We used longitudinal data from the Massachusetts Male Aging Study, a large population-based random-sample cohort of men aged 40-70 yr at baseline, to establish normative age trends for serum level of T and related hormones in middle-aged men and to test whether general health status affected the age trends. Of 1,709 men enrolled in 1987-1989, 1,156 were followed up 7-10 yr afterward. By repeated-measures statistical analysis, we estimated simultaneously the cross-sectional age trend of each hormone between subjects within the baseline data, the cross-sectional trend between subjects within the follow-up data, and the longitudinal trend within subjects between baseline and follow-up. Total T declined cross-sectionally at 0.8%/yr of age within the follow-up data, whereas both free and albumin-bound T declined at about 2%/yr, all significantly more steeply than within the baseline data. Sex hormone-binding globulin increased cross-sectionally at 1.6%/yr in the follow-up data, similarly to baseline. The longitudinal decline within subjects between baseline and follow-up was considerably steeper than the cross-sectional trend within measurement times for total T (1.6%/yr) and bioavailable T (2-3%/yr). Dehydroepiandrosterone, dehydroepiandrosterone sulfate, cortisol, and estrone showed significant longitudinal declines, whereas dihydrotestosterone, pituitary gonadotropins, and PRL rose longitudinally. Apparent good health, defined as absence of chronic illness, prescription medication, obesity, or excessive drinking, added 10-15% to the level of several androgens and attenuated the cross-sectional trends in T and LH but did not otherwise affect longitudinal or cross-sectional trends. The paradoxical finding that longitudinal age trends were steeper than cross-sectional trends suggests that incident poor health may accelerate the age-related decline in androgen levels.