Plasma EGFR T790M Monitoring in NSCLC: What a Negative Result Can — and Cannot — Tell You

The clinical question

Patients with advanced non–small cell lung cancer (NSCLC) driven by a sensitizing EGFR mutation respond well to early-generation EGFR tyrosine kinase inhibitors — until they don't. Resistance is the rule, not the exception, and the single most common resistance mechanism to those earlier drugs is a second EGFR mutation, T790M. Detecting T790M matters because it defines eligibility for a distinct therapeutic class — the third-generation EGFR inhibitors, of which osimertinib is the archetype [2].

So the question at progression is concrete: does this patient's tumor now carry T790M? Historically that meant a repeat tissue biopsy. Re-biopsy of a progressing lung lesion is invasive, sometimes unsafe, and occasionally impossible. Plasma genotyping offers an appealing alternative — a blood draw that samples tumor DNA shed into the circulation. The catch is that a blood test is only as useful as your understanding of what its results mean, and here the asymmetry between a positive and a negative result is everything.

What the assay is and how it works

The analyte is circulating tumor DNA (ctDNA): fragments of tumor-derived DNA that enter the bloodstream and float within the larger pool of cell-free DNA. In plasma genotyping for T790M, the assay interrogates that ctDNA for the specific resistance allele rather than sequencing a solid tumor specimen.

The study that anchors much of our understanding used BEAMing, a digital PCR–based method sensitive enough to detect rare mutant molecules against a background of wild-type DNA [1]. The principle is straightforward even if the chemistry isn't: you quantify how many copies of the T790M allele are present in a small volume of plasma. A positive signal means tumor DNA bearing T790M is circulating. That's a meaningful finding.

Here's the conceptual core. A plasma-positive result for T790M is highly informative — if the mutation is detected in blood, it's almost certainly present in the tumor. A plasma-negative result is a different animal entirely. It doesn't reliably exclude T790M, because the tumor may simply not be shedding enough ctDNA into the circulation for the assay to catch it. That's the defining feature of plasma genotyping in this setting, and it's why the two results carry such unequal diagnostic weight.

The evidence

The Oxnard analysis examined plasma detection of T790M in 58 patients and reported a sensitivity of roughly 70% against tissue as the reference standard [1]. Read that number carefully: it means about 30% of patients with tissue-confirmed T790M had a negative plasma result. Nearly one in three. That is the false-negative rate that defines how this test must be used.

The practical consequence follows directly. Because a plasma-positive result corresponds closely to tissue positivity, a patient with a positive blood test can be identified as T790M-positive without necessarily proceeding to a biopsy. But we can't treat a negative plasma result as definitive. A plasma-negative patient may still harbor T790M that the blood test missed, and confirming that requires tissue. This reflex-to-tissue logic — trust the positive, confirm the negative — is, in my view, the most durable lesson from the Oxnard data, more enduring than any single sensitivity figure.

Why does this matter clinically? Because T790M status gates access to a treatment class. The AURA3 trial established that osimertinib improved outcomes relative to platinum-based chemotherapy in patients with T790M-positive NSCLC that had progressed on earlier EGFR inhibitors [2]. In that trial, T790M positivity was the entry criterion — the biomarker that defined who stood to benefit from the third-generation agent. A false-negative plasma result in real-world practice therefore isn't a trivial laboratory footnote. It is a patient potentially misclassified out of eligibility for a drug class that could help them, which is precisely why the confirmatory biopsy exists.

A few caveats about the evidence base deserve emphasis. The Oxnard sensitivity estimate rests on 58 patients [1]. That's a modest cohort, and any single-study metric from a sample that size carries real statistical uncertainty around it — the true sensitivity could reasonably sit somewhat above or below 70%. The number should be read as a well-grounded estimate, not a fixed constant. The broader signal — substantial sensitivity, non-trivial false-negative rate, strong positive predictive value — is consistent and clinically actionable, but the precise figure is not carved in stone.

Where this stands

Plasma genotyping for EGFR resistance mutations moved from investigational curiosity to accepted clinical practice on the strength of exactly this reasoning. The asymmetry is the whole point. A positive plasma result is confident; a negative one is provisional.

What are the limitations? The central one is tumor DNA shedding, which varies enormously between patients and even between lesions in the same patient. Tumors with low burden, or with disease confined to sanctuary sites that shed poorly into the blood, are the ones most likely to generate a false-negative plasma result. The assay didn't fail in those cases — there simply wasn't enough signal to find. That biology is the mechanism behind the 30% miss rate, and no amount of assay refinement fully abolishes it when the input material isn't there.

This isn't a new principle in diagnostics. A test with high specificity and imperfect sensitivity is a rule-in test, not a rule-out test, and the correct algorithm follows mechanically: a plasma-first approach that reflexes to tissue when plasma is negative. The Oxnard data give that algorithm its empirical foundation [1], and the AURA3 trial supplies the reason the distinction carries therapeutic weight [2].

Open questions remain. How sensitivity varies by assay platform, by disease burden, and by metastatic pattern are all areas where the field continues to refine its expectations, and practice in this space has evolved considerably since the foundational studies. The core teaching point, however, has held up remarkably well.

Consider the patient most likely to be misled: someone with oligometastatic disease and low tumor burden at progression. That patient's plasma may read negative not because T790M is absent but because too little tumor DNA reached the tube. Stopping at the negative blood test, in that scenario, risks withholding a treatment class the tumor would actually have qualified for — and it is exactly that patient for whom the confirmatory biopsy earns its place.

References

1. Oxnard GR, Thress KS, Alden RS, et al. Association between plasma genotyping and outcomes of treatment with osimertinib (AZD9291) in advanced non–small-cell lung cancer. J Clin Oncol. 2016. PMID 27354477.
2. Mok TS, Wu Y-L, Ahn M-J, et al. Osimertinib or platinum–pemetrexed in EGFR T790M–positive lung cancer (AURA3). N Engl J Med. 2017;376(7):629–640. doi:10.1056/NEJMoa1612674. PMID 27959700.