UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, World Health Organization, Geneva, Switzerland
University of California, San Francisco General Hospital, San Francisco, CA, USA
A full list of the contributors and participating centres can be found on page 530.
Correspondence: Dr H. von Hertzen, Special Programme of Research, Development and Research Training, World Health Organisation, CH 1211 Geneva-27, Switzerland.
Objectives To compare the efficacy of two different regimens of mifepristone followed by misoprostol for medical abortion in women with menstrual delay of ≤ 35 days.
Design Double-blind, randomised controlled trial.
Setting Seventeen centres internationally.
Participants We enrolled 1589 healthy pregnant women with menstrual delay of ≤ 35 days who were requesting nonsurgical abortion.
Interventions Within gestational age strata, we randomly assigned women to receive a single oral dose of mifepristone, either 200 mg or 600 mg, followed in 48 h by misoprostol 400 μg by mouth. We concealed the allocation assignments from investigators and participants and maintained double-blinding throughout the study.
Main outcome measures Complete abortion was the principal outcome measure. We also compared rates of side effects such as abdominal pain.
Results The complete abortion rate with the lower dose of mifepristone was similar to that with the higher dose (89.3%vs 88.1%) The crude relative risk of failure to achieve complete abortion with the 200 mg dose compared with the 600 mg dose was 0.9 (95% CI 0.7 to 1.2). The likelihood of complete abortion was inversely related to gestational age, although this finding is exploratory in nature. Among failures the percentage of women with continuing pregnancies increased from 1.4% at menstrual delay of two weeks or less to 9.0% when the delay was 4–5 weeks. Low efficacy led to stopping enrolment at 29 to 35 days' menstrual delay. Stopping criteria were also met at completion of the study in the group with 22–28 days' menstrual delay. No significant differences emerged in the frequency of side effects between the two mifepristone groups.
Conclusions Both regimens had similar efficacy. Women with a menstrual delay of four to five weeks had twice the risk of failure to abort compared with those who received treatment within two weeks of the expected menses. The efficacy of the mifepristone-prostaglandin regimen was not reduced by decreasing the dose of mifepristone from 600 mg to 200 mg. The regimens of 600 mg or 200 mg of mifepristone, followed by a single oral dose of misoprostol 400 μg 48 hours later, were not sufficiently efficient in inducing abortion when the menstrual delay was > 21 days.
The antiprogestogen mifepristone followed two days later by a suitable prostaglandin analogue is registered as a nonsurgical alternative to surgical termination of early intrauterine pregnancy in France and China (up to seven weeks of amenorrhoea) and in Great Britain and Sweden (up to nine weeks of amenorrhoea). The prostaglandin used in France and China is oral misoprostol; in Great Britain and Sweden vaginal gemeprost is also used in association with mifepristone. Compared with gemeprost, misoprostol has several advantages such as lower cost and stability at room temperature, and it is associated with much less pain than gemeprost1.
The recommended dose of mifepristone in Europe is 600 mg. Multicentre trials conducted under the auspices of the World Health Organisation (WHO) have shown that the same effectiveness as seen with 600 mg can be achieved by treatment with either small repeated doses2 or 200 mg in a single dose3. In both trials, the prostaglandin analogue used was gemeprost, and it may well be that efficacy would be lower if a less potent prostaglandin such as misoprostol is employed. The use of a 200 mg dose of mifepristone in a mifepristone-misoprostol regimen would lower the cost of the treatment and would be in accordance with the general pharmacological principle of using the lowest effective dose of any drug. The lower dose should not be used, however, if it is found to be less effective.
The purpose of this multinational trial was to compare the abortifacient efficacy and side effects of two single doses of mifepristone (200 mg vs 600 mg) followed 48 hours later by an oral dose of 400 μg of misoprostol in women with an early pregnancy (≤ 35 days delay in menses). Our a priori hypothesis was that mifepristone 200 mg would have a complete abortion rate 3–4% lower than that with the higher dose.
We designed a multicentre, double-blind, randomised controlled trial of the two regimens. The protocol was approved by the institutional review board of the World Health Organisation and that of each of the participating centres. Seventeen centres participated in the trial: Beijing, Havana, Helsinki, Ho Chi Min City, Hong Kong, Ljubljana, Melbourne, Moscow, Mumbai, Shanghai, Stockholm, St. Petersburg, Szeged, Tbilisi, Tianjin, Tunis, and Yerevan.
The study population included pregnant women requesting legal termination of pregnancy who were in good general health, with a menstrual delay of ≤ 35 days from the expected onset of menses, and who were older than the age of legal consent. Other inclusion criteria included a positive pregnancy test and a uterine size on pelvic examination consistent with menstrual history. In addition, to be enrolled in the study each participant had to be willing to participate and to use other than hormonal or intrauterine contraception until the first menses after abortion. Each woman signed an informed consent form before enrolment, and had a medical and gynaecological examination, including height and weight and a cervical cytology and haemoglobin determination. Investigators requested that each woman maintain a diary card to note days of bleeding and the occurrence of side effects.
Exclusion criteria were the following: medical conditions which would contraindicate the use of either mifepristone (e.g. adrenal disease) or misoprostol (e.g. hypertension or glaucoma); a history of thromboembolism, liver disease, or pruritus of pregnancy; regular use of prescription drugs before recruitment; use of an intrauterine device; and suspected or proven ectopic pregnancy. Other criteria included heavy cigarette smoking (more than ten cigarettes per day over the past two years), breastfeeding, and irregular menses.
The trial had two treatment arms. One group received a single oral dose of mifepristone 600 mg by mouth on day one of the study, followed by a single oral dose of misoprostol 400 μg on day three (48 h later). The other group received a single oral dose of mifepristone 200 mg, again followed by oral misoprostol 400 μg on day three. After administration of misoprostol, the women were observed for three hours. Participants returned for follow up evaluations on days 15 and 43 after beginning the treatment. All had haemoglobin determinations on entry and at both follow up visits.
The primary outcome measure for the trial was complete abortion. Investigators made this determination at the first follow up visit, using ultrasonography if necessary. Complete abortion was confirmed by passage of the products of conception, a negative pregnancy test, or an empty uterus on ultrasonography, and no emergency or elective curettage during the period up to first menstruation. Incomplete abortion included those requiring curettage for completion. A missed abortion included a nonviable pregnancy (gestational sac without cardiac activity) on ultrasonography. A failed attempted abortion was a continuing pregnancy despite treatment. When the woman was thought to have a failed attempted abortion or missed abortion but ultrasonography was not available, we categorised the abortion as an unclassified failure, and the tissue obtained on vacuum aspiration was sent for histologic evaluation. Secondary outcome measures included side effects: nausea, vomiting, diarrhoea, lower abdominal pain after mifepristone and before misoprostol, after misoprostol and at any time during treatment. Another secondary outcome was the need for blood transfusion. Investigators did hourly assessments of vital signs and side effects during the three-hour observation period after misoprostol administration.
Our a priori hypothesis was that the lower dose of mifepristone would be less effective. We assumed a complete abortion rate of 96% for the higher dose and of 92% for the lower dose. With an alpha of 0.05 in a onesided test and 90% power, we anticipated that 601 participants would be needed in each group. To increase the power further and to allow for losses to follow up, a total of 1589 women was enrolled (Fig. 1).
The data were analysed centrally in Geneva. We compared baseline demographic characteristics of women in both groups using descriptive statistics and calculated crude rates for outcome measures with exact confidence intervals by the binomial distribution4. Outcomes between groups or categories of delay were compared using relative risks or odds ratios and 95% confidence intervals, χ2 tests or χ2 tests for trends as appropriate. The 95% confidence intervals for the crude relative risks were calculated using Taylor series. Odds ratios and 95% confidence intervals to adjust for centres were obtained by exact methods using StatXact software. Logistic regression was employed when adjusting for centres, and to assess interactions, which were introduced in the model one at a time due to the sparseness of the data. Diagnostics were used to assess validity of the estimates. Homogeneity of odds ratios across centres was also assessed using the Mantel-Haenszel method
The main analysis was by intention-to-treat including women eventually found to have been ineligible for the study. We kept those who deviated from the assigned regimen in the groups to which they had originally been allocated. We pooled those lost to follow up with undetermined failures in a single outcome category. When the outcome was dichotomised as to failure or success, those lost to follow up and undetermined failures were deemed failures. However, when the study was ongoing, we concluded that it had an equivalence nature. To protect the inferences from a possible increase in type II error when testing our a priori hypothesis in the main analysis, we decided to perform a secondary analysis by excluding women with unknown outcomes, those who had not taken misoprostol and those with protocol violations.
Stopping rules were established for the study before enrolling women, with the purpose of protecting them from a treatment with low efficacy. We determined in advance that the upper 95% confidence limit for the complete abortion rate in a dose group should not fall below 90%. Staff in Geneva examined the results after enrolment of every 100 women in each group. In September 1994, staff exercised this stopping rule for the highest gestational age stratum, and enrolment in this stratum stopped at a total of 131 women. Just before the end of the study stopping criteria were also met for the second highest gestational age group (22–28 days' delay in menses), but all subjects had already been recruited for the study.
The individual woman served as the unit of randomisation. We used a computer-generated random number sequence developed by staff in Geneva to assign participants to treatment groups. Each centre received assignments by random permuted blocks with a fixed block size of four. Stratified randomisation was done at all centres using four gestational age strata: menstrual delay of ≤ 14 days, 15–21 days, 22–28 days, and 29–35 days. The allocation sequence was concealed from investigators and participants by using identically-appearing pill bottles. We provided the allocation sequence to Roussel-Uclaf (Romainville, France), which then filled the bottles accordingly. Investigators received only pill bottles labelled with the study number, the name of their centre, the sequentially-assigned participant number, and the expiry date of the batch of mifepristone. Participants received the drugs only after signing informed consent and being enrolled in the study. Since each woman received the same dose of misoprostol, we did not blind this feature.
Double masking was maintained throughout the study. Each investigator received mifepristone in identically-appearing bottles containing the same number of tablets: three tablets of mifepristone 200 mg or one tablet of mifepristone 200 mg plus two placebo tablets identical in appearance. The allocation sequence was kept by staff in Geneva and was not provided to investigators in the centres. Participants and investigators were kept unaware of the dose of mifepristone each woman received. Analysts in Geneva were not blinded as to treatment.
Investigators in the 17 centres enrolled a total of 1589 women (Fig. 1). Overall, 792 women received the lower dose of mifepristone and 797 the higher dose. Seventeen women who were eventually found to have been ineligible for enrolment were kept in the analysis. Among these, eight had menstrual delay outside the study criteria, one was breastfeeding, six had missing information (neither pregnancy test at admission nor ultrasound examination, or no menstrual information), and two took prostaglandin synthetase inhibiting drugs.
Baseline characteristics of women in both treatment groups were similar in all important respects (Table 1). The women were young and had similar obstetrical histories, menstrual cycle lengths, and gestational ages. Height, weight, and blood pressure were comparable in both groups.
Table 1. Baseline characteristics of participants by treatment group. Values are given as mean (SD) or n [%].
Dose of mifepristone
200 mg(n= 792)
600 mg(n= 797)
Menstrual delay (days)
Menstrual delay (days)
Total with delay information
Both doses of mifepristone had comparable efficacy (Table 2). The crude complete abortion rate with the 200 mg dose of mifepristone was 89.3% (707/792), while that for the 600 mg dose was 88.1% (702/797). As shown in Table 3, the crude relative risk of failure to achieve complete abortion with the lower dose was 0.9 (95% CI 0.7–1.2; χ2 test: χ2= 0.45, P = 0.50).
Table 2. Outcomes of treatment by treatment group. Values are given as n (%) [95% CI].
Dose of mifepristone
200 mg (n= 792)
600 mg (n= 797)
707 (89.3) [86.9–91.3]
702 (88.1) [85.6–90.2]
27 (3.4) [2.349]
37 (4.6) [3.3–6.3]
19 (2.4) [1.5–3.7] 12
Failed attempted abortion
22 (2.8) [1.842]
15 (1.9) [1.1–3.1]
0 (0.0) [0.0–0.5]
7 (0.9) [0.4–1.8]
17 (2.1) [1.3–3.4]
24 (3.0) [1.944]
Table 3. Crude relative risks (RR) and adjusted odds ratios (OR) of failure to achieve complete abortion by treatment group and menstrual delay.
No. of women
Failure rate (%)
RR (95% CI)
OR (95% CI)*
*Adjusted for effect of centress.
† Total no. of women with information on delay = 1585.
Mifepristone dose (mg)
Menstrual delay (days)†
As a result of a secondary analysis after exclusion of women with unknown outcomes, those who had not taken misoprostol and those with protocol violations, the rates rose to 91.5% (701/766) with the 200 mg dose and 90.7% (693/764) with the 600 mg dose. The crude relative risk of failure with the lower dose was still 0.9 (95% CI 0.7–1.3). One-sided equivalence was shown within 3–4% difference or, correspondingly, within a relative risk of 1.25.
Gestational age strongly influenced the efficacy with both regimens, although this finding is not protected from bias by randomisation. With increasing delay in menses, both regimens became less effective. For both groups combined, the complete abortion rate declined from 92.2% at the earliest gestational ages to 80.3% at the latest. In terms of failure rates (Table 3), they increased from 7.8% to 19.7%, so that the likelihood of failing to abort was 2.5 times higher at four to five weeks' delay than at ≤ two weeks' delay. This trend was statistically significant (χ2 test for trends: χ2= 16.4, P < 0.01). The rate of continuing live pregnancies increased from 1.4% and 1.2% in the two earlier groups of menses delay (≤ 14 days and 15–21 days) to 2.7% in the 22–28 days' delay group to 9.0% in the 29–35 days' delay group (χ2 test for trends: χ2= 16.6, P < 0.01), respectively. After exclusion of participants without outcome information, the same pattern emerged, although the success rates were higher. The gestational age differences were more marked, with a 3.1-fold difference in failure rates between the latest and earliest gestational ages.
The median time to expulsion of the products of conception was 51 h after administration of mifepristone for both groups. This varied little by gestational age (data not shown). Expulsion of the products of conception occurred within three hours after misoprostol administration in 47.9% of those who received mifepristone 200 mg and 49.4% of those who received the 600 mg dose (χ2 test: χ2= 0.34, P = 0.56).
The adjusted analysis by exact methods or logistic regression confirmed the unadjusted results (Table 3). After adjustment for the effect of centres, the success rates were comparable for both groups [two-sided test by exact methods using StatXact: P = 0.50; likelihood ratio χ2 test from logistic regression: (χ2= 0.57, P = 0.45)]. The odds ratio of failure to abort was more than two-fold in the latest gestational age groups compared with the earliest. The effect of the centre on abortion rates was highly significant (likelihood ratio χ2 test from logistic regression: χ2= 59.9, P < 0.01), but age, weight, blood pressure, gravidity, and presence of complaints on enrolment were not significantly related to the abortion rate (likelihood ratio χ2 test from logistic regression: P > 0.20 for all variables).
Complaints occurred with similar frequency in both treatment groups (Table 4). The most common was abdominal pain, which was strongly related to the administration of misoprostol. Before misoprostol, about 40% of women in both groups had this complaint, and this frequency doubled after administration of the prostaglandin. Nausea was frequent in both groups, vomiting occurred in less than 20% of the women, and diarrhoea was uncommon. No significant differences emerged in any of the above mentioned side effects between the two treatment groups. However, there were five women needing blood transfusion in the 600 mg group and none in the 200 mg group (one-tailed Fisher's exact test: P = 0.03). The median duration of bleeding was the same in the two groups (11 days). Few women in either group required medication because of pain. After administration of misoprostol, 10.4% of those who received 200 mg mifepristone had medication, while the corresponding figure for the other group was 8.7% (χ2 test: χ2= 1.12, P = 0.29).
Table 4. Side-effects by treatment group. Values are given as n (%) [95% CI].
Dose of mifepristone
200 mg (n= 790)*
600 mg (n= 794)†
*2 women had missing information on side-effects.
†3 women had missing information on side-effects.
‡From χ2 test, significant at 5% with Bonferroni correction for multiple inferences if P <0.004.
After mifepristone and before misoprostol
441 (55.8) [52.3–59.3]
446 (56.2) [52.6–59.7]
420 (53.2) [49.6–56.7]
414 (52.1) [48.6–55.7]
At any time during treatment
531 (67.2) [63.8–70.5]
527 (66.4) [63.0–69.7]
After mifepristone and before misoprostol
132 (16.7) [14.2–19.5]
151 (19.0) [16.3–21.9]
162 (20.5) [17.7–23.5]
148 (18.6) [16.0–21.5]
At any time during treatment
219 (27.7) [24.6–31.0]
224 (28.2) [25.1–31.5]
After mifepristone and before misoprostol
26 (3.3) [2.2–4.8]
20 (2.5) [1.5–3.9]
68 (8.6) [6.7–10.8]
54 (6.8) [5.2–8.8]
At any time during treatment
81 (10.3) [8.2–12.6]
65 (8.2) [6.4–10.3]
Lower abdominal pain
After mifepristone and before misoprostol
306 (38.7) [35.3–42.2]
335 (42.2) [38.7–45–7]
654 (82.8) [80.0–85.4]
669 (84.3) [81.5–86.7]
At any time during treatment
670 (84.8) [82.1–87.2]
684 (86/1) [83.5–88.5]
Women with unclear outcome or lack of treatment compliance were infrequent (total of 12 women): three elected to have surgical abortions after receiving mifepristone and before receiving misoprostol. Three other women chose surgical abortion when they had not aborted three hours after having received misoprostol. The outcome of these six women was classified as undetermined. One woman who vomited four hours after mifepristone administration and never received misoprostol had a failed attempted abortion. Three who had complete abortions confirmed by ultrasonography at first follow up and then did not return were considered to have had a complete abortion. One woman considered to have had a complete abortion at first follow up, not confirmed by ultrasonography and without further follow up, was judged to be lost to follow up. One participant had an incomplete abortion at her first and only follow up, and she was also judged to be lost to follow up. Thirty-five women (2.2%) with no outcome information available and six with undetermined outcomes were considered lost to follow up (2.6%).
Mifepristone doses of 200 mg and 600 mg, when followed by 400 μg of oral misoprostol, have comparable abortifacient efficacy. This finding corroborates other trials1,3 which have compared single doses of mifepristone less than 600 mg followed by other prostaglandins. This may relate to the nonlinear pharmacokinetics of mifepristone when given in single doses larger than 100 mg5. Increasing the oral dose above 200 mg, which is the amount of the drug in one tablet, does not lead to higher serum levels of the drug or, as shown in this trial, to a better clinical efficacy.
Another important finding of this study, although exploratory in nature, is the effect of gestational age on efficacy. Several recent trials have indicated declining efficacy of mifepristone and oral misoprostol with advancing gestational age1,6. With a single 400 μg dose of oral misoprostol, duration of pregnancy strongly influenced efficacy, similarly with the two mifepristone doses. The efficacy of this oral regimen among women with a menstrual delay of more than 21 days was too low to justify the use of the regimen in such pregnancies.
More importantly, the continuing live pregnancy rate increased significantly with the menstrual delay. These exploratory findings are consistent with the results from a multicentre study carried out in the United States6 which stress the importance of accurate estimation of the length of gestation before treatment and the necessity of flawless follow up of the women. Recent articles referring to potential risks of uncontrolled use of the prostaglandin misoprostol7,8 also highlight the importance of appropriate service provision of medical abortion.
We believe this trial has internal validity. The randomisation scheme produced treatment groups comparable in all important respects. Allocation concealment, perhaps more important than randomisation in avoiding selection bias9, was maintained by using a centralised pharmacy. We had follow up information on 98% of participants. Losses to follow up were more frequent in the group that received the higher dose of mifepristone (24 vs 17 women), but since side effects and efficacy were similar in both groups, we doubt that the differential loss to follow up may have introduced bias. Double masking (blinding) throughout the study minimised ascertainment bias. Randomisation should have balanced both known and unknown confounding factors. Finally, these results are unlikely to be due to chance. The sample size was large enough to exclude an absolute difference in complete abortion rates as small as 3–4% between groups.
Moreover, we believe the trial has external validity as well. Since participants came from 17 centres on several continents, they represent the broad population of women who might choose nonsurgical abortion in early pregnancy. Participants were heterogeneous, with wide variations in baseline characteristics between centres. For example, women were younger in Havana and Hong Kong than in Stockholm or Yerevan. Women from Ljubljana and Melbourne weighed more than those from Ho Chi Min City or Mumbai. In Mumbai and Yerevan, fewer than 8% of women had not been pregnant before, while this figure exceeded 70% in Hong Kong and Szeged.
While associated with fewer side effects than alternative prostaglandins, 400 μg oral misoprostol appears less effective as an augmenting agent than vaginal gemeprost 1 mg. The overall complete abortion rate in this trial (89%; 95% CI 87%–90%) was lower than that achieved in our earlier study (94%; 95% CI 91%–96%) with the same doses of mifepristone followed by gemeprost 1 mg administered vaginally3. This may reflect a weaker effect of oral misoprostol on uterine contractility than gemeprost, a suboptimal dose of misoprostol, the different route of administration, or some combination of these.
If a nonsurgical abortion regimen is used in more advanced first trimester pregnancies, it requires a potent prostaglandin. At 22–35 days of menstrual delay, the use of 400 μg of oral misoprostol proved inadequate. Alternatives at this gestational age might include use of a different prostaglandin, a larger oral dose of misoprostol, or vaginal administration of misoprostol. Several recent studies suggest that vaginal administration of misoprostol to augment mifepristone abortions would be more effective than oral administration of the same dose10,11. A recent study comparing uterine contractility after oral and vaginal administration of misoprostol12 demonstrated a longer lasting and continuously increasing uterine contractility following vaginal administration in addition to a longer period of elevated plasma levels of the drug.
The authors wish to thank Roussel-Uclaf, Paris, France, for providing mifepristone tablets for the study and packaging them for the study centres.
Task Force on Post-ovulatory Methods of Fertility Regulation (World Health Organization)
Principal Investigators and Centres:
Dr Y. M. Wu (Peking Union Medical College Hospital, Beijing, People's Republic of China), Dr M. Gomez-Alzugaray (National Institute of Endocrinology, Havana, Cuba), Dr M. Haukkamaa (City Maternity Hospital and Helsinki University Central Hospital, Helsinki, Finland), Dr N. T. N. Ngoc (Hung Vuong Hospital, Ho Chi Minh City, Viet Nam), Dr P. C. Ho (University of Hong Kong, Hong Kong, Special Administrative Region of China), Dr A. Pretnar-Darovec (University Department of Obstetrics and Gynaecology, Ljubljana, Slovenia), Dr D. L. Healy (Monash University, Melbourne, Australia), Dr E. Sotnikova (Research Centre of Obstetrics, Gynaecology and Perinatology, Academy of Medical Sciences of Russia, Moscow, Russian Federation), Dr R. S. Shah (Institute for Research in Reproduction, Indian Council of Medical Research, Mumbai, India), Dr N. G. Pavlova (Institute of Obstetrics and Gynaecology named after D. O. Ott, Russian Academy of Medical Sciences, St Petersburg, Russian Federation), Dr J. K. Chen and Dr S. Song (Shanghai Institute of Planned Parenthood Research, Shanghai, People's Republic of China), Dr M. Bygdeman (Karolinska Hospital, Stockholm, Sweden), Dr L. Kovács (Albert Szent-Györgyi Medical University, Szeged, Hungary), Dr A. Khomassuridze (Zhordania Institute of Human Reproduction, Tbilissi, Georgia), Dr L. J. Song (Tianjin Municipal Research Institute for Family Planning, Tianjin, People's Republic of China), Dr R. Hamzaoui (Centre d'Etude et de Recherche en Reproduction Humaine de l'ONFP, Tunis, Tunisia), Dr S. Alexaniants (Armenian Research Centre of Maternal and Child Health Protection, Yerevan, Armenia).
Dr G. Piaggio*, Mr A. Peregoudov*, Ms M. Vucurevic*.
Manuscript prepared by:
Dr D. Grimes†, Dr H. von Hertzen*, Dr G. Piaggio*, Dr P. F. A. Van Look*
Statement of responsibility and of conflict of interest
Overall responsibility for this paper lies with Dr H. von Hertzen and Dr P. F. A. Van Look of the UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, World Health Organisation. Neither they nor the Special Programme or any of its cosponsors have a conflict of interest with respect to mifepristone in general and the outcome of this trial in particular.