Antibody against cyclic citrullinated protein (ACPA) is counted as one of the most important biomarkers in diagnosis, classification, and prognosis of rheumatoid arthritis (RA). We examined the evolution of ACPA during disease course and assess predictive value of time-weighted cumulative ACPA titer on radiographic progression in RA patients. A group of 734 patients with RA was followed longitudinally over 2 years, with annual measurements of ACPA. The cumulative titers of ACPA were calculated using the trapezoidal rule and were divided into three categories: negative, low-to-moderate, and high. Radiographs of the hands were scored with the modified Sharp score (SHS). Multivariable logistic regression models were performed to identify independent predictors over follow-up for individual patients with different combinations of risk factors. The effect size was computed by Cohen's d method. The patients with radiographic progression had a higher SHS at baseline; and smoking status, diabetes, RF positivity, and use of biologic DMARDs were independently associated with radiographic progression (all P < 0.05). As for ACPA, reversion happened more commonly in men and was associated with younger onset age and lower titer at baseline, but it had no direct relevance to radiographic outcome. In multivariable regression analysis, only high cumulative or baseline titer of ACPA had a predictive power for rapid radiographic progression (all P < 0.05), and cumulative ACPA titer was superior in terms of statistical significance (Cohen's d, 0.637 versus 0.583). High cumulative ACPA titer was independently associated with accelerated radiographic progression, especially with initiation of joint damage.