Spleen Stiffness Measurement Across the Spectrum of Liver Disease Patients in Real-World Practice

Objectives Spleen stiffness measurement (SSM) provides a non-invasive surrogate marker for clinical significant portal hypertension (CSPH). Results obtained in highly selected populations were promising but require validation across the spectrum of liver disease. We aimed to investigate the clinical applicability of SSM in a real-world setting. Methods We prospectively enrolled patients referred for liver ultrasound (January–May 2021). Patients with a portosystemic shunt, liver transplant, or extrahepatic etiology of portal hypertension were excluded. We performed liver ultrasound, liver stiffness measurement (LSM) and SSM (dedicated software, 100 Hz-probe). Probable CSPH was established if ≥1 of the following items occurred: ascites, varices, encephalopathy, splenomegaly, recanalized umbilical vein, collaterals, dilated portal veins, hypertensive gastropathy, or LSM ≥25 kPa. Results We enrolled 185 patients (53% male; age 53years [37–64], 33% viral hepatitis, 21% fatty liver disease). Of them, 31% of patients had cirrhosis (68% Child-Pugh A) and 38% of patients had signs of portal hypertension. SSM (23.8 kPa [16.2–42.3]) and LSM (6.7 kPa [4.6–12.0]) were successful and met reliability criteria in 70% and 95%, respectively. Spleen size was inversely associated with SSM failure (odds ratio: 0.66 increment/cm, 95% confidence interval: 0.52–0.82). Optimal spleen stiffness cut-off to detect probable CSPH was >26.5 kPa (likelihood ratio: 4.5, sensitivity: 83%; specificity: 82%). Spleen stiffness did not outperform liver stiffness in detecting probable CSPH (P = 1.0). Conclusions In real-world practice, reliable SSM were obtained in 70% and could potentially stratify patients between high- and low-risk of probable CSPH. However, cut-offs for CSPH might be substantially lower than previously reported. Future studies validating these results are required. Clinical trial number Netherlands Trial Register (Registration number: NL9369).

[10] In this prospective study, we measured spleen stiffness with a 100 Hz probe and SSM-dedicated software combined with a same-session abdominal ultrasonography and LSM in a heterogeneous group of patients having liver disease.We aim to determine the SSM cut-off value to identify the presence of probable CSPH in these patients covering different disease stages of a broad spectrum of liver diseases and identify factors associated with SSM failure in real-world practice.

METHODS
Institutional Review Board approval was obtained (Medical Ethics Review Committee of the Erasmus MC (Reference: MEC-2021-0056).

Study Design
We performed a single center prospective study in patients with any type of liver disease to determine a cut-off value to detect probable CSPH between January 2021 and May 2021.Next, we identified factors associated with SSM failure.This study was performed at the outpatient ultrasound clinic of a tertiary referral center.Since July 2020, SSM is implemented as standard of care at the outpatient liver ultrasound program in this center.The STAndards for the Reporting of Diagnostic accuracy studies 2015 checklist was followed (Supplementary Table 1).

Outcomes
To classify patients as having probable CSPH (because of the lack of HVPG) during study visit, we used predefined criteria, based on the Baveno consensus. 11In short, 'probable' CSPH was diagnosed when one or more of the following criteria were present: splenomegaly, ascites, recanalized umbilical vein, portosystemic collaterals, dilated portal veins (splenic vein $13 mm; portal vein $16 mm), varices, hepatic encephalopathy, portal hypertensive gastropathy, or LSM $25 kPa.A stricter definition was used to define 'definite CSPH' which was limited to the evident signs of portal hypertension and included the presence of ascites, recanalization of the umbilical vein, portosystemic collaterals, or history of varices.

Patient Identification and Selection Criteria
Patients visiting the outpatient clinic were prospectively enrolled in our study.Patients received a same-visit ultrasonography and LSM in addition to SSM.The allocated time for each examination was standardized at 30 min.For the purpose of this study, we only included patients when SSM was performed by the investigators to limit interobserver variability.Adults ($18 years) with any (suspected) liver disease, with or without the presence of advanced liver disease/cirrhosis or documented signs of portal hypertension, were eligible for study inclusion.We excluded patients with a transjugular intrahepatic portosystemic shunt, liver transplant, or extrahepatic etiology for liver disease (thrombosis, congestive hepatopathy and (congenital) vascular malformations).

Abdominal Ultrasonography
Prior to VCTE, each patient received an abdominal ultrasound examination.We used a Philips EPIQ 7 ultrasound system (Philips Medical Systems, Eindhoven, the Netherlands).Both a convex (2-5 MHz) and linear high frequency probe (2-9 MHz) were used following a standardized protocol.Grayscale ultrasound was followed by color-duplex sonography.

Transient Elastography
Following abdominal ultrasound, the three examinators performed the VCTE measurements.Abdominal ultrasound was used to augment VCTE probe placement to facilitate optimal spleen and liver stiffness measure-mentLSM.We used the FibroscanÒ (Fibroscan Expert 630, Echosens, France) to attempt liver and spleen stiffness measurementSSM.LSM was successful in case of $10 successful measurements and kPa #7 or kPa >7 and interquartile range [IQR] <30%. 12There are no reliability criteria available for SSM.To increase its reproducibility and accuracy, SSM was successful in case of $8 successful measurements and SSM kPa #10, or kPa >10 and IQR <30%.Median liver/spleen stiffness (in kPa) and IQR were reported.The controlled attenuation parameter (in dB/m) was measured, including standard deviation.Using the MyFibroScan smartphone application for disease-specific cut-offs, LSM results were converted to fibrosis stage (F0-F1: no fibrosis -F4: cirrhosis) and CAP controlled attenuation parameter measurements to steatosis stage (S0: no steatosis; S3: severe steatosis), respectively.

Data Collection
Individual patient data were collected and anonymously uploaded to clinical record forms.We used an approved and validated data management program.Data collection included patient characteristics, drug use, laboratory measurements, liver disease characteristics, ultrasound data, and VCTE results.Records were searched for decompensation events, endoscopy findings, imaging results, and pathology reports.

Statistical Analysis
To assess the performance of a single diagnostic test, $33 patients were necessary to obtain a predicted area under the receiver operating characteristic curve (AUROC) of 0.75 compared to a null hypothesis value of 0.5 (a:0.05,power 80%, allocation ratio 2).A total of 114 patients were required to assess a significant difference between the diagnostic performance of two diagnostic tests with an ex-pected AUROC of 0.8 and 0.9, respectively (estimated correlation between two tests 0.75 in both positive and negative cases, 13,14 a:0.05, power 80%, allocation ratio 2).
For dichotomous data Chi-square test was used and Fisher's exact test when expected count <5 in >25% of cells.For normally distributed continuous data, we used the Student t-test; for non-normally distributed data, the Mann-Whitney U test.We calculated correlations between continuous and dichotomous variables using point-biserial correlation (r pb ).
AUROC analysis was performed to determine the diagnostic performance of a continuous variable (SSM and LSM) in predicting the presence of probable CSPH.We determined the optimal SSM cut-off value (in kPa) for detecting probable and definite CSPH determined by the maximum value of the Youden's J statistic.These data were reported together with the likelihood ratio (LR), standard error (SE), sensitivity and specificity including 95% confidence interval (CI).To assess whether the difference between the AUROC of two diagnostic tests was significantly different, we used the Hanley McNeil methodology to allow comparing two diagnostic tests in the group of patients and controls. 15To assess whether the inclusion of the LSM $25 kPa criterion in our probable CPSH definition affected the comparison, we performed a sensitivity analysis in which we used a probable CSPH definition in which LSM $25 kPa alone was not a criterion.Multivariate logistic regression modeling was used to identify factors associated with SMM failure.Covariates with P < 0.15 in univariate analysis were included in the multivariate

Patient Identification and Characteristics
A total of 185 patients were included (Figure 1).Table 1 shows the patient characteristics.The main reason for their visit was the evaluation of liver disease severity (66%) followed by hepatocellular carcinoma screening (32%).Diabetes was diagnosed in 15% (n = 27).

Predictors for Failure of SSM
Table 3 shows the univariate analysis to identify predictors for the failure of SSM.In the multivariate model (Table 4), only spleen size was associated with SSM failure (odds ratio [OR] 0.66 increment per cm, 95% CI 0.52-0.82,P < 0.001).Of note, the other variables were not associated with SSM failure and thus excluded from the model by backward selection.In a post-hoc analysis when stratifying for LSM (LSM <10 kPa vs. LSM $10 kPa), we noted a higher SSM failure rate in patients with LSM <10 kPa (36%) than patients with LSM $10 kPa (21%).Interestingly, in both groups, spleen size was the most important predictor for SSM failure, aligning the results of the overall analysis.In this analysis, sex was an additional independent predictor for SSM failure (OR 2.78 (95% CI: 1.18-6.60),P 0.02 for LSM <10 kPa and OR 0.09 (95% CI 0.01-0.98,P 0.048) for LSM $10 kPa, respectively.

Diagnostic Performance of LSM and a Prediction Model to Detect CSPH
As with SSM, we observed a positive correlation between LSM and the presence of probable CSPH, which was statistically significant (r pb = 0.58, n = 118, P < 0.001, Figure 2A).The AUROC for liver stiffness for predicting portal hypertension was 0.86 (SE 0.04 95% CI: 0.78-0.94, Figure 2B).Interestingly, when comparing the diagnostic performance of SSM with LSM (n = 118), no significant difference was detected (P = 1.0).Results were consistent when LSM $25 kPa alone was not a criterion for probable CSPH.A prediction model including all statistically significant characteristics (P < 0.05) between groups with or without probable CSPH (Table 5) returned an AUROC of 0.89 (SE 0.03, 95% CI: 0.83-0.96,n = 122, P < 0.001).Again, compared to SSM, there was no significant improvement in the diagnostic accuracy for detecting CSPH (P = 0.62).

PREDICTORS FOR FAILURE OF SSM
Table 3 shows the univariate analysis to identify predictors for the failure of SSM.In the multivariate model (Table 4), only spleen size was associated with SSM failure (odds ratio [OR] 0.66 increment per cm, 95% CI 0.52-0.82,P < 0.001).Of note, the other variables were not associated with SSM failure and thus excluded from the model by backward selection.In a post-hoc analysis when stratifying for LSM (LSM <10 kPa vs. LSM $10 kPa), we noted a higher SSM failure rate in patients with LSM <10 kPa (36%) than patients with LSM $10 kPa (21%).Interestingly, in both groups, spleen size was the most important predictor for SSM failure, aligning the results of the overall analysis.In this analysis, sex was an additional independent predictor for SSM failure (OR 2.78 (95% CI: 1.18-6.60),P 0.02 for LSM <10 kPa and OR 0.09 (95% CI 0.01-0.98,P 0.048) for LSM $10 kPa, respectively.

Diagnostic Performance of SSM to Detect CSPH
Figure 2A shows that there was a positive correlation between SSM (kPa) and the presence of probable CSPH, which was statistically significant (r pb = 0.61, n = 118, P < 0.001).The AUROC for SSM to predict the presence of probable CSPH was 0.86 (SE 0.04 95% CI 0.79-0.94,n = 118, Figure 2B).

Diagnostic Performance of LSM and a Prediction Model to Detect CSPH
As with SSM, we observed a positive correlation between LSM and the presence of probable CSPH, which was statistically significant (r pb = 0.58, n = 118, P < 0.001, Figure 2A).The AUROC for liver stiffness for predicting portal hypertension was 0.86 (SE 0.04 95% CI: 0.78-0.94, Figure 2B).Interestingly, when comparing the diagnostic performance of SSM with LSM (n = 118), no significant difference was detected (P = 1.0).Results were consistent when LSM $25 kPa alone was not a criterion for probable CSPH.A prediction model including all statistically significant characteristics (P < 0.05) between groups with or without probable CSPH (Table 5) returned an AUROC of 0.89 (SE 0.03, 95% CI: 0.83-0.96,n = 122, P < 0.001).Again, compared to SSM, there was no significant improvement in the diagnostic accuracy for detecting CSPH (P = 0.62).

DISCUSSION
This prospective study on the applicability of SSM across the spectrum of liver disease patients to potentially stratify patients between high-and low-risk of probable CSPH in a real-world setting showed that SSM was successful in 70% of patients.Smaller spleen size was associated with SSM failure.The optimal cut-off value for SSM to detect probable CSPH was >26.5 kPa (sensitivity 83%, specificity 82% for probable CSPH and 100% sensitivity and 75% specificity for definite CSPH).In more targeted populations like liver stiffness $10 kPa and using a stricter definition for CSPH, the cut-off might be as high as 41.5 kPa (sensitivity 82%, specificity 80%).We did not detect a significant difference in the diagnostic performance of SSM compared to LSM in the studied population using a surrogate marker for probable CSPH.
This study showed a 70% success rate of SSM despite using dedicated SSM software and a 100 Hz probe.This is significantly lower than previously reported by a similar study which revealed a success rate of 92.5%. 8We hypothesize that there are two main reasons for this difference.First, in contrast to the aforementioned study, we included patients with non-advanced liver disease.As a result, the median spleen size of patients included in our study was lower by more than 2.5 cm (11.0 cm [9.7-13.1] vs. 13.6 cm [11.9-15.5],respectively).This could explain the difference in success rate as we confirmed a previous observation that smaller spleen size is independently associated with SSM failure. 16Importantly, spleen size was the most important predictor for SSM-failure, both in patients with high and low liver stiffness.In contrast to earlier studies, we did not identify body mass index, length, or body weight as independent predictors for SSM failure. 16,17As body mass index was not associated with SSM failure in our analysis, we believe not anthropometric measurements, but spleen size is the key factor associated with SSM failure.Second, eligibility criteria were used for SSM unlike most other studies, lowering our success rate from 81% to 70% as 20 SSM results failed our eligibility criteria.Considering these points, the success rate in daily practice, especially considering time restrictions, may be lower than reported in previous studies.Nonetheless, a large study that also included healthy individuals reported a very low failure rate but did not show data on the eligibility of the obtained measurements. 18Additional studies are warranted to estimate the success rate when SSM is adopted in clinical practice and determine which set of eligibility criteria are required.A recent meta-analysis included ten studies reporting the diagnostic accuracy of spleen stiffness in the context of evaluating portal hypertension in chronic liver disease. 19Three of the included studies (total n = 298) used an identical VCTE technique. 7,20,21In contrast to our cut-off of 26.5 kPa to detect probable CSPH, the cut-off for CSPH in these studies varied between 48.9 and 55.0 kPa. 7,20,21None of the aforementioned studies included patients with non-advanced liver disease.As expected, we showed that non-cirrhotic patients had a lower spleen stiffness than those with cirrhosis (median 17.7 kPa [14.7-24.6] vs. median 46.7 kPa [34.8-60.6],respectively).As a corollary, the distribution of SSM values used in our AUROC analysis is far more dispersed compared to that in aforementioned studies.Indeed, when a targeted analysis was performed, including only patients with LSM $10 kPa, the optimal threshold to detect definite CSPH increased to 41.5 kPa.Therefore, we hypothesize that this explains the difference in the diagnostic cut-off values, as a different degree of dispersion substantially influences these cut-off values. 22n important study recently tested the new Baveno VII guidelines in a multi-center retrospective trial among liver disease patients with LSM $10 kPa. 23In this study, it was demonstrated that the currently proposed Baveno VII guideline would result in a large proportion (up to 60%) of patients that could not be non-invasively assessed for the presence of CSPH.Interestingly, they provide a new algorithm using (1) LSM #15 kPa, (2) platelet count $150 Â 10 9 , and (3) SSM #40 kPa to rule out CSPH and (1) LSM >25 kPa and (2) platelet count <150 Â 10 9 and (3) SSM >40 kPa to rule in CSPH.The proposed 40 kPa SSM threshold aligns with our optimal cut-off in patients at high risk of definite CSPH.Important to note, only two out of three items needed to be present to rule out and rule in CSPH with 90% NPV and PPV in their cohort with a 62-63% CSPH prevalence.
LSM was the first VCTE technique in clinical use for patients having liver disease. 24Originally, LSM was developed as a non-invasive tool to replace liver biopsies to assess the presence of significant fibrosis. 25As liver parenchyma resistance is fundamental in portal hypertension development and an increased portosystemic pressure is reflected by increased liver stiffness, a correlation between LSM and HVPG measurement was made, enabling LSM to diagnose portal hypertension in patients with chronic liver disease. 26In contrast to the static assessment of liver parenchyma stiffness by LSM, SSM is hypothesized to provide a dynamic surrogate marker of real-time HVPG. 27ince its introduction, clinical studies have shown that SSM could outperform LSM in detecting portal hypertension and in risk stratification of patients for detecting high-risk varices. 7,16,20Most recent, a study showed that when a 100 Hz spleen dedicated probe is used, the diagnostic accuracy improves even further. 8In contrast, despite using the latest SSM software and a dedicated 100 Hz probe, SSM was not superior to LSM in detecting probable CSPH in our cohort.Our results are in line with another study in liver disease patients in which the study authors included 107 patients referred for HVPG measurement and performed both LSM and SSM, but showed no difference between these two techniques in detecting portal hypertension. 21Our finding may be explained by not limiting to patients with advanced liver disease and excluding patients with extra-hepatic causes of portal hypertension, which both could benefit the diagnostic accuracy of SSM compared to LSM.Thus, when LSM and SSM are used to evaluate the presence of portal hypertension in patients with an intrinsic liver disease, irrespective of liver disease severity, we expect that LSM and SSM would be equally effective.
Our data illustrate that spleen stiffness increases when liver disease progresses, which could, in turn, reflect the progressive hemodynamic changes in the portosystemic circulation.Future studies would need to assess the exact value of SSM in predicting clinical decompensation in patients suffering from chronic liver disease before they have developed cirrhosis or overt portal hypertension.Moreover, the full potential of the applicability of SSM to be used as a bedside tool to provide a real-time assessment of a patient's portal hypertensive state in the clinical context of evaluating transjugular intrahepatic portosystemic shunt function or the evaluation of response to nonselective blockade therapy is paramount. 28,29his study has several limitations.First, examinators were not blinded for clinical data or same-session abdominal ultrasound result.Moreover, probable CSPH was diagnosed based on a criteria set using surrogate markers of portal hypertension instead of the gold standard HVPG measurement and only a limited number of included patients had a diagnosis of compensated cirrhosis.Consequently, this resulted in under-or overreporting of the presence of CSPH in our study.However, it was deemed unethical to perform HVPG measurement without any clinical substantiation in patients without any sign of advanced liver disease.Comparably, a recent study in primary biliary cholangitis used comparable criteria to clinically assess the presence of CSPH, as our probable CSPH definition. 30Additionally, we used a stricter definition to define definite CSPH.In light of this, incorporating HVPG in our study protocol would have undoubtedly led to selection bias as patients with non-advanced liver disease would be unwilling to participate.Therefore, the current study provides a more accurate representation of patients undergoing evaluation with VCTE at an outpatient clinic.Second, the comparison of diagnostic accuracy of LSM vs. SSM could potentially be influenced as LSM $25 kPa was a criterion for diagnosing probable CSPH.But, even when LSM as criterion was removed, the diagnostic performance between LSM vs. SSM remained statistically equal.Third, due to the fact that SSM was only measured once by a single operator, no information was available on inter-and intra-observer agreement.However, recently, it was published that the inter-and intra-observer agreement was adequate. 18Fourth, we could not investigate the complex interactions between portosystemic shunts, LSM and SSM due to the limited sample size that did not allow for additional analysis in these small subgroups.Next, our multivariate analysis on predictors for SSM failure need to be interpreted with caution as it is at a risk of being underpowered as only a limited number of cases with a failed SSM could be included.Finally, despite a recent network meta-analysis showing a comparable diagnostic value for the different techniques available, 31 caution is needed when generalizing our VCTE results to other techniques that evaluate liver and spleen stiffness such as shear wave elastography or magnetic resonance elastography.Finally, although including a broad spectrum of liver diseases and several clinical stages could assess the clinical applicability in a real-world setting, cutoffs should be validated in disease-specific studies, comparable to the validation studies for LSM.
Strengths of this study include its prospective design, which combined LSM and SSM with a same-session abdominal ultrasound.Moreover, we present a novel study as we measured spleen stiffness with a spleendedicated 100 Hz probe in a heterogeneous liver disease population, including patients without advanced liver disease.The data generated by this study suggest that the cut-off value for diagnosing probable CSPH in a real-world setting of liver disease patients is different to the cut-off reported by previous studies.Lastly, to improve the reproducibility and accuracy of our data we applied reliability criteria for SSM which increased the robustness of our results.
In conclusion, reliable SSM were obtained in most patients in a cohort of patients across the spectrum of liver disease etiology and severity.In these patients, SSM is able to potentially distinguish between high and low risk of CSPH in this heterogeneous population.However, the cut-off for probable CSPH defined by clinical and imaging-based characteristics might be substantially lower than previously reported.In this real-world setting, SSM had a similar diagnostic performance as compared to LSM, and future research should focus on whether SSM has added value in individuals with modestly elevated LSM.

Figure 1
Figure 1 Flowchart showing number of patients at each stage of the study.Abbreviations: AUROC, area under the receiver operating characteristic curve; LSM, liver stiffness measurement; n, number; SSM, spleen stiffness measurement.

Figure 2 A
Figure 2 A. Distribution of individual SSM (left two columns, n = 118) and LSM (right two columns, n = 118) results, stratified by absence ('No' columns) or presence ('Yes' columns) of clinically significant portal hypertension.B. Area under the receiver operating characteristic curve of spleen (green solid line) and liver (red solid line) stiffness measurement in patients with portal hypertension.Abbreviations: LSM, liver stiffness measurement; SSM, spleen stiffness measurement.

Table 1
Baseline Characteristics of Included Patients.

Table 1 (
Continued ) a Multiple diagnoses possible.

Table 2
Details on Abdominal Ultrasound Results and Liver and Spleen Elastography Results.

Table 3
Univariate Analysis to Identify Predictors for the Failure of SSM.

Table 4
Multivariate Regression Analysis to Identify Predictors for the Failure of SSM in patients Who Underwent SSM (n = 185).

Table 5
Comparison of Characteristics Between Patients With Successful SSM Who did or did not met the Criteria for Clinically Significant Portal Hypertension.
b Multiple diagnoses possible.