We initiated the present study to predict interstitial lung involvement in rheumatoid arthritis (RA) by means of logistic regression analysis of clinical data, lung function, chest roentgenogram, and serologic parameters. Fifty-eight nonsmoking patients with RA were randomized from the rheumatologic unit and sent for investigation to the pulmonary department. Bronchoalveolar lavage (BAL) was performed in the middle lobe and the BAL fluid was considered abnormal in case of increased cell count per milliliter and/or lymphocytosis and/or neutrophil granulocytosis; these findings or combinations thereof were found in 42 (72.4 percent) of 58 cases. The patients' data that had an impact on the normality of BAL were the sex (p = 0.001), vital capacity (p = 0.028), peripheral blood T-helper cells (OKT4+) (p = 0.025), DR(+)-lymphocytes (p = 0.002), and antinuclear antibodies (p = 0.025). By means of the logistic regression analysis, it was possible to reach high significance in the prediction of interstitial lung involvement, with a sensitivity of 92.9 percent and a specificity of 75.0 percent (p less than 10(-6)). The efficiency of prediction was 87.9 percent. From these results, we conclude that interstitial lung involvement in RA is predictable from laboratory findings that have been yielded by noninvasive diagnostic techniques. These data should be used in clinical routine monitoring and they may help to facilitate the assessment of whether bronchoscopy is indicated.