Chan WS, Anand S, Ginsberg JS. Anticoagulation of Pregnant Women With Mechanical Heart ValvesA Systematic Review of the Literature. Arch Intern Med. 2000;160(2):191–196. doi:10.1001/archinte.160.2.191
The management of women with prosthetic heart valves during pregnancy poses a particular challenge as there are no available controlled clinical trials to provide guidelines for effective antithrombotic therapy. Oral anticoagulants such as warfarin sodium cause fetal embryopathy; subcutaneous administration of heparin sodium has been reported to be ineffective in preventing thromboembolic complications.
To identify the risks of maternal and fetal complications in women with mechanical heart valves treated with different anticoagulation regimens during pregnancy.
We performed a systematic review of the literature to determine pooled estimates of maternal and fetal risks associated with the 3 commonly used approaches: (1) oral anticoagulants (OA) throughout pregnancy, (2) replacing OA with heparin in the first trimester (from 6-12 weeks' gestation), and (3) heparin use throughout pregnancy. Fetal outcomes included spontaneous abortions and fetopathic effects, and maternal outcomes were major bleeding, thromboembolic complications, and death.
The use of OA throughout pregnancy is associated with warfarin embryopathy in 6.4% (95% confidence interval [CI], 4.6%-8.9%) of livebirths. The substitution of heparin at or prior to 6 weeks, and continued until 12 weeks, eliminated this risk. Overall risks for fetal wastage (spontaneous abortion, stillbirths, and neonatal deaths) were similar in women treated with OA throughout, compared with women treated with heparin in the first trimester. Maternal mortality was 2.9% (95% CI, 1.9%-4.2%). Major bleeding events occurred in 2.5% (95% CI, 1.7%-3.5%) of all pregnancies, most at the time of delivery. The regimen associated with the lowest risk of valve thrombosis (3.9%; 95% CI, 2.9-5.9%) was the use of OA throughout; using heparin only between 6 and 12 weeks' gestation was associated with an increased risk of valve thrombosis (9.2%; 95% CI, 5.9%-13.9%).
Thromboembolic prophylaxis of women with mechanical heart valves during pregnancy is best achieved with OA; however, this increases the risk of fetal embryopathy. Substituting OA with heparin between 6 and 12 weeks reduces the risk of fetopathic effects, but with an increased risk of thromboembolic complications. The use of low-dose heparin is definitely inadequate; the use of adjusted-dose heparin warrants aggressive monitoring and appropriate dose adjustment. Large prospective trials to determine the best regimen for these women are needed.
THE MANAGEMENT of pregnant women with mechanical heart valves is problematic because there are no controlled clinical trials to provide guidelines for optimal antithrombotic therapy. Regimens that have been recommended by experts include (1) warfarin sodium throughout pregnancy with unfractionated heparin sodium near term, (2) substitution of warfarin with unfractionated heparin between 6 and 12 weeks of gestation and near term, and (3) unfractionated heparin throughout pregnancy. Evidence supporting the use of these regimens is poor and is derived from case reports, case series, small cohort studies, and questionnaires.
There are concerns about the use of any of the 3 regimens. There is a reluctance to use warfarin because of the concern about adverse outcomes including warfarin embryopathy, other fetopathic effects, and spontaneous abortions.1,2 Using heparin alone during pregnancy eliminates fetal risks; however, multiple case reports of thrombosed valves causing maternal morbidity and mortality have raised serious concerns.3- 6 One plausible explanation for the failure of heparin therapy in these women is the use of inadequate doses. However, the possibility that unfractionated heparin is ineffective in pregnant women with prosthetic heart valves must be considered.
We performed a systemic overview to estimate the effectiveness and safety of these 3 regimens for the fetus and the mother. The primary objective of this review was to estimate maternal and fetal risks associated with these anticoagulation regimens. The maternal outcomes of interest included thromboembolic complications (TEC), major bleeding, and death, and fetal outcomes of interest included warfarin embryopathy, other fetopathic effects, and death.
A MEDLINE and EMBASE search was conducted of articles published between 1966 and October 1997 using MESH headings "prosthetic heart valves," "pregnancy," and "anticoagulation." We found a total of 40 relevant English-language articles consisting of questionnaires, case reports, case series, and cohort studies.3- 42 Thirty-five of these articles were obtained from the search3- 6,12- 42; the references of these articles were reviewed and 5 further articles were retrieved.7- 11
Articles were potentially eligible for analysis if they (1) were case series (≥6 patients) within an institution consisting of consecutive patients or cohort studies of pregnancies in women with mechanical heart valves, (2) defined the anticoagulation regimen used, and (3) reported outcomes of interest (see below). We excluded studies if they (1) were case reports or (2) did not follow all pregnancies to completion.
The major outcomes of interest were the frequencies of serious maternal and fetal events. Maternal events of interest included TEC, bleeding, and death. Fetal outcomes included warfarin embryopathy, other congenital abnormalities, spontaneous abortions, and fetal wastage (spontaneous abortions, stillbirths, and neonatal deaths).
The regimens were classified as follows: regimen 1, oral anticoagulants (OA) alone (including all warfarin derivatives), with or without heparin prior to delivery; regimen 2, OA substituted with heparin during the first trimester of pregnancy, with or without heparin near term; this regimen was further subdivided into (a) heparin substituted at or before 6 weeks of gestation and (b) heparin substituted after 6 weeks of gestation; regimen 3, heparin throughout pregnancy; and regimen 4, no anticoagulants, which included use of antiplatelet agents alone.
We divided all the eligible studies into "retrospective" or "prospective" depending on whether the cases of pregnancy in women with mechanical heart valves were collected in a retrospective (ie, chart reviews) or prospective manner (ie, longitudinal follow-up of patients). The rationale for such a division was to derive more accurate estimates of various complications from methodologically superior prospective studies.
All pregnancies were classified according to the anticoagulation regimen used and adverse outcomes in the eligible studies were pooled. The rates of adverse events associated with these regimens were expressed as proportions of total pregnancies. Fetal congenital anomalies were expressed as a proportion of livebirths. Maternal deaths from all causes were expressed as proportions of women at risk.
The reported frequencies of various outcomes of interest from different studies were pooled and weighted by the size of each study if tests of homogeneity revealed absence of heterogeneity. If heterogeneity existed among the different proportions, a more conservative approach that took into account differences among studies described by DerSimonian and Laird43 was used to calculate the appropriate 95% confidence intervals (CIs).
Of the 40 articles retrieved, 25 were case series (≥6 patients per study), 6 were cohort studies, 3 were questionnaires, and 6 were case reports. The 3 questionnaires,12- 14 6 case reports,3- 6,15,16 and 3 case series10,17,18 that did not fulfill the eligibility criteria were excluded from further analysis. Therefore, 28 articles (6 cohort studies and 22 case series) were included in the final analysis.7- 9,11,19- 42 Patients in 8 studies were followed up longitudinally22,23,28,30,32,35,36,42 and these studies were designated as prospective.
There were 976 women with 1234 pregnancies included in the studies conducted between 1966 to October 1997. The types and positions of the mechanical heart valves are presented in Table 1. A total of 433 women (49.7%) had cage and ball valves, more commonly the Starr-Edwards valve. The less thrombogenic and newer bi-leaflet valves, St Jude and Duromedics, were present in only 62 patients (7.1%). About two thirds (647/976) of the valves were in the mitral position.
Twenty-four studies7- 9,11,19- 31,33- 35,37,38,40,41 reported using oral anticoagulants, most commonly warfarin (Coumadin), throughout pregnancy (regimen 1). Twelve studies9,22,24,28,30,32,34- 36,38,39,42 reported the substitution of heparin in the first trimester of pregnancy (regimen 2). Five studies19,20,26,38,41 reported management of some of these pregnancies with heparin alone throughout pregnancy (regimen 3); in 2 studies19,20 adjusted-dose heparin was used, but the target activated partial thromboplastin time was not reported, and in 3 studies,26,38,41 low-dose heparin (≤15,000 U/d) was used. Seven studies8,11,20,23,24,33,41 reported no anticoagulation use throughout pregnancy, and 2 studies35,41 reported the use of antiplatelet agents only, consisting of aspirin and dipyridamole. Twenty studies8,9,19- 22,24,26- 31,33- 35,37,38,40,41 reported switching to heparin for labor and delivery. Target prothrombin times were reported in 13 studies; these varied from international normalized ratios of 2.0 to 4.0, to prothrombin time ratios of 1.5 to 2.5.
The frequencies of adverse fetal events with different anticoagulation regimens from all eligible studies are summarized in Table 2. When OA was used throughout pregnancy (regimen 1), spontaneous abortions resulted in 24.7% (196/792) of pregnancies. When heparin was substituted for OA in the first trimester (regimen 2), spontaneous abortions resulted in 24.8% (57/230) of pregnancies. Regimen 3 (use of heparin throughout) was associated with spontaneous abortions in 23.8% (5/21) of pregnancies, and regimen 4 (no anticoagulation or antiplatelet agent use alone) resulted in spontaneous abortions in 9.8% (10/102) of pregnancies. Within regimen 2, when warfarin was discontinued and heparin used at or prior to 6 weeks' gestation (regimen 2a), 14.7% of pregnancies (19/129) resulted in spontaneous abortions compared with 33.9% (19/56) if warfarin was replaced with heparin after 6 weeks of gestation (regimen 2b). The number of pregnancies reporting use of heparin throughout pregnancy was small; there were a total of 21 pregnancies, of which 16 involved the use of adjusted-dose heparin (regimen 3a) and 5 involved the use of low-dose heparin (regimen 3b). The rates of spontaneous abortions were 25.0% (4/16) and 20.0% (1/5), respectively.
The prevalence of congenital fetal anomalies in the livebirths associated with regimen 1 was 6.4% (35/549). The prevalence was 3.4% (6/174) if heparin was substituted in the first trimester; however, if warfarin was replaced by heparin at or prior to 6 weeks, there were no (0/108) fetal anomalies, compared with 11.1% (4/36) if heparin was replaced after 6 weeks. No congenital anomalies (0/17) were reported with regimen 3; and 3 cases (3/92) of congenital fetal anomalies reported with regimen 4 (bilateral accessory auricles, curved fifth finger, and unspecified multiple anomalies).
Regimen 1 resulted in an overall risk for fetal wastage of 33.6% (266/792) compared with 26.5% (61/230) for regimen 2. When warfarin was discontinued and heparin was used at or prior to 6 weeks, the risk fell to 16.3% (21/129), whereas when warfarin was discontinued and heparin used after 6 weeks, the risk increased to 35.7% (20/56). The prevalence of fetal wastage resulting from regimens 3 and 4 were 42.9% (9/21) and 19.6% (20/102), respectively.
The most common congenital fetal anomaly associated with OA use (regimens 1 and 2) was the characteristic warfarin embryopathy (nasal hypoplasia and epiphysis stippling) (29 of 41 livebirths). Central nervous system anomalies (hydrocephalus, mental retardation) were reported in 4 cases, cleft lip and cleft palate in 4, and other anomalies in 4 (single kidney-toe-finger deformity, left ventricular hypoplasia, corneal leukoma, and bilateral hand polydactyly).
Subgroup analyses of prospective studies are summarized in Table 3. The use of OA throughout pregnancy resulted in spontaneous abortion in 21.1% (52/247) of pregnancies vs 15.0% (19/127) when warfarin was discontinued and heparin used between 6 and 12 weeks of gestation, and 35.8% (19/53) when warfarin was discontinued and heparin started after 6 weeks. Congenital fetal anomalies were seen in 10.2% (14/137) when OA was used throughout pregnancy, in 0.0% (0/108) when warfarin was discontinued and heparin used at or prior to 6 weeks, and 12.1% (4/33) when warfarin was used beyond 6 weeks. Similar trends were repeated with overall risk of fetal wastage; OA use resulted in 29.7% (73/246), warfarin discontinued by 6 weeks of gestation resulted in 16.3% (21/129), and its use after 6 weeks resulted in 37.7% (20/53) of pregnancies.
Important maternal complications (Table 4) associated with various regimens include TEC and death. Thromboembolic complications were reported in 3.9% (31/788) of pregnancies in women treated with regimen 1. When regimen 2 was used, TEC occurred in 9.2% (21/229) of pregnancies. When heparin alone was used, in low or adjusted doses (regimen 3), the frequency of TEC was 7 (33.3%) of 21 pregnancies, whereas when no anticoagulation therapy was used, the frequency was 26 (24.3%) of 107. In 29 of the 52 TEC reported with regimens 1 and 2, heparin was the anticoagulant used at the time the adverse event occurred.
The overall maternal mortality was 2.9% (25/854). Maternal deaths were reported in 1.8% (10/561) and 4.2% (7/167) of women receiving regimens 1 and 2, respectively. The maternal mortality rate was 15.0% (3/20) with regimen 3 and 4.7% (5/106) with regimen 4.
When the subgroup of women enrolled in prospective studies was analyzed separately, the data were consistent with the pooled analysis. Thromboembolic complications occurred in 1.6% (4/246; 95% CI, 0.5%-4.4%) of pregnancies associated with regimen 1 and 8.2% (15/183; 95% CI, 4.8%-13.4%) of pregnancies associated with regimen 2. Maternal death occurred in 0.5% (1/182; 95% CI, 0.0%-3.5%) of pregnancies associated with regimen 1, and 4.0% (5/126; 95% CI, 1.5%-9.5%) of those associated with regimen 2.
The pooled frequency of major bleeding complications reported from all studies was 31 (2.5%) of 1234 pregnancies, with 25 occurring at delivery and 6 outside delivery. Eleven of these 31 women were receiving intravenous or adjusted-dose heparin, 14 were receiving heparin, and the regimen was unknown for 6.
There were 25 maternal deaths. The causes of maternal death were thrombosis of the valve prostheses and related complications (17 women), subacute bacterial endocarditis (3 women), hemorrhage (2 women), unknown causes (2 women), and Eisenmenger syndrome (1 woman).
Several conclusions can be made from our systematic review of antithrombotic therapy during pregnancy in women with mechanical heart valves. The risk of fetopathic effects, predominantly warfarin embryopathy, when OA was used throughout pregnancy is approximately 6%. In addition, the frequency of fetopathic effects when warfarin was replaced by heparin between 6 and 12 weeks' gestation or with heparin therapy throughout is negligible.
The risk of fetal wastage appears to be similar with the use of OA throughout pregnancy, compared with the use of OA with heparin substituted in the first trimester. It also appears that the window between 6 and 12 weeks of gestation is pivotal, since the continued exposure to warfarin resulted in an increased risk of fetal wastage by more than 50%. This trend is seen in all studies as well as the subgroup analysis of prospective studies; the latter are more likely to report first-trimester events accurately. However, the apparent advantages (reduced fetopathic effects, reduced fetal wastage) gained from the use of heparin appear to be counterbalanced by an increase in maternal complications.
Indeed, pregnant women with mechanical heart valves face serious maternal complications. While receiving anticoagulation, the estimated risk of maternal hemorrhage is 2.5%, with most episodes (80%) occurring in association with delivery. The risk of valve thrombosis appears to be lowest with the use of OA throughout pregnancy (4%), and increases with the use of heparin. The observation that most reported valve thromboses during pregnancy occur during the use of heparin further raises the concern regarding the effectiveness of heparin in pregnant women with mechanical heart valves.
Finally, our review emphasizes the fact that pregnancy in a woman with a mechanical heart valve is not benign, and carries a significant risk of maternal mortality; our estimate is 1% to 4%, and the main cause of death is complications from valve thrombosis.
An analysis of this nature has obvious limitations. The internal validity of our analysis relies on the validity of studies conducted, the accuracy of reporting in individual studies, and the accuracy of data extraction from the articles. Outcomes of pregnancies with certain management regimens, like use of OA throughout, and heparin in the first trimester, were well represented; on the other hand, outcomes of pregnancies managed with heparin alone were few. One possible reason for this is our systematic rejection of case reports, which were often those reporting failure of heparin in preventing devastating maternal sequelae. Accurate assessments of the risks involved with the use of heparin in either low doses or adjusted doses cannot be made without any available large case series in the literature.
Many factors are obviously important in determining the risk of valve thrombosis. These factors include timing of valve replacement, valve type, position of the valve, presence of arrhythmias, left atrial size, history of previous thrombosis, number of mechanical valves, and adequacy of anticoagulation.44 Many of these factors, which are also potential sources of heterogeneity in our study, cannot be adequately dealt with in our analyses because these variables were not often described fully by the authors to allow us to perform subgroup analyses or investigate variability among studies.
Valve designs have changed significantly during the period covered by these studies. There is decreased thrombogenic potential associated with newer valve types, such as the bi-leaflet St Jude valve compared with the ball-and-cage Starr-Edwards valve type; the latter is overrepresented in our analysis, which could lead to an overestimation of the risk of thrombosis.
This study highlights the dilemma that physicians and pregnant women with mechanical heart valves face. While the risk and severity of warfarin embryopathy may not be as high as previously reported, the continued use of OA until close to term can result in a high rate of fetal wastage. The use of warfarin throughout pregnancy, however, appears to confer the greatest protection to the mother against valve thrombosis and death. The use of heparin in the first trimester prior to 6 weeks of gestation avoids fetal embryopathy, but exposes the women to a period of increased risk for thrombosis. Clearly, low-dose heparin is inadequate for thromboprophylaxis during pregnancy; it is unclear whether adjusted-dose heparin to prolong mid-interval activated partial thromboplastin time results to 1.5 to 2.5 times control is adequate. Indeed, Salazar et al36 reported failure of TEC prophylaxis on adjusted-dose heparin given up to every 6 to 8 hours. The hypothesis that further increasing the dosing of heparin by increasing frequency of injections, and by keeping mid-interval activated partial thromboplastin times higher than 1.5 to 2.5 times control cannot be substantiated by this study. The risk of TEC associated with heparin use throughout pregnancy in this study is likely exaggerated because of our small numbers; and how it is modified in patients with newer, less thrombogenic valves is not known. We strongly recommend that if unfractionated heparin is used, that it be used in high doses (35,000 to 40,000 U in 24 hours) and the dose be adjusted aggressively based on a mid-interval activated partial thromboplastin time between 2.0 and 2.5.45
Low-molecular-weight heparins are evolving as an option for the treatment of venous thromboembolism in pregnancy46 and are attractive for prophylaxis in women with prosthetic heart valves because they may produce less bleeding for an equivalent antithrombotic effect and are easier to use than unfractionated heparin. If this hypothesis is correct, then low-molecular-weight heparins could be used in these women. Currently, data on the effectiveness of low-molecular-weight heparins in thromboprophylaxis of patients with prosthetic heart valves is lacking. The use of antiplatelet agents alone is definitely not a reasonable alternative in these patients; their role in combination with anticoagulants has not been evaluated in pregnancy.47
The choice of anticoagulant regimens for mechanical heart valve thromboprophylaxis during pregnancy should be made by balancing 2 risks—maternal morbidity and mortality—from TEC and fetal loss and embryopathy. The practice of substituting OA for heparin in the first trimester potentially eliminates fetal embryopathy, but appears to subject woman to an increased risk of thromboembolism. We recommend careful counseling of women with prosthetic heart valves prior to or shortly after the diagnosis of pregnancy and discussion of the risks associated with available anticoagulant options.
Accepted for publication April 13, 1999.
Dr Chan is a recipient of a research fellowship from the Medical Research Council of Canada, Pharmaceutical Manufacturers Association of Canada; Dr Anand is a recipient of a Medical Research Council of Canada Clinician-Scientist Award; and Dr Ginsberg is a career investigator of the Heart and Stroke Foundation of Ontario.
Reprints: Jeffrey S. Ginsberg, MD, FRCPC, McMaster University Medical Centre, 1200 Main St W, Room 3W15, Hamilton, Ontario, Canada L8N 3Z5.