Yu Liu1, Miaomiao Sang1, Yang Yuan1, Ziwei Du1, Wei Li2, Hao Hu1, Liang Wen3, Fenghua Wang4, Haijian Guo5, Bei Wang6, Duolao Wang7, Zilin Sun8, Shanhu Qiu9. 1. Department of Endocrinology, Zhongda Hospital, Institute of Diabetes, School of Medicine, Southeast University, No.87 Dingjiaqiao Street, Nanjing, 210009, People's Republic of China. 2. Department of Endocrinology, Suzhou Hospital of Anhui Medical University (Suzhou Municipal Hospital of Anhui Province), Suzhou, China. 3. Department of Ophthalmology, Fushun Eye Hospital, Fushun, China. 4. Beijing Tongren Eye Center, Beijing Tongren Hospital, Beijing Ophthalmology and Visual Science Key Lab, Capital Medical University, Beijing, China. 5. Jiangsu Provincial Centre for Disease Control and Prevention, Nanjing, China. 6. Key Laboratory of Environmental Medicine and Engineering of Ministry of Education, Department of Epidemiology and Health Statistics, School of Public Health, Southeast University, Nanjing, China. 7. Liverpool School of Tropical Medicine, Liverpool, UK. 8. Department of Endocrinology, Zhongda Hospital, Institute of Diabetes, School of Medicine, Southeast University, No.87 Dingjiaqiao Street, Nanjing, 210009, People's Republic of China. sunzilin1963@126.com. 9. Department of General Practice, Zhongda Hospital, Institute of Diabetes, School of Medicine, Southeast University, No.87 Dingjiaqiao Street, Nanjing, 210009, People's Republic of China. tigershanhu@126.com.
Abstract
BACKGROUND: Cluster analysis may assist in stratifying heterogeneous clinical presentations of type 2 diabetes (T2D). However, the association of cluster-based subgroups with diabetes-related outcomes such as diabetic retinopathy remains unclear. This study was aimed to address this issue with novel clusters of T2D derived from four simple parameters. METHOD: We developed a k-means clustering model in participants with newly diagnosed T2D (N = 1910) from the SENSIBLE and SENSIBLE-Addition studies, based on body mass index (BMI), waist circumference (WC), mean arterial pressure (MAP), and hemoglobin A1c (HbA1c). Diabetic retinopathy was ascertained with the protocol from the Early Treatment of Diabetic Retinopathy Study. Participants (N = 515) without diabetic retinopathy at baseline were followed-up for 3 years. Logistic regression analyses were performed to obtain the odds ratios (ORs) and 95% confidence intervals (CIs). RESULTS: Three clusters were identified, with cluster 0, 1 and 2 accounting for 48.2, 8.9 and 42.9%, respectively. Participants with T2D were featured by the lowest BMI, WC, MAP, and HbA1c in cluster 0, poor glycemic condition in cluster 1, and the highest BMI, WC, and MAP in cluster 2. Compared with cluster 0, cluster 1 was associated with increased odds of diabetic retinopathy in both the cross-sectional study (OR 6.25, 95% CI: 3.19-12.23) and the cohort study (OR 9.16, 95% CI: 2.08-40.34), while cluster 2 was not. Moreover, most participants remained their clusters unchanged during follow-up. CONCLUSIONS: Our cluster-based analysis showed that participants with poor glycemic condition rather than high blood pressure and obesity had higher risk of diabetic retinopathy.
BACKGROUND: Cluster analysis may assist in stratifying heterogeneous clinical presentations of type 2 diabetes (T2D). However, the association of cluster-based subgroups with diabetes-related outcomes such as diabetic retinopathy remains unclear. This study was aimed to address this issue with novel clusters of T2D derived from four simple parameters. METHOD: We developed a k-means clustering model in participants with newly diagnosed T2D (N = 1910) from the SENSIBLE and SENSIBLE-Addition studies, based on body mass index (BMI), waist circumference (WC), mean arterial pressure (MAP), and hemoglobin A1c (HbA1c). Diabetic retinopathy was ascertained with the protocol from the Early Treatment of Diabetic Retinopathy Study. Participants (N = 515) without diabetic retinopathy at baseline were followed-up for 3 years. Logistic regression analyses were performed to obtain the odds ratios (ORs) and 95% confidence intervals (CIs). RESULTS: Three clusters were identified, with cluster 0, 1 and 2 accounting for 48.2, 8.9 and 42.9%, respectively. Participants with T2D were featured by the lowest BMI, WC, MAP, and HbA1c in cluster 0, poor glycemic condition in cluster 1, and the highest BMI, WC, and MAP in cluster 2. Compared with cluster 0, cluster 1 was associated with increased odds of diabetic retinopathy in both the cross-sectional study (OR 6.25, 95% CI: 3.19-12.23) and the cohort study (OR 9.16, 95% CI: 2.08-40.34), while cluster 2 was not. Moreover, most participants remained their clusters unchanged during follow-up. CONCLUSIONS: Our cluster-based analysis showed that participants with poor glycemic condition rather than high blood pressure and obesity had higher risk of diabetic retinopathy.