We studied the interference of the sample matrix on digoxin radioimmunoassays using four commercial kits. Plasma samples from non-digitalized patients of the following categories were assayed: uncomplicated essential hypertension treated with spironolactone, uremia, and acute myocardial infarction (AMI). Digoxin 2.50 nmol/L was added to all samples. Digoxin in plasma from patients on spironolactone was overestimated by two of the kits (means 2.77 and 2.68 nmol/L, respectively; p less than 0.01) and underestimated in samples from uremic patients by one kit (2.32 nmol/L; p less than 0.01). The digoxin content of AMI plasma was overestimated by one kit (2.62 nmol/L; p less than 0.05). Significant differences were found between radioimmunoassays when estimating digoxin concentration in the same category of patient and within individual methods used for different categories. Precision expressed as 95% confidence intervals ranged from 0.43 to 0.80 nmol/L for the kits. Thus, deviations in recorded digoxin concentrations from the true values found, but were of secondary importance because of the relatively low precision of the assays.