A-025 Principal Component Analysis as a Clustering Method for the Identification of Populations with Risk of Diabetes Development

Abstract

Abstract Background Obesity is a chronic disease, defined by the World Health Organization as the abnormal or excessive accumulation of fat in the body. Overweight populations have grown epidemically in all age groups in recent decades and represent a major public health issue globally. Excess weight is one of the potential factors for insulin resistance, thus increasing the risk of developing type 2 diabetes. This study aimed to investigate the urea/creatinine ratio as a marker of renal function in individuals with different body mass index (BMI) levels and examine clusters with increased risk for diabetes development. Methods Study population and characteristics: This study was a 34-month retrospective analysis from a large laboratory database in the city of São Paulo, Brazil (January 2019–November 2022). Individuals ranging between 19–60 years old were selected based on the availability of BMI data and results for creatinine, urea, serum glucose, and glycated (A1C). Only outpatient results were included in the current analysis. The following parameters were included in this study: BMI (kg/m2): 18.5–24.9 (healthy), 25–29 (overweight), 30–34.9 (overweight level 1), 35–39.9 (overweight level 2), and ≥40 (overweight level 3). The definition of diabetes risk was as follows: (i) without diabetes risk: serum glucose>100 mg/dL and A1C < 5.7; (ii) transition to diabetes risk: serum glucose 101–125 mg/dL and A1C 5.7%–6.4%; (iii) diabetes risk: serum glucose >126 mg/dL and A1C > 6.4%. Statistical analysis: Principal component analysis (PCA) was performed resulting in a division of four subpopulation clusters based on the risk of diabetes development. One-way ANOVA with the Tukey test was used for intergroup comparison. For all results *P < 0.05 was considered statistically significant. Results A total of 667 participants met the inclusion criteria. From all BMI profiles, the overweight population in diabetes risk situation showed a significant difference in the ratio of urea/creatinine compared with the same BMI population without diabetes risk [33 ± 11.5 vs 38 ± 11.8 Mean ± SD] (P < 0.05). PCA analysis formed four clusters: PC1+ (cluster 1) with increased creatinine (1.9 mg/dL) & urea (63.5 mg/dL) and transition to diabetes; PC2+ (cluster 2) with normal creatinine (0.7 mg/dL) & urea (28.1 mg/dL) and diabetes risk; PC1+PC2+ (cluster 3) with increased creatinine (1.4 mg/dL), increased urea (50.44 mg/dL), and diabetes risk; PC1-PC2- (cluster 4) with normal creatinine (0.7 mg/dL), normal urea (28.7 mg/dL), and transition to diabetes. PCA analysis found that Cluster 1 had a higher urea/creatinine ratio (34.7 ± 10.6 vs 40 ± 12.3, Mean ± SD) compared to Cluster 2 (P < 0.01). Cluster 1 was made up of mostly men (80.5%) with an average age of 54 ± 8.9 years, of which 22% were normal weight and 34.2% were overweight. Cluster 2 had more women (65.10%) with an average age of 47 ± 11.7 years, 20.6% normal weight and 32.6% overweight. A total of 17.9% of overweight individuals with a transition to diabetes risk were found in PC1 (14/78) and 8.1% in PC2 (4/49). Conclusion Urea/creatinine ratio could be a useful indicator of diabetes risk in overweight individuals. PCA selection method identifies diabetes risk indexes for early care management by clinicians.