Artificial Neural Network-Based Prediction of Fatigue in Preeclampsia: Role of Inflammatory and Placenta-Related Biomarkers

Abstract

Background. Preeclampsia (PE) is associated with inflammation and many neuropsychiatric symptoms, including fatigue. Many biomarkers are correlated with fibro fatigue (FF) scores in various diseases. In the present study, inflammatory and placenta-related biomarkers are utilized to predict fatigue severity in PE using an artificial neural network (ANN).
Methods. Two PE patient groups in the study: one with high fibro fatigue (FF) scores (PE-FF≥25, n=45) and another with lower FF (PE-FF<25, n=65). Fifty pregnant women were recruited as a control group. Pentraxin-3, VEGF, hs-CRP, osteoprotegerin, TNFα, high-sensitive C-reactive protein (hs-CRP), insulin, FABP4, and TRAIL were all measured using the ELISA method.
Results. Both patient groups' serum levels of insulin resistance indices, glucose, fasting insulin, FABP4, and PTX3 are significantly higher than those of the control group. The PE-FF≥25 group had substantially greater serum levels of TRAIL, TNFα, PAI-1, and hs-CRP than both the PE-FF<25 group and the control group. FFtotal exhibits a strong association with TRAIL, OPG, TNFα, hs-CRP, and FABP4. The ANN model's findings demonstrated 100% sensitivity and specificity in differentiating between patients with PE-FF≥25 and those with PE-FF<25. hs-CRP has the highest predictive power (100%), according to the model's predictive strength. TRAIL (97.3%), FABP4 (87.9%), TNFα (64.2%), and OPG (55.8%) are next in line.
Conclusions. Fatigue is correlated with the inflammatory and placenta-related biomarkers in PE women. The top three predictors of severe fatigue using ANN are hs-CRP, TRAIL, and FABP4, which all point to an inflammatory state in PE.