OBJECTIVES: Compared 2 indirect comparison methods of proportional hazard ratio (PH) model and non-proportional hazards ratio (non-PH) model based on an economic evaluation of trastuzumab emtansine (T-DM1) versus pyrotinib as 2^nd-line therapy for patients with metastatic breast cancer (mBC).

METHODS: Based on the survival data of T-DM1 from EMILIA trial and pyrotinib from PHOEBE trial, 2 indirect comparison methods were carried out: Bayesian network meta-analysis was used to derive a fixed hazard ratio (HR) for PH model and fractional polynomials network meta-analysis was used to derive the HR changing over time for non-PH model. PH model and non-PH model were applicated to adjust the survival curve of pyrotinib. A Markov model for mBC was built to evaluate the difference of life year (LY) and quality adjusted life year (QALY) using survival parameters derived from 2 HR models. Other parameters was taken from published literatures.

RESULTS: In PH model, T-DM1 provided more 1.1 LYs and more 0.52 QALYs (-0.25 LYs and -0.2 QALYs in progression-free state and 1.35 LYs and 0.72 QALYs in post-progression state). In non-PH model, T-DM1 provided more 0.5 LYs and more 0.33 QALYs (0.24 LYs and 0.2 QALYs in progression-free state and 0.26 LYs and 0.13 QALYs in post-progression state). All results showed patients using T-DM1 could gain more benefits, but there was a wide difference of PFS curves between 2 models. Meanwhile, a crossover of PFS and OS curves was found when adjusting the survival data of pyrotinib in PH model, so data correction had to be carried out.

CONCLUSIONS:

The non-PH model is more suitable to fit the survival data of T-DM1 versus pyrotinib, and the results were more in line with the facts. Based on the case, non-PH model used in indirect comparison was recommended to adjust survival data in Markov model when PH test was failed.