Study design
This was a multicenter, observational, prospective study in which patients were enrolled by central registration from 82 participating centers in Japan. The planned study period spanned from October 2013 to September 2015. The planned observation period was from October 2013 to December 2017.
The treatment period was 104 weeks for the TCZ-SC ± conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) group, and 52 weeks for the csDMARD-alone group. As the main report of this research, we focus on reporting the comparison between treatment groups at 52 weeks.
Patients
The inclusion criteria were: diagnosis of RA according to the 2010 American College of Rheumatology (ACR)/European League Against Rheumatism (EULAR) 2010 Classification Criteria; previous treatment with more than one csDMARD; performing remunerated work as an employee of a given company or family business (i.e., PW), or performing a central role in the housework within a household (i.e., HW); Disease Activity Score in 28 joints using the erythrocyte sedimentation rate (DAS28-ESR) ≥ 3.2; biologic-naive; prescribed TCZ-SC for the first time; receiving a csDMARD (except for tofacitinib) dose increase; receiving a csDMARD (except for tofacitinib) as add-on therapy; switching to a csDMARD (except for tofacitinib) treatment from other csDMARD(s); and written informed consent. Patients with any contraindication for use of the drugs evaluated in this study and those judged as ineligible for participation in this study by the investigators were excluded.
Study oversight and conduct
Ethical approval was obtained from the institutional review board of each institution. The study was conducted in accordance with the Declaration of Helsinki and “Ethical Guidelines for Clinical Research” of the Ministry of Health, Labour and Welfare. Urgent events, such as adverse events (AEs), were reported to the research steering committee. Accordingly, the institutional review board and research steering committee determined the continuity of the patients in the study as well as that of the study itself. All patients provided written informed consent to participate in this study before being registered in the electronic data capturing system.
Study treatment
In the TCZ-SC ± csDMARDs group, the dose was prescribed by the treating physician according to the prescribing information in the package insert [16]. In the csDMARDs-alone group, the dose of each csDMARD was prescribed according to the prescribing information in the corresponding package insert. Starting a csDMARD alone or in combination, as well as dose changes, switching to other csDMARDs, or adding other csDMARDs was permitted in the TCZ-SC ± csDMARDs group. Dose changes, switching, or addition of another csDMARD was also permitted in the csDMARDs-alone group.
Assessments
The registration questionnaire was obtained at study registration. Patient demographic and disease characteristics were evaluated at baseline. The WPAI, Work Functioning Impairment Scale (WFun) [17], Frenchay Activities Index (FAI) [18], EuroQol 5 Dimension (EQ-5D) [19], Japanese Health Assessment Questionnaire Disability Index (HAQ-DI) [20, 21], and 6-item Kessler psychological distress scale (K6) [22] were assessed at baseline and at weeks 12, 24, and 52. DAS28-ESR, Clinical Disease Activity Index (CDAI) [23], and the Simplified Disease Activity Index (SDAI) [23] were assessed at baseline and at weeks 12, 24, 36, and 52. AEs were assessed continuously. The duration of assessments was approximately 52 weeks plus an additional 28 days (allowance).
Discontinuation criteria were as follows: 1) patient withdrawal; 2) physician’s decision because of AEs; 3) patients in the TCZ-SC ± csDMARDs group who switched from TCZ-SC to other biological agents; 4) patients in the csDMARDs-alone group who started treatment with biological agents, including TCZ and/or tofacitinib; and 5) other cases judged to require discontinuation by treating physicians.
Endpoints
The primary endpoint was the change in the percentage of OWI among PWs at week 52 as assessed using the WPAI. The secondary endpoints for efficacy were as follows: change in the percentage of presenteeism (in PWs), absenteeism (in PWs), and activity impairment of daily work by WPAI (PWs and HWs); change in employment rate by WPAI (PWs); changes in WFun (PWs); and changes in disease activity by DAS28-ESR, CDAI, and remission rate. WPAI parameters were scored in the following way: absenteeism = (hours absent from work due to RA) / (hours absent from work due to RA + hours actually worked); and percentage of OWI = absenteeism + [(1 − absenteeism) × presenteeism].
The secondary endpoints for QOL were as follows: changes in FAI among HWs; changes in EQ-5D; changes in HAQ-DI (some questions were replaced to accommodate Japanese lifestyle differences and have been validated/confirmed) [20]; and changes in K6 improvement factor.
Additional exploratory analyses were conducted to assess the relationship between characteristics and each assessment outcome. Safety was assessed based on all AEs reported.
Sample size calculation
Based on previous studies in Japan and the US reporting WPAI of PWs with RA as the main endpoint [5, 24], we assumed that the mean percentage of OWI (primary endpoint) was 30% to 40% at baseline. We also assumed that the percent change in OWI from baseline in the TCZ-SC and csDMARDs-alone groups at week 52 would be 40% and 15% (i.e., − 12% and −4.5% change from baseline considering a value of 30% at baseline), respectively. We used the Monte Carlo Simulation, repeated 10,000 times, to investigate the target population. Using the Wilcoxon rank sum test, we calculated the sample size to achieve a 5% two-sided significance level of 5% and 80% power. As a result, we estimated the need for a total of 160 PWs in both groups. Considering a possible drop out/discontinuation rate of 50% among PWs, we set the target population at 800 patients: 400 patients in the TCZ-SC ± csDMARDs group and 400 patients in the csDMARDs-alone group. Patient enrollment was continued until the number of PWs (excluding HWs) reached at least 200 in the TCZ-SC ± csDMARDs group.
Study population
Efficacy analysis sets were the intention-to-treat set (patients whose treatment plan was determined among registered patients, except for any patient who did not provide written informed consent, or duplicated patients) and the modified intention-to-treat (mITT) set (patients in the TCZ-SC or csDMARDs-alone groups who received TCZ-SC or corresponding csDMARDs one or more times, except for patients with significant protocol deviations such as erroneous registration, lacking data for efficacy evaluations, or lacking baseline data for propensity score estimation). The safety analysis set included all patients in the TCZ-SC or csDMARDs-alone groups who received TCZ-SC or the corresponding csDMARDs one or more times, respectively, among the registered patients in this study. All analyses were conducted using the mITT population.
Statistical analysis
In contrast to randomized controlled trials, it is difficult to compare the efficacy in an observational study because of the treatment selection bias. Therefore, we adjusted patient characteristics between groups using propensity scores. Propensity scores were estimated using a multivariate logistic regression model predicting treatment with TCZ-SC based on the following key variables: background (age, weight, disease duration, salary, education, and occupation); concomitant use of glucocorticoids and/or methotrexate, rheumatoid factor, anti-cyclic citrullinated peptide antibody; disease activity and severity (class, stage, DAS28-ESR, CDAI, and SDAI); and questionnaires (percentage of OWI, absenteeism, presenteeism, activity impairment, EQ-5D, HAQ-DI, and K6).
The mean change from baseline and differences between treatment groups were estimated by linear regression with a robust variance estimator adjusted by the inverse probability of treatment weighting (IPTW) method. The last observation carried forward (LOCF) method was used for missing data. The primary adjustment method for confounding was changed from propensity score matching to IPTW by the research steering committee based only on baseline information, excluding post-treatment measurement values, as prespecified in the protocol.
Sensitivity analysis for estimating propensity scores confirmed the robustness of the present analyses by the model selection method (backward selection) using the clinically significant factors and selected variables. Additionally, we performed the Wilcoxon rank sum test after propensity score matching, a stratified analysis with five strata based on propensity score and regression analyses adjusting for clinically significant factors. Data insufficiently adjusted by IPTW (i.e., methotrexate yes/no) were separately and additionally adjusted by using sensitivity analysis to avoid any effects on the primary statistics.
An exploratory, linear regression analysis was conducted to investigate background factors possibly related to activity impairment and OWI improvement in PWs. The absolute standard partial regression coefficient for each baseline factor was calculated to assess the treatment response to TCZ-SC and csDMARDs. All statistical analyses were performed using SAS 9.3 version (SAS Institute Inc., Cary, NC, USA).