Abstract Background: Several histopathologic classification systems have been described to characterise residual tumour in patients undergoing neoadjuvant chemotherapy (NACT) for breast cancer. Residual Cancer Burden (RCB) has demonstrated prognostic value but requires pathologists to enter specific data items into an external portal to generate a score. This process is time consuming and many measurements are subject to inter-observer variability. The Modified Miller Payne (MMP) score is a global assessment of tumour response in the breast and lymph nodes by pathologists performed as part of routine pathological assessment without the need to measure specific variables and additional data entry. We aimed to assess the performance of the more pragmatic MMP score in predicting patient outcome alongside RCB. Methods: MMP score is a modification of the original Miller Payne method classifying residual tumour into 5 groups (no/minimal response, partial response, minimum residual disease, DCIS, complete response) plus assessment of nodal response. Post treatment histology was reviewed and RCB and MMP scores calculated as per protocol. Cox proportional hazard models were fitted to survival data: two univariate models where overall survival is regressed against RCB class and against MMP score, and a bivariate model for overall survival against MMP score and a binary variable for node response. Akaike Information Criterion (AIC) and Harrell’s C-index were used to assess the performance of each prognostic model. Unweighted and weighted estimates of Cohen’s kappa were used to measure the agreement between RCB class and MMP score. In addition, MMP score categories 4 and 5 were collapsed into a single category, which for the purpose of agreement analysis is considered equivalent to RCB class 0. Results: 412 early breast cancer patients that received NACT at the Edinburgh Breast Unit between 2011-2018, were included , with median follow-up of 5 years. Tumour molecular subtypes were as follows: ER+ HER2- 37%, HER2+ 32%, Triple negative 30%. The univariate Cox model for RCB had AIC 796 and C-index 0.65 (SE 0.03), against AIC 800 and C-index 0.64 (SE 0.04) of the univariate model for MMP score. The bivariate model with MMP score and node response displays an AIC of 795.1 and C-index = 0.66 (se: 0.03). All RCB and MMP categories included in the Cox models are significant to p = 0.0001, while node response is significant up to p = 0.05. The unweighted Cohen’s kappa between RCB class and MMP score is 0.58 (95% CI: 0.52 - 0.64), which increases to 0.85 (p = 0.0001) when accounting for elements off the main diagonal (weighted kappa). Conclusion: The results indicate potentially equivalent prognostication value between MMP score and RCB class. The comparison of model performance suggests that the bivariate Cox model using MMP score and node response can fit the available survival data as well as the univariate model using RCB class. While the estimated agreement on the main diagonal is only moderate, the weighted estimate shows a strong estimated agreement between the two scoring systems. Given the practical and efficient nature of this scoring method, MMP appears to be a robust alternative to RCB. Further investigation and research should focus on the prognostic value of MMP for recurrence-free survival and on external validation of the outlined predictive models. Univariate, OS ~ RCB classHR (CI 95%)Univariate, OS ~ MMP scoreHR (CI 95%)Bivariate, OS ~ MMP score + Node responseHR (CI 95%)Baseline: RCB 3Baseline: MMP 1Baseline: MMP 1 and Node Response 1RCB 20.35 (0.21 - 0.61)MMP 20.37 (0.21 - 0.67)MMP 20.31 (0.17 - 0.58)RCB 10.24 (0.11 - 0.56)MMP 30.28 (0.15 - 0.53)MMP 30.22 (0.11 - 0.43)RCB 00.21 (0.1 - 0.4)MMP 4-50.24 (0.12 - 0.51)MMP 4-50.19 (0.09 - 0.42)Node Resp 00.51 (0.3 - 0.86) Citation Format: Peter S Hall, Alistair Ironside, Giovanni Tramonti, Alex Cavanagh, Katy Quiohilag, Olga Oikonomidou. Modified Miller-Payne score as a pragmatic and efficient alternative to Residual Cancer Burden [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P1-02-04.
Read full abstract