Hepatopulmonary syndrome: Update on recent advances in pathophysiology, investigation, and treatment


  • Conflicts of interest: No conflicts of interest exist for either of the authors of this manuscript.


Dr Josephine A Grace, Department of Medicine, Austin Health, University of Melbourne, Studley Road, Heidelberg, Vic. 3084, Australia. Email: j.grace4@pgrad.unimelb.edu.au


Hepatopulmonary syndrome (HPS) is an important cause of dyspnea and hypoxia in the setting of liver disease, occurring in 10–30% of patients with cirrhosis. It is due to vasodilation and angiogenesis in the pulmonary vascular bed, which leads to ventilation-perfusion mismatching, diffusion limitation to oxygen exchange, and arteriovenous shunting. There is evidence, primarily from animal studies, that vasodilation is mediated by a number of endogenous vasoactive molecules, including endothelin-1 and nitric oxide (NO). In experimental HPS, liver injury stimulates release of endothelin-1 and results in increased expression of ETB receptors on pulmonary endothelial cells, leading to upregulation of endothelial NO synthase (eNOS) and subsequent increased production of NO, which causes vasodilation. In addition, increased phagocytosis of bacterial endotoxin in the lung not only promotes stimulation of inducible NO synthase, which increases NO production, but also contributes to intrapulmonary accumulation of monocytes, which may stimulate angiogenesis via vascular endothelial growth factor pathway. Despite these insights into the pathogenesis of experimental HPS, there is no established medical therapy, and liver transplantation remains the main treatment for symptomatic HPS, although selected patients may benefit from other surgical or radiological interventions. In this review, we focus on recent advances in our understanding of the pathophysiology of HPS, and discuss current approaches to the investigation and treatment of this condition.


Two intriguing and incompletely understood disorders of the pulmonary vasculature can cause pulmonary dysfunction in cirrhotic patients. The more common is hepatopulmonary syndrome (HPS), in which the primary pathological process is abnormal pulmonary vasodilation. This condition represents one manifestation of generalized circulatory dysfunction in portal hypertension, which is characterized by vascular dilatation and development of a hyperdynamic circulation. The other, but far less common, pulmonary vascular disorder associated with cirrhosis is portopulmonary hypertension. Here, the pulmonary circulatory abnormality is vasoconstriction, and there is fibro-obliteration of the vascular bed, the opposite from the changes that occur in HPS. Rarely, patients can have features of both disorders.[1]

HPS is defined as the presence of the triad of an arterial oxygenation defect, intrapulmonary vasodilation, and the presence of liver disease.[2] It is usually diagnosed in patients with cirrhosis, but neither cirrhosis nor portal hypertension is a prerequisite for the diagnosis, as it has been reported in chronic non-cirrhotic hepatitis,[3] non-cirrhotic portal hypertension,[4, 5] Budd–Chiari syndrome,[6] and even in acute liver diseases, such as fulminant hepatitis A[7] and ischemic hepatitis.[8] Estimates of the prevalence of HPS are complicated by a lack of consensus in the past regarding the diagnostic criteria. In particular, the degree of gas exchange abnormality required to make the diagnosis is variable, so that even within the same group of cirrhotic patients in one study, the apparent prevalence varied from 19% to 32%.[9] Most studies have been conducted in patients with advanced liver disease undergoing assessment for liver transplantation, in whom the prevalence ranges from 16% to 33%.[10-14] Limited data suggest that a slightly lower prevalence of 10–17% exists in the overall cirrhotic population.[15, 16]

Thus, HPS represents a relatively common and important cause of pulmonary disease in patients with cirrhosis. This review will focus on recent advances in our understanding of the pathophysiology of HPS, and discuss appropriate investigation, prognosis, and treatment of patients with HPS.


The pathological findings in HPS were first described by Berthelot in 1966, who documented widespread dilatation of pulmonary microvessels encompassing the pulmonary precapillary and alveolar capillary beds.[17] This intra-pulmonary vasodilation is responsible for the three physiological mechanisms that contribute to impaired gas exchange in HPS: ventilation-perfusion mismatch, diffusion limitation, and shunting (Fig. 1). Ventilation-perfusion mismatching occurs due to overperfusion of the alveolar capillary bed, particularly in the less well-ventilated dependent lower zones, and is exacerbated by a blunted vasoconstrictor response to hypoxia.[18] Dilatation of pulmonary microvessels at the gas exchange interface increases the distance that oxygen must travel from the alveolus to equilibrate with red cells in the center of the alveolar capillary, creating a functional diffusional barrier to oxygen exchange. This diffusional barrier is exacerbated by rapid blood transit due to the hyperdynamic circulation in these patients, as alveolar–capillary equilibration for oxygen is influenced by blood transit time.[19] Patients may also have true anatomical shunting in the form of direct arteriovenous communications, which allow blood to completely bypass alveoli, resulting in mixed venous blood passing into the pulmonary veins.

Figure 1.

The pathophysiology of impaired gas exchange in hepatopulmonary syndrome. (a) Represents the healthy gas exchange unit, where oxygen diffuses from a normal alveolus into a pulmonary capillary. (b) Represents the gas exchange unit in hepatopulmonary syndrome, where dilation of the intrapulmonary capillary vessel results in diffusion limitation of oxygen from the normal alveolus, resulting in incomplete oxygenation of blood. This is exacerbated by accelerated blood flow and the presence of shunting due to direct arteriovenous communications.

The mechanisms responsible for the vascular changes in HPS remain incompletely understood; however, there are some important clinical clues. One key observation is that although the majority of cases occur in patients with cirrhosis, impaired hepatic synthetic function, and portal hypertension, it has also been reported in their absence, for example in chronic viral hepatitis without portal hypertension[3] and in non-cirrhotic portal hypertension.[4] This indicates that neither liver synthetic dysfunction nor portal hypertension is necessary for the development of the syndrome.

Another important observation comes from the field of pediatric cardiac surgery. Children who have undergone superior cavopulmonary shunt surgery, predominantly for polysplenia associated with an interrupted inferior vena cava, often develop hypoxia as adults due to the development of pulmonary arteriovenous malformations (AVMs). These AVMs cause intrapulmonary shunting, and share characteristics with some of the vascular abnormalities found in HPS, being formed by the opening up of preexisting vascular channels. A review of these patients found that the common anatomical feature was the exclusion of hepatic venous blood from the affected pulmonary circulation.[20] Furthermore, a number of case studies have demonstrated that revision surgery to rechannel hepatic venous blood into the pulmonary vasculature improved oxygenation, and in some cases led to resolution of pulmonary AVMs.[21] These findings support the theory that factors either produced by, or modified in, the liver are essential to regulate vessel tone in the pulmonary circulation.

Proposed mediators of pulmonary vasodilation in HPS

The majority of research into the molecular basis of HPS has been performed in rats that have undergone surgical bile duct ligation (BDL). These animals develop cirrhosis, portal hypertension, and HPS at 4–5 weeks after surgery. Most other animal models of cirrhosis, such as that induced by carbon tetrachloride, have not proved useful since they do not cause HPS.[22] Research into the underlying pathophysiological mechanisms has mainly focused on the roles of nitric oxide (NO), carbon monoxide (CO), endothelin-1 (ET-1), and tumor necrosis factor-α (TNF-α), and is summarized below.


NO plays a central role in the pathophysiology of systemic and splanchnic vasodilation in cirrhosis. NO is synthesized from L-arginine by the action of NO synthase (NOS), which exists in three isoforms—inducible NOS (iNOS), endothelial NOS (eNOS), and neuronal NOS. When released from endothelial cells, NO causes vasodilation by diffusing into vascular smooth muscle, where it stimulates soluble guanylyl cyclase, which in turn activates the generation of cyclic guanosine monophosphate (cGMP). cGMP activates cGMP-dependent kinase, which leads to activation of myosin light chain phosphatase, and thus vasodilation.

There are several lines of evidence that suggest NO plays a role in the pathogenesis of HPS. Levels of exhaled NO are elevated in cirrhotic patients with HPS compared with cirrhotic controls, and these levels correlate with PA-aO2.[23] Both iNOS and eNOS are upregulated in the lung in cirrhotic animals with HPS. Furthermore, inhibition of NOS with NG-nitro-L-arginine methyl ester improves oxygenation in animals with cirrhosis.[24, 25]

iNOS, endotoxemia, and TNF-α

iNOS is mainly found in smooth muscle cells of the systemic circulation, and probably does not contribute significantly to systemic vasodilation in cirrhosis.[26] However, a different situation exists in experimental HPS, where iNOS is localized to intravascular macrophages in the lung.[24] These macrophages are stimulated by endotoxemia to produce pro-inflammatory cytokines, including TNF-α, which triggers upregulation of iNOS. The role of TNF-α in HPS is highlighted by the observation that monoclonal TNF-α reduces intrapulmonary shunting and improves oxygenation,[27] while pentoxifylline, a non-specific phosphodiesterase inhibitor that blocks TNF-α synthesis, has been shown to prevent the development of HPS in BDL animals.[28]

It is, therefore, proposed that lung endotoxemia due to bacterial translocation from the gut is responsible for increased levels of TNF-α and upregulation of lung iNOS in cirrhosis.[24] Bacterial translocation is common in cirrhosis, affecting up to 70% of cirrhotic animals,[29] and 30% of patients with Child-Pugh C cirrhosis.[30] There is increased pulmonary intravascular phagocytosis in experimental cirrhosis,[24] with fivefold uptake of lipopolysaccharide in cirrhosis compared with control.[31] These results suggest that pulmonary endotoxemia is a central step in the evolution of HPS. Encouragingly, intestinal decontamination with norfloxacin normalizes iNOS expression and improves HPS in experimental cirrhosis.[29]

eNOS and ET-1

ET-1 is released by endothelial cells, and can cause both vasoconstriction and vasodilation. Both ETA and ETB receptors on vascular smooth muscle cells mediate vasoconstriction,[32] but activation of ETB receptors on endothelial cells[33] causes NO-mediated vasodilation.[32] Activation of endothelial ETB receptors in the pulmonary circulation is likely to contribute to the pathogenesis of HPS. Plasma ET-1 levels are increased in cirrhosis and are higher in patients with intrapulmonary vasodilation.[34-36] Following BDL, and prior to the development of cirrhosis, hepatic stellate cells and cholangiocytes become important sources of ET-1.[37, 38] Furthermore, ET-1 infusion into the peripheral circulation causes vasoconstriction in healthy subjects, but vasodilation in patients with advanced cirrhosis.[39] ETB receptor expression in the pulmonary vessels, but not in the systemic circulation, is increased in cirrhosis and in portal hypertension.[40] Thus, ET-1 produced by an injured liver causes activation of pulmonary ETB receptors, resulting in NO-mediated vasodilation via upregulation of pulmonary eNOS.[40, 41] In agreement with this finding, selective ETB receptor blockade or knockout inhibits pulmonary eNOS activation and improves HPS in BDL rats.[42, 43] However, there have been no clinical trials of ETB blockade in human HPS.

Carbon monoxide

CO mediates vasodilation in a similar way to NO by stimulating cGMP production in vascular smooth muscle cells. Levels of arterial carboxyhemoglobin are raised in patients with HPS,[44] suggesting that CO may contribute toward vasodilation in these patients. CO is primarily produced from degradation of heme by heme oxygenase (HO), an enzyme that exists in inducible (HO-1) and constitutive (HO-2) forms. In healthy, HO-1 is found in low levels in the lung, in both pulmonary endothelial cells and intravascular macrophages, and in experimental cirrhosis pulmonary HO-1 expression is increased.[45] Furthermore, HO inhibition can improve gas exchange and intrapulmonary vasodilation[46] in experimental HPS.

Pulmonary angiogenesis

There is evidence that pulmonary dilatation is not the only mechanism causing impaired gas exchange in HPS. Both splanchnic and pulmonary angiogenesis have been documented in experimental cirrhosis and portal hypertension.[47, 48] Hypoxia and diffusing capacity do not improve in a proportion of patients after liver transplantation, and this may be attributable to the presence of pulmonary capillary proliferation, which has been documented in post-mortem studies of patients with HPS.[17, 49, 50] Several recent studies have suggested that pulmonary angiogenesis in experimental HPS may result from accumulation of pulmonary intravascular monocytes, leading to the activation of vascular endothelial growth factor-dependent signaling pathways, as inhibition of this pathways improved gas exchange.[48, 51, 52]

The cause of this monocyte accumulation may be increased TNF-α signaling due to bacterial translocation and/or altered chemokine expression.[51, 52] Although there is no direct evidence that these abnormalities extend to the clinical setting, a recent study of patients with HPS identified upregulation of several genes involved in the control of angiogenesis, supporting the concept that patients with increased genetic risk of disordered angiogenesis might be more susceptible to developing HPS.[53]

Summary of pathophysiology of HPS

In summary, pulmonary vasodilation in experimental HPS is mediated by a number of endogenous vasoactive molecules, including ET-1 and NO (Fig. 2). Liver injury stimulates release of ET-1, which increases expression of ETB receptors in pulmonary endothelial cells. Activation of these receptors results in the upregulation of eNOS and subsequent increased production of NO, which diffuses into vascular smooth muscle, causing vasodilation. In addition, increased phagocytosis of bacterial endotoxin in the lung promotes activation of iNOS, which also contributes toward increased NO production. Bacterial translocation and subsequent monocyte accumulation may also stimulate pulmonary angiogenesis in HPS, which may be partly controlled by genetic factors. However, there remains a need for more human experimental data to support the development of new therapies targeting these proposed mechanisms.

Figure 2.

Possible mediators of pulmonary vasodilation in hepatopulmonary syndrome (HPS). Nitric oxide (NO) appears to be a key mediator of pulmonary vasodilation in hepatopulmonary syndrome. Pulmonary endothelin-B receptors are increased in the pulmonary circulation in experimental HPS, at least partly due to shear stress secondary to portal hypertension. Circulating endothelin-1 (ET-1) produced by the injured liver binds to endothelin-B receptors, stimulating endothelial NO synthase (eNOS). Phagocytosis of bacterial endotoxin by pulmonary intravascular macrophages causes release of tumor necrosis factor-α (TNF-α), which activates inducible NO synthase (iNOS), another source of NO. In addition, carbon monoxide (CO) may also play a role in producing pulmonary vasodilation, as heme oxygenase-1 (HO-1) is upregulated in pulmonary intravascular macrophages.

Investigation of the dyspneic cirrhotic patient

The presence of HPS should be considered in all patients with liver disease who complain of dyspnea, which is common in cirrhosis, but which is present in 50% of patients with HPS.[13] A more specific symptom is platypnea (dyspnea that increases from the supine to the erect position), which may be associated with orthodoxia (hypoxia that is worse when erect). Finger clubbing is very common in HPS. In one study, it was found in almost 50% of HPS patients compared with 2% in those without the disorder.[54] This striking difference implies that one should always suspect HPS in patients with chronic liver disease and clubbing. Patients with severe HPS may be sufficiently hypoxic to appear cyanosed at rest, and the rare finding of cyanosis and clubbing in a cirrhotic patient is highly suggestive of the presence of severe HPS.[54] Although some studies show that spider nevi are often seen in HPS, there is no major difference in their prevalence in cirrhotics with HPS compared with control cirrhotic patients with similar liver disease.[13]

The diagnosis of HPS depends on establishing that impaired gas exchange in a patient with liver disease is due to pulmonary vascular dilatation. In most cases, the results of arterial blood gases and a study to detect intrapulmonary shunting (see later) are sufficiently specific to do this once other intrinsic cardiorespiratory diseases are excluded.

Determination of hypoxemia

Pulse oximetry can be a useful monitoring tool in the outpatient setting, and has been proposed as a screening tool for HPS in the cirrhotic population, with a cut-off value of ≤ 97% providing a high sensitivity and moderate specificity for an arterial oxygen tension (PaO2) ≤ 70 mmHg, but is less sensitive in mild HPS.[55, 56] However, in order to confirm the diagnosis, arterial blood gas estimation should be undertaken with the patient in a sitting position, breathing room air. The degree of gas exchange abnormality that is required for the diagnosis of HPS remains controversial. The most sensitive marker is an increase in the alveolar–arterial oxygen gradient (PA-aO2). Recommended cut-off values for the diagnosis of HPS are PaO2 ≤ 80 mmHg or PA-aO2 ≥ 15 mmHg. To avoid a complex calculation to correct for the increase in PA-aO2 that occurs with age, cut-off values of PaO2 ≤ 70 mmHg or PA-aO2 ≥ 20 mmHg are suggested in patients older than 64 years[2] (Table 1).

Table 1. Diagnosis of hepatopulmonary syndrome
  1. aEither with positive transthoracic contrast echocardiogram or macroaggregated albumin lung perfusion study.
  2. Adapted from Rodriguez-Roisin et al.[2]
Diagnostic triad of hepatopulmonary syndrome
Liver disease
Alveolar–arterial oxygen gradient ≥ 15 mmHg (or ≥ 20 mmHg in patients over 64 years)
Demonstration of intrapulmonary shuntinga

Demonstration of intrapulmonary vasodilation

Two methods of defining intrapulmonary dilatation are available: contrast echocardiography, most often using microbubbles as the contrast, and radioactive lung perfusion scan using macroaggregated albumin (MAA). In a direct comparison between the two, bubble contrast echo was found to be more sensitive,[15] and is more readily available, less invasive, and can exclude intracardiac shunting. It has, therefore, become the screening tool of choice. However, it should be noted that a proportion of cirrhotic patients have intrapulmonary vasodilation detected at echocardiography without gas exchange abnormalities, and in general these patients do not appear to develop hypoxemia over time.[15, 56]

In bubble contrast echocardiography, a sample of liquid (normally saline) is vigorously shaken to produce microbubbles, and then injected into an arm vein while the cardiac chambers are visualized via a transthoracic approach. Normally, these bubbles, which are > 25 μm in diameter, are trapped in the alveolar capillary bed, where the vessels have a diameter of 5–8 μm. Therefore, their appearance in the left atrium after intravenous injection suggests that pulmonary vasodilation has allowed them to traverse the capillary bed, reaching the left side of the heart. A positive study can of course also occur due to the passage of bubbles through a cardiac defect, but in this case the bubbles appear in the left atrium much sooner (within three cycles) after their first appearance in the right atrium. In practice, an intracardiac shunt cannot be definitively excluded in a small proportion of patients with positive studies, and this may require further cardiac investigations.

MAA perfusion lung scan is performed by peripheral venous injection of MAA particles that have been radio-labeled with technetium-99 m, followed by whole body scanning to estimate the extrapulmonary shunt fraction. These radio-labeled particles have a diameter of 10–90 μm and are removed in the normal pulmonary circulation. Thus, the detection of a significant amount of radiation in the brain or kidneys suggests intrapulmonary vasodilation or intracardiac shunting. MAA scanning appears to be highly specific but less sensitive than bubble contrast echo for detecting intrapulmonary dilatation consistent with HPS, and may fail to detect the presence of intrapulmonary vasodilation in the absence of hypoxia.[15] However, its high specificity makes it useful in diagnosing HPS in patients with coexisting lung disease,[15] and it has the advantage of being quantitative.

Other investigations

Chest X-ray may be normal or may show increased vascular markings in the lower zones. High resolution computerized tomography can be helpful in selected patients to exclude intrinsic lung disease, but the absence of vascular abnormalities does not preclude the diagnosis of HPS. A reduced carbon monoxide diffusing capacity is frequently seen in cirrhotic patients and is almost universal in HPS,[10, 12] possibly reflecting diffusion limitation at the alveolus. In the absence of intrinsic lung disease, other pulmonary function tests are normal. Pulmonary angiography can be normal in HPS and is rarely required. It is, however, useful in patients in whom a large arteriovenous shunt is suspected, for example in patients that do not respond well to oxygen therapy (proposed as PaO2 < 300 mmHg on 100% inspired O2 [57]). These patients sometimes benefit from radiological coil embolization of the shunt.

Prognosis and treatment

Most studies have found that patients with HPS have an increased mortality compared with cirrhotic patients without HPS who have a similar severity of liver dysfunction.[11, 13] One study found that patients with HPS who do not undergo liver transplantation have a 23% 5-year survival from diagnosis of HPS compared with a 63% 5-year survival in matched cirrhotic controls without HPS.[58] Not surprisingly, the prognosis is worst in those patients with severe hypoxia, with most patients with PaO2 < 60 mmHg dying within 6 months.[11] These findings led to the practice of allocating additional model of end-stage liver disease (MELD) points to patients with HPS associated with PaO2 < 60 mmHg who are listed for transplant. The increased mortality in patients with HPS is related to liver failure and its complications, rather than respiratory failure.

Liver transplantation

Liver transplant remains the only effective treatment of HPS, although post-transplant survival is often reduced compared with patients without HPS. One large, prospective, multicenter trial documented a relative risk of death of 2.41 in patients with HPS after adjustment for age, sex, ethnicity, MELD, and liver transplantation.[13] Although the authors did not find any association between hypoxemia and mortality, there is evidence from other studies to suggest that PaO2 < 50 mmHg and/or MAA shunt fraction > 20% are predictive of increased mortality of up to 67% post-transplant.[58-60] These findings led to the concept of a “transplant window” for patients with HPS, in which patients with PO2 less than 60 mmHg are prioritized for transplant, while those with more severe hypoxia are excluded because of their poor post-transplant prognosis. However, it should be noted that other, albeit retrospective, studies evaluating outcomes following liver transplantation found that a preoperative diagnosis of HPS did not affect long-term mortality.[58, 61] Furthermore, more recent clinical experience would suggest that outcomes are improving with specialized postoperative care, particularly in the early post-transplant period.[62] Indeed, a recent study reported mortality of only 9% in patients with severe HPS, as defined by PaO2 < 50 mmHg.[62]

Interventional radiology

Early case reports suggested that portal decompression with transjugular intrahepatic shunt (TIPS) placement improved gas exchange and shunt fraction in HPS.[63] However, more recent case reports have been disappointing,[64, 65] and therefore the role of TIPS in the management of HPS remains unproven. Intra-arterial coil embolization of discrete pulmonary arteriovenous communications has been used successfully and may have a place in improving right to left shunt in rare patients with large fistulae amenable to radiographic intervention.[66] Coil embolization of multiple discrete arteriovenous fistulae has also been used successfully in a patient with persistent hypoxia 6 months after liver transplant.[67]

Medical therapy

An effective medical therapy for HPS has yet to be established. Oxygen is used for symptomatic relief in HPS and helps prevent hypoxic end-organ damage; however, objective evidence of beneficial effect is lacking. Interestingly, two cases were reported of improvement in liver function following oxygen treatment for HPS[68] in keeping with the concept that hypoxia may directly impair hepatic function and regeneration in this condition.

Results of small human trials of medical therapies for HPS have, in general, been disappointing. There have been several studies targeting NO, given its central role in mediating pulmonary vasodilation. Although inhibition of NO synthesis using intravenous methylene blue acutely improved oxygenation in HPS,[69] nebulized treatment with NOS inhibitor had no effect on gas exchange parameters, despite reducing cardiac output and increasing pulmonary vascular resistance.[70]

Given the possible role of TNF in HPS pathogenesis, pentoxifylline has been trialed in a small number of patients with HPS but failed to improve arterial oxygenation. However, the treatment was poorly tolerated, and only one patient was able to complete the study protocol, making it difficult to interpret the results.[71] A pilot study of intestinal decontamination with norfloxacin in patients with HPS, in an attempt to reduce endotoxemia, failed to produce any improvement in gas exchange.[72] Other therapies that have been tried without success includes somatostatin analogues[73] and indomethacin.[74] Two children with HPS improved with long-term aspirin therapy;[75] however, there have been no other studies to confirm this finding. Direct respiratory stimulation using almitrine resulted in the improvement in the alveolar–arterial oxygen gradient but not hypoxia.[76] Finally, a beneficial effect of garlic on oxygenation and dyspnea in HPS has been documented in two pilot trials,[77, 78] although the mechanism of action is unknown. No randomized controlled studies using garlic have been published.


HPS remains a fascinating pathophysiological entity that has a significant impact on both quality of life and mortality in patients with portal hypertension. While our understanding of the mechanisms of the pulmonary vasodilation that underlies the condition continues to improve, this has yet to translate to the development of effective pharmacological therapy. Liver transplantation is an effective treatment for HPS, and prompt recognition of the syndrome and timely referral are important in improving patient outcomes.