Nomogram for Predicting Postoperative Portal Venous Systemic Thrombosis in Patients with Cirrhosis Undergoing Splenectomy and Esophagogastric Devascularization

Objectives The aim of the study is to develop a nomogram for predicting postoperative portal venous systemic thrombosis (PVST) in patients with cirrhosis undergoing splenectomy and esophagogastric devascularization. Methods In total, 195 eligible patients were included. Demographic characteristics were collected, and the results of perioperative routine laboratory investigations and ultrasound examinations were also recorded. Blood cell morphological traits, including the red cell volume distribution width (RDW), mean platelet volume, and platelet distribution width, were identified. Univariate and multivariate logistic regressions were implemented for risk factor filtration, and an integrated nomogram was generated and then validated using the bootstrap method. Results A color Doppler abdominal ultrasound examination on a postoperative day (POD) 7 (38.97%) revealed that 76 patients had PVST. The results of the multivariate logistic regression suggested that a higher RDW on POD3 (RDW3) (odds ratio (OR): 1.188, 95% confidence interval (CI): 1.073–1.326), wider portal vein diameter (OR: 1.387, 95% CI: 1.203–1.642), history of variceal hemorrhage (OR: 3.407, 95% CI: 1.670–7.220), and longer spleen length (OR: 1.015, 95% CI: 1.001–1.029) were independent risk parameters for postoperative PVST. Moreover, the nomogram integrating these four parameters exhibited considerable capability in PVST forecasting. The nomogram's receiver operating characteristic curve reached 0.83 and achieved a sensitivity and specificity of 0.711 and 0.848, respectively, at its cutoff. The nomogram's calibration curve demonstrated that it was well calibrated. Conclusion The nomogram exhibited excellent performance in PVST prediction and might assist surgeons in identifying vulnerable patients and administering timely prophylaxis.


Introduction
Liver cirrhosis has become the 11 th most lethal disease worldwide, causing approximately 100 million deaths every year [1]. Portal hypertension (PH) is a typical manifestation among patients with cirrhosis and is mainly characterized clinically by hypersplenism and portosystemic collateral varices. Notably, the rupture of esophagogastric varices (EV) is a common complication that can be lethal without timely hemostasis. Certain diferences exist in the management of variceal hemorrhage between Western and Asian doctors. Medication, endoscopic ligation, and intervention therapy are frst-line treatments in Western countries [2]. However, the etiology of cirrhosis in China and Western countries is diferent: liver cirrhosis in China is mainly derived from hepatitis B virus infection, characterized by poor liver function and afecting a large population, whereas alcoholic and nonalcoholic hepatitis and hepatitis C virus infection account for most cases of cirrhosis in Western countries [3]. Splenectomy combined with esophagogastric devascularization (SED) is widely performed in China because of its great ability to decrease pressure in portal veins, ameliorating the degree of EV, and improving immune function without harming liver function [4]. For patients admitted with emergent uncontrollable bleeding, SED is quite efective in hemorrhage control and has a lower rebleeding rate than endoscopic therapy and higher economic benefts than other therapies in the long run [5]. Tus, SED is still an important treatment for PH in China.
However, portal venous systemic thrombosis (PVST) (defned as the formation of a thrombus in either the intrahepatic portal vein, extrahepatic portal vein, splenic vein, or superior mesenteric vein) is a common and serious complication following SED, with an incidence rate ranging from 18.9% to 55% [6,7]. Furthermore, the incidence rate of spontaneous PVST in the general population with cirrhosis is 5%-15% [8], indicating a prevalence of thrombophilia among patients receiving SED. In patients with PH, PVST has been associated with a high rebleeding rate and liver transplantation failure [9,10]. Although PVST may manifest as abdominal discomfort or continuous low-grade fever, it is mostly asymptomatic until complications occur, making timely prophylaxis challenging. Tus, solving how to assess the risk of early PVST after SED is crucial for the prognosis of patients with cirrhosis.
To date, studies investigating PVST following SED have reported various risk factors, including a wider portal vein diameter (PVD), poorer liver function, prolonged prothrombin time (PT), higher levels of preoperative aspartate aminotransferase (AST) and alanine aminotransferase (ALT), and postoperative thrombocytosis [11][12][13][14]. However, a consensus has yet to be achieved. A nomogram is an effective tool for the visualization and application of regression models and can be helpful for risk assessment and clinical decision-making [15]. In this study, we developed a nomogram for the risk evaluation of postoperative PVST by retrospectively analyzing the hospitalization data of patients with cirrhosis undergoing SED to provide evidence for the early prophylaxis of PVST.

Study Design and Participants.
Tis single-center retrospective case-control study conformed to the Declaration of Helsinki (as revised in 2013) and was approved by the Medical Ethics Committees of the Second Hospital of Nanjing.
Patients with cirrhosis who underwent open SED at the Department of Hepatobiliary Surgery, Second Hospital of Nanjing, between January 2013 and December 2018 were included. A detailed description of the SED procedure is provided in the following section. Te inclusion criteria were as follows: (1) patients aged between 18 and 75 years, (2) patients diagnosed with liver cirrhosis based on pathological or radiological evidence, (3) patients with PH and severe EV (varices in the form of a serpent, nodule, or tumor or moniliform with or without red signs), (4) patients with a platelet count <50 × 10 9 /L, white blood cell (WBC) count <3 × 10 9 /L, and/or with a history of variceal bleeding resulting from PH, and (5) patients with a Child-Pugh score of A or B.
Te exclusion criteria were as follows: (1) patients with hepatocellular carcinoma or other malignant diseases identifed before or during surgery, (2) patients with PVST prior to surgery, (3) patients with hematological or immune system diseases, (4) patients who had received anticoagulants prior to surgery, (5) patients with organ failure or refractory ascites, and (6) patients with missing data.

Data Collection.
Basic data including age, sex, body mass index (BMI), history of hypertension, history of diabetes mellitus, history of variceal hemorrhage, cirrhosis etiology, the presence of ascites, emergency at admission, the model for end-stage liver disease (MELD) score, Child-Pugh score, PVD, spleen thickness, spleen length, and portal venous fow velocity were recorded. Preoperative laboratory blood tests to measure WBC, red blood cells (RBCs), RDW, platelet distribution width (PDW), mean platelet volume (MPV), hematocrit, coagulation parameters, AST, ALT, total bilirubin, albumin, globin, and postoperative RDW, PDW, MPV, and hematocrit were also collected on a postoperative day (POD) 1, 3, and 7.
All the patients received an abdominal color Doppler ultrasound examination on POD7, and the diagnosis of PVST was confrmed by two experienced imaging experts. Ultrasound follow-up was conducted every 2 weeks for all patients.

Surgical
Procedure. All patients underwent SED using laparotomy. Te abdomen was opened using an L-shaped left subcostal incision, and a splenectomy was performed. Te peripheral ligaments of the spleen, including the gastrocolic ligament, splenocolic ligament, gastrosplenic ligament, splenorenal ligament, and splenophrenic ligament, were cut and suture ligated. Te splenic hilum was rigorously dissected, and the splenic artery and vein were carefully transected and ligated. Soon after splenectomy, devascularization was conducted. Te portosystemic collateral branches at the lower esophagus and fundus of the stomach, including the short gastric vein, posterior gastric vein, left inferior phrenic vein, and esophagogastric branches of the gastric coronary vein, were identifed and ligated. Te esophagus was dragged down, and the high esophageal branches of the left gastric vein were sutured at a distance of approximately 10 cm from the fundus of the stomach; the arteries accompanying these veins were divided accordingly.

Statistical Analysis.
Te normality of continuous data were analyzed using the Shapiro-Wilk test. Te Student's ttest was performed on data with normal distribution, and the Mann-Whitney U test was used for nonparametric tests. Categorical variables were compared using either the Chisquare test or Fisher's exact test. Normally distributed continuous variables are presented as the mean (standard deviation) and nonnormal variables as the median (interquartile range, IQR). Categorical variables are presented as the exact number (percentage, %). Signifcant variables in the univariate comparison were further analyzed using a multivariable stepwise logistic regression.
Te nomogram was developed based on the multivariate logistic regression model using the "rms" R package. Te bootstrap method [16] (1,000 resamples) was used for internal validation, and the corresponding calibration and receiver operating characteristic (ROC) curves were plotted using "plotROC" R packages. Te cutof value of the ROC was calculated according to the Youden index [17]. A twotailed p value of less than 0.05 was considered signifcant throughout the analysis. R software (version 3.6.1) was used for all the statistical analyses.

Participant Characteristics.
A total of 218 patients with cirrhosis were admitted and underwent SED for either the primary or secondary prevention of variceal hemorrhage at the Department of Hepatobiliary Surgery, Second Hospital of Nanjing, between January 2013 and December 2018. A total of 23 patients were excluded, in line with the exclusion criteria, as follows: 10 with preoperative PVST, 5 with hepatocellular carcinoma, 1 with thalassemia, and 7 without the required data. Te study fowchart is presented in Figure 1. Terefore, 195 eligible patients were included in the analysis. Te abdominal color Doppler ultrasound examination on POD7 (38.97%) revealed that 76 of these patients had developed PVST and they were assigned to the study group; the remaining 119 patients without PVST were assigned to the control group. Te location of the PVST is illustrated in Supplementary Table 1.
Te age of the entire cohort ranged from 23 to 72 years, with an average age of 47.8 years. Men constituted 128 (65.64%) of the patients. Te commonest cause of cirrhosis was infection with hepatitis B (140 patients), with the other causes being hepatitis C infection (34 patients), alcoholic hepatitis (14 patients), primary biliary cirrhosis (4 patients), drug-induced cirrhosis (2 patients), and schistosomiasis cirrhosis (1 patient).
All patients underwent SED using laparotomy (the surgical procedure is described above). Detailed data from the PVST and non-PVST subgroups are presented in Table 1.   Table 2. WBC, hematocrit, lymphocyte, hemoglobin, AST, and fbrinogen were eliminated from the fnal model.

Performance of the Risk Factors and Nomogram.
A ROC analysis was conducted to validate the performance of the independent risk factors and logistic model in forecasting postoperative PVST. Te ROC curves of the individual risk parameters are presented in Figure 2. Te area under the curve (AUC) and the sensitivity and specifcity at their respective cutofs are listed in Table 3. Te AUC for the RDW3, PVD, spleen length, and history of bleeding were 0.685 0.757, 0.655, and 0.60, respectively. Te model integrating the four parameters achieved an AUC of 0.83, with a sensitivity and specifcity of 0.711 and 0.849, respectively, indicating the superiority of the model over any individual factor in PVST forecasting. Te nomogram was generated using a logistic model (Figure 3). Each risk factor value in the nomogram was assigned a weighted score from the point bar at the top, and their sum was mapped to the risk bar at the bottom,   representing the PVST risk of an individual. Te bootstrapped concordance index (C-index) of the nomogram was 0.83, emphasizing its excellent discrimination capability. A calibration curve was plotted for the internal validation of the nomogram (Figure 4), which indicated that the predicted risk was in close agreement with the observed scenario.

Discussion
PH and hypersplenism are common complications of cirrhosis. Splenectomy greatly reduces pressure in the portal system and ameliorates variceal bleeding and ascites in patients with cirrhosis; however, it also increases patients' exposure to infection [18][19][20]. Splenic arterial embolism, which has the advantages of less trauma, a quicker recovery time after surgery, and a confrmed curative efect on hypersplenism, is recognized as a surrogate modality for splenectomy [21]. Unfortunately, both splenectomy and splenic arterial embolism have an elevated risk of postoperative PVST [7,22], with a recent metaanalysis demonstrating that the incidence of postoperative PVST after these two surgical procedures were statistically similar [23]. One study reported that patients undergoing SED had a relatively high incidence of postoperative PVST (18.9%-55% vs. 5%-15%) [6]. Te incidence of postoperative PVST following SED in this study was 38.97% (76/195), which is consistent with the reported incidence.
To date, multiple factors, such as postoperative thrombocytosis, a wider preoperative splenic vein diameter, faster portal blood fow, prolonged PT, larger spleen volume, and devascularization, have been reported as risk factors for postoperative PVST [11,24] but a consensus is yet to be reached. Our univariate and multivariate analyses identifed RDW3, PVD, spleen length, and a history of bleeding as independent risk factors for PVST after SED. Te nomogram integrating these four factors exhibited an excellent performance in PVST prediction, achieving an AUROC of 0.83.  0  10  20  30  40  50  60  70  80  90  100   10  12  14  16  18  20  22  24  26  28  30  8   16  20  24  28  32   . Te x-axis represents the predicted risk of PVST using the nomogram, and the y-axis represents the observed risk. 6 Canadian Journal of Gastroenterology and Hepatology Tis study identifed that preoperative PVD was an independent risk factor for postoperative PVST (OR: 1.387; 95% CI: 1.203-1.642). Previous studies [14,25] have demonstrated that a wider preoperative PVD is associated with a greater decrease in blood fow and velocity following spleen resection, resulting in longer blood retention in the portal vein. In addition, the stumps of the dissected vessels were more likely to trigger turbulence and lead to the development of PVST.
Tis study also revealed spleen length as a risk factor for PVST (OR: 1.015; 95% CI: 1.001-1.029). Spleen volume refects the severity of PH and the degree of hypersplenism; the larger the spleen is, the more serious the PH, and thus, the greater the likelihood of PVST [26]. In addition, our study demonstrated that patients with cirrhosis with a history of variceal hemorrhage are more prone to PVST (OR: 3.407; 95% CI: 1.670-7.220; p < 0.001). Variceal hemorrhage is the manifestation of a decompensated liver, and patients with a history of bleeding have a relatively poor prognosis [27]. Previous studies [28] have proposed that the hypercoagulability of blood may be more severe in patients with cirrhosis with poorer liver function because of their resistance to thrombomodulin activity, a critical factor mediating the anticoagulation process. Consistent with the fndings of this study, Xu et al. [29] demonstrated that variceal bleeding was more common in the PVSTgroup than in the non-PVST group (p � 0.006) but they failed to include this factor in their fnal model.
Te RDW is one of the parameters in routine blood tests-it is an indicator of the variability in erythrocyte size and is frequently overlooked in thrombosis events. Recently, the role of the RDW in thrombotic diseases has attracted signifcant attention from researchers. Relevant studies have demonstrated that the RDW is a powerful indicator of myocardial infarction, cerebral thrombosis, pulmonary embolism, and atrial fbrillation [30][31][32][33]. Lappegard et al. conducted a prospective study of the general population and discovered that individuals with a higher RDW were more likely to experience an incident stroke during follow-up (hazard ratio: 1.55, 95% CI: 1.16-2.06) [34]. Lippi et al. revealed that the RDW (cutof 14.6%) is an independent risk factor for deep vein thrombosis and pulmonary embolism (OR: 2.52; 95% CI: 1.42-4.47), although the underlying mechanisms have yet to be determined [35].
Increasing evidence suggests that the RDW is a feasible surrogate indicator for systemic infammatory response and oxidative stress [36][37][38]. Relevant studies have associated the RDW with typical infammatory indicators, including the C-reactive protein, erythrocyte sedimentation rate, platelet count, and cytokines such as interleukin (IL)-8 and tumor necrosis factor (TNF) [26,33]. Patients with cirrhosis often present with chronic infammation and high coagulability, and the association between the RDW and infammation may be because the elevation of infammation factors, such as TNF and IL, aggravates microcirculation hypoxia, resulting in hemodynamic changes. All these changes could suppress the maturity of erythrocytes, cause RBC heterogeneity elevation, and infuence the RDW. Tus, the elevated RDW may be attributed to the infammation response in the body.
However, the interactions between infammation and coagulation are complex. For one thing, the cytokines that trigger the infammatory cascade can also initiate the thrombosis process and inhibit the physiological mechanisms of anticoagulation. For another, infammation activity is modulated by the components of coagulation as feedback [39].
Tis study identifed that patients with postoperative PVST exhibited a higher RDW either before or after SED; however, only the RDW on POD3 was an independent risk factor for postoperative PVST (OR: 1.188; 95% CI: 1.073-1.326; p � 0.001). Tis may be because systemic infammation is at its most severe on POD3 and is indicative of a high likelihood of PVST. An excessive infammatory response may lead to severe epithelial damage, facilitating the adhesion of platelets. Furthermore, considering the higher postoperative RBC levels in the PVST group, it is hypothesized that the elevated RDW refects the hemopoiesis attributed to the cytokines secreted during infammation, and this may subsequently result in an increase in blood viscosity.
Te current consensus on the prophylaxis of postoperative PVST is to administer anticoagulants once the risk of bleeding is eliminated. Following splenectomy, the prophylactic use of anticoagulants, such as low-molecularweight heparin and rivaroxaban, can help to reduce the occurrence of PVST [40]. In our center, all the patients who had undergone SED were prescribed low-molecular-weight heparin (4,000 IU, i. h, once a day) for a week as the routine prophylaxis for PVST. Patients diagnosed with PVST using abdominal color Doppler ultrasound on POD7 were prescribed 100 mg of aspirin daily until the portal vein was recanalized. However, the rational selection of patients eligible for anticoagulant therapy is challenging because nonspecifc anticoagulation may cause coagulation dysfunction. Tus, we developed a nomogram to identify patients at high risk of postoperative PVST.
Tis study has some unavoidable limitations. First, because the study had a retrospective design and was conducted in a single center with a small sample size of only 195 patients, a selection bias may be inevitable. Second, the validation of the logistics model was based solely on interval bootstrapping methods. External large-scale sample validation and future multicenter prospective studies are warranted to test the broader application of the proposed model. Tird, most of the candidate factors included in the multivariable analysis were selected from the literature, and the independent risk factors may be afected by the inclusion of additional candidates. Last, the value of the RDW is infuenced by multiple factors not considered in this study, such as the absorbance of vitamin B12 and iron.
In summary, we proposed a nomogram integrating preoperative PVD, a history of variceal hemorrhage, spleen length, and the RDW on POD3, which exhibited an excellent capability for identifying patients vulnerable to PVST soon after SED. Our nomogram may aid surgeons in forecasting postoperative PVST and administering timely prophylactic treatment to patients at high risk of thrombosis. In the future, appropriate large-scale external validation may confrm the utility of the proposed nomogram in PVST forecasting.

Data Availability
Te datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Ethical Approval
Tis single-center, retrospective, case-control study conformed to the Declaration of Helsinki (as revised in 2013) and was approved by the Medical Ethics Committees of the Second Hospital of Nanjing.