*Values in parentheses meant X-ray intervals during the long-term extension period.

†Values were correlation coefficients between the duplicate readings of baseline and year-1 radiographs when these were reread along with the year-2 films (correlation coefficient 0.971 for year-1 films and 0.972 for baseline films).

‡Sensitivity analysis was not performed in the primary analysis.

§Missing data was imputed using the change from baseline was estimated using the percentile of the entire patient population.

¶The Markov Chain Monte Carlo method was used to impute the missing radiographic data.

**Radiographs were obtained at baseline (1 year in the open-label phase), 39 weeks (91 weeks in the open-label phase) in the double-blind period.

††Missing data was imputed using group mean change.

‡‡ICC at baseline and week 24 was 0.98 and 0.80, respectively.

§§Changes from baseline in SvdH score for these patients were substituted with the median change for all patients.

CMI, Conditional mean imputation; DMARDs, disease-modifying antirheumatic drugs; GEE, generalised estimating equation; ICC, intra-class correlation coefficient; IR, inadequate responder; LE, linear extrapolation; LOCF, last observation carried forward method; MI, multiple imputation; MMRM, mixed model for repeated measures; RCM, Random coefficients model; SDC, smallest detectable change; SDD, smallest detectable difference; SvdH, van der Heijde modification of the Sharp score; TNF, tumour necrosis factor.