OBJECTIVES : Cost-effectiveness analyses in previously untreated patients often do not appropriately consider the effect of subsequent treatment. Choice of second- and third-line treatments is likely to affect model outcomes. In cases where overall survival (OS) data are immature, other endpoints such as progression-free survival (PFS) can be used as a surrogate for OS. We aimed to assess the interaction between subsequent treatment choice and surrogacy assumptions and their impact on cost-effectiveness results in the context of ribociclib for previously untreated patients in the hormone receptor-positive, HER2-negative breast cancer treatment pathway.

METHODS : Survival data from the MONALEESA-2 trial were digitised to generate pseudo-time-to-event data for ribociclib and letrozole, to which parametric survival models were fit. These data informed a cost-effectiveness model investigating ribociclib versus letrozole in previously untreated patients. The model included drug, resource use and adverse event costs, and treatment-line-specific utilities. Upon progression, patients received two additional treatment lines (base case: everolimus plus exemestane; capecitabine). Alternative treatments, including vinorelbine and eribulin, could be chosen at second- and third-line. Surrogacy scaling factors were used to investigate the effect of a PFS–OS surrogacy relationship on model outcomes, by reducing time spent in subsequent health states so the difference in OS between populations is reduced.

RESULTS : In the base case treatment pathway, partial surrogacy (38.5%) resulted in £77,620 per quality-adjusted life-year (QALY), while full surrogacy (100%) led to £71,481/QALY. Assuming partial surrogacy (38.5%), alternative second-line treatments led to a reduction of £22,343–28,963/QALY, and alternative third-line treatments resulted in an increase of £513–2,669/QALY.

CONCLUSIONS : For appraisals of interventions at first-line, where survival data are immature, both choice of subsequent treatment and approach to surrogacy are important in capturing appropriate estimates of cost-effectiveness. Cost-effectiveness models considering interventions early in treatment pathways must correctly model later-line treatments in order to accurately estimate cost-effectiveness.