Abstract / Summary
Abstract Importance Clinical phenotyping in Sjögren's syndrome is complicated by non-specific symptoms that overlap with aging, comorbidities, and medication effects. Understanding how these confounding factors affect diagnostic accuracy is essential for valid observational research. Objective To evaluate how age, comorbidities, and xerogenic medications influence clinical phenotyping and diagnostic performance in Sjögren's syndrome. Methods Cross-sectional study of 196 women (97 Sjögren's syndrome, 99 controls). Principal component analysis identified latent phenotypic structure. Random forest classifiers with SHAP values assessed variable importance. Multivariable regression estimated independent associations with stimulated salivary flow. Results A classifier using all variables achieved AUC 0.92 but relied on diabetes, cardiovascular disease, and medication use. A systemic-only model achieved AUC 0.78. An oral-only model retained AUC 0.82. In regression, Sjögren's diagnosis (-0.52 mL/min; 95% CI -0.73 to -0.29) and xerogenic medication (-0.45 mL/min; 95% CI -0.75 to -0.15) were independently associated with lower stimulated salivary flow; age showed a smaller effect (-0.018 mL/min/year; 95% CI -0.027 to -0.010). Diagnostic accuracy declined after age 70. Interpretation Confounding factors can inflate apparent diagnostic performance in Sjögren's syndrome studies. Researchers should consider control-group heterogeneity and audit models for reliance on systemic variables rather than disease-specific oral pathology.