There is interest in using transbronchial cryobiopsies (CBs) for the diagnosis of fibrotic (chronic) hypersensitivity pneumonitis (FHP), but with little information in the literature about what features are diagnostic in CBs.
To determine, using in silico investigation, whether features supporting a diagnosis of FHP in video-assisted thoracoscopic (VATS) biopsies can be identified in CBs.
In silico circular “cryobiopsies,” 5.25 mm in diameter (21.6 mm2), were created on the slides of 15 VATS biopsy cases that had been assigned a 60% or greater confident diagnosis of FHP at a specially devised multidisciplinary discussion. Using stratified random sampling, up to 8 “cryobiopsies” per case were analyzed for the presence of giant cells/granulomas or peribronchiolar metaplasia affecting 50% or more of the bronchioles, features that had statistically supported a diagnosis of FHP on the VATS biopsies in the multidisciplinary discussion exercise.
Giant cells/granulomas were detected with very low sensitivities in the “cryobiopsies.” Using peribronchiolar metaplasia in 50% or more of bronchioles alone, the sensitivity/specificity for a diagnosis of FHP of 2 “cryobiopsies” compared to the corresponding VATS biopsy was 0.57/0.63; for 4 “cryobiopsies,” 0.86/0.75; and for 8 “cryobiopsies,” 0.83/0.71. Adding giant cells/granulomas slightly improved these numbers to 0.63/0.71 for 2 “cryobiopsies”; 1.00/0.86 for 4; and 1.00/0.80 for 8.
In the setting of a multidisciplinary discussion where FHP is part of the differential diagnostic choices, 4 actual CBs with an area of roughly 20 mm2 each should have good sensitivity and reasonable specificity for diagnosing FHP using these specific morphologic criteria.
The diagnosis of fibrotic hypersensitivity pneumonitis (also referred to as chronic hypersensitivity pneumonitis, hereafter called FHP) is frequently a difficult problem, because many of the clinical, radiologic, and pathologic features are similar to usual interstitial pneumonia/idiopathic pulmonary fibrosis (UIP/IPF). There is poor concordance across multidisciplinary discussion (MDD) groups in making a diagnosis of FHP1 and probably significant misclassification.2 This distinction is important because at least the initial approach to treatment of FHP and UIP/IPF is different (immunosuppressive versus antifibrotic agent), and in FHP cases it is, in theory, possible to stop progression of the disease if the offending antigen can be found and removed from the patient's environment.3
The pathologic features that define FHP in the literature are themselves disputed (reviewed in Churg et al4 ), and separation of FHP from UIP/IPF even using standard video-assisted thoracoscopic (VATS) biopsies often poses difficulties for the pathologist.4 There is intense interest in using transbronchial cryobiopsies (CBs) for the diagnosis of all forms of interstitial lung disease, including FHP and UIP/IPF, but with quite discordant claims about the accuracy of CBs.5,6 Little has been published on the specific features required for diagnosing FHP in CBs, and determining what pathologic features allow a diagnosis of FHP is difficult because obtaining both CBs and VATS biopsies in the same patient is generally not done. The only 2 studies that actually compared CBs to VATS biopsies in the same patients for interstitial lung disease in general (not just FHP or UIP/IPF) required an elaborate organizational scheme,5,6 which is not feasible in most institutions. Neither study provided detailed information on the criteria used for diagnosis of FHP in CBs, but, of interest, in the COLDICE study,6 the separation of FHP from UIP/IPF in the CBs was noted to be the greatest source of diagnostic discrepancy.
Wright et al7 recently reviewed a data set for 23 patients whose VATS biopsies were reported as either FHP or UIP/IPF after a conventional MDD and reanalyzed them in a special MDD where an experienced interstitial lung disease clinician re-evaluated the case blindly and assigned a likelihood of one diagnosis or the other, using only clinical features; an experienced chest radiologist did a similar exercise with only imaging features; and 2 experienced lung pathologists came to a pathologic likelihood by using only biopsy features. The group then met and after discussion generated a diagnostic confidence from 0% to 100% for either diagnosis.
In that exercise, the features that supported either FHP or UIP/IPF at a clinical, radiologic, or pathologic level were examined statistically. In the VATS biopsies the presence of giant cells/granulomas or peribronchiolar metaplasia affecting 50% or more of the bronchioles was associated with a diagnosis of FHP.
In the current report we have created multiple circular nonoverlapping 5.25-mm-diameter (21.6-mm2 area) in silico “cryobiopsies” on the FHP VATS biopsy slides from this same group of patients, and using morphometric sampling, asked whether these “cryobiopsies” were able to detect the same FHP features as the VATS biopsies, and if so, how many “cryobiopsies” were required and which features were useful.
METHODS
This retrospective cross-sectional study was approved by the Research Ethics Board of the University of British Columbia, Vancouver, Canada. We used 15 VATS biopsy cases from the previous study where the diagnostic confidence of FHP, after the special MDD described above, was 60% or greater.7 All biopsy samples were fixed by formalin inflation. Two representative slides were chosen from all of the VATS biopsy slides for each lobe by assigning a number to each slide, and then selecting 2 random numbers for each lobe by using a random number generator (these were the slides used in the study by Wright et al7 ). For the current study the same slides were examined with a microscope, using an ×4 objective (5.25-mm field diameter, 21.6-mm2 field). Avoiding the pleural surface, the tissue was divided into nonoverlapping rings, each of which served as a 5.25-mm in silico “cryobiopsy”; the area encompassed by the ring was then outlined and the rings numbered (Figure 1). The number of rings (“cryobiopsies”) per case ranged from 5 to 60.
A low-power view of a video-assisted thoracoscopic biopsy slide with many 5.25-mm-diameter in silico “cryobiopsies” outlined (hematoxylin-eosin, original magnification ×20).
A low-power view of a video-assisted thoracoscopic biopsy slide with many 5.25-mm-diameter in silico “cryobiopsies” outlined (hematoxylin-eosin, original magnification ×20).
We used stratified random sampling to ensure proper morphometric technique; stratified random sampling provides an unbiased representation of the underlying tissue. A random number was first generated to determine ring 1 for analysis. An additional 7 rings for analysis were chosen by sequential addition of the numeral 8 to the random number and those corresponding rings selected. This methodology provides both random and stratified selection of the areas to be analyzed. In order of selection, each field was then examined for the presence or absence of granulomata, and the number of airways with or without peribronchiolar metaplasia was recorded.
RESULTS
The area of the underlying VATS biopsies from which the “cryobiopsies” were drawn ranged from 0.82 to 18.4 cm2 (mean ± SD, 7.6 ± 5.3 cm2; median, 6.2 cm2). We were able to create at least 5 “cryobiopsies” in all 15 cases, 7 “cryobiopsies” in 14 cases, and 8 “cryobiopsies” in 13 cases.
Figure 2 shows the mean number of bronchioles sampled in the “cryobiopsies” and the VATS biopsies. These ranged from 3 ± 1.6 (mean ± SD) for 2 “cryobiopsies” to 7 ± 2 for 4 and to 14 ± 5 for 8; and 30 ± 19 for the VATS biopsies. To ensure that our “cryobiopsies” were random, we plotted the fraction of biopsies positive for a giant cell/granuloma for all “cryobiopsies” numbered 1, 2, up to 8, and similarly the fraction of bronchioles with peribronchiolar metaplasia for all “cryobiopsies” numbered 1 to 8. These data showed that the presence or absence of the feature of interest was effectively random (data not shown), which suggests that our sampling strategy is correct.
Mean number (±SD) of bronchioles sampled in 2, 4, and 8 “cryobiopsies” and video-assisted thoracoscopic (VATS) biopsies from all 15 cases.
Figure 3. Cumulative fraction of “cryobiopsies” positive, compared to the video-assisted thoracoscopic (VATS) biopsies from which the “cryobiopsies” are derived, for giant cells/granulomas versus “cryobiopsy” number. Even with 8 “cryobiopsies,” only 50% of the cases positive on VATS biopsies were detected.
Mean number (±SD) of bronchioles sampled in 2, 4, and 8 “cryobiopsies” and video-assisted thoracoscopic (VATS) biopsies from all 15 cases.
Figure 3. Cumulative fraction of “cryobiopsies” positive, compared to the video-assisted thoracoscopic (VATS) biopsies from which the “cryobiopsies” are derived, for giant cells/granulomas versus “cryobiopsy” number. Even with 8 “cryobiopsies,” only 50% of the cases positive on VATS biopsies were detected.
In the original set of VATS biopsies, giant cells/granulomas were seen in 10 cases. Figure 3 shows the cumulative fraction of “cryobiopsies” with giant cells, compared to the VATS biopsies. For 2 biopsies the overall detection rate was extremely low (1 of 10 cases, 10%), increased to 4 of 10 (40%) for 4 “cryobiopsies,” and increased further to 5 of 10 (50%) of the VATS biopsy rate with 5 “cryobiopsies,” but did not improve beyond 5 biopsies.
Figure 4, A through C, shows the case-matched relationship between the percentage of bronchioles showing peribronchiolar metaplasia in 2, 4, and 8 “cryobiopsies” versus the percentage of bronchioles with peribronchiolar metaplasia in the corresponding VATS biopsies. The correlation between the “cryobiopsy” and the VATS biopsy peribronchiolar metaplasia fraction was 0.31 (P = .25) for 2 biopsies, 0.62 (P = .01) for 4 biopsies, and 0.77 (P = .002) for 8 biopsies. Using 50% or more of bronchioles with peribronchiolar metaplasia as “positive,” the sensitivity/specificity values (comparing the in silico “cryobiopsies” to the VATS biopsies) for 2 “cryobiopsies” were 0.57/0.63; for 4 “cryobiopsies,” 0.86/0.75; and for 8 “cryobiopsies,” 0.83/0.71.
A, Proportion of bronchioles showing peribronchiolar metaplasia (PBM) in 2 “cryobiopsies” versus corresponding video-assisted thoracoscopic (VATS) biopsies. Square outlined by dashes indicates a 50% proportion, the cutoff point used as positive. Use of only 2 “cryobiopsies” leads to considerable inaccuracy. B, Proportion of bronchioles showing PBM in 4 “cryobiopsies” versus corresponding VATS biopsies. Square outlined by dashes indicates a 50% proportion, the cutoff point used as positive. C, Proportion of bronchioles showing PBM in 8 “cryobiopsies” versus corresponding VATS biopsies. Square outlined by dashes indicates a 50% proportion, the cutoff point used as positive.
Figure 5. A, Proportion of bronchioles showing peribronchiolar metaplasia (PBM) in 2 “cryobiopsies” versus corresponding video-assisted thoracoscopic (VATS) biopsies with additional indication of cases positive for giant cells/granulomas (shown as *). Square outlined by dashes indicates a 50% proportion, the cutoff point used for PBM as positive. Addition of giant cell/granulomas produces a slight improvement in sensitivity and specificity (see text), but use of only 2 “cryobiopsies” is still inaccurate. B, Proportion of bronchioles showing PBM in 4 “cryobiopsies” versus corresponding VATS biopsies with additional indication of cases positive for giant cells/granulomas (shown as *). Square outlined by dashes indicates a 50% proportion, the cutoff point used for PBM as positive. Addition of giant cell/granulomas produces a slight improvement in sensitivity and specificity (see text). C, Proportion of bronchioles showing PBM in 8 “cryobiopsies” versus corresponding VATS biopsies with additional indication of cases positive for giant cells/granulomas (shown as *). Square outlined by dashes indicates a 50% proportion, the cutoff point used for PBM as positive. Addition of giant cell/granulomas produces a slight improvement in sensitivity and specificity (see text).
A, Proportion of bronchioles showing peribronchiolar metaplasia (PBM) in 2 “cryobiopsies” versus corresponding video-assisted thoracoscopic (VATS) biopsies. Square outlined by dashes indicates a 50% proportion, the cutoff point used as positive. Use of only 2 “cryobiopsies” leads to considerable inaccuracy. B, Proportion of bronchioles showing PBM in 4 “cryobiopsies” versus corresponding VATS biopsies. Square outlined by dashes indicates a 50% proportion, the cutoff point used as positive. C, Proportion of bronchioles showing PBM in 8 “cryobiopsies” versus corresponding VATS biopsies. Square outlined by dashes indicates a 50% proportion, the cutoff point used as positive.
Figure 5. A, Proportion of bronchioles showing peribronchiolar metaplasia (PBM) in 2 “cryobiopsies” versus corresponding video-assisted thoracoscopic (VATS) biopsies with additional indication of cases positive for giant cells/granulomas (shown as *). Square outlined by dashes indicates a 50% proportion, the cutoff point used for PBM as positive. Addition of giant cell/granulomas produces a slight improvement in sensitivity and specificity (see text), but use of only 2 “cryobiopsies” is still inaccurate. B, Proportion of bronchioles showing PBM in 4 “cryobiopsies” versus corresponding VATS biopsies with additional indication of cases positive for giant cells/granulomas (shown as *). Square outlined by dashes indicates a 50% proportion, the cutoff point used for PBM as positive. Addition of giant cell/granulomas produces a slight improvement in sensitivity and specificity (see text). C, Proportion of bronchioles showing PBM in 8 “cryobiopsies” versus corresponding VATS biopsies with additional indication of cases positive for giant cells/granulomas (shown as *). Square outlined by dashes indicates a 50% proportion, the cutoff point used for PBM as positive. Addition of giant cell/granulomas produces a slight improvement in sensitivity and specificity (see text).
Figure 5, A through C, shows similar data, but adding the presence of giant cells/granulomas to the data for peribronchiolar metaplasia. Doing so increased the sensitivity/specificity values for 2 “cryobiopsies” to 0.63/0.71, for 4 “cryobiopsies” to 1.00/0.86, and for 8 “cryobiopsies” to 1.00/0.80.
DISCUSSION
As noted in the introduction, the diagnosis of FHP is a frequent problem in interstitial lung disease, and the important issue of separation from UIP/IPF can be very difficult.4 In some instances there is a good exposure history or imaging that favors one diagnosis or the other, but other cases cannot be resolved by clinical or radiologic findings and the diagnosis may depend on biopsy findings. However, the pathologic diagnosis of FHP is also problematic because both FHP and UIP/IPF often show patchy fibrosis, fibroblast foci, and honeycombing.4 A number of other features such as bridging fibrosis (fibrosis that extends from the bronchiole to the pleura or interlobular septa) and fibroblast foci next to bronchioles have been proposed as FHP markers,8 but a recent report9 suggests that these features may not be useful. Giant cells and granulomas are usually viewed as findings favoring FHP, given the correct morphologic/clinical/radiologic context.4,10
Peribronchiolar metaplasia refers to the development of very fine fibrosis in the alveolar walls surrounding bronchioles with metaplastic bronchiolar epithelium overlying the fibrosis (Figure 6). Peribronchiolar metaplasia is believed to be a marker of bronchiolar damage. By themselves occasional foci of peribronchiolar metaplasia can be seen in normal lungs and in many types of interstitial lung disease; however, they are particularly common in FHP, and a prior study7 on separating FHP from UIP/IPF using VATS biopsies suggested that 50% or more of bronchioles with peribronchiolar metaplasia strongly favored a diagnosis of FHP over UIP/IPF.
An example of peribronchiolar metaplasia. There is fine fibrosis in alveolar walls next to the bronchiole, with overlying metaplastic bronchiolar epithelium (hematoxylin-eosin, original magnification ×200).
An example of peribronchiolar metaplasia. There is fine fibrosis in alveolar walls next to the bronchiole, with overlying metaplastic bronchiolar epithelium (hematoxylin-eosin, original magnification ×200).
Transbronchial CB has been promoted as a faster and less invasive procedure, and potentially one associated with lower mortality, than VATS for the diagnosis of interstitial lung disease. However, whether CB is as accurate as VATS biopsy is a hotly disputed issue, and most studies reporting CB results have not had any kind of pathologic gold standard for comparison. The only 2 studies that directly compared CB and VATS biopsies in the same patients came to radically different conclusions. Romagnoli et al6 found a poor correlation between CB and VATS biopsies and a poor correlation of CB with an MDD-generated final diagnosis, whereas COLDICE5 reported exactly opposite findings. But, as noted in the introduction, although both studies reported some cases as FHP, neither actually specified the criteria used for that diagnosis in the CBs.
It should be appreciated that we are not attempting to argue here whether these are necessarily the best set of features for the diagnosis of FHP or to directly compare the criteria for FHP and UIP/IPF in CBs. Rather, given an MDD-derived diagnosis of FHP and pathologic features that statistically supported a diagnosis of FHP in the VATS biopsies in a previous study,7 we are asking whether those same features are detectable and usable in CBs. This question is of more importance than one might assume at first glance, given the lack of published diagnostic criteria for diagnosing FHP in CBs.
Since we do not have access to actual CBs and VATS biopsies from the same patients, we created in silico “cryobiopsies” from the same VATS biopsy slides used in a previous study.7 Our “cryobiopsies” are reasonably similar to real CBs in terms of size, an important parameter that has been linked to diagnostic yield.6 The study by Ravaglia et al11 reported that the mean short axis of 699 CBs was 4.57 mm and the mean long axis, 6.31 mm (albeit with very wide ranges for both numbers). If one assumes that the biopsies reported by Ravaglia et al11 are roughly elliptical, the mean biopsy area in their study is 22.6 mm2. In COLDICE5 the mean area per CB was 26 mm2 and the mean cumulative area per case was 87 mm2. These figures are quite close to the values used here for our in silico “cryobiopsies” of 21.6 mm2 per biopsy, and 86.4 mm2 for a sample of 4 “cryobiopsies.”
The question of numbers of CB and diagnostic yield has also been examined. Ravaglia et al11 concluded that the diagnostic yield for diffuse parenchymal lung disease increased with increasing number of biopsies, and a consensus document recommended that patients undergoing CB for interstitial lung disease should have at least 2 biopsies from different sites.12 In the COLDICE study5 there was agreement between CBs and VATS biopsies 70% of the time if 4 or more biopsies from 2 separate lobes were sampled by CB.
The conclusions from our study are straightforward. Despite the importance that is usually placed on the finding of giant cells/granulomas for diagnosing FHP, our “cryobiopsies” are insensitive detectors of these structures, probably because giant cells/granulomas are usually not numerous in FHP. This result is not surprising, since, as noted above, the total mean area of the VATS biopsy in our previous study was 7.6 cm2, whereas even with 8 “cryobiopsies” the total area sampled is 1.7 cm2, a 4.5-fold difference, and with 4 “cryobiopsies” the sampled area is only 0.86 cm2, a 9-fold difference. Depending on finding giant cells/granulomas alone is therefore not a very fruitful approach.
By themselves “cryobiopsies” appear to be good detectors of peribronchiolar metaplasia, probably because bronchioles are numerous and homogeneously distributed in the lung and are repeatedly sampled by “cryobiopsies,” but an adequate number of samples is required, both to avoid missing true cases with a high peribronchiolar fraction, and also to avoid overcalling a high peribronchiolar fraction because of small sample numbers. Thus two 21.6-mm2 “cryobiopsies” appear to be an inadequate sample (and lead to many spuriously positive cases), even with the addition of giant cells/granulomas, but 4 “cryobiopsies” provide good sensitivity and reasonably high specificity (compared to the VATS biopsies) counting both peribronchiolar metaplasia and giant cells/granulomas, and there is no real improvement with 8 “cryobiopsies.”
There are a number of important limitations to this study. First, we are dealing with a small number of cases, but the correlations between “cryobiopsies” and VATS biopsies are statistically significant with 4 or 8 “cryobiopsies,” suggesting that, with an adequate sample size, “cryobiopsies” reflect the findings in VATS biopsies, which is the question that we intended to address.
Second, there is an element of selection bias at work; that is, the “cryobiopsies” are derived from VATS biopsies that are known to have the features of interest (giant cells/granulomas and a high fraction of bronchioles with peribronchiolar metaplasia), so one might argue that we cannot help but find them in the “cryobiopsies.” However, this is not necessarily true because these features are not present in every microscopic field of the VATS biopsies. We took care to avoid biasing our results: the VATS slides that were used to generate the in silico “cryobiopsies” had been randomly selected,7 and the individual fields (“cryobiopsies”) were also randomly selected in a morphometrically correct fashion. Thus any given “cryobiopsy” might or might not show these features. As well, these features might be so infrequent in the VATS biopsies that they would not be detected in the cryobiopsies or would be found only rarely. This is in fact the situation for giant cells/granulomas where the detection rate in the “cryobiopsies” is very low. For peribronchiolar metaplasia the situation is somewhat the opposite: use of only 2 “cryobiopsies” tends to overestimate the fraction of airways with peribronchiolar metaplasia in a portion of cases.
These findings imply that, in the specific context discussed here, 4 actual cryobiopsies of roughly 20-mm2 area each can provide a reasonably accurate reflection of the underlying VATS biopsies, and if one accepts the criteria of giant cells/granulomas and 50% or greater peribronchiolar metaplasia as indicative of a diagnosis of FHP, then 4 such cryobiopsies can provide helpful information. However, this statement should not be taken to imply that a diagnosis of FHP should be made purely from slides. Accurate diagnosis of fibrosing interstitial pneumonias requires clinical and radiologic information, preferably in an MDD format. But in such a format, and in a clinical/radiologic setting where FHP is in the differential diagnosis, then cryobiopsies of adequate area/number can be used to support that diagnosis, using these criteria.