Key messages
-
Liver stiffness measurement measured by vibration-controlled transient elastography with cut-off of 25 kPa may detect clinically significant portal hypertension (CSPH) as previously suggested by guidelines.
-
The reliability of results is low due to problems with the methods of the currently available studies.
-
More high-quality studies are needed before elastography techniques can be used routinely to assess portal hypertension.
Why is improving the diagnosis of clinically significant portal hypertension (CSPH) important?
Portal hypertension is high blood pressure in the veins of the liver, often seen in people with chronic liver disease. This condition can lead to serious complications like fluid build-up in the abdomen (ascites), yellowing of the skin (jaundice), kidney failure, confusion, and dangerous bleeding from veins in the oesophagus. Early diagnosis is important to start treatment and to prevent these complications. The standard reliable test, called the hepatic venous pressure gradient, is invasive (involves inserting a tube into a vein), expensive, and not widely available. Therefore, we need non-invasive tests, like elastography, a type of imaging test that can measure how stiff the liver or spleen is, to indirectly assess the portal pressure.
What did we want to find out?
We wanted to know how accurate different elastography techniques and their combinations are for diagnosing CSPH in people with chronic liver disease.
What did we do?
We searched for studies of various elastography techniques that measure liver or spleen stiffness, using either specialised machines such as vibration-controlled transient elastography, or standard ultrasound and magnetic resonance machines like point shear wave elastography, two-dimensional shear wave elastography, and magnetic resonance elastography, to find out how well these tests can detect CSPH at different thresholds (cut-off values).
Main results
We included 47 studies with a total of 7817 participants. On average, about 63 out of every 100 people in these studies had CSPH. The most studied test was liver stiffness measurement by vibration-controlled transient elastography. There was a lack of studies looking at other elastography techniques or combinations of tests. There were also few studies on spleen stiffness.
Liver stiffness measurement by vibration-controlled transient elastography
We looked at 27 studies involving 3818 people. These studies used different thresholds to decide when liver stiffness measurement results suggest the presence of CSPH.
When the test is used to catch nearly everyone who has CSPH (90 out of 100 people), it may also correctly suggest that 76 out of 100 people actually do not have CSPH.
On the other hand, if the test is used in a way that avoids false results in people who do not have CSPH (correctly ruling out 90 out of 100 without the condition), it will correctly find CSPH in about 73 out of 100 people who actually do have it.
Using a threshold of 25 kPa for liver stiffness (based on 9 studies with 1553 participants), the test correctly identified 62 out of 100 people who had CSPH. However, it also correctly ruled out CSPH in almost everyone (94 out of 100 people) who did not have the condition.
While liver stiffness measurement by vibration-controlled transient elastography with a 25 kPa threshold might detect CSPH, the overall low quality of the studies means that these results are not very reliable. More high-quality research is needed before elastography can be recommended for assessing portal hypertension.
What are the limitations of the evidence?
Most studies had limitations, such as problems with how people were selected for inclusion in the study and how thresholds were defined. Only 2 out of 47 studies were assessed as having no major concerns. This lowers our confidence in the accuracy of the results.
How up-to-date is this evidence?
The evidence is current to 8 April 2024.
Read the full abstract
Clinically significant portal hypertension (CSPH) in chronic liver disease (CLD) is a key driver of decompensation, with severe portal hypertension (SPH) leading to severe complications like oesophageal bleeding. Early detection is essential for timely treatment. The current gold standard for assessing portal pressure is hepatic venous pressure gradient (HVPG), an invasive and costly procedure with limited availability. Non-invasive alternatives are increasingly needed to estimate portal pressure (e.g. by liver and spleen stiffness measurement) and guide treatment. However, the diagnostic accuracy of vibration-controlled transient elastography (VCTE), point shear wave elastography (pSWE), two-dimensional shear wave elastography (2D-SWE), and magnetic resonance elastography (MRE) remains unknown.
Objectives
Primary objectives: to assess the diagnostic accuracy of liver stiffness measurement (LSM) and spleen stiffness measurement (SSM), as well as in combination, measured by any elastography technique (VCTE, pSWE, 2D-SWE, or MRE) in the detection of CSPH in adults with CLD; and to compare the diagnostic accuracies between the individual tests. We will regard a combination of tests as positive when at least one is positive.
Secondary objectives: to assess the diagnostic accuracy of LSM and SSM, as well as in combination, measured by any elastography technique (VCTE, pSWE, 2D-SWE, or MRE) in the detection of SPH in adults with CLD; and to investigate sources of heterogeneity in the results.
Search strategy
We searched the Cochrane Hepato-Biliary Group Controlled Trials Register, the Cochrane Hepato‐Biliary Group Diagnostic Test of Accuracy Studies Register, CENTRAL, MEDLINE ALL Ovid, Embase Ovid, LILACS, Science Citation Index – Expanded (Web of Science), and Conference Proceedings Citation Index – Science (Web of Science) until 8 April 2024. We applied no restrictions on language or document type.
Selection criteria
We included studies that evaluated the diagnostic accuracy of LSM and SSM either alone or in combination, as measured by different elastography techniques (VCTE, pSWE, 2D-SWE, or MRE), for the diagnosis of CSPH and SPH in adults with CLD. We only considered studies with cross-sectional design using HVPG measurement as the reference standard.
Data collection and analysis
Two review authors independently screened studies, extracted data, and assessed the risk of bias and applicability concerns using the QUADAS-C tool. In the case of different cut-off values, we used the hierarchical summary receiver operating characteristic (HSROC) model to meta-analyse data (sensitivities and specificities) and to estimate a summary ROC (SROC) curve. In the case of common cut-off values, we used the bivariate model. We presented uncertainty of the accuracy estimates using 95% confidence intervals (CIs).
Main results
We included 47 studies (7817 participants). The most evaluated index test was LSM by VCTE for CSPH (27 studies; 3818 participants).
We judged only two studies at low risk of bias for all domains. Most showed high risk in the index test domain due to lack of prespecified thresholds and high concern for applicability in patient selection. Overall, the certainty of evidence was very low.
Due to varying thresholds across studies, we used an HSROC model to obtain summary estimates. The main findings for detecting CSPH are below.
LSM by VCTE (27 studies, 3818 participants):
Sensitivity 72.6% (95% CI 60.0% to 82.5%) at fixed specificity of 90%
Specificity 75.9% (95% CI 63.5% to 85.1%) at fixed sensitivity of 90%
SSM by VCTE (6 studies, 391 participants):
Sensitivity 72.9% (95% CI 27.1% to 95.1%) at fixed specificity of 90%
Specificity 80.6% (95% CI 64.1% to 90.7%) at fixed sensitivity of 90%
SSM by pSWE (4 studies, 248 participants):
Sensitivity 84.1% (95% CI 8.4% to 99.7%) at fixed specificity of 90%
Specificity 86.1% (95% CI 40.4% to 98.3%) at fixed sensitivity of 90%
LSM by 2D-SWE (10 studies, 767 participants):
Sensitivity 73.5% (95% CI 52.6% to 87.4%) at fixed specificity of 90%
Specificity 83.4% (95% CI 68.2% to 92.2%) at fixed sensitivity of 80% (estimation at 90% not feasible)
SSM by 2D-SWE (6 studies, 353 participants):
Sensitivity 80.0% (95% CI 59.8% to 91.5%) at fixed specificity of 90%
Estimation at fixed sensitivity not feasible
HSROC analysis was not feasible for LSM by pSWE and LSM or SSM by MRE due to insufficient data.
For LSM by VCTE at 25 kPa (9 studies, 1553 participants), sensitivity was 62.3% (95% CI 53.0% to 70.7%) and specificity 94.1% (95% CI 87.2% to 97.3%). We explored heterogeneity and observed a prevalence effect: relative specificity 0.90 (95% CI 0.83 to 0.97). Other factors could not be assessed.
For SPH as secondary objective, LSM by VCTE was the most evaluated test (8 studies; 637 participants). Sensitivity corresponding to specificity of 90% was 67.2% (95% CI 48.8% to 81.4%), while specificity corresponding to sensitivity of 90% could not be calculated by the HSROC model.
Direct comparisons between two index tests were impossible due to inconsistent numbers of participants included for each index test. Due to variability in cut-off values reported across studies, we only performed indirect comparisons using SROC curves. Thus, no reliable results emerged from comparisons or combinations across techniques.
Authors' conclusions
Liver and spleen stiffness measurements may offer a non-invasive alternative to HVPG for detecting CSPH. However, the accuracy of individual techniques remains uncertain due to very low-certainty evidence and insufficient data for reliable comparisons. No test achieved both sensitivity and specificity ≥ 90%, limiting their utility for confidently ruling in or out CSPH.
For LSM by VCTE, the most commonly studied method, HSROC modelling yielded a sensitivity of 72.6% (95% CI 60.0% to 82.5%) at a fixed specificity of 90%, based on studies using various thresholds.
In a subgroup of nine studies (1553 participants) using the predefined 25 kPa cut-off, 38% of patients with CSPH would be missed, and 6% without CSPH would be incorrectly identified.
The certainty of the evidence is very low, mainly due to high risk of bias, heterogeneity, and imprecision.
High-quality research is needed with predefined thresholds, standardised methodology, and improved reporting. Future studies should target key subpopulations (e.g. compensated CLD, specific aetiologies) and assess combinations of non-invasive tools to enhance diagnostic accuracy and clinical usefulness.
Funding
No internal or external sources of support.
Registration
https://doi.org/10.1002/14651858.CD015415
