Risk of Venous Thromboembolism With the Angiogenesis Inhibitor Bevacizumab in Cancer Patients: Title and subTitle BreakA Meta-analysis FREE

Angiogenesis, a process involving the proliferation of new blood vessels, plays a crucial role in the growth and metastasis of cancer.¹ This process is mainly driven by the vascular endothelial growth factor (VEGF), whose signaling pathway has been a target of many new antineoplastic agents including bevacizumab, sorafenib, sunitinib, and others.^{2 - 3} Bevacizumab (Avastin, Genentech Inc, South San Francisco, California), a recombinant humanized monoclonal neutralizing antibody against VEGF, has shown benefits in the treatment of many types of malignancy including colorectal cancer, non–small cell lung cancer (NSCLC), renal cell carcinoma, and breast cancer.

Earlier studies have established its effectiveness for patients with metastatic carcinoma of the colon or rectum in combination with intravenous fluorouracil-based chemotherapy^{4 - 8} ; it also showed a survival benefit as the first-line treatment for patients with advanced NSCLC in combination with carboplatin and paclitaxel.^{9 - 10} In addition, its efficacy has been evaluated in metastatic breast cancer when combined with chemotherapeutic agents, such as capecitabine or taxol.^{11 - 12} The combination of bevacizumab with taxol showed an improved progression-free survival for patients with previously untreated metastatic breast cancer,¹¹ and it has been recently approved by the US Food and Drug Administration for this use. Furthermore, bevacizumab improved the progression-free survival of patients with metastatic renal cell carcinoma when combined with interferon alfa for first-line therapy.¹³ It also is undergoing extensive evaluation for a variety of other solid tumors such as malignant mesothelioma and pancreatic cancer, or in combination with other antineoplastic agents.^{14 - 20}

Serious adverse effects are associated with the use of bevacizumab, including gastrointestinal tract perforation, wound dehiscence, hemorrhage, arterial thromboembolic events, hypertensive crisis, reversible posterior leukoencephalopathy syndrome, neutropenia and infection, nephrotic syndrome, and congestive heart failure.²¹ While bevacizumab is recognized to be associated with an increased risk of arterial thromboembolic events,²² it is controversial whether bevacizumab contributes to the development of venous thromboembolism, a common complication leading to morbidity and mortality in patients with malignancy.

Reported incidences of venous thromboembolism in patients treated with bevacizumab varied substantially from 3% to 19.4% across phase 2 and 3 randomized controlled trials (RCTs).^{7 ,13} Pooled analysis from 5 RCTs that included 1745 patients with metastatic colorectal cancer, breast cancer, or NSCLC showed that bevacizumab in combination with chemotherapy was not associated with an increased risk of venous thromboembolism (relative risk [RR], 0.89 [95% confidence interval {CI}, 0.66-1.20]; P = .44) when compared with chemotherapy alone.²² This result has propagated a view that venous thromboembolism is not one of the more serious adverse effects attributable to bevacizumab. Because the number of patients included in this analysis is limited, the contribution of bevacizumab to venous thromboembolism remains poorly defined.

Many more RCTs have been performed since then. A recent RCT showed that bevacizumab in combination with an interferon was associated with a significantly increased risk of venous thromboembolism (P = .02) in comparison with interferon treatment alone¹³ ; in addition, several other studies revealed a higher incidence of venous thromboembolism consistently associated with bevacizumab treatment even though it is not significantly different when compared with controls.^{4 ,9 ,11 ,19}

We hypothesized that sample sizes in these studies were not powered to reveal a significantly increased risk of venous thromboembolism with bevacizumab; therefore, we performed a systematic review and combined the published RCTs for a meta-analysis to evaluate the effect of bevacizumab on the occurrence of venous thromboembolism in cancer patients.

An independent review of citations from PubMed from January 1966 to January 2008 was conducted. Key words included in the search were bevacizumab, avastin, cancer, human studies, and clinical trial. The search was limited to articles published in the English language. The search strategy also used the text terms thromboembolism, angiogenesis, and vascular endothelial growth factor to identify relevant information. Abstracts and virtual meeting presentations containing the term bevacizumab or avastin from the American Society of Clinical Oncology conferences (http://www.asco.org/ASCO) held between January 2000 and January 2008 also were searched to identify relevant clinical trials. An independent search using the citation database Web of Science (developed by the Institute for Scientific Information) also was performed to ensure that no clinical trials were missed.

We reviewed each publication and only the most recent or complete report of clinical trials was included when duplicate publications were identified. Efforts also were made to contact the investigators and the manufacturer of bevacizumab when relevant data were not clear. The updated manufacturer's package insert of bevacizumab also was reviewed to identify related information.²³

The goal of this study was to determine the specific contribution of bevacizumab to the development of venous thromboembolism in cancer patients. Thus, only RCTs with a direct comparison between cancer patients treated with and without bevacizumab were included for analysis. Phase 1 trials and single-arm phase 2 trials were excluded from analysis due to a lack of controls. Specifically, clinical trials that met the following criteria were included in the meta-analysis: (1) prospective phase 2 and 3 RCTs of patients with cancer; (2) random assignment of participants to bevacizumab treatment or control (placebo or best supportive care) in addition to concurrent therapy using a chemotherapeutic agent or a biological agent; and (3) available data including event or incidence of venous thromboembolism and sample size for analysis. Quality was assessed using criteria including adequate blinding of randomization, completeness of follow-up, and objectivity of outcome measurements as previously described.²⁴ We also corresponded with study authors to clarify issues.

We extracted details on study characteristics, patient characteristics, treatment information, results, and follow-up from the selected trials. Incidences of venous thromboembolism were extracted from the safety profile in each trial. Two reviewers extracted the data independently (S.R.N. and S.W.). Any discrepancies between reviewers were resolved by consensus. Venous thromboembolism in these studies was assessed and recorded according to the National Cancer Institute's common toxicity criteria (version 2 or 3; http://ctep.cancer.gov), which has been widely adopted in cancer clinical trials. Both versions are the same regarding the grading of venous thromboembolism, which starts at grade 2. A deep venous thrombosis or cardiac thrombosis with no intervention indicated is a grade 2; a deep venous thrombosis or cardiac thrombosis with an intervention (eg, anticoagulation, lysis, filter, invasive procedure) indicated is a grade 3; and an embolic event including pulmonary embolism or life-threatening thrombus or death is a grade 4. We have included incidences of all-grade venous thromboembolism (≥grade 2) and high-grade venous thromboembolism (≥grade 3) for the analysis.

All statistical analyses were performed using version 2 of the Comprehensive Meta-analysis program (Biostat, Englewood, New Jersey). For the calculation of incidence, the number of patients with venous thromboembolism (both all and high grades) and the number of patients receiving bevacizumab were extracted from selected clinical trials. The proportion of patients with venous thromboembolism and exact 95% CIs were derived for each study. For the calculation of RR, patients assigned to bevacizumab were compared only with those assigned to control treatment in the same trial. To explore a dose-effect relationship, bevacizumab therapy was further divided into low dose (5 or 7.5 mg/kg per dose per schedule, which is equivalent to a weekly dose of 2.5 mg/kg) and high dose (10 or 15 mg/kg per dose per schedule, which is equivalent to a weekly dose of 5 mg/kg). The designation of low dose vs high dose was relatively arbitrary. We have previously shown that the risk of hypertension and proteinuria with bevacizumab is dose-dependent, with increased risks associated with the high dose of bevacizumab.²⁵

For the meta-analyses, both the fixed-effects model (weighted with inverse variance) and the random-effects model were considered.²⁶ For each meta-analysis, we assessed the between-study heterogeneity using the Cochran Q statistic and study consistency using the I² statistic. For the Cochran Q statistic analysis yielding a P value of less than .10, the assumption of homogeneity was deemed invalid²⁷ and a random-effects model was reported. The causes of heterogeneity also were explored in this context. Otherwise, results from a fixed-effects model were reported. We used the Begg and Egger tests to evaluate the presence of publication bias regarding our primary end points (RR of venous thromboembolism). A 2-tailed P value of less than .05 was judged as statistically significant.

Our search yielded a total of 209 potentially relevant clinical studies on bevacizumab in the literature (see Figure 1 for the selection process of these studies). A total of 15 RCTs were included for this analysis, encompassing 5 phase 2 trials and 10 phase 3 studies (Table 1).

Randomized treatment allocation sequences were generated in all trials. Four trials had the double-blind, placebo-controlled design^{13 ,15 - 17} ; 3 other trials had placebo controls.^{5 ,8 ,20} The incidence of venous thromboembolism from 1 trial involving colorectal carcinoma was extracted from a graph of treatment toxic effects presented by Saltz et al.²⁰ The incidence of venous thromboembolism from another trial involving lung cancer was extracted from the updated manufacturer's package insert.^{10 ,23} Two small phase 2 studies (1 on renal cell carcinoma by Yang et al²⁸ and another on breast cancer by Lyons et al²⁹ ) did not report enough information regarding venous thromboembolism and were excluded from the analysis.

No evidence of publication bias was detected for the primary end point of this study (RR of venous thromboembolism) by either the Begg or Egger test (Begg test, P = .43; Egger test, P = .59).

A total of 7956 patients from 15 phase 2 and 3 RCTs (bevacizumab, n = 4292; control, n = 3664) were included for analysis. The baseline characteristics of patients in the 15 studies are listed in Table 1. Venous thromboembolism was not listed as a baseline characteristic in any of the patients. The baseline Eastern Cooperative Oncology Group (ECOG) performance status for most of the patients was between 0 and 1. Of the 15 studies with underlying malignancies, 6 included colorectal cancer, 4 included NSCLC, 2 included breast cancer, and 1 each included pancreatic cancer, malignant mesothelioma, and renal cell carcinoma. Patients with significant cardiovascular disease, bleeding diathesis, coagulopathy, and those requiring therapeutic anticoagulation, regular nonsteroidal anti-inflammatory drugs, and aspirin (>325 mg/d) use were excluded from most of the studies at the time of enrollment. Many trials allowed patients to take low-dose warfarin (1 mg) for catheter patency or central line patency. Treatment options in all trials were randomly assigned to bevacizumab or control.

A total of 2279 patients with colorectal cancer, NSCLC, breast cancer, and renal cell cancer from 6 studies with available venous thromboembolism data were identified. Among patients who received bevacizumab, the incidences of all-grade venous thromboembolism in these studies ranged between 3% and 19.1%, with the highest being in the 2 trials of patients with advanced colorectal cancer,^{5 ,7} and the lowest incidence observed in the RCT of patients with metastatic renal cell carcinoma.¹³ Using a random-effects model, meta-analysis revealed that the summary incidence of all-grade venous thromboembolism in these patients was 11.9% (95% CI, 6.8%-19.9%; Table 2).

High-grade venous thromboembolism (grades 3-5) is associated with significant morbidity and mortality. Thirteen studies reported incidence of high-grade venous thromboembolism. A total of 3795 patients with various advanced solid tumors were treated with bevacizumab in combination with chemotherapy or a biological agent and high-grade venous thromboembolism data were available for analysis. The incidences of high-grade venous thromboembolism in these studies ranged between 2% and 17%, with the highest incidence observed in the phase 2 clinical trial of patients with malignant mesothelioma,¹⁵ and the lowest incidence observed in the phase 3 trial of patients with metastatic renal cell carcinoma.¹³ Using a random-effects model for this analysis (heterogeneity test: Q = 46.74, P < .001, I² = 74.33), the summary incidence of high-grade venous thromboembolism was 6.3% (95% CI, 4.8%-8.3%; Table 3).

Most of the included studies reported high-grade venous thromboembolism in a combined manner for grades 3 through 5. From the limited data available for separate analysis of individual grade, the incidence and RR for grade 3 venous thromboembolism was 2.9% (95% CI, 1.9%-4.3%) and 2.21 (95% CI, 1.06-4.59), respectively, as determined by meta-analysis from 3 studies.^{4 ,11 - 12} The incidence and RR for grade 4 venous thromboembolism was 2.1% (95% CI, 1.0%-4.3%) and 0.75 (95% CI, 0.44-1.27), respectively, as determined by meta-analysis of 5 trials.^{4 - 5 ,8 ,11 - 12} The event for venous thromboembolism–related death (grade 5) was rare among all these trials, with a total of 2 deaths in the treatment group and 1 death in the control group from 3 trials.^{7 ,10 ,14}

The observed high incidences of venous thromboembolism with bevacizumab may be secondary to known risk factors such as underlying malignancy, mobility, and chemotherapy. To define the particular contribution of bevacizumab to the development of venous thromboembolism and exclude the influence of confounding factors, we have determined the overall RR of venous thromboembolism from 15 RCTs in which patients were treated with or without bevacizumab in combination with concurrent standard therapy.

A meta-analysis was performed to calculate the overall RR of venous thromboembolism (combination of all grade and high grade if data for all grade were not reported) associated with bevacizumab in comparison with controls. As expected for patients with advanced malignancy, the incidences of venous thromboembolism were variable and substantial in these controls ranging from 2% to 18% for all-grade venous thromboembolism and from 0.7% to 9.5% for high-grade venous thromboembolism. No heterogeneity was found among the included studies despite apparent differences in tumor type and treatment. Using the fixed-effects model, meta-analysis showed that the summary RR of venous thromboembolism for bevacizumab vs control was 1.33 (95% CI, 1.13-1.56; P < .001) (Figure 2). Bevacizumab was therefore found to be associated with a significantly increased risk of venous thromboembolism; the risk of developing venous thromboembolism with bevacizumab was 33% greater than with a control.

In addition, we have determined overall RRs of bevacizumab vs control for all-grade and high-grade venous thromboembolism separately. As shown in Table 2, the summary RR of all-grade venous thromboembolism was 1.29 (95% CI, 1.03-1.63; P = .03), suggesting a 29% greater risk for developing all-grade venous thromboembolism with bevacizumab compared with a control. As shown in Table 3, the summary RR of high-grade venous thromboembolism was 1.38 (95% CI, 1.12-1.70; P = .002), suggesting a 38% greater risk for developing high-grade venous thromboembolism with bevacizumab compared with a control.

To further understand the role of bevacizumab in the development of venous thromboembolism in cancer patients, we explored the relationship between the dose of bevacizumab and the RR of venous thromboembolism. A meta-analysis of the RR associated with bevacizumab compared with a control based on the dose of bevacizumab (high or low) was performed using a fixed-effects model. From 11 trials containing 4858 patients (heterogeneity test: Q = 6.25, P = .79, I²<.001), high-dose bevacizumab treatment with 5 mg/kg per week was associated with a significantly increased risk of venous thromboembolism with an RR of 1.31 (95% CI, 1.02-1.68; P = .04). From 7 trials containing 3594 patients (heterogeneity test: Q = 4.97, P = .55, I²<.001), low-dose bevacizumab treatment with 2.5 mg/kg per week was associated with a significantly increased risk of venous thromboembolism with an RR of 1.31 (95% CI, 1.08-1.60; P = .007). Thus, both high and low doses of bevacizumab were associated with a significantly increased risk of venous thromboembolism.

We have further determined the risk of venous thromboembolism with bevacizumab separately according to their histology to investigate the relationship between tumor type and venous thromboembolism. The incidences and RR of all-grade venous thromboembolism with bevacizumab vary among different tumors (Table 2). The highest incidence was observed among patients with colorectal cancer; meta-analysis showed that the incidence of all-grade venous thromboembolism was 19.1% (95% CI, 16.1%-22.6%). For patients with NSCLC, the incidence of all-grade venous thromboembolism was 14.9% (95% CI, 8.2%-25.5%), while for patients with breast cancer, the incidence of all-grade venous thromboembolism was 7.3% (95% CI, 4.6%-11.5%). The lowest incidence of all-grade venous thromboembolism was seen in patients with renal cancer at 3.0% (95% CI, 1.6%-5.5%). Similarly, the RR of all-grade venous thromboembolism with bevacizumab may vary with tumor type, ranging from 1.19 (95% CI, 0.92-1.55) in colorectal cancer to 3.0 (95% CI, 1.23-7.33) in renal cell cancer.

The incidences and RR of high-grade venous thromboembolism with bevacizumab also vary among different tumors (Table 3). Among patients with colorectal cancer, the summary incidence of high-grade venous thromboembolism was 7.3% (95% CI, 5.0%-10.5%). For patients with NSCLC, the summary incidence of high-grade venous thromboembolism was 6.5% (95% CI, 5.3%-8.1%); for patients with breast cancer, the summary incidence of high-grade venous thromboembolism was 3.9% (95% CI, 2.5%-5.9%). The lowest incidence of high-grade venous thromboembolism was seen in patients with renal cancer at 2.0% (95% CI, 0.9%-4.2%). Similarly, the RR of high-grade venous thromboembolism with bevacizumab may vary with tumor type, ranging from 1.00 (95% CI, 0.58-1.72) in pancreatic cancer and 2.86 (95% CI, 0.62-13.24) in renal cell cancer.

Venous thromboembolism is a major complication of cancer, and one of the leading causes of death in cancer patients.³⁰ The association of venous thromboembolism with new agents presents a challenge for recognition because many RCTs may not be powered to reveal a significant relationship. Our meta-analysis of 15 RCTs has overcome this limitation of individual trials and demonstrated that bevacizumab may be associated with a significantly increased risk of venous thromboembolism (RR, 1.33 [95% CI, 1.13-1.56]; P < .001) in patients with a variety of metastatic solid tumors. The increased risk is observed not only for all-grade venous thromboembolism, but also for clinically significant high-grade venous thromboembolism (Tables 2 and 3). This finding will help physicians and patients to recognize the risk of venous thromboembolism with the administration of bevacizumab.

The risk of bevacizumab-associated venous thromboembolism may be underestimated by a previous pooled analysis of 5 RCTs, which showed no increased risk of venous thromboembolism with bevacizumab compared with a control.²² The failure to detect such an increase in venous thromboembolism risk is likely due to the limited number of trials included for the analysis. Because bevacizumab is increasingly used in the routine treatment of cancer patients and in the setting of clinical trials in combination with other agents, it is important for oncologists and primary care physicians to be aware of the increased incidence of venous thromboembolism associated with bevacizumab to monitor and treat it appropriately. It may be appropriate to add a black box warning for venous thromboembolism in the package insert of bevacizumab to raise awareness among patients as well as physicians.

The development of venous thromboembolism may result from the anti-VEGF effect of bevacizumab. Bevacizumab may expose subendothelial procoagulant phospholipids leading to thrombosis by inhibiting VEGF-induced endothelial regeneration.³¹ Also, bevacizumab may reduce the production of nitric oxide and prostacyclin, thus predisposing to thromboembolic events.³² Inhibition of VEGF also may increase the risk of thrombosis by increasing hematocrit and blood viscosity via overproduction of erythropoietin.³³ In addition, bevacizumab may increase the release of procoagulant from the tumor into the blood stream due to an enhanced cytotoxic effect. Bevacizumab may increase expression of proinflammatory cytokines causing damage and in situ thrombus formation.³⁴ Alternatively, the increased risk of venous thromboembolism may be secondary to prolonged survival with bevacizumab. However, the survival benefit has been observed so far only in patients with colorectal cancer and NSCLC.

Venous thromboembolism is an emerging complication of many angiogenesis inhibitors. Thalidomide and its derivative lenalidomide, which have antiangiogenic properties through the blockade of basic fibroblast growth factor and VEGF,³⁵ are associated with an increased risk of venous thromboembolism in cancer patients. A pooled analysis of phase 2 and 3 clinical trials by Bennett et al³⁶ showed that the incidences of venous thromboembolism were 12% among 4862 cancer patients treated with thalidomide and 8% among 1474 cancer patients treated with lenalidomide. The RR of venous thromboembolism in patients with solid tumors associated with thalidomide or lenalidomide is not clear. In a phase 2 RCT of prostate cancer reported by Dahut et al,³⁷ venous thromboembolism was found in 9 of 43 patients receiving the combination of thalidomide and docetaxel but in none of the 25 patients receiving docetaxel only. In multiple myeloma, a multivariate logistic regression showed that thalidomide significantly increased the risk of venous thromboembolism by 2.6 times (95% CI, 1.8-3.6 times; P < .001).³⁸ Our study showed that the summary incidence of all-grade venous thromboembolism was 11.9% (95% CI, 6.8%-19.9%) and the RR was 1.33 (95% CI, 1.13-1.56; P < .001) in patients with advanced solid tumors who had been treated with bevacizumab. It is evident that risks of venous thromboembolism are all substantial among these angiogenesis inhibitors. Combination of antiangiogenic agents such as bevacizumab and thalidomide to enhance antineoplastic activity are undergoing evaluation, and may have an increased risk of venous thromboembolism.

Many factors such as age, functional status, stage and histology of the malignant tumor, mobility, concurrent chemotherapy, and prothrombotic states are known to contribute to the development of venous thromboembolism in cancer patients. We also explored risk factors for venous thromboembolism associated with bevacizumab. Our study demonstrated that the incidence of venous thromboembolism with bevacizumab varies significantly among patients with different types of tumors; higher risk was associated with aerodigestive malignancy and mesothelioma and lower risk was associated with breast cancer and renal cell carcinoma. The difference in venous thromboembolism by cancer type also may reflect the extent of prior treatment and performance status in addition to its biology. The high risk of venous thromboembolism associated with aerodigestive cancer and mesothelioma suggests a need for prophylaxis in these patients when treated with bevacizumab.

In addition to tumor type, another potential risk factor for venous thromboembolism may be the dose of bevacizumab. A high dose of bevacizumab (5 mg/kg per week) was found to be associated with a significantly increased risk of venous thromboembolism with an RR of 1.31 (95% CI, 1.02-1.68; P = .04) in comparison with controls. However, for patients who were treated with a low dose of bevacizumab (2.5 mg/kg per week), their risk of venous thromboembolism also was significantly increased (RR, 1.31 [95% CI, 1.08-1.60]; P = .007). Apparently, even low-dose bevacizumab is associated with the increased risk of venous thromboembolism. A lack of dose effect in deep venous thrombosis suggests that the so called low dose of bevacizumab may be already reaching the saturation level to induce thrombosis; alternatively, the difference between the high and low doses of bevacizumab in thrombogenesis may be too small to detect.

The treatment of venous thromboembolism may be conducted according to a recent American Society of Clinical Oncology guideline developed by a panel of experts based on a comprehensive systematic review.³⁹ The American Society of Clinical Oncology's guidelines do not recommend routine prophylaxis in ambulatory cancer patients receiving chemotherapy with the exception of myeloma patients receiving thalidomide or lenalidomide because of potential bleeding and the relatively low incidence of venous thromboembolism in this setting. However, the absolute risk of venous thromboembolism in patients with colorectal cancer treated with bevacizumab may be comparable with patients with myeloma treated with thalidomide or its derivative as discussed above.

Our study has the following limitations. First, the National Cancer Institute's common toxicity criteria grading system for venous thromboembolism has its own limitations. There is potential overlap between grade 2 and 3 because the decision for treatment is subjective. It does not distinguish distal from proximal deep venous thrombosis, and accidental finding of venous thromboembolism. Also, the majority of trials included in this analysis reported venous thromboembolism events in combined grades (all grade, 2-5; or high grade, 3-5); in addition, the ability to detect venous thromboembolism may vary among institutions in which these trials were performed, and may cause bias of the reported incidence rates. Second, the incidences of venous thromboembolism showed significant heterogeneity among the included studies. This may reflect differences in sample sizes, tumor type, concomitant chemotherapies, and many other factors among these studies. Despite these differences, the RRs reported by all of these studies showed remarkable nonheterogeneity. In addition, calculation using the random-effects model for incidence estimation may be able to minimize the problem. Third, these studies were conducted at academic centers and major institutions, and patients in these studies all have adequate major organ function, so the conclusion from this study may not apply to patients in the community or with organ dysfunction. Fourth, the study may have a potential publication bias even though it was not detectable by our analysis. Finally, this is a meta-analysis at the study level, and confounding factors at the patient level cannot be properly assessed and incorporated into the analysis.

In conclusion, our study has shown that the novel antiangiogenic agent bevacizumab is associated with a significantly increased risk of venous thromboembolism in cancer patients who receive concurrent chemotherapy or cytokine therapy. The risk is increased with both high and low doses of bevacizumab. The absolute risk and RR of venous thromboembolism may vary with tumor type. It is imperative for physicians and patients to recognize the risk. In the event of venous thromboembolism, anticoagulation is indicated, and bevacizumab may be continued if benefits of the drug outweigh the risk. Future studies are needed to investigate the prevention and management of venous thromboembolism associated with bevacizumab.