Correlation of Radiographic Abnormalities on Computer Tomography (CT) with Broncho-Alveolar Lavage (BAL) Results. What are Our Radiologists Reading?

Background: Computed tomography (CT) of the chest is routinely performed in patients with pulmonary complications post-hematopoietic cell transplant (HCT). Radiographic abnormalities on CT frequently lead to empiric changes in patient management, without diagnostic confirmation. We report a large case series of HCT recipients with paired chest CT and broncho-alveolar lavage (BAL), examining radiographic correlates with BAL outcome. Methods: A retrospective review of our HCT database identified 518 patients (median age 52 years, 0-75 years) who underwent bronchoscopy paired with a concurrent chest CT (median time from CT to BAL: 3 days). CT scans from the 518 patients were examined by two teams of radiologists (2 radiologists/team) blinded for clinical history, scan report, and BAL outcome. Scans were evaluated for the presence of nodules (≤1 cm, >1 cm, tree in bud, ground glass, solid, cavitary, centrilobular, halo sign), airspace disease (consolidation, ground glass, focal, diffuse), the extent and distribution of these findings, as well as pleural effusion, bronchiectasis, septal lines, and air trapping. Association of radiographic abnormalities with BAL outcome were performed. Results: Common radiographic abnormalities included airspace disease (ASD) (n = 327) and pulmonary nodules (n = 285). Other than cavitary lesions, all other radiographic findings were associated with <50% incidence of identifying a pathogen on BAL (Table 1). Radiographic features with the highest positive predictive value (PPV) for pathogen identification included cavitary lesions (PPV = .56), nodules >1 cm (PPV = .47) and unilobar ASD (PPV = .43). In contrast, the presence of ground glass nodules (PPV = .28) and diffuse ASD (PPV = .19) had the lowest PPV for pathogen identification. Nodules with a solid (P = .016) or cavitary architecture (P = .034), and nodules >1 cm (P < .001) were each associated with an increased likelihood of identifying a pathogen on BAL. We were unable to detect an association between nodules or air space disease and any pathogenic subtype (fungal, viral, or bacterial). However, aspergillus pneumonia (n = 40) was associated with a centrilobar location of nodules (P = .043). CMV (n = 30) was associated with the presence of nodules with a ground glass character (P = .046), and unilobar (not diffuse) distribution (P = .043). PJP was associated with a halo sign (P = .01) and diffuse airspace disease (P = .02).Table 1ASD(n = 327)GGO(n = 274)Nod <1 cm(n = 223)Nod >1 cm(n = 81)TIB(n = 81)Halo(n = 34)Positive BAL28%26%32%47%35%41%Negative BAL72%74%68%53%65%59%Indicates presence (or absence) of a pathogen on BAL and radiographic finding.ASD, airspace disease; GGO, ground glass opacification; Nod, Nodules; TIB, Tree-in-Bud. Open table in a new tab Indicates presence (or absence) of a pathogen on BAL and radiographic finding. ASD, airspace disease; GGO, ground glass opacification; Nod, Nodules; TIB, Tree-in-Bud. Conclusion: Computed tomography findings in HCT recipients are non-specific, with low predictive value in identifying any particular pathogen.