Objective To determine serum concentrations, intra-individual variability and impact of age-related co-morbidities on serum testosterone (T), dihydrotestosterone (DHT), estradiol (E2) and estrone (E1) in older men. Design Observational, repeated measures study. Participants Men (n = 325) with 40 years and older self-reporting very good or excellent health. Measurements Standardized history, physical examination and collection of nine blood samples at fixed time intervals were measured over 3 months (three at 20 min intervals on days 1 (fasting) and 2 (non-fasting), one at days 7, 30 and 90). Serum T, DHT, E2 and E1 (n = 2900, > 99% of scheduled samples) measured by liquid chromatography-tandem mass spectrometry (LC-MS) were analysed by linear mixed model analysis with fasting, age and obesity as covariables. Results Mean serum T did not vary with age (P = 0·76) but obesity (-0·35 nm per body mass index (BMI) unit, P <0·0001) and ex-smoker status (-1·6 nm, P <0·001) had significant effects. Serum DHT was increased with age (+0·011 nm per year, P = 0·001) but decreased with obesity (-0·05 nm per BMI unit, P <0·0001). Serum E2 did not vary with age (P = 0·31) or obesity (P = 0·12). Overnight fasting increased (by 9-16%, all P <0·001) and reduced variability in morning serum T, DHT, E2 and E1. Non-fasting serum T and DHT were stable over time (day, week, month or 3 months; P > 0·28). Conclusions Serum T, DHT and E2 displayed no decrease associated with age among men over 40 years of age who self-report very good or excellent health although obesity and ex-smoking status were associated with decreased serum androgens (T and DHT) but not E2. These findings support the interpretation that the age-related decline in blood T accompanying non-specific symptoms in older men may be due to accumulating age-related co-morbidities rather than a symptomatic androgen deficiency state.