Fumihiko Miyoshi1, Kyoko Honne2, Seiji Minota2, Masato Okada3, Noriyoshi Ogawa4, Toshihide Mimura1. 1. a Department of Rheumatology and Applied Immunology, Faculty of Medicine , Saitama Medical University , Saitama , Japan. 2. b Division of Rheumatology and Clinical Immunology , Jichi Medical University , Tochigi , Japan. 3. c Immuno-Rheumatology Center, St. Luke's International Hospital , Tokyo , Japan , and. 4. d Division of Immunology and Rheumatology , Internal Medicine 3, Hamamatsu University School of Medicine , Hamamatsu , Japan.
Abstract
OBJECTIVES: The aim of the present study was to generate a novel method for predicting the clinical response to infliximab (IFX), using a machine-learning algorithm with only clinical data obtained before the treatment in rheumatoid arthritis (RA) patients. METHODS: We obtained 32 variables out of the clinical data on the patients from two independent hospitals. Next, we selected both clinical parameters and machine-learning algorithms and decided the candidates of prediction method. These candidates were verified by clinical variables on different patients from two other hospitals. Finally, we decided the prediction method to achieve the highest score. RESULTS: The combination of multilayer perceptron algorithm (neural network) and nine clinical parameters shows the best accuracy performance. This method could predict the good or moderate response to IFX with 92% accuracy. The sensitivity of this method was 96.7%, while the specificity was 75%. CONCLUSIONS: We have developed a novel method for predicting the clinical response using only background clinical data in RA patients before treatment with IFX. Our method for predicting the response to IFX in RA patients may have advantages over the other previous methods in several points including easy usability, cost-effectiveness and accuracy.
OBJECTIVES: The aim of the present study was to generate a novel method for predicting the clinical response to infliximab (IFX), using a machine-learning algorithm with only clinical data obtained before the treatment in rheumatoid arthritis (RA) patients. METHODS: We obtained 32 variables out of the clinical data on the patients from two independent hospitals. Next, we selected both clinical parameters and machine-learning algorithms and decided the candidates of prediction method. These candidates were verified by clinical variables on different patients from two other hospitals. Finally, we decided the prediction method to achieve the highest score. RESULTS: The combination of multilayer perceptron algorithm (neural network) and nine clinical parameters shows the best accuracy performance. This method could predict the good or moderate response to IFX with 92% accuracy. The sensitivity of this method was 96.7%, while the specificity was 75%. CONCLUSIONS: We have developed a novel method for predicting the clinical response using only background clinical data in RApatients before treatment with IFX. Our method for predicting the response to IFX in RApatients may have advantages over the other previous methods in several points including easy usability, cost-effectiveness and accuracy.
Entities:
Keywords:
Clinical data; Infliximab; Machine-learning; Rheumatoid arthritis; The prediction of clinical response