Study objectives: To evaluate whether transbronchial biopsy (TBB) is useful for the diagnosis of acute farmer's lung (FL) by calculating the likelihood ratios (LHRs) of (1) simple pathologic criteria and (2) an overall assessment of the biopsy specimens.
Design: Retrospective study in which a blinded analysis of 105 TBBs with adequate material from patients with parenchymal diseases (55 cases of FL matched with 50 control samples) was performed by two independent pathologists.
Setting: Respiratory clinic of a university-affiliated referral center.
Measurements: Three pathologic criteria were first studied: (1) diffuse lymphocytic infiltration (LI); (2) focal LI; and (3) granulomas. Then, an overall assessment or the TBB was done. Four diagnostic categories were considered: (1) probable FL; (2) possible FL; (3) nonspecific; and (4) alternative diagnosis. LHRs favoring the diagnosis of FL were calculated for the pathologic criteria and for each diagnostic category.
Results: For both the pathologic criteria and the overall assessments, the interobserver agreement was fair. As a pathologic criterion, "diffuse LI" was better than "loosely formed granuloma" to discriminate FL from control samples (LHR, 9.1 [confidence interval, 2.2 to 37.0] vs 1.8 [confidence interval, 0.5 to 6.9]). After the overall assessment, as many as 48.6% of the TBBs were read as nonspecific. The LHRs of the four diagnostic categories were as follows: (1) probable FL: 1.1 (observer 1) and 2.6 (observer 2); (2) possible FL: 2.2 and 1.7; (3) nonspecific: 0.9 and 0.6; and (4) alternative diagnosis: 0.4 and 0.0.
Conclusion: Hematoxylin-eosin-stained TBB specimen is of limited usefulness for the diagnosis of FL and should be reserved for patients with intermediate pretest probability of FL. Diffuse LI best discriminates FL from control samples and should be specifically sought.