A prospective study was made to seek for a convenient biomarker to predict progression of bone destruction (PBD) in early stages of rheumatoid arthritis (ERA). All participated patients had definite RA and their radiographic stages were mild less than stage II of the Steinbrocker classification, naïve for treatment of any DMARDs or corticosteroids. After the entry, they were treated according to the 2002 ACR management guideline for RA. The candidate biomarkers (RF-IgM, RF-IgG, CARF, ACPA, CRP, ESR, NTx, MMP-3, IL-6 and osteopontin) were measured at the entry. PBD was assessed radiographically by interval changes in the modified Sharp scores (ΔSHS) for 24 months. The associations between ΔSHS and baseline biomarkers were assessed statistically by multivariate regression analyses. Both the baseline ACPA and IL-6 levels correlated with PBD, suggesting that they could predict PBD in ERA.