Table 4.

ROC curves of multivariable logistic models including clinical parameters plus each parameter^a

Model including	ROC area (95% CI)
Clinical parameters + SMM/WT + HGS	0.88 (0.83–0.94)b
Clinical parameters + SMM/WT	0.88 (0.83–0.93)b
Clinical parameters + SMM to squared height	0.85 (0.78–0.91)
Clinical parameters + ECW to TBW	0.81 (0.74–0.88)
Clinical parameters + ICW to TBW	0.81 (0.74–0.89)
Clinical parameters + PA	0.81 (0.74–0.89)
Clinical parameters + HGS	0.81 (0.74–0.88)
Clinical parameters only	0.81 (0.74–0.88)

Parameters measured by bioimpedance analysis were included as continuous variables in each multivariable logistic model. Clinical parameters were age, sex, cardiovascular comorbidities (chronic heart failure or ischemic heart disease), diabetes mellitus, cardiac index, and percentage of ultrafiltration to body weight. The AUC was statistically higher in models including the percentage of SMM (kg) to WT (kg) compared with those of other models.

AUC, area under the ROC curve; CI, confidence interval; ECW, extracellular water; HGS, handgrip strength; ICW, intracellular water; PA, phase angle; ROC, receiver operating characteristic; SMM, skeletal muscle mass; SMM/WT, percentage of SMM to WT; TBW, total body water; WT, dry body weight.

^aSuch as the percentage of SMM(kg) to WT (kg) (SMM/WT) plus HGS (kg•f) (model 1), SMM/WT (%) (model 2), SMM to height² (kg/m²) (model 3), ECW to TBW (kg/kg) (model 4), ICW to TBW (kg/kg) (model 5), PA at 50 kHz (model 6), or HGS (kg•f) (model 7) and only clinical parameters (model 8).

^bROC area of the model was higher than that of the other models (p < 0.05).