TRANSFORM ctDNA Debate Centers on How Biomarkers Are Counted, Timed, and Selected
A short correspondence in Journal of Clinical Oncology lands in a place where modern oncology often becomes methodologically fragile: the interpretation of circulating tumor DNA data. The reply addresses concerns raised about TRANSFORM, and the title alone frames the dispute around three pressure points that can change a biomarker story without changing a single biologic specimen — assessment timing, event definition, and subset selection.
Those details matter because ctDNA is not a static lab value. It is a moving signal shaped by tumor burden, treatment timing, assay cadence, and the way investigators define progression or response events in relation to blood draws. A sample obtained too early can miss a molecular decline still unfolding under therapy. A sample obtained too late can blur biologic response with clinical relapse. When a trial uses ctDNA to support one interpretation of treatment effect, the timing of collection can determine whether the biomarker appears predictive, prognostic, or simply noisy.
The reply is published as correspondence rather than as a full-length trial report, and that format limits how much of the underlying scientific argument can be reconstructed from the bibliographic record alone. Even so, the subject is clinically substantial. TRANSFORM has become part of the broader conversation about how to use ctDNA in real oncology practice, where enthusiasm for molecular monitoring runs ahead of standardization. The field wants a blood test that can anticipate recurrence, stratify risk, and distinguish residual disease from durable remission. It still needs rigorous rules for when to measure, what counts as an event, and which patients belong in the analysis.
One sentence captures the practical tension: ctDNA can only be as reliable as the window through which it is measured.
Assessment timing is the first fault line. In biomarker-driven oncology, serial sampling can appear straightforward on paper and become highly unstable at the bedside or in the trial database. A study may collect blood at fixed intervals, yet treatment-related cytoreduction, marrow recovery, inflammation, or intercurrent therapy can alter signal intensity between those points. If investigators evaluate ctDNA after a clinical milestone rather than before it, the analysis can accidentally reward the wrong biology. What looks like molecular persistence may reflect a delayed draw. What looks like clearance may simply reflect a favorable sampling moment.
Event definition is the second fault line. Clinical oncology has always depended on how events are named, and ctDNA studies raise the stakes because molecular endpoints sit beside radiographic and clinical ones. A recurrence may be defined by imaging, symptoms, laboratory findings, or protocol-specific criteria, yet ctDNA may change earlier than any of those. If the event definition is too broad, the biomarker may seem less specific. If it is too narrow, genuine biologic relapse can be missed or delayed in the analysis. In practice, this affects not only the apparent strength of association but also the confidence clinicians place in any proposed surveillance strategy.
Subset selection is the third fault line, and it is often the least visible to readers. Post hoc restriction of the sample can enrich for cleaner biology while thinning the real-world relevance of the findings. In ctDNA work, excluding patients because of missing samples, early progression, inadequate material, or protocol deviations may produce a neater signal than the full trial cohort would allow. The price is interpretive uncertainty. A biomarker that performs well in a selected subset may not behave the same way in the consecutive population clinicians actually treat.
This is why correspondence in oncology is more than academic sparring. Molecular endpoints are moving toward the center of treatment development, especially in settings where response can be measured before a survival difference has matured. ctDNA has real appeal in that role because it offers a dynamic readout of disease burden and clonal persistence. But the biology is unforgiving. Tumor DNA fragments circulate at low levels, fluctuate with treatment pressure, and can disappear transiently before returning. Small analytic choices can move the signal enough to alter the clinical narrative.
The correspondence also reflects an important feature of scientific progress in precision oncology: the field is still defining the rules of evidence for liquid biopsy. Standardization remains uneven across assays, disease states, and endpoints. Without consistent sampling schedules and prespecified analytic plans, two groups can look at nearly identical data and reach very different conclusions about the same biomarker. In that setting, a reply is not a minor editorial exchange. It is part of the process of deciding whether ctDNA will remain an interesting translational marker or become a dependable clinical instrument.
For clinicians, the bedside implication is straightforward even if the methods are not. A ctDNA result should never be interpreted in isolation from the schedule on which it was obtained, the event against which it is being compared, and the cohort from which it was drawn. Those three variables can move the meaning of the biomarker as much as the assay itself. In a disease-monitoring field that already carries the promise of earlier relapse detection and more tailored treatment decisions, methodological precision is not a luxury.
It is the difference between a useful molecular signal and a misleading one.
The publication record available for this correspondence identifies it as a reply to criticism of TRANSFORM ctDNA interpretation and lists author disclosures from Bristol Myers Squibb, Celgene, Novartis, AbbVie, Kite, EMD Serono, MorphoSys, Alimera Sciences, Karyopharm Therapeutics, C4 Therapeutics, BeiGene, AstraZeneca, Incyte, Bluebird Bio, Kymera, Epizyme, Genmab, MustangBio, Ono Pharmaceutical, Century Therapeutics, Lilly, Caribou Biosciences, Janssen, Takeda, Interius Biotherapeutics, Cellectar, Seagen, Roche/Genentech, ADC Therapeutics, Foresight Diagnostics, Miltenyi Biotec, Johnson & Johnson, Gilead Sciences, and multiple funded research relationships. The accessible record does not provide the body of the reply, so the substantive arguments, if any, are not visible in the material provided here.
What remains unresolved from the accessible record is the exact evidentiary point the authors intended to settle. The title tells readers where the dispute sits; the missing text leaves the resolution, if there was one, out of view.