Multiple regression model for ascertaining the skeletal muscle mass index using grip strength and lifestyle factors in older outpatients

Abstract

[Purpose] Skeletal muscle mass index, an essential parameter for diagnosing sarcopenia, necessitates special measurement. Using clinical data that can be easily evaluated through nutrition counselling, we aimed to develop a formula to derive the skeletal muscle mass index. [Participants and Methods] This retrospective study enrolled older outpatients who visited an acute-care hospital for the periodical consultation of comorbidities. The skeletal muscle mass index was measured using the bioimpedance method. Stepwise multiple linear regression was used to clarify the relationship between the skeletal muscle mass index and various factors, including age, sex, height, body weight, the Charlson Comorbidity Index, grip strength, the Barthel Index, and lifestyle factors. [Results] Among the 142 participants of this study, we applied a prediction model that was derived as follows: skeletal muscle mass index (kg/m²)=0.361 × sex (0: female, 1: male) + 0.068 × body weight (kg) −0.065 × Charlson Comorbidity Index (score) + 0.022 × grip strength (kg) + 0.089 × balanced meals per day (3: three meals, 2: two meals, 1: one meal, or 0: no meals) + 0.101 × working activity (1: unemployed at home, 2: housework, 3: desk work, 4: desk/non-desk work, or 5: non-desk work) + 1.549 (R²=0.847). [Conclusion] Dietary habits and working activities correlated with the skeletal muscle mass index. This model may facilitate the calculation of the skeletal muscle mass index in patients whose bioimpedance data are unavailable.

INTRODUCTION

Sarcopenia is a decrease in skeletal muscle mass, muscle weakness, and deterioration in physical performance in older adults¹⁾. In recent years, clinicians and researchers have shown an increased interest in sarcopenia, especially in the field of geriatric medicine. According to an epidemiological study in Japanese older adults, the prevalence of sarcopenia is 9.6% in males and 7.7% in females²⁾. To diagnose sarcopenia, skeletal muscle mass index (SMI) is required to measure using specialized equipment like bioelectrical impedance analysis (BIA) or dual energy X-ray absorptiometry (DEXA). BIA measurement with InBody 770 (InBody Japan Co. Ltd., Tokyo, Japan) has been used in Kanazawa City Hospital to examine sarcopenia, body composition ratio, and nutritional status since 2018. Compared with DEXA, the BIA method is useful in terms of transportability and safety without exposure. However, many patients who should be examined for sarcopenia cannot always have BIA measurements taken in our hospital. The problem, in particular, is that patients must stand motionless for 1 minute during measurement with InBody 770. These measurement conditions are quite difficult for some patients, especially for those with poor activities of daily living (ADL) and/or dementia.

Several studies have explored the prediction of skeletal muscle mass using multiple linear regression analysis^3,4,5). The proposed formulas indicated that skeletal muscle mass is able to be calculated from simple clinical data: age, sex, height, body weight, grip strength, and calf circumference. However, how lifestyle factors influence skeletal muscle mass remains unclear. Lifestyles of older adults show large individual differences; therefore, formula building should include lifestyle factors, such as dietary habits or daily activities.

Our dietitians usually assess the lifestyle factors in nutrition counseling, and these data may be available to calculate SMI. This retrospective study was designed to establish the prediction formula for SMI using grip strength and lifestyle factors in older outpatients.

PARTICIPANTS AND METHODS

The participants comprised older outpatients who visited Kanazawa City Hospital for the periodical examination of comorbidities from April 2021 to June 2022. We included patients with nutrition counseling data and excluded patients under 65 years old and those without BIA data. The representative data, BIA, grip strength, and lifestyle factors were obtained on the same day of the hospital visit.

This retrospective study was approved by the ethics committee of Kanazawa City Hospital (approval number: 503-10-1) and conducted in accordance with the principles of the Declaration of Helsinki. All of the patients individually provided written informed consent before the study entry. The committee approved the requirement of opt-out at any time due to the retrospective nature of the study.

All patient data were retrospectively extracted from the hospital database, including age, sex, patient’s living arrangement, height, body weight, body mass index (BMI), Charlson comorbidity index (CCI)⁶⁾, grip strength, Barthel index, SMI, and lifestyle factors. The CCI is currently the most commonly used for comorbidity assessment. The CCI is a clinical index that predicts mortality risk according to the number and severity of comorbidities, including the 19 items with weighted scoring. Higher CCI scores indicate an increase in the number of comorbidities and poor survival prognosis. The Barthel index is a common assessment scale for physical ADL and includes ordinal values ranging from 0 (total dependence) to 100 (independence) with 5-point scales.

BIA was used to analyze body composition data, e.g., skeletal muscle mass, fat-free mass, total body water, extracellular water, intracellular water, and body cell mass⁷⁾. Our physicians prescribed BIA examination for general medical practice using InBody 770, which was performed using an 8-point tactile electrode system on 5 separate body parts (trunk and extremities) at 6 frequencies of 1, 5, 50, 250, 500, and 1,000 kHz. Patients stood on the foot-shaped electrodes and held cylindrical grip electrodes during BIA measurement. Although some other models, e.g., S10 or BWA, are able to be measured in a supine position, InBody 770 requires patients to stand motionless for at least 1 minute during measurement.

The SMI is converted from appendicular skeletal muscle mass (ASM) by dividing by height squared (kg/m²). According to the Asian working group for sarcopenia (AWGS 2019)¹⁾, cut-off values for sarcopenia using BIA are less than 7.0 kg/m² in males and less than 5.7 kg/m² in females. We usually diagnose sarcopenia based on AWGS 2019: BIA measured-SMI, grip strength, and 5-time chair standing test.

In accordance with prescriptions from our physicians, patients receive nutrition counseling at their hospital visit once every two or three months. Our dietitians routinely estimate height, body weight, grip strength, and Barthel index in each counseling. Grip strength was measured using a Takei grip strength dynamometer Grip-D TKK 5401 (Takei Scientific Co. Ltd., Niigata, Japan). In a standing position with feet shoulder-width apart, patients performed two maximal attempts with their arms positioned straight by their trunk side. The representative data of grip strength was the average of the highest value from two attempts, with either hand.

In addition, information on lifestyle factors, such as balanced meals per day and working activity, were checked at each counseling. Balanced meals were assessed by the number of nutritionally well-balanced diets daily, e.g. those who take two balanced diets and one non-balanced diet daily are categorized as “two meals”. The balanced meals were defined by the nutrition guidelines of the Ministry of Health, Labour and Welfare in Japan⁸⁾. The working activity had been originally developed for nutrition counseling based on our clinical experience. These categories were firstly divided by patient employment status and, if employed, were additionally classified by labor activity: desk work and/or non-desk work. Non-desk work includes standing, walking, carrying, and physically demanding labor almost always without sitting.

We confirmed the power calculation for multiple linear regression analysis using the G*Power software program (ver. 3.1.9.7, Heinrich Heine University, Düsseldorf, Germany) at 90% power and at the 5% significance level to explore effect size. The calculation showed us that the minimum required sample size was 123 for the multiple linear regression analysis with 6 independent predictors. We were, therefore, convinced that the number of patients required was sufficient to obtain statistically valid results.

Results were reported as means ± standard deviation (SD) for parametric variables, whereas medians and 25th-75th percentiles of interquartile range (IQR) were used to describe nonparametric variables. Correlation analyses were conducted to estimate the association between SMI and each variable. Stepwise multiple linear regression analyses were adjusted for age, sex, height, body weight, CCI, grip strength, appetite, balanced meals per day, feeding amount, feeding speed, snack frequency, meal skipping, alcohol consumption, tobacco smoking, standing duration per day, exercise frequency, working activity, and sleeping duration. The categorical variables were set to a dummy number. The reliability of the prediction formula was estimated by coefficient of determination (R²), F test, and residual analyses. Multicollinearity was estimated using the variance inflation factor (VIF); a VIF value of ≥10 was considered as the risk of multicollinearity.

All data were analyzed using IBM SPSS Statistics software program, version 24.0 for Windows (IBM Japan Ltd., Tokyo, Japan). P-values <0.05 were considered statistically significant.

RESULTS

Of the 1,724 patients who had received the nutrition counseling, 142 (78 male and 64 female participants) were enrolled in this study (Supplementary Fig. 1). All of the participants had recorded data of other variables including anthropometric data, grip strength, and lifestyle factors.

Participants’ demographic details and characteristics are displayed in Table 1. BMI is used to classify participants into three body types with the following results: 17 underweight (12.0%, BMI of <18.5), 96 healthy weight (67.6%, BMI of 18.5–24.9), 28 overweight (19.7%, BMI of 25.0–29.9), and 1 obesity (0.7%, BMI of ≥30.0). The ratio of low SMI, weak grip strength, and poor physical activity (5-times chair standing test ≥12 s) based on AWGS 2019 were 47.9%, 49.3%, and 30.3%, respectively. Together, these results showed that 40.8% of the participants were diagnosed with sarcopenia. The patients with CCI of ≥3 scores showed a slightly low rate of 24.6%. Most of the participants presented maximum or high scores on the Barthel index. Analysis of dietary habits showed that less than half of the participants had both of good appetite and three balanced meals per day. In terms of the daily activities, 66.9% of the participants had no exercise habits and three-fourths of the participants were unemployed at home with or without housework. To summarize the results in Table 1, the participants as a whole could be living almost independently; however, some participants did not have healthy diet or sufficient activities in their daily lives.

Table 1. Participant demographics and clinical characteristics (n=142).

Characteristic	Value
Age, years, mean (SD)
Total	76.5 (7.2)
Male	75.9 (7.1)
Female	77.3 (7.3)
Sex, n (%)
Male	78 (54.9)
Female	64 (45.1)
Patient’s living arrangement, n (%)
Patient’s home (living alone)	104 (73.2)
Patient’s home (not living alone)	35 (24.6)
Nursing home	3 (2.1)
Height, cm, mean (SD)
Male	164.6 (7.3)
Female	149.4 (5.8)
Body weight, kg, mean (SD)
Male	61.0 (10.4)
Female	48.5 (8.4)
BMI, kg/m2 (SD)
Male	22.5 (3.2)
Female	21.7 (3.3)
CCI, median (IQR)	1.0 (1.0-2.3)
Low: 0 scores, n (%)	31 (21.8)
Medium: 1–2 scores, n (%)	76 (53.5)
High: 3–4 scores, n (%)	30 (21.1)
Very high: ≥5 scores, n (%)	5 (3.5)
Comorbidity, n (%)
Hypertension	79 (55.6)
Dyslipidemia	56 (39.4)
Chronic liver disease	52 (36.6)
Advanced cancer	52 (36.6)
Biliary disease	46 (32.4)
Diabetes	40 (28.2)
Pancreatic disease	34 (23.9)
Chronic heart disease	28 (19.7)
Chronic kidney disease	19 (13.4)
Hyperuricemia	14 (9.9)
Sarcopenia diagnosis†, n (%)
Non-sarcopenia	84 (59.2)
Sarcopenia	58 (40.8)
SMI, kg/m2, mean (SD)
Male	6.9 (1.0)
Female	5.5 (0.9)
Grip strength, kg, mean (SD)
Male	27.4 (7.3)
Female	17.5 (5.6)
5-time chair standing test
<12 seconds, n (%)	99 (69.7)
≥12 seconds, n (%)	43 (30.3)
Barthel index, scores, median (IQR)	100 (100–100)
100 scores, n (%)	125 (88.0)
95–40 scores, n (%)	17 (12.0)
Appetite, n (%)
Good	58 (40.8)
Fair	68 (47.9)
Poor	16 (11.3)
Balanced meals per day, n (%)
Three meals	47 (33.1)
Two meals	48 (33.8)
One meal	29 (20.4)
No meals	18 (12.7)
Feeding amount, n (%)
Gluttony	15 (10.5)
Almost full	79 (55.6)
Moderate	5 (3.5)
Small	43 (30.3)
Feeding speed, n (%)
Fast	30 (21.1)
Normal	90 (63.4)
Slow	22 (15.5)
Snack frequency, n (%)
Frequently	74 (52.1)
Sometimes	33 (23.2)
Rarely	35 (24.6)
Meal skipping, n (%)
No	119 (83.8)
Yes	23 (16.2)
Alcohol consumption, n (%)
Never	91 (64.1)
Former	15 (10.6)
Current	36 (25.4)
Tobacco smoking, n (%)
Never	108 (76.1)
Former	24 (16.9)
Current	10 (7.0)
Standing duration per day, hour, median (IQR)	2 (1-3)
Exercise frequency, n (%)
No exercise habits	95 (66.9)
Once a week	15 (10.6)
Twice or 3 times a week	13 (9.2)
More than 4 times a week	19 (13.4)
Working activity, n (%)
Unemployed at home	67 (47.2)
Housework	39 (27.5)
Deskwork	16 (11.3)
Desk/non-desk work	8 (5.6)
Non-desk work	12 (8.5)
Sleeping duration, hours, median (IQR)	8.0 (7.0–8.5)

^†Based on Asian Working Group for Sarcopenia (AWGS 2019).

SD: standard deviation; BMI: body mass index; SMI: skeletal muscle mass index; CCI: Charlson comorbidity index; IQR: inter-quartile range (25th–75th percentiles).

Table 2 shows the results of correlation analyses on the relationship between SMI and independent variables. In both male and female participants, SMI was negatively correlated with age and CCI, and was positively correlated with height, body weight, grip strength, and Barthel index. Regarding lifestyle factors, appetite, feeding amount, standing duration per day, and working activity in both males and females were significantly correlated with SMI. Furthermore, feeding speed and alcohol consumption in males and balanced meals per day in females were significantly correlated with SMI.

Table 2. Correlation analysis of SMI with age, height, body weight, CCI, Barthel index and lifestyle factors.

Variable	Male r	Female r
Age (years)	−0.439**	−0.340**
Height (cm)	0.489**	0.498**
Body weight (kg)	0.840**	0.800**
CCI (scores)	−0.305**	−0.258*
Grip strength (kg)	0.611**	0.565**
Barthel index (scores)	0.365**	0.320*
Appetite (good, fair, poor)	0.247*	0.255*
Balanced meals per day (three meals, two meals, one meal, no meals)	0.160	0.228*
Feeding amount (gluttony, almost full, moderate, small)	0.378**	0.398**
Feeding speed (fast, normal, slow)	−0.321**	−0.181
Snack frequency (frequently, sometimes, rarely)	0.038	0.124
Meal skipping (no, yes)	−0.191	−0.042
Alcohol consumption (never, former, current)	0.226*	−0.054
Tobacco smoking (never, former, current)	0.050	0.020
Standing duration per day (hours)	0.275*	0.508**
Exercise frequency (no exercise habits, once a week, twice or 3 times a week, more than 4 times a week)	0.042	0.213
Working activity (unemployed at home, housework, deskwork, desk/non-desk work, non-desk work)	0.357**	0.570**
Sleeping duration (hours)	−0.117	−0.246

*p<0.05, **p<0.01. SMI: skeletal muscle mass index; CCI: Charlson comorbidity index.

Stepwise multiple linear regression analysis was performed with adjustment of variables (Table 3). Body weight and grip strength were selected as independent variables. Moreover, body weight was strongly correlated with SMI. Model 2 excluded the lifestyle factors: SMI=0.071 × body weight (kg) + 0.039 × grip strength (kg) + 1.427 (R²=0.820, adjusted R²=0.818). Although working activity and balanced meals per day were selected in models 3–6, these variables were modestly correlated with SMI. The SMI prediction equation developed was model 6: SMI (kg/m²)=0.361 × sex (0: female, 1: male) + 0.068 × body weight (kg) −0.065 × CCI (score) + 0.022 × grip strength (kg) + 0.089 × balanced meals per day (3: three meals, 2: two meals, 1: one meal, 0: no meals) + 0.101 ×working activity (1: unemployed at home, 2: housework, 3: deskwork, 4: desk/non-desk work, 5: non-desk work) + 1.549 (R²=0.847, adjusted R²=0.840). The p-value for the F test was <0.001. The absolute values of standardized residuals were all less than 3SD. The value indicating residual randomness, the Durbin–Watson ratio, was 1.992 and showed an almost ideal value of 2.0. Each VIF value in model 6 was <10; therefore, we confirmed that there was no risk of multicollinearity in the analysis.

Table 3. Stepwise multiple linear regression analysis on SMI (Model 6).

Independent variable	B	SE	β	p-value	95% CI		VIF
					Lower	Upper
Constant	1.549	0.233		<0.001	1.089	2.009
Body weight	0.068	0.005	0.667	<0.001	0.058	0.077	1.852
Grip strength	0.022	0.007	0.158	0.003	0.008	0.037	2.383
Working activity	0.101	0.035	0.110	0.004	0.032	0.169	1.251
Sex	0.361	0.112	0.156	0.002	0.140	0.582	2.058
Balanced meals per day	0.089	0.040	0.078	0.026	0.011	0.167	1.070
CCI	−0.065	0.031	−0.075	0.040	−0.126	–0.003	1.161

SMI: skeletal muscle mass index; SE: standard error; β: standardized partial regression coefficient; CI: confidence interval; VIF: variance inflation factor; CCI: Charlson comorbidity index.

DISCUSSION

The present study identified that SMI could be calculated from 6 predictors: sex, body weight, grip strength, CCI, balanced meals per day, and working activity. We found that the patients with higher grip strength, lower CCI scores, a more balanced diet, and higher activity levels in working were able to increase SMI value. These lifestyle factors contributed to improving the reliability of the prediction formula for SMI in older outpatients.

Several studies have explored the prediction formulas for SMI or ASM using grip strength. Hiruta et al. examined 1,003 patients (aged 18–96 years: 525 males and 478 females) and proposed the best equation: SMI (kg/m²)=−0.27799 × sex (0: female, 1: male) + 0.02819 × grip strength (kg) + 0.05234 × body weight (kg) + 2.84640 (R²=0.830)³⁾. Hsiao et al. investigated 1,020 healthy older adults and the equation: ASM (kg)=−9.833 + 0.397 × body weight (kg) + 4.433 × sex + 0.121 × height (cm) + 0.061 × grip strength (kg) best predicts DEXA-measured ASM (adjusted R²=0.914)⁴⁾. Although these previous studies have investigated several participants, lifestyle factors have not been analyzed, and under-middle-aged adults have been included in Hiruta’s study. However, since we limited our study to older patients, lifestyle factors, such as a decline in physical abilities or poor dietary habits, might influence the equation result.

Kawakami examined 1,262 participants ranging from middle-aged to elderly, and the most optimal ASM prediction equation developed is: ASM (kg)=2.955 × sex (0: female, 1: male) + 0.255 × body weight (kg) −0.130 × waist circumference (cm) + 0.308 × calf circumference (cm) + 0.018 × height (cm) −11.897⁵⁾. Although the previous study did not use grip strength for analysis, the coefficient of determination shows quite high value (adjusted R²=0.94). Recent trends in using calf circumference (CC) have led to a proliferation of sarcopenia studies. CC is positively correlated with ASM, and can be used as a surrogate marker of muscle mass for sarcopenia diagnosis⁹⁾. Regrettably, CC is not usually measured for outpatients in our hospital. Nevertheless, CC is easy to measure, so more research using CC is required to examine the influence on the prediction formula.

In a systematic review reported by Lunt et al., grip strength is the most studied measure and is associated with mobility, balance and ADL outcomes¹⁰⁾. Weak grip strength is a strong predictor of poor patient outcomes such as longer hospital stays, increased functional limitations, poor health-related quality of life and poor survival outcomes^11,12,13). According to a recent study conducted in a large adult population, grip strength was found to be associated with SMI, lean mass, fat mass, cardiorespiratory fitness, bone mineral density, android/gynoid ratio, disease prevalence, and physical activity levels¹⁴⁾. In older Asian adults, grip strength is an accurate proxy marker for muscle mass, particularly in males¹⁵⁾. Taken together, these studies support the notion that grip strength is strongly associated with skeletal muscle mass, muscle strength, and physical performance, and is a surrogate marker for calculating SMI. Although its influence on formula building was not as strong as body weight, our study indicated that grip strength was the second most significant variable for predicting SMI.

Few studies have been conducted on the association between comorbidities and skeletal muscle mass. Among individuals aged 50 and older in the United States, a variety of comorbid medical conditions serve as independent predictors of lower muscle strength (e.g., diabetes, coronary heart disease/congestive heart failure, vision problems) and/or modify the relationship between muscle mass and muscle strength (e.g., obesity)¹⁶⁾. A recent study has shown that older patients (aged 83.34 ± 7.32 years) with high CCI are more likely to be in the sarcopenic group than in the non-sarcopenic group, and there is a linear correlation between the CCI and SMI (r=−0.549, p<0.05)¹⁷⁾. In this study, we identified that CCI independently predicted SMI, and patients with higher CCI were prone to be in the lower SMI group, and were thus at higher risk of sarcopenia.

Although some studies have examined the association between lifestyle factors and health outcomes, there has been almost no published information on SMI prediction using lifestyle factors. Okada et al. showed that older Japanese adults with higher SMI consume more energy and nutrients and more vegetables than those with a lower SMI¹⁸⁾. In a systematic review reported by Caso et al., the food choices of independent older adults are shaped by a multitude of factors and sub-factors that may serve to promote or limit the desire and ability to consume healthy food¹⁹⁾. According to a prevention study in Germany, older adults with irregular eating, especially those single and living alone, have a poorer diet and poorer health outcomes²⁰⁾.

In a systematic review reported by Wang et al., muscle mass among older adults is predictive of ADL and instrumental ADL decline²¹⁾. Zhong et al. reported that fall experience in the previous years, low body mass index, and a low level of physical activity are associated with a higher incidence of sarcopenia²²⁾. According to a nationwide survey of the Korean population aged ≥60 years, there is a strong relationship between physical activities and SMI and grip strength, and the results suggest that high-intensity physical activities may have protective effects against sarcopenia²³⁾.

Our findings suggest that healthy dietary habits and higher intensity activities in daily living had an effect on preventing skeletal muscle loss among older outpatients. However, these predictive abilities were limited compared to body weight. A possible explanation for this might be that our nutrition counseling includes unvalidated methods. To improve the accuracy of formula building with life style factors, more reliable assessments, e.g., simplified nutritional appetite questionnaire (SNAQ)²⁴⁾ or international physical activity questionnaire (IPAQ)²⁵⁾, should be adopted in a usual nutrition counselling. Nevertheless, the lifestyle in older patients was suggested to be a necessary factor for SMI prediction. Analysis using only anthropometric factors is not satisfactory because older adults are prone to unhealthy diets and decreased physical functions. The insights gained from this study may be of assistance to predict SMI in older adults.

Although the findings should be interpreted with caution, this study has several strengths. The grip strength and lifestyle factors are easy to obtain in typical clinical settings and have the potential for use in prediction of SMI. The proposed equation may help in calculating SMI even in patients without BIA data and be useful as an adjunct to the diagnosis of sarcopenia.

Several additional limitations should be acknowledged. First, there is a possibility of selection bias due to our small sample size because the BIA examination was prescribed by a limited number of gastroenterologists in a single hospital. Meanwhile, other physicians, such as pulmonologists or cardiologists had prescribed almost no BIA examination during the research period. Thus, the findings may not be generalizable to different populations. Second, the nutrition counseling methods are generally less well-validated than those in planned prospective cohorts. Future work can include prospective validation studies to improve SMI prediction accuracy using the reliable questionnaire survey identified in the present study. Third, we could not assess cognitive functions; however, almost all of the participants could comprehend and answer the questionnaires. Fourth, it is possible that the study failed to detect a significant difference in individuals with and without psychosocial problems. However, research conducted in a single prefecture in Japan, neighborhood environmental factors had limited effects on SMI and grip strength among rural older adults²⁶⁾. Finally, it is possible that our study population included some participants with overhydration; however, dialysis patients had not been included in the prescription target of BIA.

In conclusion, our findings suggest that grip strength and lifestyle factors correlate with SMI in older outpatients. The equation may help calculate SMI in patients without BIA data. Further research is required to examine the association between SMI and lifestyle factors more closely.

Authorship

Conceptualization, supervision, data analysis and writing, H.O.; data collection, Y.O.; review and editing, H.O., Y.O., T.S., T.K., K.K., H.A. and T. Y.; project administration, E.M.; English proofreading, P.I.

Conference presentation

Main part of this research was presented at the 10th Annual Meeting of Japanese Society of Nutrition and Swallowing Physical Therapy (JSNSPT 2024).

Conflict of interest

No conflict of interest to be declared.