Patients with lung cancer harboring the EGFR mutations L858R, Ex19del, or T790M were enrolled in this study between June 1, 2014 and December 31, 2017 after obtaining written informed consent. All patients were diagnosed with NSCLC using surgical tissue samples, biopsy specimens, or cytology samples. The diagnosis was based on the General Rules for the Clinical and Pathological Classification of Lung Cancer of the Japan Lung Cancer Society (8th edition) and TNM staging system of the International Association for the Study of Lung Cancer (8th edition) (17). Written informed consent was also obtained from three healthy volunteers. The baseline EGFR mutations of lung cancers were confirmed by clinically approved methods in Japan by practically examining surgical tissues, biopsy specimens, or cytology samples. Approximately 1–2 ml of EBC was collected using RTube™ (Respiratory Research), according to the manufacturer's protocol. After collecting the EBC, 1-ml aliquots were dispensed and stored at −80°C. This study was approved by the Ethics Committee of Okayama University (Authorization number: 2221).

The lung cancer cell line H3255 harboring EGFR L858R was kindly provided by Dr William Pao (Vanderbilt University, Nashville, TN, USA) (18). The gefitinib-resistant lung cancer cell lines RPC-9 harboring EGFR Ex19del and T790M were previously established in our laboratory (19).

DNA was extracted using the QIAamp DNA Mini kit (Qiagen GmbH) according to the manufacturer's protocol. DNA qualification was performed with a NanoDrop spectrometer (Thermo Fisher Scientific, Inc.). The following primer and probe kits were purchased from Bio-Rad): ddPCR Mutation assay: EGFR p.L858R c.2573T>G (#10049550); ddPCR EGFR Exon 19 Deletions Screening kit (#12002392) and ddPCR Mutation assay: EGFR p.T790M, Human (#10049550).

ddPCR was performed at Biobank (Okayama University Hospital, Okayama, Japan) using the QX200 Droplet Digital PCR system (Bio-Rad) according to the manufacturer's protocol. The following conditions were used for ddPCR: i) An initial denaturation step at 95°C for 10 min followed by: ii) 45 cycles at 94°C for 30 sec; and iii) 45 cycles at °C for 1 min, with a 4) final enzyme deactivation step at 98°C for 10 min. The ramp rate for all steps was 2°C/sec. PCR products were then subjected to analysis with a QX-200 Droplet reader and QuantaSoft analysis software (Bio-Rad), with a EGFR Specific Multiplex DNA Reference Standard (#HD802, Horizon Discovery) used as a positive control. The accuracy of the ddPCR was confirmed using serially diluted DNA of lung cancer cell lines (Fig. S1A-F). The patient samples and control samples including EBC samples from three healthy-volunteers or elution buffer AE from the QIAamp DNA Mini kit (Qiagen) as negative controls were tested to determine the threshold (Fig. S2A-D). No droplets were detected for EGFR L858R or EGFR T790M, whereas positive reactions were observed for EGFR Ex19del in the negative control samples. Therefore, the threshold for positive results for EGFR Ex19del were analyzed with the receiver operating characteristic (ROC) using patient samples (Fig. S3). As a result, the threshold for EGFR mutation-positive was defined as follows: EGFR L858R (≥0.01 copies/µl), EGFR Ex19del (≥0.5 copies/µl), and EGFR T790M (≥0.01 copies/µl).

Statistical analysis was performed using STATA software version 15.1 (StataCorp). ROC analysis was performed to determine the optimal threshold for epidermal growth factor receptor exon 19 deletion. The 95% confidence interval (CI) was calculated by using the Clopper-Pearson exact method for binomial proportions.

Patient and healthy control characteristics are detailed in Tables I and II, respectively. The median patient age was 66 years (range, 54–81), comprising 3 males and 9 females: 10/12 were never-smokers and 9/12 were in stage IV of the disease. Four lung tumors harbored EGFR L858R, whereas 8 tumors showed the EGFR Ex19del mutation (Table I). In total, 21 samples were collected from the 12 patients. Of these 21 samples, 11 were from patients with lung cancers harboring EGFR L858R and 10 were from patients with lung cancers harboring EGFR Ex19del. The timing of EBC collection and treatment history are shown in Fig. 1.

Nine EBC samples were collected from two patients at the time of initiation of EGFR-TKI (Table III). In contrast, 12 EBC samples were collected from ten patients at the time of the second biopsy (Table IV). There were no adverse events due to EBC sampling.

The 21 EBC samples from 12 patients were analyzed by ddPCR using EGFR L858R, Ex19del. or T790M primer sets. In the four patients with lung cancer harboring EGFR L858R, 3 of 11 EBC samples (sample nos. 1–1, 8–4, and 9–4) were positive for EGFR L858R (Fig. 2A and Tables II and IV). In the 8 patients with lung cancer harboring EGFR Ex19del, 3/10 EBC samples (sample nos. 2-1, 4-1, and 11-1) were positive for EGFR Ex19del (Fig. 2B and Table IV). In the 8 patients with lung cancer harboring EGFR T790M, 2/9 EBC samples (sample nos. 5-1 and 11-1) were positive for EGFR T790M (Fig. 2C and Table IV). No association was detected between the detection of mutations and T-factor/tumor localization in the lung or quality/quantity of DNA in the EBC samples (Tables III and IV).

Consequently, the sensitivity of the EBC test for EGFR mutations was as follows: 27.3% (95% CI, 6.0–61.0%) for EGFR L858R, 30.0% (95% CI, 6.7–65.2%) for EGFR Ex19del, and 22.2% (95% CI, 2.8–60.0%) for EGFR T790M. The specificity was as follows: 80.0% (95% CI, 44.4–97.7%) for EGFR L858R, 90.9% (95% CI, 58.7–99.8%) for EGFR Ex19del, and 100% (95% CI, 73.5–100%) for EGFR T790M (Table V).

Patient number 8. A non-smoking 67-year-old woman was diagnosed with synchronous double primary lung cancer composed of adenocarcinoma harboring EGFR L858R in the right lower lobe of the lung, and adenocarcinoma harboring EGFR Ex19del in the right upper lobe of the lung. A right lower lobectomy (pathological stage T2bN0M0 cStage IIA) and right segment 3a partial resection (pathological stage T1aN0M0 cStage IA1) were performed, with four subsequent courses of adjuvant chemotherapy with a combination of cisplatin and vinorelbine. However, at 12 months post-surgery, a new lesion appeared in the middle lobe. Transbronchial biopsy of the lesion revealed an adenocarcinoma harboring EGFR L858R. The patient's Eastern Cooperative Group Performance Status was grade 1, and thus gefitinib was administered at 250 mg daily. EBC was collected 1 day prior starting gefitinib and at 2, 16 and 62 days after gefitinib initiation. Gefitinib was discontinued from days 41–62 because of liver damage, but later re-administered at a reduced dosage of 250 mg every 2 days. EBC-ddPCR detected EGFR L858R only in the fourth sampling (21 days after gefitinib cessation). The maximum therapeutic effect of gefitinib was a partial response (Response Evaluation Criteria in Solid Tumors version 1.1), with a progression-free survival of 48 months (Fig. 3A and B).

A non-smoking 56-year-old woman was diagnosed with adenocarcinoma in the left lower lobe of the lung (clinical stage T3N1M1b cStage IVB, multiple brain metastases, bone metastases). The patient's Eastern Cooperative Group Performance Status was grade 1. EGFR Ex19del was detected in the biopsied tissue, after which erlotinib was administered. However, at 12 months after initiating erlotinib, the primary tumor increased. Two subsequent cycles of cytotoxic chemotherapy with cisplatin and pemetrexed were administered, but the tumor regrew following treatment. Re-biopsy was performed on primary lung tumor and EBC was collected at the same time. In addition to the baseline EGFR Ex19del mutation, EGFR T790M was also detected in the biopsy sample. Similarly, EGFR Ex19del and T790M were detected by the EBC-ddPCR method. Subsequently osimertinib was administered and continued over 6 months, which caused a partial response (Response Evaluation Criteria in Solid Tumors version 1.1) (Fig. 3C and D).