ECE2018 Symposia Precision Medicine for diabetes (Endorsed by the European Journal of Endocrinology) (3 abstracts)
Sweden.
Diabetes is presently classified into two main forms, type 1 (T1D) and type 2 diabetes (T2D), but especially T2D is highly heterogeneous. A refined classification could represent an important step towards precision medicine in diabetes. We have carried out a data-driven cluster analysis in 15,000 T2D-patients aged 18 years or older from four different cohorts in Sweden and Finland using six variables (age at diagnosis, GAD-antibodies, BMI, HbA1c, HOMA2-B and HOMA2-IR) (Ahlqvist A et al. Lancet D&E, 2018). We thereby identified five replicable clusters of diabetes patients, three more severe forms and two milder forms with different patient characteristics and risk of diabetic complications. Cluster 1 included patients with severe autoimmune diabetes (SAID) and cluster 2 similarly insulin-deficient patients (SIDD) with poor metabolic control and high risk of diabetic retinopathy. Individuals in the most insulin-resistant cluster 3 (SIRD) had a 4-5-fold increased risk of diabetic kidney disease and hepatosteatosis compared to other clusters. The obesity-related cluster 4 (MOD) and age-related cluster 5 (MARD) showed a rather benign course of the disease. A criticism of cluster analyses is that they are rather subjective. To address this criticism we used genetics. One could think that cluster 2 included patients with misdiagnosed T1D, but this cluster did not show any association with T1D-associated HLA types. Cluster 3 showed association with SNPs associated with hepatosteatosis, whereas only clusters 4 and 5 showed clear association with established T2D SNPs. Till date we used a panel of 170 SNPs but are now performing GWAS with the hope to identify cluster-specific gene scores. In conclusion, we have been able to stratify patients into five subgroups predicting disease progression and development of diabetic complications more precisely than the current classification and we have used genetics to validate this clustering.