Nineteen studies (n=272) were eligible for inclusion in the review : 4 RCTs, of which 2 included a placebo treatment group, and 15 quasi-experimental studies.
Fixed-effect models demonstrated that the concentrations of total cholesterol (WMD -14 mg/dL, 95% CI: -17, -11), LDL cholesterol (WMD -5 mg/dL, 95% CI: -8, -1) and HDL cholesterol (WMD -4 mg/dL, 95% CI: -5, -2) decreased significantly after treatment, but that the concentration of triglyceride did not change (WMD -1 mg/dL, 95% CI: -6, 4). For each lipid subtype, the study-specific differences were heterogeneous. Combining the differences in a weighted random-effects model meant that only total cholesterol (WMD -12 mg/dL, 95% CI: -22, -2) and HDL cholesterol (WMD -4 mg/dL, 95% CI: -6, -2) decreased significantly. The exclusion of a single study with an outlying pre-treatment HDL cholesterol concentration removed the heterogeneity in the remaining studies (fixed-effect WMD -3 mg/dL, 95% CI: -4, -2).
The post-treatment minus pre-treatment difference in HDL cholesterol varied inversely with testosterone dosage 9 = -0.53, p=0.055). Therefore, decreases in HDL cholesterol were larger at lower dosages of testosterone ester. These were not explained by attrition, regression to the mean, dosing frequency or duration, concomitant elevation of plasma total testosterone, aromatisation of testosterone to estradiol, or other study or patient characteristics.