Prediction of EGFR and KRAS mutation in non-small cell lung cancer using quantitative 18 F FDG-PET / CT metrics

This study investigated the relationship between epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations in non-small-cell lung cancer (NSCLC) and quantitative FDG-PET/CT parameters including tumor heterogeneity. 131 patients with NSCLC underwent staging FDG-PET/ CT followed by tumor resection and histopathological analysis that included testing for the EGFR and KRAS gene mutations. Patient and lesion characteristics, including smoking habits and FDG uptake parameters, were correlated to each gene mutation. Never-smoker (P < 0.001) or low pack-year smoking history (p = 0.002) and female gender (p = 0.047) were predictive factors for the presence of the EGFR mutations. Being a current or former smoker was a predictive factor for the KRAS mutations (p = 0.018). The maximum standardized uptake value (SUVmax) of FDG uptake in lung lesions was a predictive factor of the EGFR mutations (p = 0.029), while metabolic tumor volume and total lesion glycolysis were not predictive. Amongst several tumor heterogeneity metrics included in our analysis, inverse coefficient of variation (1/COV) was a predictive factor (p < 0.02) of EGFR mutations status, independent of metabolic tumor diameter. Multivariate analysis showed that being a never-smoker was the most significant factor (p < 0.001) for the EGFR mutations in lung cancer overall. The tumor heterogeneity metric 1/COV and SUVmax were both predictive for the EGFR mutations in NSCLC in a univariate analysis. Overall, smoking status was the most significant factor for the presence of the EGFR and KRAS mutations in lung cancer.


INTRODUCTION
Epidermal growth factor receptor (EGFR) [1,2], Kirsten rat sarcoma viral oncogene homolog (KRAS) [3] and anaplastic lymphoma kinase (ALK) [4] are all significant biomarkers for the management of non-small-cell lung cancer (NSCLC).EGFR is a member of a larger family of closely related transmembrane receptor tyrosine kinases (TK), which activate cell growth and replication, differentiation and survival [5,6].Mutations in the TK domain of the EGFR in NSCLC predict the response to TK inhibitors such as Gefitinib and Erlotinib [7][8][9].KRAS exists downstream of EGFR and the EGFR pathway is altered by KRAS mutation [3].However KRAS mutations are associated with lack of activity of the TK inhibitors [10].ALK, the downstream serine-threonine kinase of EGFR signaling, rearranged tumors are not sensitive to EGFR TK inhibitors, but they are sensitive to ALK specific TK inhibitors such as Crizotinib [4].Akt signaling is one of the main EGFR signaling pathways and includes the upregulation of glucose transporter (GLUT) 1 and 4 transporters [11,12].As a result, Akt activation may have a close relationship with EGFR mutations and fluorodeoxyglucose (FDG) uptake in NSCLC.[13,14].While the relationship between FDG uptake and EGFR mutations in NSCLC has previously been noted to have contradictory results [15,16], and one notable study has shown that the KRAS mutations in lung cancer Research Paper www.impactjournals.com/oncotargetshowed significantly higher FDG uptake than wild type (WT) cancer [17].Tumor heterogeneity relates to both tumor development and therapeutic outcomes [18].Moreover, clonal heterogeneity can be identified within the primary tumor ahead of identification of the metastases [19].Intra-tumor heterogeneity appears to correlate to the EGFR mutations in NSCLC and may predict tumor responsiveness to TK inhibitors therapy [20,21].
FDG uptake usually is not homogeneously distributed within the tumor, which can be caused by variations in necrosis [22], cellular proliferation [23] and hypoxia [24].High intratumor heterogeneity therefore could potentially serve as a prognostic factor in NSCLC [25].
In this study, we investigate if there exists a relationship between EGFR mutations and/or KRAS mutations in NSCLC status and several FDG-PET/CT parameters such as maximum standardized uptake value (SUV max ), metabolic tumor volume (MTV), total lesion glycolysis (TLG), and tumor heterogeneity, in order to determine the FDG-PET/CT metrics that are most predictive of a gene mutation.Subjects were recruited and enrolled in this trial if they had suspected NSCLC based on a diagnostic CT scan.Subsequently, patients underwent a battery of testing that included FDG-PET/CT scanning and gene mutation testing.Gene mutations were investigated using tissues from surgically resected tumor for all patients.

Patient characteristics
The total number of enrolled patients was 182.Fiftyone cases were excluded due to any one or more of the following reasons: 1) margin of lesion not well defined (pneumonic form and central obstructive lesions on preoperative CT which was confirmed by a board-certified radiologist specializing in thoracic imaging), 2) gene mutation analysis not performed, and 3) histologic subtypes other than adenocarcinoma by pathological diagnosis.After the above exclusions, 131 patients (male: 86, female: 45, mean age ± SD: 67 ± 10, range 24-81yrs) met the eligibility criteria for this study and the clinical characteristics of these patients are listed in Table 1.In this study, lung cancer was more frequently identified in males (86/131, 65.6%), but there was no significant age difference between males and females.Lung carcinoma was primarily found in patients classified as current and former smokers (75.8%).The characteristics of identified lung lesions are shown in Table 2.The majority of patients enrolled in this study was clinical stage of IA or IB disease (92/131, 70.2%).

Background FDG uptake in normal lung parenchyma
Results are shown in Table 3.No significant difference was found between left and right lobe for upper area, middle area and the lower area respectively.FDG uptake in lower area was higher than upper area (p < 0.001) and middle area (p < 0.001).No significant factor (age, sex, smoking status, pack years and gene mutation) could be identified for the FDG uptake for normal lung.

Gene mutation analysis
EGFR gene mutations were confirmed in 32 of the 127 patients (25.2%).The KRAS gene mutation was confirmed in 31 of 126 patients (24.6%).
FDG-PET/CT parameters showing a significant difference between EGFR (+) and EGFR-WT case were SUV max , TLG, SD, 1/COV and AUC.In cases where the metabolic tumor diameter was greater than 3 cm, EGFR (+) and EGFR -WT had no significant correlation to the metabolic tumor diameter, tumor volume and the remaining tumor heterogeneity parameters.Further, no PET parameters appeared to correlate to the presence or absence of KRAS mutations (Table 4).
The univariate analysis between several parameters and EGFR and KRAS mutation are shown in Table 5. Never-smoker (i.e.no prior smoking history), lowpack-year smoking history, and female gender were significant factors for EGFR mutation and smoker (current and former) was a significant factor for KRAS mutation.The SUV max of FDG uptake in lung lesion was significant predictor, but those of MTV and TLG were not significant.Of the multiple parameters regarding tumor heterogeneity, 1/COV was the only parameter that was predictive of the EGFR mutation that was not effected or dependent on the metabolic tumor volume diameter.The multivariate analysis showed smoking status was most significant predictor for EGFR mutation in lung cancer.No parameters were identified that was predictive or significantly correlated to the KRAS mutation in lung cancer.The number of cases with each index evaluated in this study are shown in Table 6.

DISCUSSION
In the present study, we found that patients that were categorized as complete never-smoker predicted the presence of the EGFR mutation and current and former smoker predicted the presence of KRAS mutation.The SUV max of FDG uptake in lung lesion were also significant parameters, while those of MTV and TLG were not significant.Of several parameters regarding tumor heterogeneity, 1/COV was the only significant factor which was not dependent on metabolic tumor diameter.The multivariate analysis showed never-smoker smoking status was the only significant factor for EGFR mutation, and that current and former smoker status was the only significant factor for KRAS mutation in lung cancer.www.impactjournals.com/oncotargetEGFR mutations have been linked patients with adenocarcinoma, lack of prior smoking history, females, and Asians.Our results demonstrate that never-smoking (no prior smoking history) was the most significant predictive factor for presence of the EGFR mutation, which corroborates previously observed trends [26].
The frequency of KRAS mutation is not associated with age, gender and smoking history (regardless of pack years) [27].Therefore, KRAS mutation defines a distinct molecular subset of the disease.KRAS mutations were found in tumors from both former/current smokers and never smokers.They are rare in never smokers  Our interest was how tumor metabolism (inferred from PET imaging) could add significant value to predict gene mutations.The most popular metabolic parameter in lung cancer is the SUV max , but SUV max represents just a single point within the tumor even it is easy to measure.Several papers have reported relationship between FDG uptake in lung cancer and EGFR [17,[28][29][30][31] and KRAS gene mutations [17,30].Our study appears to show a positive correlation for SUV max as significant factor for predicating EGFR mutation.We did not find a significant correlation between SUVmax to KRAS.
TLG and MTV were not significant factors for predicting gene mutations (Table 4).This suggests that gene mutation can occur regardless of the size or volume of lung lesion since TLG and MTV are proportional to tumor size.Therefore conventional assessment based on tumor size appeared to be limited for the prediction of gene mutation in lung cancer.One of the novelties of our study was to evaluate the relationship between tumor heterogeneity and gene mutation (Table 4).In regards to tumor heterogeneity, prior studies commonly include both partial volume effects and noise as heterogeneity [32].Several FDG-PET/CT metrics regarding tumor heterogeneity correlated to EGFR mutation, but 1/COV appears to be the most reliable due to the lack of dependence on lesion size.
Yip et al. investigated the association between FDG-PET based radiomic features and somatic mutations in NSCLC.A significant relationship could be seen in SUVmax, MTV, minimum of SUV and several indexes obtained from texture based analysis for predicting EGFR mutations; on the other hand no index could be seen as predicting of KRAS mutations [33].
The key trend of representing tumor heterogeneity is texture based analysis [34,35].Several indexes were shown to have significant relationship with predicting EGFR gene mutations [33].We did not adapt these methods in this study, because the methodology has not been standardized in terms of the software and indexes as described in each article.The 1/COV appears to have limited reliability in their robustness and repeatability, however we adopted PET edge method for tracing the edge of FDG uptake in tumor in order to minimize the measurement variance by the observer [36,37].metastases [39].It appeared that micrometastatic disease had already occurred at the some of cases with early-stage NSCLC.
Therefore, adjuvant and/or neoadjuvant cisplatinbased chemotherapy is advised for patients with early-stage disease [40][41][42].Neoadjuvant therapy has advantages for downstaging the tumor before surgery and thus increasing the chances of a complete resection.
Several randomized clinical trials assessed the advantage of neoadjuvant treatment in patients with early stage disease.Although available data suggest a trend in survival benefit in preoperative chemotherapy, the majority of studies showed no statistically significant differences [43].
A phase II study of preoperative gefitinib in clinical stage I non-small-cell lung cancer demonstrated that tumor shrinkage was frequently seen in women, neversmokers and the EGFR expression (proven by biopsy) was a strong predictor of response [44].Neoadjuvant chemotherapy has also been explored in patients with early-stage NSCLC.It is based on the rationale to be able to decrease micrometastases at distant sites and tumor burden preoperatively to increase resectability and overall survival.
In a systematic review, Nair et al. concluded that increased tumor FDG uptake is associated with poorer survival in patients with stage I NSCLC.FDG uptake has the potential to be used as a biomarker for identifying stage I patients who are at increased risk of death or recurrence and therefore could identify candidates for participation in future trials of adjuvant therapy [45].
However, one of the limitations of neoadjuvant therapy is the inability to confirm gene mutations prior

Study population
The Institutional Review Board and the Stanford Cancer Institute Scientific Review Committee approved this project and protocol.Written informed consent was obtained from all patients before participation in the study.Inclusion criteria were: 1) greater than 18 years-old at the time of radiotracer administration and 2) suspicion of lung cancer on preoperative CT scans by a board-certified radiologist specializing in thoracic imaging.Exclusion criteria were: 1) pneumonic form and central obstructive lesions on preoperative CT which was confirmed by a board-certified radiologist specializing in thoracic imaging.

Clinical data collection
We collected the following clinical variables from each patient: age, histology, sex and smoking status.After review of the histology of NSCLC, we eliminated subtypes of adenocarcinoma including bronchioloalveolar carcinoma (BAC) as defined in the previous pathological classification for lung adenocarcinoma.Smoking status was categorized as never-smoker, former smoker or current smoker.

EGFR and KRAS mutation testing
The tumor tissues were surgically resected for all patients.Mutation testing was done for both EGFR and KRAS using multiplex PCR followed by single nucleotide mutation detection using SNaPshot technology based on dideoxy single-base extension of oligonucleotide primers [46].EGFR exons 18, 19, 20 and 21 were tested and KRAS exon 2. Mutations were combined irrespective of their location in the tested exons.Patients were categorized according to the mutation testing as EGFR mutated (EGFR+) and wild-type EGFR, and KRAS-mutated (KRAS+) and wild-type KRAS.

PET/CT protocol
FDG-PET/CT scans were acquired by using a standard clinical protocol at two sites, Stanford University Hospital (SUH) and Veterans Administration Palo Alto Health Care System (VAPAHCS).PET/CT images were acquired using either GE Discovery LS PET/CT (slice thickness, 3-5 mm) (GE Healthcare, Waukesha, WI, USA) at Stanford or GE Discovery VCT (slice thickness, 3.75 mm) (GE Health care, Waukesha, WI, USA) at VAPAHCS.At both sites, patients fasted for a minimum of 6 hours, a dose of 12 to 17 millicuries (mCi) of FDG was administered, and patients were scanned from the skull base to mid-thigh using multiple bed positions every 5 minutes approximately 45 to 60 minutes after injection.CT-attenuated data were reconstructed using ordered subset expectation maximization for both scanner sites.

Image analysis
Representative images are shown in Figure 1.Images were reviewed by two board-certified Nuclear Medicine physicians (RM, AQ) with 8 and 15 years experience respectively.MIMvista 6.2 software (MIMvista Corp, Cleveland, OH, USA) was used to select and measure structures throughout the body using the regionof-interest (ROI) tool within the software.Circular ROIs with a diameter of 10mm were drawn on transaxial FDG-PET/CT images using the fusion CT scan as an anatomical guide.Background FDG uptake measurement with 10mm ROI was conducted for the upper, middle and lower field of lung in both lungs (if a lung lesion happened to exist in the nearby lung field, the measurement was not performed due to the possibility of affected by tumor FDG uptake).For the aortic blood pool, a circular ROI with 10mm of diameter was placed centrally within the ascending aorta.For SUV measurements of malignant lesions on PET images, CT images of these lesions were used to confirm the exact location of suspected malignant lesions, with reference to diagnostic chest CT.The PETedge tool within MIMvista 6.2 was used with manual adjustment where needed by consensus of two nuclear medicine physicians for measurements of lung tumor.The longest diameter of identified FDG uptake area by the PETedge tool was measured and was regarded as the metabolic tumor diameter rather than the true diameter.The maximum SUV (SUV max ), average SUV (SUV mean ), standard deviation of SUV metabolic tumor volume (MTV) and total lesion glycolysis (TLG: product of MTV and SUV mean ) within a volume of interest (VOI) were recorded.By using these measurement results, inverse coefficient of variation [1/COV, calculated as (SUV mean /SD) ×100%] were additionally calculated as a marker of tumor heterogeneity.Additional metrics for tumor heterogeneity included the calculation of cumulative SUV-volume histograms (CSH) obtained by plotting the percent (SD), volume of a tumor with an SUV above a certain threshold against that threshold, which is varied from 0 to 100% of SUV max .The area under the curve (AUC) of this plot (AUC-CSH) is a quantitative index of uptake heterogeneity, where lower values correspond with increased heterogeneity [32].
Tumor heterogeneity analyses were performed only on cases with a metabolic tumor diameter of 10mm or more (117 of 131 cases) in order to have an adequate number of pixels within a region-or volume-of-interest.

Statistical analysis
Mann-Whitney's test was used to compare the difference of PET uptake in normal lung, and PET parameters according to the gene mutation result.We used univariate analysis and multivariate analysis to investigate the relationship between the parameters regarding FDG uptake for lung lesions and the presence of EGFR and KRAS mutations.We also used univariate analysis for the relationship between several indexes and FDG uptake in normal lungs.All statistical analyses were done with Stata 11 (Stata, College Station, TX).Calculated p-values were two-sided with a p <.05 considered statistically significant.

Figure 1 :
Figure 1: Images and measurement result of FDG PET parameter of lung tumor at right middle lobe.(A) Sagittal whole body PET image, (B) CT portion of PET/CT, (C) PET image (plotted the edge of lung tumor), and (D) fused PET and CT image, (E) cumulative SUV-volume histograms (CSH) : The area under the curve (AUC) of this plot (AUC-CSH) was 0.58.

Table 3 : Result in the FDG parameters
[38]ntage for prediction of gene mutation in lung cancer was to select suitable therapeutic strategy for the patient with lung cancer, and it will be desirable if it could be possible by less invasive method.Stiles B et al. suggested that clinical stage IA lung cancer is frequently under staged in patients[38].Goldstraw et al. reported that 30% to 70% of patients with completely resected disease experienced relapse and/or distant SUV: Standardized uptake value, SUVmax: Maximum SUV, TLG: Total lesion glycolysis, MTV: Metabolic tumor volume.SD: standard deviation, COV: coefficient of variation, AUC: area under the curve of the cumulative SUV-volume histogram, range shown in parenthesis.The

Table 4 : Association between each indexes and EGFR and KRAS mutations status based on univariate analysis (p-values)
MTV: metabolic tumor volume, TLG: total lesion glycolysis, SD: standard deviation, COV: coefficient of variation, AUC: area under the curve of the cumulative SUV-volume Histogram.Measurements within parentheses are indicated maximum metabolic tumor diameter

Table 5 : Multivariate analysis for the association between each indexes and EGRF mutation status (p-values)
COV: coefficient of variation.www.impactjournals.com/oncotarget

Table 6 : Number of cases with each index
area under the curve of the cumulative SUV-volume Histogram.Measurements within parentheses are indicated maximum metabolic tumor diameter to surgical resection.The prediction of gene mutation in lung cancer can be advantageous for selecting patients who would best benefit from neoadjuvant therapy.The limitation of this study was that we have not yet obtained the result of patient's prognosis, therefore we could not report how FDG PET/CT could predict the prognosis nor its prognostic value when used in conjunction with, or compared against, smoking status.