Abstract P5-20-12: Aromatase inhibitor failure: predictors and time to first failure among women with metastatic ER+/HER2− breast cancer in the US

Abstract

Abstract Little information is available on the real-world use of aromatase inhibitors (AIs) among post-menopausal women with metastatic ER+/HER2− breast cancer (BC). This retrospective administrative claims study evaluated potential predictors of post-index AI failure (AIF) and time to first AIF by agent among these patients. Women ≥ 55 years of age newly diagnosed with metastatic stage IV BC (index) were identified in the 2006–2010 Thomson Reuters MarketScan databases and followed until the earliest of chemotherapy, combination therapy, end of enrollment, inpatient death or end of study (03/31/2011). ER+/HER2− disease was defined as any endocrine therapy (ET: tamoxifen, fulvestrant) or AI (anastrozole - ANA, letrozole - LET, or exemestane - EXE) use in the variable length post-index period and no trastuzumab, lapatinib or toremifene use anytime in the 6-month pre-or post-index periods. Those with post-index AI use except when following chemotherapy or combination therapy, were retained for analysis. AIF post-index was defined as a switch to an alternative AI, ET, chemotherapy, or combination therapy. Logistic regression was used to assess the probability of having AIF as a function of demographic and clinical characteristics. A Cox proportional hazards model was used to assess the probability of AIF by first-line AI as a function of selected patient characteristics. Kaplan-Meier (KM) plots and log-rank tests evaluated differences in time to failure by first-line AI treatment. Among 4,249 eligible patients, 36% had ≥1 AIF. Patients with pre-index ET use were more likely to fail AI treatment post-index (OR 1.43, CI: 1.08–1.89). Conversely, those receiving AI treatment pre-index were less likely to experience post-index AI failure (OR 0.73, CI: 0.63–0.85). Patients were also less likely to fail post-index AI treatment when having undergone lumpectomy or mastectomy pre-index (OR 0.62, CI: 0.47–0.83). When found at index, metastases to the lung, liver, or bone significantly increased the likelihood of experiencing AI failure while those to the brain had no association. First-line AI treatment was present in 96% of patients (n = 4,073) and those taking ANA remained on treatment the longest prior to AIF, followed by LET and EXE. Results of the Cox model suggest that compared to first-line ANA treatment, EXE use was significantly associated with increased probability of AIF (HR 1.57, CI: 1.35–1.83), controlling for other confounders. However, the association of first-line LET and probability of AIF was not significantly different than that of ANA (HR 1.09, CI: 0.97–1.22). Compared to EXE, LET was significantly associated with a lower probability of AIF (HR 0.69, CI: 0.59–0.81). The log-rank test based on the KM estimates suggests the overall difference between the three time-to-AIF curves is significantly different (p < 0.01). In patients with metastatic ER+/HER2− BC, AI failure was significantly associated with pre-index BC treatments and sites of metastases at diagnosis. In addition, the probability of first-line AIF and time to AIF differed by agent. Further examination of the interactions between BC disease characteristics, treatment history, and AIF is needed. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P5-20-12.