Qualitative and quantitative ultrasonographic assessment of muscle and its association with comprehensive geriatric assessment results among older patients scheduled for abdominal surgery

A priori sample size calculation was conducted to determine the number of participants required to detect a moderate correlation between ultrasound-derived muscle parameters and frailty features assessed by the CGA, with a statistical power of 80% and an alpha level of 0.05. Based on this, a minimum of 84 participants was required. Patients aged 65 years old and older, hospitalized in our Department between February and August 2023, and scheduled for elective surgery were enrolled. Exclusion criteria included a history of lower limb fracture or surgery within the past 12 months, tremors, paresis, and autoimmune disorders affecting the musculoskeletal system. Written informed consent was obtained from all subjects involved in the study. The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Jagiellonian University Medical College (protocol code: 1072.6120.339.2022).

A trained and certified geriatric team assessed patients at admission to the department, performing the CGA, physical function tests, and skeletal muscle ultrasonography.

To perform the CGA, validated tools described in detail in the published literature were used^(6,7). A hydraulic dynamometer (Jamar Hydraulic Hand Dynamometer) was utilized to measure the maximum voluntary handgrip strength (HGS) of the dominant hand⁽⁸⁾. The cut-off points for frailty detection were based on existing reports⁽⁹⁾. The threshold of 3 impaired CGA domains is a recognized cut-off for frailty syndrome in our department⁽⁷⁾. HGS lower than 16 kg for women and 27 kg for men or Five Times Sit to Stand Test (FTSST) exceeding 15 seconds were the criteria for diagnosing probable sarco penia⁽³⁾.

Data were collected on the length of postoperative stay, unplanned admission to the intensive care unit (ICU), and early surgical adverse events graded according to the Clavien-Dindo system and the comprehensive complication index (CCI).

Ultrasound of the right rectus femoris (RF) muscle was performed using a Toshiba APLIO 400 ultrasound scanner (Canon; Tokyo, Japan), linear transducer (400 PLT-704AT), and built-in preset for the assessment of skeletal muscles with a penetration depth set at 5.5 cm and focus at 2.0 cm. Four trained investigators performed the ultrasound examinations and obtained measurements of muscle thickness (MT), cross-sectional area (CSA), and pennation angle (PA). Prior to study initiation, all four examiners independently acquired 15 skeletal muscle images from five pilot participants. These were then compared to ensure interrater agreement in image acquisition and measurement procedures. On the day of the examination, the same person who performed the ultrasound also conducted the CGA and entered the results into an online project database. However, CGA results remained blinded during data collection and were only unblinded at the end of the study during the final data analysis phase. Participants assumed the supine position and the ultrasound probe was placed perpendicular to the skin at the midpoint of the line connecting the anterior superior iliac spine to the upper edge of the patella. RF MT and CSA were measured in the short-axis view. The line of MT was drawn at the maximal vertical distance from the superficial to deep fascia layers, perpendicular to the femur (Fig. 1B). CSA was calculated using in-built software after manually outlining the muscle boundaries from the outermost fascia to the other edge. The above measurements were repeated during maximal voluntary RF contraction. In the long-axis view, PA was measured between the deep muscle fascia and the muscle fascicle using built-in angle-measuring software. Muscle fascicle length (FL) was estimated using the approved formula (Fig. 1A). We further analyzed ultrasound images in the ImageJ software (Version 1.52a, National Institute of Health, USA). One designated researcher, blinded to both CGA results and ultrasound size measurements, performed secondary analyses to extract muscle echogenicity and tissue texture features. Texture parameters were obtained using the gray-level co-occurrence matrix (GLCM) method (GLCM Texture Analyzer plugin v0.4, Julio E. Cabrera) (Tab. 1). A square region of interest (ROI) was placed within muscle tissue, avoiding surrounding fascia. The mean echogenicity of ROI was determined by grayscale analysis using the histogram function (Fig. 1C). All measurements were performed three times across different ultrasound images, and average values were used for analysis.

Fig. 1.

Example of ultrasound measurements of the rectus femoris (RF) muscle: A. the pennation angle (PA) was obtained in the long-axis view and measured between lines drawn along the deep muscle fascia and the muscle fascicle; b) RF thickness (MT) and cross-sectional area (CSA) were measured in the short-axis view, the former represented by the maximal vertical distance between the superficial and deep fascia layers, the latter calculated after outlining the muscle’s boundaries; C and D. echogenicity and gray-level co-occurrence matrix (GLCM) features were measured in the short-axis view: a square region of interest (ROI) was selected, and the following texture parameters were calculated using ImageJ software: mean pixel intensity, angular second moment (ASM), contrast, correlation, entropy, and inverse difference moment (IDM); C. muscle image from a patient in the “probable sarcopenia” group; D. muscle image from a patient in the “no probable sarcopenia” group

Data analysis was performed using IBM SPSS Statistics 28, Statistica 10 (StatSoft, Tulsa, OK, USA), and R Statistical Software (v4.4.2; R Core Team 2024). Fisher’s exact test and the Mann–Whitney U test were used for categorical and continuous variables, respectively. Spearman’s rank correlation assessed associations between ultrasound-based muscle measurements and CGA results. Partial correlations were adjusted for age, sex, and BMI. Multivariable linear and logistic regression analyses were used for continuous and binary outcomes, respectively, with stepwise selection to identify significant predictors. All models were adjusted for age, sex, and BMI. Model assumptions and multicollinearity (via VIFs) were assessed.

Statistical significance was set at p <0.05, with False Discovery Rate (FDR) correction for multiple comparisons. Statistical power was estimated using G*Power.

Eighty-four patients were enrolled. In over half of the study population, the indication for surgery was a diagnosis of malignant neoplasm. The CGA detected frailty syndrome in over 57% of patients. The HGS and FTSST indicated that probable sarcopenia was present in almost 40% of patients (Tab. 2).

Patients with probable sarcopenia had significantly smaller MT and CSA of the RF than patients not suspected of having sarcopenia. Measurements reflecting muscle quality did not differ between the two groups (Fig. 2). Typically, ultrasound measurements showed that male patients had larger muscles than females, consistent with known physiological differences. However, no significant sex differences were observed in ultrasonographically evaluated muscle tissue quality (Tab. 3).

Fig. 2.

Comparison of ultrasonographic muscle measurements between the “probable sarcopenia“ and “no probable sarcopenia“ groups generated using R Statistical Software (v4.4.2; R Core Team 2024); Y-axis labels represent selected ultrasonographic measurements of the rectus femoris muscle; X-axis labels indicate comparison groups: patients without probable sarcopenia (“no”) and those with probable sarcopenia (“yes”). Effect sizes with 95% CI and statistical power (%) for comparisons between “probable sarcopenia“ and “no probable sarcopenia“ groups: A. 0.62 (0.16, 1.08), 78%; B. 0.52 (0.06, 0.98), 62%; C. 0.56 (0.10, 1.02), 65%; D. 0.35 (−0.11, 0.80), 31%; E. −0.16 (−0.62, 0.29), 10%; F. −0.27 (−0.72, 0.19), 20%; * p <0.05 (Mann–Whitney U test), adjusted for multiple comparisons using the false discovery rate (FDR) procedure. MT – rectus femoris muscle thickness (mm); CSA – rectus femoris muscle cross-sectional area (cm²); PA – pennation angle (°); EI – echogenicity: grayscale intensity of muscle tissue expressed in arbitrary units (AU), 0 – black, 255 – white, with higher values indicating poorer muscle quality

Patients classified as frail according to the CGA had smaller MT and CSA of both relaxed and contracted RF muscles compared to fit patients, and their muscle echogenicity was higher, indicating poorer muscle quality ( Tab. 4).

Comparison of ultrasound measurements with the CGA, adjusted for potential confounding variables (i.e., gender, age, and BMI), revealed that MT and CSA were significantly correlated with HGS (rho = 0.23, 95% CI [0.01, 0.42] and 0.22, 95% CI [0.01, 0.42], respectively), NRS (rho = −0.30, 95% CI [−0.49, −0.09] and −0.27, 95% CI [−0.49, −0.09], respectively) and ADL (rho = 0.26, 95% CI [0.05, 0.45] and 0.31, 95% CI [0.10, 0.49], respectively). Unexpectedly, FL showed a negative correlation with HGS (rho = −0.30, 95% CI [−0.48, −0.09]). Echogenicity was correlated positively with the Charlson comorbidity index (rho = 0.22, 95% CI [0.00, 0.41]) and the number of regular medications (rho = 0.26, 95% CI [0.05, 0.45]). In GLCM texture analysis, significant correlations were observed with HGS (rho for ASM = −0.26, 95% CI [−0.45, −0.04]; contrast = −0.28, 95% CI [−0.47, −0.07]; entropy = 0.25, 95% CI [0.04, 0.44]), with the number of regular medications (rho for ASM = 0.31, 95% CI [0.10, 0.49]; entropy = −0.32, 95% CI [−0.50, −0.11]; IDM = 0.32, 95% CI [0.12, 0.50]) and I-ADL (rho for contrast = 0.23, 95% CI [0.01, 0.42]; correlation = −0.28, 95% CI [−0.47, −0.07]; IDM = 0.23, 95% CI [0.01, 0.42]). Following FDR correction for multiple comparisons, only the correlations between HGS and MT and CSA remained statistically significant.

In the sex-disaggregated analysis, preliminary partial correlations adjusted for age and BMI indicated stronger associations between muscle measurements and physical fitness in men than in women. However, after FDR correction, none of the correlations in either group remained statistically significant.

MT and CSA in both relaxed and contracted states, PA, EI, and GLCM texture features were included in multivariable linear regression models for selected outcomes CGA outcomes. Potential confounding variables (age, sex, and BMI) were also incorporated. For handgrip strength, the fitted regression model including only significant predictors was: HGS = 57.864 − 0.413*(age [years]) − 13.817*(sex [woman vs man]) + 1.254* (CSA [cm²]). The model yielded an adjusted R² of 0.53. For gait speed, only PA was a significant predictor (gait speed [m/s] = 0.785 + 0.023*(PA [°])), although the model demonstrated limited explanatory power (adjusted R² = 0.05). For the FTSST, MT during contraction was the sole significant variable (FTSST [s] = 16.047 – 0.188*(contr.MT [mm])), and the model’s power was also low (adjusted R² = 0.06). In the model for the TUG test, none of the included variables reached statistical significance.

Logistic regression analyses were performed to evaluate associations between ultrasound-derived muscle parameters and the diagnosis of frailty syndrome and probable sarcopenia. For frailty, echogenicity and sex were identified as significant predictors. The model demonstrated modest explanatory power (Nagelkerke R² = 0.26). Specifically, each one-unit increase in muscle echogenicity was associated with a 6.7% (OR = 1.07; 95% CI [1.02, 1.12]) increase in the odds of frailty, while female sex, independent of echogenicity, was associated with a 3.18-fold higher likelihood of frailty (OR = 3.18; 95% CI [1.18, 8.57]). In contrast, the model for probable sarcopenia showed lower predictive power (Nagelkerke R² = 0.12), with MT emerging as the only significant variable. Each 1-mm increase in MT was associated with a 14.3% (OR = 0.86; 95% CI [0.76, 0.97]) reduction in the odds of probable sarcopenia.

Postoperative complications occurred in 39.3% of patients, with one-third classified as severe (graded 3+ on the Clavien-Dindo scale). Patients assessed as frail by the CGA tended to experience more postoperative adverse events and had higher CCI scores than fit ones, however after FDR adjustments these differences did not remain statistically significant (Tab. 5). There were no significant associations between ultrasonographic measurements and either the occurrence or severity of postoperative complications, In contrast, significant correlations were observed between CGA results and postoperative outcomes for the ADL, I-ADL, and the Charlson comorbidity index (Fig. 3).

Fig. 3.

Correlations between postoperative course, complication severity and components of the comprehensive geriatric assessment determined using R Statistical Software (v4.4.2; R Core Team 2024). Cell values represent Spearman’s rank correlation coefficients calculated using pairwise complete observations. Only correlations that remained statistically significant after adjustment for multiple comparisons using the false discovery rate (FDR) procedure are shown. ADL – activities of daily living; I-ADL – instrumental activities of daily living; CCI (0–37) – Charlson comorbidity index; ClavienD. (0–5) – Clavien-Dindo classification grade; CCI (0–100) – comprehensive complication index; LOH (days) – length of hospitalization; LOP (days) – length of postoperative stay

Logistic regression analysis was conducted to identify predictors of postoperative complications, incorporating variables such as individual CGA components, ultrasound muscle parameters, the presence of frailty and probable sarcopenia, cancer diagnosis, age, sex, and BMI. Among all examined variables, only cancer diagnosis emerged as a statistically significant predictor. Patients with cancer had 8.61 times higher odds (OR = 8.61, 95% CI [2.28, 32.52]) of experiencing postoperative complications compared to those without cancer.

There has been rising interest in the use of ultrasound for diagnosing sarcopenia, as it is non-invasive, accessible, and time-saving, with performance comparable to gold-standard modalities^(10,11). In 2021, the SARCUS working group proposed the main parameters for ultrasonographic muscle evaluation, namely, MT, PA, FL, EI, and CSA⁽¹²⁾. MT and CSA reflect muscle size. A larger PA allows for a higher number of fibers per unit of muscle volume and is associated with increased force production potential. Longer muscle fascicles allow increased muscle shortening during contraction and improved force generation. Increased muscle echogenicity indicates infiltration of intramuscular adipose and fibrotic tissue⁽¹³⁾. The RF is one of the most commonly selected muscles in the diagnostic workup of sarcopenia, due to its high susceptibility to sarcopenic changes and its tendency to exhibit pronounced, muscle-specific atrophy⁽¹⁴⁾.

Recent systematic reviews have concluded that MT, CSA, and echogenicity are the most reliable and frequently used predictors of muscle strength and function^(5,10). Fu et al. evaluated the diagnostic accuracy of ultrasonographic parameters for sarcopenia and found that RF MT and CSA demonstrated moderate diagnostic accuracy, whereas RF EI showed low accuracy. They suggested that combining CSA and EI may improve predictive performance. Our results indicate that patients suspected of sarcopenia had smaller RF muscles. Although median RF echogenicity was higher in sarcopenic patients, the difference was not statistically significant, likely due to limited statistical power. In multivariate analysis, only MT remained a significant predictor of probable sarcopenia, while muscle quality parameters did not show predictive value. The correlation between echogenicity and measures of muscle strength and physical performance has been consistently reported in the literature. In contrast, evidence supporting similar associations for parameters such as pennation angle or muscle fiber length with clinical outcomes remains limited⁽¹⁵⁾. Similar to our observations, Bunout et al. demonstrated that PA measurement did not contribute to muscle function assessment⁽¹⁶⁾. There are no standardized cut-off values for ultrasonographic parameters in diagnosing sarcopenia, and the proposed thresholds vary considerably across studies, largely depending on ultrasound settings and the characteristics of the studied population⁽¹⁷⁾.

In our study, muscle size − but not muscle quality − was more favorable in men than in women. Previous studies confirm that older women experience increased loss of muscle mass at a higher rate than men of the same age and exhibit greater fat deposition within muscles⁽¹⁸⁾. These sex-specific differences may be explained by hormonal influences such as estrogen decline in postmenopausal women that contribute to accelerated muscle mass loss and increased intramuscular fat infiltration. Also, testosterone levels, which support muscle anabolism, are generally higher in men and decline more gradually with age, potentially preserving muscle mass and strength for longer. The lack of sex-related differences in muscle quality in our sample may be attributed to limited statistical power to detect subtle variations. Recognition of sex-specific differences is essential for establishing appropriate diagnostic cut-off values for sarcopenia and for developing tailored interventions to optimize clinical outcomes both in men and women.

Motivated by previous reports highlighting the influence of various pathological conditions on muscle texture homogeneity⁽¹⁹⁾, we aimed to assess the potential utility of GLCM analysis in muscle quality evaluation. Wilkinson et al. demonstrated that GLCM-derived texture features of muscle tissue are associated with physical performance and strength in older and chronically ill patients⁽²⁰⁾. They concluded that greater texture homogeneity, characterized by higher ASM, correlation, IDM, lower entropy, and contrast, was indicative of better muscle function. In our study, however, GLCM parameters showed only weak correlations with HGS, none of which remained statistically significant after applying FDR correction for multiple comparisons. Furthermore, these correlations were counterintuitive, with greater homogeneity in muscle texture (i.e., higher ASM and IDM values) being associated with lower HGS.

Interestingly, Tang et al. reported that texture homogeneity of the rectus femoris increased (i.e., contrast decreased) across groups ranging from non-sarcopenic to sarcopenic older adults, suggesting that generalized myosteatosis may homogenize local muscle texture. Concurrently, the sarcopenic group exhibited higher entropy, possibly reflecting disruptions in muscle architecture due to fibrosis or fat infiltration. In contrast, Wilkinson et al. proposed that fat infiltration, as indicated by increased echogenicity, results in higher contrast and entropy, implying greater heterogeneity. Similarly, Mirón-Mombiela et al. demonstrated that increased texture heterogeneity is indicative of muscle dysfunction related to frailty⁽²¹⁾. Collectively, these findings highlight inconsistencies in the interpretation of texture metrics and their practical implications. In our cohort, GLCM analysis appeared less informative and more time-consuming than conventional echogenicity assessment. While GLCM methods may be less sensitive to scanner-specific settings, their clinical utility remains uncertain, and further methodological standardization is needed before GLCM-based texture analysis can be reliably integrated into routine practice.

The gold standard for identifying frailty is the CGA, an in-depth evaluation of individual health status and care needs⁽⁷⁾. Studies have shown correlations between ultrasound-derived muscle thickness and frailty assessment tools such as the Frailty Fried Phenotype, Frailty Index, and Clinical Frailty Scale, suggesting that ultrasound can be a valuable adjunct in frailty diagnosis, although the strength of association varies depending on the instrument used^(22,23). Ultrasound has also proven to be a reliable method for assessing muscle size and detecting muscle wasting in older adults, reflecting declines in muscle function relevant to frailty⁽¹⁷⁾.

Our findings indicate that patients classified as frail by the CGA tended to have smaller RF muscles with higher echogenicity compared to their fit counterparts. Ultrasonographically measured MT and CSA demonstrated a modest positive correlation with HGS. After adjusting for age and sex, each 1 cm² increase in muscle CSA was associated with a 1.25 kg increase in HGS. Additionally, greater muscle echogenicity was independently associated with higher odds of being classified as frail. Other regression models incorporating various ultrasonographic parameters as potential predictors of CGA outcomes were either not statistically significant or exhibited limited predictive accuracy.

Muscle ultrasound has been applied less extensively in the assessment of frailty compared to sarcopenia, and its correlations with frailty features have proven less consistent. A narrative review on the topic advised against relying solely on ultrasound measures for frailty screening⁽¹⁷⁾. In our sample, subgroup analysis revealed stronger correlations between muscle ultrasound measurements and physical fitness in men compared to women. Lower muscle mass in females, combined with higher total body fat and intramuscular fat infiltration, may influence the magnitude of age- and sarcopenia-related changes, potentially attenuating correlations between ultrasound parameters and functional capacity.

Sarcopenia has been recognized as an independent risk factor for morbidity, mortality, and prolonged hospital stay after gastrointestinal oncologic surgery⁽²⁴⁾. Schneider et al. showed that muscle mass correlates with length of hospital stay and CCI, and that muscle quality predicts adverse outcomes after oncologic colon surgery⁽²⁵⁾. Some authors have considered the assessment of sarcopenia as a more resource-efficient predictor of postoperative adverse events compared to the evaluation of frailty⁽²⁶⁾. Galli et al. reported an association between major complications and reduced ultrasound-derived rectus femoris CSA in oncologic patients, demonstrating a predictive value comparable to that of CSA measured by computed tomography⁽²⁷⁾. We did not replicate these findings, as no significant associations were observed between ultrasonographic muscle measurements and postoperative outcomes. Consistent with our observations, three of six studies included in a recent narrative review on the prognostic value of muscle ultrasound in oncologic surgery found no statistically significant associations between ultrasonographic parameters and short-term postoperative outcomes⁽²⁸⁾. Furthermore, the mechanisms linking sarcopenia to adverse surgical outcomes have yet to be fully characterized⁽²⁹⁾. Proposed contributors include malnutrition and systemic inflammation, which impair wound healing and immune function, increasing the risk of infections and anastomotic leaks. Reduced muscle mass may also compromise respiratory function and delay mobilization, raising the risk of pulmonary and thromboembolic complications. Şengül Ayçiçek G et al. showed that sarcopenia was not associated with adverse outcomes after surgery for gastrointestinal malignancies, and frailty had superior predictive value⁽²⁾. Similarly, our results indicate that frail patients tended to experience a higher rate and greater severity of complications compared to fit patients. However, these differences did not remain statistically significant after FDR adjustments. In this cohort, only a small number of patients experienced more than one complication, and the majority of complications were classified as non-severe. This limited severity and frequency of events may have reduced the power to detect statistically significant differences between the groups. The CCI score demonstrated a moderate positive correlation with the severity of complications, while the I-ADL score showed a moderate negative correlation with the Clavien-Dindo grade, as well as with the length of hospital stay. These findings support the potential value of preoperative disease burden and functional independence as components of the CGA in anticipating postoperative outcomes. However, in multivariate analyses, only a cancer diagnosis emerged as a predictor of postoperative complications.

This study has several limitations. First, panoramic ultrasound imaging was unavailable, which occasionally required estimation of muscle borders when the entire muscle area was not visible within a single field of view. Second, the relatively small sample size limited the study’s ability to detect differences in certain subgroup analyses by sex and frailty, and also reduced the statistical power to predict postoperative complications. Therefore, non-significant findings in these stratified analyses should be interpreted in light of the reported effect sizes and associated statistical power. Third, patients were examined at different times of day, which may have introduced variability related to hydration status, potentially affecting muscle measurements. Fourth, although all four examiners independently acquired 15 skeletal muscle images from five pilot participants to ensure interrater agreement in both image acquisition and measurement procedures, measurement data were not retained and intraclass correlation coefficients were not calculated, precluding formal assessment of interrater reliability. Larger, more standardized studies are needed to confirm and expand upon these findings.