Prognostic biomarkers that estimate residual recurrence risk after standard adjuvant therapy remain an unmet clinical need in high‑risk hormone receptor–positive, HER2‑negative (HR+/HER2‑) early breast cancer. This study evaluated Ataraxis Breast CTX, a causal multi‑modal AI model integrating clinical variables with histopathology image features, to stratify residual risk in clinically defined high‑risk patients and to identify candidates for adjuvant treatment escalation.
The primary aims were to (1) determine whether CTX‑prognostic can stratify residual recurrence risk after standard chemoendocrine therapy, and (2) explore whether CTX‑benefit predicts treatment benefit from adjuvant everolimus.
CTX is a causal multi‑modal model that combines structured clinical variables and image features extracted from H&E slides. The model provides two distinct outputs:
CTX‑prognostic: an estimate of residual recurrence risk after standard adjuvant therapy, intended to stratify patients into higher and lower residual‑risk groups.
CTX‑benefit: an estimate intended to capture likelihood of benefit from additional therapies (for example, escalation beyond standard chemoendocrine treatment).
Both outputs were applied to a clinically high‑risk subset of patients enrolled in the UNIRAD phase III trial for validation.
Prospective‑retrospective validation used participants from the phase III UNIRAD trial. The analysis included the subset of trial participants who met three criteria: they had not received neoadjuvant chemotherapy, they had received adjuvant chemotherapy, and H&E slides were available for image analysis. The reported sample size for this cohort was n = 556.
The primary endpoint for the CTX validation in this analysis was disease‑free survival (DFS). The investigators assessed CTX‑prognostic's ability to stratify patients into distinct risk groups for DFS. An exploratory multivariable analysis evaluated whether CTX‑benefit predicted differential treatment effect for adjuvant everolimus, adjusting for clinical covariates including age, tumor size, nodal status, grade, and menopausal status. The interaction between CTX‑benefit score and everolimus treatment was tested and reported.
CTX‑prognostic stratified patients into low‑risk and high‑risk groups with notably different outcomes. Across the analyzed cohort, the reported 5‑year DFS was 93% in the low‑risk group versus 80% in the high‑risk group. This difference indicates that CTX‑prognostic identified a subgroup of patients who experienced favorable outcomes after standard chemoendocrine adjuvant therapy.
In an exploratory multivariable analysis that adjusted for age, tumor size, nodal status, grade, and menopausal status, treatment benefit from adjuvant everolimus differed significantly according to the CTX‑benefit score. The interaction term for CTX‑benefit and everolimus treatment was reported as p = 0.01, suggesting a predictive relationship in this dataset.
Because this analysis is exploratory and derived from a prospective‑retrospective evaluation, these findings indicate potential predictive utility but require further validation before clinical implementation.
The authors report that CTX identified patients who have excellent outcomes on standard chemoendocrine therapy alone and identified patients with residual risk who may be candidates for adjuvant treatment escalation. Additionally, CTX showed a predictive signal for benefit from everolimus in an exploratory adjusted analysis. The results suggest CTX may be useful to stratify clinically high‑risk HR+/HER2‑ patients by residual risk and to inform selection of adjuvant escalation strategies.
Given that the report is a preprint and the exploratory nature of some analyses, the findings should be interpreted as hypothesis‑generating and will require independent and prospective validation prior to guiding clinical care.
The underlying data derive from the UNIRAD phase III clinical trial (NCT01805271). The UNIRAD trial was conducted according to Good Clinical Practice and the Declaration of Helsinki, with local regulatory and ethics approvals reported, including approval by the French medicines agency (ANSM) and an ethics committee (Comite de Protection des Personnes Sud‑Est IV‑Lyon) in September 2012. All patients provided written informed consent for trial participation and data use. The current re‑analysis is retrospective, and the authors report that no additional informed consent was required for this retrospective study.
A trial steering committee and an independent data monitoring committee oversaw the trial and monitored safety and efficacy.
The authors state that the trial data used in this analysis are not publicly available due to institutional and ethical constraints. Access to data can be requested directly from Unicancer. The manuscript is a preprint and has not undergone peer review; the exploratory nature of the everolimus analysis and the retrospective design of this re‑analysis are limitations noted by implication from the study design.
Several authors (L.P., C.T., D.B., K.Z., K.J.G., and J.W.) are equity holders of Ataraxis AI. All other authors declared no competing interests. Funder information declared includes Ataraxis AI.
Readers should note the preprint status and declared financial interests when interpreting the reported model performance and predictive findings.