Serial cfDNA assessment of response and resistance to EGFR-TKI for patients with EGFR-L858R mutant lung cancer from a prospective clinical trial

Serial plasma samples were taken from patients enrolled in CTONG0901, a single-center trial designed to compare erlotinib with gefitinib treatment in patients with advanced NSCLC who had EGFR activating mutations in tumor tissues. Eligible patients were over 18 years of age and had histologically or cytologically confirmed stage IIIB or IV NSCLC (AJCC/UICC, version 6) with EGFR activating mutations in their tumor samples, tested by direct sequencing as previously described [17]. Previously untreated patients and those receiving any type of systemic chemotherapy regimen without prior exposure to any EGFR-TKI were recruited. Eligible patients were 1:1 randomized to receive erlotinib (150 mg, po, qd) or gefitinib (250 mg, po, qd) until disease progression, unacceptable toxicity, or discontinuation of treatment due to other reasons. The primary endpoint was progression-free survival (PFS). Secondary endpoints included objective response rate (ORR), overall survival (OS), disease control rate (DCR), post-progression survival (PPS), safety, and biomarker analyses.

This trial was conducted in Guangdong Lung Cancer Institute of Guangdong General Hospital. It adhered to the Declaration of Helsinki and the Good Clinical Practice guidelines. The protocol was approved by the ethics committee at Guangdong General Hospital (committee’s reference number: GDREC [2009]011). All patients provided written informed consent for participation, with separate consent obtained for tumor specimens and/or blood samples for biomarker analyses. The pre-planned schedule for collecting serial blood samples during the course of EGFR-TKI treatment included baseline, 1 week after treatment, 1 month after treatment, and every 8 weeks until the appearance of disease progression. At each time point, 8 mL of blood was collected. Except for 1 week after treatment, other time points were exactly the same day when patients underwent computerized tomography (CT) scans for tumor response evaluation. Tumor response was evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1. The resistance was defined as disease progression according to RECIST, and the best response was the greatest reduction in tumor burden.

Blood samples were collected in tubes containing EDTA according to a fixed schedule. Plasma was immediately separated from blood cells by centrifugation at 3000 rpm at 4 °C for 5 min. Supernatants were collected and stored at −80 °C until assays were performed. cfDNA was isolated from 4 mL plasma using a QIAamp DNA blood mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. To increase the testing sensitivity, we repeated the isolation five times and then pooled and concentrated the cfDNA from a total of 4 mL plasma for mutation testing.

We analyzed plasma only from patients containing the L858R mutation. Quantitative analyses of the L858R mutation and the endogenous reference gene antitrypsin in the cfDNA were performed by qPCR using the Roche LightCycler^® 480 real-time PCR system (Roche Life Science, Penzberg, Germany) under the following conditions: 95 °C for 10 min for 1 cycle followed by 95 °C for 15 s and 60 °C for 1 min for 50 cycles. An L858R target assay and human antitrypsin primers with the same amplification efficiency were developed by Applied Biosystems (Foster City, CA, USA). Sequences of the L858R mutation primers were as follows: forward primer, GCAGCATGTCAAGATCACAGATTT, reverse primer, CCTCCTTCTGCATGGTATTCTTTCT; MGB-probe, FAM-CAGTTTGGCCCGCCCA; antitrypsin forward primer, GACACCGAAGAGGCCAAGAA, reverse primer, GAAGATGTAATTCACCAGAGCAAAAA; and MGB-probe, FAM-TGTGTCTCTGTCAAGCTCCTTGAC. The qPCR mixture contained 5 μL 2× LightCycler^® 480 Probes Master (Roche Life Science), 0.25 μL primer mix, TaqMan^® Probe Assay (Applied Biosystems), 1.75 μL nuclease-free water, and 3 μL sample DNA or calibrator or H₂O (for the no template control). Each sample was run in duplicate. The differences between replicates were lower than 2 cycle thresholds. If the result from one sample did not meet this criterion, the procedure was repeated. DNA from NCI-H1975 cells (harboring L858R mutation) was used as positive control. Human reference genomic DNA (catalog G1471, Promega Corporation, Madison, WI) was used as negative control. PCR grade H₂O was the non-template control. Two wells of positive controls, two wells of negative control, and two wells of non-template control were included in every run. A mixture of plasma DNA with the L858R mutation was used for calibration. The level of the L858R mutation was normalized to that of the antitrypsin gene. The relative L858R copy number of each plasma sample was calculated as follows: R (L858R copy number) = 2^{− [(Ct sample L858R − Ct sample antitrypsin) − (Ct CalibratorL858R − Ct calibrator antitrypsin)]}.

In plasma at the time of disease progression, we tested the exon 20 T790M mutation using a T790M mutation detection kit (Amoy Diagnostics, Xiamen, China) according to the principle of amplification refractory mutation system (ARMS) as previously described [17].

The plasma L858R copy number is described as the median range. Ward’s hierarchical clustering method was used to categorize the dynamic types of L858R. Based upon the schedule of CT scans, we selected seven time points for Ward’s hierarchical clustering analyses: baseline, 1 week after baseline, 8 weeks before the best response, the best response, 8 weeks after the best response, 8 weeks before disease progression, and disease progression. If a patient progressed early on during treatment and there were not seven time points available for analysis, the available time points were put on corresponding points of the model and the missing time points were supplemented by expectation maximum to maintain the stability of Ward’s hierarchical clustering analyses. The entire population was classified into different groups according to Ward’s hierarchical clustering analyses. The quantities of L858R mutations between groups were compared by unequally spaced repeated measures design analysis of variance (ANOVA). The ratio of change of L858R quantity from baseline to disease progression (r) was calculated as (quantity of L858R at disease progression − quantity of L858R at baseline)/(quantity of L858R at baseline). Cut point of r to separate different groups was analyzed using maximally selected rank statistics by R software. Constituent ratio of EGFR T790M and two kinds of EGFR-TKIs between groups were compared by chi-square or continuity correction tests. PFS was defined as the time from the date of randomization to that of disease progression or of death from any cause. OS was measured from the date of randomization to the date of death from any cause. PPS was measured from the date of disease progression to the date of death from any cause. Survival was estimated using the Kaplan-Meier method and is expressed as a median value with a range and a two-sided 95 % confidence interval (CI). A two-sided log-rank test was used to compare survival between groups. The multivariate Cox proportional hazards regression model (alpha = 0.05) was used to evaluate independent predictive factors associated with PFS. A two-sided P value <.05 was considered statistically significant. SPSS version 17.0 (SPSS, Inc., Chicago, IL) software was used.

From December 2009 to July 2014, a total of 256 patients were enrolled in CTONG0901. One hundred and eight patients harbored L858R mutation in their tumors tested by Sanger sequencing, and 105 patients provided blood samples, with 25 patients providing samples only at one or two time points and 80 patients providing serial blood samples as scheduled. The levels of plasma L858R were therefore tested in 80 patients. The study flow diagram is in Fig. 1. Their characteristics are summarized in Additional file 1: Table S1.

The L858R mutation was tested by qPCR in serial plasma samples from 80 patients at each time point, as scheduled. This mutation was detected in all plasma samples. As a whole, the quantity of plasma L858R decreased and reached the lowest level at the time of best response to EGFR-TKI treatment and then increased to its highest level when the disease progressed (Fig. 2a). By using Ward’s hierarchical clustering analysis, 80 patients were classified into two groups according to their type of change. The most dramatic difference between the two groups was, in one group, the quantity of L858R increased to its highest level at the time of disease progression (ascend type, group A, n = 61), while in the other group, the quantity of L858R did not increase and maintained a stable level as the disease progressed (stable type group S, n = 19) (Fig. 2b). Baseline characteristics of patients in two groups are shown in Table 1. ANOVA analyses showed that the effect of time was significant (F = 4.98, P < 0.01); the interaction of time and group was significant (F = 11.97, P < 0.01); over time, the quantity of plasma L858R copy number showed a linearly increasing trend (F = 23.97, P < 0.01); and the between-group test indicated that the variable group was not significant (F = 3.35, P = 0.071). The estimated cut point of r was 1.07 (M = 3.5877, P = 0.0063), which meant, compared with the quantity of L858R at baseline, the quantity at disease progression increased by more than 1.07 times which was the ascend type and increased by less than 1.07 times was the stable type. ANOVA analyses showed that the quantity of plasma L858R over time had no significant difference between patients receiving erlotinib and those receiving gefitinib (F = 0.160, P = 0.690). Furthermore, chi-square test showed that the constituent ratio of the ascend type and the stable type in patients who received erlotinib or gefitinib had no significant difference, either (P = 0.946).

The cutoff date was October 22, 2015, and the median follow-up time was 20.7 months. In 80 patients, the PFS endpoint was observed in 79 patients (98.8 %), and 70 patients died (87.5 %). The median PFS was 10.4 months (95 % CI, 9.1–11.8), and the median OS was 18.6 months (95 % CI, 16.4–20.8) (Fig. 3). In group A, the median PFS was 11.1 months (95 % CI, 6.6–15.6), while in group S, it was 7.5 months (95 % CI, 1.4–13.6); the difference was statistically significant (HR = 0.55, 95 % CI, 0.32–0.93, P = 0.023) (Fig. 3a). Median OS showed a marginally statistical increase in group A (19.7 months, 95 % CI, 16.5–22.9 vs. 16.0 months, 95 % CI, 13.4–18.5, HR = 0.59, 95 % CI, 0.34–1.01, P = 0.050) (Fig. 3b). Median PPS was 5.5 months (95 % CI, 1.5~9.5) and 5.8 months (95 % CI, 2.9~8.6) (HR = 0.97, 95 % CI, 0.57–1.64, P = 0.898), respectively. The subsequent therapy was well-balanced between the two groups (Additional file 1: Table S2). There was no significant difference in median OS between groups when patients received subsequent best support care, or chemotherapy and/or local treatment. However, for patients who received subsequent other EGFR-TKIs (11 in group A, four in group S), the median OS differed significantly (38.2 months, 95 % CI, 8.9–67.5, vs. 15.8 months, 95 % CI, 2.3–29.2, P = 0.034) (Additional file 1: Table S3). The subsequent EGFR-TKIs included erlotinib, gefitinib, or icotinib which is a domestic first-generation EGFR-TKI in China. No patients received second- or third-generation EGFR-TKIs. The following variants were included in the multivariate Cox proportional hazards regression model: age (≥65, <65), pathology (adenocarcinoma, non-adenocarcinoma), weight loss (<5 %, ≥5 %), smoking status (never a smoker, smoker), family history of cancer (yes, no), Eastern Cooperative Oncology Group performance status (0–1, ≥2), line of EGFR-TKI therapy (first line, second line, or further), EGFR-TKI (gefitinib, erlotinib), and dynamic types of plasma L858R (group A, group S). The results showed that adenocarcinoma, PS 0–1, and group A were independent predictive factors associated with better PFS. Details are in Table 2.

In 80 patients, 78 plasma samples at disease progression were tested for the T790M mutation and T790M was found in 17 samples, 11 samples in group A and six samples in group S (P = 0.305). There was no significant difference between patients with and without T790M in terms of PFS of EGFR-TKI treatment or post-progression survival (PPS) after EGFR-TKI treatment (Additional file 1: Table S4).