A Phase II Study of Sorafenib in Patients with Platinum-Pretreated, Advanced (Stage IIIb or IV) Non–Small Cell Lung Cancer with a KRAS Mutation

The current standard of therapy for patients with metastatic (stage IV) non–small cell lung cancer (NSCLC) is platinum-based doublet chemotherapy. Therapeutic results are far from satisfactory: survival has reached a plateau at a median of 9 to 11 months in recently published phase III trials (1–3). Therefore, clinical research of new treatment strategies is warranted. One way to improve these results is to personalize treatment, among others based on tumor molecular characteristics. One common genetic aberration that occurs in NSCLC is conversion of the proto-oncogene KRAS to its activated oncogenic form through mutations in codons 12, 13, or 61 of the KRAS gene, located on chromosome 12. The KRAS gene encodes KRAS protein that regulates signaling pathways such as RALGEF/RAL PI3K/PTEN/AKT and RAF/MEK/ERK/MAPK. A single amino acid substitution, and in particular a single-nucleotide substitution, is responsible for an activating mutation and may lead to constitutional activation of the KRAS protein, which turns on the latter pathway involved in cellular growth and survival (4). Treatment options for patients who have KRAS-mutated tumors are limited as these are believed to be poor responders to cytotoxic chemotherapy and refractory to EGF receptor (EGFR)-tyrosine kinase inhibitors (TKI; ref. 5). Sorafenib, a multitargeted TKI, targets among others the Ras/Raf pathway. Sorafenib has been evaluated in unselected patients with advanced NSCLC both as a single agent and in conjuncture with platinum doublet chemotherapy as first-line treatment. The results of these studies are equivocal: while the single-agent studies showed some activity of sorafenib in all lines of treatment, the ESCAPE phase III trial failed to improve survival when sorafenib was added to paclitaxel carboplatin (6, 7).

In addition to treatment with sorafenib, we collected serum of all patients for analysis with VeriStrat proteomics using mass spectrometry. In NSCLC, this method was found to have prognostic value in patients treated with erlotinib (8).

We hypothesized that patients with NSCLC with a KRAS-mutated tumor may benefit from treatment with sorafenib, as this drug inhibits one of the components of the downstream pathway.

In 5 hospitals, patients were recruited. Patients with histologic-documented, locally advanced, or metastatic [stage IIIB or IV, tumor–node–metastasis (TNM) 6th edition] non-squamous NSCLC (9), harboring a KRAS mutation in codons 12, 13, or 61; who progressed after at least 1 prior platinum-based chemotherapy, were eligible for enrollment. Other inclusion criteria included age ≥ 18 years; Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0 to 2; life expectancy of at least 12 weeks; presence of ≥1 measurable lesion [according to Response Evaluation Criteria in Solid Tumors (RECIST) 1.0]; and adequate bone marrow, liver, and renal function.

Exclusion criteria included active serious infections; known history of HIV infection or chronic hepatitis B or C; history of cardiac disease and uncontrolled hypertension; history of bleeding diathesis; history of organ allograft; prior exposure to sorafenib; previous or concurrent cancer (except curatively treated >3 years before study entry); any anti-cancer treatment during the study (palliative radiotherapy was allowed); autologous bone marrow transplant; use of biologic response modifiers; investigational drug therapy outside of this trial; symptomatic metastatic brain or meningeal tumors (unless >1 month from radiotherapy and off steroids); and patients with seizure disorder requiring medication (steroids or anti-epileptics).

All patients provided written informed consent. This study was approved by the medical ethical committees of the participating hospitals (Dutch trial register.NL30000.029.09).

All eligible patients received sorafenib 400 mg twice daily as tablets. Sorafenib dose was reduced or delayed for toxicities attributed to the drug. All dose modifications followed redefined dose levels: Dose level 0 (standard dose), 400 mg twice daily; dose level 1: 400 mg once daily; and dose level 2, 400 mg every other day. If further dose reduction was required (greater than level 2), sorafenib was permanently interrupted. In case of grade ≥II skin toxicity, treatment was interrupted until grade 0 to 1. Interruption of treatment due to grade II or more toxicity was allowed up to 3 times; in case of a fourth event, treatment was permanently discontinued. In case of a hematologic toxicity, a dose medication was conducted at grade III, in patients with grade IV, treatment was interrupted until grade II toxicity and restarted at a lower dose. In patients experiencing grade III non-hematologic toxicity, treatment was interrupted until grade II and restarted at a lower dose. In case of grade IV toxicity, treatment was permanently interrupted.

It was allowed to re-escalate to the standard dose after resolution of the adverse event. In addition, temporally treatment interruption was permitted up to a maximum of 28 days. If adverse events were not recovered after 28 days, treatment was discontinued unless subject was deriving clinical benefit. Sorafenib was continued until unacceptable toxicity or tumor progression. All patients were followed for at least 6 months. In case of documented tumor progression, patients discontinued study medication and received further treatment as per investigator decision.

Tumor size was evaluated by computed tomography (CT) or MRI at baseline. Baseline evaluations more than 4 weeks before the beginning of treatment were not allowed. Target lesions were evaluated every 6 weeks on study by unidimensional measurement according to RECIST 1.0. In addition, responses were evaluated according to the Crabb criteria (10), in patients with tumor cavitation. In these patients, the longest diameter of a cavity was subtracted from the overall longest diameter of the target lesion.

Overall subject safety was evaluated by tabulating reports of adverse events that occur during the study. All adverse events and normal laboratory parameters were assessed according to the NCI CTC version 3.0 grading system.

KRAS mutation analysis was conducted as standard care in all hospitals only in non-squamous NSCLC. KRAS (exon 2 and 3) mutation analysis was conducted using DNA extracted from paraffin-embedded tumor tissue using routine protocols (11). Briefly, tumor tissue was manually macrodissected from serial sections guided by a hematoxylin eosin–stained tissue section on which the tumor was marked by a pathologist. DNA was isolated by proteinase K digestion followed by magnetic bead isolation procedure. Subsequently, DNA was subjected to high-resolution melting followed by cycle sequencing of PCR products displaying a suspect melting profile.

Pretreatment serum from each patient was obtained to predict outcome using a matrix-assisted laser desorption/ionization—time of flight/mass spectrometry (MALDI-TOF/MS) proteomic algorithm (VeriStrat; ref. 8). The treating physician was unaware of the outcome of the VeriStrat assay.

The study was designed as a single-arm phase II trial to explore the activity of sorafenib in patients with NSCLC with a KRAS codon 12, 13, or 61 mutation. The primary endpoint of the study was the disease control rate (DCR), defined as the rate of subjects with no progression (based on RECIST 1.0) after 6 weeks of treatment (12–14). To be able to discontinue the trial early if the treatment showed insufficient activity, a 2-stage design was implemented (Simon optimal design; p0 = 40%, p1 = 60%, α = 0.05, β = 0.20). A total of 46 patients would be entered and the treatment would be declared to have sufficient activity to deserve further attention if at least 24 patients were nonprogressive (= rejection of the hypothesis that p0 ≤ 40%). An evaluation was made in the first 16 patients entered, if 9 or more patients of these progressed within 6 weeks, the study was stopped and the treatment declared to be insufficiently active (= rejection of the hypothesis that p1 ≥ 60%; ref. 15). Taking into account a lost to follow-up of 5%, a total of 48 patients were planned for inclusion.

Secondary efficacy endpoints were overall response rate (ORR), defined as best response recorded from start of treatment until disease progression or recurrence (according to RECIST 1.0 and Crabb criteria); duration of response, defined as the time from first response to the time of documented disease progression or death; progression-free survival (PFS), defined as the time from date of randomization to date of first observed disease progression (radiological or clinical, whichever is earlier) or death due to any cause, provided death occurs before progression was documented. Patients without after baseline tumor assessment were regarded as progressive and censored at day 1 in the PFS analysis. Overall survival (OS) was determined from the date of start of treatment to the date of death irrespective of the cause of death; patients alive at time of survival analysis were censored. Kaplan–Meier plot curves were used to calculate the PFS and OS for the study population. Log-rank test was used to calculated differences in outcome predicted by VeriStrat. Pearson χ² test was used in calculation of the relation between response and patient characteristics, type of KRAS mutation, or VeriStrat results. Cox regression test was used to calculate the HR of factors that may influence outcome (e.g., performance score, stage of disease, and type of KRAS mutation). To calculate response and HR, the types of KRAS mutation were clustered in 4 groups: the 3 most frequent mutations (G12C, G12V, G12D) and others.

A total of 59 patients were included between May 1, 2010, and February 18, 2011. August 18, 2011, the database was closed, 6 months after inclusion of the last patient. Two patients never started treatment due to ineligibility (liver enzymes abnormalities and anemia) leaving a total of 57 patients. Patient characteristics are listed in Table 1. Mean age was 58.5 (SD = ±8.1) years. More female (41 patients) than male (16 patients) participated in this study. Three patients (5.3%) had a performance status of 2, 25 patients (43.9%) have been treated with 2 or more lines of chemotherapy.

The median duration of treatment was 9 weeks (range, 0–62 weeks). At the end of the study, 2 patients still received sorafenib after 28 and 32 weeks of treatment. Fifteen patients (22.8%) stopped treatment before the first tumor response assessment at 6 weeks. Ten of the 15 patients had documented progressive disease (PD). In six patients, PD was measured by an unscheduled CT scan. In 4 patients, treatment was stopped in the first 6 weeks because of clinical deterioration. Five of the 15 patients stopped treatment because of other reasons. One patient stopped after 4 weeks of treatment due to grade II hand foot reaction. An interim CT scan had shown stable disease. The other 4 patients had no documented response assessment and thus are accounted as PD but were censored from PFS analysis on day 1. In patient 1, protocol violation was noticed after 1 week of treatment. This patient had liver function disturbances at study entry but did receive sorafenib. When protocol violation was noticed, the patient was withdrawn from therapy, this patient was included in all analysis. Patient 2 decided to discontinue treatment after 12 days due to grade II anorexia and grade II diarrhea. In patient 3, treatment was stopped after 3 days due to a cerebrovascular accident (CVA), this was not accounted as a sorafenib-related serious adverse event (SAE) by the treating physician because the CVA occurred shortly after start of therapy. Patient 4 had to stop treatment after 3 weeks because of generalized grade III rash.

Twenty-eight patients received treatment after progression, of which 6 patients had palliative radiotherapy only.

After 6 weeks of treatment, there were 5 partial response (PR; 8.8%), 25 stable disease (43.8%), and 27 PD (47.4%). The DCR was 52.6%. As best response, 1 additional patient reached PR at 12 weeks. The ORR was 10.5%. Median duration of response was 32 weeks (range, 5–58 weeks), with 1 patient still responding at the time of this analysis. There was no significant relation between any of the baseline characteristics and response. In an exploratory analysis, there was no significant relation between types of KRAS mutation and response to sorafenib (P = 0.515). In the group with G12C mutation, there were 3 PR, 12 stable disease, and 16 PD; the group with G12V mutation had 5 stable disease and 3 PD. In the group with G12D mutation, there were 5 stable disease and 3 PD, and in the group with other kinds of KRAS mutation, there were 2 patients with PR, 3 with stable disease, and 5 with PD (Fig. 1).

In 13 patients, evaluation for tumor cavitation was not possible, due to pleural fluid or no second tumor assessment. In the remaining 44 patients, 8 patients (18.2%) developed a cavity in the tumor, of which 1 patient had a preexisting cavity. In 2 patients, response changed according to the Crabb criteria, resulting in 7 PR, 23 stable disease, and 27 PD after 6 weeks of treatment.

The median PFS was 2.3 months [95% confidence interval (CI), 1.6–3.0 months], 4 patients were censored and 2 patients are still nonprogressive (Fig. 2). The median OS was 5.3 months (95% CI, 3.6–7.0 months), 14 patients were still alive at last follow-up (Fig. 3). A PS of 2 suggests a poorer OS (HR, 1.8; 95% CI, 1.0–3.2; P = 0.049), although these are only 3 patients. Type of KRAS mutation had no influence on PFS (HR, 0.90; 95% CI, 0.71–1.15; P = 0.420) or OS (HR, 0.93; 95% CI, 0.71–1.22; P = 0.61).

Pretreatment samples for VeriStrat testing were available from 55 patients. Prediction of prognosis was good in 32 patients, poor in 21, and indeterminate in 2. Patients with a good prediction had a significantly superior median PFS than patients with poor prediction (2.6 months (95% CI, 1.8–3.3 months) versus 1.5 months (95% CI, 0.6–2.3 months; HR, 1.4; 95% CI, 1.0–1.9; P = 0.029; Fig. 4A). The median OS was 6.0 months (95% CI, 3.2–8.8 months) in patients with a good prediction and 2.5 months (95% CI, 2.1–2.9 months) in patients with a poor prediction, this difference was not significant (HR, 1.3; 95% CI, 0.9–1.7; P = 0.166; Fig. 4B).

The most common adverse events reported were fatigue (6.4%), hand–foot reaction (5.7%), dyspnea (5.6%), anorexia (3.7%) diarrhea (3.6%), and cough (3.6%). Ten patients developed a grade III–IV (probable) sorafenib-related adverse event. In Table 2, all sorafenib-related adverse events are listed.

Thirty-two dose modifications were observed in 21 patients. Among these, 13 patients interrupted treatment of which 8 patients restarted treatment at reduced dose.

In this phase II study, sorafenib showed a DCR of 52.6%. Hereby, the primary endpoint was met. The PFS and OS were however disappointing. Toxicity was as expected.

We hypothesized that patients with NSCLC with a KRAS-mutated tumor may benefit from targeted treatment with sorafenib. The results of the present study are comparable with other studies in unselected patients with NSCLC, part of the patients with a KRAS mutation seem to benefit. Five studies report results of patients with advanced NSCLC treated with sorafenib monotherapy in second or later lines of therapy (16–19). In a discontinuation phase II study, patients with NSCLC were randomized to treatment with sorafenib or placebo (16). Ninety-seven patients who had a stable disease after 2 months of treatment with sorafenib were randomized to continue sorafenib or to receive a placebo. Responding patients continued on sorafenib. The PFS was 2.0 months in placebo group versus 3.6 months for patients receiving sorafenib (P = 0.009). Median survival was 9.0 versus 11.9 months, respectively (P = 0.18). In a multicenter phase II study, 52 patients with advanced pretreated NSCLC received sorafenib monotherapy (17). The DCR was 59%, median PFS was 2.7 months, and median OS was 6.7 months. The BATTLE study enrolled 158 patients with NSCLC who failed previous treatment, to treatment with erlotinib, vandetanib, erlotinib + bexarotene, and sorafenib according to mutational status (18). Patients with a KRAS mutation receiving sorafenib had a DCR of 79% (11 of 14 patients) compared with 39% (9 of 23 patients) in patients with an EGFR mutation. This study suggests that patients with a KRAS mutation may benefit from treatment with sorafenib, although the trial result was not significant. In the most recent phase II study, 37 stage IV patients with NSCLC were treated with sorafenib monotherapy (19). All patients had received one prior line of therapy. The DCR was 65%, PFS was 3.4 months, and OS of 11.6 months. In patients with KRAS mutations, a DCR of 60% was found versus 71% in KRAS wild-type. Patients with an EGFR mutation had a DCR of 40% versus 69% in EGFR wild-type. No correlation between survival and KRAS or EGFR mutational status was found. Data of this study were published after completing our study. The low DCR in KRAS-mutated patients suggests that sorafenib monotherapy may not be the best treatment option in this patient group. This is supported by data of the MISSON trial, a phase III study in 703 patients with advanced NSCLC treated with sorafenib monotherapy or placebo as a third- or fourth-line therapy. Sorafenib showed no increase in OS compared with placebo (20). In a post hoc biomarker analysis, patients with an EGFR mutation and treated with sorafenib had a significant better median OS than EFGR-mutated patients treated with placebo (423 and 197 days, respectively; P = 0.002). In EGFR wild-type patients, OS in the treatment and placebo group was similar. Also, KRAS mutational status was not predictive of sorafenib efficacy (21).

Some authors suggest that differences in outcome are, in part, mediated by different types of KRAS mutation. Using data collected in the BATTLE trial, Ihle and colleagues (22) showed that G12C and G12V mutation had a poorer PFS than patients with other types of KRAS mutation. This was also observed in the 14 KRAS-mutated patients treated with sorafenib. We cannot confirm this finding, but the small sample size of our study impedes meaningful interpretation of the results. Also, differences in response between the types of KRAS mutation were suggested by Garassino and colleagues (23). G12D mutation was found to be most sensitive to sorafenib, whereas G12C mutation and G12V needed a higher dose of sorafenib to inhibit growth.

The VeriStrat serum proteomics assay was identified as prognostic biomarker in previous studies with EGFR-TKI (8, 24). Also, it was found to predict prolonged PFS (25). In our study, patients who had a prediction of good prognosis according to VeriStrat had a significant longer median PFS. Despite the large numerical difference in median OS between patients with a “good” prediction (6.0 months) and patients with a “poor” prediction (2.5 months), this difference was not significant, possibly due to the small patient group. This assay has potential as biomarker for patients treated with sorafenib, but this has to be further studied in a larger study group.

We selected sorafenib based on its inhibition of RAF, but sorafenib has multiple targets that can play a role in its activity in patients with KRAS-mutated NSCLC. Sorafenib inhibits the activity of targets in the tumor (CRAF, BRAF, c-KIT, RET, and FLT-3) and targets involved in the angiogenesis (VEGFR-2, VEGFR-3, and PDGFR-β). We observed cavitations in some patients after start of treatment, suggesting active inhibition of angiogenesis.

Sorafenib was not active in all patients of our study group. It is possible that inhibiting KRAS activity by inhibition of the RAF/MEK/ERK pathway alone is not sufficient to inhibit tumor growth in patients harboring a KRAS mutation. Apart from RAF, KRAS has multiple other downstream effectors, most importantly, RAL and phosphoinositide 3-kinase (PI3K). RAL is reported to have a pro-oncogenetic role and PI3K is important in cell proliferation (26, 27). There are suggestions that the types of mutation have different effectors. Ihle and colleagues (22) found that RAF and RAL are the main effectors in G12C mutation but RAF and PI3K are the main effectors in G12D mutation.

To improve result of treatment, combination therapy with an mTOR inhibitor has potential. Sunaga and colleagues (27) describe that the RAS/RAF pathway and PI3K/mTOR pathway have a 2-way interaction in patients with NSCLC and have the adaptability in finding alternative signal transduction cascades. PI3K actually is partially dependent on RAS signaling (28, 29). Preclinical studies have shown synergistic effects in mice with a combination of everolimus (mTOR inhibitor) and sorafenib (30). Phase I studies in renal cell and hepatocellular cancer have shown that the combination of everolimus and sorafenib was active and tolerable (31–33). Other phase I studies with everolimus and sorafenib are ongoing, among others in patients with lung cancer (34). Another promising combination is NVP-BEZ235 (PI3K/mTOR inhibitor) and an inhibitor of the RAS/RAF pathway. In preclinical studies, this combination is synergistic in KRAS-mutated mice (35, 36). NVP-BEZ235 is currently tested in phase I studies. It can be expected that combination therapy increases toxicity.

A newly discovered interesting mTOR inhibitor is metformin, which is commonly used in treatment of non–insulin-dependent diabetes mellitus. It appears that metformin inhibits mTOR directly in the cell by stimulation of AMPK (37–39). With a direct effect on mTOR and minor side effects (mainly ephemeral gastrointestinal symptoms), metformin is a candidate to combine with sorafenib. In an exploratory evaluation, using data of this study, none of 5 patients using metformin had PD after 6 weeks of treatment. In fact, 2 of the 5 patients had PR at 6 weeks, this response was significant compared with the non-using group (P = 0.01). A trial combining sorafenib and metformin will be initiated in our institutions. Further clinical studies have to explore whether concurrent inhibition of the PI3K pathway and RAS/RAF pathway is beneficial to patients with NSCLC harboring a KRAS mutation.

In a phase III study, combination of sorafenib with standard chemotherapy was not successful in terms of patient benefit (40). Combination therapy of sorafenib and erlotinib qualifies to enter a phase III clinical trial (41–43). In these studies and also in present study, only a subgroup of patients had derived enduring benefit from treatment with sorafenib. A search for biomarkers to identify this group is mandatory in further clinical studies with sorafenib.

Sorafenib could also be a potential treatment for patients with NSCLC with a KIF5B/RET gene fusion. This mutation is found to be prevalent in less than 2% of NSCLC (44). Thyroid cancer cell lines with a RET/PTC1 rearrangement are sensitive to sorafenib (45). It is possible that KIF5B-RET gene fusion in NSCLC may be druggable using sorafenib.

In conclusion, sorafenib has shown activity as second or more line treatment in patients with NSCLC harboring a KRAS mutation. Patients had no unexpected adverse events. Further study with sorafenib is warranted as single-agent or as combination therapy in selected patients with KRAS-mutated NSCLC.

No potential conflicts of interest were disclosed.

Conception and design: A.-M.C. Dingemans, H.J.M. Groen, S.A. Burgers, E.F. Smit

Development of methodology: A.-M.C. Dingemans, H.J.M. Groen, S.A. Burgers, E.F. Smit

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.-M.C. Dingemans, H.J.M. Groen, A. van Wijk, S.A. Burgers, P.W.A. Kunst, E. Thunnissen, D.A.M. Heideman, E.F. Smit

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.-M.C. Dingemans, W.W. Mellema, H.J.M. Groen, A. van Wijk, E. Thunnissen, E.F. Smit

Writing, review, and/or revision of the manuscript: A.-M.C. Dingemans, W.W. Mellema, H.J.M. Groen, S.A. Burgers, E. Thunnissen, D.A.M. Heideman, E.F. Smit

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.-M.C. Dingemans, W.W. Mellema, A. van Wijk, D.A.M. Heideman

Study supervision: A.-M.C. Dingemans, A. van Wijk, E.F. Smit

This study was supported by an unrestricted grant from Bayer and sorafenib was supplied for free by Bayer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.