We investigated whether analysis of cellular composition (including lymphocyte subsets) in bronchoalveolar lavage (BAL) fluid at the start of follow-up in patients with untreated sarcoidosis has any predictive value for further evolution of the disease. The outcome was evaluated by the chest roentgenograms, the lung volumes, and the single breath diffusing capacity for CO (DCO) after 22 to 36 months. In contrast to the general belief, patients who improved radiologically had a significantly higher T4 cell count (as percentage of BAL lymphocytes) (p less than 0.02) and a higher T4-T8 ratio in the initial BAL sample (9.3 vs 3.2; p less than 0.05) than those whose chest roentgenogram showed deterioration or remained unchanged. Total cell count and the percentage of lymphocytes in BAL fluid were not different between both groups. The change in DCO at the end of the follow-up period correlated positively with the baseline BAL T4 cells (Rs = 0.44; p less than 0.05) and with the BAL T4-T8 ratio (Rs = 0.51; p less than 0.03) and negatively with the baseline BAL T8 cells (Rs = -0.48; p less than 0.04). In only three patients progression of the disease necessitated steroid therapy, and they all had a low to normal T4-T8 ratio in the initial BAL sample. Bronchoalveolar lavage was repeated at least once in ten patients. Improvement of the chest roentgenograms in these patients was accompanied by a decrease of the BAL T4 cell count (as percentage of lymphocytes) and of the T4-T8 ratio. We conclude that a high lymphocyte count, a high T4 cell count (as percentage of lymphocytes), and a high T4-T8 ratio in BAL fluid reflect an intense alveolitis at the time of the procedure, but they are not indicators of poor prognosis on which therapeutic decisions can be based.