Objective To evaluate the effectiveness and safety of misoprostol versus gemeprost for the first and second-trimester induction of abortion. Methods We searched randomized controlled trials（RCTs） and quasi-randomized controlled trials of misoprostol versus gemeprost for the first and second-trimester induction of abortion operation. MEDLINE （1974 - 2003 ）, EMBASE （1974 - 2003 ）, The Cochrane Library （Issue 3, 2003 ）, CBM （1979 - 2003 ）, CNKI （1989 - 2003 ）, CMCC （1994 - 2003 ）, VIP （1989 - 2003 ） were searched. Twelve journals and reference lists of eligible studies were handsearched. Two reviewers independently screened the studies for eligibility, evaluated the quality, and extracted data from eligible studies with confirmation by cross checking. Any disputes were decided by a third person. Meta-analysis was conducted using RevMan 4.2 software. Results Seven RCTs were included, three for the first-trimester and four for the second- trimester. Statictical differences were tested between the misoprostol and gemeprost groups with OR 4. 79,95% CI 1.35 to 16.94 in laxation and dilation rate,with WMD -1.50,95% CI -3.00 to 0.00 in operation time for the first-trimester abortion. However, there was no statistical difference in the amount of bleeding during operation with WMD - 14. 40, 95% CI 30.30 to 1.30. For the second-trimester abortion, compared with gemeprost, no consistent results of gravid excluding time were obtained. Compared with gemeprost, the rate of complete abortion was higher and digest system morbidity was lower in misoprostol alone group. But there was no statistical difference between the two groups when they were both combined with mifestone.Conclusions Some evidence indicates that compared with gemeprost, misoprostol may increase laxation and dilation rate of the cervix, shorten gravid exclusion time, improve the rate of complete abortion and make the laxation and dilation range of the cerix larger than gemeprost before the first and second-trimester abortive surgery. Due to the hmited evidence indentified, we can not draw a b conclusion. More radomiased controlled trails of high quahty are required.