Risk of Thrombo-Embolic Events in Ovarian Cancer: Does Bevacizumab Tilt the Scale? A Systematic Review and Meta-Analysis

Simple Summary Thromboembolic events (TEs) are the second cause of death in cancer patients. Two forms of thromboembolic events may arise: arterial, such as ischemic stroke or myocardial infarction; and venous, such as deep vein thrombosis or pulmonary embolism. Bevacizumab is a monoclonal antibody directed against vascular endothelial-derived growth factor, and is widely used in advanced ovarian cancer. However, whether bevacizumab increases the risk of thromboembolic events in ovarian cancer is matter of debate since studies have shown conflicting results. In our systematic review and meta-analysis, we included 14 trials with bevacizumab in ovarian cancer. We found that the risk of arterial thromboembolic events more than doubled with a risk ratio of 2.45. Also the risk of venous thromboembolism increased 30% with bevacizumab treatment. Bevacizumab, therefore, can be considered an additional risk factor for selecting patients for primary prophylaxis with anticoagulants. Abstract Thromboembolic events are the second cause of death in cancer patients. In ovarian cancer, 3–10% of patients present with venous thromboembolism (VTE), but the incidence may rise to 36% along the disease course. Bevacizumab is a monoclonal antibody directed against vascular endothelial-derived growth factor, and in in vitro studies it showed a predisposition to hemostasis perturbation, including thrombosis. However, in vivo and clinical studies have shown conflicting results for its use as a treatment for ovarian cancer, so we conducted a systematic review and meta-analysis on the risk of arterial thromboembolism (ATE) and VTE in ovarian cancer patients treated with bevacizumab. The review comprised 14 trials with 6221 patients: ATE incidence was reported in 5 (4811 patients) where the absolute risk was 2.4% with bevacizumab vs. 1.1% without (RR 2.45; 95% CI 1.27–4.27, p = 0.008). VTE incidence was reported in 9 trials (5121 patients) where the absolute risk was 5.4% with bevacizumab vs. 3.7% without (RR 1.32; 95% CI 1.02–1.79, p = 0.04). Our analysis showed that the risk of arterial and venous thromboembolism increased in patients treated with bevacizumab. Thrombolic events (TEs) are probably underreported, and studies should discriminate between ATE and VTE. Bevacizumab can be considered as an additional risk factor when selecting patients for primary prophylaxis with anticoagulants.


Introduction
Thromboembolic events (TEs) frequently occur during malignancy and are the second leading cause of mortality in cancer patients [1][2][3][4]. Furthermore, TEs lead to increased

Study Selection
The studies that met the following criteria were eligible for inclusion in our metaanalysis: (1) prospective randomized controlled trials; (2) available in English; (3) patients with OC, including primary peritoneal cancer or cancer of the fallopian tube; and (4) randomization to treatment with or without bevacizumab, alone or in combination with chemotherapy, targeted therapy, or immunotherapy. Titles and abstracts that did not clearly meet the inclusion criteria were excluded. Full texts were obtained for the remaining records. In case a study had multiple reporting articles, the most recent and most relevant was used. Articles reporting subanalyses or exploratory analyses were excluded if they were not relevant to the assessed outcome parameter (i.e., arterial/venous thromboembolic risk). The results from all search engines were merged using RAYYAN ( https://rayyan.qcri.org, accessed on 1 October 2020) for further screening and selection. Screening was done by two authors (MS and NV) based on title and abstract. Full-text analysis and study selection was done by authors MS and HK. Any disagreement was resolved by consensus. When the same patient source was included in different publications (e.g., abstract on congress and full text), the most recent and relevant was used.

Data Extraction
Data were extracted by authors MS and HK, and disagreements were resolved by consensus. For all the trials we extracted the following data where possible: study characteristics (author, year of publication, journal, study phase and design); patient and tumor characteristics (study population, disease setting, tumor type/subtype, number, duration of follow-up); treatment characteristics (type of treatment, dose, duration of treatment, combination regimen); and outcome parameters (progression-free survival (PFS), OS, ATE, VTE). In case of missing or confounding data, the corresponding author was contacted.

Study Objectives and Statistical Analysis
The primary outcome was the incidence of venous and arterial thromboembolism in patients treated with bevacizumab versus without treatment.
We extracted the thrombo-embolic risk for individual studies when provided and made a separate analysis for venous and arterial thrombo-embolic risk. In case of a threearm study, we pooled the data of all bevacizumab-treated patients. Missing or confounding data were addressed to the corresponding author. Data that were not provided were left blank.
We pooled data from individual trials and calculated the risk ratio (RR) for developing a thrombo-embolic even using RevMan 5.3, provided by the Cochrane collaboration (https: //training.cochrane.org/online-learning/core-software-cochrane-reviews/revman, accessed on 9 April 2021). The summary estimates were generated using a fixed-effect model (Mantel-Haenszel method). Statistical heterogeneity was assessed with the Q-test and the I 2 statistic. I 2 values of 25, 50 and 75% were considered to indicate low, moderate, and high heterogeneity, respectively [31]. Risk ratios for VTE and ATE were calculated with 95% CIs for each study. We made a predefined subgroup analysis for treatment setting and dose of bevacizumab. For all the statistical analyses, a p < 0.05 was regarded as statistically significant, and all tests were two-sided. If >10 studies were included, a funnel plot was generated to evaluate the publication bias, and Egger's regression method was used to test the symmetry of funnel plots. Risk of bias was assessed by two reviewers (MS and HK) using the Cochrane RoB2-tool [32]. Any disagreements were resolved by consensus.

Protocol Registration
Details of the protocol for this systematic review were registered with PROSPERO (www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42020176635, accessed on 26 March 2020).

Study Selection and Characteristics
The literature search resulted in 4611 possible studies, which were screened. The selection process is illustrated in the PRISMA-flowchart in Figure 1. In all, 71 articles underwent full-text review, of which 14 were available for inclusion for a total of 6119 patients (range . Six studies were conducted in a frontline setting [16,17,[33][34][35][36], 6 trials included patients with relapsed platinum-sensitive OC [20,21,[37][38][39][40] and 2 with platinum-resistant OC [22,41]. The characteristics of all trials in the qualitative analysis are provided in Table  1. ATE and VTE incidence was reported in 5 and 9 studies, respectively, and only these were included in the quantitative analysis (see Figure 1). Since <10 studies were included in the quantitative analysis, no funnel plot was generated.
for treatment setting and dose of bevacizumab. For all the statistical analyses, a p < 0.05 was regarded as statistically significant, and all tests were two-sided. If >10 studies were included, a funnel plot was generated to evaluate the publication bias, and Egger's regression method was used to test the symmetry of funnel plots. Risk of bias was assessed by two reviewers (MS and HK) using the Cochrane RoB2-tool [32]. Any disagreements were resolved by consensus.

Protocol Registration
Details of the protocol for this systematic review were registered with PROSPERO (www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42020176635, accessed on 26 March 2020).

Study Selection and Characteristics
The literature search resulted in 4611 possible studies, which were screened. The selection process is illustrated in the PRISMA-flowchart in Figure 1. In all, 71 articles underwent full-text review, of which 14 were available for inclusion for a total of 6119 patients (range . Six studies were conducted in a frontline setting [16,17,[33][34][35][36], 6 trials included patients with relapsed platinum-sensitive OC [20,21,[37][38][39][40] and 2 with platinum-resistant OC [22,41]. The characteristics of all trials in the qualitative analysis are provided in Table 1. ATE and VTE incidence was reported in 5 and 9 studies, respectively, and only these were included in the quantitative analysis (see Figure 1). Since <10 studies were included in the quantitative analysis, no funnel plot was generated.

Arterial Thromboembolic Events
Five studies, with a total of 4811 patients (2716 in the bevacizumab group and 2095 in the control group) reported on the incidence of arterial thromboembolism (see Figure 2). With a total number of 67 events in the bevacizumab group and 23 in the control group, we calculated that the absolute risk of arterial thromboembolic events was, respectively, 2.4 and 1.1%. Pooled analysis revealed that treatment with bevacizumab significantly increased the risk of ATE (RR 2.45; 95% CI 1.27-4.27, p = 0.008). Heterogeneity between studies was moderate (I 2 = 35%). Based on subgroup analysis, this risk was present in all treatment settings: either frontline treatment, relapsed platinum-sensitive or platinumresistant (Supplementary Figure S2).

Arterial Thromboembolic Events
Five studies, with a total of 4811 patients (2716 in the bevacizumab group and 2095 in the control group) reported on the incidence of arterial thromboembolism (see Figure  2). With a total number of 67 events in the bevacizumab group and 23 in the control group, we calculated that the absolute risk of arterial thromboembolic events was, respectively, 2.4 and 1.1%. Pooled analysis revealed that treatment with bevacizumab significantly increased the risk of ATE (RR 2.45; 95% CI 1.27-4.27, p = 0.008). Heterogeneity between studies was moderate (I 2 = 35%). Based on subgroup analysis, this risk was present in all treatment settings: either frontline treatment, relapsed platinum-sensitive or platinumresistant (Supplementary Figure S2).

Venous Thromboembolic Events
Nine studies, with a total of 5121 patients (2882 in the Bevacizumab arm, and 2239 in the control arm), reported on the incidence of venous thromboembolism. With a total number of 155 events in the bevacizumab group and 83 in the control group, the absolute risk of development of VTE was 5.4 and 3.7%, respectively.
Pooled analysis revealed that treatment with bevacizumab increased the risk of venous thromboembolism (RR 1.32, 95% CI 1.02-1.79, p = 0.04) compared to no bevacizumab treatment (see Figure 3). Subgroup analysis according to treatment setting showed that the increased risk was present in the frontline setting as well as in relapsed platinum-sensitive OC (See Supplementary Figure S3).

Venous Thromboembolic Events
Nine studies, with a total of 5121 patients (2882 in the Bevacizumab arm, and 2239 in the control arm), reported on the incidence of venous thromboembolism. With a total number of 155 events in the bevacizumab group and 83 in the control group, the absolute risk of development of VTE was 5.4 and 3.7%, respectively.
Pooled analysis revealed that treatment with bevacizumab increased the risk of venous thromboembolism (RR 1.32, 95% CI 1.02-1.79, p = 0.04) compared to no bevacizumab treatment (see Figure 3). Subgroup analysis according to treatment setting showed that the increased risk was present in the frontline setting as well as in relapsed platinum-sensitive OC (See Supplementary Figure S3).

Risk of Bias
The risk of bias for 4 studies was low according to the Cochrane Risk of Bias tool [16,17,20,35]. There was some concern about 6 others because details on the randomization process and allocation concealment were lacking [22,33,34] or the outcome assessment was non-blinded [21,37,38,40,41]. Two studies [36,39] were considered high risk because of a lack of detail on randomization, outcome assessment and outcome reporting. The details are summarized in Table 2.

Risk of Bias
The risk of bias for 4 studies was low according to the Cochrane Risk of Bias tool [16,17,20,35]. There was some concern about 6 others because details on the randomization process and allocation concealment were lacking [22,33,34] or the outcome assessment was non-blinded [21,37,38,40,41]. Two studies [36,39] were considered high risk because of a lack of detail on randomization, outcome assessment and outcome reporting. The details are summarized in Table 2.

Discussion
To our knowledge, our study is the first meta-analysis that specifically addressed the arterial and venous thromboembolic risk of bevacizumab in ovarian cancer in various treatment settings. Based on our analysis, patients treated with bevacizumab had an increased risk of ATE and VTE, with a relative risk ratio of 2.45 (95% CI, 1.27-4.72) and 1.32 (95% CI, 1.02-1.72), respectively. However, we also found that the absolute risk of ATE and VTE was lower than expected in both groups as the incidence of ATE and VTE was, respectively, 2.4 and 5.4% in the bevacizumab group and 1.1 and 3.7% in the control group.
Our study is the largest meta-analysis to address the thrombogenic risk of bevacizumab in ovarian cancer. We selected 14 studies with a total of 6119 patients and made a subgroup analysis according to disease setting. This was particularly interesting as the incidence of TEs may vary among tumor types and disease setting. However, we used pooled data from the trials as we did not have access to individual patient data. We could not account for possible confounding factors according to disease and patient characteristics (e.g., prolonged immobilization or poor performance status), surgical factors (e.g., surgical effort), or prophylactic use of LMWH. Furthermore, we were confronted with scarce and heterogeneous data because ATE and VTE incidence was only reported in 5 and 9 studies, respectively, of the 14 studies overall. We decided to select all prospective trials for qualitative review, including those not reporting ATEs and VTEs, to minimize the risk of selective outcome reporting and to illustrate the fact of inadequate TE reporting: five of the studies did not report any incidence of ATE or VTE (see Table 1). Of the reporting trials, five described only an advanced (grade 3 or higher) VTE [20][21][22]34,35] and four others reported all VTEs [16,17,33,38]. Moreover, in the GOG-0213 trial, there were two cases of pulmonary embolism that led to treatment discontinuation, yet these were not reported in the adverse events table as VTE [21]. In studies with bevacizumab, adequate reporting of TE complications is mandatory, distincting ATEs from VTEs.
This retrospective series showed thromboembolic events in 12-36% of ovarian cancers [6,7,42]. In the single-arm OSCAR trial, which evaluated the use of front-line bevacizumab in advanced ovarian cancer, the incidence of TE was 9% [43]. With an absolute risk of 5.4% (with bevacizumab) and 3.7% (without bevacizumab), the incidence of VTE was lower than expected in our study. This may be related to underreporting, as illustrated above, or selection bias as patients with prior VTE or hypercoagulability might have been excluded from the cancer trials. However, these adverse events should not be neglected, as they might be life-threatening. The increased TE risk, as documented in our analysis, calls for clinical vigilance for thromboembolic complications and the collection of real-world data, to further determine TE incidence and risk factors.
Previous meta-analyses showed conflicting results for the risk of TE with bevacizumab (see Table 3). In 2007, a meta-analysis including 1745 patients with advanced breast, lung and colorectal cancer showed an increased risk of ATE yet no increased risk of VTE [44]. These results were contested by Nalluri et al., who performed a larger meta-analysis in 7956 patients with advanced breast, lung, colorectal, renal or pancreatic cancer and found an increased risk of low-grade and high-grade VTE independent of the bevacizumab dose [45]. However, this study was criticized as it included studies that did not distinguish between venous and arterial events [46][47][48]. Two other meta-analyses with mixed cancer types confirmed the increased risk of ATE with bevacizumab with a RR of 1.44 (95% CI 1.08-1.91) and 1.46 (95% CI, 1.11-1.93), respectively, but these studies did not investigate the risk for VTE [49,50]. Another meta-analysis in breast cancer found no increased risk for VTE (RR 1.02, 95% CI 0.70-1.61) or ATE (RR 1.49; 95% CI, 0. 70-3.19) [51]. The largest metaanalysis, which included 20,500 patients with multiple cancer types found an increased risk for ATE and VTE, but no subgroup analysis for cancer type was made [52]. In advanced lung cancer, a Chinese meta-analysis found an increased risk for all TEs (RR 1.74; 95% CI, 1.15-2.62) but did not make a subanalysis for VTE or ATE [53]. The increased risk was driven by the high-dose group (15 mg/m 2 Q3w), whereas the result in the low-dose group (7.5 mg/m 2 ) was not significant [53]. In ovarian cancer, six meta-analyses investigated the benefit and harm of angiogenesis inhibitors, including thromboembolic events (see Table 4). Five found an increased risk of ATE with bevacizumab with a RR ranging from 2.27 to 4.84 [55][56][57][58][59]. However, other angiogenesis inhibitors besides bevacizumab were evaluated. Another meta-analysis found an increased risk of all TEs but did not distinguish between ATE and VTE [60]. Our study is in line with the largest meta-analysis regarding cardiovascular adverse events with bevacizumab by Totzeck et al. [52], and the study by Wu et al. [59], further establishing the increased risk of VTE and ATE in ovarian cancer patients treated with bevacizumab. Based on these data, bevacizumab treatment can be considered a risk factor for VTE and ATE development.

Implications for Clinical Practice
The burden of thromboembolic complications in ovarian cancer is high, and there is a clinical need to investigate relevant risk factors for TEs to define possible prevention strategies for TE development in OC.
The Khorana risk score (KRS) was the first risk assessment model to identify ambulatory patients at risk for VTE, with a score of >3 as high risk [61]. Several other models have been designed to improve the VTE risk discrimination capacity [62][63][64][65], yet they have to be validated before routine introduction [66]. To date, the KRS remains the most validated for risk assessment and guidance for TE prophylaxis, either with low-molecular-weight heparin (LMWH) [67][68][69] or direct oral anticoagulants (DOACs) [70,71]. There are no direct comparisons between LMWH and a DOAC as a preventive strategy; however, DOACs are considered more convenient because they do not require daily injections of LMWH. In a recent meta-analysis, DOACs had lower 6-month recurrent VTE compared to LMWH (RR 0.65, 95% CI 0.42-1.01), at the price of increased major bleeding (RR 1.74; 95% CI 1.05-2.88) and non-major bleeding (RR 2.31; 95% CI 0.85-6.28) for patients receiving DOACs. There was no difference in mortality (RR 1.03; 95%CI 0.85-1.26) [72].
Current guidelines recommend primary prophylaxis for hospitalized patients and prolonged prophylaxis after surgery. In ambulatory settings, prophylaxis is recommended for high-risk patients (Khorana risk score > 2) [73][74][75][76] although studies with LMWH used a KRS of >3 as this resulted in a larger reduction in VTE incidence [77] and yielded greater costeffectiveness [78]. Primary prophylaxis could be considered for newly diagnosed advanced OC patients with BMI > 35 kg/m 2 , hemoglobin < 10 g/dL, leukocytosis > 11 × 10 6 /µL or thrombocytosis > 350,000/µL. When weighing risks and benefits, treatment with bevacizumab can be considered as an additional risk factor besides others such as ascites, clear cell histology or a prior history of VTE [79].

Conclusions
In ovarian cancer patients, an increased risk of ATE and VTE was observed in treatment with bevacizumab. The incidence of TEs was probably underreported, and arterial and venous TE should be described separately. Caution should be made when initiating bevacizumab in patients at risk for TE. Primary prophylaxis of VTE with LMWH or DOACs, based on the Khorana risk score, may reduce the TE burden in OC. When selecting patients for primary prophylaxis, treatment with bevacizumab should be considered as an additional risk factor for VTE development.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/cancers13184603/s1, Figure S1: detailed search terms and syntax, Figure S2: Forrest plots of arterial thromboembolism per disease setting, Figure S3: Forrest plots of venous thromboembolism risk per disease setting.