Venous thromboembolism and cancer: new issues for an old topic

Transcription

1 Critical Reviews in Oncology/Hematology 48 (2003) Venous thromboembolism and cancer: new issues for an old topic Mario Mandalà a,, Gianluigi Ferretti b, Marco Cremonesi a, Marina Cazzaniga a, Giuseppe Curigliano c, Sandro Barni a a Division of Medical Oncology, Treviglio Hospital, P. le Ospedale 1, Treviglio, Italy b Division of Medical Oncology A, Regina Elena Cancer Institute, Rome, Italy c Division of Medical Oncology, European Institute of Oncology, Milan, Italy Accepted 10 February 2003 Contents 1. Introduction Incidence during solid malignancy and occult cancer VTE and occult malignancy Is extensive screening justified? Pathophysiology Tissue factor Cancer procoagulant Tumour cell interaction with other blood cells VTE during chemotherapy or hormone therapy VTE in cancer patients: the role of chemotherapy VTE in cancer patients: the role of hormone therapy and chemo-endocrine therapy Conclusions Central venous catheters and incidence of thrombotic complications Incidence Risk factors Prophylaxis Treatment Treatment of venous thromboembolism in patients with solid tumours Bleeding during treatment Recurrent VTE Prevention in surgical patients Conclusions Reviewers References Biographies This paper is dedicated to Maria, Cecila amd Francesco. Corresponding author. Tel.: ; fax: address: mariomandala@tin.it (M. Mandalà) /$ see front matter 2003 Elsevier Ireland Ltd. All rights reserved. doi: /s (03)

2 66 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) Abstract Thromboembolic complications represent the second leading cause of death for cancer patients. Even though the correlation between cancer and a hypercoagulable state has been widely recognised, the pathogenesis of thromboembolism during malignancy is not yet entirely understood. The direct or indirect activation of the coagulation cascade favours neoplastic dissemination and metastasis. Disordered coagulation is encountered in up to 90% of cancer patients, although only 15% of them develop a localised acute or chronic deep venous thrombosis or a disseminated intravascular coagulation. This risk is significantly increased by chemotherapy, hormone therapy, surgery and central venous catheters. Therefore, much effort is needed to develop efficient prophylaxis and treatment, to reduce recurrence and bleeding and finally, to improve quality of life. Better knowledge about the biochemical bases of the coagulation process represents a pivotal step in cancer biology comprehension and global therapeutic management Elsevier Ireland Ltd. All rights reserved. Keywords: Review; Solid tumours; Venous thrombosis 1. Introduction In 1865, Trousseau [1] described the high incidence of venous thromboembolism (VTE) in a cohort of patients with gastrointestinal carcinoma. Many clinical and necropsy studies subsequently described a large series of thromboembolic accidents, such as deep venous thrombosis (DVT), pulmonary embolism, migrant thrombophlebitis, arterial thrombosis, non-bacterial thrombotic endocarditis and disseminated intravascular coagulation (DIC) [2 6]. The association between cancer and DVT, initially evidenced in metastatic patients, has actually been described also in patients with occult cancer, where DVT may be the first clinical manifestation of disease [7]. It is noteworthy that local growth and metastatic dissemination of malignancies are influenced by the coagulation system, especially by fibrin formation and deposition and other platelet function disorders [8]. Chemotherapy and hormone therapy may further impair the hemostatic balance by causing alterations of blood vessel walls or regulatory proteins of the coagulation cascade [9,10]. The aim of our review is to show the clinical and epidemiological interactions between thrombosis and cancer in four distinct steps: (1) DVT incidence during solid malignancy and occult cancer; (2) DVT incidence during chemotherapy or hormone therapy; (3) central venous catheters and incidence of thrombotic complications; and (4) prophylaxis and medical treatment of DVT in patients with solid tumours. 2. Incidence during solid malignancy and occult cancer The incidence of DVT in solid tumours is difficult to establish because most clinical studies have reported a wide variability based on the type of diagnostic procedure performed (clinical alone or clinical and objective testing, such as venography), the most frequent tumour histotype, the medical or surgical procedure and finally the presence of an indwelling central venous catheter (which in itself increases the risk of thrombosis of the axillary/subclavian vein). In addition, although the best way to evaluate the true incidence of clinical VTE is through prospective cohort studies, most data in the literature have been derived from retrospective and prospective studies not specifically designed to evaluate this issue. Even though pancreatic adenocarcinoma has classically been associated with a higher risk of DVT, it is reasonable that the distribution of specific cancers associated with thrombosis follows the frequency of the cancer in the general population. Recent data show that the highest incidence in men is observed in patients with lung, prostate and colorectal cancer and in women with breast, ovarian and lung cancer [11]. No studies have been able to definitively quantify the risk of developing a DVT for each histotype and/or site of the primary tumour. Most data have come from retrospective studies [5,12]. DVT incidence depends on tumour type, median patient survival, the use of chemotherapy or hormone therapy, frequency and type of surgical procedures and patient immobility. It is important to emphasise that clinical conditions associated with a higher incidence of DVT (surgical interventions, immobility, chronic obstructive lung disease, cardiac failure, use of estrogen or progesterone, pregnancy, puerperium) raise the already elevated risk of thromboembolism in cancer patients. From a clinical point of view, DVT of the lower limbs is the most common clinical manifestation of thromboembolic disease. However, pulmonary embolism, migratory thrombophlebitis, arterial thrombosis, non-bacterial thrombotic endocarditis, disseminated intravascular coagulation and thrombotic microangiopathy have all been described in cancer patients [2,3,5,6]. The activation of coagulation in patients with cancer contributes significantly to morbidity and mortality rates and may play a fundamental role in the host response to growing tumours. Patients with cancer are clearly at high risk of developing VTE, particularly during chemotherapy and surgery. This situation is aggravated by the use of venous access catheters and possibly by growth factors [13]. Data derived from large, randomised, controlled trials have been used to determine the true incidence of this complication of cancer and its treatment. In view of the morbidity and

3 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) mortality attributable to VTE in cancer, widespread utilisation of prophylactic anticoagulation therapy, which has proven safe and effective in a variety of situations, should be considered VTE and occult malignancy Whereas migratory thrombophlebitis is a clear indicator of an underlying neoplasm, the risk of cancer in patients with the more typical form of VTE has been the subject of intense debate over recent years. However, the cost-effectiveness of aggressive screening for cancer in patients with VTE has not yet been adequately defined. In 1952, Wright [6] hypothesised that bilateral recurrent venous thrombosis resistant to therapy could be the manifestation of an occult cancer in some patients. In a retrospective study, he suggested the need in such patients for an accurate physical examination, routine blood chemistry, chest X-ray, occult faecal blood test and Pap test in women [6]. Several retrospective studies have evaluated cancer incidence in patients with VTE compared with patients without VTE [7,14 17]. An overall estimation of pooled odds ratios showed a 2-fold higher risk (2.09; 95% CI, ) of a newly diagnosed malignancy in patients with compared to those without VTE. A critical analysis of these studies could reveal that it is hard to draw firm conclusions from retrospectives studies because of several methodological biases. First of all, in a retrospective study it is difficult to differentiate between idiopathic DVT and DVT secondary to well-recognised risk factors. In addition, it is reasonable to think that non-comparable search strategies have been performed in order to exclude a malignant neoplasm. Finally, selection bias may be present for non-consecutive patients. Two recent population-based retrospective studies demonstrated that the standardised incidence ratio of cancer in a large cohort of patients with VTE was substantially higher than in the general population, at least during the first 6 months after discharge [18,19]. A Danish study showed a rapid fall in the standardised incidence ratio after 6 months of follow-up and strongly suggested that a thromboembolic event in patients later given a diagnosis of cancer was the result rather than the cause of cancer. In fact, the ratio, which was 3.0 for that 6-month period, decreased to 2.2 at 1 year and to 1.1 for the remaining period of observation, which in some cases was as long as 17 years [18]. The study by Baron et al. [19] showed an increased risk even 10 years or more after the thromboembolic event. The reasons for the long-term increase in risk are not clear. However, the data are consistent with the hypotheses that premalignant changes promote thrombosis or that common factors predispose individuals to thrombosis and malignant disease. Future studies on the underlying mechanisms are likely to clarify the pathophysiology of VTE and cancer. To assess whether there are some groups of patients with DVT at higher risk of bearing an underlying malignancy, several studies have compared the incidence of cancer in patients affected by idiopathic DVT with those affected by DVT secondary to well-recognised risk factors [20 30]. Overall, cancer was more commonly found in patients with idiopathic DVT or pulmonary embolism than in patients with no risk factors for such conditions. In addition, the incidence of cancer in patients with recurrent idiopathic venous thrombosis was higher than in patients with secondary venous thrombosis; the occult cancer risk seemed to be higher when DVT presented bilaterally [24]. Therefore, there is general agreement that only this subgroup of patients has a higher risk of occult cancer. The length of time after an episode of VTE during which the risk of a newly diagnosed cancer is increased is not yet clear and whether vitamin K antagonists have an antineoplastic effect is still controversial. In a prospective randomised study on the duration of oral anticoagulation (6 weeks or 6 months) after a first episode of VTE, patients were questioned annually about any newly diagnosed cancer [31]. A first cancer was diagnosed in 111 of 854 patients (13.0%) during the follow-up. The standardised incidence ratio for newly diagnosed cancer was 3.4 (95% confidence interval, ) during the first year after the thromboembolic event and remained between 1.3 and 2.2 for the following 5 years. Cancer was diagnosed in 66 of 419 patients (15.8%) who were treated for 6 weeks with oral anticoagulants compared with 45 of 435 patients (10.3%) who were treated for 6 months (odds ratio, 1.6; 95% confidence interval, ). The difference was mainly due to the occurrence of new urogenital cancers: 28 cases in the 6-week group (6.7%) and 12 cases in the 6-month group (2.8%) (odds ratio, 2.5; 95% confidence interval, ). The difference in the incidence of cancer between the treatment groups became evident only after 2 years of follow-up and it remained significant after adjustment for sex, age and whether the thromboembolism was idiopathic or non-idiopathic. Older age at the time of the venous thrombosis and an idiopathic thromboembolism were also independent risk factors for a diagnosis of cancer. No difference in the incidence of cancer-related deaths was found Is extensive screening justified? Are survival data available today to support widespread screening for early diagnosis of those cancers associated with VTE? Which screening tests for identifying occult cancers should be used in patients with recurrent, idiopathic, bilateral DVT? Most authors agree that without definitive data to demonstrate an advantage in terms of overall survival using invasive diagnostic tests and intensive follow-up, patients should undergo only routine tests (physical examination, occult faecal blood test, chest X-ray, urological visit in men, gynaecological visit in women). The request for much more expensive examinations (CT scan, digestive endoscopy, tumour markers) should be limited only to cases with a strong clinical suspicion of occult cancer. Nevertheless, there is still insufficient evidence to support extensive

4 68 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) screening of all patients. Each patient with primary VTE should undergo a thorough clinical evaluation, including a comprehensive medical history, physical examination, routine laboratory tests and chest radiography [32]. The threshold for ordering tests if abnormalities are found during this initial evaluation should be low. A prospective trial (Screening for Occult Malignancy in Patients with Symptomatic Idiopathic Venous Thromboembolism, SOMIT) is ongoing in Italy to assess whether an extensive screening programme is able to identify early stage, treatable cancers in order to improve treatment possibilities and prognosis. 3. Pathophysiology The pathophysiologic mechanisms from which DVT originates in solid tumours are described in Virchow s classical triad, which includes coagulation disorders, vessel wall alterations and blood stasis. At different steps of malignant disorders, the homeostatic mechanisms can shift to a prothrombotic state in cancer through different pathways. Although different mechanisms, such as cytokines, vessel wall damage, platelet and monocyte activation, could contribute to clotting activation, we could consider an inflammatory process as the common pathway of these different stimuli [33]. Tumour cells can activate the coagulation cascade by a direct or indirect mechanism. The direct mechanism is mainly due to the action of two substances, tissue factor [34 37] and the so-called cancer procoagulant [38 41] Tissue factor Tissue factor is a transmembrane protein that has been identified in patients with lung, prostate, colon, breast and renal cancer, lymphoma and promyelocytic leukaemia. A direct correlation between tissue factor expression and cell differentiation has been described [36,37]. However, a limitation of some of these studies is that they evaluated tissue factor expression in tumour cells from primary or secondary cultures, in which there is a substantial modification of antigenic expression and protein function compared with the in vivo situation [42]. The tissue factor activates factor VII, forming a complex that activates factors IX and X and increases thrombin formation. Furthermore, thrombin activates factors XI, V and VIII as well as protein C in order to maintain a feedback inhibition of the coagulation cascade. The overall effect of tissue factor is the production of fibrin and platelet activation. The direct or indirect activation of the coagulation cascade favours neoplastic dissemination and metastasis, in part through the angiogenesis process [4,34,43] Cancer procoagulant Cancer procoagulant is a calcium-dependent cysteinprotease, which has been found in tumour cells and foetal tissues, but not in differentiated tumours [39]. It is present in cellular extracts obtained from patients with melanoma, colon, lung, breast and renal cancer and in patients with promyelocytic leukemia. It is able to activate factor X independently of tissue factor and factor VII [40]. The importance of cancer procoagulant in promyelocytic leukemia has been well documented by a decrease or normalisation during treatment with all transretinoic acid and differentiation in vitro [44 46] Tumour cell interaction with other blood cells Tumour cells can activate the coagulation system also indirectly, by the activation of monocytes, lymphocytes and endothelial cells. Stimulus of these cells, which act on the inflammatory-immunity network, could derive from tumour cell production of interleukin 1, vascular endothelial growth factor or tumour necrosis factor. The complement activation cascade and the immunocomplex formation can contribute to stimulate monocytes and endothelial cells, which are able to produce the tissue procoagulant factor when activated. Many authors have stressed the fact that adenocarcinoma could directly activate factor X by a non-enzymatic mechanism, since such tumours are able to produce mucin with a high content of sialic acid, which could in addition stimulate platelet-aggregating activity. Consequently, lung, pancreatic, gastrointestinal and ovarian adenocarcinomas could be particularly associated with DVT [47]. Although abnormalities of haemostasis are encountered in up to 90% of cancer patients [48], it is not yet possible to define the role of abnormal laboratory tests of haemostasis in order to predict a vein thrombosis in clinical practice. Therefore, larger prospective studies are needed to better define predictive and prognostic markers in solid tumours in order to evaluate the subgroup of patients in whom anticoagulant therapy is mandatory. 4. VTE during chemotherapy or hormone therapy Chemotherapy itself can increase the risk of thromboembolic disease by three mechanisms: (1) acute damage on vessel walls (bleomycin, carmustine, vinca alkaloids); (2) non-acute damage of the endothelium (adriamycin); and (3) a decrease of natural coagulation inhibitors (reduced level of proteins C and S with cyclophosphamide, methotrexate and fluorouracil (CMF) and reduced level of antithrombin III with l-asparaginase) VTE in cancer patients: the role of chemotherapy In a recent review of the literature, Piccioli et al. [49] reported that VTE incidence, in disease-free or metastatic breast cancer patients, ranged between 4 and 15%. The highest incidence of thromboembolic events occurs in metastatic breast cancer patients likely due to extended

5 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) disease, immobility for pathologic bone fractures, tumour cachexia and external compression of veins by the tumour mass. In a prospective randomised clinical trial, Levine et al. [50] demonstrated that the use of warfarin at the dose of 1 mg per day, maintaining the international normalised ratio (INR) between 1.3 and 1.9, significantly reduced the DVT incidence in metastatic breast cancer patients who received chemotherapy. A subsequent cost-benefit analysis also showed an economic gain using low doses of warfarin to prevent major thromboembolic complications [51]. Nevertheless, the event rate in the study was low (4.4 vs. 0.6%, in the study and control group, respectively); as a consequence, the prophylaxis should be tailored and individualised. Antithrombotic prophylaxis should be further considered in the case of adjuvant perioperative chemotherapy. Clahsen et al. [52] suggested a contributing role of perioperative chemotherapy to thromboembolic disease, especially in postmenopausal women and women undergoing mastectomy. Twenty-seven of 1292 patients assigned to the perioperative chemotherapy treatment arm (2.1%) and ten of 1332 patients without chemotherapy (0.8%) developed thromboembolic events (P = 0.004). The frequency of thromboembolic complications was higher among postmenopausal than premenopausal women (2.0 vs. 0.6%, P = 0.003). Patients who had undergone a mastectomy had a higher frequency of thromboembolic disease than those who had undergone a tumour resection (2.3 vs. 0.7%, P < 0.001). Three deaths occurred after pulmonary embolism, all of them in the perioperative chemotherapy treatment arm. The risk of DVT in early stage breast cancer patients under clinical control varies from 0.2 to 0.8%, whereas in patients who have received adjuvant therapy the risk ranges from 2 to 10%, reaching 17.6% in stage IV disease [53]. Levine et al. [54] demonstrated that chemotherapy contributes to thrombosis in patients with breast cancer. They performed a randomised trial comparing 12 weeks of chemo-hormone therapy (using cyclophosphamide, methotrexate, fluorouracil, vincristine, prednisone, doxorubicin, and tamoxifen) with 36 weeks of chemotherapy (using cyclophosphamide, methotrexate, fluorouracil, vincristine, and prednisone) in patients with stage II breast cancer. Among 205 patients randomised to treatment, there were 14 episodes of thrombosis (6.8%). These 14 episodes occurred during 979 patient months of chemotherapy; in comparison, there were no events during 2413 patient months without therapy. Monitoring with sophisticated coagulation tests during adjuvant epirubicin plus cyclophosphamide chemotherapy for breast cancer does not identify patients at high risk of DVT. Preoperatively, in patients with delayed DVT, an imbalance of haemostasis is already present; thus, thrombosis might predominantly be initiated by malignancy-induced hypercoagulability and secondarily by the influence of epirubicin plus cyclophosphamide chemotherapy. In 50 consecutive node-positive breast cancer patients, serial coagulation studies (fibrinogen method of Clauss, antithrombin III, protein C amidolytic methods, D-dimers, the enzyme-linked immunoadsorbent assay, plasminogen activator inhibitor (PAI) activity and u-pa inhibition test) and impedance plethysmography for screening of DVT were performed preoperatively and postoperatively, before each of six cycles of adjuvant chemotherapy (60 mg/m 2 epirubicin and 600 mg/m 2 cyclophosphamide) and 3 months thereafter. Seventy-two healthy women served as controls. During chemotherapy, the phlebographically proven DVT incidence was 10%. Preoperative levels of D-dimers, fibrinogen and PAI activity were significantly higher in the node-positive patients than in the healthy women, and only mean levels of the D-dimers were significantly higher in patients with than in patients without DVT. Postoperatively, only D-dimers and fibrinogen levels significantly increased, whereas such levels significantly decreased during chemotherapy. Mean D-dimer levels and PAI increased steadily in patients with DVT. Preoperatively and during chemotherapy, levels of antithrombin III and protein C were within the normal range. Only one patient with DVT had decreased protein C levels throughout chemotherapy [55]. Tables 1 and 2 show the data on DVT incidence associated with adjuvant hormone therapy and/or chemotherapy in radically resected breast cancer patients [56 67]. The limitations of these studies derive from the lack in many cases of Table 1 VTE incidence (%) in breast cancer patients who received adjuvant chemotherapy for early breast cancer (stage I II) Author(s) Study Regimens VTE incidence (%) Clahsen et al. [52] Prospective FAC perioperative 2.1 Randomized No treatment 0.8 Levine et al. [54] Prospective CMFVP+AT 4.9 (12 weeks) Randomized CMFVP (36 weeks) 8.8 IBCSG [156] Prospective CMF+AF periop. 0.5 Randomized No treatment 0 Weiss et al. [62] Prospective CMF (2 years) 3.5 Randomized CMFVP (2 years) 6.3 CMFbCG (2 years) 5.4 Pritchard et al. [60] Prospective CMF+T 13.6 Randomized T 2.6 Fisher et al. [65] Prospective Tamoxifen 1.2 Randomized MTF 4.2 CMFT 4.5 Tormey et al. [63] Prospective CMF 0 Randomized CMFPT 3.8 Wils et al. [64] Prospective E+T 0.03 Randomized T F = 5-fluorouracil, C = cyclophosphamide, A = adriamycin, V = vincristine, P = prednisone, T = tamoxifen, AF = folinic acid, M = methotrexate, E = epirubicin, bcg = bacillus Calmette Guerin.

6 70 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) Table 2 VTE incidence (%) in breast cancer patients treated with adjuvant tamoxifen Author(s) Study Regimens DVT incidence (%) Cummings et al. [59] Prospective Placebo 1.2 Randomized Tamoxifen 2.3 Saphner et al. [61] Meta-analysis Tamoxifen 2.3 Placebo 0.4 Rutqvist et al. [66] Prospective Tamoxifen 0.04 Randomized Placebo 0.04 Fisher et al. [58] Prospective Tamoxifen 0.9 Randomized Placebo 0.2 McDonald et al. [67] Prospective Tamoxifen 2.8 Randomized Placebo 2.2 a control population or the accurate instrumental diagnosis of DVT or pulmonary embolism VTE in cancer patients: the role of hormone therapy and chemo-endocrine therapy Tamoxifen raises the risk of developing a thromboembolism regardless of the presence of a neoplasm or use of chemotherapy. The NSABP study demonstrated that the risk of thrombosis in women older than 50 years and treated with tamoxifen was 2-fold that of a control population [57]. The risk of thromboembolism in patients taking tamoxifen as adjuvant treatment is 1 2% [58,59]. This risk is much higher in women undergoing chemo-endocrine therapy, reaching 13% [54,60]. Moreover, the risk is increased in postmenopausal women. Saphner et al. [61] showed that the frequency of thrombosis, venous and arterial combined, was 5.4% among patients who received adjuvant therapy and 1.6% among patients without chemotherapy (P = ). Premenopausal patients who received chemotherapy and tamoxifen had significantly more venous complications than those who received chemotherapy without tamoxifen (2.8 vs. 0.8%, P = 0.03). Postmenopausal patients who received tamoxifen and chemotherapy had an increased incidence of thromboembolism compared to those who received tamoxifen alone (8.0 vs. 2.3%, P = 0.03) or those without therapy (8.0 vs. 0.4%, P < ). Premenopausal patients who received tamoxifen and chemotherapy had a 1.6% frequency of arterial thrombosis, significantly more than patients who received chemotherapy alone (1.6 vs. 0.0%, P = 0.004). Thromboembolism related to the addition of CMF chemotherapy to tamoxifen as adjuvant therapy in this group of women represents a relatively common and serious complication that may outweigh any benefits offered by this additional therapy. Pritchard et al. [60] performed a randomised trial of tamoxifen, 30 mg per day for 2 years versus tamoxifen plus 6 months of intravenous chemotherapy with CMF for postmenopausal women with involved axillary nodes and positive estrogen receptor or progesterone receptor status following primary therapy for breast cancer. They observed one or more thromboembolic events in 48 of 353 women (13.6%) randomised to receive tamoxifen plus CMF compared to five of 352 women (2.6%) randomised to receive tamoxifen alone (P < ). Six women in the tamoxifen plus CMF arm, but none randomised to receive tamoxifen alone, suffered two thromboembolic events while on the study therapy. There were also significantly more women who developed severe (grade 3 5) thromboembolic events in the tamoxifen plus CMF arm than in the tamoxifen arm (34 vs. 5; P < ). Most thromboembolic events (39 of 54) occurred while women were actually receiving chemotherapy (P < ). Thromboembolic complications resulted in more days in hospital and more deaths than any other complication of therapy, including infection, in the trial. The mechanisms by which tamoxifen could augment the risk of thrombosis could be associated with its intrinsic estrogenic activity and reduced level of antithrombin III and protein C. However, alterations of the laboratory parameters would not completely explain the clinically assessable risk. Mannucci et al. [68] did not find any alteration of coagulation or fibrinolysis markers in non-cancer patients treated with tamoxifen (FpA, F1 + 2, TAT complex, D-dimers), although antithrombin and protein C levels were reduced. Resistance to activated protein C due to mutation Q506 of factor V increases the risk of thrombosis in patients receiving tamoxifen. Some authors believe that patients with a family history of thrombosis or with past episodes of VTE should undergo a screening before receiving tamoxifen [69]. Progesterone and its derivatives have been associated with superficial and deep thrombophlebitis in 45% of treated patients. However, the data are difficult to interpret, since most of the patients had metastatic breast cancer, a condition itself associated with thromboembolism [70,71]. Thromboembolic complications, such as pulmonary embolism, myocardial infarction and stroke, in relation to chemotherapy for germ cell cancer have primarily been described in case reports, whereas only a few studies have evaluated the incidence of thromboembolic toxicity in these patients. Germ cell cancer patients who receive chemotherapy, in particular those who have liver metastases or receive high doses of corticosteroids, are at considerable risk of developing thromboembolic complications. Weijl et al. [72] treated 179 germ cell cancer patients with a variety of combination chemotherapy regimens, primarily cisplatin-containing combination regimens. Of the 179 patients, 15 patients (8.4%) developed 18 major thromboembolic complications between the start of chemotherapy and 6 weeks after administration of the last cytostatic drug in first-line treatment. Of these 18 events, three (16.7%) were arterial events, including two cerebral ischemic strokes and 15 (83.3%) were venous thromboembolic events, including 11 pulmonary embolisms. One

7 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) (5.6%) of the 18 events was fatal. Liver metastases and the administration of high doses of corticosteroids as antiemetic therapy were identified as risk factors for the development of major thromboembolic complications Conclusions Today the benefits of chemotherapy and hormone therapy in the adjuvant setting seem to by far outweigh side effects. Patients with high tumour burden metastatic disease, past venous thrombosis and a family history of thrombophilia, especially if they have an indwelling central line, represent a subgroup that should receive a concomitant chemo-endocrine therapy with particular attention for the high risk of thrombosis. In the adjuvant setting, anticoagulation therapy should be reserved only for those patients with ascertained risk factors of VTE or a positive history of thrombosis. Most authors think that only patients at high risk should be treated with prophylactic anticoagulation therapy [73]. Finally, larger prospective studies should be performed in order to identify clinical and laboratory predictive markers for the identification of patients at risk of DVT. 5. Central venous catheters and incidence of thrombotic complications Lack of vascular access is one of the most common problems facing oncologists today. The increased use of dose-intensive and continuous infusion chemotherapy requires reliable vascular access, also for frequent blood sampling to monitor potential complications of chemotherapy treatments. In addition, the increased use of supportive care measures including intravenous antiemetics, analgesics, antibiotics, haematopoietic growth factors and hyperalimentation makes it imperative to obtain durable vascular access for an increasing number of cancer patients. For many patients, vascular access devices (VADs) obviate the need for repeated venipunctures, thereby increasing patient comfort. VADs are frequently used in patients with cancer. These devices can be divided into two groups: external catheters and subcutaneous venous access ports. Each type of device has its advantages and disadvantages, but the indications and optimal use of specific VADs remain to be defined. The subcutaneous venous access ports have the advantage of being more cosmetically acceptable, since they are implanted subcutaneously and require less maintenance. VADs have many complications, but with the exception of catheter-related bloodstream infections and thrombosis, most are rare Incidence Thrombosis remains a major complication of VADs and prospective, controlled studies are needed to clearly define the risk factors, natural history and optimal treatment of this complication [74 77]. Upper extremity DVT is frequently seen in cancer patients with VADs. The management of this complication is controversial. A recent review of the literature suggests that pulmonary embolism can be seen in approximately one-third of patients with DVT [78]. The administration of anticoagulants in cancer patients can be problematic, since many of these patients can be thrombocytopenic at the time of the diagnosis of DVT and may remain thrombocytopenic for prolonged periods of time. Although rare, the heparin-induced thrombocytopenia can be a devastating complication of heparin administration [79]. Catheter-related central venous thrombosis (CR-CVT) in cancer patients has been evaluated in a number of prospective and retrospective cohort studies, since central venous catheters are placed for chemotherapy, parenteral nutrition and other drug infusions. When surveillance venography was used systematically to detect this end point, or when autopsy data were included, the rate of CR-CVT was much higher then when venography was performed only for evaluation of symptoms suggestive for CR-CVT. Anderson et al. [80] evaluated the major complications of Hickman catheter placement (thrombosis and infection) in 168 patients with solid tumours (lung, head and neck, oesophagus, miscellaneous). Catheter-related thrombosis was clinically detected in 22 individuals and was detected at autopsy in six (total 17%). Patients with adenocarcinoma of the lung constituted a high risk group. Nine of 20 (45%) of these patients had thrombosis compared to 25, 9 and 16% of patients with squamous cell cancers of the lung, head and neck and oesophagus, respectively (P < 0.002). Thrombosis occurred despite daily heparin catheter flushing. Horne et al. [81] indicated that tunnelled venous access devices frequently cause partial venous occlusion within the first 6 weeks of catheterisation and that permanent venous damage from these devices is common, even without a history of VAD-related thrombosis. Thrombosis of the axillary-subclavian veins is a frequent event, even though there is no clinical evidence of flow obstruction; a fibrin coating of the central venous catheter is present in most of the cases. Balestreri et al. [82] evaluated phlebographically 57 oncologic patients with short or long-term central venous catheters and without clinical signs of axillary-subclavian thrombosis. Different degrees of incomplete thrombosis were found in 26 patients (45.5%) and complete thrombosis, clinically silent, was found in six patients (10.5%). A fibrin sleeve around the central venous catheter was radiologically demonstrated in 45 (78%) patients, 21 of them (46%) with a negative standard venogram. There were no significant differences between patients with long- and short-term central venous catheters Risk factors To identify factors associated with the development of clinically significant venous thrombosis in cancer patients

8 72 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) with long-term indwelling subclavian Groshong catheters, Gould et al. [83] performed a longitudinal study of the outcome of clinical practice. Some 70% of the thrombi occurred after an episode of previous catheter dysfunction; only 30% of the thrombi occurred de novo. Developing a CR-CVT was more likely in patients with a catheter inserted in the left subclavian circulation or with a previous episode of aspiration difficulty with the catheter ( ball-valve effect ) [83]. Schwarz et al. [84] reported that transcutaneous tunnelled central venous lines could be placed safely, with a considerable incidence of subsequent device-specific complications, but a high salvage rate. Factors determining outcome were related to device placement, as well as the patient s disease status. In their study, 90 days after catheter placement there was a 37% chance of device complication and a 20% possibility of catheter loss. Long-term device duration is primarily influenced by patient survival. In another study by Schwarz et al. [85], 180 patients who underwent subcutaneous intravenous port placement were followed prospectively until port removal, death, or a maximum of 1960 days. Some 90% of patients alive at 1 year and 70% of patients alive at 4 years had a functional port. Most published data about peripherally inserted central catheter lines is in the area of chemotherapy. Some studies have supported the use of peripherally inserted central catheter lines in patients receiving a variety of solutions. With an experienced team approach to catheter placement and maintenance, peripherally inserted central catheter lines provide reliable, cost-effective venous access and reduce many of the complications of central venous access in a variety of clinical settings [86]. Nightingale et al. [87] presented a prospective analysis of the insertion complications and longevity of 949 cuffed, tunnelled central venous catheters used for ambulatory chemotherapy. Catheters inserted with their tip in the superior vena cava were more at risk of removal (2.57 times) than those in the right atrium (P=0.003). Eastridge and Lefor [88] found a significantly increased incidence of thrombotic complications in patients with triple-lumen catheters (ten of 48) compared with double-lumen catheters (11 of 160), as well as a significantly (P < 0.05) decreased mean time until catheter failure (40 vs. 146 days). They also observed a significant increase in the rate of thrombosis in patients with a catheter tip above the T3 level. As regards risk factors potentially important in the development of CR-CVT, the difference in thrombosis rate does not appear to be dependent on the type of catheter adopted (Hickman, Broviac, Port-a-Cath, PICC) [83,86]. However, there is a positive correlation between the catheter size and the risk of thrombosis. In fact, single-lumen catheters present a lower risk than double or triple-lumen catheters [88]. Haire et al. [89] showed that double-lumen catheters tended to be more likely to cause total venous occlusion (nine of 23) than single-lumen catheters (one of 12) (P = 0.06, Fisher s exact test). It also seems that catheter tip position distal to the superior vena cava or right atrium carries a higher risk of thrombosis [87,88]. Most thromboses of the central veins are associated with left-sided placement of the ports, with catheter tips lying against the external wall in the upper half of the superior vena cava. This position allows a narrow contact between the catheter tip and the vessel wall, thus endothelial injuries might result from mechanical and chemical attack. Puel et al. [90] suggested that patients with left-sided ports and catheter tips lying in the upper part of the vena cava are at high risk for severe thrombotic complications. Predisposing factors in the patient population studied by Lokich et al. [91] included decreased antithrombin III levels prior to catheter insertion, intrathoracic tumour, possibly creating low flow rates and improper or suboptimal home management of the catheter. The development of subclavian vein thrombosis was not related to the chemotherapeutic agent infused or the duration of therapy [91]. Overall, these data indicate that upper extremity DVT is frequently seen in cancer patients with VADs, although the true incidence is difficult to estimate because of the definition of venous thrombosis, which has been defined by a positive venographic (66%) study [92], by the need for urokinase instillation (10%) or ipsilateral upper extremity swelling [93] or by removal (3%) of the venous catheter because of thrombosis [94] Prophylaxis The frequency of vein thrombosis in this patient population necessitates consideration of prophylactic methods in high-risk patients. To our knowledge, only three prospective randomised trials have been reported as full papers in order to investigate this important issue [95 97]. Bern et al. [95] described the results obtained in 82 assessable patients at risk for thrombosis associated with chronic indwelling central venous catheters who were prospectively and randomly assigned to receive or not to receive 1 mg of warfarin, beginning 3 days before catheter insertion and continuing for 90 days. Venograms were obtained at the onset of thrombosis symptoms or after 90 days in the study. Among patients who completed the study, four of 42 who received warfarin had venogram-proven thrombosis, whereas 15 of 40 patients completing the study while not receiving warfarin had venographically proven thrombosis and ten had symptoms of thrombosis (P = 0.001). In contrast, in patients with haematological malignancies, Heaton et al. [96] found no benefit in the routine prophylactic use of minidose warfarin for the prevention of central vein catheter thrombosis. Monreal et al. [97] conducted an open randomised trial which demonstrated the efficacy of dalteparin (2500 IU per day) for prophylaxis of catheter vein thrombosis. Overall, the authors demonstrated venogram-proven thrombosis in 6% of treated patients versus 62% in the control group [97]. Although physicians universally agree about the increased risk of thromboembolic complications in patients with venous access devices, the results of the aforementioned trials have not had much impact on clinical practice, particularly

9 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) in North America. In addition, the results of two recent studies suggest that the incidence of thrombosis has substantially decreased over the past years, owing to the improved care of catheters, so that the benefit of prophylactic anticoagulation could be questionable [98,99]. A recent study assessed the compliance rate of prophylactic low-dose warfarin prescribed in cancer patients with central venous catheters at a single institution [100]. The authors found that only 10% of patients had been prescribed prophylactic warfarin [100]. Several reports have suggested that prophylaxis in cancer patients with central venous catheter should be mandatory, and certainly further large prospective studies are needed to identify a subgroup of patients with a high risk for thrombosis and those who develop thrombotic complications even with appropriate prophylaxis with warfarin or low-molecular weight heparin [77,100] Treatment The first step for the treatment of catheter-associated thrombosis is the differential diagnosis of a catheter dysfunction due to a thrombus at the tip of the catheter or thrombosis of a venous segment. Catheter tip dysfunction has classically been treated with urokinase (5 10,000 U) or tissue plasminogen activator (2 mg) instilled for min. The procedure is likely to be successful 70 90% of the time (eventually repeated if not unsuccessful). If the problem persists, diagnostic testing for DVT or catheter malposition should be performed (ultrasonography or venography). When a venous thrombosis is identified, there is no standard therapy and a variable number of therapeutic approaches can be considered. A first approach is urokinase or tissue plasminogen activator, especially when the infusion catheter is placed in close proximity to the thrombus. Alternatively, catheter removal plus heparin and then warfarin administration may be considered. Finally, if the catheter is critical for patient care, intravenous heparin followed by warfarin can be given without catheter removal. Large and well-designed prospective studies are needed in order to identify the best approach to this patient subgroup. 6. Treatment of venous thromboembolism in patients with solid tumours The conventional treatment in cancer patients with DVT includes unfractionated heparin and oral anticoagulants. Unfractionated heparin is administered first as push (5000 IU) followed by continuous infusion (nearly 30,000 IU over 24 h, adjusted for the activated partial thromboplastin time, which is maintained at a 2-fold value compared with normal). Some patients, so-called resistant to heparin, require a regular increase in dose (up to 50,000 IU). Resistance could be due to high levels of factor VIII and fibrinogen or to a congenital lack of antithrombin. Oral anticoagulants should be started within 24 h of the first heparin administration. Heparin should be continued for at least 5 days and suspended after the therapeutic range has been reached by oral anticoagulants (INR 2-3) for at least 2 consecutive days. The concomitant comorbidities due to thrombocytopenia, infections, chemotherapy and liver metastases may create problematic situations for anticoagulant therapy in clinical practice. The routine use of anticoagulant prophylaxis during infusion of 5-fluorouracil is theoretically hampered by a potential interaction between warfarin and the drug. Prolonged 5-fluorouracil half-life and increased INR have been reported, probably due to interference with the synthesis of hepatic cytochrome 450 and impaired metabolism of warfarin and 5-fluorouracil [101,102]. During the last few years, low-molecular-weight heparin (LMWH) has been introduced in clinical practice. LMWH is obtained by depolymerisation procedures of unfractionated heparin, isolating sequences with an average of 15 sacchardic residues. Their antithrombotic activity is linked to the capacity of neutralising factor Xa, while the ability of neutralising thrombin by antithrombin is reduced. Furthermore, LMWH has subcutaneous bio-availability close to 100%, since it links weakly to serum proteins and very weakly to endothelium. Three randomised clinical trials have recently demonstrated that in patients with acute proximal DVT, LMWH administered subcutaneously twice a day (at home) has the same efficacy as unfractionated intravenous heparin administered in the hospital [ ]. In addition, the incidence of thrombotic recurrence in the cancer patient subgroup was the same with both treatments. Therefore, the administration of LMWH represents an interesting therapeutic alternative in cancer patients with acute DVT in order to improve quality of life. When we consider that cancer patients represent 10 20% of study populations in published therapeutic DVT trials, LMWH could be considered as a new first-line standard therapy. Moreover, the results of two recent cohort studies in patients with DVT and pulmonary embolism demonstrated that non-trial patients can be treated safely at home [106,107] Bleeding during treatment The estimated risk of major bleeding with warfarin therapy ranges from 0.2 to 5.2% [108,109], but there is controversy about whether patients with cancer are at increased risk of recurrences and bleeding complications during anticoagulant therapy. Several investigators have observed an increased risk of haemorrhagic complications in cancer patients [110,111], whereas others could not, or only partially, confirm this observation [112]. In a recent retrospective analysis of two prospective studies [104,105], Hutten et al. [113] found that the risk for recurrence and major bleeding in cancer patients was 3- to 6-fold that of patients without malignancy during treatment with warfarin for VTE. These

10 74 M. Mandalà et al. / Critical Reviews in Oncology/Hematology 48 (2003) findings are consistent with the results of Gitter et al. [114], who reported that malignant disease at initiation of warfarin therapy is significantly associated with an increased risk for major bleeding and recurrent thromboembolism. Wester et al. [115] and Landefeld et al. [116] also found that malignancy is associated with a higher frequency of bleeding complications Recurrent VTE The risk of recurrence in cancer patients seems to be higher than in non-cancer patients. The exact rate of failure is difficult to calculate because most of the studies have been small and retrospective. In the prospective study of Prandoni [111], the risk of recurrence, during the first 3 months of an appropriate oral anticoagulation therapy, was higher in cancer patients than in non-cancer patients (8.6 and 1.3%, respectively, P < 0.01). Recently, Heit et al. [117] confirmed Prandoni s data, reporting a 2- to 3-fold risk of recurrent thrombosis in cancer patients in a population-based cohort study (1719 patients). Hansson et al. [118] prospectively followed for years, 738 consecutive patients with an objectively verified symptomatic DVT. The reported relative risk of cancer versus no cancer patients was Overall, patients with a diagnosed cancer have a higher risk of recurrent DVT, and this knowledge could certainly help to identify patients who might benefit from prolonged prophylactic treatment in various risk situations. Considering the high risk of recurrence in cancer patients, the current general practice is to continue anticoagulant therapy until there is clinical evidence of cancer or further/additional risk for thrombosis. Patients who develop a recurrence DVT despite therapeutic INR levels represent a difficult therapeutic challenge. There is evidence that LMWH is efficacious in patients who fail on warfarin [119]. Luck et al. [120] showed that long-term therapy with LMWH may be effective in managing warfarin-failure thromboembolic disease. In patients at high risk for pulmonary embolism, an inferior vena cava filter can be placed in addition to the above measures. The use of an inferior vena cava filter alone should be considered in patients with recurrent pulmonary embolism despite adequate anticoagulation and a contraindication to anticoagulant therapy. The most common reasons are active bleeding and profound, prolonged thrombocytopenia. LMWH may also represent a promising tool for secondary prophylaxis of venous thromboembolism in cancer patients. Two recent studies clearly demonstrated that prolonged treatment with LMWH may be safer [121,122] and also more effective than oral anticoagulants for the secondary prevention of venous thromboembolism in patients with cancer [122]. Although large and prospective studies are needed to clarify this issue, in the near future LMWH could be the new standard therapy in the secondary prophylaxis of venous thromboembolism in cancer patients. LMWH could have the advantage of avoiding laboratory monitoring and in order to improve the cancer patient s quality of life, its use could be beneficial. 7. Prevention in surgical patients Cancer patients who undergo surgery are at high risk of developing a thromboembolic complication. Kakkar and Williamson [123] reported that cancer patients undergoing a surgical procedure have twice the risk of postoperative DVT and more than three times the risk of fatal pulmonary embolism than patients who undergo surgery for benign disease. In addition to the increased risk for cancer itself, a large group of patients present with additional clinical risks, such as increasing age, prolonged immobility, obesity and indwelling venous catheter. Table 3 shows the DVT incidence in cancer patients without prophylaxis compared with a control group that underwent surgical resection for benign disease [ ]. A recent consensus assessed that, without heparin prophylaxis, cancer patients develop a DVT of proximal inferior limbs in 10 20% of cases, whereas pulmonary embolism occurs in 1 5% of such patients [131]. Meta-analysis on prophylaxis before surgery demonstrated a reduced incidence of VTE in patients who received heparin prophylaxis (13.6%) compared with patients with no anticoagulation (30.6%) [132]. It has been reported that the incidence of postoperative pulmonary embolism is higher in cancer patients than in a control group that underwent surgery for benign disease [133]. The therapeutic approaches used for the post-surgical DVT include compression stockings, low-dose unfractionated heparin (5000 IU given daily for 8 12 h starting 1 2 h before the operation) and more recently LMWH. In all reports, the role of prophylaxis is unquestionable, considering patients with DVT before surgery as well. Gallus et al. [134] evaluated the effect of low-dose heparin prophylaxis on venous thrombosis in a prospective controlled study of 820 patients who underwent major elective surgery. A subgroup analysis of cancer patients revealed that low-dose unfractionated heparin is able to reduce DVT from 22% (control group) to 9% in cancer patients. Table 3 Postsurgical risk of VTE in cancer patients versus non-cancer patients Author(s) Postsurgery VTE frequency Cancer patients No. (%) Non-cancer patients No. (%) Kakkar et al. [124] 24/59 (41) 38/144 (26) Walsh et al. [125] 16/45 (35) 22/217 (10) Rosenberg et al. [126] 28/66 (42) 29/128 (23) Sue-Ling et al. [127] 12/23 (52) 16/62 (26) Allan et al. [128] 31/100 (31) 21/100 (21) Multicenter Trial 9/37 (22) 13/53 (24) Committee [129] Huber et al. [130] 42/1796 (2.34) 62/17,365 (0.36) In parenthesis, percentage.