Efficacy and Toxicity of Anti-VEGF Agents in Patients with Castration-Resistant Prostate Cancer : a Meta-analysis of Prospective Clinical Studies

selection (Figure 1). After reading the full texts and abstracts, only 19 studies were considered eligible for this systematic review (Figure 1). Overall, 16 prospective phase II trials and 3 phase III randomized controlled studies were included for this analysis with a total of 3841 patients treated with anti-VEGF agents. The number of patients per trial ranged from 15 to 1224. Details of patient demographics and study designs are included in Table 1. Efficacy of anti-VEGF agents in CRPC ORR, PSA response rate, PFS (or time to progression), and OS were available from 16, 16, 17 and 13 studies, respectively. Pooled ORR was 12.4% (95% CI, 7.7%19.4%, Figure 2) with a higher response rate of 26.4% (95%CI, 13.6-44.9%) for anti-VEGF-based combinations vs. 6.7% (95%CI, 3.5-12.7%) for anti-VEGF alone (p=0.004). Heterogeneity was significant (I2=94%; Table 1. Baseline Characteristics of 19 Included Trials in the Meta-Analysis. Author/year/phase Treatment arm Patients Median Median Median ORR, n PSA, n Patients for enrolled age PFS/TTP (months) OS (months) response analysis, n Steinbild S. et al/2007/II Sorafenib 400mg bid po 57 70 1.9 NR 0 2 57 Chi K.N. et al/2008/II Sorafenib 400mg bid po 28 67 1.8 12.25 0 1 28 Dahut W.L. et al/2008/II Sorafenib 400mg bid po 22 63.9 1.8 18.3 0 0 22 Aragon-Ching J.B. et al/2009/II Sorafenib 400mg bid po 24 66 3.7 18 1 0 24 Dror Michaelson M.et al/2009/II Sunitinib 50mg qd naïve chemotherapy 17 71 NR NR 0 1 34 Sunitinib 50mg qd Doc-resistant 17 65 NR NR 1 1 Safarinejad M.R. et al/2010/II Sorafenib 400mg bid po 64 69 2.9 14.6 7 13 64 Sponpavde G. et al/2010/II Sunitinib 50mg daily 36 69.5 4.52 10.2 0 4 36 Ogita S. et al/2012/II Bev 20mg/kg 15 70 2.8 NR NR 0 15 Dahut W.L. et al/2013/II Cediranib 20mg qd 59 68.9 3.7 10.1 6 59 Di Lorenzo G. et al/2008/II Bev 10mg/kg+Doc 20 66 4 9 3 11 20 Horti J./2009/II Vandetanib 100mg qd po+ Doc+ PDN 86 67 7.8 NR 3 17 43 Placebo +Doc +PDN 67 9.9 NR 8 29 43 Francini F. et al/2011/II Bev 10mg/kg+Doc 43 74 NR NR 8 27 43 Picus J. et al/2011/II Bev10mg/kg+estramustine+Doc 79 69 8 24 23 58 79 Beardsley E.K. et al/2012/II Sorafenib 400mg bid +bicalutamide 39 75 5.5 26.6 0 12 39 Vaishampayan U.N. et al/2014/II Bev 10mg/kg+satraplatin 31 67 7 11.2 2 9 31 Kelly W.K. et al/2012/III Bev 10mg/kg + Doc+ PDN 1050 68.6 9.9 22.6 259 364 524 Placebo +Doc +PDN 69.3 7.5 21.5 187 305 526 Heath E.I. et al/2013/II Cediranib 30mg qd +Doc +PDN 57 68 NR NR 19 15 29 Placebo +Doc +PDN 68 NR NR 17 9 28 Michaelson M.D. et al/2014/III Sunitinib +PDN 873 69 5.6 13.1 35 NR 584 Placebo +PDN 68 4.1 11.8 6 NR 289 Tannock I.F. et al/2013/III Aflibercept +Doc +PDN 1224 68 6.9 22.1 NR NR 612 Placebo +Doc +PDN 68 6.2 21.2 NR NR 612 Abbreviations: PFS, progression-free survival; TTP, time-to-progression; ORR, objective response rate; PSA, Prostate Specific Antigen; Bev, bevacizumab; Doc, docetaxel; PDN, prednisone; NR, not reported Figure 1. Selection Process for Prospective Clinical Trials Included in the Meta-Analysis Figure 2. ORR with Anti-VEGF Agents in Patients with CRPC: (A) ORR of All Trials; (B): Pooled ORR of Included Trials P<0001) and a random effect model was adopted. Similarly, the pooled PSA response rate was 32.4% with a higher PSA response rate of 52.8% (95%CI: 40.2-65.1%, Figure 3) for anti-VEGF-based combinations vs. 7.3% (95%CI, 3.6-14.2%) for anti-VEGF alone (p<0.001). For PFS, the mean was 7.1 months, and the weighted median PFS was 6.9 months; standard deviation, standard error, and variance were 2.07, 0.34, and 4.29, respectively (Figure 4). When stratifying by treatment regimens, weighted median PFS was 5.6 months for anti-VEGF monotherapy in comparison with 6.9 months for antiVEGF doublet (p<0.001). For OS, the mean was 19.4 months, and the weighted median OS was 22.1 months; standard deviation, standard error, and variance were 4.55, 0.076, and 20.7, respectively (Figure 4). Again, weighted median OS was higher for anti-VEGF doublet in comparison with anti-VEGF monotherapy (p<0.001).


Introduction
Prostate cancer is the second most frequently diagnosed cancer and the sixth leading cause of cancer death in males, accounting for 14% (903,500) of the total new cancer cases and 6% (258,400) of the total cancer deaths in males (Jemal et al, 2011). Although nearly 80% of cases are diagnosed as localized disease which could be curative by radiotherapy or surgery, there is still a relapse rate of 30-60% (Merino et al, 2011). Currently, hormonal treatment with androgen deprivation therapy is a standard approach for the majority of patients with relapsed or metastatic disease, but in most cases the duration of response is limited to 15-20 months (Eisenberger & Walsh, 1999), and the disease will become castration-resistant prostate cancer (CRPC). Although secondary hormonal manipulations (eg, corticosteroids, estrogens, and ketoconazole) can benefit a subset of these patients, this benefit is usually short-lived and until recently there is no evidence that secondary hormonal therapies

Castration-Resistant Prostate Cancer: a Meta-analysis of Prospective Clinical Studies
Wei-Xiang Qi 1 , Shen Fu 1,2 *, Qing Zhang 1 , Xiao-Mao Guo 1,2 could improve overall survival (OS) (Ryan & Small, 2003). Eventually, we should consider treatment with chemotherapy once hormonal treatments fail. Docetaxel and prednisone is the standard first line therapy for patients requiring chemotherapy (Berthold et al, 2008;Qi et al, 2011a). However, responses are not durable and almost all will progress, the prognosis of these patients is poor, with median survival no longer than 10 months. As a result, there is a need for identification of better or alternative treatment strategies for improving current standard treatment.
Angiogenesis, the generation of new blood vessels, is known to play a central role in the progression of CRPC (Folkman, 2002;Folkman, 2006;Weidner et al, 1993). Among the many mediators of new blood vessel formation, vascular endothelial growth factor (VEGF) family of ligands plays a primary role (Borre et al, 1998b;Ferrara & Davis-Smyth, 1997;Hicklin & Ellis, 2005). Inhibition of VEGF signaling pathway has proven an effective strategy for the treatment of several common solid tumors such as colorectal (Hurwitz et al, 2004;Van Cutsem et al, 2012;Grothey et al, 2013), liver (Abou-Alfa et al, 2006, ovarian (Perren et al, 2011), lung (Sandler et al, 2006;Qi et al, 2011b) and kidney cancer (Yang et al, 2003;Motzer et al, 2006;Motzer et al, 2013a;Motzer et al, 2013b). Additionally, a recent meta-analysis conducted by Wang K. (Wang et al, 2012) also demonstrates that high VEGF-expression in prostate cancer is a poor prognostic factor with statistical significance for OS (HR=2.32, 95%CI: 1.40-3.24), which suggests that VEGF might be a potential treatment target for prostate cancer patients. So far, several clinical trials have been conducted to evaluate the effectiveness and adverse effects (AEs) of anti-VEGF agents in patients with CRPC, and some reviews of novel VEGF-targeted therapies available for patients with CRPC have been also reported (Aragon-Ching & Dahut, 2009;;Merino et al, 2011;Wu et al, 2011;Loblaw et al, 2013;Mukherji et al, 2013). However, to our knowledge, no metaanalysis has been carried out to estimate the pooled efficacy and safety of anti-VEGF agents administered in CRPC patients. Therefore, the current meta-analysis attempts to analyze and combine the results of these clinical trials to increase the statistical power and improve the estimates of the effects.

Literature search for identifying related studies
We conducted an independent review of citations from PubMed between January 1, 2000, and March 31, 2014. Key words were bevacizumab, avastin, aflibercept, VEGFtrap, sorafenib, nexavar, BAY43-9006, sunitinib, sutent, SU11248, pazopanib, votrient, GW786034, vandetanib, caprelsa, ZD6474, axitinib, cediranib, tivozanib, regorafenib, cabozantinib, brivanib, ramucirumab, clinical trials and cancer. The search was limited to prospective clinical trials published in English. The search strategy also used text terms such as VEGF inhibitors, angiogenesis inhibitors and vascular endothelial growth factor receptortyrosine kinase inhibitors to identify relevant information. We also performed independent searches using Web of Science databases between January 1, 2000, and March 31, 2014, to ensure that no clinical trials were overlooked. Additionally, we searched the clinical trial registration website (http://www.ClinicalTrials.gov) to obtain information on the registered prospective trials. We also searched abstracts and virtual meeting presentations from the American Society of Clinical Oncology (http://www. asco.org/ASCO) conferences that took place between Jan 2004 and Jan 2014. Reference lists from relevant primary studies and review articles were also examined to find additional publications. Each publication was reviewed and in cases of duplicate publication only the most complete, recent, and updated report of the clinical trial was included in the meta-analysis.

Study selection
Two reviewers (W.X.Q. and S.F.) independently assess the eligibility of each article. After screening all the titles and reading the abstracts, full texts of the selected articles were reviewed to determine their eligibility for inclusion in the study. Disagreement between reviewers was resolved by consensus. Clinical trials that met the following criteria were included: (1) prospective phase II and III trials in patients with CRPC; (2) participants assigned to treatment with anti-VEGF agents (alone or in combination at any dosage or frequency); and (3) studies had to record necessary data about therapy efficacy and safety. Phase I trials were excluded because of inter study variability in drug dosing as well as the small number of patients in these trials.

Data extraction and synthesis
Two authors (W.X.Q. and S.F.) conducted the data extraction independently. It was performed according to the Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) statement (Moher et al, 1999), and any types of discrepancies were resolved by consensus. The data were extracted for each trial were: first author's name, year of publication, number of enrolled patients, type and class of anti-VEGF agents, median age, objective response rate (ORR), prostate-specific antigen (PSA) response rate (decline ≥50%), median progression free survival (PFS), and median overall survival (OS). The ORR data (our primary endpoint) and PSA response rate were pooled to produce a weighted ORR and PSA response rate. We also calculated the weighted median OS and PFS for all trials. Outcome data were also pooled for the two patient subgroups (anti-VEGF monotherapy vs. anti-VEGF-based doublets) to generate a weighted median ORR, PSA response rate, median PFS and OS. Main toxicities were evaluated accordingly.

Statistical method
Using a random-effect model, the ORR and PSA response rate to anti-VEGF therapies (all studies and anti-VEGF monotherapy vs. anti-VEGF-based studies) treatment was calculated as event rate along with a 95% confidence interval. Significant heterogeneity among the studies employed for this analysis was formally assessed by Cochran's chi-square test and the I 2 index where a p value <0.1 by chi-square test and I 2 value <0.25 indicate a low degree of heterogeneity (Zintzaras & Ioannidis, 2005). ORR, PSA response rate, PFS and OS were calculated for anti-VEGF monotherapy vs. anti-VEGF-based doublets. Median pooled PFS and OS were summarized using descriptive statistics. Statistical differences in the weighted median ORR, PSA response rate, PFS and OS for anti-VEGF monotherapy and anti-VEGF doublets were assessed by two-sided t-tests. Publication bias for ORR analysis was assessed by Begg's funnel plot (Begg & Mazumdar, 1994) and Egger's test (Egger et al, 1997). Analyses were performed using the Comprehensive Meta-Analysis (version 2.2.064) and SPSS (version 17.0) programs. GraphPad Prism (version 5.01) was used to generate the graphs (histograms) with horizontal bars for both ORR and outcome values.

Search results
Of the 648 articles retrieved, 53 were eligible after  (Figure 1). After reading the full texts and abstracts, only 19 studies were considered eligible for this systematic review (Figure 1). Overall, 16 prospective phase II trials and 3 phase III randomized controlled studies were included for this analysis with a total of 3841 patients treated with anti-VEGF agents. The number of patients per trial ranged from 15 to 1224. Details of patient demographics and study designs are included in Table 1.

Publication bias
The Begg's funnel plot showed asymmetry for ORR (p=0.22), while evaluation of publication bias showed that Egger test was significant for ORR (p=0.033). Additionally, the Begg test showed asymmetry for PSA response rate among studies (p=0.22) and Egger test showed that this was significant for PSA response rate (p<0.001). These results indicated a potential publication bias for both ORR and PSA response rate.

Discussion
Despite multiple advanced in prostate cancer therapy, treatment options for castration-resistant prostate cancer (CRPC) remains limited. Preclinical research has established a clear pathogenic role for angiogenesis in the development, maintenance, and progression of CRPC (Gustavsson et al, 2005). And two clinical studies find that VEGF detected in the plasma and urine of patients with CRPC and metastatic prostate cancer is independently associated with decreased survival rates (Duque et al, 1999;Bok et al, 2001). Also, microvessel density has been shown to be prognostically associated with survival and metastasis in multiple studies (Borre et al, 1998a;Borre et al, 1998b). Based on these results, anti-angiogenic therapy with VEGF-targeted drugs is developed as a rational investigational target in prostate cancer. Although several comprehensive and systematic reviews of the literature on novel VEGF-targeted therapies available for patients with CRPC being reported recently (Loblaw et al, 2013;Merino et al, 2011;Mukherji et al, 2013;Wu et al, 2011), this study, to our best knowledge, is the first meta-analysis conducted to estimate the pooled efficacy and safety of anti-VEGF agents in CRPC patients.
Our study includes a total of 3,841 patients from 19 prospective studies for analysis. The pooled ORR and PSA response rate is 12.4% and 32.4%, with the weighted median PFS and OS for all trials 6.9 and 22.1 months, respectively. As there is a distinct lack of homogeneity among the trials in many respects, we perform sub-group analysis according to treatment regimens. In general, our results indicate that anti-VEGF monotherapy has modest efficacy in comparison with anti-VEGF combinations. Seven out of 10 trials that reported median OS in our systematic reported are greater than those reported in anti-VEGF monotherapy trials. This is particularly remarkable for the combination of anti-VEGF agents with bicalutamide with an overall survival reaching 20 months. However, the result of this particular study needs to be taken with caution because the PFS in study is only 5.5 months (Beardsley et al, 2012), which is shorter than the combination of anti-VEGF agents with docetaxel. Interestingly, among the seven studies of anti-VEGF doublets that reported median OS, the study by Di Lorenzo G. et al (Di Lorenzo et al, 2008) has the lowest (9 months), this might be explained that bevacizumab and docetaxel doublet is used as second-line therapy for patients with docetaxel-pretreated CRPC in that study. Additionally, the pooled objective response rate (ORR) was 26.4% in 9 out of 10 anti-VEGF combinations studies that reported ORR in our systematic review and 6.7% in 8 out of 9 anti-VEGF monotherapy studies. Similarly, the pooled PSA response rate for anti-VEGF-based combinations was higher than that of anti-VEGF monotherapy (52.8% vs. 7.3%, p<0.001). Taken together, these finding are encouraging, and suggest that anti-VEGF agents in combination with some agents may prove to be a preferred synergistic treatment option for CRPC.
There are several limitations needed to be considered. First, some of the trials included are randomized controlled trials while the others were not. Hence, the evidence from some of these studies is not of the highest possible quality. Second, there are potentially important differences among included studies, including differing anti-VEGF agents, dosage and administration schedule of anti-VEGF agents, periods of study conduct and study investigators. All of these would increase the clinical heterogeneity among included trials, which also make the interpretation of a meta-analysis more problematic. Additionally, our study includes a mixed population of patients treated anti-VEGF agents-based combination therapy or anti-VEGF agents alone. Therefore, the treatment design is not the same in all arms, and it might be another source of heterogeneity. Finally, in the meta-analysis of published studies, publication bias is important because trials with positive results are more likely to be published and trials with null results tend not to be published. Our research detects publication bias using Egger test but not for Begg test.
In conclusion, this meta-analysis indicates that anti-VEGF monotherapy has a modest effect in patients with CRPC and clinical benefits gained from anti-VEGF-based doublets appear greater than anti-VEGF monotherapy. Though the use of anti-VEGF agents in the treatment of CRPC patients is considered promising by many physicians, several large phase III trials investigating bevacizumab, aflibercept and sunitinib in patients with CRPC have produced disappointing results. As prostate cancer is a clinically and molecularly heterogeneous disease, further studies are still needed to identify subgroups of patients more likely to benefit from these targeted therapies.