Surgical Treatment of Prosthetic Valve Endocarditis

Transcription

1 Surgical Treatment of Prosthetic Valve Endocarditis William A. Baumgartner, M.D., D. Craig Miller, M.D., Bruce A. Reitz, M.D., Philip E. Oyer, M.D., Stuart W. Jamieson, M.B., B.S., Edward B. Stinson, M.D., and Norman E. Shumway, M.D. ABSTRACT Prosthetic valve endocarditis (PVE) remains an uncommon but serious complication of cardiac valve replacement. Operative intervention is frequently required in its management, although timing and outcome are incompletely defined. We reviewed cases of PVE in 75 patients who underwent valve replacement at Stanford University Medical Center from August, 1966, to September, Patients were classified into two groups. The healed PVE group included those patients who had completed a planned course of antibiotic therapy (N = 12); operations performed prior to completion of such a course defined the active PVE group (N = 63). Average age for both groups was 53 years, with men predominating at a ratio of 2:l. Indications for operation included congestive heart failure (87% ), recurrent emboli (15%), and persistent sepsis (4%). Staphylococcal and streptococcal organisms accounted for 47% of late PVE cases (occurring after 6 days; N = 58). Staphylococcal organisms were the primary causative agents in 43% of cases of early PVE (occurring before 6 days; N = 7). The operative mortality rate for all patients with PVE was 23% (25% for active PVE, 8% for healed PVE); deaths were caused principally by myocardial or multiple system failure present before operation. The primary predictors of operative mortality, as determined by multiple regression analysis, were emboli, renal dysfunction, type of PVE, and valve site. Operative mortality for the subgroup with active early PVE was 57%. At the present time, 29 of 47 operative survivors with active PVE (62%) and 8 of 11 patients with healed PVE (73%) are living, with the average postoperative From the Department of Cardiovascular Surgery, Stanford University School of Medicine, Stanford, CA. Presented at the Eighteenth Annual Meeting of The Society of Thoracic Surgeons, Jan 11-13, 1982, New Orleans, LA. Address reprint requests to Dr. Baumgartner, Department of Cardiovascular Surgery, A26, Stanford University School of Medicine, Stanford, CA interval being 3.8 years (range, 2 months to 15 years). There was no correlation between duration of preoperative antibiotic therapy and either intraoperative bacteriological findings or operative outcome. The 5-year actuarial survival rate (mean f standard error of the mean) of patients with active PVE discharged from hospital was 54 f 9%. Recurredresidual endocarditis occurred in 7 patients (15%) with active PVE; actuarial analysis showed 76 f 8% of active PVE patients to be free of recurrent/ residual endocarditis at 5 years. Postoperative periprosthetic leakage developed in 12 patients (26%) with active PVE, necessitating replacement in 1; actuarial analysis showed % of patients to be free of periprosthetic leak at 5 years. The multivariate predictor of an overall unsatisfactory result (defined in terms of early and late cardiac deaths and operations for recurrenthesidual endocarditis, or periprosthetic leak, or both) was the type of endocarditis-early active PVE, late active PVE, and healed PVE. When this classification was removed from the multiple regression analysis, preoperative renal dysfunction emerged as the main predictor of an overall unsatisfactory result. Earlier surgical intervention before these risk factors develop should result in fewer complications and a higher survival rate. Prior to the advent of antibiotics, infective endocarditis was nearly always associated with a fatal outcome; subsequently, successful medical treatment became possible, resulting in survival rates varying from 5 to 9% [I, 21. Although considerable progress has been made in both medical and surgical treatment of infective endocarditis, certain subgroups of patients continue to be at high risk. Patients with prosthetic valve endocarditis (I'VE) represent one of these high-risk subgroups. The incidence of this uncommon but serious complication of valve replacement has decreased considerably since the report of Geraci's group in 1963 [3]. At present, 87 OOO3-4975/83/187-18$ by The Society of Thoracic Surgeons

2 88 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 most major centers report an overall incidence of prosthetic valve infection in the range of 2 to 4%. Only 4 to 5% of patients with I'VE are cured with medical treatment alone, even though the majority of reported cases are those associated with penicillin-sensitive organisms [2, The concept of early operation in patients with active endocarditis has recently gained support in the literature [ Older guidelines reserved operative intervention for patients with severe congestive heart failure, major valve dysfunction, or multiple septic emboli, and for those in whom the infection could not be eradicated by medical treatment. This "salvage" concept has recently been superseded by a more aggressive surgical approach, based on retrospective survivorship studies comparing medical and earlier surgical treatment [6, 7, 1, 141. The present report summarizes our experience with 75 consecutive patients undergoing surgical treatment of PVE between August, 1966, and September, The purpose of this review is to examine the objectives of surgical management of I'VE and to define the efficacy of this type of treatment when it is applied earlier in the course of the disease. Long-term results of operative intervention in PVE will also be examined and discussed, as data concerning the long-term morbidity and mortality associated with this surgical approach have not been available in the past. Patients and Methods Definitions Our classification of patients with PVE is based upon the duration of antibiotic administration prior to operative intervention. Healed PVE refers to those patients who complete a planned course of antibiotic therapy. When operation is performed prior to completion of such a course of medical treatment, the disease is defined as active PVE. Although this latter distinction is not a direct index of the activity of the infectious process itself, it correlates with both the operative risk and pathological findings at operation. Further classification according to the time of onset of PVE also proves useful. Early PVE refers to those patients in whom infection oc- curred within 6 days of initial valve replacement, while late PVE refers to those patients whose prostheses became infected 6 or more days after replacement. Clinical Characteristics Selected clinical characteristics of the 75 patients and details of their treatment are summarized in Table 1. The average age for the 63 patients with active PVE was 52 years (range, 9 to 78 years). For the 12 patients with healed PVE, the average age was 54 years (range, 34 to 72 years). There were twice as many men as women in this series, reflecting the predominance of aortic valve involvement in patients with PVE requiring operation as well as the overall predominance of men in the population undergoing aortic valve replacement. Antecedent events that may have provided a portal of entry for infection were identified in 41% of the patients with active PVE and 33% of the patients with healed PVE. Dental manipulation (usually without antibiotic prophylaxis) was the most common factor identified (46%), followed by invasive urological or other surgical procedures, and extracardiac infections. The majority of patients (6%), however, did not give a history of any episode of infection prior to the onset of PVE. Delay before definitive treatment of patients with active PVE is reflected in the average interval of 3.5 weeks between the occurrence of symptoms and the initiation of therapy. Duration of preoperative antibiotic treatment ranged from 1 day to 9 weeks (mean, 2.8 weeks). The association between duration of antibiotic treatment, operative outcome, and intraoperative bacteriological findings will be discussed later in this article. The number of patients with PVE treated surgically by the year of operation is shown in Figure 1. The increasing frequency of operation for PVE may be due to the increased number of patients at risk in the general population. No particular trend in terms of PVE class is readily apparent. The interval between original valve implantation and reoperation for PVE is shown in Figure 2; this interval ranged from 2 weeks to more than 8 years, with a mean of 28 months. Early active endocarditis, representing 11% of

3 89 Baumgartner et al: Prosthetic Valve Endocarditis Table 1. Clinical Characteristics and Treatment in 75 Patients Undergoing Operation for Prosthetic Valve Endocarditis Characteristic Mean age (yr)" Sex (M : F) Antecedent event identified (no. of patients) Mean interval between appearance of symptoms and treatmenta Mean duration of preoperative antibiotic treatmenp Mean interval between initial valve replacement and reoperation for PVE" Early PVE (N = 7) Late PVE (N = 56) "Values in parentheses are range. PVE = prosthetic valve endocarditis. Active PVE (N = 63) 52 (9-78) 1.7:l 26 (41%) 3.5 wk (2 days-38 wk) 2.8 wk (1 day-9 wk) 5.4 wk (2-8 wk) 3 mo (3-11 mo) Healed PVE (N = 12) 54 (34-72) 5: 1 4 (33%) F 12 a -- ; lo 8 a # z I ACTIVE PVE, EARLY (< 2 monthr)(n = 7) ACTIVE PVE, LATE (n = 56) HEALED PVE (n = 12) 2 n " YEARS OF OPERATION Fig 1. Number of operations for prosthetic valve endocarditis (PVE) by year ( ). Fig 2. Operative mortality for PVE as a function of duration of valve implantation. (REOP = reoperation.) patients with active PVE, occurred at a mean interval of 5.4 weeks postoperatively. Incidence of PVE appeared to be time related, with the largest number of cases occurring during the first 3 years after valve replacement. There was no strong correlation between the frequency of PVE and the type of prosthetic valve employed. The incidence of early active PVE was similar for patients with mechanical and xenograft prostheses. Since 1974, porcine xenograft bioprostheses have been used almost exclusively at our center. Infection isolated to the aortic valve was observed in 63% of patients with both active and healed PVE (Table 2). Nineteen percent of patients had infection 2 F" 18 OPERATIVE SURVIVORS 16 OPERATIVE DEATHS INTERVAL FROM INITIAL VALVE REPLACEMENT TO REOP FOR PVE (months)

4 9 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 Table 2. Previous Valve Type and Position Early Active PVE Late Active PVE Healed PVE Description (N = 7) (N = 56) (N = 12) Total Valve type Allograft Mechanical" Xenograft Valve position AVR MVR AVR + MVR AVR + MVR + TVR 1 1 TVR 1 1 "Twenty-four of 26 mechanical valves were Starr-Edwards. PVE = prosthetic valve endocarditis; AVR = aortic valve replacement; MVR = mitral valve replacement; TVR = tricuspid valve replacement. confined to the mitral valve. Twelve patients (16%) had previously undergone aortic valve replacement combined with mitral valve replacement. In this group, endocarditis was equally distributed among the two prosthetic sites at risk; there were four cases of isolated aortic valve involvement, four cases of isolated mitral valve involvement, and four cases in which there was infection present on both previously implanted prostheses. Thus, both mechanical and tissue valves exhibited susceptibility to infection. No correlation between the type of prosthetic valve and the intracardiac location of the prosthesis was apparent. Progressive, severe congestive heart failure dictated operative intervention in 87% of patients with active and healed PVE. Classification of the severity of congestive heart failure in these patients was further refined using a modified New York Heart Association (NYHA) functional classification system. Patients with severe congestive heart failure (NYHA Functional Class IV) who were in a low-output state-manifested by hypotension, decreased mental status and renal dysfunction-were reclassified as Class V. Classes I through V contained 7, 8, 42, 8 and 1 patients, respectively. Thus, 67% of patients were Functional Class I11 or IV and 13% were Class V. Defined by a creatinine level greater than 3 mg% and a blood urea nitrogen determination greater than twice the upper normal limits, renal dysfunction was present in 14 patients (22%) with active PVE. Clinically evident systemic emboli occurred in 11 patients, 4 of whom had congestive heart failure. Persistent sepsis was the indication for operation in 3 patients. Although there were similar proportions of patients requiring operation for congestive heart failure in the active and healed PVE categories, the severity and rate of progression of hemodynamic deterioration were greater in the active PVE group. Operative Techniques The basic features of our operative technique have remained unchanged throughout the period of analysis and have been reported previously [17, 181. All patients operated upon initially at Stanford and included in this study received preoperative prophylaxis. Between 1963 and 1967, methicillin, penicillin, and streptomycin were administered prior to operation and continued for a minimum of five days postoperatively. From 1967 until 1975, the antibiotic regimen was changed to a cephalosporin (Keflin) and nafcillin for a similar period. Since 1975, Keflin has been used alone for antibiotic prophylaxis. At present, cefamandole+ is administered preoperatively and then continued postoperatively for 48 hours. Two distinct types of myocardial protection were used in the 75 patients in this series. Profound topical hypothermia accompanied by moderate systemic hypothermia and low-flow, 'Mandol, Eli Ldly & Co., Indianapolis, IN.

5 91 Baumgartner et al: Prosthetic Valve Endocarditis low-pressure cardiopulmonary bypass was employed in 57 patients (76%). Single-dose cold potassium cardioplegia was used in combination with topical hypothermia in 18 patients. This latter technique, initiated in 1977, is currently used in the majority of patients. Study Methods Diagnosis of PVE was based on at least two of the following criteria: (1) two blood cultures positive for the same organism; (2) two or more of the following clinical signs: fever, new cardiac murmur, arterial embolus, or splenomegaly; and (3) histopathological evidence of infective endocarditis at operation. This latter criterion included the characteristic inflammatory and destructive changes noted at operation and upon examination of excised tissue. Endocarditis was considered the cause of late death if circumstances suggested a probable complication of infection (sepsis, arterial embolus, heart failure secondary to valvular regurgitation or obstruction, and sudden death), regardless of the elapsed time after operation for PVE. The term recurrentlresidual endocarditis was chosen because organisms were not always grown from cultures of the prostheses in those patients undergoing a second operation for PVE. Operative mortality was defined as death occurring within 3 days of operation. Current follow-up was obtained by direct contact with patients, or their primary physicians, or both. Follow-up information was obtained in 1% of surviving patients during a two-month closing interval (October and November, 1981). Continuous variables were compared using the t test technique, and categorical variables by chi-square of Fisher s exact test contingency analysis. Two-tailed p values greater than or equal to.5 were judged not to be significant. Actuarial curves were calculated using the Kaplan-Meier method [ 191 and compared using the Gehan technique [2]. All data are presented as mean plus or minus one standard deviation (SD), except for the actuarial probabilities, which are expressed as plus or minus one standard error of the mean (SEM). Linearized rates (number of events divided by number of patient-years at risk multiplied by 1) (% patient- year or events per 1 patient-years) were compared using Cox s F test [21] and expressed as plus or minus one SEM (approximately equal to 7% confidence limits). Ratios (e.g., operative mortality) are expressed with k 7% confidence limits. The PLR stepwise logistic multivariate regression program of the UCLA Biomedical Computer Programs was used in the analysis. This program investigates the relation between a binary dependent variable and a set of independent variables. The procedure estimates the parameters (PI) of the linear logistic model as described by Cox [22]: E(y) = epx/(l + e)px The program proceeds in a forward stepwise manner, entering or removing one term from the model at each step. Selection of terms to be moved into or out of the model is based on an approximate asymptotic covariance matrix. Actuarial survival curves included all deaths, regardless of cause. Only cardiac deaths were considered for the multiple regression dependent variable. Results Bacteriological and Histological Findings Organisms responsible for infection and correlation with operative mortality rates are summarized in Table 3. The number and frequency of causative organisms in this series are similar to those reported in other studies. Streptococcal organisms were the most common bacteria identified in both active and healed PVE groups. Of these, Streptococcus viridans was involved in 44% of cases. Staphylococcal organisms represented the second largest group of bacteria (21%), followed by enterococcus (8%). The remaining cases of endocarditis were due to various gram-negative bacilli and fungal organisms. Nineteen patients (25%) underwent operation without identification of a causative agent ( abacteremic PVE). All patients, however, had characteristic operative or histopathological findings at operation. Microscopic gram stains, or cultures of excised valves, or both were performed in all patients with healed PVE. Only 2 patients in this group (17%) had positive results for one or both

6 92 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 Table 3. Relationship between Etiology of Prosthetic Valve Endocarditis and Operative Mortality Active PVE Healed PVE No. of Operative No. of Operative Causative Organism Patients Mortality (%) Patients Mortality (%) Streptococcus viridans Anaerobic streptococcus 1 P-hemolytic streptococcus 1 Streptococcus bovis 2 Enterococcus 6 Group D streptococcusa 3 Staphylococcus aureus 3 Staphylococcus epidermidis 11 Diphtheroids 1 Pseudomonas aeruginosa 3 Escherichia coli Serratia marascens 1 Listeria monocytogenes 1 1 Peptococcus magnus 1 Bacteroides corrodans 1 Candida albicans 2 Actinomycetes 1 Unknown Total Not Enterococcus. PVE = prosthetic valve endocarditis. Table 4. Relationship between Valve Culture and Stain in Active Endocarditis (N = 41) No. of Test Results Patients Percent CONCORDANT Negative culture/negative stain Positive culture/positive stain 3 4 Total DISCORDANT Negative culture/positive stain 7 17 Positive culturehegative stain 5 12 Total Number of patients who had both tests performed. tests. In contrast, 32% of patients with active PVE had positive results for one or both tests; this was not statistically significant. Discordance was present in 29% of patients in whom both tests were performed (Table 4). The results for these 12 patients were almost equally di- vided among negative culture/positive stain and positive culturehegative stain groups. The relationship between duration of preoperative antibiotic therapy and results of intraoperative bacteriological findings (culture, gram stain, or both) and operative outcome is illustrated in Figure 3. Duration of preoperative antibiotic therapy did not correlate with the presence or absence of organisms in excised valves, nor was there a correlation between the length of preoperative therapy and postoperative outcome. Operative Findings and Procedures Pathological findings noted at operation are summarized in Table 5. Vegetations were the most common feature observed in both active and healed PVE. Dehiscence and perivalvular abscess were seen together in approximately one-half of patients with active PVE. The relationship between the infectious agent and abscess formation is illustrated in Figure 4. Organisms associated with a high incidence of annular

7 93 Baumgartner et al: Prosthetic Valve Endocarditis ooperatlve SURVIVORS OPERATIVE DEATHS PATIENTS WITH ABSCESS PATIENTS WITHOUT & 12 k- 1 m 8 5 z n ooom W - 3m OD ooaaum oooea s NEGATIVE.b ao. W- POSITIVE RESULTS OF INTRAOPERATIVE CULTURE/STAIN Fig 3. Relationship between duration of preoperative antibiotic therapy and results of intraoperative bacteriological findings (culture, gram stain, or both) and operative outcome. Table 5. Pathological Findings at Operation Finding Active PVE Healed PVE No. of No. of Patients Percent Patients Percent Vegetations 44/63 7 5/12 41 Dehiscence 35/ /12 25 Gross perivalvular 31/ /12 17 abscess Aneurysm 4/63 6 Leaflet destruc tiona Perforation" 5/ "Tissue valves only. PVE = prosthetic valve endocarditis. abscess formation were enterococcus (loo%), Staphylococcus epiderm idis (55 % ), Ca ndida a1 bicans (5%), and gram-negative bacteria (38%). Patients with "abacteremic" endocarditis also had a high frequency of abscess formation (44%). These findings agree with other reports [14]. Annular abscess formation also correlated significantly with the type of prosthetic valve Fig 4. Relationship between infectious agent and abscess formation in patients with PVE. employed. Sixty-five percent of mechanical prostheses removed were associated with abscess formation, compared with 36% of xenograft prostheses (p =.4). Neither abscess formation nor prosthetic valve dehiscence, however, was significantly related to operative outcome. Leaflet destruction and perforation of xenograft prostheses occurred in both the early and late active PVE groups. Endocarditis limited to the leaflets themselves was found in 2 of 39 patients (51 %) with xenograft bioprostheses. Preoperative conduction abnormalities such as left bundle branch block and left hemiblock occurred in 16 patients (25%) with active PVE. Atrioventricular conduction disturbance often signals extension of infection into the septum; indeed, 11 of these 16 patients (69%) were subsequently found to have annular abscesses. Morbidity and Mortality Causes of operative deaths are summarized in Table 6. The overall operative mortality rate was 23 & 5% ( 2 7% confidence limits; 17/75). Operative mortality rates for the active and healed PVE groups were 25 & 6% and 8 & 8%, respectively; these rates were not statistically significant. Myocardial failure, either alone or associated with multisystem failure and sepsis, was the cause of death in the majority of patients (75%). Further analysis of patients with active PVE divided into early and late categories showed operative mortality rates of 57 f 19%

8 ~~ 94 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 Table 6. Causes of Operative Mortality Cause of Death Myocardial failure 9 Multisystem failure and sepsis 3 Technical error 2 Cerebrovascular accident 2 Total 16 Active PVE (N = 63)" Healed PVE (N = 12) No. of Mortality No. of Mortality Deaths (%) Deaths ("/.I f gbec Bbtc "There were 4 deaths among the 7 patients with early active PVE, for an operative mortality of 57? 19%. There were 12 deaths among the 56 patients with late active PVE, for an operative mortality of 21? 6%. bstandard deviation. 'Not statistically significant. PVE = prosthetic valve endocarditis. for patients developing active PVE within two months of prosthetic valve implantation and 21? 6% for patients with late active PVE. This difference was also not significant. Correlation of the operative outcome with the etiological agent is summarized in Table 3. Both enterococcal and staphylococcal active PVE were associated with higher operative risks. Although the numbers of patients with gramnegative and fungal organisms are small, it appears that these organisms are also associated with increased operative risk, as reported in other series [5, 7, 23, 241. Valve position did not correlate significantly with operative mortality rate when considered as an isolated variable in our patient population. Thirteen of 47 patients (28%) with prosthetic aortic valve endocarditis died in the perioperative period, while 3 of 14 patients (21%) with prosthetic mitral valve endocarditis died; this was not statistically significant. However, when 18 covariates were considered simultaneously, Table 7. Results of Univariate and Multivariate Analyses: Predictors of Unsatisfactory Outcome Univariate Analysis Multivariate Analysis Independent Variables F Value p Value F Value p Value UNSATISFACTORY OUTCOME = OPERATIVE DEATH Emboli Renal dysfunction Valve site (aortic vs. mitral) PVE type" NYHA classification NS NS NS UNSATISFACTORY OUTCOME = ALL CARDIAC DEATHS -k REOPERATION FOR PERIPROSTHETIC LEAK OR RECURRENTIRESIDUAL ENDOCARDITIS PVE type" Renal dysfunction Emboli NYHA classification Valve site (aortic vs. mitral) NS.133 NS NS.125 NS "Classified as early active PVE, late active PVE, and healed PVE. PVE = prosthetic valve endocarditis; NYHA = New York Heart Association NS NS NS NS.8

9 95 Baumgartner et al: Prosthetic Valve Endocarditis Table 8. Causes of Late Death No. of Deaths Active PVE Healed PVE Cause (N = 47) (N = 11) Myocardial failure secondary 5 to recurrentlresidual PVE Congestive heart failure 4 1 "Sudden death" 2 2 Cerebrovascular accident 2 Aortic dissection 1 Renal failure 1 Cancer 1 Suicide 1 Myocardial infarction 1 Total 18 3 PVE = prosthetic valve endocarditis. aortic valve PVE was found to be associated with a higher operative risk than mitral valve I'VE (Table 7). Causes of late deaths of discharged patients are summarized in Table 8. The majority of deaths were due to left ventricular failure, either secondary to recurrentlresidual I'VE or as a result of the initial episode of PVE. Late functional results of discharged patients are summarized in Table 9. Cumulative follow-up was 171 patient-years for the active PVE group and 45 patient-years for the healed PVE group. Average follow-up for patients with active I'VE was 3.6 years, with a maximum follow-up of 15 years. Patients in the healed category had a similar follow-up interval of 4.1 years, with a maximum follow-up of 12.3 years. Recurrenthesidual endocarditis was seen in 7 patients (15%) with active PVE and 1 patient (9%) with healed PVE. The linearized and actuarial rates of recurrent/residual PVE were similar for both groups. Seventy-six percent of patients with active PVE were free of recurrent/ residual endocarditis at five years; similarly, 75 & 22% of patients with healed PVE were free of endocarditis at five years. Recurrenthesidual endocarditis necessitated another operation in 7 of 8 patients (88%). Reoperative mortality for this specific subset of patients was 71%. Two Table 9. Late Follow-up in Patients Discharged from Hospital Variable Cumulative follow-up (patient-yr) No. of survivors Average follow-up (yr) Maximum follow-up (yr) PERIPROSTHETIC LEAKAGE Active PVE (N = 47) Healed PVE (N = 11) No. of patientsa 12 (26) 1 (9) Linearized rate of occurrence per patient-year (%) 7. f 2' ' Percent free of leakage at 5-year follow-upb ' 8 * 18' No. requiring reoperation Linearized rate of occurrence per patient-year (%) Percent free at 5-year follow-upb f 2' 64 f 9' ' 8 f 18' RECURRENT/RESIDUAL ENDOCARDITIS No. of patientsa Linearized rate of occurrence per patient-year (%) Percent free of endocarditis at 5-year follow-upb No. requiring reoperation Linearized rate of occurrence per patient-year (%) Percent free at 5-year follow-upb 'Values in parentheses are percentages. bdischarged patients. 'Not statistically significant. PVE = prosthetic valve endocarditis. 7 (15) 4.1 & 2' 76 % 8' ' 8 f 8' 1 (9) 2.2 f 2' ' r 2' 75 f 22'

10 96 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 patients are currently alive. One of these has subsequently undergone two additional procedures for periprosthetic leaks, representing a total of four operative procedures since his initial valve replacement. Periprosthetic leaks developed in 13 of 58 survivors (22%); 11 of these 13 (85%) required subsequent reoperation. Seven of these 11 patients also had associated recurrenthesidual endocarditis. Actuarial analysis revealed that 65 f 9% of patients with active I'VE were free of periprosthetic leakage at five years. Similarly, 8 f 18% of patients with healed PVE were free of periprosthetic leakage at five years; this was not statistically significant. No correlation existed between periprosthetic leakage or recurrent/ residual endocarditis and positive culture or gram stain at time of operation. Discriminant Analyses Predictive of Unsatisfactory Outcome Univariate and multivariate discriminant regression analyses were employed to identify predictors of an unsatisfactory outcome. The dependent variable unsatisfactory outcome was defined in two ways: initial operative result (hospital death) or overall long-term outcome (hospital death, late death, or recurrenthesidual PVE or periprosthetic leakage requiring reoperation). Eighteen independent variables were employed in the multivariate analysis. These included: type of endocarditis (active or healed); organism (streptococcal, enterococcal, staphylococcal, gram-negative, fungal, culture negative); type of prosthesis initially inserted (allograft, xenograft, or mechanical); valve site (aortic, mitral, both, tricuspid); initial valve replacement due to native valve endocarditis (yes or no); sex; preoperative clinical characteristics (renal dysfunction, emboli, congestive heart failure and sepsis); modified New York Heart Association classification; and method of myocardial preservation (topical hypothermia versus topical hypothermia plus cardioplegia). Results are summarized in Table 7. Using operative death as the solitary measure of an unsatisfactory outcome, univariate analysis revealed that renal failure, emboli, and functional classification were all significant determinants. Multivariate analysis subsequently identified d Y 3 a : 2 5 z 1 U NSAT I S F AC EARLY LATE HEALED ACTIVE ACTIVE PV E PVE PV E Fig 5. Overall unsatisfactory results (defined as all cardiac deaths plus reoperation for periprosthetic leakage or recurrentlresidual endocarditis, or both) as a function of PVE classification. four independent variables predictive of operative death: (1) emboli (p =.2); (2) renal dysfunction (p =.17); (3) valve site (aortic versus mitral) (p =.18); and (4) classification of PVE (early active, late active, or healed) (p =.53). When the dependent variable unsatisfactory outcome was expanded to include all cardiac deaths related to endocarditis (early or late) and late reoperation for periprosthetic leakage, or recurrenthesidual endocarditis, or both (overall unsatisfactory outcome), univariate and multivariate analyses identified only one variable to be a significant predictor of an overall unsatisfactory outcome. This variable was the classification of PVE according to its onset and type, as described previously (Fig 5). Indeed, all patients with early active PVE had an overall unsatisfactory outcome. Clinical characteristics of these seven patients are summarized in Table 1. There were four operative deaths due to left ventricular pump failure and two late deaths due to cardiac failure as a sequela of endocarditis. One patient currently alive later required operative translocation of the aortic valve 125, 261 because of annular destruction resulting from previous necrotizing endocarditis. Further evaluation of this single significant determinant of overall unsatisfactory outcome was attempted. Multiple regression analysis was repeated, excluding the classification of

11 97 Baurngartner et al: Prosthetic Valve Endocarditis Table 1. Early (6-day) Prosthetic Valve Endocarditis in 7 Patients Year of Causative Recurrent/ Reopera tion Type of Valve Organism Residual PVE Status 1967 Aortic allograft E. coli Operative death 1969 Mitral allograft Candida No Late death 1969 Mitral allograft Unknown Yes Late death 1975 Aortic xenograft Staph. (-) Operative death 1976 Aortic xenograft Staph. (-) Operative death 1979 Aortic xenograft Peptococcus magnus No Alive 198 Aortic xenograft Staph. (+) Operative death PVE = prosthetic valve endocarditis; Staph. = staphylococcus. 1 I = NS ACTIVE PVE. DISCHARGED PATIENTS - HEALED PVE, DISCHARGED PATIENTS YEARS POSTOPERATIVELY Fig 6. Actuarial survival of patients with PVE who were discharged from hospital, divided into two groups according to type of PVE (active or healed). (SEM = standard error of the mean.) PVE. Renal dysfunction was then shown to be the solitary significant predictor of an unsatisfactory result. Since progressive renal failure is an end-organ indicator of the severity of congestive heart failure and low-output state, this fact implies that those patients with early and late active PVE who had an overall unsatisfactory result also had advanced congestive heart failure. Thirty-seven of 58 operative survivors (64%) were alive at an average of 3.8 years (range, 1 month to 15 years) postoperatively. The 5-year actuarial survival rate for patients with active PVE discharged from hospital was 54 f 9% (Fig 6). This was similar to that seen for patients with healed PVE. Thirty-four of the 37 current survivors (92%) were judged to be in NYHA Functional Class I, and the remaining 3 patients in Class Comment Therapeutic approaches to prosthetic valve endocarditis have gradually evolved during the past decade. Surgical intervention, once considered only a salvage procedure, is now advocated as a primary mode of treatment in combination with antibiotic therapy early in the course of the disease [ Older guidelines for the management of PVE reserved operation for patients with severe congestive heart failure, valvular dysfunction, or multiple septic emboli, or for those in whom medical treatment failed to eradicate the infection. The majority of patients with PVE treated surgically in the past had fulminant congestive heart failure; the results of such heroic attempts were frequently unsatisfactory, resulting in high mortality rates [4-7, 271. Postmortem examinations revealed that many patients with PVE did not respond to antibiotic therapy alone because the infection frequently extended into the surrounding annular and myocardial tissues [28, 291. These retrospective studies, combined with accrued clinical experience, subsequently allowed identification of features of PVE associated with increased mortality risks. Risk factors include congestive heart failure; new atrioventricular conduction disturbances; any degree of valvular dehiscence; relapse after appropriate antibiotic therapy; persistent fever after 7 to 1 days of antibiotic treatment; failure to achieve bloodstream sterility after 2 to 3 days of therapy; aneurysm formation (in cases of aortic PVE); emboli; early PVE; and nonstreptococcal bacteriological etiology. Focusing on operative mortality alone, multi-

12 98 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 ple regression analysis showed that embolism was the most significant determinant of operative mortality (F = 1.19), followed by renal dysfunction (F = 6.5), valve site (F = 3.59), and onset and type of PVE (F = 3.8) (see Table 7). Although functional classification per se was not found to be a significant (p =.9) multivariate determinant of operative mortality, a more powerful independent variable, renal dysfunction, was identified and probably superseded the predictive power of NYHA Functional Class. Renal failure represents an end-organ response to very advanced, decompensated congestive heart failure and low cardiac output in the majority of patients, but could have been due to immune complex glomerulonephritis in some [6]. Overall result, defined by early death, late death, and long-term complications (manifested by death from endocarditis or reoperation secondary to recurrenthesidual endocarditis, or periprosthetic leakage, or both) was most strongly influenced by PVE classification (F = 5.11). Those patients with early active PVE are at highest risk of having an overall unsatisfactory result, followed by those with late active PVE and with healed PVE. This single independent variable superseded all the predictive information inherent in every other independent variable in the multivariate analysis. When classification of PVE was removed from the analysis, renal dysfunction became the sole predictor of an overall unsatisfactory outcome (F = 3.74; p =.5). Congestive heart failure was the major indication for operation in our series, with 8% of patients categorized in NYHA Classes 111 through V. Our past experience [6, 11, as well as that reported by others [4-14, 31, has shown that operation for congestive heart failure is associated with inferior survival rates. Retrospective assessment of the efficacy of medical or surgical treatment of patients with aortic or mitral PVE strongly suggests that surgical intervention offers the best results, especially in patients who exhibit moderate or severe congestive heart failure [ Our analysis indicates that surgical intervention should be strongly considered for patients with early or late active PVE \>r for those who develop progressive azotemia during treatment. Operation performed prior to the development of renal dysfunction-a reflection of severe congestive heart failure-should then be associated with risks similar to those for isolated valve replacement in the absence of PVE. The argument for earlier operation in dealing with this group of high-risk patients is supported by the relatively low incidence of residual endocarditis following a second prosthetic valve replacement. The 15% incidence of postoperative recurrenthesidual endocarditis following prosthetic valve replacement noted in the present study compares favorably with that of other reported series [7], especially when the magnitude of follow-up is considered. The linearized rate for patients with active PVE was 4.1% per patient-year. For patients with healed prosthetic valve endocarditis, the linearized rate for recurrenthesidual endocarditis was 2.2% per patient-year. Fear that insertion of a prosthetic valve at the site of infection will almost certainly lead to recurrent PVE has not been borne out in studies of native valve endocarditis [6, 121, and similarly does not seem to apply to prosthetic valve endocarditis. There was no significant relationship between the positivity of cultures or gram stains at the time of replacement of the infected prosthesis and either the development of late recurrenthesidual endocarditis or operative survival in our series. We believe that continued medical therapy alone for patients who develop emboli, renal dysfunction, or moderate to severe congestive heart failure during antibiotic treatment is not warranted, since these complications portend inferior eariy and late overall surgical results. Cardiac conduction system abnormalities are seen primarily in patients with left-sided endocarditis (31, 321. These abnormalities, manifested as second degree or complete atrioventricular block, are most prevalent with I'VE of the aortic valve, and almost always indicate annular abscess with extension into the surrounding myocardium and septum. Of the 16 patients in our series who were noted to have preoperative conduction abnormalities, 11 (69%) were subsequently found to have annular abscesses. Organisms reported by other investigators [14] to be implicated in this destructive process have been staphylococcus and fungi; our results are

13 99 Baumgartner et at: Prosthetic Valve Endocarditis concordant, with the additional discrimination of enterococcus. These patients, however, did not exhibit a higher operative mortality rate (26% versus 25%) or a significant late mortality rate (52% versus 42%) compared with patients who did not have annular destruction. Dehiscence is one of the most common cardiac abnormalities seen in PVE, and is noted at postmortem examination in approximately 5% of patients [33]. Our findings parallel this observation, with 55% of operated patients having a partially or completely dehisced valve. This pathological finding, coupled with attendant cardiac decompensation, has been associated with mortality failures approaching 8% in the past. However, patients with dehiscence in our series did not exhibit significantly higher mortality rates compared with patients whose valve remained firmly seated (56% versus 44%). Duration of antibiotic treatment should pivot upon the clinical condition of the patient. Persistence of fever after 7 to 1 days of appropriate antibiotic treatment, relapse after appropriate therapy, and inability to achieve bloodstream sterility after 2 to 3 days of appropriate therapy are all indications for operative intervention. In our series, there was no correlation between the duration of preoperative antibiotic therapy and operative outcome. The pathology of PVE provided insight into reasons why these patients tend to be refractory to medical treatment only. Frequent involvement of the prosthetic valve annulus as well as myocardial invasion represent pathological conditions not easily remedied by antibiotic therapy alone. There was some correlation between annular abscess formation and the type of prosthesis used in our series. This disparity was particularly well illustrated when patients with mechanical valves were compared with those who had xenograft prostheses (65% versus 36%; p =.4). Conversely, in 5% of patients with xenografts, PVE was limited to the porcine leaflets, without any host tissue destruction "1 * Occurrence of an arterial embolus in a patient with known PVE and echocardiographic demonstration of residual vegetations represent another indication for early operation. In the absence of abnormal echocardiographic findings, the clinical occurrence of more than one arterial embolus, or with a single embolus that, if exacerbated by any additional embolic insult, might result in irreversible neurological or renal damage, constitutes an indication for early operation [13, 28, 351. Presence of preoperative emboli was the strongest predictor of operative death based on multiple regression analysis. With the definition of unsatisfactory outcome extended to include all early and late cardiac deaths and reoperations for periprosthetic leakage, or recurrenthesidual endocarditis, or both, PVE classification became the solitary significant determinant of overall unsatisfactory outcome. The time after initial operation of I'VE was a significant risk factor. Patients with early PVE represented a very high risk subset. In fact, this group of patients, while small in comparison with other reports [7, 1, 14, 271, had the highest operative mortality rate in our series. This discordance can be explained by the following facts. Figure 2 includes a number of patients with late endocarditis who were referred from other institutions and does not include patients with PVE who underwent medical treatment alone. The prosthetic valve remains susceptible to infection indefinitely, however, as the figure demonstrates. In addition, all of these patients had an overall unsatisfactory outcome. These findings are consistent with those previously reported [4, 5, 7, 12-14]. Staphylococcal organisms have been most commonly implicated in early PVE, although other organisms, such as gram-negative bacteria, fungi, and occasionally streptococci, have also been described [lo, 14, 271. In our series, one case of early active PVE was due to S. aureus, and two of seven (29%) were caused by S. epidermidis. The very grave prognosis of patients with staphylococcal prosthetic valve endocarditis has been well described [14]. Support for the concept of earlier operation for PVE is found in a number of recent reports. Richardson and colleagues [ 141 described 35 PVE patients in whom reoperation was required. The hospital mortality rate for these patients was compared in terms of the urgency of operation. Elective operations (next convenient operative date) had a mortality rate of 29%,

14 1 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 while urgent (next day) operations had a rate of 4% and emergency operations (immediate) were associated with a 62% mortality rate. These investigators also reported that surgical treatment had a lower mortality rate (44%) in PVE patients with moderate to severe heart failure than did medical treatment (1%) [14]. Saffle and associates [ 11 treated 6 consecutive patients who had active PVE (onset was early in 3 and late in the other 3), performing valve replacement with no operative fatalities. Wideman and co-workers (151 recently described their overall experience with rereplacement of the aortic valve. Prosthetic valve endocarditis was the indication for operation in 14% (27/2) of patients. The investigators identified preoperative functional status as expressed by NYHA functional class to be a significant incremental risk factor, supporting their conclusion of the importance of performing operation before advanced cardiac decompensation occurs. Survival of patients with fungal endocarditis treated medically is rare, although chemotherapy coupled with early operative intervention can result in early survival rates approaching 5% [23,24]. However, a permanent cure of fungal PVE has probably not been achieved either with medical treatment alone or with combined medical and surgical therapy. Long-term follow-up of patients with PVE treated operatively has not previously been available. Fifteen percent of the operative survivors of active PVE in our series developed recurrenthesidual endocarditis over an average interval of 3.6 years. This corresponds to a linearized rate of 4.1% per patient-year. By actuarial analysis, % of patients were free of recurrentkesidual endocarditis at 5 years. The majority of patients who developed I'VE again had to undergo reoperation. Although the risk of recurrenthesidual endocarditis is higher than that observed in native valve endocarditis [6], this risk seems reasonable and acceptable compared with the overall results attained using medical treatment alone [6, 7, 12-14]. Long-term follow-up (mean, 3.6 years) of 47 patients discharged from hospital with active PVE revealed 12 patients (26%) with periprosthetic leakage. This corresponds to a linearized rate of 7% per patient-year. 4ctuarial analysis showed that 65 & 9% of discharged patients were free of periprosthetic leakage at 5 years postoperatively. Eight patients had associated recurrenthesidual endocarditis. Periprosthetic leakage without PVE occurred in the remaining 4 patients, all of whom had undergone aortic valve replacement. These 4 patients are currently alive, l having undergone cardiac transplantation for progressive cardiac failure in the absence of intercedent infection. Numerous investigators [4-161 have stated that the cardiovascular status is the most important prognostic sign in patients with I'VE. This evidence, combined with a relatively low incidence of postoperative recurrenthesidual endocarditis, supports the rationale favoring early valve replacement for all high-risk patients with PVE. Having demonstrated no correlation between duration of antibiotic therapy and either the results of intraoperative cultures and gram stains or operative outcome, we believe that continuation of medical therapy in patients who develop intercedent complications (emboli, renal dysfunction, congestive heart failure) is probably imprudent, especially for patients with early active PVE and for those with late active PVE. Indication for operation in the majority of patients in our series was based on the older, established guidelines described previously. Despite this traditional surgical approach to the treatment of I'VE, our overall survival rate was 77%, even though many patients underwent emergency operation because of progressive heart failure. This salvage rate, which compares favorably with that of other reported series [lo, 13-15, 271, would unquestionably have been higher if operation had been performed prior to the development of advanced left ventricular failure, especially as manifested by renal dysfunction. Long-term survival has probably been enhanced, given that myocardial failure with its attendant complications was the cause of late death in 29% of patients discharged from hospital. In summary, previous guidelines supporting early operation for patients with active native valve endocarditis [6] can probably be applied to patients with PVE as well. These guidelines include:

15 11 Baumgartner et al: Prosthetic Valve Endocarditis 1. Initial appearance with severe heart failure caused by valvular dysfunction. 2. The development of moderate heart failure at any time after initiation of a course of antibiotic treatment, especially if associated with a newly appearing murmur of valvular regurgitation. 3. More than one clinically evident arterial embolus, or a single cerebral embolus that, if exacerbated by any additional embolic insult, would result in sufficient neurological damage to preclude rehabilitation. Further delineation of this particular guideline might include echocardiographic evidence of vegetations in a patient with a single arterial embolus. 4. Lack of improvement in the clinical state of systemic toxicity after one week of appropriate antibiotics. 5. Failure to achieve bloodstream sterility after two to three days of appropriate antibiotic therapy. 6. Evidence of progressive spread of intracardiac infection, as manifested by conduction system disturbances, aneurysm, or fistula formation. 7. Any degree of dehiscence-disruption of fixation-or interference with mechanical poppet or disc movement. 8. Infection caused by organisms not easily treated by antibiotics, including staphylococcus, gram-negative bacteria, and fungi. 9. Renal dysfunction. 1. Early PVE. 11. Selected patients with aortic PVE. These last three criteria, along with emboli, were identified in the present study as being the most powerful statistical determinants of operative death. Classification of the disease (e.g., early active PVE) was found to be the only variable predictive of overall early and late unsatisfactory results. Operation, combined with appropriate antibiotic treatment, should be performed early in the course of PVE complicated by unfavorable etiology, location, or events associated with a poor prognosis. In this study these events included emboli, renal dysfunction, onset and type of PVE, and valve site. Combined medical and sur- gical treatment probably represents the most prudent means of managing most patients with PVE in that it is associated with the best likelihood of successful outcome in the majority of these gravely ill individuals. We wish to thank Rupert Miller, Ph.D., for his biostatistical expertise and advice, and Voy Wiederhold for computer programming assistance. We also wish to thank Maryanne Mullins, Marta Gomez, Mary Ferran, Joan Miller, R.N., and Pat Gamberg, R.N., for assistance in manuscript preparation. References 1. Kaye D: Changes in the spectrum, diagnosis, and management of bacterial and fungal endocarditis. Med Clin North Am 57:941, Pelletier LL Jr, Petersdorf RG: Infective endocarditis: a review of 125 cases from the University of Washington Hospitals, Medicine 56: 287, Geraci JE, Dale AJD, McGoon DC: Bacterial endocarditis and endarteritis following cardiac operations. Wis Med J 62:32, Slaughter L, Morris JE, Starr A: Prosthetic valvular endocarditis: a 12-year review. Circulation 47:1318, Dismukes WE, Karchmer AW, Buckley MJ, et al: Prosthetic valve endocarditis: analysis of 38 cases. Circulation 48:365, Stinson EB: Surgical treatment of infective endocarditis. Prog Cardiovasc Dis 22:145, Wilson WR, Jaumin PM, Danielson GK, et al: Prosthetic valve endocarditis. Ann Intern Med 82751, Engleman RM, Chase RM, Boyd AD, et al: Lethal postoperative infections following cardiac surgery. Circulation 57:Suppl 3:31, Block PC, DeSanctis RW, Weinberg AN, Austen WG: Prosthetic valve endocarditis. J Thorac Cardiovasc Surg 6:54, Saffle JR, Gardner P, Schoenbaum SC, Wild W: Prosthetic valve endocarditis: the case for prompt valve replacement. J Thorac Cardiovasc Surg 73:416, Reitz BA, Baumgartner WA, Oyer PE, Stinson EB: Surgical treatment of infective endocarditis. In Bisno, AL (ed): Treatment of Infective Endocarditis. New York, Grune & Stratton, 1981, pp Wilson WR, Nichols DR, Thompson RL, et al: Infective endocarditis: therapeutic considerations. Am Heart J 1689, Karchmer AW, Dismukes WE, Buckley MJ, Austen WG: Late prosthetic valve endocarditis: clinical features influencing therapy. Am J Med 64:199, 1978

16 12 The Annals of Thoracic Surgery Vol 35 No 1 January Richardson JV, Karp RB, Kirklin JW, Dismukes WE: Treatment of infective endocarditis: a 1- year comparative analysis. Circulation 58:589, Wideman FE, Blackstone EH, Kirklin JW, et al: Hospital mortality of re-replacement of the aortic valve. J Thorac Cardiovasc Surg 82:692, Okies JE, Viroslav J, Williams TW: Endocarditis after cardiac valve replacement. Chest 59:198, Reitz BA, Baumgartner WA, Stinson EB: Myocardial protection by topical hypothermia. In Ionescu MI (ed): Techniques in Extracorporeal Circulation. London, Butterworth, 1981, pp Oyer PE, Shumway NE: Again, via the median sternotomy (editorial). Arch Surg 1954, Kaplan EL, Meier P: Non-parametric estimation from incomplete observations. J Am Stat Assoc 53:457, Gehan EA: Generalized Wilcoxon test for comparing arbitrary singly-censored samples. Biometrika 5:23, Cox DR: Some sample tests for Poisson variates. Biometrika 4:354, Cox DR: The Analysis of Binary Data. Methuen, London, McLeod R, Remington JS: Fungal endocarditis. In Rahimtoola SH (ed): Infective Endocarditis. New York, Grune & Stratton, 1977, pp Utley JR, Mills J, Roe BB: The role of valve replacement in the treatment of fungal endocarditis. J Thorac Cardiovasc Surg 69:255, Reitz BA, Stinson EB, Watson DC, et al: Translocation of the aortic valve for prosthetic valve endocarditis. J Thorac Cardiovasc Surg 81:212, Danielson GK, Titus JL, DuShane JW: Successful treatment of aortic valve endocarditis and aortic root abscesses by insertion of prosthetic valve in ascending aorta and placement of bypass grafts to coronary arteries. J Thorac Cardiovasc Surg 67443, Masur H, Johnson WD: Prosthetic valve endocarditis. J Thorac Cardiovasc Surg 8:31, Arnett EN, Roberts WC: Prosthetic valve endocarditis. Am J Cardiol 38:281, Anderson DJ, Bulkley BH, Hutchins CM: A clinicopathologic study of prosthetic valve endocarditis in 22 patients: morphologic basis for diagnosis and therapy. Am Heart J 94:325, Wilson WR, Danielson GK, Giuliani ER, et al: Cardiac valve replacement in congestive heart failure due to infective endocarditis. Mayo Clin Proc 54:223, Arnett EN, Roberts WC: Valve ring abscess in active infective endocarditis: frequency, location, and clues to clinical diagnosis from the study of 95 autopsy patients. Circulation 54:14, Roberts NK, Somerville J: Pathological signif- icance of electrocardiographic changes in aortic valve endocarditis. Br Heart J 31:395, Johnson WD Jr: Prosthetic valve endocarditis. In Kaye D (ed): Infective Endocarditis. Baltimore, University Park Press, 1976, pp Rossiter SJ, Stinson EB, Oyer PE, et al: Prosthetic valve endocarditis. J Thorac Cardiovasc Surg 76:795, Wilson WR, Geraci JE, Danielson GK, et al: Anticoagulant therapy and central nervous system complications in patients with prosthetic valve endocarditis. Circulation Discussion DR. TIMOTHY J. GARDNER (Baltimore, MD): Dr. Baumgartner and his associates at Stanford should be complimented on updating their evolving experience with a very difficult clinical problem. As is apparent from the present report, their overall results continue to be very good. Dr. Baumgartner s principal conclusion from this study should be emphasized: Management of prosthetic valve endocarditis needs to be dealt with in a fashion similar to that of native valve endocarditis. Excision of the prosthesis should be carried out promptly when there is actual or imminent left ventricular decompensation, when there is persistent sepsis, or when the potential for embolization exists. As unpleasant as reoperation might be, there is no justification for persisting with medical treatment when one of these conditions is present. This point is clearly established by the work of Dr. Baumgartner s group. There are two other observations in this study that are interesting and noteworthy, as both gainsay commonly held clinical assumptions. The first is that the absence or presence of annular abscess did not correlate with patient survival. The second is that both streptococcus and enterococcus were cultured in many patients who in fact had annular abscess formation. I would question two aspects of the authors data analysis, however. First is the confusing distinction made between active and healed prosthetic endocarditis. Since the incidence of positive operative cultures was similar in both groups, the so-called healed group-which was not really healed-differed from the active group only in terms of the virulence or possibly the extent of the infection and the timing of surgery. Second, merely introducing the concept of 3-day hospital mortality when discussing prosthetic valve endocarditis can be misleading. The real question is how many patients survive free of infection on a long-term basis. From my calculations of the data presented in this paper, long-term survival in the series was close to 55%. While this figure is still good, it makes the survival more comparable to that of other reported series.

17 13 Baumgartner et al: Prosthetic Valve Endocarditis One additional point. In a previous review of the experience with prosthetic valve endocarditis at Stanford presented in 1978, no difference was noted in the incidence of late infection between patients with mechanical and those with heterograft prostheses. Is this still true? Since 1974, the year in which Dr. Baumgartner and his colleagues began to implant porcine valves almost exclusively, they have encountered between 7 and ll cases of late prosthetic endocarditis annually, with the exception of a single year. Although the denominator, or the size of the patient group at risk, is not provided, this seems to be an unusually high incidence of late valve infection requiring operation. Since ventricular decompensation was present in 85% of the patients in the present report and since primary leaflet tissue infection was a common finding in patients with heterograft valves, is it not possible that a tissue valve substitute is either more susceptible to late infection, or more likely to fail when infected, than is a mechanical prosthesis? Finally, what was the total number of valve replacement patients at Stanford who developed late infection, and how many of these were successfully treated by medical therapy alone, or without reoperation? Was there any difference between the patients with mechanical valves and those with tissue valves? DR. PANAGIOTIS N. SYMBAS (Atlanta, GA): I enjoyed Dr. Baumgartner s presentation, and I congratulate him and his associates on their results. We too have recently reviewed our experience with surgically treated patients who have acute endocarditis, and we concur with most of the conclusions presented in this paper. I wish to comment on the management of partial separation of the aorta from the left ventricle or, as it is better known, partial aortic-left ventricular discontinuity, which results from extension of the infectious process of the valve cusps into the adjacent sutures, annulus, aorta, and other areas. During the last several years, we have encountered five patients with acute aortic endocarditis and aortic-left ventricular discontinuity. In two patients with minimal separation of the aorta from the left ventricle, a repair was done as has been previously described, with interrupted horizontal pledgeted mattress sutures placed through the left ventricular wall, then through the aorta and the ring of the prosthesis. In the remaining three, the gap between the aorta and the left ventricle was large. In order to repair the defect without tension and to provide a secure foundation for suturing the new valve without disturbing the coronary arteries, the defect was closed with a Dacron patch graft, and the valve was sutured to the patch. Four of the five patients are doing well 8 months to 6 years postoperatively. The fifth patient was lost to followup 3 months after operation. Have the authors encountered similar lesions in their series? If so, how have they managed these patients? Also, since the incidence of reinfection in this series was considerable, what form of antibiotic therapy did they use after the valve replacement? DR. WOLFGANG BIRCKS (Duesseldorf, W Germany): I have reported my experience with 22 cases of prosthetic valve endocarditis (J Cardiovasc Surg 13:3, 1972). At that time, almost all authors thought that prosthetic valve endocarditis was caused by an infection during the operation. We found bridging symptoms in all patients; therefore, my associates and I made no distinction between early and late prosthetic valve endocarditis. Since then-approximately 11 years-we have followed our valve replacement patients with respect to PVE. In my experience, late infection of prosthetic valves is extremely rare. I wish to ask Dr. Baumgartner one question: What is the real incidence of late PVE without bridging symptoms at the time of operation? DR. BRUNO J. MESSMER (Aachen, W Germany): I would like to ask Dr. Baumgartner to comment on the site of endocarditis in his series of bioprosthetic valves. In how many patients was true destruction of the valve cusps seen and not merely apposition of septic thrombi? We have found in our patients that the endocarditis was situated at the sewing ring, with thrombotic material growing over to the nearby cusp; however, we have rarely noted primary destruction of the cusps. DR. GEORGE J. MAGOVERN (Pittsburgh, PA): I enjoyed this paper, and I wish to second to some extent the comments of Dr. Gardner. My colleagues and I reviewed 5 patients with mechanical prostheses and noted a postoperative bacterial endocarditis rate of less than 2%. We also reviewed 35 patients with porcine xenograft prostheses; early and late the infection rate was somewhere around 6%. Next, we did routine cultures of the prostheses at the time of insertion in all patients. In over 1 instances, we could find no relationship between a culture that was positive at the time of operation and any infection that may have subsequently occurred. Thus, it is often possible to obtain a positive culture on the prosthesis at the time of operation, but this culture does not necessarily represent the late infection. I do know that our own incidence of late infection on porcine prostheses is much higher, and I wonder if that has been the experience of the authors in this series. DR. JEROME H. SACKS (Encino, CA): I wish to ask whether Dr. Baumgartner has any data to support his conclusion that prosthetic valve endocarditis due to S. aureus should be operated on early if other indications are not present. This has been propounded in the literature, but I wonder whether there are inde-

18 14 The Annals of Thoracic Surgery Vol 35 No 1 January 1983 pendent data in this series to support early operation for S. aureits endocarditis. DR. BAUMGARTNER: I thank the discussants for their thoughtful comments. We have always felt that there should be a distinction between active and healed endocarditis, though we realize that we stand alone in having this attitude. In dealing with prosthetic as well as infective native valve endocarditis, this distinction, although it is not a direct index of the activity of the infectious process, seems to correlate with the operative risk and the pathological findings at operation. Even though there was no statistically significant difference in mortality between the two groups due to the small numbers involved, there was certainly a difference on a crude percentage basis, operative mortality was 8% for the healed PVE group and 25% for the active PVE group. In answer to the questions regarding late versus early endocarditis, we break these infections down according to the 6-day differential primarily because this criterion again seems to predict operative outcome. Our small group of early PVE patients probably reflects two items. First, a number of patients in our series were referred from neighboring hospitals; at the time of referral, they had had valve replacement 6 days previously. Second, the study does not take into account those patients who were medically treated. In regard to infection following porcine valve implantation, we have had a 3.8% incidence of endocarditis after valve replacement, and porcine valves are used almost exclusively. Although the 15% incidence of recurrent endocarditis following operation is three to four times higher than that reported after initial valve replacement, this figure compares favorably with others reported in the literature. With this particular subset of patients, there was no correlation between the operative cul- tures and the development of recurrent endocarditis. There was also no significant correlation between the type of organism found at operation and the subsequent development of recurrent endocarditis. Our greatest problems in technical management of the patients in this series have been situations in which the annulus has been completely dissolved by the infectious process. Seven such patients underwent translocation of the aortic valve, which was described by Danielson and associates a number of years ago. Among these 7 patients, mortality has been quite high; only 2 are currently alive. Concerning the infective potential of the porcine valve, the majority of these valves exhibited infection limited to the leaflets themselves without any annular involvement. There was destruction or perforation of the leaflet in approximately a quarter of the patients with porcine valves. There was also no correlation between operative culture and the duration of preoperative antibiotic therapy on operative outcome in this group. The patients who were operated on in the active stage had had a previous course of antibiotics ranging from less than a day to nine weeks. In response to the question on S. aureus, Dr. Richardson and his colleagues believe that this particular organism demands immediate attention for operation. Our belief is not quite as strong as theirs, because our data do not completely support this contention. We do think that if S. aureus appears on a prosthetic valve in a patient within 6 days after operation, that patient should be strongly considered for opera tion. We have examined the echocardiographic findings in this series, but our cardiologists have only recently been able to accurately identify vegetations on the porcine valve. We believe that if vegetation is seen on a valve in a patient who has documented endocarditis, this would be an indication for operative intervention.