Thirty-day mortality and long-term survival following surgery for prosthetic endocarditis: a study from the UK heart valve registry 1

Transcription

1 European Journal of Cardio-thoracic Surgery 14 (1998) Thirty-day mortality and long-term survival following surgery for prosthetic endocarditis: a study from the UK heart valve registry 1 Maria-Benedicta Edwards a, Chandana P. Ratnatunga a, *, Caroline J. Dore b, Kenneth M. Taylor a a United Kingdom Heart Valve Registry, Department of Cardiothoracic Surgery, Hammersmith Hospital, Du Cane Road, London W12 0HS, UK b Department of Medical Statistics and Evaluation, Imperial College School of Medicine, London, UK Received 29 September 1997; revised version received 22 April 1998; accepted 28 April 1998 Abstract Objective: To assess the 30-day mortality, long-term survival and freedom from reoperation following surgery for prosthetic endocarditis (PVE). Method: A retrospective analysis of data from the UK Heart Valve Registry of 322 patients who had undergone single mechanical/ bioprosthetic valve replacement for PVE between 1 January 1986 and 31 December The mean age was 54.9 ± 12.8 years and 213 (66.1%) were males. There were 170 aortic and 152 mitral valve implantations. Eighty-five (26%) of the infected valves were bioprosthetic and 237 (74%) were mechanical. Of the new prostheses implanted 53 (17%) were bioprosthetic and 269 (83%) were mechanical. Of those with infected bioprostheses, 50 (15.2%) had mechanical valves at redo surgery, whilst 219 (68.3%) of infected mechanical prostheses were re-replaced by mechanical prostheses. The follow-up was 98% complete with a total of patient years. Results: The 30-day mortality was 63 (19.9%; 95%CI %). There were 85 late deaths. One, 5 and 10 year survival rates were 67.1% ( %), 55.0% ( %) and 37.6% ( %), respectively. Age was the only significant determinant of 30-day mortality (P = 0.04). Age (P = 0.001) and explanting of infected bioprosthesis and replacement by mechanical valve (P = 0.04) determined long-term survival (P = 0.001). The incidence of re-reoperation was 9.9%. Freedom from reoperation for PVE was 88.4, 87.3 and 87.3% at 1, 5 and 10 years, respectively. Explanting of bioprosthesis and replacement by mechanical valve (P 0.001) and reoperation within 60 days of native valve replacement (P = 0.02) were determinants of reoperation for PVE. Freedom from death or reoperation was 61.1, 50.6 and 34.2% at 1, 5 and 10 years, respectively. Age (P = 0.003), explanting of bioprosthesis and replacement by mechanical valve (P = 0.002) and the period between prosthetic re-replacement (P = 0.04) determined freedom from death or reoperation. Conclusion: Operation for PVE carries a high 30-day mortality and reduced long-term survival. There is no evidence that type of prosthesis used for re-reoperation determines survival or freedom from re-reoperation Elsevier Science B.V. All rights reserved Keywords: Prosthetic valve endocarditis; Operative mortality; Long-term survival 1. Introduction Valve re-replacement for prosthetic endocarditis (PVE) presents the patient with an increased risk and the cardiac surgeon with a greater challenge than most other forms of valvular surgery. Although the cardiac surgeon may be anxious that infection was introduced or inadequately eradicated at the time of the previous operation, the overriding concern for both the patient and the cardiac surgeon must be * Corresponding author. Tel.: ; fax: Presented at the 11th Annual Meeting of the European Association for Cardio-thoracic Surgery, Copenhagen, Denmark, September 28 October 1, the high operative and late mortality reported in this condition [1,8,23] and the ability of the surgeon to completely eradicate the infection of PVE [16]. Several series have reported the operative mortality of PVE as varying between 20 and 65% [2 4,8]. More recently, however, certain centres have reported significantly lower operative mortalities varying between 6 and 13% [5,6,16]. These results reflect the experience of individual institutions and in some cases, of individual surgeons themselves. Thankfully the incidence of PVE, however, is low [6,9,13] and the majority of cardiac surgeons cannot, therefore, draw on a sizeable personal series of experience to undertake a meaningful analysis of their own surgical outcomes in this condition. It is not always appropriate for /98/$ Elsevier Science B.V. All rights reserved PII S (98)

2 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) such surgeons to base their operative predictions and judge their performance on results of institutions and individuals who have a declared interest in the management of this condition. There is a need, therefore, to provide a benchmark of outcome analysis for PVE surgery undertaken by a heterogenous group of cardiac surgeons representative of the whole cardiac surgical community. This report from the United Kingdom Heart Valve Registry (UKHVR) attempts to address this matter. The choice of prosthesis in re-replacement for PVE is an area of debate [15]. It is believed by some that the persistence of infection is promoted by the implantation of a mechanical prosthesis [8] and that the use, if not of a homograft [19], of a bioprosthesis increases the chance of successful eradication of infection. Others in this context have advocated the use of the pulmonary autograft [12]. Little nevertheless, is known about the comparative outcomes with the use of biological and mechanical prostheses in re-replacement for PVE, both in terms of survival and of freedom from re-operation for infection. 2. Methods Table 1 Preoperative characteristics in study population undergoing surgery for PVE (n = 322) Variable Sex Males 213 (66%) Females 109 (34%) Age Mean 4.9 ± 12.8 years Range years Size of prosthesis Mean (range) 27 (19 34) mm implanted Site Aortic 170 (53%) Mitral 152 (47%) Type of prosthesis Mechanical 237 (74%) explanted Biological 85 (26%) Type of prosthesis Mechanical 269 (83%) implanted Biological 53 (17%) Year of implantation Range Interoperative interval Early (within 60 days) 90 (28%) Early (within 365 days) 217 (67%) The UKHVR is a computerised database containing certain patient, surgeon and hospital identification details together with implant and selected follow-up data for patients undergoing valve replacement surgery in the majority of United Kingdom (UK) cardiac units. The Registry, established in January 1986, is funded by the Medical Devices Agency, an agency of the UK Department of Health and currently holds data on nearly valves implanted since Currently, 41 cardiac centres contribute to the Registry. In order to maintain simplicity and completeness of data collection and, to achieve maximum co-operation from all the participating centres, the Registry does not seek to obtain sequential follow-up data on patients apart from the occurrence of re-operation and death and the cause of these two outcomes. All patients entered into the UKHVR database are tracked by the Office for National Statistics (in England, Scotland and Wales) and the Central Services Agency (in Northern Ireland). Thus, the place, date and the certified cause of death together with any relevant available post-mortem data are recorded on all deceased patients entered in the Registry. Between January 1st 1986 and December 31st patients were entered into the UKHVR, and of these, 341 patients had re-replacement for PVE. For clarity of analysis only 322 patients who had single valve re-replacement for single valve PVE were included in the study population. The demography and the charateristics of this population are described in Table 1. There were no single valve re-replacements for PVE in the tricuspid position. The imbalance between the numbers of mechanical and biological valves explanted and implanted reflect the preponderance of use of mechanical valves in all types of valve replacement recorded in the Registry (mechanical valves 67% and bioprostheses 33% of all valves implanted between 1986 and 1995). Six patients were lost to follow-up with a resulting completeness of follow-up of 98% to the 31st March This resulted in total of years of follow-up for single valve PVE with a mean follow-up of 3.3 years and a range of years. 3. Statistical analysis All data are reported as the mean ± SD unless otherwise recorded. Certain data are also accompanied by 95% confidence intervals. In order to produce an accurate analysis and to avoid mixing between and within patient data, only the outcomes of the first re-replacement for PVE were considered. The Registry has accurate 30-day and time-related mortality data because of the Office for National Statistic and The Central Services Agency who provide death certificates with date of death. All deaths which occurred more than 30 days after operation were classified as late deaths. Two definitions of early PVE were considered for the purpose of analysis, one within 60 days [26] and the other within 1 year of operation [14]. The Registry variables collected for analysis were age, sex, size of prosthesis, site of valve surgery, type of explanted valve, type of implanted valve, types of explanted and implanted valves, year of operation and the inter-operation interval between the original (first) native valve operation and the subsequent (second) operation for PVE (see Table 1). These variables were used in a Cox proportional hazards regression analysis (univariate) to identify significant determinants of 30-day mortality, survival and freedom from re-operation using the Stata Release 5 statistical soft-

3 158 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) Fig. 1. Incidence of prosthetic endocarditis (PVE) by year of implantation, (the precentage this makes of the total number of valves implanted for each year is given in numbers and included as the shaded section on top of each annual column). ware package (Stata Corporation, College Station, TX). A logistic regression analysis was also performed using these variables to predict mortality within the first 30 days after surgery. The variables were similarly used in a stepwise Cox proportional hazards multiple regression (multivariate) analysis to identify determinants of 30-day mortality, overall survival and reoperation for PVE. Kaplan Meier survival curves were used for analyses of survival and freedom from re-operation. Statistical significance was defined as a P value of less than Results The distribution of the incidence of PVE for the time period studied, together with the percentage this PVE surgery represents of the total number of valves implanted for each year is illustrated in Fig. 1. The mean annual incidence of re-operation for PVE according to the Registry data is 0.5% with an annual range of between 0.3 to 0.8%. The data record a stable incidence of PVE through the study period. There were 63 deaths within 30 days with 30 day mortality of 20% (16 24%). There were 85 late deaths. The 1, 5 and 10 year survival rates for surgery for PVE were 67% (62 72%), 55% (49 61%) and 38% (28 47%), respectively (see Fig. 2). The results of the univariate analysis to identify predictors of 30-day mortality and long-term survival are given in Table 2. Age was shown to be the only variable for 30-day mortality to achieve significance when described as a continuous variable (P = 0.043). When the data were analysed with age as a categorical variable (by constructing four groups: 19 40, 41 60, and years), the influence of age on 30-day mortality is readily discernible (see Table 3). The size of prosthesis (P = 0.07) and type of explanted prosthesis (P = 0.06) were also close to achieving significance as determinants of 30-day mortality. These conclusions for 30-day mortality were similar when the logistic regression analysis was used. The influence of age persisted as a determinant of overall survival (P = ) Fig. 2. Freedom from death following surgery for prosthetic endocarditis, using Kaplan-Meier survival estimates.

4 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) Table 2 Determinants of 30-day mortality and overall survival following surgery for PVE; univariate (Cox proportional hazards regression) analysis 30-Day mortality (P) Overall survival (P) Age (continuous) (1.02; ) (1.03; ) Sex 0.76 (1.1) 0.80 (1.0) Size (0.94; ) (1.0; ) Site 0.14 (0.7) 0.47 (0.9) Explanted prosthesis 0.08 (1.6) 0.02 (1.5) Implanted prosthesis 0.93 (1.0) 0.58 (0.9) Explanted and implanted prosthesis Year of operation (0.93; ) (0.97; ) Interval between native and PVE operation (1.00; ) (1.00; ) Hazard ratios, and 95% confidence intervals (for continuous variables) are given in parenthesis. The hazard ratios for explanted and implanted prosthesis are not given because of the four categories present for this variable. Table 4 Thirty-day mortality and overall survival (95% confidence intervals in parenthesis) following surgery for PVE according to type of (a) implanted prosthesis (P = 0.9 and P = 0.6, respectively) and (b) explanted prosthesis (P = 0.08 and P = 0.02, respectively) (a) Biological (n = 85) Mechanical (n = 237) 30-day mortality (%) 19.1 ( ) 19.8 ( ) 1 year survival (%) 71.3 ( ) 66.2 ( ) 5 year survival (%) 56.3 ( ) 54.9 ( ) 10 year survival (%) 44.0 ( ) 33.1 ( ) (b) Biological (n = 85) Mechanical (n = 237) 30-day mortality (%) 26.1 ( ) 17.4 ( ) 1 year survival (%) 60.6 ( ) 69.4 ( ) 5 year survival (%) 45.4 ( ) 58.5 ( ) 10 year survival (%) 26.8 ( ) 42.3 ( ) and is also seen in Table 3. The type of explanted prosthesis (P = 0.06) was the only other determinant to approach significance as a determinant of overall survival. Multivariate analysis of all declared variables identified age as a determinant of 30-day mortality (P = 0.04; HR = 1.02[ ]). If the analysis was specifically asked to include the year of operation it identified age (P = 0.03; HR = 1.02; ) and year of operation (P = 0.06; HR = 0.92; ) as determinants of 30- day mortality. Multivariate analysis identified age also as a determinant of overall survival (P = 0.001; HR = 1.03[ ]) in the study population. In addition, the explanting of an infected bioprosthesis (P = 0.03; HR = 1.5), the replacement by bioprosthesis (P = 0.04; HR = 0.6) and the explanting of infected bioprosthesis and replacement with mechanical valve (P = 0.04; HR = 1.5) were all identified as additional determinants of overall survival. The 30-day mortality for re-replacement of an infected bioprosthesis was 26% and of a mechanical prosthesis was 17% (P = 0.08; HR = 1.6). There was a greater difference, however, in the longer-term survival of patients with the two types of prosthesis; 10 year survival of 27% and 42% bioprosthetic and mechanical valves, respectively (P = 0.02; HR = 1.5). The 30-day mortality for re-replacement by a biological and a mechanical valve was 19% and 20%, respectively (P = 0.9). There was no difference for overall survival by type of valve used for re-replacement: 44% (27 59%) and 33% (19 48%) at 10 years for biological and mechanical valves, respectively (P = 0.6) (see Table 4). Of the 85 infected bioprostheses explanted 35 (11%) were re-replaced by a bioprostheses (Group 1) and 50 (15%) were re-replaced by mechanical valves (Group 2). Of the 237 infected mechanical valves explanted 219 (68%) were replaced by mechanical valves (Group 3) and 18 (6%) were replaced by bioprostheses (Group 4). The 30-day mortality and 1, 5 and 10 year survival rates for re-replacement for PVE for Groups 1, 2, 3 and 4 is given in Table 5. With tissue valves there were no differences in outcome between the 35 re-replaced by bioprostheses and the 50 re-replaced by mechanical valves (P = 0.4). Similarly there was no difference in overall survival between the 219 mechanical valves re-replaced by mechanical valves and the 18 mechanical valves replaced by a bioprosthesis (P = 0.15). Classifying the operation on the native valve as the first operation and the operation for PVE as the second operation, 30 of the 322 patients had a further 3rd operation and one had a 4th operation because of recurrent PVE, giving a re-operation rate of 9.9%. The freedom from re-operation for PVE was 88% (84 92%), 87% (82 91%) and 87% (82 91%) at 1, 5 and 10 years, respectively. Analysis of all the Table 3 Thirty-day mortality and overall survival following surgery for PVE (95% confidence intervals in parenthesis) according to age: (P = 0.04 and 0.002, respectively as a categorical variable) Age (in years) >70 30-Day mortality (%) 10.4 ( ) 20.3 ( ) 18.3 ( ) 39.1 ( ) 1 year survival (%) 82.8 ( ) 71.0 ( ) 65.7 ( ) 50.1 ( ) 5 year survival (%) 76.6 ( ) 62.1 ( ) 48.2 ( ) 40.1 ( ) 10 year survival (%) 48.0 ( ) 40.5 ( ) 37.4 ( )

5 160 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) Table 5 Thirty-day mortality and overall survival (95% confidence intervals in parenthesis) according to type of prosthesis explanted and implanted for PVE (P = 0.3 and P = 0.03, respectively) Group 1 Group 2 Group 3 Group 4 Bioprosthesis/ Bioprosthesis/ Mechanical/ Mechanical/ bioprosthesis (35) mechanical (50) mechanical (219) bioprosthesis (18) 30-Day mortality 23.2 ( ) 28.0 ( ) 18.0 ( ) 11.1 ( ) 1 year survival 65.0 ( ) 57.6 ( ) 68.2 ( ) 83.3 ( ) 5 year survival 49.8 ( ) 42.3 ( ) 57.6 ( ) 68.4 ( ) 10 year survival 32.4 ( ) 18.1 ( ) 36.0 ( ) 68.4 ( ) Registry variables by univariate analysis identified removal of infected bioprosthesis and replacement by mechanical valve (Group 2) (P = 0.02; HR = 2.5) as a determinant of reoperation. This was confirmed by multivariate analysis; removal of infected bioprosthesis and re-replacement by mechanical valve (P 0.001; HR = 4.5). In addition, reoperation within 60 days of native valve replacement also determined reoperation (P = 0.02; HR = 3.4) (see Table 6). The freedom from death or reoperation for PVE, following PVE surgery were 61% (56 66%), 51% (45 56%) and 34% (25 44%) at 1, 5 and 10 years and is illustrated in Fig. 3. Univariate analysis on the variables identified age (P = 0.001; HR = 1.02; ), type of explanted prosthesis (P = 0.03; HR = 1.5) and the explanting of an infected bioprosthesis and re-replacement by a mechanical valve (P = 0.02; HR = 1.6) as determinants of death or reoperation. Multivariate analysis identified age (P = 0.003; HR = 1.02; ) and explanting of infected bioprosthesis and replacement by mechanical valve (P = 0.002; HR = 1.9) and also the period between operation (P = 0.04; HR = 1.00; ) as determinants of death or reoperation. 5. Conclusion Re-operation for PVE remains a major challenge for the cardiac surgeon. Not only can infection be difficult to eradicate, but redo operations in the context of systemically ill patients invariably result in poorer survival figures than for native valve surgery [1]. This challenge is being met by more aggressive debridement of infected tissue and by reconstruction and repair of the defects produced by excision, with biological tissues [6]. Improved myocardial management protocols have also contributed to this Table 6 improvement [2] by enabling the surgeon to undertake a longer, more complicated operation. Many also believe that the use of a biological prosthesis (usually homograft or autograft) in PVE is important in preventing further reinfection [5]. The results from centres promoting these practices continue to improve with many centres claiming substantial reductions in operative mortality and substantial increases in long-term survival [3,4,16]. These reported results reflect the considerable experience and the high level of technical expertise that reside at such centres. It is unlikely that either can be readily reproduced by others even when the protocols of these centres are followed in intimate detail. The practical value of these results to most cardiac surgeons with limited experience of PVE is thus, not clear and it is probably inappropriate for such surgeons to only use the results of these centres to judge their own performance. There is also a need, therefore, for results from PVE surgery that reflect the performance of a more heterogeneous group of cardiac surgeons, representative of the entire cardiac surgical community. The UKHVR in this context reflects the experience of a large number of surgeons, from a substantial number of different units and reports the outcomes of a much less homogenous group of surgeons than a single institution. This series is the largest reported in the literature to date for surgery for PVE and needs no defending for offering a benchmark for timerelated freedom from death and reoperation following sur- Incidence of reoperation according to interoperative interval (percentages of population in parenthesis) Interoperative interval (days) >365 No reoperation 77 (85.6) 118 (92.9) 97 (92.4) Reoperation 13 (14.4) 9 (7.1) 8 (7.6) Fig. 3. Freedom from death or reoperation following surgery for prosthetic endocarditis using Kaplan-Meier survival estimates.

6 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) gery for PVE, that may reasonably be adopted by most cardiac surgeons. The use of data from the UKHVR does have certain disadvantages. Firstly, the Registry began by only collecting information on mechanical and biological (stented and stentless) prosthetic replacements. It is only from very recent times that data have been collected on homograft and autograft valve replacements. In the context of PVE surgery, this is particularly relevant, because of the growing school of opinion that promotes the use of the homograft [5,18] or autograft [12] in this condition. Such opinion suggests, that the more conventional mechanical and biological prostheses are less resistant to re-infection following surgery. Evidence from elsewhere suggests, however, that the homograft may have a comparable risk to the standard prosthesis, with regard to developing late PVE [26]. What is certain is that the information regarding homograft and autograft surgery in PVE is derived from small series. The Registry clearly is not in the position to comment on this matter, but it can, nevertheless, perform the valuable role of reporting the results of use of standard prostheses in PVE surgery. Secondly, in order to achieve maximum co-operation and data collection, the UKHVR data set was, from the very outset, deliberately restricted and simplified as much as possible. All UKHVR data analyses do not, therefore, carry pre-operative variables such as, the type of microorganism involved, the extent of coronary artery disease, the degree of left ventricular dysfunction and the degree of operative priority. Operative variables such as bypass time and additional operative procedures and, post-operative variables such as length of stay are similarly not collected. The relative importance of these variables, to those identified by the Registry in this manuscript as determinants of time-related outcome for PVE surgery, however, is unknown, although in smaller series only a handful of such variables have been shown to be influential [16]. Recognition of the value of a registry as illustrated in this report may be influential in persuading cardiac surgeons to contribute to our understanding, by recruiting their patients to the registry of their choice. The incidence of surgery for PVE (0.5% each year) recorded by the Registry is in agreement with other series that report incidences of between 0.3 and 0.8% per patient year [7,9,13,26]. This stability of incidence is, however, not in agreement with others who have found a growing incidence of surgery for PVE [16], as evidence emerges that combined medical and surgical therapy results in better outcomes than medical therapy alone [4,27] and that surgery is specifically indicated in situations such as fungal PVE [20]. It is possible that the above stability of incidence noted in the Registry data, may reflect an increased use of the homograft or autograft in place of the conventional prosthesis undetected by the Registry. Whatever the reason, it is evident from the data that a substantial proportion of the heart valves used by UK cardiac surgeons for replacement in PVE, continue to be mechanical or biological prostheses and this paper reports the results of PVE surgery for these conventional types of valves. It is of note that none of the patients in the present study underwent re-replacement for tricuspid valve PVE. This is in keeping with other series, where tricuspid PVE was rare [6,8] and may reflect a generally more conservative surgical approach to tricuspid valve surgery, tricuspid endocarditis and tricuspid PVE. The Registry 30-day mortality for replacement for PVE of 19.9% compares favourably with other series that have reported equivalent time interval mortalities of between 20 and 65% [2,8,13]. Even lower mortality rates have, however, been reported by certain centres such as the Cleveland Clinic which reported an in-hospital mortality of 13% [16], but it should be emphasised that in-hospital does not equate to 30-day mortality. With regard to late survival, Lytle et al. [16] reported a 5 and 10 year survival of 82 and 60%, respectively for patients undergoing surgery for PVE, Farina et al. [8] reported a 12 year survival rate of 52% and David [6] reported a 5-year actuarial survival rate of 61%. The Registry survival rates are comparable but poorer; 55% (49 61%) and 38% (28 47%) at 5 and 10 years, respectively. It is difficult to undertake a meaningful comparison of survival outcomes for PVE surgery with those for other forms of valve surgery. The UKHVR 30-day mortality and 1, 5 and 10 year survival for native aortic valve replacement are 4.4 and 92%, 81 and 66%, respectively and for native mitral valve replacement are 6.5 and 87%, 76 and 57%, respectively [25]. The operative mortality for valve reoperation for all causes in three large series was 10.6% [1], 8.8% [10] and 11.0% [17]. Actuarial survival following valve reoperation was 73% at 5 years [10], 57% at 7 years [22] and 38% at 10 years [3]. Recently Jault et al. [11] have reported an operative mortality of 7.6% for valve replacement for native valve endocarditis and an actuarial survival of 71% at 9 years (operative mortality excluded). The findings of the present study do, therefore, confirm the higher operative mortality and lower longer-term survival following surgery for PVE when compared to other relevant forms of valve surgery as reported by others [1]. The present study identified age as the only independent determinant of 30-day mortality although others [8,16] have failed to do so. The year of operation approached statistical significance as a determinant of 30-day mortality (P = 0.6), but as in the Cleveland Clinic series [16] failed to quite achieve it. Farina et al. [8] found that female gender determined operative mortality in a series of 64 patients, but this was not confirmed in our larger series. The same group identified sepsis at the time of surgery and others [16] identified the presence of atrio-ventricular block, coronary artery disease and severe left ventricular dysfunction as other determinants of operative mortality, although Farina et al. [8] failed to confirm the finding with left ventricular dysfunction. None of these could be confirmed by the Registry for the reasons discussed elsewhere. With regards to longer-term survival following surgery

7 162 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) for PVE, age was again found to be an important determinant. Although Farina et al. [8] failed to find any effect of age on long-term survival, our findings confirmed those of other series of valve re-replacement, whether for infective reasons [1] or not [16], of the importance of age as a determinant of survival and may amongst many factors reflect the proposed increased sensitivity to PVE in the elderly population [26]. The Registry also found that explanting an infected bioprosthesis and re-replacing it with a mechanical valve determined overall survival. This is difficult to explain satisfactorily. One explanation may be that there is a relationship between those who had infected bioprostheses explanted and age, and that it is age that is the most important determinant rather than type of prosthesis. Certainly, if the description of the type variable was made more specific to include implant and explant type, then the predictive value of age was increased (P 0.001). Furthermore, although 60% of those with infected bioprostheses who had re-replacement with mechanical valves died, in a smaller group (n = 35) 57% of those with infected bioprostheses who had re-replacement with another bioprosthesis also died (P = 0.4). The difference between the latter group and those who had mechanical valves explanted and implanted (n = 219) did not achieve statistical significance as it did with the former, because of the large difference in population size between the two groups. The occurrence of early PVE has been reported as an important determinant of mortality and survival following PVE surgery [21]. Using either definition of early PVE (either before 60 days or before 1 year after operation on native valve) we, like others [8], were unable to confirm this. Kuyvenhoven et al. [13] have reported the importance of microbiological organisms on survival with PVE. In that series, however, only 61% of the patients underwent surgery for PVE and Grover et al. [9] found no difference between gram positive and gram negative cocci in survival following surgery for PVE [9]. Pre-operative functional class [2,22], state of operation [2], mitral valve surgery [10] and year of operation [10] have also been reported as determinants of survival in re-do valve surgery, but with re-replacement for PVE causative organism, functional class and degree of operative urgency did not determine long-term survival [8,16]. The limited data collection by the Registry prevents any comment from being made upon these findings with regards to re-do PVE surgery in the present study, as discussed before. None of the other surgical PVE series has, however, identified any other variable other than age as a predictor of overall survival [8,10,13], which is hardly surprising given the importance of age in determining overall survival in re-do valve surgery [1]. The Registry was unable to identify any difference in time-related outcome by type of prosthesis used at replacement. Farina et al. [8] identified the mechanical valve as a determinant of operative mortality, but our data did not identify any difference either in 30-day mortality or overall survival between biological and mechanical prostheses, as was also reported by the Cleveland Clinic [16]. Furthermore, although Farina et al. [8] reported an increased recurrence of PVE, but not of reoperation for PVE with mechanical prostheses, the view that mechanical valves are more likely than bioprosthetic valves to prevent the complete eradication of infection is not universally supported in the literature [15,26]. Sweeney et al. [24] have reported a lower incidence of reoperation with mechanical rather than bioprosthetic valve replacement, for valve endocarditis (native and prosthetic). Similarly, our data did not identify any difference between biological and mechanical valves used at re-replacement in freedom from reoperation, suggesting that complete excision of infected tissue may be more important than type of replacement prosthesis. Our data did, however, identify the removal of an infected bioprosthesis and replacement by a mechanical valve (patients in Group 2; see Table 5) as a determinant of reoperation both by univariate and multivariate analysis. It is difficult to satisfactorily explain this in the context that our data did not identify either the type of explanted prosthesis (P = 0.26) or the type of implanted prosthesis (P = 0.24) as determinants of reoperation. It is possible that the data may be supportive of the hypothesis that mechanical valves should be avoided in re-replacement for PVE. The finding that re-replacement of explanted infected mechanical valve by a mechanical prosthesis (patients in Group 3; Table 5) did not result in a higher incidence of reoperation, when compared with reimplantation with a biological valve (Group 4), however, is not consistent with this proposal. Our finding that early reoperation (within 60 days of previous valve replacement) is a determinant for re-reoperation for PVE in contrast, underlines the importance of meticulous removal of all infected tissue prior to re-reimplantation with a prosthesis and may reflect the outcome of a particular subset of organism that are more active in early PVE, that cannot be identified by the Registry database. This finding is contrary, nevertheless, to that of a recent review of PVE [26], which suggests that there is a constant hazard for PVE, although it was acknowledged that there was a barely discernable increase in the pre-60 day hazard for PVE. When freedom from death or reoperation were considered in combination, the type of explanted prosthesis (bioprosthesis) was identified as a determinant (P = 0.04) by multivariate analysis as it was for overall survival. The predictive ability of the other two determinants (age and interval between operation) were, however, improved (P = and P = 0.03, respectively) when the details of this variable were more specific and included not only type of prosthesis explanted (biological), but also type of prosthesis implanted (mechanical). As before this is difficult to explain satisfactorily, but may in part be a reflection of the power of age as a determinant of the survival or reoperation outcomes as there was a correlation between insertion of bioprosthesis and age in our population. Our

8 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) Registry data do not demonstrate a survival difference between bioprosthetic and mechanical valves implanted at surgery for PVE and show a persistence of mechanical valves amongst standard prostheses, as the prosthesis of choice for PVE. The data presented here may be comforting to the cardiac surgeon, as there is no evidence that the choice of prosthesis influences freedom from death or reoperation following surgery for PVE and because many patients clearly can be cured of PVE with standard prosthesis re-replacement. It must be noted, however, that those patients who had infected bioprostheses explanted and mechanical prostheses implanted were identified in our series, as a subgroup with poor overall survival and a higher risk of reoperation. This remains difficult to explain satisfactorily. It is not appropriate, for this study to comment on the relative efficacies of standard prostheses in relation to homografts or autografts in PVE surgery; a carefully constructed prospective trial is the only satisfactory means of addressing this issue. In summary, data from 322 patients undergoing single valve re-replacement for single valve PVE reported to the UKHVR confirmed the high 30-day mortality and reduced longer-term survival. Survival and cure of infection is possible in many patients, however, with standard mechanical and bioprosthetic valves, and no evidence was found to substantiate the hypothesis that freedom from death and reoperation is poorer in patients receiving mechanical when compared to patients receiving bioprosthetic valves for PVE. Early reoperation (within 60 days) for PVE does influence the incidence of reoperation. Patient age stands out as a principal determinant of short- and medium-term outcome. It remains to be seen if greater use of the homograft and autograft can extend the recent improvements in the surgical management of PVE. Acknowledgements The authors wish to thank the centres who have contributed their data to this study and, who continue to support the UK Heart Valve Registry. UK Heart Valve Registry participating hospitals: Aberdeen Royal Infirmary, United Bristol Health Trust, Broadgreen Cardiothoracic Centre, Brompton Hospital, Castle Hill Hospital, Edinburgh Royal Infirmary, Freeman Group of Hospitals, Glasgow Royal Infirmary, Glenfield General Hospital, Guy s Hospital, Hammersmith Hospital, Harefield Hospital, Killingbeck Hospital, Kings College Hospital, Leeds General Infirmary, London Chest, Manchester Royal Infirmary, Middlesex Hospital, Northern General Hospital, North Staffordshire Royal Infirmary, Nottingham City Hospital, Queen Elizabeth Hospital, Royal Victoria Hospital, St. Bartholomew s Hospital, St. George s Hospital, St. Mary s Hospital, St. Thomas Hospital, Southampton General Hospital, South Cleveland Hospital, Victoria Hospital, Walsgrave Hospital, Western Infirmary, Wythenshawe Hospital. References [1] Antunes MJ. Reoperations on cardiac valves. J Heart Valve Dis 1992;1: [2] Bortolotti U, Milano A, Mossuto E, Mazzaro E, Thiene G, Casarotto D. Early and late outcome after reoperation for prosthetic valve dysfunction: analysis of 549 patients during a 26-year period. J Heart Valve Dis 1994;3: [3] Bosch X, Pomar JL, Pelletier LC. Early and late prognosis after reoperation for prosthetic valve replacement. J Thorac Cardiovasc Surg 1984;88: [4] Calderwood SB, Swinski LA, Karchmer AW, Waternaux CM, Buckley MJ. Prosthetic valve endocarditis. J Thorac Cardiovasc Surg 1986;92: [5] Camacho MT, Cosgrove DM. Homografts in the treatment of prosthetic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995; 7: [6] David TE. The Surgical treatment of patients with prosthetic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995;1: [7] De Gevigney G, Pop C, Delahaye JP. The risk of infective endocarditis after cardiac surgical and interventional procedures. Eur Heart J 1995;16:7 14. [8] Farina G, Vitale N, Piazza L, De Vivo F, de Luca L, Cotrufo M. Long term results of surgery for prosthetic valve endocarditis. J Heart Valve Dis 1994;3: [9] Grover FL, Cohen DJ, Oprian C, Henderson WG, Sethi G, Hammermeister KE. Determinants of the occurrence of and survival from prosthetic valve endocarditis. J Thorac Cardiovasc Surg 1994;108: [10] Husebye DG, Pluth JR, Piehler JM, Schaff HV, Orszulak TA, Puga FJ, Danielson GK. Reoperation on prosthetic heart valves. J Thorac Cardiovasc Surg 1983;86: [11] Jault F, Gandjbakhch I, Rama A, Nectoux M, Bors V, Vaissier E, Nataf P, Pavie A, Cabrol C. Active native valve endocarditis: determinants of operative death and late mortality. Ann Thorac Surg 1997;63: [12] Joyce F, Tingleff J, Pettersson G. The Ross Operation in the treatment of prosthetic aortic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995;7: [13] Kuyvenhoven JP, van Rijk-Zwikker GL, Hermans J, Thompson J, Huysmans HA. Prosthetic valve endocarditis: analysis of risk factors for mortality. Eur J Cardio-thorac Surg 1994;8: [14] Lytle BW. Surgical treatment of prosthetic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995;7: [15] Lytle BW. Introduction: prosthetic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995;7:1. [16] Lytle BW, Priest BP, Taylor PC, Loop FD, Sapp SK, Stewart RW, McCarthy PM, Muehrcke D, Cosgrove DM. Surgical treatment of prosthetic valve endocarditis. J Thorac Cardiovasc Surg 1996;111: [17] Lytle BW, Cosgrove DM, Taylor PC, Gill CC, Goormastic M, Golding LR, Stewart RW, Loop FD. Reoperations for valve surgery: perioperative mortality and determinants of risk for 1000 patients, [18] McGiffin DC, Galbraith AJ, McLachlan GJ. Aortic valve infection: risk factors for death and recurrent endocarditis after aortic valve replacement. J Thorac Cardiovasc Surg 1992;104: [19] McGiffin DC, Kirklin JK. The impact of aortic valve homografts on the treatment of aortic prosthetic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995;7: [20] Muehrcke DD. Fungal prosthetic valve endocarditis. Seminars Thorac Cardiovasc Surg 1995;7: [21] Otaki M. Prosthetic valve endocarditis. An analysis of the outcome in 32 cases. ASAIO J 1994;40: [22] Pansini S, Ottino G, Forsennati PG, Serpieri G, Zattera G, Casabona R, Di Summa M, Villani M, Poletti A, Morea M. Reoperations on

9 164 M.-B. Edwards et al. / European Journal of Cardio-thoracic Surgery 14 (1998) heart valve prostheses: an analysis of operative risks and late results. Ann Thorac Surg 1990;50: [23] Piehler JM, Blackstone EM, Bailey KR, Sullivan ME, Pluth JR, Weiss NG, Brookmeyer RS, Chandler JG. Reoperations on prosthetic heart valves: patient-specific estimates of in-hospital events. J Thorac Cardiovasc Surg 1995;109: [24] Sweeney MS, Ruel GJ, Denton MD, Cooley A, Ott DA, Duncan JM, Frazier OH, Livesay JJ. Comparison of bioprosthetic and mechanical valve replacement for active endocarditis. J Thorac Cardiovasc Surg 1985;90: [25] UK Heart Valve Registry Report, Crown Copyright, [26] Vlessis AA, Khaki A, Grunkenmeuer GL, Li H-H, Starr A. Risk and diagnosis and management of prosthetic valve endocarditis: a review. J Heart Valve Dis 1997;6: [27] Yu VL, Fang GD, Keys TF, Harris AA, Gentry LO, Fuchs PC, Wagener MM, Wong ES. Prosthetic valve endocarditis: superiority of surgical valve replacement versus medical therapy only. Ann Thorac Surg 1994;58: Appendix A. Conference discussion Dr J. Pomar (Barcelona, Spain): What was the percentage of bioprostheses and mechanical valves in endocarditis? You said about 26% were bioprostheses and 74% were mechanical, but among the whole set in the resists of the bioprostheses, how many of them got endocarditis and how many of them were mechanical? Do you have any idea? Miss Edwards (London, UK): The Fig. 1 gave in the presentation of 0.5% was the overall incidence of endocarditis. However, when this is disaggregated against valve type, the incidence of endocarditis for all mechanical valves registered within the database is 0.6% compared to 0.5% of all bioprosthetic valves. Dr A. Bodnar (London, UK): You said that age was the only significant factor. Tell me, please, how does the mean age of the endocarditis group compare to the mean age of your entire database, entire patient population? Miss Edwards: The mean age of the database population is 61.6 years. In comparison, the mean age for patients suffering from prosthetic endocarditis is 54.9 years. Dr C. Knott-Craig (Oklahoma City, OK, USA): I enjoyed the presentation very much. I believe it is probably one of the largest series of patients that have undergone surgery for prosthetic endocarditis, and I applaud you for presenting the data here. Was there a difference in mortality and reoperation rate between the aortic valves and the mitral valves in patients who underwent surgery for endocarditis? Miss Edwards: We did look at this variable and found there was no difference in mortality or reoperation rate in relation to site of valve surgery. Dr G. Rizzoli (Padova, Italy): The single most reknowned risk factor for prosthetic endocarditis is endocarditis of the natural valve. Did you look at the etiology of the natural valve pathology? And also I wanted to know if there are two types of prosthetic valve endocarditis, one that takes place very early, related to the natural valve infection, and one that takes place late and probably unrelated. Did you look at the time occurrence of the prosthetic endocarditis? Miss Edwards: In the first instance, until recently we did not collect information on native valve endocarditis. It is only in the last year that we have collected information on the presence of native valve endocarditis. With regard to your second question using two definitions for early and late prosthetic endocarditis of less than 60 days and less than 365 days, reoperation within 60 days was a significant determinant of freedom from subsequent reoperation for prosthetic endocarditis. However, our data do not find early endocarditis to be a determinant when freedom from death or reoperation are considered together. Dr Y. Goffin (Brussels, Belgium): You did not mention homografts. Why not? Miss Edwards: No. We have not collected data on homografts or autografts for the first 10 years. Again, it is only in the last year that we have actually started to collect this data. Dr Goffin: The second question. Did you specify or did you make the difference between cases with abscess in the annulus and the septum and without abscess? Miss Edwards: No. The Registry has no data on this issue. Dr Mark F. O Brien (Chermside, Australia): Regarding Staphylococcal endocarditis, were you able to separate the influence of such an organism compared to endocarditis due to other organisms? Miss Edwards: The registry collects very, limited preoperative data in terms of etiology. Thus we do not have data on the microbiology of endocarditis. However, in due course we hope we will be possible to extend the database to include this type of information. Dr O Brien: Already it has been shown that the presence of staphlococcal endocarditis has a distinct important influence on operative, postoperative and late events.