Criterion Validity and Sensitivity to Change of a Pediatric Bioelectrical Impedance Analysis Scale in Adolescents

Abstract

Background:

BMI measures are often used to infer body composition. Bioelectrical impedance analysis (BIA) is a more accurate way to estimate percent body fat (%BF), particularly when screening children who may be overweight. The aim of this study was to determine the validity, sensitivity to change, and diagnostic value of a BIA scale designed specifically for adolescents.

Methods:

One hundred twelve adolescents had their body fat assessed using both BIA and dual-energy X-ray absorptiometry (DEXA). Mean difference and limits of agreement (LoA) were calculated for criterion validity. Intraclass correlation coefficients (ICCs) were calculated for sensitivity to change. Sensitivity/specificity for each classification was also assessed. Data from 46 returning adolescents (6–8 months later) were then used to assess sensitivity to change of BIA compared with DEXA.

Results:

ICC for absolute agreement (range) comparing BIA and DEXA was 0.78 (0.48–0.88). The mean difference between the BIA %BF reading and DEXA was −4.05% (LoA = [4.80%, −12.90%]). Sensitivity and specificity values for the underfat, healthy, overfat, and obese classifications were 0.0/0.89, 0.79/0.46, 0.28/0.92, and 0.5/1.00, respectively. ICC for absolute agreement over time between the BIA %BF and the DEXA %BF was 0.71 (0.242–0.866).

Conclusions:

The Tanita BF-689 demonstrated poor–good agreement with DEXA when measuring %BF, poor–moderate agreement when measuring change in %BF over time, high sensitivity for classification into the healthy category and high specificity for classification into the underfat/overfat/obese categories. Compared to DEXA, the BF-689 is accurate, accessible, and efficient in classifying adolescents based on %BF.

Background

Childhood obesity has more than tripled in the United States since the 1970s¹ and is considered an epidemic both in North America and internationally.² Specifically, the proportion of children in the highest percentile of BMI has seen the most rapid increase.³ Diseases that were never previously diagnosed in children, including type 2 diabetes, sleep apnea, and fatty liver disease, are becoming more prevalent due to their direct association with obesity.¹ Obesity specifically during adolescence has been shown to be the strongest predictor of adult obesity, whereas the relationship between early childhood obesity and adult obesity is not as strong.⁴ It has been shown repeatedly that obesity is associated with increased long-term mortality and morbidity.^1,2,4,5 Therefore, addressing obesity in adolescence may be key for prevention or cardiometabolic comorbidities.⁵ To properly address adolescent obesity and thus improve future quality and length of life, it must first be correctly identified.

Currently, BMI is a widely used and commonly accepted way to measure health and wellness.⁶ However, it is becoming increasingly clear that BMI is a poor estimate of percent body fat (%BF) as it does not differentiate between fat mass and fat-free mass.⁷ It is a measure of relative weight rather than adiposity,⁸ yet professionals and nonprofessionals alike often use it to measure fatness due to ease of measurement. While any method of measuring body composition is an imperfect process, there are several alternative methods to BMI that have been previously tested, including, but not limited to, skinfold measurements and waist circumference.⁹ Skinfold and waist circumference measurements are considered simple yet informative assessments of fatness. However, they both have limitations when assessing whole body composition and correlate poorly with other reference standards in the literature.^9,10

Dual-energy X-ray absorptiometry (DEXA) is considered a more sophisticated method of assessing %BF. It is a commonly accepted reference standard for body composition, but it is costly, time consuming, and often inaccessible to the community.¹¹ DEXA scans also require exposure to low amounts of radiation and are often unable to be performed on obese individuals that exceed the maximum scan width.¹² Another method, bioelectrical impedance analysis (BIA) may be a quick and safe alternative to DEXA as it is noninvasive, portable, and inexpensive. BIA may be a viable option for daily clinical practice, if it is shown to be accurate and dependable.¹¹

Rather than using height and weight alone, BIA measures impedance of the body to a small electrical current.⁹ The currents are used with advanced and validated predictive formulas to differentiate between fat and lean mass tissue in subjects.¹² When compared with reference standards in the literature, BIA devices have been shown to have moderate-to-strong correlations for body composition measures.^11,13 However, there are large variations in reliability and validity between models^14,15 based on the device- and population-specific predictive equations that are used to both calculate and categorize %BF from the measured electrical impedance.¹¹

While most BIA devices have been created for and marketed to adults, the BF-689 (Tanita Corporation, Tokyo, Japan) is a BIA device claimed to be the world's first body composition scale for children and adolescents.¹⁶ This device has been cleared by the FDA to provide body composition assessment between the ages of 5 and 17 years.¹⁶ This tool allows parents and children to monitor body fat percentages from home and has an indicator, which provides an easy-to-read, color-coded output to depict health categories (underfat, healthy, overfat, and obese). It uses predictive equations that require age, gender, and height for each child to account for gender-specific physical developmental milestones. These specific equations are important, as no one predictive equation can be accurate across the entire pediatric age range.¹⁰ Due to being relatively new on the market, the Tanita BF-689 has yet to be proven reliable and valid for the entire purported age range when compared with reference standards. It is also unknown if the BF-689 can accurately track changes in %BF over time. One prior study showed this tool to have excellent test–retest reliability, moderate absolute agreement with DEXA, and high specificity for the overfat and obese classifications but only in children aged 5–11 years.¹² Therefore, the aim of this study was to examine agreement of the BF-689 with DEXA as a gold standard as well as to estimate its sensitivity to change in children 12–17 years of age.

Methods

This study was approved by the Institutional Review Board at Texas Woman's University before recruitment or data collection. A detailed overview of the study was provided in person to all participants by the primary investigator and all questions were answered before signing the written parental/guardian informed consent and adolescent assent forms. Adolescents 12–17 years of age were recruited through word-of-mouth and e-mail from the greater Houston area. Subjects were repeatedly given the opportunity to decline any or all testing procedures throughout data collection and all female participants were offered a urine pregnancy test before the DEXA scan. An a priori power analysis using the Shoukri et al.¹⁷ guidelines for reliability in medical research indicated that a sample size of 39 adolescents would be needed to obtain a 95% confidence level. Sixty-six adolescents enrolled in the study and completed the first data collection visit. Forty-eight participants returned for the second visit 6–10 months later giving a total of 112 samples to assess criterion validity. One participant turned 18 before the second visit and therefore could not be measured on the Tanita scale (age range = 12–17 years) during the second visit and one declined the DEXA scan. This resulted in a total of 46 returning participants to assess sensitivity to change of the BIA.

Each adolescent was assessed for %BF by both the BF-689 and the DEXA machine (Horizon W; Hologic, Inc., Bedford, MA), to be used as the gold standard. DEXA software version 13.6.0.4 with the NHANES adjustment was used to analyze the results. Participants were requested to fast for 2 hours before testing and void their bladder to optimize measurement accuracy. Every child was assessed on the BF-689 followed by DEXA within a 30-minute timeframe. Different raters were used for the BF-689 and the DEXA and participants were denied their results to maintain a double-blind study. The participants completed both tests wearing a single layer of light clothing, bare feet, and no metal jewelry of any kind.

Testing sessions began by measuring the child's height with the participant facing away from the stadiometer (Detecto, Webb City, MO), standing upright with heels together. Next, body fat was calculated using the Tanita BF-689 standardized procedures as described in the manual. Testers guided each subject's heels to the foot sensors for optimal placement. Subjects were asked to stand tall and look straight ahead throughout the duration of the measurements. The BF-689 machine displayed each child's weight in kilograms, body fat percentage, and body fat percentage classification. Classification was determined by colored indicator lights on the display screen (blue = underfat; green = healthy; orange = overfat; red = obese). Cutoffs for each classification were preprogrammed and reflect the cutoffs established by McCarthy et al.¹⁸

All DEXA scans were performed by a certified technician after proper calibration with a density phantom. Each subject was positioned in the machine by the technician lying supine with forearms pronated and hips internally rotated as per device specifications. The great toes were secured together with a thin piece of tape to maintain proper foot alignment. All subjects were instructed to lie still without speaking for the duration of the test. In addition to completing these physical measurements, each participant filled out the Pubertal Development Scale and parents filled out a demographic survey.

Statistical Analyses

An intraclass correlation coefficient (ICC) value, using a two-way random effects model,¹⁹ was calculated between BIA and DEXA readings from each participant to determine absolute agreement. Body fat classification for DEXA was based on the same age- and sex-specific charts used by the BF-689.¹⁸ Mean difference between BIA %BF and DEXA %BF and the corresponding limits of agreement (LoA) using Bland–Altman plots were calculated to assess convergent validity. LoA is often used to assess agreement between tools that measure the same construct.²⁰ Sensitivity and specificity values were calculated to examine the BF-689s ability to correctly classify children's body fat as underfat, healthy, overfat, or obese when compared with the DEXA machine. Forty-six participants were measured twice over an 8-month period and an ICC was calculated between the change in DEXA score and the change in BIA score. Correlation analyses are thought to be well suited to measure responsiveness, often between a new health-related measure, the BF-689, and a traditional clinical outcome, DEXA.²¹ All statistical analyses were completed using SPSS software (v. 25.0; SPSS, Inc., Chicago, IL). All stated ranges indicate 95% confidence intervals, and the alpha value used to test significance was p = 0.05.

Results

Table 1 shows the demographics and descriptive statistics of the subjects that participated in the visits; there were 66 participants in Visit 1 and 46 participants in Visit 2.

Table 1.

Participant Demographics

Visit 1: n = 66Visit 2: n = 46	Mean or %		SD or number		Range
Visit 1: n = 66Visit 2: n = 46	Visit 1	Visit 2	Visit 1	Visit 2	Visit 1	Visit 2
Age	14.55 years	15.0 years	1.64	1.56	[12.0–17.9]	[12.4–17.6]
Gender
Male	50%	52.2%	33	24
Female	50%	47.8%	33	22
Race
White	81.8%	80.4%	54	37
Black	6.1%	6.5%	4	3
Hispanic	6.1%	8.7%	4	4
Asian	3.0%	2.2%	2	1
Other	3.0%	2.2%	2	1
Height, cm	163.78	165.58	9.53	10.0	[146.7–188.0]	[147.0–188.0]
Weight, kg	57.67	60.83	11.86	13.43	[36.0–96.7]	[38.2–93.8]
Body fat percentile classification (DEXA)
Underfat	0%	0%	0	0
Healthy	65.2%	60.9%	43	28
Overfat	18.2%	28.3%	12	13
Obese	16.7%	10.9%	11	5
Body fat percentile classification (BIA)
Underfat	10.6%	10.9%	7	5
Healthy	71.2%	67.4%	47	31
Overfat	12.1%	13.0%	8	6
Obese	6.1%	8.7%	4	4
BMI	21.4	21.8	3.35	3.89	[16.6–34.7]	[16.2–33.4]
Pubertal developmental scale stage
Prepubertal	4.5%	2%	3	1
Early pubertal	4.5%	2%	3	1
Mid-pubertal	27.3%	29%	18	14
Late pubertal	53.0%	52%	35	25
Postpubertal	10.6%	15%	7	7

BIA, bioelectrical impedance analysis; DEXA, dual-energy X-ray absorptiometry; SD, standard deviation.

The ICC value for absolute agreement between the BF-689 and DEXA (n = 112) was 0.78 [0.48–0.88; p < 0.001]. The mean difference between the BIA %BF reading and DEXA was −4.05% (LoA = [4.80%, −12.90%]). The relationship between the differences and mean of BIA and DEXA %BF measurements are shown in a Bland–Altman plot with a linear regression line (Fig. 1). Table 2 is a contingency table comparing BIA %BF and DEXA %BF calculations for underfat, healthy, overfat, and obese children. Overall sensitivity (true positive rate) and specificity (true negative rate) were calculated for the four BIA %BF classifications. Sensitivity/specificity values for the underfat, healthy, overfat, and obese classifications were 0.0/0.89, 0.79/0.46, 0.28/0.92, and 0.5/1.00, respectively.

Figure 1.

Bland–Altman Plot with mean difference, limits of agreement, and regression line for BIA %BF and DEXA %BF reading. Slope = −0.011; t-value = −0.209, p = 0.834. %BF, percent body fat; BIA, bioelectrical impedance analysis; DEXA, dual-energy X-ray absorptiometry.

Table 2.

Contingency Table Comparing Bioelectrical Impedance Analysis and Dual-Energy X-Ray Absorptiometry Percent Body Fat Calculations for All Classifications

Categorization		Test positive on DEXA	Test negative on DEXA
Underfat	Test positive on BIA	0	12	Total: 12
	Test negative on BIA	0	100	Total: 100
		Total: 0	Total: 112	n = 112
Healthy	Test positive on BIA	56	22	Total: 78
	Test negative on BIA	15	19	Total: 34
		Total: 71	Total: 41	n = 112
Overfat	Test positive on BIA	7	7	Total: 14
	Test negative on BIA	18	80	Total: 98
		Total: 25	Total: 87	n = 112
Obese	Test positive on BIA	8	0	Total: 8
	Test negative on BIA	8	96	Total: 104
		Total: 16	Total: 96	n = 112

Forty-six subjects completed all testing requirements during both visits and were used to calculate sensitivity to change. The average BIA %BF change (standard deviation [range]) between visits was +1.2% (2.17 [−4.1 to 5.6]) and the average DEXA %BF change was −0.12% (2.02 [−6.3 to 4.1]). The ICC value for absolute agreement between BIA change and DEXA change (n = 46) was 0.71 (0.242–0.866, p < 0.001).

Discussion and Conclusion

When looking at the ICC value, this study shows good absolute agreement between the Tanita BF-689 and DEXA. Moderate absolute agreement was found between the two devices when studying change in %BF over time.¹⁹ However, this study shows poor-to-good absolute agreement between the two devices and poor-to-moderate absolute agreement when studying change over time if the 95% confidence intervals are considered as suggested by Koo and Li.¹⁹ The agreement between the devices aligns with a previous study that reported an absolute agreement of 0.79 between DEXA and BIA in children ages 5–11.¹² As there is limited previous evidence for agreement between BIA and DEXA in children of all ages, it is promising that this specific device showed similar results in both elementary school children and adolescents. Unlike prior studies,²² which used a Pearson product-moment correlation coefficient, we chose to report validity using an ICC to account for both correlation and absolute agreement. The mean difference between the BIA %BF reading and the DEXA %BF reading was found to be −4.05% (LoA = [4.80%, −12.90%]. In a systematic review of 50 recent studies,²² the mean difference between BIA and the gold standard of body composition (multicomponent models) ranged from −12.0% to 13.7%. Given this range, the calculated underestimation of %BF on the BF-689 (−4.05%), while still present, is smaller than many previous studies. In addition, this previous value is smaller than the mean difference (−6.75%) found in children ages 5–11.¹²

No significant gender or classification bias was seen with the BF-689. Likely, the lack of gender bias is due to the use of gender-specific equations to analyze %BF in the BF-689. Before any statistical analysis, a Bland–Altman plot was created to rule out any magnitude bias as previous studies on BIA devices have shown significant differences in accuracy depending on the classification groups (underweight, healthy, overfat, and obese).^11,12,22 Table 1 shows that a larger number of children were classified as overfat or obese by DEXA when compared with BIA. DEXA also classified fewer children as underfat when compared with BIA. However, Figure 1 shows a negligible slope in the regression line (−0.011). This, along with the inability to reject the null hypothesis that the average of the two mean equals 0 (t = −0.209, p = 0.834), confirms that there is no magnitude bias and that the device is appropriate to use on children of all sizes. The absence of gender or magnitude bias increases the device's usefulness in testing %BF in the entire adolescent population.

The sensitivity and specificity data also demonstrates the diagnostic value of the BF-689 to triage or categorize children into health classifications. Diagnostic tools that are sensitive are used to rule out the presence of a condition when the result is negative, while diagnostic tools that are specific are used to rule in the presence of a condition when the result is positive. In this study, the BP-689 demonstrated low sensitivity for the underfat and overfat classifications (0.00 and 0.28, respectively), moderate sensitivity for the obese classification (0.50), and high sensitivity for the healthy classification (0.79). The device demonstrated moderate specificity for the healthy classification (0.46) and high specificity for the underfat, overfat, and obese classifications (0.89, 0.92, and 1.00, respectively).

These results demonstrate that of all the tested children with adequate body composition, the BF-689 correctly categorized 79% of them as healthy. This is useful because if a child is categorized in another classification (such as underfat, overfat, or obese), we can be fairly certain that they truly are unhealthy. Being able to classify children outside of the healthy classification is important in screening for obesity as it can be the first step to identifying a need for intervention. The specificity results also show that a child categorized as underfat, overfat, or obese are correctly classified with 89%, 92%, and 100% accuracy, respectively. By combining both the sensitivity and specificity results, the device can successfully identify and classify children in their correct category potentially supporting its use as a triage instrument.

Since 46 of the participants were measured twice over the course of 6–8 months, these data were also used to calculate the sensitivity to change of the BF-689 as compared with the DEXA. Statistical analysis showed an absolute agreement of 0.71 between changes in %BF measured on both devices. This tells us that the BF-689 is a reliable way to track changes in health classification based on body composition over time, which is especially useful if the device is being used at home without a health care professional to regularly monitor body composition. As adolescents are constantly growing and developing, particularly as they go through puberty,²³ it is important that any device used to measure body composition accurately reflects these changes in classification. The BF-689 was able to track increases in %BF of up to 5.6% and decreases in %BF of up to −4.1%, showing it tracked a total %BF fluctuation of 9.6%. Therefore, this device can be used to reliably track a sizeable range in both an increase and decrease in %BF.

Previous studies have shown that pediatricians and primary care providers often feel as if they do not have enough time per check-up to thoroughly discuss obesity and weight management.²⁴ It may be more effective for parents to manage their children's weight at home, provided they have the appropriate tools and knowledge to do so. Studies have also shown that after an appointment with a physician, parents are likely to remember if their child's BMI was higher than expected but they are often unable to remember any other information or advice that was provided.²⁵ This device could allow parents to identify and track their child's body composition classification and take appropriate action if necessary. By empowering parents to be more involved in managing their children's health, we may be able to decrease the number of overweight children at risk to become overweight adults. Since the BF-689 is relatively affordable¹⁶ and easy for nonprofessionals to use, this device would be beneficial to any family interested in tracking health classification in their children. Although there is not yet strong enough evidence to use the BIA device exclusively to measure and track %BF, the device is sufficient to use when categorizing children into health classifications.

Notably, this study was not without limitations, including our sample of convenience. The participants were primarily Caucasian and fell into the healthy BMI classification. There were more adolescents in the later stages of puberty compared with the earlier stages, and no adolescents were classified as underfat on the DEXA. Between the two periods of data collection, there was a decrease in the number of participants (n = 66 on visit 1, n = 46 on visit 2) resulting in an attrition rate of 30.3%. Despite the decrease, our sample size for the sensitivity to change analysis still exceeded the 39 participants identified by our a priori power analysis. Other strengths of this study included the wide age range of the participants and an even representation of both genders.

In summary, the BF-689 was found to accurately classify adolescents 12–17 years of age based on %BF and to track changes in their classification over time. The combined sensitivity and specificity data show that this BIA device can be used to both rule in and rule out specific body fat categories for each patient. While further research with larger sample sizes and a wider range of body composition may be helpful in confirming these initial results, the results of this investigation showed that the BF-689 can be used by health care professionals and parents alike to assess %BF classification and track body composition classification changes over time in adolescents.

Footnotes

Acknowledgment

This study was funded by the Texas Physical Therapy Foundation.

Author Disclosure Statement

No competing financial interests exist.

References

Yanovski

. Pediatric obesity. An introduction. Appetite, 2015; 93:3–12.

Deckelbaum

, Williams

. Childhood obesity: The health issue. Obes Res, 2001; 9:239S–243S.

Lakshman

, Elks

, Ong

. Childhood obesity. Circulation, 2012; 126:1770–1779.

Dietz

, Gortmaker

. Preventing obesity in children and adolescents. Annu Rev Public Health, 2001; 22:337–353.

Mossberg

. 40-Year follow-up of overweight children. Lancet, 1989; 2:491–493.

Nuttall

. Body mass index: Obesity, BMI, and health: A critical review. Nutr Today, 2015; 50:117–128.

Sahoo

, Sahoo

, Choudhury

, et al. Childhood obesity: Causes and Consequences. J Family Med Prim Care, 2015; 4:187–192.

Whitlock

, Williams

, Gold

, et al. Screening and interventions for childhood overweight: A summary of evidence for the US preventive services task force. Pediatrics, 2005; 116:125.

Wells

, Fewtrell

. Measuring body composition. Arch Dis Child, 2006; 91:612–617.

10.

Dietz

, Story

, Leviton

. Issues and implications of screening, surveillance, and reporting of children's BMI. Pediatrics, 2009; 124(Suppl 1):98.

11.

Wan

, Ward

, Halim

, et al. Bioelectrical impedance analysis to estimate body composition, and change in adiposity, in overweight and obese adolescents: Comparison with dual-energy x-ray absorptiometry. BMC Pediatr, 2014; 14:249.

12.

Kabiri

, Hernandez

, Mitchell

. Reliability, validity, and diagnostic value of a pediatric bioelectrical impedance analysis scale. Child Obes, 2015; 11:650–655.

13.

Y-C

, Li

C-I

, Lin

W-Y

, et al. Percentage of body fat assessment using bioelectrical impedance analysis and dual-energy X-ray absorptiometry in a weight loss program for obese or overweight Chinese adults. PLoS One, 2013; 8:e58272.

14.

Lazzer

, Boirie

, Meyer

, et al. Evaluation of two foot-to-foot bioelectrical impedance analysers to assess body composition in overweight and obese adolescents. Br J Nutr, 2003; 90:987–992.

15.

Montagnese

, Williams

, Haroun

, et al. Is a single bioelectrical impedance equation valid for children of wide ranges of age, pubertal status and nutritional status? Evidence from the 4-component model. Eur J Clin Nutr, 2012; 67(Suppl 1):S34–S39.

16.

Tanita Corporation. BF-689 children's body fat monitor. Updated. 2018. Available at www.tanita.com/en/bf-689 Last accessed April 12, 2018.

17.

Shoukri

, Asyali

, Donner

. Sample size requirements for the design of reliability study: Review and new results. Stat Methods Med Res, 2004; 13:251–271.

18.

McCarthy

, Cole

, Fry

, et al. Body fat reference curves for children. Int J Obes (Lond), 2006; 30:598–602.

19.

Koo

, Li

. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chriopr Med, 2016; 15:155–163.

20.

Zou

, Zou

. Confidence interval estimation for the Bland-Altman limits of agreement with multiple observations per individual. Stat Methods Med Res, 2013; 22:630–642.

21.

Husted

, Cook

, Farewell

, Gladman

. Methods for assessing responsiveness. J Clin Epidemiol, 2000; 53:459–468.

22.

Talma

, Chinapaw

, Bakker

, et al. Bioelectrical impedance analysis to estimate body composition in children and adolescents: A systematic review and evidence appraisal of validity, responsiveness, reliability and measurement error. Obes Rev, 2013; 14:895–905.

23.

Loomba-Albrecht

, Styne

. Effect of puberty on body composition. Curr Opin Endocrinol Diabetes Obes, 2009; 16:10–15.

24.

Flower

, Perrin

, Viadro

, Ammerman

. Using body mass index to identify overweight children: Barriers and facilitators in primary care. Ambul Pediatr, 2007; 7:38–44.

25.

Dawson

, Taylor

, Williams

, et al. Do parents recall and understand children's weight status information after BMI screening? A randomised controlled trial. BMJ Open, 2014; 4:e004481.