.60 mmHg and DBP was
65.42 mmHg (Table 1). The confidence interval was 95%. The mean age of the par-
ticipants was 21.74.
According to Table 1, mean values of SBP, DBP, triceps skinfold (TS) and upper
arm muscle index (AMI) were different between upper arms, with statistically signif-
icant values (p < 0.05). However, in general, publications related to body composition
use TS values, precluding muscle fractions. Significant differences were not observed
in relation to length values (AL) and upper arm circumference (AC).
From these results, assuming the similarity of AL and AC and the difference
between the values of BP, TS and AMI, we chose to use n of 978 [sum of n of both
upper arms (489)]. This is justified by measures that statistically show that the right and
left upper arms are symmetrical in length and circumference, but asymmetrical in the
values of the TS and the AMI and with different BP values.
1180 L. H. Januário et al.
The mean values of the participants’blood pressure values were different inter-arm.
This asymmetry of pressure values has been widely discussed in the literature, but there
is still no consensus among publications. There were differences in pressure values
inter-arms for men and women, both for SBP (89.6% of women and 98.1% for men)
and for DBP (92.1% of women and 96.5% of men). BP was higher in the right upper
arm than in the left upper arm for both sexes [12]. Differences in BP inter-arms which
was greater in the right upper arm were also found, but without statistically significant
values [13]. Another study aimed at clarifying if both upper arms are equally good for
assessing BP in the general population found only a small difference in BP between the
upper arms in a healthy population of 484 participants aged 25–74 years [14].
A review of meta-analysis was performed with 22 articles, which related the dif-
ference in blood pressure values between upper arms, vascular diseases and increased
cardiovascular mortality and all-cause mortality. According to the authors, a difference
of 15 mmHg or greater may be a useful indicator of the risk of vascular disease and
death by [15]. An increased SBP difference inter-arms ( 6 mmHg) is associated with
atherosclerotic coronary disease burden [16,17].
The values of SBP correlated positively with AL, AC, MC and AMI, and nega-
tively with TS. The DBP values positively correlated with AL, AC, TS, and negatively
with MC and AMI. The variables that had the strongest correlations (AL, AC and AMI)
were included in multiple linear regression models (LRM) To verify the relationship of
SBP and DBP with upper arm dimension and upper arm body composition (estimated
by AMI). The analysis resulted in statistically significant models for SBP and DBP, as
presented in Eqs. 1and 2.
DBP: F3;971ðÞ¼16;057;p\0;001;
R2¼0;47
ð2Þ
SBP: F3;971ðÞ¼151;813;p\0;001;
R2¼0;319
ð3Þ
Table 1. Comparison of the mean of variables by hemi-
corps
Variables Average per half-body P value
Right Left
SBP 109.91 108.60 0.04
DBP 64.01 65.42 0.001
AL 36.49 36.23 0.221
AC 27.22 27.18 0.868
TS 18.50 17.18 0.014
MC 21.41 21.79 0.000
AMI 0.49653 0.349 0.000
Statistical and Classification and Regression Tree Analysis 1181
Model 1, which includes AL, explained 22% of the increase in SBP. Model 2,
including AL and AC, explained 29% of the increase in BP. Model 3, including AL,
AC and AMI accounted for 32% of the increase in BP.
Thus, AL, AC and AMI variables can be considered as predictors of changes SBP
and DBP in the indirect measure of BP. Mathematical Eqs. 3and 4describe these
relationships. Tables 2and 3show the coefficients used to interpret the influence of
each independent variable on the model.
SBP ¼47;207 þ0;918:AL þ0;772:AC þ15;144:AMI ð4Þ
DBP ¼50;153 þ0;336:AL þ0;227:AC 7;476:AMI ð5Þ
The values of AL, AC and AMI were predictors of changes in blood pressure values in
the indirect measurement of both SBP and DBP. The AL and AC values increase the
SBP and DBP values. However, the behavior of the AMI was different, contributing to
the increase in SBP and to the reduction of DBP.
Pressure values were correlated with the AMI. The densities of human tissues,
specifically of muscle and fat, are different. The tissue density of mammalian skeletal
muscle is estimated to be 1.06 kg/l and adipose tissue (fat) is 0.92 kg/l. The higher
density in the area of BP measurement may require more force for brachial artery
occlusion and, therefore, overestimating pressure values [18–20].
This result is similar to the study by Vaziri et al. [21]. The authors found a positive
association between the lean mass determined by MC, and the BP in university stu-
dents. However, this correlation has not been explored. Frequently, studies explore the
relationship between BP and body composition estimated by body mass index
(AMI) or the sum of skinfolds, others with AC, and still others with TS.
BP has frequently been correlated positively with body fat in cross-sectional epi-
demiological studies [22]. For the authors this may be a bias of using only one cuff size
for adults, disregarding the differences between upper arm circumferences. The cor-
relations between BP and obesity found in epidemiological studies may have been
significantly influenced by the effects of AC variation.
Table 2. Table of coefficients for SBP
SBP ßtP
AL 0.310 10.044 0.000
AC 0.296 10.639 0.000
AMI 0.175 5.920 0.000
Table 3. Table of coefficients for DBP
DBP ßtP
AL 0.158 4.330 0.000
AC 0.121 3.688 0.000
AMI −0.120 −3.445 0.001
1182 L. H. Januário et al.
On the other hand, the strong negative correlation between BP and AC of six obese
participants, which for the author suggests a difference in the behavior of cutaneous
fold values between obese and non-obese individuals [24].
However, the reproducibility of determining the difference in blood pressure values
inter-arms, lacks concordance, for which the authors suggest further studies [25].
Classification and Regression Trees (CART) methodology was also used to
develop models that can predict SBP and DBP values. CART analysis is a tree-building
technique in which several “predictor”variables are tested to determine how they
impact the “outcome”variable. The algorithm selects the predictor that provides the
best or “optimal”split, such that each of the two subgroups is more homogeneous with
respect to outcome. Each subgroup is further dichotomized into smaller and more
homogeneous groups by choosing the variable that best splits the subgroup [26]. Thus,
the CART method is able to determine the complex interactions among variables in the
final tree. In studies which the distributions of variables are not well known, CARTs
provide a model void of any assumptions about the distribution of the variables,
preventing model misspecifications.
The variables included in the CART analysis were AL, AC, AMI, TS and MC.
Using the CART hierarchically, AL, AC and AMI emerged as predictors of SBP and
DBP, which are the same set of variables used in the multiple linear regression.
Notably, TS and MC were not important predictors of BP. The trees resulted from
CART are showed at Fig. 1for SBP and Fig. 2for DBP.
Fig. 1. CART tree SBP
Statistical and Classification and Regression Tree Analysis 1183
Figure 1shows interaction between AL, AC and AMI in prediction SBP values.
The SBP mean increases when: AC > 39 and CB < 26; and when AC > 34, AL > 27
and AMI > 0.61. The SBP mean decrease when: AC < 39, AMI < 0.61 and AL < 27.
The tree at Fig. 2shows interactions between AL, AC and AMI to predict
DBP. The DBP mean increases when: AL > 33 and AMI < 0.57; and when AC > 34,
AL > 25 e o AMI < 0.59. On the other hand, DBP mean decreases when: AL < 33
and AC > 34; when AL is measured between 22 e 25, AC > 34; and when AL < 25, o
AC > 34, e o AMI 0.59.
A comparison between LRM and CART model is showed at Table 4. The pre-
diction by the SBP has a greater error than by the DBP. It also turns out that the CART
model is slightly better than linear regression models.
Fig. 2. CART tree DBP
Table 4. Predictors error
Method SBP DBP
LM 68.2883125 48.5061875
CART 68.3663755 45.1417131
1184 L. H. Januário et al.
4 Potentiality and Limitations
The main limitation of this work was the use of anthropometric measures to estimate
the body composition of the upper arms. This was due to the greater ease of access to
the equipment and the more economical budget.
The most relevant aspects of this work were the simultaneous measurement of BP,
strictly controlled, performed by a single researcher and the taking of anthropometric
measures also performed by a single researcher.
5 Theoretical and Practical Contributions
The study contributes with the technological advance in the arterial hypertension
control in both theoretical and practical perspectives. In the theoretical aspect is pre-
sented a new proposal of discussion regarding the procedure of blood pressure mea-
surement: the inclusion of the variable AMI which in practice can increase the accuracy
of the measurement.
6 Conclusion
The mean values of BP in this population of healthy young adults were different
between hemi-bodies, higher SBP in the right upper arm and greater DBP in the left
upper arm. The AL and AC values may overestimate the indirect measure of SBP and
DBP. The AMI value may overestimate the SBP and underestimate the DBP in the
indirect measure of BP. There were interactions between the upper arm dimensions
both in SBP and DBP. Further research on the relationship between the measurement of
BP and different body compositions in the upper arm is suggested, using other tech-
niques for estimating fat and muscle percentages, such as dual beam radiological
absorptiometry, as well as analyzes with other data mining techniques.
References
1. Teo, K., et al.: Prevalence of a healthy lifestyle among individuals with cardiovascular
disease in high-, middle- and low-income countries the prospective urban rural epidemiology
(PURE) study. JAMA 309(15), 1613–1621 (2013)
2. Moreira, J.P.L., Moraes, J.R., Luiz, R.R.: Prevalence of self-reported systemic arterial
hypertension in urban and rural environments in Brazil: a population-based study. Cad.
Saúde Pública 29(1), 62–72 (2013)
3. Zhou, B., et al.: Worldwide trends in blood pressure from 1975 to 2015: a pooled analysis of
1479 population-based measurement studies with 191 million participants. Lancet 389
(10064), 37–55 (2017)
4. Kaczorowski, J., Dawes, M., Gelfer, M.: Measurement of blood pressure: New develop-
ments and challenges. BCMJ 54(8), 399–403 (2012)
Statistical and Classification and Regression Tree Analysis 1185
5. Alpert, B.S., Quinn, D., Gallick, D.: Oscillometric blood pressure: a review for clinicians.
J. Am. Soc Hypertens 8(12), 930–938 (2014)
6. Forouzanfar, M., et al.: Oscillometric blood pressure estimation: past, present, and future.
IEEE Rev. Biomed. Eng. 8,44–63 (2015). Disponível em: https://www.ncbi.nlm.nih.gov/
pubmed/25993705
7. Benmira, A., et al.: From korotkoff and marey to automatic non-invasive oscillometric blood
pressure measurement: does easiness come with reliability? Expert Rev. Med. Devices 13(2),
179–189 (2016)
8. Frisancho, A.R.: Anthropometric standards for the assessment of growth and nutritional
status. The University of Michigan Press, Ann Arbor (1990)
9. NAHNES –National Health and Nutrition Examination Survey. (2007). Anthropometry
Procedures Manual. 65 p. (acesso em maio 2017). Disponível em: http://www.cdc.gov/nchs/
data/nhanes/nhanes07-08/manual_an.pdf
10. WHO (World Health Organization). Global action plan for the prevention and control of
noncommunicable diseases 2013–2020. (2013). (acesso em: 2016 nov). Disponível em.
http://apps.who.int/iris/bitstream/10665/94384/1/9789241506236_eng.pdf?ua=1
11. Sociedade Brasileira de Cardiologia: Departamento de Hipertensão Arterial. VII Diretrizes
brasileiras de hipertensão. Rio de Janeiro. Arq Bras Cardiol, 107 (supl. 3) 1–83 (2016 Set)
12. Kim, K.B., et al.: Inter-arm differences in simultaneous blood pressure measurements in
ambulatory patients without cardiovascular diseases. Korean J. Fam. Med. 34(2), 98–106
(2013)
13. Fonseca-Reyes, S., Forsyth-MacQuarrie, A.M., García de Alba-García, J.E.: Simultaneous
blood pressure measurement in both arms in hypertensive and nonhypertensive adult
patients. Blood Press Monit. 17(4), 149–154 (2012)
14. Johansson, J.K., Puukka, P.J., Jula, A.M.: Interarm blood pressure difference and target
organ damage in the general population. J. Hypertens. 32(2), 260–266 (2014)
15. Song, X., et al.: Association of simultaneously measured four-limb blood pressures with
cardiovascular function: a cross-sectional study. Biomed Eng Online 15(Suppl2), 147 (2016)
16. Her, A.Y., et al.: Association of inter-arm systolic blood pressure difference with coronary
atherosclerotic disease burden using calcium scoring. Yonsei Med. J. 58(5), 954–958 (2017)
17. Hirono, A., et al.: Development and validation of optimal cut-off value in inter-arm systolic.
blood pressure difference for prediction of cardiovascular events. J. Cardiol. 71(1), 24–30
(2017). pii: S0914-5087(17)30197-1
18. Farvid, M.S., et al.: Association of adiponectin and resistin with adipose tissue compart-
ments, insulin resistance and dyslipidaemia. Diabetes Obes. Metab. 7(4), 406–413 (2005)
19. Urbanchek, M.G., et al.: Specific force deficit in skeletal muscles of old rats is partially
explained by the existence of denervated muscle fibers. J. Gerontol. A Biol. Sci. Med. Sci.
56(5), 191–197 (2001)
20. Mendez, J., Keys, A.: Density and composition of mammalian muscle. Metabolism 9(2),
184–188 (1960)
21. Vaziri, Y., et al.: Lean Body Mass as a Predictive Value of Hypertension in Young Adults, in
Ankara, Turkey. Iran. J. Public Health 44(12), 1643–1654 (2015)
22. Abdollahian, M.: The impact of body mass index on low birth weight. In: Third International
Conference on information technology: new generations, pp. 567–572 (2013). https://doi.
org/10.1109/itng.2013.71
23. Uluaszek, S.J., Hennberg, M.: Results of epidemiological studies of blood pressure are
biased by continuous variation in arm size related to body mass. Hum. Biol. 84(4), 437–444
(2012)
1186 L. H. Januário et al.
24. Rosenthal, J.: Evaluación metodológica de la relación entre presión arterial sanguinea y
pliegue cutáneo: un enfoque epidemiológico. Rev. Saúde Pública, São Paulo (Internet), 23
(4) 322–335. (acesso em: 2017 Out 09). Disponível em: http://www.scielo.br/scielo.php?
script=sci_arttext&pid=S0034-89101989000400008&lng=pt
25. Kim, Y.H., So, W.Y.: A low arm and leg muscle mass to total body weight ratio is associated
with an increased prevalence of metabolic syndrome: The Korea National Health and
Nutrition Examination Survey 2010–2011. Technol. Health Carev. 24(5), 655–663 (2016)
26. Barlin, J.N., et al.: Classification and Regression tree (CART) analysis of endometrial
carcinoma: Seeing the forest for the trees. J. Gynecol. Oncol. 130, 452–456 (2013)
Statistical and Classification and Regression Tree Analysis 1187
