Developing a Screening Tool for Sarcopenia in Hospitalized Geriatric Patients: Estimation of Appendicular Skeletal Muscle Mass using Bioelectrical Impedance

Abstract

Background and Aims.

Sarcopenia is now a billable ICD-10 geriatric condition characterized by low appendicular skeletal muscle mass (ASMM) and low function. There is an increasing need for portable, provider-friendly, cost-effective methods for estimating ASMM. The overall goal of this project is to create and validate a regression model for obtaining ASMM from Bioelectrical Impedance Analysis (BIA) measurements using Dual-energy X-ray Absorptiometry (DXA) as the reference.

Methods.

Geriatric patients (≥65 years of age) were enrolled during an acute hospitalization. Body composition measurements were obtained through DXA and BIA devices. The ASMM prediction model was derived using stepwise multiple regression modelling. The model was 10 fold validated and tested as a screening tool (sensitivity, specificity, positive predictive and negative predictive values) using the Foundation of NIH Sarcopenia Project (FNIH) definition.

Results.

The following variables were selected by stepwise regression modeling: sex, body mass index, max grip strength, and fat mass derived by BIA. The model was internally validated with 10 fold cross validation. Using the FNIH definition, the model was found to have a sensitivity of 80%, a specificity of 91%, a positive predictive value of 73% and a negative predictive value of 93%.

Conclusions.

We have developed a screening tool that can easily be used in geriatric patients to screen for sarcopenia. Once validated with a larger sample, the developed prediction model can be used to estimate ASMM using provider-friendly measurements and can be easily implemented as a sensitive screening tool for identifying patients at risk for sarcopenia. Those identified at risk would undergo further functional testing for diagnosis and treatment for sarcopenia.

INTRODUCTION:

Over 20% of hospitalized geriatric patients suffer from sarcopenia(1, 2). This age-associated loss of skeletal muscle mass and function is exacerbated by catabolic stressors such as hospitalization, illness, or injury(3). The causes of sarcopenia are multifactorial and associated with malnutrition, endocrine changes, neurodegenerative diseases, disuse and atrophy of the muscles(4). The ability to prevent, diagnose and treat sarcopenia is important as a substantial loss of muscle can result in functional impairment and disability leading to multiple comorbidities and mortality(5). Several international consensus panels including Foundation of NIH Sarcopenia Project (FNIH)(6), International Working Group Sarcopenia Initiative (IWGS)(3), and European Working Group on Sarcopenia in Older Persons (EWGSOP2)(4, 7) have established operational definitions for sarcopenia. All three definitions utilize a measure of function (gait speed, handgrip strength) plus the measure of appendicular skeletal muscle mass (ASMM). Cross-sectional data has shown that a larger decrease in ASMM occurs with aging when compared to non-appendicular muscle mass(8), making ASMM a good representation of the extent of lost muscle mass.

With the growing awareness of the prevalence of sarcopenia and addition of the syndrome to the ICD-10 coding system(9) there is an increasing need for a portable and cost-effective approach to measuring ASMM and sarcopenia status in elderly patients. ASMM can be measured in multiple ways: dual energy X-ray absorptiometry (DXA), magnetic resonance imaging, computerized tomography (CT), anthropometry, ultrasound, and bioelectrical impedance analysis (BIA)(10, 11). DXA radiologic scans provide precise regional assessments of appendicular muscle mass and are often considered to be the reference standard for the measurement of ASMM(12). BIA devices are commonly used to estimate body composition because they are portable, non-invasive, inexpensive, safe, and require minimal maintenance and training(13). BIA devices estimate body composition by measuring tissue impedance to an alternating current passed through the body. Devices come in various complexities with all devices measuring electrical properties of the body, reactance and resistance. However, not all devices report these measures. Some models, mostly more affordable ones, implement proprietary computations based on prediction models to report only estimations of body composition. The prediction models are formulated based on studied populations and are the most accurate in these specific populations. When compared to DXA, the reference standard, BIA has been reported to be less accurate in the estimation of body composition when compared to DXA(14).

Earlier studies have validated the use of BIA estimates against DXA in determining ASMM through prediction models(15, 16). Previous models implemented measures of reactance and resistance obtained through “hand to foot” BIA devices that required the application of electrodes. Such devices increase the facility cost, require additional steps, and decrease the overall usability of BIA devices in the acute hospital settings. Currently, there are no BIA-determined prediction models for estimating ASMM in acutely ill and hospitalized geriatric patients. Moreover, there is limited research completed on determining ASMM from BIA derived fat mass (FM_BIA), measured by a single frequency “foot to foot” device (resembling bathroom weight scales). Prediction models that use single frequency BIA derived measurements could provide useful data for evaluating acutely ill geriatric patients sarcopenia status.

Thus, the overall goal of this project is to create a regression model for obtaining ASMM from a “foot to foot” BIA measurements using DXA as the reference. Our long-term goal is to develop a tool to help physicians screen for and diagnose sarcopenia in their acutely ill geriatric patients with simple and cost-effective devices.

METHODS:

Research Design.

All data was collected from a single university hospital from January 2013 to January 2018. Data was pooled from three ongoing studies to establish the model. All studies were approved by the University of Texas Medical Branch (UTMB) Institutional Review Board (IRB # 13–038, 14–0527, and 16–0146). Patients eligible for inclusion met the following criteria: 65 years or older, English speaking, able to provide consent, ability to stand upright, and testing could be completed within 72 hours of admission. Written informed consent was obtained from each subject prior to any study procedure.

Testing.

Testing occurring during acute hospitalization. Testing measures included: (1) demographics (age, gender, race; collected from electronic medical records), (2) body composition measures of fat mass (FM) and ASMM of legs and arm (dual-energy X-ray absorptiometry (GE Lunar iDXA) and estimated %FM by bioimpedance (Tanita; BF-350), (3) anthropometric measures (height, weight, mid-arm circumference, triceps skinfold thickness, calf circumference), (4) laboratory measures (creatinine, albumin, blood urea nitrogen; collected from electronic medical records), (5) physical function tests (usual gait speed and bilateral hand grip strength), (6) questionnaires (activities of daily living; ADL, and instrumental activities of daily living; IADL), and (7) chart review (admission history). Detailed protocols of baseline testing have been published previously (17).

Statistical Analysis.

Data was analyzed using IBM SPSS Version 24 and Stata Version 14. A p-value <0.05 was considered statistically significant.

Subjects Characteristics:

Descriptive statistics were reported for age, sex, ethnicity, race, grip strength and body composition measures. ASMM was calculated as the addition of lean mass of the arms and legs measured by DXA (ASMM_DXA) (12). Grip strength was established as the maximum obtained from left or right-hand attempts. Significant differences in subject characteristics between females and males were explored using unpaired t-test.

Derivation of the prediction regression:

Scatterplots for linearity were run to determine variables associated with ASMM. FM_BIA, age, gender, max grip strength, gait speed, body mass index (BMI), creatinine, albumin, blood urea nitrogen, triceps skinfold, mid arm and calf circumference were selected as potential model variables. Stepwise multiple regression modelling was chosen to derive the ASMM prediction model with the following selection criteria: p value for inclusion of ≤0.05 and exclusion of ≥0.10. The statistics assumptions of regression model including normality of error distribution, homoscedasticity of the errors, and collinearity were checked by residual plots and variance inflation factor. We also assessed whether there were outliers or influential cases by leverage statistics, Cook’s distance, and DFFITS.

Comparison of ASMM measured by DXA and regression model:

Bland-Altman analysis was used to determine the level of agreement between ASMM obtained through the regression model (ASMM_BIA) and DXA (ASMM_DXA). The mean difference, standard deviation and the 95% confidence interval were reported. The performance of the predictive model was assessed using 10 fold cross validation(18, 19). The 10 group R², mean absolute error and root mean square error were reported. We used Bland-Altman plots to visually present the agreement between the prediction of ASMM from our model and ASMM measured by DXA. Since our sample is not large, we used 10-fold cross-validation to estimate the prediction error of our model.

Testing the model as a screening tool for Sarcopenia using Foundation for the NIH Sarcopenia and European Working Group on Sarcopenia in Older People Definitions:

The Foundation of NIH Sarcopenia Project (FNIH) definition for weakness (hand grip: male<26kg; female<16kg) and low lean mass (ASMM: male<19.75 kg; female<15.02 kg) and the European Working Group on Sarcopenia in Older People (EWGSOP2) definition for weakness (hand grip: male<27kg; female<16kg) and low lean mass (ASMM:/ht² male<7 kg; female<6 kg) was used to determine the test subject’s sarcopenia status. The ASMM measure used in the definition was determined through the study’s prediction model (ASMM_BIA) and then compared to DXA derived measure of ASMM (ASMM_DXA). The sensitivity, specificity, positive and negative predictive values were calculated.

RESULTS:

Subjects Characteristics:

From January 2013 to January 2019, 229 subjects were enrolled in the study. 65 patients were excluded because the patient was discharged prior to the completion of the measurements. 25 were excluded because they were unable to stand. Of the 139 patients that met the inclusion/exclusion criteria, 14 were excluded due to missing BIA, DXA, and/or max grip data. The final sample size consisted of 125 subjects and their characteristics are shown in Table 1. The sample included 85 females and 40 males, with average age of 77 ± 7 years. Similar to our previous work in this population(20), discharge diagnosis for the sample was primarily cardiovascular (26%), pulmonary (23%) or gastrointestinal (19%). On average the subjects had a FM_BIA of 35.55%, FM_DXA of 39.50% and ASMM_DXA of 18.35kg. Height, weight, FM_BIA, FM_DXA, ASMM_BIA, and ASMM_DXA, did differ significantly between males and females.

Table 1.

Characteristics of study participants expressed as frequencies or means. Minimum and maximum values are given in brackets.

	N	Whole Sample	N	Male	N	Female	p-valuea
Age, yrs	125	78.72 ± 7.60 (65 – 98)	40	78.98 ± 6.28 (70 – 93)	85	78.60 ± 8.18 (65–98)	ns
Female, %		68.00%
White, %		85.60%		87.50%		84.71%
Hispanic, %		8.80%		5.00%		10.59%
Body Composition
Height, cm	125	163.94 ± 8.85 (142.40 – 190.50)	40	171.77 ± 8.25 (152.10 – 190.50)	85	160.25 ± 6.41 (142.40 – 177.80)	<0.001
Weight, kg	125	75.84 ± 17.08 (43.80 – 124.30)	40	86.31 ± 16.54 (62.30 – 124.30)	85	70.91 ± 15.07 (43.80 – 122.80)	<0.001
BMI, kg/m2	125	28.16 ± 5.69 (16.73 – 44.28)	40	29.25 ± 5.19 (19.44 – 40.87)	85	27.65 ± 5.86 (16.73 – 44.28)	ns
FMBIA, %	125	35.55 ± 8.81 (14.70 – 53.70)	40	29.52 ± 7.90 (14.70 – 46.50)	85	38.39 ± 7.75 (16.90 – 53.70)	<0.001
FMDXA, %	125	39.50 ± 7.55 (18.52 – 58.48)	40	35.51 ± 6.84 (18.52 – 49.05)	85	41.38 ± 7.15 (19.74 – 58.47)	<0.001
ASMMDXA, kg	125	18.35 ± 4.60 (9.41 – 33.53)	40	22.61 ± 3.86 (15.30 – 33.53)	85	16.34 ± 3.40 (9.41 – 26.76)	<0.001
ASMMBIA, kg	125	17.63 ± 3.56 (11.71 – 27.11)	40	21.46 ± 2.33 (17.62 – 27.11)	85	15.83 ± 2.43 (11.71 – 23.27)	<0.001
Physical Function
Grip strength, kg	125	19.08 ± 8.82 (1 – 38.5)	40	26.88 ± 8.21 (10 – 38.5)	85	15.41 ± 6.40 (1 – 33.7)	<0.001
Gait Speed, m/s	123	0.65 ± 0.27 (0.14 – 1.55)	40	0.69 ± 0.30 (0.18 – 1.55)	83	0.63 ± 0.26 (0.14 – 1.26)	ns
Antropoimetric Measures
Tricep Skinfold, mm	75	14.78 ± 10.79 (2.00 – 45.00)	23	12.54 ± 10.21 (2.33 – 45.00)	52	15.77 ± 10.99 (2.00 – 41.67)	ns
Mid Arm Circumference, cm	75	26.43 ± 5.41 (15.92 – 43.13)	23	28.18 ± 4.76 (19.36 – 35.18)	52	25.65 ± 5.54 (15.92 – 43.13)	ns
Calf Circumference, cm	75	34.64 ± 3.77 (26.40 – 44.53)	23	35.54 ± 3.91 (30.93 – 43.13)	52	34.23 ± 3.68 (26.40 – 44.53)	ns
Laboratory Results
BUN, mg/dL	123	23.15 ± 13.82 (5 – 78)	40	25.75 ± 13.90 (5 – 78)	83	21.90 ± 13.69 (7 – 78)	ns
Creatinine, mg/dL	123	1.19 ± 0.95 (0.32 – 9.90)	40	1.46 ± 1.43 (0.63 – 9.90)	83	1.06 ± 0.55 (0.32 – 3.60)	ns
Albumin g/dL	79	3.73 ± 0.49 (2.10 – 5.00)	26	3.61 ± 0.38 (2.90 – 4.20)	53	3.79 ± 0.53 (2.10 – 5.00)	ns

ASMM_BIA: appendicular skeletal muscle mass derived from the model; ASMM_DXA: appendicular skeletal muscle mass measured by DXA; BIA: bioelectrical impedance analysis; BMI: body mass index; BUN: Blood Urea Nitrogen; DXA: dual-x-ray absorptiometry; FM_BIA: Fat mass measured by BIA; FM_DXA: Fat mass measured by DXA.

^afor significant differences between women and men, n.s. not significant (p >0.001)

Derivation of the prediction regression:

Scatterplot analysis revealed FM_BIA, max grip strength and BMI had a linear relationship with ASMM_DXA. A stepwise multiple regression model was used to derive the ASMM_BIA. The prediction model included the following variables: gender, BIA measured FM%, max grip strength, and BMI (Table 2). The following model was calculated: ASMM_BIA= 7.1 + (−2.8* Gender (male = 0; female = 1) + (0.5* BMI) + (0.1* Max Grip Strength) + (−0.1* FM_BIA²). R² and root mean square deviation (RMSD) for the equation was 0.69 and 2.56, respectively. Table 2 demonstrates that gender, the main contributor in our model, explained 40.4% of the variance. FM_BIA explained an additional 1.3% of the variance seen in ASMM_DXA. There were no violations of statistics assumptions in our prediction model. We also did not detect any outliers or influential cases in our model.

Table 2.

Multiple linear regression model for predicting ASMM_DXA from FM_BIA, gender, BMI, and grip strength

	Coefficient	SE	p-Value	R2 cumulative	RMSD
Constant	7.145	1.555	<0.001	---	---
Gender (male = 0; female = 1)	−0.284	0.780	<0.001	0.404	3.552
BMI, kg/m2	0.504	0.061	<0.001	0.631	2.795
Grip strength, kg	0.146	0.033	<0.001	0.677	2.613
FMBIA, %	−0.108	0.044	<0.001	0.69	2.561

Equation: 7.1 + (−2.8* Gender)+(0.5* BMI)+(0.1* Max Grip Strength)+ (−0.1*BIAfat)

ASMM_DXA: appendicular skeletal muscle mass measured by DXA; BIA: bioelectrical impedance analysis; BMI:body mass index; DXA: dual-x-ray absorptiometry; FM_BIA: Fat mass measured by BIA; SE: standard error of the coefficient; RMSD: Root mean square deviation.

Comparison of ASMM measured by DXA and regression model:

The Bland-Altman plot showed the regression model to underestimate ASMM with a mean difference of −0.0036± 2.407 kg (p>0.05) when compared to the mean difference of zero (Figure 1). The 10 Fold Validation revealed a 10 group R2 of 0.690 ± 0.008, mean absolute error of 2.065 ± 0.3279 and root mean square error of 2.559 ± 0.047.

Figure 1. Bland Altman of model derived ASMM compared to DXA derived ASMM.

The solid line represent the mean difference in ASMM determined by prediction model compared and DXA. The dotted line represents the 95% confidence internal of this difference. ASMM_DXA: appendicular skeletal muscle mass measured by DXA; ASMM_BIA: appendicular skeletal muscle mass derived from prediction model

Testing the model as a screening tool for Sarcopenia using Foundation for the NIH Sarcopenia and European Working Group on Sarcopenia in Older People Definitions:

Prevalence of sarcopenia was 24% when using FNIH or EWGSOP2 cutpoints as the standard, with no differences between the two standards. The prevalence of sarcopenia in the study population determined by the prediction model was 26%. Our model had a sensitivity of 80%, a specificity of 91%, a positive predictive value of 73% and a negative predictive value of 93% (Table 3).

Table 3.

Sarcopenia Status determined using FNIH or EWGSOP2 Definitions

		DXA Determined ASMM
		(+) Sarcopenia	(−) Sarcopenia
Model Determined ASMM	(+) Sarcopenia	24	9
	(−) Sarcopenia	6	86
Sensitivity	80%	Positive Predictive Value	73%
Specificity	91%	Negative Predictive Value	93%

FNIH: Foundation of NIH Sarcopenia Project; EWGSOP2: European Working Group on Sarcopenia in Older People.

DISCUSSION:

In this study we established a new equation to estimate appendicular skeletal muscle mass in acutely ill hospitalized patients. Using either the FNIH or the EWGSOP2 definitions of sarcopenia, the model had a sensitivity of 80% and a specificity of 91% for diagnosing patients with sarcopenia.

The addition of sarcopenia into the ICD-10-CM codes, M62.84, has established this geriatric syndrome as a reportable medical disease. The growing awareness of the prevalence of sarcopenia among health professionals has resulted in an increasing need for a screening tool. The diagnosis of sarcopenia is made using ASMM and a measure of function (handgrip strength or gait speed). Estimating ASMM and diagnosing sarcopenia at hospital admission in acutely ill geriatric patients is important because hospitalization can rapidly accelerate muscle loss and lead to functional impairment if not addressed. ASMM is typically measured by DXA or other imaging method like CT or ultrasound. However, these methods are relatively expensive and difficult to use in the hospital setting due to the need for scheduling instrument time and transporting the patient to the exam room. The portability and cost-effectiveness of single frequency ‘foot to foot’ BIA devices make them a suitable tool for use in a hospital setting to quantify body composition. The use of our prediction equation allows for early diagnosis of sarcopenia.

BIA scales report body composition measures of fat free mass, fat mass and electrical resistance (resistance and reactance). The single frequency ‘foot-to-foot’ device used in the study reports FM% obtained through patented device calculations. Thus, the model presented in this study incorporated the FM% measure in the estimation of ASMM.

The variance of 69% seen with our prediction model was lower than in previous ASMM models developed by Kyle et al(15) and Sergei et al(16): 95% and 92% respectively. Both studies implemented the raw measures of resistance and reactance obtained through “hand to foot” BIA devices that required the application of electrodes. However, these devices increase the facility cost due to being more technologically complex, requiring the use of electrode pads and specialized personnel training. The decrease in variance seen in our study can be attributed to the extra proprietary computations done by “foot to foot” BIA devices in specific patient populations to determine FM as opposed to reporting the raw measurements of electrical properties of the body such resistance and reactance. Thus, resulting in an increase in the error seen in our model. In addition, the ‘foot to foot’ single frequency device used in our study only requires a patient to stand on it allowing for rapid screening in the acute hospital settings.

The model we proposed in this study is also clinically valuable as it does include patients with chronic comorbidities comparable to the modeling done by Sergi et al (16), Reiss et al (21), and Reiter et al (22). Geriatric patients with chronic conditions have a higher incidence of sarcopenia(3). Catabolic stress induced by chronic conditions contribute to the high prevalence of sarcopenia seen in the geriatric population(3).

The main limitation of our study was the exclusion of patients who could not stand unsupported on the BIA device. This criterion excluded about 10% of our sample. The exclusion of these patients likely resulted in an underestimation of the prevalence of sarcopenia in our sample. In order to determine the sarcopenic status of hospitalized patients that are unable to stand different methods would have to be considered (e.g. electrode bioimpedance instruments, ultrasound or CT scan). However, the potential to quickly and easily screen 90% of the hospitalized patients would provide clinical evidence for the implementation of specific interventions.

The study also did not control for patients’ total body water, which could have altered the electrical conductance analysis done by the device. Variability of the patients’ water content could be due to use of diuretics, edema, consumption of water prior to testing, or fullness of the bladder(23). We purposely limited the exclusion criteria of this study to increase the clinical generalizability of our results, enrolling patients whose diet, exercise, and hydration status were not controlled for and could have been affected by the underlying diseases. However, the aim of this study was not to determine the prevalence of sarcopenia in hospitalized patients but to develop a model that could be broadly applied in hospitalized geriatric patients.

In conclusion, our model allows for the estimation of ASMM in acutely ill geriatric patients using simple and cost effective ‘foot to foot’ BIA device. The clinical benefit of implementing this equation in hospitals would be to diagnose patients with sarcopenia, which is the first step for implementation of targeted interventions to reduce the risk of functional dependence. Longitudinal studies are also warranted to determine the validity of this novel method for monitoring sarcopenia progression or the efficacy of targeted interventions.

Abbreviations:

ASMM

Appendicular Skeletal Muscle Mass

BIA

Bioelectrical Impedance Analysis

BMI

Body Mass Index

BUN

Blood urea nitrogen

DXA

Dual Energy X-ray Absorptiometry

EWGSOP

European Working Group on Sarcopenia in Older Persons

FM

Fat mass

FM_BIA

Fat mass measured by Bioelectrical Impedance Analysis

FM_DXA

Fat mass measured by DXA

FNIH

Foundation of NIH Sarcopenia Project

IWGS

International Working Group Sarcopenia Initiative