This week NICE published the final appraisal determination for the use of rituximab (MabThera, Roche) for the maintenance treatment of follicular non-Hodgkin's lymphoma following response to first-line chemotherapy. As with many cancer therapies an important part of economic analysis was the extrapolation of trial data to estimate the lifetime treatment effect of treatment. Whilst the standard approach in the literature is to fit individual curves to the Kaplan-Meier plots then chose the ‘best’ fit, this is a simplistic approach.
In the non-Hodgkin's appraisal, the Liverpool Reviews and Implementation Group (LRiG) adopted a more thoughtful approach. Whilst the manufacturer used a guesstimate of the period over which rituximab provided a reduced risk of disease progression (=6 years), LRiG examined the cumulative PFS function for the control and intervention groups to identify a mixed model that described the data more closely. Their bi-phase exponential model was used to identify the point at which the harzard functions became parallel, and as such, marks the point at which therapeutic effect ends; 27 months, not 72.
HEDS and the Centre of Health Economics are currently undertaking a MRC-funded study looking at the methods used, and which are available, for extrapolating trial-based data. Mixed models, such as that used by LRiG are an important part of this work. The project will produce a set of recommendations on the favoured analytical approach(es). The study is being led by Alan Brennan.