From standard laboratory parameters to skeletal muscle mass: A novel prediction model for chronic kidney disease.

BACKGROUND AND AIMS: Sarcopenia is a critical problem in patients with chronic kidney disease (CKD). Longitudinal assessment of skeletal muscle mass is essential for early detection and intervention. However, conventional standard assessment methods for measuring skeletal muscle mass require specialized equipment, which might not be practical in routine care settings. The aim of this study was to construct a predictive model for skeletal muscle mass using readily available variables from the routine clinical care for patients with CKD.

METHODS

This cross-sectional study included patients with CKD who underwent kidney biopsy and non-contrast abdominal computed tomography. Skeletal muscle mass was estimated by the skeletal muscle index, calculated as the cross-sectional area of muscle at the third lumbar vertebra level, normalized by height. The creatinine muscle index (CMI), calculated as the product of serum creatinine and cystatin C-based estimated glomerular filtration rate (eGFR), was included as a variable in the analysis. A predictive model was developed using Least Absolute Shrinkage and Selection Operator (LASSO) regression with 10-fold cross-validation to optimize the regularization parameter (λ). Variables selected by LASSO regression were used for constructing an ordinary least squares (OLS) regression model (Model 1), followed by a second model (Model 2) that included only variables with p < 0.05. Model performance was evaluated by adjusted R2, mean absolute percentage error, root mean square error, and mean Winkler interval score with 90 % prediction coverage (PC). Bootstrap resampling was used to calculate 95 % confidence intervals (CIs).

RESULTS

Between June 2018 and March 2024, 111 patients (56 female individuals) were enrolled. The median age and creatinine-based eGFR were 55 years and 56 mL/min/1.73 m2, respectively. Model 1, derived from LASSO and OLS regressions, included log-transformed age, male sex, body mass index (BMI), CMI, serum albumin level, and log-transformed creatine kinase level. Model 2 included only male sex, BMI, and CMI. The adjusted R2 and PC were 0.749 (95 % CI: 0.663-0.835) and 92.0 % for Model 1, and 0.739 (95 % CI: 0.652-0.825) and 92.24 % for Model 2, respectively. No significant differences were observed between the models across all metrics, despite the simplicity of the variables included in Model 2.

CONCLUSION

A simple and reasonable predictive model for skeletal muscle mass was developed. This model solely comprised parameters routinely obtained in the daily clinical care for patients with CKD, allowing for longitudinal monitoring and early detection of muscle loss.