Risk of Warfarin During Pregnancy With Mechanical Valve Prostheses

Cotrufo, Maurizio MD, FECTS; De Feo, Marisa MD; De Santo, Luca S. MD; Romano, GianPaolo MD; Della Corte, Alessandro MD; Renzulli, Attilio MD, FECTS; Gallo, Ciro MD

Obstetrics & Gynecology:
Original Research

OBJECTIVE: To assess the determinants of pregnancy outcome in patients with mechanical valve prostheses anticoagulated with warfarin.

METHODS: Between January 1987 and January 2000, 52 patients with mechanical valve prostheses who had 71 pregnancies were anticoagulated with warfarin for the entire duration of pregnancy. Warfarin was withdrawn 48 hours before and 24 hours after a scheduled cesarean delivery carried out by the end of the 37th gestational week. The targeted international normalized ratio ranged between 2.25 and 4.0, depending on the prosthetic model. Exact univariate and multivariable analyses were performed to assess which among the following variables predicted poor pregnancy outcome: patient age, prosthetic model, site of implantation, average international normalized ratio, and average daily dose of warfarin.

RESULTS: Pregnancy loss occurred in 23 of 71 of pregnancies, stillbirth in five of 71, embryopathy in four of 71 (two aborted fetuses and two full‐term infants). There were no maternal deaths or thromboembolic or hemorrhagic complications. Warfarin daily dosage over 5 mg per day was a significant predictor of poor pregnancy outcome (P < .001).

CONCLUSION: The risk for pregnancy complications in patients treated with sodium warfarin is higher when the mean daily dose exceeds 5 mg.

In Brief

In pregnant women with mechanical heart valve prostheses treated with warfarin anticoagulation, the daily dose of warfarin correlates with fetal complications.

Author Information

Department of Cardio‐Thoracic and Respiratory Sciences, V. Monaldi Hospital, and Department of Public, Clinical and Preventive Medicine, Second University of Naples, Naples, Italy.

Address reprint requests to: Maurizio Cotrufo, MD, Largo Donn'Anna, 9, 80123 Naples, Italy. E‐mail: maurizio.cotrufo@unina2.it.

Received January 23, 2001. Received in revised form September 7, 2001. Accepted September 17, 2001.

Article Outline

Last‐generation mechanical heart valves have good hemodynamic performance and low rates of valve‐related complications. Although long‐term oral anticoagulation with coumarin derivatives is used safely to avoid throm‐boembolic complications in the general population, there are still concerns about its use in women of childbearing age.1,2

This topic has been a source of controversy. For many years mechanical valves and anticoagulation were considered a contraindication to pregnancy because of a high risk of maternal and fetal complications. Many authors originally suggested that bioprostheses could be used in young women.3 Later, anticoagulation was found to be necessary even in patients with mitral valve bioprostheses because of atrial fibrillation,4 and pregnancy was found to augment bioprosthetic structural malfunction.5 Parenteral heparin administration in the first trimester in patients with mechanical valves has been used6,7 with discouraging results: high rates of maternal complications such as embolism and prosthetic valve thrombosis were observed. Although venous thromboembolism, antiphospholipid antibody, and thrombophilia are commonly treated with low‐molecular‐weight heparin during pregnancy, this regimen proved ineffective in patients with mechanical heart valve prostheses, a condition with a much higher risk of thromboembolism.1,2 Recently protocols of low intensity anticoagulation with coumarin derivatives8,9 and studies suggesting that fetal adverse effects of anticoagulation might be dose‐dependent (Vitale N, De Feo M, Cotrufo M. Letters to the Editor. J Am Coll Cardiol 2000; 35:1365–6)10 led the American Heart Association to recommend warfarin administration from the first trimester to the 35th week of pregnancy. However, the package insert of coumarin derivatives still includes a statement contraindicating the administration in pregnant women with no specific recommendation regarding women with mechanical heart valves.

We conducted a retrospective investigation of maternal and fetal risk in women taking oral anticoagulants to assess the determinants of pregnancy outcome in patients with mechanical valves.

Back to Top | Article Outline


Between January 1987 and January 2000, 71 pregnancies were observed in 52 consecutive patients with mechanical heart valves. The age at the beginning of gestation ranged from 20 to 46 with a mean of 27.9 ± 6.0 years. Cardiac function before pregnancy was usually good, with all patients in New York Heart Association class I (41 cases) or II (11 cases).

Table 1 shows the distribution of the study population according to the type of prostheses implanted and the site of implantation. The valve model used depended on the surgeon's preference and generally reflected the current practice at the time of the operation.

As previously described (Vitale N, De Feo M, Cotrufo M. Letters to the Editor. J Am Coll Cardiol 2000;35: 1365–6),10,11 our anticoagulation protocol was based on sodium warfarin administration during pregnancy and remained unchanged throughout the study. Cesarean delivery was scheduled before the end of the 37th gestational week. Warfarin therapy was discontinued only 2 days before surgery and restarted 1 day after surgery. During this perioperative period, heparin was not administered, and international normalized ratios were checked daily.

The target international normalized ratio range was 2.25 to 4.0, depending on the type of prosthesis and site of implantation. It ranged between 4 and 3 for patients with first‐ or second‐generation prosthetic valves, whereas for third‐generation prostheses it ranged between 3.5 and 2.5 for aortic valves and between 3 and 2.25 for mitral valves. Throughout pregnancy, the international normalized ratio was estimated on a weekly basis at our outpatient clinic and recorded along with prescribed warfarin doses. Patients were followed up by cardiologists and obstetricians at monthly intervals until the 37th week of gestation, when they were electively hospitalized until delivery. Echocardiographic follow‐up was performed monthly to evaluate cardiac and prosthetic function. Ultrasound evaluations of the fetus were done at the third, fifth, and eighth months. Neonates underwent clinical examination soon after birth and at 4 and 12 months to ascertain or exclude the diagnosis of warfarin embryopathy. All miscarriages and stillbirths were clinically evaluated by neonatologists; no fetal pathologic or radiographic examination was performed.

Warfarin daily doses during pregnancy ranged between 2.5 and 10 mg per day with a mean of 5.81 ± 0.22 mg, and observed international normalized ratios ranged between 2.1 and 3.2 with a mean of 2.52 ± 0.025; international normalized ratio at the time of cesarean delivery ranged between 1.4 and 1.9.

Poor pregnancy outcome, defined as the occurrence of spontaneous abortion, stillbirth, or congenital birth defect, was the main end point. Predictive factors for poor pregnancy outcome were assessed in univariate and multivariable analyses of the following variables: patient age, prosthetic model, site of prosthesis implantation, average international normalized ratio, and average warfarin daily dose.

Units of analysis were pregnancies rather than women. We chose this unit because, when focusing on the effect of drug assumption on the fetus, each pregnancy should be viewed as a separate episode, irrespective of the outcome of previous pregnancies; that is, each fetus is at his own risk, even though pregnancies take place in the same woman. However, pregnancies occurring in the same woman are not completely independent, because they share some common risk factors, such as the site of implantation, the prosthetic model implanted, or preoperative abortion. Further, pregnancies after the first one were more likely to have occurred early in the study period because of longer time of exposure. Thus, to control for the possible confounding effect of repeated pregnancies, we first described differences between first and subsequent pregnancies, and eventually we adjusted for it in multivariable analysis.

Continuous variables (gestational age, average drug dosage, and international normalized ratio) were reported as mean and standard deviation (SD) and compared by Wilcoxon rank sum test. Categoric data (average drug assumption greater or less than 5 mg, prosthetic model, and site of implantation) were assessed by Fisher exact test. Analyses were performed by S‐Plus software (S‐PLUS 2000; MathSoft, Inc., Cambridge, MA, 1999). The effect of drug assumption on pregnancy outcome after adjustment by order of pregnancy (first of subsequent) was evaluated by exact logistic regression model using LogXact software to account for the small sample size (LogXact‐Turbo; CYTEL Software Corporation, Cambridge, MA, 1993).

Back to Top | Article Outline


There were 52 first pregnancies and 19 subsequent pregnancies (ten second pregnancies, four thirds, three fourths, and two fifths). Differences between order of pregnancies are reported in Table 2. Significant differences between first and subsequent pregnancies were found for warfarin dosage and international normalized ratio, with higher doses of warfarin in subsequent pregnancies, which pertained to women operated on in the early years of the study period, when protocols of anticoagulation required higher doses of warfarin. International normalized ratio values were slightly lower in pregnancies other than the first pregnancy. No significant differences were found for prosthetic model and site of implantation.

Of the 71 pregnancies, 30 poor outcomes were observed (42%) (23 spontaneous abortions, five stillbirths, and two embryopathies in full‐term infants [nasal hypoplasia and small ventricular septal defect]). Associations between study variables and pregnancy outcome are shown in Table 3. Poor pregnancy outcome was significantly related to warfarin dose; only three of 30 (10%) women had received less than 5 mg of drug daily. Gestational ages of 28 preterm pregnancy losses are reported in Table 4.

Relative effects of drug daily dose and order of pregnancy were estimated by exact logistic regression model (LogXact‐Turbo, CYTEL Software Corporation Cambridge, MA, 1993) including both as covariates (Table 5). Distribution of warfarin dose and poor outcome according to order of pregnancy and risk estimates are also in Table 5. Although the numbers are small, results were similar both in the first and subsequent pregnancies. Multivariable analysis found that warfarin daily dose greater than 5 mg was more hazardous than lower doses (odds ratio [OR] 49.4, exact 95% confidence interval [CI] 9.1, 424.1, P < .001), whereas order of pregnancy was no longer significant (OR 0.79, exact 95% CI 0.10, 5.4, P = .84).

Overall, the incidence of embryopathy in our series was 5.5% (four of 71). Three cases of skeletal malformation occurred, two in spontaneously aborted fetuses of mothers in the high‐dosage warfarin group and one in a full‐term infant in the low‐dosage group. Malformations included nasal hypoplasia in two cases (one in a spontaneously aborted fetus and one in a full‐term infant), cervical spine abnormalities in one case, and hydrocephalus associated with microphtalmia in one aborted fetus. A small ventricular septal defect was observed in the full‐term infant of a mother taking more than 5 mg of the drug daily.

There were no maternal deaths. No patient had thromboembolic or hemorrhagic complications.

Back to Top | Article Outline


The most commonly used anticoagulation protocols for patients with mechanical heart valves include sodium warfarin. The fetotoxicity of warfarin has been assessed extensively in the medical literature,12,13 but only recently have studies been performed to elucidate the mechanisms of such an effect.14,15

According to the studies published by Yacobi et al in 1976,16 although a variable fraction of sodium warfarin in the human blood is bound to albumin, the therapeutic effects on the synthesis of vitamin K–dependent coagulation proteins are due to the unbound fraction. One of the reported mechanisms affecting the rate of warfarin‐albumin binding is the competition of bilirubin for the same binding site.17

Because the molecular weight of unbound sodium warfarin is lower than 1000, it easily crosses the human placenta. A greater unbound fraction has been found in the serum of pregnant women than in nonpregnant subjects.17 Fetal warfarin concentrations seem to be particularly high, especially in early gestation.18 The anticoagulant effect of warfarin in maternal blood does not correlate with its activity in the fetus. The following factors appear to have an important effect on the prohemorrhagic action of warfarin in the fetus: high fetal bilirubin serum concentration,17 greater affinity of fetal albumin for bilirubin than for sodium warfarin,19,20 reduced drug metabolism in the immature liver,21 and poor fetal hepatic synthesis of vitamin K–dependent coagulation proteins14,22 (which cannot cross the human placenta).14 According to Pelkonen,23 fetal exposure to the drug is further increased because of the poor development (or absence) of the glucuronide conjugation enzymatic pathway in fetal livers and the subsequently limited renal elimination of warfarin hydrophilic metabolites.

The observation that prothrombin activity in fetal serum can be considerably lower than that in maternal blood has led to the hypothesis that most adverse effects in the fetuses of women treated with warfarin could result from hemorrhagic mechanisms.12,24 According to Oakley,13 not only premature deliveries and neonatal deaths, but even neurologic and ocular complications, as well as placental hemorrhages and spontaneous abortions should be considered anticoagulation‐related events. Conversely, evidence of metabolic mechanisms underlying cartilage abnormalities in warfarin embryopathy has emerged from both early and more recent studies.25,26 Therefore, the mechanisms underlying sodium warfarin fetotoxicity are still being investigated. Whether fetal injuries are determined by the molecular structure of the drug or by its interference with cartilage and bone tissue development is still being debated. Sodium warfarin interferes with vitamin K regeneration, inhibiting epoxide reductase, which plays a role in the synthesis of osteocalcin and Gla matrix protein, resulting in chondrodysplasia punctata and nasal hypoplasia.27

Experimental studies by Quick28 and Krauss et al29 showed that hemorrhages and fetal deaths during and after pregnancy in warfarin‐treated dogs and rabbits were dose‐dependent. Porreco et al30 showed the clinical harmlessness of minidoses of warfarin during pregnancy. Based on our experience, it was not necessary to adjust warfarin daily dose when a patient became pregnant in order to keep international normalized ratio within the therapeutic range.

New insight into the mechanism behind warfarin's adverse effects has emerged from a recent study by Franco et al,15 who investigated the molecular aspects of X‐linked recessive chondrodysplasia punctata, which has a phenotype virtually identical to warfarin embryopathy. Point mutations on the sulfatase gene arylsulfatase E, located within the genomic region Xp22.3, have been identified as the lesion in chondrodysplasia punctata. They found that induced expression of the gene in COS cells resulted in arylsulfatase activity that is inhibited by warfarin in a dose‐dependent fashion. This evidence suggests that warfarin embryopathy might involve a drug‐induced, dose‐dependent inhibition of the same enzyme.15

Our study demonstrated a significant statistical correlation (P < .001) between warfarin dose and fetal mortality, assuming 5 mg per day as a threshold dose. Dose dependence of warfarin embryotoxicity seems to decrease throughout gestation as the fetal liver reaches more complete development and as both albumin and vitamin K–dependent protein synthesis increases. Previous investigators have substituted heparin for the sodium warfarin during the first trimester of gestation based on the decrease in embryotoxicity in the latter part of gestation.6,7

Consistent with other studies,8,31,32 we showed that a safer and effective anticoagulant therapy can be achieved with warfarin derivatives in women with mechanical valve prostheses. There were no significant differences between outcomes with acenocoumarin and sodium warfarin.33 Maternal and fetal complications of such therapy can be limited if their dose‐dependence is considered. We suggest the following guidelines for treating pregnant women with mechanical heart valves:

1. The prosthetic model implanted (first, second, or third generation) and the site of implantation (mitral valve and/or aorta) can influence anticoagulant therapy. New bileaflet prostheses carry a lower risk of thromboembolic complications which allows lower‐dose anticoagulation.34

2. Pregnant women taking warfarin daily doses of 5 mg or less can be managed by continuing warfarin throughout pregnancy and delivery 48 hours after withdrawal of anticoagulation.

3. Patients in whom the international normalized ratio is kept within the therapeutic range by daily doses of sodium warfarin more than 5 mg should be informed that their pregnancy carries a higher risk of fetomaternal complications.35

4. A clinical evaluation should be done in each young woman requiring valve replacement procedures, to assess the dose needed to maintain the targeted international normalized ratio. If the mean dose is lower than or equal to 5 mg, the patient should receive a third‐generation mechanical prosthesis. If it is more than 5 mg, a bioprosthesis should be advised.

5. Because of the high incidence of bioprosthetic degeneration, recipients of biological valves should be informed that pregnancy should be undertaken within the first 6–8 postoperative years.3,5,6

Back to Top | Article Outline


1. Chan WS, Anand S, Ginsberg JS. Anticoagulation of pregnant women with mechanical heart valves: A systematic review of the literature. Arch Intern Med 2000;160:191–6.
2. Arnaout MS, Kazma H, Khalil A, Shasha N, Nasrallah A, Karam K, et al. Is there a safe anticoagulation protocol for pregnant women with prosthetic valves? Clin Exp Obstet Gynecol 1998;25:101–4.
3. North RA, Sadler L, Stewart AW, McCowan LM, Kerr AR, White HD. Long-term survival and valve-related complications in young women with cardiac valve replacements. Circulation 1999;99:2669–76.
4. Horstkotte D. Prosthetic valves or tissue valves–a vote for mechanical prostheses. Z Kardiol 1985;74(Suppl 6):19–37.
5. Lee CN, Wu CC, Lin PY, Hsieh FJ, Chen HY. Pregnancy following cardiac prosthetic valve replacement. Obstet Gynecol 1994;83:353–6.
6. Hanania G, Thomas D, Michel PL, Garbarz E, Age C, Millaire A, et al. Pregnancy and prosthetic heart valves: A French cooperative retrospective study of 155 cases. Eur Heart J 1994;15:1651–8.
7. Elkayan CL. Anticoagulation in pregnant women with prosthetic heart valves: A double jeopardy. J Am Coll Cardiol 1996;27:1704–6.
8. Sbarouni E, Oakley CM. Outcome of pregnancy in women with valve prostheses. Br Heart J 1994;71:196–201.
9. Suri V, Sawhney H, Vasishta K, Renuka T, Grover A. Pregnancy following cardiac valve replacement surgery. Int J Gynaecol Obstet 1999;64:239–46.
10. Vitale N, De Feo M, De Santo LS, Pollice A, Tedesco N, Cotrufo M. Dose-dependent fetal complications of warfarin in pregnant women with mechanical heart valves. J Am Coll Cardiol 1999;33:1637–41.
11. Cotrufo M, de Luca TSL, Calabrò R, Mastrogiovanni G, Lama D. Coumadin anticoagulation during pregnancy in patients with mechanical valve prostheses. Eur J Cardiothorac Surg 1991;3:300–5.
12. Luth DJ, Noller KL, Spikell JA, Danielson GK, Fish CR. Pregnancy and its complications following cardiac valve prostheses. Ann J Obstet Gynecol 1978;131:460–8.
13. Oakley CM. Valve prostheses and pregnancy. Br Heart J 1987;58:303–5.
14. Reverdieu-Moalic P, Delahousse B, Body G, Bardos P, Leroy, Gruel Y. Evolution of blood coagulation activators and inhibitors in the healthy human fetus. Blood 1996;88:900–6.
15. Franco B, Meroni G, Parenti G, Levilliers J, Bernard L, Gebbia M, et al. A cluster of sulfatase genes on Xp22.3: Mutations in condrodysplasia punctata (CDPX) and implications for warfarin embryopathy. Cell 1995;81:15–25.
16. Yacobi A, Udall JA, Levy G. Serum protein binding as a determinant of warfarin body clearance and anticoagulant effect. Clin Pharmacol Ther 1976;19:552–8.
17. Bajora R, Soorana SR, Contractor SF. Differential binding of warfarin to maternal, foetal and nonpregnant sera and its clinical implications. J Pharm Pharmacol 1996;48:486–91.
18. Mungall D, Ludden TM, Marshall J, Crawford M, Hawkins D. Relationship between steady-state sodium warfarin concentrations and anticoagulant effect. Clin Pharmacokinet 1984;9:99–100.
19. Sjoholm I, Ekman B, Kober A, Ljugstedt-Pahlman I, Seiving B, Sjodin T. Binding of drugs to human serum albumin. Mol Pharmacol 1979;16:767–77.
20. Irollo B, Dang Vu B, Nguyen Dai D, Yonger J. Dissociation and association rate constants changes following bilirubin binding affinity decreases. Dev Pharmacol Ther 1987;10:436–42.
21. Pauli RM, Lian JB, Masher DF, Suttie JW. Association of congenital deficiency of multiple vitamin K dependent coagulation factors and the phenotype of sodium warfarin embryopathy—clues to the mechanism of teratogenicity. Am J Hum Genet 1987;41:566–83.
22. Bonnar J. Haemostasis and coagulation disorders in pregnancy. In: Bloom AL, Thomas DP, eds. Haemostasis and thrombosis. London: Churchill Livingstone, 1994: 570–83.
23. Pelkonen O. Biotransformation of xenobiotics in the foetus. Pharmacol Ther 1980;10:261–81.
24. Mahairas GH, Veingold AB. Fetal hazard with anticoagulant therapy. Am J Obstet Gynecol 1963;85:234–7.
25. Barr M Jr, Burdi AR. Warfarin-associated embryopathy in a 17-week-old abortus. Teratology 1976;14:129–34.
26. Menger H, Lin AE, Toriello HV, Bernert G, Spranger JW. Vitamin K deficiency embryopathy: A phenocopy of the warfarin embryopathy due to a disorder of embryonic vitamin K metabolism. Am J Med Genet 1997;72:129–34.
27. Pauli RM. Mechanism of bone and cartilage maldevelopment in the warfarin embryopathy. Pathol Immunopathol Res 1988;7:107–12.
28. Quick A. Experimentally induced changes in the prothrombin level of the blood. Prothrombin concentration of newborn pups of a mother given dicoumarol before parturition. J Biol Chem 1946;164:371–4.
29. Krauss AP, Perlow A, Singer K. Danger of dicoumarol treatment in pregnancy. JAMA 1949;139:758–61.
30. Porreco RP, McDuffie RS Jr, Peek SD. Fixed mini-dose warfarin for prophylaxis of thromboembolic disease in pregnancy. A safe alternative for the fetus? Obstet Gynecol 1993;81:806–7.
31. Gohlke-Barwolf C, Acar J, Oakley C, Butchart E, Buckhardt D, Bodnar E, et al. Guidelines for prevention of thrombo-embolic events in valvular heart disease. Eur Heart J 1995;16:1320–30.
32. Oakley CM. Anticoagulation in pregnancy. Br Heart J 1995;74:107–11.
33. Ibarra-Perez C, Azevalo Toledo N, Alvarez de La Cadena O, Noriega Guerra L. The course of pregnancy in patients with artificial heart valves. Am J Med 1976;61:504–12.
34. Horstkotte D, Schulte HD, Bircks W, Strauer BE. Lower intensity anticoagulation therapy results in lower complication rates with the St. Jude Medical prosthesis. J Thorac Cardiovasc Surg. 1994;107:1136–45.
35. Ginsberg JS, Greer I, Hirsh J. Use of antithrombotic agents during pregnancy. Chest 2001;119:122S–31S.
© 2002 by The American College of Obstetricians and Gynecologists.