Introduction

The correct classification of interstitial lung disease (ILD) into specific subtypes requires assimilation of various clinical, radiographic, and sometimes pathologic information. Once these data are assembled, multidisciplinary discussion (MDD) involving pulmonologist, radiologist, and pathologist improves the consistency of ILD subclassification [1]. Bronchoscopy has been reported to add diagnostic information in up to 30% of patients with ILD, but it remains unclear which patients benefit from inclusion of information obtained by bronchoscopy in the diagnostic classification of ILD by MDD [2,3,4,5,6,7,8].

Recent guidelines on the diagnosis of IPF recommend bronchoscopy with BAL for ILD patients with a high-resolution computed tomography (HRCT) pattern that is probable usual interstitial pneumonia (UIP), indeterminate for UIP, or suggestive of an alternative diagnosis [2]. The IPF guidelines make no other recommendations for patient selection for bronchoscopy and do not make a recommendation for or against TBBx [2]. When a specific ILD diagnosis cannot be rendered confidently after incorporating other clinical and radiographic data, the guidelines recommend surgical lung biopsy (SLB), which has higher sensitivity but also much greater risk of morbidity and mortality than TBBx [2,3,4,5,6,7,8].

Determining clinical characteristics that help with patient selection for bronchoscopy may improve confidence of ILD classification while limiting potential adverse outcomes associated with SLB. Therefore, we sought to identify factors that were associated with change in multidisciplinary ILD diagnosis before and after incorporation of BAL and TBBx data. We hypothesized that the yield of bronchoscopy results would depend both on radiographic pattern and identification of an antigen.

Methods

We retrospectively identified ILD patients evaluated between 2011 and 2018 from the University of Texas Southwestern Medical Center (UTSW). Those who had a bronchoscopy for the diagnostic workup of their ILD were included in the study. Patients were excluded if the bronchoscopy was performed for a reason other than the diagnostic workup of their ILD such as evaluation for infection or malignancy. Patients with sarcoidosis were also excluded from this study.

Clinical data extracted from the medical record included age, gender, smoking history, potential fibrogenic antigen exposure, response to exposure removal, pulmonary function testing (PFTs), BAL cell count and differential, histopathologic interpretation of the transbronchial and Surgical Lung Biopsy (SLB), and connective tissue disease (CTD) serologies. HRCTs were evaluated by a thoracic radiologist (KB) and reported as definite usual interstitial pneumonia (UIP), probable UIP, indeterminate for UIP, or most consistent with a non-IPF diagnosis [9]. Findings from pathologic specimens were recorded by a thoracic pathologist (YB).

The sequential MDD format described here is adapted from prior studies [10, 11]. A retrospective chart review for each patient was conducted by pulmonologists with expertise in ILD (TNA, CAN). They presented cases in an MDD format to other expert pulmonologists (LS, VA, CG), radiologist (KB), and pathologists (YMB, HT) who were blinded to the aggregate clinical data. The pulmonologist who performed the chart review did not participate in the discussion or assist in providing a diagnosis.

Initially, each patient was assigned a “pre-bronchoscopy diagnosis” based on their clinical history, examination, serologic studies, and HRCT results without taking into account their BAL TBBx or SLB findings. A pre-bronchoscopy diagnosis of HP required an antigen exposure and a characteristic HRCT scan (defined as the presence of centrilobular nodules, mosaicism, and a mid or upper lung zone predominant interstitial disease [12]) or an exposure and a significant response to exposure removal (defined as a 10% improvement in forced vital capacity % predicted or radiographic improvement on follow-up CT within 3 months) [13, 14]. A pre-bronchoscopy diagnosis of IPF was made according to guidelines [2]. Patients meeting the American College of Rheumatology criteria for a defined connective tissue disease were assigned a pre-bronchoscopy diagnosis of connective tissue disease-related ILD (CTD-ILD). A diagnosis of interstitial pneumonia with autoimmune features (IPAF) was made according to a recently published research statement [15]. Those who did not meet criteria for IPF, HP, CTD-ILD, IPAF, or alternative cause of ILD were assigned a pre-bronchoscopy diagnosis of unclassifiable ILD.

Next, patients were assigned a “post-bronchoscopy diagnosis” after incorporating their bronchoscopy information in addition to the non-invasive testing. A diagnosis of IPF could be made on bronchoscopy only in the presence of patchy interstitial fibrosis and fibroblast foci or honeycomb change [16]. A BAL lymphocyte percentage of > 30% was used to support a diagnosis of HP [3, 17]. The TBBx result was considered characteristic of HP based on the presence of granulomas, particularly loose granulomas or giant cells, and at least one of the following: inflammatory bronchiolitis or a predominantly mononuclear cellular interstitial infiltrate [18,19,20].

Lastly, patients were assigned a “final diagnosis” taking into account all non-invasive testing, subsequent HRCT, bronchoscopy, SLB, and explant where available.

Statistical Analysis

Continuous variables were expressed as means and standard deviations; comparisons were made using Student’s t test or Wilcoxon signed rank sum test as appropriate. Categorical variables were expressed using counts and percentages; comparisons were made using Chi-squared test or Fisher’s exact test, where appropriate. Univariable logistic regression was performed to identify patient- and disease-specific factors that were associated with change in pre-bronchoscopy diagnosis after incorporating information obtained from the BAL and/or TBBx. These variables were chosen based on clinical relevance and included demographic features, smoking status, presence of potential fibrogenic antigen, inconsistent with UIP pattern on HRCT, and pre-bronchoscopy diagnosis of CTD-ILD or IPAF. The variables that were significantly associated with change in diagnosis (p value < 0.1) were included in multivariable model to test independent associations. All p values less than 0.05 were considered significant. Statistical analyses were performed using R version 3.3.2 statistical analysis software (www.R-project.org).

Results

Patient Characteristics

In our retrospective cohort, 245 patients had a bronchoscopy performed for the purpose of ILD classification and were included in analysis. Mean age was 58.4 years at time of ILD diagnosis, 43% were male, and 71.4% were non-Hispanic white (Table 1). A potential fibrogenic antigen exposure was identified in 55.9% of the cohort.

Table 1 Characteristics of patients with ILD who underwent bronchoscopy based on post-bronchoscopy MDD data (N = 245)
Full size table

At baseline, this cohort had mild impairment in lung function. The most common pre-bronchoscopy diagnoses in our cohort were unclassifiable (49.8%), CTD-ILD (18.4%), IPF (9%), IPAF (7.8%), and HP (7.3%) (Table 2). The majority of ILD patients that underwent bronchoscopy had a clinical suspicion of HP (75.9%) based on the presence of antigen, suggestive radiographic pattern, or absence of a clear alternative diagnosis.

Table 2 Pre-bronchoscopy, post-bronchoscopy, and final diagnosis (N = 245)
Full size table

Diagnostic Information Obtained from Bronchoscopy

BAL cell count was performed in 147 patients (60%), TBBx in 193 (78.8%), and both BAL and TBBx in 117 (47.8%). Seventy-three patients (49.7%) had a BAL lymphocyte percentage > 30%. The mean BAL lymphocyte percentage was 26.7% ± 21.9.

Data obtained from BAL or TBBx led to a change from pre-bronchoscopy diagnosis to post-bronchoscopy diagnosis in 58 (23.7%) of 245 patients (Table 2). Of those 58 patients, the pre-bronchoscopy diagnosis was unclassifiable in 54 patients (93.1%) and IPF in 4 patients (6.9%); post-bronchoscopy diagnosis included 51 cases (87.9%) of HP, 2 (3.4%) chronic eosinophilic pneumonia, 1 (1.7%) pulmonary alveolar proteinosis, 1 (1.7%) amyloid, 1 (1.7%) vasculitis, 1 (1.7%) primary biliary cirrhosis, and 1 (1.7%) IPF.

Incorporating BAL lymphocyte count > 30% alone led to a change in diagnosis for 32/147 patients (21.8%), while incorporating TBBx alone changed diagnosis in 50/195 (25.9%) (p = 0.44). Compared to BAL alone, the 117 patients who had both BAL and TBBx performed had a higher rate of change in diagnosis (21.8% vs 34.1%, respectively, p = 0.027).

Features Associated with Diagnosis Change

In the pre-specified univariable analysis, male gender (odds ratio 2.49; 95% CI 1.37–4.59), smoking status (odds ratio 2.01; 95% CI 1.11–3.67), presence of antigen (odds ratio 6.13; 95% CI 2.97–14.0), HRCT scan consistent with a non-IPF diagnosis (odds ratio 2.9; 95% CI 1.36–6.97) were positively associated with a change in diagnosis after bronchoscopy; a pre-bronchoscopy diagnosis of CTD-ILD or IPAF (odds ratio 0.03; 95% CI 0.002–0.15) was negatively associated with a change in diagnosis after bronchoscopy. In the multivariable analysis, the presence of antigen (odds ratio, 4.48; 95% CI, 1.92–11.5) and HRCT scan consistent with a non-IPF diagnosis (odds ratio, 4.71; 95% CI, 1.94–13.0) were positively associated with a change in diagnosis after bronchoscopy. A pre-bronchoscopy diagnosis of CTD-ILD or IPAF was negatively associated with a change in diagnosis after bronchoscopy (odds ratio 0.05; 95% CI 0.003–0.26) (Table 3).

Table 3 Clinical features associated with change in ILD diagnosis after incorporating bronchoalveolar lavage and/or transbronchial biopsy data
Full size table

Results Based on Antigen Identification

Compared to patients without antigen identified, patients with an exposure to a potential fibrogenic antigen were more likely to have diagnosis after BAL alone (26.2% vs 12.5%, p = 0.009) and after both BAL and TBBx (42.1% vs 14.7%, p = 0.005) (Table 4). The addition of TBBx to BAL increased the yield in patients with antigen identified from 26.2 to 40.1% (p = 0.028), but there was no difference in diagnostic yield when TBBx was added to BAL for patients without antigen identified (12.5% vs 14.7%, p = 1.0).

Table 4 Change in diagnosis after bronchoscopy based on the presence of antigen
Full size table

Bronchoscopy Yield by Radiographic Pattern

HRCT was available for evaluation in 99.6% of patients. The HRCT pattern was definite UIP in 17 patients (7.0%), probable UIP in 12 (4.9%), indeterminate in 38 (15.6%), and inconsistent in 177 (72.5%) (Table 1).

Incorporating bronchoscopy data was associated with a change in diagnosis in 4 patients (13.8%) with a probable or definite UIP pattern, 4 (10.5%) with indeterminate for UIP pattern, and 50 (28.2%) with inconsistent with UIP pattern. Among the 7 patients whose diagnosis was changed after bronchoscopy to a diagnosis other than HP, 7 (100%) had an HRCT that was consistent with a non-IPF diagnosis. For those who underwent BAL alone, there was no difference in the frequency of diagnosis change after bronchoscopy between patients with an HRCT consistent with a non-IPF diagnosis and those with a definite, probable, or indeterminate for UIP HRCT pattern (25.5% vs 12.5%, p = 0.12) (Table 5). Incorporating both BAL and TBBx data was associated with a higher frequency of diagnosis change in patients with an inconsistent HRCT (40.0%) than in those with a definite, probable, or indeterminate HRCT (19.4%) (p = 0.048).

Table 5 Yield of bronchoscopy by HRCT pattern (N = 244)
Full size table

Discussion

In this study, we examined the sequential incorporation of information obtained from bronchoscopy (BAL and/or TBBx) to the MDD process to arrive at a confident diagnosis in a cohort of well-phenotyped ILD patients. Information obtained from BAL and TBBx led to a change in diagnosis of 23.7% of patients. Yield was higher when TBBx was added to BAL. Positive predictors of a change in diagnosis after bronchoscopy included identification of antigen and HRCT consistent with a non-IPF diagnosis. A pre-bronchoscopy diagnosis of IPAF or CTD-ILD was a negative predictor of a change in diagnosis after bronchoscopy. In patients with antigen identified, a confident diagnosis of HP could be made in 40% of patients by BAL and TBBx without the need for SLB.

The yield of TBBx in the workup of ILD aids in ILD classification of approximately 20–30% of patients [4], and the addition of BAL to TBBx likely further increases the yield.[21] These studies suggest that TBBx can be a useful tool in the workup of ILD; however, unlike HRCT, the utility of bronchoscopy is not universal to all ILD patients. Reasons for the lack of uniformity are not only due to anatomic limitations and small pathologic samples from TBBx, but also stem from poor patient selection. Our study identified clinical factors that were associated with a change in multidisciplinary diagnosis after incorporation of bronchoscopy data, highlighting that the yield of minimally invasive tissue sampling is higher in enriched populations. Among the most predictive factors of a change in diagnosis after bronchoscopy in our cohort was identification of antigen, which supports the results from a prior study demonstrating that identification of antigen had the highest likelihood ratio of any factor in predicting a diagnosis of HP [22]. We also found that patients who had a radiographic pattern consistent with a non-IPF diagnosis had a stronger association with a useful bronchoscopy, which is not completely unexpected based on prior studies [2, 9, 12, 16]. However, we also found that 13% of patients with a definite or probable UIP radiographic pattern also had a change in diagnosis after incorporating information from bronchoscopy; each of these patients had an identified antigen and changed from a pre-bronchoscopy diagnosis of IPF to a post-bronchoscopy diagnosis of HP. We suggest that bronchoscopy can be useful in confirming a diagnosis of HP even for patients with a radiographic pattern that suggests IPF. Given that treatment with immunosuppression is associated with worse outcomes in IPF [23], but is the mainstay of treatment for progressive HP [24], confidently distinguishing these two disorders is of utmost importance.

It is particularly important to define the role of bronchoscopy in the workup of ILD because TBBx has a much lower complication rate than transbronchial lung cryobiopsy (TBLC) or SLB. Recent meta-analyses reported a 30- to 60-day mortality rate of 0.3–0.7% after TBLC and 1.8–2.7% after SLB; complication rates were 23.1% after TBLC [6, 25, 26]. TBBx, in contrast, has a complication rate ranging from 0.08 to 6.8% with a mortality rate ranging from 0 to 0.13% [5, 7, 25, 27]. Adding BAL to TBBx can help achieve a confident diagnosis in a sizable minority of patients (40%), thus avoiding the need for higher risk procedures such as TBLC or SLB.

Our study findings also support prior data regarding the role of lung biopsy in CTD-ILD patients. Lung biopsy, either surgical or transbronchial, is not routinely recommended in CTD-ILD patients because it is unclear whether biopsy is an independent predictor of prognosis in CTD-ILD when non-invasive data including HRCT and pulmonary function testing is available [28,29,30,31,32,33]. Further, a diagnosis of IPAF can be made without pathologic data, and many of the IPAF pathologic criteria cannot be met with confidence on a TBBx [15]. Our results suggest that bronchoscopy is unlikely to change the diagnosis in a patient with CTD-ILD or IPAF.

Strengths of our study include the serial incorporation of information obtained from bronchoscopy to the MDD process to arrive at a confident diagnosis in a cohort of well-phenotyped ILD patients across a range of diagnoses, thus limiting incorporation bias. While prior studies either focus on the yield of bronchoscopy for a particular diagnostic group or only included patients who underwent both bronchoscopy and SLB, our pragmatic cohort included patients with a variety of diagnoses, many of whom did not undergo SLB [4, 16]. As a result, we were able to diagnose conditions such as amyloid and pulmonary alveolar proteinosis that were not suspected before bronchoscopy. Although most patients in our cohort had suspected HP, clinical suspicion for HP was not an inclusion criterion, which improves our generalizability. Further, our study evaluates the additive information of TBBx to BAL, which is important as current diagnostic guidelines do not make a recommendation for or against performing TBBx [9].

There are limitations to this study that should be acknowledged. Most patients (75.9%) in our cohort were suspected of having HP; however, we did not limit our cohort to those with a single diagnosis, and we included patients who had already had a bronchoscopy prior to being referred to our center, thus limiting potential selection bias. The diagnostic evaluation of ILD patients in this study was dictated by the treating physician; therefore, not all ILD patients at our center had BAL and/or TBBx performed. We maintained strict pre-specified pre- and post-bronchoscopy diagnostic criteria for each of the subtypes of ILD, thus potentially underestimating the clinical utility of bronchoscopy itself in achieving a high-confidence ILD diagnosis while reducing risk of bias in the interpretation of results. Further, we addressed potential investigator bias by having the diagnostic determination made by a team of experts that were blinded to all clinical information except for what was presented by the expert pulmonologist that performed the chart review. Lastly, given the retrospective nature of our study, we could not accurately assess the influence of treatment at the time of bronchoscopy on BAL lymphocyte count or TBBx characteristics.

In summary, accurate ILD diagnosis is absolutely essential since the diagnosis dictates treatment plans and provides prognostic information. The diagnostic evaluation should prioritize low-risk non-invasive or minimally invasive procedures when these approaches allow for confident diagnosis. Here, we identified specific clinical features that help with patient selection for bronchoscopy in the classification of ILD. We suggest that bronchoscopy with both BAL and TBBx be performed in patients with a sensitizing antigen identified even if imaging features suggest a diagnosis of IPF and in patients with an HRCT pattern that is consistent with a non-IPF diagnosis. We also suggest that bronchoscopy be avoided in patients with a diagnosis of CTD-ILD or IPAF. Appropriate patient selection for bronchoscopy may improve ILD diagnostic confidence and avoid potential complications from more invasive and higher risk procedures.