Between April 2011 and April 2013, all patients hospitalized with COPD exacerbation at Kyungpook National University Hospital (KNUH), a tertiary referral center in Daegu, South Korea, were enrolled prospectively. The Institutional Review Board of the KNUH approved the study protocol, and informed consent was obtained from all participants. COPD was diagnosed based on available spirometric and medical records as recommended by the Global Initiative for Chronic Obstructive Lung Diseases (GOLD)23. COPD exacerbation was defined as acute deterioration requiring hospitalization from a stable condition24. On arrival at the hospital, emergency or internal medicine department physicians evaluated patients by history taking, physical examination, chest radiography, and routine laboratory tests. Patients with any specific causes identified during these examinations, including pneumonia, pneumothorax, and congestive heart failure, were excluded. After exclusions, all patients underwent computed tomography (CT) within 24 hours of admission. Patients with azotemia (serum creatinine>1.5 mg/dL) or exhibiting hypersensitivity to contrast media were excluded. As described in a previous study25, the etiologies of COPD exacerbations were determined as follows: 1) pneumonia was diagnosed when consolidation or ground-glass opacity (GGO) was observed by CT; 2) tracheobronchitis if symptoms suggestive of a respiratory infection, including increased sputum, purulent sputum, fever, or upper respiratory infection symptoms, were identified without consolidation or GGO by CT or if centrilobular nodules or a tree-in-bud pattern were observed on CT images, without consolidation or GGO; 3) congestive heart failure was diagnosed when CT findings indicated pulmonary edema with echocardiographic left LV dysfunction; 4) pulmonary embolism when a sharply delineated pulmonary arterial filling defect was located centrally within a vessel or made an acute angles at its interface with the vessel wall in at least two consecutive CT sections; and 5) undetermined etiology.

The 97 patients were allocated to two groups, that is, to an adverse outcome group (n=10) or to a control group (n=87). An adverse outcome was defined as a composite end point of 30-day mortality or the need for endotracheal intubation or inotropic support.

Baseline and clinical characteristics of the patients, including age, gender, smoking history, body mass index, comorbid conditions, and respiratory infection symptoms, were checked. Dyspnea was assessed using modified Medical Research Council (mMRC) grades26. For severity of COPD exacerbation, BAP-65 classes 27 were defined as follows: class I, patients≤65 years of age without any risk factors (blood urea nitrogen [BUN] level≥25 mg/dL, altered mental status, or pulse≥109 beats/min); class II, patients>65 years of age with no risk factor; and class III, IV, and V, patients with one, two, and three risk factors, respectively. Lengths of hospital stay were also recorded.

Pulmonary function testing (PFT) was performed as recommended by the American Thoracic Society and European Respiratory Society Guideline28. PFT data obtained within 6 months of index admission dates were used in the analysis.

Blood tests included white blood cell count, erythrocyte sedimentation rate (ESR), CRP, procalcitonin, NT-proBNP, troponin I, BUN, sodium, and albumin levels. Arterial blood gas analysis data, including partial oxygen pressure in arterial blood (PaO₂), partial carbon dioxide pressure in arterial blood (PaCO₂), inspired oxygen fraction (FiO₂), PaO₂/FiO₂ ratio, and alveolar-arterial oxygen gradient (P[A-a]O₂), were also checked.

Blood samples were obtained within 24 hours of arrival, centrifuged immediately, and stored at -80℃. Serum samples were registered and later provided by the National Biobank of Korea at the KNUH. Serum GDF-14 and H-FABP levels were measured using commercially available ELISA kits (Human GDF-15 Immunoassay; Quantikine, Minneapolis, MN, USA and Human H-FABP; Hycult Biotech, Uden, The Netherlands) according to the manufacturers' protocols.

Only 69 patients underwent echocardiography, because of operator non-availability. Transthoracic echocardiographic findings including the presence of RV dysfunction and RV systolic pressure (RVSP) were reviewed. RV dysfunction was echocardiographically defined as RV free wall hypokinesia, and RVSP was calculated using the tricuspid regurgitation flow velocity measured by Doppler echocardiography29. LV ejection fraction (LVEF) was determined by the Simpson biplane formula30.

CT scans were performed using a multidetector CT with 16 or 64 detector rows: Light Speed 16 (General Electric, Milwaukee, WI, USA) or Aquilion 64 (Toshiba Medical Systems, Tokyo, Japan). CT scans were reviewed by two radiologists (K-.M.S. and J.L.), and diagnoses were made by consensus. The presences of consolidation or GGO, bronchiolitis (centrilobular nodules or tree-in-bud pattern), bronchiectasis, bronchial wall thickening, and emphysema were checked. CT findings were defined as follows: consolidation as airspace opacification with obscuration of the underlying vasculature; GGO as mildly increased attenuation without obscuration of the underlying vasculature; tree-in-bud pattern as centrilobular nodules with either V- or Y-shaped branching linear opacities; bronchiectasis when bronchial internal diameter was greater than that of the adjacent pulmonary artery (PA); and bronchial anthracofibrosis as multiple smooth bronchial narrowing with adjacent calcified lymph nodes; bronchial wall thickening was assessed subjectively. As has been performed in previous studies31,32, diameters of main PA and ascending aorta (AA) at the level of the PA bifurcation were measured in order to obtain PA sizes and PA/AA diameter ratios. As in a previous study33, a PA/AA diameter ratio of >1 was selected as the cutoff value for pulmonary hypertension.

Statistical analyses were performed using SPSS version 12.0 (SPSS Inc., Chicago, IL, USA). Results are expressed as medians with ranges (interquartile range [IQR]) for continuous variables and numbers with percentages for categorical variables. The Mann-Whitney U test was used to compare continuous variables, whereas the chi-squared test or Fisher exact test was used to compare categorical variables. When continuous variables were converted to categorical variables, cut-off values were determined using receiver operating characteristic curves. To identify predictors of adverse outcomes, backward stepwise multiple logistic regression analysis was used using variables of p<0.1 in univariate analysis. A goodness-of-fit test used to assess the fit of logistic regression models was the Hosmer-Lemeshow test. Statistical significance was accepted for p<0.05.

Initially, 103 patients were enrolled. However, after reviewing CT scans, 6 were excluded due to underlying diseases with the potential to influence clinical features: pneumoconiosis, n=2; nonspecific interstitial pneumonia, n=1; lung cancer, n=1; tracheal cancer, n=1; and tracheal stenosis, n=1. As a result, 97 patients were included. The adverse outcome group comprised 10 patients; six died within 30 days after admission and four required inotropic support or endotracheal intubation. The remaining 87 patients were allocated to the control group.

In both groups, the majority of COPD exacerbations were caused by respiratory tract infections (pneumonia or tracheobronchitis) (Figure 1). The frequency of mental change was significantly more common in the adverse outcome group, as compared with the control group (3 [30%] vs. 5 [6%], p=0.034) (Table 1). However, no significant intergroup difference was found for BAP-65 class or the frequency of a high BAP-65 score (class IV-V). In addition, no significant difference was evident in percent predicted value of post-bronchodilator forced expiratory volume in one second (FEV₁) and post-bronchodilator FEV₁/forced vital capacity between the two groups.

Blood test results are presented in Table 2. Median ESR level tended to be lower in the adverse outcome group (14 mm/hr [IQR, 10-26 mm/hr] vs. 25 mm/hr [15-39 mm/hr]; p=0.068), and median serum CRP was significantly lower (1.26 mg/dL [0.35-2.92 mg/dL] vs. 3.54 mg/dL [1.33-8.27 mg/dL]; p=0.041). Median serum procalcitonin levels did not differ significantly. However, median serum NT-proBNP level tended to be higher in the adverse outcome group (1,799 [123-7,118] vs. 263 [69-833]; p=0.088). No significant difference was noted in plasma troponin I between the two groups. PaCO₂>37 mm Hg was significantly more frequent in the adverse outcome group (8 [80%] vs. 19 [22%]; p<0.001), and P(A-a)O₂ tended be higher in the adverse outcome group (105 mm Hg [36-157 mm Hg] vs. 52 mm Hg [34-75 mm Hg]; p=0.094).

Elevated serum GDF-15 levels (GDF-15>1,600 pg/mL) was significantly more common in the adverse outcome group (8 [80%] vs. 37 [43%]; p=0.041). In contrast, median serum H-FABP and the frequency of elevated H-FABP levels (H-FABP>755 pg/mL) were not significantly different between the two groups.

Echocardiographic parameters, including RV dysfunction, RVSP, and pulmonary hypertension, did not show any differences between the two groups (Table 3). There was no difference in LVEF between both groups. Of CT findings, no differences in PA/AA diameter ratio and signs of pulmonary hypertension (PA/AA diameter ratio>1) were observed between the two groups.

To identify predictors of adverse outcomes, we performed multiple logistic regression analysis. Of parameters with p<0.1 in the univariate analysis, serum NT-proBNP was excluded in multivariate analysis, because both NT-proBNP and GDF-15 are surrogate markers of myocardial stretch. Consequently, mental change, PaCO₂>37 mm Hg, P(A-a)O₂, and GDF-15 elevation (GDF-15>1,600 pg/mL) were selected for multivariate analysis to predict adverse outcomes in patients with COPD exacerbation (Table 4). The Hosmer-Lemeshow test indicated that the overall model fit was good (p=0.929). Multivariate analysis showed that PaCO₂>37 mm Hg and GDF-15 elevation were significant predictors of adverse outcomes (odds ratio [OR], 25.8; 95% confidence interval [CI], 2.7-243.8; p=0.005 and OR, 11.8; 95% CI, 1.2-115.3; p=0.034, respectively).