Chronic obstructive pulmonary disease (COPD) is associated with several comorbidities contributing to different phenotypes, with related mortality rates and reduced physical activity.1,3

Pulmonary rehabilitation programs (PRPs) are main components of the comprehensive management of these patients. After a PRP, patients, either in stable state or following an acute exacerbation of COPD (AECOPD), report improvements in their symptoms, health-related quality of life (HRQL), and exercise tolerance.4,5 This is also true when complex comorbidities and different phenotypes are present.6–8 However, there is no accepted selection criteria for PRPs for these patients.9

We wondered whether characteristics of patients and results of PRPs reported in randomized controlled trials (RCTs) and meta-analyses could also be observed in real health services. Therefore, the aim of this real-life retrospective observational study was to evaluate the characteristics of COPD patients undergoing an inpatient PRP and to identify the predictors of its success (if any).

The protocol was approved (2164 CE, December 5, 2017) by the Ethical Committee of the Istituti Clinici Scientifici Maugeri IRCCS, Pavia, Italy. All patients at admission to the Scientific Institute of Montescano gave informed consent to the scientific use of their data.

Out of 3311 patients admitted during the study period, there were 975 (29.4%) patients with COPD undergoing the inpatient PRP. Of these, 142 (14.6%) were transferred from acute care hospitals (group A) and 833 (85.4%) were admitted from home (group B) (Table 1). Groups 1–3 included 30.7, 51.7, and 17.6% of patients, respectively (Table 2).

Demographic, anthropometric, physiological, and clinical characteristics of patients according to provenance are described in Table 1. Patients of group A had longer LoS, lower BMI, suffered from more comorbidities as assessed by CI and SI, respectively, and lower levels of PaO2.

The LoS were different across groups of diagnosis (group 1: 26.7 (13.3), group 2: 28.7 (12.8), group 3: 24.3 (10.1) days; p=0.0002), with a statistically significant difference between group 2 and group 3 (p=0.001).

Characteristics of patients according to the recorded associated main diagnosis as well as their distribution across BMI, the GOLD classes, and levels of arterial blood gases are described in Table 2. As expected, obese patients were significantly more represented in group 3, whereas only about 10% of patients of both groups 1 and 2 were underweight. Data of lung function and arterial blood gases were available for 769 and 701 patients, respectively. Patients of group 2 were significantly older and more represented in GOLD stages III and IV, showed lower mean levels of PaO2 and higher mean levels of PaCO2 than the others.

Mean biochemical data of patients at admission are shown in Table 3.

Distribution among groups of associated diagnosis of reported components of CIRS and assessed comorbidities are shown in Table 4. Patients of group 1 suffered from less CI and SI than the other groups. Heart (valvular disease and chronic heart failure) and cardiovascular diseases had a similar prevalence in patients of all groups. Hypertension was present in more than half the patients of all groups; it was more prevalent in patients of groups 2 and 3 than in patients of group 1. The overall endocrine diseases reported, including diabetes, occurred in more than 60% of patients of all groups, and as expected occurred in the vast majority of patients of group 3. As also shown in Table 4, hypercholesterolemia and hypertriglyceridemia as assessed at admission was common across all groups, whereas hyperglycemia had a significantly greater prevalence in group 3. Heart (p=0.007) and cardiovascular (p=0.0003), as well as kidney (p=0.0001) diseases occurred more significantly among patients transferred from hospitals (group A), whereas obesity was significantly more common among patients in group B (p=0.0001).

The overall hospital mortality rate was 1.3% (5.6% vs 0.6%, in groups A and B, respectively, p=0.0001; 0.7, 2.2, and 0.0% in groups 1, 2, and 3, respectively, p=0.048), mainly due to patients with respiratory failure transferred from acute care hospitals after an AECOPD. These patients also accounted for most discharges to acute care hospitals (6.5% of all patients; 15.5% vs 4.9%, in groups A and B, respectively, p=0.0001; 4.7, 9.1, and 1.7% in groups 1, 2, and 3, respectively, p=0.001). AECOPD was the main cause of death (13 patients), or of transfer to acute care hospitals (63 patients) together with other complications (unavailable data). No death or transfer to acute care hospitals was due to the activities of PRP nor was any adverse effect observed during sessions.

Patients who died were older (79.8±6.1 vs 71.0±10.5 years, p=0.003), suffered from more severe comorbidities as assessed by means of the SI (2.0±0.3 vs 1.8±0.3, p=0.020), and at admission showed lower levels of PaO2 (51.7±12.7 vs 68.7±9.9mmHg, p=0.0001) and higher levels of PaCO2 (53.9±38.1 vs 38.8±5.6mmHg, p=0.0001) than patients discharged alive.

Besides death or transfer to acute care hospitals, from the available data dropouts from PRPs were due to voluntary withdrawal (23 patients).

Values of 6MWT before and after the PRPs were available only for 424 patients. Missing data were mainly due to lack of inclusion into database, and to a lesser extent to dropouts from PRPs or impossibility of performing the 6MWT at discharge. Changes in 6MWT after the PRPs are shown in Tables 5a and 5b. At admission, patients transferred from acute care hospitals showed reduced 6MWT as compared to patients admitted from home. Patients in both groups improved 6MWT after PRP but patients transferred from acute care hospitals showed a greater improvement in 6MWT than patients of group B (Table 5a). Patients of group 2 showed significantly lower baseline values of 6MWT than the other groups. However, patients of all groups significantly improved their exercise tolerance (Table 5b). The patients who improved their 6MWT by more than the MCID were 48.4% of all patients without any significant difference between groups: 45.5, 50.3, and 48.5% in groups 1, 2, and respectively, p=0.76. There was no difference in success of PRPs between male and female patients.

Out of all considered potential predictors, multiple regression analysis showed a significant association with baseline 6MWT only for age, male gender, serum Hb, FEV1/FVC, FVC, % predicted, and PaO2 but not for BMI, CI, SI, serum glucose, and PaCO2. All significant associations were positive except for age (inverse association). Detailed results from multiple regression analysis are reported in Table 6a.

Finally, the results of logistic regression analysis modeling the probability of an improvement in the 6MWT greater than the MCID using the same predictors as above and including baseline 6MWT are given in Table 6b. The only significant predictors were age, gender, FEV1/FVC (borderline), and baseline 6MWT. In particular, for a 1 year increase in age and 1 meter increase in baseline 6MWT, there were respectively an 8 and 1.8% increase in the probability of an improvement in the 6MWT greater than the MCID after PRP.

This real-life study shows the characteristics of patients with COPD undergoing an inpatient PRP resulting in significant improvement in exercise capacity independent of provenance or associated main diagnosis, confirming in a “real-life” setting the results of RCTs and meta-analyses.

The demographic characteristics of an OSAS occurrence in our patients are those commonly reported for COPD patients undergoing PRPs.7,8,22,23 Underweight was reported only in 10% of patients independent of associated respiratory failure. In agreement with Greening et al.,24 in our patients BMI did not influence their baseline 6MWT or their response to PRP. Obesity is observed in 17% of patients with AECOPD25 and, like our study, it has no effect on the percentage of patients achieving the MCID of 6MWT after PRP.26,27

More than a third of patients admitted to our PRP suffered from severe airway obstruction as assessed by distribution of patients in GOLD stages III or IV (Table 2). In our study, neither the associated diagnosis of respiratory failure, lower PaO2, or higher PaCO2 levels negatively influenced the success of PRP, confirmation for a larger number of patients than other previous reports.28 Therefore, the results of our study support the guideline suggestions to perform PRPs in almost all COPD stages.1,4,29

Among COPD patients, type 2 diabetes mellitus and hypertriglyceridemia showed a prevalence between 16.8–28.5 and 19.7%, respectively.30 In the study by Griffo et al.,7 hypercholesterolemia was less prevalent (22.4%), whereas the occurrence of diabetes mellitus (17.4%) was similar to the reported prevalence of diabetes mellitus in our study.7

Hypertension is observed in 28.5–64.7%7,30 of COPD patients and occurred in more than 60% of our patients. Reported heart and cardiovascular diseases occurred in about half of our patients. Heart failure is reported between 12.3 and 28.3%,30 whereas in study by Griffo et al.7 chronic heart failure and cardiovascular diseases showed 11.9 and 51.9% prevalences, respectively.

In patients with COPD, the number of comorbidities can predict mortality31 and comorbidity-attributable costs were twice those directly related to COPD.32 A systematic review of literature suggested that comorbidities might also have a negative influence on PRP outcomes.33 Our study confirms in a large COPD population the results of other studies6,7,34 that comorbidities do not preclude patients with COPD from obtaining improvements in exercise capacity. Our study confirms also the findings by Crisafulli et al.6 that the number of comorbidities is not predictive of PRP outcomes.

Confirming other studies,35 we did not observe any deaths or the need to transfer patients to acute care hospitals due to PRP nor was any adverse effect observed during PRP activities. Our PRP also improved exercise tolerance in patients recovering from an AECOPD. Meta-analyses give conflicting results on the safety and the clinical benefits of PRP after an AECOPD.36,37 There is evidence that aerobic training is safe even when conducted after 72h of hospitalization for AECOPD,38 however a RCT39 reported that early rehabilitation during hospital admissions for chronic respiratory diseases did not reduce the risk of subsequent readmissions and was associated to higher 12-month mortality.

Finally, our study confirms that baseline exercise capacity and severity of airway obstruction but not comorbidities are predictors of success of PRPs in exercise tolerance as assessed by reaching the MCID of 6MWT.6,40 However, an earlier study showed that baseline 6MWT was the only variable that correlated and significantly predicted also the post-PRP changes (delta) in exercise capacity in these patients.8

Our study has several limitations. Like all retrospective studies, it suffers from missing data such as in 6MWT or PRP adherence rate and importantly, it reports only one direct outcome of PRPs, the 6MWT, without any information on HRQL, or symptoms. On the other hand, retrospective analyses of large populations describe the real life of health services. The reported comorbidity occurrence must not be considered as ‘prevalence’, because in our study, patients did not undergo specific diagnostic tests. However, our comorbidity occurrence was comparable to the prevalence observed in studies in which the comorbidities were systematically investigated.2

In conclusion, the inpatient PRP resulted in significant improvement in patient's exercise capacity independent of whether in stable state or recovering from a recent AECOPD or of associated respiratory failure or OSAS, confirming that characteristics of patients and results of PRP reported in RCTs and meta-analyses can be observed also in a routine setting.

None.

None of the authors declare any conflict of interest with this paper.