OBJECTIVES: The two well-known population-adjusted indirect treatment comparison methods for unanchored indirect treatment comparisons in the case of no common comparator are matching-adjusted indirect treatment comparison (MAIC) and simulated treatment comparison (STC). In HTA, MAIC has been applied much more frequently than STC when evidence is from single-arm studies. We conducted a systematic literature review and it shows that there is no clear guidance on the use of unanchored STC for time-to-event data. We propose a novel unanchored STC approach for time-to-event data and evaluate its performance using a simulation study.

METHODS: The STC approach uses the regression technique to adjust for the imbalances in observed covariates. When the outcome is time-to-event, the first step is to assess the suitability of fitting a Cox regression model to the data from the study with individual patient-level data (IPD) available. For the cases where proportional hazard assumptions are not valid, we propose to use a stratified Cox model to account for heterogeneity among different groups with the flexibility of using different baseline hazard function for each stratum. The second step involves simulating covariates for the comparator study using a Gaussian copula. In the third step, we propose to predict the survival probabilities given the simulated covariates. The final step is to reconstruct the IPD for the comparator population based on the predicted survival probabilities and the published Kaplan-Meier curves. Once IPD are reconstructed for both arms, standard survival analysis could be applied to generate the relative treatment effect.

RESULTS: Our proposed unanchored STC approach for time-to-event data allows for modelling non-proportional hazards and the simulation study shows that it performs well across all scenarios.

CONCLUSIONS: We provide a novel unanchored STC approach for time-to-event data, which performs well in the simulation study. Unanchored STC should also be considered in addition to MAIC when analysing single-arm studies.