Bacterial and Viral Infection in Patients Hospitalized for Acute Exacerbation of Chronic Obstructive Pulmonary Disease: Implication for Antimicrobial Management and Clinical Outcome

Abstract

Patients with chronic obstructive pulmonary disease (COPD) exhibit frequent acute exacerbations (AE). The objectives of this study were first to evaluate the prevalence of pathogens associated to these episodes by combining conventional bacteriology and multiplex viral and bacterial PCR assays in sputum specimens, and second to determine whether C-reactive protein (CRP) value and clinical outcome could be influenced by the type of microbial agent(s) recovered from these samples. A cohort of 84 Tunisian patients hospitalized at the emergency room for AECOPD was investigated prospectively for the semi-quantitative detection of bacteria by conventional culture (the threshold of positivity was of 10⁷ CFU/ml) and for the detection of viral genome and DNA of atypical bacteria by quantitative PCR using two commercial multiplex respiratory kits (Seegene and Fast-track). The two kits exhibited very similar performances although the Seegene assay was a bit more sensitive. A large number and variety of pathogens were recovered from the sputum samples of these 84 patients, including 15 conventional bacteria, one Chlamydia pneumoniae and 63 respiratory viruses, the most prevalent being rhinoviruses (n = 33) and influenza viruses (n = 13). From complete results available for 74 patients, the presence of bacteria was significantly associated with risk of recurrence at 6 and 12 months post-infection. The combination of these different markers appears useful for delineating correctly the antimicrobial treatment and for initiating a long-term surveillance in those patients with high risk of recurrent exacerbation episodes. A prospective study is required for confirming the benefits of this strategy aimed at improving the stewardship of antibiotics.

Keywords:

• Chronic obstructive pulmonary disease
• acute exacerbation
• multiplex PCR testing
• antimicrobial stewardship
• C-reactive protein
• clinical outcome

Abbreviations

AE	=	acute exacerbation
AECOPD	=	acute exacerbation of chronic obstructive pulmonary disease
COPD	=	chronic obstructive pulmonary disease
CRP	=	C-reactive protein
FTD	=	fast track diagnostics
PaCO2	=	partial pressure of carbon dioxide in arterial blood
PaO2	=	partial pressure of oxygen in arterial blood
qPCR	=	quantitative polymerase chain reaction
SD	=	standard deviation
SG	=	Seegene
WBC	=	white cell blood count

Introduction

Chronic obstructive pulmonary disease (COPD) is an inflammatory process characterized by progressive airflow limitation and destruction of the parenchyma [1]. Patients with COPD exhibit frequent exacerbations [2]. Acute exacerbations of COPD (AECOPD) are characterized by increased sputum volume and purulence associated to worsening dyspnoea and cough in the absence of radiologic images of pneumonia [1]. They constitute the main cause of prolonged hospitalization in these patients and contribute to increase their morbidity and mortality rates [3]. They are favoured by air pollutants, allergens, irritants including smoking, climate changes and bacterial or viral pathogens.

Different biomarkers that express the early pan-airway inflammation of the respiratory tract may be useful to distinguish between AECOPD of bacterial, viral or non-infectious origin and could orientate the use of antibiotics at the early stage of hospitalization. These inflammatory biomarkers appear to be increased in the serum of patients with AECOPD. C-reactive protein (CRP) is currently the best studied biomarker in this setting. It is an acute-phase reactant secreted by the liver in response to infection, inflammation or tissue damage with well-documented sensitivity, and is commonly used for diagnosing infectious and inflammatory conditions, including in AECOPD [4].

Bacteria with respiratory tropism, including atypical respiratory pathogens such as Mycoplasma pneumoniae and Chlamydia pneumoniae, were estimated to account for 50–70% of AECOPD [5]. Ancient studies relying on serology or viral culture associated respiratory viruses to up to 30% of AECOPD [6]. With the recent development of more sensitive molecular tests such as real-time quantitative polymerase chain reaction (qPCR), the role of viruses as well as atypical fastidious-growing bacteria was better defined in AECOPD patients (for reviews, see [7,8]). The large and still increasing number of viruses associated with respiratory infection makes laboratory testing with individual virus assays challenging. In contrast, multiplex PCR assays that combine several individual targets in a single reaction are able to differentiate viral and bacterial respiratory infections and to identify co-infections in a time-saving and cost-effectiveness manner [9–11]. They are suitable for optimizing the use of antimicrobial drugs, which contributes to reduce the duration of hospital stay and to save money and antibiotic resistance [12].

In this study, we investigated prospectively a cohort of Tunisian patients hospitalized at the emergency room for AECOPD with two specific objectives: firstly evaluating the prevalence of different pathogens associated to these episodes by combining conventional bacterial quantitative culture and multiplex viral and bacterial qPCR assays in sputum specimens, and secondly determining whether CRP values and clinical outcome could be associated with the type of microbial agent(s) recovered from these samples. The final aim of the study was to propose an algorithm for optimizing the antimicrobial treatment in AECOPD patients at the early stage of hospitalization.

Subjects and methods

Study design

Inclusion criteria

The study was conducted prospectively from December 2014 to March 2015 in three University Hospitals (Monastir, Mahdia and Sousse) located in the South-East part of Tunisia. All the adult patients presenting consecutively at the Emergency room of these hospitals with AECOPD and requiring hospitalization were included into the study, as indicated on the flowchart of the study (Figure 1). The diagnosis of COPD was performed according to the criteria of the Global Initiative for Chronic Obstructive Lung Disease (GOLD) [1].

Figure 1. Flowchart of the study. COPD, chronic obstructive pulmonary disease; NAAT, nucleic acid amplification test.

AECOPD was defined as any sustained increase beyond patient’s baseline condition in any respiratory symptom that required a change of regular medication necessitating hospital admission. The diagnosis of AECOPD was first based on medical history including increase of respiratory symptoms (in particular dyspnoea) behind patient baselines conditions with the necessity to change regular medication, and hospital admission. Blood gas measures and the presence of a low blood pH (<7.5) were used as additional criteria of COPD exacerbation.

Exclusion criteria

Exclusion criteria were high fever (≥38.5 °C), prior antibiotic treatment for ≥24 h, extensive treatment with sytemic corticosteroids (>30 mg prednisolone equivalent dose for more than four days), radiological signs of pneumonia, history of severe AECOPD requiring mechanical ventilation at admission and other infectious disease requiring antibiotic therapy. The absence of radiological signs of pneumonia was confirmed in all patients by consensus of senior physicians; a second chest radiograph was done when needed. Patients were also excluded if they presented one of the following conditions: pregnancy or lactation, active drug or alcohol addiction, previous adverse reactions to quinolone derivatives, gastrointestinal trouble that could affect drug absorption, clinical evidence of hemodynamic compromise with need of vasoactive drugs, Glasgow coma scale <7.20, severe immunosuppession, severe renal failure (creatinine clearance >40 mL/min), chronic hepatic impairment or chronic lung disease other than COPD that could affect the clinical evaluation of the treatments.

Ethics

The research protocol was approved by the Ethical Committee of the University Hospital of Sousse. It was registered in the ClinicalTrials.gov database under the number NCT02067780. Written informed consent was obtained from all the patients or the nearest relative.

Management of patients

Management at entry

This study was a part of a bigger double-blinded randomized one aiming at comparing two lines of antimicrobial therapy: a two-day treatment by levofloxacin versus a seven-day treatment using the same molecule, in the prolongation of a similar study performed previously in the same setting [13]. The randomization into these two arms occurred at entry. The regulatory documents relative to this bigger double-blinded randomized clinical trial are provided as Supplementary Data.

The patients were treated according to procedures inspired from the current GOLD report [1] and described in the study mentioned above [13]. Briefly, all the patients were initially assigned to oxygen supplementation or non-invasive ventilation if blood pH was less than 7.35; those for whom this strategy failed were intubated and mechanically-ventilated in the assist-control mode. The patients received subcutaneous heparin, inhaled bronchodilators and enteral nutrition. Prophylactic treatment for stress ulcer was not systematic and intravenous sedatives were used only in intubated patients. No selective decontamination of the digestive tract was performed.

At entry, blood samples were taken systematically for standard biological management including white cell blood count (WBC), partial pressure of oxygen in arterial blood (PaO₂), partial pressure of carbon dioxide in arterial blood (PaCO₂), pH and CRP measurement. CRP concentrations were determined using an immunoturbidimetric assay (Roche Diagnostics, Indianapolis, IN, USA) consisting in an agglutination of the human CRP on latex particles coated with monoclonal anti-CRP antibody.

Microbiological management

Induced sputum was obtained at admission prior to any antimicrobial treatment according to previous recommendations including safety precautions [14]. A macroscopic examination of the specimen was performed to exclude salivary samples. Sputum samples were collected systematically at entry in sterile vials, an aliquot fraction being kept at −80 °C for further investigation. A gram stain of the sputum sample showing ten or fewer epithelial cells and more than 25 leucocytes per low-power field was required to demonstrate that the specimen was adequate for microbiological assessment. Semi-quantitative bacterial cultures were conducted at the Microbiology Laboratory of the University Hospital of Monastir as recommended in European manuals [15,16]. A result was considered indicative of bacterial infection if a bacterium of interest was isolated with a count of at least 10⁷ CFU/ml [15,16]. The susceptibility of respiratory pathogens to antimicrobials was measured by standard diffusion methods.

Follow-up of patients

Subsequent changes in clinical course, chest radiographs or standard laboratory measurements, and occurrence of adverse events were monitored for all the patients. Study treatment was prematurely discontinued if major side effects occurred or if the patient’s condition deteriorated because of worsening or persistence of clinical signs of infection requiring additional antibiotics. The decision to initiate new antibiotics in addition or replacement of levofloxacin was left to the attending physician after discussion with the medical staff. When the decision to start new antibiotics was made during the first 3 days of admission, agents directed against community-acquired organisms such as ampicillin, amoxicillin–clavulanate or erythromycin were used; the treatment was then adapted according to the results of the bacteriological investigations. Infections arising beyond the third day of admission at the emergency department were judged to be nosocomial and were usually treated with aminoglycosides and imipenem–cilastatin or ceftazidime, and secondarily adapted to culture results.

All the patients were monitored until their discharge from hospital. Data collected for each patient included age, sex, previous medical history, severity of COPD according to the GOLD classification, laboratory test results and current medical treatment. Six months and one year post-hospitalization, patients were contacted by telephone to record their outcome regarding the occurrence of new exacerbation episodes and vital status.

Detection of viruses and atypical bacteria by multiplex qPCR assays

This part of the study was conducted retrospectively on frozen aliquots of sputum samples. Two multiplex qPCR assays were tested in parallel. The Seegene One-step RP real-time PCR assay (Seegene Inc., Seoul, Korea) (SG) was performed in the Laboratory of Infectious Agents and Hygiene of the University Hospital of Saint-Etienne, France. The Fast Track Diagnostics respiratory pathogen assay (Fast Track Diagnostics, Esch-sur-Alzette, Luxembourg) (FTD) was performed in the Laboratory of Emerging Pathogens of Lyon, France.

Total nucleic acid extraction (DNA/RNA) was performed using the NucliSens easy-MAG instrument (BioMérieux, Marcy l’Etoile, France). Nucleic acids were eluted in a final volume of 200 µl then stored at −80 °C. Extracted nucleic acids were amplified by both SG and FTD assays according to the manufacturers’ instructions. The pathogens tested by each kit are listed in Table 1.

Table 1. List of respiratory pathogens detected by each of the two commercial molecular multiplex kits used in this study.

	Seegene test	Fast-track test
Viruses	Influenza A virus	Influenza A virus (FLU A)
	Influenza A-H1N1pdm09	Influenza A-H1N1pdm09 virus
	Influenza B virus	Influenza B virus
	Influenza A-H1 virus	–
	Influenza A-H3 virus	–
	Respiratory syncytial virus A	Respiratory syncytial virus A
	Respiratory syncytial virus B	Respiratory syncytial virus B
	Coronavirus NL63	Coronavirus NL63
	Coronavirus 229E	Coronavirus 229E
	Coronavirus OC43	Coronavirus OC43
	–	Coronavirus HKU-1
	Parainfluenza virus type 1	Parainfluenza virus type 1
	Parainfluenza virus type 2	Parainfluenza virus type 2
	Parainfluenza virus type 3	Parainfluenza virus type 3
	Parainfluenza virus type 4	Parainfluenza virus type 4
	Adenovirus	Adenovirus
	Enterovirus	Enterovirus
	Rhinovirus	Rhinovirus
	Human metapneumovirus	Human metapneumovirus A
	–	Human metapneumovirus B
	Bocavirus	Bocavirus
	–	Parechovirus
Bacteria	Mycoplasma pneumoniae	Mycoplasma pneumoniae
	Chlamydia pneumoniae	Chlamydia pneumoniae
	Legionella pneumophila	–

Briefly, SG was conducted in a final reaction volume of 17 µl, containing 5 µl of 5× RP MOM (MuDT Oligo Mix), 5 µl of RNase-free water, 5 µl 5× Real-time One-step Buffer and 2 µl Real-time One-step Enzyme (One-step RT-PCR Mastermix). For each panel (Mix 1, Mix2, Mix 3 and Mix4), 17 µL of mixture was then added to 8 wells of a PCR microplate (7 sample reactions plus one positive control). Eight microliters of the extracted samples and the positive control were then added to the respective wells of each primer/probe pool. PCR thermal cycling conditions were optimized on a CFX96TM Real-time PCR System (Biorad Laboratories, Hercules, CA, USA) to the following protocol: 15 minutes at 50 °C, 10 minutes at 95 °C and 40 cycles of 8 seconds at 95 °C and 34 seconds at 60 °C. A positive test result was considered a well-defined curve that crossed the threshold cycle within 40 cycles.

For the FTD assay, PCR was conducted in a final reaction volume of 25 µl, containing 12.5 µl of 2× RT-PCR buffer, 1.5 µl of each primer/probe pool and 1 µl of 25× RT-PCR Enzyme mix (Fast-track mastermix). Fifteen microliters of each mixture were then added to 14 wells of a PCR plate (12 sample reactions plus one positive and one negative control). Ten microliters of the extracted samples, the negative and the positive controls were then added to the respective wells of each primer/probe pool. PCR thermal cycling conditions were optimized on a similar thermo-cycler as for SG to the following protocol: 20 minutes at 50 °C, 15 minutes at 95 °C and 45 cycles of 10 seconds at 95 °C, followed by 60 °C for 1 minutes and 72 °C for 10 seconds. A positive test result was considered a well-defined curve that crossed the threshold cycle within 45 cycles.

Each assay included an internal control to rule out PCR inhibition.

Statistical analyses

Statistical analyses were performed using the SPSS 18.1 software for Windows (SPSS Inc, Chicago, IL, USA). Kolmogorov–Smirnov test was performed for all quantitative variables. Categorical variables were presented as number (%) and numerical variables as mean ± standard deviation (SD). Comparisons between groups were performed using chi-square test or Fisher’s exact test for categorical data, as well as unpaired t-test or Mann–Whitney U-test for normally distributed or skewed numerical data, respectively. A p value ≤0.05 was considered as statistically significant. A logistic regression analysis was conducted by constructing a model in which the binary variable (occurrence or not of a new episode of COPD within 6 or 12 months) was submitted to different explicative variables found significant or close to signification by univariate analysis; the Wald test was used to test the statistical significance.

Results

The flowchart of the whole study is depicted in Figure 1. From 100 consecutive patients with presumptive AECOPD who were screened for inclusion, 84 patients (mean age: 67.8 years, range: 50–88 years; 92.9% males) were included into the first part of the study according to the inclusion and exclusion criteria defined in the previous section for studying the pathogens involved in AECOPD. From these 84 patients, 74 patients for whom both bacterial and viral data were available were included into the second part of the analysis. As mentioned earlier, the patients analyzed in this study were issued from a bigger one for which they were randomized in two arms in terms of initial antimicrobial treatment; we checked that the patients included in the present study were equally distributed between the two arms (data not shown).

First part: microbial investigation of pathogens involved in AE of 84 COPD patients

Table 2 summarizes the demographics of these patients, together with the clinical and biological data at admission. The initial sputum specimen of these 84 patients was tested for the detection of viruses and atypical bacteria by PCR using two different molecular tests (SG and FTD), except for five samples whose volume was not sufficient for testing by the second test (Figure 1). From these 84 patients, 74 were also tested for the detection of conventional bacteria in sputum (Figure 1); the 10 missing specimens were due to a failure to obtain induced sputum of good quality. Table 3 lists the large number and variety of pathogens that were recovered from the sputum samples of these 84 patients.

Table 2. Demographics together with clinical and biological characteristics at hospital admission of the 84 patients exhibiting an exacerbation of COPD included into the study.

Patients’ characteristics	All patients
Age in years, mean (SD)	67.8 (10)
Male sex, no. (%)	78 (92.9)
Smoking history in packets/year, mean (SD)	52.3 (64)
Number of previous exacerbations in the last 12 months, mean (SD)	2.5 (1.7)
BMI in kg/m^2, mean (SD)	25.7 (4.9)
GOLD severity classification
Stage 4, no. (% )	30 (35.7)
Stage 3, no. (%)	47 (56)
Stage 2, no. (% )	7 (8.3)
PaO2 in kPa, mean (SD)^a	14 (1.4)
PaCO2, in kPa, mean (SD)^a	5.9 (1.9)
pH, mean (SD)	7.3 (0.06)
WBC in cells/mm^3*10^3, mean (SD)	12.4 (6)
CRP in mg/l, mean (SD)	58 (56)

^a These values were recorded while the patients were under oxygen supplementation.

BMI, body mass index; CRP, C-reactive protein; PaCO₂, partial pressure of carbon dioxide in arterial blood; PaO₂, partial pressure of oxygen in arterial blood; SD, standard deviation; WBC, white cell blood count.

Table 3. List of pathogens recovered from the sputum samples of the patients of the study.

Pathogen	Number
Conventional bacteriological cultures (n = 74)^a	15
– Acinetobacter baumannii	2
– Escherichia coli	1
– Klebsiella pneumoniae	2
– Haemophilus influenzae	3
– Haemophilus parainfluenzae	1
– Pseudomonas aeruginosa	3
– Staphylococcus aureus	1
– Streptococcus pneumoniae	2
Molecular testing (n = 84)
Bacteria	1
– Chlamydia pneumoniae	1
Viruses	63
– Influenza virus A/H1N1	7
– Influenza virus A/H3N2	5
– Influenza virus B	1
– Rhinovirus	33
– Respiratory syncytial virus B	1
– Parainfluenza virus type 1	2
– Parainfluenza virus type 3	1
– Human coronavirus 229E	2
– Human coronavirus OC43	2
– Human coronavirus NL63	2
– Human coronavirus HKU-1^b	2
– Human bocavirus	4
– Adenovirus	1

^a The result was considered positive for a threshold of 10⁷ UFC/ml. Ten sputum samples could not be tested by conventional culture.

^b The HKU-1 type of human coronavirus was available only in the Fast-track panel.

The two molecular kits tested on sputum specimens of 79 patients were shown to exhibit very close results with a strict concordance of pathogens in 69 cases. However, the SG test allowed the detection of 10 more pathogens including five rhinoviruses, one bocavirus, one influenza virus A/H1N1 and three coronaviruses (CoV-NL63, CoV-OC43, CoV-229E, one strain of each). By contrast, the FTD kit identified two cases of CoV-HKU-1 that is not present in the SG panel (Table 1). Interestingly, one patient was found infected with two strains of coronavirus (NL63 by SG and HKU-1 by FTD); as these two types belong to different serogroups of human coronavirus, a cross-reaction of the PCR assays can be excluded. Further analyses were based on the results of the SG kit (more strains identified and more samples tested) with the addition of the two CoV-HKU-1 cases detected only by the FTD kit.

Second part: analysis of CRP level and clinical outcome according to pathogens

The second part of the study was focused on the 74 patients for whom both bacterial and viral data were available (Figure 1). A threshold of 10⁷ CFU/ml was considered as presumptive of a bacterial infection for conventional culture. Overall, at least one bacterium at the significant rate of 10⁷ CFU/ml was detected in 15 patients (20.3%), the genome of at least one viral pathogen was detected in 45 patients (60.8%) and no pathogen was detected in 21 patients (28.4%). A bacterium–virus co-infection was present in seven patients (9.5%), as depicted in Table 4 that recapitulates the results of the 15 patients for whom more than one pathogen was detected.

Table 4. Description of the 15 pathogen associations recorded in the 74 patients tested for bacterial and viral agents in sputum specimen.

Pathogen association	Number of cases
Bacterium-bacterium	1
– Klebsiella pneumoniae + Pseudomonas aeruginosa	1
Bacterium–virus(es)	7
– Haemophilus parainfluenzae + influenza virus A/H1N1	1
– Acinetobacter baumannii + influenza virus A/H1N1	1
– Acinetobacter baumannii + respiratory syncytial virus type B	1
– Klebsiella pneumonia + rhinovirus	1
– Haemophilus influenzae + rhinovirus	1
– Escherichia coli + rhinovirus	1
– Streptococcus pneumoniae + rhinovirus + parainfluenza virus type 1	1
Virus–virus	7
– Bocavirus + rhinovirus	4
– Adenovirus + rhinovirus	1
– Influenza virus A/H3N2 + rhinovirus	1
– Parainfluenza virus type 3 + human coronavirus HKU-1	1

Table 5 illustrates the baseline characteristics and the follow-up of these patients according to the category of pathogens recovered from the sputum samples: group 1: bacterium associated or not to virus; group 2: virus only; group 3: no pathogen detected. It is interesting to notice that the level of CRP was associated with the presence of a bacterial infection by univariate analysis (p < 0.001); in contrast, this marker was lower in acute episodes associated with only viruses. The severity of infection, appreciated by the GOLD score, was also increased in exacerbation episodes associated to bacterial infections. In addition, the risk of recurrence of exacerbation episodes was significantly higher at 6 months and at one year of follow-up in patients exhibiting an initial episode associated to a bacterial infection; in contrast, this risk was particularly reduced in patients with stigmata of viral infection (Table 5).

Table 5. Baseline characteristics and follow-up of the 74 patients for whom a sputum specimen was available for conventional bacterial culture and molecular techniques, according to the category of pathogen(s) recovered from the sputum of COPD patients.

Variable	Group 1: Bacterial pathogen (n = 15)	Group 2: Viral pathogen (n = 38)	Group 3: No pathogen (n = 21)	p Value
Age in years, mean ± SD	67.6 ± 9	69.3 ± 8.5	65.3 ± 40.4	NS
Male sex, nb (%)	15 (100)	34 (89.5)	21 (100)	NS
Smoking history in packets/year, mean ± SD	72.2 ± 76.6	59 ± 70.7	41.9 ± 41.7	NS
Number of previous exacerbations in the last 12 months, mean ± SD	2.8 ± 1.7	2.7 ± 2	2.2 ± 1	NS
BMI in kg/m^2, mean ± SD	28.1 ± 4.7	25.7 ± 4.6	23.8 ± 5	NS
GOLD severity classification				0.05
– Stage 2, no. (%)	2 (13.3)	0 (0)	3 (14.3)
– Stage 3, no. (%)	6 (40)	19 (52.8)	14 (66.7)
– Stage 4, no. (% )	7 (46.7)	17 (47.2)	4 (19)
PaO2 in kPa , mean ± SD	16.6 ± 1.8	11.5 ± 0.6	17.9 ± 2.0	NS
PaCO2, in kPa , mean ± SD	5.8 ± 1.8	5.6 ± 1.9	6.02 ± 1.9	NS
pH, mean ± SD	7.38 ± 0.07	7.37 ± 0.06	7.34 ± 0.07	NS
WBC in cells/mm^3*10^3, mean ± SD	14.5 ± 8.3	12.4 ± 5.3	11.3 ± 6.4	NS
CRP in mg/l, mean ± SD	104.3 ± 70.8	40.3 ± 40.2	60.3 ± 57.7	0.001
Duration of treatment by levofloxacin				NS
– 2 days, nb (%)	4 (26.7)	21 (55.3)	9 (42.9)
– 7 days, nb (%)	11 (73.3)	17 (44.7)	12 (57.1)
Follow-up at 6 months
– Recurrence of exacerbation, n (%)	10 (66.7)	2 (5.3)	2 (9.5)	0.0001
– Death, nb (%)	3 (20)	1 (2.8)	1 (4.8)	NS
Follow-up at 12 months
– Recurrence of exacerbation nb (%)	12 (80)	4 (10.5)	3 (14.3)	0.0001
– Death, nb (%)	3 (21.4)	1 (2.8)	2 (9.5)	NS

Values are reported as mean ± SD or as frequency (%).

NS, not significant; AECOPD, acute exacerbation of chronic obstructive pulmonary disease; BMI, body mass index; CRP, C-reactive protein; PaO₂, partial pressure of oxygen in arterial blood; PaCO₂, partial pressure of carbon dioxide in arterial blood; WBC, white cell blood count.

Neither the association of a virus to a bacterial infection nor that of two or three viruses in the same sputum specimen exhibited a statistically significant correlation with a change in CRP level or an increase of further exacerbation episodes at 6 or 12 months of follow-up (data not shown) but the number of observations was low.

In order to explore more in depth the value of the different variables listed in Table 5 for predicting the occurrence of new AECOPD, a logistic regression was conducted. The explicative variables that were conserved in the model were the number of previous episodes of AECOPD, the level of CRP (<36 mg/l vs <36 mg/l), the GOLD score (4 vs 2 + 3) and the pathogen category (group 1 vs groups 2 + 3). As shown in Table 6, the latter variable was the only one to be highly predictive of recurrence, either at 6 months or at 12 months (odds ratio of 37.72 and 22.97 with a 95% confidence interval of 5.93–239.93 and 4.76–120.75, respectively; p < 0.0001 by Wald test in both cases).

Table 6. Multivariate analysis conducted for studying the relationship between the occurrence of a new episode of exacerbation of COPD (recurrence) within 6 or 12 months and different explicative parameters in the 74 patients for whom univariate results are shown in Table 5.

Variable	Recurrence at 6 months			Recurrence at 12 months
	OR [CI 95%]	p Value		OR [CI 95%]	p Value
Number of previous episodes of exacerbation	1.27 [0.81–2.01]	0.31		1.27 [0.83–1.96]	0.13
CRP (< 36 mg/l vs > 36 mg/l)	1.40 [0.23–8.42]	0.71		4.54 [0.90–22.96]	0.07
GOLD score (4 vs 2 + 3)	0.20 [0.03–1.30]	0.09		0.51 [0.11–2.40]	0.39
Pathogen category (group 1 vs groups 2 + 3)	37.72 [5.93–239-93]	<0.0001		22.97 [4.76–120.75]	<0.0001

Discussion

Respiratory infection is recognized as an important trigger for AECOPD [17–19]. A recent meta-analysis performed on 118 studies indicated that bacterial infection was involved in 49.95% of AECOPD cases [20] whereas other studies pointed out that viral pathogens were associated with about half of all exacerbations [8,11,21]. However, the whole spectrum of causative pathogens is not commonly looked for, even within published studies in which the simultaneous detection of bacteria and viruses by pertinent microbiological tools is rarely performed. Because quantitative bacteriology on induced sputum is not well standardized and that virological testing is not yet routinely recommended in patients with AECOPD [1], clinicians often prescribe antibiotics on empirical basis. This is particularly true in middle-income countries like Tunisia. The present study, which is an ancillary arm of a larger clinical trial, was undergone for evaluating what would be the benefit of a more global approach of the microbiological investigation of EACOPD, notably in terms of antimicobial stewartship.

By using multiplex molecular testing, the genome of a viral respiratory pathogen was recorded in 60.8% of our 74 patients, which is in the range of recent studies using similar tools (47% in [22], 58% in [23], 35.1% in [24] and 46.1% in [25]). In our study, HRV was the much most prevalent viral agent; in a systematic review about the relevance of respiratory viruses in AECOPD by PCR techniques, Zwaans et al. [8] noticed that HRV were the most commonly detected viruses with a pooled prevalence of 16.4%, followed by RSV (9.9%) and influenza viruses (7.8%). If the latter viruses were also well represented in our study, we observed a very low prevalence of RSV infection (one case), which is in concordance with some studies [10,22,26,27] but in large disagreement with others [11,23,28]. As for influenza viruses, the rate of RSV infections is largely dependent on the intensity of the outbreak related to this agent at the time of the study. Human coronaviruses and bocavirus were detected in our study in eight and four cases, respectively; interestingly, bocavirus was always present in association with rhinovirus. ADV and parainfluenza viruses were only detected anecdotally; however, it is noteworthy that, in a recent study comparing prospectively the upper respiratory infection-AECOPD sequence, parainfluenza virus type 3 was the only pathogen significantly detected in both events within a period of 21 days [25].

In our study, a bacterial infection was recorded in only 15 patients. This low rate of bacterial infections by comparison to the figures reported in the review of Moghoofei et al. [20] is due to the strict criteria that were applied consisting, for conventional bacteria, to a bacterial count of at least 10⁷ CFU/ml in induced sputum, as recommended in recent European guidelines [15,16]. To our opinion, qualitative PCR is not adapted to the identification of conventional bacteria because it is too sensitive and recognizes mainly colonizing agents, as illustrated by the study of Aydemir et al. who reported the detection of bacterial genome in 63.5% of patients by PCR method [29]. The use of quantitative PCR tests for measuring the bacterial load could help to overcome this difficulty in the future [30]. With regard to atypical bacteria, their implication in AECOPD could vary from 3% to 20% [31]; in this study, the multiplex PCR methods used for the detection of atypical agents’ genome identified only one case of C. pneumoniae whereas no case of M. pneumoniae or Legionella pneumophila was recorded.

Different studies based on molecular techniques have emphasized the importance of mixed infections in patients with AECOPD [10,22,23,32,33]. In the present study, co-detection of several pathogens (two bacteria, bacterium + virus or association of several viruses) was found in the sputum of more than 20% of the patients. It was not associated to an improved severity, although the number of investigated cases was relatively small. Further studies are needed for confirming that these double or triple infections do not constitute an additional factor of severity in AECOPD patients.

The measure of systemic inflammatory markers and notably of CRP may prove to be useful in the assessment of severity and identification of bacterial aetiology in AECOPD patients. The present study, in accordance with others [23,34,35], suggests that high CRP levels may be associated to bacterial infection in these patients; however, the value of this biomarker as an indicator of exacerbation recurrence within 6 or 12 months was not confirmed in the multivariate analysis. In contrast, an important lesson driven from this study was that the category of the presumed pathogen influenced the overall clinical outcome of AECOPD patients. Exacerbation recurrences at 6 and 12 months were significantly lower in patients in whom viral genome only or no pathogen at all was recorded compared to those with bacterial infection; this factor remained highly significant in the multivariate model. These original results indicate that patients with bacterial infection, associated or not with virus detection, are more prone to be readmitted to hospital following a previous episode of AE, which could have important practical implications in terms of initial antimicrobial therapy and subsequent follow-up.

We acknowledge that our study has some limitations. First, the size of the groups is relatively small, which suggests that our findings on the relationship between the category of pathogens found in sputum of AECOPD patients and the risk of new episodes of exacerbation would need to be validated in a larger study specifically orientated towards this objective. Second, highly febrile patients (≥38.5 °C) were excluded from the study in order to limit the inclusion of pneumonic patients with normal or subnormal X-ray chest. It must be noted that patients with mild episodes of AECOPD, which was the focus of this study, exhibit usually low grade fever or no fever at all. As we excluded AECOPD patients with high fever or radiological pneumonia, our findings apply only to patients with non-pneumonic exacerbation of COPD and cannot be extrapolated to patients with pneumonia for whom a specific prospective study would be needed. Third, serological tests for atypical bacteria, which remains the gold standard for documenting recent infections with these agents [36,37], were not performed systematically in our patients. However, the lack of standardization of these serological tests represents a true limitation to their use. Fourth, although HRV is the major contributor in AECOPD, no genotyping was performed despite the fact that HRV strains belonging to subspecies C were shown to be associated to highest severity in AECOPD [10].

Despite its limitations, this study contributes to highlight three messages for the monitoring of patients with AECOPD. First, even if detection is not synonymous of causality, the current multiplex molecular tests were shown able to document accurately the presence in sputum of a wide range of additional pathogens that are difficult to obtain by conventional techniques; indeed, a very good concordance was observed between the two molecular kits assayed in the study; the role of viral agents in AECOPD of low severity is predominant, which would plead for recommending the point-of-care testing of respiratory viruses in this context [38,39]. Second, the quantification of conventional bacteria in sputum remains a key element for identifying the acute episodes for which a specific antimicrobial treatment is required; in the future, molecular assays able to give this information in a short time could be useful to avoid empirical antimicrobial treatment [30], even if conventional culture would be still required to perform precise antimicrobial testing due to the frequent resistance pattern of bacteria in this setting. Third, in terms of prognosis, bacterial infections were shown by logistic regression to represent a significant predictor of recurrence of exacerbation episodes during the next year, which implies to treat correctly these infections and to follow-up these patients with a particular attention.

Conclusion

In the light of our observations and of those from previous studies, the following algorithm could be proposed for the monitoring of patients presenting at the emergency room with AECOPD with no signs of severity, by combining the results of CRP measurement to those of multiplex molecular detection of viruses and atypical bacteria and of conventional quantitative culture in sputum. In the group of patients with presence of high counts of conventional bacterium, a specific antibiotic treatment must be implemented in accordance to the results of antimicrobial testing; these patients with a documented higher risk of AECOPD would require a reinforced follow-up. In the small group of patients exhibiting an atypical bacterium by multiplex assay, a specific treatment based on quinolones or macrolides must be initiated with additional testing of the antibody response in order to confirm the reality of infection. In the other patients with no documented infection or only viral infection, the antimicrobial treatment could be avoided, which would curtail inappropriate antibiotic use and prevent the selection of resistant bacteria; in case of influenza, an early specific treatment based on anti-neuraminidase drugs could help to shorten the evolution of the acute episode. Additional studies are required to validate prospectively the present algorithm and to improve the current stewardship of antibiotics in AECOPD patients.

Declaration of interest

The authors declare no conflict of interest regarding the matter of this study.

Acknowledgements

The authors thank William Caule from the EUROBIO Company for his assistance. The physicians of the University Hospitals who participated to the recruitment of patients are acknowledged and also all the patients who agreed to enter the study. The authors are indebted to Philippe Berthelot and Béatrice Trombert-Paviot for fruitful discussion of statistical results.

Additional information

Funding

The Seegene reagents were a gift of the Eurobio Company, France. The other sources of funding were issued from the grants of the EA3064 unit from the University of Saint-Etienne, and of the Laboratory of Emerging Pathogens, Mérieux Foundation, CIRI, Inserm U1111 from the University of Lyon, France.