Impact of Circulating Tumor DNA and Copy-Number Alterations on Clinical Outcome in Relapsed/Refractory Germ Cell Tumors Treated With Salvage High-Dose Chemotherapy

In relapsed germ cell tumors, a liquid biopsy tumor fraction above threshold predicted significantly worse OS; miR-371a-3p did not.

Journal: Journal of Clinical Oncology | Published: 2026-02-19 | Type: Observational cohort study | PMID: 41712877 Authors: Urbini M et al. (IRCCS IRST "Dino Amadori," Italy; Princess Máxima Center, Netherlands; NCI Bratislava, Slovakia — multi-institutional) Funding/COI: Not listed

Summary

In men with relapsed or refractory germ cell tumors (rGCT), circulating tumor DNA — specifically the tumor fraction (TF) measured by shallow whole-genome sequencing — carried prognostic weight that the widely used miR-371a-3p marker did not. High baseline TF predicted worse overall survival in nonseminoma patients across both high-dose and conventional-dose chemotherapy cohorts. A specific cluster of copy-number alterations (gains on 9q and 11q, loss of 6q) flagged 13 patients with dramatically worse outcomes than the rest of the cohort.

Claims

• TF was detectable in 75.4% of baseline plasma samples from 69 HDCT-treated patients; miR-371a-3p was detectable in 98.4%
• High baseline TF was associated with worse overall survival in nonseminoma patients treated with HDCT (P = .021) and confirmed in the CDCT cohort (P = .039)
• miR-371a-3p levels showed no significant association with PFS or OS — not prognostic in this setting
• Gains on chromosomes 9q and 11q, and loss of 6q were linked to poor HDCT outcomes; gain of 3p25.3 (previously reported) was found in ~50% of HDCT cases
• Unsupervised CNA clustering identified 13 patients (high-risk cluster, predominantly extra-embryonic histology — yolk sac, choriocarcinoma) with significantly worse PFS and OS versus the remaining cohort
• In high-TF nonseminoma patients, HDCT appeared to confer better OS than CDCT — though this comparison is cross-institutional and not randomized
• Chromosome 12p gain was the most prevalent CNA, consistent with prior tissue-based GCT literature

Study Quality

This is a prospective sample collection / observational analysis (IGG-04 trial, IRB-approved, consent obtained) from a single HDCT-treating institution in Italy, with a comparison CDCT cohort recruited from a separate institution in Slovakia. The primary cohort (n = 69) is modest in size for a genomic biomarker study; the CDCT comparator (n = 26) is small enough that the cross-cohort OS comparison — different centers, different protocols, different patient selection pressures — cannot support causal conclusions. The sequencing method (shallow WGS at ~3× coverage analyzed via ichorCNA) is an established, validated approach for TF estimation from cfDNA. Serial sampling during treatment is a genuine methodological strength, though on-treatment samples were unavailable for 14 of 69 HDCT patients.

The CNA clustering is exploratory and hypothesis-generating. The 13-patient high-risk cluster finding is intriguing but must be treated as a signal, not a validated risk classifier. External tissue cohort validation of CNA patterns adds some credibility to the chromosomal findings.

Red Flags

• Funding and COI not disclosed — for a study potentially influencing diagnostic tool adoption, this is a meaningful gap
• HDCT vs. CDCT comparison is not randomized and not co-institutional — patients came from different countries, different eras, different referral patterns; any survival difference between treatment arms is confounded
• n = 69 (HDCT) and n = 26 (CDCT) — underpowered for the number of comparisons made, including subgroup analyses by histology
• Histologic subtype at diagnosis, not at relapse — mixed primary histologies reclassify at recurrence; using diagnosis-era subtype may mischaracterize the biology being treated
• Seminoma showing higher TF than nonseminoma is biologically counterintuitive and unexplained; authors acknowledge this needs further study
• The high-risk CNA cluster (n = 13) is too small to draw clinical conclusions from; P-values in clusters this small are unreliable without permutation testing

Strengths

• First study to assess cfDNA prognostic value specifically in relapsed/refractory GCTs — fills a genuine gap in the literature
• Directly compares two candidate biomarkers (TF vs. miR-371a-3p) head-to-head, finding they are not equivalent; the negative miRNA finding is itself informative
• Serial plasma sampling captures treatment dynamics, not just baseline snapshot
• CNA findings validated against independent external tissue cohorts
• IGG-04 is an ongoing observational trial with consent and IRB approval — data collected with future analysis in mind

Verdict

This is a well-executed, appropriately scoped pilot study that delivers a credible negative finding (miR-371a-3p is not prognostic here) alongside a positive one (high TF predicts poor OS in nonseminoma). The multi-timepoint design and head-to-head biomarker comparison make it more useful than a simple association study. But the cohort sizes are too small to validate the CNA clustering findings, the HDCT-versus-CDCT survival comparison is contaminated by institutional confounders, and the missing funding/COI disclosure is a problem. Worth reading for the methodology and the miRNA null result; the CNA risk cluster needs a much larger replication cohort before it changes anything in practice.