Development of Pulmonary Arteriovenous Fistulae in Children After Cavopulmonary Shunt
Background The cavopulmonary shunt procedure is now used for palliation of complex congenital heart lesions in infants. While pulmonary arteriovenous fistulae (PAVF) are a well-known complication of this surgery in older patients, no study of the prevalence of this condition in children and young infants has been reported.
Methods and Results We compared 29 patients with cavopulmonary shunts or total caval exclusion with 53 control subjects evaluated by contrast echocardiography at the University of California, San Francisco. The primary cardiac lesion, age at the time of surgery, type of right heart bypass procedure, provision of auxiliary pulmonary blood flow, and changes in oxygen saturation over time were compared. The prevalence of PAVF in children after cavopulmonary anastomosis is 60%, higher than previously reported. The prevalence is significantly higher in infants <6 months old and in those with a heterotaxy syndrome. The provision of an additional source of pulsatile, pulmonary blood flow appears to have little effect on the development of PAVF. Patients who developed PAVF had arterial oxygen saturations at the time of discharge from surgery similar to those who did not develop them. Those with PAVF had significantly lower arterial and pulmonary venous oxygen saturations at follow-up as a result of their intrapulmonary shunt.
Conclusions Contrast echocardiography provides a sensitive method for the detection of PAVF. While the origins, natural history, and ultimate clinical significance of PAVF in children after cavopulmonary anastomosis are unclear, surveillance by contrast echocardiography is indicated for all patients who have had this procedure because PAVF may cause significant intrapulmonary right-to-left shunting in some patients.
Over the past decade, cavopulmonary anastomosis has become the surgery of choice for treatment of single-ventricle physiology.1 2 3 The occurrence and management of pulmonary arteriovenous fistulae (PAVF) have been studied as a “late” complication of this surgery in older patients.4 5 6 7 8 9 Recently, the cavopulmonary shunt procedure has been applied to infants to avoid complications arising from aortopulmonary shunts.10 We described the fulminant development of microscopic PAVF in one such patient, which resulted in his early demise, and identified for the first time the histopathologic site for such intrapulmonary shunting.11 However, a large-scale study of the prevalence of PAVF in children who have undergone cavopulmonary anastomosis has not been reported. We studied 29 pediatric patients who had undergone cavopulmonary anastomosis or total caval exclusion and reviewed previously acquired data for 53 control patients to assess the prevalence of PAVF as detected by contrast echocardiography.
For the purpose of this study, 32 patients with cyanotic heart disease who had undergone cavopulmonary shunt or total caval exclusion from 1993 to 1994 at the University of California, San Francisco, were identified. Three patients were subsequently excluded because collateral connections from the superior to inferior vena cava were identified during the contrast echocardiogram (see below). Of the remaining 29 patients, 20 had undergone cavopulmonary (superior vena cava–to–pulmonary artery) anastomosis. Nineteen patients had flow to both lungs through their anastomoses (commonly known as a bidirectional Glenn procedure), while one had unilateral flow (a classic Glenn). Nine patients had undergone a total caval exclusion procedure, consisting of anastomosis of both superior and inferior venae cavae to the pulmonary arterial circulation (a Fontan-type procedure), as either a single (2 patients) or a staged procedure (7 patients). In the total caval exclusion group, 5 underwent connection of the inferior vena cava to the pulmonary arterial circulation by extracardiac conduit and 4 by lateral atrial tunnel.
Fifty-three children without cavopulmonary shunts were identified by review of all patients who had undergone contrast echocardiography from 1992 to 1994 in the Pediatric Echocardiography Laboratory at the University of California, San Francisco. Contrast studies were performed to rule out an intracardiac shunt before transvenous pacemaker placement, after cerebrovascular accident, or in the evaluation of congenital heart disease. Patients evaluated by peripheral contrast injection in whom an intracardiac shunt was demonstrated were excluded because the source of contrast in the left atrium could not be determined reliably. Seven control patients with intracardiac shunts had selective pulmonary arterial injection of echocardiographic contrast at catheterization.
This study was approved by the Committee for Human Research at the University of California, San Francisco. Charts were reviewed in accordance with institutional guidelines. All prospective study patients gave informed consent.
Contrast echocardiography was performed in the Pediatric Echocardiography Laboratory in accordance with institutional guidelines, as described by Van Hare and Silverman.12 Contrast consisted of 3 mL of agitated saline in patients weighing <20 kg and 6 mL in patients weighing >20 kg. When possible, 0.3 mL of the patient’s blood was mixed with 10 mL of saline before agitation. Contrast was administered as rapidly as possible through a peripheral intravenous catheter in the upper extremity or through a side-hole catheter placed in the main or branch pulmonary arteries at the time of cardiac catheterization. Experience in our laboratory indicated that rapid administration of contrast was more important than volume administered in achieving adequate delivery of contrast, as determined by opacification of the pulmonary artery. Precordial echocardiographic images were obtained using subcostal, parasternal, or suprasternal views at the time of contrast delivery. A study was considered positive if microcavitations were seen in the pulmonary venous atrium within five cardiac cycles of their injection into the pulmonary arterial circulation on at least two injections (see the Figure⇓).
Several controls were incorporated into each examination. In all patients, opacification of the pulmonary artery by microcavitations was demonstrated to ensure adequate administration of contrast. The absence of contrast in the inferior vena cava was then confirmed to exclude the possibility of venous collateral vessels through the azygous system. The absence of contrast also was demonstrated in the systemic venous atrium and coronary sinus to exclude other systemic venous collaterals. Three patients were excluded from the study because of the existence of such collateral vessels.
Oxygen Saturation Measurements
Systemic oxygen saturation data were determined by transcutaneous pulse oximetry. Pulmonary venous saturation values were measured by hemoximetry. All measurements were obtained while patients were breathing room air.
Fisher’s exact test was used to perform two-tailed tests of significance of the associations between the presence of PAVF and several historical variables.13 Relative risks for developing PAVF and predictive values were calculated by use of standard methods.14 Oxygen saturation data were compared by the nonparametric Mann-Whitney U statistic for unpaired data.15 Because only a decrease in oxygen saturation accompanying PAVF would be considered significant, a one-tailed test was used for these comparisons. The threshold for significance was set at P=.05 for all tests.
Table 1⇓ lists the features of the control and study samples. The median age and age range were similar in both groups. The male-to-female ratio in the study group was almost twice that in the control group. While a male predominance has been reported for some hypoplastic right-ventricle syndromes, such as tricuspid atresia and pulmonary atresia with intact ventricular septum,16 17 the balance between sexes observed in our study group was not statistically significant (P=.16). Heart lesions seen in the study group reflected the range of defects referred for cavopulmonary shunt or total caval exclusion. A variety of cardiac and noncardiac illnesses as well as unremarkable medical histories were encountered in the control group. Seven patients in the study group but none of the control patients had heterotaxy syndromes.
Prevalence of PAVF After Cavopulmonary Shunt
Three control patients and 15 study patients had PAVF, as demonstrated by contrast echocardiography. Table 2⇓ summarizes the clinical and laboratory findings in these patients. Table 3⇓ summarizes the same data for 14 patients who had undergone cavopulmonary shunt surgery but did not have echocardiographic evidence of PAVF. Of the 3 patients in the control group who had evidence of PAVF (6%), 1 had hepatic insufficiency and 1 had severe pulmonary disease, conditions known to result in intrapulmonary shunting.18 19 The third control patient (patient 16 in Table 2⇓), who at 1 month old had a univentricular heart palliated with an aortopulmonary shunt, was found to have contrast evidence of PAVF at 5 months old.
Compared with 3 of 53 control patients, 12 of 20 patients who underwent cavopulmonary anastomosis had echocardiographic evidence of PAVF (60%; P<.001). Of the 9 patients who had a total caval exclusion procedure, 3 had PAVF by contrast echocardiography (33%; P=.04 versus control group). We evaluated the effect on the development of PAVF of an additional source of pulsatile pulmonary blood flow, such as a banded main pulmonary artery or aortopulmonary shunt, and found that 6 of 13 patients with additional pulmonary blood flow had PAVF (46%), as did 9 of 16 patients without an additional source of pulmonary flow (53%; P=.99).
Effect of PAVF on Oxygen Saturation
Because PAVF might be expected to produce pulmonary venous desaturation,11 we compared pulmonary venous oxygen saturations obtained at cardiac catheterization in 7 patients with and 6 patients without PAVF. As anticipated, patients with echocardiographic evidence of PAVF had lower pulmonary venous saturations than those who did not (mean oxygen saturation, 90% versus 97%; P=.01). The subgroup of patients with PAVF who underwent catheterization appear to be representative of the PAVF group as a whole; they had systemic oxygen saturations similar to patients who were not catheterized, and they had the same spectrum of diagnoses (Table 2⇑). Interestingly, 3 patients (patients 3, 8, and 11) were found at catheterization to have PAVF confined to one lung lobe. In 2 of these patients, pulmonary venous saturations were measured and desaturation was confined to the vein draining the affected lobe. These findings demonstrate that PAVF, as identified by contrast echocardiography, are of functional significance.
To define the effect of PAVF on systemic oxygenation, we also compared systemic oxygen saturations of patients with and without PAVF. Systemic oxygen saturation measurements were available for 28 of the 29 patients who had undergone shunt surgery. Patients who did not develop PAVF had no change in systemic oxygen saturation between discharge and the time of study (88% versus 89%, P=.34). Patients who developed PAVF had initial arterial oxygen saturations similar to those who did not develop PAVF (86% versus 88%; P=.51). However, saturations decreased in the PAVF group between surgery and follow-up (86% versus 79%; P=.03).
We also examined whether desaturation over time would be useful in predicting the presence of PAVF in individual patients. Nine of 14 patients with PAVF for whom data were available had a decrease in saturation of at least 5%, compared with 1 of 10 patients without PAVF (P=.01). The positive predictive value of a 5% decrease in systemic oxygen saturation was 90%. However, because contrast echocardiography can detect small intrapulmonary shunts that produce little effect on systemic oxygen saturation, the negative predictive value is only 64%. Hence, while a change in systemic oxygen saturation of >5% after placement of a cavopulmonary shunt should raise suspicion that PAVF have developed, the absence of desaturation does not rule out their presence.
Effect of Age and Heterotaxy on the Development of PAVF
The group of patients who developed PAVF were, on average, significantly younger at the time of operation than those in whom they did not develop (mean, 3.0 versus 7.5 years; P=.005). Of the 20 patients who underwent cavopulmonary anastomosis, 6 were <6 months old at the time of surgery. All 6 patients had positive contrast studies. Only 6 of 14 patients >6 months old at the time of surgery had PAVF (43%; P=.04 versus patients <6 months old). Compared with older children who had a cavopulmonary anastamosis, the relative risk of developing PAVF for infants <6 months old was 2.3.
Because four of the patients who were operated on before 6 months of age had a heterotaxy syndrome, we evaluated heterotaxy as an independent risk factor for PAVF. Six of 7 study patients with heterotaxy had PAVF (86%) compared with 9 of 22 study patients with normal situs (41%). While this difference was not statistically different (P=.08), the number of patients with heterotaxy was small, and we cannot be certain that these syndromes do not pose a special risk for patients <6 months old who undergo cavopulmonary anastomosis.
The length of follow-up for patients both with and without PAVF ranged from <1 month to >9 years (mean, 30 versus 24 months, respectively; P=.48). However, we noted that patients with PAVF who had been followed for the longest periods after surgery (patients 4 and 8) had some of the largest decreases in systemic oxygen saturation. Indeed, all 6 patients with PAVF who were followed up for >2 years had a decrease in systemic oxygen saturation of ≥5% compared with 3 of 8 patients followed up for <2 years (P=.03). This indirectly suggests that the functional consequences of PAVF progress with time.
Regression of PAVF in One Patient
One patient with an interrupted aortic arch, subaortic stenosis, a ventricular septal defect, and branch pulmonary artery stenoses underwent cavopulmonary anastomosis and aortopulmonary shunt at 8 months (patient 12, Table 2⇑). Contrast echocardiography 2 years 9 months after surgery demonstrated the presence of PAVF. She subsequently underwent closure of the ventricular septal defect and repair of the pulmonary artery and left ventricular outflow tract. Her cavopulmonary shunt was left intact. Remarkably, she had a negative contrast study within 2 weeks of biventricular repair.
PAVF have been postulated to represent the persistence of minute arteriovenous communications present in normal fetal and neonatal lung that enlarge to become fistulous and cause precapillary shunting and arterial desaturation.20 Gelatin–calcium carbonate, when injected into the pulmonary arteries of stillborn human fetuses, demonstrates vascular channels that bypass the alveolar capillary bed.21 These channels may play a role in the developing pulmonary circulation before the establishment of an alveolar capillary network.21 Neonatal lungs, therefore, may be more likely to retain these primordial arteriovenous connections and thus increase the susceptibility of younger patients to the development of clinically significant PAVF.
We recently described the histopathology of microscopic PAVF in a 7-month-old infant with polysplenia and profound hypoxemia occurring 5 months after a cavopulmonary shunt.11 At autopsy, we found dilated pulmonary veins near the pleural surface and identified abnormal thin-walled vessels within the lung lobule by immunohistochemistry. The diameter of these abnormal vessels was 50 to 100 μm, approximating the size of alveoli, thereby making them inconspicuous during routine light microscopy. These vessels appear to be the anatomic site of intrapulmonary shunting.
Historically, the evaluation of PAVF has been hampered by the lack of a reliable tool for surveillance of patients at risk for PAVF development. While the use of angiography and radiolabeled microaggregated albumin to demonstrate PAVF has been described,5 8 9 both of these techniques are more cumbersome and appear to be less reliable than contrast echocardiography. Angiography is insensitive: In our study, an angiographic diagnosis could be made in only 2 of 9 patients with PAVF who had undergone angiography. The diagnostic utility of microaggregated albumin, on the other hand, is limited by the large number of false-positives,9 probably owing to the inability to identify the variety of systemic venous collateral connections seen in patients with cavopulmonary shunts.
Our findings establish contrast echocardiography as a sensitive method for the early detection of PAVF that can be performed as part of the routine echocardiographic surveillance of these patients. Specific attention, however, must be paid to technique because inadequate delivery of contrast and the presence of collateral systemic venous connections may result in false-negative and false-positive results, respectively. In our experience, the demonstration of contrast in the pulmonary arterial circulation and interrogation of the inferior vena cava for collateral delivery of contrast to the atrium can prevent these diagnostic errors.
A positive contrast echocardiogram was accompanied by pulmonary venous desaturation in a representative subgroup of our patients who underwent cardiac catheterization, demonstrating that PAVF have functional consequences. The sensitivity of contrast echocardiography is demonstrated by its ability to detect the intrapulmonary shunt before the degree of pulmonary venous desaturation is sufficient to cause a major alteration of systemic arterial saturation. Contrast echocardiography also is specific. Within our control group, only patients with conditions known to be associated with PAVF had positive contrast echocardiograms, providing a positive control for this technique independent of the study group. It is of note that one of these was a patient with single-ventricle physiology, suggesting that the underlying congenital heart diseases for which cavopulmonary shunts are performed may play an independent role in PAVF development.
The development of PAVF after cavopulmonary shunt in older children is well described. Three series composed of >100 patients >12 months old, reported a 25% prevalence of intrapulmonary shunting at 2 to 13 years after cavopulmonary anastomosis.4 5 9 Because it is now possible to achieve favorable postoperative hemodynamics in infants <12 months old by use of this procedure,10 it is important to question whether these patients are at increased risk for developing PAVF.
Our data suggest that the prevalence of PAVF in pediatric patients who have undergone cavopulmonary anastomosis is higher than previously reported. In addition, these results indicate that the risk in very young infants is significantly higher than in older children. An additional source of pulsatile pulmonary blood flow does not appear to have a significant effect on the development of PAVF in children. Although the prevalence of PAVF in patients who had a total caval exclusion procedure was lower than in children with cavopulmonary anastomoses, we were unable to determine whether this reflects regression of PAVF after conversion of superior vena cava–pulmonary artery anastomosis to total caval exclusion or that total caval exclusion was performed in an older population. Nonetheless, our results demonstrate that PAVF occur or persist in a significant proportion of pediatric patients who have undergone total caval exclusion.
It has been proposed that patients with polysplenia syndrome may be predisposed to PAVF as part of the spectrum of incomplete organ development or that PAVF may be related to the biliary disease sometimes found in these patients.22 23 The results of our study suggest that younger patients with situs abnormalities may be at increased risk for developing PAVF after cavopulmonary shunt. However, heterotaxy syndromes do not explain the development of PAVF for most of our study patients. When patients with heterotaxy are excluded from analysis, the prevalence of PAVF is still 41%.
While acquired PAVF have been described in association with fungal and parasitic infestations,19 there was no evidence of these in any of the patients included in this study. PAVF also are part of the clinical spectrum found in Rendu-Osler-Weber disease24 ; however, the associated findings of widely distributed telangiectasias and aneurysms known to occur with this autosomal-dominant disorder of systemic fibrovascular dysplasia were not seen in any of the study patients.
It is tempting to speculate that diversion of normal hepatic venous flow from the pulmonary circulation plays a role in PAVF development in patients who have undergone cavopulmonary anastomosis.11 Such a hypothesis is supported by the observation that intrapulmonary shunting seen with cirrhosis in adults and biliary atresia in children may regress after liver transplantation.25 26 This might explain the lower prevalence of PAVF in patients who had undergone total caval exclusion. Possible mechanisms by which this might occur include normal hepatic metabolism of a stimulatory factor or hepatic synthesis of a suppressive factor.
The ultimate clinical significance of the PAVF that we have described is not yet clear. Although they can produce profound hypoxemia and death,11 the usual clinical picture consists of mild desaturation with slow progression. In our study group, we found that although a decrease in systemic oxygen saturation may occur with PAVF, suggesting their presence, the level of systemic arterial oxygenation is of little value in predicting their absence. Several factors may contribute to this result. The interpretation of systemic arterial oxygen saturation data frequently is clouded by the patient’s pulmonary status, the presence of intracardiac right-to-left shunting, changes in oxygen-carrying capacity, and intrinsic variability of measurements by pulse oximetry. Further, the development of PAVF may be a regional phenomenon, blunting the effect of PAVF on systemic arterial oxygen saturation.
Because the natural history and clinical significance of PAVF are not yet certain, recommendations for clinical management must be limited at this time. While we have not altered the timing of cavopulmonary shunt placement in these patients, we now include contrast echocardiography as part of our routine surveillance of patients who have had cavopulmonary anastomosis. For those who have a positive contrast study, we recommend cardiac catheterization to evaluate the extent of pulmonary venous desaturation. On the basis of our findings, we focus particular attention on infants and on patients with long-standing cavopulmonary shunts. Because cavopulmonary shunts are performed more frequently in infants, these two populations eventually will merge, raising the concern that PAVF will become a more significant problem in the future.
PAVF that develop after cavopulmonary anastomosis might be ameliorated by subsequent incorporation of hepatic venous return into the pulmonary arterial circulation (ie, by completing a total caval exclusion). This concept is supported by regression of PAVF in one of our patients after biventricular repair. We have not, however, documented PAVF regression after total caval exclusion in any of our patients. Nonetheless, our present algorithm is to perform total caval exclusion before patients are 2 years old in this population. Clearly, a prospective study of children with single-ventricle physiology, to include evaluation before cavopulmonary shunt and serial studies after surgery, is needed to define the natural history and long-term clinical significance of PAVF in these patients.
In summary, the results of this study indicate that all children who have undergone cavopulmonary shunt are at significant risk for developing PAVF and that infants as well as patients with long-standing shunts are at especially high risk for this complication. While PAVF result in pulmonary venous desaturation, various factors confound the predictive value of monitoring systemic oxygen saturation by transcutaneous pulse oximetry. Contrast echocardiography, however, provides a sensitive and reliable method for detecting PAVF.
Dr Bernstein was supported by a postdoctoral fellowship in developmental pediatric cardiology (HL-07544), and Dr Bristow was supported by a Clinical Investigator Award (HL-02695), both from the National Institutes of Health, Bethesda, Md. We would like to thank Drs Paul Stanger, David Teitel, Michael Heymann, Abraham Rudolph, Julien Hoffman, Scott Soifer, Harold Tarnoff, and Michael Cooper for allowing us to include their patients in this study; Drs Grant Burch, Alan Flint, Colin Phoon, and Darlene Horton, as well as Doug Huffines, Maria Villegas, and Ruth Petrone for providing assistance with contrast echocardiography; and Dr Julien Hoffman for advice on statistical analysis.
- Copyright © 1995 by American Heart Association
Trusler GA, Williams WG, Cohen AJ, Rabinovitch M, Moes CAF, Smallhorn JF, Coles JG, Lightfoot NE, Freedom RM. The cavopulmonary shunt: evolution of a concept. Circulation. 1990;82:131-138.
Cloutier A, Ash JM, Smallhorn JF, Williams WG, Trusler GA, Rowe RD, Rabinovitch M. Abnormal distribution of pulmonary blood flow after the Glenn shunt or Fontan procedure: risk of development of arteriovenous fistulae. Circulation. 1985;72:471-479.
Bernstein HS, Ursell PC, Hanley FC, Brook MM, Silverman NH, Bristow J. Fulminant development of pulmonary arteriovenous fistulae in an infant following total cavopulmonary shunt. Pediatr Cardiol. 1995. In press.
Fisher RA. In: ed. Statistical Methods for Research Workers. New York, NY: Hafner; 1958:96-97.
Fletcher RH, Fletcher SW, Wagner EH. In: ed. Clinical Epidemiology: The Essentials. Baltimore, Md: Williams and Wilkins; 1988:48,197.
Buckley L, Dooley K, Fyler D. Pulmonary atresia and intact ventricular septum in New England. Am J Cardiol. 1976;37:124.
Fewtrel MS, Noble-Jamieson G, Revell S, Valente J, Friend P, Johnston P, Rasmussen A, Jamieson N, Calne RY, Barnes ND. Intrapulmonary shunting in the biliary atresia/polysplenia syndrome: reversal after liver transplantation. Arch Dis Child. 1994;70:501-504.