Erlotinib-based doublet targeted therapy versus erlotinib alone in previously treated advanced non-small-cell lung cancer: a meta-analysis from 24 randomized controlled trials

Background To assess the efficacy profile of erlotinib-based doublet targeted therapy compared with erlotinib monotherapy for previously treated patients with advanced NSCLC, a meta-analysis was performed. Patients and methods We rigorously searched PubMed, Embase, Cochrane and meeting proceedings. Phase II/III randomized trials reporting on the efficacy of erlotinib-doublet therapy versus single-agent therapy were selected. We estimated the HR for OS, PFS and the RR for ORR, DCR, 1-year SR. Phases of trials, targeted signaling pathways, EGFR-status and KRAS- status were included in subset analysis. Results 24 studies involving 6,196 patients were eligible. In general, the combination targeted therapy significantly improved PFS, ORR and DCR. There was also a trend showing improved OS and 1-year SR in doublets group, though it was not statistically significant. Subgroup analysis suggested PFS improvement in EGFR wild-type, KRAS mutant, KRAS wild-type populations. Moreover, patients treated with anti-angiogenesis or anti-MET targeted agent revealed a significant benefit in PFS. Conclusion In patients with advanced NSCLC, erlotinib-doublets target therapy (specially combination with anti-angiogenesis and anti-MET targeted agents) was associated with a statistically significantly longer PFS, greater ORR and DCR, but the combination did not improve OS and 1-year SR compared with erlotinib alone.


INTRODUCTION
Based on the most recent WHO estimate, lung cancer is a leading cause of cancer-related mortality with approximately 1•59 million deaths worldwide in 2012. [1] In China, lung cancer is estimated to account for 21.6% of all cancer deaths in 2015. [2] In patients with advanced non-small-cell lung cancer (NSCLC), platinum-doublet chemotherapy is standard treatment in the first-line setting; however, most patients ultimately progress and survived for less than 1 year. [3] Discovery and subsequent targeting of the epidermal growth factor receptor (EGFR) pathway has imparted clinical benefit and ushered in a new era of targeted therapeutic agents for patients with NSCLC. Several guidelines recommend EGFR tyrosine kinase inhibitors (TKIs), such as erlotinib, as an option of second-or thirdline treatments for advanced NSCLC, independent of the EGFR mutational status. [4] Nonetheless, prognosis remains poor; the median progression-free survival (PFS) for patients treated with erlotinib monotherapy, regardless of EFGR mutation status, is still only around 2.2 months after failure with platinum salts and overall survival was 6.7 months according to a placebo-controlled trial conducted by Shepherd et al. [5] Multiple signaling pathways recognized to play key roles in homeostatic processes have been identified as key drivers of oncogenesis through genetic and Review epigenetic aberrations, including ErbB receptor tyrosine kinases, anaplastic lymphoma kinase (ALK), insulin-like growth factor-1 receptor (IGF-1R), hepatocyte growth factor (HGF)-mesenchymal-epithelial transition factor (MET) axis, to name a few. [6] Given the heterogeneity of NSCLC and potential crosstalk between signaling pathways implicated in tumor growth, angiogenesis and metastasis, combining targeted agents could improve the efficacy over single-target agents,, which could also be necessary to reverse resistance to EGFR inhibitor therapy. [6][7][8] Several trials have been conducted to evaluate benefits of combining targeted agent with erlotinib compared with erlotinib alone, especially the agents targeting angiogenesis, MET, IGF-1R and ErbB3 signaling. However, the results from these trials were controversial and some were of small sample size. This meta-analysis intended to pool and analyze all relevant randomized phase II/III trials, which provided a more precise assessment of efficacy of erlotinib-doublet targeted therapy compared with monotherapy in subsequent lines after previously treated with standard chemotherapy. Predefined subgroup analysis was conducted to identify the potential appropriate patient population to benefit from such combined therapy.

Risk of bias
All the included trials reported ''randomization'' with 75% and 54% studies providing the conduction details of random sequence generation and allocation concealment, respectively. 10 RCTs were marked with ''open-label'' and the performance bias was assessed as "high risk". For other key domains, no high risk of bias was detected. Full details of the assessment are in Supplementary Table 1.

Efficacy outcomes
The median OS were 5.7 to 13.3 months in the combination arm versus 4.1 to 14 months in the control arm. Pooled HR for OS estimated from 22 studies was 0.96 (95% CI 0.91-1.03, p = 0.26; Figure 2). No significant heterogeneity was detected among the studies included for OS analysis (I 2 = 31%).
The median PFS of the doublets group and singleagent group were 1.3 to 5.4 months and 1.5 to 3.5 months, respectively. Considering significant heterogeneity among the studies (I 2 = 58%), a random effect model was employed to estimate the pooled HR for PFS. Pooled PFS of patients treated with erlotinib plus the other targeted agent was superior to those treated with erlotinib alone (HR 0.83, 95% CI 0.75-0.91, p = 0.0002; Figure 3).

DISCUSSION
EGFR inhibitors have been approved for the second-line treatment of advanced NSCLC, regardless of EGFR mutational status. [4] However, patients who initially benefit from EGFR-targeted therapy eventually develop resistance and have poor prolongation of survival. Currently, there are multiple trails combining molecular agents that target different signaling pathways, attempting to overcome drug resistance and optimize utilization of single-agent erlotinib.
Our meta-analysis focused on erlotinib-based doublets as subsequent treatment after disease progression with chemotherapy. We confirmed that combination therapy resulted in prolonged progression-free survival (PFS), better overall response rate (ORR) and disease control rate (DCR) as compared to erlotinib monotherapy, though similarities in overall survival and one-year survival rate were observed. Perhaps these results were not surprising because PFS, ORR and DCR were all tumorbased assessment end points, while OS analysis could be confounded by multiple factors such as cross-over, subsequent therapies and long post-progression survival. A recent study investigating trail-level associations between PFS, ORR and OS may supporting our viewpoint, which demonstrated a strong association between ORR and PFS, but no association existed between ORR and OS or between PFS and OS. [33] Pan et al. had performed a meta-analysis about similar subjects based on published data updated in November 2012, which concluded that erlotinib-based doublets regimen significantly improved ORR and DCR compared with single erlotinib, but 1-year SR was not significantly improved for doublets. [34] Though these results were consistent with ours, only five studies involving 2,100 patients were included in the metaanalysis, while our study included 24 RCTs involving 6,196 patients. Furthermore, besides dichotomous data (ORR, DCR, 1-year SR), our study pooled the HR of time-to-event data (OS, PFS) as well, taking into account both the event and the timing of the event, to evaluate the efficacy of doublets therapy.
Qi et al. also conducted a meta-analysis evaluating combined targeted agents versus single-agent erlotinib, updated in May 2012. [35] The author included eight studies involving 2,417 patients and the efficacy endpoints were OS (HR 0.90, 95% CI 0.82-0.99, p = 0.024), PFS (HR 0.83, 95% CI 0.72-0.97, p = 0.018) and ORR (OR 1.35, 95% CI: 1.01-1.80, p = 0.04), all of which were in favor of the doublet targeted therapy according to the author's analysis. Whereas, our pooled data showed no statistical difference existed in OS between two arms. Possible explanation for this inconsistency was that another sixteen trails were incorporated and the number of participants was approximately 2.5-fold in our metaanalysis; Besides, the discordance might be associated with a three-arm trail investigating combing R1507 (given weekly or every 3 weeks) with erlotinib. [13] The trail reported HR for survival data with 90% confidence interval (CI), which should be transformed to 95% CI for further meta-analysis. For example, the 90% CI of HR for OS in 'weekly' group were 0.58-1. 21 as reported yet it should be transformed to 95% CI, namely 0.54-1.30. Consequently, the revised pooled HR along with 95% CI for OS and PFS in the meta-analysis conducted by Qi et al. were 0.90 (95% CI 0.82-1.00, p = 0.04) and 0.82 (95%CI 0.71-0.95, p = 0.010). The revised p value (0.04) for pooled OS data, though statistically significant, was apparently larger than the author reported (0.024). that there was just a tendency to improve PFS and OS in doublets, except that PFS for patients with EGFRmutation or wild-type KRAS favored single agent. All of these subset results were not statistically significant. However, given that mutational status was rarely reported according the included trails in Qi's article, results must be interpreted with caution. Conversely, we performed similar subset analysis based on a relatively large number of patients and strict definitions of EGFR status, that is gene mutant or wild-type, IHC positive or negative and fluorescence in-situ hybridization (FISH) positive or negative. Significantly, PFS improvement in doublets in EGFR wild-type (p < 0.0001), KRAS mutant (p < 0.00001), KRAS wild-type (p = 0.03) was observed; While, PFS in EGFR-mutant patients showed a trend in favor of single-agent erlotinib (HR 1.09, 95%CI 0.63-1.88). The mechanism underlying these observations were unclear.

Subgroup analysis conducted by Qi et al. based on phases of trials, EGFR-status and KRAS status showed
MET, a transmembrane tyrosine kinase receptor, is central to the processes of cancer cell migration, invasion, proliferation, and metastasis. [36] MET amplification and/ or mutations are found in many human malignancies, including NSCLC, and predicts both resistane to EGFR TKIs and poor survival. [36][37][38] Thus, EGFR and MET may cooperate in driving tumorigenesis. Targeting angiogenesis is another promising strategy to improve survival in patients with many solid tumors, including NSCLC. [39] Cabozantinib is a small molecule inhibitor of multiple receptor tyrosine kinases, including MET and vascular endothelial growth factor receptor 2 (VEGFR2). Notably, encouraging results of a randomized phase II trial testing cabozantinib, erlotinib or the combination in patients with EGFR wild-type NSCLC were presented during ASCO Annual Meeting 2015. [32] Cabozantinib, co-targeting angiogenesis and MET signaling plus erlotinib showed statistically significant improvement in both OS and PFS compared with erlotinib alone. Indeed, this trail was the only one of all included trials demonstrating overall survival benefits from combining therapy. Interestingly, our subset analysis based on different signaling pathways, involving 2,095 patients in anti-angiogenesis arm and 2,158 patients in anti-MET arm, suggested significant PFS improvement in patients treated with combined targeted agents including anti-angiogenesis (sorafenib, bevacizumab, pazopanib, sunitinib) and anti-MET (tivantinib, onartuzumab) targeted agents.
It should be noted that our analysis was limited to the use of individual patient data. All the outcome estimates were taken from published data, which tended to overestimate treatments effects. Furthermore, 10 of the 24 included RCTs were marked with ''open-label'' and the performance bias was assessed as "high risk", which may decrease the quality of our meta-analysis.
Notably, according to NCCN Guidelines Version 2.2017, the standard of care in NSCLC now is to select patients based on their EGFR or ALK status. As for patients with EGFR mutation or ALK rearrangement, several targeted drugs are recommended as first line choose. Chemotherapy is an first option for EGFR or ALK negative patients. Therefore, RCTs studying erlotinib versus doublets targeted therapy are recommended being conducted in first-line setting. However, according to our update searching in PubMed database (February 5, 2017), there were only two articles reporting the efficacy of erlotinib compared to doublets in chemotherapynative patients (no additional studies based in second-line therapy were found). One is an open-label randomized phase II study compared the combination of erlotinib and bevacizumab versus erlotinib alone in patients with non-squamous NSCLC harboring EGFR mutations in first-line setting. [40] The addition of bevacizumab to erlotinib conferred a significant improvement in PFS. Another investigating erlotinib plus Linsitinib (an IGF-1R inhibitor) or placebo in chemotherapy-naive patients. [41] Considering the limited number of relevant studies in firstline setting, our meta-analysis which seems lagging in the contemporary management of NSCLC is actually of great referential value in assessing efficacy of erlotinib versus doublets in first-line therapy. Future clinical studies should be designed based on the actual data in our meta-analysis.
From this analysis, we conclude that erlotinib combined with additional targeted agent, especially anti-angiogenesis and anti-MET agent, could provide superior clinical benefit to patients with previously treated advanced NSCLC. The efficacy of combination therapy for particular selected populations, such as EGFR wildtype population, need further investigation. The absence of a biomarker to identify sensitive populations is a major hurdle for optimal utilization.

Protocol
This review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) Statement issued in 2009. No ethical approval and patient consent are required as all analysis were based on previous published studies.

Search strategy
A comprehensive and systematic search of the electronic databases (PubMed, Embase, and Cochrane) for studies published between inception and February 2, 2016 was conducted. Applicable terms, such as "erlotinib", "NSCLC", "combin*" were used in the literature search with the filter "randomized control trial". Relevant abstracts were searched and retrieved from American Society of Clinical Oncology (ASCO) databases.

Study eligibility
Studies investigating combining molecular targeted therapy based on erlotinib versus erlotinib plus placebo or erlotinib alone in patients with advanced NSCLC (stage IV or IIIB) were eligible for inclusion. Studies that satisfied all the following criteria were included: (i) patients with histologically or cytologically confirmed stage IIIB or stage IV NSCLC and previously treated with at least one chemotherapy; (ii) assessing efficacy (and safety) profile of erlotinib-doublet targeted therapy versus single-agent erlotinib; (iii) phase II/III randomized controlled trials; (iv) at least one of the following outcome measures was  extractable in an analyzable form: overall survival (OS), progression-free survival (PFS), objective response rate (ORR), disease control rate (DCR) or 1-year survival rate (SR).
The exclusion criteria were as follows: (i) duplicate reports; (ii) review articles; (iii) case reports; (iv) phase I and single-arm phase II trials owing to a lack of control groups; (v) ongoing studies; (vi) studies investigating targeted therapy as first-line treatment; (vii) studies not within the field of interest of this study.

Data extraction
Data extraction from eligible studies were performed independently by two reviewers and disagreements were resolved by discussion and consensus with a third reviewer. The following information was extracted: the first author, year, trial phase, interventions, targeted pathways, number of subjects, median age, the percentage of female, smoking history, histology, ECOG performance status, stage, prior chemotherapy regimens, median OS, median PFS, ORR, DCR, 1-year SR, and the hazard ratio (HR) along with 95% confidence interval (CI) for the comparison of OS or PFS of erlotinib-based doubletstreated patients with that of patients receiving erlotinib alone. If the HR and 95% CI was not directly reported in the article, an estimation from the survival curve was made using Tierney's method. [42]

Assessment of risk of bias in included studies
The methodological quality of RCTs was assessed using the risk of bias tool following the Cochrane Collaboration guidelines. Seven domains were employed for this part including random sequence generation, allocation concealment, blinding of participants, personnel or outcome assessment, incomplete outcome data, selective reporting and other sources of bias.

Statistical analysis
The pooled HR for time-to-event outcomes (OS, PFS) and pooled relative risk (RR) for dichotomous data (ORR, DCR, and 1-year SR was calculated using the Review Manager 5.3 software statistical software. Heterogeneity assessed with the inconsistency statistic (I 2 ) was interpreted as follows: I 2 = 0% indicates no heterogeneity, 0% < I 2 < 25% indicates the least heterogeneity, 25% ≤ I 2 < 50% indicates mild heterogeneity, 50% ≤ I 2 < 75% indicates moderate heterogeneity, and 75% ≤ I 2 indicates strong heterogeneity. [43] We employed a random-effects model in case of the existence of moderate or strong heterogeneity ( I 2 ≥ 50% ). Otherwise, a fixed-effects model was used. We pooled time-to event data using inverse variance method and dichotomous data with Mantel-Haenszel method. Subgroup analysis was performed according to phases of trials, targeted signaling pathways, EGFR-status and KRAS-status. p values < 0.05 were regarded as being statistically significant for all included studies. Publication bias was evaluated according to Begg's and Egger's test using the STATA 12.0 software statistical software.