Predictive accuracy of phase angle for short-term and stage-specific mortality in cancer

• Phase angle (PhA) values were consistently lower among 6-month non-survivors, and remained an independent predictor of 6-month mortality after multivariable adjustment. • Sex-specific PhA cutoffs associated with 6-month mortality were ≤ 5.43° for males and ≤ 4.22° for females • PhA showed fair predictive accuracy across TNM stages, with the highest accuracy in stage IV disease and a cutoff point of ≤ 4.72°. Cancer-related factors can negatively affect nutritional status and impact disease trajectory. Prognostic biomarkers such as bioelectrical impedance-derived phase angle (PhA) may help identify patients at higher risk of adverse outcomes. This study aimed to determine optimal PhA cut-off points associated with short-term mortality across different cancer stages. This multicenter cohort included 1,121 adult patients with cancer in Brazil (≥18 years; 51.2% females; mean age 60 ± 13 years; 43.6% colorectal cancer; 34.3% TNM IV). PhA was derived from raw bioelectrical impedance values: resistance (R) and reactance (Xc) from a tetrapolar single-frequency device (50 kHz). The primary outcome was 6-month mortality. PhA’s predictive accuracy was estimated using Receiver Operating Characteristic (ROC) curves, and cut-off points were estimated using the Youden index. Cox regression models examined crude and adjusted associations between PhA and mortality across TNM stages (I-II, III, IV). PhA values were significantly lower in non-survivors (mean difference: -1.02° in males; -1.21° in females). ROC analyses demonstrated fair predictive performance, with optimal thresholds of ≤ 5.43° for males (AUROC 0.74) and ≤ 4.22° for females (AUROC 0.77). Across TNM stages, PhA consistently predicted mortality, with highest accuracy in stage IV (AUROC 0.75; criterion ≤ 4.72°). In multivariate Cox models, PhA remained independently associated with 6-month mortality after adjustment for confounders. Stepwise analyses confirmed these findings across all stages. PhA is an independent predictor of short-term mortality in cancer across all TNM stages. Identified thresholds (consistently < 5.5°) may guide risk stratification and support clinical decision-making, although external validation is required.