Background and AimsTestosterone deficiency is often related to geriatric syndrome including erectile dysfunction, osteo‐porosis, depression, cognitive impairment, cardiovascular diseases and frailty. Despite the existence of many studies on short‐term exercise and serum testosterone levels, few research have focused on exercise habits from young age and testosterone values in middle‐aged male. In this study, we investigated whether belonging to an athletic club from school age could predict serum‐free testosterone (FT) levels.MethodsThe subjects were 1609 middle‐aged male outpatients aged 40 years or older (median: 61 years, interquartile range: 54–69) who visited our hospital. Participants had their FT values measured in the morning hours during the period from December 2007 to June 2009. A questionnaire survey on exercise habits was conducted at the same time as the measurements. The exercise habit questionnaire was created based on whether the patients belonged to an athletic club in (a) elementary school, (b) junior high school, (c) high school, (d) college, (e) adult life, and (f) at the time of the test.ResultsThere was only one positive response to the questionnaire among 456 patients (28% of total), followed by zero for 358 patients (22% of total). The number of patients with low‐testosterone levels (FT < 8.5 pg/mL) according to the Japanese diagnostic criteria for late‐onset of hypogonadism was 839 (52.1%). In multivariate analysis, it was shown that with low‐testosterone levels (FT < 8.5 pg/mL), age (odds ratio [OR]: 1.065, 95% confidence interval [CI]: 1.052–1.079; p < 0.001), hypertension (OR: 3.489, 95% CI: 2.728–4.462; p < 0.001), type‐2 diabetes (OR: 3.035, 95% CI: 2.296–4.01; p < 0.001), and dyslipidemia (OR: 2.039, 95% CI: 1.558–2.668; p < 0.001) were risk factors, and more than two positive responses to the questionnaire (OR: 0.886, 95% CI: 0.802–0.980; p = 0.018) were also a significant independent factor.ConclusionA sports club membership during school years may affect future testosterone levels.