Prior studies assessing the diagnostic and prognostic utility of BALL in fibrotic and non-fibrotic HP may be limited by varied diagnostic criteria and incorporation bias, hindering broad application to current practice. Our findings are derived from a cohort of well-characterized HP patients meeting at least ‘moderate’ or higher diagnostic confidence levels according to recent 2020 ATS/JRS/ALAT diagnostic guidance, without the addition of BALL as a diagnostic criterion (3). Using this approach, we found lower frequencies of diagnostic BALL in both fibrotic and non-fibrotic subtypes and higher comparative BALL counts in non-fibrotic vs fibrotic patients. There was also correlation of BALL with long-term survival in those presenting with fibrosis.
Our study confirms BALL counts are lower in fibrotic compared to non-fibrotic HP patients (median BALL 15% vs 19% respectively, P = 0.005). Higher BALL counts and increased frequency of positive cut-offs have been previously reported in non-fibrotic patients. Adams et al. reported median BALL of 46% and 19% respectively in 17 non-fibrotic and 60 fibrotic patients undergoing diagnostic bronchoscopy (5). Takei et al applied a recent Delphi consensus statement for f-HP diagnosis to consecutively presenting ILD patients (using a BALL cut-off > 40%) and found only 19% of MDD-diagnosed fibrotic patients had diagnostic BALL findings(17). Frequency of diagnostic BALL was higher at 40% in one study using a lower diagnostic cut-off of 20% (16). Median BALL was only 16% in another study of 160 chronic HP patients (85% with reported fibrosis) diagnosed according to prior criteria proposed by Schuyler and colleagues (11, 18). Caillaud and colleagues reported BALL findings in 139 HP patients from five French centers categorized according to presentations of acute, subacute, or chronic disease duration (6). These BALL findings were much higher with a mean of 42% in those classified as chronic. We again eliminated incorporation bias in our study by reviewing only patients meeting ‘moderate’ or higher diagnostic confidence levels without the contribution of BALL. With that approach, only 28% of fibrotic and 41% of non-fibrotic patients met current consensus criteria for diagnostic BALL. More so, among patients with fibrotic disease, the largest proportion were those with BALL counts less than 10% (42%), as might be observed in healthy non-smokers (19).
Depending on diagnostic cut-off and study-specific criteria for inclusion, reported frequencies of diagnostic BALL varied significantly but occurred less than 50% of the time in most studies, particularly when BALL was being assessed for diagnostic utility. This lower frequency suggests likely overlap with other ILD, as demonstrated by the findings of two recent meta-analyses(4, 10). Both found lower range sensitivity and specificity for BALL as a diagnostic predictor (depending on cut-off 20% vs 30%), with expected inverse relationship between sensitivity and specificity (higher specificity at the cost of lower sensitivity). This was particularly true among fibrotic patients where the AUC for differentiating f-HP from IPF or sarcoid was reported as poor at 0.54 and 0.44 respectively in one meta-analysis (4). Lower sensitivity and specificity with poor predictive characteristics for distinguishing HP from other ILD, particularly among fibrotic patients, suggests BALL may be more additive than characteristic or distinguishing, as further supported by our study after specific elimination of incorporation bias.
As such, in patients with suspected f-HP and often more clinically severe disease, predicting procedural yield may be relevant to minimize bronchoscopy-related risk or complications. We assessed pre-procedural parameters in those with fibrosis and found no presenting demographics, exposure type, pulmonary function, or positive precipitin serology correlated with a greater likelihood of diagnostic BALL. Only radiologic findings of centrilobular nodules and honeycombing appeared to support higher and lower likelihoods of diagnostic findings respectively. A similar correlation of BALL findings (using a 20% cut-off) with radiologic honeycombing was reported by De Sadeleer and colleagues(16). Honeycombing was found in 50.9% of those with lower BALL compared to only 13.9% of those with higher BALL. Considering the higher rate of honeycombing in their study compared to ours (36% vs 17%), no patient with honeycombing reached diagnostic BALL in our study.
BALL has also been reported to be predictive of long-term outcomes (13, 14). A study of 160 chronic HP patients found a positive association of higher BALL with improved survival (11). De Sadeleer and colleagues reported improved survival in fibrotic patients stratified by BALL cut-off > or < 20%, with stratification by honeycombing also having similarly poor survival when present. BALL was inversely correlated with honeycombing in their study, with the latter on adjusted Cox regression remaining predictive of poorer outcome while BALL was not. We reviewed univariable and multivariable predictors of all-cause mortality in our fibrotic cohort and found higher BALL at presentation to be independently predictive of survival. As BALL was also inversely colinear with honeycombing in our study, we did not adjust for honeycombing in the Cox regression model, noting again none with honeycombing had a diagnostic BALL in our cohort.
Our study has several limitations. First, its retrospective design can only assume correlation but not causation. We also did not specifically compare BALL findings in HP to other ILD subtypes, focusing on reporting BALL in carefully selected patients meeting updated and strict diagnostic guidance, categorized according to levels of diagnostic confidence. Our primary goal was also to apply this recent diagnostic approach with elimination of incorporation bias to obtain a sense of relevant BALL findings in otherwise diagnosable disease. This methodologic approach naturally selects for patients whose diagnoses were already likely to have been made without BALL and therefore cannot explore the role of BALL in increasing diagnostic confidence in those whose presenting data are inconclusive or missing. However, based on our findings bronchoscopy appears to be unjustified as positive BALL results in otherwise diagnosable disease appears to be low and likely similar or overlapping with other ILD. A majority of patients in our study also underwent biopsy to achieve inclusion criteria of ‘moderate’ or higher diagnostic confidence levels, suggesting more aggressive assessment was still pursued despite available BALL findings, which may not be typical of real-world practices (as reflected for example in the study by Adams et al (5)). This obtaining of additional biopsy is perhaps the result of lower overall BALL findings overall in our cohort or clinically suspected nonspecific elevation without other supportive criteria.