Ageing in men is associated with a gradual decline in serum testosterone levels and a concomitant loss of muscle mass, accumulation of central adiposity, impaired mobility and increased risk of bone fractures. Whether androgen treatment might be beneficial in these subjects is still under debate. We have carried out a systematic review of randomized controlled trials (RCTs) evaluating the effects of testosterone (T) administration to middle-aged and ageing men on body composition, muscle strength, bone density, markers of bone metabolism and serum lipid profile. A comprehensive search of all published randomized clinical trials was performed using the MEDLINE, Cochrane Library, EMBASE and Current Contents databases. Guided by prespecified criteria, software-assisted data abstraction and quality assessed by two independent reviewers, 29 RCTs were found to be eligible. For each investigated variable, we reported the results of pooled estimates of testosterone treatment using the random effect model of meta-analysis. Heterogeneity, reproducibility and consistency of the findings across studies were explored using sensitivity and meta-regression analysis. Overall, 1,083 subjects were evaluated, 625 randomized to T, 427 to placebo and 31 to observation (control group). Weighted mean age was 64.5 years (range 49.9--77.6) and mean serum testosterone was 10.9 nmol/l (range 7.8--19). Testosterone treatment produced: (i) a reduction of 1.6 kg (CI: 2.5--0.6) of total body fat, corresponding to -6.2% (CI: 9.2--3.3) variation of initial body fat, (ii) an increase in fat free mass of 1.6 kg (CI: 0.6--2.6), corresponding to +2.7% (CI: 1.1--4.4) increase over baseline and (iii) no change in body weight. The effects of T on muscle strength were heterogeneous, showing a tendency towards improvement only at the leg/knee extension and handgrip of the dominant arm (pooled effect size=0.3 standard mean difference (SMD), CI: -0.0 to 0.6). Testosterone improved bone mineral density (BMD) at the lumbar spine by +3.7% (CI: 1.0--6.4%) compared to placebo, but not at the femoral neck, and produced a consistent reduction in bone resorption markers (pooled effect size = -0.6 SMD, CI: -1.0 to -0.2). Testosterone also reduced total cholesterol by 0.23 mmol/l (CI: -0.37 to -0.10), especially in men with lower baseline T concentrations, with no change in low density lipoprotein (LDL)-cholesterol. A significant reduction of high density lipoprotein (HDL)-cholesterol was found only in studies with higher mean T-values at baseline (-0.085 mmol/l, CI: -0.017 to -0.003). Sensitivity and meta-regression analysis revealed that the dose/type of T used, in particular the possibility of aromatization, explained the heterogeneity in findings observed on bone density and HDL-cholesterol among studies. The present analysis provides an estimate of the average treatment effects of testosterone therapy in middle-aged men. Our findings are sufficiently strong to justify further interventional studies focused on alternative targets of androgenic treatment carrying more stringent clinical implications, in particular the cardiovascular, metabolic and neurological systems.