Blood Test and Scans Combined Reveal Hidden Patchiness in Advanced Melanoma

For patients with advanced melanoma, tracking how a tumor responds to treatment has long depended on two imperfect tools: periodic imaging scans and, more recently, blood tests that detect fragments of DNA shed by cancer cells. A new study from researchers at the University Medical Center Hamburg-Eppendorf, published in the Journal of Translational Medicine, has now put both approaches under the microscope at unprecedented resolution, and the results reveal a picture far more complicated than either method alone can capture. By following 42 patients with unresectable stage III or IV melanoma who were treated with immune checkpoint inhibitors, the team assembled one of the most detailed longitudinal datasets to date, pairing 241 circulating tumor DNA measurements with radiologic assessments and quantitative measurements of tumor volume.

The central question was deceptively simple: does the amount of tumor DNA floating in a patient’s plasma actually mirror what radiologists see on a scan? The answer, according to the analysis led by Isabel Heidrich, with Klaus Pantel and Christoffer Gebhardt as senior authors, is a qualified yes. Circulating tumor DNA levels rose and fell in broad agreement with disease status, increasing significantly as radiologic response worsened across the standard RECIST 1.1 categories, from complete remission through partial remission and stable disease to progression. The statistical association was modest but real, with a Spearman correlation coefficient of 0.31 and a p-value of 0.002 across 99 paired time points.

Yet the study’s most valuable contribution may be its honest accounting of where that agreement breaks down. When the researchers tested how well a single ctDNA measurement could identify progressive disease, the discriminative performance was only moderate, yielding an area under the receiver operating characteristic curve of 0.69. Threshold analyses exposed a fundamental trade-off that clinicians will recognize immediately. Setting the bar high, at 25 or more mutant molecules per milliliter of plasma, produced strong specificity of at least 83 percent but sensitivity of no more than 30 percent, meaning the test rarely cried wolf but also missed most progressing patients. Lower thresholds of one to two mutant molecules per milliliter hovered around 56 percent for both sensitivity and specificity, a level of accuracy that cannot stand alone.

To move beyond categorical imaging readouts, the team also compared ctDNA concentrations directly with total metastatic tumor volume, calculated quantitatively rather than judged by eye. Here the correlation was stronger: ctDNA levels tracked tumor volume with a Spearman coefficient of 0.46, highly significant at p less than 0.0001 across 73 paired time points. Crucially, timing mattered. When blood draws and imaging occurred within 30 days of each other, the correlation tightened to 0.56, while measurements separated by longer intervals showed weaker agreement. This temporal effect carries a practical lesson for clinical trials and routine care alike: a ctDNA result is most informative about tumor burden when it is anchored closely in time to an imaging reference point.

Perhaps the most striking finding, however, concerned change itself. If ctDNA is a faithful liquid biopsy of tumor dynamics, then a rise or fall in plasma DNA should parallel a corresponding change in tumor volume. It did not. Dynamic changes in ctDNA were not significantly correlated with changes in tumor volume, with a correlation coefficient of just 0.20 and a p-value of 0.25. In other words, a patient whose ctDNA dropped sharply was not reliably the patient whose tumors shrank the most. This decoupling of trajectories suggests that ctDNA shedding is governed by biology that a ruler on a scan cannot see, including cell death rates, vascularization, and the metabolic state of individual metastases.

That spatial dimension emerged forcefully from the organ-specific analyses. The strength of the association between ctDNA and tumor burden varied dramatically depending on where the metastases were located. Lymph node and peritoneal metastases showed comparatively strong relationships between ctDNA levels and disease status, while lung, liver, and brain metastases contributed far less detectable DNA to the bloodstream. This pattern of differential shedding has direct clinical implications: a patient whose disease is confined to visceral sites may have undetectable ctDNA despite substantial tumor burden, while another with bulky nodal disease may show abundant plasma DNA. A negative blood test, the study suggests, cannot be interpreted without knowing the anatomy of the disease it is meant to represent.

The researchers also quantified how often ctDNA dynamics and radiologic assessments actually agreed. Concordance was high among patients whose disease was clearly responding, reaching 85 percent in those with complete or partial remission. In stable or progressive disease, agreement fell to 72.9 percent. The discordant cases, rather than being dismissed as noise, were examined descriptively and linked to three plausible culprits: tumor heterogeneity within and across metastatic sites, differences in the timing between blood draws and scans, and treatment-related factors that may alter shedding independently of tumor cell numbers. Immune checkpoint inhibitors, which unleash immune cells to attack tumors, may cause inflammatory swelling or delayed shrinkage that complicates both imaging and biomarker readouts in the early weeks of therapy.

Technically, the study relied on a UMI-based amplicon next-generation sequencing assay capable of detecting mutations in BRAF, EGFR, KRAS, NRAS, and PIK3CA. Unique molecular identifiers allow the sequencing platform to distinguish true mutant fragments from errors introduced during amplification, a critical safeguard when hunting for a handful of mutant molecules in a sea of healthy cell-free DNA. By quantifying ctDNA as mutant molecules per milliliter rather than as a binary positive-or-negative call, the assay enabled the continuous, high-frequency sampling that made the temporal analyses possible. The retrospective design and the modest cohort size of 42 patients mean the findings should be validated prospectively, but the density of sampling, 241 time points, gives the conclusions a robustness that sparse datasets cannot match.

What emerges from the combined analysis is neither a triumph for liquid biopsy nor a rebuttal of it, but a map of its boundaries. ctDNA clearly reflects radiologic disease status and overall tumor burden, and its correlation with volume strengthens when measurements are temporally aligned. At the same time, the moderate discrimination of progression, the weak link between ctDNA dynamics and volume changes, and the pronounced heterogeneity of shedding across metastatic sites all argue against replacing scans with blood tests. Instead, the authors position ctDNA as a complementary biomarker, one that adds a molecular, real-time layer to the anatomical snapshot provided by CT and MRI. For a disease like advanced melanoma, where immunotherapy can produce deep and durable responses in some lesions while others escape, that layered view may be exactly what clinicians need to catch relapse earlier and tailor treatment with greater confidence.

Subject of Research: Longitudinal ctDNA monitoring and its relationship to radiologic response and tumor volume in advanced melanoma

Article Title: Integrated longitudinal analysis of ctDNA, radiologic response, and tumor volume reveals spatial and temporal heterogeneity in advanced melanoma

Article References: Heidrich, I., Streckenbach, A., Rautmann, C., Freiberg, H., Kött, J., Geidel, G., Rünger, A., Zell, T., Roeper, C., Hansen-Abeck, I., Abeck, F., Schneider, S. W., Smit, D. J., Pantel, K., & Gebhardt, C. (2026). Integrated longitudinal analysis of ctDNA, radiologic response, and tumor volume reveals spatial and temporal heterogeneity in advanced melanoma. Journal of Translational Medicine, 24(1), Article 1199. https://doi.org/10.1186/s12967-026-08899-0

Image Credits: AI Generated

DOI: 10.1186/s12967-026-08899-0

Keywords: ctDNA, liquid biopsy, advanced melanoma, immune checkpoint inhibitors, tumor volume, RECIST, tumor heterogeneity, next-generation sequencing, radiologic response, longitudinal monitoring, metastasis, biomarker