Imaging of sarcoidosis of the airways and lung parenchyma and correlation with lung function

Scadding staging

Scadding [2] classified postero–anterior chest radiography findings in sarcoidosis into five stages: stage 0 (normal); stage I (BHL); stage II (BHL accompanied by pulmonary infiltrates); stage III (pulmonary infiltrates without BHL); and stage IV (overt pulmonary fibrosis). The distribution of patients according to radiographic stages largely depends on geographic or ethnic origin and referral source. Overall, the frequency of each stage at presentation is reported as: stage 0, 5–15%; stage I, 25–65%; stage II, 20–40%; stage III, 10–15%; and stage IV, approximately 5% (table 1) [1, 3–5].

A major shortcoming with this staging classification is the poor reproducibility of reading. In a recent study, the overall agreement between two expert radiologists was fair (weighted κ=0.43). The two major problems of interpretation were the assessment of lymphadenopathy and the presence of fibrosis [11]. Surprisingly, the interobserver agreement was quite good in another study (weighted κ>0.8), including for the adjudication of stage IV disease [12].

Radiographic staging and disease outcome

Despite its purely descriptive nature, Scadding radiographic staging supplies major prognostic information. Although individual exceptions exist, there is clearly an inverse relationship between stage at presentation and the probability of recovery from sarcoidosis. Spontaneous resolution occurs in 60–90% of patients with stage I disease, 40–70% of those with stage II and 10–20% of those with stage III and does not occur with stage IV disease (table 1) [1, 4, 5]. Regardless of the initial stage, the majority of spontaneous remissions occur within the first 2 years after presentation. The likelihood of remission is reduced after 5 years, but patients do not necessarily show a stepwise progression from stage 0 to stage IV disease. In a large patient cohort, only 9% of stage I patients had progressed to stage II (and 1.6% to stages III or IV) at 5 years, while only 5.5% of stage II patients had progressed to stage IV [15]. Stage I disease persists after 5 years in 8% of cases, which is not inevitably a sign of ongoing activity. Stage IV usually develops after 5 or more years of disease. It is associated with substantial morbidity and higher mortality [16, 17].

Correlations of chest radiography with baseline lung function

In individual cases, radiographic findings do not, in themselves, discriminate reliably between active inflammation and fibrosis. Moreover, correlations between radiographic staging and other pulmonary parameters such as the degree of physiological impairment, dyspnoea [18] and 6-min walk test [19] are imprecise. Overall, PFTs are abnormal in about 20% of patients with stage I disease, and 40–80% of those with pulmonary infiltrates (stage II, III or IV) [3, 4, 20]. The frequency and severity of restrictive defect and reduced diffusing capacity of the lung for carbon monoxide (D_L,CO) usually increase in more advanced stages. However, even when chest radiography is normal, alterations in forced vital capacity (FVC) and D_L,CO are noted in 15–25% and 25–50% of cases, respectively [4, 20]. Although occurring at all stages of disease, airflow limitation is more frequent with progressive staging: reported frequencies of airflow limitation are: 2–25% in stage 0; 7–12% in stage I; 8–38% in stage II; 10–23% in stage III; and 45–71% en stage IV [21–23]. Stage IV disease has been found to be an independent index that relates to lower forced expiratory volume in 1 s (FEV₁)/FVC in a large prospective study [22] and some patients with stage IV exhibit severe obstructive defects [24, 25].

Exercise gas exchange measurements vary significantly with radiographic stage [26, 27]. However, resting D_L,CO has a much superior performance than chest radiography in predicting exercise-induced desaturation [26, 27]. Arterial desaturation is usually not observed in stage 0–I disease [27]. In patients with stage II–IV disease, there is no further independent improvement in variable prediction when radiographic staging is considered in addition to D_L,CO [27]. In one study, radiographic stage was found to be more significantly associated with alveolar-arterial oxygen tension difference (P_(A-a),O2) than with D_L,CO in patients with stage 0–II disease [26]. Semi-quantitative chest radiography scoring systems improve the correlations with baseline physiological impairment [13] and changes during follow-up (discussed later in this article).

Correlations between changes in serial chest radiographs and lung function

Chest radiography is inexpensive, noninvasive and easily accessible. Together with serial PFTs, it is traditionally considered the cornerstone of monitoring of sarcoidosis patients [1]. However, there are few data to validate this role of radiography.

In a prospective study of long-term corticosteroids in pulmonary sarcoidosis, changes in chest radiography were analysed against physiological parameters over 5 years. Significant correlations were demonstrated with R and F scores on the Muers scoring system. Despite low correlation coefficients, R score improved in parallel with spirometry and D_L,CO and a worsening F score was associated with deteriorating FEV₁ and D_L,CO [13]. Similarly, Baughman et al. [11] evaluated the sensitivity of chest radiography to detect treatment effect in the prospective trial of infliximab for pulmonary sarcoidosis [11]. The films were compared using two methods: the Muers scoring system and the five-point Likert scale global assessment of change (markedly worsened, worsened, unchanged, improved, and markedly improved). Interobserver agreement was good for the two methods (weighted κ=0.61 for the Likert scale). Only the R score showed a significant improvement with therapy. The initial R score was positively correlated with subsequent improvement in FVC under therapy. There was also a significant correlation between the changes in FVC and both the changes in R score and the global assessment. Correlation was better for the latter [11], suggesting that the evaluation of a simple instrument, easily applicable in routine, was at least as robust as the complex Muers scoring system.

In a large retrospective study, Zappala et al. [12] also examined two methods of serial chest radiographic scoring in relation to serial pulmonary function trends, irrespective of treatment effect [12]. Changes in Scadding radiographic stage and disease extent were condensed into a three-point scale: reduction, stability and increase. Despite fair to good interobserver agreement for each of the two methods (weighted κ>0.76 for the change in disease extent), the relationship between changes in stage and disease extent at 2 and 4 years of follow-up was poor. Changes in disease extent were linked to FEV₁, FVC and D_L,CO variations, whereas changes in stage were not. However, discordance between PFTs and disease extent data was still observed in approximately 50% of cases. In the remaining cases, isolated change in disease extent was more frequent than isolated change in PFTs, but in 6%, disease extent and PFTs exhibited change in opposing directions [12].

The utility of chest radiography in diagnosing exacerbations of pulmonary sarcoidosis was addressed by Judson et al. [28], using the profusion score of the ILO reading system. Exacerbation was defined by worsening pulmonary symptoms thought to be related to pulmonary sarcoidosis (i.e. not infection) that responded to an increase of the corticosteroid dose. A subset of these patients (66.7%) was identified with a spirometric decline of ≥10% in FVC or FEV₁ from the previous baseline visit. Interobserver agreement in the determination of profusion score appeared to be moderate (weighted κ=0.54). Although changes in profusion score tended to worsen during exacerbations in both the whole population and the spirometric decline subgroup, a large percentage of readings demonstrated no modification (32% and 27%, respectively) or an improvement (19% and 23%, respectively). As expected, a significant decrease in spirometric measurements was seen during exacerbations. There were no statistically significant correlations between changes in the profusion score and changes in FVC or FEV₁ for either the whole population or the spirometric decline subgroup [28]. The authors concluded that radiography was inadequate to reliably detect exacerbations of pulmonary sarcoidosis.

Taken together, these results highlight the weaknesses of chest radiography for the monitoring of sarcoidosis. However, a further advantage of serial chest radiographs over PFTs is the detection of pulmonary complications, including aspergilloma. From this point of view, it is the combination of the two tests that is, in essence, helpful in clinical practice.

Chest radiography as a guide for treatment decisions

Treatment for pulmonary sarcoidosis is usually dictated by the severity of symptoms and/or functional impairment, but chest radiography per se may also guide therapeutic intervention [29–31]. A prospective randomised study conducted by the British Thoracic Society provides some support for long-term corticosteroid therapy in asymptomatic patients with persistent pulmonary infiltrates for at least 6 months [29]. After adjustment for possible confounding factors, the average difference in vital capacity between treated and untreated groups at final assessment was 9% of the predicted value. In another placebo-controlled study, patients with newly detected stage I or II–III disease and (sub)normal lung function were immediately treated with oral prednisolone for 3 months, followed by inhaled budesonide for 15 months [30]. Treated patients with initial stage II–III, but not stage I, disease improved significantly more in FVC and D_L,CO compared with patients on placebo. Functional benefits were maintained after 5 years, but differences were small [31]. It remains uncertain whether modest improvement in asymptomatic stage II–III patients is clinically pertinent and justifies the morbidity associated with treatment.

Thoracic lymphadenopathy

Although not necessary in typical stage I disease, CT is more sensitive to detect enlarged lymph nodes than a chest radiograph [57]. Overall, hilar or mediastinal lymphadenopathy are encountered on CT in 47–94% of patients with sarcoidosis, irrespective of radiographic staging [35, 42, 57–60]. Lymph nodes are usually bilateral but with right-sided predominance [57–60]. In two studies based on the American Thoracic Society lymph node map, the most commonly involved nodal stations, in decreasing order of frequency, were: 4R (right lower paratracheal); 10R (right hilar); 7 (sub-carinal); 5 (sub-aortic, i.e. aorto-pulmonic window); 11R (right interlobar); and 11 L (left interlobar) [59, 60]. A median of three lymph nodes were enlarged, as defined by a maximum short-axis diameter ≥10 mm, and diameter was ≥20 mm in 29.1% of cases [59]. Lymphadenopathy is also common in collagen vascular disease (present in 70% of cases), idiopathic pulmonary fibrosis (IPF) (67%), extrinsic allergic alveolitis (53%) and organising pneumonia (36%) [59]. However, the number of enlarged lymph nodes is higher in sarcoidosis and a size greater than 20 mm in short-axis diameter increases the chance of sarcoidosis [59].

Sarcoidosis lymph nodes are usually non-necrotic and noncompressive, with nodal calcification frequent in longstanding disease. Calcification is present at presentation in 20% of cases, increasing to 44% 4 years later [42], with egg-shell aspect present in 9% [58]. The mean diameter of calcified nodes is significantly larger in sarcoidosis, calcium deposition is more commonly focal in sarcoidosis and diffuse in tuberculosis and, when present, hilar nodal calcification is much more likely to be bilateral in sarcoidosis than in tuberculosis (65% versus 8%) [58].

Lymphadenopathy may be unilateral or localised in an unusual site. Although possible in sarcoidosis, the enlargement of internal mammary and pericardial lymph nodes requires exclusion of lymphoma. Enlarged lymph nodes, essentially when calcified, and/or mediastinal fibrosis can rarely provoke extrinsic compression on adjacent organs, including bronchi [61], large pulmonary arteries [36] or veins [62, 63], the superior vena cava [64], the oesophagus [65], the left recurrent nerve [66] or the thoracic duct [67].

Pulmonary patterns

Nodular and alveolar opacities

Small nodules can coalesce into larger ones or masses, which can very rarely cavitate [73–75]. Alveolar or pseudoalveolar consolidations are seen in 12–38% of patients [35, 41, 44, 45, 54, 76]. However, a predominance of multiple large nodules/masses and multifocal consolidations is uncommon in sarcoidosis and can mimic organising pneumonia or malignancy. According to a recent series by Malaisamy et al. [74], the presentation of such a form of disease is usually acute and symptomatic with an excellent prognosis and it may particularly affect smokers [74]. Nodules/masses and consolidations are homogeneous or inhomogeneous, measure 10–80 mm in diameter, and are usually located in the mid/upper lobes, along the bronchovascular bundles or sub-pleurally, sometimes with the presence of a central air bronchogram [74, 77]. They are characterised by ill-defined contours as they fade to a micronodular pattern toward the surrounding lung (fig. 2). In addition to these lesions, other more representative abnormalities such as lymph nodes are usually associated [74, 77]. Solitary mass-like nodule and alveolar consolidation are exceedingly rare in sarcoidosis [74, 78].

• Download figure
• Open in new tab
• Download powerpoint

Figure 2–

a) Chest radiograph and b) computed tomographic image of “alveolar” sarcoidosis associated with profuse micronodular involvement and cavitary lesions on the left.

Necrotising sarcoid granulomatosis, which is defined pathologically by a sarcoid-like granulomatous reaction with vasculitis (involving both arteries and veins) and noncaseating necrosis, shares many clinical and radiological features with alveolar sarcoidosis but is traditionally viewed as a separate entity [79].

Cavitary lesions in sarcoidosis are believed to result from either ischaemic necrosis (with extrusion of hyaline material from conglomerate granulomas) or angiitis [73]. Reported rates of cavitary lesions were 3.4% and 6.8% in two CT studies, respectively [40, 75]. According to a recent series by Hours et al. [73], cavitary lesions manifest as thin-walled cysts in most cases or as cavities with thick wall or developing inside nodules or condensations [73] (fig. 2). Cavitary lesions are variable in size and, although occasionally found in isolation, they are more likely to be multiple and bilateral. They usually arise in patients with severe and active sarcoidosis. The evolution of cavitary lesions is variable. A wall thinning is usually observed under treatment whereas a wall thickening is always associated with an infectious complication [73]. Pneumothorax can occur [73, 80]. Because primary cavitary sarcoidosis is rare, granulomatosis with polyangiitis (Wegener’s) and superimposed infection should always be excluded.

“Sarcoid galaxy” and “sarcoid cluster” signs

Two terms have recently been coined in sarcoidosis to describe a particular distribution of nodules: the “sarcoid galaxy” sign (fig. 3a) and the “sarcoid cluster” (fig. 3b) sign [81]. The “sarcoid galaxy” is a large nodule, usually with irregular boundaries, encircled by a rim of numerous tiny satellite nodules. It is usually multiple throughout the lungs. Pathologically, the “sarcoid galaxy” represents innumerable coalescent granulomas that are much more concentrated in the centre of the lesion than at the periphery [75]. The “sarcoid cluster” also corresponds to rounded or long clusters of many small nodules that are close to each other but, in contrast to those of the “sarcoid galaxy”, not confluent. Pathologically, the “sarcoid cluster” represents granulomas without coalescence [82]. Although initially reported as a specific manifestation of sarcoidosis, these two signs have subsequently also been identified in other granulomatous diseases, including tuberculosis, silicosis and cryptococosis. Certain associated features help to differentiate these diseases [81].

• Download figure
• Open in new tab
• Download powerpoint

Figure 3–

Computed tomographic images of a) the “sarcoid galaxy” sign (irregular nodule resulting of the confluence of numerous micronodules), b) the “sarcoid cluster” sign (clusters of micronodules without confluence) and c) the “reversed halo” sign (ring of micronodules surrounding central ground-glass opacity, which is a very rare finding in sarcoidosis).

Ground-glass opacification

The frequency of ground-glass opacity varies from 16–83% [35, 41, 42, 44, 45, 53, 76]. The pathological significance of ground-glass opacity is not univocal, as this CT feature can result from the confluence of multiple micronodular granulomas and/or from fibrotic lesions [68]. Ground-glass opacity seems to be more common at presentation than later in the disease course [40]. Although seldom the predominant abnormality [40, 44, 45, 48, 50, 52], ground-glass opacity is usually multifocal, rather than extensive. Extensive ground-glass opacity tends to have a superior predominance, with ill-defined margins, and it is often overlaid on a background of subtle micronodularity (fig. 4) and associated with lymph nodes [83].

• Download figure
• Open in new tab
• Download powerpoint

Figure 4–

Computed tomographic images of a 35-yr-old female with pulmonary sarcoidosis and extensive ground-glass opacification. The diagnosis of sarcoidosis is suspected given the association with more characteristic signs of sarcoidosis, including a) bronchial distortion in the superior and posterior regions and b) the presence of few micronodules with a perilymphatic distribution (boxed area).

Scarring and fibrosis

Lung architectural distortion is constant in stage IV disease [35, 40], and, overall, is reported in 20–50% of patients [40, 42, 44, 76]. It includes abnormal displacement of hila, fissures, bronchovascular bundles and distortion of secondary pulmonary lobules. Posterior displacement of the main or upper lobe bronchus and volume loss (particularly in the posterior segment of upper lobes) are characteristic features of fibrotic sarcoidosis [35]. Bronchi may be deformed, angulated, crossed or stenosed [35]. Conglomerate masses often surround and encompass the bronchi and vessels. They are a frequent feature in stage IV disease (60%) and are usually central and associated with bronchial distortion [35] (fig. 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5–

a) Chest radiograph and b, c) computed tomographic images of a 50-yr-old female with obstructive pattern (forced expiratory volume in 1 s/forced vital capacity ratio 68%) showing conglomerated masses in the upper lobes with posterior and superior displacement of the main bronchus and distorted airways (angulation, dilatation). Notably, scar emphysema is observed in the anterior part of the lung with reticulations on the left. The numerous nodules and micronodules suggest probable reversibility of lesions. d) Volume rendering technique image of the bronchi improves the detection of multiple bronchial stenoses (long arrows) and traction bronchiectasis (short arrows).

In advanced fibrotic sarcoidosis, traction bronchiectasis, honeycombing, other types of cystic destruction, bullae and paracicatricial emphysema are encountered, mainly in mid/upper areas. Three distinct patterns of distribution have been recognised in stage IV disease [35]: the bronchial distortion pattern (47% of patients), with or without co-existent masses; the honeycombing pattern (29%); and the linear pattern (24%). These patterns are associated with different functional profiles (see below) [35]. In a serial CT study, ground-glass opacities and consolidations seen on the initial CT scan tended to evolve into honeycombing whereas conglomeration masses shrank and evolved into bronchial distortion [45].

When present in sarcoidosis, honeycombing usually affects the upper and perihilar regions [35] (fig. 6a), which is very dissimilar from IPF. Interestingly, a pattern of IPF-like honeycombing has been exceptionally reported in sarcoidosis, showing a striking prominence in the lower lung zones with a peripheral and subpleural distribution, or a diffuse distribution [35, 84]. Pathological examination is available for a handful of patients. A particular localisation of granulomas has been mentioned along the alveolar septa [84, 85], together with a so-called “chronic interstitial pneumonia” sometimes taking the appearance of usual interstitial pneumonia [84–87].

• Download figure
• Open in new tab
• Download powerpoint

Figure 6–

Computed tomographic (CT) images of a 67-yr-old female with fibrosing pulmonary sarcoidosis with honeycombing pattern and pulmonary hypertension. a) Honeycombing predominates in the upper and perihilar regions, and along the bronchovascular bundles. Note the various sizes of the cysts. b) Contrast-enhanced CT shows the widest diameter of the main pulmonary artery with a diameter superior to that of the ascending aorta and slight vascular compression by enlarged mediastinal lymph nodes.

Diagnostic accuracy of CT versus clinical diagnosis and chest radiography

The accuracy of HRCT for the diagnosis of DLDs depends on the underlying disease. Grenier et al. [98] evaluated supplementary information yielded by HRCT after taking into account clinical data and radiographic findings [98]. Computer-aided diagnoses were made by applying a Bayesian model in a large population of DLDs, including patients with pulmonary sarcoidosis. For sarcoidosis, the percentages of correct diagnosis with a high level of confidence were 33–42% with clinical data alone, 52–76% with clinical and radiographic findings and 78–80% when HRCT findings were integrated [98]. This relatively small difference emphasises the opinion that CT adds little information when a confident diagnosis of sarcoidosis can be made from typical clinical and radiographic presentation [99]. However, it understates the true compelling diagnostic role of CT in other less straightforward cases.

Diagnostic contribution of CT

In the appropriate clinical setting, the observation of the characteristic picture of sarcoidosis on CT (bilateral hilar lymph node enlargement with perilymphatic micronodular pattern) points to a highly confident (virtually pathognomonic) diagnosis (table 2 and fig. 1). CT can reveal these abnormalities lying below the resolution limits of chest radiography, which is important when there is a clinical suspicion of sarcoidosis with normal chest radiography. CT, in concert with several other standard diagnostic tools, increases the diagnostic likelihood of sarcoidosis, in particular when the disease is confined to one extrapulmonary organ and the biopsy is considered too risky (as in central nervous system involvement). It seems especially useful in the evaluation of elderly females with uveitis [100, 101].

In atypical clinical and/or radiographic variants of sarcoidosis, there is a broad variety of differential diagnoses. In radiological practice, the diagnosis of interstitial lung diseases is usually based on the identification of a predominant pattern on CT. In sarcoidosis, the combination of features is at least as decisive. Bilateral hilar lymphadenopathy, when large, and/or perilymphatic micronodular pattern are probably the most important findings, whether predominant or not. These signs, which can be subtle and must be looked for carefully by the radiologist, are highly prevalent except in advanced pulmonary fibrosis. They are present even in rather uncommon patterns, in particular in alveolar or ground-glass patterns (fig. 4), this association improving the level of diagnostic confidence considerably (table 2). Proximal bronchial distortion is also an important finding in fibrotic cases. Using a logical analysis of data technique, Martin et al. [83] recently demonstrated the value of combining CT features for the diagnosis of sarcoidosis with predominant ground-glass opacification.

CT as an aid to obtaining biopsy/cytology material

There is a clear association between bronchial abnormalities seen on CT and the presence of mucosal granulomas on endobronchial biopsy [37]. Similarly, de Boer et al. [38] recently demonstrated that the total extent of parenchymal disease on CT in addition to the pattern and lobar distribution, i.e. reticular pattern and ground-glass opacification, predicted the likelihood of a positive transbronchial biopsy at bronchoscopy. CT may help to guide EBUS-TBNA [39]. Lastly, CT-guided transthoracic needle biopsy of mediastinal lymph nodes [102] or lung nodules [74] is occasionally very helpful in establishing the diagnosis.

Correlations between chest CT and disease activity

Correlations between CT findings and classic markers of sarcoidosis activity, including serum angiotensin-converting enzyme (SACE) levels, bronchoalveolar lavage (BAL) lymphocytosis, and gallium scan signal, are inconclusive or discrepant [40, 41, 43, 44, 46]. Much more has been learned from studies of the reversibility of CT features (spontaneously or under treatment) on serial examinations (table 3) [42, 44, 45, 76]. Architectural distortion, traction bronchiectasis, honeycombing and bullae are consistently irreversible. Micronodules, nodules, peribronchovascular thickening and consolidation are wholly or partially reversible in most, but not all, cases. The evolution of ground-glass and linear opacities is more variable. Ground-glass opacity may steady, worsen or improve over time, reflecting the fact that it may represent either granulomas or fine fibrosis [68, 103]. A coarse texture or concomitant traction bronchiectasis increases the likelihood of underlying fibrosis [103]. Similarly, septal thickening from intense granulomatous infiltration tends to reverse, whereas irregular distorted lines are more likely to be fibrotic.

Thus, discrimination between active inflammation and irreversible fibrosis with CT may occasionally be helpful when the decision to initiate or continue potentially toxic treatment is marginal. For instance, in stage IV disease a trial of therapy might be warranted if a potentially reversible component is still visible on CT [35]. This role of CT may be supplanted by PET-CT in the future.

Correlations between chest CT and baseline lung function

Many studies have explored the relationships between CT findings and PFTs [22, 23, 35, 40, 44–48, 50–54] with variable results, probably reflecting the diversity of imaging analysis and scoring. In the study of Remy-Jardin et al. [44], lungs were divided into three zones and a percentage involvement score was assigned for abnormal parenchymal patterns (nodules, consolidation, lung distortion, septal and nonseptal linear, ground-glass opacity and honeycombing). The overall extent of disease was the summation of the scores for each type of abnormality. Profusion of septal lines was the only CT finding that correlated with initial disease activity (as assessed by SACE levels and BAL lymphocytosis). Significant but low correlation was observed between the scores of CT abnormalities and either initial FEV₁, FVC or D_L,CO, except for nodules. CT findings could not help predict the further evolution of disease activity and functional changes over time [44]. A more simple semi-quantitative CT scoring system, first described by Oberstein et al. [43], served in the correlation study of Drent et al. [48]. This consisted of coarse quantification of the extent of abnormal parenchymal patterns (thickening or irregularity of the bronchovascular bundle, nodules, septal and nonseptal lines, and consolidation, including ground-glass opacity) using a four-point scale (0 = no lesions found; 1 = up to 33%; 2 = up to 66%; and 3 = more than 66%), and the extent of focal pleural thickening and lymph node enlargement (0 = no pathological findings; 1 = minor; 2 = moderate; and 3 = pronounced changes). The total score was obtained by adding the individual subscores. Interobserver agreement was moderate for the subscores (weighted κ 0.34–0.65, with worse results for bronchovascular bundle and lymph nodes), but excellent results for the total score (intraclass correlation coefficient 0.99). All CT subscores, except lymph nodes enlargement, were correlated with FEV₁, FVC, D_L,CO, maximal arterial oxygen tension and maximal P_(A-a),O2, whereas the chest radiographic stage was not.

In other studies, CT was classified according to subjective appraisal of the predominant pattern of involvement. Lopes et al. [51] investigated the relationship between outcome measures of cardiopulmonary exercise testing (CPET) and predominant CT pattern: nodules, ground-glass opacity, and traction bronchiectasis plus honeycombing. PFTs and CPET results were close to normal only for patients with predominant nodules. Patients with predominant ground-glass opacity showed intermediate values of FEV₁, FVC and D_L,CO compared with patients of the other two groups. Interestingly, only the CPET results were able to differentiate patients with predominant ground-glass opacity and those with predominant traction bronchiectasis and honeycombing, with a significantly decreased peak oxygen uptake and breathing reserve, and increased P_(A-a),O2 for the latter [51]. As discussed previously, Abehsera et al. [35] separated three CT patterns of fibrotic pulmonary sarcoidosis, which were associated with different functional profiles. The interobserver agreement for recognising the main CT pattern was very good as observers agreed in 80% of cases (κ=0.87). Bronchial distortion pattern was associated with lower expiratory airflow rates, while honeycombing pattern was associated with restriction and lower D_L,CO, and functional impairment was relatively minor when linear pattern predominated [35].

In summary, the degree of functional alteration is usually linked to an overall CT score but most correlations are weak [40, 46, 48, 53]. Total CT score seems superior to chest radiographic staging [48] but its real clinical impact is questionable. Correlations between PFTs and the score of specific patterns are hardly interpretable because of considerable overlap among CT categories, but they sometimes give clues to the understanding of disease complications, in particular airflow limitation [22, 23, 50]. The reported frequency of airflow limitation is wide in sarcoidosis, with an estimate of 8.8% of patients in a recent prospective study [22]. It is known to portend a higher risk of mortality [25]. CT is a reliable method to identify the underlying mechanisms of airflow limitation and it enables prediction of the therapeutic response (table 4). In a case–control study, Naccache et al. [23] demonstrated that CT patterns of airway involvement (i.e. bronchial distortion, peribronchovascular thickening, air-trapping, and bronchial compression by enlarged lymph nodes) were found more frequently, scored higher, and were more often multiple in patients with airflow obstruction than in those without. Furthermore, functional improvement under treatment was observed more frequently in patients with predominant peribronchovascular thickening in comparison to those with predominant bronchial distortion. Notably, interobserver agreement was good for identifying the CT patterns of airway involvement (agreement for 89% of cases, κ=0.85) [23]. In the study by Handa et al. [22], the only CT morphological determinant of lower FEV₁/FVC was peribronchovascular thickening. Hansell et al. [50] demonstrated that the extent of reticular pattern was independently associated with several indices of airflow obstruction, including FEV₁, FEV₁/FVC, maximal expiratory flow (MEF) at 25% above residual volume (RV), MEF at 50% above RV and RV/total lung capacity (TLC). Air-trapping is associated with evidence of small-airway obstruction including maximal expiratory flow between of FVC, RV and/or RV/TLC in some studies [47, 52, 90], whereas in others it only contributes little to airflow obstruction [50, 54].

Aspergilloma

Mycetoma formation has been reported in about 2% of sarcoidosis patients, essentially those with advanced fibrocystic or cavitary disease [104]. Aspergilloma is more readily visualised using chest CT than radiography, which generally reveals a cavity containing a solid mass, with or without a characteristic air crescent. The fungus ball can be mobile within the cavity when the patient is turned from the supine to the prone position (fig. 7). In the presence of a cavitary lesion, the thickening of the wall and adjacent pleura points to aspergilloma and is sometimes the earliest sign before any changes are visible inside the cavity [73]. Aspergilloma may be multiple and bilateral [104], and this is best appreciated by CT and imperative to recognise before surgery.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7–

a) Chest radiography and b) computed tomographic image (performed in procubitus) in a 43-yr-old female with cystic lung lesions complicated with aspergilloma (fungus ball with a crescent sign in the right lung).

Pulmonary hypertension

The prevalence of pre-capillary pulmonary hypertension does not traditionally exceed 5% of all sarcoidosis patients but it largely depends on the stage of disease. Although usually attributed to the destruction of distal capillary bed by fibrotic process and/or to the resultant chronic hypoxaemia, various alternative mechanisms come into play, including extrinsic compression of large pulmonary arteries by enlarged lymph nodes or mediastinal fibrosis [4], specific granulomatous or fibrotic vascular involvement that sometimes simulates pulmonary veno-occlusive disease [36, 105], and vasoconstriction by vasoactive factors. Contrast-enhanced CT may raise the possibility of pulmonary hypertension when the widest diameter of the main pulmonary artery is larger than 30 mm or superior to that of the ascending aorta (fig. 6b) but this sign is poorly reliable [106]. Contrast-enhanced CT can also help to recognise underlying mechanisms by showing vascular compression or both extensive septal reticulations and ground glass opacity in case of pulmonary veno-occlusive disease [36]. Most importantly, contrast-enhanced CT rules out pulmonary embolism, which has been recently associated with sarcoidosis [107].

Bronchial stenosis

Significant bronchial stenosis is very rare and may occur at any stage of the disease [108]. Obstruction may be the consequence of endobronchial granulomatous involvement and extrinsic compression by enlarged lymph nodes [61]. The stenoses can be solitary or multiple, lobar or segmental and may cause or contribute to pulmonary symptoms [61, 108]. The left upper lobe (44.5%), the right upper and middle lobes (15.5% each), and the left lower lobe (11%) are most often affected. Apart from depicting atelectasis and nodal external reduction of bronchial lumen, CT is useful in determining the extent and nature of bronchial stenosis (fig. 8) [37, 108]. However, it cannot replace bronchoscopy, as it leads to false-positive results, incorrectly predicting the presence of focal bronchial abnormalities in up to 14% of patients [37, 109]. New techniques of reconstruction of the bronchial tree may improve the assessment of airway involvement.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8–

a) Chest radiography and b, c) computed tomographic images of a 60-yr-old female with stenoses of the dorsal bronchus of the right upper lobe with atelectasis. Minimum intensity projection reconstruction in the coronal plane improves the detection of bronchial stenoses (c), which are confirmed during bronchoscopy (white arrow, d).