Repeated Thromboembolic and Bleeding Events After Mechanical Aortic Valve Replacement

Repeated Thromboembolic and Bleeding Events After Mechanical Aortic Valve Replacement Filip P. Casselman, MD, Michiel L. Bots, MD, PhD, Willem Van Lommel, MD, Paul J. Knaepen, MD, Ruud Lensen, MD, PhD and Freddy E. E. Vermeulen, MD Departments of Cardiothoracic Surgery and Stichting Hartenzorg, St Antonius Ziekenhuis, Nieuwegein, Julius Center for General Practice and Patient Oriented Research, University Medical Center, Utrecht, and Department of Cardiology, Rijnstate Ziekenhuis, Arnhem, The Netherlands Background. The choice of a valve substitute in young adults requires a decision balancing the risks of longterm anticoagulation versus reoperation(s). This article analyzes the long-term risk and determinants of thromboembolic (TE) and bleeding (BLE) complications after mechanical aortic valve replacement (AVR). Methods. From December 1963 to January 1974, 249 patients survived a mechanical AVR at our institution. Mean age was 41.8 12.4 years and 81% (n 202) were male. Ball valves were implanted in 24% (n 61) and disc valves in 76% (n 188). Patients were anticoagulated with vitamin K antagonists and dipyridamole. A total of 4,855 patient years was available for analysis. Mean follow-up was 19.5 9.4 years and was 100% complete. Analyses were performed with Kaplan-Meier and multivariable Cox regression methods. Results. One hundred and two patients had one TE or BLE postoperative event and 58 patients had two postoperative events. Six patients had more than five postoperative events. Freedom from a first postoperative event was 74.8% 2.9%, 55.3% 3.5%, and 46.8% 4.0% at 10, 20, and 30 years, respectively. Freedom from a second postoperative event was 45.4% 5.4%, 29% 6.0%, and 23.2% 7.1% at 10, 20, and 30 years, respectively. Multivariate predictors for TE or BLE complications were ball valve (Odds Ratio (OR) 2.9), postoperative endocarditis (OR 2.2), and any surgery (OR 2.2). The incidence of events was highest the first 5 postoperative years. Conclusions. The risk of adverse events is highest the first 5 postoperative years. Once an event has occurred, the risk for a second event is increased. The incidence and frequency of events is substantial and should be considered in the choice of a valve substitute. (Ann Thorac Surg 2001;71:1172 80) 2001 by The Society of Thoracic Surgeons Because a considerable proportion of patients in need of an aortic valve replacement still have a 25- to 50-year life expectancy, it is important to consider the lifelong risk of thromboembolic and bleeding complications in case of a mechanical valve substitute. Some articles in literature deal with long-term follow-up after mechanical valve replacement [1 4]. However, the follow-up seldom extends beyond 25 years and is hardly ever complete, and most studies censor the patient at the first event for a given complication, thereby neglecting subsequent events. Therefore, it is extremely difficult to estimate the real incidence of thromboembolic and bleeding complications. This study focuses on the very long-term follow-up after mechanical aortic valve replacement with specific emphasis on the occurrence and frequency of thromboembolic (TE) and bleeding (BLE) complications and its determinants. Accepted for publication Nov 13, 2000. Address reprint requests to Dr Casselman, Department of Cardio- Thoracic Surgery, St Antonius Ziekenhuis, Koekoekslaan 1, 3435 CM Nieuwegein, The Netherlands; e-mail:casself@hotmail.com. Material and Methods Patient Selection December 1963 was the start of the aortic valve replacement program at our institution and we studied the first 10 years of this program (December 1963 through January 1, 1974). A total of 312 patients underwent aortic valve replacement during this time. Reoperations, urgent, or combined procedures were included. Operative mortality before 1971 was 33% (44 of 132 patients) and from 1971 to 1974 it was 10.5% (19 of 180 patients). Causes of hospital death are mentioned in Table 1. Since none of these were thromboembolic- or anticoagulation-related bleeding events, and also because of the high operative mortality in the early experience, we only included hospital survivors in the study. (At that time, hospital survivors included 30-day survivors). After these criteria were met, 249 patients were the subject of further analysis in this article. Demographics The mean age of the 249 hospital survivors was 41.8 12.4 years (range 14 to 68 years) and 81% (n 202) were male. Most of the patients had an isolated aortic valve replacement (n 242). The remainder (n 7) had simultaneous coronary artery bypass grafting. A minority 2001 by The Society of Thoracic Surgeons 0003-4975/01/$20.00 Published by Elsevier Science Inc PII S0003-4975(00)02683-7

Ann Thorac Surg CASSELMAN ET AL 2001;71:1172 80 REPEATED TE AND BLEEDING EVENTS AFTER AVR 1173 Table 1. Causes of In-Hospital Death (n 63) Cause Number of Patients % Heart failure 27 43 Intraoperative technical problems 17 27 Intraoperative stroke 6 10 Sepsis 5 8 Endocarditis 2 3 Respiratory failure 2 3 Multiple organ failure 2 3 Kidney failure 2 3 Total 63 100 of patients had aortic valve replacement as a reoperation after a previous congenital aortic valve commissurotomy (n 13), and 3 patients had a reoperation after a correction of aortic coarctation. The cause of the valve disease in the remaining patients was predominantly rheumatic. Preoperative acute endocarditis was present in 28 patients (11%). Isolated aortic stenosis was present in 12.4% of patients (n 31), whereas isolated aortic regurgitation and mixed diseases were present in 42.2% of patients (n 105) and 45.4% of patients (n 113), respectively. The procedure was elective in 94% of the patients (n 233) and urgent in the remaining patients (n 16). The replaced aortic valve was tricuspid in 202 patients (81%), bicuspid in 38 patients (15%), and unknown in 9 patients (4%). Several types of mechanical valves were used during the study. Type and number of specific valve implantations are given in Table 2. Postoperatively, patients were anticoagulated with vitamin K antagonists in association with an antiplatelet agent (mostly dipyridamole). The level of anticoagulation was followed with the thrombotest and the target (during those years) was between 10% and 6%, which corresponds with an International Normalized Ratio (INR) between 2.8 and 4.2 [5]. Patients were regularly followed by the Dutch thrombosis service, a national organization with multiple regional offices, Table 2. Types, Era of Implantation and Number of a Specific Valve Implanted (n 312) Type of Mechanical Valve Number of Patients Hospital Survivors Starr-Edwards 1000 30 10 (Dec 1963 Aug 1966) Starr-Edwards 1200/1260 37 23 (Sept 1966 June 1968) Starr-Edwards 2300 20 17 (July 1968 Oct 1969) Starr-Edwards 2310/2320 11 11 (June 1969 March 1970) Björk-Shiley AB 119 100 (Dec 1969 June 1973) Björk-Shiley ABP 95 88 (June 1971 Jan 1974) Total 312 249 specifically created to follow up on anticoagulated patients to coordinate an adequate anticoagulation level. Follow-up The first follow-up study of the patient cohort was retrospectively conducted in April and May 1975 by one author (W.V.L.). All data were collected but never published. A repeat follow-up study was carried out by another author (FPC) between May and December 1999. A new individual patient file was created according to the official guidelines [6]. All definitions of events were made according to these guidelines with the exception of hemolysis, which was defined as any raise in lactate dehydrogenase not attributable to other causes (because of the definition used in 1975). Follow-up information was obtained from the Dutch thrombosis service and any doctor(s) possibly contacted by the patient. All patient files were personally seen by an author (F.P.C.) and scrutinized for events and causes of death. The anticoagulation treatment at the time of an event, and the INR if known, were recorded. All surviving patients received and completed a questionnaire regarding their present status, past events, and the frequency of minor bleeding events, which were roughly graded as occurring weekly or more frequently, monthly, yearly, or only seldom. In addition, they were asked whether it bothered them to take anticoagulation medication. A total of 4,855 patient years was available for analysis. Mean follow-up was 19.5 9.4 years, and it was 100% complete. The New York Heart Association functional class in 91 surviving patients at follow-up was I in 39% of patients (n 35), II in 46% of patients (n 42), and III in 15% of patients (n 14). Data Analysis Data are expressed as the mean the standard deviation. Survival and event-free estimates were determined by life-table analysis [7] and expressed as proportion the standard error. Analysis was performed with the SPSS package version 8.0 (SPSS Inc, Chicago, IL). Risk factors for outcome were evaluated using Cox proportional hazard models. The first event was used as outcome. Associations are presented as hazard ratios (ie, relative risk) with corresponding 95% confidence intervals (CIs). First analyses were performed using only the risk factor of interest in the Cox univariate model. Those risk factors with associations that showed a statistical significance of less than or equal to 0.10 were included in a multivariate Cox regression model. A priori, we evaluated the following risk factors of information that were collected at base line: age, sex, year of operation (before 1970 or after 1970), hypertension (systolic pressure 160 or diastolic pressure 95 or treatment), atrial fibrillation, history of diabetes mellitus, aortic stenosis (peak gradient 75), aortic regurgitation (definition grade 1), preoperative endocarditis, type of valve (Starr-Edwards versus Björk-Shiley), and type of operation (elective or emergency). Postoperative factors that were evaluated

1174 CASSELMAN ET AL Ann Thorac Surg REPEATED TE AND BLEEDING EVENTS AFTER AVR 2001;71:1172 80 Fig 1. Overall late survival among hospital survivors (n 249). were paravalvular leaks, postoperative endocarditis, new onsets of atrial fibrillation, cardiac events and any operation (other than aortic valve reoperation). In addition, we evaluated whether base line characteristics were predictive of a recurrent anticoagulation-related complication. Estimates of the linearized incidence rate with corresponding standard errors were obtained by dividing the number of first events by the corresponding patient years of follow-up. The standard error was calculated as the square root of the incidence divided by the patient years of follow-up, assuming a Poisson distribution. A two-sided p-value less than 0.05 was considered statistically significant. Survival curves for the second and third event were estimated using the same method as for the first event. The population at risk for a second event was restricted to those who suffered a first event, irrespective of whether the first event was fatal or nonfatal. The same applies for the analyses of the third event: the population comprised the subjects who suffered a first and second event. Results Late Mortality Overall actuarial survival among hospital survivors was 80.3% 2.6%, 57.4% 3.1%, and 33.6% 4.2% at Table 3. Causes of Late Death (n 158) Cause of Death Number of Patients % Cardiac 95 60.1 Malignancy 20 12.6 COPD 5 3.2 Trauma 4 2.5 Other 27 17 Unknown 7 4.4 Total 158 100 Fig 2. Freedom from cardiac death and thromboembolic- or bleedingrelated death. (CD cardiac death; TE/BLE thromboembolic/bleeding.) postoperative years 10, 20, and 30, respectively (Fig 1). The linearized incidence rate was 3.2% 0.3% per year. Causes of death are shown in Table 3. Multivariate independent risk factors for death were age [hazard ratio increase of 1.0 per year (95% CI 1.0 to 1.1)], male gender [hazard ratio 1.7 (95% CI 1.1 to 2.7)], operation before 1970 [hazard ratio 1.6 (95% CI 1.1 to 2.4)], and postoperative endocarditis [hazard ratio 2.2 (95% CI 1.3 to 3.8)]. Diabetes and emergency operations were significantly related to mortality in the univariate model but not in the multivariate model. Freedom from cardiac death was 87.9% 2.6%, 72.9% 2.9%, and 51.8% 4.2%, at postoperative years 10, 20, and 30, respectively (Fig 2). The linearized incidence rate was 1.9% 0.2% per year. Causes of cardiac death are shown in Table 4. Multivariate independent risk factors for cardiac death were age [hazard ratio increase of 1.0 per year (95% CI 1.0 to 1.1)], and operation before 1970 [hazard ratio 1.9 (95% CI 1.2 to 3.1)]. Male sex, type of valve, and postoperative endocarditis were significantly related to cardiac mortality in the univariate model but not in the multivariate model. Freedom from valve-related mortality was 92% 1.7%, Table 4. Causes of Cardiac Death (n 95) Cause Number of Patients % Heart failure 43 45.3 Myocardial infarction 7 7.4 Arrhythmia 0 0 Valve related Sudden death 26 27.4 Valve thrombosis 4 4.2 Thromboembolic event 2 2.1 Bleeding event 8 8.4 Endocarditis 5 5.3 Total 95 100

Ann Thorac Surg CASSELMAN ET AL 2001;71:1172 80 REPEATED TE AND BLEEDING EVENTS AFTER AVR 1175 Fig 3. Freedom from a first, second, and third thromboembolic- or bleeding-related event. Note the increased slope of the multipleevents curves, which indicates an increased risk of subsequent events after the first event was encountered. (N number; TE/BLE thromboembolic/bleeding.) 83.3% 2.5%, and 75.4% 3.7% at postoperative years 10, 20, and 30, respectively. The linearized incidence rate was 0.9% 0.1% per year. Causes of valve-related mortality are also shown in Table 4. Multivariate independent risk factors for valve-related mortality were age [hazard ratio increase of 1.0 per year (95% CI 1.0 to 1.1)], and postoperative endocarditis [hazard ratio 2.1 (95% CI 1.1 to 4.2)]. Preoperative endocarditis, emergency operations, and aortic stenosis were significantly associated with valve-related mortality in the univariate model but not in the multivariate model. Freedom from TE or BLE mortality was 97.7% 1.3%, 93.2% 1.8%, and 90.5% 3.2% at postoperative years 10, 20, and 30, respectively (Fig 2). The linearized incidence rate was 0.3% 0.1% per year. Out of 14 total events, 8 were BLE events, whereas the remainder were TE phenomena (including 4 patients with valve thrombosis; see Table 4). Age was the only independent predictor for TE or BLE mortality with a hazard ratio increase of 1.1 per year (95% CI 1.0 to 1.2). Diabetes and emergency operations did not reach statistical significance in the multivariate model. Thromboembolic and Bleeding Complications Overall TE (minor and major) and BLE (major) complications are termed events and refer to any of these complications in this paragraph. One hundred and two patients experienced an event, in the absence of endocarditis, during follow-up. The majority of patients experienced one or two events, (n 102) or (n 58), respectively. Thirty patients experienced three events and 13 patients experienced four events. Six patients had more than five events and 1 patient had more than 10 events. Freedom from a first event was 74.8% 2.4%, 56.3% 3.5%, and 46.8% 4.1% at postoperative years 10, 20, and 30, respectively (Fig 3). Multivariate predictors for a first event were ball valve (hazard ratio 2.9 [95% CI 1.2 to 7.2]), postoperative endocarditis (hazard ratio 2.2 [95% CI 1.2 to 4.0]), and any surgery other than aortic valve reoperation (hazard ratio 2.2 [95% CI 1.3 to 3.7]). Atrial fibrillation and operation before 1970 did not reach statistical significance in the multivariate model. The linearized incidence rate of a first event was 3.0% 0.3% per patient year. Freedom from a second event after the first event was 45.4% 5.4%, 29% 6.0%, and 23.2% 7.1% at postoperative years 10, 20, and 30, respectively (Fig 3). None of the risk factors reached a statistical significant level in association with recurrent events. The risk of an event was highest within the first 5 years after aortic valve replacement. After 5 years the risk decreased (Table 5). VALVE THROMBOSIS. Valve thrombosis, in the absence of endocarditis, occurred six times. None of these patients were adequately anticoagulated at the time of the event. In two instances, the patients did not even take vitamin K antagonists, which they had stopped spontaneously. The linearized incidence rate of valve thrombosis was 0.1% 0.1% per patient year. Table 5 shows the occurrence of valve thrombosis during follow-up. Freedom from valve thrombosis was 98.7% 0.7%, 97.0% 1.0%, and 97.0% 1.1% at postoperative years 10, 20, and 30, respectively. THROMBOEMBOLIC PHENOMENA. A total of 140 TE phenomena (excluding valve thrombosis) took place in 77 patients who did not have endocarditis at the time of the TE event. Table 5. Incidence of Valve-Related Morbidity During Follow-up Event Follow-up Period (Years Postoperatively) 0 4.9 5 9.9 10 14.9 15 19.9 20 24.9 25 29.9 ARC (TE or BLE) 4.5 0.6 1.8 0.5 3.0 0.7 2.9 0.7 1.8 0.7 2.3 1.1 Thromb 0.3 0.1 0 0.3 0.2 0 0 0 TE 2.7 0.5 1.7 0.4 1.5 0.4 1.7 0.5 1.5 0.6 2.0 1.0 BLE 1.3 0.3 0.5 0.2 1.1 0.4 1.9 0.5 0.7 0.4 0.4 0.4 Endocarditis 0.5 0.2 0.5 0.2 0.2 0.2 0.4 0.2 0.2 0.2 0.7 0.5 AV Reoperation 1.6 0.4 0.8 0.3 0.8 0.3 0.6 0.3 0.9 0.4 1.5 0.8 a Values are 5 years of cumulative incidences (with corresponding standard errors) of the event. ARC global anticoagulation-related complication; AV aortic valve; BLE bleeding event; TE thromboembolic event; Thromb valve thrombosis.

1176 CASSELMAN ET AL Ann Thorac Surg REPEATED TE AND BLEEDING EVENTS AFTER AVR 2001;71:1172 80 Twenty-eight patients experienced a second TE event, 17 patients experienced a third TE event, and 4 patients experienced a fourth TE event. Transient ischemic attack occurred in the majority of patients: 82 events in 47 patients. No reversible ischemic neurologic deficits were noted. Stroke occurred 35 times in 28 patients. All of them resulted in some degree of permanent deficit. A minority of the TE events were peripheral emboli (seven events in 6 patients). Sixteen events were classified as other, including nine probable embolic events (according to history) and seven nonspecified. Lethal outcome was noted in 2.6% (two TE events). The INR at the time of the first event was known in 23 patients (30% of the events). Mean INR at the time of the event was 1.9 1.2 and 85% of these values were below the target base line of 2.8. Sixty-seven percent of the values were even below 2.0. Of all 140 TE events, 12.8% of the patients (n 18) were not using anticoagulant drugs at the time of the event. Of 77 patients with a first TE event, 27 patients were taking vitamin K antagonists solely and 42 patients took antiplatelet drugs in addition to vitamin K antagonists. Five patients took only antiplatelet drugs. Freedom from a first TE event was 79.9% 2.6%, 68.5% 3.3%, and 57.3% 4.3% at postoperative years 10, 20, and 30, respectively (Fig 4). Multivariate independent risk factors for first thromboembolic event, excluding valve thrombosis, were age [hazard ratio increase of 1.0 per year (95% CI 1.0 to 1.0)], operation year before 1970 [hazard ratio of 2.2 (95% CI 1.3 to 3.7)], and not using anticoagulant drugs at the time of the event [hazards ratio 4.1 (95% CI 2.1 to 8.0)]. A trial fibrillation was significantly related to first thromboembolic event in the univariate model, but not in the multivariate model. The linearized incidence rate of a first TE event was 2.0% 0.2% per patient per year. Thereafter, freedom from a second TE event was 73% 5.3%, 63.6% 6%, and 49.9% 7.8% at postoperative years 5, 10, and 15, respectively (Fig 4). In the analysis of the determinants of Fig 4. Freedom from a first, second, and third thromboembolic event. Note the increased slope of the multiple-events curves, which indicates an increased risk of subsequent events after the first event was encountered. (N number; TE thromboembolic.) Fig 5. Freedom from a major thromboembolic event. (N number; TE thromboembolic.) a recurrent TE event, none of the risk factors reached a statistically significant level. The incidence of TE events was higher within the first 5 postoperative years. After that it remained relatively constant (Table 5). Freedom from a major TE event (35 strokes in 28 patients) was 95% 1%, 87% 3%, and 85% 3% at postoperative years 10, 20, and 25, respectively (Fig 5). MAJOR BLEEDING EVENTS. A total of 72 major BLE events occurred in the absence of endocarditis. There were 47 patients who had a first major BLE event, whereas 15 patients had a second major BLE event, and 6 patients had a third major BLE event. A minority of the major BLE events was cerebral bleeding (n 22), with a variable degree of residual impairment (n 14) or lethal outcome (n 8); the remainder of BLE events was noncerebral, requiring transfusion or surgical evacuation (50 events). All the patients with major BLE events were taking oral anticoagulation at the time of the event. The INR at the time of the first major BLE event (n 47) was known in 13 cases. Mean INR was 4.7 1.8 and 69% of these values were greater than the target maximum INR of 4.2. Of 47 patients with a first BLE event, 21 patients were taking vitamin K antagonists solely and 26 patients took antiplatelet drugs in addition. Freedom from a first major BLE event was 91.1% 1.9%, 78.6% 2.4%, and 74.1% 3.4% at postoperative years 10, 20, and 30, respectively (Fig 6). Multivariate independent risk factors for a first major BLE event were increasing age (hazard ratio increase of 1.0 per year [95% CI 1.0 to 1.1]) and any surgery (hazard ratio 3.8 [95% CI 2.0 to 7.7]). Atrial fibrillation was significantly related to first major BLE event in the univariate model, but not in the multivariate model. The linearized incidence rate of a first major BLE event was 1.1% 0.2% per patient year. Freedom from a second major BLE event after the first event was 78.7% 4.2%, 69% 7.6%, and 58.3% 9.5% at postoperative years 5, 10, and 15, respectively (Fig 6). In the analysis

Ann Thorac Surg CASSELMAN ET AL 2001;71:1172 80 REPEATED TE AND BLEEDING EVENTS AFTER AVR 1177 patient year. The incidence of postoperative endocarditis was relatively constant over time (Table 5). Fig 6. Freedom from a major first, second, and third bleeding event. Note the increased slope of the curves with multiple events, indicating an increased risk of subsequent events after the first event was encountered. (N number; BLE bleeding.) of the determinants of a recurrent major BLE event, none of the risk factors reached a statistically significant level. The incidence of major BLE events was relatively constant over time (Table 5). MINOR BLEEDING EVENTS. The frequency of minor bleeding events was estimated by the 91 survivors and graded as very seldom (n 59), yearly (n 17), monthly (n 5), and weekly or more (n 10). ATTITUDE TOWARDS ANTICOAGULATION. Among 91 survivors, 13 patients would prefer not to take anticoagulation medication, whereas 75 patients did not care. The remainder had no opinion. Other Valve-Related Events VALVE DYSFUNCTION. There were no structural valve deteriorations. Leaflet obstruction due to pannus overgrowth necessitating reoperation occurred in 2 patients. PARAVALVULAR LEAK. Paravalvular leak occurred once in 28 patients, twice in 3 patients, and three times in 1 patient. The linearized incidence rate was 0.7% 0.1% per patient year. Paravalvular leak was the major cause of aortic valve reoperation (Table 6). HEMOLYSIS. Hemolysis occurred frequently with a linearized incidence rate of 4.2% 0.4% per patient year (n 118). The majority of the hemolysis events occurred within the first 5 postoperative years (n 93). However, hemolysis was only once the leading cause of reoperation (Table 6). ENDOCARDITIS. Postoperative endocarditis occurred in 20 hospital survivors. The linearized incidence rate of endocarditis in hospital survivors was 0.4% 0.1% per AORTIC VALVE REOPERATION. A total of 60 aortic valve reoperations occurred in 46 patients. Twelve patients underwent a third aortic valve reoperation and 2 patients a fourth reoperation. Causes of aortic valve reoperation are listed in Table 6. Paravalvular leak was the leading cause occurring in 60.6% of cases (n 37). Freedom of aortic valve reoperation was 88.7% 2.1%, 82.4% 2.7%, and 67.5% 6.2% at postoperative years 10, 20, and 30, respectively. The linearized incidence rate was 1% 0.1% per patient year. The incidence of aortic valve reoperation was highest within the first 5 postoperative years and beyond 25 years of follow-up (Table 5). None of the patients died at aortic valve reoperation. Other Events OTHER REOPERATIONS. A total of 25 cardiac reoperations, other than aortic valve reoperations occurred in 20 patients. They included coronary artery bypass grafting (n 9), mitral valve operation (n 9), ascending aortic replacement (n 2) and other reoperations (n 5). A total of 27 pacemakers were implanted during follow-up. A total of 199 other, noncardiac, surgical interventions took place in 100 patients. Freedom from any first surgical intervention (excluding aortic valve reoperation) was 71.2% 3.1%, 48.3% 3.7%, and 28.8% 4.5% at postoperative years 10, 20, and 30, respectively. OTHER CARDIAC EVENTS (EXCLUDING CARDIAC MORTALITY). A total of 395 cardiac events in 168 patients were noted during follow-up. They included heart failure (87 events), myocardial infarction (24 events), angina (35 events), supraventricular arrhythmias (93 events), ventricular arrhythmias (41 events), electrical cardioversions (31 events), hypertension treatment (43 events) and other (41 events). Freedom from any first nonoperative cardiac-related event was 56.7% 3.3%, 33.1% 3.3%, and 12.8% 4.8% at postoperative years 10, 20, and 30, respectively. Table 6. Causes of Aortic Valve Reoperation (60 Events in 46 Patients) Cause First Operation (No. of Events) Second or Third Reoperation (No. of Events) Paravalvular leak 29 8 Valve thrombosis or 5 1 recurrent thromboemboli Valvular dysfunction 2 0 Endocarditis 1 1 Hemolysis 1 0 Other 8 4 Total 46 14

1178 CASSELMAN ET AL Ann Thorac Surg REPEATED TE AND BLEEDING EVENTS AFTER AVR 2001;71:1172 80 Comment Overall TE and Major BLE Complications (Global Incidence) In agreement with previous publications, the global incidence of TE or BLE complications was fairly high [4, 8 12]. In this series, only 46.8% of the patients remained free from a first TE or BLE event at 30 years postoperatively. However, the linearized incidence rate for a first event in this series was 3% per patient year, which compares favorably with reported incidence rates of 3% to 5% for the Starr-Edwards valve [4, 9], and an overall incidence of approximately 3.5% for the Björk-Shiley standard valve [9]. Fifty-eight patients (23.3%) had multiple events. Although this incidence is certainly substantial, this series demonstrates that only a few patients experienced more than four events over the entire study period. However, Figure 3 demonstrates that patients who had a first event are at an increased risk for subsequent events. Although the literature usually reports TE phenomena and BLE complications separately, we also wanted to report the global rate because we want to inform patients about the global risks of any complication. In addition, patients experiencing a TE or bleeding event are by no means separate patient groups, because they overlap considerably as previously reported [13]. Interestingly, one of the risk factors for TE or BLE complications was the occurrence of any operation other than aortic valve reoperation, during follow-up (OR 2.2). Anticoagulation is commonly interrupted and the patient is protected from adverse events with intravenous heparin [11, 14]. The risk of this interruption has been estimated [12] and reported [14] to be low but nevertheless, emerges as a risk factor in this series. It is conceivable that fluctuations in levels of anticoagulation make the patient more prone to complications as both the intensity and consistency of the anticoagulation are important factors in avoiding adverse events [9, 11]. Thromboembolic Events Out of a total of 140 TE events, two had lethal outcomes and 35 were strokes with residual impairment. This high incidence of 26.4% is undoubtedly related to the fact that 85% of the known INR at the time of the TE event were below the target base line. Inadequate or stopped anticoagulation is known to be strongly associated with increased risk of TE events [11 13], as also evidenced by the hazard ratio of 4.1. On the other hand, the current first event linearized incidence rate of 2% per patient year is very comparable with previously reported rates of 2% to 2.8% per patient year for the Björk-Shiley and Starr-Edwards valve [3, 4, 15]. Equally, the 10-year freedom from a first TE event is situated around 80%, which is, however, lower than the 86.7% reported by the Mayo Clinic in their long-term follow-up study of Starr- Edwards valves [16]. As for the global anticoagulation-related complications, patients who had a first TE event were at increased risk for multiple events (Fig 4). This was already demonstrated in a previous study, which in addition proved that the interval to a second TE event depends on the severity of the first event and not on its timing [17]. Bleeding Events Outcome for BLE events was generally worse than for TE events (eight lethal events versus two) as already stated by Cannegieter and colleagues [18]. Although the INR was known only in a minority of BLE events, most of these INRs were above the target maximum, in accordance with the literature [10 14, 18, 19]. The linearized incidence rate of 1.1% for the first BLE event in this series is better than in the previously mentioned studies [3, 4, 9] where rates of 1.2% to 2.2% per patient per year were found. An exceptionally high incidence rate of 5% per patient per year was found in the study by Borkon and colleagues [15] without obvious reason. In the conference discussion of that article it was suggested that the favorable incidence in the Netherlands was probably thanks to the nationally organized thrombosis service. However, equally good or even better results have also been reported without the aid of such an organization [16]. As for the global adverse event rate and the TE events, patients with a first BLE event were at increased risk for subsequent events (Fig 6). Practical Inferences One hundred and two patients (41% of all study patients) had 37 major TE events (out of 140 events), six valve thromboses and 72 major bleeding events. This means that out of a total of 218 TE and BLE events, with the exception of minor bleeding events, 115 or 53% were major events. In addition, 6.4% of all events were lethal. This high proportion occurred despite the fact that the target INR of 2.8 to 4.2 was not much different from the current recommendations: according to the literature, the overall target INR should be between 2.5 and 4.0 [18 21]. Therefore, although INR levels were regularly followed by a specialized organization that closely adhered to the official guidelines for anticoagulation, it is clear that the incidence of adverse events for the extremely long-term is still considerable and above all, in this series, not substantially inferior to other published series. Since a lot of patients experiencing an event had an inadequate anticoagulation level, it seems imperative to increase the INR control frequency, rather than change the target INR level, in order to reduce the frequency of events. Age per year was a risk factor for both TE events and bleeding events in the current series. This is still controversial in the literature because it confirms the findings by Cannegieter and colleagues [15], but contradicts the findings of several other studies [1, 9, 10, 13, 22, 23]. The fact that younger age could be less prone to complications than older age may be an argument in favor of mechanical aortic valve replacement versus other types of valve replacement. Of all other identified risk factors, only the type of valve and careful consistent anticoagulation regimen are

Ann Thorac Surg CASSELMAN ET AL 2001;71:1172 80 REPEATED TE AND BLEEDING EVENTS AFTER AVR 1179 correctable factors. As the incidence of adverse events was highest during the first 5 years, but remained present throughout the study period, careful monitoring seems the most important determinant of outcome. Limitations of the Study This series is a retrospective analysis over a long period of time. This inevitably raises the concern of the completeness of our data collection, despite major efforts to reduce this error to a minimum. Although it is unlikely that we missed major events, the registration of minor events (especially minor bleeding events) was more difficult. The reported risk may therefore be underestimated and the proportion of major events consequently overestimated. A limitation is the lack of registration of preoperative neurologic events. Some reports indicated a history of neurologic events as a risk factor for future events [11, 18], but the original 1975 follow-up of the patient cohort did not register this information, and therefore, we were unable to evaluate this variable. However, the demographics of the patient population suggest that probably only very few patients might have had a preoperative neurologic event: mean age was 41.8 years, only 13 patients (5.2%) had an aortic valve reoperation and only 7 patients (2.8%) had simultaneous coronary artery bypass grafting (indicative of systemic arteriosclerosis). Therefore, it is unlikely that the lack of this information affected our risk estimates. Another remark concerns the valves implanted in this series. Ball valves were implanted in 24% of hospital survivors and tilting disc valves in the remainder. While these valves are known to have an increased risk of anticoagulation-related complications as compared to the more recently available bileaflet mechanical valves [3, 4, 9, 11, 12, 14, 18, 19, 24], ball valves and tilting disc valves are currently still being implanted. Moreover, some recent reports suggest a lower target INR in current bileaflet valves with consequently lower bleeding complications without increasing TE events [25, 26]. It is not doubtful, but yet unproven, that the extremely long-term incidence of anticoagulation-related complications with the currently available bileaflet prostheses will be less than the currently reported incidence. With this in mind, we believe that this series should serve as a reference for future extremely long-term follow-up studies. Finally, in this retrospective study, we recorded all medication at the time of an event, but not in between events or in patients without events. Therefore, it was impossible to evaluate the use of antiplatelet drugs in the occurrence of TE or BLE events. In conclusion, at 30 years of follow-up, 46.8% of the patients remained free of any TE or BLE event and about one fourth of the patients (23.3%) experienced multiple events. This incidence of adverse events should be considered whenever an aortic valve operation is being considered for a particular patient. References 1. Baudet EM, Puel V, McBride JT, et al. Long-term results of valve replacement with the St Jude Medical prosthesis. J Thorac Cardiovasc Surg 1995;109:858 70. 2. Debetaz LF, Ruchat P, Hurni M, et al. St Jude Medical valve prosthesis: an analysis of long-term outcome and prognostic factors. J Thorac Cardiovasc Surg 1997;113:134 48. 3. Grunkemeier GL, Starr A, Rahimtoola SH. Prosthetic heart valve performance: long-term follow-up. Curr Probl Cardiol 1992;17:329 406. 4. Lund O, Pilegaard HK, Ilkjaer LB, Nielsen SL, Arildsen H, Albrechtsen OK. Performance of the Starr-Edwards aortic cloth covered valve, track valve, and silastic ball valve. Eur J Cardiothorac Surg 1999;16:403 13. 5. Van den Besselaar AM, Broekmans AW, Loeliger EA. INR: an internationally accepted standard for the monitoring of oral anticoagulant treatment. Ned Tijdschr Geneesk 1986; 130:1975 6. 6. Edmunds LH, Clark RE, Cohn LH, Grunkemeier GL, Miller DC, Weisel RD. Guidelines for reporting morbidity and mortality after cardiac valvular operations. Eur J Cardiothorac Surg 1996;10:812 6. 7. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457 81. 8. Best JF, Hassanein KM, Pugh DM, Dunn M. Starr-Edwards aortic prosthesis: a 20-year retrospective study. Am Heart J 1986;111:136 42. 9. Edmunds LH. Thrombotic and bleeding complications of prosthetic heart valves. Ann Thorac Surg 1987;44:430 45. 10. Petty GW, Lennihan L, Mohr JP, et al. Complications of long-term anticoagulation. Ann Neurol 1988;23:570 4. 11. Isreal DH, Sharma SK, Fuster V. Antithrombotic therapy in prosthetic heart valve replacement. Am Heart J 1994;127: 400 11. 12. Cannegieter SC, Rosendaal FR, Briet E. Thromboembolic and bleeding complications in patients with mechanical heart valve prostheses. Circulation 1994;89:635 41. 13. Gitter MJ, Jaeger TM, Petterson TM, Gersh BJ, Silverstein MD. Bleeding and thromboembolism during anticoagulant therapy: a population-based study in Rochester, Minnesota. Mayo Clin Proc 1995;70:725 33. 14. Carrel TP, Klingenmann W, Mohacsi PJ, Berdat P, Althaus U. Perioperative bleeding and thromboembolic risk during non-cardiac surgery in patients with mechanical prosthetic heart valves: an institutional review. J Heart Valve Dis 1999; 8:392 8. 15. Borkon AM, Soule L, Baughman KL, et al. Ten-year analysis of the Björk-Shiley standard aortic valve. Ann Thorac Surg 1987;43:39 51. 16. Orszulak TA, Schaff HV, Puga FJ, et al. Event status of the Starr-Edwards aortic valve to 20 years: a benchmark for comparison. Ann Thorac Surg 1997;63:620 6. 17. Starr A, Grunkemeier GL. Recurrent thromboembolism: significance and management. In: Butchart EG, Bodnar E, eds. Thrombosis, embolism and bleeding. London: ICR Publishers, 1992:402 15. 18. Cannegieter SC, Rosendaal FR, Wintzen AR, Van der Meer FJM, Vandenbroucke JP, Briet E. Optimal oral anticoagulant therapy in patients with mechanical heart valves. N Engl J Med 1995;333:11 7. 19. Cannegieter SC, Torn M, Rosendaal FR. Oral anticoagulant treatment in patients with mechanical heart valves: how to reduce the risk of thromboembolic and bleeding complications. J Intern Med 1999;245:369 74. 20. Liem TK, Silver D. Coumadin: principles of use. Sem Vasc Surg 1996;9:354 61. 21. Stein PD, Alpert JS, Copeland J, et al. Antithrombotic therapy in patients with mechanical and biological prosthetic heart valves. Chest 1995;108:371S-9S. 22. Masters RG, Semelhago LC, Pipe AL, Keon WJ. Are older

1180 CASSELMAN ET AL Ann Thorac Surg REPEATED TE AND BLEEDING EVENTS AFTER AVR 2001;71:1172 80 patients with mechanical heart valves at increased risk? Ann Thorac Surg 1999;68:2169 72. 23. Gurwitz JH, Goldberg RJ, Holden A, Knapic N, Ansell J. Age-related risks of long-term oral anticoagulant therapy. Arch Intern Med 1988;148:1733 6. 24. Butchart EG, Bodnar E. The influence of prosthesis related factors. In: Butchart EG, Bodnar E, eds. Thrombosis, embolism and bleeding. London: ICR Publishers, 1992:123 244. 25. Horstkotte D, Schulte H, Bircks W, Strauer BE. Unexpected findings concerning thromboembolic complications and anticoagulation after complete 10 year follow-up of patients with St Jude Medical prosthesis. J Heart Valve Dis 1993;2: 291 301. 26. Horstkotte D, Schulte HD, Bircks W, Strauer BE. Lower intensity anticoagulation therapy results in lower complication rates with the St Jude Medical prosthesis. J Thorac Cardiovasc Surg 1994;107:1136 45. INVITED COMMENTARY This study confirms the findings of Starr and Grunkemeier (reference 17 in the article) that the probability of suffering a repeat embolism after the first event is higher than that of having a first embolic event. The generally accepted explanation of this phenomenon is that patientrelated factors would cause or promote thromboembolism in a certain subgroup of the patient population. While in the majority of cases this assumption is probably true, a device-related etiology cannot be ruled out completely. Due to accepted and inevitable tolerances in the design and the manufacturing process, replacement valves coming off the same production line are not necessarily and completely identical. There are bioprostheses of identical design which are less durable than others, and it may well be that some mechanical valves of the same design and make are more thrombogenic than others. The actual rate of repeat embolism is perhaps the result of these two factors. There is now good clinical evidence to pinpoint certain patient-related risk factors, like atrial fibrillation, atherosclerosis, hypertension and others. Further research is necessary to reveal additional, but so far unknown risk factors. Equally important is the recent emergence in Atlanta of a nondestructive laboratory methodology pursued by Ajit Yoganathan and his team to assess valve thrombogenicity. Currently, they are comparing different valve designs. It may well be possible that with further refinements those delicate flow assessments might become a routine step in the quality assurance and comparison of individual mechanical valves of the same make and design. Another important finding of the study by Dr Casselman and associates is that the implant date is an independent predictor of the thromboembolic risk. Again, this is in concordance with the prior results of Starr and Grunkemeier, and adds to the number of etiological factors by proving that in addition to valverelated and patient-related events we must accept the actual, real-life existence of treatment-related thromboembolism. One must regret however that this 36- year follow-up study, due to its retrospective nature, cannot provide any information on the beneficial or detrimental effect of adding platelet inhibitors to warfarin, although about half of the patients was and the other half was not on combined warfarin plus antiplatelet treatment. Scientifically valid results coming from this larger and otherwise well documented study could have been significant in the future structuring of anticoagulation management. Nonetheless, this article remains one of the very few providing data on truly long-term outcome after aortic valve replacement with mechanical prostheses. Endre Bodnar, MD, PhD The Journal of Heart Valve Disease Crispin House 12A South Approach, Moor Park Northwood HA6 2ET, United Kingdom e-mail: bodnarendre@cs.com. 2001 by The Society of Thoracic Surgeons 0003-4975(01)$20.00 Published by Elsevier Science Inc PII S0003-4975(01)02430-4