Liver stiffness-spleen diameter to platelet ratio score (LSPS model) predicts variceal rebleeding for cirrhotic patients

Purpose The liver stiffness- spleen diameter to platelet ratio score (LSPS model) can identify a high risk of decompensated events in cirrhotic patients. We aimed to evaluate the value of the LSPS model as a risk stratification strategy in the secondary prevention for cirrhotic patients with esophageal and gastric variceal bleeding (EGVB). Methods Consecutive EGVB patients who underwent liver stiffness measurement by acoustic radiation force impulse, platelet count and ultrasonography were enrolled between January 2013 and December 2019. We calculated the LSPS of all patients and followed up for over 2 years. The primary outcome was rebleeding. Transplant-free survival and overt hepatic encephalopathy (OHE) were the secondary outcomes. Results A total of 131 patients were analyzed. The median value of the LSPS model is 0.1879. We developed risk stratification based on the LSPS model and divided the patients into two groups: the high-LSPS (LSPS > 0.1879) group and the low-LSPS (LSPS ≤ 0.1879) group. Sixty-two (47.33%) patients suffered rebleeding, in which there were 21 (31.92%) patients with low LSPS and 41 (63.08%) patients with high LSPS (hazard ratio 2.883; 95% confidence interval, 1.723–4.822, P < 0.001). For the whole cohort, the rates of transplant-free survival and OHE were consistently similar between the two groups at 2 years. Conclusion The LSPS is a reliable, noninvasive method for the detection of a high risk of rebleeding for the secondary prevention of EGVB.


Introduction
Esophageal and gastric variceal bleeding (EGVB) is a usual decompensation event in cirrhosis. More than 50% of patients with acute EGVB under the recommended therapy experience rebleeding in 1 year, and this is one of the most common reasons for readmission [1]. Furthermore, many fatal complications can follow EGVB, such as hepatic encephalopathy, liver failure and hemorrhagic shock [2]. Endoscopic therapy combined with nonselective β-blockers (NSBBs) is regarded as 'the standard therapy' for secondary prophylaxis of rebleeding [3]. However, 20-40% of cirrhotic patients experience rebleeding after 'standard therapy' and 30-50% experience rebleeding after only endoscopic therapy [4]. Due to the adverse outcome, risk stratification to identify patients with a high risk of rebleeding and mortality should be performed before secondary prophylaxis for patients surviving acute EGVB.
The hepatic venous pressure gradient (HVPG) has been recommended as the best risk stratification tool for portal hypertension [3]. Decision-making regarding treatment based on HVPG risk stratification is usually suggested [3]. Early transjugular intrahepatic portosystemic shunt (TIPS) should be considered to effectively decrease the portal tension for those patients with HVPG ≥ 20 mmHg in acute variceal bleeding, while endoscopic therapy is the recommended therapy for patients with low HVPG [3,5]. However, the measurement of HVPG is an invasive operation in which intervention-radiologists' operating skills and equipment are needed, and the probability of transient arrhythmias, vagus nerve reactions and other adverse reactions cannot be ignored, which greatly limits the clinical application and promotion of HVPG.
Several noninvasive measurements have been extensively studied recently, such as liver stiffness, spleen stiffness [6] and biomarkers, while the performance of a single indicator was less than satisfactory [7,8]. Therefore, some models based on noninvasive measurements have been presented to improve the ability of detection and diagnosis by combining several indexes. The liver stiffness-spleen diameter to platelet ratio score (LSPS) has been studied in several studies and has a proven potency in the detection of EGVB for patients with compensated cirrhosis [9]. A meta-analysis in 2018 [10] concluded that the diagnostic efficiency in EGVB of LSPS is higher than that of liver stiffness for those in the compensated stage, while in highrisk EGVB, both LSPS and liver stiffness had moderate sensitivity and specificity.
In previous studies, the LSPS model was mainly applied in patients with viral hepatitis or compensated cirrhosis to detect clinically significant portal hypertension and the presence of EGVB [8,11] and has been rarely discussed in application for the secondary prevention of decompensated cirrhotic patients. In our study, we aimed to evaluate the value of the LSPS model as a risk stratification strategy in the secondary prophylaxis of EGVB for cirrhotic patients.

Study design and patients
We prospectively collected clinical data from cirrhotic patients with first EGVB who were admitted to the Department of Gastroenterology in Nanjing Drum Tower Hospital from January 2013 to December 2019. All patients who signed the informed consent form were monitored according to the current guidelines [3,12]. All patients included had no history of decompensated events. All patients included did not experience endoscopic therapy or covered TIPS. For this study, from the whole sample registered, the included patients had to receive standard therapy for secondary prophylaxis of EGVB. The exclusion criteria of the study were as follows: (1) complicated with hepatocellular carcinoma or extrahepatic tumor; (2) serious heart, lung, liver, renal insufficiency or serious hemorrhagic disease or local, systemic infection, hypothyroidism, Raynaud's syndrome, peripheral vascular disease, etc.; (3) patients who underwent splenectomy; (4) patients with severe portal vein thrombosis; and (5) women planning to become pregnant or already pregnant or breastfeeding.
The study protocol was approved by the ethics committees of the Nanjing Drum Tower Hospital. Two physicians regularly reviewed the data to detect errors. Two physicians were set to assess the accuracy of the data. After verification of the collected clinical variables in our study, statistical analysis was performed.
The calculation of liver stiffness-spleen diameter to platelet ratio score In our research, the LSPS was calculated according to the formula 'LSPS = liver stiffness (m/s) × spleen intercostal diameter (cm)/platelet (×10 9 /L)'. Liver stiffness and spleen stiffness were measured by acoustic radiation force impulse (ARFI). The measurements were performed using the techniques described previously [13]. A 4C-1 curved array transducer (Acuson S2000, Siemens Medical Solutions, San Jose, California, USA) was used to perform both real-time grey scale imaging and point shear wave elastography. Five valid Shear wave velocity (SWV) measurements were tested in the liver or spleen for each patient, and the mean SWV values were calculated. The abdominal ultrasound, the measurement of liver stiffness and the blood test were measured before receiving the therapy after admission. Two ultrasound doctors with over 5 years of experience supervised the process of measurement. The spleen diameter was measured by abdominal ultrasonography.

Therapy for secondary prophylaxis
Endoscopic treatments were performed for the included patients. Endoscopic variceal ligation was necessary and endoscopic injection sclerotherapy or N-butyl-lcyanoacrylate injection was performed according to the willingness of the patients, the experience of doctors and the conditions of blood vessels comprehensively. Endoscopy was performed every 4-8 weeks until the variants were eradicated or turned into F1, and endoscopic treatment was performed again if necessary. The ligator is a six-shot ligator (MET; COOK Company, Bloomington, Indiana, USA), and the main components of the hardener and tissue glue are lauromacrogol injection (lauryl alcohol polyoxyethylene) from Tianyu Pharmaceutical Company, Taizhou, Zhejiang, China and n-butyl α-cyanoacrylate adhesive from COMPONT Company, Beijing, China.
After excluding the contraindications, some patients took carvedilol, which was started at 6.25 mg every day, or propranolol, with a starting dose of 10 mg twice a day, and the amount was increased to the maximum tolerable dose gradually. If adverse reactions occurred, the dosage was reduced or discontinued. Patients whose blood pressures were less than 90/60 mmHg or heart rates were less than 55 beats per minute during the first week of initial NSBBs were defined as intolerant of NSBBs, who received endoscopic treatments only.
During the 24-month follow-up period, the patients who rebled after the mentioned therapy received consecutive endoscopic therapy or TIPS, according to the current consensus [3].

The outcome
The patients were followed up until death or liver transplantation, and the clinical data, including medical records, imaging diagnosis and laboratory tests, during this period were used for the analysis. The patients for oral NSBBs or not at the end of follow-up were counted.
The primary endpoint of the study was the rebleeding-free time. The secondary endpoints were (1) survival time and (2) hepatic encephalopathy.

Statistical analysis
Our long-term observational study followed the Strengthening the Reporting of Observational Studies in Epidemiology guideline [14]. Data were described as frequencies and percentages, means and SDs or medians and interquartile ranges, as appropriate. The baseline characteristics were compared using Fisher's exact test for categorical variables, t test for continuous variables and the Wilcoxon rank sum test for ordinal and noncontinuous variables. Rebleeding and survival were estimated using the cumulative incidence function. Survival curves were generated by Kaplan-Meier analysis. All analyses were performed using GraphPad software (version 9.0, Harvey Motulsky, San Diego, Southern California, USA) or Statistical Package for Social Sciences (version 22.0, SPSS Inc, Armonk, New York, USA), and a level of significance was established at the two-sided 5% level.

Baseline characteristics
From January 2013 to December 2019, a total of 187 patients with EGVB were registered in the LSPS-guided risk-stratification database, as shown in Fig. 1, and 56 patients met one or more exclusion criteria. Consequently, 131 patients were included in the main analysis. The follow-up time was over 24 months. The median follow-up time was 47 months. The receiver operator characteristic curve of liver stiffness, spleen stiffness and LSPS are demonstrated in Supplementary Fig. 1, Supplemental Digital Content 1, http://links.lww.com/EJGH/A835 which showed that the LSPS model was superior to spleen stiffness and liver stiffness. Based on LSPS-guided risk stratification, 65 (49.62%) patients with an LSPS higher than the median value (0.1879) of the LSPS were divided into the high-LSPS group, while the remaining 66 (50.38%) patients were divided into the low-LSPS group. There were few differences in the baseline characteristics between the two groups ( Table 1).

Comparison of the rebleeding rates between the patients in different liver stiffness-spleen diameter to platelet ratio groups
During the 24-month follow-up period, 62 (47.33%) patients experienced rebleeding, among which there were 21 (31.92%) patients with low LSPS and 41 (63.08%) patients with high LSPS. In the whole cohort, the patients in the high-LSPS group suffered significantly more rebleeding [hazard ratio 2.883; 95% confidence interval (CI), 1.723-4.822; P < 0.001], as shown in Fig. 2a. The median rebleeding time in the whole cohort was more than 2 years (27 months). In the low-LSPS group, the median rebleeding time was more than 2 years (35 months), while 18.5 months in the high-LSPS group (Fig. 2a).
A subgroup analysis was performed due to the differences in baseline etiology between the two groups. The results of subgroup analysis based on etiology are shown in Fig. 2b and c. In 61 patients with cirrhosis related to viral hepatitis, the probability of rebleeding in the high-LSPS group was significantly higher (hazard ratio 2.630; 95% CI, 1.280-5.404; P = 0.02), which was also consistent with that in the 70 cirrhotic patients without viral hepatitis (hazard ratio 3.958; 95% CI, 1.818-8.617, P < 0.001).
We performed the subgroup analysis on the basis of receiving NSBB therapy or not. Seventy-two patients took NSBBs for a long time. The results demonstrated that more patients with high LSPS experienced rebleeding regardless of whether NSBBs were used (hazard ratio 2.464; 95% CI, 1.201-5.057; P = 0.01; hazard ratio 3.790; 95% CI, 1.786-8.042; P < 0.001) (Fig. 2d and e).
The results for the whole follow-up period were consistent with the 2-year period ( Supplementary Fig. 2

Comparison of the transplant-free survival rates between the patients in different liver stiffness-spleen diameter to platelet ratio groups
Fourteen (10.69%) patients died, of whom nine (13.85%) patients belonged to the high-LSPS group and five (7.58%) belonged to the low-LSPS group. In the whole cohort and over 2 years, one (0.76%) patient in the high-LSPS group received orthotopic liver transplantation. The transplant-free survival was not significantly different between the high-LSPS group and the low-LSPS group in the whole cohort (hazard ratio 2.550; 95% CI, 0.892-7.293; P = 0.08) (Fig. 3a).
Uncontrolled rebleeding, hepatic encephalopathy caused by liver failure and multiple organ dysfunction were the most common causes of death and transplant ( Table 2).

Subgroup analysis based on nonselective β-blockers usage
Subgroup analysis was conducted based on the use of different NSBB drugs. Overall, patients with oral NSBBs had a lower rate of rebleeding than patients with NSBB intolerance (P ≤ 0.001) shown in Fig. 5a. There was no significant difference between propranolol and carvedilol in reducing rebleeding rates in our study. The patients in the high-LSPS group suffered significantly more rebleeding ignoring the NSBBs usage ( Fig. 5b-d). The results for the whole follow-up period were consistent with the 2-year period ( Supplementary Fig. 5, Supplemental Digital Content 5, http://links.lww.com/EJGH/A839).

Discussion
For secondary prophylaxis of decompensated-cirrhotic patients, risk stratification should be performed to recognize the patients with a high risk of decompensation events to facilitate chronic disease management individually, and this includes individualized therapy and regular follow-up visits. The measurement of HVPG is an effective risk-stratification tool in further decompensation. Several randomized controlled trials (RCTs) have proven that the progress of EGVB correlated with the value of HVPG [12]. A study in 2015 concluded that HVPG measurement was useful for making decisions to select the type of endoscopic therapy with NSBBs and TIPS in secondary prophylaxis [15]. However, the characteristics of invasive manipulation, such as high technical requirements and possible adverse reactions, limit the widespread application of HVPG. It is necessary to develop and validate noninvasive methods to detect high-risk patients and assess the effect of first-line therapy.
The diagnosis of compensated advanced chronic liver disease, which aims at stratifying the risk of clinically significant portal hypertension and decompensation, is based on the measurement of liver stiffness [3]. The measurement of liver stiffness, between which HVPG has good relevance, is widely applied for patients with hepatitis [16]. However, when the portal venous pressure had increased to a certain extent, the changes in liver stiffness stopped, while the parameters related to the spleen, such as spleen stiffness, spleen size and platelet count, continued to change [10]. It was concluded that spleen stiffness was better than liver stiffness in detecting EGVB for patients with compensated cirrhosis [17][18][19]. Giuffrè et al. [20] pointed out that more than 30% of compensated cirrhotic patients could safely avoid endoscopy if the spleen stiffness cutoff value is applied. Table 1. Baseline characteristics of the high-liver stiffness-spleen diameter to platelet ratio score group and the low-liver stiffness-spleen diameter to platelet ratio score group Meanwhile, the diagnostic value of complex models based on noninvasive indexes has been validated in recent years. Liver stiffness combined with spleen diameter, as a simple noninvasive scoring system, performed well in predicting EGVB and discriminating the grades of EGVB [21]. Shibata et al. pointed out that the ratio of platelets to liver stiffness could improve the accuracy of screening for high-risk EGVB requiring advanced treatment [22]. Kim et al. [9] established the LSPS model, which is based on transient elastography and liver stiffness. According to LSPS risk stratification, different prophylactic treatments should be considered for subgroups with LSPS ≥ 6.5 [9]. A retrospective study concluded that LSPS can be used to identify EGVB in patients with chronic hepatitis C [23]. LSPS based on ARFI can accurately identify high-risk-EGVB in patients with compensated cirrhosis [24]. However, noninvasive diagnostic models have been applied in cirrhosis associated with viral hepatitis, while cirrhosis related to other etiologies has rarely been revealed. Furthermore, previous studies did not include patients receiving NSBBs.
Transient elastography, which was applied in the LSPS model in most previous studies, has been widely used in the assessment of liver stiffness and spleen stiffness, but there are some technical limitations for patients with obesity, intercostal stenosis, ascites, etc. According to the EFUSUMB guidelines, transient elastography should not be used in patients with perihepatic ascites [25], which are common in patients with decompensated cirrhosis. At present, ARFI is a noninvasive, rapid, simple, objective and repeatable method for detecting the degree of liver fibrosis with good application prospects. Compared with transient elastography, it is more convenient to measure the stiffness of tissue while conducting general Ultrasound (US) without extra equipment. Meanwhile, ARFI was less affected by ascites and liver lipid deposition. Previous studies have shown that ARFI measurements of liver  stiffness have similar accuracy as transient elastography in the diagnosis of cirrhosis [26,27], while in patients with decompensated cirrhosis, there are some advantages over transient elastography.
In our study, it was shown that the LSPS model was superior to single indicators, such as spleen stiffness and liver stiffness, in identifying cirrhotic patients with a high risk of rebleeding after standard therapy. A significantly higher rate of rebleeding was observed in the high-LSPS group. From our results of subgroup analysis, the LSPS model was sufficient to assess the risk of rebleeding in secondary prophylaxis in cirrhotic patients without selected etiology, while the model is more potent in cirrhosis related to viral hepatitis. With or without NSBB therapy, the measurement of LSPS is important to evaluate the risk of rebleeding in patients.
As a result of our exclusion criteria, significantly, our conclusion cannot be extended to patients with severe portal vein thrombosis, hepatocellular carcinoma, or extrahepatic tumor and splenectomy. Due to the limitations of single-center observational studies, the inherent risk of selection bias could not be avoided. It should be noted that in our study, there were more patients with viral hepatitis cirrhosis and primary biliary cirrhosis and fewer patients with alcoholic liver disease  and nonalcoholic fatty liver disease. Furthermore, the focus on the patients' clinical symptoms was lower than expected. Despite the limitations mentioned, the study produced reliable data. More RCTs are needed to evaluate the diagnostic value of the LSPS model as a risk-stratification model in the secondary prophylaxis of EGVB. More research is suggested to evaluate the therapeutic value of the LSPS model.

Conclusion
The LSPS model is a reliable, noninvasive method for predicting a high risk of rebleeding for the secondary prophylaxis of patients with EGVB. Close follow-up is essential for patients with high LSPS. More advanced therapy, such as TIPSs, should be considered for patients with high LSPS.