A total of 96 restrospective FFPE samples were collected and assessed using the Idylla™ EGFR Assay at the Department of Pathology, Cancer Hospital, Chinese Academy of Medical Sciences (CAMS). Samples with a histological diagnosis of NSCLC and a tumor cell content of ≥10% were deemed eligible for inclusion in the study. EGFR mutational status of these samples were assessed between April 2017 and August 2018 using either ARMS-PCR or NGS (Illumina platform). Mutations detected by NGS that were beyond the scope of the Idylla™ EGFR Assay were not included in the analysis. In case of discordance, samples were retested by the Idylla™ assay, and if the results remained inconsistent, ARMS-PCR and NGS (Ion Torrent platform) were repeated for confirmation. Idylla™ tests were also repeated for the discordant cases by increasing tissue input or manual enrichment of tumor cell content via macro-dissection. Another 35 prospective samples were collected and screened for EGFR mutations with the Idylla™ system afterwards. The results were then compared with those obtained using ARMS-PCR and NGS between June 2020 and September 2020. The study was approved by the Institute Review Board of the National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College. The methods were carried out in accordance with approved guidelines. The written informed consent was obtained from all patients. This study followed the ethical standards of the institutional and/or national research committee and the 1964 Helsinki declaration and its later amendments or comparable ethical guidelines.

The Idylla™ EGFR Assay is an integrated cartridge with all sample processing buffers and PCR reagents pre-loaded. This assay is specifically designed to detect 51 mutations in exons 18–21 of the EGFR gene ( Supplementary Table 1 ). For the 96 retrospective samples, a single 8 μm FFPE tissue section containing ≥10% neoplastic cells was added in the cartridge for each test, following the instruction for use of the Idylla™ EGFR Assay. For the 35 prospective samples, a single 8 μm FFPE tissue section was used for surgical samples, while three 8 μm FFPE sections were used separately for biopsy samples. In cases where the neoplastic cell content was lower than 10%, tissue sections were macro-dissected to enrich the sample. Tissue section was sandwiched between two layers of wetted filter paper and loaded directly into the cartridge. The cartridge was then inserted into the Idylla™ system. The system completes sample processing and real-time PCR automatically and reports result of mutations directly. In the Idylla™ EGFR Assay, the control is a wild-type EGFR sequence included in the assay cartridge, and the sample of interest is the DNA extracted from the patient’s tissue sample. The difference between the Cq (Cycle of Quantification) values of the control and the sample of interest (ΔCq) is used to determine the presence or absence of a mutation. If the ΔCq falls within the reference range, a mutant signal is considered valid and a mutation is identified, otherwise, the sample is considered EGFR mutation-negative.

DNA was isolated using the QIAamp DNA FFPE Tissue Kit (Qiagen, CA, USA) and was quantified by the Qubit double-stranded DNA (dsDNA) HS assay kit on the Qubit 3.0 fluorometer (Thermo Fisher Scientific, NH, USA) following the manufacturer’s instructions.

ARMS-PCR was carried out using the National Medical Product Administration (NMPA) approved Human EGFR Mutation Detection Kit (ACCB, Beijing, China). The kit is capable of detecting 44 mutations in EGFR exon 18-21, and some of the target mutations differ from those detected by Idylla™ EGFR Assay ( Supplementary Table 1 ). In accordance with the Kit’s Instruction for Use, 15 ng of genomic DNA from each sample was used for each test. The PCR reaction was performed with the following parameters: initial denaturation at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 15 s and annealing at 60°C for 1 min. Mutation subtypes were determined by analyzing the threshold count (Ct) values of the samples, where mutations were identified when the Ct value was ≤36, following the manufacturer’s instructions. If the Ct value was between 36 and 39, the test was repeated, and if the result remained within this range, the sample was considered a possible EGFR mutant.

DNA-based hybrid capture sequencing was carried out following the protocol as previously reported (27). Genomic DNA was first fragmented using Covaris M220, and then subjected to end repair and adaptor ligation. DNA fragments ranging from 200 and 400 bp were isolated using beads hybridized with a capture-probe panel targeting all exons in 56 cancer-related genes. Subsequently, sequencing libraries were generated after PCR amplification. Indexed libraries were pooled together and then sequenced on a NextSeq N550 platform (Illumina, San Diego, USA). Sequencing data were analyzed by GATK 3.2.

Discrepancies in mutation status were resolved using the Ion Ampliseq Colon and Lung Cancer Panel on the Ion Torrent PGM platform (Thermo Fisher Scientific, NH, USA) following the protocol as previously described (28). Briefly, 10 ng of genomic DNA from each sample was PCR-amplified and then ligated to different barcodes to generate a library. The libraries were mixed and clonally amplified onto the IonSpheres (ISPs) for template preparation, and sequencing was carried out on a 318 chip using the Torrent Suite Software. Mutations were annotated through Torrent Variant Caller and viewed with Integrative Genomics Viewer. Mutations with a coverage depth of ≥1000 and a minor allele frequency (MAF) ≥5% were considered positive using the Torrent Variant Caller.

A total of 96 archival FFPE lung adenocarcinoma samples were included in this study for EGFR mutation analysis using the Idylla™ EGFR Assay as shown in Figure 1 . Among the 96 samples, 79 were previously tested with AMRS-PCR, 11 with NGS, and 7 with both ARMS-PCR and NGS. Initially, 98.96% (95/96) of the sample were successfully tested by Idylla™ EGFR.(One result was invalid due to instrument error). Therefore, a total of 95 samples were included in the concordance analysis. Patients in the 95 samples had a median age of 61 years (interquartile range 37 to 81), with 62.1% (59/95) being female ( Supplementary Table 2 ).

Idylla™ detected mutations in 82 out of the 95 samples as presented in Supplementary Table 3 and Figure 2 . Of these mutated samples, 66 had a single mutation while 16 had two mutations, resulting in a total of 98 mutations being discovered by Idylla™. Among the 66 samples with a single mutation, there were 21 with Exon 19 deletion, 25 with L858R, and 4 with Exon 20 insertion mutations, accounting for 75.8% (50/66) of all mutations. For the remaining 16 samples, 10 had L861Q, 3 had G719X, and 3 had S768I mutations. The 16 samples with two mutations comprised 6 with G719X and S768I, 6 with L858R and T790M, 2 with L858R and S768I, 1 with G719X and L861Q, and 1 with T790M and S768I mutations. The frequency of different types of mutations, from high to low, is as follows: L858R at 33.7% (33/98), Exon 19 del at 21.4% (21/97), L861Q at 11.2% (11/98), G719X and S768I both at 10.2% (10/98), T790M at 6.1% (6/98), and Exon 20 ins at 4.1% (4/98). Among these 95 samples, only two exhibited inconsistent results between the Idylla™ EGFR and the reference method. In one sample (Sample No. 91#), Idylla failed to detect any mutation, whereas the reference method identified G719X + S768I mutations (NGS also detected E709K in this sample, which falls outside the detection range of the Idylla™). In the other sample (Sample No. 88#), Idylla detected only the L858R mutation, while the reference method revealed the presence of both L858R and L861Q mutations. Among all 101 mutations detected by the reference method, Idylla™ missed a total of 3 mutations in 2 samples. The Idylla™ system achieved a sensitivity of 97.6%, a specificity of 100%, and an overall concordance of 97.9%.

The two discordant cases (91# and 88#) were re-examined using the Idylla™ assay, followed by Ion Torrent NGS and ARMS-PCR ( Table 1 ) The H&E photos of samples 88 and 91 were displayed in Figure 3C, F . Sample 91# was wild-type when re-tested with Idylla™ using only one FFPE section. However, when the number of FFPE sections was increased to two, S768I mutation was detected. Further increasing the number of sections to three or four, both G719X and S768I mutations were identified as shown in Table 2 . In the second Idylla™ test of Sample 88#, the same result was obtained as in the first test, with only L858R mutation detected. Subsequently, the neoplastic cell content of the FFPE sections was enriched through macro-dissection, and 2-4 sections were re-tested with Idylla™. However, the result remained the same for 88#, with only the L858R mutation being detected. As illustrated in Figure 3D , the median Cq value for sample 91# was 27.04, indicating that the amount of amplifiable DNA in the cartridge was less than 1.584 ng according to the manufacturer’s instructions. For sample 88#, the median Cq value for the EGFR control was 24.73 ( Figure 3A ), which suggested that the amount of amplifiable DNA in the cartridge was between 7.92 ng and 15.84 ng. However, most of the other samples had a median Cq value of less than 20.00 for the EGFR control, which corresponded to more than 396 ng of amplifiable DNA in the cartridge. The Ct values for samples 88# and 91# by ARMS-PCR were 29.21 and 29.87, respectively, which were close to the upper limit of detection of the assay. Moreover, the Ct values of samples 88# and 91# by ARMS-PCR were 37.04 for L858R ( Figure 3B ), 37.28 for G719X, and 36.62 for S768I ( Figure 3E ), suggesting that the discordance was caused by low DNA input or low mutational allele frequency.

Thirty-five prospective samples were tested using the Idylla™ EGFR Assay in parallel with ARMS-PCR and NGS. Of the 35 samples, 23 were biopsy tissue samples and 12 were surgical tissue samples. Out of the 23 biopsy tissue samples, 2 had a neoplastic cell content of 10%, which is at the minimum threshold required for the Idylla sample input. Patients in the 35 samples had a median age of 58 years (interquartile range 42 to 84), with 54.3% (19/35) being female ( Supplementary Table 2 ). Eleven patients were untreated, four had undergone chemotherapy, and three had received or were currently undergoing EGFR-TKIs. The treatment status of the remaining seventeen patients was unknown. Idylla™ detected mutations in 21 out of the 35 samples, resulting in positive rate of 60% (21/35) ( Supplementary Table 4 ). Of these mutated samples, 15 had a single mutation while 5 had two mutations, resulting in a total of 25 mutations being discovered by Idylla™. Among the 15 samples with a single mutation, there were 6 with Exon 19 deletion, 8 with L858R, and 1 with G719X mutations. Each of the remaining 6 samples had a distinct combination of two mutations. The frequency of different types of mutations, from high to low, is as follows: L858R at 44.0% (11/25), Exon 19 del at 32.0% (8/25), G719X, S768I and T790M each at 8.0% (2/27). Among these 35 samples, only 1(Sample No. 2#) exhibited inconsistent results between the Idylla™ EGFR and ARMS-PCR. Idylla detected only the L858R mutation in 2#, while ARMS-PCR revealed the presence of both L858R and T790M mutations. The Cq value of 2# in Idylla was above 26(data not shown). Two samples were found to have the 19Del variant according to NGS, but they were reported as wild-type by both Idylla™ and ARMS-PCR ( Table 3 ), as both variant types fell outside the detection range of the two methods. Additionally, the presence of C797S in cis with T790M mutation was identified by NGS in one sample. Compared to ARMS-PCR, the Idylla™ system demonstrated high accuracy with an overall agreement of 97.1% (34/35), a sensitivity of 95.2% (20/21) (95% CI, 76.2%-99.9%), and an estimated specificity of 100% (12/12) (95% CI, 76.8%-100%). When compared to NGS, including the two rare 19Del variations, the overall accuracy was 91.4% (32/35), with a sensitivity of 87% (20/23) (95% CI, 66.4%-97.2%), and a specificity of 100% (12/12) (95% CI, 73.5%-100%).