Long-Term Survival and Valve-Related Complications in Young Women With Cardiac Valve Replacements
Background—The type of cardiac valve replacement associated with the lowest health risks for young women who may undergo pregnancies is unknown. We investigated which valve type was associated with greatest patient and valve survival and the effect of pregnancy on valve loss.
Methods and Results—In this retrospective study, all women 12 to 35 years old who underwent valve replacements between 1972 and 1992 at Greenlane Hospital were identified, and follow-up was available in 93%. The 232 women were followed up for 1499 patient-years. Ten-year survival of women with mechanical (n=178), bioprosthetic (n=73), and homograft (n=72) valves was 70% (95% CI, 59% to 83%), 84% (95% CI, 72% to 99%), and 96% (95% CI, 91% to 100%), P=0.002. After adjustment for confounding variables, the relative risk (RR) of death with mechanical compared with bioprosthetic valves was 2.17 (95% CI, 0.78 to 5.88). Thromboembolic events occurred in 45% of women with mechanical valves within 5 years, compared with 13% with bioprosthetic valves, P=0.0001. Valve loss at 10 years was higher in bioprosthetic valves [82% (95% CI, 62% to 92%)] than in mechanical [29% (95% CI, 17% to 39%)] or homograft [28% (95% CI, 12% to 41%)] valves, P=0.0001. Pregnancy was not associated with increased bioprosthetic (RR, 0.96; 95% CI, 0.68 to 1.35), homograft (RR, 0.65; 95% CI, 0.37 to 1.13), or mechanical (RR, 0.54; 95% CI, 0.27 to 1.08) valve loss.
Conclusions—Although 10-year valve survival was greater with mechanical than bioprosthetic valves, mechanical valves may be associated with reduced patient survival in young women. Thromboembolic complications, often with long-term sequelae, were common with mechanical valves. Pregnancy did not increase structural deterioration or reduce survival of bioprosthetic valves.
The optimal type of heart valve replacement for young women is controversial.1 2 3 4 Each type of valve, whether mechanical, bioprosthetic, or homograft, is associated with particular problems.1 It is not known which valve replacement is associated with the lowest overall health risks to young women who may undergo pregnancies.
Women with mechanical prosthetic valves are at risk of thromboembolic events and require lifelong anticoagulation.5 6 This has the attendant risk of hemorrhage and poses significant hazards to both mother and baby in future pregnancies.2 3 7 8 9 10 Warfarin use in pregnancy is associated with a low risk of maternal thromboembolic complications but high fetal losses.2 3 7 Heparin therapy significantly improves fetal outcome, but up to one third of pregnant women with mechanical valves have thromboembolic complications, including fatal events.2 3 9
The high rate of complications in pregnancy in women with mechanical valves has led some groups to suggest that bioprosthetic valves may be the preferred valve replacement in women who are anticipating having a child.1 3 Women with bioprosthetic valves need only short-duration anticoagulation after valve replacement, provided that sinus rhythm is maintained, and usually have a successful pregnancy outcome.4 However, as young patients with bioprosthetic valves are at risk of early reoperation because of increased structural valvular deterioration,11 there is reluctance to use these valves in the young.12 Concern has also been raised that pregnancy may hasten the rate of structural deterioration of bioprosthetic valves.2 13 The controlled study reporting this was confounded by younger age in the women who underwent pregnancies.13 When this study was extended, no difference was found in the rate of structural valvular deterioration with pregnancy.4
Homograft valves are used in the aortic position and do not require long-term anticoagulation.14 There is limited information on the survival and valve complications of young women with homografts and no information on the effect of pregnancy on valve survival.15 16
In this study, we investigated which type of valve replacement (mechanical, bioprosthetic, or homograft) was associated with the greatest long-term survival in young women and compared the rate of valve complications with each valve type. We also investigated whether pregnancy was associated with increased or earlier valve loss.
Women (n=308) 12 to 35 years old who underwent aortic, mitral, or tricuspid cardiac valve replacement between January 1972 and December 1992 at Green Lane Hospital, Auckland, New Zealand, were identified in the cardiothoracic surgical database. Fifty-three women were excluded because they had received valve types that are no longer used (stent-mounted homografts or Braunwald Cutter valves; n=46) or refused consent (n=7). The hospital’s ethics committee approved the study.
We obtained information on cardiac status and outcome of pregnancies from 1972 to December 1994. Data on all women were collated from our hospital medical records (cardiac and obstetric), a national database, and the New Zealand register of deaths. Additional information on any pregnancies and consent to review medical records held by their local medical practitioner and other hospitals was obtained from 41% of women or their families.
The type of valve replacement was categorized as homograft, bioprosthetic (Hancock, Carpentier-Edwards, Medtronic Intact), or mechanical (Starr-Edwards, St Jude Medical, Medtronic Hall, Bjork-Shiley). Cardiac complications were classified according to published guidelines.17 Operative mortality was defined as death from any cause within 30 days of operation. Late deaths included all other deaths. Valve-related mortality was any death due to structural valvular deterioration, nonstructural dysfunction, thromboembolism, endocarditis, anticoagulant-related hemorrhage, or death at reoperation. Sudden, unexpected, unexplained deaths were included under valve-related death, unless autopsy proved that the death was not valve-related. Deaths due to any other cardiac conditions, in the presence of satisfactory functioning of the cardiac valve replacements, were classified as cardiac deaths.
Reoperation was defined as any operation that replaced a previous valve replacement. Valve loss was defined either as reoperation or a valve-related death. Structural valvular deterioration was any change in valve function resulting from an intrinsic abnormality that caused stenosis or regurgitation. Nonstructural valve dysfunction was any abnormality resulting in stenosis or regurgitation at the operated valve that was not intrinsic to the valve itself. Endocarditis was any infection involving a valve replacement and was diagnosed by standard clinical criteria, including positive blood cultures or evidence of endocarditis at reoperation. Thromboembolism was any valve thrombosis or embolism, excluding infection. It included any new, permanent or transient, focal or global neurological deficit (exclusive of hemorrhage) and any peripheral arterial emboli. A transient ischemic attack was a neurological deficit reversible within 24 hours, and a reversible ischemic neurological deficit was reversible within 3 weeks. Anticoagulant-related hemorrhage was any episode of internal or external bleeding that caused death, stroke, hospitalization, or operation or required transfusion.
Women with concurrent valves of different types (for example, a homograft valve at the aortic site and a mechanical valve at the mitral site) were excluded, because any association between valve type and outcome could not be attributed to a particular valve type. The women could have >1 valve of the same type simultaneously and/or could have a series of consecutive valves in the same site. Inclusion of each valve as a separate entity in the survival analyses would have resulted in duplicate counting of end points in women with multiple concurrent valves. Hence, in the primary survival analysis, the women’s histories were partitioned into valve states. A valve state began at insertion of a new valve and concluded either when the data were censored (the last follow-up information available), or the valve or valves were replaced, or the woman died. These valve states were taken as independent measures, and the survival plots were produced by the Kaplan-Meier method. Tests of significance of difference between the valve types were produced by use of the counting process formulation of Andersen and Gill,18 an extension of the Cox proportional hazards regression model that takes into account the time-dependent nature of the data. Operative deaths were excluded from survival analyses, because these deaths were not attributed to the new valve replacement. The single operative death at replacement of a valve previously operated on was classified as a late death of the initial valve replacement.
In the multivariate analyses, the mechanical valve group was compared with the bioprosthetic valve group. Homografts were excluded from these analyses, because all homografts were implanted in the aortic site and have a recognized different natural history. Potential confounding variables included in the model were valve site, year of operation, woman’s age at valve replacement, number of concurrent valves (1, 2, or 3), ethnicity (categorized as Maori, Pacific Islander, or other), and whether the woman was pregnant at any time during the life of that valve. The large number of women with concurrent valves at 2 sites resulted in difficulties in classifying women according to their valve site. In the multivariate analysis, valve site was divided into mitral (including those with concurrent valves at the aortic or tricuspid site) and aortic or tricuspid site only.
To provide data consistent with previous studies, we also analyzed the end points attributable to a particular valve by valve rather than by patient. Reoperation, structural valve deterioration, and nonstructural valve dysfunction at the mitral and aortic sites were analyzed with Kaplan-Meier 10-year survival probabilities and the log-rank statistic.
Two hundred and fifty-five women with 394 heart valve replacements were studied. Twenty-three women with coexistent valves of mixed type were excluded (Figure 1⇓). The final study group comprised 164 women with a single valve replacement, 22 with sequential valve replacements at a single site, and 46 with concurrent valve replacements of the same type. Each woman had only 1 type of valve at any time point. The 232 women had 323 valve replacements. Follow-up was available in 93% of women, for a total of 1499 patient-years.
The characteristics of the women and types of valves are shown in Table 1⇓. Seventy-one women underwent 132 pregnancies. Table 2⇓ shows patient characteristics according to the type of valve. Women with mechanical valves were older, had a greater proportion of valve replacements since 1980, and were more likely to have concurrent valves (P<0.0001).
Survival of Women
Forty-five women (19.4%) died during the study period. There were 11 operative deaths (4.7%) and 34 late deaths (14.7%). The operative death rate was 11.6% (95% CI, 3.9% to 25.1%) from 1972 to 1979, 2.9% (95% CI, 0.8% to 7.2%) from 1980 to 1989, and 4.1% (95% CI, 0.5% to 14.0%) from 1990 to 1992 (P=0.10). The late death rate was 3 per 100 patient-years. Among the late deaths, 44% were valve related, 20% were cardiac but not valve related, 11% were noncardiac, and 25% were of unknown cause. At the time of death, 28 women had mechanical valves (3 operative deaths, 25 late deaths), 8 had bioprosthetic valves (2 operative, 6 late), and 9 had homografts (6 operative, 3 late). Including operative deaths, 10-year survival was 69% in women with mechanical valves, compared with 82% in the bioprosthetic valve group and 88% in homografts (P=0.09).
There was a difference in long-term patient survival between the 3 valve types (P=0.002; Figure 2⇓). Ten-year survival (excluding operative deaths) of women with mechanical valves was 70% (95% CI, 59% to 83%) compared with 84% (95% CI, 72% to 99%) in the bioprosthetic valve group and 96% (95% CI, 91% to 100%) in women with homografts. In the multivariate analysis, women with mechanical valves were compared with those with bioprosthetic valves. After adjustment for valve site, age at valve replacement, year of operation, number of concurrent valves, ethnicity, and pregnancy, the probability of an effect of valve type on long-term survival was P=0.12. The relative risk (RR) of death in women with mechanical valves compared with bioprosthetic valves was 2.17 (95% CI, 0.78 to 5.88; Table 3⇓).
Maori and Pacific Island women had an 8-fold and 7-fold RR of death, respectively, compared with European and Asian ethnic groups (Table 3⇑). There was no interaction between race and valve type on survival (P=0.35). Pregnancy was associated with increased patient survival (RR, 0.38; 95% CI, 0.17 to 0.84). Reliable data on the postoperative NYHA classification were available in 80% of valves, and of these, 96% were in NYHA class I or II. The relationship between the type of valve replacement and patient survival did not change after adjustment for the postoperative NYHA classification. The NYHA classification was not included in the final model.
Freedom from valve loss is shown in Figure 3⇓. At 10 years, the rate of valve loss was 82% (95% CI, 62% to 92%) in women with bioprosthetic valves compared with 29% (95% CI, 17% to 39%) in those with mechanical valves and 28% (95% CI, 12% to 41%) in those with homograft valves (P=0.0001). After exclusion of homografts and adjustment for potential confounding variables, reduced survival of bioprosthetic valves compared with mechanical valves persisted (P=0.005; RR, 2.48; 95% CI, 1.35 to 4.57). No other variable was associated with increased valve loss, including valve site, age at valve replacement, or race. The data suggested a different relationship between valve type and valve loss in the mitral site compared with aortic or tricuspid sites (P=0.05), with relatively greater bioprosthetic valve loss in the mitral site.
In women with concurrent valves, valve-related death was not attributable to a particular valve. Consequently, valve loss (reoperation or valve-related death) at the aortic and mitral sites could not be separately analyzed. However, the probability of reoperation alone was analyzed according to valve site (Figure 4⇓, Table 4⇓). At the aortic site, comparison of homografts with only bioprosthetic valves showed that reoperation was less likely with homografts (P=0.03).
Effect of Pregnancy on Valve Survival
Overall, pregnancy did not influence valve loss (RR, 0.79; 95% CI, 0.61 to 1.01). The relationship between pregnancy and valve loss was analyzed separately for each valve type. Women with bioprosthetic valves had the same rate of valve loss whether or not they had undergone pregnancy (RR, 0.96; 95% CI, 0.68 to 1.35; Figure 5⇓). These findings remained after adjustment for potential confounding variables, including age. Similarly, there was no difference in valve survival between women who underwent pregnancy and those who did not in the homograft (RR, 0.65; 95% CI, 0.37 to 1.13) or mechanical valve (RR, 0.54; 95% CI, 0.27 to 1.08) groups.
Structural Valvular Deterioration
Structural valvular deterioration was most common in bioprosthetic valves (P<0.0001; Figure 6⇓) and was the major reason for replacement of these valves (91%). After adjustment for confounding variables, structural valvular deterioration remained greater in bioprosthetic than in mechanical valves (P<0.0001). The risk of structural valvular deterioration was 7-fold greater in the mitral than in aortic or tricuspid sites (RR, 7.3; 95% CI, 1.3 to 40.3). Analysis by valve showed that structural valvular deterioration at the mitral site occurred in 84% (95% CI, 63% to 93%) of bioprosthetic valves by 10 years, but none of the mechanical valves (Figure 7⇓). Pregnancy did not increase structural valvular deterioration of bioprosthetic valves (RR, 0.84; 95% CI, 0.57 to 1.24; P=0.39).
Nonstructural Valve Dysfunction
At 10 years, 95% (95% CI, 88% to 100%) of homograft valves, 96% (95% CI, 90% to 100%) of bioprosthetic valves, and 85% (95% CI, 75% to 96%) of mechanical valves were free of nonstructural valve dysfunction (P=0.09). Nonstructural valve dysfunction according to valve site is shown in Table 4⇑.
There was no difference in the rate of endocarditis with the different valve types. At 10 years, endocarditis had occurred in 22% (95% CI, 0% to 46%), 10% (95% CI, 0% to 19%), and 12% (95% CI, 3% to 21%) of women with bioprosthetic, homograft, and mechanical valves, respectively (P=0.84).
Sixty women had 94 thromboembolic events. These events were 37 strokes (39%), 46 transient ischemic attacks or reversible ischemic neurological deficits (49%), 7 emboli to other organs or limbs (7%), and 4 isolated valve thrombi (4%). Thromboembolic events were more common in women with mechanical than bioprosthetic valves (P=0.0001; Figure 8⇓). Of women with mechanical valves, 45% (95% CI, 34% to 55%) had a thromboembolic event within 5 years compared with 13% (95% CI, 3% to 22%) of women with bioprosthetic valves and 1.4% (95% CI, 0% to 5%) of women with homograft valves (P=0.0001).
After adjustment for potential confounding variables, the RR of a thromboembolic event with mechanical compared with bioprosthetic valves was 3.07 (95% CI, 1.34 to 7.04). Pregnancy was associated with a lower risk of thromboembolism (RR, 0.46; 95% CI, 0.24 to 0.89). Neither the mitral position (RR, 0.71; 95% CI, 0.26 to 1.89) nor multiple concurrent valves (RR, 0.84; 95% CI, 0.46 to 1.51) were associated with increased thromboembolism.
When the thromboembolic event occurred, 31% of women with mechanical valves and 83% of women with bioprosthetic valves were in atrial fibrillation (P=0.02). In the mechanical valve group, 78 women were prescribed warfarin (7 with antiplatelet drugs), 5 received heparin, 1 was on aspirin alone, and in 1 woman the anticoagulation regimen was unknown. An international normalized ratio (INR) or prothrombin ratio (PR) was available at the time of the thromboembolic event in 40 of the 78 women on warfarin. Among these, 28 (70%) had an INR or PR of <2.0, 4 had an INR or PR of 2 to 2.9, and in 8 women the INR or PR was ≥3.
Twenty-nine women (26 with mechanical and 3 with bioprosthetic valves) had 40 anticoagulant-related hemorrhages resulting in hospital admissions. Twelve women had menorrhagia, 9 epistaxis or hemoptysis, 9 hemorrhage after surgery that required transfusion or further surgery, 4 hemarthroses, 2 renal hematomas, 2 trauma, 1 gastrointestinal bleed, and 1 cerebral hemorrhage. At the time of hemorrhage, the type of anticoagulation was warfarin (n=34), heparin (n=5), and both warfarin and heparin (n=1). Among women treated with warfarin, the median INR was 4.7 (range, 2 to 11.9; n=22) or median PR 3.1 (range, 2.3 to 4; n=5). Fifty-five percent of women had an INR >4.
This retrospective cohort study of young women found greater valve survival of mechanical valves compared with bioprosthetic valves. The high rate of bioprosthetic valve loss, predominantly due to structural valvular deterioration, was particularly evident after 7 years. This is consistent with previous studies.4 11 19 In contrast to valve survival, the unadjusted long-term survival of women was greater in those with bioprosthetic rather than mechanical valves. After adjustment for potential confounding variables, the trend for increased patient survival in women with bioprosthetic compared with those with mechanical valves persisted.
Pregnancy was not associated with an increase in either structural valvular deterioration or valve loss of bioprosthetic, mechanical, or homograft valves. This is in keeping with a previous large controlled study, which found no increase in structural valvular deterioration of bioprosthetic valves with pregnancy.4 The association of pregnancy with greater survival of women is likely to be due to patient selection rather than pregnancy conferring a specific health benefit.
Women with mechanical valves had a high rate of thromboembolism and hemorrhagic complications. Five years after mechanical valve replacement, ≈50% of these young women had suffered a thromboembolic event, with 39% having long-term sequelae. INR values at the time of complications indicated poor compliance, either with taking warfarin or with undergoing laboratory monitoring. The importance of factoring in compliance and adherence in treatment outcome has recently been highlighted.20 21 The majority of women in this study were young Maori and Pacific Island women who had rheumatic heart disease, a disease associated with poor living conditions. Young age and lower income levels have been associated with poor compliance in other chronic diseases.22 23 24
The majority of the mechanical valves in this study were Starr-Edwards valves. These valves carry a greater risk of thromboembolic complications than the newer, low-profile St Jude Medical or Medtronic Hall valves.1 5 The relevance of our findings to the newer valves is uncertain.
The majority of women with bioprosthetic or homograft valves have a successful pregnancy outcome. In the aortic site, the probability of reoperation of homograft valves was less than for bioprosthetic valves. Although the number of bioprosthetic valves in the aortic position was small, these data would support the use of homografts in preference to bioprosthetic valves at the aortic site.
In this study, we wished to address outcomes in terms of the woman’s overall health, not just valve-related end points. Few long-term studies have specifically investigated this patient group with its unique problems. The complexity of valve combinations in many women made it difficult to attribute certain end points, including death, valve loss, thromboembolism, and hemorrhage, to a particular valve. To overcome this difficulty, our analysis was performed by “valve state.” To allow comparison with previous studies, the end points attributable to an individual valve were also analyzed by valve. Limitations of this study include failure to use an assessment of global ventricular function in our analysis. Because this was a retrospective study over 2 decades, consistent ventricular function data were not available. Postoperative NYHA classification was also not included in the multivariate analyses because few women were in class III or IV, there was insufficient power to obtain meaningful results, and 20% had no information.
In conclusion, our data suggest that although valve survival is better with mechanical valves (predominantly Starr-Edwards), these valves may result in significant morbidity in young women and higher overall mortality. Because many young women wish to have children, the increase in pregnancy complications with mechanical valves needs to be considered. A successful pregnancy outcome is likely with bioprosthetic or homograft valves. Conversely, mitral bioprosthetic valves will probably require replacement by 10 years, with the associated operative risks. In deciding which valve replacement is preferable for a particular woman, the benefits and risks need to be balanced. Large, randomized trials are necessary to investigate long-term survival of young women with valve replacements, comparing a low-profile mechanical valve with a new-generation bioprosthetic valve in the mitral site and homografts with a bioprosthetic valve in the aortic site.
This study was supported by the Green Lane Hospital Research and Educational Fund, the Auckland Medical Research Foundation, the New Zealand Health Research Council, and the New Zealand Obstetric and Gynaecological Society.
- Received November 30, 1998.
- Revision received February 26, 1999.
- Accepted March 11, 1999.
- Copyright © 1999 by American Heart Association
Sbarouni E, Oakley CM. Outcome of pregnancy in women with valve prostheses. Br Heart J. 1994;71:196–201.
Hanania G, Thomas D, Michel PL, Garbarz E, Age C, Millaire A, Acar J. Pregnancy and prosthetic heart valves: a French cooperative retrospective study of 155 cases. Eur Heart J. 1994;15:1651–1658.
Jamieson WR, Miller DC, Akins CW, Munro AI, Glower DD, Moore KA, Henderson C. Pregnancy and bioprostheses: influence on structural valve deterioration. Ann Thorac Surg. 1995;60:S282–S287.
Cannegieter SC, Rosendaal FR, Briet E. Thromboembolic and bleeding complications in patients with mechanical heart valve prostheses. Circulation. 1994;89:635–641.
Barratt-Boyes BG, Roche AH, Subramanyan R, Pemberton JR, Whitlock RM. Long-term follow-up of patients with the antibiotic-sterilized aortic homograft valve inserted freehand in the aortic position. Circulation. 1987;75:768–777.
Edmunds LH Jr, Clark RE, Cohn LH, Grunkemeier GL, Miller DC, Weisel RD. Guidelines for reporting morbidity and mortality after cardiac valvular operations: the American Association for Thoracic Surgery, Ad Hoc Liaison Committee for Standardizing Definitions of Prosthetic Heart Valve Morbidity. Ann Thorac Surg. 1996;62:932–935.
Hammermeister KE, Sethi GK, Henderson WG, Oprian C, Kim T, Rahimtoola S. A comparison of outcomes in men 11 years after heart-valve replacement with a mechanical valve or bioprosthesis: Veterans Affairs Cooperative Study on Valvular Heart Disease. N Engl J Med. 1993;328:1289–1296.
Miller NH, Hill M, Kottke T, Ockene IS. The multilevel compliance challenge: recommendations for a call to action: a statement for healthcare professionals. Circulation. 1997;95:1085–1090.
Fox R. Compliance, adherence, concordance. Circulation. 1998;97:127.