RNA Test Shows Promise for Predicting Breast Cancer Recurrence and Guiding Chemotherapy

For the hundreds of thousands of women diagnosed each year with hormone receptor-positive, HER2-negative breast cancer, one of the most consequential questions after surgery is deceptively simple: is chemotherapy actually necessary? Most of these tumors respond to endocrine therapy, and a substantial proportion of patients would derive minimal benefit from the added toxicity of cytotoxic drugs. Yet identifying precisely who falls into that category has long depended on expensive multigene assays that are not always reimbursed, leaving clinicians and patients to navigate a costly gray zone. A new study from Swiss and British researchers, published in Pathology and Oncology Research, offers a potential way forward by evaluating an RNA-based test that could bring molecular subtyping and proliferation assessment into ordinary pathology workflows at a fraction of the established price.

The research team, led by Marcus Vetter and Elena Diana Chiru of Cantonal Hospital Basel-Land and the University of Basel, together with collaborators at APIS Assay Technologies in Manchester, examined the APIS Breast Cancer Subtyping Kit, or BCSK, in 141 adults diagnosed with early luminal HER2-negative breast cancer between 2020 and 2022 at Cantonal Hospital Basel-Land and Basel University Hospital. Every patient in the cohort already had a valid Oncotype DX Recurrence Score on file and was planned to receive at least one line of adjuvant therapy, allowing the investigators to benchmark the newer assay against one of the most widely used genomic tests in breast oncology. In a subset of 59 patients, the team also performed the Prosigna PAM50 test, which yields an intrinsic molecular subtype and a risk of recurrence score.

The technical foundation of the APIS kit is reverse transcription quantitative polymerase chain reaction, a workhorse method of molecular pathology. The assay measures the expression of seven genes in formalin-fixed, paraffin-embedded tissue: ESR1, the gene encoding the estrogen receptor; PGR, encoding the progesterone receptor; ERBB2, encoding HER2; and MKI67, which encodes the proliferation marker Ki-67, alongside three additional proliferation genes, CCNA2, PCNA, and KIF23. From the relative expression of the four proliferation markers, a logistic model calculates a proliferation score between zero and one, with values below 0.5 classified as low proliferation and values above 0.5 as high proliferation. This quantitative, mRNA-level approach is designed to sidestep a well-known weakness of routine diagnostics: immunohistochemical measurement of Ki-67 is notoriously prone to preanalytical and analytical variability, as well as disagreement between observers, which can ripple into clinical decision-making.

The concordance results at the level of individual markers were striking. The BCSK showed essentially perfect agreement with immunohistochemistry for the estrogen receptor, at 99.29 percent, and for HER2, where all 141 samples were concordantly negative. Agreement for the progesterone receptor reached 86.52 percent. The proliferation marker told a more nuanced story. When Ki-67 staining from the original core needle biopsy was compared with the qPCR result from the resected specimen, positive percent agreement was only 43.55 percent, likely reflecting genuine tumor heterogeneity between a small biopsy and the larger resection. But when both the immunohistochemistry and the RNA test were performed on matched resected tissue, overall agreement rose to 71.19 percent, supporting the idea that much of the apparent discordance stems from comparing different tissue pieces rather than from flaws in the molecular assay itself.

Moving from single markers to molecular subtypes, the BCSK agreed with the Prosigna PAM50 classification in 71.2 percent of cases, an encouraging result for a test built on just seven genes. Perhaps more notable was the comparison with routine immunohistochemistry-based subtyping. The BCSK achieved an overall percent agreement of 71.2 percent with IHC-defined subtypes, whereas the PAM50 agreed with IHC in only 54.2 percent of cases. In other words, the inexpensive qPCR kit tracked conventional pathology assessments more closely than the established nanoscale gene-expression platform did. Three cases labeled luminal B by IHC were reclassified as basal-like by PAM50, and one was labeled HER2-enriched, underscoring how different platforms can meaningfully reshape a tumor’s molecular portrait.

The heart of the study lies in how the APIS proliferation score performs as a risk stratification tool when set against the two commercially validated recurrence scores. Treating all three as continuous variables, the researchers found a moderate positive correlation between the APIS score and the Prosigna risk of recurrence score, with a Spearman coefficient of 0.4787, and a somewhat weaker correlation with the Oncotype DX Recurrence Score, at 0.4006. Tellingly, the correlation between Prosigna and Oncotype DX themselves was weak in this cohort, at just 0.2255. That observation echoes earlier findings from the OPTIMA prelim trial, in which fewer than 40 percent of patients were classified alike by the two established assays, a reminder that these tools measure overlapping but distinct biological signals. Oncotype DX thresholds its proliferation component and blends it with an estrogen signaling module including PGR expression, whereas Prosigna and the APIS score lean more directly on proliferation, which transcriptomic studies suggest carries much of the prognostic information in breast cancer.

The risk distributions themselves were revealing. Within the 59-patient subset, Prosigna classified 23.7 percent of cases as low risk, 44.1 percent as intermediate, and 32.3 percent as high. Oncotype DX placed 67.8 percent in the low-risk category, 6.8 percent in the intermediate, and 25.4 percent in the high category. The APIS proliferation score, using its binary cutoff, labeled 39 percent of cases as low risk and a striking 61 percent as high risk, suggesting a more conservative stance toward stratification. During the study’s follow-up window, four patients experienced clinical relapse; two of them had been classified as high risk by both Oncotype DX and the APIS proliferation score, and one had a high Prosigna risk of recurrence score. The authors are careful to note that with so few events, no meaningful outcome analysis is possible, and the trial was not designed or powered for that purpose.

The clinical and economic arguments woven through the paper are as compelling as the analytical data. Established prognostic gene-expression assays cost between roughly 1,500 and 2,500 pounds, and reimbursement criteria exclude many patients who might still benefit from molecular risk assessment. The APIS kit, by contrast, is described as cost-neutral relative to routine immunohistochemistry, uses the same formalin-fixed paraffin-embedded blocks already collected at diagnosis, and can in principle be run and reported by local pathology laboratories rather than centralized reference facilities. Because results could be available from the initial core needle biopsy, the proliferation score might inform the very first multidisciplinary team discussion, flagging patients with very high or very low scores for whom additional costly testing may be unnecessary, while reserving full multigene assays for cases with intermediate scores near the cutoff.

The study’s limitations deserve equal billing with its promise. The cohort of 141 patients, and the subset of 59 tested with PAM50, are small and may not capture the heterogeneity of the broader breast cancer population. Follow-up was short and clinical events were rare, so the analysis speaks to analytical concordance rather than proven prognostic power. Most immunohistochemistry was performed on core needle biopsies with limited validation on matched resections, and the retrospective design meant that molecular results could not influence the treatment decisions actually made. There is also an industry dimension that readers should weigh: five co-authors are employed by APIS Assay Technologies, and the company funded the study and was involved in its design, analysis, and writing.

Even with those caveats, the study marks a meaningful proof of concept. It demonstrates that a compact, seven-gene qPCR assay can reproduce receptor status with near-perfect fidelity, approximate molecular subtyping, and generate a proliferation score that correlates with the drivers of two clinically validated recurrence tests, all at a cost structure that could democratize access to molecular profiling. If larger studies with longer follow-up confirm that the proliferation score reliably identifies patients who can safely forgo chemotherapy, this locally run, biopsy-based approach could reshape the diagnostic pathway for the roughly 65 percent of breast cancer patients whose tumors are hormone receptor-positive and HER2-negative. The message for patients and clinicians alike is one of cautious optimism: the tools for cheaper, faster, and more reproducible risk assessment are edging closer to the clinic, but the decisive validation trials are still to come.

Subject of Research: Evaluation of an RNA-based qPCR assay for molecular subtyping and proliferation-based recurrence risk prediction in early HR-positive, HER2-negative breast cancer.

Article Title: Evaluating an RNA-based test for proliferation assessment and recurrence prediction in early HR+/HER2− breast cancer

Article References: Evaluating an RNA-based test for proliferation assessment and recurrence prediction in early HR+/HER2− breast cancer. (n.d.). https://doi.org/10.3389/pore.2025.1612113

Image Credits: AI Generated

DOI: 10.3389/pore.2025.1612113

Keywords: breast cancer, RNA-based testing, molecular subtyping, proliferation score, Oncotype DX, Prosigna PAM50, Ki-67, recurrence risk, endocrine therapy, adjuvant chemotherapy, qPCR, biomarkers