Background: The combination of male gender, obstructive sleep apnoea (OSA) and obesity magnifies cardiometabolic risk. There has been no systematic study evaluating whether testosterone therapy can improve cardiometabolic health in obese men with OSA by improving body composition, visceral abdominal fat and insulin sensitivity. Objective: To assess body compositional and cardiometabolic effects of testosterone treatment in obese men with severe OSA. Design: An 18-week randomised, double-blind, placebo-controlled and parallel group trial in 67 men. Methods: Participants (age = 49±12 years, apnoea hypopnoea index = 39.9 ± 17.7 events/h, BMI = 31.3 ± 5.2 kg/m2) were placed on a hypocaloric diet and received i.m. injections of either 1000 mg testosterone undecanoate (n = 33) or placebo (n = 34) for 18 weeks. Outcomes were the changes in body composition (total muscle mass, total and abdominal fat, total body dual-energy X-ray absorptiometry and computerised tomography (CT)), weight, insulin sensitivity (homeostasis model assessment), abdominal liver fat (CT), arterial stiffness (pulse wave analysis), resting metabolic rate and respiratory quotient (indirect calorimetry) and blood lipids and metabolic syndrome from baseline to week 18. Results: After 18 weeks, testosterone treatment increased insulin sensitivity (-1.14 units, 95% confidence interval (95% CI) -2.27 to -0.01, P<0.05), reduced liver fat (0.09 Hounsfield attenuation ratio, 95% CI 0.009 to 0.17, P = 0.03) and increased muscle mass (1.6 kg, 95% CI 0.69 to 2.5, P = 0.0009) to a greater extent than placebo. Other measures of body composition and regional adiposity as well as the number of participants with metabolic syndrome did not change. Testosterone also decreased arterial stiffness (augmentation index) by 3.2% (95% CI -6.01 to -0.46%, P = 0.02) and decreased the respiratory quotient (95% CI -0.04, -0.08 to -0.001, P = 0.04) after 18 weeks compared with placebo. Conclusion: Eighteen weeks of testosterone therapy in obese men with OSA improved several important cardiometabolic parameters but did not differentially reduce overall weight or the metabolic syndrome. Longer term studies are required.