Thromboembolic Complications After Fontan Operations
Background Despite the increasing recognition of thromboembolic complications of the Fontan procedure, data characterizing such events are limited. The total cavopulmonary connection is believed to be less prone to this complication than other modifications of Fontan operations. We examined our experience with thromboembolism after Fontan operations to better characterize these events and their relation to the type of Fontan operation performed.
Methods and Results We retrospectively identified 70 patients who underwent a Fontan operation between January 1978 and March 1994. Patients were divided into three groups: (1) total cavopulmonary connection, (2) atriopulmonary connection, and (3) conduit interposition. Fourteen patients (20%) developed a thromboembolic complication during a mean (±SD) follow-up of 5.2±4.7 years. The rate of thrombosis was similar in each group. The time from Fontan operation to thrombosis averaged 6.1±5.0 years. The overall rate of thromboembolic events was 3.9 per 100 patient-years. Twelve of the 14 thrombi were located within the venous circulation, 1 was in the left ventricle, and the location of 1 was undetermined. Six of the patients (43%) were asymptomatic, 3 (21%) presented with cerebrovascular events, and 5 (36%) presented with other symptoms. Thromboembolic events occurred from the perioperative period to 15 years after surgery.
Conclusions Thromboembolic complications occur frequently after the Fontan operation and its modifications and are a cause of significant morbidity. The time of presentation varies greatly. The rate of thrombosis appears to be similar in all modifications of the Fontan operation.
The Fontan operation was introduced in 1971 by Fontan and Baudet1 as a surgical therapy for tricuspid atresia and has subsequently come into more widespread use as the definitive surgical palliation for virtually all forms of univentricular congenital heart disease. As short- and long-term complications of the surgery have been recognized, successive modifications of the original procedure have been introduced.2 3 4 5 6 7 The most recent of these is the total cavopulmonary connection, developed in part with the hope of alleviating the problem of thromboembolic complications seen after the Fontan operation, by providing streamlined venous flow into the pulmonary arteries.2
Thrombi and thromboemboli have been reported as complications of the Fontan operation since 1978, but their incidence and clinical characteristics have not been well described.8 9 10 11 12 13 14 15 16 17 18 19 20 21 Furthermore, no data are available that document the efficacy of the total cavopulmonary connection in preventing these problems. To address these issues, we have reviewed our experience with thromboembolic complications after the Fontan operation.
This study was conducted in accordance with the policies of the Human Investigation Committee of the Yale–New Haven Hospital.
We retrospectively reviewed the medical and surgical records and autopsy notes of all 70 patients who underwent a Fontan operation or one of its modifications at Yale–New Haven Hospital between January 1978 and March 1994.
For patients in whom there was a clinical suspicion of thrombus formation, echocardiography was performed transthoracically and followed by transesophageal echocardiography. On occasion, angiography, employed for hemodynamic evaluation, established the diagnosis. Lung perfusion scans were employed in selected instances but were never the sole means by which a thrombus was diagnosed.
A number of thrombi were found unexpectedly in asymptomatic patients, all diagnosed with transthoracic echocardiography, which is routinely performed annually on our patients. When necessary, these findings were confirmed by more-invasive studies as described above.
Results are expressed as mean±SD unless otherwise indicated. Comparisons between two groups were performed using an unpaired Student’s t test, Fisher’s exact test, and χ2 test. Multiple-group comparisons were performed by ANOVA and Student-Newman-Keuls tests. Life-table analysis was performed using Kaplan-Meier plots. Statistical significance was defined as P<.05.
Between January 1978 and March 1994, a total of 70 patients underwent a Fontan operation or one of its modifications. The mean patient age at the time of operation was 7.8±7.6 years (range, 0.8 to 33.9 years). There were 43 males (61%) and 27 females (39%). The anatomic diagnoses were tricuspid atresia in 34 patients (48%), pulmonary atresia with intact ventricular septum in 4 patients (6%), unbalanced atrioventricular septal defect in 4 patients (6%), and other forms of complex univentricular heart in 28 patients (40%).
Patients were divided into three groups for analysis. Group 1, which consisted of 23 patients, had total cavopulmonary connections. Within this group, 18 patients had true lateral tunnel Fontan operations, whereas 5 patients had earlier versions of the total cavopulmonary connection in which the atrial baffle was not a true lateral tunnel. All of the lateral tunnels were constructed of expanded polytetrafluoroethylene (Gore-Tex). A surgically created fenestration was made in 11 of the tunnels to allow decompression of the systemic venous atrium. Group 2 consisted of 20 patients who had direct atriopulmonary anastomoses. Of these, 9 required patch augmentation of the anastomosis. The patch material was pericardium in all patients. Group 3 contained 25 patients who had an extracardiac conduit as part of their Fontan procedures. The conduits were placed from right atrium to either right ventricle or pulmonary artery. Of the 25 conduits, 17 (68%) contained valves, whereas the remainder did not. Our other 2 patients underwent Kawashima operations in lieu of Fontan operations. This operation entails construction of a bidirectional superior vena cava–to–pulmonary artery anastomosis in patients with an interrupted inferior vena cava. By way of the azygous continuation, lower-body venous return is routed directly to the pulmonary arteries. Although the Kawashima operation is not strictly considered a Fontan operation, we have included these 2 patients in this analysis because of the anatomic and physiological resemblance of the Kawashima operation to the Fontan operation. These 2 patients are included in the summary data for patients and thromboembolic events. They have not been analyzed as a separate group because the numbers are too small for meaningful intergroup comparisons.
A similar proportion of patients in each group were female: 43% in group 1, 20% in group 2, and 40% in group 3. These differences were not statistically significant. The mean age at the time of the initial Fontan operation was 6.6±9.7 years in group 1, 7.2±6.4 years in group 2, and 9.7±6.3 years in group 3. These differences also were not significant. The conduit interposition variation was the earliest form of Fontan operation performed at our institution, and the total cavopulmonary connection is the most recent modification introduced, as reflected by the median year of surgery, which was 1981 for group 3, 1988 for group 2, and 1992 for group 1. As a consequence of this evolution in surgical preference, the patients in group 1 had the shortest follow-up, 1.9±1.5 years, as compared with 5.5±4.9 years for group 2 and 8.3±4.7 years for group 3. These differences were all significant (P<.05). The total duration of follow-up was 357 patient-years, with 43.5 patient-years for group 1, 109.3 patient-years for group 2, and 200.1 patient-years for group 3.
The perioperative mortality for our patients was 3% (death in the first 30 days after surgery), whereas cumulative mortality was 11% at 1 year, 20% at 3 years, and 26% at 5 years.
During the period covered by this study, we did not use early postoperative heparin therapy. Warfarin and aspirin were not routinely administered, either postoperatively or long term. It is our current practice to use anticoagulation therapy with older patients, and to treat those felt to be at high risk for thrombosis with warfarin therapy. Younger patients, in whom titration of warfarin can be a problem, are given aspirin. In either case, the medications are initiated several days after the Fontan operation and prior to hospital discharge.
There were a total of 14 thromboembolic events among the 70 patients, for an overall occurrence rate of 20% or 3.9 events per 100 patient-years. One of these events occurred in a patient who had undergone the Kawashima operation. Several patients had reappearance of thrombi after apparently successful initial medical therapy. These were counted as single events because of the difficulty in differentiating recurrence from persistence of thrombus.
The distribution of thromboses among the three groups is shown in Table 1⇓. The risk of thrombus formation was similar in the three groups, with a total occurrence rate that ranged from a low of 13% in group 1 to a high of 24% in group 3, and with an event rate per 100 patient-years that ranged from a low of 3.0% in group 3 to a high of 6.9% in group 1. The time from Fontan operation to development of thrombosis tended to be lower in group 1, but the difference was not statistically significant. However, the age at the time of thromboembolic event was significantly lower for patients in group 1 (4.3±2.5 years) than for group 2 (12.0±6.3 years) or group 3 (20.4±3.2 years). This reflects the combined contribution of younger age at initial surgery, shorter follow-up duration, and increased vigilance on our part during the last 3 years.
Several potential risk factors for development of thromboembolic complications were analyzed (including sex, age at time of Fontan operation, associated cardiac arrhythmia, presence of a valved conduit, and type of material used for conduit). Ventricular systolic performance was not included in this analysis because of the inaccuracy, particularly in a retrospective study, of determining ventricular function echocardiographically in patients with abnormal ventricular geometry.
The mean age at Fontan operation for patients who developed thromboses was 7.4±5.2 years, which was not significantly different from those patients free of thromboses (7.9±8.1 years). The proportion of female patients was 28% in those with thromboses and 41% in those without thromboses, which was also not significantly different.
Altogether, 34 patients had one or more arrhythmias identified either by ambulatory ECG monitoring or the resting ECG. Among these patients, 10 had evidence of sinus node dysfunction, 7 of atrioventricular conduction delay (including 3 with complete heart block), and 24 of atrial tachyarrhythmias (because some patients had more than 1 abnormality, the total is more than 34). An arrhythmia was found in 71% of patients who subsequently developed thromboses versus 43% of those who did not. This difference approached but did not achieve statistical significance (P=.08). Of interest is that 70% of patients who developed thromboses had normal sinus rhythm at the time of thrombus detection, whereas the remaining 30% had atrial tachyarrhythmias.
Of the 25 patients with conduits, 17 patients had valves within the conduit. Five of these 17 patients developed thromboses (40%), whereas only 1 of the 8 patients (12%) with a nonvalved conduit developed a thrombosis. The conduits consisted of synthetic material in 17 patients, 6 of whom (35%) developed thrombosis, whereas none of the 8 patients with homograft or pericardial conduits developed thromboses. However, although these differences are intriguing they are not statistically significant, probably because of the small size of the groups.
The majority of patients who developed thromboses (9 of 14, 64%) were taking no medications at the time of thrombus formation. Of the 5 patients who were taking some medication, 4 patients were taking digoxin, 5 a diuretic, 2 an angiotensin-converting enzyme inhibitor, and 1 warfarin. None were taking aspirin. The patient on warfarin was thought to be particularly compliant with therapy, with monthly monitoring of therapeutic effect documented over the 3 months preceding the thrombus detection, and with international normalized ratio (INR) of the prothrombin time that ranged from 1.6 to 2.4.
The earliest observed thrombus was detected 6 days after surgery, whereas a thrombus has been detected as late as 13.6 years after surgery, a time span limited by our follow-up. Two of the thrombi occurred perioperatively (ie, within 30 days of surgery). The relation between the risk of thrombosis and the time elapsed after surgery was analyzed with a Kaplan-Meier plot (Fig 1⇓). This analysis demonstrates the likelihood of remaining free of thrombosis at any given time after surgery, for all patients with survival and follow-up extending through the given time interval. There is an initial drop in the curve corresponding to the perioperative thromboses, followed by a fairly constant risk of thrombosis over time. There is no plateau; therefore, the risk of thrombosis appears to persist throughout the patient’s lifetime. The various surgical groups were not analyzed separately in this fashion because the numbers are too small to allow meaningful comparison in this manner.
Patients presented with a variety of symptoms from their thromboembolic events. Three presented with cerebrovascular events, representing 21% of the total. These events were presumed to result from paradoxic emboli. None of these patients had deliberate fenestration of their Fontan, although two were known to have hemodynamically insignificant residual atrial-level shunting at the site of atrial suture lines. Two patients presented with signs of venous obstruction. One patient, with a fenestrated lateral tunnel Fontan, developed worsening perioperative cyanosis due to partial obstruction within the lateral tunnel that caused increased right-to-left shunting across the fenestration. One patient presented with chest pain and was found to be in atrial flutter; this prompted a careful evaluation for thrombus, which was identified within the right atrium. One patient had palpitations; he was found to be in atrial fibrillation. He had no thrombus visible with echocardiography but did have pulmonary emboli identified with angiography. However, the largest of the groups, consisting of six patients (43% of the total), were asymptomatic at presentation and were found to have thrombi on routine annual echocardiograms.
The outcome of each patient who had a thrombus is detailed in Table 2⇓. Of the six patients who were asymptomatic when their thrombi were identified, five had resolution of thrombus with heparin or thrombolytic therapy, and one who presented with a large left ventricular (LV) thrombus died of a massive cerebrovascular embolus while receiving heparin. Of the three patients who presented with cerebrovascular accidents, one had virtually complete resolution of symptoms, whereas the others were left with significant hemiplegias. In two patients surgical thrombectomy was performed with a successful result in one, whereas the other (who presented in a moribund condition) died during surgery. One other patient, who had thrombi located primarily within the lateral tunnel and pulmonary emboli, had a partial response to thrombolytic therapy but died in the perioperative period of multiorgan failure after a difficult postoperative course. Thus, overall mortality was 21%.
Detection of Thrombi
Virtually all of the thrombi were found in the venous circulation. Of the 14 patients with thromboembolic events, 7 (50%) had thrombi found in the right atrium, whereas 3 others (21%) had thrombi within the lateral tunnel. One of these latter patients also had multiple small pulmonary emboli found on angiography; the lateral tunnel thrombus was presumed to be the primary source. One patient had thrombus identified within the superior vena cava; 1 patient had thrombus in the pulmonary artery with no atrial thrombus found by either transesophageal echocardiography or angiography. One patient who denied symptoms at time of presentation was found to have extremely poor LV function and had a very large thrombus found within the LV cavity at the time of routine evaluation. Also, 1 patient presented with an embolic stroke but the source of the embolism was never found (the patient who had undergone the Kawashima operation). Ultimately, after a period of anticoagulation, this patient underwent cardiac catheterization that demonstrated a collateral vessel from the azygous vein to the hepatic venous drainage. Because of the possibility that the embolus had traversed this collateral to enter the systemic circulation via the inferior vena cava, it was closed with coil embolization. This patient has been maintained on warfarin therapy since the time of embolization with no further thromboembolic events.
The year of thrombus detection is depicted in Fig 2⇓. Although the first thromboembolic event occurred in 1980, 71% of all thromboembolic events have been detected since 1992, including all the patients with asymptomatic thrombi. Over this same time period (1992 to the present), transthoracic echocardiography has successfully identified thrombi in 7 of the 10 patients. In 1 of the 3 patients in which transthoracic echocardiography was negative, transesophageal echocardiography was employed that successfully demonstrated thrombus within a lateral tunnel. This patient was dependent on mechanical ventilation, was critically ill, and had very poor transthoracic echocardiographic windows. In the other 2 patients in whom transthoracic echocardiography was negative, transesophageal echocardiography was not performed. In 1 of these (the patient who had the Kawashima operation), the thrombus was never found; in the other, pulmonary emboli were identified by angiography. Transesophageal echocardiography was also performed in 3 of the patients in whom transthoracic echocardiography did locate the thrombus; in each of these patients, transesophageal echocardiography also demonstrated the thrombus, and it is our impression that it provided substantially better imaging of the thrombus than the transthoracic route, particularly in older patients in whom transthoracic echocardiographic windows are frequently suboptimal. This comparison is demonstrated in Fig 3⇓, in which the transthoracic and transesophageal images in a patient with right atrial thrombus are compared.
There are individual case reports and one series of four patients who developed thromboembolic complications following Fontan operations but there are no large published series that examine the incidence and clinical features of such thromboembolic events.20 This issue is relevant given the widespread application of the Fontan procedure, the substantial morbidity from thromboembolic events, and the theoretical but unproven advantage of the total cavopulmonary connection in diminishing the incidence of these events.2 12
In our patients, the overall risk of developing thromboemboli after any type of Fontan operation was 3.9 events per 100 patient-years. Thrombotic events are seen as early as the perioperative period, and the risk for such events appears to persist during the follow-up period, with no plateau. This is consistent with previous reports of thrombosis, which have been described as early as 24 hours and as late as 5 years after surgery.20 21 Such studies have identified acute dehydration as a potential etiologic factor for thrombosis but have not identified other putative clinical risk factors for this complication. Our data did not clearly identify any clinical or surgical risk factors for the development of thrombosis that could be used to stratify patients, although there was a suggestion that supraventricular arrhythmia was correlated with an increased risk of subsequent thrombosis. Presumably, arrhythmias would act to increase the risk of thromboemboli by increasing stasis and turbulence within the atria and by diminishing cardiac output. A statistically significant definition of the role of arrhythmia in thrombogenesis will require a somewhat larger series.
We were not able to demonstrate a statistically conclusive link between the type of conduit employed and the risk of thrombosis. Nonetheless, it is striking that none of the 8 patients with homograft or pericardial conduits developed thromboses, whereas 35% of the 17 patients with synthetic conduits went on to develop thromboses. It is important to note that we have found that thromboemboli occur after all modifications of the Fontan operation, including the lateral tunnel, and that there does not appear to be any meaningful difference in the risk of thrombosis based on the modification of the Fontan.
Also of interest is our finding of a stroke due to a paradoxical right-to-left thromboembolic event in a patient who had undergone the Kawashima operation, which allows hepatic blood and coronary sinus blood to enter the heart directly and thereby constitutes a right-to-left shunt. However, it is not thought to present a high risk of paradoxical embolus because of the exclusion of right-to-left shunting of femoral or iliac venous blood. Moreover, although hepatic venous thrombosis has been noted in Fontan patients, it has been on the systemic venous side of the circulation where high pressure and relative stasis can be expected.22 This differs somewhat from the situation created by the Kawashima operation, in which the hepatic veins are on the pulmonary venous side of the circulation. Our patient developed collateral vessels that allowed blood from the azygous vein to enter the hepatic venous return to the heart. We surmise that this was the source of the paradoxical embolus.
Previous reports have not focused on the potential for asymptomatic thrombosis in Fontan patients, whereas in our series a high proportion of patients (43%) were asymptomatic at the time of thrombus detection and were found to have thrombi based on routine annual echocardiography.11 12 13 14 15 16 17 18 19 20 21 Of the others, three had cerebrovascular events. Of the patients with cerebrovascular events, none had purposely created surgical fenestrations, but two had residual atrial level right-to-left shunting at suture lines. This highlights the variable clinical setting in which thrombosis can occur and emphasizes the need for careful, ongoing surveillance for thromboemboli in patients who have undergone Fontan operations. These findings further suggest that our data define a minimum rate of thromboembolic complications because not all of our patients’ thromboses will necessarily have been detected in this study.
Morbidity and mortality from the thromboembolic events have been substantial. Altogether, 3 of the 14 patients (21%) who had thromboemboli died proximate to the detection of the thromboemboli. In 1 patient, death was directly attributable to embolization of a LV thrombus, whereas 1 patient died during an attempted surgical thrombectomy, and 1 patient died near the time of thrombus discovery from multiple causes (of which the thrombosis was just one contributor). Of the 11 survivors, 2 patients (18%) have a significant residual neurological disability. Even patients without permanent disability required a significant period of hospitalization for therapy. None of the patients who were asymptomatic at the time of thrombus detection has subsequently had a symptomatic thromboembolic event, supporting the role of an aggressive approach to detection and treatment.
Consistent with previous reports, most of our patients’ thromboemboli originated in the venous circulation, with the majority of these located either in the true right atrium or within the lateral tunnel.11 12 13 14 15 16 17 18 19 20 21 A single patient who had severe LV dysfunction had a LV thrombus, whereas in one patient the site of the original thrombus could not be determined but was probably within the venous circulation. The preponderance of right-sided thromboses may be strictly due to hemodynamic factors such as elevated venous pressure and flow stasis, but there are several reports that suggest that the coagulation cascade is actually abnormal in Fontan patients, and it is possible that this may predispose to thrombogenesis.23 24 The true pathogenesis of the thromboembolic events must be considered unknown at the present time.
Our detection of thrombi has increased markedly in recent years as compared with the earlier period of the study. This may simply reflect an accumulation of a larger period of follow-up, but it is likely that the increased detection also reflects an improvement in the technology of echocardiography, which is the primary means by which thrombi are detected. In addition, we have become increasingly aware of the potential for thrombosis, which undoubtedly increases the detection rate. These factors, as well as the retrospective nature of the study and the high frequency of asymptomatic thromboses, lead to the conclusion that the rate of thrombosis we report is most likely a minimum estimate of the true risk.
In this regard, we believe that routine use of transesophageal echocardiography may be beneficial in evaluating the patient for the presence of thrombosis after the Fontan procedure. In our hands, transesophageal echocardiography provided better delineation of thrombi than did transthoracic echocardiography, and in one patient detected thrombus that was not identified at all by transthoracic echocardiography. This is consistent with the reported experience of Stümper et al19 and Fyfe et al15 in which transesophageal echocardiography was able to identify intracardiac thrombi not seen by transthoracic echocardiography.
Our data suggest a considerable risk of thromboembolic complications after the Fontan operation. This risk is comparable to that seen in other conditions, such as aortic and mitral valve prostheses, for which routine anticoagulation therapy is recommended.25 Although Fontan operations are now generally performed at an age at which titration of warfarin can be full of problems, a prospective trial of the efficacy of warfarin or other antithrombotic agents in preventing thromboemboli seems indicated. Further work clarifying the pathogenesis of these thromboembolic events may also help to guide future therapy, with respect to both surgical and medical options.
In summary, we report a high incidence of thromboembolic complications after the Fontan procedure and its modifications. The risk appears to be largely independent of the particular modification performed. These thromboembolic events cause substantial morbidity and mortality. The use of synthetic material in the Fontan procedure may be associated with the development of such events, as may the presence of cardiac arrhythmias, although these relations are not firmly established by our data. Many patients will be symptomatic from thromboemboli, but a large minority are asymptomatic. This finding, as well as the attendant morbidity and mortality in symptomatic patients, merits aggressive detection of thromboemboli in all Fontan procedure patients. This may be best accomplished through diligent transthoracic or transesophageal echocardiography. The efficacy of antithrombotic therapy in this setting is unknown, but the risk of thromboemboli is sufficiently high to warrant a prospective trial of antithrombotic therapy.
Reprint requests to Charles S. Kleinman, MD, Department of Pediatrics, Section of Pediatric Cardiology, LCI 302, Yale University School of Medicine, 333 Cedar St, New Haven, CT 06510.
- Copyright © 1995 by American Heart Association