Association Between the Appendicular Extracellular-to-Intracellular Water Ratio and All-Cause Mortality: A 10-Year Longitudinal Study

Abstract

The appendicular extracellular-to-intracellular water ratio (A-E/I) is a potential marker of skeletal muscle quality, reflecting the balance of water distribution between the extracellular and intracellular compartments of the appendicular limb regions. A-E/I has been increasingly used in recent studies; however, its association with adverse outcomes remains unclear. This study investigated the potential association between A-E/I and all-cause mortality. A prospective cohort study of 8 015 middle-aged and older adults (comprised of 4 755 women, aged 45–74 years) residing in a Japanese community was conducted. The baseline assessment was performed between 2010 and 2012, and the follow-up period lasted until July 2022. A-E/I and skeletal muscle mass were measured using segmental bioelectrical impedance spectroscopy. Handgrip strength (HGS) was measured using a Smedley-type dynamometer. Lifestyle, medical history, and physical activity were assessed by questionnaire and accelerometer. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) for each quartile (Q) of A-E/I were estimated using the multivariable Cox regression model. During a 10.5-year median follow-up, the mortality rates were 8.9 and 3.6 per 1 000 person-years for men (292 deaths) and women (174 deaths), respectively. A-E/I quartiles were positively associated with all-cause mortality in both sexes (men: Q1, HR: 1.0 [95% CI: reference], Q4, HR: 1.8 [1.1–2.9], p_trend < .05; women, Q4, HR: 2.2 [1.3–3.8], p_trend < .01). This association remained significant after further adjustment for skeletal muscle mass and HGS (p_trend < .05). Our findings suggest that A-E/I serves as an early predictive marker for mortality in middle-aged and older Japanese adults.

Water is a crucial component of the human body and is indispensable for sustaining life (1). Total body water content is between 49% and 59% of body weight in adult men and 42%–48% in adult women (2). Skeletal muscle is characterized by its considerable water content compared with other organs (3). This water in skeletal muscle is divided into extracellular water (ECW) and intracellular water (ICW) (4). Previous studies indicate a marked decrease in ICW with aging, potentially reflecting a decline in muscle cell mass (5–8). Therefore, evaluating the hydration status of skeletal muscle could offer valuable insights that may help predict future health outcomes (9).

Segmental Bioelectrical Impedance Spectroscopy (S-BIS) is a noninvasive and rapid method for assessing the water content of specific body segments such as the arms and legs (5,7,10). Because the appendicular limb regions are primarily composed of skeletal muscle tissue (11), water measurements in the appendicular limb regions are commonly used to assess the hydration status of skeletal muscle (6,12–14). S-BIS can distinguish between ECW and ICW contents in the appendicular limb regions (5,7). The ECW in skeletal muscle reflects the extracellular space filled with plasma and interstitial fluid, whereas ICW is used as a proxy marker for muscle cell mass as it is contained within cells (4,5,7). Prior studies have used the appendicular ECW-to-ICW ratio (A-E/I) as a marker reflecting muscle quality, and A-E/I is higher with age (6). Higher A-E/I is also characterized by lower levels of muscle strength, gait speed, and habitual physical activity (PA) and higher levels of functional disability, independent of skeletal muscle mass (6,12,14). Therefore, A-E/I has the potential to serve as a predictive marker for future adverse health outcomes associated with skeletal muscle aging.

However, the clinical significance of measuring A-E/I remains unclear, as the majority of the supporting evidence stems from cross-sectional studies (5–7,12,14). Specifically, the extent to which A-E/I predicts the risk of mortality, a crucial health outcome, is yet to be investigated. Handgrip strength (HGS) and walking speed are well-established predictors of mortality (15). Because A-E/I is associated with HGS and walking speed, independent of skeletal muscle mass (6,12), it may also serve as a valid indicator of mortality risk. In contrast to A-E/I, whole-body E/I has been reported to be associated with mortality in patients with hemodialysis, chronic kidney disease, and cancer, but this measure may differ from the hydration status in the appendicular limb regions of the general population owing to pathological edema, pleural effusions in the trunk, and ascites (16–18). Therefore, evaluation of appendicular water content in the general population is necessary to assess skeletal muscle hydration status and its relationship to mortality risk.

To our knowledge, no study has yet examined the association between the hydration status of appendicular limb regions and mortality in the general population. Moreover, the portability and ease of measurement of A-E/I hold promise for its application as a valuable marker in clinical and public health settings, provided its connection with mortality is clarified. Therefore, this study aimed to examine the association between A-E/I and future mortality. We hypothesized that higher A-E/I would be associated with higher mortality risk independent of skeletal muscle mass and other covariates.

Method

Study Design and Population

The study was conducted in a longitudinal study design using prospective cohort data from the Japan Multi-institutional Collaborative Cohort Study in the Saga Region (Saga J-MICC study) (19,20). Details of the Saga J-MICC study have been previously reported (19,20). Briefly, the J-MICC study was initiated in 2005 by 10 research groups in Japan to investigate the interaction between genes and the environment in lifestyle-related diseases such as cancer (21). The J-MICC study planned to enroll 100 000 people from 10 research groups (21). In the Saga region, one of the 10 research groups mailed, invitations to approximately 62 000 people; all residents of the former Saga City aged 40–69 years (19). Assuming a participation rate of 20%, approximately 12 000 residents were expected to participate; actual participation was 12 078 (19,20). Of the 12 078 participants in the baseline study, 8 454 (73.6%) participated in a secondary study conducted 5 years later (Figure 1) (20). The first survey was conducted from 2005 to 2007, and the secondary survey was conducted from 2010 to 2012 (19,20). The study participants were followed up until 2022, and each year the resident record of Saga City was checked to obtain information on the dates of moving out of the city, new addresses, and death (20).

Figure 1.

Flowchart of Saga J-MICC study participants. We defined a secondary survey with the addition of bioelectrical impedance spectroscopy as the baseline for this study. A-E/I, appendicular extracellular-to-intracellular water ratio; HGS, handgrip strength; PAL, physical activity level; Saga J-MICC, Japan Multi-institutional Collaborative Cohort Study in the Saga Region.

The Saga region has several independent survey items in addition to those common to the overall J-MICC study (19,20). The secondary survey added HGS and body composition, as measured by the S-BIS, to the first survey items (questionnaire, anthropometry, laboratory data, and PA survey with an accelerometer). Therefore, in this study, we defined baseline as the secondary survey in which the S-BIS was measured (Figure 1). In summary, this study included 8 454 participants. The follow-up period was from November 1, 2010, the start date of the secondary study, to July 31, 2022, which included the most recent mortality data (Figure 1). The research protocol was approved by the ethics committees of the Saga University Faculty of Medicine (approval no. 17–11) and Nagoya University Graduate School of Medicine (approval no. 253). The purpose, content, and conditions of the study were explained in writing and orally, and the participants provided written informed consent to participate in the study.

Assessment of A-E/I

An S-BIS device (MLT-30, SK Medical, Electronics Co, Ltd., Shiga, Japan) was used to measure body water content (5,6). The device uses a logarithmic distribution of 140 frequencies from 2.5 to 350 kHz. Participants with artificial joints or cardiac pacemakers were excluded. For arm measurements, 2 injection electrodes were placed on the dorsal surfaces of both hands, with sensing electrodes placed at the center of the dorsal surfaces of both wrist joints. For leg measurements, 2 injection electrodes were placed on the dorsal surface of each foot, with the sensing electrodes placed on the center of the dorsal surface of each ankle (5,6). Arm segment length (L) was calculated using height as a proxy for arm span, and leg segment L was calculated by measuring the distance from the femoral abductor to the lateral condyle using a tape measure. Resistances at zero (R₀) and infinity (R_∞) frequencies were determined by extrapolation after fitting the spectrum of bioimpedance data to the Cole–Cole model using dedicated software (5,6). The formula for ECW (mL) is ρ_ECW × L²/R₀, where ρ_ECW is the extracellular resistivity factor (47 Ω cm) (5,6). The formula for ICW (mL) is ρ_ICW × L²/R_ICW, where ρ_ICW represents factors for intracellular resistivity (273.9 Ω cm), and R_ICW was calculated using the following equation: 1/[(1/R_∞) − (1/R₀)] (5,6). ECW and ICW were calculated for each segment (both arms and both legs) and the summed values were defined as the appendicular ECW (A-ECW) and appendicular ICW (A-ICW) (6). The A-E/I was calculated by dividing A-ECW by A-ICW (6).

Identification of Death Information and Censoring

Death information was obtained by browsing the basic annual resident registration; if there was a change in the research participants, the date of death was confirmed by making a residence record inquiry (20). The cause of death was confirmed by reviewing the death certificate at the Saga-Chubu Public Health and Welfare Office, with the approval of the Director General of the Minister of Health, Labour, and Welfare. In this study, the all-cause mortality information followed through July 31, 2022, was used in the analysis. Cause-specific mortality information was not used in this study because information collection was not completed through July 31, 2022. Surviving cases followed up until the last day (July 31, 2022), cases moving out of the city and outside Japan, cases deleted from the basic resident registration, and cases dropping out during the follow-up period (participants who provided consent for data use up to the date of dropout) were defined as censored. If a person moved out of the city once and then returned to the city, this was also defined as censored on the date of the first move. The dates of moving out and deletion were confirmed by browsing the Basic Resident Registration, and the specific date was identified through a Resident Record Inquiry.

Other Covariates

Information on age, sex, smoking, alcohol consumption, and medical history used as covariates was extracted from a self-administered questionnaire. Smoking and alcohol consumption statuses were categorized into three groups (never, former, and current). Medical history was obtained from a questionnaire to determine the presence of cancer, cardiovascular disease (CVD), diabetes, and renal failure; CVD included cerebral infarction, cerebral hemorrhage, subarachnoid hemorrhage, myocardial infarction, and angina pectoris. Body mass index (BMI, body weight/m²) was calculated from actual measurements of height (to the nearest 0.1 cm) and body weight (to the nearest 0.1 kg). Habitual PA was assessed using a validated accelerometer (Lifecorder, Suzuken, Japan) (22). The participant was instructed to wear the accelerometer attached to the waistband or belt of their clothing and aligned to the midline of either the left or right thigh, during waking hours (except during bathing or swimming) for 10 days (23). After completing the measurements, the accelerometers were returned by mail. To eliminate the possibility of PA changes due to the accelerometer being worn, accelerometer data obtained during the first 3 days were excluded from the analysis, and only data from the last 7 days when the accelerometer was worn for at least 8 hours per day, were used in the analysis (23). If the accelerometer data did not meet this criterion, it was considered as missing data. PA level (PAL) was calculated as the total energy expenditure calculated from the accelerometer divided by the basal metabolic rate (23). The calculation method has been reported in detail elsewhere (23). The skeletal muscle mass was estimated using the same S-BIS device. Electrodes were placed on the right hand and dorsum of the right foot, and whole-body skeletal muscle mass was calculated using dedicated software supplied with the MLT-30. The skeletal muscle mass estimated by the software included in this device was based on a regression equation created based on the skeletal muscle mass measured using MRI (24). The coefficient of determination and standard error of this regression equation were reported to be 0.937 and 1.85 kg, respectively (24). Skeletal muscle mass was corrected using the square of the height and the skeletal muscle mass index (SMI, kg/m²). The HGS was measured using a Smedley-type dynamometer (Grip-D, T.K.K. 5401, Takei Scientific Instruments, Niigata, Japan). HGS was recorded with the participant in a standing position with the elbows extended; HGS values were measured once on each side, and the higher HGS value (from either the right or left side) was used for analysis (25).

Statistical Analysis

We examined 8 015 participants with complete data without missing values for the main analysis. The 439 participants (5.2%) excluded from the main analysis were missing A-E/I (n = 133), missing PAL (n = 283), missing smoking status (n = 2), missing HGS (n = 19), obvious outliers in A-E/I (z-score > 10; n = 1), and 0 days of follow-up (ie, moved out of the city on the day of the baseline survey; n = 1; Figure 1). A comparison of participant demographic characteristics included in and excluded from the analysis is presented in Supplementary Table 1.

All analyses were stratified by sex. The baseline characteristics were presented as arithmetic means and standard deviations for continuous variables and as numbers and percentages for categorical variables. Comparisons of baseline characteristics by sex were made using unpaired t-tests (continuous) and chi-squared tests (categorical). A generalized linear model was used to examine the linear trends in baseline characteristics based on the A-E/I quartiles in men and women. Linear trend tests were analyzed by transforming the A-E/I quartiles into ordinal variables. Kaplan–Meier curves were drawn and log-rank tests were performed to determine the association between the A-E/I quartiles and mortality.

Cox proportional hazard models were used to calculate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for the A-E/I quartiles. The adjusted model included the following covariates: Model 1 was unadjusted; Model 2 was adjusted for age (continuous) and sex (category, overall analysis only); Model 3 was further adjusted for BMI (continuous), smoking status (category: current, former, never), alcohol consumption (category: current, former, never), PAL (continuous), cancer (category: yes or no), CVD (category: yes or no), diabetes (category: yes or no), and renal failure (category: yes or no). In Model 4, SMI was added to the covariates to examine the associations independent of skeletal muscle mass. We considered Model 4, which was adjusted for all covariates, as the main finding. To demonstrate the dose–response relationship between A-E/I (continuous) and mortality, we fitted a restricted cubic spline (RCS) curve, with 3 knots located at the 10th, 50th, and 90th percentiles (26,27). The median of the lowest quartile was used as the reference value. The plotted HRs and 95% CIs were adjusted for covariates included in Model 4.

We performed a set of sensitivity analyses to verify the robustness of our findings. First, multiple imputations based on chained equations were conducted to account for potential bias due to the complete case analysis (28,29). To generate 10 multiple imputation data sets, the self-administered questionnaire’s total PA was included in the imputation model as an auxiliary variable for the PAL because of its high missing rate (3.5%) compared to other variables (28,29). A comparison of demographic characteristics of participants with and without missing values in the PAL is presented in Supplementary Table 2. Before implementing the multiple imputation model, two participants were excluded for reasons other than missing values: an outlier in A-E/I (n = 1) and a follow-up duration of zero days (n = 1). The same Cox proportional hazards models were applied to each multiple imputation data set, and Rubin’s rules were used to combine the estimates (28). Second, stratification by age groups (under 65 and over 65 years) to examine the effect of age on the association between A-E/I and mortality. Third, we conducted the same analyses after excluding deaths that occurred within 2 years of follow-up to address reverse causality bias. Fourth, we conducted the same analysis after excluding participants who had cancer or CVD, which are the major causes of mortality in Japan (30), at baseline. Fifth, we performed an additional analysis adjusted for HGS because based on previous research findings (6,12), we expected muscle function to be an intermediate factor in the association between A-E/I and mortality. We also constructed receiver operating characteristic (ROC) curves for A-E/I, SMI, and HGS and compared their area under the curves (AUC) to assess A-E/I mortality prediction performance (31). Furthermore, we used the Akaike’s Information Criterion (AIC) and Concordance statistic (C-statistic) to determine if including A-E/I in the multivariate Cox regression model improved model fit and performance (31,32). Lower AIC values suggest better model fit, whereas higher C-statistic values indicate superior predictive performance (31,32). Sixth, in addition to A-E/I, previous studies have also utilized ECW-to-total water (TW) ratio and ICW-to-TW ratio to assess appendicular water content (5). Therefore, we also presented the results with appendicular ECW/TW (A-E/T) and appendicular ICW/TW (A-I/T), respectively, as exposure variables. Finally, participants with possible hypoalbuminemia (serum albumin level <3.5 g/dL) were excluded to address the impact of potential peripheral edema (33).

All analyses were performed using R version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). Kaplan–Meier curves and Cox proportional hazards models were computed using the “survminer” and “survival” packages. The fitting of the RCS curves and p values for nonlinearity were generated using the “rms” package. Multiple imputations and a combination of estimates were performed using the “mice” package. ROC curves and AUCs were obtained using the “timeROC” package. Statistical significance was set at p < .05.

Results

For both men and women, the median follow-up duration was 10.5 years. During this period, 292 deaths were recorded among men, indicating a mortality rate of 8.9 per 1 000 person-years, whereas 174 deaths were documented for women, indicating a mortality rate of 3.6 per 1 000 person-years.

Table 1 shows a comparison of the baseline characteristics by sex. Significant sex differences were observed, except for a history of cancer (p < .05).

Table 1.

Baseline Demographic Characteristics of Participants Based on Sex

Characteristics	Overall, N = 8 015	Men, N = 3 260	Women, N = 4 755	p Value*
Age, y	61.1 (8.1)	61.8 (8.0)	60.6 (8.1)	<.001
Women, n (%)	4 755 (59.3)	0 (0.0)	4 755 (100)
Weight, kg	58.9 (10.8)	66.9 (9.5)	53.4 (7.8)	<.001
Height, cm	159.0 (8.3)	166.3 (6.0)	153.9 (5.5)	<.001
BMI, kg/m2	22.8 (3.1)	23.7 (2.9)	22.2 (3.1)	<.001
A-ECW, kg	4.2 (0.9)	5.0 (0.7)	3.7 (0.5)	<.001
A-ICW, kg	7.7 (2.5)	10.0 (2.0)	6.0 (1.2)	<.001
A-E/I	0.58 (0.11)	0.51 (0.08)	0.63 (0.09)	<.001
SMI, kg/m2	6.8 (0.8)	7.1 (0.8)	6.7 (0.8)	<.001
HGS, kg	29.2 (8.6)	37.6 (6.2)	23.5 (4.1)	<.001
Smoking status, n (%)				<.001
Current	988 (12.3)	764 (23.4)	224 (4.7)
Former	1 871 (23.3)	1 592 (48.8)	279 (5.9)
Never	5 156 (64.3)	904 (27.7)	4 252 (89.4)
Alcohol consumption, n (%)				<.001
Current	4 310 (53.8)	2 517 (77.2)	1 793 (37.7)
Former	146 (1.8)	103 (3.2)	43 (0.9)
Never	3 559 (44.4)	640 (19.6)	2 919 (61.4)
Cancer, n (%)	663 (8.3)	289 (8.9)	374 (7.9)	.111
CVD, n (%)	466 (5.8)	310 (9.5)	156 (3.3)	<.001
Diabetes, n (%)	634 (7.9)	393 (12.1)	241 (5.1)	<.001
Renal failure, n (%)	26 (0.3)	16 (0.5)	10 (0.2)	.030
PAL	1.447 (0.090)	1.446 (0.100)	1.447 (0.082)	.005

Notes: Continuous variable, mean (standard deviation); categorical variable, number (percentage). A-ECW = appendicular extracellular water; A-E/I = appendicular extracellular-to-intracellular water ratio; A-ICW = appendicular intracellular water; BMI = body mass index; CVD = cardiovascular disease; HGS = handgrip strength; PAL = physical activity level; SMI = skeletal muscle mass index.

^*Compared by sex.

Supplementary Table 3 displays baseline characteristics by A-E/I quartiles. Among men, significant linear trends were noted in all variables, except height, A-ECW, smoking status, and renal failure (p for trend < .05). Among women, significant linear trends were observed in all variables, except A-ECW, diabetes, and renal failure (p for trend < .05).

In Figure 2, the Kaplan–Meier curves illustrate the association between the A-E/I quartiles and mortality. The log-rank test showed significant differences between A-E/I quartiles and mortality for both men and women (p values <.01).

Figure 2.

Kaplan–Meier curves plotted by sex. The log-rank test revealed statistically significant differences in the appendicular extracellular-to-intracellular water ratio (A-E/I) quartile survival curves for both men and women (p < .001).

Table 2 presents the Cox proportional hazards model used to examine the association between the A-E/I quartiles and mortality. After adjusting for all the covariates, a significant positive linear trend was observed for both men and women. Quartile 4 had significantly higher adjusted HRs compared to quartile 1:1.8 (95% CI: 1.1–2.9) for men (p for trend = .01), and 2.2 (95% CI: 1.3–3.8) for women (p for trend < .01).

Table 2.

Association of A-E/I Quartiles With Mortality: Hazard Ratios and 95% Confidence Intervals

						Model 1		Model 2		Model 3		Model 4
Quartiles	Median [min, max]	N	Person-years	Event, N	Mortality rate, per 1 000 person-years	HR	95% CI	HR	95% CI	HR	95% CI	HR	95% CI
Men, n = 3 260
Q1	0.43 [0.33, 0.46]	815	8 371	29	3.5	1.00	Reference	1.00	Reference	1.00	Reference	1.00	Reference
Q2	0.48 [0.46, 0.50]	815	8 288	50	6.0	1.75	1.10, 2.76	1.27	0.80, 2.02	1.30	0.82, 2.07	1.17	0.73, 1.88
Q3	0.53 [0.50, 0.55]	815	8 251	71	8.6	2.50	1.62, 3.85	1.53	0.98, 2.37	1.38	0.88, 2.16	1.18	0.73, 1.90
Q4	0.60 [0.55, 0.99]	815	7 821	142	18.2	5.37	3.60, 8.00	2.63	1.73, 4.00	2.24	1.44, 3.49	1.76	1.07, 2.91
p for trend						<0.001		<0.001		<0.001		0.011
Women, n = 4 755
Q1	0.53 [0.36, 0.57]	1 189	12 363	25	2.0	1.00	Reference	1.00	Reference	1.00	Reference	1.00	Reference
Q2	0.60 [0.57, 0.62]	1 189	12 341	31	2.5	1.24	0.73, 2.10	1.11	0.66, 1.89	1.07	0.63, 1.82	1.13	0.66, 1.95
Q3	0.65 [0.62, 0.68]	1 189	12 146	41	3.4	1.69	1.03, 2.77	1.37	0.83, 2.27	1.22	0.73, 2.03	1.33	0.78, 2.27
Q4	0.74 [0.68, 1.31]	1 188	12 092	77	6.4	3.19	2.03, 5.01	2.30	1.45, 3.64	1.87	1.14, 3.06	2.19	1.26, 3.79
p for trend						<0.001		0.004		0.004		0.002

Notes: Model 1: unadjusted. Model 2: adjusted for age. Model 3: Model 2 + BMI, smoking status, alcohol consumption, PAL, cancer, CVD, diabetes, and renal failure. Model 4: Model 3 + SMI. A-E/I = appendicular extracellular-to-intracellular water ratio; BMI = body mass index; CI = confidence interval; CVD = cardiovascular disease; HR = hazard ratio; PAL = physical activity level; Q = quartile; SMI = skeletal muscle mass index.

Figure 3 presents the RCS curve illustrating the dose–response relationship between A-E/I and mortality. The RCS curve showed a linear shape association for both men (p for nonlinear = .52) and women (p for nonlinear = .49).

Figure 3.

Restricted cubic spline curve illustrating the association between appendicular extracellular-to-intracellular water ratio (A-E/I) and mortality. The dose–response relationship between A-E/I and mortality was found to have a linear shape association for both sexes (men, p for nonlinear = .52; women, p for nonlinear =.49). The hazard ratio (HR) and 95 % confidence intervals (CIs) were adjusted for age, body mass index, smoking status, alcohol consumption, physical activity level, cancer, cardiovascular disease, diabetes, and renal failure and skeletal muscle mass index. The y-axis is presented on the logarithmic scale. Knot positions were set at 10%, 50%, and 90% of the distribution of A-E/I in each sex. The reference value was the median of the lowest quartile in each sex.

The results of the first sensitivity analysis using multiple imputation data in the Cox proportional hazards model showed no substantial differences from the results of the main analysis (Supplementary Table 4). Briefly, adjusted HRs for Quartile 4 compared to Quartile 1 remained significant for both men (men: 1.7, 95% CI: 1.1–2.7) and women (2.4, 95% CI: 1.4–4.0). These findings suggest that results of the main analysis were robust to missing data. Results stratified by age showed that the association was attenuated in the middle age group, where mortality rates are lower (Supplementary Table 5). Moreover, the third sensitivity analysis, which excluded deaths within 2 years, showed slightly diminished associations for men (p for trend = .06, Supplementary Table 6), whereas the association remained significant for women (p for trend < .01, Supplementary Table 6). The fourth sensitivity analysis, excluding participants with a history of cancer and CVD, demonstrated a slightly attenuated association in women (p for trend = .06, Supplementary Table 7). The results of the fifth sensitivity analysis with additional adjustment for HGS, which was inferred to be an intermediate factor, showed that the association between A-E/I and mortality was only slightly attenuated, and the association remained significant for both sexes (p for trend < .05, Supplementary Table 8). AUCs of SMI, HGS, and A-E/I for mortality were highest for A-E/I for both men and women (men: 0.70; women: 0.64, Supplementary Figure 1). The multivariate Cox proportional hazards model improved model fit and predictive performance by further including A-E/I in addition to SMI and HGS (Supplementary Table 9). Supplementary Tables 10 and 11 present the results of sensitivity analyses employing A-E/T and A-I/T as exposure variables, respectively. For both men and women, the A-E/T exhibited a positive association with mortality, and the A-I/T demonstrated a negative association (p for trend < .05). Furthermore, the association between A-E/I and mortality was robust even after excluding participants with possible hypoalbuminemia (p for trend < .05; Supplementary Table 12).

Discussion

This study investigated the association between A-E/Is and all-cause mortality in community-dwelling middle-aged and older Japanese adults. We found a significant positive association between A-E/I and all-cause mortality independent of skeletal muscle mass and muscle strength. Our findings underscore the importance of evaluating skeletal muscle quality, in addition to quantity and strength, when assessing health risks. Given that A-E/I can be measured noninvasively, quickly, and easily using relatively inexpensive portable devices, it may be a useful indicator for skeletal muscle assessment in clinical and public health settings.

Previous studies have demonstrated an association between indices of thigh, calf, and trunk muscle quality, assessed by CT, and mortality risk independent of muscle cross-sectional area (34–36). Skeletal muscle undergoes age-related changes that involve not only a reduction in the number and size of muscle fibers, but also a substitution of noncontractile tissues such as fat and connective tissue (37). Fat and connective tissue infiltration into skeletal muscle impair muscle function and negatively affect metabolic function, such as reducing glucose uptake and increasing insulin resistance, thereby promoting systemic inflammation and potentially increasing mortality risk (34,38–41). Increased A-E/I levels in aging are primarily due to a significant decrease in A-ICW (6), likely reflecting a reduction in muscle cell mass (5,8,12). In addition, a recent study reported an association between thigh E/I and fatty infiltration markers measured by magnetic resonance imaging (MRI) (13). Although elevated A-E/I may reflect deterioration of muscle quality, such as decreased muscle cell mass and fatty infiltration in the appendicular limb regions, further research is needed to elucidate the underlying causes of elevated A-E/I.

The biological mechanisms underlying the association between A-E/I and mortality are not well understood. Because previous studies have shown that A-E/I is associated with HGS and knee-extensor strength (6,12), we hypothesized that muscle strength mediates the relationship between A-E/I and mortality risk. Accordingly, we expected that adjusting for HGS would substantially weaken the association between the A-E/I and mortality. However, contrary to our expectations, the sensitivity analysis with additional adjustment for HGS only slightly attenuated the association between the A-E/I and mortality, which remained significant. Given that A-E/I is associated not only with muscle function but also with muscle quality, including fat infiltration, metabolic dysfunction such as decreased glucose uptake and increased insulin resistance may be potential mediators between A-E/I and mortality risk. Therefore, further studies examining the association between A-E/I and glucose metabolism and insulin resistance may help elucidate the biological mechanisms underlying the association between A-E/I and mortality risk.

Interestingly, the AUC, which assessed predictive performance for mortality, was higher for A-E/I than for SMI and HGS for both men and women. Furthermore, the addition of A-E/I to the multivariate model including SMI and HGS improved model fit and predictive performance. These results suggest that including A-E/I, as well as muscle strength and skeletal muscle mass, in the assessment may effectively predict future mortality. In contrast, lean body mass assessed by dual-energy X-ray absorptiometry, which is commonly used as a measure of skeletal muscle mass, cannot evaluate muscle quality and remains a controversial predictor of adverse health outcomes (42). This study provides new insights into traditional skeletal muscle assessment, which has relied on measures of function and quantity, by demonstrating that A-E/I is associated not only with muscle function but also with mortality risk.

This study has a few limitations that should be acknowledged: The first is reverse causation bias. To reduce reverse causation bias, we performed a sensitivity analysis that excluded participants who died within 2 years and another sensitivity analysis that excluded participants who had a history of cancer and CVD, the leading causes of death. For both sexes, the association was slightly attenuated compared to the results of the main analysis. This could be due to a lack of power resulting from the reduced sample size, but reverse causation bias cannot be entirely excluded. Further follow-up may yield more robust results because the study population included middle-aged individuals at a low risk of mortality. Second, we were unable to examine the association with cause-specific mortality because the collection of cause-of-death data had not yet been completed. Future studies may elucidate the potential mechanisms underlying the association between A-E/I and mortality risks by investigating the association between cancer and cardiovascular mortality. Third, the A-E/I measured in this study was not validated by the dilution method, which is the gold-standard method (43); The total body water content estimated by BIS devices has been reported to correlate well with the dilution method (43); however, the dilution method reflects only whole-body water, making it difficult to validate segment by segment (12). A regional potassium counter could potentially be used as a reference method for segmental ICW; however, it is not widely available as a measurement device (44). Fourth, although the skeletal muscle is the major component of the appendicular limb regions (11), tissues other than the skeletal muscle in the appendicular limb regions could have affected the water content estimation (10). These concerns may be addressed using CT or MRI to segment the tissue and validate its association with A-E/I. Fifth, the resistivity commonly used in previous studies was used to estimate water content (5,6,24). Thus, the estimated amount of A-ECW or A-ICW may differ depending on the resistivity used, but the association with mortality is unaffected. Nevertheless, further research is required to determine the optimal segment-specific resistivity in the general population.

In conclusion, this study provides evidence that the A-E/I is independently associated with all-cause mortality in middle-aged and older adults living in a Japanese community. The positive association between the A-E/I and mortality remained significant even after adjusting for skeletal muscle mass and HGS. These findings suggest that the A-E/I may serve as a valuable biomarker for predicting mortality risk and could potentially be used in clinical and public health settings. Further research is needed to elucidate the underlying mechanisms of this association and explore the potential benefits of interventions targeting A-E/I for improving health outcomes.

Funding

This study was supported by Grants-in-Aid for Scientific Research for Priority Areas of Cancer (No. 17015018), Innovative Areas (No. 221S0001), the Japan Society for the Promotion of Science (JSPS) KAKENHI (Grant Numbers 18390182, 20249038, 16H06277, 20H03943, 16K09058, 21K11679, 22H04923 [CoBiA], and 22K21086) from the Japanese Ministry of Education, Culture, Sports, Science and Technology, and the Hachiro Honjyo Ocha Foundation.

Conflict of Interest

None.