Several multigene-based, risk scoring methods are available and their use is recommended to support treatment decision making in breast cancer. Although the value of these tests is established, critical challenges remain that limit their use, especially regarding turnaround time for fast decision making and the associated costs. We have developed DigiStain - a novel technology that uses mid-infrared imaging to precisely measure a surrogate of tumour aneuploidy, across unstained biopsy sections. Considering that genomic context shapes the pattern and consequences of aneuploidy during cancer development and progression, aneuploidy may have valuable prognostic or predictive value. Using bespoke software, DigiStain allows a quantitative score ‘DigiStain index’ (DI) to be reproducibly extracted from an objective physical measurement of a cancer i.e., aneuploidy. This information is generated within minutes and is available at the same time as routine H&E staining. We previously showed that DI significantly correlates with tumour grade and survival. To further validate the technology, we have investigated the ability of the DI to predict 5-year survival in a retrospective analysis of patients with oestrogen receptor (ER) positive breast cancer. Methods: Clinical samples were obtained from patients (N=944) treated at the University Hospital of Nottingham UK. Using the DigiStain imager, H&E images were registered before acquiring the DI score for the section. Statistical methodology followed the recommended approach of Royston et al. 2009, The BMJ. A multivariate logistic model modelling the ‘event’ of death by 5 years was generated. The predictive performance of the logistic regression model was assessed using Area Under the receiver operating characteristic Curve (AUC) plots. The predicted vs observed event probability was determined at various cut-offs and the accuracy of the classification was measured by its sensitivity and specificity. Results: Of 944 ER+ patients with follow up information, 77 (8.2%) died in the first 5 years. The age distribution was approximately normal, ranging from 26 to 70 (mean = 54.8 years). Tumour size ranged from 0.2 to 7.5 mm (mean = 1.90 mm and median = 1.70 mm) and had normal distribution when the natural log of size was considered. DI contributed to the prognostic model with (p=0.0018). Odds ratios showed increased odds of death by 5 years with higher DI. The AUC obtained for the model was 0.7722. Due to the number of events the sample was not split into training and validation sets. Instead bootstrapping (n=500) was used to assess performance. When adjusted by bootstrapping the AUC was reduced only slightly to 0.76. Classification tables for sensitivity and specificity across different risk cut-offs and distribution of 5-year predicted probabilities showed the best distinction between groups at low levels of predicted probabilities. As expected, there was a trade-off between sensitivity and specificity when choosing an appropriate cut-off point to define high risk of death by 5 years. Conclusions: The risk score obtained using the DI is a significant prognostic factor for 5-year survival in ER+ breast cancer (p=0.0018) with good diagnostic accuracy. Analysis of 10-year survival data is ongoing and plans to expand the sample collection to include patients from other sites are underway.

Citation Format: Hemmel Amrania. A Novel, Rapid and Economical Prognostic Tool For Adjuvant therapy Decisions in Hormone Positive Breast Cancer [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P2-11-24.