Study design and population
This retrospective observational study was approved by our local ethics committee (CLEP Decision no AAA-2020-08044), which waived the need for patient consent.
All consecutive patients with RP and GPA referred to the National Referral Center for Rare Systemic Autoimmune Diseases at our hospital between February 2009 and July 2020 were included if they had tracheobronchial involvement and an available chest CT. The exclusion criterion was CT images available only in slice thickness of ≥ 5 mm.
Diagnosis criteria
There are no validated diagnostic criteria for RP. Patients were classified as having RP if they fulfilled two major criteria or one major and two minor criteria according to the Michet classification [7]. Thus, the diagnosis of RP was established on clinical grounds and required either confirmed inflammation in two of three auricular, nasal, or laryngotracheal cartilages or confirmed inflammation in one of the above cartilages and two other minor criteria which include hearing loss, ocular inflammation, vestibular dysfunction, and seronegative polyarthritis. Furthermore, it is now commonly accepted that the exclusion of differential diagnoses, which can mimic RP, is crucial, particularly GPA. Thus, positive ANCA results with PR3 or MPO specificity, involvement of the lung parenchyma, or destructive ear-nose-throat (ENT) lesions in this context additionally classify the patient as having GPA and not RP [11]. For GPA, patients had to meet the 1990 American College of Rheumatology classification criteria requiring evidence of vasculitis and/or revised Chapel Hill Nomenclature [1, 10]. We did not use any airway involvement pattern to differentiate GPA from RP patients.
CT acquisitions and image analysis
CT images were acquired on several CT scanner models without the standardization of acquisition parameters. All patients had inspiratory volumetric acquisition of at least the entire chest from the lung apices to the diaphragm during a single inspiration. When available, additional expiratory CT images were also analyzed. When several eligible chest CT scans were available, the first scan showing tracheobronchial involvement was chosen for the analysis.
Image analysis was performed independently by two radiologists (CJ and IS) with 2 and 5 years of experience in thoracic imaging, respectively. They reviewed CT images blindly to the final RP or GPA diagnosis, looking for airway wall thickening, tracheobronchial stenosis, tracheobronchial calcifications, tracheobronchomalacia, bronchiectasis, small airway involvement (mosaic attenuation and air trapping), and parenchymal abnormalities, as follows:
-
Airway wall thickness was visually assessed. In case of thickening, the following features were described: the airway segment involved, the localized or extensive nature (more than 2-cm extension) of the thickening, and the sparing of the posterior tracheal membrane.
-
Airway stenosis was defined as a narrowing of the lumen diameter of at least 50%. In case of stenosis, the segment involved and its localized or extensive nature were mentioned using the same criteria as for airway wall thickening.
-
Bronchial calcifications were considered abnormal when they were present within a thickened bronchial wall with an attenuation value of ≥250 Hounsfield units.
-
Tracheobronchomalacia was assessed only if additional expiratory images were available and defined as narrowing of at least 50% of the lumen diameter in expiration.
-
Bronchiectasis was defined as a broncho-arterial ratio >1, bronchial visibility within 1cm of the pleural surface, or lack of distal tapering [18].
-
Mosaic perfusion was defined as a patchwork of regions of differing attenuation [18]. In the setting of GPA and RP, mosaic perfusion was most likely to correspond to small airway disease which was confirmed by the presence of air trapping.
-
Air trapping was visually defined in patients with additional expiratory CT images available as at least 2 adjacent lobules or at least 5 lobules per lung failing to increase in attenuation at the end of expiration.
-
Parenchymal abnormalities included noncalcified centrilobular pulmonary nodules and parenchymal consolidations.
For lesion location, the tracheobronchial tree was divided as follows: (i) the subglottic trachea (within 10 mm below the glottic plane), (ii) the cervical trachea (below the subglottic level), (iii) the thoracic trachea, (iv) main and lobar bronchi, and (v) distal bronchi (segmental bronchi and beyond). Involvement was considered extensive if it was at least 2 cm long.
In cases of discrepancy between the two observers, a third radiologist with 7 years of experience in thoracic imaging (GC) was responsible for the decision. In addition, based on their experience, the two observers were asked to classify patients as having a CT appearance suggestive of RP, GPA, or indeterminate. All observers were blinded to the initial CT interpretation report and the final diagnosis of RP or GPA.
Statistical analysis
Statistical analysis was performed using “R” software (version 3.6.3, R Foundation, Vienna, Austria). First, interobserver variability was assessed by Cohen’s kappa coefficient. Additionally, the variability between the two observers and consensus reached between them were assessed. The coefficient was interpreted as follows: values ≤ 0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement.
For the rest of the analysis, a CT sign was considered present according to the consensus between observers. The association between imaging features and a diagnosis of GPA rather than RP was studied using a generalized linear regression model. For multivariate analysis, only the variables significantly associated with a diagnosis of GPA with a p value <0.05 were included in the model. A p value <0.05 was considered significant.
To assess the performance of the observers to differentiate between GPA and RP, CT exams classified as indeterminate were considered misdiagnoses.
