Biomarkers and imaging studies to predict portal hypertension and varices

Watch a video presentation of this article

Watch the interview with the author

Abbreviations

APRI

aspartate aminotransferase‐to‐platelet ratio index

ARFI

acoustic radiation force impulse

ARFI‐LSM

ARFI measured LSM

AUC

area under the curve

AUROC

area under the receiver operating characteristic

BA

biliary atresia

CLD

chronic liver disease

EV

esophageal varices

HCV

hepatitis C virus

HVPG

hepatic venous pressure gradient

LEV

large esophageal varices

LSM

liver stiffness measurement

LSPS

liver stiffness‐spleen diameter‐to‐platelet ratio score

MRE

magnetic resonance elastography

mSS

modified splenic stiffness

NPV

negative predictive value

PH

portal hypertension

Plt/Spl

platelet count/spleen diameter

PPV

positive predictive value

P/SZC

platelet/spleen‐length z score ratio

SSM

spleen stiffness measurement

TE

transient elastography

TE‐LSM

TE measured LSM

The development of clinically significant portal hypertension (PH) is a major milestone in the natural history of chronic liver disease (CLD) in adults and children. The gold standard assessment for portal pressure is hepatic venous pressure gradient (HVPG). PH in adults is diagnosed when HVPG is ≥6 mm Hg. If the HVPG is ≥10 mm Hg, patients have an increased risk for development of esophageal varices (EVs) and variceal bleeding, which is associated with an increased mortality risk.1 HVPG measurement is invasive and requires expertise not routinely available at every center. For these reasons, the Baveno V panel recommended development of noninvasive methods to evaluate for PH/EVs.2 Although there have been many studies for noninvasive assessment of liver fibrosis, a body of research on noninvasive tests to predict PH/EVs is emerging.

Liver Stiffness Measurement

Transient elastography (TE) is an ultrasound‐based tool that measures liver stiffness. Liver stiffness measurements (LSMs) have been shown to reflect hepatic fibrosis and have been studied to evaluate for PH/EVs.3, 4

In adults with chronic hepatitis C virus (HCV) infection, TE measured LSM (TE‐LSM) was strongly correlated to HVPG (r = 0.81; P < 0.0001).5 To detect HVPG ≥ 10 mm Hg, LSM had an area under the receiver operating characteristic (AUROC) of 0.99, cut point of 13.6 kPa, negative predictive value (NPV) 92%, and sensitivity 97%. For HVPG ≥ 12 mm Hg, the cut point, AUROC, NPV and sensitivity were 17.6 kPa, 0.92, 91%, and 94%, respectively.

There have been fewer studies of LSM to predict PH/EVs in children. Chang et al.6 studied 49 children with biliary atresia (BA), status post Kasai portoenterostomy, who underwent TE‐LSM and esophagogastroduodenoscopy. The mean LSM was higher in the 30 patients with EV than without EV (21.35 versus 9.75 kPa; P < 0.001). To detect an EV or gastric varices, LSM had an AUROC of 0.88, cut point of 9.7 kPa, sensitivity 97%, and specificity 80%. To detect grade III EVs, the cut point was 35.3 kPa (sensitivity 75%, specificity 93%). Chongsrisawat et al.7 also studied 73 BA patients status post Kasai procedure and found the mean LSM was higher in those with varices (37.72 versus 10.97 kPa; P < 0.001). Malbrunot‐Wagner et al.8 reported similar findings in 18 children with cystic fibrosis to predict EV (22.4 versus 7.9 kPa; P = 0.01).

Although TE is easy to use, it requires a dedicated machine, has operator and subject limitations, and is subject to other confounding factors.9 LSM is also measured by acoustic radiation force impulse (ARFI) and magnetic resonance elastography (MRE), which have also been studied to predict PH/EVs.

ARFI is an ultrasound‐based technique within a conventional ultrasound system. In a study of 166 adults with CLD, ARFI measured LSM (ARFI‐LSM) and ARFI‐spleen stiffness measurement (SSM) were compared with TE‐LSM and FibroTest to detect large esophageal varices (LEV).10 Although the AUROCs were similar among all methods (0.5‐0.58), multiple logistic regression analysis showed that ARFI‐SSM was superior to ARFI‐LSM to detect large EVs.

MRE has emerged as another reliable technique to assess LSM. MRE has the technical advantage over TE because a larger measurement area can be chosen. Although MRE association to liver fibrosis is well published, MRE assessment for PH/EVs is limited. In a study of 36 adults with cirrhosis, Ronot et al.11 performed HVPG, MRE‐LSM, and MRE‐SSM. Unlike MRE‐SSM, the overall LSM viscoelastic parameters did not identify patients with severe PH (HVPG ≥ 12 mm Hg) or high‐risk EV with reasonable performance.

Splenic Stiffness Measurement

Splenomegaly, an important clinical sign of PH, is a result of splanchnic congestion, tissue hyperplasia, and fibrosis.12 SSM has also been proposed as a measure of liver fibrosis and PH. Like LSM, SSM can be measured by several imaging modalities, including TE, ARFI, and MRE. Unlike LSM, TE‐SSM may require additional ultrasound assistance to determine spleen location/depth. Colecchia et al.13 compared TE‐SSM with LSM, platelet count/spleen diameter (Plt/Spl), and liver stiffness‐spleen diameter‐to‐platelet ratio score (LSPS) in 100 adult patients with HCV cirrhosis. Although all methods had significant positive correlation with HVPG, SSM had the strongest correlation (r = 0.885; P = 0.0001). In another study of 191 adults, SSM was higher in patients with EV than without EV (63.39 versus 47.78 kPa; P < 0.0001), with cut point of 46.4 kPa and AUROC of 0.781. Using LSM cut point greater than 19 kPa and SSM cut point greater than 55 kPa, the accuracy to detect EV improved to 88.52%.14

Goldschmidt et al.15 measured TE‐SSM in 62 children with CLD and 31 normal controls. SSM was feasible in children, with a greater success rate in patients with splenomegaly (90.5%) than in those with normal spleen size (66.7%). The median SSM was higher in those with splenomegaly than in those without (62.96 versus 18.4 kPa; P < 0.001). SSM was higher in patients with varices (75 versus 24 kPa; P < 0.01). In addition, variceal bleeding was not reported in children with SSM less than 60 kPa.

Because the spleen is stiffer than the liver, one limitation of TE‐SSM is that the standard machine cannot measure beyond 75 kPa. Calvaruso et al.16 proposed modified splenic stiffness (mSS) using TE software modified by the manufacturer to allow an SSM of 1.5 to 150 kPa. In 112 adults with HCV cirrhosis, the mSS cut point to detect LEV was 54 kPa, AUROC 0.82, sensitivity 80%, and specificity 70%, allowing for improved diagnostic characteristics.

ARFI can also obtain SSM at the same time as LSM and without additional ultrasound guidance. In a meta‐analysis of 12 adult studies of SSM using TE, ARFI, real‐time elastography, and virtual touch tissue quantification, the overall ability to detect EV was AUROC of 0.86, sensitivity 78%, and specificity 76%.17 Recently, Tomita et al.18 studied ARFI‐LSM and ARFI‐SSM in 28 children with BA. To diagnose high‐risk EV, ARFI‐LSM had an AUROC of 0.9, cut point of 1.94 m/second, sensitivity 100%, specificity 76.9%, and accuracy 81.3%, whereas SSM had an AUROC of 0.79, cut point of 3.14 m/second, sensitivity 100%, specificity 69%, and accuracy 75%. When LSM and SSM were combined, the performance improved (AUROC 0.92, cut point 5.13 m/second, sensitivity 100%, and specificity 84.6%).

Table 1.

Comparison of Modalities to Predict Portal Hypertension and Varices in Children

Diagnostic Study	Advantages	Disadvantages
Biomarkers	Available in all centers	Some indices require both serum and radiological measures
	No special training required
	Excellent reproducibility
TE	Most extensively studied	Requires a dedicated machine
	Can be performed in any setting	Narrow measurement field
	Any operator can be trained (i.e., does not require radiology involvement)	Measured area cannot be selected by operator
	Good reproducibility	SSM may require ultrasound guidance and may require higher stiffness values not available on standard machine
		Reproducibility/accuracy may be affected by operator's training experience, obesity, and ascites
		May be affected by cholestasis, inflammation, postprandial state, congestion
ARFI	Integrated into conventional ultrasound machine	Not available at every center
	Measurement area is larger than TE and can be chosen by operator	Requires radiology involvement
	Can be performed in patients with ascites, high body mass index	May be affected by cholestasis, inflammation, postprandial state, congestion
	Can obtain liver and spleen stiffness during the same acquisition
	Good reproducibility
MRE	Measured area is larger and can be chosen by operator	Has not been studied in children with PH/varices
	Liver and spleen stiffness can be obtained at the same time	Not available at every center
	Compression waves produced by external drivers penetrate within tissues better than shear waves	Requires radiology involvement
	Good reproducibility	Long acquisition time may not be suitable for infants/young children without sedation or anesthesia
		Difficult in claustrophobic patients
		Cannot be performed for patients with metal devices
		May be affected by cholestasis, inflammation, post‐prandial state, congestion

Unlike in LSM, very limited data on MRE‐SSM are available to predict PH/EVs. In adults, MRE is feasible and has good diagnostic performance to predict HVPG ≥ 12 mm Hg (AUROC 0.73‐0.81; P = 0.01‐0.0035) and high‐risk varices (AUROC 0.76‐093; P = 0.023‐0.001).11 Although MRE is reliable, the cost, accessibility, and need for sedation make assessment in children more challenging.

Biomarkers

Biomarkers offer hope for noninvasive, low‐cost assessments available in all centers. As with noninvasive imaging, many biomarkers were originally used to predict liver fibrosis and are now being proposed to identify PH/EVs.

In a large, multicenter study of 610 cirrhotic adults, platelet count, aspartate aminotransferase‐to‐alanine aminotransferase ratio, aspartate aminotransferase‐to‐platelet ratio index (APRI), Forns index, Lok index, FIB‐4, and FibroIndex were measured.19 Although the Lok (AUROC 0.70) and Forns indexes (AUROC 0.66) best predicted clinically significant EVs, none of the biomarkers performed well enough to clinically identify EVs (NPV < 85%, positive predictive value [PPV] < 60%). When combined, the Lok (cut point 1.5) and Forns indexes (cut point 8.8) had improved AUROC 0.8, sensitivity 84%, specificity 73.3%, PPV 57.5%, and NPV 91.4% to predict clinically relevant EVs.

Comparatively, there are few studies of biomarkers to predict PH/EVs in children. In a study of 33 patients with BA, Colecchia et al.20 found that 15 patients with EVs had statistically higher LSM, APRI, Plt/Spl, and LSPS than those without EVs. Both APRI and LSPS had good diagnostic performance with cut point, area under the curve, sensitivity, and specificity (APRI: cut point > 0.96, 0.88, 86%, 81%; LSPS: cut point ≥ 9.2, 0.96, 91%, 92%, respectively). However, with multivariate analysis, only LSPS was predictive for EVs.

In a retrospective study, Gana et al.21 measured platelet/spleen‐length z score ratio (P/SZC), platelet count, international normalized ratio, APRI, and albumin in 51 children with CLD/PH. A clinical prediction rule was developed using multivariate logistic regression with platelets, SZC, and albumin to have high sensitivity and NPV to predict those who likely had EVs. The cutoff, AUROC, sensitivity, specificity, PPV, and NPV were 130, 0.93, 94%, 81%, 0.83, and 0.94, respectively.

Recently, TE‐LSM was compared with APRI and P/SZC to predict liver fibrosis and EVs in 99 children with CLD.22 For the prediction of EVs, APRI showed the best AUROC of 0.832 in comparison with LSM, SZC, or P/SZC with AUROCs of 0.818 (P < 0.001), 0.795 (P = 0.001), and 0.760 (P = 0.004), respectively. APRI had a cutoff of 0.6, sensitivity 100%, and NPV of 100%. The LSM had a cutoff of 8.8 kPa and sensitivity 93%, whereas P/SZC had a cutoff of 0.76, sensitivity 77%, and specificity 73%.

Summary

The development of PH/EVs is a significant clinical development in CLD and is associated with increased mortality. Finding reliable, noninvasive biomarkers and imaging studies could help minimize invasive procedures and help identify patients at risk. Liver/SSMs and biomarkers are promising tools to assess for PH/EVs in adults and children. Larger studies are needed, particularly in children, to better define the use of these modalities in clinical situations.