Fifteen studies were included (n=466 enrolled, 351 randomised and 274 completed). These included six RCTs (n=249 randomised and 191 completed) and nine within-subject crossover studies (classified as the non-RCTs in the review, n=102 randomised, 83 completed). Sample size ranged from 10 to 47 for RCTs and from six to 14 for non-RCTs.
Most studies were rated as high methodological quality. Four of the six RCTs used blinding of randomisation. All RCTs were single-blinded with respect to intervention. All of the cross-over studies (the non-RCTs) used randomisation and attempted to control for confounders statistically
Gas exchange: For RCTs there was no statistically significant difference between bi-level NIPPV and control in arterial oxygen (six RCTs) or arterial carbon dioxide (six RCTs) tension. The non-RCTs found that bi-level NIPPV was associated with a statistically significant increase in arterial oxygen tension (mean difference 4.49 mmHg, 95% CI: 1.43 to 7.55, p=0.004; seven studies) and a significant reduction in arterial carbon dioxide tension (mean difference -3.52 mmHg, 95% CI: -5.93 to -1.11; eight studies) compared to control. Heterogeneity was evident.
Muscle function/work of breathing: Meta-analysis of two non-RCTs showed that bi-level NIPPV was associated with an increase in respiratory muscle strength. Heterogeneity was evident.
Health-related outcomes: Each of four RCTs showed that bi-level NIPPV was associated with a significant improvement in dyspnoea; one of two non-RCTs showed no change in dyspnoea. All three studies (two RCTs and one non-RCT) that assessed health-related quality of life reported significant improvements in at least one health-related quality of life measure.
Results for other outcomes were also reported.