ctDNA predicts breast cancer relapse—but only for some patients

For years, oncologists have eyed circulating tumor DNA (ctDNA) in blood as a potential early warning system for cancer recurrence. The theory is simple: if fragments of a tumor’s genetic material linger after treatment, the disease might not be fully gone. But evidence has been fragmented—until now.

A study presented at the 15th European Breast Cancer Conference (EBCC15) offers the strongest signal yet that ctDNA can predict relapse—but with a critical caveat. Researchers analyzed blood samples from 70 patients who’d received neoadjuvant chemotherapy (chemotherapy given before surgery to shrink tumors). When ctDNA was detected after this pre-surgical treatment, 89% of those patients later relapsed—a stark contrast to the 11% relapse rate in ctDNA-negative patients.

The methodology shift is key: prior studies often tested ctDNA at diagnosis or mid-treatment. Here, the post-chemotherapy timing isolated a high-risk subgroup. As Dr. Nicholas Turner, lead author and professor at the Institute of Cancer Research, noted, this suggests ctDNA’s predictive power may hinge on when you look for it.

Yet the study’s limits are as important as its findings. The sample size (70 patients) is small, and all had HER2-negative, estrogen receptor-positive breast cancer—a specific subset. The abstract also clarifies this was an observational analysis, not a randomized trial. In other words: it’s a compelling pattern, not proof of causality.

Clinically, nothing changes today. ctDNA testing isn’t FDA-approved for relapse prediction, and no guidelines recommend it. The American Society of Clinical Oncology (ASCO) still classifies liquid biopsy for recurrence monitoring as investigational. Even Turner’s team stresses this is a step toward personalized surveillance, not a diagnostic tool—yet.

The real bottleneck isn’t the science; it’s the logistics. ctDNA assays vary widely in sensitivity. A 2023 JAMA Oncology study found false negatives in 30% of early-stage breast cancer cases. Without standardization, broad adoption remains distant.