The effects of administration of testosterone (T) gel, resulting in plasma T levels in the low range of reference values, followed by testosterone undecanoate (TU), producing plasma T levels in the mid-normal range, were measured in 27 hypogonadal men aged 47-74 years. T gel had positive effects on the International Index of Erectile Function, the Aging Males' Symptoms Scale and International Prostate Symptoms Score and on the metabolic syndrome. The improvement was larger when TU was administered and plasma T levels were higher. The reduction in waist circumference and plasma cholesterol were larger with TU than with T gel, while the increases in plasma high-density lipoprotein and sex hormone binding globulin (an indicator of the severity of the metabolic syndrome) were larger with TU than with T gel. Both T gel and TU appeared safe on prostate parameters. Plasma haemoglobin and haematocrit were elevated but remained in the normal range. The assumption that treatment with T is adequate when achieved plasma levels of T are within the reference range is no longer tenable. Some androgen-dependent biological functions require higher plasma T levels than others, and, moreover, these thresholds differ among men.