Factors Associated With Periprosthetic Leakage Following Primary Mtral Vave ReplacemenE With Special Consideration of the Suture Technique

Transcription

1 Factors Associated With Periprosthetic Leakage Following Primary Mtral Vave ReplacemenE With Special Consideration of the Suture Technique Janardan P. Dhasmana, M.B., F.R.C.S., Eugene H. Blackstone, M.D., John W. Kirklin, M.D., and Nicholas T. Kouchoukos, M.D. ABSTRACT Among 435 patients without native valve endocarditis who were followed up to 69 months after primary mitral valve replacement, 25 developed documented periprosthetic leakage. In 10 patients this was associated with prosthetic valve infection. No evidence of prosthetic infection was found in the remaining 15 patients with documented leakages, and they form the basis of the study. Multivariate Cox regression analysis indicated that leakage in the absence of infection was strongly associated with the use of small monofilament suture (2-0 or 3-0 versus 1-0) in a continuous suture technique (92.0% actuarially leak free by 43 months versus 99.0% for continuous 1-0 monofilament sutures or pledgeted mattress sutures; p = 0.01) and with annular calcification (p = 0.01). We did not find (p > 0.2) the functional type of mitral valve lesion or its pathology, or the type and size of prosthesis used, to be incremental risk factors. Beddermann and Borst [l] concluded from their data that a continuous suture technique is ineffective for minimizing periprosthetic leakage after insertion of replacement valves in the mitral position. In our institution and in many others [2], both continuous and interrupted pledgeted technique have been used for some years for mitral replacement in the belief that both were satisfactory. In view of the uncertainty created by the conclusions of Bedderman and Borst and the importance of the matter, we undertook a study of the incremental risk factors, including suture technique, for periprosthetic leakage after mitral valve replacement. From the Department of Surgery, University of Alabama in Birmingham School of Medicine and Medical Center, Birmingham, AL. Accepted for publication Jan 6, Address reprint requests to Dr. Kirklin, Department of Surgery, University Station, Birmingham, AL Materials and Methods Study Group We examined all primary mitral valve replacements performed between January 1, 1975, and July 1, 1979, at the University of Alabama in Birmingham Medical Center, exclusive of patients whose mitral valve replacement either was associated with aortic valve replacement or was concomitant with or preceded by repair of complex congenital heart disease. From the resulting 479 cases we excluded all hospital deaths (27 patients) after finding that none of the patients who died early exhibited clinical or postmortem evidence of valve dehiscence. Sixteen hospital survivors had active native valve endocarditis at the time of their valve replacement, and we excluded these patients from the general analysis also. Among the remaining patients, 1 had a Starr-Edwards prosthesis (no periprosthetic leakage at last follow-up). This patient was excluded, leaving a study group of 435 patients in whom a Bjork-Shiley prosthesis of the supraannular type or a Carpentier- Edwards or Hancock heterograft valve was used. Of these patients, 300 had no other concomitant procedure, while 135 had one or several associated procedures. These included coronary artery bypass grafting (in 94 patients), left ventricular aneurysmectomy (in 5), tricuspid repair (in 37), closure of atrial septal defect (in 6), closure of patent ductus arteriosus (in 2), excision of left atrial myxoma (in l), left ventricular myomectomy for left ventricular outflow tract obstruction (in 5), and repair of postinfarction muscular ventricular septal defect (in 1). Definition of Documented Periprosthetic Leakage The postmortem examination or surgical finding of valve dehiscence (gap between mitral annulus and sewing ring of the prosthesis) was accepted as proof of periprosthetic leakage ( /83/ ! by The Society of Thoracic Surgeons

2 171 Dhasmana et al: Periprosthetic Leakage following MVR patients). An additional 5 patients were found at cardiac catheterization to have leakage of contrast material around the periphery of their prosthesis, while 2 others with clinically evident mitral incompetence had fluoroscopic evidence of abnormal prosthesis "rocking" 131 (and additionally in 1, large basilar excursion of the valve), which had not been present on fluoroscopic examination immediately postoperatively. In 1 of these patients, additional echocardiographic evidence [4-61 of periprosthetic leakage was obtained. We considered these 25 patients to have "documented" periprosthetic leakage. Systolic murmurs in asymptomatic patients were not accepted as evidence of leakage, since this finding has been shown to be an unreliable criterion [7, 81. Patient Follow-up Follow-up questionnaires were sent to patients and their doctors on August 1, We included in the analysis all responses received by January 6, Information was obtained for 434 of the 435 patients, 1 patient remaining untraced since hospital discharge. In 323 cases the follow-up data were provided by physicians, as either reoperation notes, postmortem reports, or reports of personal late postoperative evaluations. In 111 cases, the reports were only from the patient or relatives; however, 65 of these patients had reported seeing their physician within the last year, although we were unable to obtain a report from the physician. Excluding the untraced patient, the follow-up interval for the 341 late survivors ranged from 12 to 60 months (median, 37 months). Actuarial Analysis The product-limit life table method was used to calculate event-free actuarial estimates [9], and a logistic transformation [lo] of the exact variance [ll] was employed in calculating their 70% confidence limits (CL) (equivalent to +1 standard deviation). For periprosthetic leakage, the interval from operation to first documentation of the leak was employed (fluoroscopy, cardiac catheterization, or date of death when postmortem examination revealed periprosthetic leakage). Patients not experiencing a leak were considered "untraced" (censored) beyond the date of late death, the date of mitral valve reoperation for something other than periprosthetic leakage, or the last date of uneventful follow-up. In 10 of the 25 patients the periprosthetic leak was associated with prosthetic valve infection; for analysis of periprosthetic leakage with infection (10 patients), therefore, patients experiencing leaks not associated with infection were censored at the time their leaks were documented. For analysis of periprosthetic leakage without infection (15 patients), patients experiencing prosthetic valve infection (19 patients) with or without leakage were censored on the date infection was documented, since infected patients were no longer at risk for leakage without infection. Analysis of Variables Associated with Peripros the tic Leakage without Infection We first analyzed the association with documented leakages without infection (fifteen events) of demographic variables, variables extracted from preoperative studies, and anatomical-pathological details. For this evaluation we used simple contingency table methods and t tests for individual variables, followed by single and multivariate analyses, initially using logistic regression [121, then the Cox proportional hazards model [13] (as modified by Breslow [14]). Only the latter technique takes into account time and censoring of the study group. Nevertheless, we found that all the methods yielded the same results. Thus, we have elected to present simple contingency tables and statistics based on an N of 435, but have supplemented these data with the statistics obtained from the Cox regression model. The demographic variables examined included age and body surface area at operation, race, sex, New York Heart Association Functional Class, and "era" of operation (cardioplegia versus noncardioplegia, since about half of the patients-233, or 54%-were operated on after cardioplegia was introduced in our institution in 1977). Variables from preoperative studies included electrocardiographic evidence of left or right ventricular hypertrophy, or both; rhythm mechanism; the presence of old myocardial infarction; coronary angiographic

3 172 The Annals of Thoracic Surgery Vol 35 No 2 February 1983 evidence of right, left, or circumflex coronary artery lesions; and catheterization values for left ventricular end-diastolic pressure, valve area, and pulmonary and systemic arterial pressures. The paucity of such preoperative study information forced us to do separate analyses with this information included and excluded. Anatomical or pathological details included the predominant mitral valve lesion (stenosis, incompetence, or mixed, the latter specified whenever accompanying stenosis or incompetence was more than mild), the native valve abnormality, the presence of valvular calcium (from surgical or histological reports), the presence of annular calcium as recorded in the surgeon s operative note, the estimate of left ventricular and left atrial enlargement at the time of operation (graded from 0 = normal to 6), and the presence of a left atrial thrombus. We then added to the analysis factors relating to the operative procedure. These included associated repairs (atrial septa1 defect, aneurysmectomy, myomectomy, coronary artery bypass grafting and number of distal anastomoses, and tricuspid annuloplasty), elapsed time of bypass, myocardial ischemia time, the type of valve prosthesis employed (Bjork-Shiley versus each of the two types of heterograft), the diameter of the prosthesis, the surgeon, and details of the suturing technique, including the type of suture material used, its size (1-0,2-0, or 3-0), and the suture technique (continuous suture or pledgeted mattress sutures). To these factors we added the presence of a systolic murmur upon hospital discharge. For all multivariate analyses, a p value of 0.2 was selected as the criterion for inclusion of a factor in the model. Both forward stepwise procedures as well as backward elimination were employed in determining the final models. Details of the Suture Techniques The continuous technique (296 patients) differed in only small details from surgeon to surgeon and over the time of the study for a given surgeon. In general, one or two double-armed sutures (usually monofilament polypropylene*) were employed in a simple over-and-over Prolene; Ethicon, Inc., Somerville, NJ. single row. When one suture was used, it was begun as a pledgeted mattress suture posteriorly with the pledget on the atrial side. If two sutures were used, a mattress suture with a pledget was begun at each commissure, run in each direction, and tied together where they met, without a pledget. In 17 of the 296 patients, one or more supplemental pledgeted interrupted reinforcing sutures were added to the suture line (1 added in 11 patients, 2 in 3 patients, 4 in l patient, and an unspecified number in 2 patients). A horizontal mattress pledgeted suture technique (usually using a silicon-treated braided Dacron suture material*) was employed for the interrupted method. Tetrafluoroethylene predrilled pledgets measuring 3 X 7 mm were used on the left atrial side. In 1 patient, the anterior portion of the prosthesis was inserted using the continuous suture technique and the posterior portion using interrupted pledgeted stitches; for the analyses of suture technique, this patient was excluded. A separate analysis was made for the subgroup of patients in whom either one standard continuous technique (only polypropylene suture material without supplemental stitches) or the pledgeted mattress technique (with only silicon-treated braided Dacron sutures) was used. Additional Analyses There were 19 instances of prosthetic infection, documented at reoperation or postmortem examination (14 patients), by positive blood cultures (in 3), or by clinical evidence of septicemia (in 2). Analyses of these events were performed using logistic regression. Separate analyses were also made of the 10 patients with both prosthetic infection and periprosthetic leakage to determine whether the factors associated with such leaks were the same or different from those associated with noninfected leakages. Results Of the 435 patients who underwent primary mitral valve replacement, 25 (5.770, CL 4.6 to Tycron with integral pledgets; Davis & Geck, American Cyanamid Company, Pearl River, NY.

4 173 Dhasmana et al: Periprosthetic Leakage following MVR I 94 I MONTHS AFTER OPERATION Fig I. Actuarial analysis of freedom from PeriPrOSthetic leakage not associated with infection following primary mitral valve replacement. The fifteen events are shown as circles placed at the time the leakage was first documented. The vertical lines with crossbars represent the asymmetrical 70% confidence limits (fl standard deviation) of the actuarial estimates. The dashed extension of the curve represents follow-up beyond the last observed event. In this period, the absence of events does not allow actuarial estimate by the method used. 7.1%) developed documented periprosthetic leakage in the follow-up period. Actuarially, 93.0% of the patients (CL 91.5 to 94.3%) are free of this event beyond 43 months postoperatively. Of the 25 patients developing periprosthetic leak, 15 had no evidence of infection and 10 had prosthetic valve endocarditis. Periprosthetic Leak without Evidence of Infection Actuarial freedom from periprosthetic leakage without infection is 95.3% (CL 93.9 to 96.4%) at 43 months postoperatively (Fig 1). Of the 15 cases, 6 (40%) were documented within 6 months of operation. Eight of the 15 patients have required prosthetic valve replacement or resuture, with no hospital mortality. Two have died late postoperatively, 1 of complications of presumed air embolism at reoperation and 1 associated with documented releakage. Unfortunately, there has been little documentation of the mechanism of suture line disruption among these 8 reoperated cases. In 1, the 2-0 polypropylene suture was fractured; in 1, the sutures were said to have out of the tissue; and in a third, mention was made of friable tissues. (Subsequent to this analysis, another patient has been Seen at reoperation to have broken 2-0 polypropylene, and 1 with a Hancock heterograft valve has been seen to have the sewing ring tom away from the frame as the cause of leak- age.) Single and multivariate analyses indicate that only the use of a continuous 2-0 polypropylene suture, annular calcification, left ventricular hypertrophy on the electrocardiogram, and black race were associated with an increased risk of periprosthetic leakage without infection (Table 1). All other variables analyzed (see preceding description), including possible surgeon-to-surgeon differences, were found to be unrelated to this risk (p > 0.20). The strongest risk factor for periprosthetic leakage in the uninfected group was the use of a continuous 2-0 polypropylene suture (see Table 1). This factor was present in 13 (87%) of the 15 patients with periprosthetic leak without evidence of infection. Actuarial freedom from leakage when the 2-0 continuous polypropylene suture technique was used was 92.0% (CL 89.5

5 174 The Annals of Thoracic Surgery Vol 35 No 2 February 1983 Table 1. Risk Factors for Periprosthetic Leakage without lnfection following Primary Mitral Valve Replacement in 435 Patientsa Incremental Risk Factors Cox Coefficient + SD p Value Use of continuous f suture technique Annular calcification 1.4 k LVH on preoperative ECG 1.1 f Black race 1.2 f "Missing data in 32 patients, 2 experiencing leakage. SD = standard deviation; LVH = left ventricular hypertrophy; ECG = electrocardiogram. to 94.Oo/o), in contrast to 99.0% (Ct 97.9 to 99.5%) when either the continuous 1-0 polypropylene or interrupted pledgeted mattress suture techniques are considered (p for difference = 0.003) (Fig 2). In view of the unusual nature of the finding that size of the suture material is important, a number of other analyses were made in an effort to further test this finding. Continuous 1-0 polypropylene suture technique was associated with no periprosthetic leakage, a significantly lesser incidence than when continuous 2-0 or 3-0 suture was used, but not different (p = 0.4, Fisher's exact test) from that with the interrupted pledgeted technique (Table 2). Four different faculty surgeons used the continuous 2-0 or 3-0 (2 patients) polypropylene suture technique, and there was no difference (p = 0.995) in the incidence of periprosthetic leakage. Analysis of the group with either continu rr ALL OTHER SUTURE TECHNIQUES I- CONTINUOUS 2-0 POLYPROPYLENE 1 I p = MONTHS AFTER OPERATION Fig 2. Actuarial analysis of freedom from periprosthetic leakage not associated with infection according to whether a continuous 2-0 polypropylene suture technique was used or any of the other techniques, primarily continuous 1-0 polypropylene or interrupted pledgeted mattress sutures. The p value refers to that obtained by Cox single-variate regression. (Presentation is as in Figure 1.)

6 175 Dhasmana et al: Periprosthetic Leakage following MVR Table 2. Effect of Suture Technique on Periprosthetic Leakage without Infection following Primary Mitral Valve Replacement in 435 Patients Leakage No. of Suture Method Patients No. YO CL Significance Continuous polypropylene p (Fisher) = or 3-0 (2) YO 1 p (Cox) = Interrupted pledgeted p (Fisher) = % p (Cox) = 0.4 Unknown or mixed % CL = 70% confidence limits. Table 3. Effect of Suture Technique on Periprosthetic Leakage without Infection following Primary Mitral Valve Replacement in 435 Patients Periprosthetic Leakage No. of Suture Technique (Pure Groups) Patients No. YO CL Significance I Continuous % % p (Fisher) = 0.02 Continuous % 0-3% I Interrupted pledgeted % O.I-2.7% } p (Fisher) = 0.6 Total 373a % % 5xty-two patients in whom one or more supplemental interrupted sutures were placed are excluded. CL = 70% confidence limits. ous suture alone or interrupted pledgeted mattress sutures alone (373 patients) gave similar results, in that the incidence of periprosthetic leakage was higher with continuous 2-0 polypropylene suture than with either continuous 1-0 polypropylene or the interrupted suture technique; the incidence of periprosthetic leakage was very low and was similar with the latter two techniques (Table 3). Among the four surgeons using continuous 2-0 polypropylene only, the incidence of periprosthetic leakage was the same (p > 0.99 by chi-square test). Annular calcification increased the risk of periprosthetic leakage in both the simple (Table 4) and multivariate analysis (see Table 1). Electrocardiographic evidence of left ventricular hypertrophy and the patient being black possibly increased the risk of periprosthetic leakage (see Tables 1 and 4). Unrelated to the frequency of periprosthetic leakage was the functional type of mitral valve lesion (p = 0.3), its pathological basis (p = 0.99), or the type of prosthesis used. Periprosthetic Leakage Associated with Prosthetic Infection Among the 25 patients with documented leakage, 10 cases were associated with prosthetic valve infection. Two of these patients died before reoperation, while 8 came to prosthetic replacement or resuture with 1 hospital death. Subsequently, 1 of the 7 hospital survivors died suddenly (at 3 months); postmortem examination showed partial dehiscence of the second prosthesis and infection. Another patient is well, and a third has a scspected leak. The remaining 4 have required yet another valve replacement for repeated dehiscence, 2 dying at

7 ~ ~~ 176 The Annals of Thoracic Surgery Vol 35 No 2 February 1983 Table 4. Factors Analyzed for Association with Periprosthetic Leakage without lnfection following Primary Mitral Valve Replacement in 435 Patients Factor Present Factor Absent Variable Leakage Leakage Total No. of No. of P P No. Patients N YO CL Patients N % CL Value (Cox) Annular calcification % 4-13% % 2-4% LVH on preoperative ECG % 3-8% % 1-4% Black race % 3-13% % 2-4% Use of either heterograft prosthesis % 3-8% % 2-4% Systolic murmur on % 2-12% % 2-4% discharge Valvular calcification % 2-6% % 2-5% CL = 70% confidence limits; LVH = left ventricular hypertrophy; ECG = electrocardiogram reoperation. One of these hospital survivors is well, while the second has had the third prosthesis replaced and is now well. Thus, only 4 (40%; CL 22 to 61%) of the 10 patients suffering leakage associated with infection remain alive, and 1 of these is believed to have periprosthetic leakage again. These 10 instances of periprosthetic leakage occurred among the 19 patients (4.4% of the 435 patients; CL 3.4 to 5.6%) found to have welldocumented postoperative prosthetic valve infection, a 53% (CL 38 to 66%) incidence of periprosthetic leakage among patients with prosthetic valve endocarditis. To date, these infections have all become evident within the first two years after operation, and all the periprosthetic leaks have occurred within 6 months postoperatively. Actuarial freedom from prosthetic infection is 95.2% (CL 94.0 to 96.2%) beyond 24 months postoperatively (Fig 3). Of the 9 infected patients who did not develop leakage, 3 died of their infection without undergoing reoperation, 3 were treated medically and are well, and 3 have had prosthetic valve replacement for persistent infection. Two of these last 3 died late postoperatively of infection or suspected leakage associated with heart failure. Thus, of the 19 patients who developed prosthetic infection, 8 (42%; CL 29 to 57%) remain alive. The incidence of prosthetic valve endocar- ditis among the 435 patients was not affected by the suture technique (p for difference between continuous 2-0, continuous 1-0, and pledgeted mattress sutures = 0.6 by chi-square test). Among the 19 patients with prosthetic mitral endocarditis, the incidence of periprosthetic leakage likewise was not affected by the suture technique (p = 0.99 by chi-square test). Periprosthetic Leakage after Valve Replacement for Native Valve Endocarditis Although the 16 hospital survivors whose operation was done for native valve endocarditis were not part of the general analysis, only 1 (6%; CL 1 to 20%) developed evidence of periprosthetic leakage postoperatively (p for difference from the group without native valve endocarditis = 0.9). This occurred 2 months pos?operatively. Secondary valve replacement was done, but the patient died in the operating room. Comment We recognize that we may not have identified all patients with periprosthetic leak since we have not routinely performed left ventriculography in each patient in the study. Thus the determined incidence is a minimum one. Very probably, most instances of major leakage have been identified, since this usually leads to symptoms and reinvestigation. Any inade-

8 177 Dhasmana et al: Periprosthetic Leakage following MVR [L LL 91 ~ ~ ~ ~. " ". ~ " " " ' ~ " ' ~ ' " ' ~ " ' MONTHS AFTER OPERATION INFECTION WITH LEAKAGE = 0 INFECTION ONLY = 0 Fig 3. Actuarial analysis of prosthetic valve infection following primary mitral valve replacement. The closed circles represent 10 patients who developed periprosthetic valve leakage associated with the prosthetic infection; open circles represent the 9 patients without leakage. (Presentation is as in Figure 1.) quacy of identification of this event probably applies equally to all the subgroups and thus should not prejudice the results. The reported incidence of periprosthetic leakages varies from 0 to about 14% [15-191, though the data have not previously been presented as an actuarially calculated figure. Our incidence of documented leakage (7% actuarially) is in keeping with these figures. Our finding that native valve endocarditis had no effect on this incidence is also similar to the findings of others [20]. The development of prosthetic valve infection after mitral valve replacement has been shown by this study and by others to predispose strongly to periprosthetic leak. Further, the present study shows that this type of periprosthetic leakage carries a very poor prognosis. Our study shows that in the absence of prosthetic valve infection, small size of the suture, 2-0, when a continuous polypropylene suture was used had an important effect on the incidence of periprosthetic leakage. This may result from the cutting effect on tissues and sewing rings of the small monofilament suture [24, 251 and its tendency to break with the passage of time. These events can result in valve dehiscence, since secure fixation by tissue encapsulation or ingrowth may never occur or at least may require many years. Our study does not confirm the idea of Beddermann and Borst [l] that the continuous suture method for mitral prosthesis insertion itself predisposes to periprosthetic leakage. In fact, when their data are analyzed for the mitral position alone, the p value for the difference between a continuous 2-0 technique and interrupted pledgeted mattress sutures was 0.3. We conclude that within the time frame of our study, both the continuous 1-0 polypropylene suture technique and the individual pledgeted mattress suture technique give excellent protection against postoperative periprosthetic leakage. As a corollary, when periprosthetic leakage does occur after use of these techniques, prosthetic valve infection should be strongly suspected. The current study supports the idea of others [24, 26, 271 that annular calcification is an incremental risk factor for the development of

9 178 The Annals of Thoracic Surgery Vol 35 No 2 February 1983 periprosthetic leak postoperatively. We cannot support the idea that mitral incompetence caused by valve prolapse [24, or ischemic heart disease [2, 221 predisposes to periprosthetic leakage. Supported in part by Program Project Grant HL from the National Heart, Lung, and Blood Institute, Bethesda, MD. We wish to thank our colleagues, Drs. Robert 8. Karp, Albert D. Pacifico, and George L. Zorn, for permitting us to include patients operated on by them in this study, and for their critique of the study. Mrs. Kathy Peterson assisted us in contacting the patients and their doctors, and Mr. R. N. Brown helped with the data analysis. Dr. Edwin Bradley, Associate Professor of Biostatistics, provided statistical consultation and advice. We appreciate the skillful help of Ms. Sandy O Brien in the many editorial and graphic tasks. References 1. Beddermann C, Borst HG: Comparison of two suture techniques and materials: relationship to perivalvular leaks after cardiac valve replacement. Cardiovasc Dis, Bull Texas Heart Inst 5:354, Pluth JR, Curtis JJ: Modified continuous suture technique for mitral valve replacement. Ann Thorac Surg 25:459, Godwin RJ: Cineradiographic assessment of Bjork-Shiley aortic and mitral prosthetic heart valves. Clin Radio1 28:355, Ikaheimo M, Takkunen J: Echocardiography after mitral valve replacement and criteria of paraprosthetic regurgitation. Ann Clin Res 9:25, Horowitz MS, Goodman DJ, Hancock EW, Popp RL: Noninvasive diagnosis of complications of the mitral bioprosthesis. J Thorac Cardiovasc Surg 71:450, Barnal-Ramirez JA, Phillips JH: Echocardiographic study of malfunction of the Bjork-Shiley prosthetic heart valve in the mitral position. Am J Cardiol 4049, Willerson JT, Kastor JA, Dinsmore RE, et al: Non-invasive assessment of prosthetic mitral paravalvular and intravalvular regurgitation. Br Heart J 34:561, Cotter L, Miller HC: Clinical and haemodynamic evaluation of mounted porcine heterograft in mitral position. Br Heart J 41:412, Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. Am Stat Assoc J 53:457, Ku HH: Notes on the use of propagation of error formulas. J Res Nat Bureau Standards 70C:263, Chiang CL: Introduction to Stochastic Processes in Biostatistics. New York, Wiley, 1968, pp Walker SH, Duncan DB: Estimation of the probability of an event as a function of several independent variables. Biometrika 54:167, Cox DR: Regression models and life tables. J R Stat SOC B 34:187, Breslow NE: Covariance analysis of censored survival data. Biometrics 30:89, Bjork VO, Henze A: Ten years experience with the Bjork-Shiley tilting disc valve. J Thorac Cardiovasc Surg 78:331, Weldon CS, Ferguson TB: The elimination of periprosthetic leaks as a complication of mitral valve replacement. Ann Thorac Surg 1847, Kastor JA, Akbarian M, Buckley MJ, et al: Paravalvular leaks and hemolytic anemia following insertion of Starr-Edwards aortic and mitral valves. J Thorac Cardiovasc Surg 56:279, Swedberg J, Larsson S, Roberts D, Sudow G: Our experience with the Carpentier-Edwards bioprosthesis. Scand J Thorac Cardiovasc Surg 13:33, Maronas JM, Rufilanchas JJ, Villagra F, et al: Reoperation for dysfunction of the Bjork-Shiley mitral disc prosthesis. Am Heart J 93:316, Becker RM, Frishman W, Frater RWM: Surgery for mitral valve endocarditis. Chest 75:314, Hairston P, Lee WH Jr: Management of infected prosthetic heart valves. Ann Thorac Surg 9:229, Slaughter L, Moms JE, Starr A: Prosthetic valvular endocarditis. Circulation 47:1319, Dismukes WE, Karchmer AW, Buckley MJ, et al: Prosthetic valve endocarditis. Circulation 48:365, Favaloro RG, Effler DB, Groves LK, et al: Surgical repair of leaking prosthetic heart valves. Ann Thorac Surg 393, Fishman NH, Hutchinson JC, Roe BB: Prevention of prosthetic cardiac valve detachment. Surgery 67:867, d Allaines C, Blondeau P, Piwnica A, et al: Reinterventions sur les protheses valvulaires mitrales. Arch Ma1 Coeur 9:923, Ben-Ismail M, Descaves C, Ziza JM, Bousnina A: Des insertions de protheses valvulaires. Arch Ma1 Coeur , Singh HM, Davies LG, Rosser THL: Prosthetic valve leakage following cardiac valve replacements. Chest 61:258, Murtra M, Castells E, Puig-Massana M, et al: Continuous sutures in replacement of mitral and tricuspid valves with prostheses. J Thorac Cardiovasc Surg 69:132, 1975