To determine overall and disease-related accuracy of the clinical/imagiological evaluation for pulmonary infiltrates of unknown aetiology, compared with the pathological result of the surgical lung biopsy (SLB) and to evaluate the need for the latter in this setting.
We conducted a retrospective review of the experiences of SLB in 366 consecutive patients during the past 5 years. The presumptive diagnosis was based on clinical, imagiological and non-invasive or minimally invasive diagnostic procedures and compared with the gold standard of histological diagnosis by SLB. We considered five major pathological groups: diffuse parenchymal lung disease (DPLD), primitive neoplasms, metastases, infectious disease and other lesions. Patients with previous histological diagnosis were excluded.
In 56.0% of patients (n=205) clinical evaluation reached a correct diagnosis, in 42.6% a new diagnosis was established (n=156) by the SLB, which was inconclusive in 1.4% (n=5). The pre-test probability for each disease was 85% for DPLD, 75% for infectious disease, 64% for primitive neoplasms and 60% for metastases. Overall sensitivity, specificity, positive and negative predictive values for the clinical/radiological diagnosis were 70%, 90%, 62% and 92%, respectively. For DPLD: 67%, 90%, 76% and 85%; primitive neoplasms: 47%, 90%, 46% and 90%; metastases: 99%, 79%, 60% and 99%; infectious disease 38%, 98%, 53% and 96%.
Despite a high sensitivity and specificity of the clinical and imagiological diagnosis, the positive predictive value was low, particularly in the malignancy group. SLB should be performed in pulmonary infiltrates of unknown aetiology because the clinical/imagiological assessment missed and/or misdiagnosed an important number of patients.