OBJECTIVES: Cost-effectiveness analyses often necessitate the extrapolation of survival data beyond the observation period of a trial. In adjuvant breast cancer, risk of recurrence is high immediately following surgery but decreases over time. An extrapolation reliant solely on observed trial data may lead to estimated survival outcomes that are not reflective of survival rates observed at later timepoints. This study outlines a piecewise approach to survival extrapolation. Outcomes from this piecewise approach are then compared to standard extrapolation and observed values in the long-term follow-up of trials evaluating trastuzumab + chemotherapy (TC) in the adjuvant treatment of HER2+ breast cancer.

METHODS: The adjusted extrapolation was segmented into three separate “phases”. During phase one (time zero to year 4), a log-logistic extrapolation is fitted to invasive disease-free survival (IDFS) data from a Phase III randomised controlled trial, APHINITY. From years four to 10 (time period 2), the initial parametric extrapolation is augmented using a “cure” model. Implementation of the ‘cure’ model assumes that the proportion of patients being “cured” (no longer at risk of recurrence and only subject to background mortality) linearly increases with time from 0% at 48 months to 95% at 120 months. In phase three (year 10 onwards), 95% of event-free patients are assumed to be no longer at risk of recurrence and are susceptible to background mortality only.

RESULTS: The unadjusted extrapolation, vastly underestimates expected DFS at later timepoints. The piecewise approach produced similar hazard rates to those observed in the long-term follow-up of older adjuvant trials. IDFS/EFS values at 10 years are 75.23% and 70.00% for the adjusted extrapolation and the BCIRG-006 trial, respectively.

CONCLUSIONS: The adjustment appears to be beneficial and serves to improve the real-world accuracy of the extrapolation.