Introduction

Portal vein thrombosis (PVT) can have serious consequences depending on the location and extent of the thrombus. Complications may include hepatic ischemia, intestinal ischemia,1 and pre-sinusoidal portal hypertension with its sequela.2–4 Anticoagulation therapy for such patients may limit propagation of the thrombus and recurrence but may be associated with major bleeding particularly in those patients with varices.5 The incidence and management of PVT following liver transplantation and pancreatoduodenectomy and in the setting of acute and chronic pancreatitis have been previously described and are well accepted.1–8 In contrast, there is a paucity of data in the English literature on PVT in patients undergoing distal pancreatectomy (DP). The operative indication for DP is often pancreatic cancer, a well-known risk factor for venous thromboembolism.9,10 To evaluate the incidence, management, and outcomes of PVT in patients undergoing DP, we reviewed our recent experience with DP.

Methods

Consecutive patients who underwent DP with or without splenectomy over a 10-year period (January 1, 2001 to December 31, 2011) were included. The diagnosis of PVT was confirmed by computed tomography (CT) or ultrasonography in all patients. Those patients who had a portal venorrhaphy, portal venous reconstruction, or pre-operative PVT were excluded from analysis. Patients undergoing DP for chronic pancreatitis with pre-operative PVT were also excluded. The study was approved by the Mayo Clinic Institutional Review Board.

Data were collected from a centralized system that contains complete records of all patients treated and followed at Mayo Medical Center. The follow-up period began at the time of the initial patient encounter and ended with the most recent medical evaluation or time of death. The use and duration of warfarin post-operatively was recorded for all patients. Imaging studies relevant to portal venous patency were included for all patients. Causes of death were determined by a review of medical records, death certificates, and autopsy results whenever available. Variables studied for effect on PVT after DP included age, sex, smoking and alcohol history, BMI, history of DM II, medications (especially estrogen and other pro-coagulants), prior history of malignancy or surgery, presence of patent IMV post-operatively, mode of surgery (laparoscopic versus open) extent of pancreatic resection, location of thrombus, and presence of thrombus at multiple locations.

Major Event Definition and Adjudication

A major thrombotic event was defined as venous thrombosis occurring within the mesenteric, extra- or intrahepatic portal venous circulation. A recurrent PVT (or venous thrombosis in other locations) was distinguished from the original thrombus by comparing serial imaging modalities. In order to be classified as a new PVT, there had to be new filling defects evident on a repeat study not appreciated on the prior images or an interval study showing complete resolution of the thrombus. Warfarin sodium was the long-term anticoagulation drug of choice in all patients undergoing treatment for PVT. The goal INR during treatment for PVT was 2–3. A major hemorrhage was defined as clinically evident bleeding and a decrease in hemoglobin of ≥2 g/dL; a hemorrhage requiring transfusion of 2 units of blood; or documented intraocular, intracerebral, or retroperitoneal hemorrhage.

Medians and ranges are used to express continuous data, unless specified otherwise. Logistic regression models were used for binary outcomes; p < 0.05 was considered statistically significant. Univariate models were fit to model the time to recurrence and the time to death in the PVT patients. Age-adjusted event-free survival curves were calculated using Cox proportional hazards models. Time-based logistic regression analysis was used to determine the effect of anticoagulation on PVT. Multivariate analysis could not be performed in this study population because of small sample size. As a result, it was not possible to assess if variables identified as predictors for persistence of PVT were indeed independent predictors. Statistical analyses were performed using the JMP statistical package (JMP software, Cary, NC, USA).

Results

Over the 10-year time period of this study, 991 patients underwent DP at our institution. Of these, 21 patients (2.1 %) were diagnosed with PVT. The median age at diagnosis was 56 years (range 20–88 years), and 14 (66 %) were females. Neoplastic indications for DP were neuroendocrine neoplasms of the pancreas (n = 266), adenocarcinoma (n = 142), involvement of the pancreas by direct invasion or metastatic disease (n = 117), and cystic neoplasms of the pancreas (n = 195), while non-neoplastic etiologies included noninsulinoma pancreatogenous hypoglycemia syndrome (NIPHS, n = 70). The incidence of PVT by primary pancreatic pathology was 7.7 % in pancreatic adenocarcinoma, 1.5 % in neuroendocrine neoplasms, 2.8 % in NIPHS, 1 % in cystic neoplasm of the pancreas, and 0.8 % in DP for direct invasion or metastasis for a non-pancreatic malignancy. The median time from DP to diagnosis of PVT was 16 days (range 5–85 days). Abdominal pain was the most common symptom prompting a diagnostic imaging study in 11 (52 %) patients. Other indications for abdominal imaging included clinical features concerning for a pancreatic stump leak or intraabdominal infection in five (24 %) or for follow-up surveillance in five (24 %) patients. In all but one patient, the diagnosis of PVT was made by contrast-enhanced CT. In one patient, Doppler ultrasonography was the mode of diagnosis and follow-up imaging of choice. Seventeen patients had a diagnosis of neoplasm on histopathology, which was of pancreatic origin in all but two patients. One patient had extension of a gastric cancer to the pancreas and the other had metastatic ovarian cancer. Five of the 84 (6 %) DPs performed laparoscopically developed PVT as compared to 16 of the 632 (2.5 %) DPs carried out as an open operation, and all but one patient underwent simultaneous splenectomy. The difference in frequency of PVT after DP in patients who underwent laparoscopic or open procedure was not statistically significant. In all except 2 procedures, transection of the pancreas was at the neck. In the 2 procedures, the pancreas was transected at the body–tail junction.

The PVT was located in the main portal vein in 15 (71 %), right portal vein branch in 8 (38 %), left portal vein branch in 3 (14 %), and superior mesenteric vein in 7 (33 %) patients. In eight patients (38 %), there were multiple segments of the portal vein (PV) involved. Nine of the 21 patients had occlusive PVT. In seven (33 %) patients, the PVT was complicated by fluid collection in the operative bed of the pancreas or a clinically confirmed pancreatic leak. Enteric leak was seen in three patients who had an en block resection of surrounding organs to ensure complete resection (Table 1). PVT was diagnosed a median of 16 days (range 5–85 days) after DP.

Table 1 Pathology and PVT characteristics
Full size table

Seventeen patients (81 %) received anticoagulation therapy with warfarin sodium at the time of diagnosis of PVT. Four patients refused treatment and in one patient, anticoagulation treatment was stopped within 1 week secondary to major upper gastrointestinal bleeding. The median duration of treatment was 6 months (range 7 days–36 months). One patient was placed on life-long anticoagulation treatment because of protein S deficiency; this patient had initial progression of PVT despite therapeutic anticoagulation.

Complete resolution of the PVT was seen in eight (38 %) patients at a median time from diagnosis of 87 days (range 7 days–9 months). The most common site of persistent PVT was the sectoral branches of the portal vein, as seen in six patients. The main portal vein remained occluded with collateralization in five, and the left portal vein in two patients. In one patient, complete occlusion of the left portal vein led to atrophy of the left lateral sector with hypertrophy of the remaining liver segments (Fig. 1). Patients who had sectoral branch PVT showed a perfusion anomaly of the corresponding sector on CT as shown in Fig. 2. The presence of perfusion anomaly on CT was not a predictor for persistent PVT. Perioperative factors for non-resolution of thrombus were anesthesia time greater than 180 min (p = 0.025), BMI >30 Kg/m2 (p = 0.025), and diabetes mellitus type II (p = 0.031). Characteristics of the portal vein clot associated with non-resolution of the PVT despite anticoagulation were thrombus in an intrahepatic segment of the portal vein (p = 0.003), involvement of multiple segments simultaneously (p = 0.002), complete occlusion of the main PV with collateralization (p < 0.001), and thrombus in a sectoral branch of the right portal vein (p = 0.019). The presence of a pancreato-enteric leak was a factor in non-resolution of thrombus (p = 0.05). Because of a limited number of events in this series, multivariate analysis was not performed because of the limited number of PVTs and clinical factors assessed which precluded meaningful statistical analysis.

Fig. 1
figure 1

CT scan demonstrating left main PVT (arrow) causing atrophy of the left lobe and compensatory hypertrophy of the right lobe of the liver

Full size image
Fig. 2
figure 2

CT scan demonstrating PVT of the right sectoral branch portal vein with perfusion anomaly of the corresponding sector

Full size image

Warfarin sodium was the long-term anticoagulation medication of choice. Anticoagulation did not affect outcomes in terms of resolution of the PVT (p = 0.929) over a median follow-up period of 21 months (range 2 months–10 years). There were no recurrences of PVT in any patients after resolution and discontinuation of anticoagulation for the duration of the follow-up.

Four patients (19 %) developed upper gastrointestinal hemorrhage while on anticoagulation. Therapeutic upper endoscopy was required in two patients. Anticoagulation was discontinued in one of these patients. There were no deaths related to anticoagulation treatment.

Discussion

The principal finding of this study is that PVT is an uncommon complication in patients undergoing DP for presumed pancreatic neoplasms. In this series of nearly 1,000 patients, the overall incidence of PVT was only 2.1 %. Although pancreatic cancer was the most common indication for DP, PVT occurs infrequently in this population. Cancer-related and especially pancreatic cancer-related operations have been deemed major risk factors for venous thromboembolism; nevertheless, PVT appears to be an infrequent complication. However, from these data, it appears that pancreatic adenocarcinoma confers a greater risk for PVT compared to other forms of pancreatic malignancy as previously mentioned.

Clinical presentation of PVT is variable and depends on the extent and location of PVT. In our series, non-specific abdominal pain was the most common condition that leads to a diagnostic workup. Most patients in this series received anticoagulation therapy for a duration of 6 months, and there were no recurrences or progression of PVT noted after completion of treatment. Complete resolution was seen in seven (38 %) patients. Therapeutic anticoagulation did not affect the frequency of PVT resolution in this series, and thrombus resolution, as defined by recanalization of the portal vein, has been similarly observed in only a third of the patients who have been anticoagulated for PVT from acute and/or chronic pancreatitis.2,8

The perioperative factors for persistence of PVT in this series were BMI, presence of type II DM, and anesthesia time >180 min. Crystalloid infusion, transfusion status, and blood loss had no significant influence on thrombus resolution. Characteristics of the clot that indicated a greater likelihood of non-resolution despite therapeutic anticoagulation included presence of a thrombus in an intrahepatic segment of the portal vein, presence of thrombus in multiple segments, complete occlusion of the portal vein with collateralization, and thrombus in a sectoral branch of the right portal vein. Presence of pro-inflammatory conditions such as a pancreatic-enteric leak as a result of en block resection of organs intimately involved with the pancreas during the operation was a factor in persistence of PVT.

Inflammatory conditions have been implicated as a risk factor for PVT and are commonly seen in conditions such as the active phases of inflammatory bowel disease11,12 and acute or chronic pancreatitis. Although the presence of a pro-inflammatory state, risk of recurrence, and progression of thrombosis leading to disastrous consequences such as bowel ischemia have been cited as the indication for prolonged anticoagulation, a recent publication from this institution8 and a prospective multicenter study2 showed a low risk of recurrence and thrombus extension. Only one patient in this series of patients undergoing DP had thrombus extension while on anticoagulation and that patient had an underlying coagulopathy (protein S deficiency). None of the patients in our series suffered major morbidity or mortality as a direct result of PVT, albeit four patients did develop GI bleeding. Although risk of progression is small but real and a surgical procedure is considered as a pro-inflammatory condition, anticoagulation may be considered in these patients short-term. Because the median time to resolution of thrombus was 87 days, one may consider anticoagulation for a minimum duration of 3 months as a reasonable duration for these patients, beyond which, resolution of a thrombus is unlikely, especially in the presence of factors such as portal vein thrombus located in the right sectoral branch or other characteristics of PVT mentioned above.

Anticoagulation for PVT in the setting of pancreatitis has been associated with substantial morbidity of GI bleeding as a result of the anticoagulation and has a reported incidence of 10–26 %.2,8 The cumulative incidence of clinically significant GI hemorrhage in this series was 19 % (4 out of 22), although the rate of GI bleeding requiring active intervention was only 9 %. The cause of GI bleeding in this and most situations is a supra-therapeutic level of anticoagulation that results from either poor oral intake or variable response to warfarin during initial anticoagulation. There are new oral medications such as dabigatran (Pradaxa® Boehringer Ingelheim pharmaceuticals, Germany) that could potentially help alleviate that problem, especially with a fixed daily dose regimen, absence of constant monitoring, and a predictable response in patients with an adequate renal function.13 The safety and efficacy of the newer agents in the setting of PVT remains to be proven in clinical trials.

There are limitations to this study that warrant mention. First, patient data were collected retrospectively. These patients had clinical indications for obtaining an imaging study which identified the PVT. As a result, the true incidence of PVT after DP is likely to be underrepresented because some patients may have developed asymptomatic PVT that went undiagnosed. Second, follow-up imaging and clinical assessment was not uniform in terms of time, duration, and frequency between image assessments of the portal vein because follow-up was either at the discretion of the provider or per specific post-operative follow-up protocols. Statements regarding timing of PVT resolution are dependent on the imaging obtained. Imaging studies obtained at shorter time intervals may have permitted identification of variables associated with PVT resolution not defined by this study. Third, anticoagulation treatment and duration were not randomly assigned. Small patient numbers preclude conclusions regarding either the safety or efficacy of anticoagulation therapy in these patients long-term. Lastly, given the limited number of events, a multivariate analysis could not be carried out to demonstrate meaningful evidence of factors for clot resolution among variables mentioned above.

Conclusion

In conclusion, PVT is unusual after DP. In the absence of inherited pro-coagulant conditions such as protein S deficiency, anticoagulation does not correlate with the rate of PVT resolution. Current literature on PVT does not support prolonged anticoagulation because of the low rate of recurrence and higher rate of morbidity, mainly GI bleeding. Similarly, the rate of thrombus progression leading to disastrous consequences is low. In the absence of randomized clinical trials, because most DP are performed for a presumed malignant etiology, and given that the operative procedure itself is a pro-coagulant condition, in the absence of thrombus propagation or pro-coagulant condition (e.g., Factor V Laden, Protein C/S deficiency), decision for anticoagulation should be made on a case-to-case basis factoring in the extent of PVT and clinical circumstances leading to the diagnosis of PVT. One strategy would be to provide short-term anticoagulation to patients with PVT followed by repeat imaging study to assess the response to anticoagulation which may help guide the duration of further anticoagulation. Venous thrombosis prophylaxis should be provided in keeping with current management guidelines.