Cluster and Heatmap Analysis in Idiopathic Dilated Cardiomyopathy (IDC): Discriminatory Variables

Purpose: Cardiomyopathy (CM) is often an idiopathic diagnosis that makes specific management difficult. Pathology studies in IDC have not led to better treatments and many patients (pts) require transplantation or die of their disease. We have shown that IDC patients cluster with others of specific etiologies into subgroups. In this study, we perform hierarchical cluster analysis followed by development of heatmaps to identify descriptors which define the subgroups. Methods: We used database of endomyocardial biopsy results from pts with unexplained heart failure and normal coronary arteries. 24 pathologic parameters were scored. 18 clinical parameters of interest were also separately sought in electronic medical records using ICD-9 codes and/or written records. Clinical features of interest were coded as positive or negative/not identified. Patient outcomes were defined using state death registries. We performed hierarchical clustering using simple matching similarity measurement for binary data and Euclidian distance similarity measures for scored parameters. Variables of interest were then selected using the cluster heat map which gave us a visual picture of the patterns generated. Results: 304 pts diagnosed 1989–1994 were clustered in the analysis. 173 pts with IDC were compared to 95 pts with inflammatory CM (IFCM) including myocarditis, autoimmune disease and rheumatic heart disease, 7 pts with post-partum CM (PPCM), 13 pts with toxic CM (TCM), and 11 pts with ischemic CM (ISCM). 4 clusters were defined. Histologic and IF features contributed to clusters 1 only. EM features were distinctive in every cluster. Capillary endothelial changes occurred in clusters 1,2,4. Myocardial sarcoplasmic reticulum changes defined cluster 3. Autoimmunity detected by ICD 9 codes was a significant clinical feature in clusters 1,4. Diabetes, smoking, obesity were important clinical features in clusters 3,4. Conclusion: Autoimmunity is a common clinical feature of IDC; EM is a useful tool to define CM. Heat map designations give tools to define CM further.Tabled 1Clinical diagnosis and cluster distributionDiagnosisCluster 1Cluster 2Cluster 3Cluster 4IDC4587140ISCM1082IFCM23421119TCM00130PPCAM0431 Open table in a new tab