Prediction of phase angle based on body composition: Insights from a large population-based registry

Abstract

BACKGROUND:

Phase angle (PhA) is obtained through bioelectrical impedance analysis (BIA) and is a potential marker of nutritional health and cellular function. We aimed to develop an equation for estimating PhA using body composition parameters, focusing on skeletal muscle mass, body water, and fat mass.

MATERIALS AND METHODS:

This cross-sectional study enrolled participants from the Iranian National Obesity Registry. Data on age, gender, body composition, and BIA measurements were collected using an InBody 770 device. Exclusion criteria included pregnancy, pacemakers, implants, or edema. The study was approved by the Ethics Committee of Mashhad University of Medical Sciences.

RESULTS:

A total of 5980 participants (28.7% men, 71.3% women) were included. PhA showed the strongest correlations with extracellular water/total body water (ECW/TBW) (r = −0.84, P < 0.001), followed by intracellular water (ICW) (r = 0.62, P < 0.001) and skeletal muscle mass (SMM) (r = 0.62, P < 0.001). Linear regression models revealed a significant negative association between ECW/TBW and PhA (β = −82.81, P < 0.001), explaining 71% of variance. Adding age (β = 0.012, P < 0.001) improved the model (R² = 0.75), and SMM (β = 0.039, P < 0.001) further increased explanatory power (R² = 0.85). Including body fat mass (BFM) (β = 0.005, P < 0.001) and height (β = -0.031, P < 0.001) led to the highest model fit (R² = 0.90).

CONCLUSION:

ECW/TBW, SMM, BFM, age, and height are significant predictors of PhA, with ECW/TBW showing the strongest negative association.

Introduction

Bioelectrical impedance analysis (BIA) has identified age-related changes in body composition, with resistance (R) and reactance (Xc) being the two main components.[1] The phase angle (PhA) is derived using these values and provides key insights into the condition of cell membranes.[2] Recently, PhA has been proposed as a potential marker for assessing nutritional and cellular health,[3] and a potential screening tool to evaluate different biomarkers, cardiovascular risk, and nutritional diagnosis.[4] Recent systematic reviews and cohort studies have supported the possible use of PhA to predict adverse outcomes in metabolic diseases, psychogeriatric patients, hospitalized individuals, and patients receiving home enteral nutrition.[4,5,6,7,8] Specifically, lower values of PhA were found to be associated with increased risk of mortality, longer hospital stays, and greater nutritional risk, independent of traditional clinical parameters.[5,6] Furthermore, PhA may be a prognostic marker for patients with disease-related malnutrition and frailty.[4,7] According to research, PhA is becoming an increasingly popular and cost-effective screening tool—especially when combined with basic BIA data—for identifying individuals with impaired nutritional status across various populations.[2,4,9]

The PhA is physiologically related to age, tending to fall with age, mostly because parallel reactance decreases as muscle mass decreases.[10] Additionally, there is a corresponding rise in resistance, which is linked to the reduction in total body water (TBW) and an increase in adipose tissue.[9,11] PhA was independently correlated with age, skeletal muscle mass index, and muscle quality in both men and women, which was assessed using handgrip strength, as revealed by multiple linear regression analysis.[12] Using multivariate regression analysis, several studies have demonstrated that height has the greatest influence on PhA, with body weight and muscle mass following in significance.[13,14]

Key indicators of physical development include weight and height, as they reflect the growth of bones and muscles. These factors, when combined, contribute to an increase in fat-free mass (FFM), which, in turn, affects both resistance and PhA.[15] Body mass index (BMI) is a parameter used to categorize adiposity, which may be associated with type 2 diabetes, hypertension, and dyslipidemia.[16] The relationship between PhA and BMI remains inconclusive, with some studies reporting a positive correlation, while others have reported a negative correlation.[17] A study observed that PhA typically increased with BMI up to 30 kg/m², but then declined once BMI values exceeded 40 kg/m² in both males and females.[18] However, an increase in body fat percentage, particularly visceral fat mass, was linked to a decrease in PhA.[17]

Despite growing interest in PhA as a potential screening tool to evaluate different health-related conditions, practical challenges remain in its assessment owing to the high cost of devices and limited accessibility of the required measurement devices. There appear to be no previously published studies providing an equation for PhA using body composition parameters, especially in Iranian population. This study addresses this gap by developing a simple, accurate predictive equation for PhA based on measured body composition parameters in a large Iranian population. So, this study explores the predictive value of body composition components and some anthropometrics to predict PhA in a large Iranian sample.

Materials and Methods

Study design and setting

This cross-sectional study was conducted on participants who took part in the Iranian National Obesity Registry (IRNOR) in Mashhad from 2017 to 2023.[19,20] The study took place in the health clinics of Mashhad, Iran, and was approved by the Research Committee of the Faculty of Medicine at Mashhad University of Medical Sciences (ID: 4020599).

Study participants and sampling

The Fisher z-transformation sample size formula was used to calculate the required sample size. The body fat percentage correlation coefficient from the study by Jaremkow was used in the formula (r = −0.432).[21] Finally, the minimum sample size equal to about 90 individuals was calculated with confidence and power of 99% and 90, respectively. Since the accuracy of the study improves with larger samples, we used all available data from the obesity registry, which included 5980 individuals.

The first visit of each participant was selected for further investigation. Among these data, participants meeting the exclusion criteria were excluded from further statistical analysis. The exclusion criteria were pregnancy and any conditions that would contraindicate the use of BIA devices, which included, but were not limited to, the presence of pacemakers, artificial implants, or obvious edema in the body. At the end, we removed individuals under the age of 18 years from the dataset.

Data collection tool and technique

An InBody 770 device (Biospace Co., Ltd., Seoul, South Korea) was used to measure body composition. Body composition components derived from BIA were body fat mass (BFM), body fat percentage (BFP), fat-free mass (FFM), skeletal muscle mass (SMM), total body water (TBW), extracellular water (ECW), intracellular water (ICW), body cell mass (BCM), extracellular water/total body water ratio (ECW/TBW), soft lean mass (SLM), fat mass index (FMI), fat-free mass index (FFMI), protein, waist-to-hip ratio (WHR), and visceral fat level (VFL). Before the BIA, participants had to follow specific instructions, which were not drinking alcohol for 12 h, not drinking caffeinated drinks for 8 h, and not exercising or going to the sauna for 8 h. They were also asked to take off metal and magnetic jewellery, electronic devices, and cell phones.[22,23]

Ethical consideration

The research followed ethical standards for studies involving human subjects. The Ethics Committee of Mashhad University of Medical Sciences’ Faculty of Medicine approved the study protocol (IR.MUMS.Medical.REC.1402.249). Each participant provided written informed consent before inclusion.

Statistical analysis method

Descriptive statistics, including absolute and relative frequency tables, were used for qualitative variables, while quantitative variables were summarized via means and standard deviations. Pearson’s correlation was applied to determine the correlation between potential variables. A linear regression analysis was performed to explore the relationship between PhA and the independent variables. Variance inflation factor (VIF) and tolerance values were used to assess multicollinearity, and variables with high multicollinearity were either removed or combined, as needed. Model fit was assessed using R-squared (R²) and adjusted R-squared, and the significance of individual predictors was tested through P values. SPSS software (IBM Corporation, version 27.0.1) was used to analyze the data.

Results

This study included 5980 participants, including 28.7% (n = 1717) men and 71.3% (n = 4263) women, with a mean age of 35.4 ± 12.9 years. The average weight was 76.6 ± 21.8 kg, and the mean BMI was 28.8 ± 7.9 kg/m². Participants were categorized into BMI groups as follows: 11.8% were classified as underweight, 20.5% as normal weight, 22.7% as overweight, and 45% as obese. Table 1 shows detailed demographic and anthropometric information.

Table 1.

Baseline characteristics of the study participants

Variables		Amount
Age (year)		35.4±12.9
Gender
Men		1717 (28.7%)
Women		4263 (71.3%)
Weight (kg)		76.6±21.8
Height (cm)		163.2±9.2
Body mass index (kg/m2)		28.8±7.9
Phase angle (°)		5.3±0.7
Body cell mass		30.2±6.9
Body fat mass (kg)		30.1±15.4
Fat mass percentage (%)		37.1±12.04
Fat mass index		11.5±6.03
Fat-free mass (kg)		46.5±10.5
Skeletal muscle mass (kg)		25.5±6.3
Soft lean mass (kg)		43.8±9.9
Fat-free mass index		17.3±2.7
Visceral fat level		13.7±7.1
Total body water (kg)		34.2 7.7
Extracellular water (kg)		13.1±2.9
Intracellular water (kg)		21.1±4.8
Extracellular water/total body water		0.38±0.007
Protein (kg)		9.1±2.1
Waist to hip ratio		0.92±0.08
BMI categories
Underweight		706 (11.8%)
Normal		1227 (20.5%)
Overweight		1354 (22.7%)
Obese		2693 (45.0%)

Quantitative data are presented as the mean±standard deviation, while qualitative data are reported as frequency (percentage)

The Pearson’s correlation between PhA and body composition components and some other demographic and anthropometric factors is shown in Table 2. The strongest correlations were found for ECW/TBW (−0.84, P < 0.001), followed by ICW (0.62, P < 0.001), SMM (0.62, P < 0.001), BCM (0.62, P < 0.001), and protein (0.62, P < 0.001). There were moderate correlations with FFM (0.59, P < 0.001), SLM (0.59, P < 0.001), FFMI (0.59, P < 0.001), TBW (0.58, P < 0.001), and ECW (0.52, P < 0.001). BFP (−0.07, P < 0.001), FMI (0.07, P < 0.001), VFL (0.07, P < 0.001), BFM (0.13, P < 0.001), and age (−0.16, P < 0.001) showed the least but significant correlation with PhA.

Table 2.

The correlation between phase angle and body composition

Variables		r*		P
Age (year)		−0.16		<0.001
Weight (kg)		0.37		<0.001
Height (cm)		0.31		<0.001
Body mass index (kg/m2)		0.25		<0.001
Body cell mass		0.62		<0.001
Body fat mass (kg)		0.13		<0.001
Body fat percentage (%)		−0.07		<0.001
Fat mass index		0.07		<0.001
Fat-free mass (kg)		0.59		<0.001
Skeletal muscle mass (kg)		0.62		<0.001
Soft lean mass (kg)		0.59		<0.001
Fat-free mass index		0.59		<0.001
Visceral fat level		0.07		<0.001
Total body water (kg)		0.58		<0.001
Extracellular water (kg)		0.52		<0.001
Intracellular water (kg)		0.62		<0.001
Extracellular water/total body water		−0.84		<0.001
Protein (kg)		0.62		<0.001
Waist to hip ratio		0.26		<0.001

*Pearson correlation

The findings using different linear regression models that investigated the association among various predictors and the PhA are shown in Table 3. Only ECW/TBW was used as a predictor in the original model (Raw Model). ECW/TBW and PhA were significantly negatively associated in the model (β = −82.81, standardized β = −0.84, P < 0.001, adjusted R² = 0.71).

Table 3.

Linear regression to predict phase angle from body composition components

Model		Variables		β		Standardized β		P		R 2		Adjusted R2
Raw model		Constant		37.06		0.26		<0.001		0.71		0.71
		ECW/TBW		−82.81		−0.84		<0.001
Model 1		Constant		39.88		0.26		<0.001		0.75		0.75
		ECW/TBW		−91.28		−0.93		<0.001
		Age		0.012		0.21		<0.001
Model 2		Constant		33.69		0.22		<0.001		0.85		0.85
		ECW/TBW		−77.47		−0.79		<0.001
		Age		0.009		0.17		<0.001
		SMM		0.039		0.34		<0.001
Model 3		Constant		34.77		0.22		<0.001		0.86		0.86
		ECW/TBW		−80.17		−0.82		<0.001
		Age		0.008		0.15		<0.001
		SMM		0.033		0.29		<0.001
		BFM		0.005		0.10		<0.001
Model 4		Constant		35.94		0.20		<0.001		0.90		0.90
		ECW/TBW		−71.88		−0.73		<0.001
		Age		0.004		0.07		<0.001
		SMM		0.078		0.69		<0.001
		BFM		−0.005		−0.10		<0.001
		Height		−0.031		−0.40		<0.001

ECW/TBW=Extracellular water/Total body water, SMM=Skeletal muscle mass, BFM=Body fat mass

Age was included as a predictor in Model 1. While age had a slight but significant positive effect (β = 0.012, standardized β = 0.21, P < 0.001), ECW/TBW still showed a strong negative association with PhA (β = −91.28, standardized β = −0.93, P < 0.001). By adding age to the Raw Model, the model power increased to R² = 0.75.

Model 2 used both age and SMM in addition to ECW/TBW in the Raw Model. R² increased to 0.85 (adjusted R² = 0.85), further increasing the model’s explanatory power. SMM presented a significant positive predictor (β = 0.039, standardized β = 0.34, P < 0.001), while ECW/TBW was kept as the most significant predictor (β = −77.47, standardized β = −0.79, P < 0.001). Age continued to have a significant but marginally diminished effect (β = 0.009, standardized β = 0.17, P < 0.001).

In Model 3, BFM was added as an additional predictor to Model 2. The R² value for this model was even higher at 0.86 (adjusted R² = 0.86). SMM continued to exhibit a significant positive association (β = 0.033, standardized β = 0.29, P < 0.001), and ECW/TBW remained a significant negative predictor (β = −80.17, standardized β = −0.82, P < 0.001). BFM had a weak but positive association with PhA (β = 0.005, standardized β = 0.10, P < 0.001), while the effect of age was still significant but smaller than in previous models (β = 0.008, standardized β = 0.15, P < 0.001).

Finally, Model 4 included height in addition to all previous predictors. The best fit was indicated by this model’s higher R² value of 0.90 (adjusted R² = 0.90). In this model, SMM had a significant positive effect (β = 0.078, standardized β = 0.69, P < 0.001), whereas ECW/TBW continued to be a strong negative predictor (β = −71.88, standardized β = −0.73, P < 0.001). Height showed a significant negative association with PhA (β = −0.031, standardized β = −0.40, P < 0.001), and BFM, unlike in Model 3, showed an inverse impact (β = −0.005, standardized β = −0.10, P < 0.001).

As a final result, the following is the prediction equation for PhA with a 90% explanatory power:

Whole-body phase angle = 35.94+ (−71.88 × ECW/TBW) + (0.004 × Age (years)) + (0.078 × SMM (kg)) + (−0.005 × BFM (kg)) + (−0.031 × Height (cm))

Discussion

This study aimed to provide an equation for PhA estimation based on body composition and some other anthropometric and demographic factors as a more practical and cost-effective way of getting PhA without the use of expensive BIA devices. As far as we know, there is no specific equation to predict PhA from body composition components to discuss. So, we discuss our findings with a focus on each association that was shown in the results and literature. The results revealed that PhA is significantly influenced by multiple factors of hydration status and body composition, especially ECW/TBW, SMM, BFM, and height. These findings are consistent with the previous literature that has investigated these associations.

Among all variables analyzed, ECW/TBW demonstrated the strongest inverse correlation with PhA (r = –0.84, P < 0.001). This result was further supported by several regression models, which consistently was the most influential predictor, explaining up to 90% of the variance in PhA when combined with other covariates. This finding is supported by literature that suggests ECW/TBW as a key determinant of PhA through an inverse association.[14,24,25] Leal-Escobar et al.[25] showed a strong inverse correlation between ECW/TBW, and PhA was determined (r = –0.77, P < 0.001) in patients on renal replacement therapy. Furthermore, Gonzalez et al.[14] observed significant negative correlations with ECW/ICW in both men and women (men: β = –0.847, adjusted R² = 0.54; women: β = –0.826, adjusted R² = 0.42). The inverse association may be because an increased ECW/TBW shows fluid imbalance and increased extracellular fluid accumulation, which may result from disturbed cell membrane integrity and consequently causes reduced PhA values.[9]

In addition, ICW, SMM, and BCM were also found to have strong positive correlations with PhA (r = 0.62, P < 0.001 for each). These findings are supported by studies such as Perez et al.,[26] who found a positive correlation between PhA and ICW in cancer patients (r = 0.37, P < 0.001), and Gerken et al.,[27] who observed positive correlations between PhA and BCM (r = 0.36, P = 0.0008) in patients undergoing bariatric surgery. The positive relationship between PhA and muscle mass is well established.[28,29,30,31] Yamada et al.[29] found a significant direct association between PhA and quadriceps femoris muscle mass in older adults (r = 0.368 in men; P < 0.001), and Yasui-Yamada et al.[30] reported a correlation of r = 0.60 (P < 0.001) between PhA and muscle mass in patients with gastrointestinal and hepatobiliary-pancreatic cancer. A study by Martins et al.[32] on an adolescent population found that lean mass was statistically significant in both linear regression models (raw model: β = 0.02, P < 0.01; R² = 0.17; adjusted model for age, sexual maturation, physical activity level, and screen time: β = 0.02, P < 0.01; R² = 0.16). Since muscle tissue is rich in intracellular water, higher muscle mass directly contributes to increased PhA.[9,33] Our analysis also demonstrated moderate positive correlations between PhA and FFM, SLM, FFMI, and TBW, with correlation coefficients ranging from r = 0.52 to 0.59 (P < 0.001). The relationship between PhA and FFM is consistent with findings among different clinical conditions, such as hemodialysis patients (r = 0.303, P = 0.002),[34] children and adolescents undergoing hematopoietic stem cell transplantation (r = 0.375, P = 0.002),[35] and in stable chronic obstructive pulmonary disease (r = 0.47, P < 0.001).[36] These results further emphasize that greater lean tissue and fluid content are associated with higher PhA values in different clinical conditions.

Interestingly, BFM showed only a weak and inconsistent relationship with PhA in our regression models. While it initially appeared as a weak positive predictor (β = 0.005, standardized β = 0.10, P < 0.001), it became a negative predictor in the final model (β = –0.005, standardized β = –0.10, P < 0.001). This inconsistency mirrors the heterogeneous findings in the literature. Barrea et al.[37] observed an inverse correlation between PhA and BFM in adults with Prader–Willi syndrome (r = –0.60, P = 0.02). Another study by Streb et al.[38] found a negative significant relationship between PhA and BFP in both Raw Model (β = –0.063, P < 0.001; adjusted R² = 0.49) and adjusted model for sex, age, skin color, physical activity, and sedentary behavior (β = –0.065, P < 0.001; adjusted R² = 0.53). But, Antunes et al.[39] found a small but significant positive relationship between PhA and BFP in HIV-infected patients (β = 0.079, P = 0.007). Another study by Martins et al.[32] found that BFM was not statistically significant in adolescents in either the Raw Model (β = –0.04, P = 0.15; R² = 0.05) or the adjusted model for age, sexual maturation, physical activity level, and screen time (β = –0.05, P = 0.17; R² = 0.09). Some other studies also found no significant correlation in hemodialysis patients or healthy adults.[40,41] These discrepancies could be because of the variations in the study populations (e.g., age, gender, health status, and clinical conditions), the methodology used to measure body composition (e.g., anthropometry vs. DXA vs. BIA), and confounding variables including physical activity levels, dietary intake, hydration status, and medication.

Furthermore, results showed a statistically significant by low correlation between PhA and BMI and WHR. Studies have demonstrated a significant positive relationship between PhA and BMI.[18,41,42,43] Bosy-Westphal et al.[18] found that PhA increases with BMI until the BMI levels are less than 30 kg/m², but their association went in the opposite direction in BMI more than 40 kg/m². Similarly, Fu et al.[42] showed that there was a significant direct correlation between PhA and BMI in overweight and obese Chinese individuals. Additionally, Cimmino et al.[43] found a significant, mild positive correlation between PhA and BMI in both genders. Barrea et al.,[17] however, showed that PhA was inversely associated with BMI in obese individuals with inflammation or vitamin D deficiency. Overall, it can be concluded that PhA may show a positive correlation with BMI, though this relationship may reverse in severe obesity, possibly because of the effect of inflammation or altered water distribution. The association between PhA and WHR is not investigated vastly in the literature and the evidence is so limited. In the study by Raji et al.[41] PhA was inversely and mildly correlated with WHR (r = −0.076, P < 0.0001) in men, but the inverse association in women group was not statistically significant (r = −0.024, P = 0.22). The study by Hu et al.[44] showed no statistically significant association between PhA and WHR in type 2 diabetes mellitus patients (β = 1.26, P = 0.16); however, the coefficient showed a positive association. After all, this study revealed a statistically significant positive correlation between PhA and WHR, possibly due to an increase in cell quantity with an increase in WHR; however, the evidence for this association is contradictory and limited. So, further research is required to fully understand this issue.

Another observation in our study was the significant positive correlation between height and PhA (r = 0.31). But, when height was added to the regression model in addition to other variables, its relationship became negative (β = –0.031, standardized β = –0.40, P < 0.001), which is similar to the results of other studies.[13] Given that height and weight are indicators of somatic development, which is characterized by skeletal and muscular growth, they collectively increase the percentage of FFM, thereby affecting resistance and PhA.[13]

We showed that there is an inverse association between PhA and age. The study by Bosy-Westphal et al.[18] showed a negative correlation between PhA and age across different BMI categories. Similarly, the study conducted by Raji et al.[41] revealed a statistically significant negative correlation between PhA and age in both genders in healthy Iranian adults. Additionally, the study by Siddiqui et al.[13] examined a univariate logistic regression model to assess potential predictors of PhA. Age was a negative predictor (β = –0.19, P > 0.05) in the regression model, which was not a statistically significant predictor, possibly due to no enough sample size. The inverse association between PhA and age is possibly because of a reduction in Xc, which is caused by a loss of SMM and an increase in R, which is due to a loss in TBW at the expense of BFM accumulation with aging.[9] Figure 1 illustrates the summary of the association between PhA and its determinants.

Figure 1.

Key determinants of body phase angle and its clinical significance. Positive (+) and negative (−) signs indicate the direction of association between each factor and phase angle

As a policy brief, we recommend using PhA—a marker of cellular health—via the provided equation in clinical practice and health guidelines as a novel screening tool to assess the risk of noncommunicable diseases and nutritional health. Moreover, we recommend health-promoting counseling and nutritional interventions aimed at enhancing SMM and reducing BFM, which are crucial strategies to improve PhA.

This study provides an equation to estimate PhA with a fit of about 90%, which is by far the best fit in the literature. Additionally, this equation was derived from a large sample size, which improves the precision of the estimates and may enhance the generalizability of the results. However, in addition to these strengths of the study, we faced some limitations. We did not have access to data about medical history, dietary intake, physical activity, and other potential covariates that may interfere with the results. We recommend considering these variables in future research.

Conclusion

This study proposes an equation for PhA based on body composition components, which contains ECW/TBW, SMM, and BFM. The equation suggests that higher amounts of SMM and lower amounts of ECW/TBW and BFM lead to better PhA values. Including this prediction equation for PhA in screening guidelines may help in the early diagnosis of nutritional risk, especially in places with limited resources.

Abbreviations

BCM: Body Cell Mass; BFM: Body Fat Mass; BFP: Body Fat Percentage; BMI: Body Mass Index; DXA: Dual-Energy X-ray Absorptiometry; ECW: Extracellular Water; ECW/TBW: Extracellular Water/Total Body Water ratio; FFM: Fat-Free Mass; FMI: Fat Mass Index; FFMI: Fat-Free Mass Index; ICW: Intracellular Water; PhA: 50-kHz Whole-body Phase Angle; R: Resistance; SLM: Soft Lean Mass; SMM: Skeletal Muscle Mass; TBW: Total Body Water; VFL: Visceral Fat Level; VIF: Variance Inflation Factor; WHR: Waist-to-Hip Ratio; Xc: Reactance.

Author’s contributions

MGM, NS, AED: Conceptualization. AED, NS and MN: Methodology. AED and HE: Formal analysis. SS, NS, SA, FSH, MK and AED: Investigation. AED and MN: Writing–original draft preparation. NS and GAF: Writing–review and editing. MGM and HE: Supervision.

Data availability statement

The datasets generated during and/or analyzed during this study are available from the corresponding author on reasonable request.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

This study is part of a master’s thesis submitted by Ali Ebrahimi Dabagh (ID: 4020599) and was supported by the internal grant of Mashhad University of Medical Sciences. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.