RE: "Cost-effectiveness analysis of fulvestrant versus anastrozole as first-line treatment for hormone receptor-positive advanced breast cancer," by Ding H., etal.

Abstract

After reading the recent publication entitled “Cost-effectiveness analysis of fulvestrant versus anastrozole as first-line treatment for hormone receptor-positive advanced breast cancer,” by Ding H., et al., I would like to make some important comments. The authors state that these transition probabilities were informed using data from the FIRST study, a Phase II, randomised, open-label study of fulvestrant 500 mg versus anastrozole 1 mg in post-menopausal women with oestrogen receptor-positive and/or progesterone receptor-positive locally advanced or metastatic breast cancer who had not received any prior endocrine therapy for locally advanced or metastatic disease (Ellis et al., 2015; Robertson et al., 2012). Previous adjuvant/neoadjuvant endocrine therapy for early disease, completed more than 12 months before randomisation, was permitted. While a Phase III, randomised, double-blind study of fulvestrant 500 mg versus anastrozole 1 mg in the first-line treatment setting has been undertaken, the overall survival (OS) results are immature (Robertson et al., 2016). Estimation of the transition probabilities requires careful consideration of the survival model for each transition, taking into account the competing and sequential nature of events modelled (NICE, 2017). It typically requires the availability of patient-level data to accurately calculate the transitions. However, the authors only had Kaplan–Meier curves and summary patient data available. Furthermore, within this study, a constant transition rate is used (in line with the equation above), which is based on the median time to event. This is a significant assumption to make, and the authors do not provide a rationale for this choice. While reference to Sonnenberg and Beck (1993) refers to this technique, it must be highlighted that these methods are somewhat outdated and that methods, including parametric modelling of survival curves, have since become more developed (Bagust & Beale, 2014; Latimer, 2013). Upon looking at the FIRST study data, both OS and progression-free survival (PFS) would not appear to be decreasing at a constant rate overtime. This is particularly true for fulvestrant, as the rate of decrease slows after the median is reached. Therefore, applying a constant rate to progression and death does not appear to be valid. Such a method likely serves to underestimate the efficacy of fulvestrant and the relative efficacy of fulvestrant to anastrozole. This could therefore lead to an underestimation in quality-adjusted life-years (QALYs) gained, estimated at 0.11 QALYs gained in this study (2.65 for fulvestrant versus 2.54 for anastrozole), and hence a potential overestimation in the incremental cost-effectiveness ratio. In the absence of patient-level data, transition probabilities are difficult to calculate, and other techniques should be explored. Within partitioned survival analysis, PFS and OS curves are fitted independently, and this can be created from Kaplan–Meier curves; such methods are advocated by NICE (2017). Finally, it should be noted fulvestrant 500 mg is now approved in HR+ HER2− postmenopausal advanced breast cancer for those with no prior endocrine therapy; this is a subset of the data used here and shows greater efficacy (Ellis et al., 2015; Robertson et al., 2012). Therefore this analysis is not representative of the approved indication. In summary, we recommend that this research be treated with caution, and that further analysis into the comparative effectiveness and cost-effectiveness of fulvestrant versus anastrozole be undertaken. This work is supported by AstraZeneca. Claire Telford is an employee of AstraZeneca.