Machine learning-based prediction of relapse in rheumatoid arthritis patients using data on ultrasound examination and blood test.

Recent effective therapies enable most rheumatoid arthritis (RA) patients to achieve remission; however, some patients experience relapse. We aimed to predict relapse in RA patients through machine learning (ML) using data on ultrasound (US) examination and blood test. Overall, 210 patients with RA in remission at baseline were dichotomized into remission (n = 150) and relapse (n = 60) based on the disease activity at 2-year follow-up. Three ML classifiers [Logistic Regression, Random Forest, and extreme gradient boosting (XGBoost)] and data on 73 features (14 US examination data, 54 blood test data, and five data on patient information) at baseline were used for predicting relapse. The best performance was obtained using the XGBoost classifier (area under the receiver operator characteristic curve (AUC) = 0.747), compared with Random Forest and Logistic Regression (AUC = 0.719 and 0.701, respectively). In the XGBoost classifier prediction, ten important features, including wrist/metatarsophalangeal superb microvascular imaging scores, were selected using the recursive feature elimination method. The performance was superior to that predicted by researcher-selected features, which are conventional prognostic markers. These results suggest that ML can provide an accurate prediction of relapse in RA patients, and the use of predictive algorithms may facilitate personalized treatment options.

Positive rate and prognostic significance of the superb microvascular imaging signal in joints of rheumatoid arthritis patients in remission with normal C-reactive protein levels and erythrocyte sedimentation rates.

PURPOSE: This study aimed to evaluate the positive rate and prognostic significance of superb microvascular imaging (SMI) in rheumatoid arthritis (RA) patients in remission with normal C-reactive protein (CRP) levels and erythrocyte sedimentation rates (ESR). METHODS: The study enrolled 112 RA patients, and ultrasound (US) assessment was performed on 28 joints of each patient. RESULTS: The SMI signal-positive rates for each joint were: metacarpophalangeal (MCP) joints: 20.5%, wrist joints: 43.8%, metatarsophalangeal (MTP) joints: 17.0%, and other foot joints: 25.0%. Investigation of the prognostic significance of the SMI signal in each joint revealed that only in the MTP joints was the total score of the SMI signal in the patients with relapse significantly higher than that in the patients with remission (P = 0.01). Comparison of the receiver operating characteristics curves for predicting relapse showed that the area under the curve (AUC) of the MTP joints was the highest (AUC = 0.66) of the investigated joints. The optimal threshold for the total MTP SMI score was 1 (accuracy = 83.3%). Positive/negative data of the SMI signal in the MTP joints were not significantly associated with the values of conventional disease activity markers. CONCLUSION: In RA patients in remission with normal CRP and ESR levels, the percentage of positive SMI signal was highest in the wrist joints. However, the accuracy of the SMI signal for predicting relapse was greatest for the MTP joints, suggesting that US assessment of the MTP joints by SMI is useful for predicting relapse in these patients.

Prediction of recurrence and remission using superb microvascular imaging in rheumatoid arthritis.

PURPOSE: Ultrasound is commonly used to assess the degree of synovitis in patients with rheumatoid arthritis (RA); however, it is unclear which joints are optimal for evaluating and predicting recurrence and remission. PATIENTS AND METHODS: In 293 RA patients enrolled in the KURAMA cohort, 28 joints were assessed by ultrasound. RESULTS: Results from patients in remission in both 2015 and 2017 (Group 1, n = 152) were compared with those from patients in remission in 2015 and non-remission in 2017 (Group 2, n = 60). The SMI scores for total (3.1 vs. 6.3, P = 0.004), MCP2-5 (1.1 vs. 2.4, P = 0.03), wrist (0.9 vs. 2.1, P = 0.0003), MTP2-5 (0.4 vs. 1.0, P = 0.03), and Lisfranc joints (0.07 vs. 0.25, P = 0.04) were significantly higher for Group 2. When those in non-remission in 2015 and remission in 2017 (Group 3, n = 27) were compared with those in remission in 2015 and non-remission in both 2015 and 2017 (Group 4, n = 54), the GS-SMI combined score (3.0 vs. 5.0, P = 0.04) and SMI score (1.5 vs. 2.9, P = 0.04) for MCP2-5 joints were significantly higher for Group 4. Multivariate logistic regression analysis identified "wrist SMI score ≧ 1" as an independent prognostic factor for recurrence (odds ratio 3.08, P = 0.001) and "MCP2-5 GS-SMI combined score ≦ 4" as an independent prognostic factor for remission (odds ratio 3.25, P = 0.048). CONCLUSION: We identified the optimal joint cut-off scores for predicting recurrence and remission in RA patients. Risk-stratification therapy based on the ultrasound scores may improve outcome and quality of life for patients with RA.