EGFR mutation status in plasma and tumor tissues in non-small cell lung cancer serves as a predictor of response to EGFR-TKI treatment

ABSTRACT

Objective: Epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) can effectively control non-small cell lung cancer (NSCLC). Therefore, EGFR mutations should be detected before lung cancer patients undergo EGFR-TKI therapy. This study assessed the feasibility and predictive value of EGFR mutations in peripheral blood samples.Methods: EGFR mutations in exons 19 and 21 were analyzed in tumor tissue and plasma DNA samples from 121 NSCLC patients using amplification refractory mutation system (ARMS) and the integrated technique of mutant enriched PCR (me-PCR) and denaturing high performance liquid chromatography (DHPLC), respectively.Results: EGFR mutations were detected in 36.4% of tumor tissues and 34.7% of the plasma at a concordance rate of 85.1% (103/121). The sensitivity and specificity of plasma EGFR mutations were 77.3% and 89.6%, respectively. The gender and tumor histology of patients served as independent predictors of EGFR mutations in both tumor tissues and plasma, while CEA level was an independent predictor of EGFR mutations in the plasma. Furthermore, EGFR-TKI treatment showed a significantly higher objective response rate (ORR), median progression-free survival (mPFS), and overall survival (mOS) in patients harboring EGFR mutation than those that did not exhibit EGFR mutation (ORR: 69.4% versus 13.0% in tissues, P < 0.001; 64.5 % vs. 28.6% in the plasma, P = 0.006. mPFS: 10.4 months versus 4.1 months in tissues, P<0.001; 10.5 months vs. 5.2 months in the plasma, P=0.001. mOS: 25.7 months versus 8.3 months in tissues, P=0.005; 25.7 months vs. 13.5 months in the plasma, P=0.038).Conclusions: EGFR mutations can be detected in the plasma using the integrated technique of me-PCR and DHPLC, which enables us to predict patient response to EGFR-TKI therapy. High serum CEA levels served as an independent predictor for plasma EGFR mutations.

Introduction

Lung cancer is a leading cause of cancer-related deaths all over the world. Nearly 80% of lung cancer patients are diagnosed as non–small cell lung cancer (NSCLC).¹ Clinically, most of NSCLC patients are diagnosed at the advanced stages of disease, leading a median survival generally less than 12 months, although chemotherapy can effectively control tumor progression in some cases.² Thus, researchers are trying to develop effective treatment strategies that can help clinicians in tackling NSCLC patients. In recent years, several target therapies have been developed; they are novel tools to effectively control the progression of NSCLC in patients.³ For example, previous studies have reported that epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKI), such as gefitinib, erlotinib, icotinib and afatinib, were able to effectively treat patients with EGFR-mutated NSCLC.^2-5 While using EGFR-TKI therapy, it was found that EGFR mutations occur in exon 19 (19Del) and exon 21 (L858R); these mutations account for up to 90% of all EGFR mutations.⁶ Such mutations lead to dependence of lung tumors on EGFR activity. Therefore, EGFR-TKI therapy targets and inhibits EGFR activities in tumor cells. Thus, numerous clinical trials, including the IRESSA Pan-Asia Study (IPASS),⁷ the EURTAC trial,⁸ and INTEREST trial⁶ have indicated that EGFR mutation is the predictor to determine the efficacy of EGFR-TKI treatment.

Before the application of EGFR-TKI therapy, EGFR mutations need to be assessed in patients. The standard method involves the analysis of EGFR mutation in genomic DNA samples obtained from tumor tissues.³ In up to 70% patients with advanced stages of NSCLC, tissue sample could be either unavailable or insufficient for carrying out EGFR mutations analysis.⁸ For example, in the IPASS clinical trial, only 36% of patients could provide sufficient tumor tissues for the detection of EGFR mutations.⁷ Moreover, at UT MD Anderson Cancer Center, a leading cancer hospital in the world, the incidence of complications associated with thoracic biopsies is more than 17.1%.⁹ Therefore, for detecting EGFR mutations, oncologists could use surrogate samples, especially when sufficient tumor tissue is unavailable. Indeed, circulating free DNA (cf-DNA) in the blood originating from tumor lesions¹⁰ could be used as surrogate sample for detecting EGFR mutations. This is a less invasive source for obtaining genomic samples. Moreover, the plasma also provides an opportunity to dynamically monitor the changes in EGFR mutation when the patient is subjected to treatment. In previous studies, researchers have reported that EFGR mutation detected in the plasma and tissue has a concordance rate that varies from 58.0% to 94.19%;^11-13 this parameter is heavily dependent on ethnicity of enrolled patients, samples collection time, and detection techniques. The mutant enriched PCR (me-PCR) is a sensitive PCR-based assay. In this assay, the mutant gene products are enriched with intermittent restricted digestion of selectively isolated wild-type gene products. Furthermore, we also perform an integrated technique of denaturing and high performance liquid chromatography (DHPLC) to screen gene mutations. DHPLC is a relatively inexpensive technique that can detect gene mutation with high sensitivity and specificity.¹⁴ In addition, serum tumor markers are useful for detecting tumors at an early stage. However, these tumor markers have limited specificity. Nevertheless, for the detection of NSCLC,^15-17 researchers have considered several tumor markers, such as serum CEA, CA125, and CYFRA21; the levels of these markers can indicate EGFR active mutations. Therefore, they can be potentially used to predict the efficacy of EGFR-TKI treatment. In this study, we further assessed the feasibility in detection of EGFR mutations in the plasma using me-PCR DHPLC by comparison to those in tumor tissues using amplification refractory mutation system (ARMS). Then, we associated plasma EGFR mutation status with other tumor markers to identify novel indicators. We also evaluated the predictive value of plasma EGFR mutation status to determine the efficacy of EGFR-TKI treatment in terms of objective response rate (ORR) and patient survival.

Patients and methods

Patients

We enrolled 121 NSCLC patients, who were treated from August 2011 and August 2014 at Cancer Center of Daping Hospital, ChongQing, China. All the patients were histologically confirmed, and EFGR mutation status was determined. Before subjecting these patients to EGFR-TKI treatment, we procured their tumor tissues and plasma. From their medical records, we collected clinicopathological data, such as gender, age, tobacco smoking history, pathological types of tumors, and the clinical stage of disease. We also recorded data on the performance status (ECOG PS), tumor responses to treatment, and patient survival. Tumor response to treatment (imaging-based response) was determined by the Response Evaluation Criteria in Solid Tumors (RECIST version1.1).¹⁸ Tobacco smokers were defined as those who had smoked more than 100 cigarettes in their lifetime. This study was approved by the Medical Ethics Committee of Daping Hospital, and all the subjects had signed a written informed consent letter before participating in this study.

Assessment of serum tumor markers

Tumor markers (CEA and CA125) were measured in serum samples of 120 patients (markers were not measured in one of 121 patients) by performing electrochemical luminescence assay using a diagnostic kit (Roche Company, Indianapolis, IN, USA) according to the manufacturer's instructions. The normal value for CEA was < 5 ng/ml, while that for CA125 was < 35 U/ml. Any values beyond these limits were considered to be increased levels.

Sample collections and genomic DNA extraction

Tumor tissues were either obtained from biopsies or surgically. Then, they were fixed with 10% buffered formalin and embedded in paraffin. Tumor tissue specimens were obtained by surgery from 44 patients (36.4%), whereas bronchial biopsy or percutaneous lung biopsy was performed on 77 patients (63.6%) to obtain tumor tissues. The tumor cell content in biopsy specimen was confirmed to be more than 20% by a pathologist. Blood samples (5 ml) were collected before subjecting the patients to EGFR-TKI therapy. Genomic DNA was extracted from 5 sections of 10 µm thickness using the QIAamp DNA FFPE Tissue kit (Qiagen, Hilden, Germany); genomic DNA was also extracted from 1 ml plasma using QIAamp DNA blood mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The concentration and purity of these DNA samples were determined with a spectrophotometer (NanoDrop2000, Thermo Scientific, Waltham, MA, USA).

Amplification refractory mutation system (ARMS) assay

In this study, ARMS assay was performed to detect EGFR mutations in tissue samples using ADx-ARMS EGFR mutation test kit (Amoy Diagnostics, Xiamen, China) according to manufacturer's instructions on 7300 real-time PCR System (ABI, Foster City, CA, USA). The concentration of template DNA was adjusted to 2 ng/μL and all reactions were performed in 25 μL volumes, including 4.7 μL of template DNA, 20 μL of reaction buffer mix, and 0.3 μL of Taq polymerase. A positive result was obtained if the samples met the criterion that was defined by the manufacturer's instructions.¹⁹

Denaturing high performance liquid chromatography (DHPLC)

In this study, DHPLC was carried out using the Transgenomic Wave DNA Fragment Analysis System equipped with a DNASep column (Transgenomic, Omaha, USA). To prepare DHPLC sample, we performed the me-PCR. The first round of PCR amplification was conducted for 22 cycles (95°C for 5 min and then 30 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s, followed by one cycle of 72°C for 7 min). The PCR product was then digested with 10 U of Mse I for exon 19 and Msc I for exon 21 at 37°C for 4 h. The resulting mutant-enriched DNA samples were subjected to a second round of PCR amplification under the same PCR conditions for 40 cycles. The restriction enzyme Sau96 I was then added to digest the second round of PCR products; from these PCR products, we denatured the codon 858 (GGNCC) of exon 21 mutant at 95°C for 4 min; the temperature was lowered gradually at a rate of 1°C per minute until it reached 35°C. At this temperature, there was formation of heteroduplex. Thereafter, aliquots of all the samples were automatically loaded into the DNASep column of the WAVE DNA Fragment Analysis System. Gradients for DHPLC analysis were determined by the WAVE maker software (Transgenomic, Omaha, USA). The elution was performed at a flow rate of 0.9 ml/min using a mixture of buffers A (0.05% acetonitrile in 0.1 mol/L triethylammonium acetate) and B (25% acetonitrile in 0.1 mol/L triethylammonium acetate). Chromatograms were recorded at an absorbance of 260 nm and compared with the reference standard.

Statistical analysis

We used SPSS statistical software 20.0 (SPSS Institute, Chicago, IL, USA) to analyze the data. The Chi-square test or Fisher's exact test was performed to analyze EGFR mutation status between tumor tissue and the plasma as well as the association between the EGFR mutation status and the clinicopathological characteristics of patients. The median progression-free survival (mPFS) and median overall survival (mOS) of different groups were estimated by Kaplan-Meier method and Log-Rank test. A P value of less than 0.05 was considered to be statistically significant.

Results

Patient characteristics

In this study, we collected tumor tissue and matched plasma specimens of 121 NSCLC patients. Patients' characteristics are enlisted in Table 1. In this study, 59 patients received EGFR-TKI treatment (250 mg of gefitinib, 150 mg of erlotinib, or 375 mg of icotinib daily orally until disease progression or unacceptable toxicity). There were 80 male and 41 female patients; among them, 62 patients were older than 60 y Ninety-seven patients were diagnosed with lung adenocarcinoma, while 104 had Stage IIIb/IV lung cancer and 100 patients had metastatic disease.

Table 1.

The association of patient characteristics with EGFR mutations.

		Tumor tissue			Plasma
variables	No. of patients (%)	Mutation	Wild type	P value	Mutation	Wild type	P value
Age (years)				0.559			0.572
≥60	62 (51.2)	21	41		23	39
<60	59 (48.8)	23	36		19	40
Gender				<0.001			<0.001
Male	80 (66.1)	15	65		16	64
Female	41(33.9)	29	12		26	15
Smoking				<0.001			0.002
Ever/Current	52 (43.0)	8	44		10	42
Never	69 (57.0)	36	33		32	37
Histology				0.001			0.002
Adenocarcinoma	97 (80.2)	42	55		40	57
Others	24 (19.8)	2	22		2	22
Stage				0.235			0.111
I-IIIa	17 (14.0)	4	13		3	14
IIIb-IV	104(86.0)	40	64		39	65
T stage				0.129			0.423
T1-T2b	66 (54.5)	28	38		25	41
T3-T4	55 (45.5)	16	39		17	38
N stage				0.730			0.749
N0-N1	24 (19.8)	8	16		9	15
N2-N3	97 (80.2)	36	61		33	64
M stage				0.188			0.248
M0	21 (17.4)	5	16		5	16
M1	100 (82.6)	39	61		37	63
Brain metastasis				0.909			0.720
Yes	35 (28.9)	13	22		13	22
No	86 (71.1)	31	55		29	57
Bone metastasis				0.535			0.081
Yes	56 (46.3)	22	34		24	32
No	65 (53.7)	22	43		18	47
Other organs metastasis				0.559			0.720
Yes	40 (33.1)	16	24		13	27
No	81 (66.9)	28	53		29	52
Performance Status				0.096			0.130
0–1	101 (83.5)	40	61		38	63
≥2	20 (16.5)	4	16		4	16
CEA (ng/ml)				0.049			0.005
≥5	65 (54.2)	29	36		30	35
<5	55 (45.8)	15	40		12	43
CA125 (U/ml)				0.560			0.818
≥35	56 (46.7)	19	37		19	37
<35	64 (53.3)	25	39		23	41

Concordance of EGFR mutation status in tumor tissue and plasma

In this experimental study, we detected EGFR mutations in both tumor tissues and plasma of 121 patients; the data is presented in Table 2. Briefly, 44 patients (36.4%) had EGFR mutations in tumor tissues, including 22 (18.2%) deletion mutations in exon 19 and 22 (18.2%) L858R substitute mutations in exon 21. In these tissues, we also detected 4 synonymous substitute mutations and one T790M mutation in exon 20; these variations were regarded as wild type mutations. Forty-two (34.7%) matched plasma samples exhibited EGFR mutations, which included 23 (19.0%) deletion mutations in exon 19 and 19 (15.7%) L858R substitute mutations in exon 21 (Table 2). The concordance rate of EGFR mutations in tumor tissues and plasma was 85.1 % (103/121) Compared to the matched tissues, the sensitivity, specificity, positive predictive value, and negative predictive value of plasma EGFR mutations were 77.3%, 89.6%, 81.0%, and 87.3%, respectively.

Table 2.

The association of EGFR mutations with tumor tissue and matched plasma.

		Tissue
Correlate		Wild-type	19 Del	21L858R	Others	Total	Kappa coefficient
Plasma	Wild-type	66	6	4	3	79	0.675 (P<0.001)
		19Del	4	12	6	1		23
		21 L858R	2	4	12	1		19
	Total	72	22	22	5	121

Association of EGFR mutations with clinicopathological features of NSCLC patients

EGFR mutation status in tumor tissue and matched plasma was associated with gender, tobacco smoking history, tumor histology, and serum CEA level (Table 1). Specifically, EGFR mutation rate was significantly higher in female patients than in males (in tumor tissues, 70.7% versus 18.8%, P < 0.001; in the plasma, 63.4% vs. 20.0%, P < 0.001). Compared to smokers, EGFR mutation rate was higher in non-smokers (in tumor tissues, 52.2% versus 15.4%, P < 0.001; in the plasma, 46.4% vs. 19.2%, P =0.002). Compared to other histology of lung cancer, EGFR mutation rate was higher in adenocarcinoma cases (in tumor tissues, 43.3% versus 8.3%, P = 0.001; in the plasma, 41.2% vs. 8.3%, P =0.002). Moreover, EGFR mutation rate was higher in cases where the serum CEA levels increased (≥ 5 μg/L) (in tumor tissues, 44.6% versus 27.3%, P = 0.049; in the plasma, 46.2% vs. 21.8%, P = 0.005) (Table 1).

The multivariate analysis showed that gender and tumor histology were independent predictors for EGFR mutations in both tumor tissues and plasma (Table 3). For plasma samples, CEA level was also an independent predictor of EFGR mutations. Furthermore, we found that plasma EGFR mutation rate was significantly higher as CEA level was higher when an interquartile range of CEA levels was analyzed (Table 4). In addition, we compared the clinicopathological parameters associated with tumor burden among patients of true positive group (activated EGFR mutations in both tumor tissue and plasma samples, n = 34) and false negative group (only positive in tumor tissue samples, n = 10). In clinical terms, patients of the true positive group were at an advanced stage as compared to patients of the false negative group (Fisher's exact test, P=0.032 Table 5).

Table 3.

Multivariate analysis elucidating the association between patient characteristics and EGFR mutations.

Sample	Variable	β	SE	OR	95%CI	P value
Tumor Tissue	Sex	2.125	0.458	8.374	3.415- 20.532	<0.001
	Histology	1.610	0.809	5.002	1.024–24.425	0.047
Plasma	Sex	1.720	0.457	5.585	2.280–13.677	<0.001
	Histology	1.563	0.808	4.775	0.981–23.253	0.053
	CEA	1.135	0.462	3.111	1.258–7.691	0.014

Table 4.

The association of CEA levels with EGFR mutations in the plasma.

Interquartile range of CEA	Mutant	Wild type	Mutations rate (%)	P value
0–25%	6	24	20.0	0.034
25–50%	8	22	26.7
50–75%	12	18	40.0
75–100%	16	14	53.3

Table 5.

Patients' clinicopathological characteristics in true positive and false negative groups.

Variables	True positive group	False negative group	P value
Stage			0.032
I-IIIa	1	3
IIIb-IV	33	7
T stage			0.067
T1-T2b	19	9
T3-T4	15	1
N stage			0.355
N0-N1	5	3
N2-N3	29	7
M stage			0.069
M0	2	3
M1	32	7
Brain metastasis			1.000
Yes	10	3
No	24	7
Bone metastasis			0.150
Yes	19	3
No	15	7
Other organs metastasis			1.000
Yes	12	4
No	22	6
CEA (ng/ml)			0.271
≥5	24	5
<5	10	5
CA125 (U/ml)			1.000
≥35	15	4
<35	19	6

Association of EGFR mutation with response of patients to EGFR-TKI treatment

Among the 121 patients, 59 patients received EGFR-TKI treatment, and the efficacy of this treatment was evaluated subsequently (Table 6). The ORR of patients with or without EGFR mutations in tumor tissues was 69.4% (25/36) and 13.0% (3/23), respectively (P<0.001). Similar results were obtained in plasma samples (64.5 % versus 28.6%, P=0.006). Thus, based on EGFR mutation status either from tumor tissue or plasma, it is consistent that patients with EGFR mutation had a higher ORR as compared to those with wild-type EGFR.

Table 6.

The association of EGFR mutations with response to EGFR-TKI treatment.

Sample	EGFR mutation status	CR+PR	SD+PD	Total	OR
Tissue	Mutation	25	11	36	69.4%
(p<0.001)	Wild type	3	20	23	13.0%
	Total	28	31	59
Plasma	Mutation	20	11	31	64.5%
(p=0.006)	Wild type	8	20	28	28.6%
	Total	28	31	59

The mPFS of patients with tumor EGFR mutations was 10.4 months (95%CI, 8.3 to12.4), it was significantly longer than those without tumor EGFR mutations (mPFS = 4.1 months, 95%CI, 2.4 to 5.9, P < 0.001, Fig. 1A). Patients with plasma EGFR mutations had an mPFS of 10.5 months (95% CI, 6.1 to 14.9), while the patients without EGFR mutations had an mPFS of 5.2 months (95% CI, 2.0 to 8.4) (P=0.001, Fig. 1B). Similarly, patients with EGFR mutations in both tumor tissue and plasma samples had a longer mOS than those without EGFR mutations (in tumor tissue, 25.7 months, 95%CI, 16.1 to 35.3 vs. 8.3 months, 95%CI, 0.0 to 19.0, P = 0.005, Fig. 1C; in the plasma, 25.7 months, 95%CI, 15.6 to 35.7 versus 13.5 months, 95%CI, 7.4 to 19.5, P = 0.038, Fig. 1D).

Figure 1.

Kaplan-Meier curves in 59 patients treated with EGFR-TKI. A, Progression free survival (PFS) stratified by EGFR mutation status in tumor tissues. B, PFS stratified by EGFR mutation status in the plasma. C, Overall survival (OS) stratified by EGFR mutation status in tumor tissues. D, OS stratified by EGFR mutation status in the plasma.

Association and concordance between imaging-based response (IBR) and tumor marker response (TMR)

Among the 59 patients who received EGFR-TKI treatment, TMR and IBR were evaluated of patients with an elevated baseline tumor maker. We also detected TMR in 51 patients 2 months after they received EGFR-TKI treatment (8 patients were excluded because they were not measured at baseline or 2 months post-treatment). We compared CEA levels in 31 patients (20 patients were excluded because they had normal baseline CEA levels) at baseline and 2 months post-treatment. Furthermore, we compared CA125 levels in 27 patients (24 patients were excluded because they had normal baseline CA125 levels) at baseline and 2 months post-treatment. The TMR were defined according to a previous study as followings¹⁴: partial response (PR), post-treatment CEA decline ≥ 50% of baseline, CA125 decline ≥ 25% of baseline; progressive disease (PD), post-treatment CEA or CA125 increase > 25% of baseline; stable disease (SD), post-treatment of CEA or CA125 levels were between PD and SD. The correlation between IBR and TMR is shown in Suppl. Table 1. The concordance between IBR and 2 months post-treatment of CEA and CA125 levels were 71.0% (22/31) and 66.7% (18/27), respectively. The TMR for CEA and CA125 levels indicated that PFS was significantly better in patients with PR than those with SD or PD (P < 0.001, Fig. S1B; P = 0.002, Fig. S1C).

Discussion

The predictive value of EGFR mutations has been extensively explored in plasma and surrogate tumor tissues, and various methodologies have been used to detect the plasma EGFR mutation status in NSCLC patients.^3,14 In previous studies, it has been reported that EGFR mutation rate was between 23.7% and 55.8% in an unselected cohort of NSCLC patients.^11,19,20 In the current study, we assessed EGFR mutations in the plasma using me-PCR and DHPLC assays. We found that EGFR mutations in tumor tissue and matched plasma were 36.4% (44/121) and 34.7% (42/121), respectively. This finding agreed well with that reported in previous studies.^3,14,19 We also proved that the concordance rate of EGFR mutations in plasma and tumor tissues was 85.1 % using analytical techniques of high sensitivity and specificity. This result indicates that to assess EGFR mutations, the plasma can be an appropriate surrogate of tumor tissues in instances where the tumor tissue of patients cannot be procured.

In our current study, we detected 10 patients who had EGFR mutations in tumor tissues but did not exhibit these mutations in their matched plasma. This discrepancy can be attributed to the fact that trace amount of mutated cf-DNA is released from primary tumors, so it is beyond the detection limit of assays.²¹ Indeed, Bai et al¹⁴ have proved that the detection limit of DHPLC assay is approximately 3%, indicating the possibility of false negative results. In our current study, we performed me-PCR and DHPLC assays, which should have improved the detection rate of EGFR mutations. However, although the sensitivity of me-PCR and DHPLC in detecting EGFR mutations in plasma was 0.1% demonstrated by the previous study of our team,²² nearly 22.7% (10/44) tumor EGFR mutations were undetected in the plasma. Similar results were reported by Zhu et al¹² using droplet digital PCR; the selective sensitivity of this technique was at least 0.04%. Therefore, there could be a small number of patients whose tumor samples carried EGFR mutations but were not released into the blood stream. Furthermore, Bettegowda et al²³ have established that the plasma concentration of tumor-derived ct-DNA increases with the size and stage of tumor. Patients with metastatic tumors have higher detectable levels of ct-DNA (86% to 100%) in peripheral plasma than those with localized tumors (49% to 78%). Thus, patients' tumor burden may be another reason for this discrepancy. In contrast, our current study showed that 6 patients whose tumor tissues tested negative for EGFR mutations had detectable EGFR mutations in their matched plasma. Notably, 3 of these patients received EGFR TKI therapy, while 2 such patients had poor prognosis (PFS = 4.6 months and 2.8 months, respectively), indicating that plasma EGFR mutations could be falsely positive. Intertumor or intratumor heterogeneity may be another presumable explanation for this phenomenon. Zhang et al²⁴ have performed genomic DNA sequencing to detect EGFR mutations in 48 tumor lesions, which were obtained from 11 patients with localized lung adenocarcinoma. These tests clearly detected intratumor heterogeneity. Moreover, the types of mutations varied substantially in different tumors. Swanton et al²⁵ have proved that a single biopsy of primary tumors cannot usually reveal the true profile of genomic alterations in tumor lesions; this finding was further confirmed by de Bruin.²⁶ Hence, the tumor tissues that tested negative for EGFR mutations elicit a selection bias of tumor lesions biopsied for EGFR mutation analysis. Therefore, the data of tumor tissues and blood samples have discrepancies.

Furthermore, our current study showed that the frequency of EGFR mutations was higher in non-smoking female patients with adenocarcinoma histology, this finding agreed well with the results of previous studies.^11-14 Moreover, CEA levels can be used as predictors of response to therapies.^15-17 Zhang et al²⁷ have proved compared to ARMS in the detection of plasma EGFR mutation, high levels of serum CEA (≥ 5 ng/ml) can predict EGFR mutation with greater sensitivity(79.1 % vs. 51.2 %), although the specificity of this technique was very low (48.1%). Shoji et al²⁸ have reported that serum CEA level is closely associated with EGFR mutations in tumor tissues of patients with lung adenocarcinoma. Moreover, EGFR mutation rate increases with an elevation of serum CEA level. Our current study further confirmed these data. Furthermore, previous studies have shown that abnormally elevated CEA levels are associated with poor prognosis of NSCLC patients treated with chemotherapy.^29,30 In contrast, Jung et al¹⁷ have reported that compared to patients with low CEA levels, patients with high CEA levels (> 5 ng/ml) had a higher ORR and PFS after being subjected to EGFR-TKI treatment. This finding was further confirmed by Okamoto et al.³¹ These data indirectly indicate that CEA level is associated with the EGFR mutations rate. At a molecular level, an overexpression of serum CEA can be caused by the aberrant activation of anti-apoptotic pathways gene, which are downstream of EGFR, such as Akt, STAT3, and STAT5;^16,33 however, scientists still need to decipher the molecular mechanisms through which CEA levels are associated with EGFR mutations. For another reason, serum CEA levels have been associated with tumor burden in NSCLC patients.^30,32 A greater amount of cf-DNA was released from the primary tumors of patients with higher serum CEA. As a result, the mutations were more often detected in the plasma. With the help of subgroup analysis, we confirmed that patients of the true positive group were at a higher clinical stage than those in the false negative group, indicating that patients of the true positive group had a higher tumor burden.

In addition, our current study also proved the association between TMR and IBR of patient's 2 mo post-therapy, indicating that TMR may serve as a novel method for evaluating the response to EGFR-TKI and have a potential clinical value to reduce the number of CT test performed on NSCLC patients. Furthermore, several previous studies demonstrated that patients who had EGFR mutations showed a higher ORR to EGFR-TKI treatment.^14,19-21 Consistent with the findings of previous reports, our current data indicates that there is a statistically significant difference in ORR of patients with EGFR mutations and wild-type tumors (69.4% vs. 13.0%); the ORR results were similar in the plasma of EGFR-mutated tumors (65.6% vs. 25.9%). After undergoing EGFR-TKI treatment, patients with EGFR mutated tumor had remarkably better PFS than that of individuals with wild-type EGFR in either tumor tissues or the plasma. This result agrees well with the findings of previous reports by Li et al¹⁹ and Kimura et al.²⁰ In addition, previous studies have reported that EGFR mutation status changes after therapy, because they were quantified according to the patients' response to EGFR TKI therapy.³⁴ Bai et al¹⁴ have reported that after being subjected to chemotherapy, patients had a better survival rate provided their plasma EGFR mutation status changed from positive to negative. Recently, Mok et al³⁵ have reported that patients with plasma EGFR mutation tested negative when they were assessed in the third cycle (C3) of gemcitabine/platinum plus sequential erlotinib; such patients had better PFS (12.0 months versus 7.2 months) and OS (31.9 months vs. 18.2 months) compared to those still with plasma EGFR mutation. Thus, plasma EGFR mutation status must be dynamically monitored to predict the prognosis and early detection of drug resistance that is associated with EGFR mutation.

However, our current study does have some limitations. For example, in plasma samples, we only assessed EGFR deletion mutation in exon 19 and L858R substitute mutation in exon 21. However, it has been proved in earlier studies that T790M mutation offers resistance to EGFR-TKI therapy (2). Moreover, other tumor markers, such as CA199 and CYFRA21 were not tested in all the patients as this was a retrospective study. In summary, plasma EGFR mutations detected by me-PCR and DHPLC assays serve as important biomarkers to EGFR mutations in tumor tissues. Therefore, they can be used predict the patients' response to EGFR-TKI treatment in cases where the tissue sample is unavailable. In addition, high serum CEA level was an independent biomarker that could predict various EGFR mutations in plasma, so patients with high level CEA had to be chosen to sensitively detect EGFR mutation in plasma. In the near future, more comprehensive studies would have to be carried out to confirm our current data. Future studies would also have to determine whether CEA levels can be used to predict EGFR mutation and patients' response to EGFR-TKI treatment.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

The authors would like to thank BaojianZhao and Xu Zhang of Beijing Epigenetics Biotechnology Co., Ltd. The authors would also like to thank Qiushi Wang and Hualiang Xiao of Department of Pathology, Daping Hospital as they provided technological assistance. This study was supported in part by a grant from Wu Jieping Medical Foundation (#320.6750.12177). The authors also thank all the patients for participating in this study.