Machine learning-based predictors of RA relapse were evaluated

The authors found that the XGBoost predictor could more accurately predict RA relapse than logistic regression and random forests.

When evaluating machine learning-based predictors of relapse among affected patients Rheumatoid arthritis (RA), the researchers found that the predictor of extreme gradient enhancement (XGBoost) was the best.

XGBoost had a higher resolution (the area under the receiver operator characteristic curve [AUC] = 0.747) from the other two classifiers, logistic regression (AUC = 0.701) and random forest (AUC = 0.719).

The logistic regression model is a conventional and generalized linear model used for binary classification in clinical prediction, and the random forest model is an aggregate algorithm that combines several decision trees. Similar to Random Forest, XGBoost is a decision tree based ensemble algorithm, but instead uses gradient boosting to achieve more accurate predictions.

“The XGBoost algorithm selects a single feature when there is a significant correlation between variables, while the Random Forest ensemble selects a feature randomly and recognizes associations of different features across the model,” the study authors wrote. “Therefore, XGBoost was considered more accurate in selecting features because it can select fewer and more efficient features.”

According to the authors, whose research was published in Scientific ReportsThese results indicate that ML-based predictors can accurately predict RA relapse, and similar predictive algorithms can thus facilitate personalized treatment plans for patients.

After the exceptions, the study included 210 patients with RA who were enrolled in the KURAMA group in 2015 and had follow-up and ultrasound data in 2017. These patients were divided into two groups, with 150 patients who achieved remission in 2017 in the “remission group” and 60 patients with rheumatoid arthritis in 2017 in the “relapse” group.

Using ultrasound and blood test data, the study authors found that several clinical and biological markers associated with RA disease activity were significantly higher among patients with relapse compared with scores in patients in remission:

  • The degree of disease activity on the 28 joints CRP
  • simplified disease activity index
  • Clinical disease activity index
  • health assessment questionnaire
  • Global Patient Assessment with a Visual Analog Scale

They then applied a recursive feature exclusion selection algorithm to improve accuracy, using gender, disease duration, age, wrist microvascular imaging (SMI), metatarsophalangeal (MTP) SMI score, ESR, C-reactive protein, and rheumatoid factor. ,
Anti-cyclic peptide, citrulline, and matrix metalloproteinase-3.

When comparing the values ​​of the ten traits between the two groups, the wrist and MTP SMI scores were significantly higher in patients in the relapse group compared to patients in the remission group. The authors note, however, that length and alanine aminotransferase were significantly lower in patients who had relapsed. No other significant differences were observed.

According to the authors, these results reflect an improved model for predicting relapse in RA patients through ML.

“Data mix for the United States [ultrasound] Screening and blood testing was a unique approach to this study, and US data has been shown to be essential for prediction. “The findings may lead to a better assessment of relapse risk and enable the selection of personalized treatment strategies for patients with rheumatoid arthritis.”


Matsuo H, Kamada M, Imamura A et al. Machine learning-based prediction of relapse in rheumatoid arthritis patients using ultrasound and blood test data. science representative. Published online May 4, 2022. doi: 10.1038/s41598-022-11361-y