Amniotic Fluid Embolism: Decreased Mortality
in a Population-Based Study
WILLIAM M. GILBERT, MD, AND BEATE DANIELSEN, PhD
Objective: To examine the risk factors and pregnancy out-
comes associated with 53 cases of amniotic fluid embolism
that occurred in California during the 2-year period January
1, 1994 to December 31, 1995.
Methods: Data were obtained from a computerized data-
base that contains linked records from the vital statistics
birth certificate and hospital discharge summaries of both
mother and newborn. This database covered all singleton
deliveries that occurred in 328 civilian acute-care hospitals in
California, which represented 98% of all deliveries in Cali-
fornia. All cases of amniotic fluid embolism were examined
for other pregnancy complications.
Results: There were 1,094,248 deliveries during that 2-year
period. Fifty-three singleton gestations had the diagnosis of
amniotic fluid embolism, for a population frequency of one
per 20,646 deliveries. Fourteen women with amniotic fluid
embolism died, for a maternal mortality rate of 26.4%. There
were 35 (66%) diagnoses of disseminated intravascular co-
agulation (DIC), 38 (72%) diagnoses of hemorrhage, and 25
(47%) diagnoses of obstetric shock. Among the 14 women
who died, the frequency of DIC (79%) and hemorrhage
(71%) was not different compared with that of the survivors
(62% and 72%, respectively), but obstetric shock was higher
(86%, P 5 .02) than in survivors (33%). The average maternal
length of stay for survivors was 6.5 days (range 3–27 days,
median 5 days). The cesarean rate was 60% and the fre-
quency of fetal distress was 49%.
Conclusion: In this population-based study of reported
cases of amniotic fluid embolism, the maternal mortality
rate (26.4%) was significantly less than previously re-
ported and might reflect a more accurate population
frequency. In addition, patients who survived and pa-
tients who died had similar pregnancy complications,
suggesting that amniotic fluid embolism was present in
all cases and not limited to those who died. (Obstet
Gynecol 1999;93:973–7. © 1999 by The American College
of Obstetricians and Gynecologists.)
From the Department of Obstetrics and Gynecology & Center for
Health Services Research in Primary Care, University of California,
Davis, and Health Information Solutions, Redwood City, California.
VOL. 93, NO. 6, JUNE 1999
Amniotic fluid embolism is an unexpected and rare
complication of pregnancy often presenting with sud-
den maternal cardiovascular collapse, disseminated in-
travascular coagulation (DIC), and maternal death.1–5
Historically, the diagnosis of amniotic fluid embolism
was made when a woman presented with those find-
ings and at autopsy when fetal squamous and amniotic
fluid cells were found within the maternal pulmonary
arteries.1,6,7 The prevailing theory of the cause of amni-
otic fluid embolism is that amniotic fluid containing
fetal cells enters the uterine venous sinuses within the
endometrium or endocervix. The fetal cells and amni-
otic fluid return to the maternal heart through the
venous system and enter the lungs in sufficient quantity
to cause an embolism or severe pulmonary vasocon-
striction.1–7 The resultant hypoxia causes cardiac and
hemodynamic collapse, DIC, and usually maternal
death.
Amniotic fluid embolism has been reported to occur
in as many as one in 8000 and as few as one in 80,000
pregnancies.1,2,8 Because amniotic fluid embolism is
rare, rapidly progressive, and unpredictable, it is diffi-
cult to study. The medical literature on amniotic fluid
embolism largely consists of case reports or series
collected over many years or even decades.1– 4,8 Clark et
al4 recently described their national registry of 46 cases
of amniotic fluid embolism collected over a 5-year
period, in which the maternal mortality rate was 61%,
which is consistent with previous reports.2,3 Little im-
provement in outcome has occurred since Steiner and
Luschbaugh1 published eight fatal cases in 1941. The
Clark et al4 national registry has inherent problems with
referral bias; physicians might refer only the best- or
worst-outcome cases of amniotic fluid embolism, which
does not provide a true picture of the disease or its
mortality rate. Our study examines a large population
of patients who delivered during the 2-year period
1994 –1995 in California with the diagnosis of amniotic
fluid embolism, thus allowing us to obtain a true
population-based frequency.
0029-7844/99/$20.00 973
PII S0029-7844(99)00004-6
Materials and Methods Table 1. Demographic Characteristics of 53 Patients With
Amniotic Fluid Embolism
A unique database was used that linked maternal and Mean 6 standard
neonatal hospital discharge records to vital statistics Characteristic error or %
birth certificate records. The linkage of vital statistics
Maternal age (y) 33 6 1
was established for all single live births in acute-care Parity (after delivery) 2.6 6 0.2
civilian hospitals (n 5 328) that reported to the Califor- Gestational age (wk) 38.8 6 0.3
nia Office of Statewide Health Planning and Develop- Birth weight (g) 3307 6 86
ment between January 1, 1994 and December 31, 1995. Maternal length of stay (d)* 6.5 6 0.8†
Neonatal length of stay (d)* 5.1 6 0.8‡
This database did not include births from home deliv-
Cesarean rate 62%
eries, out-of-state deliveries, and birthing centers not Race
reporting to the Office of Statewide Health Planning Non-Hispanic white 41.5%
and Development or deliveries at military facilities (2% Asian 15.1%
of all deliveries in the state). The method successfully Hispanic 24.5%
Black 5.7%
linked 98.9% of maternal and 98.6% of neonatal hospital
Other 13.2%
discharge records with the vital statistics birth records,
* Length of stay of survivors.
resulting in an overall linkage of 97.9% of reported †
Median 5 days.
singleton deliveries. We used data for multiple deliver- ‡
Median 3 days.
ies from the maternal discharge record only and there-
fore were unable to obtain information on newborn
outcome. All information relates to the singleton preg-
nancies unless otherwise stated. A database of 1,094,248 embolism grouped by maternal outcome. Most re-
deliveries was generated. Using SAS software (SAS ported frequencies of codiagnoses were similar between
Institute, Cary, NC), the database was queried using the two groups, including hemorrhage and DIC, but
codes from the International Classification of Diseases, obstetric shock was higher in women who died.
Ninth Revision (ICD-9) and Current Procedural Termi- Only two neonates died during the initial hospital-
nology, which resulted in a specific data set for statis- ization, whereas eight (15%) were transferred to other
tical analysis. Amniotic fluid embolism and other diag- hospitals (unknown outcomes), and five (9%) records
noses were not defined beyond those given in the had insufficient outcome information recorded. A rou-
codebooks. The linked database was searched for the tine newborn discharge was reported in 38 (72%) of all
diagnosis of amniotic fluid embolism, (ICD-9 673.1) and cases and in seven (50%) cases in which the mother died
all cases were examined with respect to multiple demo- (Table 2). A normal maternal discharge from the hospi-
graphic characteristics, as well as antepartum, intrapar- tal was found in 34 of 39 (87%) survivors, and five (13%)
tum, and postpartum diagnoses. Fisher exact test was were discharged with home health referrals or to long-
used to compare the group of maternal deaths (14 term care facilities.
women) with the group of survivors (39 women). A P
value of less than .05 was assumed significant unless
otherwise stated. Table 2. Pregnancy Complications Associated With
Amniotic Fluid Embolism
Total singleton Maternal
Results population deaths Survivors
Complication (n 5 53) (n 5 14) (n 5 39) P*
There were 1,094,248 women who delivered during the
DIC 35 (66) 11 (79) 24 (62) .16
2-year period. Fifty-three singleton pregnancies had the
Hemorrhage 38 (72) 10 (71) 28 (72) .39
diagnosis of amniotic fluid embolism, for a population Shock 25 (47) 12 (86) 13 (33) .02
frequency of one per 20,646 pregnancies. The maternal Cesarean 32 (60) 9 (64) 23 (59) .58
discharge records noted that four additional cases of Fetal distress 26 (49) 10 (71) 16 (41) .17
amniotic fluid embolism were multiple gestations. Of Female sex 26 (49) 9 (64) 17 (44) .21
Normal newborn† 38 (72) 7 (50) 31 (79) .71
these four pregnancies, all had DIC, cesarean deliveries,
and hemorrhage; two mothers died and two infants had DIC 5 disseminated intravascular coagulation (International Clas-
sification of Diseases, Ninth Revision, [ICD-9] codes 641.3, 666.3);
fetal distress. The demographic characteristics of the hemorrhage (ICD-9 codes 285.1, 998.11,666); shock (ICD-9 codes 669.1).
singleton patients with amniotic fluid embolism are Data are given as n (%).
shown in Table 1. Fourteen (26.4%) of the 53 women * P represents comparison between the maternal deaths and survi-
vor groups based on Fisher exact test.
died during the delivery hospitalization. Table 2 shows †
Diagnosis of “normal newborn” diagnosis-related group (DRG)
the coincident morbidities reported with amniotic fluid 391 at hospital discharge.

974 Gilbert and Danielsen Amniotic Fluid Embolism Obstetrics & Gynecology
Discussion
In this population-based study of 1,094,248 pregnancies,
we studied maternal and neonatal outcome in 53 cases
of amniotic fluid embolism within a 2-year period. The
most surprising finding was a maternal mortality rate of
26.4% associated with the diagnosis of amniotic fluid
embolism, which is considerably less than rates re-
ported previously (61– 80%).2– 4 There are several possi-
ble explanations for the differences between the current
and previous studies. The first explanation is that our
study is the first population-based study with a suffi-
ciently large number of cases of amniotic fluid embo-
lism to be able to obtain an unbiased population-based
frequency. Prior studies were largely compiled regis-
tries or case reports, which we cannot assume to repre-
sent true population frequencies.1– 4,8 If a patient sur-
vives an amniotic fluid embolism, it might not be of
adequate interest to report a case in the literature or
refer to a registry. Second, our study included the 2
years from January 1, 1994, to December 31, 1995. Many
previous studies spanned multiple years and were done
before widespread use of intensive care units for high-
risk patients. The intensive care, multidisciplinary ap-
proach to the treatment of very sick patients is becom-
ing more widely available, especially within the past 10
years. Recently, Burrows and Khoo8 published a series
of ten cases of amniotic fluid embolism with a maternal
mortality rate of 22%.8 This decrease in maternal mor-
tality rate may reflect these recent developments in
intensive care management of these patients.
Clark et al4 published an extensive report of their
registry of 46 cases of amniotic fluid embolism collected
over a 5-year period starting in 1988. They examined
121 clinical variables with strict entry criteria, including
acute hypotension or cardiac arrest; acute hypoxia; DIC;
and onset during labor, delivery, or within 30 minutes
after delivery or pregnancy termination.4 They reported
dismal maternal outcomes with a maternal mortality
rate of 61% and a neurologically intact maternal sur-
vival rate of 15%. Neonatal outcome was only margin-
ally better, with a survival rate of undelivered fetuses at
the time of amniotic fluid embolism of 79%, in which
only 50% of those infants were normal at discharge.4
Our results are markedly different from theirs, with a
maternal mortality rate of 26.4% and a normal maternal
discharge reported in 87% of survivors. The neonatal
survival rate was 95% (38 of 40) of known outcomes,
and routine discharge was reported in 72% of all cases.
We believe the difference in the mortality and morbidity
rates results partially from the nature of the case collec-
tion process. A registry represents a collection of cases,
probably the worst cases, not a population frequency.
Our database was limited to information reported on
VOL. 93, NO. 6, JUNE 1999
the maternal and neonatal hospital discharge summa-
ries and vital statistics birth certificates. We did not
have access to individual medical records and therefore
relied on data entered at the time of hospital discharge
or death. If a mother or infant was transferred to
another facility, the outcome of the subsequent care of
that patient would not necessarily be included in the
hospital discharge record. Finally, the number of cases
in both groups (survivors and deaths) was not large
because of the rarity of the condition. Therefore, when
small groups are compared, there may be statistically
significant differences that might not hold when larger
groups are compared.
One concern regarding our database is whether am-
niotic fluid embolism is overreported. The population
frequency (one per 20,646 deliveries) is within the range
of frequencies that has been reported in the literature
(one per 8000 to one per 80,000 deliveries), indicating
that we are not overreporting the frequency of amniotic
fluid embolism. Furthermore, the codiagnoses with
amniotic fluid embolism were similar between the
group of survivors and the group that died (Table 2).
Disseminated intravascular coagulation, hemorrhage,
and fetal distress were similarly reported between
groups, but obstetric shock was higher in the group that
died. Whether this difference is a true difference be-
tween the groups that survived or died or that the
women who survived had lesser degrees of disease is
unknown. Others have reported similar frequencies of
DIC (45– 60%) in cases of fatal amniotic fluid embolism,
suggesting that the frequency of codiagnoses in our
study is consistent with the diagnosis of amniotic fluid
embolism.9 –11 One final point to consider with a mater-
nal death is that physicians and hospitals might include
every possible diagnosis to demonstrate the extreme
nature of the patient’s condition to help explain the
death. The distribution of comorbidity found with the
survivors was similar to that found with those who
died, strongly confirming the diagnosis of amniotic
fluid embolism in the survivors.
The demographic characteristics of our patients with
amniotic fluid embolism are different from those of
women with normal deliveries in California.12 Our
patients were older (mean age 33 years, Table 1), more
likely to be non-Hispanic white and less likely to be
Hispanic (in California in 1994, 36% of births were
non-Hispanic white, 7% black, 45% Hispanic, 10%
Asian, 2% other) and more likely to be multiparous
(mean parity after delivery 2.6), which is consistent
with other reports of amniotic fluid embolism.2,4 Clark
et al4 found an increase in male offspring (67%) in their
registry that was not seen in our population (51%). The
cesarean rate in this population was markedly higher
(60%) than that of the total population during this time
Gilbert and Danielsen Amniotic Fluid Embolism 975
period (22.8%) and was not different between the total
group of women with amniotic fluid embolism and
those that died (Table 2). Lau and Chui11 also found a
high (50%) cesarean rate in 10 fatal cases of amniotic
fluid embolism. Because the exact time of the onset of
amniotic fluid embolism in our cases cannot be deter-
mined from the data set, we are unable to determine
accurately the timing of the cesarean. The higher fre-
quency of cesareans is most likely the result of the
increased diagnosis of fetal distress (49%).
There were four cases of multiple gestation (four of 57
cases, 7%), which is higher than that in the general
population (1–2%). Clark et al4 did not find higher rate
of twin gestations (2%). The finding of more twin
gestations in our population is consistent with the
theory that uterine overdistension is a risk factor asso-
ciated with amniotic fluid embolism.2 If one twin rup-
tured its membranes, with the second twin’s mem-
branes still intact, the amniotic fluid could more easily
enter the maternal venous system causing an amniotic
fluid embolism. Amniotic fluid embolism has been
reported to be consistent with an allergic (anaphylactic)
reaction instead of a true embolism. Steiner and Lush-
baugh1 attributed death to an “anaphylactoid reaction”
which has been mentioned by others.1,2,13 Morgan2
believed that an allergic reaction as a cause of amniotic
fluid embolism was less likely because of inconsisten-
cies between histamine-related anaphylaxis and clinical
amniotic fluid embolism. Recently, however, there has
been renewed interest in allergy as a cause of amniotic
fluid embolism.4,14,15 The commonality between immu-
noglobulin E-mediated anaphylaxis, endotoxin-medi-
ated sepsis, and amniotic fluid embolism supports this
interest. In addition, experimentally induced amniotic
fluid embolism can be prevented by the administration
of a 5-lipoxygenase inhibitor, a leukotriene-blocking
agent.16 This finding strongly indicates the effect of
leukotrienes in amniotic fluid embolism.16 Clark et al4
proposed renaming the syndrome to “anaphylactoid
syndrome of pregnancy” because of the similarities
between anaphylaxis and amniotic fluid embolism. An
allergic reaction requires sensitization to an antigen
before the allergic response can occur. Seventy-five
percent of the women with amniotic fluid embolism
were multiparous at the time of delivery and therefore
could have been previously exposed to fetal antigens.
The exact progression of maternal signs and symp-
toms in the initial phase of amniotic fluid embolism has
been difficult to identify because of the rarity of the
condition and lack of monitoring from the onset. Fava
and Galizia17 described a case of amniotic fluid embo-
lism that occurred during cesarean using general anes-
thesia in which the patient’s oxygen saturation de-
creased from 100% to 70% within 30 seconds. The
976 Gilbert and Danielsen Amniotic Fluid Embolism
hypoxia was followed within seconds by an initial
increase in blood pressure, with subsequent decrease to
hypotensive levels, suggestive of left atrial and ventric-
ular failure.17 They concluded that an initial pulmonary
vasospasm caused the hypoxia and subsequent cardio-
vascular collapse, and treating the hypoxia and hemo-
dynamic factors was important for the patient to sur-
vive.17 Disseminated intravascular coagulation was a
common finding in most of the women with amniotic
fluid embolism (Table 2). The exact mechanism by
which DIC develops in cases of amniotic fluid embo-
lism is not known. It is unlikely that DIC is a component
of an allergic or anaphylactic reaction because normally
it does not occur with either. Tissue factor, a primary
biologic initiator of coagulation, is found in increasing
amounts in amniotic fluid as gestation advances.18
Lockwood et al18 postulated that the large quantities of
active tissue factor in amniotic fluid could explain the
changes in coagulation accompanying amniotic fluid
embolism.
References
1. Steiner PE, Luschbaugh CC. Maternal pulmonary embolism by
amniotic fluid. JAMA 1941;117:1245–54, 1341–5.
2. Morgan M. Amniotic fluid embolism. Anaesthesia 1979;34:20 –32.
3. Plauche WC. Amniotic fluid embolism. Am J Obstet Gynecol
1983;147:982–3.
4. Clark SL, Hankins GDV, Dudley DA, Dildy GA, Porter TF.
Amniotic fluid embolism: Analysis of the national registry. Am J
Obstet Gynecol 1995;172:1158 – 69.
5. Clark SL. New concepts of amniotic fluid embolism: A review.
Obstet Gynecol Surv 1990;45:360 – 8.
6. Roche WD, Norris HJ. Detection and significance of maternal
pulmonary amniotic fluid embolism. Obstet Gynecol 1974;43:729 –
31.
7. Liban E, Raz S. A clinicopathologic study of fourteen cases of
amniotic fluid embolism. Am J Clin Pathol 1969;51:477–52.
8. Burrows A, Khoo SK. The amniotic fluid embolism syndrome: 10
years’ experience at a major teaching hospital. Aust N Z J Obstet
Gynaecol 1995;35:245–50.
9. Pan Y, Li X. Analysis of 29 maternal deaths caused by amniotic
fluid embolism. Chin J Obstet Gynecol 1995;30:349 –51.
10. Lau G. Amniotic fluid embolism as a cause of sudden maternal
death. Med Sci Law 1994;34:213–20.
11. Lau G, Chui PP. Amniotic fluid embolism: A review of 10 fatal
cases. Singapore Med J 1994;35:180 –3.
12. Gilbert WM, Nesbitt TS, Danielsen B. Childbearing beyond age 40:
Pregnancy outcome in 24,032 cases. Obstet Gynecol 1999;93:9 –14.
13. Graham HK. Amniotic fluid embolism. Am J Obstet Gynecol
1955;69:905–10.
14. Benson MD, Lindberg RE. Amniotic fluid embolism, anaphylaxis,
and tryptase. Am J Obstet Gynecol 1996;175:737.
15. Benson MD. Nonfatal amniotic fluid embolism: Three possible
cases and a new clinical definition. Arch Fam Med 1993;2:989 –94.
16. Azegami M, Mori N. Amniotic fluid embolism and leukotrienes.
Am J Obstet Gynecol 1986;155:1119 –24.
17. Fava S, Galizia AC. Amniotic fluid embolism. Br J Obstet Gynaecol
1993;100:1049 –50.
18. Lockwood CJ, Bach R, Guha A, Zhou X, Miller WA, Nemerson Y.
Obstetrics & Gynecology
 Amniotic fluid contains tissue factor, a potent initiator of coagu- Received July 21, 1998.
lation. Am J Obstet Gynecol 1991;165:1335– 41. Received in revised form November 18, 1998.
Accepted November 25, 1998.

Address reprint requests to:

William M. Gilbert, MD
University of California, Davis
Department of Obstetrics/Gynecology
4860 Y Street, Suite 2500
Sacramento, CA 95817 Copyright © 1999 by The American College of Obstetricians and
E-mail: wmgilbert@ucdavis.edu Gynecologists. Published by Elsevier Science Inc.

STANDARDS FOR REPORTING TRIALS

The CONSORT listing of standards for reporting randomized trials has been
adopted as policy by Obstetrics & Gynecology. Investigators who are planning,
conducting, or reporting randomized trials should be thoroughly familiar with these
standards. A copy can be obtained by contacting: Obstetrics & Gynecology, 10921
Wilshire Boulevard, Suite 403, Los Angeles, CA 90024-3908; FAX (310) 208-2838.

VOL. 93, NO. 6, JUNE 1999 Gilbert and Danielsen Amniotic Fluid Embolism 977