Background: We developed a model to predict remissions in patients treated with biologic disease-modifying anti-rheumatic drugs (bDMARDs) and to identify important clinical features associated with remission using explainable artificial intelligence (XAI). Methods: We gathered the follow-up data of 1204 patients treated with bDMARDs (etanercept, adalimumab, golimumab, infliximab, abatacept, and tocilizumab) from the Korean College of Rheumatology Biologics and Targeted Therapy Registry. Remission was predicted at 1-year follow-up using baseline clinical data obtained at the time of enrollment. Machine learning methods (e.g., lasso, ridge, support vector machine, random forest, and XGBoost) were used for the predictions. The Shapley additive explanation (SHAP) value was used for interpretability of the predictions. Results: The ranges for accuracy and area under the receiver operating characteristic of the newly developed machine learning model for predicting remission were 52.8–72.9% and 0.511–0.694, respectively. The Shapley plot in XAI showed that the impacts of the variables on predicting remission differed for each bDMARD. The most important features were age for adalimumab, rheumatoid factor for etanercept, erythrocyte sedimentation rate for infliximab and golimumab, disease duration for abatacept, and C-reactive protein for tocilizumab, with mean SHAP values of − 0.250, − 0.234, − 0.514, − 0.227, − 0.804, and 0.135, respectively. Conclusions: Our proposed machine learning model successfully identified clinical features that were predictive of remission in each of the bDMARDs. This approach may be useful for improving treatment outcomes by identifying clinical information related to remissions in patients with rheumatoid arthritis.

Takayasu arteritis (TAK) is a systemic disease characterized by large vessel involvement. Although the vascular characteristics of TAK are well characterized, there is no well-organized study demonstrating the extravascular manifestations of TAK. We aimed to evaluate the characteristics of extravascular manifestations of TAK, and to identify the association between vascular and extravascular manifestations of TAK. TAK patients from two independent cohorts between January 2012 and October 2017 were included in the study. Patient characteristics were retrospectively collected from the electronic dataset. The computed tomography scans of all subjects were reviewed to evaluate the pattern of vascular involvement and presence of sacroiliitis. Clinical findings including uveitis, skin lesions, oral ulcers, arthritis, and inflammatory bowel disease (IBD) were reviewed. Logistic regression analysis was performed to evaluate the association between vascular and extravascular manifestations. For the 268 TAK patients, the mean age at diagnosis was 41.2 ± 14.2 years and 88.1% were female. The extravascular manifestation of TAK was observed in 19.0% of patients, the most common being arthritis including sacroiliitis (11.9%) followed by recurrent oral ulcers (8.6%) and IBD (2.6%). A multivariate logistic regression analysis revealed type IIB vascular involvement (adjusted odds ratio (OR) 2.956, 95% confidence interval (CI) 1.337–6.537, p = 0.007) and the erythrocyte sedimentation rate (ESR) (adjusted OR 1.014, 95% CI 1.003–1.025, p = 0.012) as significantly associated with the presence of axial and peripheral arthritis. Extravascular manifestations of TAK were observed in up to one-fifth of patients. The most common extravascular manifestation was arthritis, which was associated with a type IIB vascular involvement pattern and a high ESR.

Ahn SM, et al. Yonsei Med J. 2022 Feb;63(2):141-147. https://doi.org/10.3349/ymj.2022.63.2.141

Background: The purpose of this study was to stratify patients with rheumatoid arthritis (RA) according to the trend of disease activity by trajectory-based clustering and to identify contributing factors for treatment response to biologic and targeted synthetic disease-modifying anti-rheumatic drugs (DMARDs) according to trajectory groups. Methods: We analyzed the data from a nationwide RA cohort from the Korean College of Rheumatology Biologics and Targeted Therapy registry. Patients treated with second-line biologic and targeted synthetic DMARDs were included. Trajectory modeling for clustering was used to group the disease activity trend. The contributing factors using the machine learning model of SHAP (SHapley Additive exPlanations) values for each trajectory were investigated. Results: The trends in the disease activity of 688 RA patients were clustered into 4 groups: rapid decrease and stable disease activity (group 1, n = 319), rapid decrease followed by an increase (group 2, n = 36), slow and continued decrease (group 3, n = 290), and no decrease in disease activity (group 4, n = 43). SHAP plots indicated that the most important features of group 2 compared to group 1 were the baseline erythrocyte sedimentation rate (ESR), prednisolone dose, and disease activity score with 28-joint assessment (DAS28) (SHAP value 0.308, 0.157, and 0.103, respectively). The most important features of group 3 compared to group 1 were the baseline ESR, DAS28, and estimated glomerular filtration rate (eGFR) (SHAP value 0.175, 0.164, 0.042, respectively). The most important features of group 4 compared to group 1 were the baseline DAS28, ESR, and blood urea nitrogen (BUN) (SHAP value 0.387, 0.153, 0.144, respectively). Conclusions: The trajectory-based approach was useful for clustering the treatment response of biologic and targeted synthetic DMARDs in patients with RA. In addition, baseline DAS28, ESR, prednisolone dose, eGFR, and BUN were important contributing factors for 4-year trajectories.