Abstract
Background Acute exacerbations of COPD (AECOPD) and community acquired pneumonia (CAP) often coexist. Although chest radiographs may differentiate between these diagnoses, chest radiography is known to underestimate the incidence of CAP in AECOPD. In this exploratory study, we prospectively investigated the incidence of infiltrative changes using low-dose computed tomography (LDCT). Additionally, we investigated whether clinical biomarkers of CAP differed between patients with and without infiltrative changes.
Methods Patients with AECOPD in which pneumonia was excluded using chest radiography underwent additional LDCT-thorax. The images were read independently by two radiologists; a third radiologist was consulted as adjudicator. C-reactive protein (CRP), procalcitonin (PCT), and serum amyloid A (SAA) at admission were assessed.
Results Out of the 100 patients included, 24 had one or more radiographic abnormalities suggestive of pneumonia. The interobserver agreement between two readers (Cohen's κ) was 0.562 (95% CI 0.371–0.752; p<0.001). Biomarkers were elevated in the group with radiological abnormalities compared to the group without abnormalities. Median (interquartile range (IQR)) CRP was 76 (21.5–148.0) mg·L−1 compared to 20.5 (8.8–81.5) mg·L −1 (p=0.018); median (IQR) PCT was 0.09 (0.06–0.15) µg·L−1 compared to 0.06 (0.04–0.08) μg·L−1 (p=0.007); median (IQR) SAA was 95 (7–160) µg·mL−1 compared to 16 (3–89) µg·mL−1 (p=0.019). Sensitivity and specificity for all three biomarkers were moderate for detecting radiographic abnormalities by LDCT in this population. The area under the receiver operating characteristic curve was 0.66 (95% CI 0.52–0.80) for CRP, 0.66 (95% CI 0.53–0.80) for PCT and 0.69 (95% CI 0.57–0.81) for SAA.
Conclusion LDCT can detect additional radiological abnormalities that may indicate acute-phase lung involvement in patients with AECOPD without infiltrate(s) on the chest radiograph. Despite CRP, PCT and SAA being significantly higher in the group with radiological abnormalities on LDCT, they proved unable to reliably detect or exclude CAP. Further research is warranted.
Abstract
LDCT-thorax can detect additional radiological abnormalities in patients with AECOPD after excluding CAP using chest radiography. Biomarkers are significantly elevated in patients with abnormalities, but are not able to reliably exclude these changes. https://bit.ly/3KAsBap
Introduction
Acute exacerbations of COPD (AECOPD) are associated with short-term and long-term reductions in quality of life and lung function, as well as increased risk of death [1–4]. On average, a patient with COPD suffers from 1.5 exacerbations a year [5]. Moreover, AECOPD requiring hospital admission represents a significant prognostic factor for reduced survival across all stages of COPD severity [6].
Patients with COPD have an increased risk of pneumonia, making it the most frequent infectious complication in COPD [7]. To complicate matters further, AECOPD and pneumonia are symptomatically alike, making it hard to diagnose pneumonia in COPD, based on clinical signs and symptoms alone [8]. Establishing a correct diagnosis is important for the guidance of antibiotic therapy. Misdiagnosing pneumonia could have great implications for patients, whereas overdiagnosing pneumonia leads to unnecessary exposure to antibiotics, which is a major driver of antimicrobial resistance [9, 10]. Hence, clinicians have a keen interest in new biomarkers that together with clinical assessments may improve the diagnostic accuracy of pneumonia in patients with AECOPD. Potential biomarkers that are used in AECOPD to detect bacterial infection are C-reactive protein (CRP), procalcitonin (PCT) and serum amyloid A (SAA) [11, 12]. Using these biomarkers as a diagnostic tool may increase the ability to detect clinically relevant bacterial infections at an early stage of the disease. Yet, it is not known whether these biomarkers can differentiate between AECOPD and community acquired pneumonia (CAP). Therefore, current practice for detecting pneumonia in patients with AECOPD is chest radiography. One study showed that 20% of patients admitted to hospital with AECOPD had abnormalities consistent with pneumonia on chest radiography [13], although this is probably an underestimation of the true incidence of pneumonia in patients with AECOPD; chest radiography has limited sensitivity and specificity. Moreover, the interpretation of chest radiography is often complicated by pre-existing cardiopulmonary diseases [14, 15]. Chest computed tomography (CT) outclasses chest radiography and is now considered the gold standard for diagnosing CAP [16, 17]. However, CT delivers much higher radiation doses than conventional diagnostic radiography [18]. An alternative to conventional chest CT scanning is low-dose CT (LDCT), which produces acceptable image quality with lower radiation exposure. LDCT scans have been shown to detect CAP if the chest radiograph does not reveal findings that explain the patient's clinical presentation [19].
In this exploratory study, we prospectively aimed to investigate in patients with AECOPD admitted to hospital in whom pneumonia was excluded using chest radiography, whether clinical biomarkers of CAP differed between patients with and without radiological abnormalities on LDCT. In addition, we investigated the interobserver variation in LDCT reading.
Material and methods
100 consecutive patients were enrolled at the Northwest Clinics Alkmaar (Alkmaar, the Netherlands) between November 2011 and March 2014 as part of the CRP-guided Antibiotic Treatment of Acute Exacerbations of COPD Admitted to Hospital study (CATCH study) (clinicaltrials.gov NCT01232140) [20]. The local ethics boards approved the study protocol, and all patients provided written informed consent. The study population consisted of patients diagnosed with COPD stages I–IV as defined by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) [21]. Inclusion criteria dictate that patients had acute exacerbation as defined by GOLD requiring hospital admission according to GOLD guidelines [21]; age >40 years; and former or current smoking, with a minimum smoking history of 10 pack-years. Exclusion criteria were pre-treatment with oral corticosteroids exceeding a total dose of 210 mg prednisolone during the 14 days preceding the presentation with AECOPD for hospitalisation and pneumonia visualised on a chest radiograph. Immunocompromised patients, patients with active lung cancer and patients with pulmonary embolism were also excluded.
Procedures
Before informed consent was obtained, a chest radiograph was performed to rule out CAP. After informed consent was obtained, baseline blood samples were drawn and baseline variables were collected. LDCT was performed within 12 h after informed consent was obtained using a standardised protocol. LDCT was performed on a CT Philips MX 8000 (16-slice) with the following settings: 90 kV, 25 mAs, CTDI 0.8 mGy; and CT Somatom Definition Flash (128 slice Dual source CT) with the following settings: 120 kV, 20 mAs, CTDI 1.35 mGy. For both scanners, acquisition and reconstruction was <1 mm.
Image analysis
The images were independently read by two radiologists (E. Boerhout and F.J. Rietema). If there was a dispute between the radiologists considering the presence of an infiltrate, a third radiologist was consulted as adjudicator (P.A. de Jong). The readers had no knowledge of clinical or laboratory data, other than the age and sex of the patient. From a list of abnormalities (segmental, peribronchovascular or scattered ground-glass, reticular opacity or consolidation), at least one, or a combination of several abnormalities was used for the diagnosis of pneumonia assessed by LDCT scan [22].
Biomarker measurements
Serum CRP was measured by nephelometry on a Beckman Synchron DxC 800 analyser (CRP latex reagent; Beckman Coulter, Fullerton, CA, USA); CRP reference values 0–5 mg·L−1. PCT was measured using time-resolved amplified cryptate emission (TRACE) technology on a Kryptor Compact analyser (Thermofisher – BRAHMS, Heningsdorf, Germany) with PCT-sensitive Kryptor reagent (Thermofisher – BRAHMS); PCT reference values 0.00–0.10 µg·L−1. SAA was measured using an in-house sandwich ELISA. The ELISA test was calibrated against World Health Organization standard 92/680; reference values were established as <4.2 mg·L−1 [23]. The test was performed at the laboratory of medicine of the University Medical Centre Groningen (Groningen, the Netherlands).
Microbiology
For all patients, sputum culture (if possible), nasopharyngeal swab for multiplex PCR for atypical and viral pathogens was obtained at admission. Additionally, paired serum samples were taken within 1 month for serology (Serion ELISA classic; Virion, Würzburg, Germany) for the detection of Mycoplasma pneumoniae, Chlamydophila pneumoniae, Legionella pneumophila serogroup 1–7, influenza A and B virus, parainfluenza virus 1–3, respiratory syncytial virus and adenovirus.
Statistical analysis
In this pilot study, our primary objective was to establish whether levels of CRP, PCT and SSA differed between patients with and without infiltrative changes on LDCT. As we were unaware of how many patients would have infiltrative changes on their LDCT, and how this would influence biomarker levels, we were not able to make a sample size calculation. Due to financial constraints, a convenience sample of 100 consecutive patients initially underwent a LDCT scan. Statistical analysis was performed using the SPSS package programme (version 22.0). Data was presented as median (interquartile range (IQR)). In the comparison between groups, Chi-squared test and Mann–Whitney U-test were used. The interobserver agreement was measured using Cohen's κ: <0.20 indicating poor agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 good agreement and 0.81–1.00 very good agreement between two observers. Receiver operating characteristics analysis was used to analyse the diagnostic accuracy of biomarkers. Overall statistical significance was set at a two-tailed p-value <0.05.
Results
592 patients presenting at our emergency department with AECOPD were screened for inclusion. 100 (16.8%) patients were eligible and gave informed consent (figure 1). 24 (24%) patients had radiological abnormalities compatible with acute-phase lung involvement detected by LDCT without abnormalities visible on chest radiography (figure 2). Baseline characteristics of both groups are summarised in table 1 and did not show any significant differences between the groups, except for oxygen saturation which was slightly lower in the group with radiological abnormalities.
Trial profile. CATCH: CRP-guided Antibiotic Treatment of Acute Exacerbations of COPD Admitted to Hospital; LDCT: low-dose computed tomography.
A 79-year-old female presented to our emergency room with symptoms of acute exacerbation of COPD. a) Conventional chest radiograph without evidence of radiological abnormalities; b) low-dose coronal plane computed tomography of the same patient with alveolar consolidation of the right upper lobe.
Baseline characteristics
The different types of radiological abnormalities are summarised in table 2.
Types of radiological abnormalities#
Results regarding sputum cultures are presented in table 3.
Sputum culture
Additional information regarding microbiological results in relation to the presence of radiological abnormalities can be found in supplementary tables S1 and S2.
Interobserver variation
The observed proportional agreement (P0) in LDCT judgement of observed abnormalities by radiologists A and B was 84.0%. The proportional agreement by chance (Pe) was 63.5%, resulting in a κ of 0.562 (95% CI 0.371–0.752; p<0.001), reflecting moderate agreement. The proportional agreement in positive cases (69.6%) was lower than the proportional agreement in negative cases (88.3%). 52 patients were scanned on a 128-detector-row scanner and 48 on a 16-detector row scanner. On the 16-detector-row scanner the agreement was 85.4% with a κ of 0.543 (95% CI 0.249–0.816). On the 128-detector-row system, the agreement was 82.7% with a κ of 0.575 (95% CI 0.333–0.816).
Biomarkers
All biomarkers were significantly higher in the group with radiological abnormalities on the LDCT compared to those without radiological abnormalities. CRP was 20.5 (IQR 8.8–81.5) mg·L−1 in the group without radiological abnormalities compared to 76 (IQR 21.5–148.0) mg·L−1 (p=0.018) in the group with abnormalities (figure 3a), whereas PCT was 0.06 (IQR 0.04–0.08) µg·L−1 in the group without radiological abnormalities compared to 0.09 (IQR 0.06–0.15) µg·L−1 (p=0.007) in the group with radiological abnormalities (figure 3b). SAA was 16 (IQR 3–89) µg·mL−1 in the group without radiological abnormalities compared to 95 (IQR 7–160) µg·mL−1 (p=0.019) in the group with abnormalities (figure 3c). The area under the receiver operating characteristic curve for CRP was 0.66 (95% CI 0.52–0.80), for PCT 0.66 (95% CI 0.53–0.80) and for SAA 0.69 (95% CI 0.57–0.81). The optimal cut-off value to identify radiological abnormalities was 43 mg·L−1 for CRP, with a sensitivity of 0.70 and a specificity of 0.68. The optimal cut-off point for PCT to identify radiological abnormalities was 0.05 µg·L−1, with a sensitivity of 0.78 and a specificity of 0.47. The optimal cut-off point for SAA to identify radiological abnormalities was 27 µg·mL−1, with a sensitivity of 0.70 and a specificity of 0.65.
a) C-reactive protein (CRP) level in patients with and without radiological abnormalities; b) procalcitonin (PCT) level in patients with and without radiological abnormalities; c) serum amyloid A (SAA) level in patients with and without radiological abnormalities. LDCT: low-dose computed tomography.
Discussion
In this exploratory study, we show that 24% of the patients with AECOPD admitted to hospital without evidence of pneumonia on chest radiography had additional radiological abnormalities on LDCT. In patients with radiological abnormalities on LDCT, the levels of CRP, PCT and SAA were higher than in those without LDCT abnormalities. However, we were unable to reliably exclude or confirm the diagnosis of CAP using these biomarkers.
The incidence of infiltrative changes in AECOPD not detected by chest radiography was considerably higher than in an earlier study, among 47 patients suspected to have CAP; while confirming CAP in 18 patients, in eight further patients, CAP could be detected with high-resolution computed tomography (HRCT) [22]. The higher number of infiltrates found in this study may be explained by a difference in baseline population, as our study consisted of hospitalised patients with AECOPD, while the aforementioned study consisted of patients with fever and cough from the general population.
The κ-value of agreement between the two radiologists was moderate (0.562), and lower compared to some previous studies [22]. This can probably be explained by a combination of small infiltrates that are more difficult to diagnose, and the use of LDCT instead of HRCT [22].
CRP, SAA and PCT were elevated in the group with consolidations. However, they were considerably lower compared to other studies in patients with radiographic confirmed pneumonia [24, 25]. The biomarkers reported in the group without radiological abnormalities were comparable to an earlier study in patients with AECOPD [11]. CRP, SAA and PCT showed poor sensitivity and specificity for the prediction of radiological abnormalities on LDCT. In part, this might be due to the fact that all biomarkers in this study are used in the detection of bacterial infection, whereas these radiological abnormalities can also be caused by viral infections [11, 26]. This might subsequently lead to a lower level of these biomarkers compared to bacterial infection [27, 28]. To complicate matters further, definite diagnosis of viral or bacterial origin cannot be achieved using imaging features alone. However, recognition of viral pneumonia patterns may help in differentiation between viral pathogens and bacterial pathogens, with subsequent reduction of unnecessary use of antibiotics [29].
A potential strength of this study is that all data were collected prospectively, thereby not resulting in selection bias. Another strength is the fact that radiologists were blinded to clinical data except for age and gender, so this could not have influenced their judgement. A potential limitation is design of our study, as patients were allowed to have a 12-h time gap between the initial chest radiography and the LDCT. Infiltrative changes may progress over time, which may have led to progression or emergence of new radiological abnormalities. The second limitation is the prescription of antibiotics or systemic corticosteroids prior to inclusion. We cannot exclude that these pre-treated patients may have altered biomarker performance, or that remnants of infiltrates invisible to chest radiography still can be detected on CT. The third potential limitation is the fact that CT cannot discriminate between viral and bacterial CAP, as only invasive local microbiological sampling would have provided this diagnostic precision [29]. As we did not gather these data, results regarding the differences between biomarker levels and LDCT findings should be interpreted with caution. A fourth weakness is that this is an explorative study and not powered to detect differences in biomarkers. Finally, our study was performed between 2011 and 2014. CT detector and reconstruction technology is evolving continuously. Iterative reconstruction lowered noise (and dose), but altered the lung appearance in some instances. With the most recent CT advances, resolution can be further improved. It may be that the detection of pneumonia will even further improve and that we missed some very subtle cases.
AECOPD and pneumonia have considerable symptomatic overlap, and ruling out CAP using chest radiography is often difficult. We have shown that using LDCT, it is possible to detect additional radiological abnormalities in patients with AECOPD. However, the question remains what the outcome means for the individual patient, as the abnormalities found are diverse and do often not reflect a specific aetiology. Biomarkers are increased in patients with infiltrative changes on their LDCT. Yet CRP, PCT and SAA did not carry sufficient weight to confirm or exclude the diagnosis of CAP in this specific population. Therefore, further research is necessary to determine whether the presence of additional infiltrative changes on LDCT represents a categorically different disease process or rather a spectrum of infection, especially in the light of the similarity between clinical characteristics and microbiology.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material 00054-2022.SUPPLEMENT
Footnotes
Provenance: Submitted article, peer reviewed.
This study is registered at www.clinicaltrials.gov with identifier number NCT01232140. All data will be made available after de-identification, including the protocol, with no limit. Data will be made available immediately after publication. Data will be made available upon reasonable request to researchers with a methodological sound proposal. Proposals should be directed to w.boersma{at}nwz.nl.
Conflict of interest: The authors have nothing to disclose.
- Received October 14, 2021.
- Accepted April 6, 2022.
- Copyright ©The authors 2022
This version is distributed under the terms of the Creative Commons Attribution Non-Commercial Licence 4.0. For commercial reproduction rights and permissions contact permissions{at}ersnet.org
References
