Bulletin of the World Health Organization

Misoprostol to prevent and treat postpartum haemorrhage: a systematic review and meta-analysis of maternal deaths and dose-related effects

G Justus Hofmeyr a, A Metin Gülmezoglu b, Natalia Novikova b, Verena Linder b, Sandra Ferreira b & Gilda Piaggio a

a. East London Hospital Complex and University of the Witwatersrand, PB X9047, East London 5201, South Africa.
b. Development and Research Training in Human Reproduction, World Health Organization, Geneva, Switzerland.

Correspondence to GJ Hofmeyr (e-mail: gjh@global.co.za).

(Submitted: 10 June 2008 – Revised version received: 15 November 2008 – Accepted: 25 November 2008 – Published online: 20 July 2009.)

Bulletin of the World Health Organization 2009;87:666-677. doi: 10.2471/BLT.08.055715

Introduction

In low-income countries, postpartum haemorrhage is a major cause of maternal death1 and arguably the most preventable. Attempts to reduce deaths from postpartum haemorrhage have been complicated by the fact that many deaths occur in out-of-hospital settings or too quickly for the patient to be transferred to a health facility. Furthermore, prevention and treatment have depended primarily on injectable uterotonics, which are seldom available for births outside the health system. For these reasons, the use of misoprostol to prevent or treat postpartum haemorrhage has attracted considerable attention. Misoprostol, an inexpensive and stable prostaglandin E1 analogue, has been shown to stimulate uterine contractility in early pregnancy2 and at term.3 Administered orally or vaginally, it is effective for inducing abortion4,5 and labour,68 though it poses certain risks.9,10

Administration of this drug on a wide scale at the community level to prevent and treat postpartum haemorrhage is of major public health importance. The first of many randomized trials of the use of misoprostol in the third stage of labour was conducted in 1995.11 When administered in various doses and via various routes for the prevention of postpartum haemorrhage, misoprostol has been less effective than conventional injectable uterotonics.12,13 Initial reports of placebo-controlled trials have shown variable results,12 but in more recent studies misoprostol has proved better than placebo, in terms of measured blood loss, for both the prevention14,15 and treatment16 of postpartum haemorrhage. The main side effects reported have been chills and pyrexia, both of which have been dose-dependent.17

We have reviewed the pharmacological, physiological and clinical evidence surrounding the use of misoprostol for the treatment of postpartum haemorrhage.16 The oral route of administration is the fastest but also the one associated with the shortest duration of action. The rectal route has slow uptake but a prolonged duration of action. The buccal and sublingual routes have rapid intake, prolonged duration of action and the greatest total bioavailability. We concluded from the data reviewed that the most promising route of administration was the sublingual route.

Misoprostol has been widely recommended to prevent postpartum haemorrhage when other methods are not available,18 and large-scale implementation projects are currently in progress. Caution has been urged, however, because side-effects may occasionally be life-threatening.19 In a recent placebo-controlled trial in rural India in which babies were delivered by auxiliary nurse midwives at home or in village subcentres, a significant reduction in postpartum haemorrhage and other complications was obtained with misoprostol, 600 µg orally used alone (i.e. without other components of the active management of the third stage of labour – umbilical cord clamping and controlled cord traction).15 WHO has developed guidelines supporting the use of a uterotonic when the full package of active management of the third stage of labour is not practised, which can be either oxytocin, 10 IU administered parenterally, or misoprostol, 600 µg administered orally.20

Apart from its uterotonic effects, misoprostol has known pharmacologic effects on several organ systems. It inhibits platelet-activating factor21 and affects metabolic and physiological processes, including thermoregulation. Life-threatening hyperpyrexia has been reported following the use of misoprostol, 800 µg orally, after childbirth.22

Whenever a new medical intervention is promoted, one must take into account not only its clinical benefits and risks, but also the collateral benefits and risks from both use and potential misuse in the community. Before distributing misoprostol at the community level, it is important to consider its possible misuse, an example being its administration before delivery, which could lead to uterine rupture. Because of its enormous potential benefits as an effective, oral uterotonic during the third stage of labour and its likely use on a large scale worldwide, it is important to monitor misoprostol’s benefits as well as its potential risks, both direct and indirect.

In September 2005, the BMJ published the first placebo-controlled trial to suggest a clear effect of misoprostol for preventing postpartum haemorrhage.14 The study elicited several responses, however. One of them was the suggestion to regard the single maternal death in the misoprostol group as an adverse event potentially related to misoprostol use.23 This is plausible because a drug with ubiquitous pharmacologic effects might have unanticipated adverse effects when used in the third stage of labour. Another response was a call for a trial of misoprostol as a treatment for postpartum haemorrhage, with maternal death as the primary endpoint.24

In this paper we hypothesized that women on misoprostol in the third stage of labour were at increased risk of death or severe morbidity because of an as yet unexplained adverse effect of the drug on maternal homeostatic functions.25 Thus, the primary objectives of this review was to investigate maternal deaths and severe morbidity in connection with the use of misoprostol for preventing or treating postpartum haemorrhage during the third stage of labour, as well as to see whether side effects were dose-related. A secondary objective was to determine the relative effectiveness of ≥ 600 µg of misoprostol versus smaller doses, since to justify the use of the larger dose in large numbers of women, particularly in light of potential side effects, would require evidence that it is considerably more effective than the smaller dose.

Methods

Systematic review

Types of studies: We performed a systematic review of randomized controlled trials in which misoprostol versus either placebo or another uterotonic were given to women after delivery to prevent or treat postpartum haemorrhage. We looked for the following primary outcome measures: (i) maternal death, or (ii) maternal death or severe morbidity, defined as either major surgery, admission to the intensive care unit (ICU), vital organ failure or body temperature ≥ 40 °C. For dose comparisons, the outcomes selected were a blood loss of ≥ 1000 ml or of ≥ 500 ml after the diagnosis of postpartum haemorrhage, and pyrexia (i.e. a body temperature ≥ 38 °C or as defined by trial authors).

Search strategy

We followed the search strategy used by the Cochrane Collaboration’s Pregnancy and Childbirth Group. Briefly, potentially eligible trials were identified from: (i) quarterly searches of the Cochrane Central Register of Controlled Trials (CENTRAL); (ii) monthly searches of Medline; (iii) hand searches of 30 journals and the proceedings of major conferences; and (iv) weekly current awareness searches of a further 37 journals. We also searched the Cochrane Controlled Trials Register and Pubmed, without language restrictions, for “(misoprostol AND postpartum) OR (misoprostol AND haemorrhage) OR (misoprostol AND hemorrhage)” (last search February 2007). We tried to contact authors to obtain further details about any maternal deaths reported and, if no maternal death was mentioned in a paper, we asked the authors to confirm that no death had occurred.

All studies identified through the search strategy were assessed for inclusion in the analysis. We designed a data extraction form that was used by two of the authors, and any disagreements were resolved through discussion. We used Review Manager (RevMan) 4 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) to enter the data and check it for accuracy.

Assessment of methodological quality

For each trial we classified allocation concealment as follows: (i) adequate if a method such as telephone randomization or consecutively numbered sealed opaque envelopes was used; (ii) unclear if the concealment method was not described; or (iii) inadequate if an open list of random numbers, case record numbers, dates of birth, days of the week or another such method was used.

Measures of treatment effect and assessment of heterogeneity

We used fixed effects meta-analysis for combining data in the absence of significant heterogeneity. For dichotomous data, we presented the results as summary relative risks (RRs) with 95% confidence intervals (CIs). For the outcome maternal mortality, we used both RRs and Peto odds ratios (ORs) to check for discrepancies between the two summary statistics. Peto ORs are recommended when events are very infrequent. We applied tests of heterogeneity between trials using the I² statistic, a variation of the χ² statistic.

Subgroup analyses

We classified whole trials according to interaction tests as described by Deeks et al.26 and carried out the following subgroup analyses: (i) misoprostol versus placebo and versus other uterotonics; (ii) misoprostol to prevent or treat postpartum haemorrhage; and (iii) misoprostol at a dose of ≥ 600 µg or more; 400 to < 600 µg; or < 400 µg.

In the analysis tables we ranked the trials by route of administration and dose of misoprostol to check for route- or dose-related trends in effects. We coded the trial identifiers to indicate the route of administration and dose used.

Comparisons of dose-related effectiveness and side-effects

In view of inadequate data from direct randomized comparisons of different doses of misoprostol, we supplemented the direct data with adjusted indirect comparisons. In most cases, the results of adjusted indirect comparisons have proved similar to those of direct comparisons in 44 previous meta-analyses.27 Thus, to estimate the relative effectiveness of two different doses of misoprostol we combined the results of trials in which each dose of the drug was tested against the same type of control (either placebo or a uterotonic other than misoprostol).28 Three-way comparisons of 600 µg versus 400 µg of misoprostol versus placebo or another uterotonic were included only in the direct comparisons. Studies of rectally-administered misoprostol were excluded because of uncertain absorption via the rectal route.

For direct and indirect comparisons the primary outcome measures were blood loss ≥ 1000 ml and pyrexia ≥ 38 °C. As the effectiveness outcome was blood loss measured within 1 hour of drug administration, doses of misoprostol given 1 hour or more after the initial dose were not taken into account.

Most of the trials in the review were conducted in settings with routine active management of the third stage of labour. To determine if such management affected the results, we performed a sensitivity analysis from which we excluded the one trial in which active management was not practiced.15

Prospective adverse-event surveillance studies

We report on adverse events from misoprostol use as recorded in the WHO Adverse Reaction Database, maintained by the Uppsala Monitoring Centre (UMC), which is the WHO Collaborating Centre for International Drug Monitoring. The UMC collects reports of suspected adverse drug reactions from 70 participating countries. We ranked adverse events by frequency of reporting.

Results

Systematic review

We found 46 randomized controlled trials with more than 40 000 participants. They are summarily described in Table 1.

Maternal deaths

Maternal deaths were reported in five trials.14,15,38,58,69 Three other trials specified that there were no maternal deaths.30,39,71 No additional maternal deaths were identified through communication with the authors of the included trials.

Fig. 1 (available at: http://www.who.int/bulletin/volumes/87/9/08-055715/en/index.html) shows only the five trials in which maternal deaths were reported. Of 11 deaths reported in these trials,14,15,38,58,69 8 occurred in women receiving misoprostol (RR: 2.0; 95% CI: 0.68–5.83; Peto OR: 2.49; 95% CI: 0.76–8.13). Of 8 maternal deaths reported in the misoprostol group, 6 were associated with postpartum haemorrhage.14,38,58

Fig. 1. Maternal deathsa reported in randomized trials of misoprostolb versus placebo or another uterotonic to prevent or treat PPH
Severe maternal morbidity

Fig. 2 (available at: http://www.who.int/bulletin/volumes/87/9/08-055715/en/index.html) shows the data on maternal deaths and severe morbidity as reported in the randomized controlled trials of misoprostol compared with placebo or other uterotonics to prevent or treat postpartum haemorrhage. When misoprostol was compared with other uterotonics, a similar number of adverse events was reported in both prevention and treatment trials: in prevention trials, 16 of 10281 versus 16 of 10292 women, respectively (RR: 1.0; 95% CI: 0.51–1.96); in treatment trials, 1 of 32 women in the treatment as well as the control group (RR: 1.0; 95% CI: 0.07–15.30). When misoprostol was compared with placebo, slightly more adverse events were reported in the misoprostol group in both prevention and treatment trials: in prevention trials, 8 of 2070 versus 5 of 2032 women, respectively (RR: 1.46; 95% CI: 0.52–4.09); in treatment trials, 5 of 196 versus 2 of 202 women, respectively (RR: 2.06; 95% CI: 0.52–8.17). Events were too few to draw any conclusions.

Fig. 2. Randomized trialsa of misoprostolb versus placebo or another utrotonic to prevent or treat PPH

Table 2 presents descriptions of cases of maternal death and morbidity reported in the trials. In a study from India,15 one maternal death in the placebo group was not related to haemorrhage. In the WHO (2001) study,69 the causes of two deaths in the injectable uterotonic group were not mentioned. A few studies reported admission of the participants to the ICU. One of them was the WHO (2001) study, in which 9 women were admitted to the ICU: 4 of 9224 in the misoprostol group and 5 of 9231 in the injectable uterotonic group (RR: 0.80; 95% CI: 0.22–2.98).69 In a recent study from India, 4 participants were admitted to the ICU: 2 of 812 in the misoprostol group and 2 of 808 in the placebo group.15

Blood loss  1000 ml

In the initial analysis of misoprostol versus placebo (Fig. 3, available at: http://www.who.int/bulletin/volumes/87/9/08-055715/en/index.html), significant heterogeneity (I²: > 50%) was noted in the results for 600 µg (4914 women; RR: 0.92; 95% CI: 0.54–1.57; I²: 66%) and for 400 µg (3039 women; RR: 0.80; 95% CI: 0.47–1.37, random effects model; I²: 58%). The heterogeneity could not be accounted for by route of administration or trial quality. Rather, it appeared to be related to a single outlier study that contributed data to both subgroups. Removal of this study from the meta-analysis eliminated the significant heterogeneity (I²: < 50%). After exclusion of the outlier, less blood loss ≥ 1000 ml was noted with 600 µg of misoprostol (4514 women; RR: 0.77; 95% CI: 0.59–1.00, trend, fixed effects model) and with 400 µg of misoprostol (2642 women; RR: 0.63; 95% CI: 0.44–0.91) than with placebo.

Fig. 3. Randomized trials of misoprostola versus placebo to prevent PPH, effect on blood loss ≥ 1000 ml

In one small trial in which 200 µg of misoprostol were compared with placebo (in addition to an oxytocin infusion in both groups) at Caesarean section, no statistically significant difference in blood loss ≥ 1000 ml was found (misoprostol versus placebo: 352 women; RR: 1.13; 95% CI: 0.66–1.94).

When misoprostol was compared with other uterotonics (Fig. 4, available at: http://www.who.int/bulletin/volumes/87/9/08-055715/en/index.html), results showed no heterogeneity (I²: 0%). Misoprostol, at a dose of 600 to 800 µg, was less effective than other uterotonics (22 827 women; RR: 1.36; 95% CI: 1.17–1.58). Furthermore, the results obtained with 400–500 µg of misoprostol did not differ significantly from those obtained with 600 µg (9772 women; RR: 1.23; 95% CI: 0.96–1.58), even though the sample size was smaller than for the 600–800 µg doses, the direction of the trend was the same.

Fig. 4. Randomized trials of misoprostola versus another uterotonic to prevent PPH, effect on blood loss ≥ 1000 ml

In one small trial in which misoprostol at a dose of 50 µg administered sublingually was compared with injectable uterotonics, no significant difference was found (75 women; RR: 0.33; 95% CI: 0.04–2.62).

When given in addition to routine methods of treatment (Fig. 5, available at: http://www.who.int/bulletin/volumes/87/9/08-055715/en/index.html), 600–1000 µg of misoprostol in split doses orally, sublingually and/or rectally was more effective than placebo at reducing an additional blood loss ≥ 500 ml after postpartum haemorrhage (397 women; RR: 0.57; 95% CI: 0.34–0.96).

Fig. 5. Randomized trials of misoprostola versus placebo to treat PPH, effect on additional blood loss ≥ 500 ml following postpartum haemorrhage

Meta-analysis of direct and adjusted indirect data from randomized trials showed no evidence of a benefit of 600 µg over 400 µg of misoprostol for reducing blood loss ≥ 1000 ml (RR: 1.02; 95% CI: 0.71–1.48; Table 3 and Fig. 6, available at: http://www.who.int/bulletin/volumes/87/9/08-055715/en/index.html). A sensitivity analysis excluding the one trial in which active management was not practised15 produced similar results (data not shown).

Fig. 6. Randomized trials of misoprostola versus placebo or another uterotonic to prevent or treat PPH, effect on pyrexia (body temperature [t] ≥ 38 ºC or as defined by trial authorsb)
Pyrexia

Pyrexia (body temperature ≥ 38 °C or as defined by authors) was higher among women who were given misoprostol in all subgroups (Fig. 7). There was quantitative but not qualitative heterogeneity in the subgroup in which misoprostol, at a dose of 400–500 µg, was compared with other uterotonics (I² = 65%). Such heterogeneity could not be explained by route of administration or trial quality.

Fig. 7. Meta-analysisa of direct comparisons (misoprostol, 600 µg vs 400 µg) and adjusted indirect comparisons (misoprostol, 600 µg or 400 µg versus another uterotonic or placebo) for risk of PPH

In comparisons with both placebo and other uterotonics, 600 µg of misoprostol produced pyrexia more often than 400–500 µg: misoprostol, 600 µg versus placebo: 3685 women; RR: 6.71; 95% CI: 4.83–9.32; misoprostol, 400 µg versus placebo: 1996 women; RR: 3.90; 95% CI: 2.27–6,69; misoprostol, 600 µg versus other uterotonics: 22 337 women; RR: 6.76; 95% CI: 5.54–8.25; misoprostol, 400–500 µg versus other uterotonics: 8135 women; RR: 3.05; 95% CI: 2.45–3.81. In the two treatment trials, pyrexia was more common in the misoprostol group (392 women; RR: 2.78; 95% CI: 1.39–5.53).

In meta-analysis of direct and adjusted indirect data from randomized trials, pyrexia was more common in the group given 600 µg of misoprostol rather than 400 µg (RR: 2.53; 95% CI: 1.78–3.60; Table 4 and Fig. 8). There was consistency between the estimates derived from the direct and the adjusted indirect comparisons.

Fig. 8. Meta-analysisa of direct comparisons (misoprostol, 600 µg vs 400 µg) and adjusted indirect comparisons (misoprostol, 600 µg or 400 µg versus another uterotonic or placebo) for risk of pyrexia (body temperature [t] ≥ 38 ºC or as defined by trial authors)

Prospective surveillance of adverse events

The 14 most frequent adverse events associated with misoprostol use in the WHO Adverse Reaction Database were as follows (an association does not imply causality): diarrhoea (1001), abdominal pain (687), nausea (394), haemorrhage (294), abortion (209), vomiting (209), dyspepsia (162), flatulence (154), dizziness (152), menorrhagia (125), vaginal haemorrhage (153) and fever (98). Death (53) was the 23rd most common adverse event. Without details of the clinical situations in which these complications occurred, it is not possible to draw conclusions from the data.

Discussion

Misoprostol is used in the third stage of labour primarily to prevent maternal death. Because this is such a rare outcome, no trials to date have been designed to assess the effect on mortality. However, the occurrence of several deaths in misoprostol trials has raised some concern.

It is difficult to evaluate rare adverse outcomes in randomized trials. However, the systematic review of misoprostol trials included in this study provides data on more than 40 000 women. In addition, we complemented our search with adverse events reported to an international centre for such events. To our knowledge, there are no other large population-based datasets or observational studies of maternal deaths or severe adverse outcomes potentially associated with misoprostol use.

Our systematic review of maternal mortality is limited, first, by the small number of deaths associated with misoprostol use and the wide 95% confidence intervals, which are consistent with anything from a modest reduction to a large increase in maternal deaths. Second, this was a post-hoc analysis in response to an observed clustering of maternal deaths, and the statistical calculations need to be interpreted with caution. We recommend prospective surveillance of all randomized trials for further evidence for or against our working hypothesis.

The idea that a drug may have proven benefits against a life-threatening condition yet increase mortality is counterintuitive, but such a thing is plausible. For example, class 1 antiarrhythmics were routinely administered to patients after myocardial infarction because of their proven suppression of ventricular premature depolarizations, a common cause of death after infarction. However, in a double-blind, placebo-controlled trial in 3549 patients who had suffered myocardial infarction and had left ventricular dysfunction, 1 year after randomization to blinded therapy mortality among drug-treated patients was 10%, compared with 5% in the placebo-treated group (P = 0.0006).72 Monitoring for unexpected adverse effects is critical before any new medical intervention is introduced.

The cause of death was given for some of the deaths reported in the trials. Some deaths were due to severe haemorrhage, while others did not seem to be directly related to haemorrhage. In addition, deaths were too few to allow for any meaningful interpretation of the role misoprostol or other therapeutic interventions may have played in them.

Although they do not prove our working hypothesis, these results do suggest that misoprostol may have as yet unexplained adverse effects on maternal homeostatic functions in the third stage of labour. The potential for serious side effects is particularly worrisome in connection with the use of misoprostol for preventing, rather than treating, postpartum haemorrhage, because to prevent a single case of postpartum haemorrhage one would need to treat very large numbers of women.

In estimating the relative effect of 600 µg versus 400 µg of misoprostol on postpartum haemorrhage, we used meta-analysis of direct and adjusted indirect comparisons of randomized trial data to improve the precision of the findings from direct comparisons alone. This method proved useful in an empirical study of 44 meta-analyses.27 We further validated the method by applying it to the same trials for the outcome pyrexia, for which direct data were reasonably precise because pyrexia was more common than haemorrhage. We found that the findings from adjusted indirect comparisons were consistent with those of direct comparisons.

In all studies in which maternal deaths were reported, misoprostol was given at a dose of ≥ 600 µg. Physiological studies have shown uterotonic effects with doses as low as 200 µg. Our direct and indirect comparisons of data from randomized trials showed no difference in effect size between 600 µg and 400 µg doses. Since misoprostol sometimes produces dose-related side effects such as pyrexia and chills, we suggest that research be undertaken to establish the smallest dose of the drug that is effective and safe.

Because of the enormous benefits of using an effective oral uterotonic in the third stage of labour and since misoprostol could be used in this manner on a large scale worldwide, further research is essential to better assess the potential beneficial and harmful effects of the drug. However, the use of misoprostol should not detract from international efforts to ensure that all childbearing women have access to conventional uterotonics that have been proven safe and effective. ■


Acknowledgements

We thank ME Stanton and PFA Van Look for their continuing support, and PFA Van Look for reviewing the manuscript. The Effective Care Research Unit, East London, South Africa, and the Reproductive Health and Research Department at WHO, in Geneva, Switzerland, conducted the systematic review.

Funding: The project was funded by the United States Agency for International Development.

Competing interests: None declared.

References

Share