Prognostic factors for abatacept retention in patients who received at least one prior biologic agent: an interim analysis from the observational, prospective ACTION study

ACTION was a non-interventional, international, multicenter, cohort study that evaluated retention and effectiveness in patients with RA who had initiated IV abatacept in Europe (according to the abatacept Summary of Product Characteristics) and Canada (according to the abatacept Product Monograph) [ 2– 4]. Patients were enrolled prospectively, within 3 months of abatacept initiation, between May 2008 and January 2011, and were followed for up to 2 years or, if the patient discontinued abatacept treatment before the 2-year endpoint, for up to 6 months after abatacept discontinuation. A detailed description of the ACTION study design has been reported previously in the context of a 6-month interim analysis [ 7, 15].

The study enrolled adult patients who were abatacept naïve, with established moderate-to-severe RA according to American College of Rheumatology revised criteria (1987) [ 16]. ACTION was conducted in accordance with the Declaration of Helsinki [ 17], the International Conference on Harmonization Good Clinical Practice Guidelines [ 18], and the Good Epidemiological Practice Guideline [ 19]. The study protocol and patient enrollment were approved by local ethics committees and regulatory agencies in accordance with each country’s requirements (first approval received on 31 January 2008, in Munich, Germany; Additional file 1: Table S1). Enrolled patients provided written informed consent.

Here we present an interim analysis of the ACTION study in patients who were treated in four countries (Canada, Germany, Greece, and Italy). All data received up to February 2012 were considered. The analysis included only patients who had received at least one prior biologic agent (the number of patients who had received no prior biologics was insufficient to build a robust model) and who had a baseline clinical assessment no later than 8 days after the first administration of abatacept [ 20].

The dates of first and last abatacept infusions were collected, and abatacept discontinuation and reasons for discontinuation were reported by the rheumatologist at any follow-up visit. Exposure to abatacept was defined as the difference between the first and last date of abatacept infusion, plus 30 days. Patients who did not discontinue abatacept were censored at the time of last available data.

Data concerning potential prognostic factors, including known risk factors and clinically relevant variables, were collected at abatacept initiation. Continuous and categorized variables were considered in this analysis. Categorizations were based on validated cut-offs when available or on clinical expertise, previous literature, or descriptive statistics such as medians or quartiles. Sociodemographic variables were: country, age (continuous and <65 years, ≥65 years), sex, and body mass index (continuous and <25 kg/m ² [underweight/normal], 25– < 30 kg/m ² [overweight], 30– < 35 kg/m ² [obese class I], and ≥35 kg/m ² [obese class II/III]) [ 21]. Disease characteristics considered were: disease duration (continuous and ≤2 years, 3–5 years, 6–10 years, and >10 years), tender joint count (continuous out of a total score of 28), swollen joint count (continuous out of a total score of 28), C-reactive protein (CRP) quartile (<4 mg/L, 4– < 10 mg/L, 10– < 26 mg/L, and ≥26 mg/L), patient global assessment (continuous and based on median), physician global assessment (continuous and based on median), pain (continuous and based on median), DAS28 (erythrocyte sedimentation rate or CRP) (continuous and remission or low DAS [≤3.2], moderate DAS [>3.2–5.1], and high DAS [>5.1]), calculated Clinical Disease Activity Index (continuous and remission, low or moderate DAS [≤22] and high DAS [>22]), radiographic erosion, rheumatoid factor (RF) status, anti-CCP status, and RF and anti-CCP double-positive status. RF and anti-CCP peptide antibody status were based on the assessment at abatacept initiation or any assessment made previously. Comorbidity data (including cardiovascular comorbidity) were collected by assessment of medical history at abatacept initiation and were classified by System Organ Class (SOC) as per the Medical Dictionary for Regulatory Activities. All SOCs were described and considered. Also considered were: number of prior conventional (c)DMARDs (0–3, >3), number of prior anti-TNF agents (<2, ≥2), type of biologic agent received before abatacept (anti-TNF, other mechanism of action), and reason for discontinuation of last biologic agent (intolerance, primary inefficacy, secondary inefficacy, improvement in disease, and other reasons pooled together). Final considerations were: abatacept treatment pattern at initiation (monotherapy, combination with MTX, combination with other cDMARDs), prescription for MTX at initiation, and corticosteroid use before and at abatacept initiation (no corticosteroids, continuous use of corticosteroids, corticosteroids stopped at abatacept initiation, corticosteroids introduced at abatacept initiation).

Abatacept retention rates and 95 % confidence intervals (CI) were described using Kaplan–Meier estimators. Overall retention rate and discontinuation for inefficacy and intolerance were presented at 12 months. Descriptive analysis was performed for all potential prognostic factors. All potential prognostic factors were tested in univariate analysis. For variables in which the categories ‘not available’ or ‘not done’ were predefined, univariate analyses were performed with and without consideration of this category; to be considered in the multivariate model both analyses were required to show consistent significant results. Clinically relevant variables, known risk factors, and prognostic factors with p ≤ 0.20 and no collinearity (two variables were collinear if p < 0.05 and V-Cramer >0.5) were retained in the multivariate model. Known risk factors at treatment initiation, such as infections and infestations, chronic obstructive pulmonary disease (COPD), diabetes mellitus, and tobacco use, were considered in the multivariate model regardless of results in the univariate analysis. Prognostic factors of abatacept retention were identified over the whole period until the data cut-off (February 2012) by multivariate analysis using a Cox proportional hazards model with clustered data (sandwich method) to account for dependence among patients from a single site. Prognostic factors with p > 0.10 were removed by backward selection. Interactions between the retained prognostic factors were tested and included in the final model if significant ( p < 0.10). Univariate analyses were re-run for the final patient sample included in the final multivariate model. Results were presented as hazard ratios (HRs) with 95 % CIs and p-values. An HR >1 indicates higher likelihood of abatacept discontinuation, while an HR <1 indicates a lower likelihood of abatacept discontinuation. HRs are significant when the 95 % CIs do not overlap 1. A sensitivity analysis was performed, including all variables that were significant in the univariate analysis with no variable selection in the multivariate step.