Clinical correlations with EGFR circulating tumor DNA testing in all-stage lung adenocarcinoma

Abstract

BACKGROUND:

Information on genetic alterations, notably EGFR mutations, is important for guiding non-small-cell lung cancer (NSCLC) treatment. Circulating tumor DNA (ctDNA) analysis represents a less invasive alternative to tissue biopsy for analyzing mutation status, but its clinical value may vary across disease stages.

OBJECTIVE:

To explore clinical correlates of ctDNA and tissue/plasma-based EGFR mutation (EGFRm) status across all NSCLC stages.

METHODS:

Ninety patients were analyzed, representing three cohorts: newly-diagnosed early-stage, advanced-stage, and recurrent NSCLC. Relationships among clinical/surgical parameters, ctDNA, EGFRm status, and survival outcomes were analyzed.

RESULTS:

Plasma/tissue EGFRm concordance was lower in early-stage (58.6%) than in advanced-stage patients (87.5%). In early-stage patients, ctDNA levels were variable and not significantly associated with clinical/surgical parameters. In advanced-stage patients, time to EGFR-TKI treatment failure (TTF), but not overall survival (OS), was significantly longer in EGFRm-positive vs. EGFRm-negative patients. In patients with recurrent disease, 40% of plasma samples were EGFR T790M-positive at recurrence. In T790M-positive patients, we noted slight trends toward longer OS with vs. without osimertinib treatment and longer OS and TTF with second-line vs. later-line osimertinib.

CONCLUSIONS:

Our results affirm the use of ctDNA testing in advanced-stage and recurrent NSCLC. Further studies on osimertinib as early-line therapy, clinical correlates and the utility of plasma-based testing in early-stage NSCLC are warranted.

1. Introduction

Activating mutations in the epidermal growth factor receptor (EGFR) gene are among lung cancer’s most common driver mutations [1, 2]. In non-small cell lung cancer (NSCLC), EGFR mutations have been identified in 10–15% of Caucasian patients, and 30–50% of Asian patients [3, 4, 5]. The most frequently identified EGFR mutations are exon 19 deletions and exon 21point mutations [6, 7]. These mutations accounted for $>$ 50% and $>$ 35%, respectively, of EGFR mutations identified in one recent Asian NSCLC cohort [8]. In patients with EGFR-mutant (EGFRm) NSCLC, treatment with EGFR tyrosine kinase inhibitors (EGFR-TKIs), such as gefitinib, erlotinib, and afatinib, dramatically improved both disease-free survival and overall survival compared with chemotherapy [9, 10, 11].

However, even though treatment with first- and second-generation EGFR-TKIs is associated with excellent treatment responses, nearly all patients eventually develop resistance and recurrent disease. Development of EGFR-TKI resistance is most commonly associated with the presence of the EGFR T790M mutation, which has been identified in 40–60% of cases with disease recurrence [12, 13, 14]. Osimertinib, a third-generation EGFR-TKI that targets both EGFR-sensitizing and T790M resistance mutations, has been shown to prolong survival with high efficacy and safety [15, 16]. The US FDA approved osimertinib as a standard second-line treatment for T790M-positive patients and in the first-line setting for patients with EGFR mutations (exon 19 deletion or L858R) detected by any approved test.

Although tissue testing from surgical resection specimens is considered the gold standard for EGFR mutation detection, alternatives are needed in some clinically important scenarios [17]. One such scenario involves NSCLC patients who present with advanced or inoperable disease. In these circumstances, only small tissue specimens can be obtained from the biopsy under bronchoscopy, transthoracic needle biopsy, or cytology. A second common scenario involves EGFRm NSCLC patients who develop recurrent or progressive disease after EGFR-TKI treatment. In such patients, a repeat biopsy is very important to identify genetic alterations that drive resistance, especially the T790M mutation. However, it is often challenging to obtain a repeat biopsy due to the advanced stage of the disease, poorer performance status, or risk of complications from invasive interventions. Tumor heterogeneity across various metastatic sites also complicates the interpretation of test results.

Analysis of circulating tumor DNA (ctDNA) from “liquid biopsies”, such as a plasma or urine samples, is a promising method for EGFR mutation detection in NSCLC patients [18, 19]. However, due to the small size of cell-free DNA fragments, the low fraction of tumor-derived cell-free DNA, and individual variation among patients, EGFR mutation detection in ctDNA requires more sensitive detection methods than tissue testing [20].

Sensitivities for mutation detection in plasma compared with tissue samples ranged from 70–100% for EGFR-TKI-sensitizing mutations and up to 70–80% for T790M; a high concordance between different platforms has been observed [20].

Our study aimed to compare EGFR mutation status between matched tissue and plasma specimens in early-stage and advanced-stage treatment-naïve NSCLC patients and identify EGFR T790M mutations in patients previously treated with EGFR-TKIs, using a ddPCR assay. Several digital and non-digital PCR-based detection methods have been evaluated for plasma ctDNA-based EGFR mutation testing, including the Cobas and therascreen PCR assays, BEAMing dPCR, and digital droplet PCR (ddPCR) [21].

2. Methods

2.1 Patients and samples

This study was conducted between April 2015 and March 2016 at the Ramathibodi Hospital (Bangkok, Thailand), a major tertiary referral center. Patients with histologically confirmed pulmonary adenocarcinoma and available tumor tissue and blood samples were eligible for the study. Patients without adequate documentation of clinicopathologic and treatment characteristics or survival follow-up data for the planned analyses were excluded from the study.

Three cohorts of patients were analyzed, representing the newly diagnosed (I) early-stage and (II) advanced-stage disease, and (III) recurrent disease settings. The newly-diagnosed early-stage group included treatment-naïve patients with resectable stage I–IIIA tumors. For this group, the tissue samples were obtained from surgical resection specimens, and matched peripheral blood samples were collected pre-operatively (one day before surgery). The newly-diagnosed advanced-stage group included treatment-naïve patients with unresectable locally advanced or metastatic disease (stage IIIB–IV). For this group, tissue samples were obtained from small biopsies or cytology specimens from primary tumors or metastatic sites, and matched peripheral blood samples were collected before EGFR-TKI therapy. The recurrent disease group included EGFRm patients with advanced-stage/metastatic disease previously treated with EGFR-TKI. Tissue samples from these patients were obtained from small biopsies or cytology specimens from primary tumors or metastatic sites. The matched plasma samples were those collected after disease progression on EGFR-TKI therapy.

Patients were enrolled after providing written informed consent, and the study conforms with The Code of Ethics of the World Medical Association (Declaration of Helsinki). The study was approved by the Ethics Committee of Ramathibodi Hospital.

2.2 Tissue-based EGFR mutation testing

DNA was extracted from formalin-fixed paraffin-embedded (FFPE) tissue samples using the QIAamp DNA FFPE Tissue Kit (QIAGEN), according to the manufacturer’s instructions. EGFR mutation testing was performed using the therascreen EGFR RGQ PCR Kit version 2 (QIAGEN) or the Cobas EGFR Mutation Test v2 (Roche Molecular Systems), as per routine practice in our institution.

2.3 Plasma-based EGFR mutation testing

2.3.1 Sample preparation and DNA extraction

Peripheral blood samples were collected in EDTA tubes and centrifuged once at 2,500 $\times$ g for 10 minutes at 4 ${}^{\circ }$ C and again at 14,000 g for 10 minutes at 4 ${}^{\circ }$ C within 1 hour to isolate plasma, which was then stored at $-$ 80 ${}^{\circ }$ C until ctDNA extraction. Plasma ctDNA was extracted from 2–5 ml of plasma from each patient with the QIAamp Circulating Nucleic Acid kit (QIAGEN), following the manufacturer’s instructions. Extracted ctDNA from each plasma sample was eluted in 100 $\mu$ l of Tris-EDTA buffer.

2.3.2 Digital droplet PCR detection of EGFR mutations

Extracted ctDNA was tested for EGFR mutations (exon 19 deletion, L858R, T790M, S768I, L861Q, and G719X) using the Droplet Digital PCR system (Bio-Rad Laboratories). Reactions were performed in 20 $\mu$ l volumes using 10 $\mu$ l ddPCR 2X Master mix (Bio-Rad Laboratories), 1 $\mu$ l 20X Mutant primers, 1 $\mu$ l 20X Wildtype primers, 3 $\mu$ l nuclease-free water, and 5 $\mu$ l template. Each sample was then loaded into the well of a droplet generator cartridge; 20 $\mu$ l sample was transferred into the middle wells of the cartridge, taking care to avoid bubbles, and 70 $\mu$ l of droplet generation oil (Bio-Rad Laboratories) was added to the lower wells. The sample cartridge was placed into the droplet generator to generate individual droplets. Once the process was complete, 40 $\mu$ l droplets were transferred into the wells of a 96-well PCR plate, sealed, and loaded into the thermal cycler. The following program was run: 95 ${}^{\circ }$ C for 10 min, followed by 40 cycles of 94 ${}^{\circ }$ C for 30 sec and 55 ${}^{\circ }$ C for 1 in, followed by 98 ${}^{\circ }$ C for 10 min, and holding at 4 ${}^{\circ }$ C. After PCR was complete, the sealed plate was loaded into the droplet reader to detect complete ddPCR reactions in individual droplets. The data were analyzed using QuantaSoft software (Bio-Rad Laboratories). The thresholds for detection were set manually based on results from NTC wells containing water instead of DNA and the negative control using wild-type DNA. The ddPCR result provides DNA concentration of both mutant and wild-type alleles as copies/ml of plasma. The EGFR allelic fraction was determined as the mutant to wild-type signal ratio in each case.

2.4 Outcomes

Follow-up and treatment response evaluations based on clinical or imaging criteria were performed as per routine practice in our institution.

In the early-stage and advanced-stage cohorts, overall survival (OS) was defined as the time from diagnosis (EGFR mutation testing) to death from any causes or last follow-up. In early-stage patients, recurrence-free survival (RFS) was defined as the time from diagnosis to disease recurrence, death from any causes, or last follow-up. In advanced-stage patients, time to treatment failure (TTF) was defined as the time from initiation to end of first-line treatment.

In the recurrent-disease cohort, OS was defined as the time from disease recurrence (established by T790M mutation testing) to death from any causes or last follow-up. TTF was defined as the time from initiation to end of osimertinib treatment.

Table 1

Characteristics of patients

Characteristics	Newly-diagnosed		Recurrent disease ( $n=$ 32)	Total ( $n=$ 90)
	Early-stage ( $n=$ 29)	Advanced-stage ( $n=$ 29)
Age
Median (range)	64 (22–84)	64 (40–80)	64.5 (42–115)	64 (22–115)
Gender
Male	8 (27.5)	15 (51.8)	17 (53.1)	40 (44.4)
Female	21 (72.5)	14 (48.2)	15 (46.9)	50 (55.6)
Smoking status
Never smoker	21 (72.4)	19 (65.5)	22 (68.7)	62 (68.9)
Former smoker	6 (20.7)	7 (24.1)	7 (21.9)	20 (22.2)
Current smoker	2 (6.9)	3 (10.4)	3 (9.4)	8 (8.9)
Staging
IA	12 (41.5)	0	1 (3.1)	13 (14.4)
IB	5 (17.2)	0	0	5 (5.6)
IIA	5 (17.2)	0	0	5 (5.6)
IIB	2 (6.9)	0	0	2 (2.2)
IIIA	5 (17.2)	0	1 (3.1)	6 (6.7)
IIIB	0	2 (6.9)	2 (6.3)	4 (4.4)
IV	0	27 (93.1)	28 (87.5)	55 (61.1)
EGFR mutation subtype (tissue) ${}^1$
Del19	5 (17.2)	11 (37.9)	13 (40.6)	29 (32.2)
L858R	5 (17.2)	4 (13.8)	13 (40.6)	22 (24.4)
S768I	0	0	0	0
L861Q	0	0	0	0
G719X	1 (3.4)	1 (3.4)	0	2 (2.2)
T790M	0	0	0	0
S768I $+$ G719X	2 (6.9)	0	1 (3.1)	3 (3.3)
S768I $+$ L858R	1 (3.4)	0	0	1 (1.1)
L858R $+$ T790M	0	2 (6.9)	2 (6.3)	4 (4.4)
Ex20ins	0	0	2 (6.3)	2 (2.2)
Ex20ins $+$ L858R	0	0	1 (3.1)	1 (1.1)
EGFR mutation subtype (plasma) ${}^2$
Del19	1 (3.4)	8 (27.6)	4 (12.5)	13 (14.4)
L858R	0	5 (17.2)	8 (25.0)	13 (14.4)
S768I	0	0	0	0
L861Q	0	0	0	0
G719X	1 (3.4)	1 (3.4)	0	2 (2.2)
T790M	0	0	0	0
Del19 $+$ T790M	0	0	8 (25.0)	9 (10.0)
S768I $+$ G719X	0	0	0	0
S768I $+$ L858R	0	0	0	0
L858R $+$ T790M	0	1 (3.4)	4 (12.5)	5 (5.6)
Ex20ins	0	0	0	0
Ex20ins $+$ L858R	0	0	0	0
S768I $+$ G719A $+$ T790M	0	0	1 (3.1)	1 (1.1)

${}^1$ EGFR mutation status based on tissue test. ${}^2$ EGFR mutation status based on plasma test. Del19, exon 19 deletion; Ex20ins, exon 20 insertion.

2.5 Statistical analysis

For analysis of matched tissue and plasma samples, concordance rate, sensitivity and specificity, and positive and negative predictive value (PPV, NPV) were calculated with reference to tissue EGFR mutation status. Two-sample Wilcoxon rank-sum (Mann-Whitney), Kruskal-Wallis, or Spearman’s rank correlation tests were used to analyze relationships between clinical/surgical parameters and ctDNA concentration, and between these parameters and EGFR T790M status. The Kaplan-Meier method was used to estimate survival times. Log-rank tests were performed to evaluate the significance of between-group differences in survival outcomes. Univariate and multivariate Cox regression models were used to analyze relationships between clinical/surgical parameters and survival outcomes.

All statistical analyses were performed using Stata software (version 16). Significance tests were performed at the 5% significance level ( $\alpha =$ 0.05).

3. Results

Table 2

Concordance rate for EGFR mutation status in tissue and matched plasma specimens from newly-diagnosed early- or advanced-stage patients

	Concordance	Sensitivity	Specificity	PPV	NPV
Early-stage	58.6%	14.3%	100%	100%	55.6%
Advanced-stage	87.5%	83.0%	100%	100%	78.6%

PPV, positive predictive value; NPV, negative predictive value.

Table 3

Relationship between ctDNA concentration and surgical parameters in early-stage patients

Parameters	ctDNA concentration (ng/ml of plasma)	$P^{*}$
Tumor size	754–18440 (3473)	0.57
Gender		0.59
Male	1450–9840 (4252)
Female	754–18440 (3054)
Smoking status		0.68
Never smoker	754–18440 (3536)
Exsmoker	1547–9840 (3404)
Current smoker	1785–3054 (2419)
Lymphovascular invasion		0.25
Present	754–9840 (2913)
Absent	1525–18440 (3809)
N1 metastasis		0.68
Present	754–18440 (3236)
Absent	1333–9840 (4808)
N2 metastasis		0.60
Present	754–18440 (3504)
Absent	1333–9840 (2076)
Pleural invasion		0.87
Present	754–18440 (3504)
Absent	1785–7430 (3054)

${}^{*}p$ -values from two-sample Wilcoxon rank-sum test, Kruskal-Wallis test, or Spearman’s rank correlation. Results are presented as min-max (median) values. ctDNA, circulating tumor DNA.

3.1 Patient characteristics

A total of 90 patients were included in the study. Of these, 58 patients were diagnosed with early-stage (TNM stage I to IIIa; $n=$ 29) or advanced-stage disease (TNM IIIb to IV; $n=$ 29), and 32 patients had disease recurrence after prior EGFR-TKI treatment (Table 1). In early-stage patients, treatment was primarily surgical resection. Advanced-stage patients were treated with chemotherapy and/or EGFR-TKIs. Patients with recurrent disease received osimertinib ( $n=$ 8) or other treatment ( $n=$ 5) as second- and later-line therapies.

3.2 EGFR mutation status in matched plasma and tissue specimens from newly-diagnosed NSCLC patients

In early-stage patients, EGFR mutations were identified in 14 out of 29 tissue specimens (48%) (Table 1). The matched plasma samples were also positive in two of these 14 tissue EGFRm-positive cases. In advanced-stage patients, EGFR mutations were identified in 18 of 29 tissue samples (62%), and 15 of these 18 matched plasma specimens were also positive. The plasma/tissue EGFR mutation concordance rate, sensitivity and specificity, and positive and negative predictive value of plasma tests are shown in Table 2.

3.3 Early-stage patients: Relationship between ctDNA concentration and clinical/surgical parameters

Plasma ctDNA concentration showed wide variation (range: 754–18440 ng/ml of plasma, median 3473 ng/ml). There was no significant association between ctDNA concentration and disease stage or any other clinical, or surgical resection parameters analyzed, including age, gender, smoking status, tumor size, lymphovascular invasion LVI, pleural invasion (PI), or nodal (N) metastasis (Table 3).

3.4 Early-stage patients: Relationship between clinical/surgical parameters and survival outcomes

Table 4

Relationship between survival outcomes and clinical/surgical parameters in early-stage cases

	Recurrence-free survival (RFS), months				Overall survival (OS), months
	Univariate		Multivariate		Univariate		Multivariate
	mRFS	HR ( $P^{*}$ )	mRFS	HR ( $P^{*}$ )	mOS	HR ( $P^{*}$ )	mOS	HR ( $P^{*}$ )
Age	NA	1.00 (0.989)			NA	1.02 (0454)
Gender
Male	16.1	1			NA	1
Female	NA	0.37 (0.078)			54.4	0.51 (0.364)
Smoking status
Never smoker	NA	1		1	54.4	1		1
Exsmoker	NA	0.74 (0.693)		0.94 (0.944)	NA	2.41 (0.312)		2.41 (0.312)
Current smoker	NA	25.05 (0.023)		285.09 (0.001)	11.7	43.27 (0.003)		43.27 (0.003)
Tumor size
$<$ 4 cm	NA	1		1	54.4	1
$⩾$ 4 cm	27.3	3.14 (0.042)		8.78 (0.003)	NA	1.25 (0.752)
LVI
Present	16.1	4.28 (0.016)			NA	4.24 (0.080)
Absent	NA	1			54.4	1
N1 metastasis
Present	10.18	5.57 (0.004)		16.50 ( $<$ 0.001)	42.5	4.27 (0.041)
Absent	NA	1		1	54.4	1
N2 metastasis
Present	9.1	8.02 ( $<$ 0.001)			42.5	4.35 (0.045)
Absent	NA	1			54.4	1
PI
Present	11.5	2.46 (0.138)			NA	2.38 (0.238)
Absent	NA	1			54.4	1
Resection margin
R0	NA	–			54.4	–
R1/2	NA	–			NA	–

${}^{*}p$ -values from Cox regression. mRFS, median recurrence-free survival; mOS, median overall survival; HR, hazard ratio; NA, not applicable; PI, pleural invasion; LVI, lymphovascular invasion.

The median overall survival (OS) in this group of patients was 54.4 months (95% CI: 45.7 months–not estimable). Univariate and multivariate analyses of relationships between RFS or OS and various clinical/surgical parameters are shown in Table 4. In multivariate models, shorter RFS was significantly associated with current-smoker status ( $p=$ 0.001), presence of N1 nodal metastasis ( $p<$ 0.001), and tumor size $⩾$ 4 cm ( $p=$ 0.003); shorter OS was significantly associated with current-smoker status ( $p=$ 0.003).

3.5 Advanced-stage patients: Relationship between EGFR mutation status and survival outcomes

The median OS in this group was 15.5 months. Median OS in EGFRm tissue-positive patients ( $n=$ 18; 175 [95% CI: 6.1–27.9] months) was similar or slightly longer than in tissue-negative patients ( $n=$ 11; 139 [95% CI: 8.2–25.0] months) (Fig. 1A). Median OS also appeared similar in EGFRm plasma-positive ( $n=$ 15) versus plasma-negative patients ( $n=$ 14) at 14.5 (95% CI: 7.1–40.0 months) and 15.5 (95% CI: 6.3–19.7) months, respectively.

Table 5

Analysis of factors associated with the presence of EGFR T790M in patients with recurrent disease

	Plasma T790M mutation status		$P^{*}$
	Positive ( $N=$ 13)	Negative $N=$ 19
Age, years
Mean $\pm$ SD	592 $\pm$ 18.0	684 $\pm$ 157	0.134
Median range	63 (12–82)	65 (45–115)
Gender (n)			0.430
Male	8 (61.5)	9 (47.4)
Female	5 (38.5)	10 (52.6)
Smoking status (n)			0.732
Current smoker	8 (61.5)	14 (73.7)
Ex-smoker	1 (7.7)	2 (10.5)
Never smoker	4 (30.8)	3 (15.8)
ctDNA concentration, ng/ml			0.007
Mean $\pm$ SD	1045.6 $\pm$ 1579.3	236.0 $\pm$ 653.4
Median range	303.0 (13.1–4353.3)	57.4 (20.7–2918.7)
Mutant EGFR DNA concentration, copies/ml			0.004
Mean $\pm$ SD	553.5 $\pm$ 1200.1	126.1 $\pm$ 501.0
Median range	12.4 (1.0–3340)	1.08 (0–2192)

${}^{*}p$ -values from two-sample Wilcoxon rank-sum test, Kruskal-Wallis test, or Spearman’s rank correlation. ctDNA, circulating tumor DNA.

Figure 1.

Overall survival (OS) and time to treatment failure (TTF) in advanced-stage NSCLC patients as assessed by clinical or imaging criteria as per routine practice of the institution. (A) OS and (B) TTF curves plotted for all advanced-stage patients ( $n=$ 29), EGFRm negative patients ( $n=$ 11), and EGFRm positive patients ( $n=$ 18). Abbreviations: HR, hazard ratio; CI, confidence interval; Ev, events; Ce, censored; $p$ -value: significance level ( $\alpha =$ 0.05).

The median TTF was 44 (95% CI: 3.7–9.9) months. TTF showed a statistically significant relationship with tissue EGFRm status (Fig. 1B). Median TTF in EGFRm tissue-positive patients ( $n=$ 18) was significantly longer than in tissue-negative patients ( $n=$ 11), 83 (95% CI: 4.1–16.2) months versus 3.7 (95% CI: 1.7–4.4) months ( $p=$ 0.003, log-rank test). Median TTF also appeared longer among EGFRm plasma-positive ( $n=$ 15) than plasma-negative ( $n=$ 14) patients, though this difference did not reach statistical significance: 8.3 (95% CI: 3.9–12.3) months versus 3.7 (95% CI: 1.7–5.4) months ( $p=$ 0.070, log-rank test).

3.6 Recurrent disease: Response to osimertinib in T790M-positive patients and survival outcomes

Figure 2.

Overall survival (OS) and time to treatment failure (TTF) in patients with recurrent disease. (A) OS curve among EGFR T790M-positive patients ( $n=$ 13) who either received and did not receive osimertinib treatment. (B) OS and (C) TTF curves among patients treated with osimertinib as either 2nd or $⩾$ 3rd line treatment. Abbreviations: HR, hazard ratio; CI, confidence interval; Ev, events; Ce, censored; $p$ -value: significance level ( $\alpha =$ 0.05).

The median OS after disease recurrence was 12.2 months in this group ( $n=$ 32). Plasma testing identified the T790M acquired resistance mutation in 13 (4.06%) patients, but median OS appeared similar in the T790M-positive and T790M-negative groups: 15.4 (95% CI: 4.2–17.1) months versus 12.2 (95% CI: 1.3–25.5) months ( $p=$ 0.649, log-rank test), respectively. The HR (death) was 1.19 (95% CI: 0.56–2.51) (data not shown).

Eight of the 13 T790M-positive patients were treated with osimertinib. As shown in Fig. 2A, OS appeared longer among osimertinib-treated patients, with a median OS (95% CI) of 15.7 (4.1–20.4) months versus 6.3 (0.1–not estimable) months among those who did not receive osimertinib. However, the difference was not statistically significant ( $p=$ 0.265, log-rank test).

Among osimertinibtreated patients ( $n=$ 8), the median TTF was 11.1 (95% CI: 4.2–12.2) months. Four patients received osimertinib as second-line treatment, and four received osimertinib as third- or later-line therapy up to seventh-line). Both OS and TTF tended to be longer in patients who received osimertinib as second- versus later-line therapy (Fig. 2B–C), although the differences did not reach statistical significance. TTF with osimertinib was not significantly associated with age, gender, smoking status, EGFR mutation subtype, plasma DNA or EGFR concentration (data not shown).

To identify potential predictors of T790M acquired resistance and disease recurrence, we also analyzed relationships between plasma T790M status and clinical variables (Table 5). In the T790M-positive group, plasma ctDNA concentrations and mutant EGFR DNA concentrations were significantly higher than in the T790M-negative group ( $p=$ 0.007 and $p=$ 0.004, respectively).

4. Discussion

Plasma-based ctDNA testing shows promise for quantitative assessment of driver and resistance mutations in lung cancer and is becoming more widely used as an alternative to tissue-based testing. A 2018 IASLC position paper outlined two scenarios in clinical practice in which plasma-based testing may offer advantages: initial molecular diagnosis in treatment-naïve advanced EGFRm NSCLC patients, and detection of acquired resistance mutations in patients with disease progression on EGFR-TKIs and other targeted therapy [17].

On the other hand, the clinical value of plasma-based testing is less clear in early-stage EGFRm NSCLC, when ctDNA levels tend to be low and highly variable. In this study, we used quantitative digital droplet PCR, a highly sensitive detection method, to examine EGFRm status in plasma ctDNA samples from patients treated at our center in three different clinical settings: newly-diagnosed early-stage and advanced-stage NSCLC and recurrent NSCLC after first-line EGFR-TKI therapy.

Our analysis detected EGFR mutations in 71% of NSCLC patients based on tissue samples. This is generally consistent with reports from other institutions in Thailand (49.1–57.4%) [22, 23, 24], and similar to results from next-generation sequencing (NGS) mutation testing in our institution (68.1%) [24]. The most frequently detected EGFR mutation was the exon 19 deletion (32.2%), within the range documented in other studies from Thailand and other Asian countries (31.8–52.7%) [23, 25]. L858R mutations were also common (31.1% overall) in our study; this is consistent with the range of 9.3–37.6% (either alone or in combination with another mutation) reported by other studies [8, 23, 25]. The findings are also consistent with an earlier report from the EGFR Mutation Database, which estimated that 88% of all EGFR mutations occurred in exons 19 and 21 [7].

In newly-diagnosed patients with early-stage NSCLC, concordance between plasma and tissue EGFRm status was only 58.6%, with low detection sensitivity (14.3%). In these early-stage patients, we found that ctDNA levels were low and highly variable and showed no significant relationship with tumor size, lymphovascular or pleural invasion, or nodal metastasis. These findings are in broad agreement with previous analyses of ctDNA-based EGFRm detection sensitivity at different stages of the disease with a range of 10.0–22.2% for early-stage NSCLC, and 56.0–72.7% for advanced-stage NSCLC [18, 26, 27]. As the extent of disease is limited in early-stage NSCLC, even with lymphovascular invasion, the EGFR mutation levels in these patients’ samples may have been too low to detect, even with the highly-sensitive ddPCR method. Further research is needed to evaluate the value of plasma testing in early-stage NSCLC.

Concordance of EGFRm status in matched plasma and tissue was higher in advanced-stage (87.5%) than in early-stage patients. This is within the range reported by other studies in advanced NSCLC, where plasma-tissue concordance rates were 80% or above for matched samples (79.0–93.0%) [19, 20, 28, 29]. The observed sensitivity and specificity of plasma ctDNA-based EGFRm detection were similar to pooled estimates from meta-analyses (pooled sensitivity 62–75%; pooled specificity 79–96%) [18, 30, 31]. A more recent study reported 51% sensitivity and 77% specificity for plasma ctDNA-based EGFR T790M detection in advanced-stage NSCLC [32].

In 40% of the recurrent-disease patients (13/32), the T790M mutation was detected in plasma samples taken at recurrence, similar to what has been reported in Korean, Japanese, and Chinese patients (34.0–43.3%) [33, 34, 35]. Our study is the first report of the incidence of acquired T790M resistance (40.6%) in Thailand. Within the recurrent-disease group, we found significantly higher plasma ctDNA and EGFR mutant DNA levels in T790M-positive than in T790M-negative patients.

Although TTF on EGFR-TKI therapy was significantly longer in EGFRm-positive patients than in EGFRm-negative patients (whether based on tissue or plasma status), this did not appear to translate to significantly longer survival. However, among T790M-positive patients who received osimertinib for recurrent disease ( $n=$ 8), both TTF and OS appeared longer in those who received osimertinib as second-line versus later-line therapy, although the differences did not reach statistical significance in this small subgroup of patients. Initial therapy with first- or second-generation EGFR-TKIs is likely to remain the standard of care across Asia for some time; however, the use of osimertinib as early-line therapy warrants further investigation, especially in view of the longer time-on-treatment and OS observed with osimertinib versus other EGFR-TKIs in untreated EGFRm advanced NSCLC [36].

In this analysis of plasma ctDNA-based EGFR mutation testing and clinical outcomes in Thai patients with newly-diagnosed and recurrent NSCLC, the observed EGFR mutation patterns and clinical outcomes were broadly similar to other studies in Asia. Due to the small sample size in this study, it was difficult to establish firm conclusions regarding relationships between EGFR mutation status and survival outcomes, especially in subgroups. Our results affirm the use of plasma-based mutation testing in managing advanced and recurrent NSCLC, for example, in detecting acquired resistance mutations at recurrence. However, it will be essential to continue evaluating plasma-based testing in early-stage cases, especially as more sensitive and robust detection methods become available for research or clinical use.

Disclosures

Thanyanan Reungwetwattana: Research grant: AstraZeneca, Roche, Novartis, MSD; Advisory board: AstraZeneca, Roche, Novartis, MSD, Pfizer, Yuhan, Amgen, Takeda, BMS, BI; Speaker honoraria: AstraZeneca, Roche, Novartis, MSD, Pfizer, Takeda, BMS, BI.

Pimpin Incharoen: Advisory board: Novartis.

Author contributions

Conception and study design: Pimpin Incharoen, Ekaphop Sirachainan, Thanyanan Reungwetwattana.

Collection of the specimen and laboratory work: Pimpin Incharoen, Artit Jinawath, Lalida Arsa, Kaettipong Kamprerasart, Narumol Trachu, Nanamon Monnamo, Phichai Chansriwong, Ekaphop Sirachainan, Thanyanan Reungwetwattana.

Interpretation, analysis of data, and result conclusion and suggestion: Pimpin Incharoen, Artit Jinawath, Narumol Trachu, Dittapol Muntham, Ekaphop Sirachainan, Thanyanan Reungwetwattana.

Preparation of the manuscript: Pimpin Incharoen, Lalida Arsa, Kaettipong Kamprerasart, Khantong Khiewn-gam, Narumol Trachu, Dittapol Muntham, Ekaphop Sirachainan, Thanyanan Reungwetwattana.

Revision for important intellectual content: Pimpin Incharoen, Thanyanan Reungwetwattana.

Supervision: Thanyanan Reungwetwattana.