Despite the increasing complexity of bronchoalveolar lavage (BAL) studies in asthmatics there are few published data concerning the variability of inflammatory parameters measured using this technique. We studied the intrasubject variability of cellular and solute parameters in 20 clinically stable, symptomatic, mild-to-moderate asthmatics, in repeat 180 mL BAL procedures performed 1 month apart. During the study, there was no change in disease activity or medication. Mean (SD) forced expiratory volume in one second (FEV1) was 3.2 (1.09) L at the first BAL and 3.05 (0.98) L at the second. The geometric mean dose of methacholine provoking a 20% decrease in FEV1 (PD20) was 23 micrograms (range 2-1,170 micrograms) at the first BAL and 28 micrograms (range 2-400 micrograms) at the second. There was considerable variability in the BAL cellular and solute parameters measured over the two procedures. Estimates of power calculated for subsequent studies involving this type of subject group were made from the observed variability. Sample sizes of less than 15 mean that differences have to be large in order to be detected in repeat BAL samples. However, there is little improvement in the power of BAL studies for sample sizes greater than 20, indicating that there is little gain in recruiting more than this number of subjects. Thus, our study indicates that although studies of the pathophysiology underlying asthma using BAL require considerable commitment, they are practicable.