Chronic obstructive pulmonary disease (COPD) is one of the leading causes of morbidity and mortality in the industrialized world and the only leading cause of death with a rising mortality rate [1,2].
Patients with COPD are usually elderly with multiple co-existing diseases and severe limitation of function, and hospitalization with acute respiratory failure secondary to COPD requiring invasive mechanical ventilation (IMV) has a reported in-hospital mortality of up to 50% [3-16].
Despite the large number of patients with COPD and the poor short-term outcome associated with severe acute respiratory failure secondary to COPD, little is known about long-term outcomes among these patients [4,6,11,17]. In addition, virtually no data exist on simple, readily available prognostic factors for long-term mortality of patients with COPD treated with IMV. A few studies have examined the association between arterial blood gas (ABG) values and mortality of COPD patients treated with IMV [4,6,8,10], but these studies are generally inconclusive or contradictive in their findings and they are difficult to interpret because of small sample sizes and substantial heterogeneity of the studied cohorts. It has been increasingly recognized that comorbidity is associated with increased mortality among patients with chronic lung diseases [18,19], but limited data exist on how the burden of comorbidity affects long-term outcome of mechanically ventilated COPD patients [4,17].
We therefore conducted a population-based cohort study to examine: (1) 90-day and 1-yr mortality, and (2) the association between ABG values and burden of comorbidities and 90-day and 1-yr mortality of COPD patients treated with IMV.
This population-based cohort study was conducted from 1 January 1994 through 31 December 2004 among patients admitted to the intensive care unit (ICU) at Silkeborg Hospital, Aarhus County, Denmark. The hospital serves a population of about 100 000 inhabitants and has approximately 16 000 admissions annually, 500 of which leads to transferal to the seven-bed mixed surgical/medical ICU. The Danish National Health Service provides tax-supported healthcare for all inhabitants, guaranteeing free access to general practitioners and hospitals; therefore, in practise, all treatment of critically ill patients in Denmark takes place at public hospitals.
Since 1968, every Danish citizen has received a unique civil registration number, which codes for age, gender and date of birth. This number is included in all public Danish registries and permits accurate record linkage between these .
All patients admitted to the ICU at Silkeborg Hospital are recorded in a prospective clinical database. The database collects information on cause of admission (e.g. respiratory failure), main categories of diseases (e.g. respiratory) and a number of treatment variables including IMV. All data collection is done during the ICU admission by the physicians and nurses treating the patient.
From the database, we retrieved data on all patients admitted because of ARF or categorized with acute or chronic pulmonary diseases. By reviewing medical records of these patients, we identified all COPD patients, and data on their first episode of IMV was obtained. All data collection was done by two physicians (LR and SC) who were blinded with regard to study outcomes.
COPD was defined by specific changes in premorbid pulmonary test if a test was performed within 1 yr prior to ICU admission . As the majority of patients did not have a relevant pulmonary test performed within 1 yr prior to ICU admission, we also had to rely on characteristic findings on history and physical examination by the attending physician at time of ICU admission, to identify COPD patients as it is often done in studies on COPD [10,21].
The decision to initiate IMV was taken by the ICU physicians treating the patient and based on the departments' guidelines. Similarly, assessments of weaning readiness, weaning technique and extubation decisions were made at the discretion of the ICU team and based on the department guidelines.
Arterial blood gas specimens
By review of medical records, we obtained data from the last ABG specimen obtained before IMV was initiated. Only ABG measurements obtained within 6 h prior to initiation of IMV were considered relevant. All blood gas analyses were done using Radiometer ABL 3, 520 or 725 analysers (Radiometer, Copenhagen, Denmark).
Two levels of ABG values were defined (pH < 7.20, pH > 7.20; PaCO2 < 10.0, PaCO2 > 10.0; PaO2 < 7.0, PaO2 > 7.0; BE < 0.0, BE > 0.0) based on either physiological normal values or medians. We repeated the statistical analyses using three levels of ABG values. As results from these analyses did not change any of our conclusions and because of the superior statistical efficiency of the two category approaches, we present the results based on two levels of ABG values.
We used the Charlson Comorbidity Index to assess the level of comorbidity among study patients . The Charlson Index includes 19 major disease categories, among others cardiovascular diseases, diabetes and malignancies. For calculating the Charlson Index Score, a weight is assigned to each disease category and the score is the sum of these weights. The index has been adapted and validated for use of hospital discharge data for the prediction of short- and long-term mortality . By linkage to the Hospital Discharge Register of Aarhus County, we identified all previous discharge diagnoses of study patients and computed the Charlson Index Score of each patient. The Discharge Register, established in 1977, contains prospectively collected data on patients' civil registration numbers, dates of admission and discharge and up to 20 discharge diagnoses coded by physicians according to the International Classification of Diseases. Three index levels were defined: score of 0, low; score of 1–2, medium; score of 3+, high.
Diagnoses of COPD were excluded from the index as COPD defined our study cohort.
We obtained information on death, emigration and immigration by linking the cohort to the Danish Civil Registration System. This register contains information on vital status (dead, alive), date of death and residence of the entire Danish population since 1 April 1968 .
Based on the date of first ICU admission requiring IMV, we constructed Kaplan-Meier survival curves and obtained estimates of death during the following 90 days and 1 yr for each study patients, respectively. Based on the Kaplan-Meier estimates, we computed estimates of death during the following year for the main study variables – results of ABG specimens, age, gender and comorbidity score.
We used Cox's regression analysis to estimate mortality ratios (MRs) at 90 days and 1 yr after ICU admission according to the levels of PaCO2, PaO2, pH and BE, adjusted for age, gender and comorbidity, and according to the level of Charlson Comorbidity score, adjusted for age and gender.
We also examined the association between ABG values and 90-day mortality with second-degree fractional polynomial regression in order to obtain smooth presentations of the MRs as continuous functions of PaO2, PaCO2, pH and BE levels . The median values of the reference groups of PaO2, PaCO2, pH and BE were used as reference values in these analyses.
All analyses were performed using SAS version 8.2 (SAS Institute Inc, Cary, NC, USA).
Since the present study was based on medical registries and did not involve patient contact, Danish law requires no further research ethics approval.
We identified 244 patients with COPD treated with IMV during the study period. Fourteen patients (6%) were excluded because we were unable to identify ABG specimens measured within 6 h before initiation of IMV leaving 230 patients for further analyses.
The majority of patients were female (124 patients, 54%), 68% were more than 65-yr old and 172 patients (65%) had one or more comorbidities according to the Charlson Index (Table 1).
ABG specimens demonstrated severe hypercapnic respiratory failure with a median pH of 7.24 (inter quartile range (IQR): 7.15–7.34) and median PaCO2 of 9.5 kPa (IQR: 7.7–12.2). Median PaO2 was 7.5 kPa (IQR: 5.7–9.7).
Overall 90-day mortality of COPD patients treated with IMV was 30.8% and 1-yr mortality was 40.5%. The 95% CI of all adjusted 90-day and 1-yr MRs of ABG values included one, indicating that there was no strong statistical association between level of ABG values and 90-day or 1-yr mortality (Table 2). The curves of the graphical presentations of the second-degree fractional polynomial regressions were all smooth and without indications of any well-defined threshold ABG values (Fig. 1).
Ninety-day and 1-yr mortality among patients with a Charlson Index Score of more than 2 was 34.3% and 42.9% compared with 25.9% and 29.3% among patients without comorbidities, respectively (Table 1, Fig. 2). For patients with medium and high levels of comorbidity, adjusted 90-day MR was 1.3 (95% CI: 0.7–2.2) and 1.3 (95% CI: 0.6–2.7), respectively, when compared with patients without comorbidities. Adjusted 1-yr MR was 1.6 (95% CI: 0.9–2.8) among patients with a medium level of comorbidity and 1.4 (95% CI: 0.7–2.9) among patients with a high level of comorbidity when compared with patients without comorbidity.
In this population-based cohort study among 230 COPD patients treated with IMV, we found a 90-day mortality of 30.8% and a 1-yr mortality of 40.5%. Neither the levels of ABG values at time of initiation of IMV nor the levels of comorbidity were strong predictors of 90-day or 1-yr mortality among these patients.
Several issues must be considered when assessing the validity of our findings. The main strengths of the study include the uniformly organized healthcare system allowing a population-based design and the complete long-term follow-up through population-based databases, limiting the risk of referral and information bias. The relatively large sample size compared with previous studies is a further strength [4,6,11,17], but the statistical precision of the MRs was still only modest.
The main problem of breaking continuous variables into categories is that individuals placed at elevated risk at exposure will be merged with lower-risk members of their category . To our knowledge no previous study on ABG levels as predictors of poor outcome of critically ill COPD patients have addressed this issue. We therefore used fractional polynomial regression to examine whether categorization of blood gas results biased our results . The smooth graphical presentations of the fractional polynomial regression curves make it unlikely that this potential bias had major impact on our findings. Furthermore, the absence of sharp vertical changes on the curves indicates that there were no threshold ABG values in relation to long-term mortality. We used the population-based discharge registry covering all hospital discharges in our county to obtain information on the level of comorbidity among study patients. This might have lead to a more complete recording of comorbidities and, thus, a more valid estimate of how burden of comorbidity affects outcome of IMV in COPD patients than previously reported.
The limitations of this observational study include the lack of standardized procedures for when to transfer COPD patients to the ICU, if and when to initiate IMV and when to obtain ABG samples in relation to initiation of IMV. This lack of standardized procedures may have increased the variability of our results and decreased our ability to detect differences in relative mortality estimates. We relied on clinical findings by experienced physicians in diagnosing COPD in patients who did not have a pulmonary function test done within 1 yr of ICU admission. It has been shown that only 20–30% of all hospitalized COPD patients have a relevant pulmonary function test [16,21] and the functional definition used in this study is therefore often used in studies on COPD . Even though any misclassification of COPD patients was unlikely to be related to results of ABG values or level of comorbidity, misclassification may still have lessened our ability to detect differences in MR according to the levels of ABG values or comorbidity.
Despite the limitations, overall credibility of the methodology used in our study was supported by the fact that mean pre-intubation pH, arterial carbon dioxide and oxygen tensions as well as overall mortality were within the range as previously reported [4-8,11,15,17].
Ninety-day mortality and 1-yr mortality in the present study were slightly lower than previously reported [4,6,11,17]. In a 1995 US study of 362 COPD patients requiring ICU transfer, 170 were mechanically ventilated; the 90-day and 1-yr mortality was 41% and 59%, respectively . A 2002 Australian study including 74 COPD patients (85% were mechanically ventilated) reported a 90-day mortality of 35% and a 1-yr mortality of 48.6% and a recent study from Singapore including 57 mechanically ventilated COPD patients found a 1-yr mortality of 42.7% . Generally, studies on outcome of mechanical ventilation of COPD patients are difficult to compare because of differences in inclusion criteria and aetiologies of respiratory failure.
Limited data exist on how ABG affects long-term mortality of mechanically ventilated COPD patients, but a number of studies have been conducted to examine how ABG values and short-term mortality reporting contradictive results. This may at least in part be because of great heterogeneity of study populations and differences in outcome measures. Seneff and colleagues  found that higher levels of hypercapnia were associated with increasing 90-day mortality ; however, data on the association between specific physiological parameters and mortality of mechanically ventilated patients were not presented making further comparisons with the present study difficult. Carbon dioxide tension was a predictor of in-hospital mortality in an Australian cohort study including 74 patients with COPD admitted to an ICU, 63 (85%) received IMV . Neither arterial oxygen tension nor acid–base status affected mortality. The study included patients with exacerbation of COPD exclusively, and the relatively small number of patients may have lead to statistical imprecision. Afessa and colleagues  identified 180 patients with COPD referred to an ICU because of ARF; IMV was used for 153 patients . In-hospital non-survivors had lower pH, but a similar PaO2/FiO2 ratio and arterial carbon dioxide tension compared with survivors; however, comparison with the present study is complicated by the inclusion of 15% non-mechanically ventilated patients. In contrast to our findings, Ray and colleagues  in a 2006 French study reported that hypercapnia >45 mmHg was associated with substantially increased mortality among elderly patients admitted to an emergency department with acute respiratory failure. A total of 29% of patients were transferred to an ICU, no data was published on the number of patients treated with mechanical ventilation. Based on the available data, we may only speculate on the reasons for the different findings; however, hypercapnia may be easier to manage in mechanically ventilated COPD patients than in non-mechanically ventilated patients. This may explain why the level of PaCO2 was associated with poor outcome in non-mechanically ventilated patients, but not in our study of mechanically ventilated patients.
Only a very limited number of studies have been conducted to examine how burden of comorbidity affects mortality in critically ill COPD patients treated with IMV. Besides the heterogeneity of ICU cohorts and sample size problems, the interpretation of the results of the few existing studies is hampered by the different methods used in quantifying comorbidity. In line with our findings, Seneff and colleagues  found no association between level of comorbidity and in-hospital mortality of COPD patients admitted to ICUs using the APACHE III prognostic system to identify comorbidities. The APACHE prognostic system, however, appears to be inferior compared with the Charlson Index in predicting mortality in critically ill patients based on the level of comorbidity . Ai-Ping and colleagues , using the Charlson Comorbidity Index, found no association between level of comorbidity and long-term mortality of mechanically ventilated COPD patients; however, the small sample size and statistical imprecision complicate the interpretation of their results .
Despite being a negative study, our results may still have clinical implications for the treatment of critically ill COPD patients. Clinicians' estimate of the probability of survival is one of the main factors in deciding which patients should be admitted to intensive care and treated with mechanical ventilation . Thus, incorrectly using ABGs and level of comorbidities, which intuitively may seem right, to estimate the prognosis of critically ill COPD patients may lead to some patients who might otherwise survive are being denied mechanical ventilation.
In conclusion, we found that patients with COPD treated with IMV had substantial 90-day and 1-yr mortality. Neither the level of ABG values measured immediately before IMV was initiated nor the burden of comorbidity appeared to be strong determinants of long-term mortality in this patient group.
This work was made possible through financial support from the Danish Medical Research Council (grant 271-05-0511) and the Western Danish Research Forum for Health Sciences.
No financial or other potential conflicts of interest exist.
1. Hurd S. The impact of COPD on lung health worldwide: epidemiology and incidence. Chest
(2 Suppl): 1S–4S.
2. Trends in chronic bronchitis and emphysema: Morbidity and mortality
. Available at: http://www.lungusa.org/data/copd/copd1.pdf
. Accessed 23 September 2007.
3. Needham DM, Bronskill SE, Sibbald WJ, Pronovost PJ, Laupacis A. Mechanical ventilation
in Ontario, 1992–2000: incidence, survival, and hospital bed utilization of noncardiac surgery adult patients. Crit Care Med
4. Seneff MG, Wagner DP, Wagner RP, Zimmerman JE, Knaus WA. Hospital and 1-year survival of patients admitted to intensive care units with acute exacerbation of chronic obstructive pulmonary disease
5. Connors AF Jr, Dawson NV, Thomas C et al.
Outcomes following acute exacerbation of severe chronic obstructive lung disease. The SUPPORT investigators (Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments). Am J Respir Crit Care Med
(4 Pt 1): 959–967.
6. Breen D, Churches T, Hawker F, Torzillo PJ. Acute respiratory failure secondary to chronic obstructive pulmonary disease
treated in the intensive care unit: a long term follow up study. Thorax
7. Moran JL, Green JV, Homan SD, Leeson RJ, Leppard PI. Acute exacerbations of chronic obstructive pulmonary disease
and mechanical ventilation
: a reevaluation. Crit Care Med
8. Afessa B, Morales IJ, Scanlon PD, Peters SG. Prognostic factors, clinical course, and hospital outcome of patients with chronic obstructive pulmonary disease
admitted to an intensive care unit for acute respiratory failure. Crit Care Med
9. Anon JM, Garcia de LA, Zarazaga A, Gomez-Tello V, Garrido G. Mechanical ventilation
of patients on long-term oxygen therapy with acute exacerbations of chronic obstructive pulmonary disease
: prognosis and cost-utility analysis. Intensive Care Med
10. Nevins ML, Epstein SK. Predictors of outcome for patients with COPD requiring invasive mechanical ventilation
11. Raurich JM, Perez J, Ibanez J, Roig S, Batle S. In-hospital and 2-year survival of patients treated with mechanical ventilation
for acute exacerbation of COPD. Arch Bronconeumol
12. Rieves RD, Bass D, Carter RR, Griffith JE, Norman JR. Severe COPD and acute respiratory failure. Correlates for survival at the time of tracheal intubation. Chest
13. Portier F, Defouilloy C, Muir JF. Determinants of immediate survival among chronic respiratory insufficiency patients admitted to an intensive care unit for acute respiratory failure. A prospective multicenter study. The French Task Group for Acute Respiratory Failure in Chronic Respiratory insufficiency. Chest
14. Hill AT, Hopkinson RB, Stableforth DE. Ventilation in a Birmingham intensive care unit 1993–1995: outcome for patients with chronic obstructive pulmonary disease
. Respir Med
15. Jeffrey AA, Warren PM, Flenley DC. Acute hypercapnic respiratory failure in patients with chronic obstructive lung disease: risk factors and use of guidelines for management. Thorax
16. Kaelin RM, Assimacopoulos A, Chevrolet JC. Failure to predict six-month survival of patients with COPD requiring mechanical ventilation
by analysis of simple indices. A prospective study. Chest
17. Ai-Ping C, Lee KH, Lim TK. In-hospital and 5-year mortality
of patients treated in the ICU for acute exacerbation of COPD: a retrospective study. Chest
18. Kizer JR, Zisman DA, Blumenthal NP et al.
Association between pulmonary fibrosis and coronary artery disease. Arch Intern Med
19. Hole DJ, Watt GC, Davey-Smith G, Hart CL, Gillis CR, Hawthorne VM. Impaired lung function and mortality
risk in men and women: findings from the Renfrew and Paisley prospective population study. BMJ
20. Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull
21. Pauwels RA, Buist AS, Calverley PM, Jenkins CR, Hurd SS. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease
. NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop summary. Am J Respir Crit Care Med
22. de Groot V, Beckerman H, Lankhorst GJ, Bouter LM. How to measure comorbidity. A critical review of available methods. J Clin Epidemiol
23. Royston P, Altman DG, Sauerbrei W. Dichotomizing continuous predictors in multiple regression: a bad idea. Stat Med
24. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology
25. Royston P, Ambler G, Saurbrei W. The use of Fractional Polynomials to model continuous risk variables in epidemiology. Int J Epidemiol
26. Ray P, Birolleau S, Lefort Y et al.
Acute respiratory failure in the elderly: etiology, emergency diagnosis and prognosis. Crit Care
27. Poses RM, McClish DK, Smith WR, Bekes C, Scott WE. Prediction of survival of critically ill patients by admission comorbidity. J Clin Epidemiol
28. Wildman MJ, Sanderson C, Groves J et al.
Implications of prognostic pessimism in patients with chronic obstructive pulmonary disease
(COPD) or asthma admitted to intensive care in the UK within the COPD and asthma study (CAOS): multicentre observational cohort study. BMJ