Lung density on high resolution computer tomography (HRCT) reflects degree of inflammation in smokers

Cigarette smoke induces an inflammatory reaction in the lung and is a major risk factor for a number of lung diseases such as chronic obstructive pulmonary disease (COPD) and diffuse parenchymal lung diseases [1]. Further, cigarette smoking leads to elevation of both cells and soluble markers of systemic inflammation in the circulation [2]. Infiltration of macrophages and mononuclear cells in the lung leads to tissue damage and release of numerous inflammatory mediators resulting in an increased epithelial permeability and oedema in the lung interstitium [3]. Both local and systemic inflammation in smokers may be present before any significant clinical symptoms appear [2].

Computer tomography (CT) imaging of the lung can non-invasively detect and quantify lung abnormalities [4‐7]. Early changes in airways and lung parenchyma may be recognized in smoking individuals with normal pulmonary function before any signs of lung function impairment [8, 9]. In CT scans, attenuation is measured by Hounsfield Units (HU), where attenuation for water is defined as 0 HU and air as −1000 HU [10]. Several authors have focused on the region of the lung with low attenuation i.e. <−910 or −950 HU as a quantitative assessment of emphysema. These studies are usually performed on subjects with advanced emphysema or on heavy smokers included in lung cancer screening programs [10‐12]. Despite the potential of CT to non-invasively detect and quantify early subclinical pathological changes such as increased density, studies in this field are scarce.

Given the intense systemic and local inflammation induced by smoking, we asked whether this may be mirrored by alterations in the high density spectrum assessed by high resolution CT. We therefore hypothesized that lung density were associated with an inflammatory response, and to test this a group of smokers with normal pulmonary function, a matched group of healthy never-smokers and a group of patients with COPD (GOLD I-II) underwent inspiratory CT examinations. The percentage of the lung parenchyma in the high density spectrum, i.e. with attenuation between −750 to −900 HU (%HDS) was calculated within pre-defined thresholds. We also analysed blood samples and performed bronchoscopy and bronchoalveolar lavage for cell concentrations and differential cell counts as measures of systemic and local inflammation.

In this study we aimed to quantify the impact of chronic cigarette smoking on lung density on HRCT by studying smokers with normal lung function and COPD patients. We demonstrated that lung density was positively correlated to measures of local inflammation, i.e. the total cell concentration as well as concentrations of macrophages and neutrophils in BAL. In addition, females had denser lungs than males, regardless of smoking habits.

Earlier morphological studies on the lung of smokers have revealed that inflammation in small airways and lung parenchyma precedes architectural distortion and emphysema [8, 16, 22] and Remy-Jardin et al. [23] demonstrated that early subtle parenchymal inflammation in normal smokers correlated to development of emphysema. Our findings with increased lung attenuation in smokers are in line with previous reports [12, 23‐25], although few studies have been performed including healthy never-smokers. Lederer et al. [24] reported a positive correlation between lung density and cumulative and current smoking history. Shaker et al. [12] reported a significantly increased lung density in current smokers which decreased after smoking cessation [26]. During a 5-year follow-up, Soejima et al. [25] found that the density on inspiratory CT increased in smokers compared to never-smokers while it decreased after smoking cessation.

Cigarette smoking induces a dose-related inflammatory response in the lungs with accumulation of macrophages and neutrophils into the lower respiratory tract [27, 28]. In agreement with these previous reports, we found that the numbers of macrophages and neutrophils were significantly raised in smokers with normal lung function and in current smoking COPD patients. The inflammatory process seems to be reversible, at least in terms of these parameters, since COPD ex-smokers exhibited cell number in BAL fluid equal to those of never-smokers.

We demonstrated a positive correlation between lung density and cell concentration in BAL fluid both in smokers with normal lung function and in current smoking COPD patients. Lung density does not solely depend on the presence of cells in the alveolar space and in the lung interstitium but may also depend on other factors such as oedema and peribronchiolar fibrosis. We consider, however, that cell concentration in BAL fluid could be regarded as an overall indicator of the inflammatory process in the lung. The fact that DLCO was reduced in smokers with normal lung function suggests an increased barrier for oxygen diffusion due to inflammation in the alveolar interstitium rather than reduced area for gas exchange due to emphysema, while the latter may be dominant in COPD patients.

An interesting finding is that lung density in our entire group of COPD patients was lower than the smokers’ and did not differ from never-smokers. A plausible explanation may be that areas with decreased lung density in COPD patients (i.e. emphysema) and areas with increased lung density (i.e. accumulation of inflammatory cells) may occur in the same patient but in different regions of the lung. This results in an average density in the COPD-patients’ whole lung, that does not differ from never-smokers. When the COPD patients were sub-grouped in ex-smokers and current smokers, the latter had higher density. This suggests that inflammation due to smoking (i.e. increased lung density) overshadows potential reduction in lung density due to parenchymal destruction. In COPD ex-smokers, the absence of smoke-induced cell infiltration in the lower airways results in a lower density, probably due to parenchymal destruction and emphysema. This hypothesis is also supported by our finding that total cell concentration in BAL fluid was significantly lower in COPD ex-smokers than in COPD current smokers.

It is known that smokers and COPD patients have a low grade of systemic inflammation indicated by elevated serum markers [2, 29]. We found significantly higher levels of white blood cells, orosomucoid and haptoglobin but decreased level of immunoglobulin G in serum of smokers and COPD patients compared to never-smokers. In smokers and COPD patients, lung density was positively correlated with the concentration of haptoglobin in serum, but not to C reactive protein (CRP). Haptoglobin is an acute phase protein which is elevated in inflammatory and infectious conditions and may be a sensitive marker for low grade systemic inflammation. The levels of immunoglobulin G were significantly lower in smokers and in COPD current smokers compared with never-smokers, which reassures an immunosuppressive effect of cigarette smoking [30].

Previous reports have shown gender differences for lung function impairment and symptoms related to smoking [31‐33]. Dransfield et al. [34] reported that the attenuation area lower than −950 HU on CT was greater in male than in female COPD patients, suggesting more severe emphysema in men. Other investigators have reported a more rapid progression of emphysema in females [12, 32]. Gevenois et al. [35], did not find any gender differences when they measured mean lung attenuation in never-smokers. We found that women had significantly higher lung density than men regardless of smoking status, but the difference were no longer present when adjustment for height in multivariate models was made. Thus, the gender difference is probably anatomically related and suggests that women may have more lung tissue per volume compared to men. Smaller ribcage and shorter diaphragm in females are reported previously [36]. We could not observe any gender difference regarding BAL cell count or lung function parameter including DLCO. Lower percentage of BAL fluid recovery in male smokers may indicate different response to cigarette smoke between genders.

Our study has some limitations. The number of participants with COPD was low, which makes the subgroup analysis uncertain with regard to bronchitis and airways obstruction. The clinical utility of CT for characterization of airway obstruction in phenotypes with parenchymal destruction or mainly airway predominant disease has been promising in some studies. However there is a considerable overlap between these phenotypes especially in advanced stages of COPD [37, 38]. The question rises if dividing our COPD patients in two main parenchymal and airway predominant phenotypes produce more correct correlation to inflammatory measurers in BAL in this group. We deliberately included only COPD patients of GOLD stage I and II to detect subtle parenchymal changes. Secondly, measurement of small airways i.e. with diameter less than 2 mm is difficult to perform accurately. The lower attenuation limit of −900 HU considered to be rational based on a report demonstrating pathological specimens from patients with emphysema comprise more voxels with attenuation lower than −900 HU [18]. Areas with a density of -300 to -750 HU were calculated separately since this may represent pixels including tissue such as small bronchi and vessels. Thus, we consider that areas with a density of -750 and -900 HU represent lung tissue with minimal interference from emphysema and denser structures and this area can represent a good surrogate for measuring the degree of local inflammation in the lung. The inflammatory process in smokers is mainly concentrated in conductive airways (with diameter less than 0, 5 mm), lung parenchyma and alveolar space [8, 39] and cells recruited by BAL represent these areas. A lung attenuation spectrum on CT which may represent these corresponding areas is thus motivated. The fact that lung density in the spectrum between −300 HU and −750 HU did not differ between the three studied groups may indicate that the inflammatory changes with impact on lung density occur mainly in the conductive airways and in the lung parenchyma.

Lung density is influenced by a number of factors including the level of inspiration. To validate this important factor we compared the total lung volume measured by CT and the total lung capacity measured by body plethysmography. We found a good agreement in all three groups indicating that the level of inspiration had a negligible effect on our results. Lung density was negatively correlated to TLC but not age in never-smokers and this finding is in agreement with a study from Gevenois et al. [35].

The strength of our study is that we included a group of healthy never-smokers of both genders. Further, all measurements, including CT scans and BAL, were performed locally with same equipment and followed standardized procedures. We believe that our results can be valuable for selection of patients in research and clinical trials dealing with CT and smoking associated pulmonary diseases. By following attenuation trend inter-individually it may be possible to determine when the preceding inflammation “switches” to emphysema. The findings may help to interpret high resolution CT in the context of smoking and gender and highlight the heterogeneity of structural changes in COPD.

In conclusion, smokers have denser lungs than never-smokers and females have denser lungs than males. The density in smokers is associated to measures of systemic inflammation and cell concentration in BAL. The results support our hypothesis that smoking causes an inflammation in the lungs which can be quantified non-invasively by means of CT density.