Introduction

Prospective randomised controlled studies have shown that the addition of non-invasive positive pressure ventilation (NIPPV) to standard treatment reduces the need for endotracheal intubation, lowers hospital mortality and shortens length of stay in selected patients with chronic obstructive pulmonary disease (COPD) who have acute hypercapnic respiratory failure (AHRF) [1, 2, 3, 4, 5]. Nevertheless, NIPPV is not successful in all cases, with a reported failure rate of 7–50% [6]. Therefore it is important to be able to identify the patients in whom NIPPV is more likely to fail, especially if it is implemented in a respiratory ward. A COPD patient with a high likelihood of failing and for whom invasive ventilation would be appropriate is best managed in the intensive care unit (ICU) where intubation is readily available [6]. Several clinical and physiological respiratory parameters have been shown to predict NIPPV failure [7, 8, 9].

Patients with severe pulmonary dysfunction often suffer from coexistent pathologies and are likely to develop extra-pulmonary complications [10, 11, 12, 13]. A recent Dutch survey reported that 73% of stable COPD patients [14] have at least one chronic non-respiratory co-morbidity (C-NRC) and this is an independent predictor of hospitalisation for exacerbations [15]. C-NRC, usually cardiovascular, were found in 42–97% of patients with COPD exacerbation requiring hospitalisation [13, 16, 17, 18]. C-NRC impacts on short [10, 13, 16] and long-term survival [11, 18] of COPD patients hospitalised with acute decompensation. Furthermore, the development of acute non-respiratory co-morbidity (A-NRC) complicating AHRF may adversely affect hospital [10, 12, 19, 20] and 6-month [19] outcome of severe COPD exacerbations. The effect of non-respiratory co-morbidities upon short and long-term outcome with NIPPV for COPD exacerbation has not been assessed in previous studies [7, 8, 9].

We performed this study [21] in COPD patients with AHRF needing NIPPV to (a) determine the prevalence and the type of C-NRC/A-NRC, defined according to strict criteria, and (b) clarify whether hospital and 6-month outcome are independently affected by C-NRC and/or A-NRC.

Material and methods

Patients

The study protocol was approved by the local ethics committee of our hospital and was performed in accordance with the ethical standards of the 1964 Declaration of Helsinki. Informed consent was obtained in all cases from either the patient or next kin. All 625 patients admitted to our 25-bed respiratory ward in Arezzo between January 1996 and January 2001 for acute decompensation of COPD were considered for the study. Medical treatment consisted of controlled oxygen therapy; salbutamol and ipratropium bromide via aerosol every 4 h; intravenous aminophylline, corticosteroids, antibiotics; cardiovascular drugs and therapy for coexistent C-NRC if necessary. Patients were enrolled in the study if despite this treatment they fulfilled all the following criteria for NIPPV: (a) pH lower than 7.33 or PaCO2 higher than 55 mmHg; (b) PaO2/FIO2 ratio lower than 250; (c) dyspnoea at rest with respiratory rate higher than 25 breaths/min; and (d) use of accessory respiratory muscles or paradoxical abdominal breathing. Exclusion criteria were: (a) refusal of NIPPV; (b) facial deformity sufficient to affect mask fitting; (c) severe encephalopathy unrelated to hypoxaemia and/or hypercapnia; (d) overt gastro-intestinal bleeding; (e) upper airway obstruction; (f) acute ischaemic heart disease; and (g) need for urgent intubation due to cardiac or respiratory arrest, prolonged respiratory pauses and psychomotor agitation requiring sedation [3]. Most of the 625 admitted patients (n=466, 75%) improved quickly with medical therapy. Among the remaining 159 eligible patients with AHRF, 26 (16.4%) were excluded because of the need for urgent intubation and 13 (8.2%) because of refusal of NIPPV. The remaining 120 patients were enrolled in the study 6.25 h (2–47.75) after admission to our ward. Tables 1 and 2 show, respectively, the “historic” and “acute” features of the population.

Table 1 Demographic, clinical and functional data when stable (n=120) (IQR interquartile range, FEV 1 forced expiratory volume in 1 s, BMI body mass index, ADL Activities of Daily Living, LTOT long-term oxygen therapy, HMV home mechanical ventilation)
Full size table
Table 2 Clinical and physiological data during the episode of AHRF (n=120) (IQR interquartile range, NIPPV non-invasive positive pressure ventilation, AHRF acute hypercapnic respiratory failure, APACHE Acute Physiology and Chronic Health Evaluation, CPE cardiogenic pulmonary oedema)
Full size table

Non-invasive ventilation

All patients entered in the study were managed in the three-bed respiratory monitoring unit in our respiratory ward [22]. All nurses had attended a 10-h training course on NIPPV in the preceding 2 years; they were kept up to date by means of a yearly 4-h course. Mean nurse:patient ratio was 1:9. During the earliest phase of NIPPV one pulmonologist was continuously present at the patient’s bedside; later the patient was closely looked after by one nurse with one pulmonologist available on call. Electrocardiography, transcutaneous oxygen saturation and non-invasive blood pressure were monitored continuously. Intubation and transfer to ICU was promptly available in the case of NIPPV failure.

NIPPV was delivered by means of Bi-PAP Vision in S/T mode via a well fitting facial mask (Respironics, Murrysville, Pa., USA). A plateau exhalation valve was applied to the circuit to minimise CO2 rebreathing [23]. Inspiratory positive airway pressure (IPAP) was initially set at 10 cmH2O and then gradually increased to a maximum of 25 cmH2O depending on clinical and blood gas response and on the patient’s tolerance. Expiratory positive airway pressure was always set at 5 cmH2O and back-up respiratory rate at 16 breaths/min. FIO2 was adjusted to maintain oxygen saturation at 90% or higher. Patients were encouraged to use NIPPV for as long as tolerated, aiming for at least 6–8 h per day split in three periods; NIPPV was discontinued when all the following criteria were maintained for longer than 24 h while breathing oxygen: pH higher than 7.35, PaO2/FIO2 ratio above 25 and respiratory rate below 25 breaths/min.

Criteria for endotracheal intubation

NIPPV was considered to have failed if at least one of the following criteria for intubation [3] was satisfied: (a) cardiac arrest; (b) respiratory arrest or gasping or prolonged respiratory pauses; (c) psychomotor agitation requiring sedation; (d) severe difficulty clearing bronchial secretions; and (e) worsening of gas exchange (at least two of the following: pH less than 7.30 and less than 0.02 with respect to baseline, PaO2/FIO2 ratio less than 150 and less than 20 with respect to baseline; PaCO2 greater than 20% with respect to baseline) and/or level of consciousness (Kelly and Matthay [24] score above 1 with respect to baseline) during NIPPV for at least 6 h with the best ventilator setting. In the case of refusal of invasive ventilation by the patient and/or relatives family NIPPV failure was considered when the a priori criteria for intubation were met.

Measurements

The following demographic, clinical and functional data were collected when the patient was stable (within 6 months from admission): (a) age and sex; (b) forced expiratory volume in 1 s (FEV1) and arterial blood gases while breathing room air; (c) body mass index; (d) chronic cor pulmonale [18]; (e) functional limitation evaluated with Activities of Daily Living (ADL) score [19]; (f) use of long-term oxygen therapy (LTOT) or home mechanical ventilation (HMV); and (g) prevalence, type (cardiovascular and non) and score of C-NRC [25] (Appendix A). Clinical and functional data collected at admission were: (a) causes of AHRF: community-acquired pneumonia, cardiogenic pulmonary oedema, pulmonary embolism, COPD exacerbation [12]; (b) acute severity of illness assessed by Acute Physiology and Chronic Health Evaluation (APACHE III) score [26]; (c) level of consciousness measured with Kelly-Matthay score; and (d) prevalence and type (cardiovascular and non-cardiovascular) of A-NRC [12] (Appendix B). NIPPV variables were: (a) arterial blood gases while breathing O2 and after a 2-h trial of NIPPV; (b) level of IPAP; and (c) tolerance assessed by a simplified score (1=low, 2=medium, 3=good) [12]. Data collected at 6-month follow-up included: (a) number of readmissions to hospital for further COPD exacerbations and whether mechanical ventilation needed; and (b) mortality calculated from the date of enrolment in the study. Follow-up included a specialist visit with blood gases within 3 weeks after discharge; thereafter all patients were seen every 2 months, unless they deteriorated earlier.

Statistical analysis

Continuous data are expressed as mean (standard deviation) or as median (interquartile range, IQR) if distributed, respectively, normally or not, according to Kolmogorov-Smirnov test (normal distribution: p>0.05); categorical data are presented as percentages. Univariate analysis was first performed for each variable comparing patients successfully treated with NIPPV and those who were not and comparing between subjects who survived and those who had died at 6 months. Continuous variables were analyzed with the two-tailed unpaired t test (parametric data) or Mann-Whitney U test (non-parametric data). Categorical data were compared using the χ2 test or, when appropriate, Fisher’s exact test. A p value less than 0.05 was considered statistically significant.

Due to the limited number of variables that may be entered in the multivariate analysis (one for every ten events), stringent selection criteria were employed [27]. Variables with p value less than 0.01 on univariate analysis were considered eligible for multivariate analysis. To avoid multicollinearity we examined correlations between all eligible variables using Spearman’s test (significance: p<0.01), and those which were likely to be dependent on the others were excluded. Then all selected variables were entered into a stepwise forward logistic regression model (entry criteria: p<0.05); the variables with p value less than 0.05 were considered statistically significant, and their predictive value was reported as the adjusted odds ratio with 95% confidence interval. The goodness-of-fit of the model was assessed by the Hosmer-Lemeshow test [28] (significance: p>0.05). Analyses were performed using version 10.0 of the SPSS statistical software package (SPSS, Chicago Ill., USA).

Results

NIPPV was delivered for a total of 40 h (IQR 22–65.5 h; 1st day 7.5 h, 6–11) over 5.5 days (3–9). NIPPV was well tolerated (compliance score 2, IQR 2–3). IPAP applied was 16 cmH2O (IQR 14–20). NIPPV failed in 22 of 120 cases (18%) 2.5 days (IQR 1–8) after enrolment for: worsening of blood gases (after 6 h of NIPPV: pH 7.24±0.05, PaO2/FIO2 ratio 148±40, PaCO2 98.4±12.2) and/or level of consciousness in 16 (72.7%), psychomotor agitation in 6 (27.3%), secretions retention in 2 (9.1%), cardiac arrest in 2 (9.1%) and gasping in 1 (4.5%); more than one condition caused NIPPV failure in five patients. Only seven subjects were intubated as the others refused intubation and died. Four intubated patients died after 31.5 days (IQR 24.5–55.8) of ICU stay for septic shock (n=2), multiple organ dysfunction syndrome (n=1) or cardiac arrest (n=1). Only three intubated patients were discharged alive (Fig. 1). Overall in-hospital mortality was 15.8% (19/120); length of stay in hospital was 15 days (IQR 11–21). At discharge LTOT and HMV were prescribed in 20 and 8 patients, respectively.

Fig. 1
figure 1

Distribution of the study population according to NIPPV requirement, NIPPV hospital and 6-month prognosis

Full size image

Prevalence of co-morbidity

The prevalence of C-NRC and A-NRC was 20% (24/120) and 41.7% (50/120), respectively. Cardiovascular disease was the commonest co-morbidity. Cardiogenic pulmonary oedema was not counted as A-NRC if it was a cause of AHRF. Non-cardiovascular disorders were more numerous than cardiovascular among C-NRC; the contrary was found regarding A-NRC (Table 3).

Table 3 Distribution of chronic and acute non-respiratory co-morbidities (NRC) (n=120)
Full size table

In-hospital NIPPV outcome

In-hospital NIPPV failure was greater in patients with than in those without C-NRC (33.3% vs. 14.6%, p=0.043) or A-NRC (32% vs. 8.6%, p=0.001). The higher the number of A-NRC the greater was the chance of NIPPV failure (A-NRC=0:8.6%; A-NRC=1:22.9%; A-NRC>1:53.3%, p<0.001). The presence of cardiovascular A-NRC was significantly associated with NIPPV failure (32.4% vs. 12.8%, p=0.018). Five variables associated with NIPPV failure on univariate analysis (p<0.01) were eligible for entry in the multivariate analysis. Due to the significant correlations found between ADL score and FEV1 (r=0.337, p<0.001), A-NRC and APACHE III (r=0.320, p<0.001), A-NRC and Kelly-Matthay score (r=0.404, p<0.001), FEV1 and the presence of A-NRC were selected to be entered in the multivariable model and were confirmed to be independent predictors of in-hospital outcome (Hosmer-Lemeshow test: χ2=6.449, p=0.597; Table 4). The same result was obtained when the five variables selected in univariate analysis were introduced in the multivariate analysis. If NIPPV failure was expressed as need of intubation and death instead of only need for intubation, short-term outcome was slightly better (19 failures), but the variables selected by univariate and multivariate analysis were the same.

Table 4 Variables associated with in-hospital NIPPV failure (n=120) (OR odds ratio, CI confidence interval, NIPPV non-invasive positive pressure ventilation, ADL Activities of Daily Living, APACHE Acute Physiology and Chronic Health Evaluation, NRC non-respiratory co-morbidities, AHRF acute hypercapnic respiratory failure, FEV 1 forced expiratory volume in 1 s, CPE cardiogenic pulmonary oedema)
Full size table

Longer-term NIPPV outcome

All but one of the recruited patients were followed up at 6 months. Of the 100 patients discharged alive from the hospital 49 were then re-admitted for 80 further COPD decompensations, with a mortality of 17.5% (14/80). NIPPV was required in 51 of 80 readmissions (63.8%), with a failure rate of 21.6% (11/51). Intubation was performed in 13 of 80 readmissions (in 6 cases after NIPPV failure). Nine patients died at home: 6 for COPD decompensations, 3 suddenly. Overall 6-month mortality was 35.3% (42/119; Fig. 1). In the subgroup of 51 COPD decompensations subjected to repeat NIPPV after discharge, prevalence of C-NRC and A-NRC was respectively 12% and 32%; a non-significant trend towards a higher rate of NIPPV failure was found in patients with than without A-NRC (55.6% vs. 24%, p=0.083).

In the overall studied population the presence of C-NRC was associated with greater 6-month mortality (54.2% vs. 30.5%, p=0.030); a non-significant trend towards a worse longer-term outcome was observed in patients with than without A-NRC (44.9% vs. 28.6%, p=0.067). Patients with more than one A-NRC had a higher 6-month mortality rate than patients with A-NRC≤1 (66.7% vs. 30.8%, p=0.007). The impact of NIPPV outcome upon longer-term survival was not tested as the high proportion of patients died after NIPPV failure without receiving intubation (15/22) may have introduced a bias.

Four variables significantly associated with 6-month outcome on univariate analysis (p<0.01) were eligible to be entered in the multivariate analysis. Despite the significant association of all C-NRC with longer-term survival in univariate analysis only the presence of non-cardiovascular C-NRC reached the statistical cut-off to be selected for a multivariate analysis (71.4% vs. 30.5%, p=0.005). Due to the correlation between ADL score and HMV (r=−0.291, p=0.001) the presence of more than one A-NRC and of non-cardiovascular C-NRC together with ADL score were selected to be entered in the multivariate analysis and were confirmed as independent predictors of 6-month prognosis (Hosmer-Lemeshow test: χ2=2.366, p=0.669; Table 5). The same result was obtained when the four variables selected in univariate analysis were introduced in the multivariate analysis.

Table 5 Variables associated with 6-month mortality (n=119) (OR odds ratio, CI confidence interval APACHE Acute Physiology and Chronic Health Evaluation, ADL Activities of Daily Living, BMI body mass index, NRC non-respiratory co-morbidities, FEV 1 forced expiratory volume in 1s, HMV home mechanical ventilation)
Full size table

Discussion

This is the first study specifically designed to evaluate the prevalence and the impact of both chronic and acute non-respiratory co-morbidities on the prognosis of COPD patients with AHRF undergoing to NIPPV. We found that C-NRC and A-NRC are common in cases of COPD decompensation receiving NIPPV in a respiratory monitoring unit, non-cardiovascular being more common among C-NRC and cardiovascular among A-NRC. In-hospital NIPPV failure and 6-month mortality rates were higher in patients with than in those without co-morbidities; the former were more influenced by cardiovascular A-NRC the latter by non-cardiovascular C-NRC. The presence of A-NRC was an independent predictor of in-hospital NIPPV failure (together with FEV1), while the presence of more than one A-NRC and of non-cardiovascular C-NRC (together with ADL score) independently predicted 6-month mortality.

Prevalence of co-morbidity

The lower prevalence of chronic co-morbidities (20%) found in our population than in other reports in COPD exacerbations (42–97%) [13, 16, 17, 18] could be explained by the restricted criteria used for C-NRC according to the Charlson et al. [25] index. If we consider all the disorders selected by Charlson et al. including those in which the index was lower than 1, the prevalence of C-NRC in our study (64.2%) did not differ from the others. In common with other published data [10, 11, 13, 16, 17], cardiovascular diseases was the most single frequent disorder. We found at least one A-NRC in 41.7% of patients; Confalonieri et al. [13] reported a higher prevalence (62.5%) in a heterogeneous population (COPD: 59.6%), 63% of whom underwent NIPPV. In our experience, cardiovascular dysfunction was the most common complication, as also reported by other authors [3, 12].

In-hospital NIPPV outcome

The rate of NIPPV failure (18%) and of hospital mortality (15.8%) in our study was similar to other published data [6]. The chance of NIPPV failure increased from 14.6% and 8.6% in patients without to 33.3% and 32% in patients with, respectively, C-NRC and A-NRC. Few authors have assessed the effect of chronic and acute co-morbidities on the outcome of COPD with AHRF and their results are difficult to compare with our data because of the differences in the following: criteria used for defining co-morbidities, setting, severity and causes of AHRF and indications for and technique of mechanical ventilation. In series of COPD requiring conventional mechanical ventilation for an episode of AHRF in ICU chronic left heart failure [29], APACHE II correlated C-NRC and active malignancies [16] were found to be independently associated with hospital mortality. In a study conducted in a respiratory ward [10] a previous myocardial infarction was the only C-NRC significantly associated with in-hospital mortality on univariate analysis; atrial fibrillation and ventricular arrhythmias on admission were independent predictors of hospital mortality on multivariate analysis, together with an alveolar-arterial difference in PO2 greater than 41 mmHg and age. Similarly, we found that the presence of cardiovascular A-NRC was significantly correlated with NIPPV failure. In a study conducted in two respiratory high-dependency units on 137 COPD patients with AHRF [12], the presence on admission of at least one A-NRC together with a lower pH and ADL score were the strongest predictors of late failure after an initial positive response to NIPPV.

Longer-term NIPPV outcome

The 6-month mortality of 35.3% in our series was slightly higher than that reported (up to 30%) by other investigators of COPD non-invasively ventilated for AHRF and is likely to be explained by differences in the severity of illness [9, 30, 31, 32]. The proportion of deaths at 6 months increased from 30.5% and 30.8% in patients without to 54.2% and 66.7% in patient with, respectively, C-NRC and more than one A-NRC. In the largest series of 1,016 COPD patients hospitalised for acute exacerbation and managed without ventilatory support in 65.7% of cases the authors found no clear association between the number of APACHE III correlated C-NRC and 6-month mortality [18]. Conversely, a significant effect of C-NRC on long-term outcome was observed by Antonelli Incalzi et al. [11] in their retrospective study of 270 COPD exacerbations (16% requiring mechanical ventilation); chronic renal failure was the strongest independent predictor of mortality, together with cor pulmonale, FEV1 less than 590 ml, ischaemic heart disease and age. Similarly, we found that the presence of non-cardiovascular C-NRC was independently correlated with longer-term outcome.

Important limitations in our study must be underlined. Firstly, regarding C-NRC we used the Charslon et al. [25] index, which has prognostic implication in geriatric out-patients with chronic diseases but was not calibrated for acute COPD decompensations. Moreover, we evaluated prevalence and prognostic effect of each A-NRC independently of its potentially different weight. However, to our knowledge, an index taking in account both the number and the seriousness of each acute disorder is lacking. Secondly, the limited size of our series may have underestimated the effect of co-morbidities taken singularly; however, we demonstrated the significant impact of cardiovascular A-NRC and non-cardiovascular C-NRC upon short and longer-term outcome, respectively. Thirdly, our models based on the stepwise logistic regression need to be tested prospectively in a similar population of COPD with AHRF. However, the significant Hosmer-Lemeshow goodness-of-fit value indicates no large difference between observed and expected rates of NIPPV failure and longer-term mortality.

In conclusion, this is the first study designed specifically to evaluate the effect of non-respiratory co-morbidities upon short and longer-term outcome with NIPPV for acute COPD exacerbation. We demonstrated that co-morbidities, especially cardiovascular, are common in COPD with AHRF needing NIPPV and are likely to affect hospital and 6-month outcome. When selecting COPD patients for acute NIPPV, physicians should consider not just lung function and respiratory severity scores but also non-respiratory variables.