Cerebral ischemic events in patients with pancreatic cancer

Supplemental Digital Content is available in the text


Introduction
Thromboembolic events-a known complication of cancer-are critical for vulnerable patients and costly for the society. [1] Trousseau was the first to describe venous thrombosis during his own pancreatic cancer. His name was retained to define any thromboembolic event as the first manifestation of concealed cancer. [2] Since then, most reported cases and studies have described venous thrombosis in a population of patients with various oncological diseases. [3] In this context, pancreatic cancer has emerged as a leading cause with venous thrombosis observed in about 35% of affected patients in Western countries, although a lower incidence of 5% is observed in Asian populations. [4][5][6] In contrast, reports of arterial infarctions, especially at the cerebral level, are scarce. Paraneoplastic stroke has been reported to concern 0.3% of patients and still predominantly associated with pancreatic cancer. [7][8][9] Literature more often describes cerebral infarction as Trousseau syndrome revealing concealed pancreatic cancer. It is a result of a hypercoagulable state due to both the cancer and therapies, but the pathogeny is poorly understood, rendering it difficult to establish preventive measures. [10,11] A better knowledge of cerebrovascular diseases in pancreatic cancer would help understand the pathogeny, detect patients at risk of stroke, and establish preventive measures. We decided to describe consecutive cases of patients with cerebrovascular events associated with pancreatic cancer admitted in 4 stroke units during a specified period of time. We also reviewed reported cases in literature to isolate specific clinical, radiological, and biological features of stroke patients in this particular context. We discuss thrombotic mechanisms interacting with cancer cells and promoted by some chemotherapy regimens.

Methods
This was an observational retrospective multicenter study.

Statistical analysis
All analyses were descriptive and carried out by using IBM SPSS Statistics software, version 20 (SPSS Inc., Chicago, IL). The study was observational and described clinical, radiological, and biological features of patients from our centers and in literature. Continuous variables are reported as median ± standard deviation and extreme values, and categorical factors as frequency and percentage. Statistical comparison was performed between characteristics of patients from our centers and from literature using the Mann-Whitney U test for continuous variables and Fischer exact test for categorical factors. A significant difference between both groups was defined as P < 0.05.

Results
During the study period, 17 patients with pancreatic cancer were admitted in the 4 centers after a cerebral event. Diagnosis of pancreatic cancer resulted from this hospitalization in 7 cases, and had already been known for a mean time of 5.4 months for the 10 others. Eighteen cases were found in the literature, all describing diagnosis of pancreatic cancer through management of the stroke (P = 0.001). [10,[15][16][17][18][19][20][21][22] Overall, 57% of patients were male, and median age was 63 ± 14 years, which ranged from 23 to 81 years. The patients in our series were older than in the literature, with median age of 70 versus 59 years (P = 0.02) ( Table 1). In 93% of cases, pancreatic cancer was at a metastatic stage. All documented histologic types were adenocarcinoma except for 2: one patient had mucoepidermoid carcinoma and the other cystadenocarcinoma. Pancreatic location was fairly evenly distributed between the head (40%), and the body and tail (60%). All patients presented ischemic strokes, except for 2, who had transient ischemic attacks. No case of hemorrhagic stroke was found. Eighty-two per cent of our patients presented at least 1 vascular risk factor (apart from age) and 12% atrial fibrillation. Echocardiography performed in 4 cases revealed nonbacterial thrombotic endocarditis (NBTE) in 1 patient of our series, whereas it was found in 40% of documented cases of the literature. The fourth radiological pattern was the most common overall (54%) and 64% of patients presented disseminated cerebral infarctions (patterns 3 and 4). Silent other systemic infarctions were detected in 37% of the patients overall, with a predominance in patients from the literature (62 vs 25%; P = 0.01). Platelet counts were wide-ranging (180 ± 155 G/L), with a high median prothrombin time (19 ± 6 seconds, reference range: 12-16.5 seconds) and median D-dimer level (7600± 5 Â 10 7 mg/L) ( Table 2). Median activated partial thromboblastin time and fibrinogen level were normal at 29 ± 8 seconds (reference range: 23-34 seconds) and 2.3 ± 1 g/L (reference range: 1.7-4 g/L), respectively. Two patients met the criteria of DIC. No leukemoid reaction was found: the median white blood cell count was 9 ± 6 G/L, ranging from 7 to 26 G/L. Median CRP and CA19.9 levels were elevated at 63 ± 43 mg/L and 47,000 ± 70,000 U/mL, respectively. Forty-three per cent of all the patients presented elevated liver transaminases. The 7 patients diagnosed with pancreatic cancer at stroke onset in our centers received chemotherapy as opposed to none in the literature (P = 0.02). Two of the 7 were treated with folfirinox (including 5-fluorouracil, oxaliplatin, and irinotecan) and 5 with gemcitabine. In our population, 10 patients were treated with anticoagulant therapy for secondary stroke prevention (3 with unfractionated heparin, 6 with low-molecular-weight heparin [LMWH], and 1 with a vitamin K antagonist). No cerebral ischemic recurrence was documented. All patients died after a median time of 28± 14 days after stroke onset, except for 2 with a mean follow-up of 4.5 months.

Discussion
The incidence of arterial stroke complicating pancreatic cancer seems low. We identified only 17 cases in 4 centers specialized in the management of cerebrovascular diseases in a period of about 6 years. Whereas the retrospective design of this study and inclusion of patients only admitted in stroke units would certainly tend to underestimate the real incidence, an autopsy study found less than 1% of patients with pancreatic cancer presenting stroke. [23] Prospective works have demonstrated that stroke is more frequent in patients with pancreatic cancer with a cumulative incidence of 3.4% at 3 months after diagnosis, [9] and at higher risk during the first 6 months. [8] All reported cases found in the literature describe stroke as "Trousseau syndrome," which means that it precedes the diagnosis of the concealed pancreatic cancer. This diagnostic sequence has been the main focus of authors attempting to isolate special features of stroke to achieve earlier diagnosis and treatment of pancreatic cancer. Our series including consecutive cases shows that Trousseau syndrome led to cancer diagnosis in only one-third of patients. Stroke more often complicates an already known pancreatic cancer, in accordance to the results of prospective studies. [9] Unfortunately, whatever the time of stroke onset, it almost always occurs at a metastatic stage. All the patients included in this analysis presented ischemic events. Most presented infarctions in several arterial territories, many combining small and large lesions (Fig. 1). [12] This particular radiological feature is common to adenocarcinomas and is highly suggestive of a thromboembolic mechanism. [15,24] Many high-intensity transient signals were recorded on transcranial Doppler of patients with cancer-associated stroke, and these would appear to be correlated to D-dimer levels. [25] Several theories have been put forward to explain embolism in this setting. [26] Our study found a high median prothrombin time, which would suggest a coagulation disorder. Literature cases also report a higher activated partial thromboplastin time in cancerassociated stroke patients. [27] Activation of procoagulant factors (factor X, tissue factor, thrombin, and fibrin) and a decrease of anticoagulant factors (protein C and antithrombin) appear to be involved, with the latter more involved in venous thromboembolism (Fig. 2). [28] A release of inflammatory cytokines (tumor necrosis factor, interleukin [IL]-1 and IL-6) due to interaction of monocytes with malignant tumor cells could result in damage to the arterial vessel walls, activation of platelet aggregation, and high levels of fibrinogen. [1,19,29] The latter has been described as a biomarker of poor outcome. [30] Cancer is also responsible for the production of mucin, which can activate coagulation and has been found within the cerebral infarction. [31] Coagulation disturbance may be increased by activation of oncogenes (RAS, MET, EGFR) and inhibition of tumorsuppressor genes (p53, PTEN). [32] Mutations of these genes lead to the production of procoagulant agents and especially tissue factor, with activation of the extrinsic coagulation pathway, and plasminogen-activator inhibitor I. [32,33] Platelet aggregation is a result of the production of cyclooxygenase 2 and then thromboxane A 2 . [34] The resulting hypercoagulable state not only causes thromboembolic events but also enables cancer growth and development. Thrombin and factor VII activate, respectively, the protease-activated receptors 1 and 2, promoting cancer cell proliferation, adhesion, motility, survival, and angiogenesis by induced synthesis of vascular endothelial growth Table 1 Demographic, clinical, radiological characteristics and outcome of patients with cerebral ischemic events and pancreatic cancer.

Bonnerot et al (n = 17)
Literature [10,[15][16][17][18][19][20][21][22]  Continuous variables are described as median ± standard deviation (range), categorical factors as frequency (percentage). Radiological patterns: 1 = single infarction, 2 = multiple infarction in a single arterial territory, 3 = small disseminated infarctions, 4 = small and large disseminated infarctions. [12] n = number of patients, n * = number of documented cases, NBTE = nonbacterial thrombotic endocarditis diagnosed by transthoracic/transesophageal echography or autopsy. factor. Fibrin production protects cancer cells from natural killer lymphocytes, facilitates migration through the vessel wall, and consequently supports tumor development. [33,34] In addition, some studies have demonstrated that chemotherapy regimens are responsible for the occurrence of thrombotic events, including ischemic strokes. [3,35] Li et al [11] found an incidence of postchemotherapy ischemic strokes of 0.137% in 10,963 patients. Platinum-based chemotherapies, and especially cisplatin, are the most often incriminated, with an incidence of about 18% reported by Moore et al [36] and Numico et al. [37] The hypercoagulable state theory, with a decrease in protein C and increase in the von-Willebrand and tissue factor levels, is further supported. Other mechanisms have been put forward, including endothelium damage and renal dysfunction, causing hypomagnesemia and vasospasm. [38] However, our patients were treated with oxaliplatin, which causes fewer thromboembolic events than cisplatin. [36,[39][40][41] Gemcitabine is more likely to result in systemic infarctions from thrombotic microangiopathy syndromes, associated with other signs like hemolytic anemia, thrombocytopenia, and renal dysfunction. [42][43][44] Several cases of transitory thrombocytosis have been reported, but with rare thrombotic consequences. [45] Ischemia induced by 5-fluorouracil is mainly described for the myocardium, but with the same mechanisms as for cisplatin (i.e., hypercoagulable state, vasospasm, and endothelium damage). [46] Khorana et al developed a model to predict venous thromboembolism risk in patients treated with chemotherapy. Pancreatic location of the cancer is considered as a very high-risk criterion. Other risk factors are high platelet and leukocyte counts, low hemoglobin level, and elevated body mass index. [47] We did not find any leukemoid reactions. Only 2 reported cases of pancreatic cancer patients describe this biological feature in the literature, but without thrombotic consequences. [14,48] However, Khorana et al [47] consider a leukocyte count of more than 11 G/L as a risk factor of venous thrombosis in the context of cancer and chemotherapy regimen. Interaction between leukocytes and platelets with stimulation on one other and forming leucocyteplatelet complexes has already been suggested to play a role. [49] Platelet count could be high due to an inflammatory pathway or Table 2 Biological characteristics of patients with cerebral ischemic events and pancreatic cancer.

Bonnerot et al (n = 17)
Literature [10,[15][16][17][18][19][20][21][22] (n = 18) Overall (n = 35) n * n * P n * Hemoglobin 16 11 ± 2 g/dL (7-13) 2 12 ± 3 g/dL (10)(11)(12)(13)(14) 0.57 18 11 ± 2 g/dL (7-14) White blood cells 17 9 ± 6 G/L (7-26) 1 10 G/L -18 9 ± 6 G/L (7-26  low due to coagulopathy. However, only 2 patients fulfilled the DIC criteria. [19] Levels of D-dimer, a degradation product of fibrin and reflecting hypercoagulability, were found to be considerably elevated in our study. Radiological stroke patterns 3 or 4 with a D-dimer level over 1.11 mg/mL are thought to be associated with an oncological cause. [12] CRP, a well-known biomarker of inflammation, was also consistently found to be elevated in our series. High liver transaminase levels were present in less than half of the cases, and often related to hepatic metastasis. Nonbacterial thrombotic endocarditis was diagnosed in one of our patients and in 4 cases in the literature. [10,17,19,20] This common cause of embolism can be promoted by a prothrombotic state with deposition of fibrin and platelets as sterile vegetation on cardiac valves. However, transesophageal echocardiography was performed in only 2 of our patients (Fig. 3) (see Video, Supplemental Digital Content 1, http://links.lww.com/MD/B70, which shows NBTE on mitral valve during transesophageal echocardiography). We suggest that patients should be system-atically screened for NBTE in this context to confirm indication of anticoagulation therapy. [50] Other systemic infarctions were observed in only one-third of the patients, which raises the question of a specific cerebral mechanism of thrombosis (Fig. 4). Occurrence of cerebral vasculitis could promote cerebral infarction, but diagnosis requires more precise methods of angiography than 3-dimensional (3D) time-of-flight magnetic resonance imaging, which is usually performed.
On the other hand, median age of the patients we described was high, a single arterial infarction (radiological pattern 1) was found in 29% of cases, atrial fibrillation in 2 patients, and only 21% had no vascular risk factor. This raises the issue of whether the stroke was related to the cancer. Bang et al considers these as patients with "conventional stroke mechanisms," with a pathogeny and distribution of etiologies indistinguishable from patients without cancer. They thus advocate preventive treatment "according to the stroke subtype" in these patients. [12] Little is known about the best therapeutic strategy to prevent ischemic cerebral recurrences due to pancreatic cancer. Embolic  www.md-journal.com mechanisms, coagulation disorders, and possible NBTE spontaneously appear as strong arguments to use anticoagulation therapy, but no reported long series of patients with pancreatic cancer supports this theory. [50] Most of the studies about preventive treatment refer more to venous thrombosis recurrence than arterial infarction. [51,52] The use of LMWH emerges as the most effective and safest therapeutic attitude to decrease stroke recurrence. [53] Anticoagulation prevents thromboembolic events related to cancer and chemotherapy regimens, but unfortunately no gain on cancer development has been demonstrated. [54] We confirm poor outcome with rapid decline in nearly all patients after stroke onset due to progression of cancer. [55] Finally, the most relevant criteria that emerge to characterize stroke due to pancreatic cancer are as follows: a radiological pattern 4, a high D-dimer and CRP level, and an elevated prothrombin time. These investigations are usually performed in stroke patients, whatever the suspected etiology, and are thus easy to check. However, these features are not specific to the pancreatic location, as they are described in stroke patients with various oncologic diseases. [56] The main limitation of this study is the retrospective design. Even if collecting data from a series of consecutive patients is more valid than isolated reported cases, we lack systematic assessment of the patients during the stroke management. For instance, we believe that the actual rate of NBTE and number of patients with cerebral vasculitis due to inflammatory process was higher. In addition, cerebral ischemic recurrences could be greatly underestimated due to the difficulty of conducting follow-up in this context. Finally, faced with the scarcity of the condition, multicenter studies are required with carefully standardized clinical, radiological, and biological assessment to better understand the pathogeny of cerebral infarctions related to pancreatic cancer, and to identify patients at high risk of this complication. Screening for specific biomarkers could be an interesting approach.
In conclusion, cerebral ischemic events are a complication occurring more often at an advanced stage of pancreatic cancer already diagnosed at stroke onset. A radiological pattern with disseminated infarctions of different size, a high D-dimer and CRP level, and an elevated prothrombin time are strong indicators of cancer-related stroke, but not specific to the pancreatic location. In this context, the patient should be systematically explored for NBTE. Anticoagulation therapy, especially LMWH, and etiological treatments remain the best strategy to prevent cerebral ischemic recurrences. Further studies are needed to determine which of these vulnerable patients are at risk to present ischemic stroke, and the best therapy to prevent this dramatic complication.