Polymyalgia rheumatic At every visit after baseline, patients were assessed for the pres- ence of remission, which was defined as the absence of stiffness at shoulder and/or hip girdle attributable to active polymyalgia rheumatica, as judged by a blinded investigator. Pain and stiff- ness are part of the OMERACT core domain set for polymyalgia rheumatica. 22 Relapse was defined in accordance with previous studies 10 23 as recurrence of signs of active polymyalgia rheumatica, that is, the return of aching and stiffness at shoulders, hip girdle or both, as adjudicated by the blinded investigator. At every visit, an additional blinded efficacy assessor obtained patient and assessor global scores, patient pain score, morning stiffness (all on 100 mm Visual Analogue Scales (VAS)) and evaluated physical function by the ‘elevation of the upper limb score’ (semiquantitative scale), 24 the Health Assessment Questionnaire Disability Index (HAQ) and the Short Form-36 (SF-36). Inflammatory markers, CRP and ESR, on which tocilizumab has rapid effects, were determined at each study visit, but were not part of remission/relapse definitions by the investigators, who were blinded to these results to maintain overall blinding. Blinding was further ensured through provision of tocilizumab and placebo in prefilled syringes with matching presentation for the 16 weeks of active therapy. Blinding was maintained throughout the study period. Statistical analysis Based on the nature of the disease and the usual need of a slow tapering of glucocorticoids to prevent relapses, we estimated that only 20% of enrolled patients receiving placebo would be able to achieve glucocorticoid-free remission at week 16. In contrast, previous open-label phase 2 trials and proof-of-concept studies suggested a high level of disease control for patients treated with tocilizumab, 17–20 guiding an expected 80% of patients to be in glucocorticoid-free remission at week 16 when treated with tocilizumab. Based on these estimates and a 1:1 sampling ratio, we calculated that a total of 24 patients (12 in the tocilizumab group and 12 in the placebo group) were needed to design a randomised controlled phase 2/3 clinical trial, which would provide 80% power at a significance level of p<0.05 using Fish- er’s exact test on the primary outcome (rate of glucocorticoid- free remission at week 16). We conservatively estimated 30% of patients as potential drop outs/lost to follow-up, and increased the sample size accordingly despite the planned non-responder imputation approach, by which all patients who started treat- ment would be included in the analysis. The targeted recruit- ment was therefore 32 patients. The primary and key secondary endpoints were tested between the groups using either Fisher’s exact tests for categor- ical variables, Kruskal-Wallis tests for non-normally distributed continuous data or Kaplan-Meier estimator for time-to-event data. To control for type I error of the secondary endpoints, we applied a strategy of hierarchical testing, by which hypothesis testing continues until reaching the first non-significance. The pre-determined hierarchy for testing secondary endpoints was: proportion of subjects in glucocorticoid-free remission at week 12 → proportion of subjects in glucocorticoid-free remission at week 24 → time to first relapse → cumulative dose of pred- nisone at week 16 → cumulative dose of prednisone at week 24 → proportion of subjects with increased ESR >20 mm/hour, or increased CRP levels >5 mg/L at week 24 → patient pain (VAS) at week 16 → patient global assessment of disease activity (VAS) at week 16 → evaluator global assessment (VAS) at week 16 → SF-36 at week 16 → HAQ at week 16. Other secondary outcomes as stated in the protocol, including safety assessment, were considered exploratory.
Analyses were based on an intent-to-treat approach using non-responder imputation for binominal endpoints, and the last observation carried forward method for continuous endpoints, for all visits after patients have dropped out of the study or were lost to follow- up. In the Kaplan-Meier analysis, all patients were included in the estimate of time-to-relapse, censoring patients without event (ie, without relapse) or those lost to follow-up. The statistical analysis plan was registered and submitted as part of the study protocol before study initiation and was not amended thereafter. Statistical analysis was performed by HR and AK. Figure 1 Screening, randomisation and follow-up of patients at week 16 (primary endpoint) and week 12/24 (secondary endpoints). AE, adverse event; SAE, serious AE. From 20 November 2017 to 28 October 2019, 39 participants were screened. First baseline visit was on 27 November 2017. The last visit (week 24) was on 2 June 2020. Of 39 screened patients three patients were excluded and 36 were enrolled, of whom 19 were randomly assigned to receive tocilizumab and 17 to receive placebo; 84% of patients in the tocilizumab group and 65% in the placebo group completed the trial through week 24 (figure 1). Patient characteristics at baseline were not statistically significantly different between the groups (table 1). RESULTS Patients Primary and secondary outcomes A total of 63.2% (12/19) in the tocilizumab group and 11.8% (2/17) in the placebo group achieved glucocorticoid-free remis- sion at week 16 (p=0.002, figure 2A); this state was maintained after withdrawal of tocilizumab at week 16 in 91.7% (11/12) of patients over additional 8 weeks of blinded follow-up until week 24. With respect to achieving the primary endpoint (at week 16), these numbers correspond to an OR of 12.9 (95% CI: 2.2 to 73.6) in favour of tocilizumab. Hierarchical testing of secondary endpoints revealed statis- tical significance in favour of tocilizumab versus placebo for glucocorticoid-free remission at weeks 12 and 24 (p=0.02 for both, figure 2A); time to first relapse (p=0.007); and cumulative prednisone dose at week 16 (p=0.003) and week 24 (p=0.001) (table 2; online supplemental figure 1). Using time-to-event anal- ysis (Kaplan-Meier estimator) time to first relapse was also in favour of tocilizumab (log-rank test p=0.007, figure 2B). The estimated mean time to first relapse was 130 days (±13) in the tocilizumab group and 82 days (±11) in the placebo group. The median cumulative prednisone dose over 16 weeks amounted to 727 mg (IQR 721–842) in the tocilizumab group
Bonelli M, et al . Ann Rheum Dis 2022; 81 :838–844. doi:10.1136/annrheumdis-2021-221126
840
Powered by FlippingBook