Author + information
- 1Department of Emergency Medicine, the Second Affiliated Hospital, Xi'an Jiaotong University, Xi'an, China
- 2Shaanxi Engineering Research Center of Medical and Health BIG DATA, Xi'an Jiaotong University School of Management, Xi'an, China
- 3Department of Cardiology, the Second Affiliated Hospital, Xi'an, China
Carotid plaque evaluation is useful in cardiac clinics to predict coronary artery disease for patients with various cardiovascular risk factors, such as hypertension. As far as heterogeneity is concerned for hypertension as a common condition, clustering analysis was applied to identify phenotypically distinct hypertensive group with different severity of carotid plaque.
In this study, a total of 513 patients were continuously recruited and evaluated with carotid ultrasound from April 2012 to June 2013. Clustering analysis was performed on 27 baseline clinical variables including demographic characteristics, medical history, laboratory values and echocardiographic presentation. As is necessary for clustering analysis, the missing values were globally replaced with mean values. Carotid plaque was classified into Grade 0 (normal or no observable plaque), Grade 1 (mild stenosis, 1%-24% narrowing), and Grade 2 (moderate stenosis, ≥25% narrowing). After clustering analysis, differences among cluster centers were using the χ2 test for binary variables and ANOVA for continuous data.
Three clusters were eventually identified. Cluster 1 (n=246) was the largest cluster than the others. All of these patients were males, youngest on average (59.0±13.2) and most likely to be smokers (67.9%). They had the lowest mean levels of total cholesterol (4.5mmol/L) and low density lipoprotein (LDL, 2.6mmol/L), while the 24h-systolic BP (SBP, 137mmHg), 24h-diastolic BP (DBP, 82mmHg), daytime SBP (139mmHg), daytime DBP (83mmHg) and night-time DBP (77mmHg) were the highest than other two clusters. Surprisingly, we found the majority of patients having moderate stenosis belong to this cluster, including 32 (80.0%) patients. Cluster 2 (n=97) patients tended to be older (65.4±11.9) with highest levels of total cholesterol (5.0mmol/L), LDL (3.0mmol/L) and fasting glucose (6.8mmol/L). Furthermore, cluster 2 patients had the highest rates of mild stenosis (68.0%) and diabetes mellitus (100%) while had the lowest percentage of smokers (12.4%) and circadian decline rate of SBP (4.4±7.5). Cluster 3 included the lowest percent of male (0%) and diabetes mellitus (0%) with a mean age of 62.9. Compared with other two clusters, this group had the lowest 24h-SBP (131mmHg), daytime SBP (132mmHg), night-time SBP (125mmHg) and fasting glucose (4.7mmol/L). In addition, the rate of no observable plaque was highest among the three clusters.
Clustering analysis of diverse variables identified 3 distinct categories of hypertensive patients with carotid plaque. These results demonstrated that clustering techniques allow visualization of complex multivariate data and may provide new hypothesis for clinical study on carotid plaque.