This study shows that it is feasible to integrate NGS testing from liquid biopsies in molecular tumor assessment made at baseline in late-stage NSCC using an in-house approach and an outsourced test. Importantly, both panels were able to assess genomic mutations and the allele frequency was highly correlated (Fig. 1, Additional file 1: Figure S5; Additional file 2: Table S2). It is noteworthy that the in-house test was based on 2 ml of plasma while the outsourced test required significantly more sample input (17 ml of whole blood). Consequently, the in-house test might be favorable when blood samples from the patients are very limited, especially as the in-house test was able to report mutations down to 0.1% allele frequency highlighting that the limited plasma volume did not affect the sensitivity of the assay. Nevertheless, some mutations were not confirmed across all tests, most importantly, those with a very low allele frequency in plasma. Based on the small sample cohort it very challenging to determine if those mutations that were not confirmed by other tests are false-positive calls or, alternatively, are due to an underlying intratumoral heterogeneity that hindered their detection in tissue NGS. However, our results are in line with a recent study comparing ctDNA and tissue DNA sequencing by Sabari et al. which highlighted a rather limited concordance between plasma and tissue based sequencing  but are lower than another study by Guibert et al. .
Importantly, one targetable EGFR mutation was detected in tissue only, while another EGFR mutation was detected at very low frequency using the in-house liquid biopsy assay only. This needs further investigation as the detection of targetable mutations is of utmost importance for patient care.
Furthermore, only the in-house test reported a MET I1010T mutation that was not a somatic mutation of the tumor but a germline polymorphism that is known to impact the function of the protein . Consequently, it would have been beneficial if this polymorphism was also reported by the outsourced test. However, the mutation admittedly has no current therapeutic impact.
Importantly, tissue-based NGS or sequential molecular testing from tissue section was not possible in 8/24 (33%) patients due to a low amount of extracted DNA or the absence of tissue left for the analysis. Here, the liquid biopsy sequencing was the only test that was performed to look for druggable mutations. Indeed, in two of the patients where no tissue testing was possible, an EGFR exon 19 deletion has been detected and a targeted treatment was initiated, highlighting the usability of this approach. However, EGFR mutations can easily be detected using PCR-based assays, so the same results could have been obtained using such an approach . However, with the development of novel targeted treatment options in lung cancer, the NGS approach might become more relevant in order to cover all the requested biomarkers. Additionally, plasma-based NGS is able to not only detect targetable mutations at baseline but also minimal residual disease at progression and it might consequently be beneficial as a prognostic tool [16, 17].
In our study, the turn-around time of the outsourced test was significantly shorter than the in-house testing as patient inclusion especially at the beginning of the study was slow, requiring longer duration to initiate a sequencing run. However, higher patient number will certainly dramatically reduce the turn-around time of the in-house testing. Nevertheless, the outsourced test might be more beneficial when a lower patient number is expected to be tested. However, testing results are often urgent and the in-house test should allow processing the whole sequencing run in 3 days, if needed, which is not possible with the outsourced test.
Importantly, our study is limited in patient number and we consequently cannot draw universal conclusions based on our results. Additionally, the sensitivity of the assay is strongly dependent on the amount of isolated nucleic acids which was very variable between the patients. In this regard it has been shown that the amount of circulating tumor DNA is dependent on tumor burden and disease stage which might explain the reduced sensitivity in some patients . Interestingly, the recommended biomarkers in late-stage NSCC with an approved targeted treatment are EGFR, BRAF, ALK, ROS1, and more recently NTRK . The latter three are genomic rearrangements that are assessed using extracted RNA rather than extracted DNA. The in-house NGS LB assays used in the study includes DNA and RNA extracted from plasma for enhanced detection of rearrangements, but no patients with relevant alterations in these biomarkers were detected in our study which is not surprising considering their low prevalence . In a recent study conducted at the MD Anderson Cancer Center (Houston, Texas), it was concluded that NGS from LB can completely replace tissue-based testing at a reduced TAT but only a few genomic rearrangements were included in the study population . Recently, the feasibility of the detection of ALK rearrangements from plasma samples has been demonstrated using the same outsourced test as reported in the present study which highlights the current advanced in implementing liquid biopsy-based sequencing at baseline . However, the sensitivity of LB for the detection of genomic rearrangements, in comparison to tissue-based testing, needs urgent confirmation.
To initiate a treatment with immune-checkpoint inhibitors (ICIs), like pembrolizumab, the expression of PD-L1 in tumor cells must be assessed and it has to be superior to 50% . Indeed, first line pembrolizumab treatment was initiated in 5/24 (21%) patients in our cohort, based on a tissue biopsy and not the LB results as no PD-L1 assay had been validated in daily practice using blood samples until now. Consequently, the treatment selection in our cohort was mainly driven by PD-L1 IHC on tissue sections and not by the data obtained from LB, thus highlighting the current limitations on the use of plasma-based NGS alone in NSCC patients. However, NGS from plasma certainly has some advantages over current PCR-based methods, as it allows one to assess other biomarkers, like the tumor mutational burden that has been demonstrated to be predictive for response under immunotherapy .