Discussion
In this contemporary observational study of more than 3,000 patients admitted with an AE-COPD to 38 ICUs in the United States, we found that after adjusting for severity of illness using propensity score matching, receipt of NIV was associated with lower ICU and in-hospital mortality and shorter LOS compared with IMV. Patients who failed NIV had the worst outcomes, and SAPS II was independently associated with the need for intubation.
Several current guidelines recommend early use of NIV in patients hospitalized with moderate to severe AE-COPD based on the results of RCTs which found that NIV improves outcomes compared with supportive care alone. For example, the Canadian Practice guidelines published in 2011 give a level 1A recommendation for NIV use in severe exacerbation of COPD versus standard therapy (no ventilation). However, the guidelines state: "we make no recommendation about the use of NIV versus conventional mechanical ventilation in patients who have a severe COPD exacerbation that requires ventilator support, because of insufficient evidence." There are only two RCTs which compared NIV with IMV; the small number of studies may reflect the general current belief that avoiding invasive ventilation is strongly desired. These trials, which included patients with severe ARF who failed standard treatment, found that NIV lowered hospital-acquired pneumonia but did not reduce mortality or LOS and had a high rate of NIV failure.
Our results are consistent with those from a survey study in 42 ICUs in France in 1997 which found that NIV use was associated with lower risk of death than IMV (10% vs 28%). In our study in the propensity-matched analysis, NIV therapy was associated with a 41% lower risk of death than IMV. Our results are also similar with the findings from the study by Tsai et al, a recent retrospective study of 67,000 ED visits for AE-COPD with ARF in United States; in this study, patients treated with NIV alone had a mortality rate of 8% compared with 16% in patients treated with IMV alone. Although we did reach some of the same conclusions as this study, our target population was different, and our methodology was much stronger and overcomes a number of the biases in the Nationwide Inpatient Sample, on which the work by Tsai et al was based.
In our study, the percentage of patients treated with NIV who avoided endotracheal intubation was 86%, and the SAPS II was predictive of NIV failure, which concurs with prior research. In a prospective randomized study, Antonelli et al reported that a SAPS II greater than 34 was independently associated with the need for endotracheal intubation. Consistent with the findings of Antonelli et al, we also found that the likelihood of NIV failure rose sharply at higher SAPS II. Patients who transition from NIV to IMV had a high ICU and hospital mortality in keeping with the results from other studies
Study Strengths and Limitations
This study has several strengths. To the best of our knowledge, this is the largest contemporary U.S. multicenter study of patients admitted to intensive care with respiratory failure secondary to an acute COPD exacerbation and provides an estimate of the ventilation practices and outcomes in the ICU for these patients. All the data were collected prospectively by trained, on-site investigators, the advance directive status was known, and the primary diagnosis was assigned at the time of admission, not retrospectively by coders. The dataset comprises clinical and laboratory data, and the SAPS II is calculated allowing better adjustment for the severity of illness than other observational studies with administrative datasets.
This study has also several limitations. First, although a mixture of teaching and nonteaching and small and large hospitals were included in this database, hospitals participating in the APACHE project may not be representative of all hospitals in the United States. Second, we did not have spirometric data. However, all diagnoses of COPD were assigned by an intensivist, and our comparative analysis included only ventilated patients with a diagnosis of COPD and so their COPD was most likely severe. Third, although we adjusted for the SAPS II and other patient characteristics using propensity score matching, the survival benefit we observed with NIV use assumes that all patients treated with NIV or IMV had similar degrees of severity of respiratory failure and this may not have been true. This was an observational study, and although we adjusted for severity via SAPS II and other physiological variables, we cannot exclude that unmeasured confounders at the patient and hospital level distorted our findings. In addition, the initial decision for NIV or IMV was based on physician judgment rather than on unified criteria. Fourth, unlike most randomized studies in this area, the population included in this study is of patients receiving ventilatory support for an exacerbation of COPD and not necessarily with hypercapnic acidotic exacerbation. Lastly, we examined in-hospital mortality because we did not have any information regarding the survival of patients with COPD after discharge though we note that 90-day mortality has been reported to be high.