To assess the efficacy of treatment of parasitological excretion of cysts and trophozoites and symptoms of patients with giardiasis, a systematic review of published randomized clinical trials was conducted through extensive searches in Medline, Embase and Current Contents from 1966 till 1996 as well as manual reviews of 28 journals. The methodological quality of all trials was assessed by guidelines of the Cochrane Collaboration. Thirty-one trials were included, only one of which had no serious methodological flaws. The mean score of parasitological examination was 4.8 out of a possible 15. There was a considerable effect in cure rate of treatment versus placebo (odds 9.3, 95% CI 4.69-18.4), but all 3 trials in this comparison had serious flaws. Metronidazole treatment over more than 3 days seems to achieve a better parasitological cure rate than other long treatment courses (pooled odds 2.6, 95% 1.7-3.8), but trials are clinically and statistically heterogeneous. Single-dose therapy is as effective as longer treatment courses (pooled odds 0.67, 95% 0.31-1.44). Within the single-dose regimens tinidazole (2 g) reaches a higher parasitological cure rate than other short therapies (pooled odds 55, 95% CI 3.7-8.3) with relatively few side-effects. Placebo-controlled trials with parasitological and clinical outcomes are needed.