P11-04

Development of a model to predict the severity of systemic lupus erythematosus using LIFE Study data

Kiyohiro TOYOFUKU *1Koichiro KATO1, 3Haruhisa FUKUDA2

1Department of Applied Chemistry, Graduate School of Engineering, Kyushu University
2Department of Health Care Administration and Management, Graduate School of Medical Sciences, Kyushu University
3Center for Molecular Systems, Kyushu University
( * E-mail: toyofuku.kiyohiro.864@s.kyushu-u.ac.jp )

Systemic lupus erythematosus (SLE) is an autoimmune disease that causes inflammation and tissue damage in various organs throughout the body. It is one of the designated intractable diseases, and it is estimated that there are approximately 60,000 to 100,000 patients throughout Japan.
Although advances in steroids and immunosuppressive drugs have greatly improved the prognosis of life, quality of life (QOL) is still a problem due to organ damage caused by SLE and complications associated with long-term steroid use. Therefore, it is considered important to prevent worsening of the disease and to maintain a stable disease state for a long period of time with minimal steroid use, with the goal of preventing organ damage and improving health-related QOL. In Japan, SLE has been treated with immunosuppressive agents (tacrolimus and mizoribine), but in recent years, important drugs for SLE treatment such as hydroxychloroquine, mycophenolate mofetil (immunosuppressive agent), and belimumab anifrolumab (biological agent) have become covered by insurance one after another, making it possible to treat SLE as in Europe and the US. This has made it possible to provide SLE treatment similar to that in Europe and the United States. Currently, Japan has a large number of therapeutic agents to choose from, including conventional therapies, but the treatment methods suitable for Japanese patients and the optimal treatment strategy to prevent worsening of the disease (relapse) and to maintain a stable state (remission) have not yet been clarified. Therefore, we planned a data-driven investigation to find out which factors are important for the maintenance of disease severity and remission in SLE. By conducting this study, we believe that new aspects of SLE in the Japanese population will be clarified, which will help improve patient outcomes and maintain QOL.
To collect the data of SLE patients, we use a database produced by the Longevity Improvement & Fair Evidence (LIFE) Study. In this study, SLE patients were identified from LIFE Study data using ICD10 codes and extracted by linking information on specific health checkups and medications. Since patients with severe SLE generally receive high doses of steroids and immunosuppressive drugs, we scored SLE severity based on the drugs administered. Next, these data were arranged into a time series, and patients were classified into three severity classes based on the shape of the time series. Then, a random forest classification model was constructed using patient information (e.g., laboratory values) as explanatory variables and the severity classes calculated from the time series as objective variables, and explanatory AI (SHAP) was used to search for important factors. The results suggest that there is an association between blood LDL cholesterol level and the risk for severe disease.