Five‐Year CT‐Based Skeletal Muscle Decline Curves in Cirrhosis According to Liver Functional Reserve and Routinely Available Clinical Factors Suguru Miida, Hiroteru Kamimura, Hitoshi Yoshiji, Masahito Shimizu, Takumi Kawaguchi, Tetsuya Tsuji, Saori Endo, Hiroki Natsui, Hiroki Maruyama, Hiroyuki Abe, Tadashi Namisaki, Takao Miwa, Dan Nakano, Michiyuki Kawakami, Shuji Terai Abstract ABSTRACTBackgroundSarcopenia is common in cirrhosis and is associated with poor outcomes. We aimed to identify factors associated with longitudinal skeletal muscle loss and to explore estimation of future L3 skeletal muscle index (L3‐SMI) using routinely available clinical variables.MethodsThis retrospective longitudinal study included 175 patients with cirrhosis who underwent serial abdominal CT over five annual time points between 2010 and 2022. The primary analysis used a linear mixed‐effects model including annual time point, baseline Child–Pugh class, and their interaction. An exploratory linear regression model estimated Year 5 L3‐SMI.ResultsIn the primary model, L3‐SMI declined by −1.144 cm2/m2/year in baseline Child–Pugh class A. The time‐by‐class B interaction was not significant (β = 0.024; 95% CI, −0.250 to 0.298; p = 0.865), whereas class C was associated with an additional annual decline of −1.812 cm2/m2/year compared with class A (95% CI, −2.735 to −0.890; p = 0.0001). This association remained after adjustment for age, sex, baseline HCC status, and the time‐by‐HCC interaction. In the exploratory Year 5 model, baseline L3‐SMI was the dominant predictor; estimates for class C were interpreted cautiously because only five patients were baseline class C.ConclusionsBaseline hepatic functional reserve, particularly Child–Pugh class C, was associated with faster annual CT‐based L3‐SMI decline. The Year 5 prediction model is exploratory and requires external validation. Early sarcopenia assessment may facilitate risk stratification in patients with impaired hepatic reserve. Trial Registration: UMIN Clinical Trials Registry: UMIN000051577. Introduction Sarcopenia is a progressive skeletal muscle disorder associated with frailty, falls, functional decline, impaired quality of life, and mortality [, ]. In cirrhosis, sarcopenia is common and prognostically important. A meta‐analysis of 22 cohort studies reported a prevalence of 37.5% and an independently increased mortality risk in patients with cirrhosis []. A recent CT‐based meta‐analysis of 39 studies confirmed a 44% prevalence and approximately twofold higher mortality risk, supporting the clinical relevance of EASL/AASLD CT‐based criteria []. These findings highlight the need to identify patients at high risk of progressive muscle loss. Liver rehabilitation, which includes exercise therapy, nutritional support, education, and psychosocial care, is increasingly recognized as a strategy to improve physical function, symptoms, prognosis, and quality of life in patients with steatotic liver disease, cirrhosis, and hepatocellular carcinoma (HCC) []. However, effective implementation requires longitudinal risk stratification. The Japanese Society of Hepatology recommends grip strength, bioelectrical impedance analysis, and CT‐derived skeletal muscle area at the third lumbar vertebra normalized by height squared as the L3 skeletal muscle index (L3‐SMI) for sarcopenia assessment []. The EASL guidelines also recognize cross‐sectional imaging as an objective method for sarcopenia assessment in cirrhosis []. The EASL/AASLD CT cutoffs were proposed by Carey et al. [], and their prognostic value has been supported by recent meta‐analysis []. Despite the usefulness of CT‐based L3‐SMI, repeated quantification may not be feasible in all clinical settings, and determinants of individual skeletal muscle trajectories remain incompletely defined. Cirrhosis‐related sarcopenia is influenced by disease‐specific mechanisms such as hyperammonaemia, altered autophagy, impaired protein metabolism, and reduced anabolic signaling []. We therefore aimed to identify factors associated with progressive skeletal muscle decline in cirrhosis and to develop a clinically usable framework for estimating longitudinal L3‐SMI using routine clinical information and CT‐based muscle assessment. Patients and Methods Study Design and Population This longitudinal retrospective study included patients with cirrhosis followed at Niigata University Medical and Dental Hospital between January 2010 and December 2022. Cirrhosis was diagnosed by board‐certified hepatologists based on clinical, laboratory, and imaging findings. Patients were eligible if they had continuous follow‐up for at least 5 years and at least one abdominopelvic CT examination annually during the 5‐year observation period. A total of 175 patients met these criteria. Clinical Care and Cohort Characteristics CT was performed primarily for HCC surveillance or post‐treatment monitoring, and laboratory tests and physical examinations were performed approximately concurrently with CT. Hepatic functional reserve was assessed using Child–Pugh class, modified albumin‐bilirubin (mALBI) grade, and Model for End‐Stage Liver Disease (MELD) score. Some patients had a history of locoregional therapy, systemic chemotherapy, or surgery for HCC, but all had an Eastern Cooperative Oncology Group performance status of 0. Alcohol intake was defined as regular consumption exceeding 20 g/day of ethanol for women and 30 g/day for men. Because serial CT was required, this cohort should be interpreted as a clinically relevant subgroup of patients with cirrhosis undergoing HCC surveillance or controlled post‐treatment follow‐up, rather than as an unselected cirrhosis population. Although completion of 5‐year follow‐up and preserved performance status reduced the likelihood of uncontrolled cancer‐related cachexia, the influence of previous HCC, recurrence, or anti‐cancer treatment on skeletal muscle trajectories could not be fully excluded. Skeletal Muscle Mass Assessment Skeletal muscle area at L3 was measured on axial CT images using SliceOmatic software (version 5.0, TomoVision Inc., Montreal, Canada). Muscles were quantified within −29 to +150 HU [], with manual correction when necessary. Three physicians blinded to clinical information independently reviewed the images, and the final L3 muscle area was determined by consensus. Parameter Definitions and Longitudinal Assessment The L3‐SMI derived from the first CT was defined as baseline L3‐SMI or L3‐SMI (1y), and Child–Pugh class at that time was recorded as the baseline Child–Pugh grade. CT was performed at approximately 1, 2, 3, 4, and 5 years from the first annual CT assessment. Because CT intervals varied, L3‐SMI values were standardized to yearly time points by linear interpolation between consecutive CT examinations, as previously reported []. The primary longitudinal outcome was annualized change in L3‐SMI across the five annual CT time points. The worst Child–Pugh grade and worst mALBI grade were used only for exploratory descriptive analyses, not as baseline predictors. The annualization procedure and analytical hierarchy are shown in Figure . MELD Score The MELD score was calculated using total bilirubin, creatinine, and the international normalized ratio as follows: MELD = 3.78 × ln(bilirubin [mg/dL]) + 11.2 × ln(INR) + 9.57 × ln(creatinine [mg/dL]) + 6.43. Values < 1.0 were set to 1.0, and creatinine was capped at 4.0 mg/dL, including dialysis. MELD is commonly used to assess liver disease severity and transplant priority []. MELD scores were categorized into quartiles: Q1 < 9.485, Q2 9.485 to < 12.171, Q3 12.171 to < 15.178, and Q4 ≥ 15.178. Physical Activity Assessment Physical activity (PA) was assessed in 57 patients using the short International Physical Activity Questionnaire []. Weekly energy expenditure was calculated as MET‐min/week using 8.0, 4.0, and 3.3 MET for vigorous activity, moderate activity, and walking, respectively []. Participants were classified as low (< 600 MET‐min/week), moderate (600–3000 MET‐min/week), or high (> 3000 MET‐min/week). Hepatic functional deterioration was defined as progression from Child–Pugh A to B/C or from Child–Pugh B to C during follow‐up. Statistical Analysis Linear mixed‐effects models were fitted with L3‐SMI as the dependent variable and a patient‐specific random intercept to account for repeated measurements within individuals. To directly evaluate differences in longitudinal skeletal muscle decline, the primary model included annual time point, baseline Child–Pugh class, and the time‐by‐Child–Pugh interaction as fixed effects. Additional models were adjusted for age and sex, and a sensitivity model further included baseline HCC status and the time‐by‐HCC interaction. Alternative longitudinal models used baseline mALBI grade or MELD quartiles instead of Child–Pugh class to assess whether the association between liver functional reserve and annual L3‐SMI decline was consistent across different liver reserve indices. Sex was coded as female = 0 and male = 1; Child–Pugh class A, mALBI Grade 1, and MELD Q1 were used as the respective reference categories. Individual least‐squares slopes of L3‐SMI were calculated for each patient and summarized according to the maximum Child–Pugh class observed during follow‐up; because this classification was determined after baseline, these analyses were treated as exploratory and descriptive and were not used for causal inference. An exploratory linear regression model was used to estimate Year 5 L3‐SMI using baseline L3‐SMI, sex, BMI, age, and baseline Child–Pugh class. The original prediction equation derived in the initial analysis was retained for continuity, but the model was considered hypothesis‐generating and not used for primary inference. Variables were selected a priori based on clinical availability, biological plausibility, and avoidance of overfitting. Comparisons presented in Table were descriptive and exploratory; p‐values were not adjusted for multiplicity and should therefore be interpreted cautiously. Linear mixed‐effects models were fitted using R version 4.4.1, and other analyses were performed using GraphPad Prism version 9.0. A two‐sided p‐value < 0.05 was considered statistically significant. In addition, exploratory cross‐sectional linear models were used to examine associations between baseline SMI and age or liver reserve indices, with adjustment for sex, albumin, prothrombin time, creatinine, loop diuretic use, and l‐carnitine supplementation. These adjusted analyses were reported in Table and were kept separate from the primary longitudinal inference. Unadjusted cross‐sectional age–SMI associations according to liver reserve index are shown in Figure . Ethical Considerations This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Niigata University (approval number: 2023‐0009). The requirement for consent for retrospective data use was waived through an opt‐out approach, and written informed consent was obtained from patients who completed the IPAQ. Results Patient Background and L3‐SMI According to Liver Functional Reserve Table summarizes the baseline characteristics of the 175 patients. The most common etiology was hepatitis C virus infection, followed by alcohol‐related liver disease, metabolic dysfunction‐associated steatohepatitis, and hepatitis B virus infection. Table presents baseline characteristics according to baseline Child–Pugh class. These comparisons were descriptive and unadjusted for multiplicity. Only five patients had baseline Child–Pugh class C. As Child–Pugh class progressed, albumin, cholinesterase, and prothrombin time decreased, whereas total bilirubin, ALBI score, MELD score, and FIB‐4 index increased. Significant intergroup differences were also observed for hemoglobin, platelet count, LDH, ALP, LDL‐cholesterol, and HDL‐cholesterol, whereas AST, ALT, and sex‐specific L3‐SMI did not differ significantly; sodium was slightly lower in class B than in class A in this unadjusted descriptive comparison. Exploratory unadjusted cross‐sectional associations between age and L3‐SMI according to Child–Pugh class, mALBI grade, and MELD quartiles are shown in Figure . In exploratory adjusted cross‐sectional models, older age was inversely associated with baseline SMI (β approximately −0.512 to −0.529 across the Child–Pugh, mALBI, and MELD models), whereas male sex and albumin were positively associated with baseline SMI. Loop diuretic use was consistently associated with lower baseline SMI (β approximately −1.233 to −1.286). These adjusted models are summarized in Table . Primary Longitudinal Mixed‐Effects Analysis of Annual L3‐SMI Decline The primary mixed‐effects model included annual time point, baseline Child–Pugh class, and their interaction terms (Panel A of Table ). In baseline Child–Pugh class A, L3‐SMI declined by −1.144 cm2/m2/year (95% CI, −1.276 to −1.011; p < 0.0001). The time‐by‐Child–Pugh class B interaction was not significant (β = 0.024; 95% CI, −0.250 to 0.298; p = 0.865), whereas the time‐by‐Child–Pugh class C interaction was significant (β = −1.812; 95% CI, −2.735 to −0.890; p = 0.0001). The estimated annual declines were approximately −1.144, −1.120, and −2.956 cm2/m2/year for baseline Child–Pugh classes A, B, and C, respectively. These standardized slope‐based trajectories are illustrated in Figure . The direction of the findings was unchanged after adjustment for age and sex (Panel A of Table ). In a sensitivity model including baseline HCC status and the time‐by‐HCC interaction, the time‐by‐Child–Pugh class C interaction remained significant (β = −1.963; 95% CI, −2.901 to −1.026; p < 0.0001), whereas the time‐by‐HCC interaction did not reach significance (β = −0.225; 95% CI, −0.476 to 0.027; p = 0.080) (Panel B of Table ). Models using baseline mALBI grade or MELD quartiles showed overall annual decline, but their time interaction terms were less consistent than those for baseline Child–Pugh class (Panels B and C of Table ). Sensitivity analyses are summarized in Figure . Exploratory Comparison of Individual Slopes Using the Maximum Child–Pugh Class Observed During Follow‐Up In an exploratory descriptive analysis, individual L3‐SMI slopes were estimated by least‐squares regression and summarized according to the maximum Child–Pugh class observed during follow‐up. Because this classification was determined during the observation period, it was not used as a baseline predictor and should not be interpreted causally. Using this descriptive classification, the cohort included 77 Child–Pugh A, 62 Child–Pugh B, and 36 Child–Pugh C patients. Among men, annual L3‐SMI decline was −0.991 ± 0.727, −1.296 ± 0.940, and −2.485 ± 2.140 cm2/m2/year in Child–Pugh A, B, and C, respectively. Among women, corresponding values were −0.928 ± 0.606, −1.021 ± 1.042, and −2.017 ± 0.951 (Figure ). Alcohol Etiology and Physical Activity Among men who remained Child–Pugh A, muscle decline was numerically greater in alcohol‐related cirrhosis than in other aetiologies, but the difference was not significant (−1.09 ± 0.82 vs. −0.94 ± 0.69 cm2/m2/year; p = 0.51) (Figure a). PA data were available for 57 patients: 30 low, 17 moderate, and 10 high. Hepatic functional deterioration occurred in 22.2% (6/27) of the moderate/high PA group and 40.0% (12/30) of the low PA group (p = 0.149) (Figure b). Exploratory Model for Estimating Year 5 L3‐SMI Table shows the exploratory linear regression model for estimating Year 5 L3‐SMI. Baseline L3‐SMI was significantly associated with Year 5 L3‐SMI (β = 0.731; 95% CI, 0.317–1.145; p = 0.0006). Compared with Child–Pugh class A, baseline Child–Pugh class C was associated with lower predicted Year 5 L3‐SMI (β = −23.085; 95% CI, −40.821 to −5.349; p = 0.0111), whereas class B was not significant (β = −4.077; 95% CI, −11.022 to 2.868; p = 0.2480). Sex, BMI, and age were retained as clinically relevant covariates but were not individually significant. Because only five patients had baseline Child–Pugh class C, estimates involving class C in the prediction model were interpreted cautiously and not used for primary inference. The regression equation was retained from the original exploratory model: Predicted L3‐SMI (5y) = −1.191 + 0.731 × L3‐SMI (1y)—2.609 (if male) + 0.018 × BMI—0.033 × age—4.077 (if Child–Pugh B)—23.085 (if Child–Pugh C). Apparent model performance in the derivation cohort showed R 2 0.92, adjusted R 2 0.91, and RMSE 6.46 cm2/m2. Sensitivity analyses showed that the Child–Pugh term was not significant when classes B and C were combined (β = 0.389; p = 0.645) or when class C was excluded (Child–Pugh B: β = 0.710; p = 0.401). Therefore, the equation is presented as an exploratory risk‐stratification framework rather than as a validated prediction model. An illustrative example based on this exploratory equation is shown in Figure . Discussion To our knowledge, this is the first longitudinal study to evaluate long‐term rates of skeletal muscle decline using CT in patients with cirrhosis with a primary mixed‐effects framework that directly tested annual L3‐SMI decline according to baseline liver functional reserve. The present cohort should be interpreted as a real‐world cirrhosis cohort under HCC surveillance and/or controlled post‐treatment follow‐up rather than as a fully unselected cirrhosis population. Sensitivity analysis adjusted for baseline HCC status and the time‐by‐HCC interaction, and the association between baseline Child–Pugh class C and faster annual L3‐SMI decline remained. However, detailed time‐dependent information on HCC recurrence and anti‐cancer treatment was not uniformly available; residual confounding by HCC‐related factors cannot be fully excluded. Skeletal muscle mass physiologically declines by approximately 1.4% annually in the general older population []. By contrast, more rapid skeletal muscle loss has been reported in patients with cirrhosis []. This accelerated decline is likely attributable to disease‐specific mechanisms, including protein‐energy malnutrition caused by reduced dietary intake, malabsorption, impaired hepatic protein synthesis, and metabolic dysregulation [, , , ]. In addition, with the increasing prevalence of obesity worldwide, sarcopenic obesity has become a highly relevant phenotype in chronic liver disease []. In the present cohort, the mean BMI was ≥ 24 kg/m2 in Child–Pugh A and B and ≥ 27 kg/m2 in Child–Pugh C, suggesting that obesity may coexist with skeletal muscle depletion in patients with cirrhosis []. A consistent finding across the longitudinal models was that L3‐SMI declined over time and that older age was associated with lower L3‐SMI. In the primary Child–Pugh interaction model, baseline Child–Pugh class C showed significantly faster annual decline than class A, whereas class B did not. This supports the clinical importance of both aging and advanced hepatic dysfunction in skeletal muscle depletion among patients with cirrhosis. The association between hepatic functional reserve and L3‐SMI decline depended on the scoring system. Baseline Child–Pugh class showed the clearest interaction with time, whereas the interaction terms for mALBI grade and MELD quartiles were less consistent. The Child–Pugh classification may capture clinically relevant features such as ascites and encephalopathy in addition to laboratory parameters, whereas mALBI and MELD may reflect partly different dimensions of liver function []. The sex‐specific findings warrant careful interpretation. In the baseline stratified analysis, sex‐specific L3‐SMI did not differ significantly across Child–Pugh classes, although numerical differences were observed. In the exploratory sex‐specific slope analysis, men with maximum Child–Pugh class C showed numerically steeper annual declines in L3‐SMI than those with class A or B, whereas the pattern was less clear among women. These exploratory findings should be interpreted cautiously and do not establish a sex‐specific interaction. Nevertheless, the observed numerical pattern suggests that the relationship between hepatic dysfunction and muscle loss may differ by sex and warrants evaluation in larger cohorts. Men with cirrhosis may be particularly vulnerable to rapid muscle depletion despite greater absolute muscle mass, whereas women may have lower baseline muscle reserve and distinct hormonal influences on muscle maintenance [, ]. Future studies should consider sex‐specific patterns when evaluating sarcopenia progression in cirrhosis. Albumin was consistently positively associated with baseline SMI in the exploratory adjusted cross‐sectional models. This finding is consistent with the close relationship between hepatic synthetic function, nutritional status, and preservation of skeletal muscle mass. Albumin contributes to oncotic pressure, amino acid transport, and protein turnover, and hypoalbuminemia may reflect or contribute to a catabolic state that promotes muscle loss in cirrhosis [, , , ]. Accordingly, lower albumin may be closely linked to skeletal muscle depletion, supporting nutritional assessment and optimization of protein–energy balance in cirrhosis. Loop diuretic use was also consistently associated with lower baseline SMI in the exploratory adjusted cross‐sectional models. This association is biologically plausible and consistent with previous reports suggesting that loop diuretics may contribute to muscle loss through electrolyte depletion, mitochondrial dysfunction, or increased urinary nitrogen excretion [, , ]. However, because we assessed only the presence or absence of loop diuretic use and did not evaluate dose or duration, causality cannot be inferred. These findings support careful evaluation of diuretic therapy in patients at risk of sarcopenia and future studies examining dose‐dependent effects and alternative agents such as mineralocorticoid receptor antagonists or tolvaptan. In the subgroup analysis of men who remained Child–Pugh A during follow‐up, alcohol‐related cirrhosis was associated with numerically greater muscle loss, but the difference was not statistically significant. Because this subgroup was small, the analysis should be interpreted cautiously. Nonetheless, the direction of the trend is biologically plausible, given the known roles of ethanol‐related mitochondrial dysfunction and endotoxin‐associated pathways in alcohol‐related muscle wasting [, ]. Our findings do not establish alcohol etiology as an independent determinant of muscle decline, but they suggest that this relationship warrants further investigation in larger cohorts with more detailed longitudinal assessment of drinking behavior. Physical activity may also have an important protective role. In the IPAQ subgroup, hepatic functional deterioration was numerically more frequent in the low‐activity group than in the moderate/high‐activity group, although the difference did not reach significance. This negative result should be interpreted in the context of limited sample size. The observed trend is compatible with evidence that physical activity in cirrhosis is generally lower than in healthy controls [] and supports further prospective evaluation of structured exercise interventions. With the increasing prevalence of MASH‐related cirrhosis [], exercise therapy is receiving growing attention. In fatty liver disease, exercise can improve steatosis and metabolic dysfunction, and current guidelines support weight reduction and lifestyle intervention in overweight or obese patients [, , , , , ]. In addition, experimental and epidemiologic studies suggest that regular physical activity may contribute to a lower risk of HCC and liver‐related mortality [, ]. At the same time, caution remains necessary in patients with advanced cirrhosis because exercise may increase portal pressure or aggravate renal dysfunction in susceptible individuals [, ]. Nevertheless, prior interventional studies suggest that carefully selected and appropriately supervised exercise programs can be implemented safely in patients with compensated cirrhosis [, ]. In this context, a clinically useful model for identifying patients at high risk of skeletal muscle loss could help guide the timing and intensity of individualized exercise‐based interventions. Our exploratory prediction framework suggests that estimation of future L3‐SMI from baseline clinical status may help identify patients who require individualized intervention. We retained the original exploratory prediction equation to preserve continuity with the derivation analysis, but the revised interpretation clearly positions it as preliminary and hypothesis‐generating. Baseline L3‐SMI was the strongest predictor of predicted 5‐year L3‐SMI, and baseline Child–Pugh class C was associated with a markedly lower predicted value; however, only five patients were baseline Child–Pugh class C. In addition, sensitivity analyses combining classes B and C or excluding class C did not show a significant Child–Pugh effect in the prediction model. Accordingly, the equation should not be regarded as a validated clinical prediction tool, and external validation is required before broader clinical implementation. This study has some limitations. First, this was a single‐center retrospective study of patients who underwent serial CT mainly for HCC surveillance or posttreatment monitoring. Although baseline HCC status was included in sensitivity analysis, detailed time‐dependent data on HCC recurrence, anti‐cancer treatment type, timing, and cumulative treatment burden were not uniformly available. Second, only five patients had baseline Child–Pugh class C; estimates involving baseline class C should therefore be interpreted cautiously, particularly in the exploratory Year 5 prediction model. Third, disease duration was not incorporated into the analyses. Fourth, the physical activity and alcohol subgroup analyses were limited by sample size. Finally, multiple variables were compared descriptively in Table without adjustment for multiplicity. Future prospective multicenter studies with larger numbers of patients with advanced cirrhosis, detailed HCC treatment trajectories, and external validation are needed. In addition, the adjusted cross‐sectional models for baseline SMI, including albumin and loop diuretic use, were exploratory and should not be interpreted as evidence of longitudinal causality. Conclusions In this study, baseline Child–Pugh class C was associated with faster annual decline in CT‐based L3‐SMI in patients with cirrhosis, and this association remained after adjustment for baseline HCC status. The Year 5 prediction model was exploratory and mainly driven by baseline L3‐SMI, underscoring the importance of early sarcopenia assessment. These findings provide a clinically usable framework for risk stratification and support CT‐based muscle assessment to identify patients with impaired hepatic reserve who may require more intensive monitoring and individualized management.