We have compared the pharmacokinetics and pharmacodynamics of the three commonly used testosterone formulations in a prospective, randomized cross-over clinical trial. Plasma free and total testosterone and their ratio (proportion of unbound testosterone), sex hormone-binding globulin (SHBG), oestradiol, LH and FSH were measured in 15 hypogonadal men (nine hyper- and six hypogonadotrophic) who underwent, in a randomized sequence, three treatment periods each separated by an intervening washout period. The treatments were: (i) intramuscular injection of 250 mg mixed testosterone esters at 2-weekly intervals, (ii) oral testosterone undecanoate 120 mg bd, and (iii) subcutaneous testosterone pellets (6 x 100 mg). Pellet implantation gave the most prolonged effect with free and total testosterone levels being elevated for up to 4 months. This was accompanied by prompt and sustained suppression of plasma LH and FSH, an increase in plasma levels of oestradiol but no change in SHBG levels. In contrast, intramuscular injections induced marked but reproducible week-to-week fluctuations in free and total testosterone, which resulted in a small decrease in plasma SHBG levels, less marked suppression of LH and FSH and a smaller increase in plasma levels of oestradiol. Oral testosterone undecanoate produced the most variable plasma levels of free and total testosterone with a peak in the first treatment week and a fall thereafter and, despite maintenance of testosterone levels within the physiological range, there was no significant suppression of plasma levels of LH and FSH, and oestradiol levels were unchanged but levels of SHBG and total cholesterol were decreased. Free testosterone levels were increased disproportionately during testosterone treatment as the proportion of unbound testosterone was increased by all three treatments. All three testosterone preparations lowered plasma levels of urea and all were without biochemical or haematological toxicity. Reported sexual function was better maintained and side-effects were fewer with parenteral compared with oral treatments. The marked decrease in SHBG and cholesterol levels during oral testosterone undecanoate, when compared with parenteral treatments, occurred despite lesser androgenic effects (suppression of gonadotrophin levels and reported sexual function), which suggests that the liver is exposed to excessive androgenic load via the portal vein during oral treatment with testosterone esters. It is concluded that testosterone pellets give the closest approximation to zero-order (steady-state) delivery conditions for up to 4 months after a single insertion.(ABSTRACT TRUNCATED AT 400 WORDS)