Portal vein thrombosis (PVT) is a relatively common complication in patients with cirrhosis, estimated to occur up to 28% of patients.1 Cirrhosis itself is now recognized as a hypercoagulable state which, in combination with low portal venous flow in the setting of portal hypertension, constitutes the main risks for PVT.2,3 Additionally, some genetic factors, such as the prothrombin gene mutation, also appear to contribute to PVT risk, although hypercoagulable disorders are uncommon overall.4,5 Other patient factors in cirrhosis that increase risk for PVT are not otherwise known. The impact of PVT on the natural progression of liver disease and portal hypertension are variable and remain contreversial.5,6 As a consequence, there is no consensus on the optimal management of PVT in patients with cirrhosis. Anticoagulation for PVT has been described in a few relatively small studies with good safety outcomes and recanalization rates ranging from 39% to 75%.5,6 However, there are no large studies establishing safety of anticoagulation in this population, with clinicians and patients having to weigh the risks of progressive or unresolved PVT against the at least theoretically increased bleeding risks with anticoagulation.
Portal vein thrombosis is estimated to be present in 9.7% ± 4.5% of patients undergoing liver transplantation (LT).7 Portal vein thrombosis presents unique challenges in patients undergoing LT due to technical needs for clot removal or alternate vascular reconstructions.8 Increased technical challenges of performing LT with PVT, with longer operative time, transfusion requirements, and reoperation have also been described.8,9 The argument can therefore be made that patients with PVT and listed for LT may significantly benefit from treatment of PVT before LT. At present, there is little data describing the impact or safety of anticoagulation for PVT before LT on waitlist or post-LT outcomes. Portal vein thrombosis has been associated with worse outcomes after LT, especially with complete thrombosis of the main portal vein trunk with extension below the confluence of the splenic and superior mesenteric veins.7 However, some single-center studies have described no impact on survival, particularly if physiological portal vein reconstruction can be achieved.10-12 Despite risks of worse outcomes, LT carries a favorable survival benefit in patients with PVT.13 Ultimately, a clearer understanding of the risk factors for pre-LT PVT and its impact on post-LT outcomes could help inform the assessment of risk/benefit and goals of anticoagulation in patients awaiting LT.
The aims of this study were to: (a) determine the impact of PVT at LT on post-LT patient and graft survivals; (b) assess the impact of PVT on waitlist outcomes; (c) identify the risk factors associated with development of PVT on the waitlist for LT; and (d) compare pre- and post-LT outcomes associated with persistent, absent, resolving, or developing PVT on the waitlist.
MATERIALS AND METHODS
We retrospectively studied all patients on the waitlist for LT or undergoing LT between 2002 and September 2013 as reported in the Organ Procurement and Transplant Network (OPTN) database. This period corresponds to inception of model of end-stage liver disease (MELD)–based listing prioritization and a relatively modern era for LT. Exclusion criteria for the analysis of waitlist outcomes were: age younger than 18 years, listing for multiorgan transplantation and previous LT. Additional exclusion criteria for post-LT outcomes analysis included: multiorgan transplantation, live donor LT, LT using split grafts, and organ donation after cardiac death (factors which may be associated with early post LT risk or technically different procedures). The presence or absence of PVT was reported at the time of transplant candidate registration (ie, time of listing for LT based on imaging data) and at the time of transplant recipient registration (ie, LT and may also include intraoperative findings). This allowed for grouping of patients according to the presence or absence of PVT at the time of listing and LT. Taking the reporting of the presence or absence of PVT at these time points at face value, we inferred the absence, persistence, resolution, or development of new PVT during the interval between listing and LT.
The analysis was performed in a sequence reflecting the stated aims. Patient demographics and clinical characteristics, donor characteristics, and transplant factors were compared in patients with and without PVT at LT. The primary endpoints were post-LT patient and graft survivals. The risks of post-LT death and graft failure were analyzed by Kaplan-Meier curves and life tables, with comparisons using the log rank test. The risks of death and graft failure associated with PVT were analyzed by univariate and multivariate to an increased risk for mainly early outcomes with PVT on Kaplan-Meier analysis. Simple and multiple logistic regressions were subsequently performed to study the predictors of 90-day mortality and graft failure. Models assessing the impact of PVT on post-LT patient and graft survival were adjusted for patient demographics, underlying liver disease, liver disease severity, patient condition at the time of LT, and donor risk index (forward conditional models, factors entered if P < 0.1 on univariate analysis, with significance threshold of P < 0.05). Pre-LT factors reflecting severity of disease/debilitation and those that may differ in patients with and without PVT were specifically analyzed, including markers of portal hypertension (ascites and hepatic encephalopathy), intensive care at LT, transjugular intrahepatic portosystemic shunt, and liver malignancy (though known tumor thrombus cases would not be knowingly transplanted).
A descriptive analysis was then performed comparing demographic and clinical factors in patients with and without PVT at listing, with direct comparison of waitlist removal codes for deceased donor LT, for reasons of death or being too sick for LT. In the analysis of risk factors for developing PVT on the waitlist, demographic and clinical factors at listing in patients without PVT at listing were analyzed on simple and multiple logistic regressions with an endpoint of PVT at LT. Waitlist time, listing, and LT MELD, and 90-day patient and graft survivals after LT were compared in patients with inferred persistent PVT (PVT at listing and LT), resolving PVT (PVT at listing but not LT), new PVT (PVT absent at listing but present at LT), and no PVT (no PVT at listing or LT).
Finally, propensity testing was performed for PVT at LT, and the propensity score for PVT was included in univariate and multivariate models testing the association of PVT with 90-day mortality and graft failure. Data were reported as numbers (percentages) for categorical variables, and median (interquartile range) for continuous variables. Categorical variables were compared using the χ2 test, and continuous variables using the Mann–Whitney test. All tests of significance were 2 sided with a P value less than 0.05 considered significant. All analyses were performed using SPSS 21 (IBM, New York, NY). The study was approved by the Institutional Review Board at Indiana University.
The OPTN database reported 50,393 adult recipients undergoing primary LT between 2002 and 2013. Portal vein thrombosis was reported as present at LT in 3321 (6.6%), absent in 45 249 (89.8%), and data were missing in 1823 (3.6%) patients. Demographic and clinical characteristics are compared in patients with and without PVT in Table 1. A diagnosis of fatty liver disease or cryptogenic cirrhosis, diabetes mellitus and higher body mass index were more frequent in patients with PVT. Patients with PVT were also slightly older, and more frequently men, and of Hispanic ethnicity. The MELD scores were numerically similar but patients with PVT were more frequently reported to have ascites and transjugular intrahepatic portosystemic shunts. The proportion of patients receiving intensive care and LT and donor risk index were similar in patients with and without PVT. Of note, multiple comparisons achieved statistical significance despite clinically trivial differences in median values, which is explained by the large sample size. A power analysis for continuous variables, using an α level of 0.05 and a power level of 0.80, the sample sizes of 3321 and 45 249 yielded statistically significant but miniscule effect size of 0.05. A similar power analysis for categorical variables demonstrated statistical significance for a miniscule effect size between 0.00016 and 0.00023, depending on the number of categories. That comparisons for some variables were statistically significant despite similar median values is explained by the non-normal distribution of many variables, where median values rather than slightly different mean values were reported.
Patient survival at 90 days, 1 year and 5 years after LT in patients with PVT was 91.5%, 88.6%, and 69.7% versus 95.1%, 92.8%, and 74% in patient without PVT at LT (P < 0.001). Graft survival at 90 days, 1 year, and 5 years after LT in patients with PVT was 88.4%, 80.7%, and 65.3% versus 92.8%, 86.1%, and 69.7% in patients without PVT at LT (P < 0.001). In univariate and multivariate Cox regression models, PVT at LT was an independent predictor of mortality (hazard ratio, 1.21; 95% confidence interval [95% CI], 1.11-1.31; P < 0.001) and graft failure (hazard ratio, 1.24; 95% CI, 1.15-1.33; P < 0.001) (Table 2). Patient and graft survivals with and without PVT at LT were divergent within the first 90 days after LT, (Figure 1) but were otherwise numerically and statistically similar in patients surviving longer than 180 days (1- and 5-year patient survivals 95.4% and 78.7% vs 96.2% and 79.8%, respectively (P = 0.1), and 1- and 5-year graft survivals 94.7% and 76.6% vs 95.5%, and 77.3%, respectively (P = 0.3)). On simple and multiple logistic regression analysis, PVT at LT was an independent predictor of 90-day mortality (odds ratio [OR], 1.7; 95% CI, 1.45-1.99; P < 0.001) and graft failure (OR, 1.72; 95% CI, 1.51-1.97; P < 0.001) (Table 3).
For descriptive purposes, outcomes were examined across MELD quartiles and according to intensive care status at LT. Patients with PVT and the highest quartile of MELD (MELD > 27) or receiving intensive care at LT had significantly higher 90-day mortality (range, 16.8%-21.4%) and graft failure (range, 19.1%-25.2%) rates compared to patients without PVT or in lower MELD quartiles or not requiring intensive care at LT (Table 4).
Analysis of the waitlist during the study period data identified 108 911 listings for LT in 95 122 adults, excluding listings for multiorgan transplant or resulting in multiorgan or live donor transplant. For patients with multiple listings, the final listing was included in the analysis. Patients with missing information on PVT at listing (9827) could not be categorized and were excluded. The waitlist removal codes amongst the 85 295 studied patients were, deceased donor LT in 47 711 (55.9%), death or too sick to transplant in 22 448 (26.3%), improved condition in 4221 (4.9%), unspecified in 6805 (8%), and other codes in the remaining 4110 (4.8%) patients (LT at other centers 2132 [2.5%], lost to follow-up 954 [1.1%], refused LT 657 [0.8%], and others). Portal vein thrombosis was reported as present at listing in 2819 (3.3%) patients, with 1660 (58.9%) undergoing deceased donor LT, as compared with 46 051 (55.8%) of 82 476 patients without PVT at listing (P = 0.001). The percentage of patients removed for death or being too ill to transplant was similar for those with and without PVT at listing 27.1% versus 26.3%, respectively (P = 0.3). Patients with and without PVT at listing were compared (Table S1, SDC, https://links.lww.com/TP/B162). The presence of PVT at listing was associated with similar clinical and demographic factors associated with PVT at LT, although MELD at listing was lower in patients without PVT.
Though most (60.4%) patients with PVT at listing who received LT also had PVT at LT, this was not uniform. Portal vein thrombosis status at LT was missing in 2.2% of patients with and 2.2% of patients without PVT at listing. Figure 2 describes patient grouping according to presence or absence of PVT at both listing and LT. Six hundred thirty-four of 1603 (39.6%) patients with PVT at listing did not have PVT reported at LT. Conversely, 2205 (4.9%) of 44 568 patients without PVT at listing were reported to have PVT at LT. The latter patients (newly reported PVT at LT) were compared with 42 363 patients without PVT at listing or LT (Table S2, SDC,https://links.lww.com/TP/B162). Patients with “new” PVT at LT more frequently had a diagnosis of fatty liver cryptogenic cirrhosis, ascites, diabetes, and Hispanic ethnicity. They also had slightly higher body mass index, older age, and frequency of male sex. Simple and multiple logistic regressions for the endpoint of PVT at LT in patients without PVT at listing are described in Table 5. Because waitlist time was different for patients with and without PVT at LT, waitlist time was analyzed as a covariate (in increments of 180-day intervals). Multiple factors were independently associated with PVT at LT; however, a diagnosis of fatty liver or cryptogenic cirrhosis was the most important categorical variable for risk of PVT at LT (OR, 1.7), followed by presence of ascites at listing (OR, 1.39).
Waitlist time, MELD at listing and LT, rate of MELD change, and 90-day patient and graft survival rates were compared in the 4 subsets of patients grouped according to the presence or absence of PVT at listing and LT (Table 6). Numerically speaking, patients without PVT at listing or transplant had the highest 90-day patient and graft survivals, and patients with PVT at listing and transplant had the lowest 90-day patient and graft survivals. Patients with PVT at only listing or transplant had numerically virtually identical survival rates which fell between survival rates of the former 2 groups. Patients with “new” PVT at transplant had significantly lower 90-day survival rates compared to those without PVT at listing or transplant, and a trend toward significantly better survival compared to patients with PVT at listing and transplant. Median MELD score at listing in this subgroup was also lowest of the subgroups; however, the rate of MELD change (adjusting for the longer median waitlist time) was highest. It was significantly higher when compared to patients with no PVT at listing or transplant, with a trend toward significance when compared to patients with PVT at listing and transplant.
Propensity testing was performed based on the predictors of PVT at LT (Table 5). The propensity score for PVT at LT in the study group was 0.075 (95% CI, 0.074-0.076). Transplant centers were not identified in this data set, therefore the effect of transplant volume or center “size” could not be determined and limited propensity testing in small versus large centers. When entering the propensity score for PVT in multivariate logistic regression modeling, in lieu of the representative variables, PVT remained an independent predictor of 90 day mortality (OR, 1.69; 95% CI, 1.22-2.34; P = 0.002) and graft failure (OR, 1.61; 95% CI, 1.21-2.15; P = 0.001).
The main finding of this analysis of the OPTN data was that PVT represented an independent risk for 90-day mortality and graft failure after LT. Although PVT was associated with overall mortality and graft failure risk on Cox regression analysis as previously described,13 the present study highlights the impact on early outcomes (90 days) as there was little to no risk in patients surviving longer than 180 days after LT or to long-term outcomes. This allowed for a more salient understanding of risk of PVT at LT. It also better accounted for competing risk factors in the analysis, for example, hepatitis C infection which was associated with increased overall risk on univariate and multivariate Cox regression analyses, but was associated with better 90-day outcomes on simple and multiple logistic regressions. Given the impact on early endpoints, PVT notably amplified risk in the most ill and debilitated LT candidates as demonstrated in patients in the highest MELD quartile group and those receiving intensive care at LT. Although there is no doubt of the survival benefit of LT in these higher-risk patients, these data underscore a transplant team's need for preparation for a challenging course while possibly counseling patients and families on risks and anticipated outcomes. Given that these patients are likely poor candidates for anticoagulation, if validated this data suggests that treatment of PVT needs to be considered at earlier time points if possible in LT candidates.
Notably, these findings are at odds with a number of single center studies that did not demonstrate universally worse outcomes for LT with PVT.10,12,14,15 However, these reports came from high volume centers, reporting on low- and high-grade PVT at LT in a large surgical experience. These studies reported higher PVT incidence at LT compared to OPTN data, in the range of 12.6% to 24%, suggesting an underreporting of PVT at LT to OPTN. A reporting bias of more severe grades of PVT to OPTN, which carry a higher risk for poor outcomes, could explain some of the discrepancies in described outcomes. In one of these studies, patients with nonphysiologic portal inflow after LT were noted to have worse outcomes, and even establishing physiologic portal inflow required varying surgical techniques including: thrombectomies most commonly, but also interpositional vein grafts, and mesoportal jump grafts.12 In another large-center study, portal vein conduits or thromboendovenectomy were associated with similar patient and graft survival compared to LT without PVT.15 Single-center data are also not uniform in their estimates of risk, with 1 large center demonstrating that complete PVT with extension into the splanchnic veins (ie, high-grade PVT) carried higher (33%) mortality rate.16 Arguably high volume centers may benefit by virtue of their large experience from added surgical expertise in managing PVT at the time of LT, and may be better equipped to manage complicated postoperative courses. Thus, a potential bias for reporting higher-grade PVT, hence higher risk, in the OPTN database, and potential differences in technical expertise at some centers, may explain these seemingly contradictory observations. Additional considerations, such as pre-LT treatment of PVT, patient selection for LT, and availability of mutlivisceral transplantation (at select centers) may differ among centers yet may significantly impact outcomes. The OPTN data only speak to aggregate LT outcomes and risk related to reported PVT, and transplant center volumes were not available for analysis in this data set.
We queried data from an ongoing study at our center on pre-LT PVT to discern any reporting bias. We reviewed the reported PVT status to the OPTN in 44 patients with known PVT undergoing LT between 2010 and 2014. In10 patients with only branch PVT, 7 (70%) were not reported, compared with 16 of 34 (47%) with clot involving the main portal vein (P = 0.2). On the other hand, 2 of 8 (25%) patients with superior mesenteric and/or splenic vein involvement (ie, more extensive PVT) compared with 21 of 36 (58%) of patients with only portal vein involvement were not reported to the OPTN (P = 0.08). Although this is a very limited preliminary data, from only 1 center, it supports the proposed bias for reporting more extensive PVT to the OPTN.
Although the reliability of PVT reporting at listing and LT in the OPTN database may be questioned, this analysis lends credibility to the reporting, given the distinct clinical phenotype of the resulting subgroups, particularly as noted in patients without PVT at listing but with PVT at LT (group C). This group with inferred new PVT between listing and LT had the longest median waitlist times, and higher rates of MELD score increase during this interval. On the other hand, PVT reported at listing was associated with higher listing MELD scores, but not with higher rates of MELD score increase at LT. Longer waitlist time was associated with PVT risk but was accounted for as a variable in the analysis. Together these data suggest that newly reported PVT at LT may be associated with liver disease progression and may warrant therapeutic considerations. Data were missing regarding the presence of PVT at waitlist removal in patients not undergoing LT. However, PVT at listing was not associated with increased waitlist mortality or lower rates of LT, as described previously.13
A number of risk factors for PVT at LT were identified in patients without PVT at listing. Among these fatty liver disease was strongly associated with a number of the other risk factors, including cryptogenic cirrhosis (thought to most likely represent undiagnosed fatty liver disease), diabetes mellitus, and obesity. A diagnosis of fatty liver or cryptogenic liver disease was associated with the highest OR of PVT (1.5), and lends support to the suggested association of fatty liver and obesity with increased thrombotic risk.17-19 These same risk factors were also associated with presence of PVT at listing, lending further credibility to their association with risk for PVT. Ascites was also associated with risk of “new” PVT, which concurs with the association of increased risk of PVT with lower portal flow in the setting of more severe portal hypertension.3
Finally, the data suggest variability in early post-LT outcomes, depending on the presence or absence of PVT at listing and LT. In this analysis, patients with newly reported PVT at LT (group C) had lower 90-day patient and graft survivals than those without reported PVT at listing or LT (group D), but a trend toward better survival compared to patients with PVT at listing and LT (group A). The reasons for this cannot be discerned from the data, but we can only speculate that acute PVT may be more amenable to portal vein clot removal intraoperatively compared to established long-standing clot with chronic vascular changes. Patients with PVT at listing but not at LT (group B) had numerically better 90-day survival compared to those with persistent PVT at LT (group A) though this did not reach statistical significance, albeit a comparison in smaller patient groups.
The strengths of the study include the large sample size of listed and transplanted patients in the MELD era, and the novelty of analysis examining predictors of PVT at listing and LT and the associated 90-day post-LT outcomes. However, a number of important limitations have to be considered. The study is retrospective with an inherent selection bias because all patients were listed for, or underwent LT, and the impact of PVT on LT candidacy could not be examined. The OPTN data also lack granularity and detail of single-center studies, and importantly does not have defined causes of graft failure and death uniformly recorded. It is also difficult to rule out reporting error related to the presence of, and critically, the extent of PVT. We suspect that more extensive PVT would have been more likely to be reported to the OPTN; therefore, these data may be skewed to reflect high-grade PVT, and has to be interpreted with that caveat. The study also suffers from lack of data on anticoagulation therapy for PVT before LT, and information on physiological or nonphysiological portal vein reconstruction at LT. Lastly, causes of waitlist mortality or removal for deterioration, and causes of post-LT death or graft failure, were not known.
In summary, PVT is a risk factor for early (90 days) mortality and graft loss, but not long-term outcomes in patients undergoing LT. Fatty or cryptogenic liver disease, obesity, diabetes mellitus, and ascites are risk factors for PVT developing on the waitlist. Large multicenter studies with detailed data on PVT, its extent, pre-LT treatment, and post-LT outcomes are needed to better define the associated risks and to preface investigations for the optimal therapeutic or preventative strategies for PVT in LT candidates.
The authors thank James Slaven, Department of Biostatistics, Indiana University School of Medicine, for his critical review of the statistical design, analysis and interpretation, and for performing the power analysis.
1. Ponziani FR, Zocco MA, Garcovich M, et al. What we should know about portal vein thrombosis in cirrhotic patients: a changing perspective. World J Gastroenterol
. 2012; 18 (36): 5014–5020.
2. Tripodi A, Anstee QM, Sogaard KK, et al. Hypercoagulability in cirrhosis: causes and consequences. J Thromb Haemost
. 2011; 9 (9): 1713–1723.
3. Zocco MA, Di Stasio E, De Cristofaro R, et al. Thrombotic risk factors in patients with liver cirrhosis: correlation with MELD scoring system and portal vein thrombosis development. J Hepatol
. 2009; 51 (4): 682–689.
4. Amitrano L, Guardascione MA, Brancaccio V, et al. Risk factors and clinical presentation of portal vein thrombosis in patients with liver cirrhosis. J Hepatol
. 2004; 40 (5): 736–741.
5. Werner KT, Sando S, Carey EJ, et al. Portal vein thrombosis in patients with end stage liver disease awaiting liver transplantation: outcome of anticoagulation. Dig Dis Sci
. 2013; 58 (6): 1776–1780.
6. Francoz C, Valla D, Durand F. Portal vein thrombosis, cirrhosis, and liver transplantation. J Hepatol
. 2012; 57 (1): 203–212.
7. Rodriguez-Castro KI, Porte RJ, Nadal E, et al. Management of nonneoplastic portal vein thrombosis in the setting of liver transplantation: a systematic review. Transplantation
. 2012; 94 (11): 1145–1153.
8. Yerdel MA, Gunson B, Mirza D, et al. Portal vein thrombosis in adults undergoing liver transplantation: risk factors, screening, management, and outcome. Transplantation
. 2000; 69 (9): 1873–1881.
9. Lendoire J, Raffin G, Cejas N, et al. Liver transplantation in adult patients with portal vein thrombosis: risk factors, management and outcome. HPB (Oxford)
. 2007; 9 (5): 352–356.
10. John BV, Konjeti R, Aggarwal A, et al. Impact of untreated portal vein thrombosis on pre and post liver transplant outcomes in cirrhosis. Ann Hepatol
. 2013; 12 (6): 952–958.
11. Levi Sandri GB, Lai Q, Berloco PB, et al. Portal vein thrombosis before liver transplant does not alter postoperative patient or graft survival. Exp Clin Transplant
. 2014; 12 (3): 238–240.
12. Hibi T, Nishida S, Levi DM, et al. When and why portal vein thrombosis matters in liver transplantation: a critical audit of 174 cases. Ann Surg
. 2014; 259 (4): 760–766.
13. Englesbe MJ, Schaubel DE, Cai S, et al. Portal vein thrombosis and liver transplant survival benefit. Liver Transpl
. 2010; 16 (8): 999–1005.
14. Molmenti EP, Roodhouse TW, Molmenti H, et al. Thrombendvenectomy for organized portal vein thrombosis at the time of liver transplantation. Ann Surg
. 2002; 235 (2): 292–296.
15. Nikitin D, Jennings LW, Khan T, et al. Twenty years’ follow-up of portal vein conduits in liver transplantation. Liver Transpl
. 2009; 15 (4): 400–406.
16. Manzanet G, Sanjuan F, Orbis P, et al. Liver transplantation in patients with portal vein thrombosis. Liver Transpl
. 2001; 7 (2): 125–131.
17. Northup PG, Argo CK, Shah N, et al. Hypercoagulation and thrombophilia in nonalcoholic fatty liver disease: mechanisms, human evidence, therapeutic implications, and preventive implications. Semin Liver Dis
. 2012; 32 (1): 39–48.
18. Assy N, Bekirov I, Mejritsky Y, et al. Association between thrombotic risk factors and extent of fibrosis in patients with non-alcoholic fatty liver diseases. World J Gastroenterol
. 2005; 11 (37): 5834–5839.
19. Ayala R, Grande S, Bustelos R, et al. Obesity is an independent risk factor for pre-transplant portal vein thrombosis in liver recipients. BMC Gastroenterol
. 2012; 12: 114.