PubMedDate not supplied
Our objective was to assess the capacity of clinical and laboratory information to predict findings in the lung biopsy in interstitial lung diseases (ILD). We studied 121 patients with ILD as a cohort recruited in our institute from 1983 to 1987 with the diagnosis of hypersensitivity pneumonitis (HP) and usual interstitial pneumonia (UIP). Histologic diagnosis (HP vs UIP) and degree of fibrosis (< 50% of the biopsy surface vs > or = 50%) were used as the gold standard to compare a series of clinical and laboratory variables in the initial assessment. We used a stepwise logistic regression model to predict the biopsy results. The model was calculated in half of the patients selected by random sampling, and the calculated model was tested in the other half of the patients. Variables found to predict degree of fibrosis were (with relative risk RR and 95% confidence interval): a radiographic pattern of honeycombing (RR 5.0 from 0.9-29), digital clubbing (RR 8 from 1.4-48) and gender (RR 2.9 from 0.4-20). This model classified correctly 72% of the biopsies, with a sensitivity of 0.38, a specificity of 0.85 and a kappa of 0.25 +/- 0.19 (p = 0.17 NS). For histologic diagnosis (NIU vs NH), the model included gender (RR 6.6, 1.3-33), honeycombing (RR 1.6, from 0.4-6.0), digital clubbing (RR 4.6, from 1.2-18), and vital capacity expressed as percent of predicted (RR 0.96, from 0.92-1.00).(ABSTRACT TRUNCATED AT 250 WORDS)