Type 2 diabetes mellitus (T2DM) is an inflammatory condition. C-reactive protein (CRP) is a well-known inflammatory marker for demonstrating systemic inflammation in T2DM. We aimed to compare the diagnostic value of CRP and C-reactive protein to albumin ratio (CAR) in discriminating T2DM patients and healthy controls and to investigate the association of CAR and glycemic control markers. This cross-sectional retrospective study included 173 T2DM patients and 104 healthy control subjects. Serum albumin and CRP levels were measured with spectrophotometric and nephelometric methods, respectively. CAR was calculated by dividing CRP level by albumin level. The median (interquartile range, IQR) CAR in the T2DM patients was significantly higher than those in the controls (0.15 (0.26) vs. 0.06 (0.20), p = 0.002, respectively). In receiver operating characteristics curve analysis (ROC), CAR > 0.0653 had 77.21% sensitivity and 52.44% specificity in discriminating T2DM. The area under ROC curve for CAR was 0.623 (95% CI =0.544–0.701, p=0.002), which was similar to CRP. A positive correlation was found between CAR and body mass index, fasting blood glucose and HbA1c levels. As a result, CAR demonstrated poor sensitivity and specificity for discriminating T2DM patients from healthy subjects and it has little diagnostic utility in diabetes.
1 September 2025 in International Journal of Statistics in Medical Research
Introduction: This study aims to determine the stages of Alzheimer's disease (AD) using different machine learning algorithms, and compares the performance of these models. Methods: Demographic, genetic, and neurocognitive inventory data from the National Alzheimer's Coordinating Center (NACC) database as well as brain volume/thickness data from magnetic resonance imaging (MRI) scans were used. Deep Neural Networks, Ordinal Logistic Regression, Random Forest, Gaussian Naive Bayes, XGBoost, and LightGBM models were used to identify four different ordinal stages of AD. Results: Although the performance measures of the developed models were similar, the highest classification rate of AD stages was achieved by the Random Forest model (accuracy: 0.86; F1 score: 0.86; AUC: 0.95). The outputs of the model with the best performance were explained by the SHapley Addictive exPlanations (SHAP) method. Conclusions: This indicates that non-invasive markers and machine learning models can be used effectively in early diagnosis and decision support systems to predict stages of AD.