Use of Noninvasive Tests to Assess Clinically Significant Portal Hypertension and Response to Beta-Blockers

Abstract

Hepatic venous pressure gradient is the gold standard for determining the presence of clinically significant portal hypertension and response to nonselective beta-blocker therapy. Of late, noninvasive tests (NITs) using algorithms based on blood tests, measurement of liver and splenic stiffness and radiomics have been found to reliably reflect stages of liver fibrosis and presence of portal hypertension. There is limited, but emerging evidence that NITs could be useful in predicting response to treatment with beta-blockers. In this review, we discuss the present status of the NITs in monitoring response and predicting nonresponse to nonselective beta-blockers therapy in patients with cirrhosis and clinically significant portal hypertension. This update also focuses on the benefits of NITs as they could be used dynamically with ease over prolonged periods.

Graphical abstract

Introduction

The last decade has witnessed a rising trend in liver-related diseases due to various etiologies, mainly related to the increasing prevalence of metabolic syndrome–associated liver disease and alcohol-associated liver disease. These liver diseases are potentially preventable but requiring careful monitoring and surveillance to predict disease progression. The progression to cirrhosis signifies a turning point in the course of liver disease, as the presence and severity of cirrhosis correlate with prognosis across all stages and etiologies.¹ In the Baveno VI consensus,² the term compensated advanced chronic liver disease (cACLD) was coined to identify the group of patients with chronic liver disease at a high risk of developing clinically significant portal hypertension (CSPH). This stratification uses a liver stiffness measurement (LSM) value ≥ 10 kilopascals (kPa). Baveno VII further stated that LSM values between 10 and 15 kPa are suggestive, while > 15 kPa is highly suggestive of cACLD.³ The natural history of cirrhosis progresses from an asymptomatic stage of compensated cirrhosis to a symptomatic stage of decompensated cirrhosis, which is marked by complications of portal hypertension (PH) and liver dysfunction. Patients of compensated cirrhosis may or may not have CSPH. Decompensated cirrhosis is marked by the development of complications such as ascites, variceal bleeding, hepatic encephalopathy, spontaneous bacterial peritonitis, hepatorenal and hepatopulmonary syndromes.³ Decompensation carries with it an increased risk of disease-associated mortality. When compared to the general population, patients of compensated and decompensated cirrhosis, carry a 5- and 10-fold increase in risk of mortality, with a poor 5-year survival of 67% and 45%.⁴ Thus, the prime focus of management in patients with cirrhosis is prevention of progression to decompensation.

The gold standard for assessing portal pressures is hepatic venous pressure gradient (HVPG), with CSPH being defined by pressure ≥ 10 mmHg. CSPH identifies patients with compensated cirrhosis at high risk of decompensation and presence of varices. Studies have shown that treatment of patients with compensated cirrhosis and CSPH with long-term nonselective beta-blockers (NSBBs) could increase decompensation-free survival.^3,5 Moreover, the response to nonselective beta-blocker (NSBB) prevents further decompensation in patients with single decompensation. Regular monitoring of these patients is needed to assess their response. The invasive nature, logistic challenges and associated cost of HVPG measurement make it impractical to identify and monitor portal pressures, despite its accuracy.⁶ Therefore, the use of repeated HVPG measurement to determine response to NSBB is not always practically possible. A number of noninvasive tests (NITs) have been proposed and studied to predict CSPH and monitor response to NSBB. However, the search for tests that accurately replicate HVPG is ongoing. The scarcity of data and lack of common consensus/guidelines hinder the applicability of these tests in standard practice. In this review, we focus on understanding the role of NITs in assessing the response to beta-blocker therapy in patients with CSPH.

HVPG: a Gold Standard

PH is characterized by a pathological increase in the portal pressure gradient, which is calculated as the difference between portal vein pressure and inferior vena cava pressure. An alternate and easier-to-obtain approximation of the portal pressure gradient is the HVPG.⁷ HVPG is measured via hepatic vein catheterization, determined by the difference between wedged and free hepatic venous pressure. Based on HVPG, PH can be classified as subclinical (HVPG 6–9 mmHg) or clinically significant (HVPG ≥ 10 mmHg). Portal pressures drive complications, including ascites, variceal bleeding, hepatic encephalopathy and others.⁸ Prognostic implications of HVPG are mentioned in Table A1. For every 1 mmHg rise in HVPG, risk of mortality increased by 3%.⁹

Assessment of CSPH

NITs for diagnosis of PH and monitoring response to therapy, is an area of intense research. In the Baveno VI consensus conference² it was agreed that NITs are acceptable as the standard of care for ruling in/out PH. Noninvasive methods for diagnosis of CSPH, include indices such as platelet count, aspartate aminotransferase–platelet ratio index, aspartate aminotransferase to alanine aminotransferase ratio, Fib-4, Lok Index, Forns Index, Fibroindex, portosystemic collaterals on cross-sectional imaging, hepatofugal flow on doppler ultrasound studies and others, although these methods have limited sensitivity.¹⁰ These tests have varied sensitivities and have been listed in Table 1. The strengths, limitations of the various modalities are tabulated in Table A2. Studies have found that liver-related clinical events were higher in patients with a higher baseline value of NITs and HVPG. In a study on patients of nonalcoholic steatohepatitis, Sanyal et al found that the relative risk of liver-related clinical event increased by 68% with a 0.5 unit increase in the enhanced liver fibrosis (ELF) score and a 13% increase for every 2 kPa increase in LSM by vibration-controlled transient elastography.¹⁴

Table 1.

NITs and Their Sensitivity to Predict High-Risk Esophageal Varices

NITs	Formula	Cut-off	Sensitivity	Specificity	NPV	PPV
Platelets11		<89,000	54.5%	74.9%	79.9%	48.6%
APRI11	(AST/ULN)/PLT ∗ 100	1.5	56.9%	56.5%	75.8%	35.4%
AAR11	AST/ALT	1.1	70.7%	56.5%	82.2%	59.5%
Fib-411	Age ∗ AST/(PLT ∗ ALT ˇ [1/2])	4.3	70.7%	55.7%	81.8%	40.3%
Lok index11	−5.56 to 0.0089 ∗ PLT + 1.26 ∗ AST/ALT + 5.27 ∗ INR	1.5	71.3%	68.3%	84.2%	50%
Forn’s index11	7.811 − 3.131 × ln (PLT) + 0.781 × ln (GGT) + 3.467 × ln(age) − 0.014 × (cholesterol)	8.8	70.5%	62.4%	82%	58.2%
Fibro-index11	1.738–0.064∗PLT +0.005∗AST + 0.463∗GGT	2.5	50.6%	75.5%	74.7%	50%
LSM12		27.5 kPa	88%	53%	90%	45%
SSM13		>45 kPa	92.1%	64.6%	95.9%	47.3%

AAR, aminotransferase to alanine aminotransferase ratio; APRI, aspartate aminotransferase–platelet ratio index; GGT, gamma glutamyl transferase; PLT, Platelet count; ULN, upper limit of normal; NPV, negative predictive value; PPV, positive predictive value.

Elastography

Elastography-based imaging techniques utilize the intrinsic elastic properties of tissues, which change due to specific pathological or physiological processes. The tissue response can be elicited by providing mechanical stimulus that can be static/quasistatic or dynamic.¹⁵ Based on the mechanical stimulus, elastography techniques can be classified as follows:^15,16

• a)Strain imaging (static/quasistatic stimulus)

strain elastography, acoustic radiation force impulse (ARFI)

• b)
  Shear-wave imaging (dynamic stimulus)

  • •
    Transient:

    • One-dimensional transient elastography (TE), point shear wave elastography (SWE), and 2-dimensional shear wave elastography (2D-SWE)].
  • •
    Continuous:

    • MR elastography.

LSM

LSM by transient elastography (TE), a reliable surrogate for CSPH, is measured in kilopascals (kPa) and correlates with the degree of fibrosis.¹⁷ The Baveno VI consensus, for untreated hepatitis B or hepatitis C cACLD, stated that a LSM of > 20–25 can be used for identification of CSPH. Subsequently, the Baveno VII consensus recommended a LSM ≥ 25 kPa rules in CSPH with a specificity of > 90% in certain etiologies with lower cut-offs in combination with platelet counts, predicting CSPH with a specificity of at least 60%.³

Etiology of cirrhosis should be kept in consideration while interpreting LSM values. A study done by Vizzutti et al¹⁸ in patients with hepatitis C–related cirrhosis, found that LSM correlated well with HVPG up to pressures of 10–12 mmHg. Beyond this, the strength of correlation reduced. While a cut-off of ≥ 25 kPa was sufficient to rule in CSPH in patients of other etiologies, including nonobese NASH, it had a PPV of only 62.8% in obese patients of NASH. Inclusion of body mass index with LSM and platelets for evaluation of CSPH in obese patients improved prediction of CSPH in the ANTICIPATE-NASH model.^19,20

The cut-offs found in various studies are listed in Table 2.

Table 2.

Sensitivity and Specificity of LSM in Diagnosing CSPH

Author, year	Technique	Cut-off	Performance
Kumar A et al, 201721	TE	21.46 kPa	Sens 79% Spec 67%
Salavrakos M et al, 201922	TE	30.6	Sens 81% Spec 94%
Zhu YL et al, 201923	2D SWE	16.1 kPa	Sens 78% Spec 72%
Thiele et al, 202024	2D SWE	Rule out < 14 kPA	Sens 91% Spec 34%
Pon et al, 202120	TE	≥ 25 kPa	LSM ≥25 kPa was the best cut-off to rule in CSPH.
Kumar A et al, 202225	TE	22.8 kPa	Sens 79% Spec 88%
Grgurevic I et al, 202226	2D SWE TE	20.1 kPa 20.2 kPa	Sens 80.5% Spec 94.3% Sens 77.5% Spec 86.1%
Jindal A et al, 202327	TE	≥ 25 kPa	LSM > 25 predicted CSPH in etiologies except HBV
Odriozola et al, 202328	TE	> 27.6 kPa	Sens 44% Spec 96%
Hrisov B et al, 202329	2D SWE	2.49 m/s for HREV	Sens 100% Spec 100%

Spleen Stiffness Measurement (SSM)

Spleen stiffness measurement (SSM) is a reliable marker for diagnosing CSPH. It is known that splenic stiffness increases in PH due to congestion caused by a functional outflow obstruction, leading to enlargement and hyperactivation (hypersplenism) of the lymphoid tissue.³⁰ SSM also offers the advantage of serving as a reliable measure of the hemodynamic changes associated with PH, which is a significant limitation of LSM.³¹ Spleen stiffness (SS) can be assessed using various methods, as outlined in Table A2. The applicability of SSM via TE is limited by its dependence on spleen size. Most patients with severe PH exhibit a significantly stiffer spleen compared to the liver, resulting in many patients presenting falsely low maximal values with the 50 Hz LSM probe.³¹ A 100 Hz probe has been developed, especially for SSM which records a maximum value of 100 kPa.³² A major limitation of SSM measurement is the high failure rates in small spleens due to the difficulty in location. Further, the lack of an XL probe makes evaluation difficult in obese patients and in those with ascites.

Multiple studies have evaluated SSM as a marker for PH and CSPH found that a SSM ≥ 50 kPa correlated well with CSPH.^33,34 The cut-offs for different methods of SSM estimation and its sensitivity and specificity are tabulated in Table 3. The Baveno VII consensus of 2022, laid down that SSM by TE could be used to rule in/rule out CSPH in patients with cACLD due to untreated viral hepatitis. A cut-off of < 21 kPa and > 50 kPa was used to rule out and rule in CSPH, respectively. The addition of SSM to our arsenal of NITs further reduces the need for endoscopy, with those having ≤ 40 kPa by TE have a low probability of having high-risk varices.³

Table 3.

Predictive Performance of SS in the Diagnosis of CSPH

Author, year	Technique	Cut-off	Sensitivity & specificity
Colecchia A et al, 201235	TE	52.8 kPa	Sens 98.5% Spec 97.1%
Takuma Y et al, 201336	pSWE	3.18 m/s	Sens 98.5%
Zykus R et al, 201537	TE	47.6 kPa	Sens 77.3% Spec 79.2
Takuma Y et al, 201638	pSWE	3.1 m/s	Sens 97.1%
Jansen C et al, 201639	2D SWE	35.6 kPa	Spec >90%
Elkrief L et al, 201740	2D SWE	34.7 kPa	Sens 95% Spec 52%
Tseng Y et al, 201841	TE	48.9 kPaa	Sens 76% Spec 100%
Colecchia A et al, 201842	TE	≤ 46 kPa to rule out	Sens 97.8% Spec 43.8%
Zhu YL et al, 201923	2D SWE	25.3 kPa	Sens 85% Spec 79%
Nagai K et al, 202143	TE	100 Hz: 43.8 kPa 50 Hz: 31.8 kPa	Sens 93% Spec 82% Sens 90% Spec 67%
Lantinga et al, 202344	TE	> 26.5 kPa	Sens 83% Spec 82%
Odriozola et al, 202328	TE	> 49.9 kPa	Sens 69% Spec 95%
Jeong-Ju et al, 202445	TE	100 Hz: 38.9 kPa	Sens 92%

pSWE, point shear wave elastography.

^aFor severe portal hypertension.

Magnetic Resonance Elastography (MRE)

Magnetic resonance elastography (MRE) measures liver and SS noninvasively. This magnetic resonance imaging (MRI)-based test uses a vibrating source to generate low-frequency mechanical waves in the tissue. It captures wave propagation through phase-contrast MRI and analyzes the data to visualize tissue stiffness across larger areas, offering a broader perspective than ultrasound.⁴⁶ Normal liver parenchyma has been found to have a cut-off value of < 3 kPa, with variable results for SS, with 1 study having a mean of 3.6 kPa⁴⁷ and another having 4.3 kPa.⁴⁸ Technical failure of MRE has been described due to presence of iron overload, high body mass index or massive ascites.⁴⁹ MRE further has the advantage of fibrosis not being affected by steatosis, as is the case in ultrasound-based techniques.⁵⁰ MRE for prediction of CSPH has had varied cut-offs, sensitivities, and specificities, as tabulated in Table 4. A 2022 study compared liver and SS between 3D, 2D MRE, and SWE, correlating with HVPG for noninvasive CSPH diagnosis. Among 36 patients, SS by 3D MRE showed the strongest correlation with HVPG (r = 0.686, P < .001) and the highest diagnostic accuracy for CSPH (area under curve [AUC] = 0.911), while SWE performed poorly (AUC = 0.583). LS performed worse than SS across all modalities. SS by 3D MRE predicted CSPH better than SWE SS or LS (P ≤ .21).

Table 4.

Predictive Performance of MRE in the Diagnosis of CSPH

Author, year	Technique	Cut-off	Sensitivity & specificity
Matsui et al, 201851	MRE (3 T)	LSM 4.2	Sens 90% Spec 71%
Ambulge et al, 201852	MRE	LSM 4.8	Sens 72% Spec 71%
Dillman et al, 201953	MRE	LSM 2.8 SSM 7.3	Sens 92% Spec 94%
Han et al, 202054	MRE	LSM 6.4	Sens 67% Spec 79%
Jhang et al, 202155	MRE (1.5 T)	SSM 9.53	Sens 84% Spec 73%

MRE carries the advantage of not being operator dependent. However, its limited availability and high costs are significant barriers to widespread use. Additionally, standardization is needed for the cut-offs that define CSPH based on MRE.

Computed Tomography (CT)

Computed tomography (CT) is widely used to evaluate chronic liver disease. It detects splenomegaly, ascites, gastroesophageal varices, and porto-systemic shunts. Radiomics, a quantitative evaluation method, noninvasively assesses PH. In a study by Romero-Cristóbal et al., found that the volume index—liver volume segmental ratio and spleen volume—predicted CSPH with an AUC of 0.83.⁵⁶ Sartoris et al. studied liver surface nodularity to assess PH using liver surface nodularity score and found it superior to spleen volume, diameter and NITs such as aspartate aminotransferase–platelet ratio index and Fib4.⁵⁷ HVPG and CT-based perfusion parameters were compared in another study which found, a positive correlation of hepatic arterial perfusion parameters (r = 0.553, P = .008) with HVPG.⁵⁸ Contrast-enhanced CT-based vascular geometric model was studied for evaluation of HVPG and was found to have a high diagnostic performance.⁵⁹

Artificial Intelligence

Machine learning models are emerging to predict portal pressures from clinical and laboratory data. Deep learning, a form of artificial intelligence, excels at identifying radiological images. The Convolutional Neural Network automates identification of characteristic lesions. A study used Convolutional Neural Network on CT and MRI based images for detection of CSPH compared to HVPG and found both modalities to perform excellently (area under the receiver operator curve (AUROC) 0.888–1).⁶⁰ Another study found CT-based HVPG quantification to outperform other NITs.⁶¹ A recent study evaluated 1232 patients for prediction of severity of PH, based on the 3P and 5P machine learning models concluded that the 5P model was good predictor of CSPH (HVPG ≥ 10 mmHg) and severe PH (HVPG ≥ 16 mmHg) with the combination of 3P/5P and LSM improving the predictive power of the model. The 3P model included parameters platelet count, serum bilirubin and international normalized ratio (INR). The 5P model additionally included cholinesterase, gamma-glutamyl transferase and activated partial thromboplastin time instead of INR.⁶²

Machine learning models are primarily software based, offering the benefit of not being constrained by hardware, thanks to the extensive availability of CT, MRI, TE, cell counters, and biochemistry analyzers. Furthermore, these models deliver results quickly with high reproducibility. In settings without HVPG access, machine learning models can serve as a dependable preliminary diagnostic modality for CSPH.

Radiomics

Radiomics, a branch of artificial intelligence, offers a quantitative approach in medical imaging aimed at enhancing clinicians' data access through sophisticated mathematical analysis. By employing radiological imaging, it generates a substantial dataset enriched through advanced calculations that examine signal intensity distributions and pixel relationships. This method quantifies textural information using analytical techniques from artificial intelligence.^63,64 It is based on the concept that radiological images contain information related to disease processes that the human eye cannot see.⁶⁵ Studies have evaluated CSPH using radiomics models. Iranmanesh P et al. found that the liver/spleen volume ratio and presence of perihepatic ascites were the best predictors of CSPH with sensitivity and specificity of 92% and 79%.⁶⁶ CHESS 1701 utilized both HVPG and Contrast-Enhanced CT scan to develop a noninvasive radiomics model to predict HVPG, called rHVPG, which showed a concordance index (C-Index) of 0.849.⁶⁷ Data of 136 patients have also been used to predict the severity of esophageal varices using CT. Rad score was calculated using right lobe (RL) and left lobe (LL) regions of interest with LL-Rad-score performing better than the RL-Rad-score in predicting severity of esophageal varices. Addition of the cross-sectional area of varices to LL-Rad score improved predictive performance.⁶⁸ Other studies have also found utility in radiomics to predict HVPG with an AUROC of 0.866.⁶⁹

Biomarkers

Liver fibrosis is marked by the excessive accumulation of extracellular matrix (ECM) in the liver, resulting in scarring and eventual cirrhosis. Biomarkers derived from blood can be either direct markers that relate to ECM synthesis or breakdown, while indirect markers indicate liver function. A deeper understanding of the processes that cause fibrosis has paved the way for discovering new biomarkers that hold potential for diagnosing liver fibrosis and CSPH.

The ELF score is a direct serum biomarker score used to assess the degree of fibrosis in the liver and has been used to predict CSPH. The combination of the macrophage activation marker sCD163 and ELF score predicted CSPH in cirrhotic patients with a AUROC of 0.91.⁷⁰ Significantly higher ELF scores were noted in patients with CSPH at a cut-off >11.4 with an AUROC of 0.68.⁷¹

A study published by Benz F et al, found that bone sialoprotein may be used as a marker of PH. Bone sialoprotein, a glycophospoprotein that aids in activation of natural killer cells, neutrophils and macrophages was found to inversely correlate with HVPG.⁷² Chemokine CXCL9 has also been studied for the prediction of complications of PH with higher portal venous levels of CXCL9 being associated with hepatic and renal dysfunction, though it showed no direct correlation with portal pressures.⁷³ Breakdown products of the ECM have also been evaluated for the diagnosis of CSPH with C4M (type IV collagen), C5M (type V collagen), ELM (elastin) and Pro-C5 (type V collagen) levels being significantly elevated in those with CSPH, when compared to those without.^74,75 Osteopontin predicted CSPH with a sensitivity of 75% and specificity of 63% at a cut-off of 80 ng/ml.⁷⁶ One study also looked at and found an inverse correlation of HVPG with miRNA-122 levels.⁷⁷ A study by Liu S et al. examined the diagnostic role of Golgi protein 73 (GP73) and found significant correlation with HVPG (r = 0.45, P < .001). Combining it with INR and platelet (IP73) increased the AUROC to 0.85, with 81.9% sensitivity and 77.9% specificity at a cut-off of 0.⁷⁸

Serum markers for fibrosis are well established. There is a need of more studies for its utilization as a marker of CSPH, along with identification of a cut-offs for the same. If proven to be a reliable marker for CSPH in head-to-head trials, serum biomarkers would herald easy diagnosis even without expensive machinery.

Multimodal Approach

PH prediction is best achieved through combination of NITs. A scoring system that combines liver and SS from 2D SWE and spleen size, with a score >0.34, outperforms individual parameters in predicting high-risk varice (HRV), achieving an AUROC of 0.93.⁷⁹ A combination of Liver stiffness, Spleen size and platelet count, the LS-spleen diameter-to-platelet ratio (LSPS) score, at a cut-off of > 5.5 was found to be able to predict high-risk varices with an accuracy of 90%.⁸⁰ The ANTICIPATE study found that LSPS of >2.65 was associated with 80% risk of CSPH, with values < 1.33 being associated with <5% risk of varices needing treatment.¹⁹ Cho et al., found SS-spleen size-to-platelet ratio risk score (SSPS) to predict HRV at a cut-off of >4.4.⁸¹ Yang et al., developed a prediction model based on LSM and SSM and found that a cut-off of > 0.27 had a sensitivity of 100%, specificity of 82% in predicting HRVs.⁸² Baveno VII states LSM of ≤ 15 kPa along with platelet count of ≥1.5 lakh/cumm, or SSM ≤ 40 kPa safely ruled out CSPH, while a LSM of ≥ 25 kPa ruled in CSPH.

A study published by Jachs et al., evaluated the utility of sequential Baveno VII–von Willebrand factor antigen–platelet ratio (VITRO) algorithm for stratifying patients in the “grey zone”, which were unclassifiable based on Baveno VII alone and found that CSPH could be ruled-in in 73% and ruled out in 70% of the patients.⁸³ Another study by the same author evaluated the prognostic capability of ANTICIPATE ± NASH and VITRO to HVPG for prediction of decompensation and CSPH, and found that the performance of ANTICIPATE ± NASH-CSPH-probability (at a cut-off >60%) and VITRO (cut-off of ≥2.5) was comparable to HVPG ≥10 mmHg.⁸⁴

Combination approach for diagnosis of CSPH has been found to have a higher accuracy at predicting CSPH. The most commonly utilized are LSM, SSM and platelet count which has been prescribed by the Baveno VII consensus. Other prediction models still require validation before utilization in clinical practice.

A brief overview of these various modalities has been tabulated in Table 5.

Table 5.

Summary of NITs Used for CSPH

Modality	Strengths	Limitations
LSM	-Predicts CSPH with good specificity (≥90%) at a cut-off ≥25 kPa. -Sensitivity of 79% and specificity of 88% in diagnosing CSPH (AUC 0.91) (meta-analysis). -High specificity and sensitivity (≥60%) when combined with platelet count.	-Cannot estimate HVPG value accurately (r 0.36–0.86). -Decreases in precision for certain etiologies like NASH, particularly in obese patients. -High failure rates with high BMI, ascites, or space occupying lesions affecting interpretation.
SSM	-Offers insight into hemodynamic changes due to PH, addressing a limitation of LSM. -Can predict high-risk varices (SSM ≥ 50 kPa) with specificity > 90% in certain studies. -Superior to LSM for diagnosing varices needing treatment	-Dependent on spleen size; not suitable for small spleens or obese patients -Limited by probe capabilities (standard probe maxes out at 100 kPa, mSS up to 150 kPa). -Lack of XL probe for larger patients
MRE	-Noninvasive and operator-independent, measures liver and SS. -Correlates well with HVPG, more accurate than TE in diagnosing liver fibrosis. -Predicts high-risk varices with 85.9% sensitivity and 72.1% specificity for LS; 84.6% sensitivity and 78.7% specificity for SS	-High cost and limited availability. -Prone to technical failure (eg, iron overload, high BMI, or massive ascites). -Standardization of cut-offs for CSPH is still needed.
CT	-Widely used for chronic liver disease assessment -Liver surface nodularity score correlates with HVPG (r = 0.75), shows high specificity (87%) and sensitivity (80%) for CSPH. -Predictive performance for HVPG using perfusion parameters (r = 0.553) seen in some studies.	-Lack of sensitivity in some cases for early PH stages. -Requires contrast-enhanced CT, which may not be suitable for all patients.
Artificial intelligence (AI)	-Machine learning models (eg, Convolutional Neural Network) show excellent performance in detecting CSPH from CT/MRI images (AUC 0.888–0.998). -Fast results with good reproducibility, enhancing clinical workflow in centers without HVPG access. -Improves diagnostic yield when combined with other modalities (eg, 2D-SWE, LSM, 3P/5P models).	-Software-dependent; requires significant computational power and data for training. -Needs large datasets for training and validation, and may not be universally applicable across populations. -Limited by the quality of available imaging data and algorithms.
Radiomics	-Quantifies textural information in imaging, offering a sophisticated method for assessing CSPH (eg, liver/spleen volume ratio, ascites). -High predictive performance for CSPH (AUROC 0.91, sensitivity 92%) using CT radiomics. -Can predict esophageal varices severity with high accuracy (C-index 0.95).	-Requires high-quality imaging and sophisticated software; may be resource-intensive. -Generalizability may be limited, and may require external validation in different cohorts. -Needs further studies for standardization of techniques and cut-off values.
Biomarkers	-Noninvasive serum markers for fibrosis (eg, ELF, GP73) provide additional diagnostic utility. -ELF and GP73 correlate with CSPH and HVPG with good specificity (eg, GP73 sensitivity 81.9%, specificity 77.9%). -Emerging biomarkers (eg, BSP, CXCL9, miRNA-122) show promise for PH and CSPH prediction.	-Limited validation for CSPH; most biomarkers are still in research stages. -Biomarker performance varies, with sensitivity for CSPH prediction being moderate (eg, ELF score sensitivity 42%). -Cut-offs for biomarkers like ELF are still not standardized. -Inconsistent results across different studies, requiring more research and validation.
Multimodal approach	-Combining tests (eg, LSM, SSM, platelet count) offers higher predictive accuracy for CSPH and esophageal varices. -Models like LSPS and SSPS improve prediction of high-risk varices with high AUROC (0.93). -Multimodal scoring systems help in stratifying patients who may be in the “grey zone” or unclassifiable by single tests.	-Some multimodal models still need further validation. -Risk of overfitting or complexity in clinical practice. -Limited evidence in nonviral etiologies. -Requires more studies to optimize and standardize cut-offs.

AUE, area under curve; BMI, body mass index; BSP, bone sialoprotein.

NSBBs for CSPH

NSBBs have been effectively used in patients with CSPH to prevent variceal bleeding, thereby reducing mortality.^2,85 These have also increased decompensation-free survival,^5,86 reduced the risk of spontaneous bacterial peritonitis⁸⁷ and hepatocellular carcinoma.⁸⁸ NSBB are recommended for both primary and secondary prophylaxis of variceal bleeding, targeting an HVPG reduction of 10% for primary and 20% for secondary prophylaxis, with an absolute value preferably below 12 mmHg.² Propranolol, Nadolol, and Carvedilol are the frequently used NSBB, with Carvedilol providing the greatest reduction.⁸⁹ However, due to the lack of a proper and practical measurement method for HVPG, a heart rate reduction to 55–60 beats per minute is regarded as adequate beta-blockade and optimal response. Unfortunately, nonresponders have a high risk of variceal bleeding. Therefore, it is crucial to accurately and promptly monitor the response to NSBB therapy. Although HVPG is safe and well tolerated,^8,90 its invasiveness, limited availability restrict its use and clinical application to only select centers, highlighting the need for the development of NITs for the monitoring of NSBB therapy.

Assessing Response to Beta-Blockers in Patients with CSPH

Liver stiffness and SS measurement

LSM and SSM have been used individually for the diagnosis and monitoring of CSPH.³ As NSBBs are prescribed to patients with CSPH, there is a need to reliably and noninvasively monitor treatment response. Studies have had varied results, as highlighted in Table A2. LSM does not correlate with CSPH at pressures >12 mmHg, while SSM is a dynamic marker of the change in portal pressures. A combination of LSM and SSM could help in monitoring patients with CSPH at varied pressures. Whether this combination can be used to positively correlate with the response to NSBB therapy is an area of research.

Elastography measures the stiffness of the liver tissue, defined by Young’s modulus (E), which is product of the density of the liver and the velocity of shear waves, where ⍴ is the density and V_s is the shear velocity.⁹¹

$$E=3\rho V_s^2$$

The concept of elastography in determining portal pressures is based on the following factors:

• (a)Increase in liver stiffness due to fibrosis
• (b)Splenic congestion due to an increase in splanchnic venous return and portal pressures

In patients who develop cirrhosis, there is an increase in the resistance detected by LSM, which holds true for estimating portal pressure until the increase in portal pressures is also influenced by flow, since pressure is the product of resistance and flow. An increase in flow causes splenic congestion, as determined by SSM. Upon initiation of beta-blockers, portal flow is modulated, thereby reducing splenic congestion, making SSM a good marker for the response to beta-blockers. On the other hand, once beta-blockers are initiated, portal pressures are driven more by hepatic resistance rather than portal flow, thereby improving correlation with LSM (Figure 1).

Figure 1.

Concept diagram showing (A) normal physiology, (B) hemodynamic changes in cirrhosis and portal hypertension, and (C) effect of NSBB on portal hemodynamics. PV, portal vein; SMV, superior mesenteric vein; SV, splenic vein; IMV, inferior mesenteric vein.

Change in LSM (Δ LS) has been assessed to reflect HVPG response to NSBB in 2 studies. The first was by Reiberger et al, a study of 122 patients, which found that the correlation of LSM (by TE) with HVPG was noted more with HVPG responders (r = 0.864) as compared to nonresponders (r = 0.535) when baseline HVPG was ≥ 12 mmHg. Δ LS; however, were similar in both HVPG responders and nonresponders, making it an unreliable measure to assess the dynamicity of PH.⁹² Another study by Choi at al., used 2D-SWE to assess the correlation of Δ LS with HVPG in 23 patients and found a strong correlation between change in HVPG (Δ HVPG) and Δ LS (r = 0.863).⁹³

Changes in SSM (ΔSS), similarly has been assessed to reflect change in portal pressures after beta-blockers. In a study by Kim et al., ΔSS (measured by ARFI) was found to parallel the HVPG changes in patients on NSBBs prophylaxis. This study employed a maximum dose of 25 mg per day carvedilol for the derivation cohort and 12.5 mg per day for the validation cohort. It demonstrated that ΔSS, (Odds ratio [OR] 0.039; 95% confidence interval [CI] 0.008–0.135; P < .0001) as an NIT, could help to identify hemodynamic responders to NSBB therapy, in patients of cirrhosis with HRV. They derived a prediction model (0.0490−2.8345 × ΔSS), and tested it with an external validation cohort where the sensitivity, specificity, and accuracy for predicting hemodynamic response with this ΔSS cut-off were 0.848, 0.733 and 0.794.⁹⁴ Marasco et al. in their small study of 20 patients used propranolol in 9 patients and carvedilol in 11 patients (maximum dose of 80 mg per day and 12.5 mg per day, respectively) and found that ≥ 10% reduction in SSM accurately correlated with HVPG reduction ≥ 10% (sensitivity of 100% and specificity of 60% (AUROC 0.973)) and thereby identified responders. ΔLS did not correlate with ΔHVPG.⁹⁵ In contradiction to the previous 2 studies, a prospective trial of 31 patients by Binzberger et al. did not find any correlation between ΔLS or ΔSS (by ARFI) with ΔHVPG (r = 0.4 and r = −0.29, respectively) to initiated NSBB therapy. The maximum dose of Carvedilol used was 25 mg per day.⁹⁶ Danielsen et al. measured LS and SS using 2D MR elastography for assessment of response to intravenous propranolol (maximum dose of 15 mg). They demonstrated no correlation between changes in the elastography parameters and response to therapy.⁹⁷ A recent study by Giuffrè used a noninvasive machine learning–based approach to determine NSBB response. In their study, participants with HRV were enrolled and received a maximum carvedilol dose of 12.5 mg per day with serial follow-up elastography at 3, 6, and 12 months. Response was defined as downstaging of the grade of varices at 12 months while nonresponse was upstaging in the grade of varices or an episode of variceal bleed. They found that a decrease in SS by >11.5%, LS by >16.8% and heart rate (HR) by >25.3% at 3 months from baseline were the best predictors of NSBB response (AUROC 0.82, 0.85, and 0.83 respectively), with a model incorporating all 3 having highest discrimination with a AUROC of 0.96, sensitivity off 94.2%, and specificity of 100%.⁹⁸ Studies showing correlation of ΔLS or ΔSS with Δ HVPG are tabulated in Table 6.

Table 6.

Studies Correlating Change in Liver Stiffness and SS With Changes in HVPG

Author	Sample size	Compared variables	Significant finding	Conclusion
Reiberger et al., 201291	122	LSM by TE vs HVPG	LS correlation with portal pressure in patients under NSBB treatment: R = 0.764; P < .0001 The improvement of the correlation of LS and HVPG was mainly noted in NSBB responders R = 0.864; P < .0001	LS, is unreliable to assess changes in HVPG.
Choi et al., 201492	23	LSM by 2D SWE vs HVPG	LS significantly correlated with HVPG during first and second measurements (r = 0.501, r = 0.527) ΔHVPG and ΔLS (r = 0.863, P < .001).	ΔLS strongly correlated with ΔHVPG. Patients with a decrease in LS of more than 2% tended to show improvement of portal hypertension.
Kim et al., 201993	106	LSM and SSM by ARFI vs HVPG	ΔSS was the only significant predictor of hemodynamic response (odds ratio 0.039; 95% confidence interval; P < .0001)	Dynamic changes in SS exhibited good performance in predicting hemodynamic response to NSBB prophylaxis in patients with high-risk esophageal varices.
Marasco et al., 202094	20	LSM and SSM by TE vs HVPG	≥ 10% reduction in SSM accurately correlate with a HVPG reduction ≥ 10% (sensitivity 100% and specificity 60% and AUROC 0.973)	SSM reduction ≥ 10% is reliable for the assessment of hemodynamic response to NSBB as primary prophylaxis.
Danielsen et al., 202196	36	LS and SS by 2D MR elastography vs HVPG	IV NSBB significantly decreased SS by 6.9% (CI: 3.5–10.4, P < .001) ΔSS was not related to the HVPG response (P = .75).	No correlation between changes in the elastographic parameters and response to beta-blocker therapy.
Binzberger et al, 202295	31	LSM and SSM by ARFI vs HVPG	ΔLS, ΔSS and ΔHVPG showed no correlation (r = 0.4, −0.298; P = .114) AUC-ROC: ΔLS - 57.35% ΔSS - 42.23%	No significant correlation between ARFI of the spleen with the HVPG both at baseline and at the follow-up examination Very low ability of ΔLSM and ΔSSM to identify a response to medication compared to ΔHVPG

ARFI, acoustic radiation force impulse; AUC-ROC, area under the receiver operating characteristic curve; AUE, area under curve; HVPG, hepatic venous pressure gradient; LS, liver stiffness; LSM, liver stiffness measurement; MR, magnetic resonance; NSBB, non-selective beta blockers; SS, spleen stiffness; SSM, spleen stiffness measurement; SWE, shear wave elastography; TE, transient elastography.

Predicting Risk of Bleed and Rebleed

Presence of and bleeding from esophageal varices portends a poor prognosis in patients of PH. HVPG is the gold standard for assessing bleeding risk, with values above 12 mmHg predicting bleed. Few studies have used LSM, SSM as a predictor for variceal bleed or rebleed. A study by Buechter et al, retrospectively studied the incidence of variceal bleed among patients who had undergone LSM, SSM and found that both predicted risk of variceal bleed. In their study, they found a LSM cut-off of 20.8 kPa had a sensitivity of 84% and specificity of 80% in predicting the risk of bleed, while a SSM cut-off of 42.6 kPa had a sensitivity and specificity of 89% and 64% respectively. Combining the 2 had a sensitivity of 100% but specificity of only 55%.⁹⁹ In a study by Agarwal S et al,¹⁰⁰ HVPG and LSM were measured in patients presenting with variceal bleed and they found, over a 4-year period that postbleed LSM > 30 kPa and HVPG > 15 mmHg identified patients at high risk of rebleed. LSPS has been used to predict bleed and rebleed. In a study by Kim et al, LSPS was found to predict risk of variceal bleed with an AUROC of 0.9 when LSPS was > 6.5.¹⁰¹ A study by Wang et al, found that a high LSPS (>0.18) predicted rebleed in patients who had a prior episode of bleed.⁵⁹ A study published by Yoo Jeong et al., found a SSM cut-off of 38.9 kPa to predict future incidence of variceal bleed in their population of 257 patients. In a median follow-up of 18 months, 4 patients with SSM ≥ 38.9 kPa bled, while none ≤ 38.9 kPa did.⁴⁵

There is a lack of use of elastography to forecast the risk of variceal bleeding and re-bleeding. Assessing thresholds through NITs to predict these decompensating events could alleviate the substantial 6-week mortality rate associated with it.

Lacunae/Future Direction

With advancement in technology, patient and physician preference is leaning toward a noninvasive approach toward diagnosis and treatment. The wide spectrum of modalities for assessment of CSPH, while encouraging, also make it challenging due to lack of standardizations. There is a need for consolidating the literature and formation of guidelines based on technology which is widely available and easily interpretable. Though NITs are an area of intense research, studies on NSBB response are few and far, each having used different modalities for the assessment of LS and SS. Further, the published data are in the form of small sample sizes, owing in part to the reluctance of patients for a repeat HVPG measurement. Larger, prospective trials are needed to be able to define cut-offs for hemodynamic response to beta-blockers. Combination of LSM and SSM for diagnosis of CSPH and monitoring response to treatment, may be assessed for higher accuracy. Evaluation of these parameters in cirrhosis of different etiologies would widen its utilization in the field of hepatology. Using a single, widely available and clinically applicable modality for assessment of LSM and SSM, such as TE, would permit the accumulation of more homogenous data, thereby facilitating a more streamlined formulation of guidelines/expert consensus. Head-to-head comparisons are a must for the selection of the best modalities in each clinical scenario. Even today, NITs are used as a screening test or a guidance for patients to undergoing endoscopy. Future research should focus on a more definitive diagnosis based on NITs, permitting initiation of therapies based on it alone and omitting the need for screening endoscopies.

Conclusion

While LSM and SSM have been extensively validated for the diagnosis of CSPH, their role in monitoring NSBB therapy is getting recognized. However, this would require large prospective trials before it can be recommended for clinical practice. Combination of NITs may perform better than any standalone test.

Though the journey is long and the goal is distant, we anticipate that NITs will become an alternative to HVPG in the near future. If proven reliable, they would become a useful tool for serial and dynamic measurement of reduction in portal pressure to predict treatment response.