Main

In non-small-cell lung cancer (NSCLC), KRAS is the most frequently mutated oncogene, accounting for 26% and 11% of adenocarcinoma tumours in Western and Asian patients, respectively (Dearden et al, 2013). Although some studies have shown KRAS mutations in NSCLC are associated with poor prognosis, including a meta-analysis of 28 studies (Mascaux et al, 2005; Johnson et al, 2013), conflicting results from other studies (recently reviewed in Martin et al, 2013) mean the prognostic impact of KRAS mutations in NSCLC remains unclear. Recently, with a number of agents in clinical development targeting KRAS mutant NSCLC, such as MEK inhibitors (Martin et al, 2013), it is important to discern the prognostic and predictive role of KRAS mutations.

In the phase II Biomarker-integrated Approaches of Targeted Therapy for Lung Cancer Elimination study (BATTLE; Kim et al, 2011), median progression-free survival (PFS) for patients with KRAS G12C or G12V mutations (1.84 months) was significantly shorter (P=0.046) compared with all other KRAS mutations (3.35 months) or wild-type KRAS (1.95 months; Ihle et al, 2012). In vitro analysis of cell lines carrying KRAS G12C or G12V mutations showed significantly decreased levels of phosphorylated AKT compared with cell lines carrying other KRAS mutations (P=0.009) or wild-type KRAS (P=0.020; Ihle et al, 2012), but no differences in the levels of phosphorylated MEK or p38 were noted. Conversely, immunoblot analysis of cell lines carrying KRAS G12D showed elevated levels of phosphorylated AKT compared with cell lines carrying wild-type KRAS (Ihle et al, 2012). These observations suggest that tumours harbouring KRAS G12C or G12V mutations may have a greater dependence upon RAS/RAF/MEK/ERK signalling pathways for survival and consequently may have increased sensitivity to MEK inhibitors compared with tumours carrying other KRAS mutations, such as KRAS G12D.

We previously reported the primary and secondary outcomes of a placebo-controlled phase II trial (NCT00890825) of second-line selumetinib (AZD6244, ARRY-142886), an orally available, potent, selective, non-ATP-competitive MEK1/2 inhibitor,+docetaxel treatment in patients with KRAS mutant NSCLC (Jänne et al, 2013). Median overall survival (OS; primary endpoint) in the selumetinib group was 9.4 months and 5.2 months in the placebo group (hazard ratio (HR) 0.80, 80% confidence interval (CI): 0.56–1.14). Treatment with selumetinib+docetaxel significantly improved PFS, objective response rate (ORR) and patient-reported outcomes compared with docetaxel alone, but was associated with increased adverse events relative to docetaxel+placebo.

Here, we report the results of a retrospective analysis of the impact of different KRAS codon mutations or combinations of codon mutations on efficacy outcomes in this phase II trial.

Materials and methods

Methodology of this trial (clinicaltrials.gov NCT00890825) has been previously described (Jänne et al, 2013). In brief, this double-blind, placebo-controlled, randomised, global phase II study included patients with histologically or cytologically confirmed, locally advanced or metastatic, KRAS mutant NSCLC (stage IIIB/IV), following failure of first-line chemotherapy. Treatment consisted of intravenous docetaxel 75 mg m2 on day 1 of every 21-day cycle in combination with either selumetinib hydrogen sulphate capsules 75 mg orally twice daily or matched placebo until disease progression or unacceptable toxic effects occurred.

KRAS mutation status was confirmed before randomisation by an AstraZeneca-appointed central laboratory with TheraScreen K-RAS ARMS Mutation Kit (Qiagen (formerly DxS Ltd), Manchester, UK), or by an AstraZeneca-agreed local laboratory with agreed methods, or an accredited commercial laboratory with proprietary-accredited methods.

Tumour response was based on investigator assessment of target and non-target lesions using computed tomography or magnetic resonance imaging at baseline, week 6, week 12 and then every 12 weeks, relative to date of randomisation. Change in tumour size was measured as the sum of longest diameters for target lesions assessed as part of Response Evaluation Criteria In Solid Tumors (RECIST version 1.0, Therasse et al, 2000) analysis.

Based on the hypothesis that KRAS G12C or G12V mutations predict for better response to selumetinib+docetaxel compared with other KRAS mutations, the following KRAS mutation groups (MGs) were assessed: MG1 (KRAS G12C or G12V); MG2 (all KRAS mutations other than G12C or G12V). Analyses were also performed in patient populations carrying KRAS G12C, G12V and G12D, respectively.

A retrospective statistical analysis was undertaken on data to investigate MG associations within the endpoints of OS, PFS, change in tumour size at week 6 and ORR. OS and PFS were analysed by Cox proportional hazard modelling, while tumour size was analysed by an additive linear model. Both analyses incorporated model terms to account for World Health Organization performance status, sex, histology and smoking status. Analysis of MG effect upon response was undertaken using a model that incorporated a term for treatment:MG interactions. Significant MG associations were defined as nominal P-values of P<0.05 for the treatment:MG interaction term within the model. Analysis of specific codon effect upon response was undertaken using a model that incorporated a term for treatment:codon interactions. ORR and associated CIs were calculated and significant MG associations were assessed by contingency table analysis.

Results

Eighty-three patients with KRAS mutant tumours were randomised. Baseline characteristics, reported previously (Jänne et al, 2013), were generally well-balanced between the two treatment groups (baseline demographics by MG are summarised in Supplementary Table 1). The most common KRAS mutations were G12C (46%), G12D (22%) and G12V (11%; Supplementary Table 2). Post-progression treatments were generally well balanced between MG1 and MG2 (Supplementary Table 3).

The median OS for the selumetinib+docetaxel arm vs the placebo+docetaxel arm in MG1 and MG2 was 9.6 vs 4.4 months (HR 0.69, 80% CI: 0.44–1.09) and 8.6 vs 7.1 months (HR 0.96, 80% CI: 0.54–1.70), respectively (Figure 1A). The weak trend towards longer OS in MG1 compared with MG2 (P=0.48) may be largely driven by patients with KRAS G12C mutations (HR 0.59, 80% CI: 0.35–1.00) (Figure 2A).

Figure 1
figure 1

Survival analysis to Overall survival (A) and progression-free survival (B). Abbreviations: CI=confidence interval; HR=hazard ratio; MG=mutation group.

Figure 2
figure 2

Cox model KRAS mutation subgroup treatment effects on (A) overall survival and (B) progression-free survival. Abbreviations: CI=confidence interval; HR=hazard ratio; MG=mutation group.

The median PFS for the selumetinib+docetaxel arm vs the placebo+docetaxel arm in MG1 and MG2 was 5.7 vs 1.4 months (HR 0.48, 80% CI: 0.31–0.74) and 4.9 vs 2.6 months (HR 0.72, 80% CI: 0.44–1.16), respectively (Figure 1B). The weak trend towards longer PFS in MG1 compared with MG2 (P=0.24) may be largely driven by patients with KRAS G12V mutations (HR 0.22, 80% CI: 0.07–0.72; Figure 2B). Analysis of PFS and OS in the placebo+docetaxel arm by MG is shown in Supplementary Table 4.

The ORR by KRAS mutation type in the selumetinib+docetaxel arm showed a numerically higher rate in MG1 compared with MG2 (46% and 26%, respectively; P=0.189 (χ2-test for association); Figure 3). No confirmed responses were recorded in the placebo+docetaxel arm. The weak trend towards greater ORR in MG1 compared with MG2 may be largely driven by patients with KRAS G12V mutations (ORR: 67%) (Figure 3).

Figure 3
figure 3

Objective response rate (RECIST version 1.0; partial responses only) by KRAS mutation subgroup. Abbreviations: CI=confidence interval; MG=mutation group; RECIST=Response Evaluation Criteria In Solid Tumors.

Changes in tumour size between MG1 and MG2 were similar at week 6 (least squares mean difference of −18% and −19%, respectively (P=0.77); Supplementary Figures 1 and 2) and week 12 (least squares mean difference of −23% and −30%, respectively (P=0.72)). Further analysis at week 6 suggested tumours harbouring KRAS G12V may have had a better response: G12V (n=9) −62%; reduction across all codons (n=81) −18% (Supplementary Figure 2).

Discussion

In this retrospective analysis of KRAS mutation types from a previously reported phase II trial, the hypothesis that MG1 will have a better response to selumetinib+docetaxel than MG2 was evaluated. The ‘direction’ of the results agrees with this hypothesis for PFS, OS and ORR but not change in tumour size at week 6. The trends in PFS and ORR may be driven largely by patients with KRAS G12V mutations, whereas the trend in OS may be driven largely by patients with KRAS G12C mutations. It should be noted that these trends are weak and not statistically significant. Patients with KRAS G12V mutations may respond better in terms of change in tumour size at week 6; however, as patient numbers are small, larger clinical studies would be required to test robustly the hypothesis that specific codon status can condition response to therapy.

Although the original study was not powered to detect treatment effects in KRAS mutation subtypes, we hypothesised that different KRAS codon mutations are not biologically equivalent and may vary in the degree to which they confer a dependence upon RAS/RAF/MEK/ERK signalling. In the BATTLE trial, median PFS for patients with KRAS G12C or G12V mutations was significantly shorter than for those patients with all other KRAS mutations or wild-type KRAS (Ihle et al, 2012). This observation is consistent with the results from this study, where MG1 in the placebo+docetaxel arm had shorter PFS and OS compared with MG2. In a recent study of 677 patients with metastatic or recurrent KRAS mutant lung cancers, no difference in median OS was observed for those patients harbouring G12C/V mutations compared with others (Yu et al, 2015); however, owing to the observational nature of the Yu et al (2015) study, it is difficult to draw comparisons with outcomes from clinical studies, such as ours.

In vitro analysis from BATTLE of lung cancer cell line phosphorylation patterns suggested that tumours harbouring KRAS G12C or G12V mutations have greater dependency on RAS/RAF/MEK/ERK signalling compared with other KRAS mutations (Ihle et al, 2012). This biological difference may confer greater sensitivity to MEK inhibitors; a hypothesis supported by the results presented here, which highlight how different KRAS mutations may lead to different signal transduction cascades, resulting in altered drug sensitivity. In a recently presented retrospective analysis, patients with KRAS G12V or G12A mutation-positive advanced NSCLC had longer PFS when treated with first-line platinum chemotherapy+taxane compared with those treated with platinum chemotherapy+pemetrexed or gemcitabine, respectively (Mellema et al, 2014). Additionally, in a study of NSCLC cell lines, KRAS G12V mutations were associated with increased sensitivity to cisplatin compared with wild-type KRAS (Garassino et al, 2011).

The small number of patients analysed in this phase II trial means it is underpowered for this analysis. Additionally, analysis of MG1 does not take into account any predictive differences between KRAS G12C and G12V mutation subgroups. No trends observed in favour of our hypothesis reached statistical significance and should be regarded as interesting observations worthy of further investigation. In the absence of full genotyping data, it cannot be determined whether results were confounded by other oncogenic aberrations, such as mutations in TP53 and LKB1 (Chen et al, 2012; Sherwood et al, 2014). However, the possibility that KRAS G12C or G12V mutations have greater sensitivity to selumetinib warrants further investigation in future clinical trials. A randomised, double-blind, placebo-controlled phase III trial (SELumetinib Evaluation as Combination Therapy-1 (SELECT-1) study; NCT01933932) of selumetinib+docetaxel in second-line KRAS mutant advanced NSCLC is now recruiting patients, and will provide an opportunity to further evaluate the impact of specific KRAS mutations.