The Impact of Chemotherapy and Survival Prediction by Machine Learning in Early Elderly Triple Negative Breast Cancer (eTNBC): a Population Based Study from the SEER Database

Purpose: We aimed to analysis the impact of chemotherapy and establish prediction models of prognosis in early elderly triple negative breast cancer (eTNBC) by using machine learning.

Methods: We enrolled 4,696 patients in SEER Database who were 70 years or older, diagnosed with primary early TNBC(larger than 5 mm), from 2010 to 2016. The propensity-score matched method was utilized to reduce covariable imbalance. Univariable and multivariable analyses were used to compare breast cancer-specific survival(BCSS) and overall survival(OS). Nine models were developed by machine learning to predict the 5-year OS and BCSS for patients received chemotherapy.

Results: Compared to matched patients in no-chemotherapy group, multivariate analysis showed a better survival in chemotherapy group. Stratified analyses by stage demonstrated that patients with stage II and stage III other than stage I could benefit from chemotherapy. Further investigation in stage II found that chemotherapy was a better prognostic indicator for patients with T2N0M0 and stage IIb, but not in T1N1M0. Patients with grade III could achieve a better survival by receiving chemotherapy, but those with grade I and II couldn't. With 0.75 in 5-year BCSS and 0.81 in 5-year OS for AUC, the LightGBM outperformed other algorithms.

Conclusion: For early eTNBC patients with stage I, T1N1M0 and grade I-II, chemotherapy couldn't improve survival. Therefore, de-escalation therapy might be appropriate for selected patients. The LightGBM is a trustful model to predict the survival and provide precious systemic treatment for patients received chemotherapy.