Maxime Verhoeven

108 Chapter 6 as cluster variable, to account for repeated measurements over time within patients. The number of imputed datasets was based on the percentages of missing values over time. By using Rubin’s rule, analysis results based on the imputed datasets were pooled to provide an overall result. 12 A second sensitivity analysis was performed to exclude the influence of acute phase reactants (APRs) on the outcome, as it is known that TCZ reduces APRs specifically. 13 However, validated disease activity indices without APR include the visual analogue scale (VAS) physician, which was not assessed during the PTFU. As an alternative, the unweighted components of the DAS28, except for ESR, were analysed separately using linear mixed effects models to study the differences in outcomes between the initial strategy groups over time using observed data. Square root transformation was applied to skewed SJC28 and tender joint count (TJC)28 data. RESULTS Of the 317 patients who had started in U-Act-Early, 226 patients were included in the PTFU, of whom 85% completed the study, Figure 1. During this period one patient died due to a brain stem infarction (TCZ strategy group) and one patient due to squamous cell carcinoma of the cervix (MTX strategy group); these complications were deemed not to be related to the treatment. The patient disposition during the U-Act-Early trial is shown in Supplementary Figure S1. Figure 1 Patient disposition during the 3-year post-trial follow-up. UAE= U-Act-Early; TCZ+MTX= tocilizumab + methotrexate initiation strategy group; TCZ= tocilizumab + placebo-methotrexate initiation strategy group; MTX= methotrexate + placebo-tocilizumab initiation strategy group; Withdr. consent= withdrawal of consent; AE= adverse event; Lost= lost to follow-up; EOS= end of study; follow-up= post-trial follow-up.