Abstract
Rationale: Increased bacterial concentration (load) in the lower airways and new bacterial strain acquisition have been posited as mechanisms for chronic obstructive pulmonary disease (COPD) exacerbations. Bacterial concentrations are higher during exacerbation than during stable disease; however, these studies are cross sectional and devoid of strain typing.
Objectives: To determine if the increased bacterial concentrations function as a separate mechanism for exacerbation induction independent of new strain acquisition.
Methods: In a prospective, longitudinal cohort of patients with COPD, the relationship between exacerbation occurrence, sputum bacterial concentrations, and new strain acquisition was examined.
Measurements and Main Results: Clinical information, quantitative sputum cultures, and molecular typing of potential bacterial pathogen isolates. Over 81 months, 104 subjects completed 3,009 clinic visits, 560 (19.6%) during exacerbations and 2,449 (80.4%) during stable disease. Among preexisting strains, sputum concentrations of Nontypeable Haemophilus influenzae and Haemophilus haemolyticus were not different in exacerbation versus stable disease. Moraxella catarrhalis (stable, 108.38 ± 0.13 [mean ± SEM] vs. exacerbation, 107.78 ± 0.26; p = 0.02) and Streptococcus pneumoniae (stable, 108.42 ± 0.21 vs. exacerbation, 107.76 ± 0.52; p = 0.07) concentrations were lower during exacerbations compared with stable periods. Concentrations of new strains of H. influenzae (stable, 107.28 ± 0.15 vs. exacerbation, 107.76 ± 0.17; p = 0.04) and M. catarrhalis (stable, 107.85 ± 0.15 vs. exacerbation, 108.37 ± 0.14; p = 0.02), were increased during exacerbations; however, the differences were small.
Conclusions: Change in bacterial load is unlikely to be an important mechanism for exacerbations. Better understanding of the host–pathogen interaction, rather than enumerating bacteria in respiratory samples, is required to provide new insights into bacterial infection in COPD.
Bacterial concentrations in the lower airways are higher during exacerbations of chronic obstructive pulmonary disease; however, it is not known if the increased bacterial concentrations are independent of acquisition of new strains and function as a separate mechanism for exacerbation induction.
Sputum concentrations of preexisting bacterial strains were not higher during exacerbations. Among new strains, small increases were seen. These results demonstrate that change in bacterial load is unlikely to be an important mechanism for exacerbations of COPD.
Recently, new strain acquisition has been demonstrated as a mechanism of exacerbations; however, exacerbations are observed in the presence of preexisting bacterial strains—those isolated from the sputum before the onset of exacerbations (5). Furthermore, before the discovery of the relationship between new strain acquisition and exacerbations, the prevailing hypothesis of exacerbation mechanism was increased bacterial concentration (load) in the lower airways, with consequent increase in airway inflammation resulting in increased respiratory symptoms (6, 7). In fact, in several recent investigations, bacterial concentrations have been shown to be higher during exacerbation than during stable disease (7–9). However, many of these studies are cross sectional in nature, in which patients with exacerbations are compared with those with stable disease (8). In these studies, as well as in the studies that have employed longitudinal sampling, strain typing has not been performed (7–9). Therefore, from these studies, it is not possible to determine the relationship between increased bacterial concentrations observed at exacerbation and acquisition of new strains, and elucidate whether these are separate mechanisms for exacerbation induction.
We hypothesized that increased bacterial concentrations in the lower airway of patients with COPD is associated with exacerbation, independent of the acquisition of new strains of bacteria. In a prospective, longitudinal study of the dynamics of bacterial infection in COPD, quantitative bacterial cultures of sputum samples obtained from a cohort of patients with COPD were performed on a monthly basis, as well as during exacerbations. The relationship between exacerbation occurrence, bacterial concentrations in sputum, and acquisition of new strains was examined. The following hypotheses were tested: (1) Are bacterial concentrations in sputum related to the presence of exacerbation? (2) Are bacterial concentrations higher in exacerbation after accounting for the characterization of the strain as new or preexisting? If an increase in bacterial concentration is an independent mechanism for exacerbations, one would anticipate that there would be a relationship between higher bacterial concentrations and exacerbations among both new and preexisting strains. (3) Is there a relationship between bacterial concentrations and the occurrence of exacerbation after accounting for the acquisition of new strains as a confounding factor? Some of the results of these studies have been previously reported in the form of an abstract (10).
The study protocol was approved by the Veterans Affairs Western New York Healthcare System institutional review board. This prospective cohort study has been described earlier in several publications (5, 11, 12). A total of 104 patients with COPD were enrolled between March 1994 and December 2000. The patients were seen monthly and whenever they had symptoms suggestive of an exacerbation. At each visit, clinical information, sputum, and serum samples were collected. Details of the study clinic are described in the online supplement.
Sputum sample processing and potential bacterial pathogen quantification is described in the online supplement. Potentially pathogenic bacteria included: Haemophilus spp., including Nontypeable Haemophilus influenzae, H. haemolyticus, and H. parainfluenzae, Moraxella catarrhalis, Streptococcus pneumoniae, Pseudomonas aeruginosa, Staphylococcus aureus, and gram-negative enteric rods. Nontypeable H. influenzae is hereafter referred to as simply H. influenzae.
Accurate colony counts could be determined from concentrations from 104 to 109 cfu/ml of homogenized sputum. Exact colony counts and, therefore, exact sputum concentrations were not available for potentially pathogenic bacteria in all sputum samples. The most common reason was that the potentially pathogenic bacterial colonies were mixed with normal flora and were, therefore, not discrete enough for accurate counts. Another reason was the presence of too many colonies to count on the highest dilution. Only rarely was the colony count data not recorded. For each of these instances, estimated counts of 1 log range were obtained as follows: for instances in which too many colonies were present at the highest dilution, bacterial colony count was estimated to lie between 1 × 109 and 1 × 1010 cfu/ml. For the instances in which potentially pathogenic bacterial colonies were not discrete enough for accurate counts, bacterial colony count was estimated to lie between the 1 log range corresponding to 1 and 10 colonies for the plate with the highest dilution of sputum, where at least one colony was seen. For missing data, the same estimates described above were used.
In certain circumstances, neither exact nor estimated counts were available. This included many instances of isolation of S. pneumoniae, when present at relatively low concentrations, because of the colony morphology resemblance between this pathogen and colonies of commensal α-streptococci. In addition, colony counts were not performed for gram-negative bacilli (including Pseudomonas) and Staphylococcus spp.
Molecular typing of strains of H. influenzae, S. pneumoniae, and M. catarrhalis was performed as described previously (5). Each strain was categorized as old or new based on molecular typing. A new strain was a strain that had not been previously isolated from sputum samples obtained from previous visits since the patient's enrollment. An old strain was one that had been isolated from sputum obtained from previous visits since the patient's enrollment. On further characterization of H. influenzae strains, we found that a substantial proportion were actually H. haemolyticus, many of which were nonhemolytic (13). These strains were analyzed separately from the H. influenzae strains.
Two different methods of data analysis were used. In the first instance, all colony counts were pooled and concentrations compared for each species for new and preexisting strains, separately. Concentrations used in all analyses were logarithm transformed. For estimated counts, logarithm-transformed concentrations were calculated at a random point of the log interval estimate. Because of multiple isolations of individual strains of pathogens from individual patients, generalized estimating equations (GEEs) were used for this analysis (14). The outcome variable was the presence of an exacerbation. The input variables included bacterial concentrations and whether the strain was new or preexisting.
In a second analysis, instances in which the same strain was isolated repeatedly from the same patient were determined. Of these instances, isolation of the same strain from a patient with an exacerbation and during stable state was analyzed further with paired t tests. If more than one visit in the stable state or exacerbation state were described for an individual strain, average values for these visits were used.
From 1994 to 2000, over 81 months, 104 subjects completed 3,009 clinic visits. The demographic and clinical characteristics of these subjects are described in Table 1. Of the 3,009 clinic visits, 560 (19.6%) were during exacerbations and 2,449 (80.4%) were during stable COPD. Table 2 shows the number of isolates for each of the potential pathogens isolated from March 1994 to December 2000, and the number of isolates for which titers (exact and estimated) were available and were included in this study. Isolates for which titers were excluded or not available were as follows: (1) isolates obtained during repeated visits for the same exacerbation: H. influenzae, 8; H. haemolyticus, 1; S. pneumoniae, 1; M. catarrhalis, 1; and (2) isolates for which adequate information was not available to obtain either an exact or estimated titer: S. pneumoniae, 33; M. catarrhalis, 6; H. parainfluenzae, 5.
Characteristics | Value |
|---|---|
| Mean age (range), yr | 67.1 (45–85) |
| Sex, n | |
| Male | 102 |
| Female | 2 |
| Race, n | |
| White | 87 |
| African American | 17 |
| Mean time since diagnosis (range), yr | 11.2 (0–54) |
| Smoking status on enrollment, n | |
| Ex-smokers | 70 |
| Current smokers | 34 |
| Mean smoking (range), pack-years | 82.6 (10–185) |
| Mean FEV1, L (range) | 1.56 (0.47–4.07) |
| Mean FEV1% predicted (range) | 46.0 (15–99) |
| GOLD severity, no. of subjects | |
| Chronic bronchitis only | 5 |
| 1 | 1 |
| 2 | 35 |
| 3 | 41 |
| 4 | 22 |
Pathogen | No. of Isolates | No. of Isolates Included* | No. with Exact Concentrations | No. with Estimated Concentrations |
|---|---|---|---|---|
| HI | 375 | 367 | 262 | 105 |
| HH | 190 | 189 | 150 | 39 |
| SP | 90 | 56 | 38 | 18 |
| MC | 194 | 187 | 140 | 47 |
| HP | 1,551 | 1,546 | 1,330 | 216 |
Different bacterial species may be present at different concentrations in the airway in COPD (Figure 1). This was determined by pair-wise comparisons of the mean concentrations among species. S. pneumoniae and M. catarrhalis exhibited the highest concentrations, significantly greater than all the other bacterial species, but equivalent to each other. Sputum concentrations of H. haemolyticus and H. parainfluenzae were significantly lower than all the other bacterial species, but equivalent to each other. Concentrations of H. influenzae were significantly higher than those of the other two Haemophilus species.
Figure 1. Sputum concentrations for each pathogen (mean ± SEM) in all sputum samples. Double-headed arrows denote significant differences (p < 0.05; generalized estimating equation with Tukey's adjustment for multiple comparisons). Pathogen: HH = Haemophilus haemolyticus (n = 189); HI = nontypeable Haemophilus influenzae (n = 367); HP = Haemophilus parainfluenzae (n = 1546); MC = Moraxella catarrhalis (n = 187); SP = Streptococcus pneumoniae (n = 56).
[More] [Minimize]Generalized estimating equations were constructed for each of the following bacterial pathogens: H. influenzae, H. haemolyticus, H. parainfluenzae, S. pneumoniae, and M. catarrhalis. Strain typing information was available for all these pathogens, with the exception of H. parainfluenzae.
We first determined whether bacterial concentrations in sputum are related to the presence of exacerbation. If all strains, new and preexisting, were examined together, we found no significant difference between concentrations in sputum during stable disease and exacerbation for H. influenzae, H. haemolyticus, and M. catarrhalis (Figure 2). An inverse relationship between exacerbation occurrence and bacterial concentrations in sputum was seen for S. pneumoniae and H. parainfluenzae. Concentrations of these pathogens were actually lower during exacerbations than during stable disease (S. pneumoniae: stable, 108.42 ± 0.17 [mean ± SEM] vs. exacerbation, 107.62 ± 0.36, p = 0.048; H .parainfluenzae: stable, 106.43 ± 0.02 vs. exacerbation, 106.29 ± 0.05; p = 0.02), although the differences between exacerbation and stable visits were less than a log (10-fold).
Figure 2. Bacterial concentration (mean ± SEM) in sputum for each pathogen in stable (striped bars) and exacerbation (dotted bars) visits. *Significant differences (p < 0.05). Pathogen: HH = H. hemolyticus (n = 158 stable, n = 31 exacerbation); HI = nontypeable H. influenzae (n = 288 stable, n = 79 exacerbation); HP = H. parainfluenzae (n = 1282 stable, n = 264 exacerbation); MC = M. catarrhalis (n = 120 stable, n = 67 exacerbation); SP = S. pneumoniae (n = 40 stable, n = 16 exacerbation).
[More] [Minimize]To test the major hypothesis of this study, we then determined if bacterial concentrations were associated with the occurrence of exacerbation after controlling for new strain acquisition using GEE models. In one approach, concentrations in sputum during stable disease and exacerbation were compared separately for new and preexisting isolates. As shown in Table 3, among new strains of H. influenzae and M. catarrhalis, but not for H. haemolyticus or S. pneumoniae, increased concentrations were indeed seen during exacerbation compared with during stable visits (H. influenzae: stable, 107.28 ± 0.15 vs. exacerbation, 107.76 ± 0.17; p = 0.04; M. catarrhalis: stable, 107.85 ± 0.15 vs. exacerbation, 108.37 ± 0.14; p = 0.02). However, the difference in concentrations between stable and exacerbation visits in each instance was about 0.5 log.
Pathogen/Clinical Status | No. of Strains | Mean Log Concentration (SEM) | p Value |
|---|---|---|---|
| New strains | |||
| HI | |||
| Stable | 78 | 7.28 (0.15) | 0.04 |
| Exacerbation | 44 | 7.76 (0.17) | |
| MC | |||
| Stable | 63 | 7.85 (0.15) | 0.02 |
| Exacerbation | 51 | 8.37 (0.14) | |
| SP | |||
| Stable | 16 | 8.32 (0.30) | 0.26 |
| Exacerbation | 7 | 7.53 (0.53) | |
| HH | |||
| Stable | 82 | 6.42 (0.10) | 0.79 |
| Exacerbation | 21 | 6.38 (0.112) | |
| Preexisting strains | |||
| HI | |||
| Stable | 210 | 7.71 (0.09) | 0.31 |
| Exacerbation | 35 | 7.99 (0.20) | |
| MC | |||
| Stable | 57 | 8.38 (0.13) | 0.02 |
| Exacerbation | 16 | 7.76 (0.26) | |
| SP | |||
| Stable | 24 | 8.42 (0.21) | 0.07 |
| Exacerbation | 9 | 7.76 (0.52) | |
| HH | |||
| Stable | 76 | 6.45 (0.10) | 0.98 |
| Exacerbation | 10 | 6.38 (0.29) |
Among preexisting strains, no differences in sputum concentrations of H. influenzae and H. haemolyticus were found during exacerbation compared with stable disease (Table 3). M. catarrhalis was actually present at significantly lower concentrations during exacerbations as compared with stable periods (stable, 108.38 ± 0.13 vs. exacerbation, 107.78 ± 0.26; p = 0.02), and a similar trend was seen for S. pneumoniae (stable, 108.42 ± 0.21 vs. exacerbation, 107.76 ± 0.52; p = 0.07) (Table 3). As observed earlier, all differences were again less than a log.
In a second analytic approach, a GEE model considering the occurrence of exacerbation as the outcome with concentration as independent variable and new strain acquisition as a confounder was determined for each pathogen. In these models, bacterial titers of H. haemolyticus (p = 0.80) and M. catarrhalis (p = 0.20) had no relationship to the occurrence of exacerbation. Higher concentrations of H. influenzae were associated with exacerbation after accounting for new strain acquisition (p = 0.03; odds ratio, 1.27; 95% confidence interval, 1.03–1.56), but, for S. pneumoniae, lower concentrations were found at exacerbation (p = 0.02; odds ratio, 0.88; 95% confidence interval, 0.80–0.98). This analysis confirmed some of the findings of the GEE models described previously here.
These analyses could not be conducted for H. parainfluenzae, as strain typing has not been performed for this potential pathogen.
In our longitudinal collection of samples, we often encountered a strain that would be isolated during exacerbations and during stable visits from the same patient. These instances were an even more rigorous test of the hypothesis that bacterial concentrations are associated with exacerbations, as they represent a change in clinical status while the host and pathogen apparently remain the same. Results shown in Table 4 demonstrate that only H. influenzae (stable, 107.51 ± 0.19 vs. exacerbation, 108.08 ± 0.19; p = 0.02) was isolated at about 0.5 log higher concentration during exacerbation than during stable disease, with a trend toward lower concentrations during exacerbation for S. pneumoniae, and no differences observed for M. catarrhalis and H. haemolyticus. This analysis confirmed some of the findings of the GEE models described previously here.
Pathogen/Clinical Status | No. of Strains | Mean Concentration (SEM) | p Value |
|---|---|---|---|
| HI | |||
| Stable | 34 | 7.51 (0.19) | 0.02 |
| Exacerbation | 34 | 8.08 (0.19) | |
| MC | |||
| Stable | 15 | 7.79 (0.28) | 0.67 |
| Exacerbation | 15 | 7.89 (0.29) | |
| SP | |||
| Stable | 7 | 8.81 (0.34) | 0.14 |
| Exacerbation | 7 | 8.50 (0.33) | |
| HH | |||
| Stable | 13 | 6.09 (0.15) | 0.16 |
| Exacerbation | 13 | 6.38 (0.16) |
If increases in bacterial concentration (or load) in the lower airway was an independent mechanism of inducing exacerbations, we should have found increased bacterial concentrations during exacerbation when compared with stable disease among the preexisting strains (5). On the contrary, we found either no difference or, in some instances, even lower concentrations of these preexisting strains during exacerbation, demonstrating that change in bacterial load is unlikely to be an important mechanism for exacerbations. One can speculate that these exacerbations are likely caused by other pathogens (viruses or atypical bacteria) or environmental stimuli, and the resultant inflammatory response nonspecifically inhibits the preexisting strains and, therefore, is responsible for lowering their concentration (15).
Increased concentrations of H. influenzae and M. catarrhalis were seen during exacerbations associated with new strains in comparison to when these new strains were causing colonization. However, this difference in bacterial load, and similar differences observed in other studies, although statistically significant, is unlikely to be of enough magnitude to be biologically significant (9). These differences are within 1 log (10-fold), whereas the total bacterial load in the airways is in the order of 7–8 logs. Therefore, these differences are actually quite small; for example, a 0.5 log difference is 7% of the total bacterial load (16). It is likely that these differences in bacterial load reflect the outcome of the host–pathogen interaction rather than an independent mechanism of exacerbation.
The strengths of this study are inclusion of molecular tying and longitudinal sampling. Molecular typing allowed discrimination of new strains from preexisting strains, as well as providing us with the ability to account for new strain acquisition when examining the relationship of bacterial concentrations to exacerbation. Our results were different from those of previous studies, which have found increased bacterial concentrations during exacerbations of COPD as compared with stable disease, but did not include molecular typing of bacterial strains, demonstrating the importance of accounting for strain acquisition (7–9). Longitudinal sampling in this study meant that the populations of patients contributing exacerbation and stable samples were the same. Therefore, potential confounding by different baseline characteristics of the populations being compared in previous studies with cross-sectional design was avoided (8). The paired analysis of patients when they have had the same strain isolated during exacerbation and stable disease is especially powerful in controlling for confounding variables, and did not demonstrate an effect of bacterial load, with the exception of a small difference in the case of H. influenzae.
We did find differences in concentrations of pathogens among species. Pathogens with the most clearly established role in exacerbations of COPD, H. influenzae, S. pneumoniae, and M. catarrhalis, were present in higher concentrations compared with pathogens that are more likely to be colonizers (H. parainfluenzae and H. haemolyticus) (11–13, 17, 18). This may reflect differences in virulence and adaptation to the human respiratory tract among the pathogenic species. These differences among potential pathogens should be taken in to account when interpreting studies examining bacterial concentrations and airway inflammation in COPD (7, 19).
Limitations of our study include the use of estimated concentrations in several instances instead of exact bacterial concentrations. However, our estimate is within a log of the actual concentration, and the statistical methods employed take in to account that these are estimates. Another limitation is that the bacterial concentrations were measured in sputum samples rather than bronchoscopically obtained samples, such as protected specimen brushings or bronchoalveolar lavage (8, 20, 21). The necessity for repeated sampling in our study design made sputum samples the only practical way to obtain these measurements. Our data are robust for Haemophilus spp. and M. catarrhalis; however, because of the limited number of observations with the pneumococcus, more observations are needed with this pathogen. In addition, other potential pathogens in exacerbations of COPD, such as Pseudomonas aeruginosa and Staphylococcus aureus, are not addressed in this study. The presence of viruses and atypical bacteria was not determined in this study. Such information could have added to the understanding of the mechanism of exacerbations, especially those with preexisting bacterial strains.
Studies of airway inflammation (unpublished observations) and immune response seem to support the concept that preexisting strains of the bacteria studied in this work are an infrequent cause of exacerbations (17). This study supports this concept, and, even if preexisting strains were to cause exacerbations, the mechanism does not appear to be an increase in bacterial load. Alternative mechanisms by which preexisting strains could cause exacerbations include alteration of their antigenic structure to evade the immune response or alteration of the airway milieu by another infection (e.g., a virus that alters the interaction between the host and the colonizing pathogen) (9, 16, 22). Such a process would be consistent with the multiple hit hypothesis recently proposed to explain variations in inflammation in airways diseases such as COPD (23). These mechanisms need investigation to further our understanding of host–pathogen interaction between colonizing bacterial strains and the host in COPD. However, this study does not preclude a pathogenic role for bacterial colonization in causing airway inflammation and inducing ongoing lung damage (19, 24, 25).
Are there clinical implications of the findings of this study? Quantitative cultures of sputum are rarely done in the clinical management of COPD. Semiquantitative sputum cultures are often performed and reported. Results of this study imply that these semiquantitative results are not useful in determining the etiologic role of the isolated pathogen. This study has obvious connotations for future research in this field. The host–pathogen interaction that underlies exacerbations of COPD is more complicated than simple changes in concentrations of bacteria. Studies limited to isolating and enumerating bacteria from respiratory samples are unlikely to provide us with new insights into bacterial infection in COPD. Better understanding of the host–pathogen interaction in COPD, including virulence factors of the bacterial pathogen and the alterations in the innate and adaptive lung defenses that allow bacteria to persist in the lower airway, are more likely to be fruitful.
The authors thank Adeline Thurston for secretarial assistance.
| 1. | Soler-Cataluna JJ, Martinez-Garcia MA, Roman Sanchez P, Salcedo E, Navarro M, Ochando R. Severe acute exacerbations and mortality in patients with chronic obstructive pulmonary disease. Thorax 2005;60:925–931. |
| 2. | Andersson F, Borg S, Jansson SA, Jonsson AC, Ericsson A, Prutz C, Ronmark E, Lundback B. The costs of exacerbations in chronic obstructive pulmonary disease (COPD). Respir Med 2002;96:700–708. |
| 3. | Veeramachaneni SB, Sethi S. Pathogenesis of bacterial exacerbations of COPD. COPD 2006;3:109–115. |
| 4. | Sethi S, Murphy TF. Acute exacerbations of chronic bronchitis: new developments concerning microbiology and pathophysiology–impact on approaches to risk stratification and therapy. Infect Dis Clin North Am 2004;18:861–882. |
| 5. | Sethi S, Evans N, Grant BJB, Murphy TF. Acquisition of a new bacterial strain and occurrence of exacerbations of chronic obstructive pulmonary disease. N Engl J Med 2002;347:465–471. |
| 6. | Gompertz S, O'Brien C, Bayley DL, Hill SL, Stockley RA. Changes in bronchial inflammation during acute exacerbations of chronic bronchitis. Eur Respir J 2001;17:1112–1119. |
| 7. | 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. |
| 8. | 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. |
| 9. | Wilkinson TM, Hurst JR, Perera WR, Wilks M, Donaldson GC, Wedzicha JA. Effect of interactions between lower airway bacterial and rhinoviral infection in exacerbations of COPD. Chest 2006;129:317–324. |
| 10. | Sethi S, Eschberger K, Lobbins P, Sethi RS, Agarwal S, Grant BJ, Murphy TF. Sputum bacterial titers and acute exacerbations of COPD [abstract]. American Thoracic Society Annual Meeting, San Diego, CA; 2005. Poster j84, p. a404. |
| 11. | Sethi S, Muscarella K, Evans N, Klingman KL, Grant BJB, Murphy TF. Airway inflammation and etiology of acute exacerbations of chronic bronchitis. Chest 2000;118:1557–1565. |
| 12. | Murphy TF, Brauer AL, Grant BJ, Sethi S. Moraxella catarrhalis in chronic obstructive pulmonary disease: Burden of disease and immune response. Am J Respir Crit Care Med 2005;172:195–199. |
| 13. | Murphy TF, Brauer AL, Sethi S, Kilian M, Cai X, Lesse AJ. Haemophilus haemolyticus: a human respiratory tract commensal to be distinguished from Haemophilus influenzae. J Infect Dis 2007;195:81–89. |
| 14. | Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986;42:121–130. |
| 15. | Aaron SD, Angel JB, Lunau M, Wright K, Fex C, Le Saux N, Dales RE. Granulocyte inflammatory markers and airway infection during acute exacerbation of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2001;163:349–355. |
| 16. | Sethi S. Coinfection in exacerbations of COPD: a new frontier. Chest 2006;129:223–224. |
| 17. | Sethi S, Wrona C, Grant BJB, Murphy TF. Strain-specific immune response to Haemophilus influenzae in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2004;169:448–453. |
| 18. | Bogaert D, van der Valk P, Ramdin R, Sluijter M, Monninkhof E, Hendrix R, de Groot R, Hermans PW. Host–pathogen interaction during pneumococcal infection in patients with chronic obstructive pulmonary disease. Infect Immun 2004;72:818–823. |
| 19. | Hill AT, Campbell EJ, Hill SL, Bayley DL, Stockley RA. Association between airway bacterial load and markers of airway inflammation in patients with stable chronic bronchitis. Am J Med 2000;109:288–295. |
| 20. | Monso E, Ruiz J, Rosell A, Manterola J, Fiz J, Morera J, Ausina V. Bacterial infection in chronic obstructive pulmonary disease: a study of stable and exacerbated outpatients using the protected specimen brush. Am J Respir Crit Care Med 1995;152:1316–1320. |
| 21. | Soler N, Torres A, Ewig S, Gonzalez J, Celis R, El-Ebiary M, Hernandez C, Rodriguez-Roisin R. Bronchial microbial patterns in severe exacerbations of chronic obstructive pulmonary disease (COPD) requiring mechanical ventilation. Am J Respir Crit Care Med 1998;157:1498–1505. |
| 22. | Hiltke TJ, Sethi S, Murphy TF. Sequence stability of the gene encoding outer membrane protein p2 of nontypeable Haemophilus influenzae in the human respiratory tract. J Infect Dis 2002;185:627–631. |
| 23. | Pavord ID, Birring SS, Berry M, Green RH, Brightling CE, Wardlaw AJ. Multiple inflammatory hits and the pathogenesis of severe airway disease. Eur Respir J 2006;27:884–888. |
| 24. | Sethi S, Maloney J, Grove L, Wrona C, Berenson CS. Airway inflammation and bronchial bacterial colonization in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2006;173:991–998. |
| 25. | Wilkinson TM, Patel IS, Wilks M, Donaldson GC, Wedzicha JA. Airway bacterial load and FEV1 decline in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2003;167:1090– 1095. |
| 26. | Global Initiative for Chronic Obstructive Lung Disease. Available from: http://www.goldcopd.com (accessed March 10, 2007). |
