Risk factors for portopulmonary hypertension in patients with cirrhosis: a prospective, multicenter study

Tricuspid regurgitation pressure gradient (TRPG) measurement by echocardiography is recommended as the most objective examination to detect portopulmonary hypertension (PoPH). This study aimed to identify factors associated with a high TRPG in patients with cirrhosis and develop a scoring model for identifying patients who are most likely to benefit from echocardiography investigations. A total of 486 patients who underwent echocardiography were randomly allocated to the derivation and validation sets at a ratio of 2:1. Of the patients, 51 (10.5%) had TRPG ≥ 35 mmHg. The median brain natriuretic peptide (BNP) was 39.5 pg/mL. Shortness of breath (SOB) was reported by 91 (18.7%) patients. In the derivation set, multivariate analysis identified female gender, shortness of breath, and BNP ≥ 48.9 pg/mL as independent factors for TRPG ≥ 35 mmHg. The risk score for predicting TRPG ≥ 35 mmHg was calculated as follows: − 3.596 + 1.250 × gender (female: 1, male: 0) + 1.093 × SOB (presence: 1, absence: 0) + 0.953 × BNP (≥ 48.9 pg/mL: 1, < 48.9 pg/mL: 0). The risk score yielded sensitivity of 66.7%, specificity of 75.3%, positive predictive value of 25.5%, negative predict value of 94.3%, and predictive accuracy of 74.4% for predicting TRPG ≥ 35 mmHg. These results were almost similar in the validation set, indicating the reproducibility and validity of the risk score. This study clarified the characteristics of patients with suspected PoPH and developed a scoring model for identifying patients at high risk of PoPH, which may be used in selecting patients that may benefit from echocardiography.


Introduction
In chronic liver diseases, such as chronic hepatitis B or C virus (HBV or HCV) infection, alcoholic liver disease, fatty liver disease, primary biliary cholangitis, and autoimmune disease, hepatic inflammation persists over a long period contributing to the development of liver fibrosis. Without appropriate therapeutic intervention, chronic liver disease may progress to cirrhosis. A proportion of patients with cirrhosis develop portal hypertension and various extrahepatic complications including ascites, hepatic encephalopathy, and esophagogastric varices. Portal hypertension may be complicated by portopulmonary hypertension (PoPH), a rare but serious condition characterized by increased pulmonary vascular resistance associated with increased risks of right heart failure and death [1][2][3][4]. Masanori Atsukawa and Akihito Tsubota have contributed equally to the preparation of this manuscript.

3
The prevalence of PoPH in patients with cirrhosis or portal hypertension is reportedly 1-6% [5][6][7][8][9][10][11]. A study in a population in the USA demonstrated that 66 (5.3%) of 1,235 patients on the waiting list for liver transplantation had a diagnosis of PoPH, with no correlation observed between PoPH and model for end-stage liver disease (MELD) scores, an index of liver functional reserve [12]. A separate study from the USA reported that female gender and autoimmune hepatitis were risk factors for PoPH and that the MELD score was not correlated with the incidence of PoPH [13], corroborating the above report. In contrast, studies from Eastern countries (Japan and China) have reported higher rates of PoPH in HBV-and/or HCV-positive patients [9][10][11], indicating that the incidence and characteristics of PoPH in patients with cirrhosis may differ among countries/regions or races.
Pulmonary arterial hypertension (PAH) has been defined as a mean pulmonary artery pressure (mPAP) of ≥ 25 mmHg [14,15]. However, the World Symposium on Pulmonary Hypertension in 2018 proposed a revision of the diagnostic criterion for PAH from an mPAP of ≥ 25 mmHg to > 20 mmHg [16]. Pulmonary hypertension (PH) was classified into five groups according to the Nice Classification, with PH with a pulmonary artery wedge pressure (PAWP) of ≤ 15 mmHg classified as PH group 1 (i.e., PAH). In this group, PoPH is defined as PAH associated with the presence of portal hypertension [14][15][16].
The prognosis of patients with PoPH is poor. The 5-year survival rate of untreated patients with PoPH is reportedly 14%, with 54% of patients dying within one year of diagnosis [17]. Compared to patients with idiopathic PAH, patients with PoPH have been reported to have a poorer prognosis due to delays in diagnosis and treatment [18]. Meanwhile, survival has reportedly been prolonged by the development of drugs that reduce the pulmonary artery pressure, with a 5-year survival rate of 67% in patients with PoPH who underwent liver transplantation [19].
Therefore, there is global consensus regarding the importance of early diagnosis of PoPH and initiation of treatment for the management of patients with cirrhosis or portal hypertension [20]. However, the incidence of PoPH in patients with cirrhosis is relatively low and there are no established guidelines concerning screening methods for PoPH in patients with cirrhosis. Moreover, the diagnosis of PoPH is often clinically challenging partly as electrocardiographic findings are normal in 37% of patients with PoPH [21] and respiratory symptoms such as shortness of breath may be absent, particularly in the early stage of the disease. Therefore, there is an urgent clinical need for the development of simple and sensitive screening methods for PoPH in patients with cirrhosis. The definitive diagnosis of PoPH requires right heart catheterization; however, the use of such invasive procedures in all patients with suspected portal hypertension is challenging in routine clinical practice. Non-invasive examinations have been reported as auxiliary methods for predicting PoPH. Subjective symptoms (such as shortness of breath), serum brain natriuretic peptide (BNP) measurement, and measurement of the main pulmonary artery diameter have been reported to have utility in the early detection of PoPH [22][23][24]. Measurement of the tricuspid regurgitation pressure gradient (TRPG) values by echocardiography is recommended as the most objective examination for the detection of PoPH [15,20].
This prospective, multicenter study aimed to identify factors associated with a high TRPG in patients with cirrhosis and develop a screening method for identifying patients most likely to benefit from investigation with echocardiography.

Subjects
This was a prospective, 21-multicenter study conducted in Japan (UMIN registration no. 000040426). Between January 2020 and March 2022, 630 Japanese patients with cirrhosis were consecutively recruited into the present study. The inclusion criteria for the analysis were (1) age ≥ 20 years, (2) diagnosis of cirrhosis, and (3) consent for TRPG measurement using Doppler echocardiography. The exclusion criteria were: (i) uncontrollable hepatocellular carcinoma; (ii) chronic respiratory disease; (iii) left ventricular heart disease, such as a left ventricular ejection fraction < 50%, moderate or severe valvular disease of the left heart, and left ventricular outflow tract obstruction; (iv) atrial fibrillation; (v) chronic pulmonary embolism; (vi) congenital heart disease; (vii) human immunodeficiency virus infection; (viii) blood disease; (ix) sarcoidosis; (x) thyroid disease; and (xi) connective tissue disease. Eligible patients were randomly allocated to the derivation and validation sets at a ratio of 2:1 to construct and validate a risk score model for predicting TRPG ≥ 35 mmHg. Patients were given the option to abstain from participating in this prospective study.

Evaluation of liver functional reserve
Cirrhosis was diagnosed through imaging (abdominal computed tomography and/or ultrasonography) or liver biopsy. Child-Pugh classification and ALBI grade were used to assess liver functional reserve. ALBI grade was calculated based on serum albumin and total bilirubin values using the following formula: ALBI score = [log10 bilirubin (µmol/L) × 0.66] + [albumin (g/L) × − 0.085] [25].

Evaluation of TRPG values
Doppler echocardiography was performed according to commonly used guidelines [15,26]. TRPG values were estimated by calculating the right ventricular to right atrial pressure gradient during systole; that is, the modified Bernoulli equation was used to calculate the gradients from the velocities: TRPG (mmHg) = 4 × [tricuspid regurgitant velocity (TRV; m/sec)] 2 . An mPAP value of 25 mmHg on right heart catheterization was considered equivalent to a TRPG value of 36 mmHg on echocardiography [27]. However, the cut-off value of TRPG for predicting PoPH varies among previous reports, ranging from 31.36 to 46.24 mmHg [20]. With reference to the above reports [20,27], the cut-off value for predicting PoPH was defined as TRPG ≥ 35 mmHg in the present study.

Pulmonary artery catheterization
In patients with high TRPG who consented to examination for a definitive diagnosis of PoPH, right heart catheterization was performed to measure pulmonary arterial pressure (PAP), as previously described [22,28]. The procedure was performed using the cuff method to determine pulmonary artery wedge pressure (PAWP). A pulmonary arterial catheter (Thermodilution catheter; Nipro, Osaka, Japan) was connected to a monitor for pressure and cardiac output (CO) recording, with the latter being determined using the thermodilution technique. A cold 10 mL saline solution of known temperature was injected into the right atrium from the proximal port of a thermistor-tipped catheter inserted into the pulmonary artery. Blood was cooled while the injection solution passed into the pulmonary artery via the ventricle. The resultant temperature drop was measured using a thermistor at the catheter tip, after which a thermodilution curve and CO were calculated using a computer. Measurements were repeated five times consecutively, with the average CO recorded after the exclusion of the lowest and highest values. Considering the hemodynamic changes with respiration, pressure measurements were consistently performed at the end of expiration. Pressures were measured in mmH 2 O, with zero references for correction at the midthoracic level. Thereafter, water column pressure was converted into mercury column pressure. Pulmonary vascular resistance (PVR) was calculated using the following equation: PVR = (PAP − PAWP)/CO.

Statistical analyses
Continuous variables are presented as medians and ranges. Categorical variables are presented as numbers. Fisher's exact test and Mann-Whitney U test were used to compare two groups, as appropriate. Multiple regression analysis was used to identify factors that were significantly and independently associated with TRPG ≥ 35 mmHg. The best risk score model for predicting TRPG ≥ 35 mmHg was constructed based on the final-step results. A receiver operating characteristic (ROC) curve was generated to analyze the cut-off values of BNP and risk score that most rationally predicted TRPG ≥ 35 mmHg. Sensitivity, specificity, and positive and negative predictive values (PPV and NPV, respectively) were calculated based on the aforementioned optimal cut-off values. A p value of < 0.05 was regarded as statistically significant. All statistical analyses were performed using IBM SPSS Statistics Version 23.0 (IBM Japan, Tokyo, Japan).

Patient characteristics
A total of 630 patients with cirrhosis underwent Doppler echocardiography to measure TRPG. Sixty-one patients were excluded from this analysis due to a lack of data on BNP measurements. A further 83 patients were excluded due to the presence of left heart failure or atrial fibrillation. Thus, 486 patients were analyzed in this study. The patients comprised 265 males and 221 females, with a median age of 72 (range 22-92) years. The median Child-Pugh score was 6 (range 5-13) points. The numbers of patients with Child-Pugh class A, B, and C were 307, 125, and 54, respectively. The median ALBI score was − 2.42 (range − 3.67-0.14). The median TRPG value was 22.0 (range 4.0-91.2) mmHg. Of the 486 patients, 51 (10.5%) had TRPG ≥ 35 mmHg ( Supplementary Fig. 1). The median BNP level was 39.5 (range 3.3-712.0) pg/mL (Supplementary Fig. 2). Shortness of breath was reported by 91 (18.7%) patients. The 486 patients were randomly allocated to the derivation (n = 324) and validation (n = 162) sets, without significant differences in baseline characteristics between both sets ( Table 1). Figure 1 shows a comparison of baseline characteristics between patients with TRPG < 35 mmHg and ≥ 35 mmHg in the derivation set. There were no significant differences in Child-Pugh score, ALBI score, or prothrombin time between the two groups ( Fig. 1a-c). BNP levels were higher in patients with TRPG ≥ 35 mmHg (p = 1.22 × 10 −2 ; Fig. 1d). The proportion of females among patients with TRPG ≥ 35 mmHg was higher than that among patients with TRPG < 35 mmHg (p = 6.73 × 10 −4 ; Fig. 1e). The prevalence of reported shortness of breath was higher in patients with TRPG ≥ 35 mmHg (p = 1.17 × 10 −3 ; Fig. 1f).

An optimal cut-off BNP value for predicting TRPG ≥ 35 mmHg and prevalence of TRPG ≥ 35 mmHg according to the cut-off BNP value in the derivation set
Using the ROC curve analysis, the optimal cut-off value of BNP for predicting TRPG ≥ 35 mmHg was 48.9 pg/mL [area under the curve (AUC), 0.639; sensitivity, 62.8%; specificity, 60.5%] (Fig. 2a). Next, patients were divided into two groups using this cut-off BNP value. The prevalence of TRPG ≥ 35 mmHg in patients with BNP ≥ 48.9 pg/mL was significantly higher than that in patients with BNP < 48.9 pg/ mL (6.9% vs.17.4%; p = 6.48 × 10 −3 ; Fig. 2b).

An optimal risk score to predict TRPG ≥ 35 mmHg in the derivation set
The optimal cut-off value of the risk score for predicting TRPG ≥ 35 mmHg was − 1.550 (AUC, 0.757; sensitivity, 66.7%; specificity, 75.3%; Fig. 2c). Next, patients were divided into two groups using this cut-off risk score. The prevalence of TRPG ≥ 35 mmHg in the patients with a risk score ≥ − 1.550 was significantly higher than that in patients with a risk score < − 1.550 (5.2% vs. 25.5%; p = 7.34 × 10 −7 ; Fig. 2b).

Diagnostic performance of risk score for predicting TRPG ≥ 35 mmHg in the derivation and validation sets
The cut-off risk score in the derivation set yielded sensitivity of 66.7%, specificity of 75.3%, PPV of 25.5%, NPV of 94.3%, and predictive accuracy of 74.4% (Table 3a). These results were confirmed using the validation set. The cut-off risk score in the validation set yielded sensitivity of 46.7%, specificity of 70.4%, PPV of 14.0%, NPV of 92.9%, and predictive accuracy of 68.5% (Table 3b). These results were almost similar between the derivation and validation sets, indicating the reproducibility and validity of the risk score.

Characteristics including risk scores in patients diagnosed with PoPH by pulmonary artery catheterization
Of the 51 patients with TRPG ≥ 35 mmHg, 15 consented to pulmonary artery catheterization for a definitive diagnosis of PoPH. A risk score ≥ − 1.550 and 2 or more risk factors were observed in 13 (86.7%) patients, while the remaining 2 patients had a risk score < − 1.550 and only 1 factor (shortness of breath) (Supplementary Table 1). As a result, all 15 patients were definitively diagnosed with PoPH via pulmonary artery catheterization.

Discussion
PoPH is a serious complication of cirrhosis and portal hypertension; however, the diagnosis of PoPH in routine clinical practice is challenging due to a low prevalence and subclinical progression. The 6th World Symposium of Pulmonary Hypertension recommended a diagnostic algorithm for PH based on past history, symptoms and laboratory results suggestive of PH, and echocardiography. These guidelines further recommended examinations for left heart failure and pulmonary disorders and consultation with a specialist in cases where PH is strongly suspected [20]. Ideally, all patients with cirrhosis or portal hypertension should undergo echocardiography for investigation of PoPH [20]; however, the cost-effectiveness of this approach would be low in clinical settings. Therefore, the present study aimed to clarify the characteristics of patients with suspected PoPH and to identify patients who may benefit from further investigation with echocardiography. The results of the present study revealed that shortness of breath, female gender, and high BNP levels were significant, independent factors for predicting TRPG ≥ 35 mmHg (i.e., PoPH). These risk factors are easily obtainable without invasive procedures (such as pulmonary artery catheterization). A simple noninvasive  risk score using such factors is convenient and helpful in early PoPH diagnosis and treatment in clinical practice. This study constructed a risk score model for predicting TRPG ≥ 35 mmHg with a derivation set and verified its diagnostic performance with a validation set. The risk score yielded high specificity, NPV, and predictive accuracy in both sets, although the sensitivity and PPV were not high. High specificity and NPV made the risk score a useful tool to avoid excessive amounts of diagnostic tests for PoPH in patients with cirrhosis. Meanwhile, low sensitivity and PPV were probably affected by subclinical progression and low prevalence of PoPH, respectively. To improve the diagnostic performance of the risk score, shortness of breath should be more carefully interviewed or examined. Alternatively, the risk score model should be reconstructed by adding stronger predictors that have not yet been elucidated. Further study is needed to reconfirm the model comprising the three risk factors in another large-scale cohort. Principal symptoms of PAH include exertional dyspnea, fatigue, and chest pain [20]. Although these symptoms are not specific to PAH, medical consultations with inquiry regarding shortness of breath are important opportunities to detect PoPH [22,29]. The recent guidelines advocated from Europe recommended the presence of unexplained exertional dyspnea (such as shortness of breath) in screening for PAH in patients with portal hypertension [30]. This study demonstrated that shortness of breath was most strongly associated with high TRPG (i.e., PoPH). Thus, patients with shortness of breath even in the absence of cardiovascular disease and associated symptoms should undergo echocardiography. However, 19% of patients with PoPH do not report shortness of breath [22]. The present study including 486 patients revealed that 45.1% (23/51) of patients with TRPG ≥ 35 mmHg reported shortness of breath. If shortness of breath is absent, it is important to consider other factors (such as BNP levels and gender, as shown in this study) so as not to overlook patients with high TRPG. BNP measurements have no utility in diagnosing PoPH, although elevated BNP levels are indicative of right heart overload and poor overall prognosis [20]. However, it has been reported that BNP measurement has utility in the early diagnosis of PoPH in patients with biliary atresia, with a cut-off value of BNP for predicting asymptomatic PoPH of 29.1 pg/mL [23]. A study comprising 157 patients with decompensated cirrhosis or portal hypertension reported that BNP levels were significantly higher in patients with high right ventricular systolic pressures, thereby predicting PoPH [8]. The aforementioned guidelines from Europe reported that a BNP value of ≥ 50 pg/mL in PAH (including PoPH) indicates intermediate risk (5-20%) of estimated 1-year mortality; thus, the cut-off BNP value of 48.9 pg/mL in this study could be rational and practical for identifying patients with cirrhosis complicated by suspected PoPH [30]. In this study, BNP levels were significantly and independently associated with high TRPG. However, there are few reports on the diagnostic performance of BNP in PoPH. Given that BNP is an objective indicator that can be easily measured, further studies are required to evaluate the potential utility of BNP measurement in the diagnosis of PoPH.
A study from Japan reported that female gender, higher shortest diameter of the inferior vena cava, presence of portosystemic shunts of ≥ 5 mm, and lower blood urea nitrogen levels were independent factors related to PoPH in 130 patients with portal hypertension [24]. It has also been reported from studies conducted in other countries that the incidence of PoPH is significantly higher in female patients than in male patients [10,13,31,32]. These results are consistent with the findings of the present study that female gender was independently associated with TRPG ≥ 35 mmHg. Factors related to sex hormone metabolism have been posited as factors influencing the higher prevalence of PoPH in females. As for heritable PAH, bone morphogenetic receptor type 2 (BMPR2) mutations have been shown to be a risk factor for the onset of PAH [33,34], with BMPR2 mutations more frequent in females [35]. CYP1B1 metabolizes estrogen and estrone, and the CYP1B1 Asn453Ser polymorphism has been shown to be associated with PAH in females with BMPR2 mutations [36]. Sex hormone abnormalities caused by variants in the genes that encode estrogen receptor 1 and aromatase (CYP19A1), which produces estradiol, estrone, and 16α-hydroxyestrone, are potentially associated with PoPH in patients with advanced liver disease [31].
PoPH is primarily observed in patients with cirrhosis complicated by portal hypertension. There have been a number of reports on the relationships of liver functional reserve and etiology with PoPH. Several studies have reported that liver functional reserve is not correlated with the frequency of complications related to PoPH [6,12,13,32]. The present study also found no correlation between high TRPG and Child-Pugh class or ALBI grade, which are commonly used indices for the assessment of liver functional reserve. Concerning the relationship between PoPH and etiology, PoPH has been reported to be frequently observed in patients with autoimmune hepatitis or primary biliary cholangitis [6,8,31,32,37]. Further, there have been reports that PoPH is often observed in HCV-positive or HBV-positive patients in China [9,10] and in HCV-positive patients in Japan [11]. Such differences may be attributable to differences in the primary etiology of cirrhosis among countries or regions. In this study, patients with autoimmune hepatitis or primary biliary cholangitis accounted for the highest proportion (15.4%) of patients in the high TRPG group (Supplementary Fig. 3); however, etiology was not associated with high TRPG. This finding requires further validation in cohorts from countries and regions other than Japan.
There are several limitations of the present study. First, a definitive diagnosis of PoPH had not been made in many patients as right heart catheterization was not performed in all patients. Accordingly, we simply analyzed the characteristics of patients suspected to have PoPH based on TRPG values. Notably, this study revealed that 13 of 15 patients with a high-risk score fulfilled the diagnosis criteria of PoPH using right heart catheterization, although with a small number of patients. Next, while this study was prospective, individual physicians may have performed echocardiography mainly in patients with suspected PoPH. In addition, subjects were selected based on the presence of cirrhosis rather than portal hypertension. Therefore, selection bias increasing the frequency of patients with high TRPG values could not be definitively excluded. As it is difficult to perform hepatic vein catheterization for determining the presence and severity of portal hypertension in all patients, the association between the presence or severity of PoPH and the severity of portal hypertension remains unclear.

Conclusions
In conclusion, we clarified the characteristics of patients with suspected PoPH (TRPG ≥ 35 mmHg) in patients with cirrhosis who underwent echocardiography. We developed a scoring model for identifying patients at high risk of PoPH, which may have utility in selecting patients with cirrhosis that may benefit from echocardiography in clinical practice.