ABSTRACT Between 1961 and 1978, 6,602 valves were replaced in 5,660 patients. Reoperation for periprosthetic leakage was performed in 105 patients

Results of Reoperation for Periprosthetic Leakage Thomas A. Orszulak, M.D., Hartzell V. Schaff, M.D., Gordon K. Danielson, M.D., James R. Pluth, M.D., Francisco J. Puga, M.D., and Jeffrey M. Piehler, M.D. ABSTRACT Between 1961 and 1978, 6,602 valves were replaced in 5,660 patients. Reoperation for periprosthetic leakage was performed in 105 patients (1.6% of the valves); early mortality was 5.7%. At reoperation, 52% of patients were in New York Heart Association Classes I11 and IV, whereas 72% had been in Classes 111 and IV prior to the primary valve replacement. Seventy-five patients had aortic periprosthetic leaks, which were distributed equally around the annulus. Among these patients, 41 (55%) had aortic valve rereplacement and 34 (45%) had suture repair. At 5 years, the survival was 94% and the event-free survival was 7l%. Seventeen patients had multiple aortic valve reoperations. Hospital mortality for the second reoperation was 5.8%. At late follow-up, 19 patients had murmurs of residual or recurrent aortic periprosthetic leakage. Twenty-nine patients had mitral periprosthetic leaks, most often near the anterior leaflet. Eight patients had mitral valve rereplacement, and 21 had suture repair. At 5 years, the survival was 75% and the event-free survival was 52%. Four patients underwent multiple reoperations. At late follow-up, 5 of the 29 patients in the mitral valve group had murmurs of residual or recurrent periprosthetic leakage. One patient had tricuspid valve rereplacement. The low hospital mortality and the good late results have encouraged us to recommend an aggressive approach in the correction of periprosthetic leakage in all symptomatic and selected asymptomatic patients. Although the risks of cardiac valve replacement have decreased steadily during the past decade, certain technical features remain problematic. Periprosthetic leakage producing valvular insufficiency may lead to progressive congestive From the Division of Thoracic, Cardiovascular, Vascular, and General Surgery, Mayo Clinic and Mayo Foundation, Rochester, MN. Presented at the Twenty-ninth Annual Meeting of the Southern Thoracic Surgical Association, Hilton Head Island, SC, NOV 4-6, 1982. Address reprint requests to Dr. Orszulak, Mayo Clinic, 200 First St SW, Rochester, MN 55905. heart failure. Frequently, patients with periprosthetic leakage are observed expectantly, until serious cardiac decompensation develops. This strategy is based on the assumptions that reoperation for valve replacement has a greater risk than the primary operation and that reoperation is less likely to be successful. To define the operative risks and long-term survival better, we reviewed the entire Mayo Clinic experience with reoperation for periprosthetic leakage. Patients and Methods The records of all patients with periprosthetic leakage who were seen at the Mayo Clinic between 1961 and 1978 were reviewed. A total of 226 patients had the clinical diagnosis of perivalvular insufficiency; 190 of the patients had their initial valve replacement at our institution. One hundred five patients subsequently underwent reoperation for correction of periprosthetic leakage, and data on these patients form the basis of this report. The records were analyzed to identify possible risk factors for the development of periprosthetic leakage, as well as methods of repair and late survival. At reoperation, the mean age of the 105 patients (79 male and 26 female patients) was 50.6 years (range, 5 to 71 years). Two patients were less than 18 years of age: one was 5 and the other was 6 years old. Seventy-five patients had prior aortic valve replacement (AVR), 29 had mitral valve replacement (MVR), and l had tricuspid valve replacement. During this same interval, 6,602 valves were replaced in 5,660 patients. Thus, the incidence of reoperation for periprosthetic leakage was 1.6%. At the initial valve replacement, 30 patients had aortic stenosis, and 26 patients had aortic stenosis and insufficiency. Pure aortic insufficiency necessitated AVR in 16 patients. Among the patients undergoing atrioventricular valve replacement, 12 had mitral insufficiency, 12 had mixed mitral stenosis and insufficiency, and 1 had tricuspid insufficiency. 584

585 Orszulak et al: Reoperation for Periprosthetic Leakage Time from Valve Replacement to Diagnosis of Periprosthetic Leakage Time (mo) ~~ ~ <1 2-6 7-12 13-24 25-36 36-48 >48 Total "In 5 patients, the interval could not be determined. No. of Patients 16 26 18 16 3 7 14 100" Seventy-two percent of patients were in New York Heart Association (NYHA) Classes I11 and IV prior to the original operation. Surgical techniques at the initial valve replacement varied during the years of this study, but interrupted unreinforced braided nonabsorbable and silk sutures were used in 48% of patients. At the primary valve replacement, the Starr-Edwards valve was used in 62 patients, the Braunwald- Cutter valve in 27 patients, the Bjork-Shiley prosthesis in 5, a porcine heterograft in 3, and other prostheses in 8. At discharge from the hospital after valve replacement, 22 patients (21%) had murmurs consistent with periprosthetic leakage (11, AVR; 11, MVR). Forty percent of periprosthetic leaks were diagnosed within the first 6 months, and an additional 32% had been diagnosed within 2 years after operation (Table). At diagnosis of periprosthetic leakage, 52% of the patients were in NYHA Classes I11 and IV. Cardiac catheterization was performed in only 34 of the 105 patients. Seven of these patients had additional valvular diseases. The primary indications for reoperation were heart failure and hemolysis (87.6%). In 14 patients, the periprosthetic leak was found incidentally at reexploration for elective poppet replacement and strut debridement with Braunwald-Cutter valves. Pathological findings at reoperation were reviewed in an attempt to identify features potentially predisposing to the development of periprosthetic leakage. In the AVR group, periprosthetic leakage was noted equally in all three cusps of the annulus, whereas in the MVR group, the leak was more prominent in the annulus of the anterior leaflet. Review of the operative notes identified 3 patients with broken sutures and 33 in whom intact sutures had pulled through the annulus or in whom the annulus had "failed to heal"; 12 patients had prior history of endocarditis, and 3 of these were suspected of having infection at reoperation. Correction of mitral prosthetic valve leaks was accomplished by suture repair in 72% of patients, and valve replacement was performed in 28% of patients. For aortic periprosthetic leaks, suture repair was performed in 45% of patients (34/75), and the remainder had valve replacement (55%). In the 1 patient with tricuspid valve leakage, the valve was replaced. When the leak was repaired, pledgeted mattress sutures were used. Associated operations included four MVR procedures in patients with aortic valve leaks and five tricuspid valve replacements in patients with mitral valve leaks. Three patients also underwent saphenous vein coronary artery bypass grafting. The most frequent postoperative complication at reoperation was arrhythmia. There were 6 hospital deaths (5.7%). Of the hospital survivors, 76% were free of evidence of periprosthetic leakage at late follow-up. Among the 24 with periprosthetic leakage subsequent to first reoperation, 14 (58%) later required further valvular operations. The patients with persistent leakage included 16 who, at initial reoperation, underwent valve replacement (2, MVR; 14, AVR) and 8 who underwent repair (4, mitral valve; 4, aortic valve). Among the original 226 patients with perivalvular insufficiency, 21 required multiple valve operations for recurrent or persistent periprosthetic leakage. At the second reoperation, there was 1 hospital death (4.8%). Of 16 late survivors, 4 required a third reoperation; 2 of these patients had murmurs of persistent periprosthetic leakage but remain in NYHA Classes I and 11. Seven were in NYHA CIasses I and 11, and 3 were in Class 111; all 10 were without evidence of periprosthetic leakage. Among the 4 patients who required a third operation, there was no hospital mortality but 2 late cardiac deaths occurred, 1 at two months and the other at three years. Both deaths were the result of

586 The Annals of Thoracic Surgery Vol 35 No 6 June 1983 01 I 0 1 2 3 4 5 t (years1 Fig 1. Five-year survival after reoperation for periprostketic leakage. Numbers of patients are given in parentheses. congestive heart failure due to persistent periprosthetic leakage. In both instances, endocarditis was confirmed as the cause of the leak at the first or the second operation. The 4 patients who required a third operation included 3 who had AVR and 1 who had MVR. At exploration, all patients had notable detachment of their prostheses. One patient had five reoperations for periprosthetic leakage. At follow-up to 14 years, there were 68 survivors. Fifty-nine of the surviving patients (86.8%) were in NYHA Classes I and I1 at follow-up. Fourteen survivors (20.6%) continued to have murmurs that were consistent with periprosthetic leaks. Survival was calculated utilizing the Kaplan- Meier method for absolute and event-free survival; an event could be myocardial infarction, congestive heart failure, multiple reoperations for periprosthetic leakage, embolic episodes, or stroke. The 5-year absolute survival was 94% postoperatively for aortic periprosthetic leakage and 75% postoperatively for mitral peripros- thetic leakage (Fig 1). Of the late deaths, 12 were cardiac in origin. During a comparable period, the 5-year survival after primary valve replacement with Starr-Edwards valves was 75% for the aortic group and 68% for the mitral group. The 5-year event-free survival for patients with periprosthetic leakage was 71% for the aortic group and 52% for the mitral group. At 5 years in the primary valve replacement group, eventfree survival was 60% and 54% for aortic and mitral valves, respectively. Twenty-seven patients with Braunwald- Cutter valves were included in the group of 105. Fourteen underwent elective reoperation to prevent poppet escape and were observed incidentally to have periprosthetic leaks. Data on these patients were subsequently deleted from the survival figures and, when excluded, did not alter the survival rates. Among the 27 Braunwald-Cutter valves, 24 were in the aortic valve position, and 17 of these underwent repair, not replacement, for correction. Comment Periprosthetic leakage is a potentially fatal complication of cardiac valve replacement. Its recog-

587 Orszulak et al: Reoperation for Periprosthetic Leakage Fig 2. Use of pledgeted mattress sutures for placement of both aortic and mitral valves. Pledgets are placed on the ventricular side of the annulus, creating a sandwich of pledgets, annulus, and sewing ring (inset). nition dates to early laboratory experience and clinical use of prosthetic valves [l-111. The reported incidence of such leaks ranges from 0 [3] to 43% [6], and several papers dealing with causes and preventive techniques have been published. The goal of the present study was to delineate common perioperative factors that identify patients who are at risk of developing this complication and to examine long-term survival following repair of periprosthetic leakage. Factors that reportedly predispose to periprosthetic leakage include: (1) annular calcification [3-5, 7, 10, 12, 131, (2) infection [lo, 111, (3) annuloprosthetic disproportion 11, 3, 5, 7, (4) excessive tension on sutures, or annulus, or both [I, 3-5, 71, (5) poor suture placement and insufficient number of sutures [2, 5, 71, (6) suture breakage or deterioration [2,5, 71, and (7) inadequate fibrous ingrowth [4-71. Calcium is a frequent inhabitant of a diseased valve and its annulus. In the present series, 76% of the native valves were calcified. There are two reasons for the association of calcium with periprosthetic leaks. The first is that large amounts of calcium remaining in the annulus make it difficult to place the valve properly in solid circumferential apposition with the sewing ring. Calcium fracture at the suture tract, suture break due to calcium [12], or small crevices between calcium fragments allow for valve rocking and the devel- opment of periprosthetic leaks. The second reason is that radical debridement may effectively remove calcium but leave friable, thin tissue with poor tensile strength. It is important that there be a smooth bed for the sewing ring of the prosthesis, and that calcium be removed as thoroughly as possible [3-5, 7, 10, 12-14]. The presence of calcium places emphasis on suture type and technique for satisfactory, permanent valve replacement. One method that provides solid fixation utilizes pledgeted nonabsorbable mattress sutures placed such that a sandwich is constructed of pledget, annulus, and sewing ring (Fig 2). Although slightly more time-consuming, this construction is strong and permanent. Cardioplegia has permitted more time to perform this maneuver safely. In cardiac operations, infection can be disastrous. In the present series, endocarditis was implicated in 11% of patients undergoing operation for perivalvular leakage, and should be suspected strongly in patients with repeated leaks. Tissue cultures should be taken from the annulus since the gross appearance may show no abnormality. The 21 patients in our study who required multiple valve replacements included 9 with clinical evidence of endocarditis. Patient-valve mismatch ordinarily refers to the relationship of prosthetic valve orifice area to the patient s size [5, 7, 10-131. A similar mismatch between the prosthetic valve and the annulus may potentiate a periprosthetic leak. Frequently, the mismatch occurs with a prosthesis that is too large for an unmodified annulus.

588 The Annals of Thoracic Surgery Vol 35 No 6 June 1983 Squeezing a large prosthesis into a small annulus may result in tilting of the valve, with a portion of the sewing ring positioned above and not in contact with the annulus, or the sutures being pulled through when the valve is not properly seated. This mismatch may allow the valve to rock, thereby creating or enlarging a periprosthetic leak. A patient with a small annulus should be treated with an appropriatesized prosthesis or formal patch enlargement of the annulus. The alternative situation of placing a prosthesis that is too small for an annulus may be less frequent but also is a potential factor in periprosthetic leakage, requiring that tissue be gathered toward the sewing ring. This is totally dependent on the strength of the sutures and tissues. Any minor tear in the tissue by the suture can lead to periprosthetic leakage. This situation may be most common in MVR for mitral insufficiency, where the leaflets are often thin and delicate. The annulus, although not calcified, is greatly dilated and larger than a standard prosthesis. Unless this difference is recognized and reinforcement with pledgets is used, the sutures may tear away from the annulus, leaving a periprosthetic leak. We could not, from our data, determine the frequency of this problem in the present series. Suture placement is very important [2-8, 12, 15-17]. Bagirov and Umovist [15] demonstrated that the depth and width of the fibrous annulus does not exceed 2 mm, and deeper bites only penetrate soft myocardium or vulnerable structures. This finding supports the need for wellplaced, reinforced sutures rather than multiple, deep, interrupted sutures. As described by Beddermann and Borst [2], single running sutures are associated with the highest rate of leaks in AVR or MVR. The integrity of the suture material for valve replacement was examined to assess the role of suture fracture in the development of periprosthetic leakage. In our series, 3 patients were found to have one or more broken silk sutures. It is impossible to determine when the fractures occurred in relation to the development or enlargement of the leak. There is concern regarding the long-term integrity of silk compared with braided nonabsorbable sutures. In abdominal aortic and peripheral vascular surgery, silk sutures predispose to anastornotic aneurysms. A late deterioration of silk may also result in a late periprosthetic leak [18, 191, and we currently favor nonabsorbable braided synthetic suture. There has been considerable debate regarding the degree of fibrous ingrowth in prosthetic valve placement [I, 4, 61. Prior impressions were that fibrous ingrowth would provide firm, lasting fixation of a prosthetic heart valve. Clinical evidence has demonstrated that fibrous ingrowth alone may be inadequate to secure the valve sewing ring completely. Postmortem studies have shown that healing of prosthetic valves is similar to that seen after excision of left ventricular aneurysm [I, 6, 10,20,21]. Once the sutures are removed, the ventriculotomy is opened easily, and postmortem findings in patients with prosthetic heart valves show a shelling-out of the prosthesis despite years of postoperative survival* [6]. Fibrous tissue above and below the prosthetic valve sewing ring forms a C, with little, if any, true ingrowth into the fabric. Emphasis should continue to be placed on the suture material and technique, without heavy reliance on the delayed tissue ingrowth. When a periprosthetic leak develops, there may be some reluctance to recommend reoperation unless the clinical status of the patient has deteriorated considerably. With a mortality of 5.7% at reoperation and a 76% chance of cure, we believe that an approach to repair of periprosthetic leakage similar to that used in treatment of primary valvular insufficiency should be employed, that is, recommending repair for all symptomatic patients and for selected asymptomatic patients with evidence of left ventricular dysfunction. Our death and cure rates are comparable to those in primary valve replacement for a series of Starr-Edwards prostheses during a similar period. The late 5-year survival figures for patients with aortic and mitral periprosthetic valve leaks (94% and 75%, respectively) were better than those for patients having primary valve replacement (75% and 68%, respectively). The eventfree survival is similar in that freedom from thromboemboli alone was 79% for patients with Edwards WD: Personal communication, 1982.

589 Orszulak et al: Reoperation for Periprosthetic Leakage valves in the aortic position and 80% for those with valves in the mitral position. Patients who required a second operation for repair of periprosthetic leakage had an early mortality of 4.7%. Although the total numbers are small (N = 21), this finding suggests that the risk at reoperation is acceptable and that repeat operation should be considered for persistent perivalvular insufficiency following the initial attempt at repair. Reoperation is safer for patients in lower NYHA classes [22]. Recent work by Husebye and colleagues [23] illustrates that the mortality in all reoperations when patients are in NYHA Class IV and operated on urgently or on an emergency basis ranges from 55 to 66%. An indication for reoperation in patients with periprosthetic leakage may be only its presence, as even small defects can produce symptoms. Any obvious sign of patient deterioration, or any subtle sign of isolated progressive cardiac enlargement, warrants consideration of repair of periprosthetic leakage. Given the low hospital mortality and good late results in our series, we recommend an aggressive approach in the repair of periprosthetic leakage in all symptomatic patients and in selected asymptomatic patients. References 1. Edwards WS, Smith L: Aortic valve replacement with a subcoronary ball valve. Surg Forum 9:309, 1958 2. Beddermann C, Borst HG: Comparison of two suture techniques and materials: relationship to perivalvular leaks after cardiac valve replacement. Cardiovasc Dis (Bull Tex Heart Inst) 5:354, 1978 3. Weldon CS, Ferguson TB: The elimination of periprosthetic leaks as a complication of mitral valve replacement. Ann Thorac Surg 1847,1974 4. Frater RWM: Late disruption of aortic valve prostheses. Am J Surg 124:637, 1972 5. Favaloro RG, Effler DB, Groves LK, et al: Surgical repair of leaking prosthetic heart valves. Ann Thorac Surg 3:503, 1967 6. Wheat MW Jr, Linhart JW, Bartley TD, et al: Aortic ball-valve (Starr-Edwards) prostheses: a follow-up evaluation. Ann Thorac Surg 2:166, 1966 7. Herr R, Starr A, McCord CW, Wood JA: Special problems following valve replacement: embolus, leak, infection, red cell damage. Ann Thorac Surg 1:403, 1965 8. 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, 1968 9. Singh HM, Davies LG, Rosser THL: Prosthetic valve leakage following cardiac valve replacements. Chest 61:258, 1972 10. Roberts WC, Bulkley BH, Morrow AG: Pathologic anatomy of cardiac valve replacement: a study of 224 necropsy patients. Prog Cardiovasc Dis 15:539, 1973 11. Roberts WC, Hammer WJ: Cardiac pathology after valve replacement with a tilting disc prosthesis (Bjork-Shiley type): a study of 46 necropsy patients and 49 Bjork-Shiley prostheses. Am J Cardiol 371024, 1976 12. Danielson GK, Cooper E, Ifuku M: Severance of mitral prosthesis fixation sutures by annulus calcification. J Thorac Cardiovasc Surg 535358, 1967 13. Bulkley BH, Morrow AG, Roberts WC: Calcification of prosthetic valve anuli: a late complication of cardiac valve replacement. Am Heart J 87129, 1974 14. Roberts WC, Morrow AG: Causes of early postoperative death following cardiac valve replacement: clinicopathologic correlations in 64 patients studied at necropsy. J Thorac Cardiovasc Surg 54:422, 1967 15. Bagirov AM, Umovist MN: The question of the anatomical basis for the technique of suturing an artificial aortic valve. Klin Khir 721, 1966 16. Girardet RE, Wheat MW Jr: Technique of aortic valve replacement. J Thorac Cardiovasc Surg 7146, 1976 17. Seidel W, Brautigam E, Tauber G: Untersuchungen zur Festigkeit der Einheilung verschiedener Kunststoffgewebe von Herzklappenprosthesen. Thoraxchirurgie 17:288, 1969 18. Starr DS, Weatherford SC, Lawrie GM, et al: Suture material as a factor in the occurrence of anastomotic false aneurysms: an analysis of 26 cases. Arch Surg 114:412, 1979 19. Holmlund DEW: Suture technic and sutureholding capacity: a model study and a theoretical analysis. Am J Surg 134:616, 1977 20. Braunwald NS, Bonchek LI: Controlled tissue ingrowth on prosthetic cardiac valves: a new means of preventing thromboembolism. Rev Surg 23: 300, 1966 21. Loop FD, Effler DB: Left ventricular aneurysm. In Sabiston DC Jr, Spencer FC (eds): Gibbon s Surgery of the Chest. Third edition. Philadelphia, Saunders, 1976, p 1390 22. Shemin RJ, Gaudiani VA, Conkle DM, et al: Prosthetic aortic valves: indications for and results of reoperation. Arch Surg 114:63, 1979 23. Husebye DG, Pluth JR, Piehler JM, et al: Reoperation on prosthetic heart valves: an analysis of risk factors in 551 cases. J Thorac Cardiovasc Surg (in press, 1982)