Supplementary Material for: Long-Term Noninvasive Ventilation in Chronic Obstructive Pulmonary Disease: Association between Clinical Phenotypes and Survival

Background: Long-term noninvasive ventilation (LTNIV) is widely used in patients with chronic hypercapnic respiratory failure (CHRF) related to COPD. Prognosis of these patients is however poor and heterogenous. Research Question: In COPD patients under LTNIV for CHRF, is it possible to identify specific phenotypes which are predictive of probability of pursuing NIV and survival? Study Design and Methods: A latent class analysis was performed in a COPD population under LTNIV included in a comprehensive database of patients in the Geneva Lake area, to determine clinically relevant phenotypes. The observation period of this subgroup of COPD was extended to allow assessment of survival and/or pursuit of NIV for at least 2 years after inclusion. A logistic regression was conducted to generate an equation accurately attributing an individual patient to a defined phenotype. The identified phenotypes were compared on a series of relevant variables, as well as for probability of pursuing NIV or survival. A competitive risk analysis allowed to distinguish death from other causes of cessation of NIV. Results: Two phenotypes were identified: a “respiratory COPD” profile with very severe airway obstruction, a low or normal body mass index, and a low prevalence of comorbidities and a “systemic COPD” profile of obese COPDs with moderate airway obstruction and a high rate of cardiovascular and metabolic comorbidities. The logistic regression correctly classified 95.7% of patients studied. Probability of pursuing NIV and survival were significantly related to these phenotypes, with a poorer prognosis for “respiratory COPD.” Probability of death 5 years after implementing NIV was 22.3% (95% CI: 15.4–32.2) for “systemic COPD” versus 47.2% (37.4–59.6) for “respiratory COPD” (p = 0.001). Conclusion: The two distinct phenotypes of COPD under LTNIV for CHRF identified appear to be strongly related to prognosis and require further validation in other cohort studies.