State of the Literature

We included 68 unique studies (76 publications) in this review, including four randomized controlled trials (RCTs), two prospective and 14 retrospective cohort studies, 10 pre-post studies (defined as studies that compare PPH management and/or outcomes before and after an intervention, such as introduction of a new protocol), four case-control studies, and 34 case series. Most studies were conducted in Europe (n = 33), and 18 were conducted in the United States or Canada, 13 in Asia, and three in Australia or New Zealand and one in Argentina (Table 5). No studies were of good quality for effectiveness outcomes. We considered 23 studies as fair quality for effectiveness outcomes and 38 as poor (including case series, which we considered poor quality by default). Seven studies provided only harms data. Among the 50 studies reporting harms, we considered 11 as good quality for harms reporting and the remainder as poor quality.

While a number of studies were classified as prospective or retrospective studies using our study classification algorithm (Appendix G), few cohort studies provided comparative analyses between the groups, and many were confounded by indication in that women who received interventions such as massive transfusion or hysterectomy likely had more severe cases of postpartum hemorrhage (PPH). Given the lack of data from randomized or controlled studies of PPH management, we present data from cohort studies and case series and note potential confounding as appropriate.

Overall, it appears that 50 deaths occurred in the included studies addressing non-systems level interventions out of roughly 152,264 participants (note that 139617 of these participants were included in a large database study reporting harms following methylergonovine maleate given in the peripartum hospitalization⁶⁰). Only one death was potentially linked to PPH management: a woman who was given methotrexate in conjunction with embolization died from methotrexate-related nephrotoxicity.⁹⁵ The remaining deaths appear to be the result of PPH and its sequelae rather than interventions used for management.

Summary of Key Findings

Findings are summarized below by Key Question (KQ).

Strength of the Evidence

Overall the evidence to answer questions about PPH management did not reach standards for high strength of evidence. The strength of evidence (SOE) tables summarize the total number of studies and the number of participants within those studies noting the study designs and quality (Tables 26-32). The tables also provide the assessment of the study limitations, consistency of findings across studies, directness of the evidence, precision of the estimate, and presence of reporting bias. We included case series in our assessment of SOE for harms and success rates of interventions, and we rated SOE for outcomes we considered to be clinically significant, consistently defined, and plausibly linked to the intervention.

SOE is insufficient for all outcomes of oxytocin and other uterotonics, misoprostol, tranexamic acid, carboprost tromethamine, thrombomodulin, and rFVIIa for PPH management due to the study sizes and lack of studies addressing each agent (Table 26). As noted, we identified few studies of medications meeting our review criteria; however, a number of studies of misoprostol and oxytocin have been conducted in developing countries. Four recent systematic reviews of interventions for PPH, including two Cochrane reviews, assessed uterotonics including misoprostol. We summarize these reviews fully in the Findings in Relation to What is Known section below and provide a brief summary here. In one Cochrane review, oxytocin infusion was more effective and caused fewer side effects when used as first-line therapy for the treatment of primary PPH compared with misoprostol.¹³³ When used after prophylactic uterotonics, misoprostol and oxytocin infusion had similar effects. The review concluded that adding misoprostol for women receiving treatment with oxytocin did not appear beneficial. In another Cochrane review differences in maternal mortality and morbidity, except for fever, did not differ significantly between misoprostol and control groups.¹³⁴ The investigators concluded that misoprostol did not increase or decrease morbidity or mortality, with the exception of fever, and the lowest effective dose should be used. In another review of misoprostol vs. placebo, misoprostol did not reduce PPH risk significantly compared with placebo.¹³⁵ In the fourth review and meta-analysis, higher doses of misoprostol (600 vs. 400 micrograms) were no more effective at preventing blood loss.⁵⁰

Table 26

Strength of the evidence for studies addressing medications.

The SOE for outcomes related to transfusion and uterine balloon tamponade is insufficient (Table 27). While there were three fair quality studies of transfusion, two of these were so confounded that we could not confidently ascertain their outcomes. There is low SOE for embolization controlling bleeding without additional procedures or surgeries.

Table 27

Strength of the evidence for studies addressing other medical interventions and procedures.

There is insufficient SOE for the success of uterine compression sutures (Table 28). There is low SOE for ligation controlling bleeding without further procedures or surgeries and insufficient SOE for all hysterectomy outcomes.

Table 28

Strength of the evidence for studies of surgical interventions.

Table 29 outlines the SOE for studies of combination interventions. Two studies assessed length of stay; however, we considered the SOE for the effect of intervention to be insufficient given the small sample sizes and inconsistency in interventions.

Table 29

Strength of the evidence for studies of combination interventions.

The SOE for harms of interventions for management of PPH can be found in Table 30. Generally SOE was insufficient given diversity of harms reported in single studies. However, SOE rose above insufficient for selected harms related to embolization and hysterectomy due to the greater number of studies and more consistent reporting of adverse events. As noted, few studies of uterotonics met our inclusion criteria; however, harms reported in recent systematic reviews of uterotonics for PPH treatment included shivering and fever (see Findings in Relation to What's Known section for full summary). In one review, oral misoprostol was associated with a significant increase in vomiting and shivering compared with either oxytocin or rectal misoprostol.¹³³ In another review, differences in maternal mortality and morbidity, except for fever, did not differ significantly between misoprostol and control groups.¹³⁴ Risk of fever was increased in misoprostol groups and was highest in studies with a misoprostol dose of 600 μg or more. In another review of misoprostol vs. placebo, shivering and fever were significantly more common in misoprostol arms.¹³⁵ A fourth review noted more adverse effects related to misoprostol vs. placebo.⁵⁰

Table 30

Strength of the evidence for harms of interventions for management of PPH.

While evidence in the current review was insufficient to comment on the association between rFVIIa and thrombolic events, studies in other populations have suggested increased risk of arterial events. In one review of RCTs in non-hemophilia patients, the pooled relative risk of thrombolic events across studies of prophylactic and therapeutic uses of rFVIIa was 1.45 (95% CI: 1.02 to 2.05).¹³⁶ Another review of fertility outcomes following embolization, ligation, and sutures concluded that the techniques reviewed did not appear to compromise fertility, but the number and quality of studies was limited.¹³⁷

SOE is insufficient for all outcomes and harms in studies of interventions for anemia after PPH given the few studies, small number of participants, and differences in intervention approaches (Table 31).

Table 31

Strength of the evidence for interventions for anemia after PPH.

Overall the SOE for any systems-level intervention on any outcome is insufficient or moderate as the observational data are biased and a single, very large trial suggest that at least one clearly described and implemented program did not change risk of severe hemorrhage or meaningfully modify processes of care or overall maternal outcomes (Table 32). SOE is moderate that these multicomponent interventions did not change specific outcomes such as severity of PPH, transfusion, hysterectomy, and ICU admission.

Table 32

Strength of the evidence for studies addressing multicomponent, systems-level interventions.

Findings in Relation to What Is Already Known

Findings in recent (2009-present) systematic reviews and meta-analyses of interventions to manage PPH are largely in line with findings reported here in that while reviews reported some positive effects, studies included in the reviews typically had significant limitations that precluded firm conclusions. Reviewers noted a lack of high quality literature, small sample sizes, limited followup, and a preponderance of observational studies of procedures or surgical approaches given the urgent nature of PPH. We summarize findings of reviews of pharmacologic studies conducted in developing nations as the current review contains few comparable studies of pharmacologic agents. We also summarize recent reviews of procedures and surgical approaches.

Few drug studies met our inclusion criteria, which specified studies must be conducted in the high-resource countries where care would be applicable to that in the United States. Four recent reviews, however, have addressed uterotonics, primarily in lower resource settings. Overall, these reviews had conflicting findings about the effectiveness of misoprostol; however, this medication was consistently associated with adverse effects, particularly fever and shivering.

One 2014 Cochrane review assessed the effectiveness and safety of any intervention used for the treatment of primary PPH.¹³³ The uterotonic interventions included in the search strategy (search dates: up to August 2013) were ergonovine, oxytocin, and prostaglandin medications. Seven RCTs evaluated misoprostol. Four RCTs (1,881 participants) compared misoprostol with placebo given in addition to other conventional uterotonics. Adjunctive use of misoprostol (600-1000 micrograms) with simultaneous administration of other uterotonics did not provide additional benefit for maternal mortality, serious maternal morbidity, admission to intensive care, or hysterectomy. Three RCTs (1,851 participants) compared oral misoprostol with oxytocin infusion (n=2 RCTs) or rectal misoprostol (n=1 RCT) as primary PPH treatment. Primary outcomes including maternal mortality, hysterectomy, ICU admission, and serious maternal morbidity did not differ between the groups. Oral misoprostol, however, was associated with a significant increase in vomiting and shivering compared with either oxytocin or rectal misoprostol. No RCTs of ergonovine or carboprost tromethamine met the inclusion criteria. The investigators concluded that, overall, the clinical trials included in the review were not adequately powered to assess impact on the primary outcome measures. Compared with misoprostol, oxytocin infusion was more effective and caused fewer side effects when used as first-line therapy for the treatment of primary PPH. When used after prophylactic uterotonics, misoprostol and oxytocin infusion had similar effects. Adding misoprostol for women receiving treatment with oxytocin does not appear beneficial.

A 2013 Cochrane review (search dates: up to January 2013) assessed maternal deaths in studies of misoprostol for prevention and treatment of PPH and included 78 RCTs reporting on 59,216 women; only seven of these studies focused on treatment vs. prevention, and most studies were conducted in low-resource countries.¹³⁴ Overall, differences in maternal mortality and morbidity, except for fever, did not differ significantly between misoprostol and control groups. Risk of fever was increased in misoprostol groups and was highest in studies with a misoprostol dose of 600 μg or more. The investigators concluded that misoprostol does not increase or decrease morbidity or mortality, with the exception of fever, and the lowest effective dose should be used.

In another review (search dates: not specified) including three RCTs (2,346 participants) of misoprostol vs. placebo, misoprostol did not reduce PPH risk significantly compared with placebo, and shivering and fever were significantly more common in misoprostol arms.¹³⁵A review of maternal deaths and dose-related effects of misoprostol included 46 trials with more than 40,000 participants. The investigators found more adverse effects related to misoprostol than placebo and no evidence, in a meta-analysis, that higher doses of misoprostol (600 vs. 400 micrograms) were more effective at preventing blood loss. Fever was higher among women given misoprostol and occurred more frequently with higher doses (600 vs. 400-500 micrograms)⁵⁰

One review (search dates: not specified) evaluating uterine tamponade in resource-poor settings included 13 observational studies and reported successful treatment of PPH in 234 of 241 women.⁴¹Most women had oxytocin and ergometrine or other medications prior to tamponade, and the tamponade device varied among studies. Another systematic review (search dates: 1950-2012) assessed menstrual and fertility outcomes after uterine-sparing interventions for PPH.¹³⁷ Studies included in the review addressed embolization (n = 17), ligation (n = 5), and compression sutures (n = 6). Overall, 183 of 235 women who desired another pregnancy were able to conceive, and 553 of 606 resumed normal menstruation within 6 months of birth. Within each intervention type, most women who wanted to conceive were able to do so: 86 percent (24/28) of women who had sutures (21 total term live births, 0 preterm births, pregnancy losses, or cases of recurrent PPH), 85 percent (33/39) of women who had ligation (68 total term live births, 1 preterm birth, 23 pregnancy losses, 8 cases of recurrent PPH), and 75 percent (126/168) of women who had embolization (136 total term live births, 4 preterm births, 30 pregnancy losses, 18 cases of recurrent PPH). The investigators conclude that the techniques reviewed do not appear to compromise fertility, but the number and quality of studies was limited.

One review (search dates: up to August 2009) evaluated emergency postpartum hysterectomy for PPH performed within 48 hours of birth and included 24 studies reporting on 981 cases of hysterectomy (73% cesarean births, 78% multiparous) in women in developed nations.⁴² More than half (55.8%) of women received uterotonics or other surgical interventions prior to hysterectomy, and 43.6 percent had blood transfusion. Ten percent of women required another surgery after hysterectomy to control bleeding (ligation, adnexectomy, laparotomy). Harms were reported in four studies in the review and included fever (n = 135 cases), DIC (116 cases), infection (83 cases), genitourinary morbidity (68 cases), pulmonary morbidity (60 cases), gastrointestinal morbidity (25 cases), neurologic morbidity (16 cases), renal morbidity (8 cases), and cardiovascular morbidity (8 cases). Overall, morbidity did not differ between women undergoing total vs. subtotal hysterectomy.

Finally, one recent review (search dates: not specified) examined effects of PPH guideline implementation and included seven studies (6 cohort studies and one RCT).¹³⁸ Studies were conducted in the United States, Europe, South America, and Pakistan. The incidence of PPH (diagnosed using variable criteria across studies) after guideline implementation declined in four studies and increased in three. The investigators concluded that guidelines can have positive effects on decreasing PPH incidence but note significant flaws among the studies.

Applicability

We set inclusion criteria intended to identify studies with applicability to women being treated for primary or secondary PPH. Studies differed in terms of study population and outcome measures. Most studies did not make direct comparisons between treatments or characterize populations well in terms of severity of PPH and prior management strategies. This lack of direct comparison of treatment options hinders our ability to understand what treatments are most effective and in what order they should be used, both of which are paramount questions for clinicians. We summarize overall applicability below, and Appendix F contains applicability tables for individual interventions.

Overall, findings of studies in the review are generally applicable to the population of women who would be experiencing PPH in hospitals in high-resource nations. Most studies were conducted in Europe or the United States in tertiary care centers. Studies frequently included a number of women with PPH who were transferred from smaller or community hospitals, which can occur when women with PPH requiring additional treatment are stable enough to be moved to facilities with interventional radiology or other services. More women had PPH after cesarean birth than vaginal birth in the 50 studies reporting mode of birth (estimated 6,304 vaginal and 7,924 cesarean births among the 14,228 births for which mode was clearly reported). The most common cause of PPH was atony, which aligns with the most frequent cause of PPH in the larger community and literature. Studies of pharmacologic agents typically included women with mild to moderate to PPH while studies of procedures or surgical approaches generally included women with more severe PPH that had not been controlled with first-line therapies such as uterotonics.

Uterotonics, blood products, and iron supplements studied are generally widely available; however, the accessibility to procedures such as embolization may be limited in smaller community hospitals. Similarly, community hospitals may lack personnel with experience with arterial ligation and compression sutures. Comparators across studies with more than one group were typically either no specific treatment (e.g., rFVIIa or no rFVIIa) or another treatment (e.g., embolization or ligation) and are likely confounded by patient and provider characteristics that may have affected the choice of intervention. For example, patients with more severe hemorrhage likely received more aggressive treatment, and providers could only offer the options available in their facilities. Outcomes addressed across studies were appropriate and clinically relevant; however, few studies reported on longer term outcomes such as future fertility or on patient-centered outcomes such as quality of life.

The populations included in the systems-level interventions both in the United States and Europe reflect those typical of similar size and type (rural, academic, etc.) obstetric units in current labor and delivery environments in the United States. Likewise the interventions designed and implemented in these studies were informed by processes of identifying evidence and crafting guidance that conforms to typical quality improvement and outcomes based research. The content of the interventions is feasible to implement across a full range of settings and the approaches to measuring outcomes are applicable to practice. Overall the systems-level interventions assessed have good applicability to current practice in the United States.

Implications for Clinical and Policy Decisionmaking

A limited body of evidence addresses interventions for managing PPH. Few studies addressed medications commonly used to treat PPH, precluding our ability to draw conclusions about their effectiveness. Success rates for uterine balloon tamponade or surgeries are typically above 60 percent (e.g., success of uterine balloon tamponade as the initial second-line therapy in one study was 86%; success rates for ligation as the first second-line intervention to control bleeding ranged from 36 to 96%). Studies of embolization suggested that it may be associated with a median rate of successful control of bleeding without the need for additional procedures or surgeries of 89 percent, with a wide range of success (58% to 98%) across studies; however, few studies clearly provided data on the success of these procedures and surgeries as the initial second-line approach, so rates are based on a small number of cases. Adverse events and longer term outcomes associated with procedures and surgical interventions are also not well-understood. Some studies reported menstrual changes and infertility rates higher than the general population rates after embolization. Studies of other procedures and surgical interventions did not consistently report fertility data. At this point, the evidence is insufficient to comment on the effectiveness and harms of most interventions for most outcomes.

Thus, given the mixed and insufficient evidence, clinicians will likely need to continue to make individual decisions about the care of women with PPH based on each woman's clinical situation and the management options available in the setting. Embolization, for example, requires an interventional radiologist and may not be widely available. Transportation to a radiology suite may also lead to treatment delays. Choice of some interventions may be guided by the availability of skilled clinicians or may naturally follow cesarean birth (when the abdomen is already open) vs. vaginal birth. This body of evidence does not provide clear answers to the key clinical questions of what interventions to use and in what order.

Limitations of the Comparative Effectiveness Review Process

We included studies published in English only. In our scan of the non-English language literature published since 1990 and located via our MEDLINE search, we determined that the majority would not meet our review criteria. Given the high percentage of non-eligible items in this scan (90%), we feel that excluding non-English studies did not introduce significant bias into the review. We also included only studies conducted very high human development countries as determined by the World Health Organization as these studies have systems of care most relevant to the United States. We recognize that this criterion eliminated many studies of first-line uterotonics such as misoprostol that have been conducted in developing or low resource nations. We provide a summary of recent systematic review of those studies to supplement our analysis (See Findings in Relation to What's Known section above).

Limitations of the Evidence Base

There are a number of limitations in the studies that we reviewed. There is not a universally agreed management strategy for PPH. Medications were typically used as the initial treatment; however, the specific drugs, dosages, and order varied. The selection of interventions, including which interventions were performed and in which order, was also inconsistent. Management was not well described in many studies, especially for women who transferred from other hospitals. Methods for estimating blood loss, when reported, varied and were limited. Overall, it was difficult to ascertain confidently the complete trajectory of care of women in many of the studies we reviewed, which compromises our ability to draw meaningful comparisons. As noted, few studies that met our criteria addressed commonly used uterotonics such as oxytocin; however, prior systematic reviews that have included studies in developing countries have reported similar effects on bleeding for misoprostol and oxytocin and benefits for misoprostol in reducing blood loss with side effects including fever.

Procedures and surgical interventions also differed across studies. For example, materials used for embolization varied, as did the sites of embolization and ligation. There is no clear trigger for starting subsequent interventions, so success rates have limited reliability. It may be that women would have recovered after the first line treatment if time allowed. In addition, there is the potential for cumulative effects of multiple interventions that cannot be measured. Outcomes other than controlling bleeding can be difficult to assess. For example, transfusion could be an adverse outcome if treatment was not sufficient and timely to halt bleeding rapidly. Alternately early transfusion can be the appropriate intervention; therefore, it is sometimes hard to know whether to classify transfusion as an adverse outcome. There are also challenges for measuring harms. In some cases, it can be difficult to assess if harms are due to PPH or management interventions and how much each contributed, especially to deaths. There is a significant lack of truly comparative studies and randomized studies, which would be ideal yet are complex to conduct with a life-threatening condition such as PPH. Studies were typically conducted or data collected over long time frames (median study duration = 5 years, range 6 months to 29 years), and it is likely that interventions and patient characteristics would have changed over time, but few studies account for secular changes such as the introduction of new interventions.

In the systems-level interventions, a natural tension exists between the desire to implement robust interventions and the challenges of understanding which components may have value. In the case of these interventions, it is particularly challenging because lower quality studies with looser measures of outcomes were more likely to see intervention effects. The literature about systems-level intervention is limited by lack of analyses that seek to adjust for secular trends and changes in confounders, such as proportion of births by cesarean and trends in rising body-mass index. Likewise lack of multivariable modeling may obscure the influence of elements of care, such as induction of labor, and comorbidities, such as chorioamnionitis, that could identify which predictors may be exerting substantial influence and inform new approaches to diminishing risk of PPH.

Research Gaps

Future research needs around management of PPH are both clinical and methodologic. Priorities for future research include the following:

• Reaching consensus on definitions and criteria for PPH and first-line management strategies to promote consistency within the literature.
• Standardizing a definition of PPH, potentially with gradations of severity, to allow for meaningful comparison of outcomes.
• Conducting more rigorously controlled studies of all interventions for PPH management, especially medication studies in light of the fact that these are considered first-line management, and few studies in developed/high resource nations addressed agents commonly in use. While studies in the PPH population are likely to be retrospective, studies should clearly describe first-line management and timing of management to clarify the course of care. Studies must report a priori study size calculation to ensure that the number of subjects will be adequate to show a difference (if the study is designed for superiority). In addition, comparative studies must declare within the design and methods section whether the study is a superiority trial or a non-inferiority trial.
• Conducting cluster randomized control trials of intervention bundles that address order of medications, order and timing of manual interventions such as uterine massage and bimanual compression, number of times to repeat medications prior to moving on to second-line interventions, hemodynamic monitoring, and supportive care such as transfusion.
• Clearly identifying the trajectory of care, including which interventions were used and the order and timing of interventions.
• Identifying markers that can inform the decision to move to an alternate intervention.
• Investigating the effectiveness of agents used to control bleeding in other clinical areas and of new medications to address PPH. It is likely that new agents would be compared with or added to existing agents and not compared with placebo.
• Conducting additional RCTs or controlled studies of treating anemia after PPH is stabilized.
• Conducting additional prospectively designed and reported studies that report data from large national databases. These studies can describe effects in larger population samples and may be valuable for identifying longer term harms--for example, effects on breastfeeding, psychological trauma, and future fertility.
• Replicating the intrauterine balloon tamponade study that found it was effective in reducing invasive interventions.
• Using and clearly reporting objective methods to diagnose PPH and evaluate management including accurate measurement of blood loss. Visual estimation of blood loss is too imprecise to be used in research.
• Dedication to prospective objective measures such as estimated blood loss, time course of intervention, and use of intervention components.
• Greater capture and multivariable adjustment, including meta-regression, for known risk factors and confounders to allow better understanding of the attributable impact, if any, of the intervention.
• Attention to the possibility that effect modifiers hide efficacy in some groups, which means that studies will need to be powered and specify a priori stratified analyses by candidate effect modifiers, such as grand multiparity, route of birth, induction, prolonged oxytocin infusion, or infection in labor.
• Prespecifying harms, differentiating harms of interventions from sequelae of PPH wherever possible, and studying longer term effects of procedures and surgical interventions.
• Using multivariate modeling. The size of the study populations in systems-level interventions can clearly support multivariate modeling and could serve to drive better understanding of the general lack of effectiveness. In particular, such data are well-suited to use of risk-adjustment models, and adjusting for these underlying differences in study population characteristics would allow comparison not only across time periods but across studies.
• Attention to the possibility that systems-level interventions are working against a biologically determined risk of PPH, meaning that within a specific population with particular characteristics there is an irreducible level of risk and event rates cannot be driven below that “floor.” If this floor were demonstrated with risk adjustment methods, this finding would fundamentally change the focus of study design and care. A floor would suggest that we need very large pragmatic trials aimed not at reducing the occurrence of PPH but at diminishing associated morbidity, mortality, personal harm and distress, and costs. The systems-level intervention studies available now cannot fully inform this goal, but primary meta-analyses of the highest quality cohorts with risk adjustment could determine if the evidence seen in some of the included studies that suggest benefits are worth pursing on a larger scale, including a scale large enough to separate the influence of candidate components to determine their individual contributions to improvements in care.

Conclusions

A limited body of evidence addresses interventions for managing PPH. The most effective treatments and the order in which to use treatments remain unclear. Diagnosis of PPH is subjective, which makes it difficult to compare the severity of PPH and determine the comparability of participants within and across studies. The trajectory of care, rationale for choice of intervention, and component of care ultimately responsible for controlling bleeding are also frequently unclear because of the need for rapid intervention in an emergency situation. Few studies included in this review addressed pharmacologic or medical management, including transfusion for supportive management of ongoing PPH, and the evidence reviewed is insufficient to comment on effects of such interventions. The success of uterine-sparing techniques, such as uterine balloon tamponade, embolization, uterine compression sutures, and uterine and other pelvic artery ligation, in controlling bleeding without the need for additional procedures or surgeries ranged from 36 to 98 percent; however, these data come from a limited number of studies with a small number of participants. Harms of interventions are diverse and not well-understood. Some studies reported an association between rFVIIa and thromboembolic events, but sample sizes were small. Some studies with longer term followup reported adverse effects on future fertility and menstrual changes in women undergoing embolization. Need for re-operation was reported after hysterectomy. Evidence is insufficient to assess the effects of interventions for anemia after PPH is stabilized, and systems-level interventions showed little benefit in reducing the incidence or severity of PPH or the need for transfusion or hysterectomy. Further research is needed across all interventions for PPH management, especially pharmacologic interventions, which are frequently used as first-line therapies.