Three different forms of testosterone (T) replacement therapy were compared; they were the intramuscular injection of mixed testosterone esters 250 mg; the subcutaneous implantation of 6 X 100 mg pellets of fused testosterone; and the oral administration of testosterone undecanoate (TU) 80 mg twice daily. Six hypogonadal males were treated with oral TU for an eight week period, during which time serial serum hormonal estimations were performed over 10 h at the initiation and after four and eight weeks of therapy. Serum T levels showed marked variability both between subjects and within the same subject on different occasions. We attribute this to variability in absorption of TU, which is formulated in oleic acid. The overall mean T level calculated from the areas under the profiles of TU was 12.0 nmol/l. Hormone responses to injected T esters were studied in nine hypogonadal males. Serum T rose to supraphysiological peak concentrations (mean 71 nmol/l) 24-48 h after an injection, followed by an exponential decay to reach baseline concentrations after 2-3 weeks. The overall calculated mean T level in subjects receiving testosterone esters 250 mg every three weeks was 27.7 nmol/l. Subcutaneous implantation of testosterone in six hypogonadal men produced a gradual rise in serum T followed by a slow decline, with T levels remaining within the normal range for 4-5 months. The calculated overall mean T level over 21 weeks after implantation was 17.0 nmol/l. Serum oestradiol (E2) levels remained within the normal male range throughout the study periods on both TU and T implant therapy but showed a supraphysiological peak (mean 347 pmol/l) 24-48 h after a T injection. 5 alpha-dihydrotestosterone (DHT) levels appeared to parallel those of T on the three forms of therapy, with DHT:T ratios being highest for TU therapy. This was also true for the target organ metabolite 5 alpha-androstane-3 alpha,17 beta-diol. At the doses studied drug costs were similar for T implantation (every 5 months) and T ester injections (every 3 weeks), but were 7-8 times higher for TU (80 mg twice a day). We conclude that T implantation remains overall the most physiological form of androgen replacement therapy, is generally well accepted and attended by few side effects; TU may have a useful role in the initial phases of therapy.