A Novel Predictive Model of Diabetic Macular Edema Based on Plasma Cytokines via Machine Learning Algorithms

Objective: The study aimed to investigate the change of plasma cytokines and to build a prediction model for diabetic macular edema (DME) in type 2 diabetes mellitus (T2DM). Methods: In pilot cohort, plasma samples were selected from 15 DM patients and 9 DME patients for screening changes of angiogenesis and inflammation related cytokines by the human cytokine antibody array. These screening cytokines were further validated by ELISA in validation cohort, which containing 100 T2DM with DME (DME group) and 100 T2DM without DME (T2DM group) patients. Pearson correlation were calculated to investigate relationship between cytokines and clinical characteristics. Finally, a prediction model based on plasma cytokines was built by machine learning algorithms. Results: In pilot cohort, 12 plasma cytokines were significant higher, among which there was obvious increase of platelet-derived growth factor-BB (PDGF-BB), tissue inhibitors of metalloproteinase 1 (TIMP-1), angiopoietin 1 (ANG-1), CXC-chemokine ligand 16 (CXCL16) and vascular endothelial cell growth factor receptor 2 (VEGF R2) in DME group (Foldchange > 6 and adjusted p < 0.05), comparing with T2DM group. PDGF-BB, TIMP-1, ANG-1, CXCL16 and VEGFR2 were further confirmed significant elevated in validation cohort. Machine learning algorithms employing random forests (RF) yielded the best performance. In the test set, accuracy, sensitivity, specificity, AUROC and f1 score of RF model were 83.3%, 86.2%, 80.6%, 0.83 and 0.83, respectively. Conclusion: Plasma cytokines were closely association with DME. The prediction model consisting of 4 plasma cytokines in the study was a novel tool for predicting DME in T2DM patients. Funding Statement: This study was funded by Science and Technology Committee of Tongzhou District (Number: KJ2019CX014-25). Declaration of Interests: The authors declare they have no conflict of interest. Ethics Approval Statement: In our research, all procedures complied with the Helsinki Declaration for investigation of human subjects. The study received ethical approval from the competent Institutional Review Boards of Beijing Luhe Hospital. All patient-related procedures were performed strictly following the approval of the internal review and ethics boards of the indicated hospitals.