Lung Per-Driver Plasma Concordance: Reading a Blood Genotype Driver by Driver in NSCLC

The problem that shows up before the tissue does

Here's a situation any thoracic team recognizes. A patient with newly diagnosed advanced non-small cell lung cancer (NSCLC) needs an actionable driver identified, the tissue block is still in the histology queue, and a plasma next-generation sequencing (NGS) result lands first — and it's negative. Now what? Does that negative rule out a targetable alteration, or does it just mean the tumor wasn't shedding enough DNA into the circulation that day?

That question is the whole game in plasma genotyping. And the honest answer is that it depends almost entirely on which driver you're asking about. Concordance between plasma and tissue is not a single number you can apply across the board. It's driver-specific, and — this is the part worth internalizing — it's directional.

What the assay actually measures

Circulating tumor DNA (ctDNA) is the tumor-derived fraction of the cell-free DNA (cfDNA) that normally circulates in plasma. Tumor cells shed short, fragmented DNA into the blood; a liquid biopsy extracts that cfDNA and interrogates it for the same alterations we'd look for in tissue — activating mutations like EGFR, and gene rearrangements like ALK, ROS1, and RET fusions.

Two broad assay types appear in the evidence below. The first is targeted allele-specific PCR, exemplified by the cobas EGFR platform, which detects a predefined set of EGFR variants [2]. The second is capture-based or hybrid-capture NGS, which can detect point mutations and, with appropriate design, fusion events [3][4][6].

One interpretive rule governs everything that follows. A plasma-positive result is highly trustworthy — if the specific assay detects the alteration, the tumor almost certainly carries it, because these platforms are built for high specificity. A plasma-negative result is the problem child. It can mean the tumor lacks the alteration, or it can mean the tumor simply wasn't shedding detectable ctDNA. That asymmetry is why the direction of concordance matters so much.

EGFR: where plasma genotyping earns its keep

EGFR is the driver with the deepest plasma evidence, and it's where the method performs most convincingly.

In the IFUM study, plasma EGFR genotyping was evaluated against tissue in a Caucasian NSCLC population treated with gefitinib, establishing early that activating EGFR mutations could be reliably detected in circulation [1]. The FASTACT-2 analysis sharpened the numbers considerably. Using the cobas assay on plasma against matched tissue, investigators reported overall concordance of 88%, with blood sensitivity of 75% and specificity of 96% [2].

Read those figures the way a diagnostician should. Specificity of 96% means a plasma-positive EGFR result rarely misleads you. Sensitivity of 75% means roughly one in four tissue-positive cases won't show up in plasma — the familiar shedding problem. That's the textbook case for how to use these assays: a positive result is clinically informative and may establish eligibility for an EGFR-directed therapy class, while a negative result doesn't close the question and reflexes you to tissue.

ALK and ROS1: the directional asymmetry, in sharp relief

The fusion drivers are where I'd ask you to slow down, because this is where naïve reading of a single concordance percentage goes wrong.

For ALK, a capture-NGS plasma study reported sensitivity of 54.2% against tissue, with specificity of 100% [3]. Hold onto those two numbers — they tell opposite stories. Perfect specificity means a plasma ALK fusion call is essentially never spurious. Sensitivity near half means plasma misses a large fraction of genuine ALK-positive tumors.

A later ctDNA NGS analysis made the directionality explicit, and this directional asymmetry is worth pausing on. Among cases with a fusion detected in plasma, 93% were confirmed in tissue (plasma→tissue concordance), whereas among tissue-confirmed fusions, only 54% were detected in plasma (tissue→plasma concordance) [4]. Same patients, two very different numbers — depending on which assay you treat as the reference. The lesson is clean: trust a plasma fusion call, but never let a plasma-negative stand in for a tissue-negative in a fusion-driven tumor.

ROS1 behaves the same way, and the picture here is less clear than EGFR would suggest because the cohorts are smaller. In a plasma NGS versus tissue evaluation specifically in lung cancer, ROS1 fusions were detectable in circulation, reinforcing that a plasma-positive ROS1 result is actionable while a negative one carries the same shedding caveat seen with ALK [5]. For a rare fusion, the practical consequence is magnified: fewer events means detection estimates are less precise, and reflex to tissue after a plasma-negative is not optional.

RET: detection that's real but incomplete

RET fusions got a rigorous look in a three-assay round-robin study using ctDNA from the LIBRETTO-431 population. Across the assays, plasma detection of known RET fusions ranged from 60% to 67.6%, and partner-level concordance — agreement on the specific fusion partner — ranged from 81% to 89% [6].

This dataset is instructive on two fronts. First, it reproduces the fusion pattern: incomplete detection in plasma, so a negative won't rule out a RET fusion. Second — and this is the subtler point — it quantifies something the single-assay literature glosses over, which is that even when two platforms agree a RET fusion is present, they don't always agree on the partner gene, and that variability sits at a level clinicians rarely see acknowledged. For a clinical validation question, agreement on the fusion partner is not a cosmetic detail; it's part of what "detected" really means.

KRAS G12C and MET exon 14: where the primary evidence isn't

I want to be explicit about a gap rather than paper over it. KRAS G12C and MET exon 14 skipping are both actionable NSCLC drivers with dedicated targeted therapy classes. But among the references available here, there's no standalone primary study reporting driver-specific plasma-versus-tissue concordance for either.

So I won't give you a number. It would be reasonable to expect KRAS G12C — a point mutation — to behave more like EGFR than like the fusions, and MET exon 14 skipping to carry its own detection challenges tied to how splice-altering events are called in cfDNA. But "reasonable to expect" isn't evidence. Until driver-specific concordance data for these two alterations are in hand, the right posture is to treat plasma results for them as not yet characterized to the standard we have for EGFR, ALK, ROS1, and RET.

Where this leaves us at the bench

If you take one thing from the data above, make it the directional asymmetry. The plasma→tissue direction is strong across these drivers — a plasma-positive call, whether it's an EGFR mutation or an ALK, ROS1, or RET fusion, is highly specific and clinically meaningful, and may establish eligibility for the corresponding targeted therapy class [2][3][4][6]. The tissue→plasma direction is where the method thins out, because sensitivity for the fusions runs near 54–68% [3][4][6], and a tumor that isn't shedding simply won't register.

That's why the plasma-negative result in the opening scenario can't close the case. In a fusion-driven tumor especially, a negative plasma genotype is an incomplete study, not a ruled-out driver, and the reflex to tissue is the whole point of understanding concordance in the first place. The RET round-robin adds a final wrinkle worth carrying forward: even "positive" isn't always a single thing, since assays that agree a fusion exists can still disagree on the partner [6]. Knowing which drivers your plasma assay reads well — and which it reads poorly, or hasn't yet been shown to read at all — is the difference between using a liquid biopsy and over-reading it.

References

1. Douillard JY, et al. Gefitinib treatment in EGFR mutated Caucasian NSCLC: circulating-free tumor DNA as a surrogate for determination of EGFR status (IFUM). Br J Cancer. 2014. PMID 24263064; DOI 10.1038/bjc.2013.721.
2. Mok T, et al. Detection and dynamic changes of EGFR mutations from circulating tumor DNA as a predictor of survival outcomes in NSCLC patients treated with first-line intercalated erlotinib and chemotherapy (FASTACT-2). Clin Cancer Res. 2015. PMID 25829397; DOI 10.1158/1078-0432.CCR-14-2594.
3. Cui S, et al. Use of capture-based next-generation sequencing to detect ALK fusion in plasma cell-free DNA of NSCLC patients. Oncotarget. 2017. PMID 27926526; DOI 10.18632/oncotarget.13741.
4. Mondaca S, et al. Detection of ALK fusions by ctDNA next-generation sequencing and concordance with tissue. Lung Cancer. 2021. PMID 34311346; DOI 10.1016/j.lungcan.2021.06.018.
5. Dagogo-Jack I, et al. ROS1 plasma next-generation sequencing versus tissue in lung cancer. J Thorac Oncol. 2019. PMID 30664990; DOI 10.1016/j.jtho.2019.01.009.
6. Paweletz CP, et al. A three-assay round-robin evaluation of RET fusion detection in ctDNA (LIBRETTO-431). Clin Cancer Res. 2025. PMID 39887260; DOI 10.1158/1078-0432.CCR-24-3747.