In oncology drug development for regulatory approval, it is critical to show direct evidence of clinical benefit (eg overall survival [OS]) or improvement in an established surrogate for clinical benefit (eg progression free survival). In some countries, the OS benefit of the new therapy is directly linked to medical reimbursements or payments. However, the presence of treatment switching to post-randomization medication creates difficulties in estimating the true effectiveness and cost-effectiveness of the new therapy.
Various methods exist to account for treatment switching eg marginal structural models with inverse probability of censoring weights,structural nested model, rank preserving structural failure time model,two-stage adjustment model and semi-competing risks model. We consider the intrinsic connection among all these methods through viewing them under the illness-death, multi-state model framework. Under this framework, the clinical data, study design and model assumptions necessary for each of the models can be taken into consideration systematically, thus leading to more robust conclusions and potentially facilitating the drug approval and reimbursement process.