Knight, Marian MBChB, DPhil; Tuffnell, Derek MBChB; Brocklehurst, Peter MBChB, MSc; Spark, Patsy BSc; Kurinczuk, Jennifer J. MBChB, MD; on behalf of the UK Obstetric Surveillance System
Amniotic-fluid embolism is a leading cause of maternal mortality, with recent rates appearing to have increased in some countries1,2 although not in others.3,4 There were no changes in diagnostic criteria or methods of case ascertainment to account for the observed rise in the number of deaths in the United Kingdom and Australia, but the rarity of the disorder makes it particularly difficult to investigate whether the larger number of deaths reflects an increase in the incidence of the condition, an increase in the case fatality, or a chance finding. Two recent retrospective analyses of incidence using large national hospital databases in the United States and Canada,3,4 did not find a temporal trend in the occurrence of fatal or nonfatal cases.
Prospective surveillance of amniotic-fluid embolism has been undertaken in the United Kingdom through two routes. In a passive system, between 1997 and 2004, cases were reported on a voluntary basis to the United Kingdom Amniotic Fluid Embolism Register5; since 2005, cases have been reported actively on a national basis through the routine monthly mailing of the UK Obstetric Surveillance System.6 The aim of this study was to estimate the current incidence of amniotic-fluid embolism in the United Kingdom and to investigate whether there is any evidence to suggest that disease incidence is changing. In addition, we sought to describe risk factors, current management, case fatality, and other outcomes.
MATERIALS AND METHODS
We carried out a prospective, population-based cohort and nested case-control study. We identified cases through the monthly mailing of the UK Obstetric Surveillance System6,7 between February 2005 and February 2009. Clinicians were asked to report any woman diagnosed with amniotic-fluid embolism with symptoms and signs consistent with amniotic-fluid embolism or any woman in whom amniotic-fluid embolism was confirmed by biopsy or postmortem examination (Box 1).
The UK Obstetric Surveillance System methodology has been described in detail elsewhere.6 In brief, every month, the UK Obstetric Surveillance System case-notification cards were sent to nominated reporting clinicians in each hospital in the United Kingdom with a consultant-led maternity unit, with a tick box to indicate whether they had seen a woman with amniotic-fluid embolism. They also were asked to return cards indicating a “nil report” so that we could monitor card-return rates and confirm the denominator to calculate the incidence. Of note, any women with amniotic-fluid embolism delivering in other settings in the United Kingdom, including either in a midwife-led unit or at home, would be transferred to a consultant-led unit and therefore would be captured by the UK Obstetric Surveillance System. When a clinician reported a case, they were sent a data-collection form asking for details of presentation, management, and outcomes. All data were collected anonymously.
Data were double entered into a customized database. Two of the authors (D.T., M.K.) reviewed all cases in which postpartum hemorrhage was reported as the first symptom of amniotic-fluid embolism against the diagnostic criteria (Box 1) to determine independently whether amniotic-fluid embolism was the most likely diagnosis.
The Centre for Maternal and Child Enquiries, which has collected information on all maternal deaths in the United Kingdom for more than 50 years, provided information (year of birth and date of diagnosis) about any maternal deaths from amniotic-fluid embolism occurring during the study period. These were compared with maternal deaths reported through the UK Obstetric Surveillance System; no additional cases were identified.
All analyses were carried out using STATA 10 software (StataCorp LP, College Station, TX). Incidence with 95% confidence intervals (CIs) was calculated using the most recently available birth data (2005 to 2007) as a proxy for February 2005 to February 20098–10; the total estimated number of maternities (women delivering) was 3,049,100.
We investigated the potential factors underlying amniotic-fluid embolism using an exploratory logistic regression analysis to estimate odds ratios (ORs) and 95% CIs. We used national data where available to conduct a univariable analysis and information from 1,227 women in the nested control group to conduct an adjusted analysis because national data do not have sufficient information on potential confounders. The women in the control group were identified by the UK Obstetric Surveillance System reporters as the two women delivering in the same hospital immediately before other UK Obstetric Surveillance System cases.11 A full regression model was developed by including both explanatory and potential confounding factors in a core model if there was a preexisting hypothesis or evidence to suggest they were causally related to amniotic-fluid embolism, for example, induction of labor. Continuous variables were tested for departure from linearity by the addition of quadratic terms, and potential interactions were tested by the addition of interaction terms between all variables in the model and subsequent likelihood-ratio testing on removal. P<.05 was considered evidence of a significant interaction or departure from linearity. The association of amniotic-fluid embolism with cesarean delivery was examined in a regression model including only those women who had amniotic-fluid embolism after delivery to exclude cases in which cesarean delivery was likely to be a consequence rather than a cause of the amniotic-fluid embolism. The analysis had 80% power at the 5% level of statistical significance to detect an OR for induction of labor of 2.6 or greater.
To estimate the proportion of cases attributable to specific causes, we calculated population-proportional attributable risks using adjusted ORs and the proportion of cases exposed.12 The UK Obstetric Surveillance System general methodology (04/MRE02/45) and this study (04/MRE02/46) were approved by the London Multicentre Research Ethics Committee.
All 229 eligible U.K. hospitals contributed data to the UK Obstetric Surveillance System during the study period (100% participation). Data collection was complete for 97% of cases (Fig. 1).
There were 60 confirmed cases in an estimated 3,049,100 maternities,8–10 representing an estimated incidence of 2.0 cases per 100,000 maternities (95% CI 1.5–2.5). There was no significant change in incidence over the 4 years of the study, although the study has low power to detect any difference owing to small case numbers (data not shown). The nine women whose cases were excluded all had primary postpartum hemorrhage with no evidence of additional features, such as cardiac or respiratory compromise or early coagulopathy, to suggest amniotic-fluid embolism.
Characteristics of women with amniotic-fluid embolism are shown in Table 1. Unadjusted analysis suggested a linear increase in risk of amniotic-fluid embolism with increasing age (OR 1.12 for every one year increase in age, 95% CI 1.07–1.18). For ease of presentation we have shown the data in binary groups (OR 2.59, 95% CI 1.53–4.41 in women aged 35 and over compared with women aged under 35). There was a nonsignificant association with maternal ethnicity (OR 1.18, 95% CI 0.61–2.27 in ethnic minority women compared with white women). The adjusted analysis revealed a significant interaction between ethnicity and age (P<.001), suggesting that the association with age varies between different ethnic groups. We therefore have presented stratified results that show the highest odds in ethnic-minority women older than 35 years (OR 9.85, 95% CI 3.57–27.2) (Table 2). In the U.K. population, the proportions of different ethnicities in the ethnic-minority group were Asian 47%, black 31%, mixed race 6%, Chinese 5%, and other ethnic-minority groups 10%. We were unable to identify any significantly raised risk associated with a particular minority ethnicity, but note that this may be the result of small group sizes and limited study power. After adjustment, no other sociodemographic factors significantly affected the risk of amniotic-fluid embolism.
The occurrence of amniotic-fluid embolism was significantly associated with induction of labor and multiple pregnancy (Table 2). Corresponding population proportional attributable risks are 35% for induction of labor, 13% for ethnic-minority women 35 years or older and 7% for multiple pregnancy.
Twenty-six women had amniotic-fluid embolism after delivery; 19 (73%) of these occurred after cesarean delivery. The adjusted OR for delivery by cesarean in women who had amniotic-fluid embolism after delivery was 8.84 (95% CI 3.70–21.1). Ten of the 19 women who had amniotic-fluid embolism after cesarean delivery were not in labor at the time of the cesarean. Five of these had elective cesarean operations (one at maternal request and four after previous cesarean deliveries), and four had emergency cesarean deliveries (three for fetal distress and one for suspected placenta previa); for one woman, the reason for cesarean delivery is unknown. The population-proportional attributable risk associated with cesarean delivery (before amniotic-fluid embolism) was 62%.
Information on time of event was missing for one woman. Thirty-three women (56%) had symptoms or signs of embolism at or before delivery; the remainder had symptoms or signs after delivery. The amniotic-fluid embolism occurred at a median gestation of 39 weeks (range 28–42) (Fig. 2) and presented clinically within a 6-hour range around delivery (range 2 hours, 18 minutes before delivery to 4 hours after delivery). Symptoms and signs at presentation are shown in Table 3. All the women had at least one, and 16 (27%) women had at least four of the five cardinal features of shortness of breath, hypotension, maternal hemorrhage, coagulopathy, and premonitory symptoms. Fetal distress occurred before maternal collapse in 19 cases (32%); in six (32%) of these cases, the woman was delivered urgently as a consequence of fetal distress and collapsed after delivery.
The majority of women (55, 92%) had ruptured membranes at or before the time of presentation with amniotic-fluid embolism. In those women who had amniotic-fluid embolism before delivery, the amniotic-fluid embolism presented a median of 45 minutes after membrane rupture (range 0 minutes to 6 days after); in four women the amniotic-fluid embolism presented at the time of membrane rupture, and a further three women presented within 10 minutes of membrane rupture. An obstetrician was present at the time of presentation in 28 cases (47%) and an anesthetist in 27 (45%); neither was present in 26 cases (43%) (Table 1).
Twenty-eight women (85%) who presented with amniotic-fluid embolism at or before delivery were delivered by cesarean. A range of management strategies were used after amniotic-fluid embolism in addition to supportive therapies (Table 4). Supportive therapies alone, defined as fluids and blood products only, were used in 35 women (58%); 15 women (25%) had hysterectomy to control hemorrhage, and 15 women (25%), including one who also had a hysterectomy, were treated with factor VIIa. In total, five women (8%) had a combination of different treatment strategies in addition to supportive therapies. Women were transfused a median of 6.5 units of blood (range 0–62 units).
Twelve women died (case fatality rate 20%, 95% CI 11–32%) (Table 1). All the women died within 1 day of the amniotic-fluid embolism at a median of 1 hour, 40 minutes after the acute event (range 0 minutes to 23 hours, 18 minutes). Women who died were significantly more likely to be from an ethnic-minority group than were those who survived (OR 4.64, 95% CI 1.11–19.5); this association persisted after adjustment for differences in age, socioeconomic status, body mass index, and parity (adjusted OR 11.8, 95% CI 1.40–99.5). There were no other significant differences between fatal and nonfatal cases, although this analysis has limited power to detect differences. Eight women who survived had other severe morbidities: four had cerebral neurologic injury, two suffered a thrombotic event, one had septicemia, and one had renal failure. Forty-five of the 48 women who survived (94%) were admitted to intensive care. Only 3 of the 12 women who died (25%) were admitted to intensive care; the remainder died before admission. The median length of stay in intensive care was 3 days (range 1–39 days).
There was one stillbirth and four neonatal deaths (all due to asphyxia before birth) among the 37 neonates for whom outcomes are known and who were born to mothers who had amniotic-fluid embolism at or before delivery, giving a perinatal mortality rate of 135 per 1,000 total births (95% CI 45–288). Neither mother nor neonate survived in three cases. None of the neonates born to mothers who had amniotic-fluid embolism after delivery died.
The incidence of amniotic-fluid embolism as estimated by this 4-year prospective national study is 2.0 per 100,000 maternities (95% CI 1.5–2.5), which is significantly lower than the rate documented in retrospective reviews of population-based hospital discharge databases in Canada (6.1 cases per 100,000 deliveries, 95% CI 5.3–7.14) and the United States (7.7 cases per 100,000 births, 95% CI 6.7–8.73). As we demonstrated, amniotic-fluid embolism is difficult to diagnose; of 86 nonduplicate cases reported, further examination of the records revealed a more likely diagnosis in 26 (30%). Hospital discharge databases do not allow for examination of sufficiently detailed clinical information to either confirm or refute the diagnosis of amniotic-fluid embolism and, therefore, are likely to be subject to a degree of overreporting. Neither of the previous studies undertook validation of the diagnosis by examining medical charts. The authors of the Canadian report cite their observed associations of amniotic-fluid embolism with complications including eclampsia, placenta previa, and abruption as evidence of overreporting; similar associations were observed in the U.S. study.
Although a possibility, we believe that our lower incidence estimate is unlikely to be due to underreporting of cases because the UK Obstetric Surveillance System is an active, prospective surveillance system in which we require negative reports, ie, participating hospitals return a report card every month, regardless of whether there are cases to report. We thus can be confident that reports of “no cases” from an individual hospital are true negative reports. Furthermore, all of the fatal cases reported to the Centre for Maternal and Child Enquiries in the study period had been reported to the UK Obstetric Surveillance System. If the true incidence in the U.K. population were the same as that estimated in Canada (6.1 per 100,000), there would have been 186 cases in total, implying that we had missed 126 nonfatal cases and that the case fatality rate was only 6% (12 of 186); levels of both missing cases and fatalities that seem highly improbable.
A third explanation for the threefold difference in incidence estimates is that this reflects a true difference in incidence in the United Kingdom compared with North America. Given that the population-proportional attributable risk associated with cesarean delivery is 62% and that the cesarean-delivery rate in the United States is 50% higher than in the United Kingdom, one might indeed anticipate a higher rate of amniotic-fluid embolism in the United States than in the United Kingdom.
Induction of labor was associated with a population-attributable risk of 35% in our study, suggesting that, assuming causality, if induction of labor were no longer performed, 35% of cases of amniotic-fluid embolism could be prevented. The practice of inducing labor, with its many potential benefits, clearly will continue, and amniotic-fluid embolism remains a very rare complication; nevertheless, clinicians should be aware of both the risks and benefits of induction because more restricted use may result in a decrease in the number of women suffering a potentially fatal amniotic-fluid embolism.
Both Canadian and U.S. studies show an increased rate of cesarean delivery among women with amniotic-fluid embolism, which the authors of the U.S. study interpret as causal. However, as the investigators discuss, one of the limitations of administrative databases is that they do not indicate whether the amniotic-fluid embolism occurred before, during, or after delivery. Our prospectively collected data include timing of all the events, and we found that 56% of women had amniotic-fluid embolism at or before delivery, with 44% after delivery. More than 80% of women who had amniotic-fluid embolism at or before delivery subsequently were delivered by cesarean; clearly, in these cases, the cesarean delivery was a consequence and not a cause of amniotic-fluid embolism. However, we also found more than an eightfold increase in the odds of amniotic-fluid embolism associated with cesarean delivery in the group who had amniotic-fluid embolism after delivery, suggesting that there indeed may be a causal relationship. The clinical message from this study thus echoes other studies in noting that cesarean delivery is not risk-free7 and that all risks and benefits should be assessed before a decision for cesarean delivery is made.
Older maternal age was also an important factor in all the studies, although, in our population, having adjusted for confounders, the increase in risk with age was limited to ethnic-minority women. The small number of older ethnic-minority women with amniotic-fluid embolism means that we do not have sufficient power to investigate this association further, but continued surveillance through the UK Obstetric Surveillance System may allow us to address this in the future. In the meantime, clinicians should be aware of the increased risk when caring for older pregnant women who are ethnic-minorities.
All the women in our study experienced at least one of the cardinal features of shortness of breath, hypotension, hemorrhage, coagulopathy, and premonitory symptoms at the time of amniotic-fluid embolism, and all cases were consistent with the reporting criteria used in a U.S. amniotic-fluid embolism registry.13 Premonitory symptoms such as numbness, tingling, and agitation were the most common initial presenting feature, although acute fetal compromise was the first feature noted in 36% of antenatal cases. Fetal compromise was noted first as a fetal heart rate abnormality in the majority of cases; clinicians should consider amniotic-fluid embolism as part of the differential diagnosis of a sudden, unexplained deterioration in the fetal heart rate even in the absence of maternal symptoms or signs.
The mainstay of the management of amniotic-fluid embolism remains supportive therapy, and this study shows that high-quality supportive care can result in a good maternal outcome in the majority of cases. The results also highlight the use of a number of therapies for amniotic-fluid embolism previously described only in case reports. Coagulopathy is a common feature of amniotic-fluid embolism (occurring in more than 60% of our cases), and, therefore, use of recombinant factor VIIa, a procoagulant leading to increased thrombin formation, has been proposed, with individual reports of its success.14–16 In our study, 14 women were treated with factor VIIa for coagulopathy; 13 survived.
Seven women were treated with exchange transfusion or plasma exchange, a therapy first described more than 20 years ago.17 Authors have suggested variously that clinical improvement is related to the removal of red cell debris and hemoglobin from the circulation17 or that the process removes amniotic fluid and cytokines from the circulation and corrects metabolic acidosis.18 Thus, these therapies should be regarded as an extension of supportive care and not as a substitute. All seven women treated with these therapies survived, although the small numbers involved means that we should not assume that this form of therapy is more effective than any other; survival to the point where these techniques are possible may put the women in a more favorable prognosis group.
One fifth of women with amniotic-fluid embolism and 1 in 10 of their newborns died. Maternal fatality was comparable with that in the Canadian and U.S. studies (13% [95% CI 9–19%] and 22% [95% CI 16–28%], respectively). This confirms the view that maternal fatality in large, unselected populations is lower than previously suggested. We found that women who died were significantly more likely to come from ethnic-minority groups than were survivors, an association that could not be explained by differences in age, socioeconomic status, body mass index, or parity. In previous UK Obstetric Surveillance System studies,11 we found that the incidence of severe maternal morbidity is higher among ethnic-minority women, and we postulate that this may be due to differences in underlying medical conditions or access to care. It is possible that these factors also may play a role in the increased risk of fatality from amniotic-fluid embolism for these women.
1. Lewis GE, editor. The Confidential Enquiry into Maternal and Child Health (CEMACH). Saving mothers lives: reviewing maternal deaths to make motherhood safer 2003–2005. London (UK): CEMACH; 2007.
2. Sullivan E, Hall B, King J. Maternal deaths in Australia 2003–2005. Canberra (Australia): Australian Institute of Health and Welfare; 2008.
3. Abenhaim HA, Azoulay L, Kramer MS, Leduc L. Incidence and risk factors of amniotic fluid embolisms: a population-based study on 3 million births in the United States. Am J Obstet Gynecol 2008;199:49.e1–8.
4. Kramer MS, Rouleau J, Baskett TF, Joseph KS. Amniotic-fluid embolism and medical induction of labour: a retrospective, population-based cohort study. Lancet 2006;368:1444–8.
5. Tuffnell DJ. United Kingdom Amniotic Fluid Embolism Register. BJOG 2005;112:1625–9.
6. Knight M, Kurinczuk JJ, Tuffnell D, Brocklehurst P. The UK Obstetric Surveillance System for rare disorders of pregnancy. BJOG 2005;112:263–5.
7. Knight M, Kurinczuk JJ, Spark P, Brocklehurst P. Cesarean delivery and peripartum hysterectomy. Obstet Gynecol 2008;111:97–105.
8. Office for National Statistics. Key population and vital statistics (2005). Newport (UK): Office for National Statistics; 2007.
9. Office for National Statistics. Key population and vital statistics (2006). Newport (UK): Office for National Statistics; 2008.
10. Office for National Statistics. Key Population and Vital Statistics (2007). Newport (UK): Office for National Statistics; 2009.
11. Knight M, Kurinczuk JJ, Spark P, Brocklehurst P, UKOSS. Inequalities in maternal health: national cohort study of ethnic variation in severe maternal morbidities. BMJ 2009;338:b542.
12. Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health 1998;88:15–9.
13. Clark SL, Hankins GD, Dudley DA, Dildy GA, Porter TF. Amniotic fluid embolism: analysis of the national registry. Am J Obstet Gynecol 1995;172:1158–67.
14. Kahyaoglu I, Kahyaoglu S, Mollamahmutoglu L. Factor VIIa treatment of DIC as a clinical manifestation of amniotic fluid embolism in a patient with fetal demise. Arch Gynecol Obstet 2009;280:127–9.
15. Lim Y, Loo CC, Chia V, Fun W. Recombinant factor VIIa after amniotic fluid embolism and disseminated intravascular coagulopathy. Int J Gynaecol Obstet 2004;87:178–9.
16. Prosper SC, Goudge CS, Lupo VR. Recombinant factor VIIa to successfully manage disseminated intravascular coagulation from amniotic fluid embolism. Obstet Gynecol 2007;109:524–5.
17. Dodgson J, Martin J, Boswell J, Goodall HB, Smith R. Probable amniotic fluid embolism precipitated by amniocentesis and treated by exchange transfusion. Br Med J (Clin Res Ed) 1987;294:1322–3.
18. Kaneko Y, Ogihara T, Tajima H, Mochimaru F. Continuous hemodiafiltration for disseminated intravascular coagulation and shock due to amniotic fluid embolism: report of a dramatic response. Intern Med 2001;40:945–7.