Home non-invasive ventilation in severe COPD: in whom does it work and how?

Background: Not all hypercapnic COPD patients benefit from home NIV, and mechanisms through which NIV improves clinical outcomes remain uncertain. We aimed to identify ‘responders’ to NIV, denoted by a beneficial effect of NIV on PaCO2, health-related quality of life (HRQL) and survival, and investigated whether NIV achieves its beneficial effect through an improved PaCO2. Methods: We used individual patient data from previous published trials collated for a systematic review. Linear mixed effect models were conducted to compare the effect of NIV on PaCO2, HRQL and survival, within subgroups defined by patient and treatment characteristics. Secondly, we conducted a causal mediation analysis to investigate whether the effect of NIV is mediated by a change in PaCO2. Findings: Data of 1142 participants from 16 studies were used. Participants treated with lower pressure support (PS <14 vs ≥14cmH2O) and with lower adherence (<5 vs ≥5h) had less improvement in PaCO2 (mean difference (MD) -0.30kPa, p<0.001 and -0.29kPa, p<0.001 respectively) and HRQL (standardised MD 0.10, p=0.002 and 0.11, p=0.02 respectively), but this effect did not persist to survival. PaCO2 improved more in patients with severe dyspnoea (MD -0.30, p=0.02), and HRQL improved only in participants with <3 exacerbations (SMD 0.52, p=0.03). The mediation analysis showed that the effect on HRQL is mediated partially (23%) by change in PaCO2. Interpretation: Higher PS and daily usage of NIV result in greater improvements in PaCO2 and HRQL. Importantly, we demonstrated that the beneficial effect of home NIV on HRQL is partially mediated through reduced PaCO2.