Abstract P4-09-06: Analyzing the changes in the HR+/HER2- metastatic breast cancer (mBC) landscape since the arrival of CDK4/6 inhibitors with machine learning and visual analytics

Abstract

Abstract BACKGROUND: The real-world impact of CDK4/6 inhibitor use on HR+/HER2- mBC treatment sequencing, treatment response, and therapeutic utilization is limited. This study sought to describe treatment sequencing and response pre- and post-FDA approval of CDK4/6 inhibitors by conducting a traditional observational study approach and supplementing with a machine-learning methodology. METHODS: Female patients ≥18yrs were identified in the MarketScan Commercial and Medicare Supplement databases with continuous enrollment for at least 12 months pre-index, ≥2 medical claims for a BC diagnosis and ≥2 medical claims for metastatic disease (earliest=index date), and who had no treatment at any time with HER2 targeted therapy. Patients were excluded if they had received CDK4/6 or everolimus treatment prior to index. Treatment was stratified by line of therapy pre- and post- first CDK4/6 approval (February 3, 2015). We identified treatment patterns with standard distributions and visual analytics. Response to therapy and utilization was analyzed with prediction models and IBM Watson® machine learning population comparison models. RESULTS: A total of 19,558 patients were eligible for the study with a mean age at diagnosis of mBC of 62 (SD 13). Prior to first CDK4/6 inhibitor approval, anastrazole and letrozole monotherapies were most likely to be identified as both first and second line treatment. Following approval of the first CDK4/6 inhibitor in 2015, CDK4/6 inhibitors were observed as a first line treatment in 25% of patients, and as second line treatment in 24% of patients. Of patients diagnosed following the first CDK4/6 inhibitor approval, 44% of patients were exposed to endocrine therapy and 10% were exposed to chemotherapy in the pre-index period. Patients receiving CDK4/6 inhibitors in combination with endocrine therapy as first line treatment were observed to have a longer progression free survival time than patients receiving endocrine monotherapy. Visual analytics demonstrate a large variation in treatment sequencing, especially after first line therapy. Machine learning and prediction models identified a strong secular bias in the use of CDK4/6 and a signal for improved treatment response in patients with no exposure to endocrine therapies in the pre-index period. CONCLUSIONS: The mBC treatment landscape has changed significantly with the introduction of CDK4/6 inhibitors, which may be expected to impact long-term outcomes. Visual analytics and machine learning approaches can improve clinical insight. These approaches can help with identifying patient and clinical characteristics that predict response and utilization as well as appropriate treatment selections across lines of therapy. Citation Format: Meade D, Hensley Alford S, Mahatma S, Malangone-Monaco E, Boulanger L, Tkacz J, Turner SJ, Chen S, Manson S, Mejia S, Buleje I, Madan P, Kim G, Fowler R, Basar A. Analyzing the changes in the HR+/HER2- metastatic breast cancer (mBC) landscape since the arrival of CDK4/6 inhibitors with machine learning and visual analytics [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P4-09-06.