A post hoc analysis of real-world data from the Abatacept Best Care (ABC) study has identified three distinct patterns of disease activity response over 12 months among patients with rheumatoid arthritis treated with subcutaneous abatacept. Published in Advances in Rheumatology, the research utilized a growth mixture model to classify patients into Rapid Responders, Late Responders, and Non-Responders based on their Clinical Disease Activity Index (CDAI) scores. The study found that approximately 85% of the 256 patients analyzed showed either rapid or late reduction in disease activity, suggesting that many patients who do not achieve immediate remission may still benefit from continued treatment.
The ABC study was a real-world observational study that enrolled 290 adult patients with moderate to severely active rheumatoid arthritis, defined as a baseline CDAI score greater than 10. Patients were initiated on subcutaneous abatacept as part of routine clinical care, with treating physicians randomized to receive formal instruction on the Treat to Target approach or to continue managing patients through standard routine care. For the current trajectory analysis, 256 patients were included who had at least two CDAI measurement time points, one at baseline and at least one subsequent measurement. The cohort had a mean age of 60.1 years, and 74.6% were female, with a mean baseline CDAI score of 30.1, indicating high disease activity at the start of treatment.
Researchers employed a growth mixture model, a statistical technique that classifies individuals into unobservable subgroups based on changes in a measured variable over time. This method allowed for the estimation of missing data without making strong assumptions about the distribution of the data. The model was selected by comparing fits of different polynomial degrees and class numbers, with a cubic model and three distinct trajectory groups providing the best fit to the data without overfitting.
The analysis identified three trajectory groups: Rapid Responders, comprising 49% of the sample; Late Responders, comprising 38%; and Non-Responders, comprising 14%. Rapid Responders were characterized by a swift decline in CDAI scores, with a restricted mean survival time to low disease activity or remission of 5.6 months. Late Responders showed a more gradual improvement, with a restricted mean survival time to low disease activity or remission of 9.0 months. Non-Responders did not achieve low disease activity or remission within the 12-month follow-up period, although they did show an average improvement in CDAI score of about 10 points.
Baseline characteristics differed significantly among the three groups. Rapid Responders had lower baseline scores for multiple measures of disease severity, including tender joint count, swollen joint count, patient global assessment, physician global assessment, and the Disease Activity Score 28. They also reported lower levels of pain, fatigue, and disability as measured by the Health Assessment Questionnaire-Disability Index. In contrast, Non-Responders had a higher proportion of baseline involvement of the shoulder, knee, and elbow joints, and a lower proportion of anti-cyclic citrullinated peptide antibody positivity. These differences suggest that the severity and specific joint involvement at the start of treatment may influence the trajectory of response to abatacept.
Using multivariable logistic regression, the researchers identified several baseline factors independently associated with being a Responder, defined as either a Rapid or Late Responder. Lower baseline physician and patient global assessment scores, lower swollen joint count, and lower comorbidity as measured by the Rheumatic Disease Comorbidity Index were associated with higher odds of being a Responder. Additionally, being biologic-naïve, meaning having no prior treatment with a biological disease-modifying anti-rheumatic drug, was associated with a higher likelihood of response. Prior treatment with a biologic was associated with lower odds of being classified as a Responder, indicating that previous exposure to similar therapies might influence the response to abatacept.
A key finding of the study was that CDAI scores at 3 months were not predictive of the 12-month response trajectory. When the growth mixture model was applied to only the baseline and 3-month CDAI scores, the classification of patients into trajectory groups showed only moderate agreement with the 12-month classification. Specifically, 27% of patients classified as Rapid Responders at 3 months were reclassified as Late Responders or Non-Responders at 12 months. This suggests that early changes in disease activity may not be a reliable indicator of long-term treatment success, and that patients who do not achieve remission or low disease activity within the first three months may still benefit from continuing abatacept.
The study also examined the individual components of the CDAI score to determine if they could be used to predict trajectory groups. The CDAI is a composite score that includes tender joint count, swollen joint count, physician global assessment, and patient global assessment. The researchers found that while changes in all four components were statistically significant over time, the individual components did not separate patient trajectories as effectively as the full CDAI score. When growth mixture models were applied to the individual components, the resulting groupings were degenerate and showed poor agreement with the classifications based on the full CDAI score. This indicates that the composite nature of the CDAI is important for capturing the overall trajectory of disease activity.
The authors noted several limitations of the study. As an observational study, it is subject to biases such as selection bias and bias by indication, where treatment decisions are influenced by patient characteristics. The follow-up visits were conducted every three months, which may be less frequent than in randomized controlled trials but is more reflective of real-world clinical practice. The results may be limited to patients with longstanding rheumatoid arthritis being treated with abatacept, and treatment decisions should be individualized. However, the use of a novel statistical technique combining fully-Bayesian growth mixture models and Bayesian imputation for missing data strengthens the robustness of the findings.
This research contributes to the understanding of how patients with rheumatoid arthritis respond to abatacept over time. By identifying distinct trajectory groups and their associated baseline characteristics, the study provides insights that may help clinicians set more realistic expectations for patients and optimize treatment decisions. The finding that early response does not necessarily predict long-term outcome is particularly relevant for clinical practice, as it suggests that patients who do not achieve immediate remission may still benefit from continued treatment. This could inform decisions about when to switch or discontinue therapy, potentially improving long-term patient outcomes and reducing the risk of irreversible joint damage.
The study highlights the importance of considering the trajectory of disease activity over time rather than relying on isolated assessments at specific time points. By using a growth mixture model, the researchers were able to capture the dynamic nature of treatment response and identify subgroups of patients with different patterns of improvement. This approach may have implications for future clinical trials and real-world studies, as it allows for a more nuanced evaluation of treatment effectiveness. The findings also underscore the need for further research to identify predictive markers of clinical response, which could help guide treatment decisions and improve patient outcomes in rheumatoid arthritis.
The ABC study was registered on ClinicalTrials.gov under the number NCT03274141, with registration occurring on September 6, 2017.
The data from this study are available from the corresponding author upon reasonable request, and Bristol Myers Squibb shares data from studies that meet their data sharing criteria with qualified researchers.
Subject of Research: CDAI response trajectories and predictors in rheumatoid arthritis patients treated with abatacept
Article Title: CDAI trajectories in patients with rheumatoid arthritis treated with abatacept in real-world
Article References: Pope, J. E., Sampalis, J. S., Haraoui, B., Rampakakis, E., Keating, D., Allum, F., Trepanier, M.-O., & Bessette, L. (2026). CDAI trajectories in patients with rheumatoid arthritis treated with abatacept in real-world. Advances in Rheumatology, 66(1), Article 54. https://doi.org/10.1186/s42358-026-00554-y
Image Credits: AI Generated
DOI: 10.1186/s42358-026-00554-y
Keywords: rheumatoid arthritis, abatacept, CDAI trajectories, growth mixture model, treat to target, real-world data, biologic-naïve, disease activity, clinical response, post hoc analysis, CDAI, trajectories
