Rationale: Exacerbations of chronic obstructive pulmonary disease (COPD) are heterogeneous with respect to inflammation and etiology.
Objectives: Investigate biomarker expression in COPD exacerbations to identify biologic clusters and determine biomarkers that recognize clinical COPD exacerbation phenotypes, namely those associated with bacteria, viruses, or eosinophilic airway inflammation.
Methods: Patients with COPD were observed for 1 year at stable and exacerbation visits. Biomarkers were measured in sputum and serum. Viruses and selected bacteria were assessed in sputum by polymerase chain reaction and routine diagnostic bacterial culture. Biologic phenotypes were explored using unbiased cluster analysis and biomarkers that differentiated clinical exacerbation phenotypes were investigated.
Measurements and Main Results: A total of 145 patients (101 men and 44 women) entered the study. A total of 182 exacerbations were captured from 86 patients. Four distinct biologic exacerbation clusters were identified. These were bacterial-, viral-, or eosinophilic-predominant, and a fourth associated with limited changes in the inflammatory profile termed “pauciinflammatory.” Of all exacerbations, 55%, 29%, and 28% were associated with bacteria, virus, or a sputum eosinophilia. The biomarkers that best identified these clinical phenotypes were sputum IL-1β, 0.89 (area under receiver operating characteristic curve) (95% confidence interval [CI], 0.83–0.95); serum CXCL10, 0.83 (95% CI, 0.70–0.96); and percentage peripheral eosinophils, 0.85 (95% CI, 0.78–0.93), respectively.
Conclusions: The heterogeneity of the biologic response of COPD exacerbations can be defined. Sputum IL-1β, serum CXCL10, and peripheral eosinophils are biomarkers of bacteria-, virus-, or eosinophil-associated exacerbations of COPD. Whether phenotype-specific biomarkers can be applied to direct therapy warrants further investigation.
Exacerbations of chronic obstructive pulmonary disease (COPD) are a major health burden worldwide, and affect a vulnerable population at risk of significant comorbidities. COPD exacerbations are heterogeneous with respect to etiology and inflammation and biomarkers are required to phenotype this heterogeneity.
We have shown that there are biologic COPD exacerbation clusters that are clinically indistinguishable, and that biomarkers can be used to identify specific clinical phenotypes during exacerbations of COPD (specifically those associated with bacteria, virus, and sputum eosinophilia). Bacterial and eosinophilic clinical exacerbation phenotypes can be identified from stable state. Our data further delineate the heterogeneity during COPD exacerbations and may identify populations that appropriately require corticosteroids and antibiotics at the onset of an exacerbation.
Acute exacerbations of chronic obstructive pulmonary disease (COPD) are associated with substantial morbidity and mortality (1, 2). Exacerbations are typically associated with increased neutrophilic and to a lesser extent eosinophilic airway inflammation (3, 4). Respiratory viral and bacterial infections have been implicated in causing most exacerbations (5–7), but how these infections alter lower airway inflammation and relate to treatment response is not completely understood. This heterogeneity translates that at present clinicians have limited tools to phenotype exacerbations. During stable state a sputum eosinophilia is associated with corticosteroid responsiveness (8–10), whereas the presence of a high bacterial load and sputum purulence has favorable outcomes with antibiotics (11–15). These findings suggest that it is possible to identify clinically important COPD exacerbation phenotypes. This is crucial because systemic corticosteroids and antibiotics have marginal efficacy (16–21) and the potential to cause adverse events in an already vulnerable population.
We hypothesize that approaches aimed at the identification of COPD exacerbation phenotypes may allow for better prognostic, therapeutic, and mechanistic applications (22–24). In this study we investigated whether during exacerbations of COPD there are (1) definable biologic phenotypes using unbiased mathematical tools (namely factor and cluster analysis); (2) identifiable biomarkers associated with clinical phenotypes, specifically those associated with bacteria, viruses, or sputum eosinophilia; and (3) exacerbation phenotypes that can be predicted from stable state.
Patients with a physician diagnosis of COPD and a post-bronchodilator FEV1/FVC ratio of less than 0.7 as per global initiative for chronic obstructive lung disease (GOLD) criteria (1) were recruited from the Glenfield Hospital, Leicester, United Kingdom, and through local advertising. All patients fulfilled the inclusion criteria of age greater than 40 years, GOLD stage I–IV, and greater than or equal to one exacerbation in the preceding 12 months defined as the requirement of emergency health care (25). Patients were excluded if there was a documented inability to produce sputum after the induced sputum procedure, a current or previous history of asthma, currently active pulmonary tuberculosis, or any other clinically relevant lung disease other than COPD. The presence of comorbidity, reported atopy to common aeroallergens, or reversibility on lung function testing was not an exclusion criterion. All patients gave informed written consent and the study was approved by the local ethics committee.
This was a prospective observational study. Patients were seen at stable state and during exacerbations for the duration of 1 year. Stable visits including the baseline visits were 8 weeks free from an exacerbation. All patients were given daily diary cards to complete, and asked to contact the research team if there was an increase in symptoms of breathlessness, sputum volume, and purulence. Exacerbations were defined according to Anthonisen criteria (14) and health care use (25). Exacerbation data recording and sampling were only performed in patients who had not received prior oral corticosteroids or antibiotics. Patients were all clinically assessed (including chest radiograph, temperature recording, and blood gas analysis if clinically necessary) to exclude other causes of breathlessness. Patients with an exacerbation of COPD were then treated according to guidelines (2).
At all visits, patients underwent pre– and post–400-μg albuterol bronchodilator spirometry (Vitalograph, Buckingham, Buckinghamshire, UK); induced or spontaneous sputum collection (26); and measurements of symptoms and health quality assessments using the Visual Analog Scale (27) and the Chronic Respiratory Disease Interviewer-Administered Standardized Questionnaire (CRQ) (28). Sputum was collected and analyzed for bacteria (29–31) (using standard routine culture, CFU, and real-time quantitative polymerase chain reaction [qPCR]), for viruses by PCR (32), and processed to produce cytospins for cell differential and supernatant for fluid phase measurements (33). A broad panel of serum and sputum biomarkers were measured using the Meso-Scale-Discovery (MSD, Gaithersburg, MD) platform standard preprepared plates (MSD, MD) and single ELISA at stable and exacerbation visits (see Table E1 in the online supplement).
Bacteria-associated exacerbations were defined as a positive bacterial pathogen on routine culture (Haemophilus influenzae, Moraxella catarrhalis, Streptococcus pneumoniae, Staphylococcus aureus, or Pseudomonas aeruginosa) or a total aerobic CFU count greater than or equal to 107 cells (12, 15). qPCR bacterial detection methods were not used to define bacteria-associated exacerbations in this study.
A virus-associated exacerbation was defined as one that had a positive sputum viral PCR, whether in isolation or in combination with a positive bacterial pathogen on routine culture. A sputum eosinophil–associated exacerbation was defined as the presence of more than 3% nonsquamous cells.
Statistical analyses were performed using PRISM version 4 (GraphPAD PRISM, La Jolla, CA) and SPSS version 16 (IBM, Chicago, IL). Parametric and nonparametric data are presented as mean (SEM) and median (interquartile range). Log transformed data are presented as geometric mean (95% confidence interval [CI]). Multivariate modeling using principal component analysis in sputum biomarkers was used to explore biomarker pattern expression at exacerbations. No adjustments for multiple comparisons have been made across biomarkers.
Factor analysis, a mathematical method that discovers patterns of relationships within large datasets, was used to identify factors in sputum mediators at exacerbations thereby demonstrating biologic factors independent of each other and of any clinical expression. This method using unsupervised principal component analysis demonstrated three factors accounting for 75% of the total variance (see Table E2). Cluster analysis, an unbiased mathematical method, allows one to classify groups on similar chosen characteristics alone. Thus, after demonstrating three biologic factors, we used hierarchical cluster analysis to generate four biologic clusters for exacerbation events and cases. Clinical characteristics for all exacerbation events were tabulated for each biologic cluster. One-way analysis of variance, Kruskal-Wallis test, and the chi-square test were used to compare the clinical characteristics between cluster groups.
For comparison of clinical and biomarker changes between baseline and exacerbation visits the paired t test or Wilcoxon matched pairs test was used. For comparison of exacerbations associated with or without bacteria, virus, and sputum eosinophilia the t test and Mann-Whitney test were used, respectively. To determine suitable biomarkers, the receiver operating characteristic curves were plotted for (1) exacerbation versus stable state; (2) bacteria- versus nonbacteria-associated exacerbations; (3) virus- versus nonvirus-associated exacerbations; and (4) sputum eosinophilia– (> 3% nonsquamous cells) versus nonsputum eosinophilia–associated exacerbations.
Validation of the identified biomarkers for bacteria-, virus-, and sputum eosinophil–associated exacerbation was performed in a further 89 exacerbation events from an independent cohort of subjects with COPD. These subjects with COPD were recruited to enter a prospective study with identical inclusion and exclusion criteria and study design as the current study. Stable and exacerbation visits were treated in accordance to the main study.
A P value of less than 0.05 was taken as the threshold of statistical significance.
One hundred fifty-six patients were enrolled; 145 (101 men and 44 women) completed the first visit and 115 completed 12 months (Figure 1; see Figure E1). At baseline 3%, 48%, 30%, and 19% had GOLD I, II, III, and IV, respectively. Most patients recruited were current or exsmokers (142 of 145), with a mean (range) pack-year history of 49 (10–153) with an absolute and percentage mean (SEM) reversibility to inhaled bronchodilator on study entry of 47 ml (11 patients) and 4% (one patient), respectively. Skin prick testing or serum-specific IgE to a wide panel of aeroallergens confirmed that 20% were atopic. Bacterial colonization, defined as the presence of a potentially pathogenic microorganism (H. influenzae, M. catarrhalis, S. pneumoniae, S. aureus, or P. aeruginosa) in a standard culture technique (29), was present in 28% of patients at baseline. Using qPCR a bacterial pathogen (H. influenzae, M. catarrhalis, S. pneumoniae, or S. aureus) was detected in 86% of patients at the baseline stable visit. A virus was detected in 5% of subjects at study entry, whereas eosinophilic airway inflammation (> 3% nonsquamous cells) was present in 27% of patients. Baseline and exacerbation clinical characteristics are shown in Table 1 (see Table E3).
|Study Entry||Study Entry||Exacerbation||P Value|
|Male, n (%)||101 (70)||FEV1,L†||1.33 (0.05)||1.10 (0.04)||<0.001|
|Age*||69 (43–88)||FEV1% predicted†||52 (2)||42 (1)||<0.001|
|Age at diagnosis*||62 (38–83)||Reversibility, ml||47 (11)||37 (11)||0.50|
|Current smokers, n (%)||42 (29)||FEV1/FVC ratio, %†||52 (2)||50 (1)||0.65|
|Exsmokers, n (%)||100 (69)||CRQTOTAL, units||4.11 (0.10)||3.12 (0.08)||<0.001|
|Pack-year history*||49 (10–153)||VASTOTAL, mm||142 (6)||239 (6)||<0.001|
|Exacerbation rate in previous 12 mo||3 (0.2)||Peripheral leukocyte count (×109 cells/L)‡||8.2 (7.9–8.6)||9.3 (8.9–9.8)||<0.001|
|Maintenance prednisolone, n (%)||9 (6)||Peripheral neutrophil count (×109 cells/L)‡||5.3 (5–5.6)||6.3 (6–6.7)||<0.001|
|Prednisolone dosage, mg*||6 (4–10)||Peripheral eosinophil count (×109 cells/L)‡||0.21 (0.18–0.23)||0.19 (0.17–0.22)||0.84|
|Inhaled corticosteroid use, n (%)||125 (86)||Total sputum cell count (×106 cells/g sputum)‡||3.8 (3.1–4.7)||6.4 (5.2–7.8)||<0.001|
|Inhaled corticosteroid dose, μg§||1,540 (59)||Sputum neutrophil count, %||68 (2)||74 (2)||0.02|
|Inhaled long-acting β agonist use, n (%)||110 (76)||Sputum eosinophil count, %‡||1.2 (1–1.6)||1.1 (0.9–1.5)||0.58|
A total of 182 exacerbation events were captured from 86 patients; of these 21 exacerbations warranted hospitalization. There was a reduction in the FEV1 and CRQ from baseline to exacerbation (FEV1 [L] 1.33 vs. 1.10; mean difference 0.24; 95% CI, 0.12–0.36; P < 0.001) (CRQ [units] 4.11 vs. 3.12; mean difference 0.99; 95% CI, 0.74–1.23; P < 0.001). The magnitude of these changes was independent of smoking status, sex, GOLD severity (1), or Anthonisen criteria (14). Hospitalized exacerbations were associated with a greater decline in lung function compared with exacerbations that were not hospitalized (ΔFEV1 [ml] −355 vs. −131; mean difference 224; 95% CI of difference, −356 to −92; P < 0.001), but not health status decline (ΔCRQ [units] −1.25 vs. −0.91; mean difference 0.34; 95% CI of difference, −0.83 to 0.15; P = 0.18).
Serum and sputum mediator data were available in 148 exacerbation events from 75 patients. Serum biomarkers that increased during an exacerbation were IL-6, tumor necrosis factor (TNF) receptors I and II, serum amyloid-A, C-reactive protein (CRP), procalcitonin, and serum eosinophil cationic protein (Table E4A). Sputum biomarkers that increased were IL-1β, TNF-α, TNFRI, TNFRII, IL6, CCL5, and CCL4 (Table E4B). No single biomarker had a receiver operating curve area under the curve greater than 0.70 in determining an exacerbation from stable state (Figure E2). Of all sputum and serum biomarkers measured there was a significantly increased level of serum TNF-α and CRP in patients who were hospitalized (CRP median [IQR] 56 (102) vs. 8 (14); P = 0.002) (serum TNF-α geometric mean 4.3 [95% CI, 3.4–5.4] vs. 3.4 [95% CI, 3.2–3.6]; P = 0.02).
Factor analysis identified three biologic factors at exacerbation representing proinflammatory, Th1, and Th2 factors as determined by their cytokine expression profiles (Table E3). Cluster analysis using the highest loading biomarker from each factor (TNFRII, CXCL11, and CCL17) revealed four biologic cluster populations for exacerbation events. Three clusters were termed as “bacteria-predominant,” “eosinophil-predominant,” and “virus-predominant.” A fourth cluster demonstrated low sputum mediator concentrations and had fewer events associated with known etiology and was termed “pauciinflammatory.” Factor mean scores were plotted for each cluster (Figures 2A, and E3A). Biologic cluster ellipsoids were calculated and plotted for all exacerbation events to schematically represent biologic clusters of COPD exacerbations in three dimensions (Figure 2B, Figure E3B). Exacerbation event characteristics of these biologic clusters are presented in Table 2 (Table E5). The baseline characteristics for each subject within each biologic cluster are shown in Tables 2 and E5. Cluster membership was determined using either a patient's first exacerbation event or the dominant cluster in patients with multiple exacerbations. The intraclass correlation coefficient of the biologic clusters for patients with repeated exacerbations was 0.73. Each biologic cluster was found to be differentially related to inflammation and etiology, but was otherwise clinically indistinguishable.
|Cluster 1: Bacteria-predominant||Cluster 2: Eosinophil-predominant||Cluster 3: Virus-predominant||Cluster 4: Pauciinflammatory||P Value|
|Number (%)||52 (35)||44 (30)||36 (24)||16 (11)||—|
|Sputum TNFRII (pg/ml)*||1,722 (1,402–2,117)||353 (287–433)||1,254 (969–1,623)||77 (41–147)||<0.0001|
|Sputum CXCL11 (pg/ml)*||3.1 (2.2–4.3)||10.9 (7.7–15.5)||799 (415–1,539)||17.3 (5.6–53.1)||<0.0001|
|Sputum CCL17 (pg/ml)*||5.5 (4.5–6.7)||34.8 (27.3–44.5)||23.5 (16.2–34.1)||4.7 (3.5–6.3)||<0.0001|
|Bacterial exacerbation, % (95% CI)||86 (73–92)||29 (18–45)||44 (28–61)||31 (12–58)||<0.0001|
|Viral exacerbation, % (95% CI)||22 (13–35)||10 (3–23)||57 (39–73)||30 (10–61)||<0.0001|
|Eosinophilic exacerbation, % (95% CI)||6 (1–16)||60 (45–74)||28 (16–44)||27 (10–52)||<0.0001|
|Δ FEV1, ml†||−132 (−251 to −35)||−110 (−230 to −31)||−232 (−340 to −124)||−280 (−524 to −36)||0.32|
|Δ CRQ, units †||−0.9 (−1.2 to −0.6)||−0.9 (−1.3 to −0.5)||−0.9 (−1.4 to −0.4)||−1 (−1.9 to −0.1)||0.99|
|Δ VASTOTAL, mm†||79 (42–116)||80 (41–119)||120 (86–154)||73 (38–108)||0.39|
|Number, (%)||28 (37)||19 (25)||20 (27)||8 (11)||—|
|Male, n (%)||18 (64)||14 (74)||14 (70)||7 (88)||0.63|
|Age, yrs‡||69 (52–84)||68 (45–88)||70 (49–84)||69 (61–85)||0.62|
|Current smokers, n (%)||8 (29)||8 (42)||4 (20)||3 (38)||0.48|
|Pack-years smoked‡||44 (10–122)||50 (20–106)||47 (10–134)||72 (23–120)||0.11|
|Exacerbation rate in previous 12 mo||3.8 (0.5)||4.3 (0.5)||4 (0.7)||4.9 (1.2)||0.58|
|Exacerbation rate during study||3.8 (0.3)||3.6 (0.4)||3.2 (0.3)||3.1 (0.5)||0.64|
|Inhaled corticosteroid dose, μg§||1,507 (147)||1,567 (133)||1,470 (160)||1,150 (188)||0.55|
|Residual volume, %||134 (8)||150 (9)||120 (8)||146 (23)||0.11|
|TLCO % predicted||56 (5)||59 (5)||57 (6)||46 (7)||0.62|
|FEV1% predicted, baseline||53 (3)||51 (5)||53 (5)||40 (7)||0.34|
|FEV1/FVC ratio (%)||51 (2)||47 (2)||50 (3)||47 (5)||0.67|
|CRQTOTAL, units||4.14 (0.20)||3.90 (0.22)||4.10 (0.26)||3.66 (0.50)||0.74|
|VASTOTAL, mm||178 (15)||142 (18)||124 (18)||147 (37)||0.14|
|Total sputum cell count (×106 cells/g)*||8.3 (5.5–12.5)||2.3 (1.6–3.2)||2.5 (1.2–5.3)||3.5 (1.2–10.7)||0.002|
|Sputum neutrophil count, %||75 (5)||53 (4)||68 (4)||81 (6)||0.003|
|Sputum eosinophil count, %*||1 (0.6–1.6)||3.1 (1.4–6.6)||1 (0.5–1.9)||0.5 (0.2–1)||0.012|
|Bacterial colonization, % (95% CI)||63 (48–77)||27 (15–43)||11 (3–29)||38 (18–61)||0.001|
Fifty-five percent of exacerbations were bacteria-associated exacerbations (positive bacterial pathogen on routine culture or CFU ≥ 107). Blood and sputum neutrophils were increased. Total bacterial load (16S) was higher in patients with a bacteria-associated exacerbation than those without (geometric mean 7.67 [95% CI, 4.27 to 1.48] vs. 2.88 [95% CI, 1.78 to 4.78]; P = 0.001). There was no difference in the 16S signal across exacerbations of Anthonisen type (analysis of variance; P = 0.64). Using qPCR, acquisition of a new species occurred in 15% of exacerbations. Clinical assessments of change in FEV1, symptoms of sputum production, and sputum purulence had an area under the receiver operating characteristic curve of 0.45 (95% CI, 0.35–0.55), 0.50 (95% CI, 0.40–0.60), and 0.58 (95% CI, 0.48–0.68), respectively. The most suitable biomarker for determining bacteria-associated exacerbations was sputum IL-1β with an area under the receiver operating characteristic curve of 0.89 (95% CI, 0.83–0.95). A cutoff of 125 pg/ml had a sensitivity of 90% and a specificity of 80% (Figures 3A and E4A). The best serum biomarker was CRP with an area under the receiver operating characteristic curve of 0.65 (95% CI, 0.57–0.74). A serum CRP cutoff of 10 mg/L had a sensitivity of 60% and specificity of 70%.
Twenty-nine percent of exacerbations were associated with a virus, most commonly rhinovirus. Virus-associated exacerbations had a larger fall in % FEV1 compared with nonvirus-associated exacerbations (−17% vs. −9%; mean difference −8%; 95% CI, −16 to −1; P = 0.04). Clinical assessments of change in FEV1, symptoms of cough and breathlessness, had an area under the receiver operating characteristic curve of 0.43 (95% CI, 0.32–0.53), 0.62 (95% CI, 0.52–0.72), and 0.51 (95% CI, 0.41–0.62), respectively. The best marker for distinguishing the presence of a virus at exacerbation was serum CXCL10 (IP-10), with an area under the receiver operating characteristic curve of 0.76 (95% CI, 0.67–0.86). A serum CXCL10 cut off of 56 pg/ml gave a sensitivity of 75% and specificity of 65% (Figures 3B and E4B). For exacerbations associated with virus alone the area under the receiver operating characteristic curve for serum CXCL10 improved to 0.83 (95% CI, 0.70–0.96).
A sputum eosinophilia was observed in 28% of exacerbations. The most sensitive and specific measure to determine a sputum eosinophilia at exacerbation was the percentage peripheral blood eosinophil count with an area under the receiver operating characteristic curve of 0.85 (95% CI, 0.78–0.93). A cutoff of 2% peripheral blood eosinophils had a sensitivity of 90% and specificity of 60% for identifying a sputum eosinophilia of greater than 3% at exacerbation (Figure 3C, Figure E4C).
In summary, the etiologic and inflammatory causes of exacerbation events were as follows: bacteria alone 37%, virus alone 10%, sputum eosinophilia alone 17%, bacteria plus virus 12%, bacteria plus sputum eosinophilia 6%, virus plus sputum eosinophilia 3%, bacteria plus virus plus sputum eosinophilia 1%, and none 14%.
Multivariate modeling using combinations of two or three biomarkers for the detection of bacteria-, virus-, and eosinophil-associated exacerbations did not significantly improve on the single mediators alone (data not shown). Differential clinical and biomarker expression for exacerbations associated with bacteria, virus, and sputum eosinophilia are shown in Tables E4–E6.
The odds ratio for a bacteria or an eosinophil-associated exacerbation was 4.9 (95% CI, 2.4–9.9; P < 0.001) or 2.7 (95% CI, 1.3–5.7; P = 0.01) if the patient had a bacterial pathogen on diagnostic routine culture or a sputum eosinophilia on greater than or equal to one occasion at stable state. The odds ratio for a virus-associated exacerbation if the patient had a virus at stable state was 0.5 (95% CI, 0.1–3.9; P = 0.67).
In an independent study of COPD exacerbations, 89 subjects (57 men and 32 women) with a mean (range) age of 68 (46–86) years and mean (SEM) FEV1% predicted of 46 (2) percent sputum IL-1β and serum CXCL10 was measured using a commercial ELISA (R&D Systems, Abingdon, UK). The area under the receiver operating characteristic curve for percentage blood eosinophils to identify a sputum eosinophil–associated exacerbation was 0.95 (95% CI, 0.87–1.00) with a cutoff of 2% having a sensitivity and specificity of 90% and 60%. The area under the receiver operating characteristic curve for sputum IL-1β and serum CRP to identify a bacteria-associated exacerbation was 0.73 (95% CI, 0.61–0.85) and 0.70 (95% CI, 0.59–0.82); a sputum IL-1β cutoff of 130 pg/ml had a sensitivity and specificity of 80% and 60%, and a serum CRP cutoff of 10 mg/L had sensitivity and specificity of 65%. The area under the receiver operating characteristic curve for serum CXCL10 to identify a virus-associated exacerbation was 0.65 (95% CI, 0.52–0.78) with a cutoff of 145 pg/ml having a sensitivity and specificity of 70% and 60%.
Further details and results are available in the online supplement.
In this study we have used two methods to investigate biomarkers in COPD. The first using unbiased statistical tools, free from bias and independent of clinical expression, identified biologic COPD exacerbation phenotypes and characterized exacerbations into four biologic clusters, The second method used current clinical exacerbation phenotypes of COPD related to potential etiology and inflammation, namely exacerbations that are associated with bacteria, virus, or a sputum eosinophilia. Interestingly, we were unable to define biomarkers for exacerbations per se, despite a generalized increase in systemic and airway inflammation (34–36). The biologic exacerbation clusters were bacterial-, viral-, or eosinophilic-predominant, and a fourth was associated with limited changes in the inflammatory profile and was termed “pauciinflammatory.” These clusters were remarkably similar to our clinical exacerbation phenotypes. We identified biomarkers for our clinical exacerbation phenotypes that were then validated in an independent cohort. The bacteria- and sputum eosinophil–associated exacerbations rarely coexisted, and were reliably predicted from stable state suggesting fundamental differences in their immunopathogenesis. Therefore, in addition to identifying potential biomarkers to direct therapy, these exacerbation clinical phenotypes are likely to represent distinct pathophysiologic entities with specific biomarker signatures.
Biomarker profiling in COPD exacerbations has the potential to further the understanding of disease mechanisms (22), whereas phenotypic approaches lend to prognostic and therapeutic strategies (37, 38). Using factor and cluster analysis, a novel approach of characterizing COPD and exacerbations (23), we were able to reduce an extensive panel of measured sputum biomarkers into three factors, from which we determined four biologic clusters. This analytic strategy is free from investigator bias. These biologic clusters could not be distinguished clinically or by Anthonisen criteria (14) and the exacerbation severity was similar across the clusters. Importantly, using factor analysis we have shown differential inflammatory profiles between the bacteria-predominant, eosinophil-predominant, virus-predominant, and pauciinflammatory clusters. In patients with multiple exacerbations the biologic clusters were repeatable, and exacerbations associated with bacteria or a sputum eosinophilia but not viruses could be predicted from stable state. Therefore, our data are consistent with the view that bacterial and eosinophilic exacerbations may reflect instability within a complex and inherently unstable system, whereas viral exacerbations are more likely to represent acquisition of a new pathogen. It is likely both of these mechanisms drive exacerbations, but critically we have determined biologic clusters and clinical phenotypes that may respond to different management strategies, which can potentially be identified using biomarker profiles.
The inflammatory profile of a COPD exacerbation is typically neutrophilic, but eosinophilic airway inflammation also exists, and is associated with a favorable response to corticosteroid therapy (8–10). Eosinophilia in inflammatory airways disease is associated with increased all-cause mortality (39, 40) and may highlight different genetic, biologic, and pathologic disease processes. Importantly, the sputum differential rather than total eosinophil count has consistently been shown to be associated with important clinical outcomes (9, 10). We found that the peripheral percentage eosinophil count was a sensitive biomarker of a sputum eosinophilia. Current guidelines recommend the use of systemic corticosteroids for COPD exacerbations, although the magnitude of the benefit is marginal and their use associated with significant side effects (18). Our findings raise the possibility that targeting corticosteroid therapy in a subgroup of exacerbations dependent on the peripheral eosinophil count may reduce inappropriate use of systemic corticosteroids.
Bacteria are considered to play a role in up to 50% of exacerbations (7). Current guidelines propose sputum purulence to guide antibiotic therapy (13). Sputum purulence is sensitive for detecting bacterial culture or high bacterial yields at exacerbation in COPD (12). However, the use of sputum purulence alone is confounded by its presence at stable state and chronic bacterial colonization (41), possibly as a consequence of poor bacterial clearance (42). Furthermore, the change in sputum purulence or sputum production symptoms in our cohort was not sensitive or specific for identifying a bacteria-associated exacerbation. The most sensitive and specific assay for determining bacteria-associated exacerbations was sputum IL-1β. This extends previous findings that bronchoalveolar lavage IL-1β was a good biomarker for ventilator-associated pneumonia (43), and suggests that this airway marker may suitably determine bacterial infections, above that of serum CRP or procalcitonin whose use could not be demonstrated in this study or in others (34, 35). Sputum IL-1β could thus be used as a biomarker to correctly identify bacteria-associated exacerbations but would require the development of a rapid near patient test to be of use in clinical practice.
Viruses have been implicated as a major cause of COPD exacerbations and are detected in approximately half of severe COPD exacerbations (5, 6). The total sputum eosinophil count has been proposed as a potential biomarker of a viral exacerbation (5). Here we also found that the total absolute sputum eosinophil count was increased in virus-associated exacerbations, but not the differential sputum eosinophil count, suggesting the association was largely a consequence of a change in the total cell count. The application of clinical symptoms in combination with serum CXCL10 (IP-10) has been proposed as a possible biomarker for rhinovirus infection at exacerbation (44). This study confirms that serum CXCL10 levels as a potential predictor of a virus-associated exacerbation, independent of a requirement for symptom evaluation. Novel antiviral approaches are in development and CXCL10 is thus a promising biomarker to direct future antiviral therapy.
One potential criticism is that this is a single-center study and therefore our findings need to be replicated across multiple centers, and validated prospectively to identify the biologic clusters and our proposed biomarkers for the clinical exacerbation phenotypes; nonetheless, this approach may represent a new paradigm in the management of COPD exacerbations. Importantly, we have replicated the biomarkers peripheral blood eosinophils, sputum IL-1β, and serum CXCL10 in a validation cohort. Peripheral blood eosinophils remained a strong marker of a sputum eosinophilia. Sputum IL-1β and serum CXCL10 were measured using a different platform but remained significant albeit weaker predictive markers of identifying a bacteria- or virus-associated exacerbation. In our study a statistical analytic limitation was that we did not correct for repeated measures and assessed changes in biomarkers in paired or unpaired tests; however, we examined two methods to investigate biomarkers in COPD exacerbations, using unbiased statistical tools to demonstrate four biologic clusters and analysis of biomarkers to look at predefined clinical exacerbation subgroups, and further used multivariate analysis to determine that combinations of markers did not improve our predictive model. The presence of coinfection with virus and bacteria in our study was lower than that previously reported (5), but may reflect differences in the severity of exacerbations. The relationship between virus and bacterial infection in exacerbations, however, remains poorly understood (5, 45). We chose to define a bacteria-associated exacerbation based on a positive routine culture or a high bacterial load; however, the causal links between the presence of bacteria and exacerbations has not been rigorously confirmed, and evidence for efficacy of antibiotics in treatment is conflicting (13, 19–21, 46). Developments in molecular bacterial identification of bacteria (47) and emerging microbiomics (48) are beginning to redefine the microbiota of the airway in health and disease and will likely change the view of what defines a bacterial infection. Improvements in viral detection and the identification of new respiratory viruses are also changing the understanding of the associations between exacerbations and these agents. Further work is required before therapeutic implications and interpretative criteria can be established for these sensitive detection methods. Whether identification of a pauciinflammatory biologic cluster and a proportion of subjects without clear evidence of a cause for their exacerbation reflect the insensitivity of our chosen cutoffs for definitions or a real entity requires further clarifications.
In conclusion, COPD exacerbations are heterogeneous. This phenotypic heterogeneity can be defined. Using unbiased statistical tools we have determined four biologic exacerbation clusters that relate to identifiable patterns of inflammation and potential causative pathogens. We have defined sensitive and specific biomarkers to identify predefined clinical exacerbation phenotypes, which need to be tested in randomized prospective studies of targeted therapy. These subgroups are independent and suggest that the mechanisms driving their exacerbations are distinct and may be amenable to more specific interventions, potentially moving the management of COPD exacerbations toward the realization of phenotype-specific management.
The authors thank all the research volunteers who participated in the study, and also the following people for their valuable assistance throughout the study: J. Aniscenko, M. Bourne, R. Braithwaite, D. Burke, J. Footitt, E. Goldie, J. Goldie, N. Goodman, S. Gupta, B. Hargadon, I. Rushby, M. Shelley, A. Singapuri, D. Vara, R. Walton, and S. Winpress.
|1.||Rabe KF, Hurd S, Anzueto A, Barnes PJ, Buist SA, Calverley P, Fukuchi Y, Jenkins C, Rodriguez-Roisin R, van Weel C, et al.. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: GOLD executive summary. Am J Respir Crit Care Med 2007;176:532–555.|
|2.||Halpin D. NICE guidance for COPD. Thorax 2004;59:181–182.|
|3.||Bhowmik A, Seemungal TA, Sapsford RJ, Wedzicha JA. Relation of sputum inflammatory markers to symptoms and lung function changes in COPD exacerbations. Thorax 2000;55:114–120.|
|4.||Saetta M, Di SA, Maestrelli P, Turato G, Ruggieri MP, Roggeri A, Calcagni P, Mapp CE, Ciaccia A, Fabbri LM. Airway eosinophilia in chronic bronchitis during exacerbations. Am J Respir Crit Care Med 1994;150:1646–1652.|
|5.||Papi A, Bellettato CM, Braccioni F, Romagnoli M, Casolari P, Caramori G, Fabbri LM, Johnston SL. Infections and airway inflammation in chronic obstructive pulmonary disease severe exacerbations. Am J Respir Crit Care Med 2006;173:1114–1121.|
|6.||Seemungal T, Harper-Owen R, Bhowmik A, Moric I, Sanderson G, Message S, Maccallum P, Meade TW, Jeffries DJ, Johnston SL. Respiratory viruses, symptoms, and inflammatory markers in acute exacerbations and stable chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2001;164:1618–1623.|
|7.||Sethi S, Murphy TF. Infection in the pathogenesis and course of chronic obstructive pulmonary disease. N Engl J Med 2008;359:2355–2365.|
|8.||Shim C, Stover DE, Williams MH. Response to corticosteroids in chronic bronchitis. J Allergy Clin Immunol 1978;62:363–367.|
|9.||Pizzichini E, Pizzichini MM, Gibson P, Parameswaran K, Gleich GJ, Berman L, Dolovich J, Hargreave FE. Sputum eosinophilia predicts benefit from prednisone in smokers with chronic obstructive bronchitis. Am J Respir Crit Care Med 1998;158:1511–1517.|
|10.||Brightling CE, Monteiro W, Ward R, Parker D, Morgan MD, Wardlaw AJ, Pavord ID. Sputum eosinophilia and short-term response to prednisolone in chronic obstructive pulmonary disease: a randomised controlled trial. Lancet 2000;356:1480–1485.|
|11.||White AJ, Gompertz S, Bayley DL, Hill SL, O'Brien C, Unsal I, Stockley RA. Resolution of bronchial inflammation is related to bacterial eradication following treatment of exacerbations of chronic bronchitis. Thorax 2003;58:680–685.|
|12.||Stockley RA, O'Brien C, Pye A, Hill SL. Relationship of sputum color to nature and outpatient management of acute exacerbations of COPD. Chest 2000;117:1638–1645.|
|13.||Ram FS, Rodriguez-Roisin R, Granados-Navarrete A, Garcia-Aymerich J, Barnes NC. Antibiotics for exacerbations of chronic obstructive pulmonary disease. Cochrane Database Syst Rev 2006;(2):CD004403.|
|14.||Anthonisen NR, Manfreda J, Warren CP, Hershfield ES, Harding GK, Nelson NA. Antibiotic therapy in exacerbations of chronic obstructive pulmonary disease. Ann Intern Med 1987;106:196–204.|
|15.||van d V, Monninkhof E, van der Palen J, Zielhuis G, van HC, Hendrix R. Clinical predictors of bacterial involvement in exacerbations of chronic obstructive pulmonary disease. Clin Infect Dis 2004;39:980–986.|
|16.||Aaron SD, Vandemheen KL, Hebert P, Dales R, Stiell IG, Ahuja J, Dickinson G, Brison R, Rowe BH, Dreyer J. Outpatient oral prednisone after emergency treatment of chronic obstructive pulmonary disease. N Engl J Med 2003;348:2618–2625.|
|17.||Davies L, Angus RM, Calverley PM. Oral corticosteroids in patients admitted to hospital with exacerbations of chronic obstructive pulmonary disease: a prospective randomised controlled trial. Lancet 1999;354:456–460.|
|18.||Niewoehner DE, Erbland ML, Deupree RH, Collins D, Gross NJ, Light RW, Anderson P, Morgan NA. Effect of systemic glucocorticoids on exacerbations of chronic obstructive pulmonary disease. Department of Veterans Affairs Cooperative Study Group. N Engl J Med 1999;340:1941–1947.|
|19.||Puhan MA, Vollenweider D, Latshang T, Steurer J, Steurer-Stey C. Exacerbations of chronic obstructive pulmonary disease: when are antibiotics indicated? A systematic review. Respir Res 2007;8:30.|
|20.||Puhan MA, Vollenweider D, Steurer J, Bossuyt PM, Ter RG. Where is the supporting evidence for treating mild to moderate chronic obstructive pulmonary disease exacerbations with antibiotics? A systematic review. BMC Med 2008;6:28.|
|21.||Rothberg MB, Pekow PS, Lahti M, Brody O, Skiest DJ, Lindenauer PK. Antibiotic therapy and treatment failure in patients hospitalized for acute exacerbations of chronic obstructive pulmonary disease. JAMA 2010;303:2035–2042.|
|22.||Han MK, Agusti A, Calverley PM, Celli BR, Criner G, Curtis JL, Fabbri LM, Goldin JG, Jones PW, Macnee W. Chronic obstructive pulmonary disease phenotypes: the future of COPD. Am J Respir Crit Care Med 2010;182:598–604.|
|23.||Weatherall M, Shirtcliffe P, Travers J, Beasley R. Use of cluster analysis to define COPD phenotypes. Eur Respir J 2010;36:472–474.|
|24.||Tashkin DP. Frequent exacerbations of chronic obstructive pulmonary disease: a distinct phenotype? N Engl J Med 2010;363:1183–1184.|
|25.||Rodriguez-Roisin R. Toward a consensus definition for COPD exacerbations. Chest 2000; 117(5, Suppl 2)398S–401S.|
|26.||Bhowmik A, Seemungal TA, Sapsford RJ, Devalia JL, Wedzicha JA. Comparison of spontaneous and induced sputum for investigation of airway inflammation in chronic obstructive pulmonary disease. Thorax 1998;53:953–956.|
|27.||Brightling CE, Monterio W, Green RH, Parker D, Morgan MD, Wardlaw AJ, Pavord D. Induced sputum and other outcome measures in chronic obstructive pulmonary disease: safety and repeatability. Respir Med 2001;95:999–1002.|
|28.||Guyatt G. Measuring health status in chronic airflow limitation. Eur Respir J 1988;1:560–564.|
|29.||Health Protection Agency 2009. Investigation of bronchoalveolar lavage, sputum and associated specimens. National Standard Method BSOP 2009;57:3.|
|30.||Pye A, Stockley RA, Hill SL. Simple method for quantifying viable bacterial numbers in sputum. J Clin Pathol 1995;48:719–724.|
|31.||Creer DD, Dilworth JP, Gillespie SH, Johnston AR, Johnston SL, Ling C, Patel S, Sanderson G, Wallace PG, McHugh TD. Aetiological role of viral and bacterial infections in acute adult lower respiratory tract infection (LRTI) in primary care. Thorax 2006;61:75–79.|
|32.||Bisgaard H, Zielen S, Garcia-Garcia ML, Johnston SL, Gilles L, Menten J, Tozzi CA, Polos P. Montelukast reduces asthma exacerbations in 2- to 5-year-old children with intermittent asthma. Am J Respir Crit Care Med 2005;171:315–322.|
|33.||Kelly MM, Keatings V, Leigh R, Peterson C, Shute J, Venge P, Djukanovic R. Analysis of fluid-phase mediators. Eur Respir J Suppl 2002;37:24s–39s.|
|34.||Hurst JR, Donaldson GC, Perera WR, Wilkinson TM, Bilello JA, Hagan GW, Vessey RS, Wedzicha JA. Use of plasma biomarkers at exacerbation of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2006;174:867–874.|
|35.||Bozinovski S, Hutchinson A, Thompson M, Macgregor L, Black J, Giannakis E, Karlsson AS, Silvestrini R, Smallwood D, Vlahos R, et al.. Serum amyloid A is a biomarker of acute exacerbations of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2008;177:269–278.|
|36.||Hurst JR, Perera WR, Wilkinson TM, Donaldson GC, Wedzicha JA. Systemic and upper and lower airway inflammation at exacerbation of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2006;173:71–78.|
|37.||Burgel PR, Paillasseur JL, Caillaud D, Tillie-Leblond I, Chanez P, Escamilla R, Court-Fortune I, Perez T, Carre P, Roche N. Clinical COPD phenotypes: a novel approach using principal component and cluster analyses. Eur Respir J 2010;36:531–539.|
|38.||Hurst JR, Vestbo J, Anzueto A, Locantore N, Mullerova H, Tal-Singer R, Miller B, Lomas DA, Agusti A, Macnee W, et al.. Susceptibility to exacerbation in chronic obstructive pulmonary disease. N Engl J Med 2010;363:1128–1138.|
|39.||Hospers JJ, Schouten JP, Weiss ST, Postma DS, Rijcken B. Eosinophilia is associated with increased all-cause mortality after a follow-up of 30 years in a general population sample. Epidemiology 2000;11:261–268.|
|40.||Hospers JJ, Schouten JP, Weiss ST, Rijcken B, Postma DS. Asthma attacks with eosinophilia predict mortality from chronic obstructive pulmonary disease in a general population sample. Am J Respir Crit Care Med 1999;160:1869–1874.|
|41.||Rosell A, Monso E, Soler N, Torres F, Angrill J, Riise G, Zalacain R, Morera J, Torres A. Microbiologic determinants of exacerbation in chronic obstructive pulmonary disease. Arch Intern Med 2005;165:891–897.|
|42.||Taylor AE, Finney-Hayward TK, Quint JK, Thomas CM, Tudhope SJ, Wedzicha JA, Barnes PJ, Donnelly LE. Defective macrophage phagocytosis of bacteria in COPD. Eur Respir J 2010;35:1039–1047.|
|43.||Conway MA, Kefala K, Wilkinson TS, Moncayo-Nieto OL, Dhaliwal K, Farrell L, Walsh TS, Mackenzie SJ, Swann DG, Andrews PJ, et al.. Diagnostic importance of pulmonary interleukin-1beta and interleukin-8 in ventilator-associated pneumonia. Thorax 2010;65:201–207.|
|44.||Quint JK, Donaldson GC, Goldring JJ, Baghai-Ravary R, Hurst JR, Wedzicha JA. Serum IP-10 as a biomarker of human rhinovirus infection at exacerbation of COPD. Chest 2010;137:812–822.|
|45.||Sethi S. Coinfection in exacerbations of COPD: a new frontier. Chest 2006;129:223–224.|
|46.||Sethi S. The problems of meta-analysis for antibiotic treatment of chronic obstructive pulmonary disease, a heterogeneous disease: a commentary on Puhan et al. BMC Med 2008;6:29.|
|47.||Sethi S, Evans N, Grant BJ, Murphy TF. New strains of bacteria and exacerbations of chronic obstructive pulmonary disease. N Engl J Med 2002;347:465–471.|
|48.||Hilty M, Burke C, Pedro H, Cardenas P, Bush A, Bossley C, Davies J, Ervine A, Poulter L, Pachter L, et al.. Disordered microbial communities in asthmatic airways. PLoS ONE 2010;5:e8578.|
Supported by the Medical Research Council (United Kingdom) and AstraZeneca jointly as a “Biomarker Call Project”; C.E.B. is a Wellcome Trust Senior Clinical Fellow, and GlaxoSmithKline supported the measurement of surfactant protein D. The research was performed in laboratories partly funded by the European Regional Development Fund (ERDF 05567). The Medical Research Council, Wellcome Trust, and the European Regional Development Fund had no involvement in the design of the study, data collection, analysis and interpretation of the data, in the writing of the manuscript, or in the decision to submit the manuscript.
Author Contributions: S.M. and S.T. were involved in the recruitment of volunteers and in data collection. C.R., V.M., K.H., H.P., A.D., and K.L. were involved in data collection and interpretation. M.M., P.R., P.D., P.N., M.J., and M.S. were involved in study design, data collection, and interpretation. R.H.G. and P.H. were involved in study design and data interpretation. M.R.B., D.A.L., S.L.J., P.V., and I.D.P. were involved in the design of the study, data collection, and interpretation. M.B. and C.E.B. were involved in the study design, volunteer recruitment, data collection, data interpretation, and data analysis, and had full access to the data and are responsible for the integrity of the data and final decision to submit. All authors contributed to the writing of the manuscript and have approved the final version for submission.
Originally Published in Press as DOI: 10.1164/rccm.201104-0597OC on June 16, 2011
This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org