|Home | About | Journals | Submit | Contact Us | Français|
Determine whether EGFR and KRAS mutations carry prognostic significance in non-small cell lung cancer (NSCLC) patients with brain metastases treated with stereotactic radiosurgery.
Ninety-four NSCLC patients with brain metastases initially treated with stereotactic radiosurgery were retrospectively reviewed. Both EGFR and KRAS mutation status were recorded in 67 patients: EGFR+/KRAS- status in 9 patients, EGFR-/KRAS+ in 15 patients, and EGFR-/KRAS- in 43 patients. Survival was determined using the Kaplan-Meier method. Cox regression was used to assess the effects of patient factors on overall survival, local control, and distant brain control – all from time of brain metastasis diagnosis.
Median overall survival from time of brain metastasis diagnosis was 30.6 months for EGFR+/KRAS- patients, 9.8 months for EGFR-/KRAS+ patients, and 19.1 months for EGFR-/KRAS- patients (p=0.094). Local control at 2 years was 100% for EGFR+/KRAS- patients, 66.7% for EGFR-/KRAS+ patients, and 97.2% for EGFR-/KRAS- patients (p=0.399). Distant brain control at 12 months was achieved in 66.7% of EGFR+/KRAS- patients, 30.0% of EGFR-/KRAS+ patients, and 73.7% of EGFR-/KRAS- patients (p=0.039). On multivariate analysis, the most important predictors of mortality were baseline DS-GPA>2 (HR=0.27; p=0.001), EGFR mutation positivity (HR=0.30; p=0.054), and KRAS mutation positivity (HR=2.12; p=0.056); the most important predictors of distant brain failure were KRAS status (HR=4.44; p=0.004) and extracranial disease (HR=3.28; p=0.058); there was no statistically significant multivariate model identified for local control.
In NSCLC patients with brain metastases, KRAS mutations portend higher rates of distant brain failure. Our data also suggests that EGFR portends better overall survival and KRAS portends worse overall survival, though this still needs to be verified by a larger study.
Cellular markers hold great promise for improving diagnostic and prognostic accuracy in cancer care. They may also be predictive of treatment response to targeted agents. In non-small cell lung cancer (NSCLC), epidermal growth factor receptor (EGFR) and v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations have emerged as important events in cancer cell biology.
EGFR is a tyrosine kinase receptor that activates cellular transcription via BRAF/MEK/ERK proteins, leading to increased proliferation, metastasis, and anti-apoptosis.1 EGFR mutations in NSCLC cell lines have been shown to predict higher tumor response to EGFR tyrosine kinase inhibitors (TKIs) erlotinib and gefitnib.2,3,4 Although progression-free survival is improved by these TKIs, EGFR mutations portend better survival in patients, regardless of TKI use.5 EGFR mutations are more commonly found in women, non-smokers, Asian patients, and with adenocarcinoma NSCLC subtypes.6,7
KRAS, a signaling protein (GTPase) downstream of EGFR, acts on the same cascade affecting cellular proliferation, metastasis, and anti-apoptosis. KRAS mutations, normally occurring in codons 12 and 138, are not affected by EGFR tyrosine kinase inhibitors erlotinib and gefitinib.9,10 Yet, the effect of KRAS on overall survival of NSCLC patients is not clear.8 Some studies have shown KRAS mutations to be negative prognostic factors in certain subsets of patients (codon 12 point mutation11, stage I NSCLC12, advanced NSCLC13, and adenocarcinoma14, while other studies have failed to show statistical significance in outcomes.15,16 KRAS mutations, as opposed to EGFR mutations, are much more common in smokers and those from western countries.7,17
While the predictive value of EGFR mutation status to TKI therapy and the negative prognostic value of KRAS mutation status have been confirmed18, few papers have clarified the specific value of these mutations in NSCLC patients at the time of brain metastasis diagnosis. Two retrospective studies have shown the positive effect of EGFR mutation status in NSCLC patients with brain metastases on survival (study by Eichler et al. in which 83% of patients received WBRT either alone or with craniotomy or SRS)19 and on clinical response to WBRT (study by Gow et al. in which all patients received WBRT).20 Johung et al. have shown superior local control after gamma knife treatment in patients with tyrosine kinase-activated tumors compared with patients with no detected mutation or a KRAS mutation.21 We sought to examine a large, institutional database to retrospectively determine the association of EGFR and KRAS mutations with outcomes, including local brain failure, distant brain failure, and overall survival, in NSCLC patients with brain metastases.
After receiving approval from our institutional review board, we retrospectively identified all NSCLC patients with brain metastases that had been treated at our institution with Gamma Knife stereotactic radiosurgery (SRS) from June 2009 to March 2010. Gamma Knife patients who had previously been treated with LINAC based radiosurgery were included, although their outcomes were measured from the time of initial brain metastasis diagnosis. Patients that had previously been treated with neurosurgery or WBRT were excluded.
Appropriateness for SRS was determined at weekly multidisciplinary conferences where each patient’s clinical history and radiographical findings were reviewed by members of the radiation oncology, neurosurgery, neuro-oncology, and neuroradiology departments present. On the day of treatment, patients underwent thin section (1 mm slice thickness, 1 mm spacing), axial three-dimensional fast spoiled gradient-echo gadolinium-enhanced magnetic resonance imaging (MRI) of the brain to generate images for planning purposes. The radiosurgical treatment was carried out with the Leksell Gamma Knife Perfexion stereotactic radiosurgery system. Patients were usually discharged the same day of treatment.
After Gamma Knife treatment, patients returned every two months for standard follow-up, including routine MRI of the brain with and without gadolinium contrast. If a new brain metastasis appeared during follow-up imaging, patients were typically offered SRS, resection, whole brain radiation therapy (WBRT), or supportive care based on their overall clinical status.
Tumor marker status was based on surgical biopsies obtained from the primary tumor. Patients with histologically-confirmed non-small cell lung cancer were routinely assessed with gene mutation analysis at our institution during the study period.
Data were analyzed using the Stata 12.0 statistical software program. All statistical tests were based on a two-sided significance level, and P≤0.05 was considered to indicate statistical significance. The three end points examined in our analysis were overall survival, local brain control, and distant brain control. Raw differences in these endpoints between patients harboring and not harboring mutations were assessed using the Fisher’s Exact test. ANOVA was used to compare the groups of patients for total tumor treated volume and largest single site treated volume.
For survival analysis, all patients alive at the time of analysis were censored at the date of last follow-up. Additionally, an online death record database was also used to determine whether patients had passed away after their final follow-up. Survival durations were calculated from the date of brain metastasis diagnosis using the method of Kaplan-Meier. Survival curves were compared using the log-rank test.
Finally, univariate and multivariate Cox regression analyses were performed for all endpoints. Covariables included in the models were: EGFR and KRAS mutation status (both compared to EGFR-/KRAS- patients), gender, ethnicity, age at brain metastasis diagnosis, whether primary disease was controlled, whether extracranial disease was present, baseline Karnofsky performance scale (KPS), baseline recursive partitioning analysis (RPA), baseline diagnostic-specific graded prognostic assessment (DS-GPA), initial treatment modality of brain metastasis, and treated volume of largest single site. All factors with a P-value <= 0.25 on univariate analysis were included in the multivariate assessment. The proportional hazards assumption was checked using the Schoenfeld residuals.
From June 2009 to March 2010, 112 NSCLC patients underwent SRS for newly diagnosed brain metastases. Eighteen patients who had previously been treated with neurosurgery or whole brain radiation therapy were excluded. Of the remaining 94 patients initially treated with SRS, both EGFR and KRAS mutation status were documented in 67 patients: 9 were positive for an EGFR mutation but negative for a KRAS mutation (EGFR+/KRAS-), 15 were negative for an EGFR mutation but positive for a KRAS mutation (EGFR-/KRAS+), and 43 were negative for both EGFR and KRAS mutations (EGFR-/KRAS-). No patient was found to have both mutations. This cohort included 33 women and 34 men, and was 90% Caucasian (Table 1). Median age at brain metastasis diagnosis was 64.5 years.
Median baseline KPS was 90, median baseline RPA was 2, and median baseline DS-GPA was 2. Of the 67 patients who received SRS upfront for their brain metastases, 63 were treated first with Gamma Knife Perfexion. The remaining four patients received LINAC-based radiosurgery prior to being treated with salvage Gamma Knife SRS between June 2009 and March 2010. The three groups of patients did not have significant differences in median total volume of tumor treated (p=0.856) or median volume of largest single site treated (p=0.755): 0.57 cc and 0.45 cc, respectively, for EGFR+/KRAS- patients vs. 0.69 cc and 0.38 cc for EGFR-/KRAS+ patients vs. 0.92 cc and 0.59 cc for EGFR-/KRAS- patients. The three groups of patients also did not have significant differences in number of lesions treated (Fisher’s exact p=0.780, chi2 p=0.802). All data regarding size of lesions treated and number of lesions treated is shown in Table 2.
The median follow-up duration was 12.4 months (range 1.3-42.2) for all patients. Kaplan-Meier estimates of median overall survival from time of brain metastasis diagnosis were 30.6 months for EGFR+/KRAS- patients vs. 9.8 months for EGFR-/KRAS+ patients vs. 19.1 months for EGFR-/KRAS- patients (p=0.094). Kaplan-Meier curves are shown in Figure 1 for overall survival in EGFR and KRAS mutations.
Local control from time of brain metastasis diagnosis at 24 months was 100% in EGFR+/KRAS- patients vs. 66.7% in EGFR-/KRAS+ patients vs. 97.2% in EGFR-/KRAS- patients (p=0.399). The median time to local failure was not reached in any of the three groups.
Distant control from time of brain metastasis diagnosis at 12 months was 66.7% in EGFR+/KRAS- patients vs. 30.0% in EGFR-/KRAS+ patients vs. 73.7% in EGFR-/KRAS- patients (p=0.039). Median time to distant brain failure was not reached in EGFR+/KRAS- patients, was 7.66 months in EGFR-/KRAS+ patients, and was not reached in EGFR-/KRAS- patients. Kaplan-Meier curves are shown below in Figure 2 for distant brain control in EGFR and KRAS mutations.
On univariate Cox regression analysis of overall survival from date of brain metastasis diagnosis, baseline KPS, KPS <= 80, baseline DS-GPA, and baseline DS-GPA > 2 were all found to be significantly associated with mortality (Table 3). Multivariate Cox regression analysis of overall survival from the time of diagnosis revealed EGFR status (HR=0.30; p=0.054), KRAS status (HR=2.12; p=0.056), and baseline DS-GPA > 2 (HR=0.27; p=0.001) as the most important predictors of mortality (Table 4).
Univariate Cox regression analysis revealed no statistically significant factors associated with local brain failure from time of brain metastasis diagnosis (Table 3). There was no statistically significant multivariate model identified for brain failure.
On univariate Cox regression analysis of distant brain failure from time of brain metastasis diagnosis, whether primary disease was controlled and KRAS status were found to be significantly associated with distant brain failure (Table 3). On multivariate analysis of distant brain failure, KRAS status (HR=4.44; p=0.004) and extracranial disease (HR=3.28; p=0.058) were shown to be the most important predictors of distant brain failure (Table 4).
No violations of the proportional hazards assumption were observed in any of the above multivariate analyses.
Advanced stage non-small cell lung cancer patients are living increasingly longer because of advances in systemic therapies. The predictive and prognostic value of genetic mutations has started guiding individual approach to patient management. The present study aimed to elucidate the role of EGFR and KRAS mutation statuses in NSCLC patients already diagnosed with brain metastases.
In terms of distant brain failure rates, we found that based on survival analysis and multivariate Cox proportional hazards regression, patients with the KRAS mutation had significantly worse outcomes. This finding, in conjunction with worse overall survival, is likely related to the absence of effective targeted agents in this patient population. No prognostic relationship was proven with distant brain failure and EGFR mutations.
Our multivariate data is suggestive that patients with an EGFR gene mutation likely have better overall survival from the time of brain metastasis diagnosis, and that patients with a KRAS gene mutation likely have worse overall survival from the time of brain metastasis diagnosis, compared to those patients with neither EGFR nor KRAS gene mutations. However, overall survival data for EGFR and KRAS mutation status both barely missed the p=0.05 threshold for statistical significance (p=0.054 and p=0.056 respectively). Thus, another study with a larger sample size would be helpful in statistically verifying this study’s observed effect of EGFR and KRAS gene mutation status on overall survival at the time of brain metastasis diagnosis.
Neither mutation was statistically confirmed to have an association with local brain failure. It is possible that the lack of association was driven by low event numbers. Given the effectiveness of ablative radiosurgery, it is not surprising that most tumors respond regardless of their inherent biology, which in turn results in few local failures available for analysis even at a large tertiary center. Given this reality, we expect that biomarkers would be most useful in predicting outcomes outside of the radiosurgical field.
Overall, these findings demonstrate that these biomarkers may have prognostic value for patients even at a late stage in the disease. Because patients and their physicians are confronted with a wide spectrum of treatment options for brain metastases (surgical resection, SRS, WBRT, systemic therapy, and supportive care), biomarker characterization may provide a useful tool with which to navigate therapeutic dilemmas in this context.
Our study has important limitations that should be considered when evaluating our data. First, as with all retrospective studies there are inherent biases present. Second, patients of only one institution were studied, forcing us to consider the generalizability of our findings to other settings, where patients of different acuity might be seen and practitioners might have different practice patterns. Third, despite our moderate sample size (n=67), several variables barely missed the p<0.05 threshold in both Cox and multivariate analyses. Therefore it is likely that if our sample size had been larger, we might have found more significant relationships. Fourth, mutation status was only tested from sample of primary tumors, not from samples of brain metastases. In NSCLC patients, Kalikaki et al. have shown discordance in mutation status between primary site samples and distant metastasis samples in 28% of EGFR mutations and in 24% of KRAS mutations.22 Finally, by virtue of including only patients who received SRS upfront, our analysis cannot be generalized to all NSCLC patients with brain metastases (especially those who received WBRT alone for their metastases). This omission in patient selection might have preselected for better outcomes overall.
Although EGFR and KRAS have generally been thought of as mutations influencing early stages of cancer, our study demonstrates that KRAS mutations and EGFR mutations carry prognostic significance even in the late stages of non-small cell lung cancer once brain metastases are diagnosed. At time of brain metastasis diagnosis, mutation in the KRAS gene portends worse distant brain control into the future. Our data is also suggestive that EGFR mutations portend better overall survival and KRAS mutations portend worse overall survival at time of brain metastasis diagnosis, although additional studies are needed to statistically verify this effect.