Secondary Logo

Journal Logo

Original Investigation

Bioelectrical Impedance for Accuracy Detecting Body Composition Changes during an Activity Intervention

Schneider, Patrick L.1; Bassett, David R. Jr2; Thompson, Dixie L.2; Crouter, Scott E.2

Author Information
Translational Journal of the ACSM: October 1, 2017 - Volume 2 - Issue 19 - p 122-128
doi: 10.1249/TJX.0000000000000041
  • Free



It is well established that obesity is a major public health problem in the United States and other developed countries. Current data indicate that more than one-third (34.9% or 78.6 million) of U.S. adults are obese (15). The risks associated with obesity include but are not limited to type 2 diabetes mellitus, dyslipidemia, respiratory dysfunction, certain types of cancer, and hypertension (16). Given these risks, accurate assessment of body composition is an important component of a health risk appraisal. In addition, although there are many types of weight loss programs available, the most appropriate goal for an overweight or obese individual should be to lose fat mass (FM) while preserving fat-free mass (FFM). Therefore, accurate tracking of body composition changes by researchers, those in the health and fitness industry, and individual consumers is vital to determine the success of weight loss programs. In addition, the popularity of smart phones and tablets with apps to track behavioral and physiological measures further validates the need for the accurate tracking of these categorical measures. For example, products such as the Fitbit Aria, Withings Body Cardio, and the Garmin Index Smart Scale are body composition analyzers that can be synced to smart phones or tablets to allow for the tracking of body composition values over time. Of particular importance is the ability of body composition analyzers to properly track gains or losses in body fat in response to a diet and/or exercise program.

Some of the more common methods for assessing body composition are hydrostatic weighing (HW), air displacement plethysmography (ADP), bioelectrical impedance analysis (BIA), dual-energy x-ray absorptiometry (DXA), and anthropometry. HW is generally considered to be the “gold standard” for body composition assessment. However, other more efficient and less expensive methods have been shown to have similar accuracy compared with HW in cross-sectional comparisons of healthy adults (2,18). In particular, ADP has compared favorably with HW in previous studies (2,11,13,18) and can serve as a criterion method, suitable for validating other methods that are efficient and less expensive.

The BOD POD (ADP) (Life Measurement, Inc., Concord, CA) is a more sophisticated method of body composition assessment that is commonly used in research settings. It has demonstrated excellent validity and reliability when compared with HW and DXA in adults (10,14). It has also demonstrated effectiveness at tracking body composition changes during periods of weight loss (23) and weight gain (9). In addition, based on the scientific evidence available, a recent publication deemed the BOD POD the best instrument available for tracking body composition from infancy into adulthood (3).

BIA is a quick, easy, noninvasive means of assessing body composition by passing an electrical current through the body and measuring the resistance (or impedance) to current flow. It has become popular in wellness screenings, weight loss clinics, fitness facilities, and similar settings over the past couple of decades. In general, cross-sectional studies involving normal weight individuals show that while BIA tends to underestimate percent body fat (%BF), the values do not differ significantly from those of other previously validated methods such as HW, ADP, and DXA (2,13). However, at the extremes of adiposity, estimates of body composition by BIA may be less accurate, overestimating %BF in the lean and underestimating it in the obese (1,17).

Although some studies have compared various methods of body composition assessment in diverse populations (4,8,12), none to our knowledge have compared the accuracy of BIA in tracking changes in %BF in response to a physical activity intervention in overweight men and women. This could be important because physical activity may result in losses in FM, while at the same time increasing FFM. Thus, the purpose of this study was to assess the accuracy of the Tanita TBF-305 body fat analyzer in detecting body composition changes in comparison with ADP throughout a 36-wk, 10,000 steps per day intervention in previously sedentary, overweight men and women.



Fifty-six sedentary, overweight or obese, middle-aged men and women between 30 and 55 yr of age volunteered to participate in the study. Participants were not dieting for weight loss and were relatively weight stable (within 4% of their current body mass) for the 3 months before the intervention. To be included in the study, participants had to meet the following criteria: age between 30 and 60 yr, body mass index between 25 and 45 kg·m−2, ability to follow instructions and record data, and ability to walk 1 mile without pain or discomfort. The exclusion criteria were as follows: history of myocardial infarction, angina, stroke, heart failure, or uncontrolled cardiac arrhythmias; resting blood pressure greater than 180 mm Hg systolic and/or 100 mm Hg diastolic; current cigarette smoking; exceeding 7300 steps per day or 90 min of moderate intensity activity per week; and currently taking weight loss medication or prescription medication that might impair physical performance or affect metabolism. Before their participation, the study was verbally explained, and each participant read and signed an informed consent form that had been approved by the University of Tennessee and the University of Tennessee Medical Center Institutional Review Boards. A health history questionnaire was completed by all participants before the study to screen for any contraindications to exercise.

Study Design

This study was part of a longitudinal study using a repeated-measures experimental design that has been reported previously (19). The intervention lasted 36 wk. The 2 wk before the intervention served as a baseline period, and data from this phase were used to screen participants to verify that they had low levels of physical activity (≤7300 steps per day). Participants were instructed to avoid changing their normal activities during the 2-wk baseline period. All participants were given a waist-mounted Digi-Walker SW-200 electronic pedometer (Yamax, Inc., Tokyo, Japan) along with instructions on the proper use of the pedometer (20). The participants were instructed to wear the pedometer each day during all waking hours, except when showering or swimming. All participants were asked to record the number of steps taken per day along with the time the pedometer was put on and the time it was taken off.

After the baseline period, preliminary measures were obtained. At this point, all participants were given a physical activity prescription that led up to a goal of accumulating at least 10,000 steps per day. The first week they were encouraged to take 7000 steps per day, the second week 8000 steps per day, the third week 9000 steps per day, and by the fourth week they were instructed to take at least 10,000 steps per day. Participants also completed a 3-d dietary record at baseline, 20 wk, and 36 wk. The dietary record was explained to each participant in detail, including how to properly describe the food item, the amount eaten, and how it was prepared. All items consumed within the 3-d period were recorded by the participant and analyzed using Nutritionist Pro software (First Data Bank Inc., San Bruno, CA).

Body Composition Measures

Body composition was assessed using ADP (BOD POD® Body Composition System, Life Measurement, Inc.) and BIA (Tanita TBF-305; Tanita Corporation of America, Inc., Skokie, IL) at baseline, at 20 wk, and at 36 wk with ADP serving as the criterion measure. Every attempt was made to measure body weight and body composition variables at approximately the same time of day (between 6:00 and 10:00 a.m.) at each testing time point. All measurements were made after an overnight fast. In addition, participants were asked to refrain from exercise for 12 h before the measurements were made, to consume water in their typical fashion, and to empty their bladder immediately before testing.

ADP determines body volume, which is then used to calculate body density. The Siri equation (21) was used to convert body density to %BF. The participants wore a Lycra/nylon swimsuit and swim cap during testing. Body volume was corrected for thoracic gas volume using a prediction equation based on age, sex, and height (11).

BIA is based on the principle that lean tissue, which contains large amounts of water and electrolytes, conducts current more readily than fat, which is anhydrous (22). The Tanita TBF-305 uses a foot-to-foot, pressure contact electrode system in which a small current is sent through the anterior portion of the footpad electrodes, and the voltage drop is measured in the posterior portion of the footpad electrodes. This method estimates body composition by measuring the resistance or impedance to flow of this electrical current. The impedance value is inserted into a proprietary equation along with body mass, height, and sex to estiate %BF, FFM, and total body water.

Data Analysis

All statistical analyses were performed using SPSS 22 for Windows (SPSS Inc., Chicago, IL). For all analyses, an alpha of 0.05 was used to denote statistical significance. Paired samples t-tests were used to compare the baseline characteristics between sexes. A repeated-measures ANOVA (measurement time–sex) was used to compare dietary intake measures.

For the body composition variables, a three-way repeated-measures ANOVA (measurement time–device–sex), controlling for baseline body mass index, was initially performed to compare the changes in body composition measures (body mass, %BF, FM, and FFM). Subsequently, a two-way repeated-measures ANOVA (measurement period–device) was used to compare changes in the body composition measures for each sex separately. Huynh–Feldt corrections were used whenever the sphericity assumption was violated. Bonferroni adjustments were used for multiple within-group comparisons. Pearson correlation coefficients (r) were used to explore the association between body composition measures from the two devices between sexes. Paired t-tests were used to compare body composition measures between ADP and BIA, for each sex separately, at each measurement time point.

Modified Bland–Altman plots were constructed to show the distribution of the individual scores around zero at each measurement time point for the difference in the individual changes in %BF from the beginning to the end of the intervention (ADP change in %BF minus BIA change in %BF) and also for differences in %BF (ADP %BF minus BIA %BF). In this manner, the mean difference (ADP minus BIA) can be illustrated, and the 95% prediction interval (PI) (i.e., 95% confidence interval for the individual observations) can also be shown. Individual error scores that have a tight PI around zero signify a more accurate device. Data points below zero signify an overestimation by BIA compared with ADP, whereas points above zero signify an underestimation.


Fifty-six participants (19 men, 37 women) met all study criteria and started the intervention, and 36 (14 men, 22 women) completed the study. The baseline characteristics of those who completed the 36-wk intervention are presented in Table 1. There was no significant interaction between dietary measures and sex, nor was there a significant main effect for time, indicating that both groups showed unchanging dietary intake measures over the course of the intervention. Dietary intake among men were 2498 kcal·d−1 (baseline), 2315 kcal·d−1 (20 wk), and 2711 kcal·d−1 (36 wk). For women, it was 2243 kcal·d−1 (baseline), 2133 kcal·d−1 (20 wk), and 2021 kcal·d−1 (36 wk).

Baseline Characteristics of the Participants that Completed the 36-wk Intervention.

The %BF values were significantly correlated between the two measurement devices at each time point (P ≤ 0.001), although the correlations were stronger in men. For men and women, the correlations (r) between ADP and BIA %BF values were 0.904 and 0.639, respectively, at baseline; 0.896 and 0.708, respectively, at 20 wk; and 0.850 and 0.716, respectively, at 36 wk.

There was a significant three-way interaction for %BF between time (baseline, 20 wk, and 36 wk measurements), device (ADP vs BIA), and sex (P = 0.019). Separate analyses of men and women showed a significant time–device interaction in women (P < 0.001), but not men (P = 0.656). Table 2 shows the body composition measures (body mass, %BF, FM, and FFM) for each device at each time point (baseline, 20 wk, and 36 wk), and Figure 1 shows the change in %BF for each device and sex across the measurement time points. Among women, BIA %BF estimates increased 0.5% and 0.3% from baseline to 20 wk and from baseline to 36 wk, respectively. By contrast, ADP %BF estimates decreased 1.9% and 1.5% from baseline to 20 wk and from baseline to 36 wk, respectively. Among men, BIA %BF estimates decreased 1.5% and 2.3% from baseline to 20 wk and from baseline to 36 wk, respectively, whereas ADP showed similar decreases of 1.9% and 2.8% over the same periods.

Mean ± SD Body Composition Variables Estimated by ADP and BIA at Baseline, 20 wk, and 36 wk in those that Completed the Intervention (Males, n = 14; Females, n = 22).
Figure 1:
Comparison of estimated changes in percent body fat by ADP and BIA over the course of the 36-wk intervention in men (A) (n = 14) and women (B) (n = 22).

Among men, BIA-estimated %BF was, on average, 1.4%, 1.0%, and 0.9% lower than ADP estimates at baseline, 20 wk, and 36 wk, respectively (P > 0.05). For women, there was a significant interaction between the ADP and the BIA %BF estimates across time (P < 0.001). Among women, BIA-estimated %BF was significantly lower than ADP by 3.0% at baseline (P = 0.003) but was within 0.6% and 1.2% at 20 and 36 wk, respectively (P > 0.05).

Figure 2 shows the modified Bland–Altman plots that compare the difference in the individual changes in body fat between the two devices from the baseline to 36 wk. Although the mean bias was lower for males than females (0.5% vs 1.8%), the 95% PI values were higher for males than females (−5.2% to 6.2% vs −2.5% to 6.0%). Figure 3 shows the modified Bland–Altman plots at baseline, 20 wk, and 36 wk. Although the mean bias was similar for males (range, 0.9% to 1.4%) and females (range, 0.6% to 3.0%), the 95% PIs were lower for males (95% PI range, −6.6% to 8.4%) compared with females (95% PI range, −7.4% to 11.1%) at each time point.

Figure 2:
Modified Bland–Altman plots for the change in percent body fat (%BF) estimated by ADP and BIA for males from baseline to 36 wk (A) and females from baseline to 36 wk (B). Solid horizontal line, mean error score; dashed lines, 95% PI (i.e., 95% confidence intervals of the individual observations).
Figure 3:
Modified Bland–Altman plots for percent body fat (%BF) estimated by ADP and BIA for males at baseline (A), females at baseline (B), males at 20 wk (C), females at 20 wk (D), males at 36 wk (E), and females at 36 wk (F). Solid horizontal line, mean bias; dashed lines, 95% PI (i.e., 95% confidence intervals of the individual observations).

Figure 4 shows the relationship between body composition changes from baseline to 36 wk for BIA and ADP. According to this figure, 86% of males and 45% of females were accurately classified with respect to the direction of %BF change from baseline to 36 wk (both devices showed increases or both showed decreases in %BF).

Figure 4:
A comparison of body composition changes from baseline to 36 wk. Open symbols represent %BF measurements that went in the same direction (both devices showed increases or both showed decreases), and closed symbols represent measurements that went in opposite directions (one device showed an increase and the other decreased or remained unchanged, or vice versa).


Results from the current study showed that BIA was effective at tracking body composition changes on a group level in men, but not women. Among women, while ADP was showing decreases in %BF values from baseline to 20 wk, BIA was showing increases during this same period. Likewise, from 20 to 36 wk, ADP showed a subtle increase in %BF whereas BIA showed a small decrease. Thus, from the baseline to 36 wk, ADP showed a decrease in %BF whereas BIA showed an increase over the same period. Among men, BIA tracked changes similarly with ADP from baseline to 20 wk and from 20 to 36 wk. However, as shown in Figure 2, which compared the difference in the individual changes in %BF between devices from baseline to 36 wk, there was a lot of variation in BIA's accuracy in tracking body composition changes at the individual level in both sexes.

Although numerous studies have compared various body composition assessment devices against one another in cross-sectional studies, few have done so in longitudinal studies. Similar to the current study, Miyatake et al. (12) showed that BIA was unable to accurately assess body composition changes among Japanese women in response to exercise-induced weight loss. In their study, changes in %BF, FM, and FFM between BIA and ADP were poorly correlated (r = 0.386, r = 0.556, and r = 0.215, respectively). Frisard et al. (4) compared the ability of ADP, BIA, and DXA (criterion method) to assess body composition changes in response to a commercial weight loss intervention (Weight Watchers®) in men and women and found that correlations for %BF and FM between BIA and DXA were relatively low before weight loss (r2 = 0.61 and 0.35, respectively) and improved after weight loss (r2 = 0.76 and 0.83, respectively). The regression coefficients for ADP were consistently high when compared with DXA before and after weight loss (r2 > 0.80). However, their study only measured %BF values at two points (pre- and postweight loss) and did not examine the ability of these methods to detect changes in %BF. Furthermore, they did not distinguish between sexes making comparisons with the current study difficult. In contrast to our study, Jebb et al. (8) showed that among overweight and obese women, leg-to-leg bioelectrical impedance, a BIA device comparable with that used in the current study, tracked body fat changes similarly to ADP and DXA over a 12-wk weight loss intervention and a subsequent 1-yr follow-up, which involved weight regain. BIA underestimated FM changes compared with ADP during the 12-wk period of weight loss and the 1-yr follow-up, a trend that was similar to our study.

The results of the current study showed that BIA-estimated %BF consistently tracked changes on a group level when compared with ADP in men. However, consistency in the measurement of %BF between the two methods was lacking in women. There are a few of potential explanations for this variation among sexes. First, as expected, the women had significantly higher %BF values than the men. Previously, it was pointed out that BIA may be less accurate at the extremes of adiposity. Thus, the combination of these two factors may have contributed to the lack of agreement between the two measures (BIA and ADP) among females but not males. Second, the women in this study experienced greater changes in FFM than the men according to ADP. Previous studies have indicated that there is large variation in the water content of FFM (7), and given the dependence of BIA on hydration status, increases in FFM may affect the estimation of %BF. In the current study, women increased FFM while men remained essentially unchanged. Third, variations in hydration status could affect the estimation of %BF by BIA. Thus, water retention due to menstrual cycle phases could have affected the ability of BIA to accurately track body composition changes. Previous data showed that water retention may cause changes in FFM during different phases of the menstrual cycle (5). However, the effect of menstrual cycles on hydration status and the subsequent effect on estimates of body composition using BIA remain unclear. Although Gleichauf et al. (5) showed small, but statistically significant changes in body weight, FFM, and impedance related to hydration status during different phases of menses (despite no change in %BF), Gualdi-Russo and Toselli (6) found no significant differences in impedance during the menstrual cycle. Future research projects of this nature should consider standardizing the time at which %BF is measured as it pertains to a female's menstrual cycle.

Although BIA accurately tracked changes on a group level in men but not women, Figure 3, which compared the two devices at baseline, 20 wk, and 36 wk, shows that both devices performed similarly on an individual basis although the mean bias and 95% PI values were slightly smaller in men than women. Given that the 95% PI ranged from −6.6% to 8.4% in males and from −7.4% to 11.1% in females, the use of BIA to track individual changes in %BF should be done with caution.

A limitation of the current study may be the use of ADP as the criterion measure. Although HW and DXA are generally considered to be among the most accurate instruments for estimating body composition, studies have shown that ADP has compared favorably with HW and DXA in children and adults (2,10,11,13,14,18). In addition, it is helpful to observe how a less expensive, noninvasive measure of %BF such as BIA compares with an expensive, more established method such as ADP.

The implications of this research extend to the research community, those in the health and fitness industry, and the individual consumer. For those conducting research that involves tracking body composition changes among groups in response to an intervention, it appears that this particular type of BIA analyzer may have some use for tracking changes in groups of men, but not women. It is important to emphasize, however, that the limitations of estimating body composition using BIA devices leaves them less than ideal as a criterion measurement technique. As for those in the fitness industry who are more likely to deal with participants on an individual basis, caution should be exercised when using BIA to track body composition changes in either sex. Individual users of this device must recognize that the day-to-day variability in measurement can lead to incorrect conclusions about changes in body composition. Thus, other tools such as circumference measurements (e.g., waist circumference) and total body mass should be used as additional indicators of change. This research has important implications given the potential for misrepresenting the %BF change in response to an exercise-induced weight loss intervention. According to BIA, the %BF values for women as a group in this study show virtually no change, whereas ADP showed a significant reduction in %BF after 20 wk. In fact, BIA showed a mean increase of 0.5% in %BF after the first 20 wk of the intervention, whereas ADP showed a mean decrease of 1.9% decrease among women. On an individual level, BIA was more likely to misclassify the direction of the change in body composition in women compared with men (55% of women were misclassified vs 14% of men). From a behavioral change perspective, this type of discrepancy between methods at the group and individual level could discourage women from adhering to an exercise program, if they were to be told their %BF increased in response to an exercise program when it actually decreased.


These results suggest that BIA is an acceptable method for tracking %BF changes on a group level in overweight and obese men, throughout a physical activity intervention. However, there was greater variation in the accuracy of BIA in quantifying change in %BF among women as a group when compared with ADP. On an individual basis, there were large variations in BIA's accuracy in tracking changes over the course of the intervention as well as its accuracy in estimating %BF at any given time point in both men and women. Thus, researchers, health and fitness professionals, and consumers should be cautious when using BIA to track body composition on an individual basis. Although potential explanations for these results have been suggested, more research is necessary to determine which factors contribute to these differences in the accuracy of BIA in assessing %BF changes between sexes. Lastly, future research should examine multifrequency BIA analyzers to track body composition changes in comparison with an acceptable criterion method.

There were no funding sources for this study. The results of the present study do not constitute endorsement by the American College of Sports Medicine.


1. Bedogni G, Agosti F, De Col A, Marazzi N, Tagliaferri A, Sartorio A. Comparison of dual-energy X-ray absorptiometry, air displacement plethysmography and bioelectrical impedance analysis for the assessment of body composition in morbidly obese women. Eur J Clin Nutr. 2013;67:1129–32.
2. Biaggi RR, Vollman MW, Nies MA, et al. Comparison of air-displacement plethysmography with hydrostatic weighing and bioelectrical impedance analysis for the assessment of body composition in healthy adults. Am J Clin Nutr. 1999;69:898–903.
3. Fields DA, Gunatilake R, Kalaitzoglou E. Air displacement plethysmography: cradle to grave. Nutr Clin Pract. 2015;30:219–26.
4. Frisard MI, Greenway FL, Delany JP. Comparison of methods to assess body composition changes during a period of weight loss. Obes Res. 2005;13:845–54.
5. Gliechauf CN, Roe DA. The menstrual cycle's effect on the reliability of bioimpedance measurements for assessing body composition. Am J Clin Nutr. 1989;50:903–7.
6. Gualdi-Russo E, Toselli S. Influence of various factors on the measurement of multifrequency bioimpedance. Homo. 2002;53:1–16.
7. Hewett MJ, Going SB, Williams DP, Lohman TG. Hydration of the fat-free body mass in children and adults: implications for body composition assessment. Am J Physiol. 1993;265:E88–95.
8. Jebb SA, Siervo M, Murgatroyd PR, Evans S, Frühbeck G, Prentice AM. Validity of the leg-to-leg bioimpedance to estimate changes in body fat during weight loss and regain in overweight women: a comparison with multi-compartment models. Int J Obes (Lond). 2007;31:756–62.
9. Kilduff LP, Lewis S, Kingsley MI, Owen NJ, Dietzig RE. Reliability and detecting change following short-term creatine supplementation: comparison of two-component body composition methods. J Strength Cond Res. 2007;21:378–84.
10. McCrory MA, Gomes TD, Bernauer EM, Molé PA. Evaluation of a new air displacement plethysmograph for measuring human body composition. Med Sci Sports Exerc. 1995;27:1686–91.
11. McCrory MA, Molé PA, Gomez TD, Dewey KG, Bernauer EM. Body composition by air-displacement plethysmography by using predicted and measured thoracic gas volumes. J Appl Physiol (1985). 1998;84:1475–9.
12. Miyatake N, Takenami S, Kawasaki Y, Fujii M. Comparison of air displacement plethysmograph and bioelectrical impedance for assessing body composition changes during weight loss in Japanese women. Diabetes Obes Metab. 2005;7:268–72.
13. Moon JR, Tobkin SE, Smith AE, et al. Percent body fat estimations in college men using field and laboratory methods: a three-compartment model approach. Dynamic Medicine [Internet]. 2008. [cited 2017 March 10];7(7). Available from: doi:10.1 186/1476-5918-7-7.
14. Nunez C, Kovera AJ, Pietrobelli A, et al. Body composition in children and adults by air displacement plethysmography. Eur J Clin Nutr. 1999;53:382–7.
15. Ogden CL, Carroll MD, Kit BK, Flegal KM. Prevalence of childhood and adult obesity in the United States, 2011–2012. JAMA. 2014;311:806–14.
16. Pi-Sunyer FX. Medical hazards of obesity. Ann Intern Med. 1993;119:655–60.
17. Ratamess N. Body composition status and assessment. In: Swain DP, editor. ACSM's Resource Manual for Guidelines for Exercise Testing and Prescription. 7th ed. Baltimore (MD): Lippincott Williams & Wilkins; 2014, p. 296.
18. Sardinha LB, Lohman TG, Teixeira PJ, Guedes DP, Going SB. Comparison of air displacement plethysmography with dual-energy X-ray absorptiometry and 3 field methods for estimating body composition in middle-aged men. Am J Clin Nutr. 1998;68:786–93.
19. Schneider PL, Bassett DR Jr, Thompson DL, Pronk NP, Bielak KM. Effects of a 10,000 steps per day goal in overweight adults. Am J Health Promot. 2006;21:85–9.
20. Schneider PL, Crouter SE, Bassett DR. Pedometer measures of free-living physical activity: comparison of 13 models. Med Sci Sports Exerc. 2004;36:331–5.
21. Siri NE. Body composition from fluid spaces and density. In: Brozek J, Henschel A, editors. Techniques for Measuring Body Composition. Washington (DC): National Academy of Science; 1961. pp. 223–4.
22. Wagner DR, Heyward VH. Techniques of body composition assessment: a review of laboratory and field methods. Res Q Exerc Sport. 1999;70:135–49.
23. Weyers AM, Mazzetti SA, Love DM, Gómez AL, Kraemer WJ, Volek JS. Comparison of methods for assessing body composition changes during weight loss. Med Sci Sports Exerc. 2002;34:497–502.
Copyright © 2017 by the American College of Sports Medicine