Low-molecular-weight heparin or warfarin for anticoagulation in pregnant women with mechanical heart valves: what are the risks? A retrospective observational study


Dr S Basude, Department of Obstetrics, St Michael’s Hospital, Southwell Street, BS2 8EG, Bristol, UK. Email sneha.basude@uhbristol.nhs.uk


Please cite this paper as: Basude S, Hein C, Curtis S, Clark A, Trinder J. Low-molecular-weight heparin or warfarin for anticoagulation in pregnant women with mechanical heart valves: what are the risks? A retrospective observational study. BJOG 2012;119:1008–1013.

The management of anticoagulation in pregnant women with mechanical heart valves is complex. The maternal and fetal outcomes of 32 pregnancies in 15 women on three different anticoagulation regimens were compared. Anticoagulation with low-molecular-weight heparin (n = 4), warfarin (n = 22) and combination therapy (n = 6) resulted in adverse maternal events in four (100%), three (50%) and three (14%) women, and resulted in fetal losses in one (25%), 17(77%) and three (50%) pregnancies, respectively. Whereas the rate of fetal loss in the warfarin group was high, all women in the LMWH and half of those in the combination group had serious adverse maternal events, including valve thrombosis, maternal death and postpartum haemorrhage.


The management of pregnant women with mechanical heart valves is complex and problematic. Patients with prosthetic mechanical valves are at risk of thromboembolic complications, including thrombosis of the prosthetic valve with resultant valve obstruction, and/or regurgitation and systemic embolisation. The risk of thrombosis is influenced by the site and type of mechanical valve, and by the presence of atrial fibrillation, left ventricular dysfunction and previous thromboembolism.1 This risk is further increased by the hypercoagulable state induced by pregnancy. Selecting the optimal anticoagulant therapy for the mother that has minimal adverse effects on the fetus is difficult.

Warfarin has been advocated as the anticoagulant of choice in non-pregnant patients with mechanical valves.2 However, warfarin crosses the placenta, and can therefore adversely affect the fetus by inducing embryopathy and fetal loss.3 There is also a risk of haemorrhage at any stage in the pregnancy. The alternative strategy, which protects the fetus, is to use heparin for anticoagulation throughout pregnancy. However, unfractionated heparin has been shown to result in unacceptably high maternal morbidity.4 Low-molecular-weight heparin (LMWH), used with or without aspirin (75–100 mg), has been increasingly used as the anticoagulant of choice in pregnant women, but with limited safety data. Various regimens have been proposed to optimise the use of LMWH in pregnant women with mechanical heart valves.4 It has been suggested that LMWH, with dose adjustment to maintain the anti-Xa levels between 1.0 and 1.2 iu/l, provides effective anticoagulation in pregnant women with mechanical heart valves.5

The aim of this study was to compare the incidence of adverse maternal and fetal outcomes in women with mechanical heart valves on the three different anticoagulation regimens in use at our centre.


All pregnancies in women with mechanical heart valves, seen in the combined cardiac antenatal clinic at University Hospitals Bristol between January 2003 and September 2011, were identified using cardiac and obstetric medicine databases and included in the study. Case notes and laboratory results were reviewed.

In the combined cardiac antenatal clinic, the women were counselled, either pre-conception or upon referral, regarding the maternal and fetal risks associated with different anticoagulation regimens. Warfarin was advised as the drug of choice for maternal safety in all cases, based on the evidence at that time, regarding the efficiency of LMWH in preventing thrombosis. Perceived compliance, lifestyle, valve position and type were all taken into account when counselling the women regarding anticoagulant options (Table S1). The women were made aware of the increased association of warfarin with fetal loss. The final decision was based on patient preference.

A choice of three treatment options was offered:

  •  regimen 1 – LMWH and low-dose aspirin (LMWHa);
  •  regimen 2 – warfarin;
  •  regimen 3 – combination (LMWHa from 6 to 13 completed weeks of gestation, followed by warfarin substitution for the majority of the pregnancy until 36 weeks of gestation, when LMWH was then resumed).

Low-molecular-weight heparin (LMWH) was initially administered as a twice-daily dose of 1 mg/kg. The dose was adjusted based on peak anti-Xa levels (4–6 h after administration of LMWH).1 Weekly monitoring and dose adjustment was carried out until optimum anti-Xa levels were achieved (1–1.2 iu/l), followed by two to four weekly monitoring thereafter. Women in the warfarin group continued to take warfarin throughout the pregnancy until 36 weeks of gestation, when warfarin was substituted with LMWH to reduce the risk of haemorrhagic complications. The warfarin dose was adjusted to maintain a valve-specific target international normalised ratio (INR) of 2.5–3.5. The women in the combination group were monitored, dependent on the anticoagulant, with dose adjustment as described above.

Data were obtained on type and site of mechanical valve, deterioration of valve function or ventricular function during pregnancy, anticoagulation regimen and dose, INR (whilst on warfarin) and peak anti-Xa levels (whilst on LMWH). Notes were reviewed to obtain information on various maternal risk factors. Women were assessed as ‘compliant’ with anticoagulation therapy if they reported meticulous self-dosing and if continuous therapeutic anti-Xa/INR levels were obtained (or if an immediate anticoagulant dose increase occurred, following subtherapeutic levels).

Maternal outcome measures included: thromboembolic events, deterioration of valve or ventricular function, significant haemorrhage (of more than 1000 ml) and death. Primary haemorrhage was defined as bleeding occurring within 24 hours of delivery or miscarriage, and secondary haemorrhage was defined as bleeding occurring after 24 hours and up to 12 weeks post delivery. Fetal/neonatal outcome measures included: livebirth rate, miscarriage, stillbirth and neonatal death.


A total of 32 pregnancies in 15 women were identified. The mean age of the women was 30 years (21–37 years). All but two pregnancies were in white women (n = 30), with one pregnancy in an Asian woman and the other in an Afro-Caribbean woman. Of these pregnancies, 44% (n = 14) occurred in women with aortic valves, 37.5% (n = 12) occurred in women with mitral valves, 12.5% (n = 4) occurred in women with tricuspid valves, and 6% (n = 2) had both aortic and mitral mechanical valves. The most common prosthetic valve was the bileaflet type (14 St Jude, 11 Carbomedics and five Sorin). One woman (who had two pregnancies) had tilting disc valves (Medtronic-Hall).

The LMWHa regimen was chosen by 13% (n = 4), the warfarin regimen was chosen by 69% (n = 22) and the combination regimen was chosen by 19% (n = 6) of the women. LMWHa was commenced by 4–6 weeks of gestation in all four women on regimen 1. All four women in the LMWHa group were compliant. Target anti-Xa levels were achieved by 5–16 weeks of gestation. Compliance in the warfarin group was 95% (21/22). Only 66% (4/6) of women in the combination group were compliant.

Maternal outcome

Morbidity and mortality

There were ten adverse maternal events (Table 1): a maternal death, three thrombotic complications, five haemorrhagic complications and a deterioration of ventricular function in one woman. The maternal death was in the LMWHa group. She had a mitral valve replacement (Sorin) 4 years prior to the index pregnancy. She was known to be heterozygous for the prothrombin gene mutation and had a body mass index (BMI) of 36. This was her fourth pregnancy (previously, she had three-first-trimester miscarriages whilst on warfarin 6 mg). The LMWHa regimen was commenced at 5 weeks of gestation. She had weekly monitoring of peak anti-Xa levels, which were all between 1.0 and 1.3 iu/l. The peak anti-Xa levels taken 6 and 4 days prior to her death were 1.0 and 1.3 iu/l, respectively (Enoxaparin 100 mg, twice daily). At 7 weeks of gestation she presented to the emergency department with a fatal parietal intracerebral haemorrhage. The postmortem examination did not identify any specific vascular cause for the intracerebral bleed.

Table 1.   Fetal and maternal outcome in the three groups of anticoagulation treatment
  1. *Valve thrombosis.

  2. **TIA, Transient ischemic attack.

  3. ***First-trimester loss secondary to maternal death.

  4. The rows in Bold are for the total figures for maternal and fetal adverse outcomes whilst the normal font is the broken down data.

Total pregnancies n (%)4 (12.5%)22 (68.75%)6 (18.75%)
Total maternal adverseevents n (%)4(100%)3(13.6%)3(50%)
Maternal mortality100
Haemorrhage (>1000 ml)131
Worsening cardiac function001
Adverse fetal outcomen (%)1***(25%)17(77%)3(50%)
First-trimester loss1***161
Second-trimester loss012
Third-trimester loss000


Thrombotic events occurred in three pregnancies: of these, two were in the LMWHa group and one was in the combination group, whilst taking LMWHa (Table 1). One of the women in the LMWHa group had a Carbomedics (27-mm) mitral valve. Enoxaparin was increased from 60 mg (1 mg/kg) to 90 mg twice daily to achieve therapeutic anti-Xa levels. The patient reached 16 weeks of gestation by the time a consistent anti-Xa level of more than 1.0 iu/l was achieved. This woman presented in premature labour with pulmonary oedema secondary to mitral valve thrombosis at 35 weeks of gestation. Vaginal delivery was expedited, and an emergency mitral valve replacement was performed in the immediate postnatal period.

The second woman had Ebstein’s anomaly, with a St Jude (35-mm) tricuspid valve. She was compliant with LMWH and all anti-Xa levels were within the therapeutic range. She was delivered at 34 weeks of gestation because of increasing tricuspid velocities and a new diastolic murmur: postpartum fluoroscopy showed that the valve leaflets were not moving, and she underwent repeat tricuspid valve replacement at 5 days postpartum. Upon surgery the valve was found to have significant pannus ingrowth and fresh thrombosis adherent to it.

The woman in the combination group had a Carbomedics (21-mm) aortic valve. She had a transient ischaemic attack at 9 weeks of gestation while on LMWH. She confirmed that she had not been compliant with the administration of aspirin. Following a transient ischemic attack (TIA) she was recommenced on warfarin and miscarried 2 weeks later.

Deterioration of cardiac function

One woman in the combination group had a St Jude (21-mm) aortic valve. She did not attend many of her hospital appointments and was difficult to monitor. She was induced at 35 weeks of gestation because of clinical and echocardiogram evidence of deterioration in left ventricular function, with a concomitant drop in peak velocity across the prosthesis. Cardiac function improved with supportive care following delivery and there was no evidence of valve thrombosis.


Significant haemorrhage occurred in five pregnancies. Of these, one occurred in the LMWHa group, three in the warfarin group and one in the combination group (Table 1). The three cases of primary haemorrhage in the warfarin group occurred in the same patient in each of her three pregnancies (one vaginal delivery and two-first-trimester miscarriages). She had a St Jude mitral valve and was taking 5 mg warfarin during each pregnancy. Two patients had a secondary haemorrhage: one occurred 6 days after normal vaginal delivery following the induction of labour at 38 weeks of gestation in the LMWHa group, on the second day of warfarin treatment; and the other occurred in the combination group after a second-trimester loss at 17 weeks of gestation, whilst on warfarin.

Fetal outcome

Miscarriage occurred in one pregnancy in the LMWHa group (secondary to maternal death), 77% (17/22) of pregnancies in the warfarin group and 50% (3/6) of pregnancies in the combination group (Table 1). The daily warfarin doses taken by the women (in the warfarin and combination groups) ranged from 4 to 13 mg. Only 21% of women on warfarin (6/28 women) were on less than or equal to 5 mg of warfarin, although none of these were in the combination group. The fetal loss in this group was 66%. In the women taking 6–9 mg of warfarin, the fetal loss was 72% in the warfarin group (n = 11) and 25% in the combination group (n = 4). The fetal loss rate was 100% in women taking 10 mg or more of warfarin, in both the warfarin group (n = 5) and the combination group (n = 2). There were three-second-trimester miscarriages: one in the warfarin group (6 mg) and two in the combination group.

Of the three loses in the combination group, one was in the first trimester when the woman was recommenced on warfarin following the diagnosis of a transient ischaemic attack. The two losses in the second trimester were at 17 and 18 weeks of gestation (whilst on 10- and 13-mg warfarin doses, respectively). Only one of these fetuses underwent a postmortem, which showed intraventricular haemorrhage and multiple periventricular haemorrhages. There were no fetal losses after 20 weeks of gestation in our study.

Discussion and conclusions

Our data demonstrate a high rate of both maternal and fetal adverse outcomes in women with mechanical heart valves, despite evidence of therapeutic anticoagulation.

Of most concern is the high rate of serious adverse maternal events in the LMWHa group. All women in this group suffered significant morbidity or mortality (one maternal death from intracerebral haemorrhage, two valve thromboses needing emergency valve replacement and one postpartum haemorrhage). This was in spite of evidence of compliance and therapeutic anti-Xa levels, and with the additional use of aspirin.

Based on previous evidence, women with mechanical valves have been advised that the best anticoagulant in terms of preventing maternal complications is warfarin, but at the cost of increased fetal loss.1,6 Although LMWH is safe for the fetus, there have been significant concerns regarding maternal safety (9% risk of valve thrombosis).1 More recent reports claim that LMWH is safe for pregnant women receiving therapeutic anticoagulation.5,7 Quinn 5reported no adverse maternal events in all but one of the 12 pregnancies in women on LMWH with therapeutic peak anti-Xa levels, and attributed the one adverse event to subtherapeutic anti-Xa levels. McLintock 7reported five thrombotic events in 34 women, and attributed all of these to subtherapeutic anti-Xa levels or non-compliance. Whereas these reports are encouraging, our data suggests that clinicians should only proceed to advise LMWH with caution. This advice is shared by other authors.1

Currently, monitoring the peak antithrombotic activity using anti-Xa levels is recommended to ensure therapeutic anticoagulation with LMWH.1 There are some crucial points to consider. Firstly, the correct timing of venepuncture following LMWH administration is important for interpretation, but can be difficult to achieve in practice, especially in the community setting. Secondly, there is limited evidence regarding assurance of therapeutic levels in pregnancy. Barbour 8demonstrated subtherapeutic pre-dose anti-Xa levels in pregnant women with therapeutic peak anti-Xa levels. It may therefore be important to monitor and maintain pre-dose as well as peak anti-Xa levels in the therapeutic range. However, there was a fatal haemorrhagic complication in the LMWHa group in our study, and therefore expert consensus would be of benefit prior to recommending minimum trough levels. Thirdly, LMWH has anti-Xa and anti-IIa activities, which can vary in different preparations, and therefore measuring anti-Xa levels alone may not be sufficient to measure the antithrombotic activity.9 Although anti-Xa assays may provide some clue to LMWH pharmacokinetics in individual subjects, only limited information on the antithrombotic effect and bleeding risk can be deduced from this measurement.

Our data show that the potential maternal benefits of warfarin are accompanied by a 77% fetal loss rate, which is higher than recently published data.3 The losses were mostly in the first trimester and may reflect the intense surveillance of patients from an early gestation in our study. The recent European Society of Cardiology Guidelines on Management of Cardiovascular Disease in Pregnancy recognises the risk of warfarin-associated early fetal loss, and recommends the use of the combination regimen.1 It is interesting to note that of the three fetal losses in the combination group in our study, two were in the second trimester, after recommencing warfarin. Furthermore, 50% of the women in the combination group had adverse maternal events (one worsening ventricular function, one haemorrhage and one transient ischaemic attack). This raises questions about the safety and efficacy of the combination regimen.

There is no ideal anticoagulant for pregnant women with mechanical heart valves. Whereas warfarin is associated with a high fetal loss rate, our results indicate that serious adverse maternal events occur in women taking LMWH and aspirin, despite attaining therapeutic peak anti-Xa levels. Our data are observational, and the potential for confounding cannot be excluded. The conclusions are limited by the small number of women in this study, and therefore further data are required, regarding means of monitoring, therapeutic levels, and safety and efficacy, to enable appropriate counselling for these women. Ideally, a multicentre randomised trial is needed to provide good-quality evidence. However, because of the consequences of the Heparin in Pregnancy-Cardiac Valve Thromboprophylaxis (HIP-CAT) trial,10 it will be difficult to undertake a randomised controlled trial of LMWH in pregnant women with mechanical valves, but data could be pooled between units, preferably recording peak and pre-dose levels to allow larger observational studies to be performed.

These data should be considered, not just for informing women regarding anticoagulant choice when pregnant, but also prior to pregnancy when valve replacement surgery is planned in young women. Cardiologists and surgeons should be aware of these risks when recommending replacement with a mechanical valve in women of childbearing age.

Disclosure of interests

None to declare.

Contribution to authorship

SB was responsible for assessing data and writing the article. CH was responsible for collecting the data. SC assisted in data collection and advised on article preparation. AC reviewed the article. JT was responsible for the initial project idea, advised on the article and supervised the project.

Details of ethics approval

This project falls under the definition of ‘service evaluation’ by the National Research Ethics Service (NRES) standards, and is therefore not subject to ethics approval: Defining Research, http://www.nres.nhs.uk/applications/is-your-project-research/


None required.


We would like to acknowledge Graham Stuart (Consultant Cardiologist in Adult Congenital Heart Disease) and Susan Sellers (Consultant Obstetrician) for initiating the cardiac obstetric service at the University Hospitals Bristol, and for providing the clinical care for many of these patients.

Commentary on ‘Reconsidering the counseling we provide to women concerning anticoagulation in pregnancy with mechanical heart valves’

The article by Basude et al. brings to light once more the paucity of information we have on anticoagulation in women with mechanical cardiac valves during pregnancy, and the large differences in practice between teams.

Using the right anticoagulation during pregnancy is a complicated issue. Basude et al. present their surprisingly high rate (100%) of maternal adverse outcome in women treated with low-molecular-weight heparin (LMWH), although anti-Xa activity was monitored and kept within optimal ranges. Women on warfarin also experienced a high rate of adverse outcome, mainly fetal. The bottom line is an overall poor outcome: their results seem to indicate that there is no ideal anticoagulation regimen for pregnant women with mechanical valves.

This problem needs to be addressed on a larger scale. Adult cardiology has greatly improved the life of women with mechanical valves, allowing them to have normal adult lives with full-time jobs, and to conceive pregnancies.

It is therefore time to reconsider pregnancy outcome with modern treatments, and find the correct balance between counselling for termination of pregnancy and being over-confident with pregnancy outcome. We need data to inform such discussions. The series by Basude et al. should lead other centres to review their cases and publish them, including a detailed description of their anticoagulation management, to increase our understanding, to enable us to provide better counselling, and ultimately to improve the management of these women before and during pregnancy.

Disclosure of interests

Nothing to disclose.

J Nizard

Department of Obstetrics and Gynaecology, GH Pitié Salpêtrière – UPMC

Paris 6, Paris, France