Article Text
Abstract
Background: The hepatopulmonary syndrome (HPS) is defined as the triad of liver disease, arterial deoxygenation, and pulmonary vascular dilatation. The reported prevalence of HPS in cirrhotic patients varies between 4% and 19%, and various threshold values defining arterial deoxygenation have been used and recommended previously. However, it is not known how the prevalence of HPS differs using different cut off values for arterial deoxygenation.
Methods: We studied 127 patients for the presence of HPS using transthoracic contrast echocardiography for detection of pulmonary vasodilation, pulmonary function tests, and blood gas analysis.
Results: Ninety eight patients were included in the study, of whom 33 (34%) had a positive contrast echocardiography. Using an increased alveolar-arterial difference for the partial pressure of oxygen (AaDO2) as an indication of hypoxaemia, the prevalence of HPS was considerably higher (>15 mm Hg, 32%; >20 mm Hg, 31%; and >age related threshold, 28%) than using reduced partial pressure of arterial oxygen (PaO2) as a threshold (<80 mm Hg, 19%; <70 mm Hg, 15%; and <age related threshold, 15%). For AaDO2 as the cut off, the positive predictive value for a diagnosis of HPS was low (34%, 37%, and 53%, respectively). In contrast, PaO2 as a cut off had considerably higher positive predictive values (44%, 93%, and 94%, respectively). Introducing PaO2 <65 mm Hg as the cut off, the positive predictive value increased to 100%. Dyspnoea was more often present in patients with “clinically significant” HPS (57%) compared with “subclinical HPS” (8%), and patients without HPS (6%). The Child-Pugh score correlated significantly with the severity of HPS. Two of five liver transplanted patients with “subclinical HPS” had embolic brain infarcts, possibly induced by venous emboli passing through dilated intrapulmonary vessels.
Conclusions: Defining arterial hypoxaemia in HPS by different, previously used, cut off values for arterial oxygenation leads to a wide variation in the prevalence of HPS in the same sample of cirrhotic patients.
- liver cirrhosis
- portal hypertension
- hypoxaemia
- pulmonary vasodilation
- contrast echocardiography
- HPS, hepatopulmonary syndrome
- PaO2, partial pressure of arterial oxygen
- AaDO2, alveolar-arterial difference for the partial pressure of oxygen
- PaCO2, partial pressure of arterial carbon dioxide
- FEV1, forced expiratory volume in one second
- TLC, total lung capacity
- OLT, orthotopic liver transplantation
Statistics from Altmetric.com
- HPS, hepatopulmonary syndrome
- PaO2, partial pressure of arterial oxygen
- AaDO2, alveolar-arterial difference for the partial pressure of oxygen
- PaCO2, partial pressure of arterial carbon dioxide
- FEV1, forced expiratory volume in one second
- TLC, total lung capacity
- OLT, orthotopic liver transplantation
The hepatopulmonary syndrome (HPS) is defined as the triad of liver disease, pulmonary gas exchange abnormalities leading to arterial deoxygenation, and widespread pulmonary vascular dilatation.1,2 Whereas both acute and chronic liver diseases have been associated with HPS, most commonly it presents in patients with cirrhosis. Portal hypertension seems to be the predominant factor related to this syndrome. The hallmark of pulmonary vascular changes in HPS are dilated vessels at the precapillary and capillary level and direct arteriovenous communications.1 This causes right to left shunting of blood flow, mismatch between ventilation and perfusion, and diffusion limitation.2 Pulmonary features include digital clubbing, cyanosis, dyspnoea, platypnoea, and orthodeoxia; the latter two are defined as dyspnoea and arterial deoxygenation induced by the upright position and relieved by recumbency.3
A diagnosis of HPS is established when the following three points are fulfilled.4
Chronic liver disease, usually complicated by portal hypertension.
Arterial hypoxaemia, defined by a reduced partial pressure of arterial oxygen (PaO2) or more accurately by an increased alveolar-arterial difference in the partial pressure of oxygen (AaDO2). The latter includes determination of the partial pressure of arterial carbon dioxide (PaCO2) which is often low in cirrhotic patients as a result of hyperventilation.2
Intrapulmonary vascular dilatation, detected either by two dimensional contrast echocardiography or macroaggregated albumin lung perfusion scan.3
In the literature, the prevalence of pulmonary vasodilation, detected by transthoracic contrast echocardiography, varies widely (5–47%)5–11 and a definitive diagnosis of HPS varies between 4% and 19%5–8,10–13 in cirrhotic patients. Various threshold values defining arterial hypoxaemia have been recommended and used in previous publications: PaO2 <70 mm Hg,5,7,8,11 PaO2 <80 mm Hg,14 AaDO2 >15 mm Hg,12,15 AaDO2 >20 mm Hg,16 and AaDO2 >age related threshold value.6,13 Consequently, these different threshold values may result in a different prevalence of HPS. Thus the aims of our study were: (1) to clarify how the prevalence of HPS in a large sample of cirrhotic patients varies according to previously used different cut off values for arterial hypoxaemia; and (2) to determine the predictive value of those cut off values in the diagnosis of HPS.
PATIENTS AND METHODS
Patients
The study protocol was approved by the institutional ethics committee of the University of Vienna and written informed consent was obtained from each patient. A total of 127 patients with biopsy proven cirrhosis, evaluated for liver transplantation or transjugular intrahepatic portosystemic shunt, underwent transthoracic contrast echocardiography, arterial blood gas analysis on room air, lung function tests, and chest radiograph. Nine patients were excluded because of inadequate echocardiographic image quality, 12 patients were excluded because of unfeasible lung function tests, and eight patients were excluded because of impaired lung function tests, defined as forced expiratory volume in one second (FEV1) or total lung capacity (TLC) <66% of predicted.5,17,18 Mean age of the remaining 98 patients was 56 years (range 33–82); 65 (66%) were men and 33 (34%) were women. No patient had evidence of portopulmonary hypertension, assessed by echocardiography,19 including Doppler measurements.20–22
Procedure
Contrast echocardiography
Agitated saline was used as a contrast medium which creates a stream of microbubbles after intravenous injection.1 In healthy individuals, these microbubbles, greater than 15 μm in diameter,16,23 opacify the right heart chambers only because they are filtered in the pulmonary capillary bed and do not appear at the left side of the heart. The distinction between intrapulmonary or intracardiac shunt is made by the time of appearance of the microbubbles in the left heart chambers: in intracardiac shunt the microbubbles appear generally within three heartbeats after appearance in the right heart chambers and in intrapulmonary shunt they appear 4–6 heartbeats after their initial appearance in the right side of the heart.
Arterial blood gas analysis
Arterial blood gas samples were obtained by percutaneous radial artery puncture with the subject in a seated position breathing room air, and were analysed with a standard blood gas analyser (BGElectrolytes, Instrumentation Lab. Inc., USA). AaDO2 was calculated using the alveolar gas equation.24
Lung function tests
FEV1 was obtained using a computerised spirometer (Autobox 6200; Sensor Medics, Yorba Linda, California, USA) according to standard procedures.25 The best two of three acceptable tracings for FEV1 did not vary by more than 5%. TLC was measured by a body plethysmograph (Autobox 6200; Sensor Medics).
Statistical analysis
Group comparisons were performed using the Mann-Whitney U test and Fisher’s exact test. The prevalence of HPS among cirrhotics and positive predictive values for a diagnosis of HPS with corresponding confidence intervals were calculated for different cut off values of PaO2 and AaDO2.
RESULTS
Patient characteristics according to findings on contrast echocardiography (table 1)
Thirty three of 98 patients (34%) had intrapulmonary vasodilation, as evidenced by contrast echocardiography. Demographic data of the subjects with and without a positive contrast echocardiography are shown in table 1. Patients with a positive contrast echocardiogram had more severe cirrhosis, as assessed by a significantly higher Child-Pugh score,26 compared with patients with a negative contrast echocardiography. Mean total bilirubin was significantly higher, and mean values for serum albumin, prothrombin time, and erythrocyte count were significantly lower in cirrhotics with a positive contrast echocardiography. Oesophageal varices were found more often in the group with a positive contrast echocardiography but the difference did not reach statistical significance. Lung function values (FEV1 and TLC) were slightly lower in cirrhotics with a positive contrast echocardiography compared with those with negative contrast echocardiograms. The parameters of arterial oxygenation, PaO2 and AaDO2, were highly significantly different, with lower values in the group with a positive contrast echocardiography. PaCO2 was reduced due to hyperventilation, with no difference between the two groups. The frequency of smokers was similar in both groups: seven patients (21%) in the group with a positive contrast echocardiography and 11 patients (17%) in the group with a negative contrast echocardiography. Ascites was present in four (12%) and 13 (20%) patients with and without a positive contrast echocardiography, respectively (NS). Chest radiograph showed interstitial markings, predominately in the lower lung fields, in seven (21%) and two (3%) patients with and without a positive contrast echocardiography, respectively (p<0.01). Small pleural effusions were seen in two (6%) contrast echo positive patients and in seven (11%) contrast echo negative patients (NS).
Prevalence of HPS (table 2)
PaO2 and AaDO2 values of patients are plotted in fig 1. Vertical lines were added to show the previously published cut off values for PaO2 and AaDO2, and also PaO2 values of 60 and 65 mm Hg. The different cut off values of PaO2 and AaDO2 which, together with pulmonary vasodilation defined HPS previously, led to a wide range of prevalence rates for this syndrome (15–32%; see table 2). The prevalence of HPS was lowest when PaO2 <70 mm Hg was used as the cut off value (15%) and increased more than twofold when AaDO2 >15 mm Hg (32%) and >20 mm Hg (31%) were used. Overall, prevalence was higher when AaDO2 cut off values were used (28–32%) compared with PaO2 cut off values (15–19%). Introduction of an age related threshold value for PaO2 led to a prevalence of 15%. As expected, PaO2 <65 mm Hg and <60 mm Hg, threshold values for more severe hypoxaemia which were not used previously in definitions of HPS, resulted in the lowest prevalence rates (13% and 12%, respectively).
Positive predictive value for HPS (table 2)
Positive predictive values for different cut off values of arterial oxygenation for a diagnosis of HPS are presented in table 2. The positive predictive value for a diagnosis of HPS was highest (100%) with PaO2 <65 mm Hg and <60 mm Hg, and slightly lower (93% and 94%) with PaO2 <70 mm Hg and <age related threshold value, respectively. Using other oxygenation cut off values (PaO2 <80 mm Hg, AaDO2 >15 mm Hg, AaDO2 >20 mm Hg, and >age related threshold value) resulted in very low positive predictive values for a diagnosis of HPS.
Negative predictive value for HPS and overall accuracy (table 2)
Negative predictive values for exclusion of HPS ranged from 73% to 78% when PaO2 was used as a cut off, and increased to 83% for AaDO2 >20 mm Hg and to 87% for AaDO2 >age related threshold value. Overall accuracy (relative part of correctly classified patients) was highest for PaO2 <age related threshold value (81%), followed by PaO2 <70 mm Hg (80%), and PaO2 <65 mm Hg (80%).
Frequency of a positive contrast echocardiography according to different grades of gas exchange abnormalities
In the patient group with moderate to severe gas exchange abnormalities, the frequency of a positive contrast echocardiography was high: 14/15 (93%) patients with PaO2 <70 mm Hg, 15/16 (94%) patients with PaO2 <age related threshold value, 13/13 (100%) patients with PaO2 <65 mm Hg, and 9/9 (100%) patients with PaO2 <60 mm Hg had a positive contrast echocardiography.
In contrast, in the normoxaemic group (PaO2 >70 mm Hg) with an elevated AaDO2 (>age related threshold value), 13/36 (36%) patients had a positive contrast echocardiography (fig 2A).
In the group with normal gas exchange, defined by AaDO2 <age related threshold value, 6/47 (13%) patients had a positive contrast echocardiography (fig 2B).
Clinical features of patients with “clinically significant” and “subclinical” HPS (table 3)
For clinical purposes, we subdivided the patient group with intrapulmonary vasodilation (assessed by a positive contrast echocardiography) and abnormal gas exchange into two groups: “clinically significant” HPS, characterised by hypoxaemia (PaO2 <70 mm Hg) and “subclinical” HPS, characterised by normoxaemia (PaO2 ≥70 mm Hg but AaDO2 >age related threshold). Fourteen patients with “clinically significant” HPS had a higher frequency of dyspnoea at rest (57%) compared with 13 “subclinical” HPS patients (8%) and the patient group without HPS (6%; p<0.001). Spider naevi were found more often in the “clinically significant” HPS group (79%) compared with the “subclinical” HPS group (54%) and the group without HPS (39%; p<0.05). The frequency of palmar erythema was not significantly different between the three groups. The Child-Pugh score was highest in “clinically significant” HPS patients (11.6 (1.9)), followed by the “subclinical” HPS patients (10.5 (3.1)); it was lowest in the patient group without HPS (9 (2.5); p<0.05).
Nineteen patients (19%) underwent orthotopic liver transplantation (OLT): two from the “clinically significant” HPS group, five from the “subclinical” HPS group, and 12 from the group without HPS. Two patients with “subclinical” HPS experienced cerebral embolic events. The first patient had two episodes of paresis of the right arm four years after transplantation; no cardiac or arterial source for cerebral embolism could be detected (no arrhythmias, normal transoesophageal echo, no patent foramen ovale, normal sonography of the carotid and vertebral arteries). The second patient had embolic brainstem infarct three years after OLT and no source for the emboli could be detected. Wound infection (abdominal wall abscess) occurred in one patient with “subclinical” HPS. None of the transplanted patients had respiratory problems in the postoperative period except for one “subclinical” HPS patient who needed prolonged weaning from the ventilator. Mortality was 37% (3/8) in the HPS group compared with 17% (2/12) in the group without HPS. Both transplanted patients with “clinically significant” HPS died two and 10 months after OLT because of progression of their underlying disease (OLT was performed because of hepatocellular carcinoma; they developed multiple metastasis formation after several weeks). One patient with “subclinical” HPS died 6.5 months after OLT due to an ileus, and two transplanted patients without HPS died 7.5 and 9 months after OLT (relapse of hepatocellular carcinoma; unknown reason).
DISCUSSION
Our study shows that defining hypoxaemia in HPS by different, previously used, cut off values for arterial oxygenation leads to wide variation in the prevalence of HPS in the same sample of cirrhotics. This clearly demonstrates that previously published data on the prevalence of HPS using different oxygenation cut off values cannot be compared directly.
Available data of previous publications on the frequency of positive contrast echocardiograms and the prevalence of HPS are provided in table 4 and grouped according to their preferred cut off value defining arterial hypoxaemia. The prevalence of HPS in our patients using PaO2 <70 mm Hg as the cut off value (15%) was within the reported range of 5–17.5%.5,7,8,11 Using PaO2 <80 mm Hg, the prevalence in our study (19%) was higher compared with the study of Vedrinne and colleagues (8%).10 This can be explained by the higher rate of hypoxaemic patients in our study population (43% of all patients had PaO2 values <80 mm Hg) compared with those in Vedrinne et al’s study10 (24% of all patients had PaO2 values <80 mm Hg). With AaDO2 >15 mm Hg as the cut off, the prevalence in our patients (31%) was also higher compared with the studies of Rolla and colleagues15 (20%) and Martinez-Palli and colleagues12 (19%). In the study of Rolla et al, fewer patients had a positive contrast echocardiography (27%) and patients were less hypoxaemic (mean PaO2 85.1 mm Hg v 80.5 mm Hg in our patients; mean AaDO2 17.8 mm Hg v 31.2 mm Hg in our patients). In the abstract by Martinez-Palli et al, no detailed data are provided on the frequency of positive contrast echocardiograms. In addition, AaDO2 >age related threshold was associated with a higher prevalence in our study (28%) compared with the studies of Stoller and colleagues13 (4%) and Aller and colleagues6 (16%). In Stoller et al’s study, HPS was diagnosed in four of 98 patients but the frequency of intrapulmonary vasodilation and arterial blood gas analysis in the remaining 94 patients were not provided. In the study of Aller and colleagues6 a different formula for calculation of AaDO2 may have been the reason for the different prevalence values. Also, the contrast agent used may have contributed to the different prevalence data. Krowka and colleagues7 and Stoller and colleagues13 in part used indocyanine green dye solution which provides microbubbles with diameters of up to 90 μm, while we and others (Hopkins and colleagues,8 Abrams and colleagues,5 Rolla and colleagues,15 and Aller and colleagues6) used saline solution which creates microbubbles of 15–180 μm. Vedrinne and colleagues10 used a modified fluid gelatine solution which creates microbubbles of 10±2 μm. The frequency of intrapulmonary vasodilation in our patients (34%), measured by transthoracic contrast echocardiography, was within the range reported in the literature (5–47 %, see table 3).
Figure 1 clearly illustrates that most of the study patients, those with as well as those without a positive contrast echocardiography, had AaDO2 values >15 mm Hg (90/98; 92%) and >20 mm Hg (80/98; 82%); this frequency was reduced with AaDO2 >age related threshold (51/98; 52%). In contrast, the frequency of abnormal PaO2 values was lower: 43% of all study patients had a PaO2 <80 mm Hg, 16% had a PaO2 <70 mm Hg or <age related threshold value, and 12% had a PaO2 <60 mm Hg. This higher frequency of patients with abnormal values of AaDO2 compared with PaO2 may be explained by the higher sensitivity of AaDO2 in describing the oxygenation abnormality because it includes the PaCO2 determination (which is low as a result of hyperventilation). Thus PaO2 may be normal yet as a result of hypocapnia there is an increased AaDO2.1,2
The resulting positive predictive values of all three AaDO2 cut off values for a diagnosis of HPS were low (34%, 37%, and 53%, respectively). In contrast, PaO2 cut off values generally showed a higher positive predictive value. Using PaO2 <80 mm Hg as the cut off, the positive predictive value was still low (44%) but increased with PaO2 <70 mm Hg (93%) and <age related threshold value (94%). No patient without a positive contrast echocardiography had PaO2 values below 65 mm Hg; thus cut off levels for more severe hypoxaemia (PaO2 <65 mm Hg and <60 mm Hg) led to a positive predictive value of 100%. Consequently, cirrhotics with PaO2 <70 mm Hg or <age related threshold value had a higher probability of HPS. Moreover, when PaO2 was below 65 mm Hg, a diagnosis of HPS was definitely established.
In the patient group with a normal PaO2 (>70 mm Hg) but elevated AaDO2, the frequency of positive contrast echocardiography was relatively low (36%). This frequency may be increased when using transoesophageal contrast echocardiography which has been shown previously to be superior to the transthoracic approach in the detection of intrapulmonary vasodilation in cirrhotics.6,10 In the study of Vedrinne and colleagues,10 the frequency of positive contrast echocardiography and the prevalence for HPS in cirrhotics increased by 5% and 19%, respectively, when the transoesophageal approach was used compared with the transthoracic procedure (see table 4). In concordance, using the transoesophageal technique, Aller et al reported an increase of 14% and 6% for detection of intrapulmonary vasodilation and prevalence for HPS (see table 4).6 Therefore, it seems reasonable that some of our patients with normoxaemia but elevated AaDO2 and a negative transthoracic contrast echocardiography would have had a positive transoesophageal contrast echocardiography. However, most studies investigating HPS have used transthoracic contrast echocardiography for the detection of intrapulmonary vasodilation,5,7,8,11–13,15 and all reviews on HPS describe transthoracic contrast echocardiography as the method of assessment of intrapulmonary vasodilation,1–4,16,27–29 apart from lung perfusion scan. Moreover, for our screening study with a relatively high patient number (127 patients) the usefulness of the transoesophageal technique was limited due to the imposed risk as an invasive procedure in patients who had a history of variceal bleeding. However, the correlation between intrapulmonary vasodilation and mild gas exchange abnormalities seems to be weak and other reasons for an elevated AaDO2 should be considered in these patients. In fact, of our 23 patients with normal PaO2 but elevated AaDO2 and negative contrast echocardiography, six had mild ascites, two had small pleural effusions, and two had mild pulmonary function test abnormalities (FEV1 and/or TLC between 66% and 85% of predicted), which might contribute to the elevated AaDO2.
Interstitial markings were observed significantly more often in contrast echo positive patients. They are typical of HPS,29–33 are predominately localised in the lower lung fields, and may reflect pulmonary vascular dilatations.
Patients with normal gas exchange may also show a positive contrast echocardiography (6/47; 13% in our study); this is in agreement with previous studies.5–8,10,11,15
As expected, a high percentage of patients with “clinically significant” HPS felt dyspnoeic (57%) whereas dyspnoea was seldom noted in the patient groups with “subclinical” HPS (8%) and without HPS (6%). It has been noted previously that patients with cutaneous spider naevi have more profound gas exchange abnormalities and more intrapulmonary vasodilation,34–36 suggesting that spider naevi might be a cutaneous marker of intrapulmonary vascular dilatations.3 Our study supports these previous reports with small patient numbers, and showed that there was a significant correlation between cutaneous spider naevi and the severity of HPS. Conflicting data exist in the literature regarding the correlation between HPS and the severity of liver disease. Whereas a study by Abrams and colleagues17 showed significantly lower PaO2 values, higher AaDO2 values, and greater shunt fractions in Child-Pugh A cirrhosis compared with Child-Pugh B and C classes, another study by Vachiéry and colleagues37 showed that hypoxaemic cirrhotics had a significantly higher Child-Pugh score. Our study clearly showed a significant correlation between the severity of HPS and Child-Pugh score. Patients with “subclinical” HPS bear the risk of developing “clinically significant” HPS during the course of their disease but the risk is unknown and should be determined in a prospective follow up study. Deterioration of gas exchange is not the only risk in these patients; they may develop other complications of intrapulmonary vasodilation, such as venous emboli passing through dilated intrapulmonary vessels into the systemic circulation. Two of our 13 “subclinical” HPS patients had embolic brain infarcts after OLT with no cardiac or arterial source detected for the cerebral embolism. Cerebral embolism due to venous emboli has been described previously in a few patients with HPS after OLT.38–40 For the development of wound infections after OLT, as observed in one of our patients, HPS may also play an important pathogenetic role.40
In conclusion, the use of various cut off values defining arterial hypoxaemia in HPS described in previously published studies leads to a wide range of prevalence rates for HPS in patients with cirrhosis. PaO2 values below 70 mm Hg or below the age related threshold value predict the presence of HPS with high probability in the absence of intrinsic cardiopulmonary diseases, whereas a PaO2 value <65 mm Hg definitely predicts a diagnosis of HPS. The severity of HPS is clearly correlated with the degree of liver disease. Whereas dyspnoea is often noted in patients with “clinically significant” HPS, patients with “subclinical” HPS may have complications not directly associated with impaired gas exchange but caused by dilated intrapulmonary vessels, such as venous emboli passing through the pulmonary circulation to the brain.
Acknowledgments
The authors thank the nursing staff of the intensive care unit, Department of Internal Medicine IV, University of Vienna, Austria, for help in performing the study.