Circulating miR-21-5p and miR-223-3p as diagnostic biomarkers for rheumatoid arthritis in a sub-Saharan African cohort: a hospital-based case-control study.

Background

Diagnosing rheumatoid arthritis (RA) remains challenging, particularly in resource-constrained settings. Circulating microRNAs (miRNAs) have emerged as candidate non-invasive biomarkers; however, their expression in African populations has not been systematically evaluated.

Methods

This hospital-based study drew on a parent cohort of 170 participants (85 patients with rheumatoid arthritis and 85 healthy controls) recruited at Kenyatta National Hospital, Kenya. Circulating miRNAs were quantified in a subsample of 40 participants (20 RA, 20 controls). Two miRNAs (miR-21-5p and miR-223-3p) were measured in plasma using stem-loop RT-qPCR with miR-16-5p as the endogenous control. Relative expression was assessed using the comparative cycle threshold (ΔΔCt) method. Diagnostic performance was evaluated by receiver operating characteristic analysis with bootstrap confidence intervals and bootstrap internal validation, and all comparisons were additionally adjusted for age and sex.

Results

Both miRNAs were significantly downregulated in RA patients relative to controls (miR-21-5p fold change 0.21; miR-223-3p fold change 0.35; both P < 0.001), and both differences persisted after adjustment for age and sex (miR-21-5p adjusted P = 0.006; miR-223-3p adjusted P = 0.019). miR-21-5p showed the stronger individual discrimination (AUC 0.900, 95% CI 0.772-0.990) compared with miR-223-3p (AUC 0.857, 95% CI 0.717-0.960). A two-miRNA logistic model achieved an apparent AUC of 0.925 (95% CI 0.820-0.997), corresponding to an optimism-corrected AUC of 0.911 after bootstrap internal validation; this did not significantly exceed miR-21-5p alone (ΔAUC 0.025, P = 0.299). No statistically significant difference was detected between the two-miRNA model and CRP, ACPA or RF, although the study was not powered for equivalence testing. Within RA patients, no correlation between miRNA expression and clinical variables survived correction for multiple testing.

Conclusion

Circulating miR-21-5p and miR-223-3p are markedly downregulated in Kenyan patients with rheumatoid arthritis, and miR-21-5p discriminated patients from healthy controls with an apparent AUC of 0.900 (95% CI 0.772-0.990); the two-miRNA logistic model had an apparent AUC of 0.925 and an optimism-corrected AUC of 0.911. Within-group analyses suggest that age alone is unlikely to account for the between-group difference in miR-21-5p (age explained approximately 1% of the observed difference), although age and disease status are strongly collinear and the age distributions overlap very little. These findings identify circulating miR-21-5p as a credible candidate diagnostic biomarker in a sub-Saharan African population, in which no such evaluation has previously been reported. Consistent with the accepted sequence of biomarker development, establishing clinical diagnostic accuracy now requires validation against early disease and clinically relevant disease controls.