Reliability of Multifrequency Bioelectrical Impedance Analysis to Quantify Body Composition in Patients After Musculoskeletal Trauma

Abstract

Background

Changes in body composition, especially loss of lean mass, commonly occur in the orthopedic trauma population due to physical inactivity and inadequate nutrition. The purpose of this study was to assess inter-rater and intra-rater reliability of a portable bioelectrical impedance analysis (BIA) device to measure body composition in an orthopedic trauma population after operative fracture fixation. BIA uses a weak electric current to measure impedance (resistance) in the body and uses this to calculate the components of body composition using extensively studied formulas.

Methods

Twenty subjects were enrolled, up to 72 hours after operative fixation of musculoskeletal injuries and underwent body composition measurements by two independent raters. One measurement was obtained by each rater at the time of enrollment and again between 1-4 hours after the initial measurement. Reliability was assessed using intraclass correlation coefficients (ICC) and minimum detectable change (MDC) values were calculated from these results.

Results

Inter-rater reliability was excellent with ICC values for body fat mass (BFM), lean body mass (LBM), skeletal muscle mass (SMM), dry lean mass (DLM), and percent body fat (PBF) of 0.993, 0.984, 0.984, 0.979, and 0.986 respectively. Intra-rater reliability was also high for BFM, LBM, SMM, DLM, and PBF, at 0.994, 0.989, 0.990, 0.983, 0.987 (rater 1) and 0.994, 0.988, 0.989, 0.985, 0.989 (rater 2). MDC values were calculated to be 4.05 kg for BFM, 4.10 kg for LBM, 2.45 kg for SMM, 1.21 kg for DLM, and 4.83% for PBF.

Conclusion

Portable BIA devices are a versatile and attractive option that can reliably be used to assess body composition and changes in lean body mass in the orthopedic trauma population for both research and clinical endeavors.

Level of Evidence: III

Introduction

Sarcopenia is the age-related decline in skeletal muscle mass and strength.¹ Physical activity is important for maintaining bone density and muscle function. Bedrest and physical inactivity lead directly to loss of lean mass.^2-4 Even short periods of inactivity^5,6 can lead to substantial loss of muscle mass. Decreased function is required during the recovery phase after musculoskeletal trauma.⁷ Low lean body mass (LBM) after musculoskeletal trauma results in a higher rate of complications, poorer outcomes, and increased mortality.^8-11

There is a current focus on nutritional interventions and early rehab programs to prevent loss of LBM after orthopedic trauma. Accurate and precise longitudinal measurements of body composition in these individuals is critical when assessing the efficacy of interventions targeted to maintain LBM in future clinical trials.^12,13 There are many methods currently being used to assess body composition (whole-body DXA, CT, MRI, PET scan, air displacement plethysmography), but many of these methods are not suitable for clinical trials because they are indirect measures of LBM, require highly trained technicians, expose subjects to radiation, are expensive to use, and/or are not available at the bedside early after injury.^14,15

The purpose of this study was to (1) Assess the interrater and intra-rater reliability of bioelectrical impedance (BIA) to measure body composition in patients after operative fracture fixation and (2) Quantify standard error of measurement (SEM) and minimum detectable change (MDC) in these same patients, using the portable, bedside, non-invasive InBodyS10 device (InBody USA, Cerritos, CA).^16,17 Having a reliable BIA device for body composition measurement will allow for accurate assessment of body composition for musculoskeletal trauma patients in the research setting.

Methods

Subjects

Approval was obtained from our Institutional Review Board. Written informed consent was obtained from all participants. We enrolled 20 adult participants after operative fixation of pelvic or extremity fractures. Operative fixation included both temporary fracture fixation procedures, such as external fixation, as well as definitive fracture fixation with plates and screws or intramedullary nail. Participants with multiple injuries requiring multiple operations were eligible for enrollment after their first operative fixation procedure. The exclusion criteria were (1) BIA measurements not able to be completed within 72 hours of operative fixation, (2) BIA measurements not able to be obtained due to limits of patient positioning or splints/casts obstructing electrode placement, or (3) previous cardiac pacemaker implantation as this is contraindicated with BIA.

Measurement Technique

BIA body composition measurements were obtained in the supine position in accordance with the manufacturer’s recommendations; arms separated from the trunk and legs shoulder width apart without the thighs touching (Figure 1A).¹⁸ Body composition measurements using the InBody S10 device can be obtained using two different types of electrodes, known as touch type electrodes and adhesive type electrodes. To compare the reliability of the BIA device using both types of electrodes, 10 subjects were measured using the touch type electrodes and 10 subjects were measured using the adhesive type electrodes. The touch type electrodes were placed according to the manufacturer recommendations, with the finger clips placed on the first and third digits bilaterally and the ankle clips placed just posterior to the ankle malleoli bilaterally. Placement of the adhesive type electrodes was modified from the manufacturer recommendations to be better suited for orthopedic trauma patients who frequently have upper and/or lower extremity casts. For the upper extremity electrodes, the red electrode was placed over the head of the second metacarpal and the black electrode over the head of the fifth metacarpal. Similarly, for the lower extremity electrodes, the red electrode was placed over the head of the first metatarsal and the black electrode over the head of the fifth metatarsal (Figure 1B). Weight, height, age, and gender were recorded from the electronic health record to be entered into the device before obtaining the measurements.

Figure 1.

(A) Proper supine body positioning for InBody S10 measurements, obtained from user’s manual.¹³ (B) modified adhesive type electrode placement on the bilateral upper and lower extremities.

Reliability

Body composition measurements were completed by two reviewers (B.K. and A.M.). Each subject was measured a total of four times, twice by each reviewer. The first measurement by each reviewer was obtained immediately following the informed consent and enrollment process. The electrodes were completely removed from the patient between measurements and the reviewers were blinded to results of the other reviewer. The second two measurements were obtained at least 1 hour after, but no more than 4 hours after, the first set of measurements. If the patient had any food or drink, took any medications, or participated in physical therapy between the two sets of measurements, this was recorded. All measurements were obtained in a single day and, thus, no follow-up was required.

Statistical Analysis

Inter-rater and intra-rater reliability of BIA body composition measurements were assessed using the intraclass correlation coefficient (ICC). Overall ICC values were calculated using all subjects, as well as ICC values for upper extremity and lower extremity injuries separately. The inter-rater reliability data was then used to calculate a Standard Error of Measurement (SEM). The SEM was then subsequently used to determine the Minimal Detectable Change with a 90% confidence interval (MDC90) and Minimal Detectable Change with a 95% confidence interval (MDC95) for the overall inter-rater reliability and the lower extremity inter-rater reliability data.

Results

A total of 20 orthopedic trauma patients, aged 27-88 years, were enrolled from October 2021 to December 2021 and underwent BIA body composition measurements by two trained members of the research team. The demographics of the participating subjects as well as their associated injuries are listed in Table 1. During statistical analysis it was determined that height and weight was entered incorrectly for one subject, so results of only 19 subjects were used for statistical analysis.

Table 1.

Patient Demographics and BIA Measurements

	Total (n=19)	Male (n=13)	Female (n=6)
Age (years)	59.5 ± 17.6	57.3 ± 17.0	64.3 ± 18.0
Height (cm)	172.4 ± 10.9	178.5 ± 7.3	159.4 ± 21.6
Weight (kg)	86.6 ± 20.5	85.0 ± 19.8	90.1 ± 21.6
BMI	29.3 ± 7.3	26.5 ± 5.4	35.2 ± 7.4
BFM (kg)	28.4 ± 17.1	21.8 ± 12.7	42.9 ± 16.6
LBM (kg)	58.1 ±12.0	63.1 ± 10.6	47.2 ± 6.1
SMM (kg)	31.7 ± 7.1	34.6 ± 6.2	25.2 ± 3.9
DLM (kg)	15.3 ± 3.1	16.6 ± 2.7	12.5 ± 1.6
PBF (%)	30.9 ± 14.5	23.9 ± 11.1	46.0 ± 7.9
Injuries:
- Upper Ext. Only:	1 (5%)
- Lower Ext. Only:	9 (47%)
- Upper + Lower Ext:	4 (21%)
- Pelvis Only:	1 (5%)
- Pelvis + Ext:	4 (21%)

BMI, Body Mass Index; BFM, Body Fat Mass; LBM, Lean Body Mass; SMM, Skeletal Muscle Mass; DLM, Dry Lean Mass; PBF, Percent Body Fat.

BIA Outcomes

As can be seen from Table 1, the average BMI was 29.3, but broken down further one (5%) subject was underweight (BMI < 18.5), six (32%) subjects had a healthy BMI (18.5-24.9), three (16%) subjects had an overweight BMI (25-29.9), and nine (47%) subjects had an obese BMI (>30). Overall, BIA body composition measurements for Body Fat Mass (BFM), Lean Body Mass (LBM), Skeletal Muscle Mass (SMM), Dry Lean Mass (DLM), and Percent Body Fat (PBF) were 28.4 kg, 58.1 kg, 31.7 kg, 15.3kg, and 30.9%, respectively. The values for BFM, LBM, SMM, DLM, and PBF in men were 21.8 kg, 63.1 kg, 34.6 kg, 16.6 kg, and 23.9% and for women were 42.9 kg, 47.2 kg, 25.2 kg, 12.5 kg, and 46.0%, respectively.

Reliability

Overall, the inter-rater and intra-rater reliability was excellent in all categories analyzed. The inter-rater reliability ICC values for BFM, LBM, SMM, DLM, and PBF were 0.993, 0.984, 0.984, 0.979, and 0.986 respectively. For intra-rater reliability, ICC values for rater 1 were 0.994, 0.989, 0.990, 0.983, 0.987 and for rater 2 were 0.994, 0.988, 0.989, 0.985, 0.989 for BFM, LBM, SMM, DLM, and PBF respectively (Table 2). Reliability remained excellent when the data was divided into lower extremity injuries (n=15) and upper extremity injuries (n=8). Inter-rater ICC values for injured lower limbs were 0.918, 0.987, and 0.985 for LBM, BFM, and PBF respectively. Similarly, for upper extremity injuries, inter-rater ICC values were 0.984, 0.995, and 0.993 for LBM, BFM, and PBF respectively. These values, along with the upper and lower extremity intra-rater reliability values are listed in Table 3.

Table 2.

Overall BIA Reliability Data Showing Inter-rater and Intra-rater ICC Values With 95% Confidence Intervals and Associated P-values

	Inter-rater		Intra-rater
	ICC (95% CI)	p-value	Rater 1 ICC (95% CI)	p-value	Rater 2 ICC (95% CI)	p-value
BFM:	0.993 (0.981- 0.997)	<0.001	0.994 (0.986- 0.998)	<0.001	0.994 (0.9860.998)	<0.001
LBM:	0.984 (0.958- 0.994)	<0.001	0.989 (0.973- 0.996)	<0.001	0.988 (0.9700.995)	<0.001
SMM:	0.984 (0.960- 0.994)	<0.001	0.990 (0.975- 0.996)	<0.001	0.989 (0.9720.996)	<0.001
DLM:	0.979 (0.948- 0.992)	<0.001	0.983 (0.957- 0.993)	<0.001	0.985 (0.9630.994)	<0.001
PBF:	0.986 (0.964- 0.995)	<0.001	0.987 (0.969- 0.995)	<0.001	0.989 (0.9730.996)	<0.001

BFM, Body Fat Mass; LBM, Lean Body Mass; SMM, Skeletal Muscle Mass; DLM, Dry Lean Mass; PBF, Percent Body Fat

Table 3.

Inter- and Intra-rater Reliability ICC Values Stratified by Lower Extremity and Upper Extremity Injuries

	Inter-rater		Intra-rater
	ICC (95% CI)	p-value	Rater 1 ICC (95% CI)	p-value	Rater 2 ICC (95% CI)	p-value
Lower Extremity
- LBM:	0.918 (0.756- 0.972)	<0.01	0.990 (0.973- 0.997)	<0.01	0.992 (0.9770.997)	<0.01
- BFM:	0.987 (0.963- 0.996)	<0.01	0.991 (0.974- 0.997)	<0.01	0.993 (0.9810.998)	<0.01
- PBF:	0.985 (0.956- 0.995)	<0.01	0.987 (0.965- 0.995)	<0.01	0.990 (0.9720.997)	<0.01
Upper Extremity
- LBM:	0.984 (0.927- 0.996)	<0.01	0.980 (0.919- 0.995)	<0.01	0.994 (0.9770.999)	<0.01
- BFM:	0.995 (0.970- 0.999)	<0.01	0.998 (0.993- 1.000)	<0.01	1.000 (0.9991.000)	<0.01
- PBF:	0.993 (0.963- 0.999)	<0.01	0.997 (0.989- 0.999)	<0.01	1.000 (0.9991.000)	<0.01

LBM, Lean Body Mass; BFM, Body Fat Mass; PBF, Percent Body Fat.

Minimal Detectable Change

Inter-rater reliability data was used to calculate SEM and MDC values. Using the overall inter-rater reliability data, the MDC95 values were the following: 4.05 kg for BFM, 4.10 kg for LBM, 2.45 kg for SMM, 1.21 kg for DLM, and 4.83% for PBF. When the inter-rater reliability data from only lower extremity injuries was used, the following MDC95 values were obtained: 2.31 kg for LBM, 0.88 kg for BFM, and 37.66% for PBF. These values are listed in Table 4.

Table 4.

Minimum Detectable Change (MDC) Values Calculated From All Subjects and From the Subjects With Lower Extremity Injuries

	ICC	p-value	S.D.	SEM	MDC95	MDC90
All Subjects:
- BFM:	0.993	<0.001	17.45	1.46	4.05	3.44
- LBM:	0.984	<0.001	11.68	1.477	4.1	3.47
- SMM:	0.984	<0.001	6.99	0.884	2.45	2.08
- DLM:	0.979	<0.001	3.01	0.436	1.21	1.02
- PBF:	0.986	<0.001	14.74	1.744	4.83	4.10
Lower Ext. Injuries
- LBM:	0.918	<0.001	0.833	0.833	2.31	1.94
- BFM:	0.987	<0.001	0.316	0.316	0.88	0.73
- PBF:	0.985	<0.001	13.587	13.587	37.66	31.61

ICC, Intraclass Correlation Coefficient; SD, Standard Deviation; SEM, Standard Error of Measurement.

Discussion

This study assessed the intra-rater and inter-rater reliability of a portable BIA device to measure body composition after operative fracture fixation and to determine the minimal detectable change in these same patients. Reliability values were excellent for both interrater and intra-rater reliability in this population with an average age of 59 years and an average BMI of 29.3. For total body measurements, rater 1 had an average ICC of 0.989, rater 2 an average of 0.989, and inter-rater ICC average of 0.985. LBM measurements were one of the most reliable, with ICC values of 0.984, 0.989, and 0.988 for inter-rater, rater 1, and rater 2 reliability, respectively. Reliability values remained excellent when patients were stratified into upper and lower extremity injuries. For all subjects, MDC90 values were found to be 3.44 kg, 3.47 kg, and 4.10 kg for BFM, LBM, and PBF, respectively.

Reliability of BIA for various body composition measurements has been studied in other populations and using various BIA devices. In 1985, Lukaski et al. developed a method to use BIA to assess fat free mass. They measured 37 healthy male volunteers on successive days using a four-electrode method and had a correlation coefficient (r) of 0.98 across all their measurements.¹⁹ In a systematic review looking at body fat measurement in children and adolescents, Talma et al. found strong reliability results for BIA measurements completed on the same day and on consecutive days. For same day measurements, they found intra-rater reliability of 0.99 and inter-rater reliability of 0.97. For consecutive day measurements, they found a reliability value of 0.97.²⁰ Schubert et al. conducted a reliability study of various laboratory methods to assess body composition in young adults, with BIA being one of the techniques assessed. This was a highly controlled study in which food intake, hydration, and physical activity were all limited prior to measurements. They calculated r values of 0.93, 0.98, and 0.94 for percent body fat, fat free mass, and fat mass respectively.²¹ Vasold et al. completed a reliability study using three different low-cost BIA devices looking at fat free mass. In this study, reliability for measures of fat free mass ranged from 0.991-0.996 for all subjects, 0.973-0.985 when stratified for males, and 0.921-0.991 for females.²² The inter-rater and intra-rater ICC values in our study involving orthopedic trauma patients were comparable to the reliability values found in these other studies across various populations, with values often above 0.97. With confidence that measurements will be precise across periods of time and separate reviewers, BIA can be a powerful tool for body composition measurements in the research setting.²³

Bioelectric impedance is only one method in which body composition can be measured, so it is important to determine how reliable BIA is compared to other methods of measurement. Underwater weighing (UWW) was one of the first body composition measurements used due to its simplicity. In 1985, while testing two different tanks used for UWW, Williams et al. found percent body fat correlation coefficients of 0.989 and 0.981 for the two different tanks.²⁴ Whole body DXA has long been the gold standard for body composition measurement, and Shiel et al. in their reliability study using DXA found ICC values of 0.996 for both whole body fat mass and wholebody lean mass. When the data was stratified regionally to arms, trunk and legs, the ICC values were lower but were consistently above 0.96.²⁵ Air displacement plethysmography (ADP) is another standard for body composition measurements, although it is not used as often as DXA. In a study of 283 women assessing ADP, Tucker et al. found an ICC value of 0.991. When they completed a third measurement and used the two closest values, the ICC improved to 0.998.²⁶ Lastly, A-mode ultrasound is a another method of body composition analysis that is being studied currently. Specifically in trauma subjects, Hendrickson et al. had ICC values of 0.96, 0.98, and 0.99 for percent body fat, fat mass, and fat free mass, respectively.¹³ In a comprehensive study of 32 healthy adult patients, Schubert et al. studied the reliability and validity of several of the most common body comp. measurement techniques in a single study. For UWW reliability values were 0.964, 0.993, and 0.968; for ADP were 0.973, 0.992, and 0.975; for DXA were 0.996, 0.994, and 0.997; and for BIA were 0.983, 0.997, and 0.986 for percent body fat, fat free mass, and fat mass, respectively.²¹ As can be seen, reliability of BIA in this study using orthopedic trauma patients is comparable to the other methods used to measure body composition, including those that are the gold standard. Due to its high reliability, and because it is portable and easy to use, BIA is a feasible option for body composition analysis in the trauma population. Most trauma patients are limited in their mobility due to their injuries, and thus a device that can be brought to the bedside for measurement is essential. BIA using the InBody S10 meets these requirements.

Finally, it is worth discussing the accuracy of BIA devices in predicting lean body mass (LBM). Despite many articles in the literature validating BIA against DXA and other reference standards,^23,27-29 questions of BIA to accurately measure body composition are continuously raised.³⁰ BIA measures LBM (or fat free mass) by determining the total body water content and then using the assumption that LBM is uniformly and constantly hydrated at a certain percentage.³¹ Of course this assumption can become inaccurate in certain populations and disease states, most notably elderly patients, obese patients, and patients with significant volume changes such as occurs in heart failure, liver disease, or kidney disease. It is also reasonable to assume that the hydration status of a patient around the time of BIA measurement can affect the values. In fact, Tinsley et al. found hydration status to be an independent variable predicting discrepancies between DXA and BIA measurements.³² This has led many to believe that BIA tends to either underpredict or overpredict LBM, depending on the circumstances. In a retrospective study of over 3600 measurements, Achamrah et al. found BIA underpredicted LBM by 0.8-2.5 kg, as compared to DXA as the gold standard, at BMI’s below 18.5 kg/m2, and overpredicted by 0.6-5.6 kg at BMI’s above 18.5 kg/m2.³³ In contrast, Ling et al. found that BIA, as compared to DXA as the gold standard, slightly underpredicted at normal and overweight BMI’s but overpredicted at obese BMI’s in a healthy middle-age adult population.³⁴ In hemodialysis patients, who have a frequently changing hydration status, Furstenberg et al., measured an average LBM of 46.215 using BIA and 45.691 using DXA.³⁵ In an outpatient Thai hemodialysis population, Jayanama et al. found a LBM of 43.77 using DXA, 42.6 using the portable InBody S10, and 45.08 using a stationary BIA device.³⁶ Lastly, comparing genders and body segments, Leahy et al. found that BIA overestimated LBM by 0.6 kg in men and 2.0 kg in women as compared to DXA as the gold standard in a healthy population aged 18-29 years. Dividing the body into segments, they report that BIA overestimated LBM in the trunk by 5.8 kg in men and 6.2 kg in women. For extremities, BIA underestimated LBM by 0.2-0.8 kg in men and 0.2-0.7 kg in women.³⁷ From this literature, it is clear that BIA has good validity for measuring LBM and does not predictably under or overpredict LBM, regardless of the population being studied. Importantly for orthopedic trauma patients, the measurement of LBM by BIA in the extremities appears to be quite accurate.

Limitations

There are several limitations of this study that are worth discussion. The first limitation of this study is the small population size and lack of ethnic diversity. Although we completed this project using 20 subjects, we felt this was an appropriate number for an initial reliability study using BIA technology in a specific orthopedic trauma fracture fixation population. This study was completed at a single Midwest academic center and thus diversity was limited to that of the general Midwest population. As portable BIA devices gain popularity in measuring body composition in the orthopedic population, reliability should be periodically reassessed as the population size and diversity of patients measured increases. Secondly, only a single BIA device was used in this study and no gold standard body composition device was used for validation. In future studies, it will be beneficial to assess whether reliability is different using different BIA devices, especially the portable options used in this study and the less portable options that are currently on the market. A portable device was preferred in this study as trauma patients are inherently less mobile given the nature of their injuries. Although BIA has been extensively validated against DXA and other techniques, it will need to be validated specifically in the orthopedic trauma population for future work. Third, all subjects enrolled were in the trauma population and we did not complete any measurements on a healthy population for comparison. As we did not compare to a healthy control population, we are unable to assess whether the trauma itself (with localized edema and fluid shifts) changes the measurements obtained by BIA devices. Relatedly, we excluded subjects who were unable to assume the general body positioning outlined in the InBodyS10 user manual, so we are unable to assess how large deviations from this body positioning affect the reliability of the measurements. If the reliability decreases with significant changes in body positioning, this may be one limitation to using BIA in the trauma population. Lastly, we only performed measurements on a single day for each patient, thus differences across longer periods of time cannot be assessed with this current study.

Conclusion

In conclusion, the results of this study show portable BIA to have a high inter-rater and intra-rater reliability for total and segmental muscle mass in an orthopedic trauma population after operative fixation. MDC values were calculated from this study to assist with powering and analyzing future research endeavors using this technology. Our findings are consistent with previous BIA reliability studies conducted on varying populations.^19,21,22 Due to its portability, ease of use, low cost, lack of radiation exposure, and high reliability, BIA is an attractive option to assess LBM in orthopedic trauma patients in the clinical setting and in the research setting. Future research in this area should aim to assess the validity of BIA devices specifically in the orthopedic trauma population.