Study population

The study population consisted of patients with chronic HCV infection consecutively enrolled in the ongoing prospective PITER cohort (the Italian Platform for the study of Viral Hepatitis Therapy) from 40 centers specialized in liver and infectious diseases from teaching hospitals and general hospitals throughout Italy.¹⁷ The data were accurately and prospectively collected in dedicated electronic case report forms at pre-defined time points, that is, before DAA therapy and during the follow-up, following the clinical practice of each center. The Child-Pugh-Turcotte (CPT) stage, albumin-bilirubin (ALBI) score,¹⁸ Baveno expanded,¹⁹ Fibrosis-4 (FIB-4) index,²⁰ and RESIST score²¹ were calculated using well-established formulas. Following the PITER protocol, decompensated cirrhosis was diagnosed based on the presence or appearance of ascites and/or gastrointestinal bleeding due to portal hypertension and/or hepatic encephalopathy and/or jaundice and/or spontaneous bacterial peritonitis.⁸ The presence of hepatic steatosis by abdominal ultrasound (abdominal-US) examinations associated with hypertension, cardiovascular disease, type 2 diabetes, and body mass index (BMI) >25 kg/m² were used as surrogate markers of metabolic syndrome in patients who did not report alcohol consumption.

Exclusion and inclusion criteria and duration of follow-up

We evaluated patients with pre-treatment liver cirrhosis included in the PITER cohort between June 2014 and March 2020 for inclusion in this study. Cirrhosis was defined by liver biopsy (Metavir ≥4 or Ishak score ≥6) or Transient Elastometry (liver stiffness measurement [LSM] >12.5 kPa) or biochemical and/or imaging indications of clinically significant portal hypertension (i.e., presence of esophageal and/or gastric varices and/or platelet [PLT] count ≤150,000/μL with spleen enlargement). A PLT count of ≤120,000/μL and an albumin level of ≤3.5 g/dL were chosen as the cut-off for determining the risk of liver disease progression due to their better performance in predicting severe liver cirrhosis in our study population, as also reported in previous studies.^{22, 23}

As per the initial Italian regulations on treating patients with HCV-related liver cirrhosis without an indication for LT, only those belonging to CPT stages A and B at pre-treatment evaluation were eligible for antiviral treatment.²⁴ Thus, no CPT stage C patient at the time of the treatment start was available for this analysis. Patients with a pre-treatment compensated CPT-A or B class or with a previous decompensation history but under a stable clinical condition and having a compensated liver disease within the last 12 months before DAA treatment were included.

Patients previously diagnosed with PVT or with a diagnosis of malignant PVT and those who either underwent or were awaiting LT were excluded. A total of 6 patients with malignant PVT were excluded from the present analysis (4 with pre-therapy hepatocellular carcinoma (HCC) and 2 with post-therapy HCC). Patients with less than 12 weeks of follow-up after the EOT (i.e., in whom SVR could not be evaluated) were excluded.

For each patient, the duration of follow-up was calculated between the EOT and the last follow-up.

Screening for portal vein thrombosis

The results of abdominal ultrasound examinations, routinely performed for HCC screening every 6 months for patients with cirrhosis, were prospectively collected. The PVT was defined as the presence of hyper-echogenic material within the portal vein main trunk and/or intrahepatic branches, determined by ultrasound examinations or through the visualization of non-contrasted material at the venous phase of the CT scan or T2 sequence at MRI. PVT was classified as complete when blood flow in the portal vein was absent and as partial when the lumen was only partially occluded, and the flow was still present at Doppler evaluation. Only PVT occurring after the EOT was considered to be de novo PVT.

As per the current guidelines, all patients with a new diagnosis of PVT based on ultrasound examinations underwent second-line imaging (either CT scan or MRI).^{5, 6} In patients with a history of HCC, the malignant nature of thrombosis was excluded according to the following criteria: the presence of thrombus enhancement, PVT adjacent to HCC, and an increase in the diameter of the portal vein.²⁵

Data collection

The case report forms recorded detailed demographic, clinical, and laboratory characteristics. The outcomes collected during the follow-up after viral eradication included de novo PVT, development of hepatic decompensation, de novo HCC, LT, and patient survival. In patients who experienced PVT, data on the extension of thrombosis were also collected.

Statistical analysis

The main characteristics of the patients are reported as the median and interquartile range (IQR) or as proportions (N and %) for continuous and categorical variables, respectively. Albumin, alanine aminotransferase (ALT), bilirubin, creatinine, International Normalized Ratio (INR), PLT, and LSM were evaluated using clinically relevant threshold cut-offs below or above which the risk could be assumed to be homogeneously distributed.

The Mann-Whitney U test was used for analyzing continuous variables to assess differences between distributions, and the Chi-squared test was used for comparing proportions. The Wilcoxon matched-pairs signed-rank test was used to assess the differences between matched pairs of observations.

To evaluate the effect of demographic data and clinical variables on the risk of development of PVT, a Fine-Gray competing-risks regression model was applied, considering death or LT as competing events (i.e., events that could prevent the event of interest). Potential multicollinearity was tested among all the variables included in the final models. The possible presence of multicollinearity was checked by estat vce, corr STATA command. A correlation coefficient <0.5 indicated predictor variables in the same model are not correlated, and they can independently predict the value of the dependent variable.

Univariate and multivariate analyses were performed. In the multivariable analysis, a forward stepwise selection method was used to obtain robust results. The results are presented as a sub-hazard ratio (subHR) and the corresponding 95% Confidence Interval (95% CI).

For all analyses, differences were considered to be statistically significant at p < 0.05. All statistical analyses were performed on STATA version 16.1 (Stata Corp, College Station, TX, USA).

Baseline characteristics of the study population based on the occurrence of PVT

The baseline characteristics of 1609 patients who achieved the SVR and were followed up for a median time of 38.3 months (IQR: 25.1–48.7 months) after EOT are shown in Table 1. Of these, 32 patients (2.0%) experienced PVT; the median time of PVT development was 36.1 months after EOT; IQR: 19.1–46.2 months). PVT was complete in 9 patients and partial in 23 patients. Of these patients, 21 (65%) were treated with anticoagulation. The median duration of anticoagulant therapy was 7 months (IQR: 3.0–24.0). Of the treated patients, 3 (14%) achieved complete resolution of PVT and 5 (24%) achieved a partial response (i.e., any improvement in PVT extension without complete resolution). In 13 (62%) patients, PVT either remained stable or progressed despite anticoagulant treatment.

Age, sex, BMI distribution, alcohol use, and presence of surrogate markers of metabolic syndrome were similar among patients who achieved the SVR and developed PVT and those who did not develop PVT. In contrast, a PLT count of ≤120,000/μL, albumin levels ≤3.5 mg/dL, bilirubin >1.1 mg/dL, a history of liver decompensation, and ALBI, Baveno, FIB-4, and RESIST scores were all significantly different among patients who developed PVT versus those who did not (indicating greater severity of advanced liver disease in PVT patients vs. non-PVT patients). The use of non-selective beta-blockers (NSBBs) was more frequent in patients who developed PVT versus those who did not. The proportion of patients who were administered anticoagulants was similar between the two groups (p = 0.262, data not shown). A transjugular intrahepatic portosystemic shunt was reported in four patients who did not develop PVT and in none of the patients who developed PVT (data not shown). The overall comorbidity pattern and those that might be associated with liver disease progression were not different among patients who developed PVT and those who did not (Figure S1).

Pre-treatment predictive factors for the occurrence of PVT

Considering death or LT as competing risk events, the pre-treatment variables independently associated with the development of PVT in patients who achieved the SVR were evaluated (Table 2). No multicollinearity in the parameters chosen to be evaluated as predictive factors of PVT (all correlation coefficients were <0.5) was found. The presence of esophageal varices and an ALBI grade ≥2 were significantly associated with the risk of PVT development. Similar results were obtained after excluding the data on patients with a history of previous decompensation from the analysis (Table S1).

Prospective evaluation of the predictors of PVT measured after viral eradication

The changes in the biochemical markers of chronic liver disease severity (PLT, albumin, and bilirubin measurement) before antiviral treatment (baseline) and after a median time of 6 months after EOT (IQR: 4.7–7.9 months, with no difference among patients who developed PVT and those who did not: p = 0.977) according to the PVT development are shown in Figure 1. For these three parameters, there were significant variations (all p < 0.001) in the values evaluated after viral eradication versus pre-treatment values in patients who did not develop PVT, whereas in those who developed PVT, PLT count, albumin, and bilirubin levels remained unchanged (p = 0.759, p = 0.054, and p = 1.000, respectively).

The biochemical characteristics evaluated post-EOT (independently and combined with the ALBI or RESIST scores) have shown that despite viral eradication, patients who developed PVT remained with parameters of more severe liver disease compared with those who did not (Table 3).

Post-treatment predictive factors related to the occurrence of PVT

Post-treatment evaluated variables associated with the development of PVT in patients who achieved the SVR are shown in Table 4. The correlation coefficients suggested weak correlations (all coefficients were <0.5), indicating no multicollinearity in the parameters chosen for the regression model used. The presence of esophageal varices and post-treatment ALBI grade ≥2 were significantly associated with the risk of PVT. For patients with esophageal varices, the adjusted (independent) risk of developing PVT was 9.32 times higher than the risk of developing PVT in patients with no esophageal varices. For patients with ALBI grade ≥2, the adjusted (independent) risk of developing thrombosis was 5.50 times higher than the risk of patients with ALBI grade 1; similar results were obtained after excluding patients with previous decompensation from the analysis (Table S2).

The competing risk regression curves showed that for all patients in whom the presence of esophageal varices was reported, the cumulative PVT incidence rates in patients with post-treatment ALBI grade ≥2 versus those with post-treatment ALBI grade 1 were 27% versus 6% at 60 months and 43% versus 10% at the end of the observation period (80 months) (Figure 2).

The outcomes of patients according to the development of PVT after the SVR are shown in Table S3. Patients who experienced PVT had higher rates of liver decompensation and mortality.