Abstract
BACKGROUND: Cancer survivorship is increasing, and many people living with and beyond cancer seek supportive care such as Traditional Chinese Medicine (TCM), where tongue inspection is routinely used. This exploratory study examined whether visually assessed tongue features differ between cancer survivors and those without a self-reported history of cancer. METHODS: Matched case-control (N = 120): 60 adults diagnosed with cancer vs. 60 without a self-reported cancer history. Three standardized tongue photographs were taken for each participant. Three blinded TCM practitioners scored 20 predefined features. Logistic models with Benjamini-Hochberg false discovery rate (FDR) control estimated associations; Fleiss' κ quantified inter-rater reliability. Nested logistic models explored discrimination; 10-fold cross-validation assessed area under the curve (AUC) and calibration. RESULTS: Six features were found to be associated with cancer presence with an FDR of q < 0.05: peeled coating (odds ratio (OR) = 3.66, 95% confidence intervals (CI) 1.40-9.60), spider sublingual veins (OR = 2.74, 95% CI 1.53-4.92), thick coating (OR = 2.56, 95% CI 1.50-4.36), purple body (OR = 2.29, 95% CI 1.44-3.63), tooth marks (OR = 1.87, 95% CI 1.19-2.95), and thin coating (OR = 0.38, 95% CI 0.22-0.65; inverse). Reliability was slight to moderate (κ ≈ 0.01-0.48). Discrimination was modest (AUC 0.676-0.736). CONCLUSION: Selected tongue features, particularly spider sublingual veins and tooth marks, were associated with current cancer in exploratory analyses. However, discrimination was modest, and inter-rater reliability varied across features. These findings should be interpreted cautiously and do not support diagnostic utility. Standardized definitions, rater training, and external validation in larger prospective, multi-center cohorts are required before clinical translation.