**Background:** Testosterone is a fundamental male sex hormone produced by testicular Leydig cells. Levels peak in the 20s–30s and gradually decline with age. Low testosterone may predispose older men to poor outcomes in COVID-19. Chronic sleep changes are common in older adults, and testosterone production is dependent on sleep with a circadian rhythm. However, large-scale population-based studies exploring the effect of sleep duration on testosterone levels in older men have been lacking. The authors aimed to investigate this association using data from the West China Health and Aging Trend (WeCHAT) study.
**Methods:** This cross-sectional analysis included 1628 men aged ≥60 years from the WeCHAT study, an ongoing prospective multicenter cohort initiated in 2018 across four western Chinese provinces (Sichuan, Yunnan, Guizhou, Xinjiang) involving 7538 people from 18 ethnic groups. Data were collected via electronic questionnaire and in-person interviews. Sleep duration was self-reported in response to: "During the past month, how many sleep time did you get at night (average hours for one night)?" Total testosterone (TT) was measured from morning blood samples. Covariates included BMI, age, ethnic group, education level, marital status, smoking history, alcohol use, cognitive performance (Chinese version of 10-point Short Portable Mental Status Questionnaire), nutritional status (Mini Nutritional Assessment Short Form), physical frailty (modified Fried frailty phenotype), and chronic diseases (diabetes, hypertension, COPD, coronary heart disease, stroke). A generalized additive model with smooth curve fitting (penalized spline method) was used to assess non-linear associations. An inflection point was calculated using a recursive algorithm, and two-phase piecewise linear regression was constructed. Multivariable linear regression models were adjusted in two stages: Model I (BMI, age, ethnic group, education level, marital status, smoking history, alcohol use) and Model II (all Model I variables plus cognitive performance, nutritional status, physical frailty, chronic diseases). Subgroup analyses by BMI and sensitivity analyses were performed.
**Key Results:** The mean age was 68.5 ± 5.9 years, mean BMI was 24.59 ± 3.72 kg/m², and median TT was 428.00 (314.00–554.25) ng/dL. The association between sleep duration and TT was non-linear. The inflection point was 9.5 hours (P for log-likelihood ratio test: <0.001). There was a significant threshold effect: on the left of the inflection point (<9.5 h), effects were insignificant; on the right (≥9.5 h), longer sleep was significantly associated with higher TT (β = 640.92, 95% CI 380.42–901.42; P <0.001). Based on the 9.5 h cut-off, 1477 men were in the short-sleep group and 151 in the long-sleep group. In linear regression, long sleep was significantly associated with higher TT compared to short sleep in Model I (β = 436.48, 95% CI 181.81–691.15; P <0.001) and Model II (β = 617.84, 95% CI 227.45–1008.23; P = 0.002). When sleep duration was analyzed as a continuous variable, increased sleep duration was significantly associated with increased TT (β = 75.04, 95% CI 8.21–141.86; P = 0.028). In subgroup analysis by BMI, longer sleep in the BMI <25 kg/m² group was significantly associated with increased TT (β = 134.98, 95% CI 7.90–262.06; P = 0.038), but no significant association was found in the BMI ≥25 kg/m² group. A significant interaction between BMI and sleep duration on TT was identified (P for interaction = 0.010).
**Clinical Implications:** This study provides population-based evidence that longer sleep duration (≥9.5 h) is associated with higher testosterone levels in older men with normal BMI (<25 kg/m²). The findings suggest that ensuring adequate sleep may be an effective non-pharmacological intervention to improve androgen concentrations in older men with low testosterone. However, due to the cross-sectional design, causal relationships cannot be established. Limitations include self-reported sleep duration (potential recall bias), unavailability of sex hormone-binding globulin and free testosterone measurements, and the inability to draw causal inferences. The study is notable for its large multicenter sample from a prospective cohort, comprehensive data collection, and standardized laboratory measurements in a representative sample of older men in Western China.