Relationship of CT-quantified emphysema, small airways disease and bronchial wall dimensions with physiological, inflammatory and infective measures in COPD

Chronic obstructive pulmonary disease (COPD) is a heterogeneous disease characterised by the progressive development of airflow limitation, leading to functional impairment and associated symptoms [1]. The clinical features and natural history of the condition vary considerably and traditional methods of measuring airflow obstruction do not reflect the complexity of the condition or the underlying pulmonary pathology. This has limited our understanding of COPD and consequently, there have not been the improvements in outcome seen in other chronic diseases and there are no significant disease-modifying medications. Other tools are required to help explain the heterogeneity and provide further insights into approaches to study and manage the condition.

Computed tomography (CT) can image the key pathological changes seen in COPD, including emphysema and remodelling of the large and small airways [2]. These pathologies both contribute to airflow obstruction and therefore CT has the potential to provide vital insights into the exact nature of the underlying pathophysiology. Newly developed automated image assessment techniques enable segmentation of the lung parenchyma and airways from the chest wall and surrounding structures, thus allowing quantitative analysis of emphysema, bronchial wall dimensions and small airways disease to characterise this heterogeneous disease more effectively.

Emphysema can be visualised as low attenuation areas on CT and by applying a density mask to the lung parenchyma, the percentage of voxels below − 950 Hounsfield (%LAA_<− 950) can be used to assess the quantity of emphysema. This method shows strong associations with histological measures of emphysema as well as physiological markers of disease [3‐5]. Using CT imaging, three-dimensional reconstructions of the bronchial tree can be generated down to the fifth or sixth generation airway, allowing analysis of airway wall dimensions. Thus far, studies have shown equivocal results, with some demonstrating thicker airway walls in COPD [6‐8] and others demonstrating thinner walls [3, 9]. This may be partly explained by the variation in methods used to describe airway wall dimensions and therefore a standardised parameter called Pi10 has been developed. This predicts the square root of the wall area for a hypothetical airway with an internal perimeter of 10 mm, giving a single value for bronchial wall thickening of the entire bronchial tree and has shown an inverse correlation with forced expiratory volume in one second (FEV1) [3]. Due to limited resolution, CT cannot image the small airways directly and instead quantification of air trapping can be used as a surrogate marker. A number of CT-derived methods exist to measure air trapping with one of the most widely used being the ratio of the mean lung density (MLD) in expiration to inspiration. In COPD, this measure has been found to correlate with a number of physiological and functional parameters [10‐12].

Recent analysis from large COPD cohort studies have added significantly to our understanding of the use of quantitative CT imaging in COPD, in particular exploring the relationships with pulmonary physiology [3, 13], functional ability [14, 15] and tracking the longitudinal progression of disease over time [13]. However, these large-scale studies allow only limited characterisation of their cohorts, with fewer detailed assessments possible. Furthermore, few studies have assessed biological markers of COPD, including measures of inflammation and infection. This further work is required to confirm CT analysis as an imaging biomarker of disease that can be a useful research tool to provide insights into the disease and as a clinical measure to guide management of patients with COPD. In the present study we performed quantitative CT analysis on a well-characterised cohort of patients with moderate to very severe COPD and explored the relationship with multiple clinical, physiological, microbiological and inflammatory indices to identify the utility of imaging to stratify disease to better predict outcomes and disease response.

Of 152 patients screened, 127 were enrolled onto the study. 122 of these had complete CT data as 5 of the scans could not be quantitatively analysed fully for technical reasons. All 122 had complete data for demographics, spirometry and CRP. 117 subjects had gas transfer data, 113 had plethysmography data, 120 had 6MWT parameters, 121 blood white cells and procalcitonin and 109 had fibrinogen. 106 subjects had sufficient sputum collected for microbiology culture while 98 had microbiology PCR performed. As sputum was prioritised for microbial analysis, only 75 subjects had enough for white cell differential and of these, 66 were deemed good quality samples (< 30% squamous cells). Baseline data for demographics, pulmonary physiology, biological and CT data are shown in Table 1. The majority of patients had moderate or severe airflow obstruction. There was no significant difference in age, gender, BMI or smoking status between the GOLD groups. As expected there was significantly lower FEV1% and FEF75–25% in the more severe GOLD groups. There was also significantly lower gas transfer and higher RV and RV/TLC ratio in subjects with severe and very severe COPD compared to moderate COPD. Subjects with very severe COPD had significantly lower 6MWD than subjects with either moderate or severe COPD. Subjects with severe and very severe COPD desaturated during exertion significantly more than those with moderate disease. There was no significant difference in sputum or blood inflammatory markers between the GOLD groups.

When assessing the CT parameters, there was significantly more emphysema and air trapping in severe and very severe COPD subjects compared to subjects with moderate COPD (Table 1). Pi10 was significantly raised in very severe COPD compared to subjects with severe COPD. Only 8 subjects had clinically significant bronchiectasis present on their CT scans and even then this was relatively mild with a median bronchiectasis score of 2.5. Given this low number of subjects with bronchiectasis, further analysis was not possible on this CT parameter. There was a significant positive association between %LAA_<− 950 and E/I MLD (rho = 0.47, p < 0.001) (Additional file 1: Figure S1). There was also an inverse relationship between %LAA_<− 950 and Pi10 (rho = − 0.36, p < 0.001) and a much weaker association between E/I MLD and Pi10 (rho = 0.18, p 0.045).

In the AERIS study we used sophisticated, cutting-edge techniques to explore the heterogeneity of COPD, resulting in an extremely well-characterised cohort of subjects. This was complemented by using CT analysis to understand the key structural changes identified in COPD and describing the relationship with a number of key physiological, functional and biological markers of disease. The main findings from the present analysis were the demonstration of a novel independent association between emphysema and desaturation on exertion, but interestingly not on exercise capacity. We also showed that CT-derived measurements of emphysema and small airways disease were independently associated with FEV1%. These CT measures also had strong associations with other pulmonary function tests, with emphysema having the strongest association with gas transfer and small airways disease having the strongest association with RV/TLC. None of the CT parameters were associated with sputum inflammatory cells or microbiology.

In this study, subjects exhibited a wide range of emphysema and small airways disease, with those in more severe GOLD categories having larger amounts, which is consistent with previous studies [3]. For dimensions of larger airways this was less clear-cut with only those in the most severe GOLD group having larger airway wall dimensions than severe COPD but not moderate COPD. Other studies have also shown inconsistent results when using Pi10 as a marker of bronchial wall thickening in COPD [3, 9, 21]. Bronchiectasis was not particularly common in this cohort, which is inconsistent with previous work that have shown higher rates in COPD populations [18, 22, 23]. The reasons for this are unknown, however subjects from this study were recruited from an area with low TB prevalence and most patients were recruited from specialist COPD clinics where clinically relevant bronchiectasis would have been detected and referred to specialist care. When assessing the association between these CT measures, a significant positive association was found between emphysema and small airways disease. However, the scatterplots demonstrated significant variability of small airways disease at lower levels of emphysema. It has been hypothesised that small airways disease may be a precursor to emphysema although more longitudinal work is required to understand this. There was also a weak inverse association between Pi10 and emphysema. Interestingly there was no real relationship between Pi10 and small airways disease, suggesting that structural abnormalities of the large and small airways are unrelated in COPD.

CT measures of emphysema and small airways disease had independent associations with airflow obstruction, which is in agreement with previous studies [3]. E/I MLD had the strongest association with FEV1%, suggesting small airways disease is the largest contributor to airflow obstruction in COPD. Emphysema and small airways disease also had significant associations with gas transfer, although in multivariate analysis %LAA_<− 950 was the only CT parameter had independent associations. As gas transfer is a measure of the disruption of the alveolar-capillary membrane it is unsurprising that emphysema was the only CT measure that showed this association and is consistent with prior studies [11, 24‐30]. E/I MLD was the only CT measure to independently predict RV% and RV/TLC ratio. RV and RV/TLC ratio are measures of pulmonary air trapping, therefore confirming E/I MLD as a an accurate marker of small airways disease, which has also been shown by other studies [10, 24, 31, 32]. Pi10 did not show an independent associations with any lung function parameters.

It is vital to understand the pathologies that contribute to poor exercise in COPD, as lower exercise capacity during the 6MWT has been linked with increased mortality [33‐37]. We found no associations between the 6MWD and CT parameters, with only FEV1%, BMI and gender showing an independent association. In addition, FEV1% was the only variable that significantly predicted which subjects walked less than 350 m. Previous work has suggested that subjects with increased emphysema [15, 27, 38, 39] and air trapping [11, 14, 38] walk shorter distances at the six minute walk test, although this is not a universal finding [40, 41]. The reasons for these apparent discrepancies are unknown but in our cohort there was significant variability in the 6MWD, which is likely to be influenced by a number of complex pulmonary and extra-pulmonary manifestations of disease. In addition, co-morbidities and psychological factors can have an effect on walking distance which were not investigated as part of this analysis. Desaturation during the 6MWT has also been linked with mortality in COPD patients [34, 42, 43]. On univariate analysis we found an association between desaturation and %LAA_<− 950 and E/I MLD but not Pi10. On multivariate analysis this association only remained significant for %LAA_<− 950. We also found that subjects who desaturated more than 5% on exertion had significantly lower FEV1% and higher %LAA_<− 950 and on multivariate analysis, %LAA_<− 950 was the only variable which independently predicted whether a subjects desaturated. We used > 5% to define significant desaturation as this has previously been linked to increasing mortality [43]. To the best of our knowledge the association between desaturation and CT parameters has not been previously investigated and our results suggest that emphysema, but not airways remodelling are directly linked to oxygen desaturation during exertion in COPD. This study does not provide any mechanistic explanations for this although a plausible suggestion would be the ventilation/perfusion mismatch that is caused by alveolar destruction seen in emphysema.

Few previous studies have investigated the link between CT parameters and biological markers of inflammation and microbiology. Although multiple studies have shown increased airway neutrophils in COPD, we found no association between any of the quantitative CT markers and sputum white cell differential. Previous studies have also demonstrated no link between emphysema and airway neutrophils [12, 32, 44, 45] and inconsistent results with sputum eosinophils [44, 46]. As opposed to this, a study using PET-CT demonstrated a relationship between pulmonary uptake of FDG, a marker of neutrophilic inflammation, and emphysema severity [47]. Prior work has also shown inconsistent results when investigating the link between CT parameters of gas trapping and airway neutrophil counts [12, 32], although these studies differed from ours as they utilised BAL sampling. Histological studies have also shown equivocal results regarding neutrophilia within the distal airways [48, 49]. The reason why our study and others fail to show a consistent link between neutrophils and emphysema or small airways disease is unknown, however it may be that the concentration of airway neutrophils at a single-time point is more influenced by other factors such as airway microbiology or smoking status. There was a suggestion of a negative relationship between emphysema and systemic inflammation, although the association was rather weak and so the relevance of this finding is unknown, which is further highlighted by a previous study showing increased fibrinogen but not CRP in emphysematous subjects [50].

We have previously shown in this cohort that colonisation of the airways with potentially pathogenic bacteria is important as it is related to exacerbations [17], making it vital to understand the underlying features which pre-dispose to this. We found few differences in CT parameters between subjects where bacteria were detected in sputum and those where it was not, by either culture or PCR. Few previous studies have investigated the link between structural changes identified on CT and microbiology, although one study supported our findings by demonstrating no difference in sputum bacterial culture between patients who had emphysema on CT and those who did not [46]. It is perhaps unsurprising that bacterial detection did not influence these structural changes as colonisation or invasion is likely to be due to abnormalities within the epithelial mucosal surface or immunological deficiencies that cannot be detected on CT. There are also obvious limitations on relying on sputum culture and PCR for bacterial detection and so further work is required to explore the relationship between the microbiome and structural changes within the lung.

There were a number of limitations to this study. Firstly, a number of different quantitative analysis techniques exist with which to measure each facet of structural lung disease, however we chose some of the most widely used and best-validated techniques. There is particular uncertainty for using Pi10 as a measure of airway wall dimensions, as simplifying the complexity of the bronchial tree into one number may be sub-optimal. Further work is required to understand which technique represents the best measure of airway wall dimensions in COPD. More recently, co-registration techniques such as parametric response mapping have been developed to estimate emphysema and small airways disease [51]. It has been proposed that this analysis method may be more accurate than those employed in this study although there is little direct evidence to support this currently. As is standard, we measured air trapping on CT by scanning subjects at maximal expiration. In some subjects, especially those with more severe COPD it may have been difficult to obtain residual volume which could have introduced errors. Furthermore, physiological and CT measurements of air trapping may not only reflect small airways disease but air trapping secondary to emphysema and therefore further work is required to validate CT measures by comparing them with more specific physiological measures of small airways disease. Another limitation is the multiple comparisons made in this study, however we found far more significant associations than would be expected by chance alone. For a CT study the sample size was relatively small, which may have limited the statistical power of the study, although the benefit of this was that it allowed us to perform substantial and in-depth phenotyping of the cohort. In addition, the results reported in this study were cross-sectional and some of the variables may vary over time and therefore further longitudinal studies are required to understand this.

In conclusion, CT-derived measurements of emphysema and E/I MLD had independent associations with FEV1%, providing further evidence that both of these structural abnormalities are directly associated with the airflow obstruction that defines COPD. We also provided further validation for their use, as CT-measures of emphysema showed strong associations with gas transfer and CT-measure of gas trapping with RV. Furthermore, quantitative CT measures of emphysema have independent associations with desaturation on exertion, a measure which has been linked to mortality and therefore may help explain the mechanism by which emphysema is related to mortality. In contrast to this, quantitative CT parameters did not show any convincing associations with biological markers of disease. This suggests that while quantitative CT measures of emphysema and small airways disease are useful tools at reflecting pulmonary physiology and functional status they are less effective at reflecting disease biology and therefore may not be a useful tool in assessing disease activity.