Chronic obstructive pulmonary disease (COPD) continues to be one of the leading causes of death and disability worldwide,1 ranking fourth in the world's leading causes of death. Several therapeutic options have emerged, but to date, only smoking cessation and long-term oxygen therapy (LTOT) have been shown to increase survival.2 However, patients with more severe COPD continue to be a challenge due to their greater disability and consumption of medical resources, noting that once hypercapnia develops, 2-year mortality increases by approximately 30–40%.3

In COPD patients with chronic hypercapnic failure, nocturnal noninvasive positive pressure ventilation (NIPPV) has been postulated as a potentially beneficial therapy. There are several theories that explain why NIPPV could be beneficial in these types of patients, including improvement in sleep and time efficiency,4 amelioration of nocturnal hypoventilation,5 recovery of inspiratory muscle function6 and a decrease in hyperinflation leading to an improvement in respiratory mechanics.5,7 However, the effects of chronic NIPPV in stable hypercapnic COPD patients in randomized clinical trials, retrospective studies and systematic reviews are contradictory in terms of benefits on survival rates, reductions in hospital admissions, decreased carbon dioxide retention and improvements in symptoms or quality of life.8–15 This may be due to long recruitment periods, heterogeneity of the centers involved and their therapeutic strategies, variability in the exclusion criteria, small numbers of patients and uncontrolled designs.

Despite the lack of clear evidence from randomize control trials (RCTs), the number of COPD patients on home noninvasive ventilation (HNIV) has increased progressively worldwide. RCTs have high internal validity but often represent fewer than 5% of patients treated in routine care.16 Therefore, it is difficult to extrapolate the results to the general population, and furthermore, there are retrospective studies based on daily clinical practice showing that long-term NIPPV reduces hospital readmissions after acute exacerbation in addition to other benefits,13 so other study designs could be very enlightening. Additionally, in the real world, it is sometimes difficult to define whether NIPPV has been prescribed due to worsening of COPD or has been prescribed for other causes in patients who have COPD.

We created a curated cohort of all COPD patients ventilated in Catalonia to describe patient characteristics from multiple perspectives and to perform a cluster analysis of this cohort to evaluate its different clinical and mortality evolution.

In the real world, the number of COPD patients receiving HNIV is growing. However, RCTs show controversial results to support the indication of HNIV to treat chronic respiratory failure in COPD patients. To our knowledge, this is the first well-characterized cohort including 247 ventilated severe COPD patients with large administrative data and long-term follow-up, allowing us to describe what happens in real practice.

The main result of this COPD severe cohort was the high heterogeneity of the HNIV prescription, with more than half of the patients having coexisting obesity (59.9%) and sleep apnea (55.1%). Moreover, 25% of patients were on HNIV only, without long-term oxygen therapy. The latter were frequently obese (33%) and with sleep apnea (40%). Selecting the appropriate variables, cluster analysis determined four representative distinct groups of patients: Cluster One (1) was composed of obese smokers with a high proportion using LTOT; Cluster Two (2) had the most severe COPD patients; Cluster Three (3) represented obese males with sleep apnea and less severe COPD; and Cluster Four (4) represented older and comorbid patients with heart failure. These four groups demonstrated a completely different mortality pattern in the long-term follow-up, with the very severe COPD group and the older comorbid patients (Clusters 2 and 4) showing the worst survival.

Studies showing heterogeneity in the prescription of HNIV and in the profile of patients in our country19 and even in Europe,20–21 Canada and the United States22 have been published for over 20 years, independently related to the characteristics of the health care system or the economic situation in each country. Additionally, the number of patients receiving ventilatory support at home is growing,23–24 even though the evidence base for this intervention and its impact is still patchy. This substantial growth is even more evident in patients with obesity hypoventilation syndrome (OHS) and chronic obstructive pulmonary disease.25

Obesity-related disease and COPD are the two most common indications for HNIV in most surveys.26–27 A long follow-up in a HNIV cohort26 showed differential mortality patterns regarding the initial indication for this technique. In this study, patients with OHS had better survival, whereas the group with pulmonary disease (mostly represented by COPD) and motor neuron disease patients showed the worst survival. This goes in line with our results: Clusters (1) and (3) were obese males with a higher prevalence of sleep apnea that showed significantly better survival. These two groups represent two distinct phenotypes of obesity hypoventilation that coexist with COPD, and these overlapping patients form an additional phenotype in which hypoxemia and hypercapnia predominate. In a case series, the use of HNIV in these types of patients was shown to improve survival,25 and obesity is a marker of better survival in patients with COPD under HNIV.26,28 This fact, in addition to the fact that 25% of these patients were not using LTOT, seriously question whether the real indication for HNIV in these patients is obesity-related disease and not COPD.

Miravitlles et al.29 suggested that the COPD prevalence in Spain in adults aged 40–80 years was approximately 10%. According to these data, more than 400,000 people have COPD in Catalonia. Vela et al.30 estimated that 9% of these patients with COPD (more than 38,000 patients) would be in the high-risk group due to the severity of the disease, multimorbidity and social conditions. With these data, Cluster (2) (very severe COPD patients) represents a very small percentage of patients with HNIV in Catalonia (<3%) and an insignificant percentage of all high-risk COPD patients. In our study, Cluster (2) showed survival outcomes similar to those reported in several RCTs of HNIV in these patients.8,31 In addition, patients in Cluster (2) had worse lung function, the highest proportion of patients using LTOT, and the highest admissions to the ICU: all factors known to have a high impact on the survival of patients with COPD. The other cluster in our cohort with worse survival was Cluster (4), and it is totally plausible because factors such as age,26,32 comorbidities26 (including cardiovascular and heart failure) and even the use of cardiovascular agents33 have demonstrated worse survival in patients using HNIV.

This study suggests some key elements in the analysis of HNIV. Health care systems have increasingly more data available, and it makes sense (and would certainly be ethical) to use these data to understand the real world. With the available data, it is possible to study the variability in more depth and to attempt to understand the causes. With real-world data, the magnitude of the problems can be measured more accurately. Long-term ventilation of patients with severe COPD is a relatively minor problem in HNIV as a whole.

As Caneiras et al.34 showed, clinical and health resource consumption data are not enough to understand the impact of HNIV. We need to combine this information with patient experience analysis using a qualitative-quantitative approach.35

“Labels” linked to physiopathological diagnoses are not useful. The idea of clusters (groups of patients with common characteristics and needs) may be more useful for the management of patients with complex chronic diseases and many comorbidities. In this context, clusters make it possible to differentiate the cause (COPD as a cause of ventilation) from COPD as a comorbidity (which may contribute to the severity of a clinical situation). In addition, comparisons between hospitals and between territories to study variability are easier if clusters are used.

This study has several limitations. First, this was a retrospective study of COPD patients on HNIV in Catalonia. In this line, exacerbations and hospitalizations due to COPD were not recorded prospectively and were only available in the previous two years (2016–2018). Although HNIV compliance was not recorded, acceptable compliance (at least 4h) is required in daily clinical practice to continue with this treatment. However, this was a multicenter study including all COPD patients on HNIV in Catalonia. Moreover, this study included large amount of data that included reliable administrative information from the AQuAS and, along with the long-term follow-up, makes our results representative and generalizable. Further research is needed to validate the clusters identified in this study due to the small sample size.

In conclusion, in a well-characterized cohort of severe COPD patients with a long follow-up, high heterogeneity in the prescription of HNIV was demonstrated. Cluster analysis allowed us to identify different COPD patient survival groups, including an overlapping phenotype of obesity-related disease and sleep apnea that showed better survival.

Supported by a grant from Ramon Pla Armengol Foundation and Barcelona Respiratory Network (BRN).

No conflicts exist for any of the authors.