Validation and refinement of a biomarker panel for frailty assessment and prediction of muscle weakness in older adults

Fig. 1. Distribution of J-CHS scores in pre-frail and frail subjects. Proportions of the five J-CHS components—shrinking, weakness, exhaustion, slowness, and low activity—in pre-frail (blue) and frail (red) subjects are shown as observed (solid lines) and expected (dashed lines) values. An asterisk indicates a statistically significant difference between the observed and the expected values for each criterion (P < 0.05), with the corresponding J-CHS components underlined.

Fig. 2. Associations between frailty-related factors and each J-CHS criterion. Box plots show the associations between each J-CHS criterion (1 or 0; x axis) and the frailty-related factors (y axis), displaying the minimum, lower quartile, mean, upper quartile, and maximum values. Statistical significance was assessed by logistic regression, with P < 0.05 considered significant. An asterisk indicates P < 0.05; NS indicates not significant. Statistical significance was evaluated for each criterion, and the sum of the frailty-related factors is shown in the right column as the total J-CHS score.

Fig. 3. Biomarkers associated with each J-CHS criterion. a All possible combinations of the six biomarkers were considered for model construction using five-fold cross-validation. The detailed combinations, along with the mean AUCs and 95% confidence intervals (CIs) of the top five models are shown. “Base” represents the model adjusted only for age, sex, and body mass index (BMI). An asterisk indicates a statistically significant difference between the Rank 2 model and the Base model (P < 0.05). b One clinical variable (skeletal muscle mass index [SMI]; shown in black bold letters) and two age-related factors (apelin and growth differentiation factor 15 [GDF15]; shown in white bold letters) were associated with the five J-CHS criteria.

Fig. 4. Application of identified biomarkers to prospective data. In the prospective cohort of robust subjects, participants were stratified into high-risk (pink) and low-risk (light blue) groups for shrinking (a) and weakness (b) on the basis of the identified biomarkers: apelin and growth differentiation factor 15 (GDF15) for shrinking; and skeletal muscle mass index (SMI), apelin, and GDF15 for weakness. The optimal cutoff values were determined by using the minimum P value from the log-rank test on Kaplan–Meier curves. No statistically significant differences were observed between the high- and low-risk groups with the combination of apelin and GDF15 for shrinking (a), but statistically significant between-group differences were detected with the combination of SMI, apelin, and GDF15 for weakness (b).