© Borgis - New Medicine 1/2010, s. 2-6
*Petr Kutáč
Reliability of body composition measurement by the BIA Method (bioelectric impedance)
Centre of Human Motion Diagnostics – Department of Physical Education, Pedagogical Faculty, University of Ostrava, Czech Republic
Head of Department: Doc. PhDr. Vojtěch Gajda, CSc.
Summary
Introduction. The basic condition for credibility of the measured values for all the used methods is their standardization. Aside from complying with standard measuring conditions it includes especially knowledge of the method´s validity and reliability.
Aim. The aim of this study is to estimate the specific reliability of body composition measurement by the BIA Method (using TANITA 418 MA device), by means of its parallel form generated by multiple repetition of the given measurement.
Material and methods. The experimental sample group comprised of 91 students of Physical Education (53 men and 38 women) whose average age was 20.5 and 20.1 years, respectively. A tetrapolar bioimpedance weighing machine TANITA 418 MA was used for measurement of body fat ratio by the BIA method. For estimation of the specific reliability we used a calculation of correlation coefficient between two parallel forms of measurement. All the statistic tests were realized on a significance level α/symbol> = 0.05.
Results. The values of Pearson correlation coefficient (rXX') exceeded 0.970 in all tested parameters. The value of body weight was calculated as 1.000. The values of standard test error (SE) for monitored parameters oscillated between 0.06-0.12.
Conclusion. With regard to the high values of rXX' and low values of SE we are able to assess the impact ratio of the external intervention affecting the changes in body composition even in relatively small changes of the monitored parameters. The basic condition for this is a strict complying with the standard measurement conditions.
Introduction
Apart from the body height, we can include among the basic anthropometric characteristics also the body weight. For this reason, the body weight is, in most studies, one of the main characteristics of the monitored sample. However, the body weight is a relatively complex parameter which is composed of a series of fractions (components) (1, 2, 3). Therefore, when we want to assess whether an individual has an appropriate body weight, using only some weight-height index without assessment of the individual body weight fractions can be, in many cases, misleading, and it can serve only as a draft assessment, rather for average population. In the sport sphere we can find cases when the BMI levels indicate overweight in the monitored sportsman, sometimes on the verge of obesity. These individuals have a low fat ratio though, and their increased weight is caused by an intensive development of the muscle-skeletal apparatus (4, 5, 6, 7). Therefore it is necessary to examine the proportion of the individual fractions (body composition) (8). Body composition is presently a largely monitored parameter, because it is considered to be an indicator of the individual´s health condition, his nutrition level and physical fitness (9, 10). In sports, it can – to a certain degree – determine the readiness level of the sportsman´s organism for the training and race body burden. And therefore we can consider it to be one of the key aspects of the sport performance. By means of monitoring the body composition changes we can assess the influence of the physical exercise (training burden) on the sportsman´s organism. For this reason the body composition determination is an integral part of the sportsmen diagnostics during functionality examinations (11, 12, 13, 14). The most often monitored fraction is body fat. In sport practice, some trainers and specialists consider the 1% change of body fat ratio significant (8, 15).
When determining body composition it is always necessary to measure a series of parameters by means of various methods which can be divided into laboratory methods and field methods. Selected laboratory methods serve at the same time as reference methods. In conditions of a field practice they are demanding in terms of technical equipment, staff skills requirements, organization and last, but not least, costs (16, 17, 18). Therefore, in field practice we use field methods which allow field examination of the large samples – they are not demanding in terms of organization, staff training and they are also affordably priced. Anthropometry can be also included into these methods and, especially, bioelectric impedance method (BIA), which is now very popular. This method works on the basis of different flow of a low-intensity electric current in various biological structures. It is based on the principle of different electric characteristics of tissues, fat and mainly body water. The current flows through water and through electrolyte components in fat free mass and the resulting resistance is therefore proportional to its volume (17). Individual fractions of the body weight are calculated from the values of impedance on the basis of regression equations (19).
The basic condition for credibility of the measured values at all used methods is their standardization. Apart from complying with the standard measurement conditions (which are described both in treatises and in measuring device manuals), it includes especially knowledge of a method´s validity and reliability (20).
The presented paper deals with the BIA method realized by means of Tanita BC – 418 device. The issue of validity of the BIA method realized by means of this device has been already solved (21), therefore this paper focuses mainly on the reliability of the method.
Aim
The main aim of this study is to estimate the specific reliability of body composition measurement by the BIA Method (using TANITA 418 MA device), by means of its parallel form generated by multiple repetition of the given measurement. And, consequently, to assess the factual acceptability of errors for this measurement in monitored parameters of body composition.
Material and methods
The experimental sample group comprised of 91 students specializing in Physical Education and Sports at Pedagogical Faculty, University of Ostrava. These individuals, with regard to their specialization, can be considered as representatives of a healthy population. The total of tested students was 91 (53 men and 38 women). Average age was 20.5 years in men and 20.1 years in women. A tetrapolar bioimpedance weighing machine TANITA 418 MA was used for measurement of body fat ration by the BIA method. The measurement complied with all recommended rules. It was performed in a STANDARD regime because there was not detected 10 or more hours of weekly motion activities in any tested person (which is a condition for selecting an ATHLETIC regime).
Body composition is represented by values of body weight, body fat and total body water. Fat-free mass was not detected because its value was calculated as a difference of body weight and body fat.
The estimation of the specific reliability was based on the classical model of test theory (22), where a calculation of correlation coefficient between two parallel forms of measurement was used for the reliability estimation. We used a method of multiple repetition of the given measurement for creating of the parallel form. And we used the same theory for calculation of the test error.
When evaluating results of body fat measurement and total body water measurement the resulting values for three groups were assessed: 1st group consisted only of men, 2nd group consisted only of women and 3rd group consisted of all the tested persons regardless of their sex. It allowed to assess a diagnostic error of the method between groups with a different ratio of a fat fraction. First and second measurements were performed consecutively (within the range of 10 minutes) so that the result could not be affected by the external factors (e.g. water intake etc.).
The statistical computing was performed by means of PASW SPSS 18.0 statistical software. All the statistic tests were realized on a significance level α = 0.05.
The study protocol was approved by the Ethics and Research Committee of the University of Ostrava. All participants signed an informed consent form.
Results
a) Body fat measurement
Table 1 presents the results of the first and second measurements of body fat ratio in individual groups (test-retest), estimation of a reliability coefficient and of a test error (SE).
Table 1. Body fat.
| | Male (n=53) | Female (n=38) | Male + Female (n=91) |
| 1. (%) | 2. (%) | 1. (%) | 2. (%) | 1. (%) | 2. (%) |
| M | 12.52 | 12.16 | 22.02 | 22.27 | 16.49 | 16.38 |
| SD | 4.19 | 4.13 | 4.34 | 4.48 | 6.31 | 6.55 |
| rXX' | 0.971 | 0.979 | 0.987 |
| SE (%) | 0.12 | 0.09 | 0.08 |
n – frequency, M – mean, SD – standard deviation, 1. – first measurement, 2. – second measurement,
rXX' – Pearson correlation coefficient, SE – standard error of the test
The differences in mean values of body fat ratio between Measurement 1 and 2 were minimal for all selected groups – less than 0.5%. They oscillated in a range 0.11%-0.36%.
The highest estimated value of the reliability coefficient was detected for Group 3 (male + female) and the lowest for Group 1 (male). However, the detected differences are negligible. The same applies for standard errors of measurements (SE). The lowest was again detected for Group 3 (male + female) and the highest for Group 1 (male).
b) Total body water measurement (TBW)
Table 2 presents results of TBW Measurement 1 and 2 in individual groups and also the estimation of the reliability coefficient (rXX') performed by the selected procedure for TWB, including standard error of measurements (SE).
Table 2. Total body water.
| | Male (n=53) | Female (n=38) | Male + Female (n=91) |
| 1. (%) | 2. m. (%) | 1. m (%) | 2. m (%) | 1. m. (%) | 2. m (%) |
| M | 63.96 | 64.27 | 57.08 | 56.93 | 61.12 | 61.22 |
| SD | 3.05 | 3.06 | 3.18 | 3.29 | 4.62 | 4.79 |
| rXX' | 0.971 | 0.978 | 0.987 |
| SE (%) | 0.08 | 0.06 | 0.06 |
The differences in mean values of total body water ratio between Measurement 1 and 2 was, just as for body fat measurement, in all selected groups minimal and oscillated almost in the same range (0.31%-0.10%).
Also the estimations of the reliability coefficient (rXX') and of the TWB measurement errors practically do not differ from the results detected for body fat.
c) Body weight measurement
Table 3 presents results of Body Weight Measurement 1 and 2, the estimation of the reliability coefficient (rXX') and of the test error (SE).
Table 3. Body weight.
| Female (n=38) | Male + Female (n=91) |
| 1. m (kg) | 2. m (kg) | 1. m. (kg) | 2. m (kg) |
| M | 78.59 | 78.55 | 61.05 | 61.03 |
| SD | 9.70 | 9.67 | 8.83 | 8.80 |
| rXX' | 1.000 | 1.000 |
The mean values of body weight detected from the first and second measurement are almost the same in men and women. The differences are only 0.04 kg in men and 0.02 in women. Also the reliability of the measurement is almost perfect (see rXX' values). This almost error-free measurement is caused by the selected way of measuring. Body weight is measured directly, by weighing, in contrast to the other parameters (body fat, total body water) which are detected by means of the BIA method and a device software.
Discussion
The obtained values of reliability coefficients for all monitored parameters indicate a high specific reliability of the evaluated method – the lowest value is 0.971. The above mentioned measurement errors allow us to determine the approximate boundary error (dmax = 2SE) which is for all the monitored parameters lower than 0.20% (only for body fat in the male group it is 0.24%). To get an explorative answer to the question if the amount of body fat affects the estimated specific reliability we divided all the tested persons into 6 categories by the body fat ratio (regardless of sex). The values of body fat for each category oscillated in a range of 5%. This limitation is set by the treatises where the values of body fat are very often presented in a range of 2-5% for single sport branches (23, 24, 25, 26, 27). The obtained values of Pearson correlation coefficient are presented in Graph 1.
On the basis of this graph we can deduce whether the estimations of the reliability coefficients in our case show dependance on the body fat ratio. To verify the statistic significance we used a Spearmen correlation coefficient. Its value (0.371) was not statistically significant which confirms our presupposition made on the basis of the graph.
The obtained estimations of the reliability coefficient (rXX') are consistent with the results of other authors who presents their results in a range of 0.740-0.998, depending on the sex, age and employed methodology (28, 29, 30, 31). However, the direct comparison of our results with other authors is problematic because the authors used not only different bioimpedance devices and different sample groups, but also a different methodical procedure. Another significant difference can be the interval between input and output measurement. The treaties (17, 20) and the manual for Tanita 418 MA device present the different errors for measurements performed in one day (1-2%) and for measurements performed in the interval of several days (3.5%). This fact was apparent especially in our SE values and dmax which are in comparison to the above mentioned authors very low. The low values of SE and dmax calculated by our team result from the employed methodics (test-retest method repeated in immediate sequence). This procedure allowed us to determine the method´s reliability without influence of any external factors (among which belongs especially various level of hydratation). The presented results fully correspond with the results obtained at detection of the measurement error by means of typical (standard) error of measurement (17, 3).
Fig. 1. Pearson correlation coefficient value in dependance on body fat ratio.
Conclusion
With regard to the high values of the estimated reliability coefficient and to the low values of the standard test error we can the reliability of body composition measurement by the BIA method (using the above mentioned device) consider to be highly specifically reliable. We are able to assess the impact ratio of the external intervention (training burden, change of the nutrition habits, etc.) affecting the changes in body composition in the monitored individual even for relatively small changes of the monitored parameters, under the proviso that the measurement conditions are strictly met. At the BIA measurement it means especially to meet the principles relating to the organism hydratation, nutrition and motion activities. This study is limited by the employed sample group. Therefore the presented findings apply for the class of age ?adultus' with the regular motion activity.
Piśmiennictwo
1. Riegerová J, Přidalová M, Ulbrichová M: Aplikace fyzické antropologie v tělesné výchově a sportu (příručka funkční antropologie). Olomouc, Hanex 2006. 2. Malina RM., Bouchard C, Or OB: Growth, maturation, and physical aktivity. Champaign, IL, Human Kinetics 2004. 3. Kutáč P: Základy kinantropometrie: Ostrava, Pedagogická fakulta Ostravské univerzity v Ostravě 2009. 4. Sigmund M, Dostálová I: Somatické charakteristiky mladých hráčů ledního hokeje ve věku 15-18 let. Česká antropologie 1999; 49: 43-46. 5. Hoff J, Svendsen LH, Helgerud J: Lactate production and elimination in ice hockey players during an elite series match. Corpus, Psyche & Societas 2001; 8: 45-55. 6. Quinney HAet al.: A 26 year physiological description of a National Hockey League team. Applied Physiology, Nutrition & Metabolism 2008; 33: 753-760. 7. Neumayer G: Physical and Physiological Factors Associated with Success in Professional Alpine Skiing. International Journal of Sports Medicine 2003; 24: 571-575. 8. Bunc V: Body composition as a determining factor in the aerobic fitness and physical performance of czech childrens. Acta Gymnica 2006; 36: 39-45. 9. Zanovec M et al.: Self-Reported Physical Activity Improves Prediction of Body Fatness in Young Adults. Medicine & Science in Sports & Exercise 2009; 41: 328-335. 10. Nogueira JA, Nacedí da costa TH: Gender Differences in Physical Activity, Sedentary Behavior, and Their Relation to Body Composition in Active Brazilian Adolescents. Journal of Physical Activity & Health 2009; 6: 93-98. 11. Gil S, Ruiz F, Irazusta J: Selection of young soccer players in terms of antropometric and physiological factors. Journal of Sports Medicine & Physical Fitness 2007; 47: 25-32. 12. Bauer PW et al.: Cross Validation of Fat Free Mass Prediction Models For Elite Female Gymnasts. Pediatric Exercise Science 2005; 17: 337-344. 13. Stuelcken M, Pyne D, Sinclair P: Anthropometric characteristics of elite cricket fast bowlers. Journal of Sports Sciences 2007; 25: 1587-1597. 14. Sanchez MC, Sanz D, Zabala M: Anthropometric characteristics, body composition and somatotype of elite junior tennis players. Britsh Journal of Sports Medicine 2007; 41: 793-799. 15. Bunc V: Physiological and functional characteristic of adolescent athletes in several sports: implications for talent identification. In Silva, MC., Malina, R. (eds) Children and Sport in Organized Sports. Coimbra: I. da Universitade, 2004. 16. Wang Z: Human body composition models and methodology: Theory and experiment. Grafisch Service Centrum Van Gils, Wageningen 1997. 17. Heymsfield SB, Timothy GL, Wang Z, Scott G: Human Body Composition. Champaign, IL, Human Kinetics 2005. 18. McArdle WD, Katch FI., Katch VL: Exercise Physiology. Energy, Nutrition, & Human Performance. Philadelphia, Lippincott Williams & Wilkins 2007. 19. Bunc V: Estimation of body composition by multi frequency bioimpedance measurement in children. Annals of the New York Academy of Sciences, 2000; 904: 203-204. 20. Heyward VH, Wagner DR: Applied body composition assessment. Champaign, IL, Human Kinetics 2004. 21. Kutáč P et al.: Validity of measuring body composition by means of the BIA Metod. New Medicine 2008; 12: 89-93. 22. Měkota K, Blahuš P: Motorické testy v těkesné výchově: Praha, SPN 1983. 23. Soric M, Misigoj-Durakovic M, Pedisic Z: Dietary Intake and Body Composition of Prepubescent Female Aesthetic Athletes. International Journal of Sport Nutrition & Exercise Metabolism, 2008; 18: 343-354. 24. Norton KI, Craig NP, Withers RT, Whitingham NO: Assessing the body fat of athletes. Australian Journal of Science & Medicine in Sport 1994; 26: 6-13. 25. Bunc V, Psota R: Physiological profile of very young soccer players. Journal of Sports Medicine & Physical Fitness 2001; 41: 337-341. 26. Dostálová I, Přidalová M, Kudrna Z: Evalution of body constitution and body fractions of water polo players. Slovenská antropológia 2005; 8: 46-49. 27. Bandyopadhyay A: Body Composition, Hematological Profiles and Cardiorespiratory Fitness in Female Swimmers of West Bengal, India. International Journal of Applied Sports Sciences 2008; 20: 10-21. 28. Kilduff LP, Lewis S, Kingsley MI: Reliability and Detecting Change Following Short-Term Kreatine Supplementation: Comparasion of Two-Component Body Composition Methods. Journal of strength and reserch: the research journal of the NSCA, 2007; 21: 378-384. 29. Jackson AS, Pollock ML, Graves JE, Mahar MT: Reliability and validity of bioelectrical impedance in determining body composition. Journal of Applied Physiology 1988; 64: 529-534. 30. Kettanech A et al.: Reliability of bioimpedance analysis compared with other adiposity measurements in children: The FLVS II Study. Diabetes & Metabolism 2005; 31: 534-541. 31. Jesensky-Squires NE et al.: Validity and reliability of body composition analysers in children and adults. British Journal of Nutrition 2008; 100: 859-865.
