Patient characteristics and clinical outcomes

The study included 276 patients with recurrent or refractory ascites treated with TIPS. The median age of the patients was 57 years, with 173 (62.7%) being male and 103 (37.3%) being female. One hundred forty patients had hepatitis B virus and received continued antiviral therapy. Hepatitis B virus DNA was detected in 49 (35.0%) patients. A total of 246 patients had varices, among whom 155 patients had a history of variceal bleeding. All of these patients received non-selective beta-blockers or endoscopic therapy according to the guideline for portal hypertension [8]. Partial portal vein thrombosis was diagnosed in 107 patients before TIPS placement, and they all received anticoagulation therapy according to the treatment guideline [8]. All patients were administered furosemide (20–120 mg/d) and spironolactone (40–160 mg/d) before TIPS. The median duration of diuretic use was 24.0 (8.0, 48.0) months. Additionally, all patients underwent at least one LVP procedure, and 162 patients underwent more than three LVP procedures within 12 months. During the TIPS procedure, a gastrorenal shunt was found in eight (2.9%) patients and a splenorenal shunt was found in five (1.8%) patients; all of these patients underwent shunt embolisation. The follow-up period was 21.6 (7.5, 41.6) months. The details of patients’ characteristics are shown in Table 1.

Table 1 Demographic and baseline of all patients and group with PPG threshold of 7 and 11 mmHg

During the follow-up period, a total of 151 (54.7%) patients died, with 122 (44.2%) of them experiencing liver-related death. Out of the total number of patients, 73 (26.4%) experienced a recurrence of ascites that was uncontrollable with diuretics, while only 11 (4.0%) underwent LVP. The median time to the recurrence of ascites was 2.9 (1, 14.7) months. Among the 68 patients with reduced ascites, five resumed the use of diuretics. Only 28 (10.2%) patients experienced hemorrhage. Among the 136 patients who experienced OHE after TIPS placement, 25 (18.4%) required hospitalization more than once, and three of these patients underwent stent reduction. Meanwhile, 27 (10.2%) patients experienced shunt dysfunction. The outcomes are summarized in Table 2.

Table 2 Clinical outcomes during the follow-up periodImmediate PPG after TIPS procedure

The PPG was decreased from 23 (19, 27) mmHg to 9 (7, 10) mmHg post-TIPS procedure in all patients, indicating a reduction of 62% (53%, 72%). Based on the information from density plots and the ROC curves, the thresholds for lower mortality may be a post-TIPS PPG of > 7, 8, or 9 mmHg, or a decrease rate of < 75%, and the potential thresholds for ascites control may be < 10 or 11 mmHg for post-TIPS PPG, or a 65% decrease in PPG (Fig. 1A–B). Supplementary Fig. 2–4 displays potential thresholds for other clinical outcomes. The potential thresholds were used as categorical variables for univariable and multivariable regression analyses.

Fig. 1

The density plots of post-TIPS PPG and forest plots of interactive analysis. a, b Distributions of post-TIPS PPG in patients with ascites and liver-related death are shown with density plots, and (c) the results of interactive analysis show no interaction between thresholds of post-TIPS PPG and types of ascites. TIPS, transjugular intrahepatic portosystemic shunt; PPG, portal pressure gradient; CI, confidence intervals; HR, hazard ratio

The effect of PPG reduction on clinical outcomes

The univariable and multivariable analyses of liver-related death revealed that a post-TIPS PPG < 7 mmHg was identified as a risk factor for mortality (p = 0.003, HR = 1.803, 95% CI 1.220–2.665) (Supplementary Table 1). Considering other secondary endpoints as influencing factors one by one, we found that the recurrence of ascites (p = 0.033, HR = 1.520, 95% CI 1.003–2.235), hemorrhage (p = 0.019, HR = 1.911, 95% CI 1.115–3.278), and OHE (p = 0.009, HR = 1.666, 95% CI 1.136–2.443) were also risk factors for mortality (Supplementary Table 2). Meanwhile, a post-TIPS PPG < 11 mmHg was a protective factor for ascites control (p = 0.012, HR = 0.524, 95% CI 0.316–0.868) (Supplementary Table 3). No specific PPG threshold was identified as an independent factor for other outcomes (Supplementary Table 4–6). The interactive analysis revealed no significant interaction between the PPG < 11 mmHg and recurrent/refractory ascites (p = 0.449) or between PPG < 7 mmHg and recurrent/refractory ascites (p = 0.488, Fig. 1C).

The 7 mmHg as a potential PPG threshold for survival

According to the post-TIPS PPG of 7 mmHg, 202 patients had post-TIPS PPG ≥ 7 mmHg and 74 had PPG < 7 mmHg (Table 1). Patients with post-TIPS PPG ≥ 7 mmHg had significantly lower mortality than those with PPG < 7 mmHg (51.0% vs 66.6%, p = 0.004, HR = 1.752, 95% CI 1.202–2.555), and the difference persisted after accounting for competing events (p = 0.015, HR = 1.605, 95% CI 1.098–2.346, Fig. 2A, B). There was no significant difference between the two groups categorized by thresholds of post-PPG < 7 mmHg in recurrence of ascites (35.7% vs 36.3%, p = 0.729, HR = 1.097, 95% CI 0.650–1.852), hemorrhage rate (16.5% vs 12.7%, p = 0.778, HR = 0.878, 95% CI 0.356–2.169), OHE (55.6% vs 65.7%, p = 0.535, HR = 1.128, 95% CI 0.771–1.652), or shunt dysfunction (21.6% vs. 7.5%, p = 0.138, HR = 0.403, 95% CI 0.121–1.341) (Supplementary Fig. 5). The Fine–Gray tests demonstrated robust results (Supplementary Fig. 6).

Fig. 2

Differences in liver-related mortality between groups with a PPG threshold of 7 mmHg. The differences are showed using Kaplan–Meier curve (a) and adjusted with competing risk analysis (b) in all patients, which are also performed after PSM (c–d). PPG, portal pressure gradient; PSM, propensity score matching; CI, confidence intervals; HR, hazard ratio

After 2:1 PSM, there were 117 patients in ≥ 7 mmHg group and 70 patients in < 7 mmHg group (Table 3). The significant differences were found in sex, age, aspartate aminotransferase, creatinine, and pre-TIPS PPG between two groups in baseline, and these characteristics were comparable after PSM (Tables 1 and 3, and Supplementary Fig. 7). The survival curve indicated a difference between the two groups in liver-related death (≥ 7 mmHg vs < 7 mmHg, 55.3% vs 66.7%, p = 0.015, HR = 1.702, 95% CI 1.111–2.607 Fig. 2C), while no significant difference was observed in other clinical outcomes between two groups (recurrence of ascites: 35.5% vs 37.5%, p = 0.647, HR = 1.143, 95% CI 0.645–2.026; hemorrhage, 13.6% vs 13.1%, p = 0.771, HR = 1.163, 95% CI 0.422–3.206; OHE, 60.6% vs 64.4%, p = 0.941, HR = 1.016, 95% CI 0.664–1.555; shunt dysfunction, 14.8% vs 7.8%, p = 0.481, HR = 0.628, 95% CI 0.172–2.288) (Supplementary Fig. 8). The competing risk analysis indicated robust results of significant difference between two groups in liver-related death (p = 0.025, HR = 1.633, 95% CI 1.065–2.503 Fig. 2D), and no difference was observed in other clinical outcomes (Supplementary Fig. 9).

Table 3 Demographic and baseline of patients with PPG threshold of 7 and 11 mmHg after PSM§The 11 mmHg as a potential PPG threshold for ascites control

There were 61 patients with post-TIPS PPG ≥ 11 mmHg and 215 with PPG of < 11 mmHg (Table 1). Patients with post-TIPS PPG ≥ 11 mmHg had a significantly higher incidence of recurrence of ascites compared with another group (44.6% vs 33.7%, p = 0.023, HR = 0.560, 95% CI 0.340–0.925), and similar patterns were observed after using competing risk analysis (p = 0.045, HR = 0.591, 95% CI 0.356–0.981, Fig. 3A, B). There was no significant difference between the two groups categorized by thresholds of post-PPG < 11 mmHg in liver-related death (52.4% vs 56.0%, p = 0.974, HR = 0.993, 95% CI 0.648–1.523), hemorrhage rate (13.3% vs 16.5%, p = 0.574, HR = 1.320, 95% CI 0.502–3.472), OHE (52.2% vs 58.8%, p = 0.266, HR = 1.270, 95% CI 0.833–1.936), or shunt dysfunction (28.8% vs 15.7%, p = 0.668, HR = 0.828, 95% CI 0.350–1.959), and the competing risk analysis indicated robust results (Supplementary Fig. 10–11).

Fig. 3

Differences in recurrence of ascites between groups with a PPG threshold of 11 mmHg. The differences are shown using Kaplan–Meier curve (a) and adjusted with competing risk analysis (b) in all patients, which are also performed after PSM (c–d). PPG, portal pressure gradient; PSM, propensity score matching; CI, confidence intervals; HR, hazard ratio

After 1:2 PSM, there were 56 patients with post-TIPS PPG ≥ 11 mmHg and 97 with PPG of < 11 mmHg (Table 3). The sex, age, aspartate aminotransferase, gamma-glutamyl transferase, ascites types, and pre-TIPS PPG of patients had significant differences between two groups in baseline, and these characteristics except ascites types were comparable after PSM (Tables 1 and 3, and Supplementary Fig. 12). The survival curve showed a significant statistical difference in ascites incidence post-TIPS (≥ 11 mmHg vs < 11 mmHg, 46.5% vs 32.3%, p = 0.013, HR = 0.468, 95% CI 0.258–0.851), and the difference was also observed with competing risk analysis (p = 0.048, HR = 0.549, 95% CI 0.305–0.991, Fig. 3C, D). However, patients in groups categorized by 11 mmHg of post-TIPS PPG had a similar cumulative incidences of liver-related death (52.3% vs 52.5%, p = 0.602, HR = 0.875, 95% CI 0.529–1.446), hemorrhage (14.2% vs 16.9%, p = 0.827, HR = 1.126, 95% CI 0.390–3.245), OHE (55.2% vs 57.4%, p = 0.274, HR = 1.302, 95% CI 0.811–2.089), and shunt dysfunction (31.6% vs 15.0%, p = 0.355, HR = 0.627, 95% CI 0.233–1.688), and the Fine–Gray test indicated robust results (Supplementary Fig. 13–14).

Clinical benefit for PPG7-11 patients

We compared the clinical outcomes of three groups of patients based on post-TIPS PPGs of < 7 mmHg, 7–11 mmHg, and ≥ 11 mmHg. The results indicated that patients with a PPG < 7 mmHg had a significantly higher liver-related mortality rate than that of the other two groups (< 7 mmHg vs 7–11 mmHg: 66.6% vs 50.5%, p = 0.002, HR = 0.531, 95% CI 0.353–0.798; < 7 mmHg vs ≥ 11 mmHg: 66.6% vs 52.4%, p = 0.107, HR = 0.670, 95% CI 0.412–1.091). However, patients with a PPG ≥ 11 mmHg had a higher rate of ascites recurrence than that of the other two groups (< 7 mmHg vs ≥ 11 mmHg: 36.3% vs 44.6%, p = 0.242, HR = 1.443, 95% CI 0.781–2.667; 7–11 mmHg vs ≥ 11 mmHg: 32.2% vs 44.6%, p = 0.014, HR = 1.983, 95% CI 1.150–3.421). Conversely, patients with PPGs within the range of 7–11 mmHg had lower mortality and ascites recurrence rates (Supplementary Fig. 15 and 16).

The results of the interaction analysis showed no significant interaction between different medical centers and the PPG threshold of < 7 mmHg in terms of survival (p = 0.384). Furthermore, no significant interaction between different medical centers and the PPG threshold of ≥ 11 mmHg in terms of ascites control (p = 0.319) was observed. Therefore, the results of the thresholds were not influenced by a center effect.