Introduction

Lung cancer is the leading cause of cancer-related mortality worldwide, with non-small cell lung cancer (NSCLC) being the predominant form of the disease [1, 2]. The majority of patients are diagnosed at advanced stages for which few treatment options remain [3]. Although the effect of platinum-based doublet chemotherapy was insufficient, it was still a standard first-line treatment for advanced NSCLC in these years [4–7]. Additionally, NSCLC could be caused by the accumulation of genetic alterations, among which, the most common are Kirsten rat sarcoma viral oncogene (KRAS) mutations (22 %), epidermal growth factor receptor (EGFR) mutations (17 %), and anaplastic lymphoma kinase (ALK) rearrangement (7 %) [8]. EGFR tyrosine kinase inhibitors (EGFR-TKIs) were considered to be the first-line therapy for advanced NSCLC patients harboring EGFR exon 19 deletions and exon 21 L858R mutations [9–12]. Gene translocation involving ALK and echinoderm microtubule-associated protein-like 4 (EML4) of NSCLC revealed an extraordinary response to crizotinib [13, 14]. However, neither targeting drugs nor cytotoxic chemotherapy showed obvious curative effect in patients with KRAS mutation [15].

KRAS mutations, most frequently occur on exon 2 and exon 3, affect genes which encode protein harboring single amino acid substitutions primarily at residues G12, G13, or Q61 [16, 17]. The high prevalence of KRAS mutation made it an appealing biomarker for investigation, but its guidance for clinical medication was limited. NSCLC patients harboring KRAS mutation were associated with worse prognosis and lower response to EGFR target therapy including EGFR-TKIs and anti-EGFR monoclonal antibodies than those with wild-type tumors [18–24]. Nevertheless, the effect of KRAS mutation status on front-line chemotherapy response in advanced NSCLC was ambiguous. Several studies have found that chemo-naive patients with KRAS mutation had lower objective response rate (ORR) and/or shorter progression-free survival (PFS) following first-line treatment with conventional chemotherapy compared with wild-type patients [25, 26], although this has not been shown in all reports [27, 28].

Therefore, whether KRAS mutation status is associated with responsiveness to front-line chemotherapy in advanced NSCLC is still not clear. A comprehensive analysis of the various outcomes is warranted. Thus, we sought to perform a meta-analysis incorporating all available evidences to evaluate the clinical outcome according to the KRAS mutation status in patients with advanced NSCLC treated with front-line conventional chemotherapy.

Materials and methods

Literature search

All relevant articles were retrieved by searching PubMed, Embase, and the Central Registry of Controlled Trials of the Cochrane Library using a combination of the terms “KRAS,” “Kirsten rat sarcoma viral oncogene,” “mutation,” “Lung,” “non-small cell lung cancer,” “NSCLC,” and “chemotherapy.” An additional search through Google Scholar and a manual search through reference lists of relevant reviews were additionally performed. Two authors (ZY and WM) carried out the search independently. No restriction by language or year was set in the search.

Inclusion and exclusion criteria

Eligible studies should meet the following criteria: (i) Random control trials (RCTs), prospective studies or retrospective studies which investigate or report a subset of chemo-naïve patients with local advanced or metastatic (IIIB or IV) NSCLC using cytotoxic chemotherapy without combination of other targeting drugs (e.g., anti-EGFR monoclonal antibodies); (ii) prior neoadjuvant or adjuvant chemotherapy in patients with recurrence after surgery was permitted if it had elapsed from last administration to relapse at least 6 months; (iii) KRAS mutation analysis was performed on available tumor tissue samples instead of circulating free DNA in serum in first-line chemotherapy treatment cohort; (iv) at least one of the primary outcomes was available. Studies failed to meet the inclusion criteria will be excluded.

Outcomes measures, data extraction, and quality assessment

Primary outcomes for this meta-analysis were ORR, namely partial response (PR) plus complete response (CR), 6-month PFS rate, and 1-year PFS rate. The data collection and assessment of methodological quality followed the QUORUM and the Cochrane Collaboration guidelines (http://www.cochrane.de). The data on study type, treatment regimens, KRAS mutation features, ORR, and PFS rate were extracted by two investigators (KS and FW) independently. Figures were electronically digitized, and Kaplan–Meier curves were downloaded by appropriate software (Engauge Digitizer, ver 2.12, Mark Mitchell 2002, free software down loaded from http://sourceforge.net). Two reviewers (YY and WX) used the Jadad scale to assess the quality of included RCTs and a modified Newcastle–Ottawa scale to assess other studies. Discrepancies were discussed by all investigators to reach consensus.

Statistical analysis

In consideration of any potential heterogeneity, we conducted this meta-analysis with a random-effect model. The results were reported as pooled odds radios (ORs) with the corresponding 95 % confidence interval (CI). Subgroup and sensitivity analysis were stratified for literature type, KRAS mutation analysis method, therapeutic regimen, patient origin, and EGFR mutation status in KRAS wild-type patients. An OR >1 reflected a better ORR or PFS rate in the KRAS mutant arm. Statistical heterogeneity across studies was assessed with a forest plot and the inconsistency statistic (I2). Statistical significance was considered at P < 0.05. All calculations were performed using REVIEW MANAGER (version 5.0 for Windows; the Cochrane Collaboration, Oxford, UK).

Publication bias

An extensive search strategy was made to minimize the potential for publication bias. Graphical funnel plots were generated to visually assess a publication bias. The statistical methods to detect funnel plot asymmetry were the rank correlation test of Begg and Mazumdar and the regression asymmetry test of Egger [29, 30].

Results

Eligible studies

We identified 656 records according to the search strategy and focused on 17 potentially relevant studies [25–28, 31–43]. However, seven articles were not eligible for mutation detection in non-tissue sample [37–39] or involving patients at early stage (I to IIIA) [40–43]. We finally included ten studies [25–28, 31–36] (two RCTs and eight retrospective studies) involving 1,677 advanced NSCLC patients who had been tested for KRAS mutations in first-line chemotherapy treatment cohort. Figure 1 summarizes the flow chart. Among these studies, chemotherapy regimens were platinum-based doublets at standard dose, namely cisplatin/carboplatin plus one of the third-generation agents (including gemcitabine, paclitaxel, docetaxel, vinorelbine, and pemetrexed), or some non-platinum-based regimens. Regimens were not specific in four retrospective studies [25, 32, 35, 36] so that they were excluded in subgroup analysis stratified for therapeutic regimen. Detecting approaches for KRAS mutation included direct sequencing, nested polymerase chain reaction (PCR), real-time PCR (RT–PCR), which were also a subgrouping factor. Additionally, we considered time to progression (TTP) as PFS in studies by Eberhard [31] and Lee [25]. Table 1 summarizes the characteristics of all involved studies.

Fig. 1
figure 1

Profile summarizing the trial flow

Full size image
Table 1 KRAS mutation as a predictive marker of first-line chemotherapy in advanced NSCLC
Full size table

Objective response rate

According to all the literature with available data on 1,677 patients, KRAS mutants had lower ORR than wild-type patients (25.1 vs. 34.4 %) significantly (OR 0.67, 95 % CI, 0.50–0.88; P = 0.004; heterogeneity: χ 2 = 9.12, P = 0.43, I 2 = 1 %; Fig. 2a). Subgroup analysis showed that data from retrospective studies (OR 0.67, 95 % CI, 0.49–0.92; P = 0.01; heterogeneity: χ 2 = 7.73, P = 0.36, I 2 = 9 %) and those using PCR methods in mutation detecting (OR 0.58, 95 % CI, 0.37–0.89; P = 0.01; heterogeneity: χ 2 = 0.86, P = 0.65, I 2 = 0 %) had similar tendency of significantly inferior ORR in KRAS mutants, while no significant difference in terms of ORR between the mutant group and wild-type group if stratified by therapeutic regimen (pemetrexed-based vs. gemcitabine-based vs. taxane-based vs. vinorelbine-based regimens and carboplatin-based vs. cisplatin-based regimens), patient origin (Asia vs. Non-Asia area), and EGFR mutation status in KRAS wild-type patients (EGFR mutation negative vs. not available for EGFR mutation) (Table 2).

Fig. 2
figure 2

Meta-analysis on following outcomes among advanced NSCLC patients receiving first-line chemotherapy according to KRAS mutation status (a ORR; b 6-month PFS rate; c 1-year PFS rate) CI confidence interval; NSCLC non-small cell lung cancer; ORR objective response rate; PFS progression-free survival

Full size image
Table 2 Subgroup analysis on objective response rate among advanced NSCLC patients receiving first-line chemotherapy according to KRAS mutation status
Full size table

Six-month and one-year PFS rate

Data on PFS rate were available only in 891 patients enrolled in six trials. Patients with positive KRAS mutation had lower pooled 6-month (51.0 vs. 56.8 %) and 1-year (10.3 vs. 13.3 %) PFS rate than wild-type patients, but there was no significant difference between the two groups (6-month PFS rate: OR 0.75, 95 % CI, 0.54–1.04; P = 0.08; heterogeneity: χ 2 = 3.03, P = 0.70, I 2 = 0 %; Fig. 2b and 1-year PFS rate: OR 0.75, 95 % CI, 0.47–1.21; P = 0.25; heterogeneity: χ 2 = 3.98, P = 0.55, I 2 = 0 %; Fig. 2c). Subgroup analysis stratified by literature type, mutation analysis method, and patient origin consistently revealed no significant difference between the mutant group and wild-type group (Tables 3, 4).

Table 3 Subgroup analysis on 6-month PFS rate among advanced NSCLC patients receiving first-line chemotherapy according to KRAS mutation status
Full size table
Table 4 Subgroup analysis on 1-year PFS rate among advanced NSCLC patients receiving first-line chemotherapy according to KRAS mutation status
Full size table

Assessment of heterogeneity and publication bias

As described above, the statistical heterogeneity was moderate. In addition, sensitivity analysis by leaving any study out did not alter the tendency of the general results. There was no publication bias for outcome measures, with asymmetrical appearance on funnel plot analysis (Fig. 3) and all P values >0.05 in Begg’s test and Egger’s test.

Fig. 3
figure 3

Funnel plots of SE by Log OR (a ORR; b 6-month PFS rate; c 1-year PFS rate). OR odds ratios; ORR objective response rate; PFS progression-free survival; SE standard error

Full size image

Discussion

Previous pooled analysis of the prognostic and predictive effects of KRAS mutation status in early-stage resected NSCLC after cisplatin-based adjuvant chemotherapy (ACT) revealed that there was no significant difference in overall survival (OS) and disease-free survival (DFS) based on KRAS. Moreover, patients with tumors harboring codon 12 mutations seemed to derive no benefit from ACT, whereas the presence of codon 13 mutations was associated with significantly worse OS and DFS with ACT [44]. However, whether KRAS mutation could predict the efficacy of first-line chemotherapy in advanced NSCLC was controversial based on previous small-size reports. A meta-analysis that could incorporate all available results, including subgroup data from RCTs as well, was a good way to address our concerns.

In the current study, we found that ORR was significantly higher in wild-type patients than KRAS mutants after first-line chemotherapy. In addition, similar results were found for PFS but the difference did not reach significance. An interesting in vitro study showed different sensitivity for specific chemotherapy among wild-type and KRAS-overexpressing clones from the human NSCLC cell line [45]. Nevertheless, no significant difference was found in ORR if stratified for therapeutic regimen as above research [45] in subgroup analysis. Therefore, more efforts should be made to evaluate the curative effect of specific chemotherapy in NSCLC patients harboring different types of KRAS mutation as well as those in wild type.

Subgroup analysis revealed significant difference in terms of ORR between the mutant group and wild-type group in data from retrospective studies and those using PCR methods in mutation detecting. The above results admit of following interpretations. Firstly, the sample size was too small in RCT cohort to get the significant results. Secondly, the commercial real-time PCR kit could provide greater sensitivity than direct sequencing to detect KRAS mutations so that it might avoid loss of positive detection to some degree [46]. As a result, the lower efficacy of chemotherapy in KRAS-mutated chemo-naïve patients with advanced NSCLC could be demonstrated obviously in PCR cohort.

Notably, this is the first study to comprehensively answer the impact of KRAS mutation on chemotherapy in advanced NSCLC, which only focused on first-line chemotherapy without combination of other target treatments in order to minimize the crossover effects. However, there existed several limitations. First, our meta-analysis was based on subgroup data extracted from RCTs and retrospective studies, which somehow compromised the evidence level. Besides, more than half of included studies could not exclude the occurrence of EGFR mutation in KRAS wild-type cohort. Additionally, KRAS exons identified as mutant were heterogeneous among included articles, but we were unable to assess whether 2 or 3 exon alterations, as well as different codons’ mutations, had different impact on chemotherapy. Finally, we cannot differentiate the respective impact of KRAS mutation on cell cycle-specific agents (third-generation agents) and nonspecific antineoplastic agents (platinum).

Nonetheless, regardless of above limitations, this comprehensive analysis confirmed the association between KRAS mutation status and ORR in advanced NSCLC. The result led to an important hint that the response to chemotherapy in KRAS-mutated patients was worse than what we acknowledged. The combination of chemotherapy and target therapy might be superior toward KRAS mutants in ongoing clinical trials compared with chemotherapy alone. Encouragingly, a randomized phase II study reported selumetinib, a selective inhibitor of mitogen-activated protein kinase kinase (MEK), in combination with docetaxel had a significant improvement in ORR and PFS compared with docetaxel and placebo in the second-line treatment of advanced NSCLC patients with KRAS mutation [47]. Similar combination therapy in chemo-naïve KRAS mutant patients with NSCLC is warranted in further studies.

In conclusion, this meta-analysis showed that advanced NSCLC patients with KRAS mutations had significantly lower ORR and potentially lower 6-month/1-year PFS rate compared with wild-type patients after first-line chemotherapy.