Review Article | Published:

# Bioelectrical impedance analysis for body composition assessment: reflections on accuracy, clinical utility, and standardisation

## Abstract

Bioelectrical impedance analysis is an extremely popular method for assessment of body composition. Despite its wide-spread use over the past thirty years, its accuracy and clinical value is still questioned. Most frequently, criticisms focus on its purported poor absolute accuracy and that different impedance analysers or prediction equations fail to measure body composition identically. This perspective review highlights that the magnitude of errors associated with impedance methods are not dissimilar to those observed for so-called gold standard methods. It is argued that the focus on statistically significant but small differences between methods can obscure operational equivalence and that such differences may be of minor clinical significance. Finally, the need for better standardization of protocols and the need for consensus on what is a minimal clinically important difference between methods is highlighted.

## Historical perspective and rise in popularity

Bioelectrical impedance analysis first came to prominence as a method for the analysis of body composition following the publication of Lukaski and colleagues seminal paper in 1985 [1] although the use of impedance to predict or estimate total body water (TBW) had been proposed some years before by Thomasset [2] and Hoffer et al. [3]. Bioimpedance analysis is based on the principle that the volume of a conductor (in the human body this is the highly conductive body water) is proportional to conductor length and inversely proportional to its electrical resistance as defined by

$${\mathrm{Volume}} = \rho \frac{{L^2}}{R}$$
(1)

where ρ is the resistivity (ohm cm) of the conductor, L is conductor length (cm, for whole body measurements in humans, stature is used as a surrogate for the unknown true conductive length), and R is the electrical resistance of the conductor (ohm). For more information on bioelectrical impedance theory the reader is referred to [4]. Development of the method between the work of Hoffer and Lukaski was hindered by the lack of suitable commercially available impedance devices. (Note that although body composition is predicted from measurements of electrical resistance, the method is referred to as an impedance technique. BIA devices typically measure electrical impedance, total opposition to flow of an alternating electrical current, and from this measurement derive resistance, opposition to current flow due to the inherent resistivity of body fluids). Lukaski’s paper reported, using a newly commercially available impedance device, highly significant correlations (r = 0.95–0.98) between the impedance quotient, H2/R, and fat-free mass (FFM), total body water (TBW), and body cell mass (BCM measured as total body potassium).

The intervening years since 1985 have seen an exponential rise in the popularity of the impedance technique as indicated by publications. This is illustrated in Fig. 1 which shows the numbers of publications per year since 1969 listed on PubMed using the search term “(Bioelectrical impedance) OR Bioimpedance AND (Body composition)” currently averaging approximately 350 publications per year. This period has also seen technological innovation, new analytical methods for body composition prediction and an increase in the number of companies manufacturing impedance devices such that the impedance technique is now a technological family (Fig. 2). In 1985, the device used by Lukaski et al. measured impedance at a single current frequency of 50 kHz (single frequency BIA, SFBIA); now devices are available that measure at multiple fixed frequencies (multi-frequency BIA, MFBIA) and over a range of frequencies (bioelectrical impedance spectroscopy, BIS). The BIS technique takes advantage of the frequency-dependent nature of current flow through the body, low frequency current flows only through the extracellular water space (ECW), while high frequency current, generally considered to be >50 kHz, flows through both extra- and intracellular water (ICW), to predict body water compartment volumes. Measurements may now be made across the whole body, from the wrist to the ankle, across body segments, i.e. the limbs and the trunk [5] or be highly focal measuring across a small body region such as the muscle bed of the calf [6]. Data analytical techniques have advanced from simple empirical predictive approaches based on statistical regression techniques [7] to biophysical-model-based methods for BIS [8] to index-based approaches in which impedance parameters per se are used as indices of body composition, e.g., impedance ratios [9].

Technological innovation does not fully explain the rise in popularity of impedance technology; other contributing factors include, relatively low-cost instrumentation, portability of devices, ease of use, non-invasive, and rapidity of measurement providing convenience to both the user and subject. Impedance-based assessment of body composition has found wide application in both research and clinical settings, including, nutritional management in organ (liver, heart, kidney) failure, nutritional management in, for example, sarcopenia, public health, obesity management, and as an aid to drug dosing [10, 11]. Despite these positive attributes, impedance technology would only have found wide adoption if it were reliable, precise and accurate.

### Impedance technology: questioning its accuracy for body composition assessment

Many publications attest to the reliability, precision and accuracy of impedance technologies [7, 10, 12]. A suite of statistical procedures that are typically used to evaluate performance have been widely adopted [13, 14]. Typically, performance is assessed on the basis of the magnitude of the correlation between body composition assessed by impedance against that measured by a reference method, such as dual-energy X-ray absorptiometry (DXA) or dilution-measured TBW, with agreement between methods assessed using limits of agreement analysis (LOA) [15]. Correlations greater than 0.95 are common for assessment of TBW or FFM with absolute errors being small (1–2%) although LOA may be large (±5–10%). A large number (more than 250 papers listed on PubMed between 1985 and 2018) of cross-validation studies have been undertaken in a variety of settings confirming the utility value of impedance technology, questions continue to be raised about the accuracy of BIA for body composition analysis; with comments such as “a systematic error in underestimating FFM” and “clinically significant errors occurred in FM and % body fat estimates” being common. It is instructive to ask why, in spite of the overwhelming body of evidence attesting to the value of impedance techniques, concerns remain. Three key issues are apparent:

• misunderstanding of, or an unwillingness to accept, the limitations of the impedance method and a desire for performance beyond these limits;

• distinction between statistically significant differences and clinically relevant differences and

• lack of standardization of impedance protocols.

### Limitations of the impedance technique

Quantitative estimation of body composition using bioimpedance technology is an indirect technique. BIA does not measure body composition; impedance devices measure an electrical response of the body, resistance, when exposed to an electric current. The measured resistance is then transformed into a prediction of TBW by an algorithm that incorporates in some form the relationship given in Eq 1. This transformation necessarily involves assumption of a value for the resistivity, which is either empirically determined as in BIS [8] or implicit in regression procedures used in SFBIA approaches for body composition assessment [13, 14]. Resistivity is not constant for all body tissues and fluid pools [16]. The value used in these algorithms is a whole body or segment average. It may also involve estimation of intermediate impedance parameters such as the estimation of resistance at zero and infinite frequency by extrapolation in BIS [17]. Further assumptions are invoked if body composition parameters other than TBW are required. Fat-free mass is calculated by assuming an hydration fraction for FFM, typically 0.73, for a healthy adult. Assumed values for apparent resistivities and hydration fractions are population mean values that are inserted into prediction algorithms but applied to impedance measurements in an individual. Inevitably, the inappropriate use of assumed coefficients and estimation of impedance parameters contribute to increasing prediction error (Fig. 3). Consequently, it must be recognized that impedance techniques are likely to be inherently less accurate than the reference methods against which they are cross-validated and unrealistic expectations tempered. Furthermore, if body fat is the measure of interest, this is most commonly determined by subtraction of predicted FFM from measured body weight. This, except in the case of exceedingly high %body fat, means subtracting a large quantity (FFM) from a somewhat larger quantity, body weight, to determine a proportionally smaller quantity (FM). Where %body fat is low, the error associated with FFM estimates, can lead to proportionally large errors in FM prediction.

### Effective performance: statistics versus clinical utility

As noted above, the common approach to assessing predictive power of impedance methods is to compare predictive methods using correlation and LOA analyses. Table 1 presents summary results of an unpublished study developing and validating an SFBIA prediction equation for TBW in fifty healthy 6–8 year-old Asian children. TBW measured by deuterium dilution was the reference method and single frequency whole body impedance was measured and a prediction equation developed. The predictive performance of this equation is compared to a number of published prediction equations for similarly aged children. The locally-raised prediction appears to be the most accurate, predicted data not being significantly different from measured values, exhibiting a high correlation, small bias and pure error. Limits of agreement are, however, relatively large at approximately ±10%. The equations of Davies and Gregory [18] and Nielsen et al [19]. perform particularly poorly. A number of points deserve consideration. Notably, the two equations of Davies and Gregory [18] and Nielsen et al. [19]. although derived in children of a similar age were Caucasian not Asian, highlighting the population specificity of prediction equations and the need to be vigilant in one’s choice of the most appropriate prediction equation. In this regard, prediction algorithms incorporated by a manufacturer into device-specific software and of unknown provenance should be treated with caution. The intrinsic equation and those of Horlick et al. [20], de Lorenzo et al. [21], and Wickramasinghe et al. [22] perform similarly although the latter exhibits a low concordance correlation. A reasonable conclusion would be to dismiss the suitability of all but the locally derived predictive equation. The preponderant approach is to focus on disagreement between methods rather than their equivalence.

Statistical testing of equivalence between methods has been customary in other settings [23,24,25] but has not, to date, found use in body composition research. Equivalence testing of the data in Table 1 shows that equivalence varies from 6.2–19.4% for the Horlick et al. [20] prediction equation through to that of Davies and Gregory [18]. Equivalence ratings for the Horlick et al. [20] and de Lorenzo et al. [21] equations are both within 2% of the value for the local equation despite having weaker concordance correlations, larger biases and wider LOA. Indeed, the Horlick et al. equation has a marginally smaller equivalence percentage.

It is generally considered that the BIS biophysical approach, which does not rely upon population-specific prediction equations, should exhibit greater accuracy and predictive performance than equation-based methods [26]. Seoane et al. [27] compared BIS with a number of popular empirically-based prediction equations and found that improvements were marginal. Using mean absolute percentage error as the overall measure of performance (MAPE), BIS models produced values of 4.6–4.9% with three of the four empirical equations not much larger, 6.3–7.9%. The comparable equivalence percentages were, BIS, 7.0–7.7% and Equations 10.0–11.8%.

An additional consideration often overlooked is the implicit assumption that a reference method that is used for comparison is totally accurate and error-free. Clearly, this is not the case. All experimental methods are subject to errors including, technical errors of measurement, analytical errors and both systematic and random errors. In a careful analysis of fluid volume estimations in hemodialysis patients by BIA, direct isotopic measurements (TBK) and dilution methods, Raimann et al. [28] concluded: “Bioimpedance can be of great help in clinical medicine for the monitoring of body fluid volumes and nutritional markers such as muscle mass and intracellular volume. The errors in precision and accuracy are evident, but are of comparable magnitude to the errors found between the measurements of so-called ‘gold-standard’ techniques”.

It could be argued that these small absolute differences between methods are clinically immaterial raising the issue of clinical relevance of a statistical difference. Researchers who overwhelmingly are those that develop prediction equations tend to focus on statistical significance whereas clinicians and clinical researchers may be more interest in what is considered “important” or “worthwhile” in practice, i.e., clinically important differences [29]. Unfortunately, these are not well defined, particularly in body composition research while in other disciplines this is well recognized and variably referred to as “minimal clinical important difference” (MCID), “clinically meaningful difference” (CMD) or “minimally important changes” (MIC) [30, 31]. There seems to be few attempts to adopt such an approach in the body composition field. One study where MCID was used is that of Warkentin et al. [32] who found that weight loss of >10% in the severely obese produces minimal clinically important differences cardiometabolic health. However, Rothberg [33], in a commentary on this study noted that “Whether this degree of weight loss produces meaningful improvements in health-related quality of life (HRQOL) is unclear.

### Standardization of methods

The importance of quality control and standardization of analytical methods has long been recognized in other branches of biomedical science, notably clinical chemistry and laboratory medicine where organizations such as the International Federation of Clinical Chemistry and Laboratory Medicine [34]. In 1994, a technology assessment conference on the then newly-emerging BIA technology was convened under the auspices of the National Institutes of Health; the proceedings being published in American Journal of Nutrition in 1996 [35]. The consensus statement [36] for the conference concluded “Therefore, the panel recommends that a committee of appropriate scientific experts and instruments manufacturers be formed with the goal of setting instruments standards and procedural methods”. Since that time, precious little progress has been made in achieving standardization of BIA technology [37, 38] or indeed body composition methods in general [39]. When reviewing reporting of factors critical to best practice in impedance measurements using lead-style devices in children, Brantlov et al. [37] found that none of 71 reviewed publications reported the hydration state of study participants while 37% reported steps taken to prepare the skin for attachment of electrodes and only 25% provided details of steps taken to minimize participant movements during measurement, a particular problem with young children. It is clear that greater effort must be made to develop standard operating procedures for impedance measurements. The variety of measurement devices and protocols (Fig. 2), however, makes this challenging.

## Concluding remarks

BIA offers many advantages for body composition assessment in health and disease [40] but its very simplicity can lead to uncritical application and disappointment that results are not as accurate as may be desired. Improvements in accuracy may be possible but gains made over the past 30 years have been marginal. A fundamental problem is that BIA is a predictive method that inherently requires simplifications and assumptions based on population mean values yet considered as being accurately applicable to all subjects. Improved standardization of protocols for measurement is essential. Acceptance of the limitations of the method is required, unrealistic expectations of precision and accuracy tempered and clinically acceptable limits of accuracy need to be defined. BIA technology is here to stay and its future is assured but may lie in real-time physiological monitoring rather than “snapshots” of body composition at discrete points in time [41, 42].

## References

1. 1.

Lukaski HC, Johnson PE, Bolonchuk WW, Lykken GI. Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am J Clin Nutr. 1985;41:810–7.

2. 2.

Thomasset M. Bioelectric properties of tissue. Impedance measurement in clinical medicine. Significance of curves obtained. Lyon Med. 1962;94:107–18.

3. 3.

Hoffer EC, Meador CK, Simpson DC. Correlation of whole-body impedance with total body water volume. J Appl Physiol. 1969;27:531–4.

4. 4.

Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, et al. Bioelectrical impedance analysis—Part I: review of principles and methods. Clin Nutr. 2004;23:1226–43.

5. 5.

Ward LC. Segmental bioelectrical impedance analysis: an update. Curr Opin Clin Nutr Metab Care. 2012;15:424–9.

6. 6.

Zhu F, Leonard EF, Levin NW. Body composition modeling in the calf using an equivalent circuit model of multi-frequency bioimpedance analysis body composition modeling in the calf using an equivalent circuit model of multi-frequency bioimpedance analysis. Physiol Meas. 2005;26:s133–43.

7. 7.

Mulasi U, Kuchnia AJ, Cole AJ, Earthman CP. Bioimpedance at the bedside: current applications, limitations, and opportunities. Nutr Clin Pract. 2015;30:180–93.

8. 8.

Ward LC, Isenring E, Dyer JM, Kagawa M, Essex T. Resistivity coefficients for body composition analysis using bioimpedance spectroscopy: effects of body dominance and mixture theory algorithm. Physiol Meas. 2015;36:1529–49.

9. 9.

Lukaski HC, Kyle UG, Kondrup J. Assessment of adult malnutrition and prognosis with bioelectrical impedance analysis: phase angle and impedance ratio. Curr Opin Clin Nutr Metab Care. 2017;20:330–9.

10. 10.

Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, et al. Bioelectrical impedance analysis-part II: utilization in clinical practice. Clin Nutr. 2004;23:1430–53.

11. 11.

Earthman CP. Body composition tools for assessment of adult malnutrition at the bedside. J Parenter Enter Nutr. 2015;39:787–822.

12. 12.

Sergi G, De Rui M, Stubbs B, Veronese N, Manzato E. Measurement of lean body mass using bioelectrical impedance analysis: a consideration of the pros and cons. Aging Clin Exp Res. 2017;29:591–7.

13. 13.

Guo SS, Chumlea WC, Cockram DB. Use of statistical methods to estimate body composition. Am J Clin Nutr. 1996;64:428S–435S.

14. 14.

Tronstad C, Pripp AH. Statistical methods for bioimpedance analysis. J Electr Bioimpedance. 2014;5:14–27.

15. 15.

Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;327:307–10.

16. 16.

Geddes L, Baker LE. The specific resistance of biological material—a compendium of data for the biomedical engineer and physiologist. Med Biol Eng. 1967;5:271–93.

17. 17.

Cornish BH, Thomas BJ, Ward LC. Improved prediction of extracellular and total body water using impedance loci generated by multiple frequency bioelectrical impedance analysis. Phys Med Biol. 1993;38:337.

18. 18.

Davies P, Gregory J. Body water measurements in growth disorders. Arch Dis Child. 1991;66:1467.

19. 19.

Nielsen BM, Dencker M, Ward L, Linden C, Thorsson O, Karlsson MK, et al. Prediction of fat-free body mass from bioelectrical impedance among 9- to 11-year-old Swedish children. Diabetes Obes Metab. 2007;9:521–39.

20. 20.

Horlick M, Arpadi SM, Bethel J, Wang J, Moye J Jr, et al. Bioelectrical impedance analysis models for prediction of total body water and fat-free mass in healthy and HIV-infected children and adolescents. Am J Clin Nutr. 2002;76:991–9.

21. 21.

De Lorenzo A, Di Campli C, Andreoli A, Sasso GF, Bonamico M, Gasbarrini A. Assessment of body composition by bioelectrical impedance in adolescent patients with celiac disease. Am J Gastroenterol. 1999;94:2951–5.

22. 22.

Wickramasinghe VP, Lamabadusuriya SP, Cleghorn GJ, Davies PS. Assessment of body composition in Sri Lankan children: validation of a bioelectrical impedance prediction equation. Eur J Clin Nutr. 2008;62:1170–7.

23. 23.

Lakens D. Equivalence tests: a practical primer for t-tests, correlations, and meta-analyses. Soc Psychol Personal Sci. 2017;8:355–62.

24. 24.

Richter SJ, Richter C. A method for determining equivalence in industrial applications. Qual Eng. 2002;14:375–80.

25. 25.

Dixon PM, Saint-Maurice PF, Kim Y, Hibbing P, Bai Y, Welk GJ. A primer on the use of equivalence testing for evaluating measurement agreement. Med Sci Sports Exerc. 2018;50:837–45.

26. 26.

Matthie JR. Bioimpedance measurements of human body composition: critical analysis and outlook. Expert Rev Med Devices. 2008;5:239–61.

27. 27.

Seoane F, Abtahi S, Abtahi F, Ellegård L, Johannsson G, Bosaeus L, et al. Mean expected error in prediction of total body water. A true accuracy comparison between bioimpedance spectroscopy and single frequency regression equations. Biomed Res Int. 2015; 2015:656323.

28. 28.

Raimann JG, Zhu F, Wang J, Thijssen S, Kuhlmann MK, Kotanko P, et al. Comparison of fluid volume estimates in chronic hemodialysis patients by bioimpedance, direct isotopic, and dilution methods. Kidney Int. 2013;85:1–11.

29. 29.

Page P. Beyond statistical significance: clinical interpretation of rehabilitation research literature. Int J Sports Phys Ther. 2014;9:726–36.

30. 30.

Copay AG, Subach BR, Glassman SD, Polly DW Jr, Schuler TC. Understanding the minimum clinically important difference: a review of concepts and methods. Spine J. 2007;7:541–6.

31. 31.

Jaeschke R, Guyatt G, Sackett D. Users’ guides to the medical literature diagnostic test. JAMA. 1994;271:703–7.

32. 32.

Warkentin LM, Majumdar SR, Johnson JA, Agborsangaya CB, Rueda-Clausen CF, Sharma AM, et al. Weight loss required by the severely obese to achieve clinically important differences in health-related quality of life: two-year prospective cohort study. BMC Med. 2014;12:175.

33. 33.

Rothberg AE. Weight loss ≥ 10% is required by the severely obese to achieve minimal clinically important differences in health-related quality of life. Evid Based Med. 2015;20:69.

34. 34.

Federation TI, Chemistry C, Group W, et al. International Federation of Clinical Chemistry (IFCC). J Clin Chem 2002;2001:1–5.

35. 35.

Yanovski S, Hubbard V, Heymsfield S, Lukaski HC. Bioelectrical impedance analysis in body composition measurement. NIH Technology Assessment Statement. Am J Clin Nutr. 1996;64:387S–532S.

36. 36.

National Institutes of Health. Bioelectrical impedance analysis in body composition measurement: National Institutes of Health Technology Assessment Conference Statement. Am J Clin Nutr. 1996;64:524s–532s.

37. 37.

Brantlov S, Jødal L, Lange A, Rittig S, Ward LC. Standardisation of bioelectrical impedance analysis for the estimation of body composition in healthy paediatric populations: a systematic review. J Med Eng Technol. 2017;41:1–20.

38. 38.

Brantlov S, Ward LCLC, Jødal L, Rittig S, Lange A. Critical factors and their impact on bioelectrical impedance analysis in children: a review. J Med Eng Technol. 2017;41:22–35.

39. 39.

Wootton S, Durkin K, Jackson A. Quality control issues related to assessment of body composition. Food Nutr Bull. 2014;35:S79–85.

40. 40.

Jackson AA, Johnson M, Durkin K. Wootton s. Body composition assessment in nutrition research: value of BIA technology. Eur J Clin Nutr. 2013;67:S71–8.

41. 41.

Villa F, Magnani A, Merati G, Castiglioni P. Feasibility of long-term monitoring of multifrequency and multisegment body impedance by portable devices. IEEE Trans Biomed Eng. 2014;61:1877–86.

42. 42.

Asogwa C, Lai D. A review on opportunities to assess hydration in wireless body area networks. Electronics. 2017;6:1–16.

## Author information

### Affiliations

1. #### School of Chemistry and Molecular Biosciences, The University of Queensland, St Lucia, QLD, Australia

• Leigh C. Ward

### Conflict of interest

The author provides consultancy services to ImpediMed Ltd. ImpediMed Ltd. had no involvement in the inception or execution of this manuscript.

### Corresponding author

Correspondence to Leigh C. Ward.