Caris Life Sciences, a leading TechBio company, has published a study in Cancer Research Communications demonstrating the ability of a multimodal, multitask deep learning model to estimate late distant recurrence risk in hormone receptor-positive (HR+) early breast cancer.
The study, titled ‘Development and Validation of a Multimodal-Multitask Deep Learning Approach for Estimating Late Distant Recurrence Risk in Hormone Receptor–Positive Early Breast Cancer’, was conducted by a team of researchers from across the public and private sectors, including Caris Life Sciences and two leading cooperative research organizations, the NSABP Foundation/NRG Oncology and the ECOG-ACRIN Cancer Research Group.
HR+ breast cancer represents approximately 70–80% of all breast cancer diagnoses and is associated with a prolonged risk of recurrence that can persist well beyond the initial five years of endocrine therapy, which is standard of care.
Recurrence can happen at the original tumour site or further away in the body (distant recurrence). While extended endocrine therapy for an additional five years may reduce this risk, it comes with a trade-off of prolonged, challenging side effects. Identifying which patients are most likely to benefit remains a significant clinical challenge.
The multimodal AI model integrates digitised haematoxylin and eosin (H&E) pathology images with clinicopathologic data to generate risk predictions for late distant recurrence in both node-positive and node-negative HR+ breast cancer.
It was developed using banked tumour specimens contributed by 2,271 patients in the clinical trial.
Through a public-private partnership with ECOG-ACRIN, it was externally validated in an independent cohort of 4,300 banked specimens from patients who participated in the landmark study.
In the cohort, the study identified patients with substantially different outcomes, with a 10-year absolute distant recurrence risk difference of nearly 8% between high- and low-risk groups. External validation in the independent cohort confirmed the model’s prognostic performance and demonstrated that it independently predicted late distant recurrence risk, even after accounting for established clinical risk factors and the oncotype DX recurrence score.
Exploratory analyses further suggested that patients classified as high risk experienced greater absolute benefit from extended letrozole therapy compared to those classified as low risk, supporting the model’s potential utility in informing discussions regarding extended endocrine therapy.
Existing genomic assays provide valuable prognostic insights but may be limited by cost, accessibility, and turnaround time. The findings from this study suggest that AI-based analysis of routinely available pathology slides and clinical data could offer a scalable and accessible alternative or complement to existing tools.
By leveraging widely available diagnostic data, this approach may enable oncologists to better identify patients at elevated risk of late recurrence and support more personalized discussions regarding the benefits and risks of extended endocrine therapy.
In early May, Caris launched Caris MI Clarity, the first prognostic test designed to deliver insight into both early and late distant recurrence risk (years 0 through five and five through 15) for postmenopausal patients with HR+/HER2-negative, node-negative early-stage breast cancer at the time of diagnosis.
The new version includes decision support, not just prognosis. Adding information for chemotherapy decision support, identifying which patients are likely to benefit from chemo.
Extended endocrine therapy decision support, informing treatment beyond the first five years. Late-window ordering in years three to five, so recurrence risk can be reassessed during treatment, not only at diagnosis. MI Clarity unifies early and late distant risk into a single test, replacing two existing expensive tests with multi-week turnaround times.
