Biomarkers to Guide the use of Antibiotics for Acute Exacerbations of COPD (AECOPD)

A Systematic Review and Meta-Analysis

George Hoult; David Gillespie; Tom M. A. Wilkinson; Mike Thomas; Nick A. Francis


BMC Pulm Med. 2022;22(194) 

In This Article


This systematic review appraised 39 studies evaluating 61 biomarkers. Four biomarkers (serum CRP and PCT, and sputum IL-8 and TNF-α) show potential for use in differentiating bacterial from non-bacterial AECOPD. The strongest evidence was found for serum CRP, where studies performed on non-ICU inpatient populations largely suggested greater CRP concentrations in patients presenting with bacterial AECOPD. High heterogeneity in meta-analysis did not allow for reliable quantitative synthesis for serum CRP or PCT, and meta-analysis was not possible for IL-8 or TNF-α due to the low number of studies. The available evidence suggests that serum WBC count is not useful as a marker of bacterial AECOPD. The evidence for sputum IL-1β, IL-6, MPO and NE as biomarkers for detecting bacterial AECOPD is inconclusive, and there was insufficient evidence available for other biomarkers evaluated.

Strengths and Limitations

Strengths of this review include the broad aims and inclusion criteria, the systematic approach to searching and selection of studies, and the use of meta-analysis to pool data. We were able to identify 39 relevant studies from 1105 titles identified by our search strategy. However, most studies had small sample sizes with fewer than 50 bacterial exacerbation events, and only a quarter of the studies had more than 100 exacerbation events of any aetiology in their analysis. Furthermore, some of the studies did not present information in a way that enabled us to perform meta-analysis, which further limited our capabilities in drawing reliable conclusions. The quality of the evidence was mixed, with a large number of studies having a high or unclear risk of bias in terms of patient selection, but most studies being at low risk of bias in terms of the index test used. A large proportion of studies included patients that were taking antibiotics prior to giving samples, which could have reduced positive culture rates and so underestimated the number of bacterial exacerbations and therefore artificially reduced the ability of the biomarkers to correctly predict a bacterial cause based on culture positivity. In addition to this, some patients were already receiving corticosteroid treatment prior to giving samples which could reduce the concentrations of inflammatory biomarkers (such as CRP) and therefore affect the ability of these biomarkers to positively identify a bacterial cause. Furthermore, two studies did not describe how they defined 'bacterial exacerbation' and two studies examining the utility of CRP[19,37] used CRP > 50 mg/L as part of their definition of bacterial AECOPD, leading to classification bias. Further limitations include the fact that initial screening was only done by one reviewer (with uncertainties discussed with a second reviewer), only two databases were searched, and that we excluded studies not published in English.

A major challenge in synthesising and interpreting these studies is the variation in setting and severity of participants. Most studies involving out-patients, and some inpatient studies, found CRP levels < 40 mg/L in those with no evidence of bacterial infection. However, six studies reported mean or median CRP levels greater than 40 mg/L (46.4 mg/L–105.6 mg/L) in culture negative patients, and correspondingly higher CRP concentrations in culture positive patients (45.7 mg/L–106.7 mg/L). The patients in these cohorts often had more severe COPD and exacerbations with worse respiratory symptoms than many of the other studies. In addition, three studies were performed on patients admitted to ICU, of which mean CRP of bacterial AECOPD was 56–106.7 mg/L, and 53–105.6 mg/L in non-bacterial AECOPD.[11–13] This variation in underlying severity likely contributed to the high level of heterogeneity found in the meta-analysis, however sub-group analyses by setting did not sufficiently reduce heterogeneity.

Mean PCT concentrations in non-bacterial AECOPD were > 0.2 ng/mL in 3 studies, although two of the studies were performed in ICU, where in one 60% of their cohort had GOLD stage III/IV disease[13] and in the other 75% of the cohort with non-bacterial AECOPD had GOLD stage IV disease.[12] Mean PCT was still reported higher in bacterial versus non-bacterial AECOPD in two of the three studies despite relatively high PCT concentrations in the non-bacterial AECOPD groups.[13,35]

Variation in the definitions used for COPD, exacerbations, and bacterial exacerbations is likely to have introduced further heterogeneity. Most of the studies used different definitions which made drawing comparisons between them difficult, and any analysis across them is limited. Despite international classification schemes being developed such as GOLD and Anthonisen[49] criteria, many studies decided to use other parameters to measure COPD & exacerbation severity. Moreover, there was large variation in the way that bacterial exacerbation was defined. The most common approach was to use a certain level of growth (CFU/mL) of a known pathogen from sputum culture, but this approach can be biased by sampling, transport and culturing techniques, and fails to differentiate between colonisation and acute infection. Indeed, almost all approaches failed to differentiate airway colonisation from acute infection. This is significant as the presence of bacteria does not indicate disease, but rather the combination of dysbiosis with key pathogens, such as non-typeable Haemophilus influenzae, that leads to inflammation and acute exacerbation events. Furthermore, many patients experience co-infection with more than one organism, including co-infection with bacterial and viral pathogens. In these patients it is difficult to determine the relative effects of the different pathogens, and therefore the relative need for antibiotic treatment.