To study the diagnostic accuracy of transient elastography (TE) for detecting clinically significant portal hypertension (CSPH) in Indian patients with cirrhotic portal hypertension.

This retrospective study was conducted at the Institute of Liver, Gastroenterology, and Pancreatico-Biliary Sciences, Sir Ganga Ram Hospital, New Delhi, on consecutive patients with cirrhosis greater than 15 years of age who underwent hepatic venous pressure gradient (HVPG) and TE from July 2011 to May 2016. Correlation between HVPG and TE was analyzed using the Spearman’s correlation test. Receiver operating characteristic (ROC) curves were prepared for determining the utility of TE in predicting various stages of portal hypertension. The best cut-off value of TE for the diagnosis of CSPH was obtained using the Youden index.

The study included 326 patients [median age 52 (range 16-90) years; 81% males]. The most common etiology of cirrhosis was cryptogenic (45%) followed by alcohol (34%). The median HVPG was 16.0 (range 1.5 to 30.5) mmHg. Eighty-five percent of patients had CSPH. A significant positive correlation was noted between TE and HVPG (rho 0.361, P < 0.001). The area under ROC curve for TE in predicting CSPH was 0.740 (95%CI: 0.662-0.818) (P < 0.01). A cut-off value of TE of 21.6 kPa best predicted CSPH with a positive predictive value (PPV) of 93%.

TE has a fair positive correlation with HVPG; thus, TE can be used as a non-invasive modality to assess the degree of portal hypertension. A cut-off TE value of 21.6 kPa identifies CSPH with a PPV of 93%.

The end result of ongoing injury to the liver due to any cause is hepatic fibrosis, leading to architectural changes and often cirrhosis[1]. Progressive hepatic fibrosis is the most important factor leading to parenchymal dysfunction and the development of portal hypertension. The measurement of portal hypertension is important, as a progressive increase in portal pressure is believed to predict various complications of cirrhosis[2-4]. Hepatic venous pressure gradient (HVPG) is the ideal method for the measurement of portal hypertension and the prediction of complications[5]. Porto-systemic collaterals develop at HVPG greater than 10 mmHg[6], and variceal bleeding could occur from varices when the pressure increases to greater than 12 mmHg[7]. An HVPG greater than 10 mmHg is used as the cut-off for “clinically significant portal hypertension” (CSPH)[8]. However, HVPG is an invasive procedure that requires care and training.

Many non-invasive direct and indirect tests have been reported that are able to predict the presence of CSPH in patients with cirrhosis with considerable accuracy. The ideal non-invasive diagnostic test for portal hypertension should be simple, inexpensive, widely accessible and reliable in measurement and interpretation and provide clinically reliable and relevant information about the degree of portal hypertension. Transient elastography (TE) is a novel, non-invasive, ultrasound-based technology that allows the measurement of liver stiffness. Established evidence indicates that TE has good sensitivity and specificity for diagnosing liver fibrosis and cirrhosis and has been popular over the past few years[9,10]. Recently, many European studies have reported a fairly good correlation between liver stiffness and portal hypertension, suggesting that it could be a good non-invasive tool for evaluation of portal hypertension[11]. However, none of the studies are from South Asia where the etiological profile of cirrhosis is different from other regions of the world.

In the present study, we aimed to identify a possible correlation between TE and HVPG in Indian cirrhosis patients and to investigate whether TE can serve as a non-invasive diagnostic test to identify patients who have CSPH with a reliable cut-off TE value.

In the present study, we showed that in patients with cirrhosis, TE has a fair positive correlation with HVPG, and TE can thus be used as a non-invasive modality to assess the degree of portal hypertension. The TE values increase progressively as portal pressure increases from normal through SCPH, CSPH, SPH and VSPH. We also found that a cut-off TE value of 21.6 kPa has 93% positive predictive value in diagnosing CSPH.

Numerous previous studies have correlated TE with HVPG (Table 4). In these studies, the AUROC curve for prediction of CSPH varied between 0.78 and 0.99. The optimal cut-offs ranged between 13.6 and 65.3 kPa with PPV typically greater than 80%. However, most of these studies used a small number of patients, and almost all of these studies were performed in Europe with none from South Asia where the etiological profile of cirrhosis is different from the West. The largest of these studies was by Reiberger et al[13] from Austria, who retrospectively correlated TE and HVPG in 502 patients. They identified a very high correlation of TE with HVPG (r = 0.799; P < 0.0001). Compared with their study, our correlation coefficient was lower (r = 0.361). Two possible reasons could explain this difference in the strength of correlation between their study and ours. The first reason is the difference in the etiological profiles of patients in the two studies. Their study had more patients with a viral etiology (56%) that typically exhibits a better correlation, whereas our study had more patients with a NASH/cryptogenic and alcohol etiology, which typically has poorer correlation. The second possible reason for the difference in the strength of correlation between their study and ours was that their patients had less severe liver disease compared with our patients. The mean HVPG of their cohort was 12.6 (± 7.6) mmHg, whereas the median HVPG in our patients was 16.0 mmHg. Their median liver stiffness value was 26.4 kPa, whereas that in our study was 36 kPa. TE and HVPG correlate better when liver disease is less advanced. As observed from the scatterplot of our study (Figure 1) and also from the scatterplot of Reiberger’s study[13], the slope of the trend line on the left side of the graph when the disease is less severe is steeper, indicating a better correlation compared with the right side when the trend line is flatter, indicating a poorer correlation at higher TE values. Vizzutti et al[14] also found that the correlation was excellent for HVPG values less than 10 or 12 mmHg (r = 0.81 and r = 0.91, respectively); however, the linear regression analysis was not optimal for HVPG values ≥ 10 mmHg (r = 0.59) or ≥ 12 mmHg (r = 0.37). In advanced portal hypertension, it is not only the liver fibrosis but also the extrahepatic factors, such as the increase in blood flow due to hyperdynamic circulation, that contribute to portal hypertension[15].

There was an urgent need for a South Asian study on the correlation of TE - HVPG because the results of Western studies may not apply to the South Asian population where the etiological and the clinical profile of chronic liver disease differs from the West. Alcohol consumption and the prevalence of diabetes [a major risk factor for nonalcoholic fatty liver disease (NAFLD)] has been steadily increasing in South Asia[16]. The International Diabetes Federation has revised its estimates of the number of people with the diabetes in South-East Asia from 50 million in 2009 to 78.3 million in 2015, with a projection of 140 million by 2040[17,18]. Although cirrhosis mortality has been steadily decreasing globally and especially in the West, it has been steadily increasing in India since 1980. With 188,575 liver cirrhosis deaths in 2010, India ranks number one in the world in liver cirrhosis deaths, accounting for almost one-fifth (18%) of global liver cirrhosis deaths[19]. With such a huge population of liver cirrhosis patients in India whose etiology differs from the West, an easy and non-invasive modality for portal hypertension estimation is urgently needed for treating these patients. Our study attempts to address this need for Indian and South Asian populations.

Most of the complications of cirrhosis are typically related more to CSPH compared with any other factor. Although HVPG measurement is the gold standard for diagnosing CSPH, it is not in common use because it is invasive, not widely available, and expensive, thus hindering its broad use in diagnostic and therapeutic algorithms in patients with cirrhosis with CSPH. Our study found that a TE value of 21.6 kPa is a good cut-off for predicting CSPH with a PPV of 93%. Seventeen additional studies have calculated the cut-off TE values for diagnosing CSPH (Table 4). These studies were performed in patients with different disease severities and different proportions of viral cirrhosis, and their cut-off values ranged from 13.6 kPa to 65.3 kPa. The weighted mean of cut-off from all these studies was 21.9 kPa, which is very similar to what we obtained in our study. Adding our study to the pool of studies, the weighted mean for the best cut-off TE value for diagnosing CSPH is 21.8 kPa.

Notably, Llop et al[20] from Spain provided two cut-offs of TE instead of one to predict CSPH. They showed that a cut-off of 13.6 kPa was sufficiently sensitive to exclude CSPH, and a cut-off of 21 kPa was sufficiently specific to include CSPH. They suggested that values in between these limits (which were found in 35% of their patients) were not useful. Some authors believe that the use of a single TE cut-off, although simple, limits the value of TE to predict CSPH. The use of at least two cut-offs reproduces the clinical thinking in which a diagnostic test commonly provides three outputs: the condition is included, excluded or “further tests are needed”[15]. Thus, TE, using these three outputs, might be useful to select these suspicious patients with cirrhosis for HVPG measurement. However, we believe that multiple cut-offs may lead to confusion, and the single cut-off with a high PPV is best for guiding primary physicians in the community to make treatment decisions.

A recent meta-analysis[11], which included 5 studies of the diagnostic accuracy of TE for significant portal hypertension, also indicated that TE had a high accuracy for the detection of significant portal hypertension. The hierarchical summary receiver-operating characteristic (HSROC) for the diagnosis of significant portal hypertension by TE was 0.93 (95%CI: 0.90-0.95). The Fagan plot analysis showed that TE could be used to diagnose significant portal hypertension (when pre-test probability = 50%), with 81% probability of correctly diagnosing significant portal hypertension following the ‘‘positive’’ measurement. Furthermore, a ‘‘negative’’ measurement was also informative, as significant portal hypertension was present in only 11% of patients. However, when the pre-test probability of significant portal hypertension was as low as 25%, the probability of correctly identifying significant portal hypertension decreased markedly. This finding suggests that an accurate selection of patients is necessary to exploit the performance of TE at its best[11].

Other newer and promising noninvasive modalities are being developed for diagnosing portal hypertension, such as two-dimensional shear wave elastography (2D-SWE)[21-25]; acoustic radiation force impulse (ARFI)[26-28]; VITRO Score (Von Willebrand Factor Antigen/Thrombocyte Ratio)[29]; aspartate aminotransferase/platelet ratio index (APRI)[30]; spleen elsatography[23-27,31]; and serum tests, such as Fibrosis-4, and Lok score[32]. These tests have their own advantages and disadvantages. However, to date, very few studies have been performed on them for correlation with HVPG, so their routine use cannot be recommended outside of clinical trials.

There could be a few limitations in our study. First, it is a retrospective study, so the study may suffer from selection bias. We included only those patients who underwent HVPG and TE during the study period; hence, our patients may not represent the entire population of patients with cirrhosis, as most included patients have moderate to severe portal hypertension. A prospective study design, which includes all consecutive patients of cirrhosis, regardless of degree of portal hypertension, would have been a better study design and more representative of the cirrhotic population of the community. A second limitation could be the lack of follow-up. Follow-up data on complications of portal hypertension would have further validated our results of TE cut-off for CSPH.

In conclusions, our study demonstrated that in patients with cirrhosis, TE has a fair positive correlation with HVPG; thus, TE can be used as a non-invasive modality to assess the degree of portal hypertension. The TE values increase progressively as portal pressure increases from normal through SCPH, CSPH, SPH and VSPH. A cut-off TE value of 21.6 kPa has 93% positive predictive value in diagnosing CSPH. This cut-off will be very useful in diagnosing CSPH and making appropriate treatment decisions in places where HVPG is not available or when patients are unwilling to undergo HVPG due to its invasiveness. As a reliable and non-invasive procedure, TE is a promising and worthy tool to translate into routine clinical practice for detecting CSPH. TE could be integrated in the detection of CSPH in untreated patients for portal hypertension. Further large prospective studies are needed to prospectively validate the findings of our study and also to determine whether TE can be used in monitoring the hemodynamic response and the effect of drugs reducing portal pressure.