Predictive and Prognostic Biomarkers for Patients Treated with Anti-EGFR Agents in Lung Cancer: A Systemic Review and Meta-Analysis

Background: Several studies have investigated predictive and prognostic biomarkers for patients treated with anti-epidermal growth factor receptor (EGFR) agents in lung cancer. However, the conclusion is controversial. Materials and Methods: A meta-analysis was conducted to evaluate the associations of mutant K-ras, PIK3CA and PTEN deficiency with the efficacy of anti-EGFR agents in lung cancer. The primary endpoint was objective response rate (ORR). The secondary endpoints were overall survival (OS) and progression-free survival (PFS). Results: A total of 61 studies were included in the final meta-analysis. The result showed that K-ras mutation was a good predictor for ORR (RR=0.42, 95%CI, 0.33-0.55, p =0.000) and an effective prognostic marker for OS (HR=1.37, 95%CI, 1.15-1.65, p =0.001) and PFS (HR=1.33, 95%CI, 1.05-1.69, p =0.019). However, PTEN deficiency or PIK3CA mutation did not show any significance predictive value for ORR (PTEN, RR=0.82, 95%CI, 0.56-1.19, p =0.286; PIK3CA, RR=1.08, 95%CI, 0.17-6.66, P=0.938). And PTEN deficiency or expression of PIK3CA did not show significance prognostic value for OS (PTEN, HR=0.88, 95%CI, 0.31-2.46,P=0.805; PIK3CA, HR=0.79, 95%CI: 0.23-2.68, P=0.706). Conclusions: Our meta-analysis showed that K-ras mutation may be an effective predictor in lung cancer patients treated with anti-EGFR agents. Whereas, the predictive and prognostic value of PTEN deficiency and PIK3CA mutation need to be further investigated. network of EGFR and the mutations are common at codon12, codon13 of exon2 (Qi


Introduction
Lung cancer is the leading cause of cancer-related death in both sexes worldwide (Siegel et al., 2013). Although the therapeutic options have made great progress for these patients, the overall outcome remains poor. Since the standard chemotherapy for non-small-cell lung cancer (NSCLC), platinum-based doublet chemotherapy, has reached an efficacy plateau, anti-epidermal growth factor receptor (EGFR) agents became a standard optional for advanced lung cancer (Dingemans et al., 2011;Guan et al., 2013). As most kinds of targeted agents, the clinicians need to find effective biomarkers to identify appreciate patients. Although EGFR mutation has been shown as a promising biomarker for lung cancer patients treated with anti-EGFR agents, there exists 20% patients with EGFR mutations do not respond to anti-EGFR agents and 10% patients without EGFR mutations response to anti-EGFR-agents (Hirsch et al., 2007;Zucali et al., 2008). To identify patients who may benefit from the treatment of anti-EGFR agents, convenient predictors urgently need to be identified.
K-ras plays a vital role in the downstream signaling agents in lung cancer patients, we performed the metaanalysis with up-to-date data.

Inclusion criteria
The inclusion criteria to obtain eligible studies: i) evaluating the K-ras mutations, loss of PTEN expression, PIK3CA mutations status and response to anti-EGFR agents in lung cancer; ii) sufficient data on ORR, OS, PFS stratified by corresponding mutation status; iii) studies with full text articles; iv)when the same author or group reported results obtained from the same patient population in several publications, only the most recent report or most informative one was included; Studies included in our analysis are assessed by two reviewers. When it came to discrepancies, they decided to include or exclude studies after joint review.

Statistical analysis
The primary endpoint was ORR and the secondary endpoints were OS and PFS. The ORR was defined as the sum of partial response (PR) and complete response (CR). We conducted heterogeneity analysis to find an appropriate calculation model. Heterogeneity assumption was assessed by the I 2 statistic. An I 2 ≥50% indicated a lack of between-study heterogeneity (JP et al., 2003) and directed the analysis to be conducted in a fixed-effects model. Meta-regression was performed to find the source of heterogeneity. Sensitivity analyses were conducted by removing one study each time (Yang et al., 2013). Potential publication bias was evaluated by Egger's funnel plots and if the funnel plot showed asymmetry, it suggested a possible publication bias. And p≤0.05 used to assess the heterogeneity suggested statistically significantly bias in two-tailed level. All the statistical tests were performed with Stata 12.0 software.
The median age ranged from 56 to 75 years, the percent of female from 17.3% to 83.8%, and the percent of nonsmoker from 4.1% to 97.3%. Most patients were received erlotinib and (or) gefitinib. In 46 studies known response criteria, the response to treatment was evaluated mostly according to RECIST or WHO criteria.

Combined multiple biomarkers on the clinical outcomes
Three articles (Endoh et al., 2006;Fidler et al., 2011;Ludovini et al., 2012) studied the predictive and prognostic values of the combination biomarkers. V. Ludovini et al. (2012) reported that patients with the combination of KRAS, PIK3CA, MET, and non-sensitizing EGFR mutations had worse OS. MJ Fidler et al (Fidler et al., 2011) investigated and found that PFS and OS were significantly shorter in patients who meet CEN7<4 copies per cell, PTEN loss, PIK3CA gain and EGFR wild-type. Additionally, high levels of PIK3CA gain and PTEN loss had strongly significantly shorter PFS and OS. Hideki Endoh et al. (2006) found that PTEN expression and PIK3CA mutation did not correlate with the response to gefitinib, but high expressions of PIK3CA and PTEN were associated with prolonged survival. Although these articles showed that the predictive and prognostic powers of the combined markers were stronger than single markers, it lack of sufficient data to conduct meta-analysis.

Discussion
In our meta-analysis, we reviewed the literature systematically on the predictive and prognostic values of K-ras mutation, PIK3CA mutation and loss of PTEN in lung cancer patients treated with anti-EGFR agents. 63 studies were brought into in the final meta-analysis.
Of the 48 studies, K-ras mutation showed significantly predictive value of anti-EGFR agents in lung cancer patients. The ORRs of mutant K-ras and wild-type K-ras suggested the former probably lack of sensitivity to anti-EGFR agents. Furthermore, we analyzed the sensitivity of EGFR-TKIs and anti-EGFR MoAbs separately. The results showed that mutant K-ras patients were insensitive to EGFR-TKIs not anti-EGFR MoAbs. But it was noteworthy that the studies referred with anti-EGFR MoAbs also used other agents three chemotherapy and one

Loss of PTEN with ORR (C), OS (D) and PFS (E) in Lung Cancer Patients Treated with anti-EGFR Agents
EGFR-TKIs. The conclusion about anti-EGFR MoAbs might be affected by chemotherapy and could not display the real curative effect on mutant K-ras patients.
Eleven studies presented data on the association between K-ras mutation with OS, and eight studies on PFS. The results suggested that K-ras mutation possessed the prognostic value when lung cancer patients used with anti-EGFR agents. It was found that mutant K-ras patients had a significantly shorter OS and shorter PFS than wild-type K-ras patients. It indicated that K-ras mutant patients had poor survival benefits than wild-type K-ras patients. According to the incompatible relation between K-ras mutation with EGFR-mutation, K-ras mutation may be a valid predictive and prognostic biomarker of anti-EGFR agents to complement EGFR mutations to select insensitive patients to anti-EGFR agents.
In the four studies, the ORRs of mutant PIK3CA and wild-type PIK3CA lung cancer patients were about 50.00%(3/6) and 74.75%(74/99), respectively. The analysis incorporated with corresponding RR indicated that the predictive value of PIK3CA mutation was insignificant. Only 1 study referred to OS of PIK3CA mutations. In that study, the relevant results statistically significantly showed that mutant PIK3CA patients had a shorter OS.
Two studies focused on the relationship between the expression of PIK3CA with OS, but the result did not show significantly prognostic value of PIK3CA expression. There was one study on PFS of PIK3CA expression which showed us low expression of PIK3CA patients had a shorter PFS.
Three studies showed that the ORRs of high and low expression of PTEN patients were about 43.48%(30/69) and 53.15%(59/111), respectively. The results indicated the predictive value of loss of PTEN was insignificant. In another way, three and two studies concentrated on OS and PFS of loss of PTEN respectively, the results both showed the insignificant prognostic value of loss of PTEN in lung cancer treated with anti-EGFR agents.
Although our meta-analysis showed PTEN and PIK3CA had no significantly association with outcomes of anti EGFR-agents, but in the studies focused on the predictive and prognostic values of the combination biomarkers, their conclusions suggested PTEN, PIK3CA, EMT and so on still had the promising predictive and prognostic values.
EGFR mutation plays a vital role in predicting the outcomes of anti-EGFR agents in lung cancer patients. It reported that patients with EGFR mutations had better response to anti-EGFR agents than wild-type EGFR patients in lung cancer (Brugger et al., 2011;Johnson et al., 2013;Kerner et al., 2013;Kim et al., 2013). However, there still exists 10% patients without EGFR mutations response to anti-EGFR agents (Hirsch et al., 2007). K-ras, PTEN and PIK3CA as important downstream signaling molecules of EGFR network were studied in predicting outcomes of anti-EGFR agents recently.
Chen Mao et al. (2010) summarized 22 studies from 2005 to 2009 to evaluate the association between K-ras mutation and resistance to the treatment of anti-EGFR agents in patients with NSCLC. They found that K-ras mutation may be a negative predictive biomarker. The report was very deep and consultative to the clinical usefulness of anti-EGFR agents. However after 2009, more large sample relevant studies emerged up. Based on this, we collected and reintegrated the relevant data systematically to conduct the analysis comprehensively. The analysis consisted of predictive and prognostic value. We perfected the conclusion about the relationship between ORR with K-ras mutation. Additionally, the associations between OS/PFS and K-ras mutant patients were further analyzed to explore the survival benefits.
Since PIK3CA and PTEN regulate phosphorylation of Akt, it is rational to hypothesize that alteration of these genes might influence the response with anti-EGFR agents (Endoh et al., 2006;Giaccone et al., 2006;Fidler et al., 2011;Ludovini et al., 2012). Recently, several articles reported that PIK3CA mutation and loss of PTEN might have the association with resistance to anti-EGFR agents in lung cancer patients but the conclusions were controversial. To the best of our knowledge, there did not have one meta-analysis focused on these two genes. We analyzed how PIK3CA expression, PIK3CA mutation and loss of PTEN affected the outcomes of anti-EGFR agents.
Our results did not show PTEN and PIK3CA had significant predictive or prognostic values, but some reasons may lead to the conclusion. It was difficult to determine the cutoff line of PTEN expression. The studies involved in PTEN had different classification to be divided into groups (Endoh et al., 2006;Fidler et al., 2011). It may resulted in discrepancies of combining and analyzing data. De Roock W et al (De Roock et al., 2010) investigated the relationship between PIK3CA mutation and the outcomes of colorectal cancer patients treated with anti-EGFR agents, and found it was exon20 exon9 showed significantly shorter PFS than did those wild-type for this exon. Zu-Yao Yang et al (Yang et al., 2013) conducted the relevant colorectal meta-analysis also showed that PIK3CA exon20 mutation was significantly associated with worse PFS, OS, and ORR of wild-type K-ras colorectal cancer patients treated with anti-EGFR agents. It may give us hints that PIK3CA exon20 and exon9 should be analyzed separately. Unfortunately, the data collected did not compare PIK3CA exon20 and exon9, but it was a research orientation to further study.
We collected the trials estimated the predictive and prognostic values of combination of biomarkers. The studies all suggested that the combination biomarkers showed stronger predictive power. Due to the types of combination biomarkers were various, it was impossible to integrate to conduct meta-analysis.
There were several limitations should be taken into consideration in our meta-analysis: the present meta-analysis was not based on individual patient data; heterogeneity existed in the trial design; there is no subgroup analysis about ethnicity, smoking status or gender since the data is not sufficient; the predictive and prognostic value of combination of the three biomarkers were not estimate. However, the combination biomarkers to predict the outcomes of anti-EGFR agents may be a enlightening idea.
In conclusion, our meta-analysis showed that K-ras mutation may be an effective predictor in lung cancer patients treated with anti-EGFR agents. Whereas, the predictive and prognostic value of PTEN deficiency and PIK3CA mutation need to be further investigated..