OBJECTIVE: To determine the effect of 900 mg/day of dried garlic powder (standardised to 1.3% allicin) in reducing total cholesterol. DESIGN: Double-blind, randomised six-month parallel trial. SUBJECTS: 115 individuals with a repeat total cholesterol concentration of 6.0-8.5 mmol/l and low-density lipoprotein (LDL) cholesterol of 3.5 mmol/l or above after six weeks of dietary advice. INTERVENTION: The active treatment group received dried garlic tablets (standardised to 1.3% allicin) at a dosage of 300 mg three times daily. The control group received a matching placebo. OUTCOME MEASURES: Primary end-point: total cholesterol concentration; secondary end-points: concentrations of LDL and high-density lipoprotein cholesterol, apolipoproteins (apo) A1 and B, and triglycerides. RESULTS: There were no significant differences between the groups receiving garlic and placebo in the mean concentrations of serum lipids, lipoproteins or apo A1 or B, by analysis either on intention-to-treat or treatment received. In a meta-analysis which included the results from this trial, garlic was associated with a mean reduction in total cholesterol of -0.65 mmol/l (95% confidence intervals: -0.53 to -0.76). CONCLUSIONS: In this trial, garlic was less effective in reducing total cholesterol than suggested by previous meta-analyses. Possible explanations are publication bias, overestimation of treatment effects in trials with inadequate concealment of treatment allocation, or a type 2 error. We conclude that meta-analyses should be interpreted critically and with particular caution if the constituent trials are small.