Chronic obstructive pulmonary disease (COPD) is the commonest indication for domiciliary non-invasive ventilation (NIV) in Europe.1 However, there is a paucity of evidence to support its use. A number of short-term randomised controlled trials (RCTs) failed to show any consistent benefit from NIV.2^–5 In a crossover study comparing 3 months of bilevel ventilation plus long term oxygen therapy (LTOT) with LTOT alone,6 there were small improvements in arterial blood gas tensions during spontaneous breathing by day and in health-related quality of life (QOL) with NIV. In a longer term RCT,7 20 patients randomised to NIV were compared with 24 controls; there was a reduction in the need for hospitalisation in the first 3 months in the NIV group but this effect disappeared by 1 year. One-year survival (78%) was similar in both groups. Dyspnoea, measured using the Borg scale, was reduced in the NIV group. In a larger RCT from Italy,8 90 stable patients on oxygen for more than 6 months were randomly assigned to continuing oxygen alone or oxygen and bilevel ventilation. There was no change in survival, lung or inspiratory muscle function, exercise tolerance or sleep quality score in either group. By contrast, the arterial carbon dioxide tension (Paco₂) measured on usual oxygen and resting dyspnoea scores improved. Health-related QOL, as assessed by the Maugeri Foundation Respiratory Failure Questionnaire (MRF 28), improved in the NIV plus oxygen group, but there was no difference in St George’s Respiratory Questionnaire (SGRQ) scores. NIV has also been used as an adjunct to pulmonary rehabilitation with two RCTs showing greater improvements in exercise capacity and QOL when NIV, used overnight, was added to a pulmonary rehabilitation programme.9 10

In this issue of Thorax McEvoy et al11 report the results of an RCT of NIV in patients with chronic stable hypercapnic ventilatory failure due to COPD (see page 561). The study was powered for survival. The patients had severe airflow limitation with a forced expiratory volume in 1 s (FEV₁) approximately 25% predicted and severely impaired QOL. The sleep-related increase in transcutaneous carbon dioxide tension (Tcco₂) was reduced in the NIV + LTOT group compared with the group receiving LTOT alone (12.6 vs 18.8 mm Hg). Over the course of the study there were 40 deaths in the NIV group compared with 46 in the LTOT group. At 24 months, 47% of the LTOT-treated patients had died compared with 32% of the NIV-treated patients. By 3.5 years the survival curves had come together. At follow-up at 1 year there was no change in Paco₂ during spontaneous breathing by day on oxygen at the same flow rate in either group or in the FEV₁. QOL as measured by the SGRQ was unchanged, but statistically significant differences were observed in several subscales of the SF36 and the profile of mood questionnaires, suggesting that patients treated with NIV had poorer general and mental health, and reported less vigour and more confusion and bewilderment.

It is interesting to compare these data with those previously published (table 1). The rationale behind the use of NIV and the way that it was applied in these studies is different. An understanding of how NIV “works” is critical in determining the therapeutic end point of NIV, but this remains unresolved.12^–14 Competing theories include resting of chronically fatigued respiratory muscles, reducing the load against which the respiratory muscle pump must work or restoring the central drive to breathe. There is evidence that restoration of central chemosensitivity12 and a reduction in load15 are important, and the ventilator needs to be set to have an effect upon these parameters. In negative studies, if important physiological parameters are not affected, it cannot be concluded that NIV really had no effect. A major problem with previous studies is that there has either been no clear physiological target behind the choice of ventilator mode and settings or the evidence that an effect has been achieved is weak. Casanova et al7 adjusted pressures to decrease accessory muscle use, reduce the sensation of dyspnoea and reduce respiratory rate by 20%. Oxygen was titrated to maintain saturations of >90%. No measurements were made of carbon dioxide tensions or data provided confirming the degree to which accessory muscle use or respiratory rate were reduced. No evidence therefore is provided to confirm a physiological effect. Clini et al8 used pressure support ventilation delivered by a spontaneous/timed bilevel ventilator through a nasal mask. Oxygen was added to achieve a target saturation of >90%. All patients underwent a 10-night NIV trial in hospital before randomisation. Patients were deemed to be compliant when they used the ventilator for 5 h/night and, among the compliant patients, the average was high at 9.2 h/night. No data were provided as to the proportion of “compliant” patients or the average compliance overall. The effectiveness of NIV had to be proven by a 5% decrease in Paco₂ after 1 h of continuous support and by night time oximetry. NIV was considered effective when the patient spent >90% of the recording time with saturations >90% during NIV. However, oximetry cannot be used to assess the adequacy of ventilation when patients are receiving supplemental oxygen.16 Again, the evidence of a physiological effect from NIV when it was administered (ie, during sleep) is not convincing.

In the study by McEvoy et al,11 patients assigned to NIV were admitted for 3–4days to hospital for education and familiarisation with a patient-triggered bilevel positive pressure device. Bilevel positive airway pressure was gradually increased during daytime and night time trials to the maximum tolerated with a target difference between inspiratory positive airway pressure (IPAP) and expiratory positive airway pressure (EPAP) of ⩾10 cm H₂O. The level of EPAP was determined during a polysomnographic study, with the level being increased to abolish snoring and obstructive events. NIV was considered successfully established when at least 3 h of sleep was confirmed on NIV at an IPAP–EPAP difference of at least 5 cm H₂O. There was no requirement to change oxygen saturation or Tcco₂, although NIV did reduce the maximum sleep-related rise in Tcco₂ from 16.5 to 12.6 mm Hg. If restoration of central drive is important, the patient needs to spend the night, on average, with a respiratory alkalosis. This will allow the excretion of bicarbonate and restoration of central drive. This requires the carbon dioxide tension to fall with NIV; if only the highest rises in carbon dioxide tension are reduced, this aim will not be achieved. The results of the bicarbonate concentration during daytime arterial blood gas measurements are not reported in this study, but would be expected to fall if carbon dioxide was reduced during NIV overnight.

Diaz and co-workers15 17 performed short-term (3 weeks) RCTs of daytime NIV compared with a control group who received sham ventilation. They gradually increased IPAP to the maximum that could be tolerated during the day and achieved an average of 18 cm H₂O. This resulted in an approximate doubling of the tidal volume and halving of the respiratory rate. This would be expected to produce a substantial increase in alveolar ventilation and improve arterial blood gas tensions, although these data were not presented. They also showed a 48% reduction in the pressure time product of the diaphragm, indicating substantial unloading of the diaphragm. At the end of 3 weeks of ventilation there were improvements in both the Pao₂ (+8.6 mm Hg) and the Paco₂ (−8.4 mm Hg) during spontaneous breathing. There was a 93 ml (13%) increase in forced expiratory volume in 1 s (FEV₁) and a reduction in hyperinflation, gas trapping and intrinsic positive end expiratory pressure (PEEP). Tidal volume during spontaneous breathing increased by 181 ml. In a separate study with a similar design17 they showed an increase in the 6-minute walking distance (mean 76 m) and a reduction in dyspnoea scores. There were similar changes in arterial blood gas tensions during spontaneous breathing and an 80 ml increase in FEV₁ which correlated with the improved walking distance.

The study by Windisch et al,18 while uncontrolled, also provides some important insights into how NIV might be best used in patients with COPD. NIV was initiated in hospital and the level of inspiratory pressure support was gradually increased aiming for normocapnia, based on previous work showing that high inflation pressures could be tolerated resulting in improved Paco₂ during spontaneous breathing19 and also in QOL.20 Using this strategy, they showed that Paco₂ fell during spontaneous breathing by a mean of 7 mm Hg, Pao₂ increased by a mean of 6 mm Hg and FEV₁ improved by a mean of 140 ml. The 2-year survival rate was 86%.

Taken together, these studies suggest that benefit is seen in daytime parameters (physiological and QOL) when there is a reduction in carbon dioxide during NIV. Given that NIV is usually applied during sleep, overnight monitoring is mandatory. End-tidal carbon dioxide measurements are of no use in patients with obstructive lung disease and monitoring of transcutaneous or arterial carbon dioxide tensions is required. Most of these studies used bilevel ventilators in spontaneous mode. Ventilators without a timed back-up are significantly cheaper and, in the absence of any data confirming that a timed back-up is needed, these should be the machine of choice for domiciliary NIV in COPD. Higher inflation pressures appear to be necessary to affect physiological parameters.

The observation of improvements in FEV₁— often larger than those obtained with bronchodilators—is intriguing. For instance, in the UPLIFT Trial21 the mean improvement in FEV₁ with tiotropium was 87–103 ml over 4 years. In the TORCH Study22 the mean increase in FEV₁ over 3 years with the combination of salmeterol and fluticasone was 92 ml. Diaz et al15 and Windisch et al19 both reported comparable increases in FEV₁ after NIV. This raises the possibility that NIV has an effect on the airways themselves; possible mechanisms include a reduction in airway oedema23 or even stretching open of chronically fibrosed airways. If correct, it is likely that this will require a level of pressure to be delivered to the airways for a period of time. If this is an important mechanism, logic would suggest that the higher the pressure and the longer it is applied to the airways, the better. This is entirely speculative, but the improvements in FEV₁ were only seen in these studies,15 23 both of which used higher inflation pressures.

Although McEvoy et al showed an improvement in survival, this was apparently at the cost of worsening QOL with worse scores in generic questionnaires at 12 months.11 However, there are a number of difficulties with this. As pointed out by the authors, more patients in the non-invasive group completed QOL questionnaires than in the control group. It is also notable that only 50–60% of patients repeated the QOL questionnaire at 12 months. These factors may introduce “survivor” effects, which may introduce significant bias if patients with poor QOL do not feel well enough or choose not to fill in questionnaires. Second, many different domains of QOL were measured and no adjustment was made for multiple measures; there is a possibility that—by chance alone—one or more might be statistically significant. Finally, there is question about the choice of questionnaire. The disease-specific questionnaire (SGRQ) did not show a deterioration in QOL. Although the SGRQ was developed in patients with milder disease, changes in QOL have been demonstrated in patients with COPD receiving NIV.6 The generic questionnaires did show a deterioration, and the SF36 has been shown to be reliable and responsive in COPD.24 25 Which measures should be believed? In other studies,6 26 improvements in QOL have been shown to correlate with a reduction in bicarbonate, suggesting that effective control of nocturnal hypoventilation is important in improving quality of life. McEvoy et al showed an attenuation in the sleep-related rise in carbon dioxide, but the failure to lower carbon dioxide may explain the absence of a beneficial effect on QOL.

One other potential role for domiciliary NIV in patients with COPD is after discharge for an episode in which NIV has been used acutely in hospital.27 These patients have a poor prognosis,28 and there are some uncontrolled data to suggest that NIV may reduce the need for hospitalisation subsequently and be economically advantageous.29 In the study by McEvoy et al,11 hospitalisation rates did not differ between the two groups. However, it is difficult to interpret these data as the NIV + LTOT group had 7474 more days on the trial with a median follow-up of 28.5 months compared with 20.5 months in the LTOT alone group. It may not be appropriate to compare raw rates in a progressive disease such as COPD; patients are probably more likely to be admitted when the condition is more advanced. In the study by Clini et al,8 hospital admissions did not differ between the groups during follow-up. However, in a post hoc analysis comparing the year before with the first year of the trial, hospital admissions decreased by 45% in the NIV + oxygen group and increased by 27% in the oxygen group. This issue warrants further investigation in future trials.

What can be concluded about the current role for NIV in patients with stable COPD following the study of McEvoy et al? First, that ventilation for an average of 4 h/night with modest inflation pressures applied with the intent of reducing sleep-disordered breathing reduced the highest peaks of Tcco₂ and improved survival. The benefit was lost by 3.5 years. Second, this improvement in survival may have been at the expense of QOL. The results from different studies using different questionnaires are conflicting and future studies should use questionnaires validated in patients with COPD with chronic respiratory failure receiving mechanical ventilation such as the SRI or MRF288 20; if multiple questionnaires are used there must be a clearly defined analysis plan from the outset. Third, future trials must have a clear therapeutic end point during NIV and it must be confirmed objectively that this is achieved. Available data suggest that this should be a demonstration of a reduction in the carbon dioxide level overnight. This should be to a level that allows net excretion of bicarbonate and restoration of central drive. It is likely that higher pressures will be needed to achieve this. The possibility that NIV has a direct effect on the airways needs to be explored in future studies and higher pressures may be important in this regard. The case for domiciliary NIV in patients with COPD remains unproven, and most patients with COPD should probably only be offered NIV in the context of a clinical trial. However, there is sufficient evidence of benefit to suggest that trials of domiciliary NIV for COPD should be funded, as COPD continues to be a major cause of morbidity and incurs high costs for healthcare providers. In the absence of such trials, it is reasonable for hypercapnic patients with COPD, already established on LTOT, to be offered domiciliary NIV.