Hepatic Venous Blood and the Development of Pulmonary Arteriovenous Malformations in Congenital Heart Disease
Background Pulmonary arteriovenous malformations (PAVMs) are a known complication after some types of cavopulmonary anastomoses (CVPAs). Their cause is unknown, but they may be related to the absence of pulsatile flow or the presence or absence of circulating factors. These PAVMs are diffuse and are presumed to be progressive and irreversible.
Methods and Results All patients with congenital heart disease (CHD) seen at Children’s Hospital, Boston, Mass, between 1970 and 1993 were reviewed. We report on the 10 patients with CHD who were found to have developed PAVMs, as diagnosed by cardiac catheterization. Diagnoses included heterotaxy syndrome/polysplenia, with interrupted inferior vena cava and hepatic veins draining to the right atrium (n=6); heterotaxy/asplenia (n=1); corrected transposition with pulmonary stenosis (n=1); and biliary atresia and associated CHD (n=2). PAVMs were diagnosed 0.1 to 7.0 years (median, 3.5 years) after creation of a CVPA that resulted in exclusion of hepatic venous flow from one or both lungs in 8 of the 10 patients; the remaining 2 patients had normal drainage of hepatic veins to the lungs but had biliary atresia. In all, the common anatomic feature was the exclusion of normal hepatic venous return from the affected pulmonary arterial circulation. All patients with interrupted inferior vena cava, azygous continuation to the superior vena cava, and hepatic veins draining to the right atrium (polysplenia syndrome) were reviewed to determine the incidence of PAVMs in those with CVPA (ie, hepatic venous flow excluded from the pulmonary arteries) and without CVPA. Six of 28 (21%) of those with versus 1 of 56 (1.8%) of those without CVPA developed PAVMs (P=.004). The 1 patient without CVPA who had PAVMs also had biliary atresia. Among patients with CVPA, the probability of developing PAVMs was 15% and 28% at 3 and 5 years, respectively, after CVPA. The histological and angiographic appearances of PAVMs after CVPA are similar to those seen in PAVMs associated with hepatic cirrhosis.
Conclusions We postulate that PAVMs after CVPA are related to the diversion of normal hepatic venous flow from the pulmonary circulation. In this sense, these PAVMs may be analogous to those associated with liver disease, which have been found to resolve after liver transplantation. Redirection of hepatic flow to the pulmonary bed in some patients with CHD and PAVMs may lead to reversibility of the PAVMs.
The development of pulmonary arteriovenous malformations (PAVMs) after the classic Glenn1 anastomosis (superior vena cava [SVC] to right pulmonary artery [PA]) is a previously described phenomenon with a reported incidence up to 25%.2 Although embolization therapy can successfully treat isolated PAVMs, effective therapy can be elusive when the disease is diffuse.3 4 The cause of the formation of these PAVMs remains unknown, but maldistribution of pulmonary blood flow to upper and lower lobes and nonpulsatile blood flow have been among the factors implicated in their pathogenesis.2 5 6 Angiographically, they appear as small, diffuse angiomatoid formations (Fig 1⇓) and are similar in angiographic appearance to those seen in cirrhotic liver disease.7 Rapid transit of contrast media is seen from pulmonary arterial-to-venous systems and is associated with pulmonary venous desaturation. Histologically, PAVMs associated with liver disease are characterized by diffuse dilation of precapillary and capillary vessels.7 No histological descriptions of PAVMs after cavopulmonary anastomosis (CVPA) were found in our review of the literature. The natural history of this condition is not well described, but our experience is that it is a progressive and irreversible condition leading to increasing cyanosis.
In contrast to the experience with classic Glenn anastomoses, the reported incidence of PAVMs after other types of cavopulmonary anastomoses has been quite low. In the 1970s, the Fontan operation (atriopulmonary anastomosis)8 and later the modified Fontan operation,9 which incorporates almost all systemic venous return to the pulmonary arterial bed by the technique of total cavopulmonary anastomoses, became the palliative procedure of choice for most patients who had single ventricles and met the necessary anatomic and hemodynamic criteria. The distribution of pulmonary blood flow and the lack of pulsatile blood flow after the Fontan operation are similar to those noted after the Glenn anastomosis.10 However, PAVMs have not been reported after the Fontan operation and were absent in a series of Fontan patients examined specifically for the presence of PAVMs.2 The incidence of PAVMs after bidirectional cavopulmonary Glenn anastomoses (BDGs) has not been reported but also seems to be extremely low. However, this finding may be confounded by the short natural history of patients with a BDG, since the majority of such patients proceed to Fontan operations within 6 to 18 months after the creation of a BDG.
One exception to this sequence of BDG and early Fontan is patients with the polysplenia form of heterotaxy syndrome who have an interrupted inferior vena cava (IVC). In this form of heart disease, the IVC is interrupted, with only the hepatic veins entering the lower portion of the right atrium (RA). The remainder of the lower-body systemic venous return enters the SVC via the azygous system. Thus, in these cases, an SVC-to-PA anastomosis (BDG) directs all venous return to the pulmonary bed except for coronary sinus and hepatic venous blood. Two patients reported to have developed PAVMs after modified Fontan operations had similar anatomies and had BDG anastomoses that included all venous return except hepatic and coronary sinus blood.11
In this report, we describe all 10 patients with congenital heart disease (CHD) at our hospital who developed PAVMs as diagnosed by cardiac catheterization. A majority of these patients were diagnosed with a heterotaxy syndrome with interrupted IVC. This experience led to a review of all our patients with interrupted IVC, which is also included in this report. In all patients who developed PAVMs, the common anatomic feature was the exclusion of normal hepatic venous flow from the affected pulmonary vasculature. The angiographic and histological appearance of these PAVMs is similar to that seen in PAVMs associated with liver disease. We postulate that the occurrence of PAVMs in patients with CHD is related to the diversion of normal hepatic venous flow away from the pulmonary circulation. In this sense, PAVMs in CHD may be analogous to those associated with hepatic cirrhosis and may thus share a common pathogenesis.
From the computerized cardiology data base at Children’s Hospital, Boston, Mass, all patients diagnosed with both PAVMs and CHD between 1970 and 1993 were identified. Additionally, all patients with interrupted IVC anatomy were identified. Medical charts and echocardiographic and catheterization data were reviewed retrospectively, with close attention paid to the possible diagnoses of PAVMs and the patterns of hepatic and pulmonary blood flow. Follow-up data were obtained from charts and correspondence with the patients’ referring physicians. All patients diagnosed with PAVMs were included in this study.
The diagnosis of PAVM was made when there was (1) resting pulmonary venous desaturation without evidence of parenchymal lung disease on chest roentgenogram and/or (2) demonstration of PAVMs by selective pulmonary angiography. Pulmonary arteriograms of patients presumed to have PAVMs and those without were interpreted in a blinded fashion by two independent observers (V.M. and J.F.K.).
Interpretation of Pulmonary Arteriograms
To confirm previous diagnoses of PAVMs, pulmonary arteriograms of patients with and without previous diagnoses of PAVMs were interpreted in a blinded fashion by two independent observers. Rapid arterial-to-venous transit time and a reticular appearance of the lung parenchyma were used in establishing the diagnosis (see Fig 1⇑). All patients with previous diagnoses of PAVMs were again thought to have PAVMs by both observers. Locations of PAVMs are given in Table 1⇓. PAVMs were seen in both the upper and lower lobes of lungs. The classic reticular pattern was noted only in the patients previously thought to have PAVMs. This pattern may represent concomitant filling of arteries and veins. The transit time from arterial to venous circulation was assessed subjectively and was noted to be more rapid in all patients with PAVMs. In addition, the lack of the capillary phase disappearance of dye was noted only in patients thought to have PAVMs.
Autopsy slides were available in two patients who had developed PAVMs after CVPA and were reviewed with pathologists at our institution. Evidence of abnormal pulmonary parenchymal and vascular changes was noted compared with normal control subjects as previously identified.12
Actuarial survival curves were generated for two groups of patients with interrupted IVCs: those with CVPA and those without. A log-rank test13 compared the probability of freedom from PAVMs in patients with and without CVPA. In addition, Fisher’s exact test13 was used to compare the proportion of patients who developed PAVMs in the two groups. Values of P<.05 were considered statistically significant for both tests.
Ten patients with CHD were found to have developed PAVMs over the 23-year period reviewed. Clinical characteristics of these patients, including anatomic arrangement, are shown in Table 1⇑. Six patients had heterotaxy syndrome with interrupted IVC, azygous continuation to the SVC, and hepatic veins draining to the RA. Three of these patients (patients 1 through 3) underwent anastomoses of the cranial end of the SVC to the PA, ie, a BDG, as their definitive palliative procedure, and they developed bilateral diffuse PAVMs between 3 and 4 years after surgery. One patient (patient 4) had a left SVC anastomosed to the left PA and the RA anastomosed to the right PA but had preferential blood flow from the RA to the right PA, as demonstrated by angiography. This patient developed PAVMs in the left lung only, whereas the right lung (which continued to receive hepatic venous blood) remained unaffected. Patient 5 had an interrupted IVC draining to a left SVC, and a left classic Glenn procedure was performed because of pulmonary vascular obstructive disease in the right lung after a Waterston shunt. Hepatic veins drained to the RA, which then emptied into the systemic arterial circulation. PAVMs developed in the left lung 4 years after the Glenn anastomosis, whereas the right lung was without PAVMs. Patient 6 also had similar anatomy but underwent a procedure that incorporated the hepatic veins into the pulmonary circulation as well. Secondary to postoperative complications, the patient was catheterized 1 month after surgery, and it was found by contrast distribution that most, if not all, of the hepatic venous flow was streaming to the right lung. PAVMs were noted in the left lung only. The seventh patient had a bidirectional Glenn anastomosis, and because of aortic oxygen saturations in the 85% to 90% range, a Fontan procedure was postponed. The IVC and hepatic veins continued to drain to the RA. PAVMs were found bilaterally 3.5 years after surgery. Patient 8 had a right classic Glenn shunt and developed many small PAVMs in the right lung 4 years after surgery. The final 2 patients had complex CHD but did not undergo any type of CVPA. These patients both had biliary atresia, and both developed bilateral PAVMs that were angiographically similar to those seen in patients with CVPAs. These developed despite normal anatomic and pulsatile flow of hepatic venous blood to the lungs.
In summary, 4 patients developed bilateral PAVMs after BDG; 2 patients developed unilateral PAVMs after CVPAs that directed hepatic venous flow to the contralateral lung; 2 patients were found to have unilateral PAVMs after classic Glenn procedures to the ipsilateral lung; and 2 patients had PAVMs associated with biliary atresia. As seen in Table 1⇑, the median time of diagnosis after surgery was 3.5 years. Of note, 8 of 10 patients had the diagnosis made more than 2 years after hepatic vein “diversion.” The majority of patients were diagnosed relatively recently, and follow-up has been brief, the median follow-up being 0.7 years, with a range of 0.1 to 7.5 years.
Hemodynamic and Laboratory Data
Hemodynamic and laboratory data at the time of diagnosis are shown in Table 2⇓. The median aortic saturation was 77% (range, 50% to 90%), the median hemoglobin was 17.3 mg/dL (range, 9.0 to 18.9 mg/dL), and the median mean PA pressure was 14 mm Hg (range, 10 to 20 mm Hg). Of note, 2 patients were found to have PAVMs despite the presence of pulsatile flow in the PAs. Pulmonary venous saturations were measured in 7 of 10 patients (Table 2⇓). Data on hepatic function were available in 8 patients. Bilirubin levels ranged from 0.8 to 3.2 mg/dL, with a mean of 1.6 mg/dL in the patients without biliary atresia. The 2 patients with biliary atresia had significant elevation of serum bilirubin.
Histological sections of the lungs were available for study in 2 patients with PAVMs after CVPA, both deceased. The lungs in patient 1 had a PA barium sulfate–gelatin injection under nonstandard conditions, precluding vascular morphometric analysis. Pulmonary arteries and veins were patent and appeared to be of adequate caliber. The lungs in patient 6 were examined after routine formalin fixation without injection of the PA tree. The external diameter of respiratory and terminal bronchiolar arteries and percentage medial wall thickness of PAs 200 to 300 μm in diameter and septal veins 100 to 200 μm in diameter were determined according to the method of Haworth and Hislop.12 The respiratory bronchiolar arteries were dilated (95 μm observed, 64 μm expected), as were the terminal bronchiolar arteries (165 μm observed, 120 μm expected) (Fig 2⇓, top). The mean percentage arterial wall medial thickness of arteries 200 to 300 μm in diameter was 5.9% and essentially similar to the control value of 5.8%.12 The mean percentage wall thickness of veins 100 to 200 μm in diameter was increased, with a value of 6.6% compared with the control value of 4.3±1.4%.12 Some areas of pleura contained clusters of predominantly thin-walled vessels of indeterminate origin showing a thin elastic lamina and thin smooth-muscle-cell collar (Fig 2⇓, bottom). Some ancient recanalized PA thromboemboli were present in the left lung.
Incidence of PAVM in Polysplenia
Given that 6 patients had the common anatomy of interrupted IVC with azygous continuation to the SVC and that a genetic abnormality associated with heterotaxy syndrome has recently been identified,14 we considered whether there might be an association between heterotaxy syndrome and PAVMs. Therefore, we reviewed all patients with interrupted IVC to determine whether this anatomy alone promoted the development of PAVMs.
A total of 84 patients were identified who had interrupted IVC anatomy. Of the 84 patients, 56 (67%) did not have any type of CVPA; only 1 of 56 (1.8%) of those without CVPA developed PAVMs. The 1 patient without CVPA who developed PAVMs also had biliary atresia. The remaining 28 patients had CVPAs that excluded hepatic venous blood from one or both lungs; 6 of 28 (21%) of these patients were found to have PAVMs (P=.004) (Fig 3⇓). Long-term clinical follow-up was available for 55 of the 56 patients without CVPA (median, 8 years) and for 23 of the 28 with CVPA (median, 4 years). The incidence of PAVMs in patients with long-term follow-up is 1 of 55 (1.8%) in those without CVPA and 6 of 23 (26%) in those with CVPA (P=.004). Actuarial analysis of these patients (Fig 4⇑, top) shows a 15% and 28% probability of developing PAVMs 3 and 5 years, respectively, after CVPA. Fig 4⇑, bottom, compares the development of PAVMs in interrupted IVC patients with and without CVPA and shows a significant increase in probability for patients with CVPAs (P=.002). Thus, in our series, patients with interrupted IVC alone did not develop PAVMs but did have a higher probability of PAVMs after exclusion of hepatic venous flow to the lungs.
We have described a series of 10 patients with CHD who developed PAVMs and whose clinical description led to a common denominator: the exclusion of normal hepatic venous blood flow from the affected pulmonary circulation. Bilateral PAVMs were seen when both lungs were deprived of normal hepatic venous flow. In all cases in which one lung received hepatic venous blood and the remaining lung did not, only the latter developed PAVMs. The median duration from the time of CVPA to diagnosis of PAVM was 3.5 years. Two patients did show angiographic evidence of PAVMs less than 2 years after surgery, one with biliary atresia and a second at an early postoperative catheterization carried out because of an abnormal postoperative course. This patient died of a fatal arrhythmia soon after the catheterization.
Although the development of PAVMs was noted after the classic Glenn procedure, we found no reported cases of PAVMs after the Fontan operation, which directs hepatic and IVC blood, in addition to SVC blood, to the lungs. The majority of the patients in our series who developed PAVMs had a unique anatomic arrangement: heterotaxy/polysplenia syndrome having an interrupted IVC with azygous continuation to a SVC but hepatic veins still draining to the RA. Two patients with similar anatomy have been reported11 who developed diffuse, bilateral PAVMs after a “modified Fontan procedure,” known as the Kawashima operation.15 The Kawashima procedure involves an anastomosis of the cranial end of the SVC to the PA (similar to a bidirectional Glenn procedure) and diverts all lower-body venous return to the PAs except hepatic venous return. These patients frequently have arterial oxygen saturations in the 85% to 90% range and have previously been thought to require no further intervention. In contrast, the long-term outcome after the usual type of BDG has been difficult to determine, since most patients with BDGs undergo subsequent completion to a Fontan type of circulation. In most patients who develop PAVMs after CVPA, detection is not for several years.
Although many patients with PAVMs after Glenn procedures have been reported, we have not found any histological descriptions of the pulmonary parenchyma in this condition. The pulmonary parenchymal changes associated with PAVMs in one of the patients in our series who had CVPA consisted of diffuse dilation of precapillary vessels. These resemble the “PAVMs” seen in patients with liver disease.7
A recent report of a patient with polysplenia who did not have a CVPA but did develop bilateral PAVMs raises the possibility of an inherent predisposition to PAVMs in patients with polysplenia.16 This patient had hypoplasia of the intrahepatic portal vein branches and a portal-to-systemic shunt; thus, mesenteric venous return bypassed the liver and went directly to the heart and lungs without metabolic reductions or additions. We nonetheless reviewed all our patients with polysplenia who did not have CVPA to determine whether polysplenia alone contributed to the development of PAVMs. However, only 1 of 56 patients with polysplenia syndrome but without CVPA developed PAVMs, and that patient had biliary atresia.
Although the first reported case of PAVMs was in 1897,17 the pathogenesis of PAVMs remains unclear. Hemodynamic alterations have been suggested, but in general, most intracardiac, PA (mean), and pulmonary capillary wedge pressures are normal.18 Normal hemodynamic findings were present in our series of patients as well. Maldistribution of pulmonary blood flow has been suggested as a possible cause of PAVMs.2 It has been shown that passive blood return to the lungs after a Glenn anastomosis results in flow to the lower lobes increased over normal compared with the upper lobes.2 Although this maldistribution is present in some patients after the Fontan operation,2 PAVMs have not been a complication of Fontan operations. Absence of pulsatile pulmonary blood flow has also been implicated in the development of PAVMs; however, our two patients with biliary atresia developed bilateral PAVMs in the presence of pulsatile blood flow.
These data strongly suggest that normal hepatic venous blood may play a role in the prevention of PAVMs. Patients with hepatic cirrhosis develop vascular dilations7 similar to those seen in our patients at the pulmonary precapillary and capillary levels,19 which are thought to cause diffusion-perfusion defects resulting in cyanosis. Pulmonary arteriograms in patients with liver disease and PAVMs are similar to the arteriograms in our patients. Although abnormal vasoactive agents have been found in hepatic venous blood in patients with cirrhosis,20 21 most of our patients had normal livers, thus implicating the absence of a normal factor rather than the presence of an abnormal one. Given the dilation of vessels seen histologically, the putative hepatic product may be involved in vasomotor control rather than angiogenic control. Since only the minority of patients with hepatic vein exclusion develop PAVMs, any causal relation between the former and latter must be complex, although it is possible that echocardiographic or radionuclide studies would reveal an even higher incidence of PAVMs. In a recent report, two patients with biliary atresia who had developed diffuse PAVMs underwent orthotopic liver transplantation and within 3 months were found to have resolution of intrapulmonary shunting,22 suggesting that return of normal hepatic blood flow to the lungs may reverse these lesions.
These data seem to permit several inferences. Angiographically, PAVMs should be suspected when pulmonary venous return is evident more rapidly than normal after contrast injection into the PAs. Patients with interrupted IVCs, and possibly others, who have had BDG-type operations diverting hepatic venous blood away from the lungs should be followed closely for the development of PAVMs. Patients who undergo BDG operations as a long-term palliation may benefit from adjunctive procedures that provide some source of normal hepatic blood to the lungs. The recent evidence of reversibility of PAVMs after liver transplantation in patients with cirrhosis raises the possibility that redirection of hepatic venous flow to the pulmonary bed in some patients with CHD and PAVMs may reverse the arteriovenous malformations.
Deepak Srivastava, MD, is a fellow of the National Institutes of Health Pediatric Scientist Development Program. The authors would like to thank Lynne Reid, MD, Department of Pathology, Children’s Hospital, for her review and consultation concerning the histological evaluation.
- Received September 22, 1994.
- Revision received February 15, 1995.
- Accepted February 25, 1995.
- Copyright © 1995 by American Heart Association
Glenn WW. Circulatory bypass of the right side of the heart, IV: shunt between SVC and distal RPA: report of clinical application. N Engl J Med. 1958;259:117-120.
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.
Samanek M, Oppelt A, Kasalicky J, Voreskova M. Distribution of pulmonary blood flow after cavopulmonary anastomosis (Glenn operation). Br Heart J. 1969;31:511-516.
Fontan F, Baudet E. Surgical repair of tricuspid atresia. Thorax. 1971;26:240-248.
DiSessa TG, Child JS, Perloff JK, Wu L, Williams RG, Laks H, Friedman WF. Systemic venous pulmonary arterial flow pattern after Fontan’s procedure for tricuspid atresia or single ventricle. Circulation. 1984;70:898-902.
Dawson-Saunders B, Trapp RG. Basic and Clinical Biostatistics. Norwalk, Conn: Appleton & Lange; 1990.
Seo JW, Brown NA, Ho SY, Anderson RH. Abnormal laterality and congenital cardiac anomalies. Circulation. 1992;86:642-650.
Churton T. Multiple aneurysm of pulmonary artery. Br Med J. 1897;1:1223.
Burke CM, Safai C, Nelson DP, Raffin TA. Pulmonary arteriovenous malformations: a critical update. Annu Rev Respir Dis. 1986;134:334-339.
Berthelot P, Walker JG, Sherlock S, Reid L. Arterial changes in the lungs in cirrhosis of the liver: lung spider nevi. N Engl J Med. 1966;274:291-298.
Snell AM. The effects of chronic disease of the liver on the composition and physiochemical properties of blood: changes in serum proteins: reduction in the oxygen saturation of the arterial blood. Ann Intern Med. 1935;9:690-711.
Laberge JM, Brandt ML, Lebecque P, Moulin D, Veykemans F, Paradis K, Pelletier L, La Croix J, Blanchard H, de Ville de Goyet J, Weber A, Vliers A, Ott J. Reversal of cirrhosis-related pulmonary shunting in two children by orthotopic liver transplantation. Transplantation. 1992;53:1135-1138.