The Efficacy of Bevacizumab Compared with Other Targeted Drugs for Patients with Advanced NSCLC: A Meta-Analysis from 30 Randomized Controlled Clinical Trials

Background The extent of the benefit of bevacizumab combined with chemotherapy in the treatment of advanced non-small-cell lung cancer (NSCLC) is still unclear. We performed this meta-analysis to compare the efficacy of bevacizumab with other commonly used targeted drugs for different patients with advanced NSCLC. Methods We searched PubMed, Cochrane Library, EMBASE and abstracts from the proceedings of the American Society of Clinical Oncology (ASCO), and identified 30 randomized controlled clinical trials published within 1999 to 2011 for meta-analysis. Results The outcomes of treatment efficacy included response rate, PFS and OS. Comparing bevacizumab (15 mg/kg) with chemotherapy to standard chemotherapy alone, for chemotherapy-naïve patients, the pooled OR of response rate was 2.741(95%CI: 2.046, 3.672), the pooled HR for disease progression was 0.645 (95%CI: 0.561, 0.743), and the pooled HR for death was 0.790 (95%CI: 0.674, 0.926), respectively. In addition, the adjusted HR for previously-treated patients was 0.680 (95%CI: 0.492, 0.942) comparing bevacizumab combined with chemotherapy to standard chemotherapy alone. Conclusions Bevacizumab accompanied by chemotherapy was found to significantly improve patients' response rate, progression free survival (PFS), and overall survival (OS) among chemotherapy-naïve patients compared to other targeted drugs in the treatment of non-small cell lung carcinoma (NSCLC).


Introduction
Lung cancer has become the most common cancer and the leading cause of cancer death in the world [1,2]. Non-small cell lung cancer accounts for at least 85% in all lung cancer cases [3], presenting as local advanced disease in approximately 25-30% of cases and as metastatic disease in approximately 40-50% of cases [4]. Various epidemiological studies have shown that the 5-year survival rate for patients with NSCLC is extremely low, ranging from 5% to 15% [2]. For NSCLC patients with local advanced or metastatic disease, chemotherapy, radiation and supportive treatment are the principal therapies given the fact that these patients are not able to tolerate surgical operations. However, standard first-line chemotherapy has limited efficacy for NSCLC patients, with an objective response rate about 30%, median survival time 8-9 months and 1-year survival rate 30-40% [5], all of which call for a more effective and safer therapy for lung cancer.
In general, aberrant biological pathways in tumorigenesis result in the disfunction of a protein molecule or a gene fragment, mostly at the molecular level. Accordingly, recent clinical trials have focused on targeted therapies designed to interfere with specific aberrant biological pathways as a new treatment option for NSCLC [6]. Studies, including a recent meta-analysis report, have showed that the use of chemotherapy plus Bevacizumab (at a dose of 15 mg/kg, every 3 weeks) increases two year survival rate for patients diagnosed with advanced lung cancer compared to chemotherapy alone [7,8]. The main agents that have been investigated so far in NSCLC treatment are epidermal growth factor receptor (EGFR) family (tyrosine kinase) inhibitors (gefitinib Table 1. Baseline characteristics of the thirty trials.   and erlotinib), monoclonal antibodies targeting EGFR (cetuximab), and anti-VEGF monoclonal antibody (bevacizumab). In different clinical trials, the hazard ratios for PFS and OS of bevacizumab use ranged from 0.55 to 0.85 and from 0.71 to 1.03, respectively [9][10][11][12][13][14]. In terms of gefitinib use, the ranges of hazard ratios for PFS and OS were from 0.30 to 1.09 and from 0.77 to 1.64, respectively [15][16][17], which overlapped those of bevacizumab. Similarly, controversial and inefficient results have been reported for other targeted drugs in studies with small sample size and/or different inclusion and exclusion criteria.
In this study we performed an updated meta-analysis to systematically study the efficacy of bevacizumab combined with chemotherapy for advanced NSCLC patients. Our meta-analysis is different from the previous ones in that we target to provide information for future research in comparisons between bevacizmab and other targeted drugs. Information used in the study was obtained from reported and unreported randomized controlled clinical trial studies, and targeted drugs included gefitinib, erlotinib and cetuximab. Our meta-analysis has a higher power in testing efficacy compared to previously reported individual clinical trials, and will help make evidence-based clinical decisions for the treatment of NSCLC.

Searching method
An electronic search of the PubMed database, the Cochrane Library, and the EMBASE was performed, with the keywords ((non-small-cell lung cancer) OR nsclc) AND (target* therapy). The published language was limited to English and the years were limited from 1999 to 2011. MeSH terms searching was performed in PubMed. The American Society of Clinical Oncology (ASCO) Annual Meeting abstracts were also searched from 2000 to 2011. At the same time, the reference of related systematic reviews and clinical trials were screened.

Inclusion Criteria
The relevant clinical trials were manually selected carefully based on the following criteria: (1) randomized controlled trial (RCT); (2) patients with confirmed stage IIIB, stage IV or recurrent NSCLC based on historical or cytological evidence; (3) placebo-controlled or other types of superiority trial as well as noninferiority trial; (4) Information collected including response rate, hazard ratio for progression free survival and overall survival, along with their 95% CIs or relevant data.
When searched references referred to same studies, the most recently published papers were chosen.

Efficacy indicators
Objective response rate (ORR) is defined as the proportion of complete response (CR) plus partial response (PR) among evaluable patients. Progression free survival (PFS) is defined as the duration of time from random assignment to documented disease progression or death, whichever occurs first. Overall survival (OS) is defined as the time from random assignment to death, irrespective of the cause of death. For patients with no event observed, the time to censor refers to the time to last follow-up. The treatment efficacy of targeted drug compared to alternative drugs was measured by odds ratio for response rate (OR ORR ), and

Quality assessment
The methodological quality of trials was evaluated using the Jadad scale [a 5-point scale assessing randomization (0-2 points), double-blinding (0-2 points), and follow-up (0-1 points)] [18]. The Jadad scale has a total range from 0 to 5, and clinical trials are defined as 'good' when the scale is 3-5 [18]. Two reviewers independently assessed trial quality, and disagreements were resolved by consensus.

Data extraction
Two investigators searched the publications independently using standardized data-abstraction forms. When the two investigators discovered different results, an independent expert in oncology made the final decision of study conclusions. Information collected from these publications included first author, year of publication, targeted treatment, chemotherapy regimens, number of centers, number of patients, patient characteristics, study design (blinded or not), and the outcomes. Outcomes collected from these studies included response rate, median PFS and OS, hazard ratios for PFS and OS (HR PFS or HR OS ) and their 95% confidence intervals (CIs), and adverse events. In addition, patient character- istics collected from these studies included median age, the percentage of female, percentage of stage IV patients, ECOG performance status, and whether EGFR expression as entry criteria, When HRs were not reported in collected papers, we computed HRs and its confidence intervals assuming an exponential distribution of the survival curve. In the estimation of HRs, we applied the published methodology [19] on the graphic software package Engauge to estimate the logarithm transformed HR and variance from the Kaplan-Meier curves.

Statistical analysis
Analyses were performed in intention-to-treat (ITT) population. We first tested the statistical heterogeneity between trials (meaningful differences between studies) using the chi-squared Q-test based on the fixed-effect model. The clinical trials were considered heterogeneous when the P value of the chi-squared Qtest was less than 0.10, or when I 2 was greater than 50%. When the analyses showed heterogeneity between different clinical trials, a random effect model was applied to accommodate the heterogeneity [20]. The pooled odds ratios for response rate (OR ORR ), HRs for PFS and OS (HR PFS or HR OS ) were calculated. We decided to present three primary measures to show the treatment effect from different angles because PFS and OS can better describe the efficacy of a targeted drug than response rate. In addition, it is not uncommon to detect discrepancy between a clear benefit in PFS and a vague benefit in OS for lung cancer patients [21][22][23]. Furthermore, we estimated and tested the difference of treatment effect between bevacizumab combined with chemotherapy and other targeted drugs using the meta-regression model. The crude and riskadjusted 95% confidence interval were reported when the models included/excluded patient characteristics. To demonstrate whether the progression free survival was associated with stable disease (SD) or objective response rate (ORR) to the medication, or both, we performed the additional analysis of logarithm transformed outcomes (HR PFS ) against use of bevacizumab and OR ORR , controlling for patient characteristics (median age, mean ECOG performance score) and study design (chemotherapy type for the control group). Similarly, logarithm transformed HR OS was modeled against HR PFS and bevacizumab.
In addition to the above tests, we performed imputation study to test the influence of each individual study using the leave-one-out strategy [20]. Finally, we performed the funnel plot as well as Begg's and Egger's tests to examine potential publication bias.
We performed subgroup analysis in this study based on patient treatment status using the meta-regression models. Chemotherapy-naïve patients were defined as those with no prior chemotherapy and no previous treatment with EGFR-targeted drugs or monoclonal antibodies. Previously-treated patients were defined as patients progressed or recurred after at least one previous chemotherapy regimen.
All the analyses were performed using STATA 11.0. The study was written according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [24].

Results
The flowchart of our study is shown in Figure1. From 1,329 published papers and abstract that we found, 967 were excluded from this study based on our inclusion/exclusion criteria. In addition, 309 articles were further excluded if they were already review papers or comments. Among the 53 articles that were left from the above exclusion criteria, five articles were excluded since they were duplicate reports. Finally, 15 additional articles were excluded since they did not report outcomes relevant to our study. Our final sample included 15,650 patients collected from 30 randomized clinical trials.
In chemotherapy-naïve patient, a meta-regression analysis showed that the overall lnHR PFS was negatively associated with the lnOR ORR (b = 20.251, P = 0.001; Figure 5 and Table 2). The subgroup analyses based on patient treatment status showed that the treatment of bevacizumab for previously-treatment patients was statistically different from those of other targeted drugs in terms of disease progression(P = 0.027). For HR OS , we found similar results for both chemotherapy-naïve patients and previously-treated patients (b = 0.374, P = 0.009; and b = 0.685, P = 0.020, Figure 5 and Table 2). Trials applying bevacizumab were marked in red and grey shaded areas with the confidence band for the regression line. The size of the circles represented the weight of each trial in the regression procedure.
The Begg's funnel tests were conducted to demonstrate the influence of publication bias (figure 6). The p-values were 0.301, 0.159 and 0.851, respectively.

Discussion
Our meta-analyses showed that compared to other commonly used targeted drugs, chemotherapy with bevacizumab significantly improved patients' response rate, PFS and OS. The above findings were similar to previous findings [46]. In addition, bevacizumab provided significantly higher OR ORR , lower HR PFS , and lower HR OS among chemotherapy-naïve patients, and lower HR PFS among previous treated patients. It was also found that in EGFRmutated patients, gefitinib significantly improved OR ORR and reduces HR PFS . However, in general patients with EGFR status untested, bevacizumab showed a clear benefit in OR ORR , HR PFS , as well as HR OS , compared with gefitinib. These findings were consistent with previous publications [30].
Generally, mechanism of action of anticancer drugs was causing cancer cell death or blocking cancer cell growth. Objective response rate (ORR), which refers to the proportion of CR+PR, reflects the treatment effect by causing cancer cell death. On the other hand, SD reflects the treatment effect by blocking cancer cell The Efficacy of Bevacizumab for Advanced NSCLC PLOS ONE | www.plosone.org growth. Our meta-regression models were performed to decompose the two treatment mechanisms among NSCLC patients by introducing ln(OR ORR ) together with the bevacizumab indicator into the model. In these models we identified differences between the two types of targeted drugs in the contribution of blocking cell growth by estimating the adjusted bevacizumab effect, controlling the effect on contribution of killing tumor cells (OR ORR ).
From the results (table 2), we found that in previously-treated patients, although bevacizumab was not outstanding in promoting beneficial events such as CR and PR, it surpassed other targeted drugs in maintaining the pharmacodynamic effect. This finding was consistent with the mechanism of bevacizumab which was slowing down the vessel growth instead of causing cell death. As we can see in figure 5, several trials with treatment group applying bevacizumab (marked in red) fall below the regression line, indicating that there are other factors contributing to the prolongation of PFS in spite of the elevation of ORR. The contribution of SD in PFS time is greater in the treatment group than in the control group.
We presented three primary measures (ORR, PFS and OS) to show the treatment effect of different targeted drugs. Response rate is greatly affected by the original volume of the solid tumor, average duration of administration, and the clinical stage of patients, while PFS and OS time can be greatly affected by the frequency of follow-up. These are possible reasons of having only one clinical trial (E4599) with significant overall survival benefit. Another possible reason of the negative findings in overall survival time may be the low power to detect significance due to small valid sample size. Simple meta-regression in this study showed significantly positive correlation between ln(HR OS ) and ln(HR PFS ) in both chemotherapy-naïve and previously treated patients, indicating that given a clear benefit in PFS, benefit in OS is much likely to be detected with a larger sample size (figure 5). In other words, we can eliminate the accelerated growth of tumor cells after disease progression which would result in a clear benefit in PFS but not in OS. Our finding that the crude but not the adjusted HR OS of bevacizumab was significantly lower than that of other three drugs in chemotherapy-naïve patients indicated that the advantage in chemotherapy-naïve patients was mainly attributed to the elevation of ORR and prolongation of PFS. The finding that neither crude HR OS nor adjusted HR OS of bevacizumab was significantly different from those of other targeted drugs in previously treated patients may be explained by the complex and severity of patients.
Selection of target is essential in targeted therapies; therefore whether EGFR is mutated or not is of great significance in clinical decision. However, a considerable number of patients are unable to provide adequate tissue samples for accurate genotyping in practice. Our study showed that the benefit from bevacizumab was independent of EGFR status among a relatively large number of patients especially for those of first-line treatment. Such an effect was not able to detect for patients in second-line or third-line treatment, which suggests that patients may be more likely to show better response to the anti-angiogenic drug at early stage. Based on these findings, we would recommend early use of bevacizumab.
Limitations exist in this study. First, our meta-analysis cohort is heterogeneous regarding chemotherapies of the controls, and this may lead to unreliable findings. To address this issue, we performed an imputation study with leave-one-out strategy. The imputation analysis showed that the results had only slight difference when any single trial was removed from the metaanalysis, which indicates robustness of our study. Secondly, our analysis included a number of steps to minimize the potential for publication bias, including the Begg's test and Egger's test. The symmetrical distributions presented in Funnel plot showed a small number of outliers, which may result from the limit of published language. Third, with limited data information, our study was not able to control for heterogeneity of EGFR status in testing the treatment effect of different medications. However, literature shows that bevacizumab is an anti-VEGF mAb with a high affinity for VEGF [47]; therefore the treatment effect would not differ from the EGFR status of patients. In addition, when gefitinib was used, patients with EGFR mutated were found to have better treatment effects than those with unknown EGFR status (composed of both patients with EGFR mutation and those without EGFR mutation) [15,34]. Given the fact that we found better treatment effect of bevacizumab comparing to gefitinib for patients with unkonwn EGFR status, we believe bevacizumab should show better treatment effect than gefitinib for patients without EGFR mutation.
Our study included clinical trials with only slightly different enrollment criteria and patient demographics. However patient characteristics (age, gender, ECOG performance status) were found not to be balanced between groups in a small number of trials. Such patient level difference may lead to heterogeneity in the meta-analysis. We carefully included aggregated patient characteristics into our meta regression level to control for heterogeneity in our study. Inconsistency of chemotherapies of the control group did exist in this analysis, which could not be eliminated due to the study background. Further analysis with Bayesian method might solve this problem [48].
Finally, the clinical trials collected in this study show high heterogeneity. Due to the relative small sample size, our analysis may not be considered as strong evidence of treatment effect as other meta-analysis although we controlled for patient characteristics as well as study design. A large RCT(s) or individual-patient data meta-analysis may be needed in the future to further examine the treatment difference.
In conclusion, we found from this meta-analysis study that for chemotherapy-naïve patients, the advantage of bevacizumab in HR OS is mainly due to the elevation of ORR and prolongation of PFS. In addition, compared with other targeted drugs mentioned, chemotherapy with bevacizumab significantly improved patients' response rate, PFS and OS, especially for chemotherapy-naïve patients.