Malignancy, whether in the form of a solid tumor or hematologic neoplasm, constitutes a significant risk for hypercoagulability. Venous and arterial thromboembolic diseases are important causes of death and morbidity in individuals with cancer.^1,2,3,4 It is estimated that venous clots complicate the clinical course of about 20% of all patients diagnosed with or treated for a malignancy,^5,6 and even that figure may underestimate the problem, given autopsy findings of thrombosis in as many as 50% of patients with cancer.⁷

Not only does known malignancy pose an increased risk of later venous thromboembolism (VTE), but there are also substantial data supporting thromboembolism as an epiphenomenon of cancer, occasionally occurring months to years ahead of the other clinical manifestations of malignancy.⁷ The questions of how clinicians should respond to such a warning is, in itself, the subject of considerable research.

Although the relationship of hypercoagulability and neoplasm has been known for centuries,⁸ our understanding of the pathophysiologic mechanisms underlying this relationship continues to evolve. The linkage is multifactorial; with evidence supporting multiple mechanisms, including active clotting factors released by the neoplastic cells and by the healthy host cells in response to the malignancy; effects of the cytotoxic therapies instituted; or by the therapeutic interventions of cancer treatment, including immobilization, infection, frequent hospitalization, and surgery.

This chapter aims to provide readers with an introduction to some of the key points in the relationship between malignancy and thromboembolism. It begins with an overview of the historical context of this association and then outlines the accumulation of data supporting the broad clinical impact and incidence of hypercoagulability in this special population, an endeavor that requires estimating the incidence of VTE in individuals with cancer as well as the incidence of otherwise occult cancer in individuals with embolism. The next section attempts to clarify what is currently known about pathophysiology of hypercoagulability in cancer, including the risk conferred by certain tumor cell characteristics and the growing understanding of treatment-related risk. It summarizes the current recommendations on the prophylaxis of patients to prevent clots and the data underpinning the current standards for treatment in these patients. Finally, there will be a short review of the evidence that suggests a survival benefit of anticoagulation in patients with cancer, independent of a decrease in clot risk.

The 19th century French internist, epidemiologist, and lecturer Armand Trousseau is widely credited with the clinical observation that hypercoaguability is often associated with both solid tumors and malignancies of the hematopoetic systems. Trousseau’s conclusions were likely informed by the 1823 writings of Jean-Baptiste Bouillaud, a professor of the Hopital de la Charite, who years before had described three cancer patients with “caillot fibrineau,” or fibrin clots induced by the cancer process. Indeed, the relationship may well have been known long before. One expert in the field has even tracked a reference to the relationship to the writings of an Indian surgeon in the year 1000 BC.⁹

No matter who is acknowledged as the original observer, it is certainly Trousseau’s work that brought the syndrome of hypercoagulability and malignancy into the medical parlance. In 1865, in a series of lectures at the Hotel-Dieu in Paris, Dr. Trousseau articulated his experience with phlegmasia alba dolens (or painful white clot) and cachexia.⁸

Among the key points that Dr. Trousseau made is that phlegmasia is often diagnosed in patients without other evidence of cancer. But he concluded:

In an example of the blind irony of medical history, about 18 months after giving his famous lecture, Trousseau self-diagnosed his own terminal gastric cancer on the basis of left lower extremity edema. As recounted by a colleague: “It was on January 1, 1867 when I went to give him my best wishes for the new year that Trousseau told me with resigned sadness—’I am lost; a phlegmasia which showed itself that night leaves me no doubt about the nature of my affliction.'”¹⁰

The eponymous syndrome is variably construed in the medical literature. The strictest definition of “Trousseau’s syndrome” is the condition of a migratory thrombophlebitis that precedes a subsequent diagnosis of cancer.¹¹ This is, however, only one of multiple manifestations of hypercoagulability in malignancy. In a seminal review published in the late 1970s, Sack et al,¹² using 10 case studies as a basis for their arguments, outlined the variety of presentations that can arise within the hypercoagulable state of malignancy. These investigators emphasize that “a wide range of clinical abnormalities may develop as a consequence of changes in blood coagulation and platelets. This is particularly apparent in the group of patients whose course was sufficiently chronic to allow the expression of various possible manifestations.”

In their review, Sack et al¹² outline the variety of pathophysiologic mechanisms that lead to disseminated intravascular coagulation (DIC) in malignancy, touching on research that identified the glycoproteins that are aberrantly secreted in mucin-producing tumors, tumor cells themselves, and “humeral agents,” which had been implicated. Their work laid the foundation for our current understanding of multiple overlapping mechanisms that evolve early in the neoplastic state and that are vulnerable to the myriad of interactions between host, disease, and treatment. The acute perceptions of Trousseau and the other researchers who have continued his investigations of the relationships between cancer and clot continue to provoke the many questions that remain at play in the clinic currently.

Understanding the true incidence of hypercoagulability in cancer is intimately related to the definitions one chooses of disease manifestation. The classic Trousseau’s syndrome, a migratory superficial thrombophlebitis, is rare and characterized by relapsing migratory superficial venous clots. Much more common manifestations of the hypercoagulability of cancer include deep venous thrombosis (DVT), nonbacterial endocarditis, DIC, thrombotic microangiopathy, and arterial thrombosis. Most of the data on the incidence of hypercoagulability is focussed on VTE, although small case series do look at the less common manifestations.

Incidence estimations remain difficult not only because of the variety of definitions of what constitutes a hypercoaguable event but also because of the great heterogeneity of conditions under the rubric of malignancy. It is well documented that the incidence of clot differs significantly among tumor types, and there is also great variability within tumor types based on staging; the incidence is reproducibly higher in individuals with advanced disease than in those with local disease even if the primary tumor type is the same. Finally, given the interplay of treatments—chemotherapy, hormonal therapy, and targeted anticancer agents and hypercoagulability— the incidence also varies between untreated versus treated patients within a tumor type. These considerations have led to much debate about the best way to calculate incidence of true hypercoagulability in the setting of cancer.

Incidence of Venous Thromboembolism in Patients with Known Cancer

Data for estimating the risk or incidence of clot in patients with cancer are typically gathered in one of three investigative settings—(1) the single-institution study in which a series of patients with a single tumor type or some other unifying characteristic are counted and then the incidence of VTE is calculated; (2) the hospital-based administrative dataset in which incidence is drawn from diagnostic codes; and (3) from autopsy data. This third method typically demonstrates a higher incidence, as expected, because reports are not limited to clinically diagnosed VTE and some clots found on autopsy are epiphenomenon of the cause of death, rather predating the death. Autopsy studies, which theoretically capture both symptomatic and asymptomatic clotting events, report among the highest incidences.^13,14

Overview analyses have estimated the annual incidence of symptomatic venous embolic event, usually a lower-extremity DVT or a pulmonary embolism (PE), at one in 200 in patients with cancer.^15,16,17,18 Clinical data collected in Minnesota since 1966 estimate the incidence of a first episode of DVT or PE in the general population to be about 117 in 100,000.¹⁸ In a population-based case-control study published in 2000, Heit et al¹⁹ analyzed 625 Minnesota patients with a first VTE. A known malignancy was associated with a fourfold increase in the risk of VTE; subjects with cancer and getting treatment had a 6.5-fold increase in their risk compared with baseline.¹⁹ When looking at the population of “cancer patients,” in general, the rates of VTE vary significantly depending on the tumor type, stage, and treatment modality.^{12,20,21,22,23} In a study of hospitalized patients using a database of discharge codes for reporting outcome, researchers looked at more than 40 million hospitalized cancer patients for the 20 years of 1979 to 1999. Over that 20-year period, they found that 2% also carried a diagnosis of VTE, which was twice the rate of hospitalized patients without a diagnosis of malignancy. These data also provide an overview of the types of cancer that are most likely to be associated with VTE, as well as the risk gained with older age, as is illustrated in Figure 5-1.²⁴

These data also illustrate the interesting finding that the burden of VTE in cancer seems to be increasing. According to the data published by Stein et al²⁴ in their review of hospitalized cases, the rate of VTE and hospitalization began to increase in the late 1980s, increasing from about 2% to about 4% currently.

Not only is the risk of first VTE higher in cancer patients, but the marginal risk for a second clot is also much higher, with estimates of recurrent VTE in patients with cancer at about 30.0 events per 100 patient years compared with 12.8 events per 100 patient years in those without cancer.²⁵

When looking at the rates of VTE in patients with various cancers, cohort studies from single institutions typically report the rate to be about 4% to 10%, with higher incidence occurring when additional risk factors, such as surgery, inherited thrombophilia, or immobilization accrue.^26,27 Most studies demonstrate an increase in rates for patients with metastatic disease and among patients receiving chemotherapy.²⁸ Given the relatively low incidence of various types of cancer, the largest studies are the most useful for determining which tumor type has a higher risk of embolic phenomenon. One report, conducted by researchers in the Netherlands, linked patient records from the Cancer Registry in the midwestern part of the Netherlands with the database of two anticoagulation clinics to find cases of symptomatic DVT.²⁸ These investigators looked at more than 66,000 patients with cancer and found the cumulative incidence of venous thrombosis in the first 6 months after diagnosis to be 12.3 in 1000 patients with cancer. (95% confidence interval [CI], 11.5–13.0). Patients with tumors of the bone, ovary, brain, and pancreas had the highest cumulative incidence of DVT. This compares with an incidence of DVT in the noncancer population of the Netherlands of two per 1000 per year.

In a U.S. study looking at records for more than 1 million hospitalized cancer patients, the subgroups of cancer patients with the highest rates included those with black ethnicity (5.1% per hospitalization) and those receiving chemotherapy (4.9%). Sites of cancer with the highest rates of VTE included the pancreas (8.1%), kidney (5.6%), ovary (5.6%), lung (5.1%), and stomach. Among the hematologic malignancies, patients with myeloma had the highest rate of VTE (5%).²⁹ These data also underscore that the extent of tumor and treatment seems to matter when considering risk. In the Danish study, patients with distant metastases and those undergone chemotherapy had a twofold increased risk compared with those without metastases or not using chemotherapy.

In the U.S. population, similar findings have been published. In 1999, Levitan et al³⁰ published a large-scale study of the Medicare database and found a statistically significant difference in patients hospitalized for a DVT or PE and the coexistence of a diagnosis of malignancy. The Levitan study³⁰ found rates of initial VTE highest in patients with pancreatic, brain, and ovarian cancer. In a more recent study, Chew et al²⁶ used the California Cancer Registry and linked it to the California Patient Discharge Data Set to assemble a large cohort of cancer cases and determined, with information based on site, histologic type, and stage, the incidence and time course of the development of acute VTE. They found, in a 3-year study period, a total of 235,149 eligible cancer cases. Their findings included a strong association between metastatic-stage cancer at the time of diagnosis and the incidence of thromboembolism. In multivariate models, there was a strong and consistent relationship between metastatic disease at the time of diagnosis and development of thromboembolism for most of the cancer types analyzed. The highest 2-year cumulative incidence was in metastatic pancreatic cancer, with an incidence of 5.4%, but the real difference, researchers found was not in the type of cancer, but rather in the extent of cancer. “Compared with patients with localized disease, the relative risk of developing symptomatic thromboembolism was more than 20-fold higher for metastatic melanoma, ninefold higher for metastatic bladder cancer, and five- to sixfold higher among patients with metastatic breast or uterine cancer,”²⁶ they wrote in their conclusion.

As mentioned above, autopsy studies typically lead to even higher rates of thrombosis for various tumor types. A 2006 study of nearly 24,000 Swedish autopsies—covering about 84% of in-hospital deaths in that country between 1970 and 1982—investigated the relationship between adenocarcinoma and PE. They calculated an overall prevalence of 23% for PE among patients with adenocarcinoma.³¹ A different Swedish study, looking at all cancers, showed the highest prevalence of PE to be in patients with ovarian carcinoma, cancer of the extrahepatic bile duct, or stomach cancer.³²

Patients with hematologic malignancies also carry an increased risk for VTE. The classic association is with acute promyelocytic leukemia (APL), in which thrombosis can be a presenting symptom in about 9.6%.³³ But the incidence rates range from 3.87% to 5.79% in patients with non-M3 acute myeloid leukemia (AML) or acute lymphoblastic leukemia (ALL)³⁴ and is higher still when individuals are treated with chemotherapy, including asparaginase. The rates are also elevated in patients with non-Hodgkin’s lymphoma, in whom the relative risk over patients without cancer has been estimated at 1.8. Studies looking primary central nervous system (CNS) lymphoma have quoted rates of VTE as high as 18% to 60%.³⁵ There is also increased risk in patients with multiple myeloma. Finally, it has long been known that in myeloproliferative diseases, especially essential thrombocytosis and polycythemia vera, both venous and arterial clotting can be a manifestation of uncontrolled disease, and this bears out in the larger population studies.²⁴

Many of the studies above limit their investigation to the classical presentation of DVT or PE. But there are also data on the relationships between certain cancers and the less common phenomena that are included under the heading of hypercoagulability. Migratory superficial thrombophlebitis is often associated with adenocarcinoma, and in one study, pancreatic, lung, prostate, stomach cancer, and acute leukemia were all seen in patients presenting with the diagnosis.¹² A study looking at 51 patients with cerebral ischemia and active cancer documented nonbacterial endocarditis (the formation of aseptic vegetations on the cardiac valves) in 18% of these subjects. Patients with this finding tended to be younger and had a greater frequency of hematologic malignancies than patients with cerebral ischemia and no endocarditis.³⁶

In summary, about 5% to 8% of cancer patients develop VTE, most typically DVT or PE. Rates are lower in those with limited disease but higher in patients receiving chemotherapy and advanced disease. It is important to recognize that cancer risk is not uniform across subgroups and that rates are higher in patients with certain tumor subtypes or getting certain chemotherapies.

Just as there has been considerable attention paid to the increased incidence of VTE in patients with malignancy, there has also been significant research into the increased incidence of cancer in individuals diagnosed with idiopathic VTE. The classic Trousseau’s syndrome of migratory thrombophlebitis rarely occurs outside of the setting of a coincident cancer. But for the more common syndromes of DVT or PE, epidemiologic work for the past 3 decades has been conducted on how to determine which individuals with idiopathic thromboembolic disease have or will develop cancer.^{25,37,38,39,40,41,42,43,44,45,46,47,48} The most recent and most inclusive meta-analysis estimates that there is a threefold excess risk of occult cancer in patients with VTE.⁴⁹ The details about how long that risk persists and what extent cancer screening at the time of diagnosis can change outcome remains unknown.

In the 1980s and 1990s, several studies prospectively analyzed the likelihood that individuals with an idiopathic embolic event would develop subsequent cancer.^37,40,42,45 In one of the earliest studies, Monreal and colleagues⁴² consented 113 patients with either idiopathic or secondary DVT to undergo laboratory work, clinical examination, endoscopy, and abdominal imaging to look for occult cancer. Seven of 31 patients with idiopathic DVT were found to have cancer, five of 82 patients with a secondary DVT.⁴² In 1992, Italian investigators followed 250 consecutive patients with symptomatic, venographically confirmed DVT for 2 years after their presentation. A total of 105 of these subjects had clots associated with well-known, noncancer risk factors. The remaining 145 were considered to be idiopathic DVTs. During the 2 years of follow-up, overt cancer developed in 1.9% of the patients with secondary venous thrombosis and 7.6% of the patients in the “idiopathic group.”⁴⁵ Other investigators using similar methods have found subsequent cancer rates ranging from 7.3%⁵⁰ to as high as 25% in patients with idiopathic DVT.^37,41 Variations in incidence among these studies likely reflect alternate definitions for idiopathic and secondary cases and variation in the intensity of cancer surveillance in each study. In a pooled analysis of these studies, the odds ratio for subsequent cancer in patients presenting with idiopathic VTE compared with secondary VTE was 4.8.⁴¹ A summary of these findings is presented in Table 5-1.

Although nearly all of these studies demonstrate an increased risk of subsequent cancer in the first 1 to 2 years after a diagnosis of idiopathic VTE, they are limited in power and design and have not provided definitive guidance on what workup should ensue in a patient with idiopathic VTE. Nevertheless, larger, population-based studies and database analyses have confirmed these prospective data. Using a Swedish registry of inpatients and linking it to a cancer database, Baron et al⁵¹ assessed cancer incidence among more than 61,000 patients who were admitted to Swedish hospitals with a diagnosis of VTE between the years 1865 and 1983. They found that the standardized incidence ratios (SIR; the observed number of cases divided by the expected number of cases in the age-matched normal population) for cancer during the first year after a diagnosis of VTE was 3.2, with particularly high rates of finding of polycythemia vera and cancers of the liver, pancreas, ovary, and brain.⁵¹ Of note, in their study, not only was there a large increase in the risk for diagnosis of virtually all cancers in the year after VTE diagnosis, but in subsequent years, a persistent 30% increase in risk remained. A study of Danish registry data has similar data, although generally lower standardized incidence ratios were reported.⁴⁷ In their 2005 study, researchers from the University of California at Davis used the California Cancer Registry to identify cases of common malignancies and linked them to a hospital discharge database to identify venous thromboembolic events in the year before the cancer diagnosis. Looking at more than 528,000 cancer cases, there was a SIR of 1.3, although the ratio was higher (2.3) for individuals with metastatic-stage cancer at diagnosis. Almost all of the unexpected VTE cases were associated with a diagnosis of metastatic-stage cancer within 4 months.⁴⁸

Given this understanding, researchers have investigated whether there is a basis for intensive search for malignancy in patients with unprovoked DVT. A good screening strategy would need to be safe and would need to improve the outcome of the VTE or the outcome of cancer. Testing such a strategy is best done with a randomized trial of screening versus no screening in patients with confirmed VTE, although trials of this nature are problematic.⁵² There has only been one study that published at attempt to randomize patients to two different screening arms.⁴⁴ This study, dubbed SOMIT (for extensive Screening for Occult Malignancy in Idiopathic venous Thromboembolism), failed to meet adequate accrual to demonstrate their endpoints. In this study, subjects with acute idiopathic VTE were randomized to a strategy of extensive screening or to no further testing. They were followed for 2 years. A total of 201 patients were enrolled, and 99 patients received screening, including computed tomography (CT) scans and ultrasonography of the abdomen and pelvis, endoscopies and sputum testing, tumor markers, Pap smears, and transrectal ultrasonography. This battery of testing identified otherwise occult cancer in a total of 13 (13.1%) patients of the 99 in the extensive screening group, and in this group, a single (1.0%) malignancy became apparent during follow-up. In the control group, a total of 10 (9.8%) malignancies became symptomatic (relative risk [RR], 9.7; 95% CI, 1.3–36.8; P < .01). Overall, malignancies identified in the extensive screening group were at an earlier stage, and the mean delay to diagnosis was reduced from 11.6 to 1.0 months (P < .001). However, the authors were unable to determine from their data whether or not the screening improved mortality among patients who later developed cancer.

In 2008, the Annals of Internal Medicine published a systematic review of evidence of 34 studies in an effort to answer the question of whether screening for cancer at the time of VTE improves overall outcome.²⁵ This analysis looked at 9516 patients with any diagnosis of VTE who had been included in these studies. They found the period prevalence of undiagnosed cancer to be about 6% at baseline in unprovoked VTE. If one extends that to 12 months from the diagnosis of VTE, the prevalence increases to 10%. This review showed that use of an extensive screening strategy, including CT of the abdomen and pelvis, detects more malignant conditions than a limited screening strategy without reported complications. However, the authors concluded that there was insufficient evidence to determine whether detecting those new malignancies results in a statistically significant change in early-stage, previously undiagnosed cases of cancer or a decrease in cancer-related mortality. What also remains unknown is whether detection reduces morbidity, ameliorates quality of life, is cost effective, or improves survival.

General practice, based on this body of literature to date, is that all patients who present with idiopathic VTE should have a full history and physical examination. All age-appropriate cancer screening should be completed. They should have basic blood work and, in smokers, a chest radiograph. Abnormalities in any of these tests should be investigated and followed. This procedure is expected to detect 90% of occult cancers.³⁹ All the authors on this topic stress that better trials are needed but acknowledge the associated challenges.

Virchow’s Triad

More than 150 years ago, Rudolph Virchow delineated three of the fundamental pathologies of clots—venous injury, stasis of blood, and a state of hypercoagulability.⁵³ This fundamental paradigm remains true today, and in malignancy, one can see all aspects of this triad playing a role in the increased clotting risk. For example, chemotherapy is known to cause injury to the endothelial lining, and there is increased stasis of blood in patients with large tumor burden or compressive disease. Certainly, many of the aspects of the hypercoagulability of malignancy have less to do with the microbiology of the tumor and more to do with the state of chronic illness that the host faces. Immobility secondary to fatigue, hospitalization, or surgery are all very common in cancer patients. Venous compression from tumor bulk also occurs. Infection, and the common surgeries—often large, visceral surgeries—also contribute to embolic risk. Cancer patients often have long-term indwelling catheters, an independent risk factor for embolism. Finally, good evidence suggests that several chemotherapy agents are prothrombotic as well. Layered on top of these risks is the biologic interaction of the tumor cell and the host environment, which drives the molecular hypercoagulable state in the case of cancer. An overview of these risks is presented in Figure 5-2.

Tumor Procoagulants

The pathophysiology of embolism in malignancy seems to be broader than the local effects. Trousseau’s observations called attention to nonlocal effects of the malignant cell manifested by changes to the host clotting system. Based on his observations of a relationship to gastrointestinal (GI) tumors, the phenomenon of VTE was originally linked to mucin-secreting tumors of the GI tract, and research initially proceeded along these lines. Since then, there has been confirmation that mucin-producing tumors indeed are thrombogenic.^11,54 However, it is now recognized that clot risk not only increases with mucin-secreting adenocarcinomas but also with a number of other tumor types. Although some groups are focusing on mucins and clots, research has also highlighted at least two other tumor-associated secreted factors that may predispose hosts to clots.