Cluster analyses in a sample of COPD patients

Abstract

Defining phenotype of COPD patients has been necessary to optimize pharmacological treatment and recognize mortality risk factors. The aim of this study was to identify clusters in a sample of 166 COPD patients (GOLD stage: I:31, II:58, III:30, IV:47). All individuals underwent to medical history and physical examination, clinical and laboratory assessments and pre-and post-bronchodilator spirometry. Comorbidities were registered based on medical chart diagnoses and clinical evaluation and the burden classified according to Charlson index. Arterial blood gases, inflammatory mediators, exercise capacity (six minute walk distance-6MWD) and quality of life (St. George's Respiratory Questionnaire-SGRQ) were evaluated. All statistical analyses were performed using Viscovery Profiler 5.3 by Viscovery Software GmbH (www.viscovery.net;Vienna,Austria). The variables included in the analyses were: Charlson index, smoking history, 6MWD, SGRQ scores (impact, symptoms and activity), baseline dyspnea index (BDI), PaO₂, PaCO₂, FEV₁, hematocrit, lymphocytes, albumin, interleukin 6 (IL6) and C-reactive protein (CRP), glycemia, triglycerides, LDL, HDL and body mass index (BMI). We identified six different clusters. C1: comorbidity [SGRQ impact score, Charlson index] 36.31%, C2: quality of life impairment [SGRQ activity, impact, symptoms scores] 23.81%, C3: metabolic profile [LDL, triglycerides and 6MWD] 16.07%, C4: severe COPD [BDI, FEV₁, PaO₂, SGRQ activity score and 6MWD] 9.52%, C5: inflammatory systemic [IL6 and 6MWD] 10.12% and C6: obstructive sleep apnea [glycemia, BMI and PaCO₂] 4.17%. In conclusion, the findings show that COPD patients can be classified in different clusters that could interfere in the management and outcomes during follow-up.