Serum ferritin patterns among ferritin-tested patients across six autoimmune rheumatic disease cohorts: distributions, laboratory associations, and exploratory population context

Background Serum ferritin integrates iron status, inflammation, and tissue injury but is selectively ordered in clinical care. We compared ferritin patterns among tested patients for six autoimmune rheumatic diseases and assessed potential confounding by recorded infections or malignancies. Methods We retrospectively studied 557 ferritin-tested patients at Nanjing Drum Tower Hospital (2018–2022), using a common baseline and laboratory-matching framework. Within-disease associations were assessed using Spearman’s correlation. The primary Gaussian right-censored regression model was adjusted for age and sex (n = 497); laboratory-adjusted models were exploratory. In post hoc sensitivity analyses, clinician-entered infection or malignancy diagnoses were identified using encounter-linked diagnosis text screening without independent chart re-adjudication. Censored-model diagnostics and an alternative error distribution were evaluated. National Health and Nutrition Examination Survey (NHANES) served only as an exploratory population context. Results Ferritin distributions differed across the six cohorts (P = 0.00197); systemic lupus erythematosus (SLE) had the highest original median (264.9 ng/mL). The primary SLE-versus-rheumatoid arthritis (RA) conditional ferritin ratio was 2.58 (95% CI, 1.76–3.77). The baseline-encounter screening flagged 131 patients, leaving 426 (381 with known sex); the SLE-versus-RA ratio decreased to 1.97 (95% CI, 1.30–2.99). The broader all-encounter screening approach excluded 199 patients, leaving 358 (331 with known sex), with a ratio of 1.86 (95% CI, 1.20–2.89). After baseline-encounter exclusion, ferritin–lactate dehydrogenase (LDH) correlations persisted in SLE (rho = 0.495) and systemic sclerosis (SSc) (rho = 0.606), although their magnitudes were attenuated. Residual diagnostics suggested some upper-tail deviation from Gaussian assumptions; a logistic latent-error sensitivity model yielded directionally similar disease contrasts. Conclusions In this selectively ferritin-tested cohort, the SLE-related ferritin association and SLE/SSc ferritin–LDH correlations persisted after exclusion based on documented infection or malignancy, but effect attenuation highlights their vulnerability to clinical confounding. Diagnosis-text screening cannot establish the absence of concurrent conditions. The findings remain exploratory and cannot support disease-specific thresholds or causal claims.