A network analysis of clinical characteristics of patients with COPD: partial results

Aiming to identify associations of clinical outcomes and their conditional independences using network analysis, 121 patients with chronic obstructive pulmonary disease (COPD) were evaluated regarding anthropometrics, lung function, exercise capacity, whole-body composition, peripheral muscle function, functional status, comorbidities, dyspnea and objectively measured daily physical activity in two time-points: before and after a 12-week high-intensity exercise training program (ET). Network-analysis was as by Peter-Clark algorithms, based on partial correlations and conditional independences of outcomes. Thus, two models were constructed (figure 1). In model 1 (pre-ET), despite being conditional to fat-mass index and sedentarism, time in activities >3METs was independent of steps/day and gender; sedentarism and steps/day were conditional to each other; and functional status was conditional to age and dyspnea. In model 2 (post-ET), changing time in activities >3METs was conditional to gender and changes in sedentarism, independent of changes in fat-free mass index and steps/day; changes in steps/day and sedentarism were conditional to each other. In conclusion, relevant conditional independences between patients’ clinical characteristics were demonstrated both before and after an exercise training program in COPD. These interrelationships strongly support the concept of patient-tailored interventions.